Merge tag 'i3c/for-7.3' of git://git.kernel.org/pub/scm/linux/kernel/git/i3c/linuxgrafted

Pull I3C updates from Alexandre Belloni:
 "A new driver this cycle, for the AMD AXI I3C controller.

  There are also new features such as IBI wakeup support, SETAASA device
  discovery and ACPI support for the the DAA methods, meaning we can now
  communicate with the SPD devices on DDR5 modules.

  New driver:
   - AMD AXI I3C master controller

  Subsystem:
   - I3C Common Command Code (CCC) handling improvements, especially
     around GET CCCs
   - SETAASA device discovery support
   - ACPI support for all existing DAA methods like SETDASA, SETNEWDA as
     well as I2C devices on I3C bus
   - IBI-based wakeup support

  Drivers:
   - dw: SETAASA support
   - mipi-i3c-hci: advertise IBI wakeup capability, AMD_PT I3C
     controller support, PIO queue management support for HCI v1.2
   - renesas: improve suspend to RAM support, add runtime PM support"

* tag 'i3c/for-7.3' of git://git.kernel.org/pub/scm/linux/kernel/git/i3c/linux: (70 commits)
  i3c: dw: reduce do_daa time if there's no client
  i3c: mipi-i3c-hci: Add support for AMD_PT I3C controller
  i3c: mipi-i3c-hci: Add PIO queue management support for HCI v1.2
  i3c: mipi-i3c-hci: Fix missing STAT_IBI_STATUS_THLD in PIO mode
  i3c: dw: rename "pclk" to "apb" to match dt-binding
  i3c: dw: make struct dw_i3c_cmd smaller
  i3c: dw: use COMMAND_PORT_TRANSFER_ARG instead of hardcoding
  i3c: renesas: Don't register devices when ENTDAA times out
  i3c: master: dw-i3c-master: fix OD timing for first broadcast
  i3c: mipi-i3c-hci: Advertise IBI wakeup capability
  i3c: mipi-i3c-hci: Factor out i3c_hci_sysdev()
  i3c: mipi-i3c-hci-pci: Propagate I3C wakeup requirements to PCI
  i3c: master: Reject IBI requests from non-IBI-capable devices
  i3c: master: Add helper to query bus wakeup requirements
  i3c: master: Report wakeup events for IBIs
  i3c: master: Support IBI-based wakeup capability
  i3c: master: Fix potential UAF in i3c_device_match()
  i3c: master: Fix potential UAF in i3c_device_uevent()
  i3c: Make dev->desc locking assumptions explicit
  i3c: master: Fix use-after-free of master->this
  ...

18 files changed, 35247 insertions, 0 deletions

diff --git a/drivers/net/ethernet/sun/Kconfig b/drivers/net/ethernet/sun/Kconfig
new file mode 100644
index 000000000..b0d3f9a29
--- /dev/null
+++ b/drivers/net/ethernet/sun/Kconfig
@@ -0,0 +1,108 @@
+# SPDX-License-Identifier: GPL-2.0
+#
+# Sun network device configuration
+#
+
+config NET_VENDOR_SUN
+	bool "Sun devices"
+	default y
+	depends on SUN3 || SBUS || PCI || SUN_LDOMS
+	help
+	  If you have a network (Ethernet) card belonging to this class, say Y.
+
+	  Note that the answer to this question doesn't directly affect the
+	  kernel: saying N will just cause the configurator to skip all
+	  the questions about Sun network interfaces. If you say Y, you will be
+	  asked for your specific card in the following questions.
+
+if NET_VENDOR_SUN
+
+config HAPPYMEAL
+	tristate "Sun Happy Meal 10/100baseT support"
+	depends on (SBUS || PCI)
+	select CRC32
+	help
+	  This driver supports the "hme" interface present on most Ultra
+	  systems and as an option on older Sbus systems. This driver supports
+	  both PCI and Sbus devices. This driver also supports the "qfe" quad
+	  100baseT device available in both PCI and Sbus configurations.
+
+	  To compile this driver as a module, choose M here: the module
+	  will be called sunhme.
+
+config SUNBMAC
+	tristate "Sun BigMAC 10/100baseT support"
+	depends on SBUS
+	select CRC32
+	help
+	  This driver supports the "be" interface available as an Sbus option.
+	  This is Sun's older 100baseT Ethernet device.
+
+	  To compile this driver as a module, choose M here: the module
+	  will be called sunbmac.
+
+config SUNQE
+	tristate "Sun QuadEthernet support"
+	depends on SBUS
+	select CRC32
+	help
+	  This driver supports the "qe" 10baseT Ethernet device, available as
+	  an Sbus option. Note that this is not the same as Quad FastEthernet
+	  "qfe" which is supported by the Happy Meal driver instead.
+
+	  To compile this driver as a module, choose M here: the module
+	  will be called sunqe.
+
+config SUNGEM
+	tristate "Sun GEM support"
+	depends on PCI
+	select CRC32
+	select SUNGEM_PHY
+	help
+	  Support for the Sun GEM chip, aka Sun GigabitEthernet/P 2.0.  See also
+	  <http://docs.oracle.com/cd/E19455-01/806-3985-10/806-3985-10.pdf>.
+
+config CASSINI
+	tristate "Sun Cassini support"
+	depends on PCI
+	select CRC32
+	help
+	  Support for the Sun Cassini chip, aka Sun GigaSwift Ethernet. See also
+	  <http://docs.oracle.com/cd/E19113-01/giga.ether.pci/817-4341-10/817-4341-10.pdf>.
+
+config SUNVNET_COMMON
+	tristate "Common routines to support Sun Virtual Networking"
+	depends on SUN_LDOMS
+	depends on INET
+	default m
+
+config SUNVNET
+	tristate "Sun Virtual Network support"
+	default m
+	depends on SUN_LDOMS
+	depends on SUNVNET_COMMON
+	help
+	  Support for virtual network devices under Sun Logical Domains.
+
+config LDMVSW
+	tristate "Sun4v LDoms Virtual Switch support"
+	default m
+	depends on SUN_LDOMS
+	depends on SUNVNET_COMMON
+	help
+	  Support for virtual switch devices under Sun4v Logical Domains.
+	  This driver adds a network interface for every vsw-port node
+	  found in the machine description of a service domain.
+	  Linux bridge/switch software can use these interfaces for
+	  guest domain network interconnectivity or guest domain
+	  connection to a physical network on a service domain.
+
+config NIU
+	tristate "Sun Neptune 10Gbit Ethernet support"
+	depends on PCI
+	select CRC32
+	help
+	  This enables support for cards based upon Sun's
+	  Neptune chipset.
+
+endif # NET_VENDOR_SUN
diff --git a/drivers/net/ethernet/sun/Makefile b/drivers/net/ethernet/sun/Makefile
new file mode 100644
index 000000000..9a5249dee
--- /dev/null
+++ b/drivers/net/ethernet/sun/Makefile
@@ -0,0 +1,14 @@
+# SPDX-License-Identifier: GPL-2.0
+#
+# Makefile for the Sun network device drivers.
+#
+
+obj-$(CONFIG_HAPPYMEAL) += sunhme.o
+obj-$(CONFIG_SUNQE) += sunqe.o
+obj-$(CONFIG_SUNBMAC) += sunbmac.o
+obj-$(CONFIG_SUNGEM) += sungem.o
+obj-$(CONFIG_CASSINI) += cassini.o
+obj-$(CONFIG_SUNVNET_COMMON) += sunvnet_common.o
+obj-$(CONFIG_SUNVNET) += sunvnet.o
+obj-$(CONFIG_LDMVSW) += ldmvsw.o
+obj-$(CONFIG_NIU) += niu.o
diff --git a/drivers/net/ethernet/sun/cassini.c b/drivers/net/ethernet/sun/cassini.c
new file mode 100644
index 000000000..9a8b5cf2d
--- /dev/null
+++ b/drivers/net/ethernet/sun/cassini.c
@@ -0,0 +1,5212 @@
+// SPDX-License-Identifier: GPL-2.0+
+/* cassini.c: Sun Microsystems Cassini(+) ethernet driver.
+ *
+ * Copyright (C) 2004 Sun Microsystems Inc.
+ * Copyright (C) 2003 Adrian Sun (asun@darksunrising.com)
+ *
+ * This driver uses the sungem driver (c) David Miller
+ * (davem@redhat.com) as its basis.
+ *
+ * The cassini chip has a number of features that distinguish it from
+ * the gem chip:
+ *  4 transmit descriptor rings that are used for either QoS (VLAN) or
+ *      load balancing (non-VLAN mode)
+ *  batching of multiple packets
+ *  multiple CPU dispatching
+ *  page-based RX descriptor engine with separate completion rings
+ *  Gigabit support (GMII and PCS interface)
+ *  MIF link up/down detection works
+ *
+ * RX is handled by page sized buffers that are attached as fragments to
+ * the skb. here's what's done:
+ *  -- driver allocates pages at a time and keeps reference counts
+ *     on them.
+ *  -- the upper protocol layers assume that the header is in the skb
+ *     itself. as a result, cassini will copy a small amount (64 bytes)
+ *     to make them happy.
+ *  -- driver appends the rest of the data pages as frags to skbuffs
+ *     and increments the reference count
+ *  -- on page reclamation, the driver swaps the page with a spare page.
+ *     if that page is still in use, it frees its reference to that page,
+ *     and allocates a new page for use. otherwise, it just recycles the
+ *     page.
+ *
+ * NOTE: cassini can parse the header. however, it's not worth it
+ *       as long as the network stack requires a header copy.
+ *
+ * TX has 4 queues. currently these queues are used in a round-robin
+ * fashion for load balancing. They can also be used for QoS. for that
+ * to work, however, QoS information needs to be exposed down to the driver
+ * level so that subqueues get targeted to particular transmit rings.
+ * alternatively, the queues can be configured via use of the all-purpose
+ * ioctl.
+ *
+ * RX DATA: the rx completion ring has all the info, but the rx desc
+ * ring has all of the data. RX can conceivably come in under multiple
+ * interrupts, but the INT# assignment needs to be set up properly by
+ * the BIOS and conveyed to the driver. PCI BIOSes don't know how to do
+ * that. also, the two descriptor rings are designed to distinguish between
+ * encrypted and non-encrypted packets, but we use them for buffering
+ * instead.
+ *
+ * by default, the selective clear mask is set up to process rx packets.
+ */
+
+#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
+
+#include <linux/module.h>
+#include <linux/kernel.h>
+#include <linux/types.h>
+#include <linux/compiler.h>
+#include <linux/slab.h>
+#include <linux/delay.h>
+#include <linux/init.h>
+#include <linux/interrupt.h>
+#include <linux/vmalloc.h>
+#include <linux/ioport.h>
+#include <linux/pci.h>
+#include <linux/mm.h>
+#include <linux/highmem.h>
+#include <linux/list.h>
+#include <linux/dma-mapping.h>
+
+#include <linux/netdevice.h>
+#include <linux/etherdevice.h>
+#include <linux/skbuff.h>
+#include <linux/skbuff_ref.h>
+#include <linux/ethtool.h>
+#include <linux/crc32.h>
+#include <linux/random.h>
+#include <linux/mii.h>
+#include <linux/ip.h>
+#include <linux/tcp.h>
+#include <linux/mutex.h>
+#include <linux/firmware.h>
+
+#include <net/checksum.h>
+
+#include <linux/atomic.h>
+#include <asm/io.h>
+#include <asm/byteorder.h>
+#include <linux/uaccess.h>
+#include <linux/jiffies.h>
+
+#define CAS_NCPUS            num_online_cpus()
+
+#define cas_skb_release(x)  netif_rx(x)
+
+/* select which firmware to use */
+#define USE_HP_WORKAROUND
+#define HP_WORKAROUND_DEFAULT /* select which firmware to use as default */
+#define CAS_HP_ALT_FIRMWARE   cas_prog_null /* alternate firmware */
+
+#include "cassini.h"
+
+#define USE_TX_COMPWB      /* use completion writeback registers */
+#define USE_CSMA_CD_PROTO  /* standard CSMA/CD */
+#define USE_RX_BLANK       /* hw interrupt mitigation */
+#undef USE_ENTROPY_DEV     /* don't test for entropy device */
+
+/* NOTE: these aren't useable unless PCI interrupts can be assigned.
+ * also, we need to make cp->lock finer-grained.
+ */
+#undef  USE_PCI_INTB
+#undef  USE_PCI_INTC
+#undef  USE_PCI_INTD
+#undef  USE_QOS
+
+#undef  USE_VPD_DEBUG       /* debug vpd information if defined */
+
+/* rx processing options */
+#define USE_PAGE_ORDER      /* specify to allocate large rx pages */
+#define RX_DONT_BATCH  0    /* if 1, don't batch flows */
+#define RX_COPY_ALWAYS 0    /* if 0, use frags */
+#define RX_COPY_MIN    64   /* copy a little to make upper layers happy */
+#undef  RX_COUNT_BUFFERS    /* define to calculate RX buffer stats */
+
+#define DRV_MODULE_NAME		"cassini"
+#define DRV_MODULE_VERSION	"1.6"
+#define DRV_MODULE_RELDATE	"21 May 2008"
+
+#define CAS_DEF_MSG_ENABLE	  \
+	(NETIF_MSG_DRV		| \
+	 NETIF_MSG_PROBE	| \
+	 NETIF_MSG_LINK		| \
+	 NETIF_MSG_TIMER	| \
+	 NETIF_MSG_IFDOWN	| \
+	 NETIF_MSG_IFUP		| \
+	 NETIF_MSG_RX_ERR	| \
+	 NETIF_MSG_TX_ERR)
+
+/* length of time before we decide the hardware is borked,
+ * and dev->tx_timeout() should be called to fix the problem
+ */
+#define CAS_TX_TIMEOUT			(HZ)
+#define CAS_LINK_TIMEOUT                (22*HZ/10)
+#define CAS_LINK_FAST_TIMEOUT           (1)
+
+/* timeout values for state changing. these specify the number
+ * of 10us delays to be used before giving up.
+ */
+#define STOP_TRIES_PHY 1000
+#define STOP_TRIES     5000
+
+/* specify a minimum frame size to deal with some fifo issues
+ * max mtu == 2 * page size - ethernet header - 64 - swivel =
+ *            2 * page_size - 0x50
+ */
+#define CAS_MIN_FRAME			97
+#define CAS_1000MB_MIN_FRAME            255
+#define CAS_MIN_MTU                     60
+#define CAS_MAX_MTU                     min(((cp->page_size << 1) - 0x50), 9000)
+
+#if 1
+/*
+ * Eliminate these and use separate atomic counters for each, to
+ * avoid a race condition.
+ */
+#else
+#define CAS_RESET_MTU                   1
+#define CAS_RESET_ALL                   2
+#define CAS_RESET_SPARE                 3
+#endif
+
+static char version[] =
+	DRV_MODULE_NAME ".c:v" DRV_MODULE_VERSION " (" DRV_MODULE_RELDATE ")\n";
+
+static int cassini_debug = -1;	/* -1 == use CAS_DEF_MSG_ENABLE as value */
+static int link_mode;
+
+MODULE_AUTHOR("Adrian Sun <asun@darksunrising.com>");
+MODULE_DESCRIPTION("Sun Cassini(+) ethernet driver");
+MODULE_LICENSE("GPL");
+MODULE_FIRMWARE("sun/cassini.bin");
+module_param(cassini_debug, int, 0);
+MODULE_PARM_DESC(cassini_debug, "Cassini bitmapped debugging message enable value");
+module_param(link_mode, int, 0);
+MODULE_PARM_DESC(link_mode, "default link mode");
+
+/*
+ * Work around for a PCS bug in which the link goes down due to the chip
+ * being confused and never showing a link status of "up."
+ */
+#define DEFAULT_LINKDOWN_TIMEOUT 5
+/*
+ * Value in seconds, for user input.
+ */
+static int linkdown_timeout = DEFAULT_LINKDOWN_TIMEOUT;
+module_param(linkdown_timeout, int, 0);
+MODULE_PARM_DESC(linkdown_timeout,
+"min reset interval in sec. for PCS linkdown issue; disabled if not positive");
+
+/*
+ * value in 'ticks' (units used by jiffies). Set when we init the
+ * module because 'HZ' in actually a function call on some flavors of
+ * Linux.  This will default to DEFAULT_LINKDOWN_TIMEOUT * HZ.
+ */
+static int link_transition_timeout;
+
+
+
+static u16 link_modes[] = {
+	BMCR_ANENABLE,			 /* 0 : autoneg */
+	0,				 /* 1 : 10bt half duplex */
+	BMCR_SPEED100,			 /* 2 : 100bt half duplex */
+	BMCR_FULLDPLX,			 /* 3 : 10bt full duplex */
+	BMCR_SPEED100|BMCR_FULLDPLX,	 /* 4 : 100bt full duplex */
+	CAS_BMCR_SPEED1000|BMCR_FULLDPLX /* 5 : 1000bt full duplex */
+};
+
+static const struct pci_device_id cas_pci_tbl[] = {
+	{ PCI_VDEVICE(SUN, PCI_DEVICE_ID_SUN_CASSINI) },
+	{ PCI_VDEVICE(NS, PCI_DEVICE_ID_NS_SATURN) },
+	{ }
+};
+
+MODULE_DEVICE_TABLE(pci, cas_pci_tbl);
+
+static void cas_set_link_modes(struct cas *cp);
+
+static inline void cas_lock_tx(struct cas *cp)
+{
+	int i;
+
+	for (i = 0; i < N_TX_RINGS; i++)
+		spin_lock_nested(&cp->tx_lock[i], i);
+}
+
+/* WTZ: QA was finding deadlock problems with the previous
+ * versions after long test runs with multiple cards per machine.
+ * See if replacing cas_lock_all with safer versions helps. The
+ * symptoms QA is reporting match those we'd expect if interrupts
+ * aren't being properly restored, and we fixed a previous deadlock
+ * with similar symptoms by using save/restore versions in other
+ * places.
+ */
+#define cas_lock_all_save(cp, flags) \
+do { \
+	struct cas *xxxcp = (cp); \
+	spin_lock_irqsave(&xxxcp->lock, flags); \
+	cas_lock_tx(xxxcp); \
+} while (0)
+
+static inline void cas_unlock_tx(struct cas *cp)
+{
+	int i;
+
+	for (i = N_TX_RINGS; i > 0; i--)
+		spin_unlock(&cp->tx_lock[i - 1]);
+}
+
+#define cas_unlock_all_restore(cp, flags) \
+do { \
+	struct cas *xxxcp = (cp); \
+	cas_unlock_tx(xxxcp); \
+	spin_unlock_irqrestore(&xxxcp->lock, flags); \
+} while (0)
+
+static void cas_disable_irq(struct cas *cp, const int ring)
+{
+	/* Make sure we won't get any more interrupts */
+	if (ring == 0) {
+		writel(0xFFFFFFFF, cp->regs + REG_INTR_MASK);
+		return;
+	}
+
+	/* disable completion interrupts and selectively mask */
+	if (cp->cas_flags & CAS_FLAG_REG_PLUS) {
+		switch (ring) {
+#if defined (USE_PCI_INTB) || defined(USE_PCI_INTC) || defined(USE_PCI_INTD)
+#ifdef USE_PCI_INTB
+		case 1:
+#endif
+#ifdef USE_PCI_INTC
+		case 2:
+#endif
+#ifdef USE_PCI_INTD
+		case 3:
+#endif
+			writel(INTRN_MASK_CLEAR_ALL | INTRN_MASK_RX_EN,
+			       cp->regs + REG_PLUS_INTRN_MASK(ring));
+			break;
+#endif
+		default:
+			writel(INTRN_MASK_CLEAR_ALL, cp->regs +
+			       REG_PLUS_INTRN_MASK(ring));
+			break;
+		}
+	}
+}
+
+static inline void cas_mask_intr(struct cas *cp)
+{
+	int i;
+
+	for (i = 0; i < N_RX_COMP_RINGS; i++)
+		cas_disable_irq(cp, i);
+}
+
+static void cas_enable_irq(struct cas *cp, const int ring)
+{
+	if (ring == 0) { /* all but TX_DONE */
+		writel(INTR_TX_DONE, cp->regs + REG_INTR_MASK);
+		return;
+	}
+
+	if (cp->cas_flags & CAS_FLAG_REG_PLUS) {
+		switch (ring) {
+#if defined (USE_PCI_INTB) || defined(USE_PCI_INTC) || defined(USE_PCI_INTD)
+#ifdef USE_PCI_INTB
+		case 1:
+#endif
+#ifdef USE_PCI_INTC
+		case 2:
+#endif
+#ifdef USE_PCI_INTD
+		case 3:
+#endif
+			writel(INTRN_MASK_RX_EN, cp->regs +
+			       REG_PLUS_INTRN_MASK(ring));
+			break;
+#endif
+		default:
+			break;
+		}
+	}
+}
+
+static inline void cas_unmask_intr(struct cas *cp)
+{
+	int i;
+
+	for (i = 0; i < N_RX_COMP_RINGS; i++)
+		cas_enable_irq(cp, i);
+}
+
+static inline void cas_entropy_gather(struct cas *cp)
+{
+#ifdef USE_ENTROPY_DEV
+	if ((cp->cas_flags & CAS_FLAG_ENTROPY_DEV) == 0)
+		return;
+
+	batch_entropy_store(readl(cp->regs + REG_ENTROPY_IV),
+			    readl(cp->regs + REG_ENTROPY_IV),
+			    sizeof(uint64_t)*8);
+#endif
+}
+
+static inline void cas_entropy_reset(struct cas *cp)
+{
+#ifdef USE_ENTROPY_DEV
+	if ((cp->cas_flags & CAS_FLAG_ENTROPY_DEV) == 0)
+		return;
+
+	writel(BIM_LOCAL_DEV_PAD | BIM_LOCAL_DEV_PROM | BIM_LOCAL_DEV_EXT,
+	       cp->regs + REG_BIM_LOCAL_DEV_EN);
+	writeb(ENTROPY_RESET_STC_MODE, cp->regs + REG_ENTROPY_RESET);
+	writeb(0x55, cp->regs + REG_ENTROPY_RAND_REG);
+
+	/* if we read back 0x0, we don't have an entropy device */
+	if (readb(cp->regs + REG_ENTROPY_RAND_REG) == 0)
+		cp->cas_flags &= ~CAS_FLAG_ENTROPY_DEV;
+#endif
+}
+
+/* access to the phy. the following assumes that we've initialized the MIF to
+ * be in frame rather than bit-bang mode
+ */
+static u16 cas_phy_read(struct cas *cp, int reg)
+{
+	u32 cmd;
+	int limit = STOP_TRIES_PHY;
+
+	cmd = MIF_FRAME_ST | MIF_FRAME_OP_READ;
+	cmd |= CAS_BASE(MIF_FRAME_PHY_ADDR, cp->phy_addr);
+	cmd |= CAS_BASE(MIF_FRAME_REG_ADDR, reg);
+	cmd |= MIF_FRAME_TURN_AROUND_MSB;
+	writel(cmd, cp->regs + REG_MIF_FRAME);
+
+	/* poll for completion */
+	while (limit-- > 0) {
+		udelay(10);
+		cmd = readl(cp->regs + REG_MIF_FRAME);
+		if (cmd & MIF_FRAME_TURN_AROUND_LSB)
+			return cmd & MIF_FRAME_DATA_MASK;
+	}
+	return 0xFFFF; /* -1 */
+}
+
+static int cas_phy_write(struct cas *cp, int reg, u16 val)
+{
+	int limit = STOP_TRIES_PHY;
+	u32 cmd;
+
+	cmd = MIF_FRAME_ST | MIF_FRAME_OP_WRITE;
+	cmd |= CAS_BASE(MIF_FRAME_PHY_ADDR, cp->phy_addr);
+	cmd |= CAS_BASE(MIF_FRAME_REG_ADDR, reg);
+	cmd |= MIF_FRAME_TURN_AROUND_MSB;
+	cmd |= val & MIF_FRAME_DATA_MASK;
+	writel(cmd, cp->regs + REG_MIF_FRAME);
+
+	/* poll for completion */
+	while (limit-- > 0) {
+		udelay(10);
+		cmd = readl(cp->regs + REG_MIF_FRAME);
+		if (cmd & MIF_FRAME_TURN_AROUND_LSB)
+			return 0;
+	}
+	return -1;
+}
+
+static void cas_phy_powerup(struct cas *cp)
+{
+	u16 ctl = cas_phy_read(cp, MII_BMCR);
+
+	if ((ctl & BMCR_PDOWN) == 0)
+		return;
+	ctl &= ~BMCR_PDOWN;
+	cas_phy_write(cp, MII_BMCR, ctl);
+}
+
+static void cas_phy_powerdown(struct cas *cp)
+{
+	u16 ctl = cas_phy_read(cp, MII_BMCR);
+
+	if (ctl & BMCR_PDOWN)
+		return;
+	ctl |= BMCR_PDOWN;
+	cas_phy_write(cp, MII_BMCR, ctl);
+}
+
+/* cp->lock held. note: the last put_page will free the buffer */
+static int cas_page_free(struct cas *cp, cas_page_t *page)
+{
+	dma_unmap_page(&cp->pdev->dev, page->dma_addr, cp->page_size,
+		       DMA_FROM_DEVICE);
+	__free_pages(page->buffer, cp->page_order);
+	kfree(page);
+	return 0;
+}
+
+#ifdef RX_COUNT_BUFFERS
+#define RX_USED_ADD(x, y)       ((x)->used += (y))
+#define RX_USED_SET(x, y)       ((x)->used  = (y))
+#else
+#define RX_USED_ADD(x, y) do { } while(0)
+#define RX_USED_SET(x, y) do { } while(0)
+#endif
+
+/* local page allocation routines for the receive buffers. jumbo pages
+ * require at least 8K contiguous and 8K aligned buffers.
+ */
+static cas_page_t *cas_page_alloc(struct cas *cp, const gfp_t flags)
+{
+	cas_page_t *page;
+
+	page = kmalloc_obj(cas_page_t, flags);
+	if (!page)
+		return NULL;
+
+	INIT_LIST_HEAD(&page->list);
+	RX_USED_SET(page, 0);
+	page->buffer = alloc_pages(flags, cp->page_order);
+	if (!page->buffer)
+		goto page_err;
+	page->dma_addr = dma_map_page(&cp->pdev->dev, page->buffer, 0,
+				      cp->page_size, DMA_FROM_DEVICE);
+	return page;
+
+page_err:
+	kfree(page);
+	return NULL;
+}
+
+/* initialize spare pool of rx buffers, but allocate during the open */
+static void cas_spare_init(struct cas *cp)
+{
+	spin_lock(&cp->rx_inuse_lock);
+	INIT_LIST_HEAD(&cp->rx_inuse_list);
+	spin_unlock(&cp->rx_inuse_lock);
+
+	spin_lock(&cp->rx_spare_lock);
+	INIT_LIST_HEAD(&cp->rx_spare_list);
+	cp->rx_spares_needed = RX_SPARE_COUNT;
+	spin_unlock(&cp->rx_spare_lock);
+}
+
+/* used on close. free all the spare buffers. */
+static void cas_spare_free(struct cas *cp)
+{
+	struct list_head list, *elem, *tmp;
+
+	/* free spare buffers */
+	INIT_LIST_HEAD(&list);
+	spin_lock(&cp->rx_spare_lock);
+	list_splice_init(&cp->rx_spare_list, &list);
+	spin_unlock(&cp->rx_spare_lock);
+	list_for_each_safe(elem, tmp, &list) {
+		cas_page_free(cp, list_entry(elem, cas_page_t, list));
+	}
+
+	INIT_LIST_HEAD(&list);
+#if 1
+	/*
+	 * Looks like Adrian had protected this with a different
+	 * lock than used everywhere else to manipulate this list.
+	 */
+	spin_lock(&cp->rx_inuse_lock);
+	list_splice_init(&cp->rx_inuse_list, &list);
+	spin_unlock(&cp->rx_inuse_lock);
+#else
+	spin_lock(&cp->rx_spare_lock);
+	list_splice_init(&cp->rx_inuse_list, &list);
+	spin_unlock(&cp->rx_spare_lock);
+#endif
+	list_for_each_safe(elem, tmp, &list) {
+		cas_page_free(cp, list_entry(elem, cas_page_t, list));
+	}
+}
+
+/* replenish spares if needed */
+static void cas_spare_recover(struct cas *cp, const gfp_t flags)
+{
+	struct list_head list, *elem, *tmp;
+	int needed, i;
+
+	/* check inuse list. if we don't need any more free buffers,
+	 * just free it
+	 */
+
+	/* make a local copy of the list */
+	INIT_LIST_HEAD(&list);
+	spin_lock(&cp->rx_inuse_lock);
+	list_splice_init(&cp->rx_inuse_list, &list);
+	spin_unlock(&cp->rx_inuse_lock);
+
+	list_for_each_safe(elem, tmp, &list) {
+		cas_page_t *page = list_entry(elem, cas_page_t, list);
+
+		/*
+		 * With the lockless pagecache, cassini buffering scheme gets
+		 * slightly less accurate: we might find that a page has an
+		 * elevated reference count here, due to a speculative ref,
+		 * and skip it as in-use. Ideally we would be able to reclaim
+		 * it. However this would be such a rare case, it doesn't
+		 * matter too much as we should pick it up the next time round.
+		 *
+		 * Importantly, if we find that the page has a refcount of 1
+		 * here (our refcount), then we know it is definitely not inuse
+		 * so we can reuse it.
+		 */
+		if (page_count(page->buffer) > 1)
+			continue;
+
+		list_del(elem);
+		spin_lock(&cp->rx_spare_lock);
+		if (cp->rx_spares_needed > 0) {
+			list_add(elem, &cp->rx_spare_list);
+			cp->rx_spares_needed--;
+			spin_unlock(&cp->rx_spare_lock);
+		} else {
+			spin_unlock(&cp->rx_spare_lock);
+			cas_page_free(cp, page);
+		}
+	}
+
+	/* put any inuse buffers back on the list */
+	if (!list_empty(&list)) {
+		spin_lock(&cp->rx_inuse_lock);
+		list_splice(&list, &cp->rx_inuse_list);
+		spin_unlock(&cp->rx_inuse_lock);
+	}
+
+	spin_lock(&cp->rx_spare_lock);
+	needed = cp->rx_spares_needed;
+	spin_unlock(&cp->rx_spare_lock);
+	if (!needed)
+		return;
+
+	/* we still need spares, so try to allocate some */
+	INIT_LIST_HEAD(&list);
+	i = 0;
+	while (i < needed) {
+		cas_page_t *spare = cas_page_alloc(cp, flags);
+		if (!spare)
+			break;
+		list_add(&spare->list, &list);
+		i++;
+	}
+
+	spin_lock(&cp->rx_spare_lock);
+	list_splice(&list, &cp->rx_spare_list);
+	cp->rx_spares_needed -= i;
+	spin_unlock(&cp->rx_spare_lock);
+}
+
+/* pull a page from the list. */
+static cas_page_t *cas_page_dequeue(struct cas *cp)
+{
+	struct list_head *entry;
+	int recover;
+
+	spin_lock(&cp->rx_spare_lock);
+	if (list_empty(&cp->rx_spare_list)) {
+		/* try to do a quick recovery */
+		spin_unlock(&cp->rx_spare_lock);
+		cas_spare_recover(cp, GFP_ATOMIC);
+		spin_lock(&cp->rx_spare_lock);
+		if (list_empty(&cp->rx_spare_list)) {
+			netif_err(cp, rx_err, cp->dev,
+				  "no spare buffers available\n");
+			spin_unlock(&cp->rx_spare_lock);
+			return NULL;
+		}
+	}
+
+	entry = cp->rx_spare_list.next;
+	list_del(entry);
+	recover = ++cp->rx_spares_needed;
+	spin_unlock(&cp->rx_spare_lock);
+
+	/* trigger the timer to do the recovery */
+	if ((recover & (RX_SPARE_RECOVER_VAL - 1)) == 0) {
+#if 1
+		atomic_inc(&cp->reset_task_pending);
+		atomic_inc(&cp->reset_task_pending_spare);
+		schedule_work(&cp->reset_task);
+#else
+		atomic_set(&cp->reset_task_pending, CAS_RESET_SPARE);
+		schedule_work(&cp->reset_task);
+#endif
+	}
+	return list_entry(entry, cas_page_t, list);
+}
+
+
+static void cas_mif_poll(struct cas *cp, const int enable)
+{
+	u32 cfg;
+
+	cfg  = readl(cp->regs + REG_MIF_CFG);
+	cfg &= (MIF_CFG_MDIO_0 | MIF_CFG_MDIO_1);
+
+	if (cp->phy_type & CAS_PHY_MII_MDIO1)
+		cfg |= MIF_CFG_PHY_SELECT;
+
+	/* poll and interrupt on link status change. */
+	if (enable) {
+		cfg |= MIF_CFG_POLL_EN;
+		cfg |= CAS_BASE(MIF_CFG_POLL_REG, MII_BMSR);
+		cfg |= CAS_BASE(MIF_CFG_POLL_PHY, cp->phy_addr);
+	}
+	writel((enable) ? ~(BMSR_LSTATUS | BMSR_ANEGCOMPLETE) : 0xFFFF,
+	       cp->regs + REG_MIF_MASK);
+	writel(cfg, cp->regs + REG_MIF_CFG);
+}
+
+/* Must be invoked under cp->lock */
+static void cas_begin_auto_negotiation(struct cas *cp,
+				       const struct ethtool_link_ksettings *ep)
+{
+	u16 ctl;
+#if 1
+	int lcntl;
+	int changed = 0;
+	int oldstate = cp->lstate;
+	int link_was_not_down = !(oldstate == link_down);
+#endif
+	/* Setup link parameters */
+	if (!ep)
+		goto start_aneg;
+	lcntl = cp->link_cntl;
+	if (ep->base.autoneg == AUTONEG_ENABLE) {
+		cp->link_cntl = BMCR_ANENABLE;
+	} else {
+		u32 speed = ep->base.speed;
+		cp->link_cntl = 0;
+		if (speed == SPEED_100)
+			cp->link_cntl |= BMCR_SPEED100;
+		else if (speed == SPEED_1000)
+			cp->link_cntl |= CAS_BMCR_SPEED1000;
+		if (ep->base.duplex == DUPLEX_FULL)
+			cp->link_cntl |= BMCR_FULLDPLX;
+	}
+#if 1
+	changed = (lcntl != cp->link_cntl);
+#endif
+start_aneg:
+	if (cp->lstate == link_up) {
+		netdev_info(cp->dev, "PCS link down\n");
+	} else {
+		if (changed) {
+			netdev_info(cp->dev, "link configuration changed\n");
+		}
+	}
+	cp->lstate = link_down;
+	cp->link_transition = LINK_TRANSITION_LINK_DOWN;
+	if (!cp->hw_running)
+		return;
+#if 1
+	/*
+	 * WTZ: If the old state was link_up, we turn off the carrier
+	 * to replicate everything we do elsewhere on a link-down
+	 * event when we were already in a link-up state..
+	 */
+	if (oldstate == link_up)
+		netif_carrier_off(cp->dev);
+	if (changed  && link_was_not_down) {
+		/*
+		 * WTZ: This branch will simply schedule a full reset after
+		 * we explicitly changed link modes in an ioctl. See if this
+		 * fixes the link-problems we were having for forced mode.
+		 */
+		atomic_inc(&cp->reset_task_pending);
+		atomic_inc(&cp->reset_task_pending_all);
+		schedule_work(&cp->reset_task);
+		cp->timer_ticks = 0;
+		mod_timer(&cp->link_timer, jiffies + CAS_LINK_TIMEOUT);
+		return;
+	}
+#endif
+	if (cp->phy_type & CAS_PHY_SERDES) {
+		u32 val = readl(cp->regs + REG_PCS_MII_CTRL);
+
+		if (cp->link_cntl & BMCR_ANENABLE) {
+			val |= (PCS_MII_RESTART_AUTONEG | PCS_MII_AUTONEG_EN);
+			cp->lstate = link_aneg;
+		} else {
+			if (cp->link_cntl & BMCR_FULLDPLX)
+				val |= PCS_MII_CTRL_DUPLEX;
+			val &= ~PCS_MII_AUTONEG_EN;
+			cp->lstate = link_force_ok;
+		}
+		cp->link_transition = LINK_TRANSITION_LINK_CONFIG;
+		writel(val, cp->regs + REG_PCS_MII_CTRL);
+
+	} else {
+		cas_mif_poll(cp, 0);
+		ctl = cas_phy_read(cp, MII_BMCR);
+		ctl &= ~(BMCR_FULLDPLX | BMCR_SPEED100 |
+			 CAS_BMCR_SPEED1000 | BMCR_ANENABLE);
+		ctl |= cp->link_cntl;
+		if (ctl & BMCR_ANENABLE) {
+			ctl |= BMCR_ANRESTART;
+			cp->lstate = link_aneg;
+		} else {
+			cp->lstate = link_force_ok;
+		}
+		cp->link_transition = LINK_TRANSITION_LINK_CONFIG;
+		cas_phy_write(cp, MII_BMCR, ctl);
+		cas_mif_poll(cp, 1);
+	}
+
+	cp->timer_ticks = 0;
+	mod_timer(&cp->link_timer, jiffies + CAS_LINK_TIMEOUT);
+}
+
+/* Must be invoked under cp->lock. */
+static int cas_reset_mii_phy(struct cas *cp)
+{
+	int limit = STOP_TRIES_PHY;
+	u16 val;
+
+	cas_phy_write(cp, MII_BMCR, BMCR_RESET);
+	udelay(100);
+	while (--limit) {
+		val = cas_phy_read(cp, MII_BMCR);
+		if ((val & BMCR_RESET) == 0)
+			break;
+		udelay(10);
+	}
+	return limit <= 0;
+}
+
+static void cas_saturn_firmware_init(struct cas *cp)
+{
+	const struct firmware *fw;
+	const char fw_name[] = "sun/cassini.bin";
+	int err;
+
+	if (PHY_NS_DP83065 != cp->phy_id)
+		return;
+
+	err = request_firmware(&fw, fw_name, &cp->pdev->dev);
+	if (err) {
+		pr_err("Failed to load firmware \"%s\"\n",
+		       fw_name);
+		return;
+	}
+	if (fw->size < 2) {
+		pr_err("bogus length %zu in \"%s\"\n",
+		       fw->size, fw_name);
+		goto out;
+	}
+	cp->fw_load_addr= fw->data[1] << 8 | fw->data[0];
+	cp->fw_size = fw->size - 2;
+	cp->fw_data = vmalloc(cp->fw_size);
+	if (!cp->fw_data)
+		goto out;
+	memcpy(cp->fw_data, &fw->data[2], cp->fw_size);
+out:
+	release_firmware(fw);
+}
+
+static void cas_saturn_firmware_load(struct cas *cp)
+{
+	int i;
+
+	if (!cp->fw_data)
+		return;
+
+	cas_phy_powerdown(cp);
+
+	/* expanded memory access mode */
+	cas_phy_write(cp, DP83065_MII_MEM, 0x0);
+
+	/* pointer configuration for new firmware */
+	cas_phy_write(cp, DP83065_MII_REGE, 0x8ff9);
+	cas_phy_write(cp, DP83065_MII_REGD, 0xbd);
+	cas_phy_write(cp, DP83065_MII_REGE, 0x8ffa);
+	cas_phy_write(cp, DP83065_MII_REGD, 0x82);
+	cas_phy_write(cp, DP83065_MII_REGE, 0x8ffb);
+	cas_phy_write(cp, DP83065_MII_REGD, 0x0);
+	cas_phy_write(cp, DP83065_MII_REGE, 0x8ffc);
+	cas_phy_write(cp, DP83065_MII_REGD, 0x39);
+
+	/* download new firmware */
+	cas_phy_write(cp, DP83065_MII_MEM, 0x1);
+	cas_phy_write(cp, DP83065_MII_REGE, cp->fw_load_addr);
+	for (i = 0; i < cp->fw_size; i++)
+		cas_phy_write(cp, DP83065_MII_REGD, cp->fw_data[i]);
+
+	/* enable firmware */
+	cas_phy_write(cp, DP83065_MII_REGE, 0x8ff8);
+	cas_phy_write(cp, DP83065_MII_REGD, 0x1);
+}
+
+
+/* phy initialization */
+static void cas_phy_init(struct cas *cp)
+{
+	u16 val;
+
+	/* if we're in MII/GMII mode, set up phy */
+	if (CAS_PHY_MII(cp->phy_type)) {
+		writel(PCS_DATAPATH_MODE_MII,
+		       cp->regs + REG_PCS_DATAPATH_MODE);
+
+		cas_mif_poll(cp, 0);
+		cas_reset_mii_phy(cp); /* take out of isolate mode */
+
+		if (PHY_LUCENT_B0 == cp->phy_id) {
+			/* workaround link up/down issue with lucent */
+			cas_phy_write(cp, LUCENT_MII_REG, 0x8000);
+			cas_phy_write(cp, MII_BMCR, 0x00f1);
+			cas_phy_write(cp, LUCENT_MII_REG, 0x0);
+
+		} else if (PHY_BROADCOM_B0 == (cp->phy_id & 0xFFFFFFFC)) {
+			/* workarounds for broadcom phy */
+			cas_phy_write(cp, BROADCOM_MII_REG8, 0x0C20);
+			cas_phy_write(cp, BROADCOM_MII_REG7, 0x0012);
+			cas_phy_write(cp, BROADCOM_MII_REG5, 0x1804);
+			cas_phy_write(cp, BROADCOM_MII_REG7, 0x0013);
+			cas_phy_write(cp, BROADCOM_MII_REG5, 0x1204);
+			cas_phy_write(cp, BROADCOM_MII_REG7, 0x8006);
+			cas_phy_write(cp, BROADCOM_MII_REG5, 0x0132);
+			cas_phy_write(cp, BROADCOM_MII_REG7, 0x8006);
+			cas_phy_write(cp, BROADCOM_MII_REG5, 0x0232);
+			cas_phy_write(cp, BROADCOM_MII_REG7, 0x201F);
+			cas_phy_write(cp, BROADCOM_MII_REG5, 0x0A20);
+
+		} else if (PHY_BROADCOM_5411 == cp->phy_id) {
+			val = cas_phy_read(cp, BROADCOM_MII_REG4);
+			val = cas_phy_read(cp, BROADCOM_MII_REG4);
+			if (val & 0x0080) {
+				/* link workaround */
+				cas_phy_write(cp, BROADCOM_MII_REG4,
+					      val & ~0x0080);
+			}
+
+		} else if (cp->cas_flags & CAS_FLAG_SATURN) {
+			writel((cp->phy_type & CAS_PHY_MII_MDIO0) ?
+			       SATURN_PCFG_FSI : 0x0,
+			       cp->regs + REG_SATURN_PCFG);
+
+			/* load firmware to address 10Mbps auto-negotiation
+			 * issue. NOTE: this will need to be changed if the
+			 * default firmware gets fixed.
+			 */
+			if (PHY_NS_DP83065 == cp->phy_id) {
+				cas_saturn_firmware_load(cp);
+			}
+			cas_phy_powerup(cp);
+		}
+
+		/* advertise capabilities */
+		val = cas_phy_read(cp, MII_BMCR);
+		val &= ~BMCR_ANENABLE;
+		cas_phy_write(cp, MII_BMCR, val);
+		udelay(10);
+
+		cas_phy_write(cp, MII_ADVERTISE,
+			      cas_phy_read(cp, MII_ADVERTISE) |
+			      (ADVERTISE_10HALF | ADVERTISE_10FULL |
+			       ADVERTISE_100HALF | ADVERTISE_100FULL |
+			       CAS_ADVERTISE_PAUSE |
+			       CAS_ADVERTISE_ASYM_PAUSE));
+
+		if (cp->cas_flags & CAS_FLAG_1000MB_CAP) {
+			/* make sure that we don't advertise half
+			 * duplex to avoid a chip issue
+			 */
+			val  = cas_phy_read(cp, CAS_MII_1000_CTRL);
+			val &= ~CAS_ADVERTISE_1000HALF;
+			val |= CAS_ADVERTISE_1000FULL;
+			cas_phy_write(cp, CAS_MII_1000_CTRL, val);
+		}
+
+	} else {
+		/* reset pcs for serdes */
+		u32 val;
+		int limit;
+
+		writel(PCS_DATAPATH_MODE_SERDES,
+		       cp->regs + REG_PCS_DATAPATH_MODE);
+
+		/* enable serdes pins on saturn */
+		if (cp->cas_flags & CAS_FLAG_SATURN)
+			writel(0, cp->regs + REG_SATURN_PCFG);
+
+		/* Reset PCS unit. */
+		val = readl(cp->regs + REG_PCS_MII_CTRL);
+		val |= PCS_MII_RESET;
+		writel(val, cp->regs + REG_PCS_MII_CTRL);
+
+		limit = STOP_TRIES;
+		while (--limit > 0) {
+			udelay(10);
+			if ((readl(cp->regs + REG_PCS_MII_CTRL) &
+			     PCS_MII_RESET) == 0)
+				break;
+		}
+		if (limit <= 0)
+			netdev_warn(cp->dev, "PCS reset bit would not clear [%08x]\n",
+				    readl(cp->regs + REG_PCS_STATE_MACHINE));
+
+		/* Make sure PCS is disabled while changing advertisement
+		 * configuration.
+		 */
+		writel(0x0, cp->regs + REG_PCS_CFG);
+
+		/* Advertise all capabilities except half-duplex. */
+		val  = readl(cp->regs + REG_PCS_MII_ADVERT);
+		val &= ~PCS_MII_ADVERT_HD;
+		val |= (PCS_MII_ADVERT_FD | PCS_MII_ADVERT_SYM_PAUSE |
+			PCS_MII_ADVERT_ASYM_PAUSE);
+		writel(val, cp->regs + REG_PCS_MII_ADVERT);
+
+		/* enable PCS */
+		writel(PCS_CFG_EN, cp->regs + REG_PCS_CFG);
+
+		/* pcs workaround: enable sync detect */
+		writel(PCS_SERDES_CTRL_SYNCD_EN,
+		       cp->regs + REG_PCS_SERDES_CTRL);
+	}
+}
+
+
+static int cas_pcs_link_check(struct cas *cp)
+{
+	u32 stat, state_machine;
+	int retval = 0;
+
+	/* The link status bit latches on zero, so you must
+	 * read it twice in such a case to see a transition
+	 * to the link being up.
+	 */
+	stat = readl(cp->regs + REG_PCS_MII_STATUS);
+	if ((stat & PCS_MII_STATUS_LINK_STATUS) == 0)
+		stat = readl(cp->regs + REG_PCS_MII_STATUS);
+
+	/* The remote-fault indication is only valid
+	 * when autoneg has completed.
+	 */
+	if ((stat & (PCS_MII_STATUS_AUTONEG_COMP |
+		     PCS_MII_STATUS_REMOTE_FAULT)) ==
+	    (PCS_MII_STATUS_AUTONEG_COMP | PCS_MII_STATUS_REMOTE_FAULT))
+		netif_info(cp, link, cp->dev, "PCS RemoteFault\n");
+
+	/* work around link detection issue by querying the PCS state
+	 * machine directly.
+	 */
+	state_machine = readl(cp->regs + REG_PCS_STATE_MACHINE);
+	if ((state_machine & PCS_SM_LINK_STATE_MASK) != SM_LINK_STATE_UP) {
+		stat &= ~PCS_MII_STATUS_LINK_STATUS;
+	} else if (state_machine & PCS_SM_WORD_SYNC_STATE_MASK) {
+		stat |= PCS_MII_STATUS_LINK_STATUS;
+	}
+
+	if (stat & PCS_MII_STATUS_LINK_STATUS) {
+		if (cp->lstate != link_up) {
+			if (cp->opened) {
+				cp->lstate = link_up;
+				cp->link_transition = LINK_TRANSITION_LINK_UP;
+
+				cas_set_link_modes(cp);
+				netif_carrier_on(cp->dev);
+			}
+		}
+	} else if (cp->lstate == link_up) {
+		cp->lstate = link_down;
+		if (link_transition_timeout != 0 &&
+		    cp->link_transition != LINK_TRANSITION_REQUESTED_RESET &&
+		    !cp->link_transition_jiffies_valid) {
+			/*
+			 * force a reset, as a workaround for the
+			 * link-failure problem. May want to move this to a
+			 * point a bit earlier in the sequence. If we had
+			 * generated a reset a short time ago, we'll wait for
+			 * the link timer to check the status until a
+			 * timer expires (link_transition_jiffies_valid is
+			 * true when the timer is running.)  Instead of using
+			 * a system timer, we just do a check whenever the
+			 * link timer is running - this clears the flag after
+			 * a suitable delay.
+			 */
+			retval = 1;
+			cp->link_transition = LINK_TRANSITION_REQUESTED_RESET;
+			cp->link_transition_jiffies = jiffies;
+			cp->link_transition_jiffies_valid = 1;
+		} else {
+			cp->link_transition = LINK_TRANSITION_ON_FAILURE;
+		}
+		netif_carrier_off(cp->dev);
+		if (cp->opened)
+			netif_info(cp, link, cp->dev, "PCS link down\n");
+
+		/* Cassini only: if you force a mode, there can be
+		 * sync problems on link down. to fix that, the following
+		 * things need to be checked:
+		 * 1) read serialink state register
+		 * 2) read pcs status register to verify link down.
+		 * 3) if link down and serial link == 0x03, then you need
+		 *    to global reset the chip.
+		 */
+		if ((cp->cas_flags & CAS_FLAG_REG_PLUS) == 0) {
+			/* should check to see if we're in a forced mode */
+			stat = readl(cp->regs + REG_PCS_SERDES_STATE);
+			if (stat == 0x03)
+				return 1;
+		}
+	} else if (cp->lstate == link_down) {
+		if (link_transition_timeout != 0 &&
+		    cp->link_transition != LINK_TRANSITION_REQUESTED_RESET &&
+		    !cp->link_transition_jiffies_valid) {
+			/* force a reset, as a workaround for the
+			 * link-failure problem.  May want to move
+			 * this to a point a bit earlier in the
+			 * sequence.
+			 */
+			retval = 1;
+			cp->link_transition = LINK_TRANSITION_REQUESTED_RESET;
+			cp->link_transition_jiffies = jiffies;
+			cp->link_transition_jiffies_valid = 1;
+		} else {
+			cp->link_transition = LINK_TRANSITION_STILL_FAILED;
+		}
+	}
+
+	return retval;
+}
+
+static int cas_pcs_interrupt(struct net_device *dev,
+			     struct cas *cp, u32 status)
+{
+	u32 stat = readl(cp->regs + REG_PCS_INTR_STATUS);
+
+	if ((stat & PCS_INTR_STATUS_LINK_CHANGE) == 0)
+		return 0;
+	return cas_pcs_link_check(cp);
+}
+
+static int cas_txmac_interrupt(struct net_device *dev,
+			       struct cas *cp, u32 status)
+{
+	u32 txmac_stat = readl(cp->regs + REG_MAC_TX_STATUS);
+
+	if (!txmac_stat)
+		return 0;
+
+	netif_printk(cp, intr, KERN_DEBUG, cp->dev,
+		     "txmac interrupt, txmac_stat: 0x%x\n", txmac_stat);
+
+	/* Defer timer expiration is quite normal,
+	 * don't even log the event.
+	 */
+	if ((txmac_stat & MAC_TX_DEFER_TIMER) &&
+	    !(txmac_stat & ~MAC_TX_DEFER_TIMER))
+		return 0;
+
+	spin_lock(&cp->stat_lock[0]);
+	if (txmac_stat & MAC_TX_UNDERRUN) {
+		netdev_err(dev, "TX MAC xmit underrun\n");
+		cp->net_stats[0].tx_fifo_errors++;
+	}
+
+	if (txmac_stat & MAC_TX_MAX_PACKET_ERR) {
+		netdev_err(dev, "TX MAC max packet size error\n");
+		cp->net_stats[0].tx_errors++;
+	}
+
+	/* The rest are all cases of one of the 16-bit TX
+	 * counters expiring.
+	 */
+	if (txmac_stat & MAC_TX_COLL_NORMAL)
+		cp->net_stats[0].collisions += 0x10000;
+
+	if (txmac_stat & MAC_TX_COLL_EXCESS) {
+		cp->net_stats[0].tx_aborted_errors += 0x10000;
+		cp->net_stats[0].collisions += 0x10000;
+	}
+
+	if (txmac_stat & MAC_TX_COLL_LATE) {
+		cp->net_stats[0].tx_aborted_errors += 0x10000;
+		cp->net_stats[0].collisions += 0x10000;
+	}
+	spin_unlock(&cp->stat_lock[0]);
+
+	/* We do not keep track of MAC_TX_COLL_FIRST and
+	 * MAC_TX_PEAK_ATTEMPTS events.
+	 */
+	return 0;
+}
+
+static void cas_load_firmware(struct cas *cp, cas_hp_inst_t *firmware)
+{
+	cas_hp_inst_t *inst;
+	u32 val;
+	int i;
+
+	i = 0;
+	while ((inst = firmware) && inst->note) {
+		writel(i, cp->regs + REG_HP_INSTR_RAM_ADDR);
+
+		val = CAS_BASE(HP_INSTR_RAM_HI_VAL, inst->val);
+		val |= CAS_BASE(HP_INSTR_RAM_HI_MASK, inst->mask);
+		writel(val, cp->regs + REG_HP_INSTR_RAM_DATA_HI);
+
+		val = CAS_BASE(HP_INSTR_RAM_MID_OUTARG, inst->outarg >> 10);
+		val |= CAS_BASE(HP_INSTR_RAM_MID_OUTOP, inst->outop);
+		val |= CAS_BASE(HP_INSTR_RAM_MID_FNEXT, inst->fnext);
+		val |= CAS_BASE(HP_INSTR_RAM_MID_FOFF, inst->foff);
+		val |= CAS_BASE(HP_INSTR_RAM_MID_SNEXT, inst->snext);
+		val |= CAS_BASE(HP_INSTR_RAM_MID_SOFF, inst->soff);
+		val |= CAS_BASE(HP_INSTR_RAM_MID_OP, inst->op);
+		writel(val, cp->regs + REG_HP_INSTR_RAM_DATA_MID);
+
+		val = CAS_BASE(HP_INSTR_RAM_LOW_OUTMASK, inst->outmask);
+		val |= CAS_BASE(HP_INSTR_RAM_LOW_OUTSHIFT, inst->outshift);
+		val |= CAS_BASE(HP_INSTR_RAM_LOW_OUTEN, inst->outenab);
+		val |= CAS_BASE(HP_INSTR_RAM_LOW_OUTARG, inst->outarg);
+		writel(val, cp->regs + REG_HP_INSTR_RAM_DATA_LOW);
+		++firmware;
+		++i;
+	}
+}
+
+static void cas_init_rx_dma(struct cas *cp)
+{
+	u64 desc_dma = cp->block_dvma;
+	u32 val;
+	int i, size;
+
+	/* rx free descriptors */
+	val = CAS_BASE(RX_CFG_SWIVEL, RX_SWIVEL_OFF_VAL);
+	val |= CAS_BASE(RX_CFG_DESC_RING, RX_DESC_RINGN_INDEX(0));
+	val |= CAS_BASE(RX_CFG_COMP_RING, RX_COMP_RINGN_INDEX(0));
+	if ((N_RX_DESC_RINGS > 1) &&
+	    (cp->cas_flags & CAS_FLAG_REG_PLUS))  /* do desc 2 */
+		val |= CAS_BASE(RX_CFG_DESC_RING1, RX_DESC_RINGN_INDEX(1));
+	writel(val, cp->regs + REG_RX_CFG);
+
+	val = (unsigned long) cp->init_rxds[0] -
+		(unsigned long) cp->init_block;
+	writel((desc_dma + val) >> 32, cp->regs + REG_RX_DB_HI);
+	writel((desc_dma + val) & 0xffffffff, cp->regs + REG_RX_DB_LOW);
+	writel(RX_DESC_RINGN_SIZE(0) - 4, cp->regs + REG_RX_KICK);
+
+	if (cp->cas_flags & CAS_FLAG_REG_PLUS) {
+		/* rx desc 2 is for IPSEC packets. however,
+		 * we don't it that for that purpose.
+		 */
+		val = (unsigned long) cp->init_rxds[1] -
+			(unsigned long) cp->init_block;
+		writel((desc_dma + val) >> 32, cp->regs + REG_PLUS_RX_DB1_HI);
+		writel((desc_dma + val) & 0xffffffff, cp->regs +
+		       REG_PLUS_RX_DB1_LOW);
+		writel(RX_DESC_RINGN_SIZE(1) - 4, cp->regs +
+		       REG_PLUS_RX_KICK1);
+	}
+
+	/* rx completion registers */
+	val = (unsigned long) cp->init_rxcs[0] -
+		(unsigned long) cp->init_block;
+	writel((desc_dma + val) >> 32, cp->regs + REG_RX_CB_HI);
+	writel((desc_dma + val) & 0xffffffff, cp->regs + REG_RX_CB_LOW);
+
+	if (cp->cas_flags & CAS_FLAG_REG_PLUS) {
+		/* rx comp 2-4 */
+		for (i = 1; i < MAX_RX_COMP_RINGS; i++) {
+			val = (unsigned long) cp->init_rxcs[i] -
+				(unsigned long) cp->init_block;
+			writel((desc_dma + val) >> 32, cp->regs +
+			       REG_PLUS_RX_CBN_HI(i));
+			writel((desc_dma + val) & 0xffffffff, cp->regs +
+			       REG_PLUS_RX_CBN_LOW(i));
+		}
+	}
+
+	/* read selective clear regs to prevent spurious interrupts
+	 * on reset because complete == kick.
+	 * selective clear set up to prevent interrupts on resets
+	 */
+	readl(cp->regs + REG_INTR_STATUS_ALIAS);
+	writel(INTR_RX_DONE | INTR_RX_BUF_UNAVAIL, cp->regs + REG_ALIAS_CLEAR);
+
+	/* set up pause thresholds */
+	val  = CAS_BASE(RX_PAUSE_THRESH_OFF,
+			cp->rx_pause_off / RX_PAUSE_THRESH_QUANTUM);
+	val |= CAS_BASE(RX_PAUSE_THRESH_ON,
+			cp->rx_pause_on / RX_PAUSE_THRESH_QUANTUM);
+	writel(val, cp->regs + REG_RX_PAUSE_THRESH);
+
+	/* zero out dma reassembly buffers */
+	for (i = 0; i < 64; i++) {
+		writel(i, cp->regs + REG_RX_TABLE_ADDR);
+		writel(0x0, cp->regs + REG_RX_TABLE_DATA_LOW);
+		writel(0x0, cp->regs + REG_RX_TABLE_DATA_MID);
+		writel(0x0, cp->regs + REG_RX_TABLE_DATA_HI);
+	}
+
+	/* make sure address register is 0 for normal operation */
+	writel(0x0, cp->regs + REG_RX_CTRL_FIFO_ADDR);
+	writel(0x0, cp->regs + REG_RX_IPP_FIFO_ADDR);
+
+	/* interrupt mitigation */
+#ifdef USE_RX_BLANK
+	val = CAS_BASE(RX_BLANK_INTR_TIME, RX_BLANK_INTR_TIME_VAL);
+	val |= CAS_BASE(RX_BLANK_INTR_PKT, RX_BLANK_INTR_PKT_VAL);
+	writel(val, cp->regs + REG_RX_BLANK);
+#else
+	writel(0x0, cp->regs + REG_RX_BLANK);
+#endif
+
+	/* interrupt generation as a function of low water marks for
+	 * free desc and completion entries. these are used to trigger
+	 * housekeeping for rx descs. we don't use the free interrupt
+	 * as it's not very useful
+	 */
+	/* val = CAS_BASE(RX_AE_THRESH_FREE, RX_AE_FREEN_VAL(0)); */
+	val = CAS_BASE(RX_AE_THRESH_COMP, RX_AE_COMP_VAL);
+	writel(val, cp->regs + REG_RX_AE_THRESH);
+	if (cp->cas_flags & CAS_FLAG_REG_PLUS) {
+		val = CAS_BASE(RX_AE1_THRESH_FREE, RX_AE_FREEN_VAL(1));
+		writel(val, cp->regs + REG_PLUS_RX_AE1_THRESH);
+	}
+
+	/* Random early detect registers. useful for congestion avoidance.
+	 * this should be tunable.
+	 */
+	writel(0x0, cp->regs + REG_RX_RED);
+
+	/* receive page sizes. default == 2K (0x800) */
+	val = 0;
+	if (cp->page_size == 0x1000)
+		val = 0x1;
+	else if (cp->page_size == 0x2000)
+		val = 0x2;
+	else if (cp->page_size == 0x4000)
+		val = 0x3;
+
+	/* round mtu + offset. constrain to page size. */
+	size = cp->dev->mtu + 64;
+	if (size > cp->page_size)
+		size = cp->page_size;
+
+	if (size <= 0x400)
+		i = 0x0;
+	else if (size <= 0x800)
+		i = 0x1;
+	else if (size <= 0x1000)
+		i = 0x2;
+	else
+		i = 0x3;
+
+	cp->mtu_stride = 1 << (i + 10);
+	val  = CAS_BASE(RX_PAGE_SIZE, val);
+	val |= CAS_BASE(RX_PAGE_SIZE_MTU_STRIDE, i);
+	val |= CAS_BASE(RX_PAGE_SIZE_MTU_COUNT, cp->page_size >> (i + 10));
+	val |= CAS_BASE(RX_PAGE_SIZE_MTU_OFF, 0x1);
+	writel(val, cp->regs + REG_RX_PAGE_SIZE);
+
+	/* enable the header parser if desired */
+	if (&CAS_HP_FIRMWARE[0] == &cas_prog_null[0])
+		return;
+
+	val = CAS_BASE(HP_CFG_NUM_CPU, CAS_NCPUS > 63 ? 0 : CAS_NCPUS);
+	val |= HP_CFG_PARSE_EN | HP_CFG_SYN_INC_MASK;
+	val |= CAS_BASE(HP_CFG_TCP_THRESH, HP_TCP_THRESH_VAL);
+	writel(val, cp->regs + REG_HP_CFG);
+}
+
+static inline void cas_rxc_init(struct cas_rx_comp *rxc)
+{
+	memset(rxc, 0, sizeof(*rxc));
+	rxc->word4 = cpu_to_le64(RX_COMP4_ZERO);
+}
+
+/* NOTE: we use the ENC RX DESC ring for spares. the rx_page[0,1]
+ * flipping is protected by the fact that the chip will not
+ * hand back the same page index while it's being processed.
+ */
+static inline cas_page_t *cas_page_spare(struct cas *cp, const int index)
+{
+	cas_page_t *page = cp->rx_pages[1][index];
+	cas_page_t *new;
+
+	if (page_count(page->buffer) == 1)
+		return page;
+
+	new = cas_page_dequeue(cp);
+	if (new) {
+		spin_lock(&cp->rx_inuse_lock);
+		list_add(&page->list, &cp->rx_inuse_list);
+		spin_unlock(&cp->rx_inuse_lock);
+	}
+	return new;
+}
+
+/* this needs to be changed if we actually use the ENC RX DESC ring */
+static cas_page_t *cas_page_swap(struct cas *cp, const int ring,
+				 const int index)
+{
+	cas_page_t **page0 = cp->rx_pages[0];
+	cas_page_t **page1 = cp->rx_pages[1];
+
+	/* swap if buffer is in use */
+	if (page_count(page0[index]->buffer) > 1) {
+		cas_page_t *new = cas_page_spare(cp, index);
+		if (new) {
+			page1[index] = page0[index];
+			page0[index] = new;
+		}
+	}
+	RX_USED_SET(page0[index], 0);
+	return page0[index];
+}
+
+static void cas_clean_rxds(struct cas *cp)
+{
+	/* only clean ring 0 as ring 1 is used for spare buffers */
+        struct cas_rx_desc *rxd = cp->init_rxds[0];
+	int i, size;
+
+	/* release all rx flows */
+	for (i = 0; i < N_RX_FLOWS; i++) {
+		struct sk_buff *skb;
+		while ((skb = __skb_dequeue(&cp->rx_flows[i]))) {
+			cas_skb_release(skb);
+		}
+	}
+
+	/* initialize descriptors */
+	size = RX_DESC_RINGN_SIZE(0);
+	for (i = 0; i < size; i++) {
+		cas_page_t *page = cas_page_swap(cp, 0, i);
+		rxd[i].buffer = cpu_to_le64(page->dma_addr);
+		rxd[i].index  = cpu_to_le64(CAS_BASE(RX_INDEX_NUM, i) |
+					    CAS_BASE(RX_INDEX_RING, 0));
+	}
+
+	cp->rx_old[0]  = RX_DESC_RINGN_SIZE(0) - 4;
+	cp->rx_last[0] = 0;
+	cp->cas_flags &= ~CAS_FLAG_RXD_POST(0);
+}
+
+static void cas_clean_rxcs(struct cas *cp)
+{
+	int i, j;
+
+	/* take ownership of rx comp descriptors */
+	memset(cp->rx_cur, 0, sizeof(*cp->rx_cur)*N_RX_COMP_RINGS);
+	memset(cp->rx_new, 0, sizeof(*cp->rx_new)*N_RX_COMP_RINGS);
+	for (i = 0; i < N_RX_COMP_RINGS; i++) {
+		struct cas_rx_comp *rxc = cp->init_rxcs[i];
+		for (j = 0; j < RX_COMP_RINGN_SIZE(i); j++) {
+			cas_rxc_init(rxc + j);
+		}
+	}
+}
+
+#if 0
+/* When we get a RX fifo overflow, the RX unit is probably hung
+ * so we do the following.
+ *
+ * If any part of the reset goes wrong, we return 1 and that causes the
+ * whole chip to be reset.
+ */
+static int cas_rxmac_reset(struct cas *cp)
+{
+	struct net_device *dev = cp->dev;
+	int limit;
+	u32 val;
+
+	/* First, reset MAC RX. */
+	writel(cp->mac_rx_cfg & ~MAC_RX_CFG_EN, cp->regs + REG_MAC_RX_CFG);
+	for (limit = 0; limit < STOP_TRIES; limit++) {
+		if (!(readl(cp->regs + REG_MAC_RX_CFG) & MAC_RX_CFG_EN))
+			break;
+		udelay(10);
+	}
+	if (limit == STOP_TRIES) {
+		netdev_err(dev, "RX MAC will not disable, resetting whole chip\n");
+		return 1;
+	}
+
+	/* Second, disable RX DMA. */
+	writel(0, cp->regs + REG_RX_CFG);
+	for (limit = 0; limit < STOP_TRIES; limit++) {
+		if (!(readl(cp->regs + REG_RX_CFG) & RX_CFG_DMA_EN))
+			break;
+		udelay(10);
+	}
+	if (limit == STOP_TRIES) {
+		netdev_err(dev, "RX DMA will not disable, resetting whole chip\n");
+		return 1;
+	}
+
+	mdelay(5);
+
+	/* Execute RX reset command. */
+	writel(SW_RESET_RX, cp->regs + REG_SW_RESET);
+	for (limit = 0; limit < STOP_TRIES; limit++) {
+		if (!(readl(cp->regs + REG_SW_RESET) & SW_RESET_RX))
+			break;
+		udelay(10);
+	}
+	if (limit == STOP_TRIES) {
+		netdev_err(dev, "RX reset command will not execute, resetting whole chip\n");
+		return 1;
+	}
+
+	/* reset driver rx state */
+	cas_clean_rxds(cp);
+	cas_clean_rxcs(cp);
+
+	/* Now, reprogram the rest of RX unit. */
+	cas_init_rx_dma(cp);
+
+	/* re-enable */
+	val = readl(cp->regs + REG_RX_CFG);
+	writel(val | RX_CFG_DMA_EN, cp->regs + REG_RX_CFG);
+	writel(MAC_RX_FRAME_RECV, cp->regs + REG_MAC_RX_MASK);
+	val = readl(cp->regs + REG_MAC_RX_CFG);
+	writel(val | MAC_RX_CFG_EN, cp->regs + REG_MAC_RX_CFG);
+	return 0;
+}
+#endif
+
+static int cas_rxmac_interrupt(struct net_device *dev, struct cas *cp,
+			       u32 status)
+{
+	u32 stat = readl(cp->regs + REG_MAC_RX_STATUS);
+
+	if (!stat)
+		return 0;
+
+	netif_dbg(cp, intr, cp->dev, "rxmac interrupt, stat: 0x%x\n", stat);
+
+	/* these are all rollovers */
+	spin_lock(&cp->stat_lock[0]);
+	if (stat & MAC_RX_ALIGN_ERR)
+		cp->net_stats[0].rx_frame_errors += 0x10000;
+
+	if (stat & MAC_RX_CRC_ERR)
+		cp->net_stats[0].rx_crc_errors += 0x10000;
+
+	if (stat & MAC_RX_LEN_ERR)
+		cp->net_stats[0].rx_length_errors += 0x10000;
+
+	if (stat & MAC_RX_OVERFLOW) {
+		cp->net_stats[0].rx_over_errors++;
+		cp->net_stats[0].rx_fifo_errors++;
+	}
+
+	/* We do not track MAC_RX_FRAME_COUNT and MAC_RX_VIOL_ERR
+	 * events.
+	 */
+	spin_unlock(&cp->stat_lock[0]);
+	return 0;
+}
+
+static int cas_mac_interrupt(struct net_device *dev, struct cas *cp,
+			     u32 status)
+{
+	u32 stat = readl(cp->regs + REG_MAC_CTRL_STATUS);
+
+	if (!stat)
+		return 0;
+
+	netif_printk(cp, intr, KERN_DEBUG, cp->dev,
+		     "mac interrupt, stat: 0x%x\n", stat);
+
+	/* This interrupt is just for pause frame and pause
+	 * tracking.  It is useful for diagnostics and debug
+	 * but probably by default we will mask these events.
+	 */
+	if (stat & MAC_CTRL_PAUSE_STATE)
+		cp->pause_entered++;
+
+	if (stat & MAC_CTRL_PAUSE_RECEIVED)
+		cp->pause_last_time_recvd = (stat >> 16);
+
+	return 0;
+}
+
+
+/* Must be invoked under cp->lock. */
+static inline int cas_mdio_link_not_up(struct cas *cp)
+{
+	u16 val;
+
+	switch (cp->lstate) {
+	case link_force_ret:
+		netif_info(cp, link, cp->dev, "Autoneg failed again, keeping forced mode\n");
+		cas_phy_write(cp, MII_BMCR, cp->link_fcntl);
+		cp->timer_ticks = 5;
+		cp->lstate = link_force_ok;
+		cp->link_transition = LINK_TRANSITION_LINK_CONFIG;
+		break;
+
+	case link_aneg:
+		val = cas_phy_read(cp, MII_BMCR);
+
+		/* Try forced modes. we try things in the following order:
+		 * 1000 full -> 100 full/half -> 10 half
+		 */
+		val &= ~(BMCR_ANRESTART | BMCR_ANENABLE);
+		val |= BMCR_FULLDPLX;
+		val |= (cp->cas_flags & CAS_FLAG_1000MB_CAP) ?
+			CAS_BMCR_SPEED1000 : BMCR_SPEED100;
+		cas_phy_write(cp, MII_BMCR, val);
+		cp->timer_ticks = 5;
+		cp->lstate = link_force_try;
+		cp->link_transition = LINK_TRANSITION_LINK_CONFIG;
+		break;
+
+	case link_force_try:
+		/* Downgrade from 1000 to 100 to 10 Mbps if necessary. */
+		val = cas_phy_read(cp, MII_BMCR);
+		cp->timer_ticks = 5;
+		if (val & CAS_BMCR_SPEED1000) { /* gigabit */
+			val &= ~CAS_BMCR_SPEED1000;
+			val |= (BMCR_SPEED100 | BMCR_FULLDPLX);
+			cas_phy_write(cp, MII_BMCR, val);
+			break;
+		}
+
+		if (val & BMCR_SPEED100) {
+			if (val & BMCR_FULLDPLX) /* fd failed */
+				val &= ~BMCR_FULLDPLX;
+			else { /* 100Mbps failed */
+				val &= ~BMCR_SPEED100;
+			}
+			cas_phy_write(cp, MII_BMCR, val);
+			break;
+		}
+		break;
+	default:
+		break;
+	}
+	return 0;
+}
+
+
+/* must be invoked with cp->lock held */
+static int cas_mii_link_check(struct cas *cp, const u16 bmsr)
+{
+	int restart;
+
+	if (bmsr & BMSR_LSTATUS) {
+		/* Ok, here we got a link. If we had it due to a forced
+		 * fallback, and we were configured for autoneg, we
+		 * retry a short autoneg pass. If you know your hub is
+		 * broken, use ethtool ;)
+		 */
+		if ((cp->lstate == link_force_try) &&
+		    (cp->link_cntl & BMCR_ANENABLE)) {
+			cp->lstate = link_force_ret;
+			cp->link_transition = LINK_TRANSITION_LINK_CONFIG;
+			cas_mif_poll(cp, 0);
+			cp->link_fcntl = cas_phy_read(cp, MII_BMCR);
+			cp->timer_ticks = 5;
+			if (cp->opened)
+				netif_info(cp, link, cp->dev,
+					   "Got link after fallback, retrying autoneg once...\n");
+			cas_phy_write(cp, MII_BMCR,
+				      cp->link_fcntl | BMCR_ANENABLE |
+				      BMCR_ANRESTART);
+			cas_mif_poll(cp, 1);
+
+		} else if (cp->lstate != link_up) {
+			cp->lstate = link_up;
+			cp->link_transition = LINK_TRANSITION_LINK_UP;
+
+			if (cp->opened) {
+				cas_set_link_modes(cp);
+				netif_carrier_on(cp->dev);
+			}
+		}
+		return 0;
+	}
+
+	/* link not up. if the link was previously up, we restart the
+	 * whole process
+	 */
+	restart = 0;
+	if (cp->lstate == link_up) {
+		cp->lstate = link_down;
+		cp->link_transition = LINK_TRANSITION_LINK_DOWN;
+
+		netif_carrier_off(cp->dev);
+		if (cp->opened)
+			netif_info(cp, link, cp->dev, "Link down\n");
+		restart = 1;
+
+	} else if (++cp->timer_ticks > 10)
+		cas_mdio_link_not_up(cp);
+
+	return restart;
+}
+
+static int cas_mif_interrupt(struct net_device *dev, struct cas *cp,
+			     u32 status)
+{
+	u32 stat = readl(cp->regs + REG_MIF_STATUS);
+	u16 bmsr;
+
+	/* check for a link change */
+	if (CAS_VAL(MIF_STATUS_POLL_STATUS, stat) == 0)
+		return 0;
+
+	bmsr = CAS_VAL(MIF_STATUS_POLL_DATA, stat);
+	return cas_mii_link_check(cp, bmsr);
+}
+
+static int cas_pci_interrupt(struct net_device *dev, struct cas *cp,
+			     u32 status)
+{
+	u32 stat = readl(cp->regs + REG_PCI_ERR_STATUS);
+
+	if (!stat)
+		return 0;
+
+	netdev_err(dev, "PCI error [%04x:%04x]",
+		   stat, readl(cp->regs + REG_BIM_DIAG));
+
+	/* cassini+ has this reserved */
+	if ((stat & PCI_ERR_BADACK) &&
+	    ((cp->cas_flags & CAS_FLAG_REG_PLUS) == 0))
+		pr_cont(" <No ACK64# during ABS64 cycle>");
+
+	if (stat & PCI_ERR_DTRTO)
+		pr_cont(" <Delayed transaction timeout>");
+	if (stat & PCI_ERR_OTHER)
+		pr_cont(" <other>");
+	if (stat & PCI_ERR_BIM_DMA_WRITE)
+		pr_cont(" <BIM DMA 0 write req>");
+	if (stat & PCI_ERR_BIM_DMA_READ)
+		pr_cont(" <BIM DMA 0 read req>");
+	pr_cont("\n");
+
+	if (stat & PCI_ERR_OTHER) {
+		int pci_errs;
+
+		/* Interrogate PCI config space for the
+		 * true cause.
+		 */
+		pci_errs = pci_status_get_and_clear_errors(cp->pdev);
+
+		netdev_err(dev, "PCI status errors[%04x]\n", pci_errs);
+		if (pci_errs & PCI_STATUS_PARITY)
+			netdev_err(dev, "PCI parity error detected\n");
+		if (pci_errs & PCI_STATUS_SIG_TARGET_ABORT)
+			netdev_err(dev, "PCI target abort\n");
+		if (pci_errs & PCI_STATUS_REC_TARGET_ABORT)
+			netdev_err(dev, "PCI master acks target abort\n");
+		if (pci_errs & PCI_STATUS_REC_MASTER_ABORT)
+			netdev_err(dev, "PCI master abort\n");
+		if (pci_errs & PCI_STATUS_SIG_SYSTEM_ERROR)
+			netdev_err(dev, "PCI system error SERR#\n");
+		if (pci_errs & PCI_STATUS_DETECTED_PARITY)
+			netdev_err(dev, "PCI parity error\n");
+	}
+
+	/* For all PCI errors, we should reset the chip. */
+	return 1;
+}
+
+/* All non-normal interrupt conditions get serviced here.
+ * Returns non-zero if we should just exit the interrupt
+ * handler right now (ie. if we reset the card which invalidates
+ * all of the other original irq status bits).
+ */
+static int cas_abnormal_irq(struct net_device *dev, struct cas *cp,
+			    u32 status)
+{
+	if (status & INTR_RX_TAG_ERROR) {
+		/* corrupt RX tag framing */
+		netif_printk(cp, rx_err, KERN_DEBUG, cp->dev,
+			     "corrupt rx tag framing\n");
+		spin_lock(&cp->stat_lock[0]);
+		cp->net_stats[0].rx_errors++;
+		spin_unlock(&cp->stat_lock[0]);
+		goto do_reset;
+	}
+
+	if (status & INTR_RX_LEN_MISMATCH) {
+		/* length mismatch. */
+		netif_printk(cp, rx_err, KERN_DEBUG, cp->dev,
+			     "length mismatch for rx frame\n");
+		spin_lock(&cp->stat_lock[0]);
+		cp->net_stats[0].rx_errors++;
+		spin_unlock(&cp->stat_lock[0]);
+		goto do_reset;
+	}
+
+	if (status & INTR_PCS_STATUS) {
+		if (cas_pcs_interrupt(dev, cp, status))
+			goto do_reset;
+	}
+
+	if (status & INTR_TX_MAC_STATUS) {
+		if (cas_txmac_interrupt(dev, cp, status))
+			goto do_reset;
+	}
+
+	if (status & INTR_RX_MAC_STATUS) {
+		if (cas_rxmac_interrupt(dev, cp, status))
+			goto do_reset;
+	}
+
+	if (status & INTR_MAC_CTRL_STATUS) {
+		if (cas_mac_interrupt(dev, cp, status))
+			goto do_reset;
+	}
+
+	if (status & INTR_MIF_STATUS) {
+		if (cas_mif_interrupt(dev, cp, status))
+			goto do_reset;
+	}
+
+	if (status & INTR_PCI_ERROR_STATUS) {
+		if (cas_pci_interrupt(dev, cp, status))
+			goto do_reset;
+	}
+	return 0;
+
+do_reset:
+#if 1
+	atomic_inc(&cp->reset_task_pending);
+	atomic_inc(&cp->reset_task_pending_all);
+	netdev_err(dev, "reset called in cas_abnormal_irq [0x%x]\n", status);
+	schedule_work(&cp->reset_task);
+#else
+	atomic_set(&cp->reset_task_pending, CAS_RESET_ALL);
+	netdev_err(dev, "reset called in cas_abnormal_irq\n");
+	schedule_work(&cp->reset_task);
+#endif
+	return 1;
+}
+
+/* NOTE: CAS_TABORT returns 1 or 2 so that it can be used when
+ *       determining whether to do a netif_stop/wakeup
+ */
+#define CAS_TABORT(x)      (((x)->cas_flags & CAS_FLAG_TARGET_ABORT) ? 2 : 1)
+#define CAS_ROUND_PAGE(x)  (((x) + PAGE_SIZE - 1) & PAGE_MASK)
+static inline int cas_calc_tabort(struct cas *cp, const unsigned long addr,
+				  const int len)
+{
+	unsigned long off = addr + len;
+
+	if (CAS_TABORT(cp) == 1)
+		return 0;
+	if ((CAS_ROUND_PAGE(off) - off) > TX_TARGET_ABORT_LEN)
+		return 0;
+	return TX_TARGET_ABORT_LEN;
+}
+
+static inline void cas_tx_ringN(struct cas *cp, int ring, int limit)
+{
+	struct cas_tx_desc *txds;
+	struct sk_buff **skbs;
+	struct net_device *dev = cp->dev;
+	int entry, count;
+
+	spin_lock(&cp->tx_lock[ring]);
+	txds = cp->init_txds[ring];
+	skbs = cp->tx_skbs[ring];
+	entry = cp->tx_old[ring];
+
+	count = TX_BUFF_COUNT(ring, entry, limit);
+	while (entry != limit) {
+		struct sk_buff *skb = skbs[entry];
+		dma_addr_t daddr;
+		u32 dlen;
+		int frag;
+
+		if (!skb) {
+			/* this should never occur */
+			entry = TX_DESC_NEXT(ring, entry);
+			continue;
+		}
+
+		/* however, we might get only a partial skb release. */
+		count -= skb_shinfo(skb)->nr_frags +
+			+ cp->tx_tiny_use[ring][entry].nbufs + 1;
+		if (count < 0)
+			break;
+
+		netif_printk(cp, tx_done, KERN_DEBUG, cp->dev,
+			     "tx[%d] done, slot %d\n", ring, entry);
+
+		skbs[entry] = NULL;
+		cp->tx_tiny_use[ring][entry].nbufs = 0;
+
+		for (frag = 0; frag <= skb_shinfo(skb)->nr_frags; frag++) {
+			struct cas_tx_desc *txd = txds + entry;
+
+			daddr = le64_to_cpu(txd->buffer);
+			dlen = CAS_VAL(TX_DESC_BUFLEN,
+				       le64_to_cpu(txd->control));
+			dma_unmap_page(&cp->pdev->dev, daddr, dlen,
+				       DMA_TO_DEVICE);
+			entry = TX_DESC_NEXT(ring, entry);
+
+			/* tiny buffer may follow */
+			if (cp->tx_tiny_use[ring][entry].used) {
+				cp->tx_tiny_use[ring][entry].used = 0;
+				entry = TX_DESC_NEXT(ring, entry);
+			}
+		}
+
+		spin_lock(&cp->stat_lock[ring]);
+		cp->net_stats[ring].tx_packets++;
+		cp->net_stats[ring].tx_bytes += skb->len;
+		spin_unlock(&cp->stat_lock[ring]);
+		dev_consume_skb_irq(skb);
+	}
+	cp->tx_old[ring] = entry;
+
+	/* this is wrong for multiple tx rings. the net device needs
+	 * multiple queues for this to do the right thing.  we wait
+	 * for 2*packets to be available when using tiny buffers
+	 */
+	if (netif_queue_stopped(dev) &&
+	    (TX_BUFFS_AVAIL(cp, ring) > CAS_TABORT(cp)*(MAX_SKB_FRAGS + 1)))
+		netif_wake_queue(dev);
+	spin_unlock(&cp->tx_lock[ring]);
+}
+
+static void cas_tx(struct net_device *dev, struct cas *cp,
+		   u32 status)
+{
+        int limit, ring;
+#ifdef USE_TX_COMPWB
+	u64 compwb = le64_to_cpu(cp->init_block->tx_compwb);
+#endif
+	netif_printk(cp, intr, KERN_DEBUG, cp->dev,
+		     "tx interrupt, status: 0x%x, %llx\n",
+		     status, (unsigned long long)compwb);
+	/* process all the rings */
+	for (ring = 0; ring < N_TX_RINGS; ring++) {
+#ifdef USE_TX_COMPWB
+		/* use the completion writeback registers */
+		limit = (CAS_VAL(TX_COMPWB_MSB, compwb) << 8) |
+			CAS_VAL(TX_COMPWB_LSB, compwb);
+		compwb = TX_COMPWB_NEXT(compwb);
+#else
+		limit = readl(cp->regs + REG_TX_COMPN(ring));
+#endif
+		if (cp->tx_old[ring] != limit)
+			cas_tx_ringN(cp, ring, limit);
+	}
+}
+
+
+static int cas_rx_process_pkt(struct cas *cp, struct cas_rx_comp *rxc,
+			      int entry, const u64 *words,
+			      struct sk_buff **skbref)
+{
+	int dlen, hlen, len, i, alloclen;
+	int off, swivel = RX_SWIVEL_OFF_VAL;
+	struct cas_page *page;
+	struct sk_buff *skb;
+	void *crcaddr;
+	__sum16 csum;
+	char *p;
+
+	hlen = CAS_VAL(RX_COMP2_HDR_SIZE, words[1]);
+	dlen = CAS_VAL(RX_COMP1_DATA_SIZE, words[0]);
+	len  = hlen + dlen;
+
+	if (RX_COPY_ALWAYS || (words[2] & RX_COMP3_SMALL_PKT))
+		alloclen = len;
+	else
+		alloclen = max(hlen, RX_COPY_MIN);
+
+	skb = netdev_alloc_skb(cp->dev, alloclen + swivel + cp->crc_size);
+	if (skb == NULL)
+		return -1;
+
+	*skbref = skb;
+	skb_reserve(skb, swivel);
+
+	p = skb->data;
+	crcaddr = NULL;
+	if (hlen) { /* always copy header pages */
+		i = CAS_VAL(RX_COMP2_HDR_INDEX, words[1]);
+		page = cp->rx_pages[CAS_VAL(RX_INDEX_RING, i)][CAS_VAL(RX_INDEX_NUM, i)];
+		off = CAS_VAL(RX_COMP2_HDR_OFF, words[1]) * 0x100 +
+			swivel;
+
+		i = hlen;
+		if (!dlen) /* attach FCS */
+			i += cp->crc_size;
+		dma_sync_single_for_cpu(&cp->pdev->dev, page->dma_addr + off,
+					i, DMA_FROM_DEVICE);
+		memcpy(p, page_address(page->buffer) + off, i);
+		dma_sync_single_for_device(&cp->pdev->dev,
+					   page->dma_addr + off, i,
+					   DMA_FROM_DEVICE);
+		RX_USED_ADD(page, 0x100);
+		p += hlen;
+		swivel = 0;
+	}
+
+
+	if (alloclen < (hlen + dlen)) {
+		skb_frag_t *frag = skb_shinfo(skb)->frags;
+
+		/* normal or jumbo packets. we use frags */
+		i = CAS_VAL(RX_COMP1_DATA_INDEX, words[0]);
+		page = cp->rx_pages[CAS_VAL(RX_INDEX_RING, i)][CAS_VAL(RX_INDEX_NUM, i)];
+		off = CAS_VAL(RX_COMP1_DATA_OFF, words[0]) + swivel;
+
+		hlen = min(cp->page_size - off, dlen);
+		if (hlen < 0) {
+			netif_printk(cp, rx_err, KERN_DEBUG, cp->dev,
+				     "rx page overflow: %d\n", hlen);
+			dev_kfree_skb_irq(skb);
+			return -1;
+		}
+		i = hlen;
+		if (i == dlen)  /* attach FCS */
+			i += cp->crc_size;
+		dma_sync_single_for_cpu(&cp->pdev->dev, page->dma_addr + off,
+					i, DMA_FROM_DEVICE);
+
+		/* make sure we always copy a header */
+		swivel = 0;
+		if (p == (char *) skb->data) { /* not split */
+			memcpy(p, page_address(page->buffer) + off,
+			       RX_COPY_MIN);
+			dma_sync_single_for_device(&cp->pdev->dev,
+						   page->dma_addr + off, i,
+						   DMA_FROM_DEVICE);
+			off += RX_COPY_MIN;
+			swivel = RX_COPY_MIN;
+			RX_USED_ADD(page, cp->mtu_stride);
+		} else {
+			RX_USED_ADD(page, hlen);
+		}
+		skb_put(skb, alloclen);
+
+		skb_shinfo(skb)->nr_frags++;
+		skb->data_len += hlen - swivel;
+		skb->truesize += hlen - swivel;
+		skb->len      += hlen - swivel;
+
+		skb_frag_fill_page_desc(frag, page->buffer, off, hlen - swivel);
+		__skb_frag_ref(frag);
+
+		/* any more data? */
+		if ((words[0] & RX_COMP1_SPLIT_PKT) && ((dlen -= hlen) > 0)) {
+			hlen = dlen;
+			off = 0;
+
+			i = CAS_VAL(RX_COMP2_NEXT_INDEX, words[1]);
+			page = cp->rx_pages[CAS_VAL(RX_INDEX_RING, i)][CAS_VAL(RX_INDEX_NUM, i)];
+			dma_sync_single_for_cpu(&cp->pdev->dev,
+						page->dma_addr,
+						hlen + cp->crc_size,
+						DMA_FROM_DEVICE);
+			dma_sync_single_for_device(&cp->pdev->dev,
+						   page->dma_addr,
+						   hlen + cp->crc_size,
+						   DMA_FROM_DEVICE);
+
+			skb_shinfo(skb)->nr_frags++;
+			skb->data_len += hlen;
+			skb->len      += hlen;
+			frag++;
+
+			skb_frag_fill_page_desc(frag, page->buffer, 0, hlen);
+			__skb_frag_ref(frag);
+			RX_USED_ADD(page, hlen + cp->crc_size);
+		}
+
+		if (cp->crc_size)
+			crcaddr = page_address(page->buffer) + off + hlen;
+
+	} else {
+		/* copying packet */
+		if (!dlen)
+			goto end_copy_pkt;
+
+		i = CAS_VAL(RX_COMP1_DATA_INDEX, words[0]);
+		page = cp->rx_pages[CAS_VAL(RX_INDEX_RING, i)][CAS_VAL(RX_INDEX_NUM, i)];
+		off = CAS_VAL(RX_COMP1_DATA_OFF, words[0]) + swivel;
+		hlen = min(cp->page_size - off, dlen);
+		if (hlen < 0) {
+			netif_printk(cp, rx_err, KERN_DEBUG, cp->dev,
+				     "rx page overflow: %d\n", hlen);
+			dev_kfree_skb_irq(skb);
+			return -1;
+		}
+		i = hlen;
+		if (i == dlen) /* attach FCS */
+			i += cp->crc_size;
+		dma_sync_single_for_cpu(&cp->pdev->dev, page->dma_addr + off,
+					i, DMA_FROM_DEVICE);
+		memcpy(p, page_address(page->buffer) + off, i);
+		dma_sync_single_for_device(&cp->pdev->dev,
+					   page->dma_addr + off, i,
+					   DMA_FROM_DEVICE);
+		if (p == (char *) skb->data) /* not split */
+			RX_USED_ADD(page, cp->mtu_stride);
+		else
+			RX_USED_ADD(page, i);
+
+		/* any more data? */
+		if ((words[0] & RX_COMP1_SPLIT_PKT) && ((dlen -= hlen) > 0)) {
+			p += hlen;
+			i = CAS_VAL(RX_COMP2_NEXT_INDEX, words[1]);
+			page = cp->rx_pages[CAS_VAL(RX_INDEX_RING, i)][CAS_VAL(RX_INDEX_NUM, i)];
+			dma_sync_single_for_cpu(&cp->pdev->dev,
+						page->dma_addr,
+						dlen + cp->crc_size,
+						DMA_FROM_DEVICE);
+			memcpy(p, page_address(page->buffer), dlen + cp->crc_size);
+			dma_sync_single_for_device(&cp->pdev->dev,
+						   page->dma_addr,
+						   dlen + cp->crc_size,
+						   DMA_FROM_DEVICE);
+			RX_USED_ADD(page, dlen + cp->crc_size);
+		}
+end_copy_pkt:
+		if (cp->crc_size)
+			crcaddr = skb->data + alloclen;
+
+		skb_put(skb, alloclen);
+	}
+
+	csum = (__force __sum16)htons(CAS_VAL(RX_COMP4_TCP_CSUM, words[3]));
+	if (cp->crc_size) {
+		/* checksum includes FCS. strip it out. */
+		csum = csum_fold(csum_partial(crcaddr, cp->crc_size,
+					      csum_unfold(csum)));
+	}
+	skb->protocol = eth_type_trans(skb, cp->dev);
+	if (skb->protocol == htons(ETH_P_IP)) {
+		skb->csum = csum_unfold(~csum);
+		skb->ip_summed = CHECKSUM_COMPLETE;
+	} else
+		skb_checksum_none_assert(skb);
+	return len;
+}
+
+
+/* we can handle up to 64 rx flows at a time. we do the same thing
+ * as nonreassm except that we batch up the buffers.
+ * NOTE: we currently just treat each flow as a bunch of packets that
+ *       we pass up. a better way would be to coalesce the packets
+ *       into a jumbo packet. to do that, we need to do the following:
+ *       1) the first packet will have a clean split between header and
+ *          data. save both.
+ *       2) each time the next flow packet comes in, extend the
+ *          data length and merge the checksums.
+ *       3) on flow release, fix up the header.
+ *       4) make sure the higher layer doesn't care.
+ * because packets get coalesced, we shouldn't run into fragment count
+ * issues.
+ */
+static inline void cas_rx_flow_pkt(struct cas *cp, const u64 *words,
+				   struct sk_buff *skb)
+{
+	int flowid = CAS_VAL(RX_COMP3_FLOWID, words[2]) & (N_RX_FLOWS - 1);
+	struct sk_buff_head *flow = &cp->rx_flows[flowid];
+
+	/* this is protected at a higher layer, so no need to
+	 * do any additional locking here. stick the buffer
+	 * at the end.
+	 */
+	__skb_queue_tail(flow, skb);
+	if (words[0] & RX_COMP1_RELEASE_FLOW) {
+		while ((skb = __skb_dequeue(flow))) {
+			cas_skb_release(skb);
+		}
+	}
+}
+
+/* put rx descriptor back on ring. if a buffer is in use by a higher
+ * layer, this will need to put in a replacement.
+ */
+static void cas_post_page(struct cas *cp, const int ring, const int index)
+{
+	cas_page_t *new;
+	int entry;
+
+	entry = cp->rx_old[ring];
+
+	new = cas_page_swap(cp, ring, index);
+	cp->init_rxds[ring][entry].buffer = cpu_to_le64(new->dma_addr);
+	cp->init_rxds[ring][entry].index  =
+		cpu_to_le64(CAS_BASE(RX_INDEX_NUM, index) |
+			    CAS_BASE(RX_INDEX_RING, ring));
+
+	entry = RX_DESC_ENTRY(ring, entry + 1);
+	cp->rx_old[ring] = entry;
+
+	if (entry % 4)
+		return;
+
+	if (ring == 0)
+		writel(entry, cp->regs + REG_RX_KICK);
+	else if ((N_RX_DESC_RINGS > 1) &&
+		 (cp->cas_flags & CAS_FLAG_REG_PLUS))
+		writel(entry, cp->regs + REG_PLUS_RX_KICK1);
+}
+
+
+/* only when things are bad */
+static int cas_post_rxds_ringN(struct cas *cp, int ring, int num)
+{
+	unsigned int entry, last, count, released;
+	int cluster;
+	cas_page_t **page = cp->rx_pages[ring];
+
+	entry = cp->rx_old[ring];
+
+	netif_printk(cp, intr, KERN_DEBUG, cp->dev,
+		     "rxd[%d] interrupt, done: %d\n", ring, entry);
+
+	cluster = -1;
+	count = entry & 0x3;
+	last = RX_DESC_ENTRY(ring, num ? entry + num - 4: entry - 4);
+	released = 0;
+	while (entry != last) {
+		/* make a new buffer if it's still in use */
+		if (page_count(page[entry]->buffer) > 1) {
+			cas_page_t *new = cas_page_dequeue(cp);
+			if (!new) {
+				/* let the timer know that we need to
+				 * do this again
+				 */
+				cp->cas_flags |= CAS_FLAG_RXD_POST(ring);
+				if (!timer_pending(&cp->link_timer))
+					mod_timer(&cp->link_timer, jiffies +
+						  CAS_LINK_FAST_TIMEOUT);
+				cp->rx_old[ring]  = entry;
+				cp->rx_last[ring] = num ? num - released : 0;
+				return -ENOMEM;
+			}
+			spin_lock(&cp->rx_inuse_lock);
+			list_add(&page[entry]->list, &cp->rx_inuse_list);
+			spin_unlock(&cp->rx_inuse_lock);
+			cp->init_rxds[ring][entry].buffer =
+				cpu_to_le64(new->dma_addr);
+			page[entry] = new;
+
+		}
+
+		if (++count == 4) {
+			cluster = entry;
+			count = 0;
+		}
+		released++;
+		entry = RX_DESC_ENTRY(ring, entry + 1);
+	}
+	cp->rx_old[ring] = entry;
+
+	if (cluster < 0)
+		return 0;
+
+	if (ring == 0)
+		writel(cluster, cp->regs + REG_RX_KICK);
+	else if ((N_RX_DESC_RINGS > 1) &&
+		 (cp->cas_flags & CAS_FLAG_REG_PLUS))
+		writel(cluster, cp->regs + REG_PLUS_RX_KICK1);
+	return 0;
+}
+
+
+/* process a completion ring. packets are set up in three basic ways:
+ * small packets: should be copied header + data in single buffer.
+ * large packets: header and data in a single buffer.
+ * split packets: header in a separate buffer from data.
+ *                data may be in multiple pages. data may be > 256
+ *                bytes but in a single page.
+ *
+ * NOTE: RX page posting is done in this routine as well. while there's
+ *       the capability of using multiple RX completion rings, it isn't
+ *       really worthwhile due to the fact that the page posting will
+ *       force serialization on the single descriptor ring.
+ */
+static int cas_rx_ringN(struct cas *cp, int ring, int budget)
+{
+	struct cas_rx_comp *rxcs = cp->init_rxcs[ring];
+	int entry, drops;
+	int npackets = 0;
+
+	netif_printk(cp, intr, KERN_DEBUG, cp->dev,
+		     "rx[%d] interrupt, done: %d/%d\n",
+		     ring,
+		     readl(cp->regs + REG_RX_COMP_HEAD), cp->rx_new[ring]);
+
+	entry = cp->rx_new[ring];
+	drops = 0;
+	while (1) {
+		struct cas_rx_comp *rxc = rxcs + entry;
+		struct sk_buff *skb;
+		int type, len;
+		u64 words[4];
+		int i, dring;
+
+		words[0] = le64_to_cpu(rxc->word1);
+		words[1] = le64_to_cpu(rxc->word2);
+		words[2] = le64_to_cpu(rxc->word3);
+		words[3] = le64_to_cpu(rxc->word4);
+
+		/* don't touch if still owned by hw */
+		type = CAS_VAL(RX_COMP1_TYPE, words[0]);
+		if (type == 0)
+			break;
+
+		/* hw hasn't cleared the zero bit yet */
+		if (words[3] & RX_COMP4_ZERO) {
+			break;
+		}
+
+		/* get info on the packet */
+		if (words[3] & (RX_COMP4_LEN_MISMATCH | RX_COMP4_BAD)) {
+			spin_lock(&cp->stat_lock[ring]);
+			cp->net_stats[ring].rx_errors++;
+			if (words[3] & RX_COMP4_LEN_MISMATCH)
+				cp->net_stats[ring].rx_length_errors++;
+			if (words[3] & RX_COMP4_BAD)
+				cp->net_stats[ring].rx_crc_errors++;
+			spin_unlock(&cp->stat_lock[ring]);
+
+			/* We'll just return it to Cassini. */
+		drop_it:
+			spin_lock(&cp->stat_lock[ring]);
+			++cp->net_stats[ring].rx_dropped;
+			spin_unlock(&cp->stat_lock[ring]);
+			goto next;
+		}
+
+		len = cas_rx_process_pkt(cp, rxc, entry, words, &skb);
+		if (len < 0) {
+			++drops;
+			goto drop_it;
+		}
+
+		/* see if it's a flow re-assembly or not. the driver
+		 * itself handles release back up.
+		 */
+		if (RX_DONT_BATCH || (type == 0x2)) {
+			/* non-reassm: these always get released */
+			cas_skb_release(skb);
+		} else {
+			cas_rx_flow_pkt(cp, words, skb);
+		}
+
+		spin_lock(&cp->stat_lock[ring]);
+		cp->net_stats[ring].rx_packets++;
+		cp->net_stats[ring].rx_bytes += len;
+		spin_unlock(&cp->stat_lock[ring]);
+
+	next:
+		npackets++;
+
+		/* should it be released? */
+		if (words[0] & RX_COMP1_RELEASE_HDR) {
+			i = CAS_VAL(RX_COMP2_HDR_INDEX, words[1]);
+			dring = CAS_VAL(RX_INDEX_RING, i);
+			i = CAS_VAL(RX_INDEX_NUM, i);
+			cas_post_page(cp, dring, i);
+		}
+
+		if (words[0] & RX_COMP1_RELEASE_DATA) {
+			i = CAS_VAL(RX_COMP1_DATA_INDEX, words[0]);
+			dring = CAS_VAL(RX_INDEX_RING, i);
+			i = CAS_VAL(RX_INDEX_NUM, i);
+			cas_post_page(cp, dring, i);
+		}
+
+		if (words[0] & RX_COMP1_RELEASE_NEXT) {
+			i = CAS_VAL(RX_COMP2_NEXT_INDEX, words[1]);
+			dring = CAS_VAL(RX_INDEX_RING, i);
+			i = CAS_VAL(RX_INDEX_NUM, i);
+			cas_post_page(cp, dring, i);
+		}
+
+		/* skip to the next entry */
+		entry = RX_COMP_ENTRY(ring, entry + 1 +
+				      CAS_VAL(RX_COMP1_SKIP, words[0]));
+#ifdef USE_NAPI
+		if (budget && (npackets >= budget))
+			break;
+#endif
+	}
+	cp->rx_new[ring] = entry;
+
+	if (drops)
+		netdev_info(cp->dev, "Memory squeeze, deferring packet\n");
+	return npackets;
+}
+
+
+/* put completion entries back on the ring */
+static void cas_post_rxcs_ringN(struct net_device *dev,
+				struct cas *cp, int ring)
+{
+	struct cas_rx_comp *rxc = cp->init_rxcs[ring];
+	int last, entry;
+
+	last = cp->rx_cur[ring];
+	entry = cp->rx_new[ring];
+	netif_printk(cp, intr, KERN_DEBUG, dev,
+		     "rxc[%d] interrupt, done: %d/%d\n",
+		     ring, readl(cp->regs + REG_RX_COMP_HEAD), entry);
+
+	/* zero and re-mark descriptors */
+	while (last != entry) {
+		cas_rxc_init(rxc + last);
+		last = RX_COMP_ENTRY(ring, last + 1);
+	}
+	cp->rx_cur[ring] = last;
+
+	if (ring == 0)
+		writel(last, cp->regs + REG_RX_COMP_TAIL);
+	else if (cp->cas_flags & CAS_FLAG_REG_PLUS)
+		writel(last, cp->regs + REG_PLUS_RX_COMPN_TAIL(ring));
+}
+
+
+
+/* cassini can use all four PCI interrupts for the completion ring.
+ * rings 3 and 4 are identical
+ */
+#if defined(USE_PCI_INTC) || defined(USE_PCI_INTD)
+static inline void cas_handle_irqN(struct net_device *dev,
+				   struct cas *cp, const u32 status,
+				   const int ring)
+{
+	if (status & (INTR_RX_COMP_FULL_ALT | INTR_RX_COMP_AF_ALT))
+		cas_post_rxcs_ringN(dev, cp, ring);
+}
+
+static irqreturn_t cas_interruptN(int irq, void *dev_id)
+{
+	struct net_device *dev = dev_id;
+	struct cas *cp = netdev_priv(dev);
+	unsigned long flags;
+	int ring = (irq == cp->pci_irq_INTC) ? 2 : 3;
+	u32 status = readl(cp->regs + REG_PLUS_INTRN_STATUS(ring));
+
+	/* check for shared irq */
+	if (status == 0)
+		return IRQ_NONE;
+
+	spin_lock_irqsave(&cp->lock, flags);
+	if (status & INTR_RX_DONE_ALT) { /* handle rx separately */
+#ifdef USE_NAPI
+		cas_mask_intr(cp);
+		napi_schedule(&cp->napi);
+#else
+		cas_rx_ringN(cp, ring, 0);
+#endif
+		status &= ~INTR_RX_DONE_ALT;
+	}
+
+	if (status)
+		cas_handle_irqN(dev, cp, status, ring);
+	spin_unlock_irqrestore(&cp->lock, flags);
+	return IRQ_HANDLED;
+}
+#endif
+
+#ifdef USE_PCI_INTB
+/* everything but rx packets */
+static inline void cas_handle_irq1(struct cas *cp, const u32 status)
+{
+	if (status & INTR_RX_BUF_UNAVAIL_1) {
+		/* Frame arrived, no free RX buffers available.
+		 * NOTE: we can get this on a link transition. */
+		cas_post_rxds_ringN(cp, 1, 0);
+		spin_lock(&cp->stat_lock[1]);
+		cp->net_stats[1].rx_dropped++;
+		spin_unlock(&cp->stat_lock[1]);
+	}
+
+	if (status & INTR_RX_BUF_AE_1)
+		cas_post_rxds_ringN(cp, 1, RX_DESC_RINGN_SIZE(1) -
+				    RX_AE_FREEN_VAL(1));
+
+	if (status & (INTR_RX_COMP_AF | INTR_RX_COMP_FULL))
+		cas_post_rxcs_ringN(cp, 1);
+}
+
+/* ring 2 handles a few more events than 3 and 4 */
+static irqreturn_t cas_interrupt1(int irq, void *dev_id)
+{
+	struct net_device *dev = dev_id;
+	struct cas *cp = netdev_priv(dev);
+	unsigned long flags;
+	u32 status = readl(cp->regs + REG_PLUS_INTRN_STATUS(1));
+
+	/* check for shared interrupt */
+	if (status == 0)
+		return IRQ_NONE;
+
+	spin_lock_irqsave(&cp->lock, flags);
+	if (status & INTR_RX_DONE_ALT) { /* handle rx separately */
+#ifdef USE_NAPI
+		cas_mask_intr(cp);
+		napi_schedule(&cp->napi);
+#else
+		cas_rx_ringN(cp, 1, 0);
+#endif
+		status &= ~INTR_RX_DONE_ALT;
+	}
+	if (status)
+		cas_handle_irq1(cp, status);
+	spin_unlock_irqrestore(&cp->lock, flags);
+	return IRQ_HANDLED;
+}
+#endif
+
+static inline void cas_handle_irq(struct net_device *dev,
+				  struct cas *cp, const u32 status)
+{
+	/* housekeeping interrupts */
+	if (status & INTR_ERROR_MASK)
+		cas_abnormal_irq(dev, cp, status);
+
+	if (status & INTR_RX_BUF_UNAVAIL) {
+		/* Frame arrived, no free RX buffers available.
+		 * NOTE: we can get this on a link transition.
+		 */
+		cas_post_rxds_ringN(cp, 0, 0);
+		spin_lock(&cp->stat_lock[0]);
+		cp->net_stats[0].rx_dropped++;
+		spin_unlock(&cp->stat_lock[0]);
+	} else if (status & INTR_RX_BUF_AE) {
+		cas_post_rxds_ringN(cp, 0, RX_DESC_RINGN_SIZE(0) -
+				    RX_AE_FREEN_VAL(0));
+	}
+
+	if (status & (INTR_RX_COMP_AF | INTR_RX_COMP_FULL))
+		cas_post_rxcs_ringN(dev, cp, 0);
+}
+
+static irqreturn_t cas_interrupt(int irq, void *dev_id)
+{
+	struct net_device *dev = dev_id;
+	struct cas *cp = netdev_priv(dev);
+	unsigned long flags;
+	u32 status = readl(cp->regs + REG_INTR_STATUS);
+
+	if (status == 0)
+		return IRQ_NONE;
+
+	spin_lock_irqsave(&cp->lock, flags);
+	if (status & (INTR_TX_ALL | INTR_TX_INTME)) {
+		cas_tx(dev, cp, status);
+		status &= ~(INTR_TX_ALL | INTR_TX_INTME);
+	}
+
+	if (status & INTR_RX_DONE) {
+#ifdef USE_NAPI
+		cas_mask_intr(cp);
+		napi_schedule(&cp->napi);
+#else
+		cas_rx_ringN(cp, 0, 0);
+#endif
+		status &= ~INTR_RX_DONE;
+	}
+
+	if (status)
+		cas_handle_irq(dev, cp, status);
+	spin_unlock_irqrestore(&cp->lock, flags);
+	return IRQ_HANDLED;
+}
+
+
+#ifdef USE_NAPI
+static int cas_poll(struct napi_struct *napi, int budget)
+{
+	struct cas *cp = container_of(napi, struct cas, napi);
+	struct net_device *dev = cp->dev;
+	int i, enable_intr, credits;
+	u32 status = readl(cp->regs + REG_INTR_STATUS);
+	unsigned long flags;
+
+	spin_lock_irqsave(&cp->lock, flags);
+	cas_tx(dev, cp, status);
+	spin_unlock_irqrestore(&cp->lock, flags);
+
+	/* NAPI rx packets. we spread the credits across all of the
+	 * rxc rings
+	 *
+	 * to make sure we're fair with the work we loop through each
+	 * ring N_RX_COMP_RING times with a request of
+	 * budget / N_RX_COMP_RINGS
+	 */
+	enable_intr = 1;
+	credits = 0;
+	for (i = 0; i < N_RX_COMP_RINGS; i++) {
+		int j;
+		for (j = 0; j < N_RX_COMP_RINGS; j++) {
+			credits += cas_rx_ringN(cp, j, budget / N_RX_COMP_RINGS);
+			if (credits >= budget) {
+				enable_intr = 0;
+				goto rx_comp;
+			}
+		}
+	}
+
+rx_comp:
+	/* final rx completion */
+	spin_lock_irqsave(&cp->lock, flags);
+	if (status)
+		cas_handle_irq(dev, cp, status);
+
+#ifdef USE_PCI_INTB
+	if (N_RX_COMP_RINGS > 1) {
+		status = readl(cp->regs + REG_PLUS_INTRN_STATUS(1));
+		if (status)
+			cas_handle_irq1(dev, cp, status);
+	}
+#endif
+
+#ifdef USE_PCI_INTC
+	if (N_RX_COMP_RINGS > 2) {
+		status = readl(cp->regs + REG_PLUS_INTRN_STATUS(2));
+		if (status)
+			cas_handle_irqN(dev, cp, status, 2);
+	}
+#endif
+
+#ifdef USE_PCI_INTD
+	if (N_RX_COMP_RINGS > 3) {
+		status = readl(cp->regs + REG_PLUS_INTRN_STATUS(3));
+		if (status)
+			cas_handle_irqN(dev, cp, status, 3);
+	}
+#endif
+	spin_unlock_irqrestore(&cp->lock, flags);
+	if (enable_intr) {
+		napi_complete(napi);
+		cas_unmask_intr(cp);
+	}
+	return credits;
+}
+#endif
+
+#ifdef CONFIG_NET_POLL_CONTROLLER
+static void cas_netpoll(struct net_device *dev)
+{
+	struct cas *cp = netdev_priv(dev);
+
+	cas_disable_irq(cp, 0);
+	cas_interrupt(cp->pdev->irq, dev);
+	cas_enable_irq(cp, 0);
+
+#ifdef USE_PCI_INTB
+	if (N_RX_COMP_RINGS > 1) {
+		/* cas_interrupt1(); */
+	}
+#endif
+#ifdef USE_PCI_INTC
+	if (N_RX_COMP_RINGS > 2) {
+		/* cas_interruptN(); */
+	}
+#endif
+#ifdef USE_PCI_INTD
+	if (N_RX_COMP_RINGS > 3) {
+		/* cas_interruptN(); */
+	}
+#endif
+}
+#endif
+
+static void cas_tx_timeout(struct net_device *dev, unsigned int txqueue)
+{
+	struct cas *cp = netdev_priv(dev);
+
+	netdev_err(dev, "transmit timed out, resetting\n");
+	if (!cp->hw_running) {
+		netdev_err(dev, "hrm.. hw not running!\n");
+		return;
+	}
+
+	netdev_err(dev, "MIF_STATE[%08x]\n",
+		   readl(cp->regs + REG_MIF_STATE_MACHINE));
+
+	netdev_err(dev, "MAC_STATE[%08x]\n",
+		   readl(cp->regs + REG_MAC_STATE_MACHINE));
+
+	netdev_err(dev, "TX_STATE[%08x:%08x:%08x] FIFO[%08x:%08x:%08x] SM1[%08x] SM2[%08x]\n",
+		   readl(cp->regs + REG_TX_CFG),
+		   readl(cp->regs + REG_MAC_TX_STATUS),
+		   readl(cp->regs + REG_MAC_TX_CFG),
+		   readl(cp->regs + REG_TX_FIFO_PKT_CNT),
+		   readl(cp->regs + REG_TX_FIFO_WRITE_PTR),
+		   readl(cp->regs + REG_TX_FIFO_READ_PTR),
+		   readl(cp->regs + REG_TX_SM_1),
+		   readl(cp->regs + REG_TX_SM_2));
+
+	netdev_err(dev, "RX_STATE[%08x:%08x:%08x]\n",
+		   readl(cp->regs + REG_RX_CFG),
+		   readl(cp->regs + REG_MAC_RX_STATUS),
+		   readl(cp->regs + REG_MAC_RX_CFG));
+
+	netdev_err(dev, "HP_STATE[%08x:%08x:%08x:%08x]\n",
+		   readl(cp->regs + REG_HP_STATE_MACHINE),
+		   readl(cp->regs + REG_HP_STATUS0),
+		   readl(cp->regs + REG_HP_STATUS1),
+		   readl(cp->regs + REG_HP_STATUS2));
+
+#if 1
+	atomic_inc(&cp->reset_task_pending);
+	atomic_inc(&cp->reset_task_pending_all);
+	schedule_work(&cp->reset_task);
+#else
+	atomic_set(&cp->reset_task_pending, CAS_RESET_ALL);
+	schedule_work(&cp->reset_task);
+#endif
+}
+
+static inline int cas_intme(int ring, int entry)
+{
+	/* Algorithm: IRQ every 1/2 of descriptors. */
+	if (!(entry & ((TX_DESC_RINGN_SIZE(ring) >> 1) - 1)))
+		return 1;
+	return 0;
+}
+
+
+static void cas_write_txd(struct cas *cp, int ring, int entry,
+			  dma_addr_t mapping, int len, u64 ctrl, int last)
+{
+	struct cas_tx_desc *txd = cp->init_txds[ring] + entry;
+
+	ctrl |= CAS_BASE(TX_DESC_BUFLEN, len);
+	if (cas_intme(ring, entry))
+		ctrl |= TX_DESC_INTME;
+	if (last)
+		ctrl |= TX_DESC_EOF;
+	txd->control = cpu_to_le64(ctrl);
+	txd->buffer = cpu_to_le64(mapping);
+}
+
+static inline void *tx_tiny_buf(struct cas *cp, const int ring,
+				const int entry)
+{
+	return cp->tx_tiny_bufs[ring] + TX_TINY_BUF_LEN*entry;
+}
+
+static inline dma_addr_t tx_tiny_map(struct cas *cp, const int ring,
+				     const int entry, const int tentry)
+{
+	cp->tx_tiny_use[ring][tentry].nbufs++;
+	cp->tx_tiny_use[ring][entry].used = 1;
+	return cp->tx_tiny_dvma[ring] + TX_TINY_BUF_LEN*entry;
+}
+
+static inline int cas_xmit_tx_ringN(struct cas *cp, int ring,
+				    struct sk_buff *skb)
+{
+	struct net_device *dev = cp->dev;
+	int entry, nr_frags, frag, tabort, tentry;
+	dma_addr_t mapping;
+	unsigned long flags;
+	u64 ctrl;
+	u32 len;
+
+	spin_lock_irqsave(&cp->tx_lock[ring], flags);
+
+	/* This is a hard error, log it. */
+	if (TX_BUFFS_AVAIL(cp, ring) <=
+	    CAS_TABORT(cp)*(skb_shinfo(skb)->nr_frags + 1)) {
+		netif_stop_queue(dev);
+		spin_unlock_irqrestore(&cp->tx_lock[ring], flags);
+		netdev_err(dev, "BUG! Tx Ring full when queue awake!\n");
+		return 1;
+	}
+
+	ctrl = 0;
+	if (skb->ip_summed == CHECKSUM_PARTIAL) {
+		const u64 csum_start_off = skb_checksum_start_offset(skb);
+		const u64 csum_stuff_off = csum_start_off + skb->csum_offset;
+
+		ctrl =  TX_DESC_CSUM_EN |
+			CAS_BASE(TX_DESC_CSUM_START, csum_start_off) |
+			CAS_BASE(TX_DESC_CSUM_STUFF, csum_stuff_off);
+	}
+
+	entry = cp->tx_new[ring];
+	cp->tx_skbs[ring][entry] = skb;
+
+	nr_frags = skb_shinfo(skb)->nr_frags;
+	len = skb_headlen(skb);
+	mapping = dma_map_page(&cp->pdev->dev, virt_to_page(skb->data),
+			       offset_in_page(skb->data), len, DMA_TO_DEVICE);
+
+	tentry = entry;
+	tabort = cas_calc_tabort(cp, (unsigned long) skb->data, len);
+	if (unlikely(tabort)) {
+		/* NOTE: len is always >  tabort */
+		cas_write_txd(cp, ring, entry, mapping, len - tabort,
+			      ctrl | TX_DESC_SOF, 0);
+		entry = TX_DESC_NEXT(ring, entry);
+
+		skb_copy_from_linear_data_offset(skb, len - tabort,
+			      tx_tiny_buf(cp, ring, entry), tabort);
+		mapping = tx_tiny_map(cp, ring, entry, tentry);
+		cas_write_txd(cp, ring, entry, mapping, tabort, ctrl,
+			      (nr_frags == 0));
+	} else {
+		cas_write_txd(cp, ring, entry, mapping, len, ctrl |
+			      TX_DESC_SOF, (nr_frags == 0));
+	}
+	entry = TX_DESC_NEXT(ring, entry);
+
+	for (frag = 0; frag < nr_frags; frag++) {
+		const skb_frag_t *fragp = &skb_shinfo(skb)->frags[frag];
+
+		len = skb_frag_size(fragp);
+		mapping = skb_frag_dma_map(&cp->pdev->dev, fragp, 0, len,
+					   DMA_TO_DEVICE);
+
+		tabort = cas_calc_tabort(cp, skb_frag_off(fragp), len);
+		if (unlikely(tabort)) {
+			/* NOTE: len is always > tabort */
+			cas_write_txd(cp, ring, entry, mapping, len - tabort,
+				      ctrl, 0);
+			entry = TX_DESC_NEXT(ring, entry);
+			memcpy_from_page(tx_tiny_buf(cp, ring, entry),
+					 skb_frag_page(fragp),
+					 skb_frag_off(fragp) + len - tabort,
+					 tabort);
+			mapping = tx_tiny_map(cp, ring, entry, tentry);
+			len     = tabort;
+		}
+
+		cas_write_txd(cp, ring, entry, mapping, len, ctrl,
+			      (frag + 1 == nr_frags));
+		entry = TX_DESC_NEXT(ring, entry);
+	}
+
+	cp->tx_new[ring] = entry;
+	if (TX_BUFFS_AVAIL(cp, ring) <= CAS_TABORT(cp)*(MAX_SKB_FRAGS + 1))
+		netif_stop_queue(dev);
+
+	netif_printk(cp, tx_queued, KERN_DEBUG, dev,
+		     "tx[%d] queued, slot %d, skblen %d, avail %d\n",
+		     ring, entry, skb->len, TX_BUFFS_AVAIL(cp, ring));
+	writel(entry, cp->regs + REG_TX_KICKN(ring));
+	spin_unlock_irqrestore(&cp->tx_lock[ring], flags);
+	return 0;
+}
+
+static netdev_tx_t cas_start_xmit(struct sk_buff *skb, struct net_device *dev)
+{
+	struct cas *cp = netdev_priv(dev);
+
+	/* this is only used as a load-balancing hint, so it doesn't
+	 * need to be SMP safe
+	 */
+	static int ring;
+
+	if (skb_padto(skb, cp->min_frame_size))
+		return NETDEV_TX_OK;
+
+	/* XXX: we need some higher-level QoS hooks to steer packets to
+	 *      individual queues.
+	 */
+	if (cas_xmit_tx_ringN(cp, ring++ & N_TX_RINGS_MASK, skb))
+		return NETDEV_TX_BUSY;
+	return NETDEV_TX_OK;
+}
+
+static void cas_init_tx_dma(struct cas *cp)
+{
+	u64 desc_dma = cp->block_dvma;
+	unsigned long off;
+	u32 val;
+	int i;
+
+	/* set up tx completion writeback registers. must be 8-byte aligned */
+#ifdef USE_TX_COMPWB
+	off = offsetof(struct cas_init_block, tx_compwb);
+	writel((desc_dma + off) >> 32, cp->regs + REG_TX_COMPWB_DB_HI);
+	writel((desc_dma + off) & 0xffffffff, cp->regs + REG_TX_COMPWB_DB_LOW);
+#endif
+
+	/* enable completion writebacks, enable paced mode,
+	 * disable read pipe, and disable pre-interrupt compwbs
+	 */
+	val =   TX_CFG_COMPWB_Q1 | TX_CFG_COMPWB_Q2 |
+		TX_CFG_COMPWB_Q3 | TX_CFG_COMPWB_Q4 |
+		TX_CFG_DMA_RDPIPE_DIS | TX_CFG_PACED_MODE |
+		TX_CFG_INTR_COMPWB_DIS;
+
+	/* write out tx ring info and tx desc bases */
+	for (i = 0; i < MAX_TX_RINGS; i++) {
+		off = (unsigned long) cp->init_txds[i] -
+			(unsigned long) cp->init_block;
+
+		val |= CAS_TX_RINGN_BASE(i);
+		writel((desc_dma + off) >> 32, cp->regs + REG_TX_DBN_HI(i));
+		writel((desc_dma + off) & 0xffffffff, cp->regs +
+		       REG_TX_DBN_LOW(i));
+		/* don't zero out the kick register here as the system
+		 * will wedge
+		 */
+	}
+	writel(val, cp->regs + REG_TX_CFG);
+
+	/* program max burst sizes. these numbers should be different
+	 * if doing QoS.
+	 */
+#ifdef USE_QOS
+	writel(0x800, cp->regs + REG_TX_MAXBURST_0);
+	writel(0x1600, cp->regs + REG_TX_MAXBURST_1);
+	writel(0x2400, cp->regs + REG_TX_MAXBURST_2);
+	writel(0x4800, cp->regs + REG_TX_MAXBURST_3);
+#else
+	writel(0x800, cp->regs + REG_TX_MAXBURST_0);
+	writel(0x800, cp->regs + REG_TX_MAXBURST_1);
+	writel(0x800, cp->regs + REG_TX_MAXBURST_2);
+	writel(0x800, cp->regs + REG_TX_MAXBURST_3);
+#endif
+}
+
+/* Must be invoked under cp->lock. */
+static inline void cas_init_dma(struct cas *cp)
+{
+	cas_init_tx_dma(cp);
+	cas_init_rx_dma(cp);
+}
+
+static void cas_process_mc_list(struct cas *cp)
+{
+	u16 hash_table[16];
+	u32 crc;
+	struct netdev_hw_addr *ha;
+	int i = 1;
+
+	memset(hash_table, 0, sizeof(hash_table));
+	netdev_for_each_mc_addr(ha, cp->dev) {
+		if (i <= CAS_MC_EXACT_MATCH_SIZE) {
+			/* use the alternate mac address registers for the
+			 * first 15 multicast addresses
+			 */
+			writel((ha->addr[4] << 8) | ha->addr[5],
+			       cp->regs + REG_MAC_ADDRN(i*3 + 0));
+			writel((ha->addr[2] << 8) | ha->addr[3],
+			       cp->regs + REG_MAC_ADDRN(i*3 + 1));
+			writel((ha->addr[0] << 8) | ha->addr[1],
+			       cp->regs + REG_MAC_ADDRN(i*3 + 2));
+			i++;
+		}
+		else {
+			/* use hw hash table for the next series of
+			 * multicast addresses
+			 */
+			crc = ether_crc_le(ETH_ALEN, ha->addr);
+			crc >>= 24;
+			hash_table[crc >> 4] |= 1 << (15 - (crc & 0xf));
+		}
+	}
+	for (i = 0; i < 16; i++)
+		writel(hash_table[i], cp->regs + REG_MAC_HASH_TABLEN(i));
+}
+
+/* Must be invoked under cp->lock. */
+static u32 cas_setup_multicast(struct cas *cp)
+{
+	u32 rxcfg = 0;
+	int i;
+
+	if (cp->dev->flags & IFF_PROMISC) {
+		rxcfg |= MAC_RX_CFG_PROMISC_EN;
+
+	} else if (cp->dev->flags & IFF_ALLMULTI) {
+	    	for (i=0; i < 16; i++)
+			writel(0xFFFF, cp->regs + REG_MAC_HASH_TABLEN(i));
+		rxcfg |= MAC_RX_CFG_HASH_FILTER_EN;
+
+	} else {
+		cas_process_mc_list(cp);
+		rxcfg |= MAC_RX_CFG_HASH_FILTER_EN;
+	}
+
+	return rxcfg;
+}
+
+/* must be invoked under cp->stat_lock[N_TX_RINGS] */
+static void cas_clear_mac_err(struct cas *cp)
+{
+	writel(0, cp->regs + REG_MAC_COLL_NORMAL);
+	writel(0, cp->regs + REG_MAC_COLL_FIRST);
+	writel(0, cp->regs + REG_MAC_COLL_EXCESS);
+	writel(0, cp->regs + REG_MAC_COLL_LATE);
+	writel(0, cp->regs + REG_MAC_TIMER_DEFER);
+	writel(0, cp->regs + REG_MAC_ATTEMPTS_PEAK);
+	writel(0, cp->regs + REG_MAC_RECV_FRAME);
+	writel(0, cp->regs + REG_MAC_LEN_ERR);
+	writel(0, cp->regs + REG_MAC_ALIGN_ERR);
+	writel(0, cp->regs + REG_MAC_FCS_ERR);
+	writel(0, cp->regs + REG_MAC_RX_CODE_ERR);
+}
+
+
+static void cas_mac_reset(struct cas *cp)
+{
+	int i;
+
+	/* do both TX and RX reset */
+	writel(0x1, cp->regs + REG_MAC_TX_RESET);
+	writel(0x1, cp->regs + REG_MAC_RX_RESET);
+
+	/* wait for TX */
+	i = STOP_TRIES;
+	while (i-- > 0) {
+		if (readl(cp->regs + REG_MAC_TX_RESET) == 0)
+			break;
+		udelay(10);
+	}
+
+	/* wait for RX */
+	i = STOP_TRIES;
+	while (i-- > 0) {
+		if (readl(cp->regs + REG_MAC_RX_RESET) == 0)
+			break;
+		udelay(10);
+	}
+
+	if (readl(cp->regs + REG_MAC_TX_RESET) |
+	    readl(cp->regs + REG_MAC_RX_RESET))
+		netdev_err(cp->dev, "mac tx[%d]/rx[%d] reset failed [%08x]\n",
+			   readl(cp->regs + REG_MAC_TX_RESET),
+			   readl(cp->regs + REG_MAC_RX_RESET),
+			   readl(cp->regs + REG_MAC_STATE_MACHINE));
+}
+
+
+/* Must be invoked under cp->lock. */
+static void cas_init_mac(struct cas *cp)
+{
+	const unsigned char *e = &cp->dev->dev_addr[0];
+	int i;
+	cas_mac_reset(cp);
+
+	/* setup core arbitration weight register */
+	writel(CAWR_RR_DIS, cp->regs + REG_CAWR);
+
+#if !defined(CONFIG_SPARC64) && !defined(CONFIG_ALPHA)
+	/* set the infinite burst register for chips that don't have
+	 * pci issues.
+	 */
+	if ((cp->cas_flags & CAS_FLAG_TARGET_ABORT) == 0)
+		writel(INF_BURST_EN, cp->regs + REG_INF_BURST);
+#endif
+
+	writel(0x1BF0, cp->regs + REG_MAC_SEND_PAUSE);
+
+	writel(0x00, cp->regs + REG_MAC_IPG0);
+	writel(0x08, cp->regs + REG_MAC_IPG1);
+	writel(0x04, cp->regs + REG_MAC_IPG2);
+
+	/* change later for 802.3z */
+	writel(0x40, cp->regs + REG_MAC_SLOT_TIME);
+
+	/* min frame + FCS */
+	writel(ETH_ZLEN + 4, cp->regs + REG_MAC_FRAMESIZE_MIN);
+
+	/* Ethernet payload + header + FCS + optional VLAN tag. NOTE: we
+	 * specify the maximum frame size to prevent RX tag errors on
+	 * oversized frames.
+	 */
+	writel(CAS_BASE(MAC_FRAMESIZE_MAX_BURST, 0x2000) |
+	       CAS_BASE(MAC_FRAMESIZE_MAX_FRAME,
+			(CAS_MAX_MTU + ETH_HLEN + 4 + 4)),
+	       cp->regs + REG_MAC_FRAMESIZE_MAX);
+
+	/* NOTE: crc_size is used as a surrogate for half-duplex.
+	 * workaround saturn half-duplex issue by increasing preamble
+	 * size to 65 bytes.
+	 */
+	if ((cp->cas_flags & CAS_FLAG_SATURN) && cp->crc_size)
+		writel(0x41, cp->regs + REG_MAC_PA_SIZE);
+	else
+		writel(0x07, cp->regs + REG_MAC_PA_SIZE);
+	writel(0x04, cp->regs + REG_MAC_JAM_SIZE);
+	writel(0x10, cp->regs + REG_MAC_ATTEMPT_LIMIT);
+	writel(0x8808, cp->regs + REG_MAC_CTRL_TYPE);
+
+	writel((e[5] | (e[4] << 8)) & 0x3ff, cp->regs + REG_MAC_RANDOM_SEED);
+
+	writel(0, cp->regs + REG_MAC_ADDR_FILTER0);
+	writel(0, cp->regs + REG_MAC_ADDR_FILTER1);
+	writel(0, cp->regs + REG_MAC_ADDR_FILTER2);
+	writel(0, cp->regs + REG_MAC_ADDR_FILTER2_1_MASK);
+	writel(0, cp->regs + REG_MAC_ADDR_FILTER0_MASK);
+
+	/* setup mac address in perfect filter array */
+	for (i = 0; i < 45; i++)
+		writel(0x0, cp->regs + REG_MAC_ADDRN(i));
+
+	writel((e[4] << 8) | e[5], cp->regs + REG_MAC_ADDRN(0));
+	writel((e[2] << 8) | e[3], cp->regs + REG_MAC_ADDRN(1));
+	writel((e[0] << 8) | e[1], cp->regs + REG_MAC_ADDRN(2));
+
+	writel(0x0001, cp->regs + REG_MAC_ADDRN(42));
+	writel(0xc200, cp->regs + REG_MAC_ADDRN(43));
+	writel(0x0180, cp->regs + REG_MAC_ADDRN(44));
+
+	cp->mac_rx_cfg = cas_setup_multicast(cp);
+
+	spin_lock(&cp->stat_lock[N_TX_RINGS]);
+	cas_clear_mac_err(cp);
+	spin_unlock(&cp->stat_lock[N_TX_RINGS]);
+
+	/* Setup MAC interrupts.  We want to get all of the interesting
+	 * counter expiration events, but we do not want to hear about
+	 * normal rx/tx as the DMA engine tells us that.
+	 */
+	writel(MAC_TX_FRAME_XMIT, cp->regs + REG_MAC_TX_MASK);
+	writel(MAC_RX_FRAME_RECV, cp->regs + REG_MAC_RX_MASK);
+
+	/* Don't enable even the PAUSE interrupts for now, we
+	 * make no use of those events other than to record them.
+	 */
+	writel(0xffffffff, cp->regs + REG_MAC_CTRL_MASK);
+}
+
+/* Must be invoked under cp->lock. */
+static void cas_init_pause_thresholds(struct cas *cp)
+{
+	/* Calculate pause thresholds.  Setting the OFF threshold to the
+	 * full RX fifo size effectively disables PAUSE generation
+	 */
+	if (cp->rx_fifo_size <= (2 * 1024)) {
+		cp->rx_pause_off = cp->rx_pause_on = cp->rx_fifo_size;
+	} else {
+		int max_frame = (cp->dev->mtu + ETH_HLEN + 4 + 4 + 64) & ~63;
+		if (max_frame * 3 > cp->rx_fifo_size) {
+			cp->rx_pause_off = 7104;
+			cp->rx_pause_on  = 960;
+		} else {
+			int off = (cp->rx_fifo_size - (max_frame * 2));
+			int on = off - max_frame;
+			cp->rx_pause_off = off;
+			cp->rx_pause_on = on;
+		}
+	}
+}
+
+static int cas_vpd_match(const void __iomem *p, const char *str)
+{
+	int len = strlen(str) + 1;
+	int i;
+
+	for (i = 0; i < len; i++) {
+		if (readb(p + i) != str[i])
+			return 0;
+	}
+	return 1;
+}
+
+
+/* get the mac address by reading the vpd information in the rom.
+ * also get the phy type and determine if there's an entropy generator.
+ * NOTE: this is a bit convoluted for the following reasons:
+ *  1) vpd info has order-dependent mac addresses for multinic cards
+ *  2) the only way to determine the nic order is to use the slot
+ *     number.
+ *  3) fiber cards don't have bridges, so their slot numbers don't
+ *     mean anything.
+ *  4) we don't actually know we have a fiber card until after
+ *     the mac addresses are parsed.
+ */
+static int cas_get_vpd_info(struct cas *cp, unsigned char *dev_addr,
+			    const int offset)
+{
+	void __iomem *p = cp->regs + REG_EXPANSION_ROM_RUN_START;
+	void __iomem *base, *kstart;
+	int i, len;
+	int found = 0;
+#define VPD_FOUND_MAC        0x01
+#define VPD_FOUND_PHY        0x02
+
+	int phy_type = CAS_PHY_MII_MDIO0; /* default phy type */
+	int mac_off  = 0;
+
+#if defined(CONFIG_SPARC)
+	const unsigned char *addr;
+#endif
+
+	/* give us access to the PROM */
+	writel(BIM_LOCAL_DEV_PROM | BIM_LOCAL_DEV_PAD,
+	       cp->regs + REG_BIM_LOCAL_DEV_EN);
+
+	/* check for an expansion rom */
+	if (readb(p) != 0x55 || readb(p + 1) != 0xaa)
+		goto use_random_mac_addr;
+
+	/* search for beginning of vpd */
+	base = NULL;
+	for (i = 2; i < EXPANSION_ROM_SIZE; i++) {
+		/* check for PCIR */
+		if ((readb(p + i + 0) == 0x50) &&
+		    (readb(p + i + 1) == 0x43) &&
+		    (readb(p + i + 2) == 0x49) &&
+		    (readb(p + i + 3) == 0x52)) {
+			base = p + (readb(p + i + 8) |
+				    (readb(p + i + 9) << 8));
+			break;
+		}
+	}
+
+	if (!base || (readb(base) != 0x82))
+		goto use_random_mac_addr;
+
+	i = (readb(base + 1) | (readb(base + 2) << 8)) + 3;
+	while (i < EXPANSION_ROM_SIZE) {
+		if (readb(base + i) != 0x90) /* no vpd found */
+			goto use_random_mac_addr;
+
+		/* found a vpd field */
+		len = readb(base + i + 1) | (readb(base + i + 2) << 8);
+
+		/* extract keywords */
+		kstart = base + i + 3;
+		p = kstart;
+		while ((p - kstart) < len) {
+			int klen = readb(p + 2);
+			int j;
+			char type;
+
+			p += 3;
+
+			/* look for the following things:
+			 * -- correct length == 29
+			 * 3 (type) + 2 (size) +
+			 * 18 (strlen("local-mac-address") + 1) +
+			 * 6 (mac addr)
+			 * -- VPD Instance 'I'
+			 * -- VPD Type Bytes 'B'
+			 * -- VPD data length == 6
+			 * -- property string == local-mac-address
+			 *
+			 * -- correct length == 24
+			 * 3 (type) + 2 (size) +
+			 * 12 (strlen("entropy-dev") + 1) +
+			 * 7 (strlen("vms110") + 1)
+			 * -- VPD Instance 'I'
+			 * -- VPD Type String 'B'
+			 * -- VPD data length == 7
+			 * -- property string == entropy-dev
+			 *
+			 * -- correct length == 18
+			 * 3 (type) + 2 (size) +
+			 * 9 (strlen("phy-type") + 1) +
+			 * 4 (strlen("pcs") + 1)
+			 * -- VPD Instance 'I'
+			 * -- VPD Type String 'S'
+			 * -- VPD data length == 4
+			 * -- property string == phy-type
+			 *
+			 * -- correct length == 23
+			 * 3 (type) + 2 (size) +
+			 * 14 (strlen("phy-interface") + 1) +
+			 * 4 (strlen("pcs") + 1)
+			 * -- VPD Instance 'I'
+			 * -- VPD Type String 'S'
+			 * -- VPD data length == 4
+			 * -- property string == phy-interface
+			 */
+			if (readb(p) != 'I')
+				goto next;
+
+			/* finally, check string and length */
+			type = readb(p + 3);
+			if (type == 'B') {
+				if ((klen == 29) && readb(p + 4) == 6 &&
+				    cas_vpd_match(p + 5,
+						  "local-mac-address")) {
+					if (mac_off++ > offset)
+						goto next;
+
+					/* set mac address */
+					for (j = 0; j < 6; j++)
+						dev_addr[j] =
+							readb(p + 23 + j);
+					goto found_mac;
+				}
+			}
+
+			if (type != 'S')
+				goto next;
+
+#ifdef USE_ENTROPY_DEV
+			if ((klen == 24) &&
+			    cas_vpd_match(p + 5, "entropy-dev") &&
+			    cas_vpd_match(p + 17, "vms110")) {
+				cp->cas_flags |= CAS_FLAG_ENTROPY_DEV;
+				goto next;
+			}
+#endif
+
+			if (found & VPD_FOUND_PHY)
+				goto next;
+
+			if ((klen == 18) && readb(p + 4) == 4 &&
+			    cas_vpd_match(p + 5, "phy-type")) {
+				if (cas_vpd_match(p + 14, "pcs")) {
+					phy_type = CAS_PHY_SERDES;
+					goto found_phy;
+				}
+			}
+
+			if ((klen == 23) && readb(p + 4) == 4 &&
+			    cas_vpd_match(p + 5, "phy-interface")) {
+				if (cas_vpd_match(p + 19, "pcs")) {
+					phy_type = CAS_PHY_SERDES;
+					goto found_phy;
+				}
+			}
+found_mac:
+			found |= VPD_FOUND_MAC;
+			goto next;
+
+found_phy:
+			found |= VPD_FOUND_PHY;
+
+next:
+			p += klen;
+		}
+		i += len + 3;
+	}
+
+use_random_mac_addr:
+	if (found & VPD_FOUND_MAC)
+		goto done;
+
+#if defined(CONFIG_SPARC)
+	addr = of_get_property(cp->of_node, "local-mac-address", NULL);
+	if (addr != NULL) {
+		memcpy(dev_addr, addr, ETH_ALEN);
+		goto done;
+	}
+#endif
+
+	/* Sun MAC prefix then 3 random bytes. */
+	pr_info("MAC address not found in ROM VPD\n");
+	dev_addr[0] = 0x08;
+	dev_addr[1] = 0x00;
+	dev_addr[2] = 0x20;
+	get_random_bytes(dev_addr + 3, 3);
+
+done:
+	writel(0, cp->regs + REG_BIM_LOCAL_DEV_EN);
+	return phy_type;
+}
+
+/* check pci invariants */
+static void cas_check_pci_invariants(struct cas *cp)
+{
+	struct pci_dev *pdev = cp->pdev;
+
+	cp->cas_flags = 0;
+	if ((pdev->vendor == PCI_VENDOR_ID_SUN) &&
+	    (pdev->device == PCI_DEVICE_ID_SUN_CASSINI)) {
+		if (pdev->revision >= CAS_ID_REVPLUS)
+			cp->cas_flags |= CAS_FLAG_REG_PLUS;
+		if (pdev->revision < CAS_ID_REVPLUS02u)
+			cp->cas_flags |= CAS_FLAG_TARGET_ABORT;
+
+		/* Original Cassini supports HW CSUM, but it's not
+		 * enabled by default as it can trigger TX hangs.
+		 */
+		if (pdev->revision < CAS_ID_REV2)
+			cp->cas_flags |= CAS_FLAG_NO_HW_CSUM;
+	} else {
+		/* Only sun has original cassini chips.  */
+		cp->cas_flags |= CAS_FLAG_REG_PLUS;
+
+		/* We use a flag because the same phy might be externally
+		 * connected.
+		 */
+		if ((pdev->vendor == PCI_VENDOR_ID_NS) &&
+		    (pdev->device == PCI_DEVICE_ID_NS_SATURN))
+			cp->cas_flags |= CAS_FLAG_SATURN;
+	}
+}
+
+
+static int cas_check_invariants(struct cas *cp)
+{
+	struct pci_dev *pdev = cp->pdev;
+	u8 addr[ETH_ALEN];
+	u32 cfg;
+	int i;
+
+	/* get page size for rx buffers. */
+	cp->page_order = 0;
+#ifdef USE_PAGE_ORDER
+	if (PAGE_SHIFT < CAS_JUMBO_PAGE_SHIFT) {
+		/* see if we can allocate larger pages */
+		struct page *page = alloc_pages(GFP_ATOMIC,
+						CAS_JUMBO_PAGE_SHIFT -
+						PAGE_SHIFT);
+		if (page) {
+			__free_pages(page, CAS_JUMBO_PAGE_SHIFT - PAGE_SHIFT);
+			cp->page_order = CAS_JUMBO_PAGE_SHIFT - PAGE_SHIFT;
+		} else {
+			printk("MTU limited to %d bytes\n", CAS_MAX_MTU);
+		}
+	}
+#endif
+	cp->page_size = (PAGE_SIZE << cp->page_order);
+
+	/* Fetch the FIFO configurations. */
+	cp->tx_fifo_size = readl(cp->regs + REG_TX_FIFO_SIZE) * 64;
+	cp->rx_fifo_size = RX_FIFO_SIZE;
+
+	/* finish phy determination. MDIO1 takes precedence over MDIO0 if
+	 * they're both connected.
+	 */
+	cp->phy_type = cas_get_vpd_info(cp, addr, PCI_SLOT(pdev->devfn));
+	eth_hw_addr_set(cp->dev, addr);
+	if (cp->phy_type & CAS_PHY_SERDES) {
+		cp->cas_flags |= CAS_FLAG_1000MB_CAP;
+		return 0; /* no more checking needed */
+	}
+
+	/* MII */
+	cfg = readl(cp->regs + REG_MIF_CFG);
+	if (cfg & MIF_CFG_MDIO_1) {
+		cp->phy_type = CAS_PHY_MII_MDIO1;
+	} else if (cfg & MIF_CFG_MDIO_0) {
+		cp->phy_type = CAS_PHY_MII_MDIO0;
+	}
+
+	cas_mif_poll(cp, 0);
+	writel(PCS_DATAPATH_MODE_MII, cp->regs + REG_PCS_DATAPATH_MODE);
+
+	for (i = 0; i < 32; i++) {
+		u32 phy_id;
+		int j;
+
+		for (j = 0; j < 3; j++) {
+			cp->phy_addr = i;
+			phy_id = cas_phy_read(cp, MII_PHYSID1) << 16;
+			phy_id |= cas_phy_read(cp, MII_PHYSID2);
+			if (phy_id && (phy_id != 0xFFFFFFFF)) {
+				cp->phy_id = phy_id;
+				goto done;
+			}
+		}
+	}
+	pr_err("MII phy did not respond [%08x]\n",
+	       readl(cp->regs + REG_MIF_STATE_MACHINE));
+	return -1;
+
+done:
+	/* see if we can do gigabit */
+	cfg = cas_phy_read(cp, MII_BMSR);
+	if ((cfg & CAS_BMSR_1000_EXTEND) &&
+	    cas_phy_read(cp, CAS_MII_1000_EXTEND))
+		cp->cas_flags |= CAS_FLAG_1000MB_CAP;
+	return 0;
+}
+
+/* Must be invoked under cp->lock. */
+static inline void cas_start_dma(struct cas *cp)
+{
+	int i;
+	u32 val;
+	int txfailed = 0;
+
+	/* enable dma */
+	val = readl(cp->regs + REG_TX_CFG) | TX_CFG_DMA_EN;
+	writel(val, cp->regs + REG_TX_CFG);
+	val = readl(cp->regs + REG_RX_CFG) | RX_CFG_DMA_EN;
+	writel(val, cp->regs + REG_RX_CFG);
+
+	/* enable the mac */
+	val = readl(cp->regs + REG_MAC_TX_CFG) | MAC_TX_CFG_EN;
+	writel(val, cp->regs + REG_MAC_TX_CFG);
+	val = readl(cp->regs + REG_MAC_RX_CFG) | MAC_RX_CFG_EN;
+	writel(val, cp->regs + REG_MAC_RX_CFG);
+
+	i = STOP_TRIES;
+	while (i-- > 0) {
+		val = readl(cp->regs + REG_MAC_TX_CFG);
+		if ((val & MAC_TX_CFG_EN))
+			break;
+		udelay(10);
+	}
+	if (i < 0) txfailed = 1;
+	i = STOP_TRIES;
+	while (i-- > 0) {
+		val = readl(cp->regs + REG_MAC_RX_CFG);
+		if ((val & MAC_RX_CFG_EN)) {
+			if (txfailed) {
+				netdev_err(cp->dev,
+					   "enabling mac failed [tx:%08x:%08x]\n",
+					   readl(cp->regs + REG_MIF_STATE_MACHINE),
+					   readl(cp->regs + REG_MAC_STATE_MACHINE));
+			}
+			goto enable_rx_done;
+		}
+		udelay(10);
+	}
+	netdev_err(cp->dev, "enabling mac failed [%s:%08x:%08x]\n",
+		   (txfailed ? "tx,rx" : "rx"),
+		   readl(cp->regs + REG_MIF_STATE_MACHINE),
+		   readl(cp->regs + REG_MAC_STATE_MACHINE));
+
+enable_rx_done:
+	cas_unmask_intr(cp); /* enable interrupts */
+	writel(RX_DESC_RINGN_SIZE(0) - 4, cp->regs + REG_RX_KICK);
+	writel(0, cp->regs + REG_RX_COMP_TAIL);
+
+	if (cp->cas_flags & CAS_FLAG_REG_PLUS) {
+		if (N_RX_DESC_RINGS > 1)
+			writel(RX_DESC_RINGN_SIZE(1) - 4,
+			       cp->regs + REG_PLUS_RX_KICK1);
+	}
+}
+
+/* Must be invoked under cp->lock. */
+static void cas_read_pcs_link_mode(struct cas *cp, int *fd, int *spd,
+				   int *pause)
+{
+	u32 val = readl(cp->regs + REG_PCS_MII_LPA);
+	*fd     = (val & PCS_MII_LPA_FD) ? 1 : 0;
+	*pause  = (val & PCS_MII_LPA_SYM_PAUSE) ? 0x01 : 0x00;
+	if (val & PCS_MII_LPA_ASYM_PAUSE)
+		*pause |= 0x10;
+	*spd = 1000;
+}
+
+/* Must be invoked under cp->lock. */
+static void cas_read_mii_link_mode(struct cas *cp, int *fd, int *spd,
+				   int *pause)
+{
+	u32 val;
+
+	*fd = 0;
+	*spd = 10;
+	*pause = 0;
+
+	/* use GMII registers */
+	val = cas_phy_read(cp, MII_LPA);
+	if (val & CAS_LPA_PAUSE)
+		*pause = 0x01;
+
+	if (val & CAS_LPA_ASYM_PAUSE)
+		*pause |= 0x10;
+
+	if (val & LPA_DUPLEX)
+		*fd = 1;
+	if (val & LPA_100)
+		*spd = 100;
+
+	if (cp->cas_flags & CAS_FLAG_1000MB_CAP) {
+		val = cas_phy_read(cp, CAS_MII_1000_STATUS);
+		if (val & (CAS_LPA_1000FULL | CAS_LPA_1000HALF))
+			*spd = 1000;
+		if (val & CAS_LPA_1000FULL)
+			*fd = 1;
+	}
+}
+
+/* A link-up condition has occurred, initialize and enable the
+ * rest of the chip.
+ *
+ * Must be invoked under cp->lock.
+ */
+static void cas_set_link_modes(struct cas *cp)
+{
+	u32 val;
+	int full_duplex, speed, pause;
+
+	full_duplex = 0;
+	speed = 10;
+	pause = 0;
+
+	if (CAS_PHY_MII(cp->phy_type)) {
+		cas_mif_poll(cp, 0);
+		val = cas_phy_read(cp, MII_BMCR);
+		if (val & BMCR_ANENABLE) {
+			cas_read_mii_link_mode(cp, &full_duplex, &speed,
+					       &pause);
+		} else {
+			if (val & BMCR_FULLDPLX)
+				full_duplex = 1;
+
+			if (val & BMCR_SPEED100)
+				speed = 100;
+			else if (val & CAS_BMCR_SPEED1000)
+				speed = (cp->cas_flags & CAS_FLAG_1000MB_CAP) ?
+					1000 : 100;
+		}
+		cas_mif_poll(cp, 1);
+
+	} else {
+		val = readl(cp->regs + REG_PCS_MII_CTRL);
+		cas_read_pcs_link_mode(cp, &full_duplex, &speed, &pause);
+		if ((val & PCS_MII_AUTONEG_EN) == 0) {
+			if (val & PCS_MII_CTRL_DUPLEX)
+				full_duplex = 1;
+		}
+	}
+
+	netif_info(cp, link, cp->dev, "Link up at %d Mbps, %s-duplex\n",
+		   speed, full_duplex ? "full" : "half");
+
+	val = MAC_XIF_TX_MII_OUTPUT_EN | MAC_XIF_LINK_LED;
+	if (CAS_PHY_MII(cp->phy_type)) {
+		val |= MAC_XIF_MII_BUFFER_OUTPUT_EN;
+		if (!full_duplex)
+			val |= MAC_XIF_DISABLE_ECHO;
+	}
+	if (full_duplex)
+		val |= MAC_XIF_FDPLX_LED;
+	if (speed == 1000)
+		val |= MAC_XIF_GMII_MODE;
+	writel(val, cp->regs + REG_MAC_XIF_CFG);
+
+	/* deal with carrier and collision detect. */
+	val = MAC_TX_CFG_IPG_EN;
+	if (full_duplex) {
+		val |= MAC_TX_CFG_IGNORE_CARRIER;
+		val |= MAC_TX_CFG_IGNORE_COLL;
+	} else {
+#ifndef USE_CSMA_CD_PROTO
+		val |= MAC_TX_CFG_NEVER_GIVE_UP_EN;
+		val |= MAC_TX_CFG_NEVER_GIVE_UP_LIM;
+#endif
+	}
+	/* val now set up for REG_MAC_TX_CFG */
+
+	/* If gigabit and half-duplex, enable carrier extension
+	 * mode.  increase slot time to 512 bytes as well.
+	 * else, disable it and make sure slot time is 64 bytes.
+	 * also activate checksum bug workaround
+	 */
+	if ((speed == 1000) && !full_duplex) {
+		writel(val | MAC_TX_CFG_CARRIER_EXTEND,
+		       cp->regs + REG_MAC_TX_CFG);
+
+		val = readl(cp->regs + REG_MAC_RX_CFG);
+		val &= ~MAC_RX_CFG_STRIP_FCS; /* checksum workaround */
+		writel(val | MAC_RX_CFG_CARRIER_EXTEND,
+		       cp->regs + REG_MAC_RX_CFG);
+
+		writel(0x200, cp->regs + REG_MAC_SLOT_TIME);
+
+		cp->crc_size = 4;
+		/* minimum size gigabit frame at half duplex */
+		cp->min_frame_size = CAS_1000MB_MIN_FRAME;
+
+	} else {
+		writel(val, cp->regs + REG_MAC_TX_CFG);
+
+		/* checksum bug workaround. don't strip FCS when in
+		 * half-duplex mode
+		 */
+		val = readl(cp->regs + REG_MAC_RX_CFG);
+		if (full_duplex) {
+			val |= MAC_RX_CFG_STRIP_FCS;
+			cp->crc_size = 0;
+			cp->min_frame_size = CAS_MIN_MTU;
+		} else {
+			val &= ~MAC_RX_CFG_STRIP_FCS;
+			cp->crc_size = 4;
+			cp->min_frame_size = CAS_MIN_FRAME;
+		}
+		writel(val & ~MAC_RX_CFG_CARRIER_EXTEND,
+		       cp->regs + REG_MAC_RX_CFG);
+		writel(0x40, cp->regs + REG_MAC_SLOT_TIME);
+	}
+
+	if (netif_msg_link(cp)) {
+		if (pause & 0x01) {
+			netdev_info(cp->dev, "Pause is enabled (rxfifo: %d off: %d on: %d)\n",
+				    cp->rx_fifo_size,
+				    cp->rx_pause_off,
+				    cp->rx_pause_on);
+		} else if (pause & 0x10) {
+			netdev_info(cp->dev, "TX pause enabled\n");
+		} else {
+			netdev_info(cp->dev, "Pause is disabled\n");
+		}
+	}
+
+	val = readl(cp->regs + REG_MAC_CTRL_CFG);
+	val &= ~(MAC_CTRL_CFG_SEND_PAUSE_EN | MAC_CTRL_CFG_RECV_PAUSE_EN);
+	if (pause) { /* symmetric or asymmetric pause */
+		val |= MAC_CTRL_CFG_SEND_PAUSE_EN;
+		if (pause & 0x01) { /* symmetric pause */
+			val |= MAC_CTRL_CFG_RECV_PAUSE_EN;
+		}
+	}
+	writel(val, cp->regs + REG_MAC_CTRL_CFG);
+	cas_start_dma(cp);
+}
+
+/* Must be invoked under cp->lock. */
+static void cas_init_hw(struct cas *cp, int restart_link)
+{
+	if (restart_link)
+		cas_phy_init(cp);
+
+	cas_init_pause_thresholds(cp);
+	cas_init_mac(cp);
+	cas_init_dma(cp);
+
+	if (restart_link) {
+		/* Default aneg parameters */
+		cp->timer_ticks = 0;
+		cas_begin_auto_negotiation(cp, NULL);
+	} else if (cp->lstate == link_up) {
+		cas_set_link_modes(cp);
+		netif_carrier_on(cp->dev);
+	}
+}
+
+/* Must be invoked under cp->lock. on earlier cassini boards,
+ * SOFT_0 is tied to PCI reset. we use this to force a pci reset,
+ * let it settle out, and then restore pci state.
+ */
+static void cas_hard_reset(struct cas *cp)
+{
+	writel(BIM_LOCAL_DEV_SOFT_0, cp->regs + REG_BIM_LOCAL_DEV_EN);
+	udelay(20);
+	pci_restore_state(cp->pdev);
+}
+
+
+static void cas_global_reset(struct cas *cp, int blkflag)
+{
+	int limit;
+
+	/* issue a global reset. don't use RSTOUT. */
+	if (blkflag && !CAS_PHY_MII(cp->phy_type)) {
+		/* For PCS, when the blkflag is set, we should set the
+		 * SW_REST_BLOCK_PCS_SLINK bit to prevent the results of
+		 * the last autonegotiation from being cleared.  We'll
+		 * need some special handling if the chip is set into a
+		 * loopback mode.
+		 */
+		writel((SW_RESET_TX | SW_RESET_RX | SW_RESET_BLOCK_PCS_SLINK),
+		       cp->regs + REG_SW_RESET);
+	} else {
+		writel(SW_RESET_TX | SW_RESET_RX, cp->regs + REG_SW_RESET);
+	}
+
+	/* need to wait at least 3ms before polling register */
+	mdelay(3);
+
+	limit = STOP_TRIES;
+	while (limit-- > 0) {
+		u32 val = readl(cp->regs + REG_SW_RESET);
+		if ((val & (SW_RESET_TX | SW_RESET_RX)) == 0)
+			goto done;
+		udelay(10);
+	}
+	netdev_err(cp->dev, "sw reset failed\n");
+
+done:
+	/* enable various BIM interrupts */
+	writel(BIM_CFG_DPAR_INTR_ENABLE | BIM_CFG_RMA_INTR_ENABLE |
+	       BIM_CFG_RTA_INTR_ENABLE, cp->regs + REG_BIM_CFG);
+
+	/* clear out pci error status mask for handled errors.
+	 * we don't deal with DMA counter overflows as they happen
+	 * all the time.
+	 */
+	writel(0xFFFFFFFFU & ~(PCI_ERR_BADACK | PCI_ERR_DTRTO |
+			       PCI_ERR_OTHER | PCI_ERR_BIM_DMA_WRITE |
+			       PCI_ERR_BIM_DMA_READ), cp->regs +
+	       REG_PCI_ERR_STATUS_MASK);
+
+	/* set up for MII by default to address mac rx reset timeout
+	 * issue
+	 */
+	writel(PCS_DATAPATH_MODE_MII, cp->regs + REG_PCS_DATAPATH_MODE);
+}
+
+static void cas_reset(struct cas *cp, int blkflag)
+{
+	u32 val;
+
+	cas_mask_intr(cp);
+	cas_global_reset(cp, blkflag);
+	cas_mac_reset(cp);
+	cas_entropy_reset(cp);
+
+	/* disable dma engines. */
+	val = readl(cp->regs + REG_TX_CFG);
+	val &= ~TX_CFG_DMA_EN;
+	writel(val, cp->regs + REG_TX_CFG);
+
+	val = readl(cp->regs + REG_RX_CFG);
+	val &= ~RX_CFG_DMA_EN;
+	writel(val, cp->regs + REG_RX_CFG);
+
+	/* program header parser */
+	if ((cp->cas_flags & CAS_FLAG_TARGET_ABORT) ||
+	    (&CAS_HP_ALT_FIRMWARE[0] == &cas_prog_null[0])) {
+		cas_load_firmware(cp, CAS_HP_FIRMWARE);
+	} else {
+		cas_load_firmware(cp, CAS_HP_ALT_FIRMWARE);
+	}
+
+	/* clear out error registers */
+	spin_lock(&cp->stat_lock[N_TX_RINGS]);
+	cas_clear_mac_err(cp);
+	spin_unlock(&cp->stat_lock[N_TX_RINGS]);
+}
+
+/* Shut down the chip, must be called with pm_mutex held.  */
+static void cas_shutdown(struct cas *cp)
+{
+	unsigned long flags;
+
+	/* Make us not-running to avoid timers respawning */
+	cp->hw_running = 0;
+
+	timer_delete_sync(&cp->link_timer);
+
+	/* Stop the reset task */
+#if 0
+	while (atomic_read(&cp->reset_task_pending_mtu) ||
+	       atomic_read(&cp->reset_task_pending_spare) ||
+	       atomic_read(&cp->reset_task_pending_all))
+		schedule();
+
+#else
+	while (atomic_read(&cp->reset_task_pending))
+		schedule();
+#endif
+	/* Actually stop the chip */
+	cas_lock_all_save(cp, flags);
+	cas_reset(cp, 0);
+	if (cp->cas_flags & CAS_FLAG_SATURN)
+		cas_phy_powerdown(cp);
+	cas_unlock_all_restore(cp, flags);
+}
+
+static int cas_change_mtu(struct net_device *dev, int new_mtu)
+{
+	struct cas *cp = netdev_priv(dev);
+
+	WRITE_ONCE(dev->mtu, new_mtu);
+	if (!netif_running(dev) || !netif_device_present(dev))
+		return 0;
+
+	/* let the reset task handle it */
+#if 1
+	atomic_inc(&cp->reset_task_pending);
+	if ((cp->phy_type & CAS_PHY_SERDES)) {
+		atomic_inc(&cp->reset_task_pending_all);
+	} else {
+		atomic_inc(&cp->reset_task_pending_mtu);
+	}
+	schedule_work(&cp->reset_task);
+#else
+	atomic_set(&cp->reset_task_pending, (cp->phy_type & CAS_PHY_SERDES) ?
+		   CAS_RESET_ALL : CAS_RESET_MTU);
+	pr_err("reset called in cas_change_mtu\n");
+	schedule_work(&cp->reset_task);
+#endif
+
+	flush_work(&cp->reset_task);
+	return 0;
+}
+
+static void cas_clean_txd(struct cas *cp, int ring)
+{
+	struct cas_tx_desc *txd = cp->init_txds[ring];
+	struct sk_buff *skb, **skbs = cp->tx_skbs[ring];
+	u64 daddr, dlen;
+	int i, size;
+
+	size = TX_DESC_RINGN_SIZE(ring);
+	for (i = 0; i < size; i++) {
+		int frag;
+
+		if (skbs[i] == NULL)
+			continue;
+
+		skb = skbs[i];
+		skbs[i] = NULL;
+
+		for (frag = 0; frag <= skb_shinfo(skb)->nr_frags;  frag++) {
+			int ent = i & (size - 1);
+
+			/* first buffer is never a tiny buffer and so
+			 * needs to be unmapped.
+			 */
+			daddr = le64_to_cpu(txd[ent].buffer);
+			dlen  =  CAS_VAL(TX_DESC_BUFLEN,
+					 le64_to_cpu(txd[ent].control));
+			dma_unmap_page(&cp->pdev->dev, daddr, dlen,
+				       DMA_TO_DEVICE);
+
+			if (frag != skb_shinfo(skb)->nr_frags) {
+				i++;
+
+				/* next buffer might by a tiny buffer.
+				 * skip past it.
+				 */
+				ent = i & (size - 1);
+				if (cp->tx_tiny_use[ring][ent].used)
+					i++;
+			}
+		}
+		dev_kfree_skb_any(skb);
+	}
+
+	/* zero out tiny buf usage */
+	memset(cp->tx_tiny_use[ring], 0, size*sizeof(*cp->tx_tiny_use[ring]));
+}
+
+/* freed on close */
+static inline void cas_free_rx_desc(struct cas *cp, int ring)
+{
+	cas_page_t **page = cp->rx_pages[ring];
+	int i, size;
+
+	size = RX_DESC_RINGN_SIZE(ring);
+	for (i = 0; i < size; i++) {
+		if (page[i]) {
+			cas_page_free(cp, page[i]);
+			page[i] = NULL;
+		}
+	}
+}
+
+static void cas_free_rxds(struct cas *cp)
+{
+	int i;
+
+	for (i = 0; i < N_RX_DESC_RINGS; i++)
+		cas_free_rx_desc(cp, i);
+}
+
+/* Must be invoked under cp->lock. */
+static void cas_clean_rings(struct cas *cp)
+{
+	int i;
+
+	/* need to clean all tx rings */
+	memset(cp->tx_old, 0, sizeof(*cp->tx_old)*N_TX_RINGS);
+	memset(cp->tx_new, 0, sizeof(*cp->tx_new)*N_TX_RINGS);
+	for (i = 0; i < N_TX_RINGS; i++)
+		cas_clean_txd(cp, i);
+
+	/* zero out init block */
+	memset(cp->init_block, 0, sizeof(struct cas_init_block));
+	cas_clean_rxds(cp);
+	cas_clean_rxcs(cp);
+}
+
+/* allocated on open */
+static inline int cas_alloc_rx_desc(struct cas *cp, int ring)
+{
+	cas_page_t **page = cp->rx_pages[ring];
+	int size, i = 0;
+
+	size = RX_DESC_RINGN_SIZE(ring);
+	for (i = 0; i < size; i++) {
+		if ((page[i] = cas_page_alloc(cp, GFP_KERNEL)) == NULL)
+			return -1;
+	}
+	return 0;
+}
+
+static int cas_alloc_rxds(struct cas *cp)
+{
+	int i;
+
+	for (i = 0; i < N_RX_DESC_RINGS; i++) {
+		if (cas_alloc_rx_desc(cp, i) < 0) {
+			cas_free_rxds(cp);
+			return -1;
+		}
+	}
+	return 0;
+}
+
+static void cas_reset_task(struct work_struct *work)
+{
+	struct cas *cp = container_of(work, struct cas, reset_task);
+#if 0
+	int pending = atomic_read(&cp->reset_task_pending);
+#else
+	int pending_all = atomic_read(&cp->reset_task_pending_all);
+	int pending_spare = atomic_read(&cp->reset_task_pending_spare);
+	int pending_mtu = atomic_read(&cp->reset_task_pending_mtu);
+
+	if (pending_all == 0 && pending_spare == 0 && pending_mtu == 0) {
+		/* We can have more tasks scheduled than actually
+		 * needed.
+		 */
+		atomic_dec(&cp->reset_task_pending);
+		return;
+	}
+#endif
+	/* The link went down, we reset the ring, but keep
+	 * DMA stopped. Use this function for reset
+	 * on error as well.
+	 */
+	if (cp->hw_running) {
+		unsigned long flags;
+
+		/* Make sure we don't get interrupts or tx packets */
+		netif_device_detach(cp->dev);
+		cas_lock_all_save(cp, flags);
+
+		if (cp->opened) {
+			/* We call cas_spare_recover when we call cas_open.
+			 * but we do not initialize the lists cas_spare_recover
+			 * uses until cas_open is called.
+			 */
+			cas_spare_recover(cp, GFP_ATOMIC);
+		}
+#if 1
+		/* test => only pending_spare set */
+		if (!pending_all && !pending_mtu)
+			goto done;
+#else
+		if (pending == CAS_RESET_SPARE)
+			goto done;
+#endif
+		/* when pending == CAS_RESET_ALL, the following
+		 * call to cas_init_hw will restart auto negotiation.
+		 * Setting the second argument of cas_reset to
+		 * !(pending == CAS_RESET_ALL) will set this argument
+		 * to 1 (avoiding reinitializing the PHY for the normal
+		 * PCS case) when auto negotiation is not restarted.
+		 */
+#if 1
+		cas_reset(cp, !(pending_all > 0));
+		if (cp->opened)
+			cas_clean_rings(cp);
+		cas_init_hw(cp, (pending_all > 0));
+#else
+		cas_reset(cp, !(pending == CAS_RESET_ALL));
+		if (cp->opened)
+			cas_clean_rings(cp);
+		cas_init_hw(cp, pending == CAS_RESET_ALL);
+#endif
+
+done:
+		cas_unlock_all_restore(cp, flags);
+		netif_device_attach(cp->dev);
+	}
+#if 1
+	atomic_sub(pending_all, &cp->reset_task_pending_all);
+	atomic_sub(pending_spare, &cp->reset_task_pending_spare);
+	atomic_sub(pending_mtu, &cp->reset_task_pending_mtu);
+	atomic_dec(&cp->reset_task_pending);
+#else
+	atomic_set(&cp->reset_task_pending, 0);
+#endif
+}
+
+static void cas_link_timer(struct timer_list *t)
+{
+	struct cas *cp = timer_container_of(cp, t, link_timer);
+	int mask, pending = 0, reset = 0;
+	unsigned long flags;
+
+	if (link_transition_timeout != 0 &&
+	    cp->link_transition_jiffies_valid &&
+	    time_is_before_jiffies(cp->link_transition_jiffies +
+	      link_transition_timeout)) {
+		/* One-second counter so link-down workaround doesn't
+		 * cause resets to occur so fast as to fool the switch
+		 * into thinking the link is down.
+		 */
+		cp->link_transition_jiffies_valid = 0;
+	}
+
+	if (!cp->hw_running)
+		return;
+
+	spin_lock_irqsave(&cp->lock, flags);
+	cas_lock_tx(cp);
+	cas_entropy_gather(cp);
+
+	/* If the link task is still pending, we just
+	 * reschedule the link timer
+	 */
+#if 1
+	if (atomic_read(&cp->reset_task_pending_all) ||
+	    atomic_read(&cp->reset_task_pending_spare) ||
+	    atomic_read(&cp->reset_task_pending_mtu))
+		goto done;
+#else
+	if (atomic_read(&cp->reset_task_pending))
+		goto done;
+#endif
+
+	/* check for rx cleaning */
+	if ((mask = (cp->cas_flags & CAS_FLAG_RXD_POST_MASK))) {
+		int i, rmask;
+
+		for (i = 0; i < MAX_RX_DESC_RINGS; i++) {
+			rmask = CAS_FLAG_RXD_POST(i);
+			if ((mask & rmask) == 0)
+				continue;
+
+			/* post_rxds will do a mod_timer */
+			if (cas_post_rxds_ringN(cp, i, cp->rx_last[i]) < 0) {
+				pending = 1;
+				continue;
+			}
+			cp->cas_flags &= ~rmask;
+		}
+	}
+
+	if (CAS_PHY_MII(cp->phy_type)) {
+		u16 bmsr;
+		cas_mif_poll(cp, 0);
+		bmsr = cas_phy_read(cp, MII_BMSR);
+		/* WTZ: Solaris driver reads this twice, but that
+		 * may be due to the PCS case and the use of a
+		 * common implementation. Read it twice here to be
+		 * safe.
+		 */
+		bmsr = cas_phy_read(cp, MII_BMSR);
+		cas_mif_poll(cp, 1);
+		readl(cp->regs + REG_MIF_STATUS); /* avoid dups */
+		reset = cas_mii_link_check(cp, bmsr);
+	} else {
+		reset = cas_pcs_link_check(cp);
+	}
+
+	if (reset)
+		goto done;
+
+	/* check for tx state machine confusion */
+	if ((readl(cp->regs + REG_MAC_TX_STATUS) & MAC_TX_FRAME_XMIT) == 0) {
+		u32 val = readl(cp->regs + REG_MAC_STATE_MACHINE);
+		u32 wptr, rptr;
+		int tlm  = CAS_VAL(MAC_SM_TLM, val);
+
+		if (((tlm == 0x5) || (tlm == 0x3)) &&
+		    (CAS_VAL(MAC_SM_ENCAP_SM, val) == 0)) {
+			netif_printk(cp, tx_err, KERN_DEBUG, cp->dev,
+				     "tx err: MAC_STATE[%08x]\n", val);
+			reset = 1;
+			goto done;
+		}
+
+		val  = readl(cp->regs + REG_TX_FIFO_PKT_CNT);
+		wptr = readl(cp->regs + REG_TX_FIFO_WRITE_PTR);
+		rptr = readl(cp->regs + REG_TX_FIFO_READ_PTR);
+		if ((val == 0) && (wptr != rptr)) {
+			netif_printk(cp, tx_err, KERN_DEBUG, cp->dev,
+				     "tx err: TX_FIFO[%08x:%08x:%08x]\n",
+				     val, wptr, rptr);
+			reset = 1;
+		}
+
+		if (reset)
+			cas_hard_reset(cp);
+	}
+
+done:
+	if (reset) {
+#if 1
+		atomic_inc(&cp->reset_task_pending);
+		atomic_inc(&cp->reset_task_pending_all);
+		schedule_work(&cp->reset_task);
+#else
+		atomic_set(&cp->reset_task_pending, CAS_RESET_ALL);
+		pr_err("reset called in cas_link_timer\n");
+		schedule_work(&cp->reset_task);
+#endif
+	}
+
+	if (!pending)
+		mod_timer(&cp->link_timer, jiffies + CAS_LINK_TIMEOUT);
+	cas_unlock_tx(cp);
+	spin_unlock_irqrestore(&cp->lock, flags);
+}
+
+/* tiny buffers are used to avoid target abort issues with
+ * older cassini's
+ */
+static void cas_tx_tiny_free(struct cas *cp)
+{
+	struct pci_dev *pdev = cp->pdev;
+	int i;
+
+	for (i = 0; i < N_TX_RINGS; i++) {
+		if (!cp->tx_tiny_bufs[i])
+			continue;
+
+		dma_free_coherent(&pdev->dev, TX_TINY_BUF_BLOCK,
+				  cp->tx_tiny_bufs[i], cp->tx_tiny_dvma[i]);
+		cp->tx_tiny_bufs[i] = NULL;
+	}
+}
+
+static int cas_tx_tiny_alloc(struct cas *cp)
+{
+	struct pci_dev *pdev = cp->pdev;
+	int i;
+
+	for (i = 0; i < N_TX_RINGS; i++) {
+		cp->tx_tiny_bufs[i] =
+			dma_alloc_coherent(&pdev->dev, TX_TINY_BUF_BLOCK,
+					   &cp->tx_tiny_dvma[i], GFP_KERNEL);
+		if (!cp->tx_tiny_bufs[i]) {
+			cas_tx_tiny_free(cp);
+			return -1;
+		}
+	}
+	return 0;
+}
+
+
+static int cas_open(struct net_device *dev)
+{
+	struct cas *cp = netdev_priv(dev);
+	int hw_was_up, err;
+	unsigned long flags;
+
+	mutex_lock(&cp->pm_mutex);
+
+	hw_was_up = cp->hw_running;
+
+	/* The power-management mutex protects the hw_running
+	 * etc. state so it is safe to do this bit without cp->lock
+	 */
+	if (!cp->hw_running) {
+		/* Reset the chip */
+		cas_lock_all_save(cp, flags);
+		/* We set the second arg to cas_reset to zero
+		 * because cas_init_hw below will have its second
+		 * argument set to non-zero, which will force
+		 * autonegotiation to start.
+		 */
+		cas_reset(cp, 0);
+		cp->hw_running = 1;
+		cas_unlock_all_restore(cp, flags);
+	}
+
+	err = -ENOMEM;
+	if (cas_tx_tiny_alloc(cp) < 0)
+		goto err_unlock;
+
+	/* alloc rx descriptors */
+	if (cas_alloc_rxds(cp) < 0)
+		goto err_tx_tiny;
+
+	/* allocate spares */
+	cas_spare_init(cp);
+	cas_spare_recover(cp, GFP_KERNEL);
+
+	/* We can now request the interrupt as we know it's masked
+	 * on the controller. cassini+ has up to 4 interrupts
+	 * that can be used, but you need to do explicit pci interrupt
+	 * mapping to expose them
+	 */
+	if (request_irq(cp->pdev->irq, cas_interrupt,
+			IRQF_SHARED, dev->name, (void *) dev)) {
+		netdev_err(cp->dev, "failed to request irq !\n");
+		err = -EAGAIN;
+		goto err_spare;
+	}
+
+#ifdef USE_NAPI
+	napi_enable(&cp->napi);
+#endif
+	/* init hw */
+	cas_lock_all_save(cp, flags);
+	cas_clean_rings(cp);
+	cas_init_hw(cp, !hw_was_up);
+	cp->opened = 1;
+	cas_unlock_all_restore(cp, flags);
+
+	netif_start_queue(dev);
+	mutex_unlock(&cp->pm_mutex);
+	return 0;
+
+err_spare:
+	cas_spare_free(cp);
+	cas_free_rxds(cp);
+err_tx_tiny:
+	cas_tx_tiny_free(cp);
+err_unlock:
+	mutex_unlock(&cp->pm_mutex);
+	return err;
+}
+
+static int cas_close(struct net_device *dev)
+{
+	unsigned long flags;
+	struct cas *cp = netdev_priv(dev);
+
+#ifdef USE_NAPI
+	napi_disable(&cp->napi);
+#endif
+	/* Make sure we don't get distracted by suspend/resume */
+	mutex_lock(&cp->pm_mutex);
+
+	netif_stop_queue(dev);
+
+	/* Stop traffic, mark us closed */
+	cas_lock_all_save(cp, flags);
+	cp->opened = 0;
+	cas_reset(cp, 0);
+	cas_phy_init(cp);
+	cas_begin_auto_negotiation(cp, NULL);
+	cas_clean_rings(cp);
+	cas_unlock_all_restore(cp, flags);
+
+	free_irq(cp->pdev->irq, (void *) dev);
+	cas_spare_free(cp);
+	cas_free_rxds(cp);
+	cas_tx_tiny_free(cp);
+	mutex_unlock(&cp->pm_mutex);
+	return 0;
+}
+
+static struct {
+	const char name[ETH_GSTRING_LEN];
+} ethtool_cassini_statnames[] = {
+	{"collisions"},
+	{"rx_bytes"},
+	{"rx_crc_errors"},
+	{"rx_dropped"},
+	{"rx_errors"},
+	{"rx_fifo_errors"},
+	{"rx_frame_errors"},
+	{"rx_length_errors"},
+	{"rx_over_errors"},
+	{"rx_packets"},
+	{"tx_aborted_errors"},
+	{"tx_bytes"},
+	{"tx_dropped"},
+	{"tx_errors"},
+	{"tx_fifo_errors"},
+	{"tx_packets"}
+};
+#define CAS_NUM_STAT_KEYS ARRAY_SIZE(ethtool_cassini_statnames)
+
+static struct {
+	const int offsets;	/* neg. values for 2nd arg to cas_read_phy */
+} ethtool_register_table[] = {
+	{-MII_BMSR},
+	{-MII_BMCR},
+	{REG_CAWR},
+	{REG_INF_BURST},
+	{REG_BIM_CFG},
+	{REG_RX_CFG},
+	{REG_HP_CFG},
+	{REG_MAC_TX_CFG},
+	{REG_MAC_RX_CFG},
+	{REG_MAC_CTRL_CFG},
+	{REG_MAC_XIF_CFG},
+	{REG_MIF_CFG},
+	{REG_PCS_CFG},
+	{REG_SATURN_PCFG},
+	{REG_PCS_MII_STATUS},
+	{REG_PCS_STATE_MACHINE},
+	{REG_MAC_COLL_EXCESS},
+	{REG_MAC_COLL_LATE}
+};
+#define CAS_REG_LEN 	ARRAY_SIZE(ethtool_register_table)
+#define CAS_MAX_REGS 	(sizeof (u32)*CAS_REG_LEN)
+
+static void cas_read_regs(struct cas *cp, u8 *ptr, int len)
+{
+	u8 *p;
+	int i;
+	unsigned long flags;
+
+	spin_lock_irqsave(&cp->lock, flags);
+	for (i = 0, p = ptr; i < len ; i ++, p += sizeof(u32)) {
+		u16 hval;
+		u32 val;
+		if (ethtool_register_table[i].offsets < 0) {
+			hval = cas_phy_read(cp,
+				    -ethtool_register_table[i].offsets);
+			val = hval;
+		} else {
+			val= readl(cp->regs+ethtool_register_table[i].offsets);
+		}
+		memcpy(p, (u8 *)&val, sizeof(u32));
+	}
+	spin_unlock_irqrestore(&cp->lock, flags);
+}
+
+static struct net_device_stats *cas_get_stats(struct net_device *dev)
+{
+	struct cas *cp = netdev_priv(dev);
+	struct net_device_stats *stats = cp->net_stats;
+	unsigned long flags;
+	int i;
+	unsigned long tmp;
+
+	/* we collate all of the stats into net_stats[N_TX_RING] */
+	if (!cp->hw_running)
+		return stats + N_TX_RINGS;
+
+	/* collect outstanding stats */
+	/* WTZ: the Cassini spec gives these as 16 bit counters but
+	 * stored in 32-bit words.  Added a mask of 0xffff to be safe,
+	 * in case the chip somehow puts any garbage in the other bits.
+	 * Also, counter usage didn't seem to mach what Adrian did
+	 * in the parts of the code that set these quantities. Made
+	 * that consistent.
+	 */
+	spin_lock_irqsave(&cp->stat_lock[N_TX_RINGS], flags);
+	stats[N_TX_RINGS].rx_crc_errors +=
+	  readl(cp->regs + REG_MAC_FCS_ERR) & 0xffff;
+	stats[N_TX_RINGS].rx_frame_errors +=
+		readl(cp->regs + REG_MAC_ALIGN_ERR) &0xffff;
+	stats[N_TX_RINGS].rx_length_errors +=
+		readl(cp->regs + REG_MAC_LEN_ERR) & 0xffff;
+#if 1
+	tmp = (readl(cp->regs + REG_MAC_COLL_EXCESS) & 0xffff) +
+		(readl(cp->regs + REG_MAC_COLL_LATE) & 0xffff);
+	stats[N_TX_RINGS].tx_aborted_errors += tmp;
+	stats[N_TX_RINGS].collisions +=
+	  tmp + (readl(cp->regs + REG_MAC_COLL_NORMAL) & 0xffff);
+#else
+	stats[N_TX_RINGS].tx_aborted_errors +=
+		readl(cp->regs + REG_MAC_COLL_EXCESS);
+	stats[N_TX_RINGS].collisions += readl(cp->regs + REG_MAC_COLL_EXCESS) +
+		readl(cp->regs + REG_MAC_COLL_LATE);
+#endif
+	cas_clear_mac_err(cp);
+
+	/* saved bits that are unique to ring 0 */
+	spin_lock(&cp->stat_lock[0]);
+	stats[N_TX_RINGS].collisions        += stats[0].collisions;
+	stats[N_TX_RINGS].rx_over_errors    += stats[0].rx_over_errors;
+	stats[N_TX_RINGS].rx_frame_errors   += stats[0].rx_frame_errors;
+	stats[N_TX_RINGS].rx_fifo_errors    += stats[0].rx_fifo_errors;
+	stats[N_TX_RINGS].tx_aborted_errors += stats[0].tx_aborted_errors;
+	stats[N_TX_RINGS].tx_fifo_errors    += stats[0].tx_fifo_errors;
+	spin_unlock(&cp->stat_lock[0]);
+
+	for (i = 0; i < N_TX_RINGS; i++) {
+		spin_lock(&cp->stat_lock[i]);
+		stats[N_TX_RINGS].rx_length_errors +=
+			stats[i].rx_length_errors;
+		stats[N_TX_RINGS].rx_crc_errors += stats[i].rx_crc_errors;
+		stats[N_TX_RINGS].rx_packets    += stats[i].rx_packets;
+		stats[N_TX_RINGS].tx_packets    += stats[i].tx_packets;
+		stats[N_TX_RINGS].rx_bytes      += stats[i].rx_bytes;
+		stats[N_TX_RINGS].tx_bytes      += stats[i].tx_bytes;
+		stats[N_TX_RINGS].rx_errors     += stats[i].rx_errors;
+		stats[N_TX_RINGS].tx_errors     += stats[i].tx_errors;
+		stats[N_TX_RINGS].rx_dropped    += stats[i].rx_dropped;
+		stats[N_TX_RINGS].tx_dropped    += stats[i].tx_dropped;
+		memset(stats + i, 0, sizeof(struct net_device_stats));
+		spin_unlock(&cp->stat_lock[i]);
+	}
+	spin_unlock_irqrestore(&cp->stat_lock[N_TX_RINGS], flags);
+	return stats + N_TX_RINGS;
+}
+
+
+static void cas_set_multicast(struct net_device *dev)
+{
+	struct cas *cp = netdev_priv(dev);
+	u32 rxcfg, rxcfg_new;
+	unsigned long flags;
+	int limit = STOP_TRIES;
+
+	if (!cp->hw_running)
+		return;
+
+	spin_lock_irqsave(&cp->lock, flags);
+	rxcfg = readl(cp->regs + REG_MAC_RX_CFG);
+
+	/* disable RX MAC and wait for completion */
+	writel(rxcfg & ~MAC_RX_CFG_EN, cp->regs + REG_MAC_RX_CFG);
+	while (readl(cp->regs + REG_MAC_RX_CFG) & MAC_RX_CFG_EN) {
+		if (!limit--)
+			break;
+		udelay(10);
+	}
+
+	/* disable hash filter and wait for completion */
+	limit = STOP_TRIES;
+	rxcfg &= ~(MAC_RX_CFG_PROMISC_EN | MAC_RX_CFG_HASH_FILTER_EN);
+	writel(rxcfg & ~MAC_RX_CFG_EN, cp->regs + REG_MAC_RX_CFG);
+	while (readl(cp->regs + REG_MAC_RX_CFG) & MAC_RX_CFG_HASH_FILTER_EN) {
+		if (!limit--)
+			break;
+		udelay(10);
+	}
+
+	/* program hash filters */
+	cp->mac_rx_cfg = rxcfg_new = cas_setup_multicast(cp);
+	rxcfg |= rxcfg_new;
+	writel(rxcfg, cp->regs + REG_MAC_RX_CFG);
+	spin_unlock_irqrestore(&cp->lock, flags);
+}
+
+static void cas_get_drvinfo(struct net_device *dev, struct ethtool_drvinfo *info)
+{
+	struct cas *cp = netdev_priv(dev);
+	strscpy(info->driver, DRV_MODULE_NAME, sizeof(info->driver));
+	strscpy(info->version, DRV_MODULE_VERSION, sizeof(info->version));
+	strscpy(info->bus_info, pci_name(cp->pdev), sizeof(info->bus_info));
+}
+
+static int cas_get_link_ksettings(struct net_device *dev,
+				  struct ethtool_link_ksettings *cmd)
+{
+	struct cas *cp = netdev_priv(dev);
+	u16 bmcr;
+	int full_duplex, speed, pause;
+	unsigned long flags;
+	enum link_state linkstate = link_up;
+	u32 supported, advertising;
+
+	advertising = 0;
+	supported = SUPPORTED_Autoneg;
+	if (cp->cas_flags & CAS_FLAG_1000MB_CAP) {
+		supported |= SUPPORTED_1000baseT_Full;
+		advertising |= ADVERTISED_1000baseT_Full;
+	}
+
+	/* Record PHY settings if HW is on. */
+	spin_lock_irqsave(&cp->lock, flags);
+	bmcr = 0;
+	linkstate = cp->lstate;
+	if (CAS_PHY_MII(cp->phy_type)) {
+		cmd->base.port = PORT_MII;
+		cmd->base.phy_address = cp->phy_addr;
+		advertising |= ADVERTISED_TP | ADVERTISED_MII |
+			ADVERTISED_10baseT_Half |
+			ADVERTISED_10baseT_Full |
+			ADVERTISED_100baseT_Half |
+			ADVERTISED_100baseT_Full;
+
+		supported |=
+			(SUPPORTED_10baseT_Half |
+			 SUPPORTED_10baseT_Full |
+			 SUPPORTED_100baseT_Half |
+			 SUPPORTED_100baseT_Full |
+			 SUPPORTED_TP | SUPPORTED_MII);
+
+		if (cp->hw_running) {
+			cas_mif_poll(cp, 0);
+			bmcr = cas_phy_read(cp, MII_BMCR);
+			cas_read_mii_link_mode(cp, &full_duplex,
+					       &speed, &pause);
+			cas_mif_poll(cp, 1);
+		}
+
+	} else {
+		cmd->base.port = PORT_FIBRE;
+		cmd->base.phy_address = 0;
+		supported   |= SUPPORTED_FIBRE;
+		advertising |= ADVERTISED_FIBRE;
+
+		if (cp->hw_running) {
+			/* pcs uses the same bits as mii */
+			bmcr = readl(cp->regs + REG_PCS_MII_CTRL);
+			cas_read_pcs_link_mode(cp, &full_duplex,
+					       &speed, &pause);
+		}
+	}
+	spin_unlock_irqrestore(&cp->lock, flags);
+
+	if (bmcr & BMCR_ANENABLE) {
+		advertising |= ADVERTISED_Autoneg;
+		cmd->base.autoneg = AUTONEG_ENABLE;
+		cmd->base.speed =  ((speed == 10) ?
+					    SPEED_10 :
+					    ((speed == 1000) ?
+					     SPEED_1000 : SPEED_100));
+		cmd->base.duplex = full_duplex ? DUPLEX_FULL : DUPLEX_HALF;
+	} else {
+		cmd->base.autoneg = AUTONEG_DISABLE;
+		cmd->base.speed = ((bmcr & CAS_BMCR_SPEED1000) ?
+					    SPEED_1000 :
+					    ((bmcr & BMCR_SPEED100) ?
+					     SPEED_100 : SPEED_10));
+		cmd->base.duplex = (bmcr & BMCR_FULLDPLX) ?
+			DUPLEX_FULL : DUPLEX_HALF;
+	}
+	if (linkstate != link_up) {
+		/* Force these to "unknown" if the link is not up and
+		 * autonegotiation is enabled. We can set the link
+		 * speed to 0, but not cmd->duplex,
+		 * because its legal values are 0 and 1.  Ethtool will
+		 * print the value reported in parentheses after the
+		 * word "Unknown" for unrecognized values.
+		 *
+		 * If in forced mode, we report the speed and duplex
+		 * settings that we configured.
+		 */
+		if (cp->link_cntl & BMCR_ANENABLE) {
+			cmd->base.speed = 0;
+			cmd->base.duplex = 0xff;
+		} else {
+			cmd->base.speed = SPEED_10;
+			if (cp->link_cntl & BMCR_SPEED100) {
+				cmd->base.speed = SPEED_100;
+			} else if (cp->link_cntl & CAS_BMCR_SPEED1000) {
+				cmd->base.speed = SPEED_1000;
+			}
+			cmd->base.duplex = (cp->link_cntl & BMCR_FULLDPLX) ?
+				DUPLEX_FULL : DUPLEX_HALF;
+		}
+	}
+
+	ethtool_convert_legacy_u32_to_link_mode(cmd->link_modes.supported,
+						supported);
+	ethtool_convert_legacy_u32_to_link_mode(cmd->link_modes.advertising,
+						advertising);
+
+	return 0;
+}
+
+static int cas_set_link_ksettings(struct net_device *dev,
+				  const struct ethtool_link_ksettings *cmd)
+{
+	struct cas *cp = netdev_priv(dev);
+	unsigned long flags;
+	u32 speed = cmd->base.speed;
+
+	/* Verify the settings we care about. */
+	if (cmd->base.autoneg != AUTONEG_ENABLE &&
+	    cmd->base.autoneg != AUTONEG_DISABLE)
+		return -EINVAL;
+
+	if (cmd->base.autoneg == AUTONEG_DISABLE &&
+	    ((speed != SPEED_1000 &&
+	      speed != SPEED_100 &&
+	      speed != SPEED_10) ||
+	     (cmd->base.duplex != DUPLEX_HALF &&
+	      cmd->base.duplex != DUPLEX_FULL)))
+		return -EINVAL;
+
+	/* Apply settings and restart link process. */
+	spin_lock_irqsave(&cp->lock, flags);
+	cas_begin_auto_negotiation(cp, cmd);
+	spin_unlock_irqrestore(&cp->lock, flags);
+	return 0;
+}
+
+static int cas_nway_reset(struct net_device *dev)
+{
+	struct cas *cp = netdev_priv(dev);
+	unsigned long flags;
+
+	if ((cp->link_cntl & BMCR_ANENABLE) == 0)
+		return -EINVAL;
+
+	/* Restart link process. */
+	spin_lock_irqsave(&cp->lock, flags);
+	cas_begin_auto_negotiation(cp, NULL);
+	spin_unlock_irqrestore(&cp->lock, flags);
+
+	return 0;
+}
+
+static u32 cas_get_link(struct net_device *dev)
+{
+	struct cas *cp = netdev_priv(dev);
+	return cp->lstate == link_up;
+}
+
+static u32 cas_get_msglevel(struct net_device *dev)
+{
+	struct cas *cp = netdev_priv(dev);
+	return cp->msg_enable;
+}
+
+static void cas_set_msglevel(struct net_device *dev, u32 value)
+{
+	struct cas *cp = netdev_priv(dev);
+	cp->msg_enable = value;
+}
+
+static int cas_get_regs_len(struct net_device *dev)
+{
+	struct cas *cp = netdev_priv(dev);
+	return min_t(int, cp->casreg_len, CAS_MAX_REGS);
+}
+
+static void cas_get_regs(struct net_device *dev, struct ethtool_regs *regs,
+			     void *p)
+{
+	struct cas *cp = netdev_priv(dev);
+	regs->version = 0;
+	/* cas_read_regs handles locks (cp->lock).  */
+	cas_read_regs(cp, p, regs->len / sizeof(u32));
+}
+
+static int cas_get_sset_count(struct net_device *dev, int sset)
+{
+	switch (sset) {
+	case ETH_SS_STATS:
+		return CAS_NUM_STAT_KEYS;
+	default:
+		return -EOPNOTSUPP;
+	}
+}
+
+static void cas_get_strings(struct net_device *dev, u32 stringset, u8 *data)
+{
+	 memcpy(data, &ethtool_cassini_statnames,
+					 CAS_NUM_STAT_KEYS * ETH_GSTRING_LEN);
+}
+
+static void cas_get_ethtool_stats(struct net_device *dev,
+				      struct ethtool_stats *estats, u64 *data)
+{
+	struct cas *cp = netdev_priv(dev);
+	struct net_device_stats *stats = cas_get_stats(cp->dev);
+	int i = 0;
+	data[i++] = stats->collisions;
+	data[i++] = stats->rx_bytes;
+	data[i++] = stats->rx_crc_errors;
+	data[i++] = stats->rx_dropped;
+	data[i++] = stats->rx_errors;
+	data[i++] = stats->rx_fifo_errors;
+	data[i++] = stats->rx_frame_errors;
+	data[i++] = stats->rx_length_errors;
+	data[i++] = stats->rx_over_errors;
+	data[i++] = stats->rx_packets;
+	data[i++] = stats->tx_aborted_errors;
+	data[i++] = stats->tx_bytes;
+	data[i++] = stats->tx_dropped;
+	data[i++] = stats->tx_errors;
+	data[i++] = stats->tx_fifo_errors;
+	data[i++] = stats->tx_packets;
+	BUG_ON(i != CAS_NUM_STAT_KEYS);
+}
+
+static const struct ethtool_ops cas_ethtool_ops = {
+	.get_drvinfo		= cas_get_drvinfo,
+	.nway_reset		= cas_nway_reset,
+	.get_link		= cas_get_link,
+	.get_msglevel		= cas_get_msglevel,
+	.set_msglevel		= cas_set_msglevel,
+	.get_regs_len		= cas_get_regs_len,
+	.get_regs		= cas_get_regs,
+	.get_sset_count		= cas_get_sset_count,
+	.get_strings		= cas_get_strings,
+	.get_ethtool_stats	= cas_get_ethtool_stats,
+	.get_link_ksettings	= cas_get_link_ksettings,
+	.set_link_ksettings	= cas_set_link_ksettings,
+};
+
+static int cas_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
+{
+	struct cas *cp = netdev_priv(dev);
+	struct mii_ioctl_data *data = if_mii(ifr);
+	unsigned long flags;
+	int rc = -EOPNOTSUPP;
+
+	/* Hold the PM mutex while doing ioctl's or we may collide
+	 * with open/close and power management and oops.
+	 */
+	mutex_lock(&cp->pm_mutex);
+	switch (cmd) {
+	case SIOCGMIIPHY:		/* Get address of MII PHY in use. */
+		data->phy_id = cp->phy_addr;
+		fallthrough;
+
+	case SIOCGMIIREG:		/* Read MII PHY register. */
+		spin_lock_irqsave(&cp->lock, flags);
+		cas_mif_poll(cp, 0);
+		data->val_out = cas_phy_read(cp, data->reg_num & 0x1f);
+		cas_mif_poll(cp, 1);
+		spin_unlock_irqrestore(&cp->lock, flags);
+		rc = 0;
+		break;
+
+	case SIOCSMIIREG:		/* Write MII PHY register. */
+		spin_lock_irqsave(&cp->lock, flags);
+		cas_mif_poll(cp, 0);
+		rc = cas_phy_write(cp, data->reg_num & 0x1f, data->val_in);
+		cas_mif_poll(cp, 1);
+		spin_unlock_irqrestore(&cp->lock, flags);
+		break;
+	default:
+		break;
+	}
+
+	mutex_unlock(&cp->pm_mutex);
+	return rc;
+}
+
+/* When this chip sits underneath an Intel 31154 bridge, it is the
+ * only subordinate device and we can tweak the bridge settings to
+ * reflect that fact.
+ */
+static void cas_program_bridge(struct pci_dev *cas_pdev)
+{
+	struct pci_dev *pdev = cas_pdev->bus->self;
+	u32 val;
+
+	if (!pdev)
+		return;
+
+	if (pdev->vendor != 0x8086 || pdev->device != 0x537c)
+		return;
+
+	/* Clear bit 10 (Bus Parking Control) in the Secondary
+	 * Arbiter Control/Status Register which lives at offset
+	 * 0x41.  Using a 32-bit word read/modify/write at 0x40
+	 * is much simpler so that's how we do this.
+	 */
+	pci_read_config_dword(pdev, 0x40, &val);
+	val &= ~0x00040000;
+	pci_write_config_dword(pdev, 0x40, val);
+
+	/* Max out the Multi-Transaction Timer settings since
+	 * Cassini is the only device present.
+	 *
+	 * The register is 16-bit and lives at 0x50.  When the
+	 * settings are enabled, it extends the GRANT# signal
+	 * for a requestor after a transaction is complete.  This
+	 * allows the next request to run without first needing
+	 * to negotiate the GRANT# signal back.
+	 *
+	 * Bits 12:10 define the grant duration:
+	 *
+	 *	1	--	16 clocks
+	 *	2	--	32 clocks
+	 *	3	--	64 clocks
+	 *	4	--	128 clocks
+	 *	5	--	256 clocks
+	 *
+	 * All other values are illegal.
+	 *
+	 * Bits 09:00 define which REQ/GNT signal pairs get the
+	 * GRANT# signal treatment.  We set them all.
+	 */
+	pci_write_config_word(pdev, 0x50, (5 << 10) | 0x3ff);
+
+	/* The Read Prefetch Policy register is 16-bit and sits at
+	 * offset 0x52.  It enables a "smart" pre-fetch policy.  We
+	 * enable it and max out all of the settings since only one
+	 * device is sitting underneath and thus bandwidth sharing is
+	 * not an issue.
+	 *
+	 * The register has several 3 bit fields, which indicates a
+	 * multiplier applied to the base amount of prefetching the
+	 * chip would do.  These fields are at:
+	 *
+	 *	15:13	---	ReRead Primary Bus
+	 *	12:10	---	FirstRead Primary Bus
+	 *	09:07	---	ReRead Secondary Bus
+	 *	06:04	---	FirstRead Secondary Bus
+	 *
+	 * Bits 03:00 control which REQ/GNT pairs the prefetch settings
+	 * get enabled on.  Bit 3 is a grouped enabler which controls
+	 * all of the REQ/GNT pairs from [8:3].  Bits 2 to 0 control
+	 * the individual REQ/GNT pairs [2:0].
+	 */
+	pci_write_config_word(pdev, 0x52,
+			      (0x7 << 13) |
+			      (0x7 << 10) |
+			      (0x7 <<  7) |
+			      (0x7 <<  4) |
+			      (0xf <<  0));
+
+	/* Force cacheline size to 0x8 */
+	pci_write_config_byte(pdev, PCI_CACHE_LINE_SIZE, 0x08);
+
+	/* Force latency timer to maximum setting so Cassini can
+	 * sit on the bus as long as it likes.
+	 */
+	pci_write_config_byte(pdev, PCI_LATENCY_TIMER, 0xff);
+}
+
+static const struct net_device_ops cas_netdev_ops = {
+	.ndo_open		= cas_open,
+	.ndo_stop		= cas_close,
+	.ndo_start_xmit		= cas_start_xmit,
+	.ndo_get_stats 		= cas_get_stats,
+	.ndo_set_rx_mode	= cas_set_multicast,
+	.ndo_eth_ioctl		= cas_ioctl,
+	.ndo_tx_timeout		= cas_tx_timeout,
+	.ndo_change_mtu		= cas_change_mtu,
+	.ndo_set_mac_address	= eth_mac_addr,
+	.ndo_validate_addr	= eth_validate_addr,
+#ifdef CONFIG_NET_POLL_CONTROLLER
+	.ndo_poll_controller	= cas_netpoll,
+#endif
+};
+
+static int cas_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
+{
+	static int cas_version_printed = 0;
+	unsigned long casreg_len;
+	struct net_device *dev;
+	struct cas *cp;
+	u16 pci_cmd;
+	int i, err;
+	u8 orig_cacheline_size = 0, cas_cacheline_size = 0;
+
+	if (cas_version_printed++ == 0)
+		pr_info("%s", version);
+
+	err = pci_enable_device(pdev);
+	if (err) {
+		dev_err(&pdev->dev, "Cannot enable PCI device, aborting\n");
+		return err;
+	}
+
+	if (!(pci_resource_flags(pdev, 0) & IORESOURCE_MEM)) {
+		dev_err(&pdev->dev, "Cannot find proper PCI device "
+		       "base address, aborting\n");
+		err = -ENODEV;
+		goto err_out_disable_pdev;
+	}
+
+	dev = alloc_etherdev(sizeof(*cp));
+	if (!dev) {
+		err = -ENOMEM;
+		goto err_out_disable_pdev;
+	}
+	SET_NETDEV_DEV(dev, &pdev->dev);
+
+	err = pci_request_regions(pdev, dev->name);
+	if (err) {
+		dev_err(&pdev->dev, "Cannot obtain PCI resources, aborting\n");
+		goto err_out_free_netdev;
+	}
+	pci_set_master(pdev);
+
+	/* we must always turn on parity response or else parity
+	 * doesn't get generated properly. disable SERR/PERR as well.
+	 * in addition, we want to turn MWI on.
+	 */
+	pci_read_config_word(pdev, PCI_COMMAND, &pci_cmd);
+	pci_cmd &= ~PCI_COMMAND_SERR;
+	pci_cmd |= PCI_COMMAND_PARITY;
+	pci_write_config_word(pdev, PCI_COMMAND, pci_cmd);
+	if (pci_try_set_mwi(pdev))
+		pr_warn("Could not enable MWI for %s\n", pci_name(pdev));
+
+	cas_program_bridge(pdev);
+
+	/*
+	 * On some architectures, the default cache line size set
+	 * by pci_try_set_mwi reduces performance.  We have to increase
+	 * it for this case.  To start, we'll print some configuration
+	 * data.
+	 */
+#if 1
+	pci_read_config_byte(pdev, PCI_CACHE_LINE_SIZE,
+			     &orig_cacheline_size);
+	if (orig_cacheline_size < CAS_PREF_CACHELINE_SIZE) {
+		cas_cacheline_size =
+			(CAS_PREF_CACHELINE_SIZE < SMP_CACHE_BYTES) ?
+			CAS_PREF_CACHELINE_SIZE : SMP_CACHE_BYTES;
+		if (pci_write_config_byte(pdev,
+					  PCI_CACHE_LINE_SIZE,
+					  cas_cacheline_size)) {
+			dev_err(&pdev->dev, "Could not set PCI cache "
+			       "line size\n");
+			goto err_out_free_res;
+		}
+	}
+#endif
+
+
+	/* Configure DMA attributes. */
+	err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64));
+	if (err) {
+		dev_err(&pdev->dev, "No usable DMA configuration, aborting\n");
+		goto err_out_free_res;
+	}
+
+	casreg_len = pci_resource_len(pdev, 0);
+
+	cp = netdev_priv(dev);
+	cp->pdev = pdev;
+#if 1
+	/* A value of 0 indicates we never explicitly set it */
+	cp->orig_cacheline_size = cas_cacheline_size ? orig_cacheline_size: 0;
+#endif
+	cp->dev = dev;
+	cp->msg_enable = (cassini_debug < 0) ? CAS_DEF_MSG_ENABLE :
+	  cassini_debug;
+
+#if defined(CONFIG_SPARC)
+	cp->of_node = pci_device_to_OF_node(pdev);
+#endif
+
+	cp->link_transition = LINK_TRANSITION_UNKNOWN;
+	cp->link_transition_jiffies_valid = 0;
+
+	spin_lock_init(&cp->lock);
+	spin_lock_init(&cp->rx_inuse_lock);
+	spin_lock_init(&cp->rx_spare_lock);
+	for (i = 0; i < N_TX_RINGS; i++) {
+		spin_lock_init(&cp->stat_lock[i]);
+		spin_lock_init(&cp->tx_lock[i]);
+	}
+	spin_lock_init(&cp->stat_lock[N_TX_RINGS]);
+	mutex_init(&cp->pm_mutex);
+
+	timer_setup(&cp->link_timer, cas_link_timer, 0);
+
+#if 1
+	/* Just in case the implementation of atomic operations
+	 * change so that an explicit initialization is necessary.
+	 */
+	atomic_set(&cp->reset_task_pending, 0);
+	atomic_set(&cp->reset_task_pending_all, 0);
+	atomic_set(&cp->reset_task_pending_spare, 0);
+	atomic_set(&cp->reset_task_pending_mtu, 0);
+#endif
+	INIT_WORK(&cp->reset_task, cas_reset_task);
+
+	/* Default link parameters */
+	if (link_mode >= 0 && link_mode < 6)
+		cp->link_cntl = link_modes[link_mode];
+	else
+		cp->link_cntl = BMCR_ANENABLE;
+	cp->lstate = link_down;
+	cp->link_transition = LINK_TRANSITION_LINK_DOWN;
+	netif_carrier_off(cp->dev);
+	cp->timer_ticks = 0;
+
+	/* give us access to cassini registers */
+	cp->regs = pci_iomap(pdev, 0, casreg_len);
+	if (!cp->regs) {
+		dev_err(&pdev->dev, "Cannot map device registers, aborting\n");
+		goto err_out_free_res;
+	}
+	cp->casreg_len = casreg_len;
+
+	pci_save_state(pdev);
+	cas_check_pci_invariants(cp);
+	cas_hard_reset(cp);
+	cas_reset(cp, 0);
+	if (cas_check_invariants(cp))
+		goto err_out_iounmap;
+	if (cp->cas_flags & CAS_FLAG_SATURN)
+		cas_saturn_firmware_init(cp);
+
+	cp->init_block =
+		dma_alloc_coherent(&pdev->dev, sizeof(struct cas_init_block),
+				   &cp->block_dvma, GFP_KERNEL);
+	if (!cp->init_block) {
+		dev_err(&pdev->dev, "Cannot allocate init block, aborting\n");
+		goto err_out_iounmap;
+	}
+
+	for (i = 0; i < N_TX_RINGS; i++)
+		cp->init_txds[i] = cp->init_block->txds[i];
+
+	for (i = 0; i < N_RX_DESC_RINGS; i++)
+		cp->init_rxds[i] = cp->init_block->rxds[i];
+
+	for (i = 0; i < N_RX_COMP_RINGS; i++)
+		cp->init_rxcs[i] = cp->init_block->rxcs[i];
+
+	for (i = 0; i < N_RX_FLOWS; i++)
+		skb_queue_head_init(&cp->rx_flows[i]);
+
+	dev->netdev_ops = &cas_netdev_ops;
+	dev->ethtool_ops = &cas_ethtool_ops;
+	dev->watchdog_timeo = CAS_TX_TIMEOUT;
+
+#ifdef USE_NAPI
+	netif_napi_add(dev, &cp->napi, cas_poll);
+#endif
+	dev->irq = pdev->irq;
+	dev->dma = 0;
+
+	/* Cassini features. */
+	if ((cp->cas_flags & CAS_FLAG_NO_HW_CSUM) == 0)
+		dev->features |= NETIF_F_HW_CSUM | NETIF_F_SG;
+
+	dev->features |= NETIF_F_HIGHDMA;
+
+	/* MTU range: 60 - varies or 9000 */
+	dev->min_mtu = CAS_MIN_MTU;
+	dev->max_mtu = CAS_MAX_MTU;
+
+	if (register_netdev(dev)) {
+		dev_err(&pdev->dev, "Cannot register net device, aborting\n");
+		goto err_out_free_consistent;
+	}
+
+	i = readl(cp->regs + REG_BIM_CFG);
+	netdev_info(dev, "Sun Cassini%s (%sbit/%sMHz PCI/%s) Ethernet[%d] %pM\n",
+		    (cp->cas_flags & CAS_FLAG_REG_PLUS) ? "+" : "",
+		    (i & BIM_CFG_32BIT) ? "32" : "64",
+		    (i & BIM_CFG_66MHZ) ? "66" : "33",
+		    (cp->phy_type == CAS_PHY_SERDES) ? "Fi" : "Cu", pdev->irq,
+		    dev->dev_addr);
+
+	pci_set_drvdata(pdev, dev);
+	cp->hw_running = 1;
+	cas_entropy_reset(cp);
+	cas_phy_init(cp);
+	cas_begin_auto_negotiation(cp, NULL);
+	return 0;
+
+err_out_free_consistent:
+	dma_free_coherent(&pdev->dev, sizeof(struct cas_init_block),
+			  cp->init_block, cp->block_dvma);
+
+err_out_iounmap:
+	mutex_lock(&cp->pm_mutex);
+	if (cp->hw_running)
+		cas_shutdown(cp);
+	mutex_unlock(&cp->pm_mutex);
+
+	vfree(cp->fw_data);
+
+	pci_iounmap(pdev, cp->regs);
+
+
+err_out_free_res:
+	pci_release_regions(pdev);
+
+	/* Try to restore it in case the error occurred after we
+	 * set it.
+	 */
+	pci_write_config_byte(pdev, PCI_CACHE_LINE_SIZE, orig_cacheline_size);
+
+err_out_free_netdev:
+	free_netdev(dev);
+
+err_out_disable_pdev:
+	pci_disable_device(pdev);
+	return -ENODEV;
+}
+
+static void cas_remove_one(struct pci_dev *pdev)
+{
+	struct net_device *dev = pci_get_drvdata(pdev);
+	struct cas *cp;
+	if (!dev)
+		return;
+
+	cp = netdev_priv(dev);
+	unregister_netdev(dev);
+
+	vfree(cp->fw_data);
+
+	mutex_lock(&cp->pm_mutex);
+	cancel_work_sync(&cp->reset_task);
+	if (cp->hw_running)
+		cas_shutdown(cp);
+	mutex_unlock(&cp->pm_mutex);
+
+#if 1
+	if (cp->orig_cacheline_size) {
+		/* Restore the cache line size if we had modified
+		 * it.
+		 */
+		pci_write_config_byte(pdev, PCI_CACHE_LINE_SIZE,
+				      cp->orig_cacheline_size);
+	}
+#endif
+	dma_free_coherent(&pdev->dev, sizeof(struct cas_init_block),
+			  cp->init_block, cp->block_dvma);
+	pci_iounmap(pdev, cp->regs);
+	free_netdev(dev);
+	pci_release_regions(pdev);
+	pci_disable_device(pdev);
+}
+
+static int __maybe_unused cas_suspend(struct device *dev_d)
+{
+	struct net_device *dev = dev_get_drvdata(dev_d);
+	struct cas *cp = netdev_priv(dev);
+	unsigned long flags;
+
+	mutex_lock(&cp->pm_mutex);
+
+	/* If the driver is opened, we stop the DMA */
+	if (cp->opened) {
+		netif_device_detach(dev);
+
+		cas_lock_all_save(cp, flags);
+
+		/* We can set the second arg of cas_reset to 0
+		 * because on resume, we'll call cas_init_hw with
+		 * its second arg set so that autonegotiation is
+		 * restarted.
+		 */
+		cas_reset(cp, 0);
+		cas_clean_rings(cp);
+		cas_unlock_all_restore(cp, flags);
+	}
+
+	if (cp->hw_running)
+		cas_shutdown(cp);
+	mutex_unlock(&cp->pm_mutex);
+
+	return 0;
+}
+
+static int __maybe_unused cas_resume(struct device *dev_d)
+{
+	struct net_device *dev = dev_get_drvdata(dev_d);
+	struct cas *cp = netdev_priv(dev);
+
+	netdev_info(dev, "resuming\n");
+
+	mutex_lock(&cp->pm_mutex);
+	cas_hard_reset(cp);
+	if (cp->opened) {
+		unsigned long flags;
+		cas_lock_all_save(cp, flags);
+		cas_reset(cp, 0);
+		cp->hw_running = 1;
+		cas_clean_rings(cp);
+		cas_init_hw(cp, 1);
+		cas_unlock_all_restore(cp, flags);
+
+		netif_device_attach(dev);
+	}
+	mutex_unlock(&cp->pm_mutex);
+	return 0;
+}
+
+static SIMPLE_DEV_PM_OPS(cas_pm_ops, cas_suspend, cas_resume);
+
+static struct pci_driver cas_driver = {
+	.name		= DRV_MODULE_NAME,
+	.id_table	= cas_pci_tbl,
+	.probe		= cas_init_one,
+	.remove		= cas_remove_one,
+	.driver.pm	= &cas_pm_ops,
+};
+
+static int __init cas_init(void)
+{
+	if (linkdown_timeout > 0)
+		link_transition_timeout = linkdown_timeout * HZ;
+	else
+		link_transition_timeout = 0;
+
+	return pci_register_driver(&cas_driver);
+}
+
+static void __exit cas_cleanup(void)
+{
+	pci_unregister_driver(&cas_driver);
+}
+
+module_init(cas_init);
+module_exit(cas_cleanup);
diff --git a/drivers/net/ethernet/sun/cassini.h b/drivers/net/ethernet/sun/cassini.h
new file mode 100644
index 000000000..0c24547a4
--- /dev/null
+++ b/drivers/net/ethernet/sun/cassini.h
@@ -0,0 +1,2900 @@
+/* SPDX-License-Identifier: GPL-2.0+ */
+/* $Id: cassini.h,v 1.16 2004/08/17 21:15:16 zaumen Exp $
+ * cassini.h: Definitions for Sun Microsystems Cassini(+) ethernet driver.
+ *
+ * Copyright (C) 2004 Sun Microsystems Inc.
+ * Copyright (c) 2003 Adrian Sun (asun@darksunrising.com)
+ *
+ * vendor id: 0x108E (Sun Microsystems, Inc.)
+ * device id: 0xabba (Cassini)
+ * revision ids: 0x01 = Cassini
+ *               0x02 = Cassini rev 2
+ *               0x10 = Cassini+
+ *               0x11 = Cassini+ 0.2u
+ *
+ * vendor id: 0x100b (National Semiconductor)
+ * device id: 0x0035 (DP83065/Saturn)
+ * revision ids: 0x30 = Saturn B2
+ *
+ * rings are all offset from 0.
+ *
+ * there are two clock domains:
+ * PCI:  33/66MHz clock
+ * chip: 125MHz clock
+ */
+
+#ifndef _CASSINI_H
+#define _CASSINI_H
+
+/* cassini register map: 2M memory mapped in 32-bit memory space accessible as
+ * 32-bit words. there is no i/o port access. REG_ addresses are
+ * shared between cassini and cassini+. REG_PLUS_ addresses only
+ * appear in cassini+. REG_MINUS_ addresses only appear in cassini.
+ */
+#define CAS_ID_REV2          0x02
+#define CAS_ID_REVPLUS       0x10
+#define CAS_ID_REVPLUS02u    0x11
+#define CAS_ID_REVSATURNB2   0x30
+
+/** global resources **/
+
+/* this register sets the weights for the weighted round robin arbiter. e.g.,
+ * if rx weight == 1 and tx weight == 0, rx == 2x tx transfer credit
+ * for its next turn to access the pci bus.
+ * map: 0x0 = x1, 0x1 = x2, 0x2 = x4, 0x3 = x8
+ * DEFAULT: 0x0, SIZE: 5 bits
+ */
+#define  REG_CAWR	               0x0004  /* core arbitration weight */
+#define    CAWR_RX_DMA_WEIGHT_SHIFT    0
+#define    CAWR_RX_DMA_WEIGHT_MASK     0x03    /* [0:1] */
+#define    CAWR_TX_DMA_WEIGHT_SHIFT    2
+#define    CAWR_TX_DMA_WEIGHT_MASK     0x0C    /* [3:2] */
+#define    CAWR_RR_DIS                 0x10    /* [4] */
+
+/* if enabled, BIM can send bursts across PCI bus > cacheline size. burst
+ * sizes determined by length of packet or descriptor transfer and the
+ * max length allowed by the target.
+ * DEFAULT: 0x0, SIZE: 1 bit
+ */
+#define  REG_INF_BURST                 0x0008  /* infinite burst enable reg */
+#define    INF_BURST_EN                0x1     /* enable */
+
+/* top level interrupts [0-9] are auto-cleared to 0 when the status
+ * register is read. second level interrupts [13 - 18] are cleared at
+ * the source. tx completion register 3 is replicated in [19 - 31]
+ * DEFAULT: 0x00000000, SIZE: 29 bits
+ */
+#define  REG_INTR_STATUS               0x000C  /* interrupt status register */
+#define    INTR_TX_INTME               0x00000001  /* frame w/ INT ME desc bit set
+						      xferred from host queue to
+						      TX FIFO */
+#define    INTR_TX_ALL                 0x00000002  /* all xmit frames xferred into
+						      TX FIFO. i.e.,
+						      TX Kick == TX complete. if
+						      PACED_MODE set, then TX FIFO
+						      also empty */
+#define    INTR_TX_DONE                0x00000004  /* any frame xferred into tx
+						      FIFO */
+#define    INTR_TX_TAG_ERROR           0x00000008  /* TX FIFO tag framing
+						      corrupted. FATAL ERROR */
+#define    INTR_RX_DONE                0x00000010  /* at least 1 frame xferred
+						      from RX FIFO to host mem.
+						      RX completion reg updated.
+						      may be delayed by recv
+						      intr blanking. */
+#define    INTR_RX_BUF_UNAVAIL         0x00000020  /* no more receive buffers.
+						      RX Kick == RX complete */
+#define    INTR_RX_TAG_ERROR           0x00000040  /* RX FIFO tag framing
+						      corrupted. FATAL ERROR */
+#define    INTR_RX_COMP_FULL           0x00000080  /* no more room in completion
+						      ring to post descriptors.
+						      RX complete head incr to
+						      almost reach RX complete
+						      tail */
+#define    INTR_RX_BUF_AE              0x00000100  /* less than the
+						      programmable threshold #
+						      of free descr avail for
+						      hw use */
+#define    INTR_RX_COMP_AF             0x00000200  /* less than the
+						      programmable threshold #
+						      of descr spaces for hw
+						      use in completion descr
+						      ring */
+#define    INTR_RX_LEN_MISMATCH        0x00000400  /* len field from MAC !=
+						      len of non-reassembly pkt
+						      from fifo during DMA or
+						      header parser provides TCP
+						      header and payload size >
+						      MAC packet size.
+						      FATAL ERROR */
+#define    INTR_SUMMARY                0x00001000  /* summary interrupt bit. this
+						      bit will be set if an interrupt
+						      generated on the pci bus. useful
+						      when driver is polling for
+						      interrupts */
+#define    INTR_PCS_STATUS             0x00002000  /* PCS interrupt status register */
+#define    INTR_TX_MAC_STATUS          0x00004000  /* TX MAC status register has at
+						      least 1 unmasked interrupt set */
+#define    INTR_RX_MAC_STATUS          0x00008000  /* RX MAC status register has at
+						      least 1 unmasked interrupt set */
+#define    INTR_MAC_CTRL_STATUS        0x00010000  /* MAC control status register has
+						      at least 1 unmasked interrupt
+						      set */
+#define    INTR_MIF_STATUS             0x00020000  /* MIF status register has at least
+						      1 unmasked interrupt set */
+#define    INTR_PCI_ERROR_STATUS       0x00040000  /* PCI error status register in the
+						      BIF has at least 1 unmasked
+						      interrupt set */
+#define    INTR_TX_COMP_3_MASK         0xFFF80000  /* mask for TX completion
+						      3 reg data */
+#define    INTR_TX_COMP_3_SHIFT        19
+#define    INTR_ERROR_MASK (INTR_MIF_STATUS | INTR_PCI_ERROR_STATUS | \
+                            INTR_PCS_STATUS | INTR_RX_LEN_MISMATCH | \
+                            INTR_TX_MAC_STATUS | INTR_RX_MAC_STATUS | \
+                            INTR_TX_TAG_ERROR | INTR_RX_TAG_ERROR | \
+                            INTR_MAC_CTRL_STATUS)
+
+/* determines which status events will cause an interrupt. layout same
+ * as REG_INTR_STATUS.
+ * DEFAULT: 0xFFFFFFFF, SIZE: 16 bits
+ */
+#define  REG_INTR_MASK                 0x0010  /* Interrupt mask */
+
+/* top level interrupt bits that are cleared during read of REG_INTR_STATUS_ALIAS.
+ * useful when driver is polling for interrupts. layout same as REG_INTR_MASK.
+ * DEFAULT: 0x00000000, SIZE: 12 bits
+ */
+#define  REG_ALIAS_CLEAR               0x0014  /* alias clear mask
+						  (used w/ status alias) */
+/* same as REG_INTR_STATUS except that only bits cleared are those selected by
+ * REG_ALIAS_CLEAR
+ * DEFAULT: 0x00000000, SIZE: 29 bits
+ */
+#define  REG_INTR_STATUS_ALIAS         0x001C  /* interrupt status alias
+						  (selective clear) */
+
+/* DEFAULT: 0x0, SIZE: 3 bits */
+#define  REG_PCI_ERR_STATUS            0x1000  /* PCI error status */
+#define    PCI_ERR_BADACK              0x01    /* reserved in Cassini+.
+						  set if no ACK64# during ABS64 cycle
+						  in Cassini. */
+#define    PCI_ERR_DTRTO               0x02    /* delayed xaction timeout. set if
+						  no read retry after 2^15 clocks */
+#define    PCI_ERR_OTHER               0x04    /* other PCI errors */
+#define    PCI_ERR_BIM_DMA_WRITE       0x08    /* BIM received 0 count DMA write req.
+						  unused in Cassini. */
+#define    PCI_ERR_BIM_DMA_READ        0x10    /* BIM received 0 count DMA read req.
+						  unused in Cassini. */
+#define    PCI_ERR_BIM_DMA_TIMEOUT     0x20    /* BIM received 255 retries during
+						  DMA. unused in cassini. */
+
+/* mask for PCI status events that will set PCI_ERR_STATUS. if cleared, event
+ * causes an interrupt to be generated.
+ * DEFAULT: 0x7, SIZE: 3 bits
+ */
+#define  REG_PCI_ERR_STATUS_MASK       0x1004  /* PCI Error status mask */
+
+/* used to configure PCI related parameters that are not in PCI config space.
+ * DEFAULT: 0bxx000, SIZE: 5 bits
+ */
+#define  REG_BIM_CFG                0x1008  /* BIM Configuration */
+#define    BIM_CFG_RESERVED0        0x001   /* reserved */
+#define    BIM_CFG_RESERVED1        0x002   /* reserved */
+#define    BIM_CFG_64BIT_DISABLE    0x004   /* disable 64-bit mode */
+#define    BIM_CFG_66MHZ            0x008   /* (ro) 1 = 66MHz, 0 = < 66MHz */
+#define    BIM_CFG_32BIT            0x010   /* (ro) 1 = 32-bit slot, 0 = 64-bit */
+#define    BIM_CFG_DPAR_INTR_ENABLE 0x020   /* detected parity err enable */
+#define    BIM_CFG_RMA_INTR_ENABLE  0x040   /* master abort intr enable */
+#define    BIM_CFG_RTA_INTR_ENABLE  0x080   /* target abort intr enable */
+#define    BIM_CFG_RESERVED2        0x100   /* reserved */
+#define    BIM_CFG_BIM_DISABLE      0x200   /* stop BIM DMA. use before global
+					       reset. reserved in Cassini. */
+#define    BIM_CFG_BIM_STATUS       0x400   /* (ro) 1 = BIM DMA suspended.
+						  reserved in Cassini. */
+#define    BIM_CFG_PERROR_BLOCK     0x800  /* block PERR# to pci bus. def: 0.
+						 reserved in Cassini. */
+
+/* DEFAULT: 0x00000000, SIZE: 32 bits */
+#define  REG_BIM_DIAG                  0x100C  /* BIM Diagnostic */
+#define    BIM_DIAG_MSTR_SM_MASK       0x3FFFFF00 /* PCI master controller state
+						     machine bits [21:0] */
+#define    BIM_DIAG_BRST_SM_MASK       0x7F    /* PCI burst controller state
+						  machine bits [6:0] */
+
+/* writing to SW_RESET_TX and SW_RESET_RX will issue a global
+ * reset. poll until TX and RX read back as 0's for completion.
+ */
+#define  REG_SW_RESET                  0x1010  /* Software reset */
+#define    SW_RESET_TX                 0x00000001  /* reset TX DMA engine. poll until
+						      cleared to 0.  */
+#define    SW_RESET_RX                 0x00000002  /* reset RX DMA engine. poll until
+						      cleared to 0. */
+#define    SW_RESET_RSTOUT             0x00000004  /* force RSTOUT# pin active (low).
+						      resets PHY and anything else
+						      connected to RSTOUT#. RSTOUT#
+						      is also activated by local PCI
+						      reset when hot-swap is being
+						      done. */
+#define    SW_RESET_BLOCK_PCS_SLINK    0x00000008  /* if a global reset is done with
+						      this bit set, PCS and SLINK
+						      modules won't be reset.
+						      i.e., link won't drop. */
+#define    SW_RESET_BREQ_SM_MASK       0x00007F00  /* breq state machine [6:0] */
+#define    SW_RESET_PCIARB_SM_MASK     0x00070000  /* pci arbitration state bits:
+						      0b000: ARB_IDLE1
+						      0b001: ARB_IDLE2
+						      0b010: ARB_WB_ACK
+						      0b011: ARB_WB_WAT
+						      0b100: ARB_RB_ACK
+						      0b101: ARB_RB_WAT
+						      0b110: ARB_RB_END
+						      0b111: ARB_WB_END */
+#define    SW_RESET_RDPCI_SM_MASK      0x00300000  /* read pci state bits:
+						      0b00: RD_PCI_WAT
+						      0b01: RD_PCI_RDY
+						      0b11: RD_PCI_ACK */
+#define    SW_RESET_RDARB_SM_MASK      0x00C00000  /* read arbitration state bits:
+						      0b00: AD_IDL_RX
+						      0b01: AD_ACK_RX
+						      0b10: AD_ACK_TX
+						      0b11: AD_IDL_TX */
+#define    SW_RESET_WRPCI_SM_MASK      0x06000000  /* write pci state bits
+						      0b00: WR_PCI_WAT
+						      0b01: WR_PCI_RDY
+						      0b11: WR_PCI_ACK */
+#define    SW_RESET_WRARB_SM_MASK      0x38000000  /* write arbitration state bits:
+						      0b000: ARB_IDLE1
+						      0b001: ARB_IDLE2
+						      0b010: ARB_TX_ACK
+						      0b011: ARB_TX_WAT
+						      0b100: ARB_RX_ACK
+						      0b110: ARB_RX_WAT */
+
+/* Cassini only. 64-bit register used to check PCI datapath. when read,
+ * value written has both lower and upper 32-bit halves rotated to the right
+ * one bit position. e.g., FFFFFFFF FFFFFFFF -> 7FFFFFFF 7FFFFFFF
+ */
+#define  REG_MINUS_BIM_DATAPATH_TEST   0x1018  /* Cassini: BIM datapath test
+						  Cassini+: reserved */
+
+/* output enables are provided for each device's chip select and for the rest
+ * of the outputs from cassini to its local bus devices. two sw programmable
+ * bits are connected to general purpose control/status bits.
+ * DEFAULT: 0x7
+ */
+#define  REG_BIM_LOCAL_DEV_EN          0x1020  /* BIM local device
+						  output EN. default: 0x7 */
+#define    BIM_LOCAL_DEV_PAD           0x01    /* address bus, RW signal, and
+						  OE signal output enable on the
+						  local bus interface. these
+						  are shared between both local
+						  bus devices. tristate when 0. */
+#define    BIM_LOCAL_DEV_PROM          0x02    /* PROM chip select */
+#define    BIM_LOCAL_DEV_EXT           0x04    /* secondary local bus device chip
+						  select output enable */
+#define    BIM_LOCAL_DEV_SOFT_0        0x08    /* sw programmable ctrl bit 0 */
+#define    BIM_LOCAL_DEV_SOFT_1        0x10    /* sw programmable ctrl bit 1 */
+#define    BIM_LOCAL_DEV_HW_RESET      0x20    /* internal hw reset. Cassini+ only. */
+
+/* access 24 entry BIM read and write buffers. put address in REG_BIM_BUFFER_ADDR
+ * and read/write from/to it REG_BIM_BUFFER_DATA_LOW and _DATA_HI.
+ * _DATA_HI should be the last access of the sequence.
+ * DEFAULT: undefined
+ */
+#define  REG_BIM_BUFFER_ADDR           0x1024  /* BIM buffer address. for
+						  purposes. */
+#define    BIM_BUFFER_ADDR_MASK        0x3F    /* index (0 - 23) of buffer  */
+#define    BIM_BUFFER_WR_SELECT        0x40    /* write buffer access = 1
+						  read buffer access = 0 */
+/* DEFAULT: undefined */
+#define  REG_BIM_BUFFER_DATA_LOW       0x1028  /* BIM buffer data low */
+#define  REG_BIM_BUFFER_DATA_HI        0x102C  /* BIM buffer data high */
+
+/* set BIM_RAM_BIST_START to start built-in self test for BIM read buffer.
+ * bit auto-clears when done with status read from _SUMMARY and _PASS bits.
+ */
+#define  REG_BIM_RAM_BIST              0x102C  /* BIM RAM (read buffer) BIST
+						  control/status */
+#define    BIM_RAM_BIST_RD_START       0x01    /* start BIST for BIM read buffer */
+#define    BIM_RAM_BIST_WR_START       0x02    /* start BIST for BIM write buffer.
+						  Cassini only. reserved in
+						  Cassini+. */
+#define    BIM_RAM_BIST_RD_PASS        0x04    /* summary BIST pass status for read
+						  buffer. */
+#define    BIM_RAM_BIST_WR_PASS        0x08    /* summary BIST pass status for write
+						  buffer. Cassini only. reserved
+						  in Cassini+. */
+#define    BIM_RAM_BIST_RD_LOW_PASS    0x10    /* read low bank passes BIST */
+#define    BIM_RAM_BIST_RD_HI_PASS     0x20    /* read high bank passes BIST */
+#define    BIM_RAM_BIST_WR_LOW_PASS    0x40    /* write low bank passes BIST.
+						  Cassini only. reserved in
+						  Cassini+. */
+#define    BIM_RAM_BIST_WR_HI_PASS     0x80    /* write high bank passes BIST.
+						  Cassini only. reserved in
+						  Cassini+. */
+
+/* ASUN: i'm not sure what this does as it's not in the spec.
+ * DEFAULT: 0xFC
+ */
+#define  REG_BIM_DIAG_MUX              0x1030  /* BIM diagnostic probe mux
+						  select register */
+
+/* enable probe monitoring mode and select data appearing on the P_A* bus. bit
+ * values for _SEL_HI_MASK and _SEL_LOW_MASK:
+ * 0x0: internal probe[7:0] (pci arb state, wtc empty w, wtc full w, wtc empty w,
+ *                           wtc empty r, post pci)
+ * 0x1: internal probe[15:8] (pci wbuf comp, pci wpkt comp, pci rbuf comp,
+ *                            pci rpkt comp, txdma wr req, txdma wr ack,
+ *			      txdma wr rdy, txdma wr xfr done)
+ * 0x2: internal probe[23:16] (txdma rd req, txdma rd ack, txdma rd rdy, rxdma rd,
+ *                             rd arb state, rd pci state)
+ * 0x3: internal probe[31:24] (rxdma req, rxdma ack, rxdma rdy, wrarb state,
+ *                             wrpci state)
+ * 0x4: pci io probe[7:0]     0x5: pci io probe[15:8]
+ * 0x6: pci io probe[23:16]   0x7: pci io probe[31:24]
+ * 0x8: pci io probe[39:32]   0x9: pci io probe[47:40]
+ * 0xa: pci io probe[55:48]   0xb: pci io probe[63:56]
+ * the following are not available in Cassini:
+ * 0xc: rx probe[7:0]         0xd: tx probe[7:0]
+ * 0xe: hp probe[7:0] 	      0xf: mac probe[7:0]
+ */
+#define  REG_PLUS_PROBE_MUX_SELECT     0x1034 /* Cassini+: PROBE MUX SELECT */
+#define    PROBE_MUX_EN                0x80000000 /* allow probe signals to be
+						     driven on local bus P_A[15:0]
+						     for debugging */
+#define    PROBE_MUX_SUB_MUX_MASK      0x0000FF00 /* select sub module probe signals:
+						     0x03 = mac[1:0]
+						     0x0C = rx[1:0]
+						     0x30 = tx[1:0]
+						     0xC0 = hp[1:0] */
+#define    PROBE_MUX_SEL_HI_MASK       0x000000F0 /* select which module to appear
+						     on P_A[15:8]. see above for
+						     values. */
+#define    PROBE_MUX_SEL_LOW_MASK      0x0000000F /* select which module to appear
+						     on P_A[7:0]. see above for
+						     values. */
+
+/* values mean the same thing as REG_INTR_MASK excep that it's for INTB.
+ DEFAULT: 0x1F */
+#define  REG_PLUS_INTR_MASK_1          0x1038 /* Cassini+: interrupt mask
+						 register 2 for INTB */
+#define  REG_PLUS_INTRN_MASK(x)       (REG_PLUS_INTR_MASK_1 + ((x) - 1)*16)
+/* bits correspond to both _MASK and _STATUS registers. _ALT corresponds to
+ * all of the alternate (2-4) INTR registers while _1 corresponds to only
+ * _MASK_1 and _STATUS_1 registers.
+ * DEFAULT: 0x7 for MASK registers, 0x0 for ALIAS_CLEAR registers
+ */
+#define    INTR_RX_DONE_ALT              0x01
+#define    INTR_RX_COMP_FULL_ALT         0x02
+#define    INTR_RX_COMP_AF_ALT           0x04
+#define    INTR_RX_BUF_UNAVAIL_1         0x08
+#define    INTR_RX_BUF_AE_1              0x10 /* almost empty */
+#define    INTRN_MASK_RX_EN              0x80
+#define    INTRN_MASK_CLEAR_ALL          (INTR_RX_DONE_ALT | \
+                                          INTR_RX_COMP_FULL_ALT | \
+                                          INTR_RX_COMP_AF_ALT | \
+                                          INTR_RX_BUF_UNAVAIL_1 | \
+                                          INTR_RX_BUF_AE_1)
+#define  REG_PLUS_INTR_STATUS_1        0x103C /* Cassini+: interrupt status
+						 register 2 for INTB. default: 0x1F */
+#define  REG_PLUS_INTRN_STATUS(x)       (REG_PLUS_INTR_STATUS_1 + ((x) - 1)*16)
+#define    INTR_STATUS_ALT_INTX_EN     0x80   /* generate INTX when one of the
+						 flags are set. enables desc ring. */
+
+#define  REG_PLUS_ALIAS_CLEAR_1        0x1040 /* Cassini+: alias clear mask
+						 register 2 for INTB */
+#define  REG_PLUS_ALIASN_CLEAR(x)      (REG_PLUS_ALIAS_CLEAR_1 + ((x) - 1)*16)
+
+#define  REG_PLUS_INTR_STATUS_ALIAS_1  0x1044 /* Cassini+: interrupt status
+						 register alias 2 for INTB */
+#define  REG_PLUS_INTRN_STATUS_ALIAS(x) (REG_PLUS_INTR_STATUS_ALIAS_1 + ((x) - 1)*16)
+
+#define REG_SATURN_PCFG               0x106c /* pin configuration register for
+						integrated macphy */
+
+#define   SATURN_PCFG_TLA             0x00000001 /* 1 = phy actled */
+#define   SATURN_PCFG_FLA             0x00000002 /* 1 = phy link10led */
+#define   SATURN_PCFG_CLA             0x00000004 /* 1 = phy link100led */
+#define   SATURN_PCFG_LLA             0x00000008 /* 1 = phy link1000led */
+#define   SATURN_PCFG_RLA             0x00000010 /* 1 = phy duplexled */
+#define   SATURN_PCFG_PDS             0x00000020 /* phy debug mode.
+						    0 = normal */
+#define   SATURN_PCFG_MTP             0x00000080 /* test point select */
+#define   SATURN_PCFG_GMO             0x00000100 /* GMII observe. 1 =
+						    GMII on SERDES pins for
+						    monitoring. */
+#define   SATURN_PCFG_FSI             0x00000200 /* 1 = freeze serdes/gmii. all
+						    pins configured as outputs.
+						    for power saving when using
+						    internal phy. */
+#define   SATURN_PCFG_LAD             0x00000800 /* 0 = mac core led ctrl
+						    polarity from strapping
+						    value.
+						    1 = mac core led ctrl
+						    polarity active low. */
+
+
+/** transmit dma registers **/
+#define MAX_TX_RINGS_SHIFT            2
+#define MAX_TX_RINGS                  (1 << MAX_TX_RINGS_SHIFT)
+#define MAX_TX_RINGS_MASK             (MAX_TX_RINGS - 1)
+
+/* TX configuration.
+ * descr ring sizes size = 32 * (1 << n), n < 9. e.g., 0x8 = 8k. default: 0x8
+ * DEFAULT: 0x3F000001
+ */
+#define  REG_TX_CFG                    0x2004  /* TX config */
+#define    TX_CFG_DMA_EN               0x00000001  /* enable TX DMA. if cleared, DMA
+						      will stop after xfer of current
+						      buffer has been completed. */
+#define    TX_CFG_FIFO_PIO_SEL         0x00000002  /* TX DMA FIFO can be
+						      accessed w/ FIFO addr
+						      and data registers.
+						      TX DMA should be
+						      disabled. */
+#define    TX_CFG_DESC_RING0_MASK      0x0000003C  /* # desc entries in
+						      ring 1. */
+#define    TX_CFG_DESC_RING0_SHIFT     2
+#define    TX_CFG_DESC_RINGN_MASK(a)   (TX_CFG_DESC_RING0_MASK << (a)*4)
+#define    TX_CFG_DESC_RINGN_SHIFT(a)  (TX_CFG_DESC_RING0_SHIFT + (a)*4)
+#define    TX_CFG_PACED_MODE           0x00100000  /* TX_ALL only set after
+						      TX FIFO becomes empty.
+						      if 0, TX_ALL set
+						      if descr queue empty. */
+#define    TX_CFG_DMA_RDPIPE_DIS       0x01000000  /* always set to 1 */
+#define    TX_CFG_COMPWB_Q1            0x02000000  /* completion writeback happens at
+						      the end of every packet kicked
+						      through Q1. */
+#define    TX_CFG_COMPWB_Q2            0x04000000  /* completion writeback happens at
+						      the end of every packet kicked
+						      through Q2. */
+#define    TX_CFG_COMPWB_Q3            0x08000000  /* completion writeback happens at
+						      the end of every packet kicked
+						      through Q3 */
+#define    TX_CFG_COMPWB_Q4            0x10000000  /* completion writeback happens at
+						      the end of every packet kicked
+						      through Q4 */
+#define    TX_CFG_INTR_COMPWB_DIS      0x20000000  /* disable pre-interrupt completion
+						      writeback */
+#define    TX_CFG_CTX_SEL_MASK         0xC0000000  /* selects tx test port
+						      connection
+						      0b00: tx mac req,
+						            tx mac retry req,
+							    tx ack and tx tag.
+						      0b01: txdma rd req,
+						            txdma rd ack,
+							    txdma rd rdy,
+							    txdma rd type0
+						      0b11: txdma wr req,
+						            txdma wr ack,
+							    txdma wr rdy,
+							    txdma wr xfr done. */
+#define    TX_CFG_CTX_SEL_SHIFT        30
+
+/* 11-bit counters that point to next location in FIFO to be loaded/retrieved.
+ * used for diagnostics only.
+ */
+#define  REG_TX_FIFO_WRITE_PTR         0x2014  /* TX FIFO write pointer */
+#define  REG_TX_FIFO_SHADOW_WRITE_PTR  0x2018  /* TX FIFO shadow write
+						  pointer. temp hold reg.
+					          diagnostics only. */
+#define  REG_TX_FIFO_READ_PTR          0x201C  /* TX FIFO read pointer */
+#define  REG_TX_FIFO_SHADOW_READ_PTR   0x2020  /* TX FIFO shadow read
+						  pointer */
+
+/* (ro) 11-bit up/down counter w/ # of frames currently in TX FIFO */
+#define  REG_TX_FIFO_PKT_CNT           0x2024  /* TX FIFO packet counter */
+
+/* current state of all state machines in TX */
+#define  REG_TX_SM_1                   0x2028  /* TX state machine reg #1 */
+#define    TX_SM_1_CHAIN_MASK          0x000003FF   /* chaining state machine */
+#define    TX_SM_1_CSUM_MASK           0x00000C00   /* checksum state machine */
+#define    TX_SM_1_FIFO_LOAD_MASK      0x0003F000   /* FIFO load state machine.
+						       = 0x01 when TX disabled. */
+#define    TX_SM_1_FIFO_UNLOAD_MASK    0x003C0000   /* FIFO unload state machine */
+#define    TX_SM_1_CACHE_MASK          0x03C00000   /* desc. prefetch cache controller
+						       state machine */
+#define    TX_SM_1_CBQ_ARB_MASK        0xF8000000   /* CBQ arbiter state machine */
+
+#define  REG_TX_SM_2                   0x202C  /* TX state machine reg #2 */
+#define    TX_SM_2_COMP_WB_MASK        0x07    /* completion writeback sm */
+#define	   TX_SM_2_SUB_LOAD_MASK       0x38    /* sub load state machine */
+#define	   TX_SM_2_KICK_MASK           0xC0    /* kick state machine */
+
+/* 64-bit pointer to the transmit data buffer. only the 50 LSB are incremented
+ * while the upper 23 bits are taken from the TX descriptor
+ */
+#define  REG_TX_DATA_PTR_LOW           0x2030  /* TX data pointer low */
+#define  REG_TX_DATA_PTR_HI            0x2034  /* TX data pointer high */
+
+/* 13 bit registers written by driver w/ descriptor value that follows
+ * last valid xmit descriptor. kick # and complete # values are used by
+ * the xmit dma engine to control tx descr fetching. if > 1 valid
+ * tx descr is available within the cache line being read, cassini will
+ * internally cache up to 4 of them. 0 on reset. _KICK = rw, _COMP = ro.
+ */
+#define  REG_TX_KICK0                  0x2038  /* TX kick reg #1 */
+#define  REG_TX_KICKN(x)               (REG_TX_KICK0 + (x)*4)
+#define  REG_TX_COMP0                  0x2048  /* TX completion reg #1 */
+#define  REG_TX_COMPN(x)               (REG_TX_COMP0 + (x)*4)
+
+/* values of TX_COMPLETE_1-4 are written. each completion register
+ * is 2bytes in size and contiguous. 8B allocation w/ 8B alignment.
+ * NOTE: completion reg values are only written back prior to TX_INTME and
+ * TX_ALL interrupts. at all other times, the most up-to-date index values
+ * should be obtained from the REG_TX_COMPLETE_# registers.
+ * here's the layout:
+ * offset from base addr      completion # byte
+ *           0                TX_COMPLETE_1_MSB
+ *	     1                TX_COMPLETE_1_LSB
+ *           2                TX_COMPLETE_2_MSB
+ *	     3                TX_COMPLETE_2_LSB
+ *           4                TX_COMPLETE_3_MSB
+ *	     5                TX_COMPLETE_3_LSB
+ *           6                TX_COMPLETE_4_MSB
+ *	     7                TX_COMPLETE_4_LSB
+ */
+#define  TX_COMPWB_SIZE             8
+#define  REG_TX_COMPWB_DB_LOW       0x2058  /* TX completion write back
+					       base low */
+#define  REG_TX_COMPWB_DB_HI        0x205C  /* TX completion write back
+					       base high */
+#define    TX_COMPWB_MSB_MASK       0x00000000000000FFULL
+#define    TX_COMPWB_MSB_SHIFT      0
+#define    TX_COMPWB_LSB_MASK       0x000000000000FF00ULL
+#define    TX_COMPWB_LSB_SHIFT      8
+#define    TX_COMPWB_NEXT(x)        ((x) >> 16)
+
+/* 53 MSB used as base address. 11 LSB assumed to be 0. TX desc pointer must
+ * be 2KB-aligned. */
+#define  REG_TX_DB0_LOW         0x2060  /* TX descriptor base low #1 */
+#define  REG_TX_DB0_HI          0x2064  /* TX descriptor base hi #1 */
+#define  REG_TX_DBN_LOW(x)      (REG_TX_DB0_LOW + (x)*8)
+#define  REG_TX_DBN_HI(x)       (REG_TX_DB0_HI + (x)*8)
+
+/* 16-bit registers hold weights for the weighted round-robin of the
+ * four CBQ TX descr rings. weights correspond to # bytes xferred from
+ * host to TXFIFO in a round of WRR arbitration. can be set
+ * dynamically with new weights set upon completion of the current
+ * packet transfer from host memory to TXFIFO. a dummy write to any of
+ * these registers causes a queue1 pre-emption with all historical bw
+ * deficit data reset to 0 (useful when congestion requires a
+ * pre-emption/re-allocation of network bandwidth
+ */
+#define  REG_TX_MAXBURST_0             0x2080  /* TX MaxBurst #1 */
+#define  REG_TX_MAXBURST_1             0x2084  /* TX MaxBurst #2 */
+#define  REG_TX_MAXBURST_2             0x2088  /* TX MaxBurst #3 */
+#define  REG_TX_MAXBURST_3             0x208C  /* TX MaxBurst #4 */
+
+/* diagnostics access to any TX FIFO location. every access is 65
+ * bits.  _DATA_LOW = 32 LSB, _DATA_HI_T1/T0 = 32 MSB. _TAG = tag bit.
+ * writing _DATA_HI_T0 sets tag bit low, writing _DATA_HI_T1 sets tag
+ * bit high.  TX_FIFO_PIO_SEL must be set for TX FIFO PIO access. if
+ * TX FIFO data integrity is desired, TX DMA should be
+ * disabled. _DATA_HI_Tx should be the last access of the sequence.
+ */
+#define  REG_TX_FIFO_ADDR              0x2104  /* TX FIFO address */
+#define  REG_TX_FIFO_TAG               0x2108  /* TX FIFO tag */
+#define  REG_TX_FIFO_DATA_LOW          0x210C  /* TX FIFO data low */
+#define  REG_TX_FIFO_DATA_HI_T1        0x2110  /* TX FIFO data high t1 */
+#define  REG_TX_FIFO_DATA_HI_T0        0x2114  /* TX FIFO data high t0 */
+#define  REG_TX_FIFO_SIZE              0x2118  /* (ro) TX FIFO size = 0x090 = 9KB */
+
+/* 9-bit register controls BIST of TX FIFO. bit set indicates that the BIST
+ * passed for the specified memory
+ */
+#define  REG_TX_RAMBIST                0x211C /* TX RAMBIST control/status */
+#define    TX_RAMBIST_STATE            0x01C0 /* progress state of RAMBIST
+						 controller state machine */
+#define    TX_RAMBIST_RAM33A_PASS      0x0020 /* RAM33A passed */
+#define    TX_RAMBIST_RAM32A_PASS      0x0010 /* RAM32A passed */
+#define    TX_RAMBIST_RAM33B_PASS      0x0008 /* RAM33B passed */
+#define    TX_RAMBIST_RAM32B_PASS      0x0004 /* RAM32B passed */
+#define    TX_RAMBIST_SUMMARY          0x0002 /* all RAM passed */
+#define    TX_RAMBIST_START            0x0001 /* write 1 to start BIST. self
+						 clears on completion. */
+
+/** receive dma registers **/
+#define MAX_RX_DESC_RINGS              2
+#define MAX_RX_COMP_RINGS              4
+
+/* receive DMA channel configuration. default: 0x80910
+ * free ring size       = (1 << n)*32  -> [32 - 8k]
+ * completion ring size = (1 << n)*128 -> [128 - 32k], n < 9
+ * DEFAULT: 0x80910
+ */
+#define  REG_RX_CFG                     0x4000  /* RX config */
+#define    RX_CFG_DMA_EN                0x00000001 /* enable RX DMA. 0 stops
+							 channel as soon as current
+							 frame xfer has completed.
+							 driver should disable MAC
+							 for 200ms before disabling
+							 RX */
+#define    RX_CFG_DESC_RING_MASK        0x0000001E /* # desc entries in RX
+							 free desc ring.
+							 def: 0x8 = 8k */
+#define    RX_CFG_DESC_RING_SHIFT       1
+#define    RX_CFG_COMP_RING_MASK        0x000001E0 /* # desc entries in RX complete
+							 ring. def: 0x8 = 32k */
+#define    RX_CFG_COMP_RING_SHIFT       5
+#define    RX_CFG_BATCH_DIS             0x00000200 /* disable receive desc
+						      batching. def: 0x0 =
+						      enabled */
+#define    RX_CFG_SWIVEL_MASK           0x00001C00 /* byte offset of the 1st
+						      data byte of the packet
+						      w/in 8 byte boundaries.
+						      this swivels the data
+						      DMA'ed to header
+						      buffers, jumbo buffers
+						      when header split is not
+						      requested and MTU sized
+						      buffers. def: 0x2 */
+#define    RX_CFG_SWIVEL_SHIFT          10
+
+/* cassini+ only */
+#define    RX_CFG_DESC_RING1_MASK       0x000F0000 /* # of desc entries in
+							 RX free desc ring 2.
+							 def: 0x8 = 8k */
+#define    RX_CFG_DESC_RING1_SHIFT      16
+
+
+/* the page size register allows cassini chips to do the following with
+ * received data:
+ * [--------------------------------------------------------------] page
+ * [off][buf1][pad][off][buf2][pad][off][buf3][pad][off][buf4][pad]
+ * |--------------| = PAGE_SIZE_BUFFER_STRIDE
+ * page = PAGE_SIZE
+ * offset = PAGE_SIZE_MTU_OFF
+ * for the above example, MTU_BUFFER_COUNT = 4.
+ * NOTE: as is apparent, you need to ensure that the following holds:
+ * MTU_BUFFER_COUNT <= PAGE_SIZE/PAGE_SIZE_BUFFER_STRIDE
+ * DEFAULT: 0x48002002 (8k pages)
+ */
+#define  REG_RX_PAGE_SIZE               0x4004  /* RX page size */
+#define    RX_PAGE_SIZE_MASK            0x00000003 /* size of pages pointed to
+						      by receive descriptors.
+						      if jumbo buffers are
+						      supported the page size
+						      should not be < 8k.
+						      0b00 = 2k, 0b01 = 4k
+						      0b10 = 8k, 0b11 = 16k
+						      DEFAULT: 8k */
+#define    RX_PAGE_SIZE_SHIFT           0
+#define    RX_PAGE_SIZE_MTU_COUNT_MASK  0x00007800 /* # of MTU buffers the hw
+						      packs into a page.
+						      DEFAULT: 4 */
+#define    RX_PAGE_SIZE_MTU_COUNT_SHIFT 11
+#define    RX_PAGE_SIZE_MTU_STRIDE_MASK 0x18000000 /* # of bytes that separate
+							 each MTU buffer +
+							 offset from each
+							 other.
+							 0b00 = 1k, 0b01 = 2k
+							 0b10 = 4k, 0b11 = 8k
+							 DEFAULT: 0x1 */
+#define    RX_PAGE_SIZE_MTU_STRIDE_SHIFT 27
+#define    RX_PAGE_SIZE_MTU_OFF_MASK    0xC0000000 /* offset in each page that
+						      hw writes the MTU buffer
+						      into.
+						      0b00 = 0,
+						      0b01 = 64 bytes
+						      0b10 = 96, 0b11 = 128
+						      DEFAULT: 0x1 */
+#define    RX_PAGE_SIZE_MTU_OFF_SHIFT   30
+
+/* 11-bit counter points to next location in RX FIFO to be loaded/read.
+ * shadow write pointers enable retries in case of early receive aborts.
+ * DEFAULT: 0x0. generated on 64-bit boundaries.
+ */
+#define  REG_RX_FIFO_WRITE_PTR             0x4008  /* RX FIFO write pointer */
+#define  REG_RX_FIFO_READ_PTR              0x400C  /* RX FIFO read pointer */
+#define  REG_RX_IPP_FIFO_SHADOW_WRITE_PTR  0x4010  /* RX IPP FIFO shadow write
+						      pointer */
+#define  REG_RX_IPP_FIFO_SHADOW_READ_PTR   0x4014  /* RX IPP FIFO shadow read
+						      pointer */
+#define  REG_RX_IPP_FIFO_READ_PTR          0x400C  /* RX IPP FIFO read
+						      pointer. (8-bit counter) */
+
+/* current state of RX DMA state engines + other info
+ * DEFAULT: 0x0
+ */
+#define  REG_RX_DEBUG                      0x401C  /* RX debug */
+#define    RX_DEBUG_LOAD_STATE_MASK        0x0000000F /* load state machine w/ MAC:
+							 0x0 = idle,   0x1 = load_bop
+							 0x2 = load 1, 0x3 = load 2
+							 0x4 = load 3, 0x5 = load 4
+							 0x6 = last detect
+							 0x7 = wait req
+							 0x8 = wait req statuss 1st
+							 0x9 = load st
+							 0xa = bubble mac
+							 0xb = error */
+#define    RX_DEBUG_LM_STATE_MASK          0x00000070 /* load state machine w/ HP and
+							 RX FIFO:
+							 0x0 = idle,   0x1 = hp xfr
+							 0x2 = wait hp ready
+							 0x3 = wait flow code
+							 0x4 = fifo xfer
+							 0x5 = make status
+							 0x6 = csum ready
+							 0x7 = error */
+#define    RX_DEBUG_FC_STATE_MASK          0x000000180 /* flow control state machine
+							 w/ MAC:
+							 0x0 = idle
+							 0x1 = wait xoff ack
+							 0x2 = wait xon
+							 0x3 = wait xon ack */
+#define    RX_DEBUG_DATA_STATE_MASK        0x000001E00 /* unload data state machine
+							 states:
+							 0x0 = idle data
+							 0x1 = header begin
+							 0x2 = xfer header
+							 0x3 = xfer header ld
+							 0x4 = mtu begin
+							 0x5 = xfer mtu
+							 0x6 = xfer mtu ld
+							 0x7 = jumbo begin
+							 0x8 = xfer jumbo
+							 0x9 = xfer jumbo ld
+							 0xa = reas begin
+							 0xb = xfer reas
+							 0xc = flush tag
+							 0xd = xfer reas ld
+							 0xe = error
+							 0xf = bubble idle */
+#define    RX_DEBUG_DESC_STATE_MASK        0x0001E000 /* unload desc state machine
+							 states:
+							 0x0 = idle desc
+							 0x1 = wait ack
+							 0x9 = wait ack 2
+							 0x2 = fetch desc 1
+							 0xa = fetch desc 2
+							 0x3 = load ptrs
+							 0x4 = wait dma
+							 0x5 = wait ack batch
+							 0x6 = post batch
+							 0x7 = xfr done */
+#define    RX_DEBUG_INTR_READ_PTR_MASK     0x30000000 /* interrupt read ptr of the
+							 interrupt queue */
+#define    RX_DEBUG_INTR_WRITE_PTR_MASK    0xC0000000 /* interrupt write pointer
+							 of the interrupt queue */
+
+/* flow control frames are emitted using two PAUSE thresholds:
+ * XOFF PAUSE uses pause time value pre-programmed in the Send PAUSE MAC reg
+ * XON PAUSE uses a pause time of 0. granularity of threshold is 64bytes.
+ * PAUSE thresholds defined in terms of FIFO occupancy and may be translated
+ * into FIFO vacancy using RX_FIFO_SIZE. setting ON will trigger XON frames
+ * when FIFO reaches 0. OFF threshold should not be > size of RX FIFO. max
+ * value is 0x6F.
+ * DEFAULT: 0x00078
+ */
+#define  REG_RX_PAUSE_THRESH               0x4020  /* RX pause thresholds */
+#define    RX_PAUSE_THRESH_QUANTUM         64
+#define    RX_PAUSE_THRESH_OFF_MASK        0x000001FF /* XOFF PAUSE emitted when
+							 RX FIFO occupancy >
+							 value*64B */
+#define    RX_PAUSE_THRESH_OFF_SHIFT       0
+#define    RX_PAUSE_THRESH_ON_MASK         0x001FF000 /* XON PAUSE emitted after
+							 emitting XOFF PAUSE when RX
+							 FIFO occupancy falls below
+							 this value*64B. must be
+							 < XOFF threshold. if =
+							 RX_FIFO_SIZE< XON frames are
+							 never emitted. */
+#define    RX_PAUSE_THRESH_ON_SHIFT        12
+
+/* 13-bit register used to control RX desc fetching and intr generation. if 4+
+ * valid RX descriptors are available, Cassini will read 4 at a time.
+ * writing N means that all desc up to *but* excluding N are available. N must
+ * be a multiple of 4 (N % 4 = 0). first desc should be cache-line aligned.
+ * DEFAULT: 0 on reset
+ */
+#define  REG_RX_KICK                    0x4024  /* RX kick reg */
+
+/* 8KB aligned 64-bit pointer to the base of the RX free/completion rings.
+ * lower 13 bits of the low register are hard-wired to 0.
+ */
+#define  REG_RX_DB_LOW                     0x4028  /* RX descriptor ring
+							 base low */
+#define  REG_RX_DB_HI                      0x402C  /* RX descriptor ring
+							 base hi */
+#define  REG_RX_CB_LOW                     0x4030  /* RX completion ring
+							 base low */
+#define  REG_RX_CB_HI                      0x4034  /* RX completion ring
+							 base hi */
+/* 13-bit register indicate desc used by cassini for receive frames. used
+ * for diagnostic purposes.
+ * DEFAULT: 0 on reset
+ */
+#define  REG_RX_COMP                       0x4038  /* (ro) RX completion */
+
+/* HEAD and TAIL are used to control RX desc posting and interrupt
+ * generation.  hw moves the head register to pass ownership to sw. sw
+ * moves the tail register to pass ownership back to hw. to give all
+ * entries to hw, set TAIL = HEAD.  if HEAD and TAIL indicate that no
+ * more entries are available, DMA will pause and an interrupt will be
+ * generated to indicate no more entries are available.  sw can use
+ * this interrupt to reduce the # of times it must update the
+ * completion tail register.
+ * DEFAULT: 0 on reset
+ */
+#define  REG_RX_COMP_HEAD                  0x403C  /* RX completion head */
+#define  REG_RX_COMP_TAIL                  0x4040  /* RX completion tail */
+
+/* values used for receive interrupt blanking. loaded each time the ISR is read
+ * DEFAULT: 0x00000000
+ */
+#define  REG_RX_BLANK                      0x4044  /* RX blanking register
+							 for ISR read */
+#define    RX_BLANK_INTR_PKT_MASK          0x000001FF /* RX_DONE intr asserted if
+							 this many sets of completion
+							 writebacks (up to 2 packets)
+							 occur since the last time
+							 the ISR was read. 0 = no
+							 packet blanking */
+#define    RX_BLANK_INTR_PKT_SHIFT         0
+#define    RX_BLANK_INTR_TIME_MASK         0x3FFFF000 /* RX_DONE interrupt asserted
+							 if that many clocks were
+							 counted since last time the
+							 ISR was read.
+							 each count is 512 core
+							 clocks (125MHz). 0 = no
+							 time blanking */
+#define    RX_BLANK_INTR_TIME_SHIFT        12
+
+/* values used for interrupt generation based on threshold values of how
+ * many free desc and completion entries are available for hw use.
+ * DEFAULT: 0x00000000
+ */
+#define  REG_RX_AE_THRESH                  0x4048  /* RX almost empty
+							 thresholds */
+#define    RX_AE_THRESH_FREE_MASK          0x00001FFF /* RX_BUF_AE will be
+							 generated if # desc
+							 avail for hw use <=
+							 # */
+#define    RX_AE_THRESH_FREE_SHIFT         0
+#define    RX_AE_THRESH_COMP_MASK          0x0FFFE000 /* RX_COMP_AE will be
+							 generated if # of
+							 completion entries
+							 avail for hw use <=
+							 # */
+#define    RX_AE_THRESH_COMP_SHIFT         13
+
+/* probabilities for random early drop (RED) thresholds on a FIFO threshold
+ * basis. probability should increase when the FIFO level increases. control
+ * packets are never dropped and not counted in stats. probability programmed
+ * on a 12.5% granularity. e.g., 0x1 = 1/8 packets dropped.
+ * DEFAULT: 0x00000000
+ */
+#define  REG_RX_RED                      0x404C  /* RX random early detect enable */
+#define    RX_RED_4K_6K_FIFO_MASK        0x000000FF /*  4KB < FIFO thresh < 6KB */
+#define    RX_RED_6K_8K_FIFO_MASK        0x0000FF00 /*  6KB < FIFO thresh < 8KB */
+#define    RX_RED_8K_10K_FIFO_MASK       0x00FF0000 /*  8KB < FIFO thresh < 10KB */
+#define    RX_RED_10K_12K_FIFO_MASK      0xFF000000 /* 10KB < FIFO thresh < 12KB */
+
+/* FIFO fullness levels for RX FIFO, RX control FIFO, and RX IPP FIFO.
+ * RX control FIFO = # of packets in RX FIFO.
+ * DEFAULT: 0x0
+ */
+#define  REG_RX_FIFO_FULLNESS              0x4050  /* (ro) RX FIFO fullness */
+#define    RX_FIFO_FULLNESS_RX_FIFO_MASK   0x3FF80000 /* level w/ 8B granularity */
+#define    RX_FIFO_FULLNESS_IPP_FIFO_MASK  0x0007FF00 /* level w/ 8B granularity */
+#define    RX_FIFO_FULLNESS_RX_PKT_MASK    0x000000FF /* # packets in RX FIFO */
+#define  REG_RX_IPP_PACKET_COUNT           0x4054  /* RX IPP packet counter */
+#define  REG_RX_WORK_DMA_PTR_LOW           0x4058  /* RX working DMA ptr low */
+#define  REG_RX_WORK_DMA_PTR_HI            0x405C  /* RX working DMA ptr
+						      high */
+
+/* BIST testing ro RX FIFO, RX control FIFO, and RX IPP FIFO. only RX BIST
+ * START/COMPLETE is writeable. START will clear when the BIST has completed
+ * checking all 17 RAMS.
+ * DEFAULT: 0bxxxx xxxxx xxxx xxxx xxxx x000 0000 0000 00x0
+ */
+#define  REG_RX_BIST                       0x4060  /* (ro) RX BIST */
+#define    RX_BIST_32A_PASS                0x80000000 /* RX FIFO 32A passed */
+#define    RX_BIST_33A_PASS                0x40000000 /* RX FIFO 33A passed */
+#define    RX_BIST_32B_PASS                0x20000000 /* RX FIFO 32B passed */
+#define    RX_BIST_33B_PASS                0x10000000 /* RX FIFO 33B passed */
+#define    RX_BIST_32C_PASS                0x08000000 /* RX FIFO 32C passed */
+#define    RX_BIST_33C_PASS                0x04000000 /* RX FIFO 33C passed */
+#define    RX_BIST_IPP_32A_PASS            0x02000000 /* RX IPP FIFO 33B passed */
+#define    RX_BIST_IPP_33A_PASS            0x01000000 /* RX IPP FIFO 33A passed */
+#define    RX_BIST_IPP_32B_PASS            0x00800000 /* RX IPP FIFO 32B passed */
+#define    RX_BIST_IPP_33B_PASS            0x00400000 /* RX IPP FIFO 33B passed */
+#define    RX_BIST_IPP_32C_PASS            0x00200000 /* RX IPP FIFO 32C passed */
+#define    RX_BIST_IPP_33C_PASS            0x00100000 /* RX IPP FIFO 33C passed */
+#define    RX_BIST_CTRL_32_PASS            0x00800000 /* RX CTRL FIFO 32 passed */
+#define    RX_BIST_CTRL_33_PASS            0x00400000 /* RX CTRL FIFO 33 passed */
+#define    RX_BIST_REAS_26A_PASS           0x00200000 /* RX Reas 26A passed */
+#define    RX_BIST_REAS_26B_PASS           0x00100000 /* RX Reas 26B passed */
+#define    RX_BIST_REAS_27_PASS            0x00080000 /* RX Reas 27 passed */
+#define    RX_BIST_STATE_MASK              0x00078000 /* BIST state machine */
+#define    RX_BIST_SUMMARY                 0x00000002 /* when BIST complete,
+							 summary pass bit
+							 contains AND of BIST
+							 results of all 16
+							 RAMS */
+#define    RX_BIST_START                   0x00000001 /* write 1 to start
+							 BIST. self clears
+							 on completion. */
+
+/* next location in RX CTRL FIFO that will be loaded w/ data from RX IPP/read
+ * from to retrieve packet control info.
+ * DEFAULT: 0
+ */
+#define  REG_RX_CTRL_FIFO_WRITE_PTR        0x4064  /* (ro) RX control FIFO
+						      write ptr */
+#define  REG_RX_CTRL_FIFO_READ_PTR         0x4068  /* (ro) RX control FIFO read
+						      ptr */
+
+/* receive interrupt blanking. loaded each time interrupt alias register is
+ * read.
+ * DEFAULT: 0x0
+ */
+#define  REG_RX_BLANK_ALIAS_READ           0x406C  /* RX blanking register for
+						      alias read */
+#define    RX_BAR_INTR_PACKET_MASK         0x000001FF /* assert RX_DONE if #
+							 completion writebacks
+							 > # since last ISR
+							 read. 0 = no
+							 blanking. up to 2
+							 packets per
+							 completion wb. */
+#define    RX_BAR_INTR_TIME_MASK           0x3FFFF000 /* assert RX_DONE if #
+							 clocks > # since last
+							 ISR read. each count
+							 is 512 core clocks
+							 (125MHz). 0 = no
+							 blanking. */
+
+/* diagnostic access to RX FIFO. 32 LSB accessed via DATA_LOW. 32 MSB accessed
+ * via DATA_HI_T0 or DATA_HI_T1. TAG reads the tag bit. writing HI_T0
+ * will unset the tag bit while writing HI_T1 will set the tag bit. to reset
+ * to normal operation after diagnostics, write to address location 0x0.
+ * RX_DMA_EN bit must be set to 0x0 for RX FIFO PIO access. DATA_HI should
+ * be the last write access of a write sequence.
+ * DEFAULT: undefined
+ */
+#define  REG_RX_FIFO_ADDR                  0x4080  /* RX FIFO address */
+#define  REG_RX_FIFO_TAG                   0x4084  /* RX FIFO tag */
+#define  REG_RX_FIFO_DATA_LOW              0x4088  /* RX FIFO data low */
+#define  REG_RX_FIFO_DATA_HI_T0            0x408C  /* RX FIFO data high T0 */
+#define  REG_RX_FIFO_DATA_HI_T1            0x4090  /* RX FIFO data high T1 */
+
+/* diagnostic assess to RX CTRL FIFO. 8-bit FIFO_ADDR holds address of
+ * 81 bit control entry and 6 bit flow id. LOW and MID are both 32-bit
+ * accesses. HI is 7-bits with 6-bit flow id and 1 bit control
+ * word. RX_DMA_EN must be 0 for RX CTRL FIFO PIO access. DATA_HI
+ * should be last write access of the write sequence.
+ * DEFAULT: undefined
+ */
+#define  REG_RX_CTRL_FIFO_ADDR             0x4094  /* RX Control FIFO and
+						      Batching FIFO addr */
+#define  REG_RX_CTRL_FIFO_DATA_LOW         0x4098  /* RX Control FIFO data
+						      low */
+#define  REG_RX_CTRL_FIFO_DATA_MID         0x409C  /* RX Control FIFO data
+						      mid */
+#define  REG_RX_CTRL_FIFO_DATA_HI          0x4100  /* RX Control FIFO data
+						      hi and flow id */
+#define    RX_CTRL_FIFO_DATA_HI_CTRL       0x0001  /* upper bit of ctrl word */
+#define    RX_CTRL_FIFO_DATA_HI_FLOW_MASK  0x007E  /* flow id */
+
+/* diagnostic access to RX IPP FIFO. same semantics as RX_FIFO.
+ * DEFAULT: undefined
+ */
+#define  REG_RX_IPP_FIFO_ADDR              0x4104  /* RX IPP FIFO address */
+#define  REG_RX_IPP_FIFO_TAG               0x4108  /* RX IPP FIFO tag */
+#define  REG_RX_IPP_FIFO_DATA_LOW          0x410C  /* RX IPP FIFO data low */
+#define  REG_RX_IPP_FIFO_DATA_HI_T0        0x4110  /* RX IPP FIFO data high
+						      T0 */
+#define  REG_RX_IPP_FIFO_DATA_HI_T1        0x4114  /* RX IPP FIFO data high
+						      T1 */
+
+/* 64-bit pointer to receive data buffer in host memory used for headers and
+ * small packets. MSB in high register. loaded by DMA state machine and
+ * increments as DMA writes receive data. only 50 LSB are incremented. top
+ * 13 bits taken from RX descriptor.
+ * DEFAULT: undefined
+ */
+#define  REG_RX_HEADER_PAGE_PTR_LOW        0x4118  /* (ro) RX header page ptr
+						      low */
+#define  REG_RX_HEADER_PAGE_PTR_HI         0x411C  /* (ro) RX header page ptr
+						      high */
+#define  REG_RX_MTU_PAGE_PTR_LOW           0x4120  /* (ro) RX MTU page pointer
+						      low */
+#define  REG_RX_MTU_PAGE_PTR_HI            0x4124  /* (ro) RX MTU page pointer
+						      high */
+
+/* PIO diagnostic access to RX reassembly DMA Table RAM. 6-bit register holds
+ * one of 64 79-bit locations in the RX Reassembly DMA table and the addr of
+ * one of the 64 byte locations in the Batching table. LOW holds 32 LSB.
+ * MID holds the next 32 LSB. HIGH holds the 15 MSB. RX_DMA_EN must be set
+ * to 0 for PIO access. DATA_HIGH should be last write of write sequence.
+ * layout:
+ * reassmbl ptr [78:15] | reassmbl index [14:1] | reassmbl entry valid [0]
+ * DEFAULT: undefined
+ */
+#define  REG_RX_TABLE_ADDR             0x4128  /* RX reassembly DMA table
+						  address */
+#define    RX_TABLE_ADDR_MASK          0x0000003F /* address mask */
+
+#define  REG_RX_TABLE_DATA_LOW         0x412C  /* RX reassembly DMA table
+						  data low */
+#define  REG_RX_TABLE_DATA_MID         0x4130  /* RX reassembly DMA table
+						  data mid */
+#define  REG_RX_TABLE_DATA_HI          0x4134  /* RX reassembly DMA table
+						  data high */
+
+/* cassini+ only */
+/* 8KB aligned 64-bit pointer to base of RX rings. lower 13 bits hardwired to
+ * 0. same semantics as primary desc/complete rings.
+ */
+#define  REG_PLUS_RX_DB1_LOW            0x4200  /* RX descriptor ring
+						   2 base low */
+#define  REG_PLUS_RX_DB1_HI             0x4204  /* RX descriptor ring
+						   2 base high */
+#define  REG_PLUS_RX_CB1_LOW            0x4208  /* RX completion ring
+						   2 base low. 4 total */
+#define  REG_PLUS_RX_CB1_HI             0x420C  /* RX completion ring
+						   2 base high. 4 total */
+#define  REG_PLUS_RX_CBN_LOW(x)        (REG_PLUS_RX_CB1_LOW + 8*((x) - 1))
+#define  REG_PLUS_RX_CBN_HI(x)         (REG_PLUS_RX_CB1_HI + 8*((x) - 1))
+#define  REG_PLUS_RX_KICK1             0x4220  /* RX Kick 2 register */
+#define  REG_PLUS_RX_COMP1             0x4224  /* (ro) RX completion 2
+						  reg */
+#define  REG_PLUS_RX_COMP1_HEAD        0x4228  /* (ro) RX completion 2
+						  head reg. 4 total. */
+#define  REG_PLUS_RX_COMP1_TAIL        0x422C  /* RX completion 2
+						  tail reg. 4 total. */
+#define  REG_PLUS_RX_COMPN_HEAD(x)    (REG_PLUS_RX_COMP1_HEAD + 8*((x) - 1))
+#define  REG_PLUS_RX_COMPN_TAIL(x)    (REG_PLUS_RX_COMP1_TAIL + 8*((x) - 1))
+#define  REG_PLUS_RX_AE1_THRESH        0x4240  /* RX almost empty 2
+						  thresholds */
+#define    RX_AE1_THRESH_FREE_MASK     RX_AE_THRESH_FREE_MASK
+#define    RX_AE1_THRESH_FREE_SHIFT    RX_AE_THRESH_FREE_SHIFT
+
+/** header parser registers **/
+
+/* RX parser configuration register.
+ * DEFAULT: 0x1651004
+ */
+#define  REG_HP_CFG                       0x4140  /* header parser
+						     configuration reg */
+#define    HP_CFG_PARSE_EN                0x00000001 /* enab header parsing */
+#define    HP_CFG_NUM_CPU_MASK            0x000000FC /* # processors
+						      0 = 64. 0x3f = 63 */
+#define    HP_CFG_NUM_CPU_SHIFT           2
+#define    HP_CFG_SYN_INC_MASK            0x00000100 /* SYN bit won't increment
+							TCP seq # by one when
+							stored in FDBM */
+#define    HP_CFG_TCP_THRESH_MASK         0x000FFE00 /* # bytes of TCP data
+							needed to be considered
+							for reassembly */
+#define    HP_CFG_TCP_THRESH_SHIFT        9
+
+/* access to RX Instruction RAM. 5-bit register/counter holds addr
+ * of 39 bit entry to be read/written. 32 LSB in _DATA_LOW. 7 MSB in _DATA_HI.
+ * RX_DMA_EN must be 0 for RX instr PIO access. DATA_HI should be last access
+ * of sequence.
+ * DEFAULT: undefined
+ */
+#define  REG_HP_INSTR_RAM_ADDR             0x4144  /* HP instruction RAM
+						      address */
+#define    HP_INSTR_RAM_ADDR_MASK          0x01F   /* 5-bit mask */
+#define  REG_HP_INSTR_RAM_DATA_LOW         0x4148  /* HP instruction RAM
+						      data low */
+#define    HP_INSTR_RAM_LOW_OUTMASK_MASK   0x0000FFFF
+#define    HP_INSTR_RAM_LOW_OUTMASK_SHIFT  0
+#define    HP_INSTR_RAM_LOW_OUTSHIFT_MASK  0x000F0000
+#define    HP_INSTR_RAM_LOW_OUTSHIFT_SHIFT 16
+#define    HP_INSTR_RAM_LOW_OUTEN_MASK     0x00300000
+#define    HP_INSTR_RAM_LOW_OUTEN_SHIFT    20
+#define    HP_INSTR_RAM_LOW_OUTARG_MASK    0xFFC00000
+#define    HP_INSTR_RAM_LOW_OUTARG_SHIFT   22
+#define  REG_HP_INSTR_RAM_DATA_MID         0x414C  /* HP instruction RAM
+						      data mid */
+#define    HP_INSTR_RAM_MID_OUTARG_MASK    0x00000003
+#define    HP_INSTR_RAM_MID_OUTARG_SHIFT   0
+#define    HP_INSTR_RAM_MID_OUTOP_MASK     0x0000003C
+#define    HP_INSTR_RAM_MID_OUTOP_SHIFT    2
+#define    HP_INSTR_RAM_MID_FNEXT_MASK     0x000007C0
+#define    HP_INSTR_RAM_MID_FNEXT_SHIFT    6
+#define    HP_INSTR_RAM_MID_FOFF_MASK      0x0003F800
+#define    HP_INSTR_RAM_MID_FOFF_SHIFT     11
+#define    HP_INSTR_RAM_MID_SNEXT_MASK     0x007C0000
+#define    HP_INSTR_RAM_MID_SNEXT_SHIFT    18
+#define    HP_INSTR_RAM_MID_SOFF_MASK      0x3F800000
+#define    HP_INSTR_RAM_MID_SOFF_SHIFT     23
+#define    HP_INSTR_RAM_MID_OP_MASK        0xC0000000
+#define    HP_INSTR_RAM_MID_OP_SHIFT       30
+#define  REG_HP_INSTR_RAM_DATA_HI          0x4150  /* HP instruction RAM
+						      data high */
+#define    HP_INSTR_RAM_HI_VAL_MASK        0x0000FFFF
+#define    HP_INSTR_RAM_HI_VAL_SHIFT       0
+#define    HP_INSTR_RAM_HI_MASK_MASK       0xFFFF0000
+#define    HP_INSTR_RAM_HI_MASK_SHIFT      16
+
+/* PIO access into RX Header parser data RAM and flow database.
+ * 11-bit register. Data fills the LSB portion of bus if less than 32 bits.
+ * DATA_RAM: write RAM_FDB_DATA with index to access DATA_RAM.
+ * RAM bytes = 4*(x - 1) + [3:0]. e.g., 0 -> [3:0], 31 -> [123:120]
+ * FLOWDB: write DATA_RAM_FDB register and then read/write FDB1-12 to access
+ * flow database.
+ * RX_DMA_EN must be 0 for RX parser RAM PIO access. RX Parser RAM data reg
+ * should be the last write access of the write sequence.
+ * DEFAULT: undefined
+ */
+#define  REG_HP_DATA_RAM_FDB_ADDR          0x4154  /* HP data and FDB
+						      RAM address */
+#define    HP_DATA_RAM_FDB_DATA_MASK       0x001F  /* select 1 of 86 byte
+						      locations in header
+						      parser data ram to
+						      read/write */
+#define    HP_DATA_RAM_FDB_FDB_MASK        0x3F00  /* 1 of 64 353-bit locations
+						      in the flow database */
+#define  REG_HP_DATA_RAM_DATA              0x4158  /* HP data RAM data */
+
+/* HP flow database registers: 1 - 12, 0x415C - 0x4188, 4 8-bit bytes
+ * FLOW_DB(1) = IP_SA[127:96], FLOW_DB(2) = IP_SA[95:64]
+ * FLOW_DB(3) = IP_SA[63:32],  FLOW_DB(4) = IP_SA[31:0]
+ * FLOW_DB(5) = IP_DA[127:96], FLOW_DB(6) = IP_DA[95:64]
+ * FLOW_DB(7) = IP_DA[63:32],  FLOW_DB(8) = IP_DA[31:0]
+ * FLOW_DB(9) = {TCP_SP[15:0],TCP_DP[15:0]}
+ * FLOW_DB(10) = bit 0 has value for flow valid
+ * FLOW_DB(11) = TCP_SEQ[63:32], FLOW_DB(12) = TCP_SEQ[31:0]
+ */
+#define  REG_HP_FLOW_DB0                   0x415C  /* HP flow database 1 reg */
+#define  REG_HP_FLOW_DBN(x)                (REG_HP_FLOW_DB0 + (x)*4)
+
+/* diagnostics for RX Header Parser block.
+ * ASUN: the header parser state machine register is used for diagnostics
+ * purposes. however, the spec doesn't have any details on it.
+ */
+#define  REG_HP_STATE_MACHINE              0x418C  /* (ro) HP state machine */
+#define  REG_HP_STATUS0                    0x4190  /* (ro) HP status 1 */
+#define    HP_STATUS0_SAP_MASK             0xFFFF0000 /* SAP */
+#define    HP_STATUS0_L3_OFF_MASK          0x0000FE00 /* L3 offset */
+#define    HP_STATUS0_LB_CPUNUM_MASK       0x000001F8 /* load balancing CPU
+							 number */
+#define    HP_STATUS0_HRP_OPCODE_MASK      0x00000007 /* HRP opcode */
+
+#define  REG_HP_STATUS1                    0x4194  /* (ro) HP status 2 */
+#define    HP_STATUS1_ACCUR2_MASK          0xE0000000 /* accu R2[6:4] */
+#define    HP_STATUS1_FLOWID_MASK          0x1F800000 /* flow id */
+#define    HP_STATUS1_TCP_OFF_MASK         0x007F0000 /* tcp payload offset */
+#define    HP_STATUS1_TCP_SIZE_MASK        0x0000FFFF /* tcp payload size */
+
+#define  REG_HP_STATUS2                    0x4198  /* (ro) HP status 3 */
+#define    HP_STATUS2_ACCUR2_MASK          0xF0000000 /* accu R2[3:0] */
+#define    HP_STATUS2_CSUM_OFF_MASK        0x07F00000 /* checksum start
+							 start offset */
+#define    HP_STATUS2_ACCUR1_MASK          0x000FE000 /* accu R1 */
+#define    HP_STATUS2_FORCE_DROP           0x00001000 /* force drop */
+#define    HP_STATUS2_BWO_REASSM           0x00000800 /* batching w/o
+							 reassembly */
+#define    HP_STATUS2_JH_SPLIT_EN          0x00000400 /* jumbo header split
+							 enable */
+#define    HP_STATUS2_FORCE_TCP_NOCHECK    0x00000200 /* force tcp no payload
+							 check */
+#define    HP_STATUS2_DATA_MASK_ZERO       0x00000100 /* mask of data length
+							 equal to zero */
+#define    HP_STATUS2_FORCE_TCP_CHECK      0x00000080 /* force tcp payload
+							 chk */
+#define    HP_STATUS2_MASK_TCP_THRESH      0x00000040 /* mask of payload
+							 threshold */
+#define    HP_STATUS2_NO_ASSIST            0x00000020 /* no assist */
+#define    HP_STATUS2_CTRL_PACKET_FLAG     0x00000010 /* control packet flag */
+#define    HP_STATUS2_TCP_FLAG_CHECK       0x00000008 /* tcp flag check */
+#define    HP_STATUS2_SYN_FLAG             0x00000004 /* syn flag */
+#define    HP_STATUS2_TCP_CHECK            0x00000002 /* tcp payload chk */
+#define    HP_STATUS2_TCP_NOCHECK          0x00000001 /* tcp no payload chk */
+
+/* BIST for header parser(HP) and flow database memories (FDBM). set _START
+ * to start BIST. controller clears _START on completion. _START can also
+ * be cleared to force termination of BIST. a bit set indicates that that
+ * memory passed its BIST.
+ */
+#define  REG_HP_RAM_BIST                   0x419C  /* HP RAM BIST reg */
+#define    HP_RAM_BIST_HP_DATA_PASS        0x80000000 /* HP data ram */
+#define    HP_RAM_BIST_HP_INSTR0_PASS      0x40000000 /* HP instr ram 0 */
+#define    HP_RAM_BIST_HP_INSTR1_PASS      0x20000000 /* HP instr ram 1 */
+#define    HP_RAM_BIST_HP_INSTR2_PASS      0x10000000 /* HP instr ram 2 */
+#define    HP_RAM_BIST_FDBM_AGE0_PASS      0x08000000 /* FDBM aging RAM0 */
+#define    HP_RAM_BIST_FDBM_AGE1_PASS      0x04000000 /* FDBM aging RAM1 */
+#define    HP_RAM_BIST_FDBM_FLOWID00_PASS  0x02000000 /* FDBM flowid RAM0
+							 bank 0 */
+#define    HP_RAM_BIST_FDBM_FLOWID10_PASS  0x01000000 /* FDBM flowid RAM1
+							 bank 0 */
+#define    HP_RAM_BIST_FDBM_FLOWID20_PASS  0x00800000 /* FDBM flowid RAM2
+							 bank 0 */
+#define    HP_RAM_BIST_FDBM_FLOWID30_PASS  0x00400000 /* FDBM flowid RAM3
+							 bank 0 */
+#define    HP_RAM_BIST_FDBM_FLOWID01_PASS  0x00200000 /* FDBM flowid RAM0
+							 bank 1 */
+#define    HP_RAM_BIST_FDBM_FLOWID11_PASS  0x00100000 /* FDBM flowid RAM1
+							 bank 2 */
+#define    HP_RAM_BIST_FDBM_FLOWID21_PASS  0x00080000 /* FDBM flowid RAM2
+							 bank 1 */
+#define    HP_RAM_BIST_FDBM_FLOWID31_PASS  0x00040000 /* FDBM flowid RAM3
+							 bank 1 */
+#define    HP_RAM_BIST_FDBM_TCPSEQ_PASS    0x00020000 /* FDBM tcp sequence
+							 RAM */
+#define    HP_RAM_BIST_SUMMARY             0x00000002 /* all BIST tests */
+#define    HP_RAM_BIST_START               0x00000001 /* start/stop BIST */
+
+
+/** MAC registers.  **/
+/* reset bits are set using a PIO write and self-cleared after the command
+ * execution has completed.
+ */
+#define  REG_MAC_TX_RESET                  0x6000  /* TX MAC software reset
+						      command (default: 0x0) */
+#define  REG_MAC_RX_RESET                  0x6004  /* RX MAC software reset
+						      command (default: 0x0) */
+/* execute a pause flow control frame transmission
+ DEFAULT: 0x0XXXX */
+#define  REG_MAC_SEND_PAUSE                0x6008  /* send pause command reg */
+#define    MAC_SEND_PAUSE_TIME_MASK        0x0000FFFF /* value of pause time
+							 to be sent on network
+							 in units of slot
+							 times */
+#define    MAC_SEND_PAUSE_SEND             0x00010000 /* send pause flow ctrl
+							 frame on network */
+
+/* bit set indicates that event occurred. auto-cleared when status register
+ * is read and have corresponding mask bits in mask register. events will
+ * trigger an interrupt if the corresponding mask bit is 0.
+ * status register default: 0x00000000
+ * mask register default = 0xFFFFFFFF on reset
+ */
+#define  REG_MAC_TX_STATUS                 0x6010  /* TX MAC status reg */
+#define    MAC_TX_FRAME_XMIT               0x0001  /* successful frame
+						      transmission */
+#define    MAC_TX_UNDERRUN                 0x0002  /* terminated frame
+						      transmission due to
+						      data starvation in the
+						      xmit data path */
+#define    MAC_TX_MAX_PACKET_ERR           0x0004  /* frame exceeds max allowed
+						      length passed to TX MAC
+						      by the DMA engine */
+#define    MAC_TX_COLL_NORMAL              0x0008  /* rollover of the normal
+						      collision counter */
+#define    MAC_TX_COLL_EXCESS              0x0010  /* rollover of the excessive
+						      collision counter */
+#define    MAC_TX_COLL_LATE                0x0020  /* rollover of the late
+						      collision counter */
+#define    MAC_TX_COLL_FIRST               0x0040  /* rollover of the first
+						      collision counter */
+#define    MAC_TX_DEFER_TIMER              0x0080  /* rollover of the defer
+						      timer */
+#define    MAC_TX_PEAK_ATTEMPTS            0x0100  /* rollover of the peak
+						      attempts counter */
+
+#define  REG_MAC_RX_STATUS                 0x6014  /* RX MAC status reg */
+#define    MAC_RX_FRAME_RECV               0x0001  /* successful receipt of
+						      a frame */
+#define    MAC_RX_OVERFLOW                 0x0002  /* dropped frame due to
+						      RX FIFO overflow */
+#define    MAC_RX_FRAME_COUNT              0x0004  /* rollover of receive frame
+						      counter */
+#define    MAC_RX_ALIGN_ERR                0x0008  /* rollover of alignment
+						      error counter */
+#define    MAC_RX_CRC_ERR                  0x0010  /* rollover of crc error
+						      counter */
+#define    MAC_RX_LEN_ERR                  0x0020  /* rollover of length
+						      error counter */
+#define    MAC_RX_VIOL_ERR                 0x0040  /* rollover of code
+						      violation error */
+
+/* DEFAULT: 0xXXXX0000 on reset */
+#define  REG_MAC_CTRL_STATUS               0x6018  /* MAC control status reg */
+#define    MAC_CTRL_PAUSE_RECEIVED         0x00000001  /* successful
+							  reception of a
+							  pause control
+							  frame */
+#define    MAC_CTRL_PAUSE_STATE            0x00000002  /* MAC has made a
+							  transition from
+							  "not paused" to
+							  "paused" */
+#define    MAC_CTRL_NOPAUSE_STATE          0x00000004  /* MAC has made a
+							  transition from
+							  "paused" to "not
+							  paused" */
+#define    MAC_CTRL_PAUSE_TIME_MASK        0xFFFF0000  /* value of pause time
+							  operand that was
+							  received in the last
+							  pause flow control
+							  frame */
+
+/* layout identical to TX MAC[8:0] */
+#define  REG_MAC_TX_MASK                   0x6020  /* TX MAC mask reg */
+/* layout identical to RX MAC[6:0] */
+#define  REG_MAC_RX_MASK                   0x6024  /* RX MAC mask reg */
+/* layout identical to CTRL MAC[2:0] */
+#define  REG_MAC_CTRL_MASK                 0x6028  /* MAC control mask reg */
+
+/* to ensure proper operation, CFG_EN must be cleared to 0 and a delay
+ * imposed before writes to other bits in the TX_MAC_CFG register or any of
+ * the MAC parameters is performed. delay dependent upon time required to
+ * transmit a maximum size frame (= MAC_FRAMESIZE_MAX*8/Mbps). e.g.,
+ * the delay for a 1518-byte frame on a 100Mbps network is 125us.
+ * alternatively, just poll TX_CFG_EN until it reads back as 0.
+ * NOTE: on half-duplex 1Gbps, TX_CFG_CARRIER_EXTEND and
+ * RX_CFG_CARRIER_EXTEND should be set and the SLOT_TIME register should
+ * be 0x200 (slot time of 512 bytes)
+ */
+#define  REG_MAC_TX_CFG                 0x6030  /* TX MAC config reg */
+#define    MAC_TX_CFG_EN                0x0001  /* enable TX MAC. 0 will
+						      force TXMAC state
+						      machine to remain in
+						      idle state or to
+						      transition to idle state
+						      on completion of an
+						      ongoing packet. */
+#define    MAC_TX_CFG_IGNORE_CARRIER    0x0002  /* disable CSMA/CD deferral
+						   process. set to 1 when
+						   full duplex and 0 when
+						   half duplex */
+#define    MAC_TX_CFG_IGNORE_COLL       0x0004  /* disable CSMA/CD backoff
+						   algorithm. set to 1 when
+						   full duplex and 0 when
+						   half duplex */
+#define    MAC_TX_CFG_IPG_EN            0x0008  /* enable extension of the
+						   Rx-to-TX IPG. after
+						   receiving a frame, TX
+						   MAC will reset its
+						   deferral process to
+						   carrier sense for the
+						   amount of time = IPG0 +
+						   IPG1 and commit to
+						   transmission for time
+						   specified in IPG2. when
+						   0 or when xmitting frames
+						   back-to-pack (Tx-to-Tx
+						   IPG), TX MAC ignores
+						   IPG0 and will only use
+						   IPG1 for deferral time.
+						   IPG2 still used. */
+#define    MAC_TX_CFG_NEVER_GIVE_UP_EN  0x0010  /* TX MAC will not easily
+						   give up on frame
+						   xmission. if backoff
+						   algorithm reaches the
+						   ATTEMPT_LIMIT, it will
+						   clear attempts counter
+						   and continue trying to
+						   send the frame as
+						   specified by
+						   GIVE_UP_LIM. when 0,
+						   TX MAC will execute
+						   standard CSMA/CD prot. */
+#define    MAC_TX_CFG_NEVER_GIVE_UP_LIM 0x0020  /* when set, TX MAC will
+						   continue to try to xmit
+						   until successful. when
+						   0, TX MAC will continue
+						   to try xmitting until
+						   successful or backoff
+						   algorithm reaches
+						   ATTEMPT_LIMIT*16 */
+#define    MAC_TX_CFG_NO_BACKOFF        0x0040  /* modify CSMA/CD to disable
+						   backoff algorithm. TX
+						   MAC will not back off
+						   after a xmission attempt
+						   that resulted in a
+						   collision. */
+#define    MAC_TX_CFG_SLOW_DOWN         0x0080  /* modify CSMA/CD so that
+						   deferral process is reset
+						   in response to carrier
+						   sense during the entire
+						   duration of IPG. TX MAC
+						   will only commit to frame
+						   xmission after frame
+						   xmission has actually
+						   begun. */
+#define    MAC_TX_CFG_NO_FCS            0x0100  /* TX MAC will not generate
+						   CRC for all xmitted
+						   packets. when clear, CRC
+						   generation is dependent
+						   upon NO_CRC bit in the
+						   xmit control word from
+						   TX DMA */
+#define    MAC_TX_CFG_CARRIER_EXTEND    0x0200  /* enables xmit part of the
+						   carrier extension
+						   feature. this allows for
+						   longer collision domains
+						   by extending the carrier
+						   and collision window
+						   from the end of FCS until
+						   the end of the slot time
+						   if necessary. Required
+						   for half-duplex at 1Gbps,
+						   clear otherwise. */
+
+/* when CRC is not stripped, reassembly packets will not contain the CRC.
+ * these will be stripped by HRP because it reassembles layer 4 data, and the
+ * CRC is layer 2. however, non-reassembly packets will still contain the CRC
+ * when passed to the host. to ensure proper operation, need to wait 3.2ms
+ * after clearing RX_CFG_EN before writing to any other RX MAC registers
+ * or other MAC parameters. alternatively, poll RX_CFG_EN until it clears
+ * to 0. Similarly, HASH_FILTER_EN and ADDR_FILTER_EN have the same
+ * restrictions as CFG_EN.
+ */
+#define  REG_MAC_RX_CFG                 0x6034  /* RX MAC config reg */
+#define    MAC_RX_CFG_EN                0x0001  /* enable RX MAC */
+#define    MAC_RX_CFG_STRIP_PAD         0x0002  /* always program to 0.
+						   feature not supported */
+#define    MAC_RX_CFG_STRIP_FCS         0x0004  /* RX MAC will strip the
+						   last 4 bytes of a
+						   received frame. */
+#define    MAC_RX_CFG_PROMISC_EN        0x0008  /* promiscuous mode */
+#define    MAC_RX_CFG_PROMISC_GROUP_EN  0x0010  /* accept all valid
+						   multicast frames (group
+						   bit in DA field set) */
+#define    MAC_RX_CFG_HASH_FILTER_EN    0x0020  /* use hash table to filter
+						   multicast addresses */
+#define    MAC_RX_CFG_ADDR_FILTER_EN    0x0040  /* cause RX MAC to use
+						   address filtering regs
+						   to filter both unicast
+						   and multicast
+						   addresses */
+#define    MAC_RX_CFG_DISABLE_DISCARD   0x0080  /* pass errored frames to
+						   RX DMA by setting BAD
+						   bit but not Abort bit
+						   in the status. CRC,
+						   framing, and length errs
+						   will not increment
+						   error counters. frames
+						   which don't match dest
+						   addr will be passed up
+						   w/ BAD bit set. */
+#define    MAC_RX_CFG_CARRIER_EXTEND    0x0100  /* enable reception of
+						   packet bursts generated
+						   by carrier extension
+						   with packet bursting
+						   senders. only applies
+						   to half-duplex 1Gbps */
+
+/* DEFAULT: 0x0 */
+#define  REG_MAC_CTRL_CFG               0x6038  /* MAC control config reg */
+#define    MAC_CTRL_CFG_SEND_PAUSE_EN   0x0001  /* respond to requests for
+						   sending pause flow ctrl
+						   frames */
+#define    MAC_CTRL_CFG_RECV_PAUSE_EN   0x0002  /* respond to received
+						   pause flow ctrl frames */
+#define    MAC_CTRL_CFG_PASS_CTRL       0x0004  /* pass valid MAC ctrl
+						   packets to RX DMA */
+
+/* to ensure proper operation, a global initialization sequence should be
+ * performed when a loopback config is entered or exited. if programmed after
+ * a hw or global sw reset, RX/TX MAC software reset and initialization
+ * should be done to ensure stable clocking.
+ * DEFAULT: 0x0
+ */
+#define  REG_MAC_XIF_CFG                0x603C  /* XIF config reg */
+#define    MAC_XIF_TX_MII_OUTPUT_EN        0x0001  /* enable output drivers
+						      on MII xmit bus */
+#define    MAC_XIF_MII_INT_LOOPBACK        0x0002  /* loopback GMII xmit data
+						      path to GMII recv data
+						      path. phy mode register
+						      clock selection must be
+						      set to GMII mode and
+						      GMII_MODE should be set
+						      to 1. in loopback mode,
+						      REFCLK will drive the
+						      entire mac core. 0 for
+						      normal operation. */
+#define    MAC_XIF_DISABLE_ECHO            0x0004  /* disables receive data
+						      path during packet
+						      xmission. clear to 0
+						      in any full duplex mode,
+						      in any loopback mode,
+						      or in half-duplex SERDES
+						      or SLINK modes. set when
+						      in half-duplex when
+						      using external phy. */
+#define    MAC_XIF_GMII_MODE               0x0008  /* MAC operates with GMII
+						      clocks and datapath */
+#define    MAC_XIF_MII_BUFFER_OUTPUT_EN    0x0010  /* MII_BUF_EN pin. enable
+						      external tristate buffer
+						      on the MII receive
+						      bus. */
+#define    MAC_XIF_LINK_LED                0x0020  /* LINKLED# active (low) */
+#define    MAC_XIF_FDPLX_LED               0x0040  /* FDPLXLED# active (low) */
+
+#define  REG_MAC_IPG0                      0x6040  /* inter-packet gap0 reg.
+						      recommended: 0x00 */
+#define  REG_MAC_IPG1                      0x6044  /* inter-packet gap1 reg
+						      recommended: 0x08 */
+#define  REG_MAC_IPG2                      0x6048  /* inter-packet gap2 reg
+						      recommended: 0x04 */
+#define  REG_MAC_SLOT_TIME                 0x604C  /* slot time reg
+						      recommended: 0x40 */
+#define  REG_MAC_FRAMESIZE_MIN             0x6050  /* min frame size reg
+						      recommended: 0x40 */
+
+/* FRAMESIZE_MAX holds both the max frame size as well as the max burst size.
+ * recommended value:  0x2000.05EE
+ */
+#define  REG_MAC_FRAMESIZE_MAX             0x6054  /* max frame size reg */
+#define    MAC_FRAMESIZE_MAX_BURST_MASK    0x3FFF0000 /* max burst size */
+#define    MAC_FRAMESIZE_MAX_BURST_SHIFT   16
+#define    MAC_FRAMESIZE_MAX_FRAME_MASK    0x00007FFF /* max frame size */
+#define    MAC_FRAMESIZE_MAX_FRAME_SHIFT   0
+#define  REG_MAC_PA_SIZE                   0x6058  /* PA size reg. number of
+						      preamble bytes that the
+						      TX MAC will xmit at the
+						      beginning of each frame
+						      value should be 2 or
+						      greater. recommended
+						      value: 0x07 */
+#define  REG_MAC_JAM_SIZE                  0x605C  /* jam size reg. duration
+						      of jam in units of media
+						      byte time. recommended
+						      value: 0x04 */
+#define  REG_MAC_ATTEMPT_LIMIT             0x6060  /* attempt limit reg. #
+						      of attempts TX MAC will
+						      make to xmit a frame
+						      before it resets its
+						      attempts counter. after
+						      the limit has been
+						      reached, TX MAC may or
+						      may not drop the frame
+						      dependent upon value
+						      in TX_MAC_CFG.
+						      recommended
+						      value: 0x10 */
+#define  REG_MAC_CTRL_TYPE                 0x6064  /* MAC control type reg.
+						      type field of a MAC
+						      ctrl frame. recommended
+						      value: 0x8808 */
+
+/* mac address registers: 0 - 44, 0x6080 - 0x6130, 4 8-bit bytes.
+ * register           contains                   comparison
+ *    0        16 MSB of primary MAC addr        [47:32] of DA field
+ *    1        16 middle bits ""                 [31:16] of DA field
+ *    2        16 LSB ""                         [15:0] of DA field
+ *    3*x      16MSB of alt MAC addr 1-15        [47:32] of DA field
+ *    4*x      16 middle bits ""                 [31:16]
+ *    5*x      16 LSB ""                         [15:0]
+ *    42       16 MSB of MAC CTRL addr           [47:32] of DA.
+ *    43       16 middle bits ""                 [31:16]
+ *    44       16 LSB ""                         [15:0]
+ *    MAC CTRL addr must be the reserved multicast addr for MAC CTRL frames.
+ *    if there is a match, MAC will set the bit for alternative address
+ *    filter pass [15]
+
+ *    here is the map of registers given MAC address notation: a:b:c:d:e:f
+ *                     ab             cd             ef
+ *    primary addr     reg 2          reg 1          reg 0
+ *    alt addr 1       reg 5          reg 4          reg 3
+ *    alt addr x       reg 5*x        reg 4*x        reg 3*x
+ *    ctrl addr        reg 44         reg 43         reg 42
+ */
+#define  REG_MAC_ADDR0                     0x6080  /* MAC address 0 reg */
+#define  REG_MAC_ADDRN(x)                  (REG_MAC_ADDR0 + (x)*4)
+#define  REG_MAC_ADDR_FILTER0              0x614C  /* address filter 0 reg
+						      [47:32] */
+#define  REG_MAC_ADDR_FILTER1              0x6150  /* address filter 1 reg
+						      [31:16] */
+#define  REG_MAC_ADDR_FILTER2              0x6154  /* address filter 2 reg
+						      [15:0] */
+#define  REG_MAC_ADDR_FILTER2_1_MASK       0x6158  /* address filter 2 and 1
+						      mask reg. 8-bit reg
+						      contains nibble mask for
+						      reg 2 and 1. */
+#define  REG_MAC_ADDR_FILTER0_MASK         0x615C  /* address filter 0 mask
+						      reg */
+
+/* hash table registers: 0 - 15, 0x6160 - 0x619C, 4 8-bit bytes
+ * 16-bit registers contain bits of the hash table.
+ * reg x  -> [16*(15 - x) + 15 : 16*(15 - x)].
+ * e.g., 15 -> [15:0], 0 -> [255:240]
+ */
+#define  REG_MAC_HASH_TABLE0               0x6160  /* hash table 0 reg */
+#define  REG_MAC_HASH_TABLEN(x)            (REG_MAC_HASH_TABLE0 + (x)*4)
+
+/* statistics registers. these registers generate an interrupt on
+ * overflow. recommended initialization: 0x0000. most are 16-bits except
+ * for PEAK_ATTEMPTS register which is 8 bits.
+ */
+#define  REG_MAC_COLL_NORMAL               0x61A0 /* normal collision
+						     counter. */
+#define  REG_MAC_COLL_FIRST                0x61A4 /* first attempt
+						     successful collision
+						     counter */
+#define  REG_MAC_COLL_EXCESS               0x61A8 /* excessive collision
+						     counter */
+#define  REG_MAC_COLL_LATE                 0x61AC /* late collision counter */
+#define  REG_MAC_TIMER_DEFER               0x61B0 /* defer timer. time base
+						     is the media byte
+						     clock/256 */
+#define  REG_MAC_ATTEMPTS_PEAK             0x61B4 /* peak attempts reg */
+#define  REG_MAC_RECV_FRAME                0x61B8 /* receive frame counter */
+#define  REG_MAC_LEN_ERR                   0x61BC /* length error counter */
+#define  REG_MAC_ALIGN_ERR                 0x61C0 /* alignment error counter */
+#define  REG_MAC_FCS_ERR                   0x61C4 /* FCS error counter */
+#define  REG_MAC_RX_CODE_ERR               0x61C8 /* RX code violation
+						     error counter */
+
+/* misc registers */
+#define  REG_MAC_RANDOM_SEED               0x61CC /* random number seed reg.
+						   10-bit register used as a
+						   seed  for the random number
+						   generator for the CSMA/CD
+						   backoff algorithm. only
+						   programmed after power-on
+						   reset and should be a
+						   random value which has a
+						   high likelihood of being
+						   unique for each MAC
+						   attached to a network
+						   segment (e.g., 10 LSB of
+						   MAC address) */
+
+/* ASUN: there's a PAUSE_TIMER (ro) described, but it's not in the address
+ *       map
+ */
+
+/* 27-bit register has the current state for key state machines in the MAC */
+#define  REG_MAC_STATE_MACHINE             0x61D0 /* (ro) state machine reg */
+#define    MAC_SM_RLM_MASK                 0x07800000
+#define    MAC_SM_RLM_SHIFT                23
+#define    MAC_SM_RX_FC_MASK               0x00700000
+#define    MAC_SM_RX_FC_SHIFT              20
+#define    MAC_SM_TLM_MASK                 0x000F0000
+#define    MAC_SM_TLM_SHIFT                16
+#define    MAC_SM_ENCAP_SM_MASK            0x0000F000
+#define    MAC_SM_ENCAP_SM_SHIFT           12
+#define    MAC_SM_TX_REQ_MASK              0x00000C00
+#define    MAC_SM_TX_REQ_SHIFT             10
+#define    MAC_SM_TX_FC_MASK               0x000003C0
+#define    MAC_SM_TX_FC_SHIFT              6
+#define    MAC_SM_FIFO_WRITE_SEL_MASK      0x00000038
+#define    MAC_SM_FIFO_WRITE_SEL_SHIFT     3
+#define    MAC_SM_TX_FIFO_EMPTY_MASK       0x00000007
+#define    MAC_SM_TX_FIFO_EMPTY_SHIFT      0
+
+/** MIF registers. the MIF can be programmed in either bit-bang or
+ *  frame mode.
+ **/
+#define  REG_MIF_BIT_BANG_CLOCK            0x6200 /* MIF bit-bang clock.
+						   1 -> 0 will generate a
+						   rising edge. 0 -> 1 will
+						   generate a falling edge. */
+#define  REG_MIF_BIT_BANG_DATA             0x6204 /* MIF bit-bang data. 1-bit
+						     register generates data */
+#define  REG_MIF_BIT_BANG_OUTPUT_EN        0x6208 /* MIF bit-bang output
+						     enable. enable when
+						     xmitting data from MIF to
+						     transceiver. */
+
+/* 32-bit register serves as an instruction register when the MIF is
+ * programmed in frame mode. load this register w/ a valid instruction
+ * (as per IEEE 802.3u MII spec). poll this register to check for instruction
+ * execution completion. during a read operation, this register will also
+ * contain the 16-bit data returned by the transceiver. unless specified
+ * otherwise, fields are considered "don't care" when polling for
+ * completion.
+ */
+#define  REG_MIF_FRAME                     0x620C /* MIF frame/output reg */
+#define    MIF_FRAME_START_MASK            0xC0000000 /* start of frame.
+							 load w/ 01 when
+							 issuing an instr */
+#define    MIF_FRAME_ST                    0x40000000 /* STart of frame */
+#define    MIF_FRAME_OPCODE_MASK           0x30000000 /* opcode. 01 for a
+							 write. 10 for a
+							 read */
+#define    MIF_FRAME_OP_READ               0x20000000 /* read OPcode */
+#define    MIF_FRAME_OP_WRITE              0x10000000 /* write OPcode */
+#define    MIF_FRAME_PHY_ADDR_MASK         0x0F800000 /* phy address. when
+							 issuing an instr,
+							 this field should be
+							 loaded w/ the XCVR
+							 addr */
+#define    MIF_FRAME_PHY_ADDR_SHIFT        23
+#define    MIF_FRAME_REG_ADDR_MASK         0x007C0000 /* register address.
+							 when issuing an instr,
+							 addr of register
+							 to be read/written */
+#define    MIF_FRAME_REG_ADDR_SHIFT        18
+#define    MIF_FRAME_TURN_AROUND_MSB       0x00020000 /* turn around, MSB.
+							 when issuing an instr,
+							 set this bit to 1 */
+#define    MIF_FRAME_TURN_AROUND_LSB       0x00010000 /* turn around, LSB.
+							 when issuing an instr,
+							 set this bit to 0.
+							 when polling for
+							 completion, 1 means
+							 that instr execution
+							 has been completed */
+#define    MIF_FRAME_DATA_MASK             0x0000FFFF /* instruction payload
+							 load with 16-bit data
+							 to be written in
+							 transceiver reg for a
+							 write. doesn't matter
+							 in a read. when
+							 polling for
+							 completion, field is
+							 "don't care" for write
+							 and 16-bit data
+							 returned by the
+							 transceiver for a
+							 read (if valid bit
+							 is set) */
+#define  REG_MIF_CFG                    0x6210 /* MIF config reg */
+#define    MIF_CFG_PHY_SELECT           0x0001 /* 1 -> select MDIO_1
+						  0 -> select MDIO_0 */
+#define    MIF_CFG_POLL_EN              0x0002 /* enable polling
+						  mechanism. if set,
+						  BB_MODE should be 0 */
+#define    MIF_CFG_BB_MODE              0x0004 /* 1 -> bit-bang mode
+						  0 -> frame mode */
+#define    MIF_CFG_POLL_REG_MASK        0x00F8 /* register address to be
+						  used by polling mode.
+						  only meaningful if POLL_EN
+						  is set to 1 */
+#define    MIF_CFG_POLL_REG_SHIFT       3
+#define    MIF_CFG_MDIO_0               0x0100 /* (ro) dual purpose.
+						  when MDIO_0 is idle,
+						  1 -> transceiver is
+						  connected to MDIO_0.
+						  when MIF is communicating
+						  w/ MDIO_0 in bit-bang
+						  mode, this bit indicates
+						  the incoming bit stream
+						  during a read op */
+#define    MIF_CFG_MDIO_1               0x0200 /* (ro) dual purpose.
+						  when MDIO_1 is idle,
+						  1 -> transceiver is
+						  connected to MDIO_1.
+						  when MIF is communicating
+						  w/ MDIO_1 in bit-bang
+						  mode, this bit indicates
+						  the incoming bit stream
+						  during a read op */
+#define    MIF_CFG_POLL_PHY_MASK        0x7C00 /* transceiver address to
+						  be polled */
+#define    MIF_CFG_POLL_PHY_SHIFT       10
+
+/* 16-bit register used to determine which bits in the POLL_STATUS portion of
+ * the MIF_STATUS register will cause an interrupt. if a mask bit is 0,
+ * corresponding bit of the POLL_STATUS will generate a MIF interrupt when
+ * set. DEFAULT: 0xFFFF
+ */
+#define  REG_MIF_MASK                      0x6214 /* MIF mask reg */
+
+/* 32-bit register used when in poll mode. auto-cleared after being read */
+#define  REG_MIF_STATUS                    0x6218 /* MIF status reg */
+#define    MIF_STATUS_POLL_DATA_MASK       0xFFFF0000 /* poll data contains
+							 the "latest image"
+							 update of the XCVR
+							 reg being read */
+#define    MIF_STATUS_POLL_DATA_SHIFT      16
+#define    MIF_STATUS_POLL_STATUS_MASK     0x0000FFFF /* poll status indicates
+							 which bits in the
+							 POLL_DATA field have
+							 changed since the
+							 MIF_STATUS reg was
+							 last read */
+#define    MIF_STATUS_POLL_STATUS_SHIFT    0
+
+/* 7-bit register has current state for all state machines in the MIF */
+#define  REG_MIF_STATE_MACHINE             0x621C /* MIF state machine reg */
+#define    MIF_SM_CONTROL_MASK             0x07   /* control state machine
+						     state */
+#define    MIF_SM_EXECUTION_MASK           0x60   /* execution state machine
+						     state */
+
+/** PCS/Serialink. the following registers are equivalent to the standard
+ *  MII management registers except that they're directly mapped in
+ *  Cassini's register space.
+ **/
+
+/* the auto-negotiation enable bit should be programmed the same at
+ * the link partner as in the local device to enable auto-negotiation to
+ * complete. when that bit is reprogrammed, auto-neg/manual config is
+ * restarted automatically.
+ * DEFAULT: 0x1040
+ */
+#define  REG_PCS_MII_CTRL                  0x9000 /* PCS MII control reg */
+#define    PCS_MII_CTRL_1000_SEL           0x0040 /* reads 1. ignored on
+						     writes */
+#define    PCS_MII_CTRL_COLLISION_TEST     0x0080 /* COL signal at the PCS
+						     to MAC interface is
+						     activated regardless
+						     of activity */
+#define    PCS_MII_CTRL_DUPLEX             0x0100 /* forced 0x0. PCS
+						     behaviour same for
+						     half and full dplx */
+#define    PCS_MII_RESTART_AUTONEG         0x0200 /* self clearing.
+						     restart auto-
+						     negotiation */
+#define    PCS_MII_ISOLATE                 0x0400 /* read as 0. ignored
+						     on writes */
+#define    PCS_MII_POWER_DOWN              0x0800 /* read as 0. ignored
+						     on writes */
+#define    PCS_MII_AUTONEG_EN              0x1000 /* default 1. PCS goes
+						     through automatic
+						     link config before it
+						     can be used. when 0,
+						     link can be used
+						     w/out any link config
+						     phase */
+#define    PCS_MII_10_100_SEL              0x2000 /* read as 0. ignored on
+						     writes */
+#define    PCS_MII_RESET                   0x8000 /* reset PCS. self-clears
+						     when done */
+
+/* DEFAULT: 0x0108 */
+#define  REG_PCS_MII_STATUS                0x9004 /* PCS MII status reg */
+#define    PCS_MII_STATUS_EXTEND_CAP       0x0001 /* reads 0 */
+#define    PCS_MII_STATUS_JABBER_DETECT    0x0002 /* reads 0 */
+#define    PCS_MII_STATUS_LINK_STATUS      0x0004 /* 1 -> link up.
+						     0 -> link down. 0 is
+						     latched so that 0 is
+						     kept until read. read
+						     2x to determine if the
+						     link has gone up again */
+#define    PCS_MII_STATUS_AUTONEG_ABLE     0x0008 /* reads 1 (able to perform
+						     auto-neg) */
+#define    PCS_MII_STATUS_REMOTE_FAULT     0x0010 /* 1 -> remote fault detected
+						     from received link code
+						     word. only valid after
+						     auto-neg completed */
+#define    PCS_MII_STATUS_AUTONEG_COMP     0x0020 /* 1 -> auto-negotiation
+						          completed
+						     0 -> auto-negotiation not
+						     completed */
+#define    PCS_MII_STATUS_EXTEND_STATUS    0x0100 /* reads as 1. used as an
+						     indication that this is
+						     a 1000 Base-X PHY. writes
+						     to it are ignored */
+
+/* used during auto-negotiation.
+ * DEFAULT: 0x00E0
+ */
+#define  REG_PCS_MII_ADVERT                0x9008 /* PCS MII advertisement
+						     reg */
+#define    PCS_MII_ADVERT_FD               0x0020  /* advertise full duplex
+						      1000 Base-X */
+#define    PCS_MII_ADVERT_HD               0x0040  /* advertise half-duplex
+						      1000 Base-X */
+#define    PCS_MII_ADVERT_SYM_PAUSE        0x0080  /* advertise PAUSE
+						      symmetric capability */
+#define    PCS_MII_ADVERT_ASYM_PAUSE       0x0100  /* advertises PAUSE
+						      asymmetric capability */
+#define    PCS_MII_ADVERT_RF_MASK          0x3000 /* remote fault. write bit13
+						     to optionally indicate to
+						     link partner that chip is
+						     going off-line. bit12 will
+						     get set when signal
+						     detect == FAIL and will
+						     remain set until
+						     successful negotiation */
+#define    PCS_MII_ADVERT_ACK              0x4000 /* (ro) */
+#define    PCS_MII_ADVERT_NEXT_PAGE        0x8000 /* (ro) forced 0x0 */
+
+/* contents updated as a result of autonegotiation. layout and definitions
+ * identical to PCS_MII_ADVERT
+ */
+#define  REG_PCS_MII_LPA                   0x900C /* PCS MII link partner
+						     ability reg */
+#define    PCS_MII_LPA_FD             PCS_MII_ADVERT_FD
+#define    PCS_MII_LPA_HD             PCS_MII_ADVERT_HD
+#define    PCS_MII_LPA_SYM_PAUSE      PCS_MII_ADVERT_SYM_PAUSE
+#define    PCS_MII_LPA_ASYM_PAUSE     PCS_MII_ADVERT_ASYM_PAUSE
+#define    PCS_MII_LPA_RF_MASK        PCS_MII_ADVERT_RF_MASK
+#define    PCS_MII_LPA_ACK            PCS_MII_ADVERT_ACK
+#define    PCS_MII_LPA_NEXT_PAGE      PCS_MII_ADVERT_NEXT_PAGE
+
+/* DEFAULT: 0x0 */
+#define  REG_PCS_CFG                       0x9010 /* PCS config reg */
+#define    PCS_CFG_EN                      0x01   /* enable PCS. must be
+						     0 when modifying
+						     PCS_MII_ADVERT */
+#define    PCS_CFG_SD_OVERRIDE             0x02   /* sets signal detect to
+						     OK. bit is
+						     non-resettable */
+#define    PCS_CFG_SD_ACTIVE_LOW           0x04   /* changes interpretation
+						     of optical signal to make
+						     signal detect okay when
+						     signal is low */
+#define    PCS_CFG_JITTER_STUDY_MASK       0x18   /* used to make jitter
+						     measurements. a single
+						     code group is xmitted
+						     regularly.
+						     0x0 = normal operation
+						     0x1 = high freq test
+						           pattern, D21.5
+						     0x2 = low freq test
+						           pattern, K28.7
+						     0x3 = reserved */
+#define    PCS_CFG_10MS_TIMER_OVERRIDE     0x20   /* shortens 10-20ms auto-
+						     negotiation timer to
+						     a few cycles for test
+						     purposes */
+
+/* used for diagnostic purposes. bits 20-22 autoclear on read */
+#define  REG_PCS_STATE_MACHINE             0x9014 /* (ro) PCS state machine
+						     and diagnostic reg */
+#define    PCS_SM_TX_STATE_MASK            0x0000000F /* 0 and 1 indicate
+							 xmission of idle.
+							 otherwise, xmission of
+							 a packet */
+#define    PCS_SM_RX_STATE_MASK            0x000000F0 /* 0 indicates reception
+							 of idle. otherwise,
+							 reception of packet */
+#define    PCS_SM_WORD_SYNC_STATE_MASK     0x00000700 /* 0 indicates loss of
+							 sync */
+#define    PCS_SM_SEQ_DETECT_STATE_MASK    0x00001800 /* cycling through 0-3
+							 indicates reception of
+							 Config codes. cycling
+							 through 0-1 indicates
+							 reception of idles */
+#define    PCS_SM_LINK_STATE_MASK          0x0001E000
+#define        SM_LINK_STATE_UP            0x00016000 /* link state is up */
+
+#define    PCS_SM_LOSS_LINK_C              0x00100000 /* loss of link due to
+							 recept of Config
+							 codes */
+#define    PCS_SM_LOSS_LINK_SYNC           0x00200000 /* loss of link due to
+							 loss of sync */
+#define    PCS_SM_LOSS_SIGNAL_DETECT       0x00400000 /* signal detect goes
+							 from OK to FAIL. bit29
+							 will also be set if
+							 this is set */
+#define    PCS_SM_NO_LINK_BREAKLINK        0x01000000 /* link not up due to
+							receipt of breaklink
+							C codes from partner.
+							C codes w/ 0 content
+							received triggering
+							start/restart of
+							autonegotiation.
+							should be sent for
+							no longer than 20ms */
+#define    PCS_SM_NO_LINK_SERDES           0x02000000 /* serdes being
+							initialized. see serdes
+							state reg */
+#define    PCS_SM_NO_LINK_C                0x04000000 /* C codes not stable or
+							 not received */
+#define    PCS_SM_NO_LINK_SYNC             0x08000000 /* word sync not
+							 achieved */
+#define    PCS_SM_NO_LINK_WAIT_C           0x10000000 /* waiting for C codes
+							 w/ ack bit set */
+#define    PCS_SM_NO_LINK_NO_IDLE          0x20000000 /* link partner continues
+							 to send C codes
+							 instead of idle
+							 symbols or pkt data */
+
+/* this register indicates interrupt changes in specific PCS MII status bits.
+ * PCS_INT may be masked at the ISR level. only a single bit is implemented
+ * for link status change.
+ */
+#define  REG_PCS_INTR_STATUS               0x9018 /* PCS interrupt status */
+#define    PCS_INTR_STATUS_LINK_CHANGE     0x04   /* link status has changed
+						     since last read */
+
+/* control which network interface is used. no more than one bit should
+ * be set.
+ * DEFAULT: none
+ */
+#define  REG_PCS_DATAPATH_MODE             0x9050 /* datapath mode reg */
+#define    PCS_DATAPATH_MODE_MII           0x00 /* PCS is not used and
+						   MII/GMII is selected.
+						   selection between MII and
+						   GMII is controlled by
+						   XIF_CFG */
+#define    PCS_DATAPATH_MODE_SERDES        0x02 /* PCS is used via the
+						   10-bit interface */
+
+/* input to serdes chip or serialink block */
+#define  REG_PCS_SERDES_CTRL              0x9054 /* serdes control reg */
+#define    PCS_SERDES_CTRL_LOOPBACK       0x01   /* enable loopback on
+						    serdes interface */
+#define    PCS_SERDES_CTRL_SYNCD_EN       0x02   /* enable sync carrier
+						    detection. should be
+						    0x0 for normal
+						    operation */
+#define    PCS_SERDES_CTRL_LOCKREF       0x04   /* frequency-lock RBC[0:1]
+						   to REFCLK when set.
+						   when clear, receiver
+						   clock locks to incoming
+						   serial data */
+
+/* multiplex test outputs into the PROM address (PA_3 through PA_0) pins.
+ * should be 0x0 for normal operations.
+ * 0b000          normal operation, PROM address[3:0] selected
+ * 0b001          rxdma req, rxdma ack, rxdma ready, rxdma read
+ * 0b010          rxmac req, rx ack, rx tag, rx clk shared
+ * 0b011          txmac req, tx ack, tx tag, tx retry req
+ * 0b100          tx tp3, tx tp2, tx tp1, tx tp0
+ * 0b101          R period RX, R period TX, R period HP, R period BIM
+ * DEFAULT: 0x0
+ */
+#define  REG_PCS_SHARED_OUTPUT_SEL         0x9058 /* shared output select */
+#define    PCS_SOS_PROM_ADDR_MASK          0x0007
+
+/* used for diagnostics. this register indicates progress of the SERDES
+ * boot up.
+ * 0b00       undergoing reset
+ * 0b01       waiting 500us while lockrefn is asserted
+ * 0b10       waiting for comma detect
+ * 0b11       receive data is synchronized
+ * DEFAULT: 0x0
+ */
+#define  REG_PCS_SERDES_STATE              0x905C /* (ro) serdes state */
+#define    PCS_SERDES_STATE_MASK           0x03
+
+/* used for diagnostics. indicates number of packets transmitted or received.
+ * counters rollover w/out generating an interrupt.
+ * DEFAULT: 0x0
+ */
+#define  REG_PCS_PACKET_COUNT              0x9060 /* (ro) PCS packet counter */
+#define    PCS_PACKET_COUNT_TX             0x000007FF /* pkts xmitted by PCS */
+#define    PCS_PACKET_COUNT_RX             0x07FF0000 /* pkts recvd by PCS
+							 whether they
+							 encountered an error
+							 or not */
+
+/** LocalBus Devices. the following provides run-time access to the
+ *  Cassini's PROM
+ ***/
+#define  REG_EXPANSION_ROM_RUN_START       0x100000 /* expansion rom run time
+						       access */
+#define  REG_EXPANSION_ROM_RUN_END         0x17FFFF
+
+#define  REG_SECOND_LOCALBUS_START         0x180000 /* secondary local bus
+						       device */
+#define  REG_SECOND_LOCALBUS_END           0x1FFFFF
+
+/* entropy device */
+#define  REG_ENTROPY_START                 REG_SECOND_LOCALBUS_START
+#define  REG_ENTROPY_DATA                  (REG_ENTROPY_START + 0x00)
+#define  REG_ENTROPY_STATUS                (REG_ENTROPY_START + 0x04)
+#define      ENTROPY_STATUS_DRDY           0x01
+#define      ENTROPY_STATUS_BUSY           0x02
+#define      ENTROPY_STATUS_CIPHER         0x04
+#define      ENTROPY_STATUS_BYPASS_MASK    0x18
+#define  REG_ENTROPY_MODE                  (REG_ENTROPY_START + 0x05)
+#define      ENTROPY_MODE_KEY_MASK         0x07
+#define      ENTROPY_MODE_ENCRYPT          0x40
+#define  REG_ENTROPY_RAND_REG              (REG_ENTROPY_START + 0x06)
+#define  REG_ENTROPY_RESET                 (REG_ENTROPY_START + 0x07)
+#define      ENTROPY_RESET_DES_IO          0x01
+#define      ENTROPY_RESET_STC_MODE        0x02
+#define      ENTROPY_RESET_KEY_CACHE       0x04
+#define      ENTROPY_RESET_IV              0x08
+#define  REG_ENTROPY_IV                    (REG_ENTROPY_START + 0x08)
+#define  REG_ENTROPY_KEY0                  (REG_ENTROPY_START + 0x10)
+#define  REG_ENTROPY_KEYN(x)               (REG_ENTROPY_KEY0 + 4*(x))
+
+/* phys of interest w/ their special mii registers */
+#define PHY_LUCENT_B0     0x00437421
+#define   LUCENT_MII_REG      0x1F
+
+#define PHY_NS_DP83065    0x20005c78
+#define   DP83065_MII_MEM     0x16
+#define   DP83065_MII_REGD    0x1D
+#define   DP83065_MII_REGE    0x1E
+
+#define PHY_BROADCOM_5411 0x00206071
+#define PHY_BROADCOM_B0   0x00206050
+#define   BROADCOM_MII_REG4   0x14
+#define   BROADCOM_MII_REG5   0x15
+#define   BROADCOM_MII_REG7   0x17
+#define   BROADCOM_MII_REG8   0x18
+
+#define   CAS_MII_ANNPTR          0x07
+#define   CAS_MII_ANNPRR          0x08
+#define   CAS_MII_1000_CTRL       0x09
+#define   CAS_MII_1000_STATUS     0x0A
+#define   CAS_MII_1000_EXTEND     0x0F
+
+#define   CAS_BMSR_1000_EXTEND    0x0100 /* supports 1000Base-T extended status */
+/*
+ * if autoneg is disabled, here's the table:
+ * BMCR_SPEED100 = 100Mbps
+ * BMCR_SPEED1000 = 1000Mbps
+ * ~(BMCR_SPEED100 | BMCR_SPEED1000) = 10Mbps
+ */
+#define   CAS_BMCR_SPEED1000      0x0040  /* Select 1000Mbps */
+
+#define   CAS_ADVERTISE_1000HALF   0x0100
+#define   CAS_ADVERTISE_1000FULL   0x0200
+#define   CAS_ADVERTISE_PAUSE      0x0400
+#define   CAS_ADVERTISE_ASYM_PAUSE 0x0800
+
+/* regular lpa register */
+#define   CAS_LPA_PAUSE	           CAS_ADVERTISE_PAUSE
+#define   CAS_LPA_ASYM_PAUSE       CAS_ADVERTISE_ASYM_PAUSE
+
+/* 1000_STATUS register */
+#define   CAS_LPA_1000HALF        0x0400
+#define   CAS_LPA_1000FULL        0x0800
+
+#define   CAS_EXTEND_1000XFULL    0x8000
+#define   CAS_EXTEND_1000XHALF    0x4000
+#define   CAS_EXTEND_1000TFULL    0x2000
+#define   CAS_EXTEND_1000THALF    0x1000
+
+/* cassini header parser firmware */
+typedef struct cas_hp_inst {
+	const char *note;
+
+	u16 mask, val;
+
+	u8 op;
+	u8 soff, snext;	/* if match succeeds, new offset and match */
+	u8 foff, fnext;	/* if match fails, new offset and match */
+	/* output info */
+	u8 outop;    /* output opcode */
+
+	u16 outarg;  /* output argument */
+	u8 outenab;  /* output enable: 0 = not, 1 = if match
+			 2 = if !match, 3 = always */
+	u8 outshift; /* barrel shift right, 4 bits */
+	u16 outmask;
+} cas_hp_inst_t;
+
+/* comparison */
+#define OP_EQ     0 /* packet == value */
+#define OP_LT     1 /* packet < value */
+#define OP_GT     2 /* packet > value */
+#define OP_NP     3 /* new packet */
+
+/* output opcodes */
+#define	CL_REG	0
+#define	LD_FID	1
+#define	LD_SEQ	2
+#define	LD_CTL	3
+#define	LD_SAP	4
+#define	LD_R1	5
+#define	LD_L3	6
+#define	LD_SUM	7
+#define	LD_HDR	8
+#define	IM_FID	9
+#define	IM_SEQ	10
+#define	IM_SAP	11
+#define	IM_R1	12
+#define	IM_CTL	13
+#define	LD_LEN	14
+#define	ST_FLG	15
+
+/* match setp #s for IP4TCP4 */
+#define S1_PCKT         0
+#define S1_VLAN         1
+#define S1_CFI          2
+#define S1_8023         3
+#define S1_LLC          4
+#define S1_LLCc         5
+#define S1_IPV4         6
+#define S1_IPV4c        7
+#define S1_IPV4F        8
+#define S1_TCP44        9
+#define S1_IPV6         10
+#define S1_IPV6L        11
+#define S1_IPV6c        12
+#define S1_TCP64        13
+#define S1_TCPSQ        14
+#define S1_TCPFG        15
+#define	S1_TCPHL	16
+#define	S1_TCPHc	17
+#define	S1_CLNP		18
+#define	S1_CLNP2	19
+#define	S1_DROP		20
+#define	S2_HTTP		21
+#define	S1_ESP4		22
+#define	S1_AH4		23
+#define	S1_ESP6		24
+#define	S1_AH6		25
+
+#define CAS_PROG_IP46TCP4_PREAMBLE \
+{ "packet arrival?", 0xffff, 0x0000, OP_NP,  6, S1_VLAN,  0, S1_PCKT,  \
+  CL_REG, 0x3ff, 1, 0x0, 0x0000}, \
+{ "VLAN?", 0xffff, 0x8100, OP_EQ,  1, S1_CFI,   0, S1_8023,  \
+  IM_CTL, 0x00a,  3, 0x0, 0xffff}, \
+{ "CFI?", 0x1000, 0x1000, OP_EQ,  0, S1_DROP,  1, S1_8023, \
+  CL_REG, 0x000,  0, 0x0, 0x0000}, \
+{ "8023?", 0xffff, 0x0600, OP_LT,  1, S1_LLC,   0, S1_IPV4, \
+  CL_REG, 0x000,  0, 0x0, 0x0000}, \
+{ "LLC?", 0xffff, 0xaaaa, OP_EQ,  1, S1_LLCc,  0, S1_CLNP, \
+  CL_REG, 0x000,  0, 0x0, 0x0000}, \
+{ "LLCc?", 0xff00, 0x0300, OP_EQ,  2, S1_IPV4,  0, S1_CLNP, \
+  CL_REG, 0x000,  0, 0x0, 0x0000}, \
+{ "IPV4?", 0xffff, 0x0800, OP_EQ,  1, S1_IPV4c, 0, S1_IPV6, \
+  LD_SAP, 0x100,  3, 0x0, 0xffff}, \
+{ "IPV4 cont?", 0xff00, 0x4500, OP_EQ,  3, S1_IPV4F, 0, S1_CLNP, \
+  LD_SUM, 0x00a,  1, 0x0, 0x0000}, \
+{ "IPV4 frag?", 0x3fff, 0x0000, OP_EQ,  1, S1_TCP44, 0, S1_CLNP, \
+  LD_LEN, 0x03e,  1, 0x0, 0xffff}, \
+{ "TCP44?", 0x00ff, 0x0006, OP_EQ,  7, S1_TCPSQ, 0, S1_CLNP, \
+  LD_FID, 0x182,  1, 0x0, 0xffff}, /* FID IP4&TCP src+dst */ \
+{ "IPV6?", 0xffff, 0x86dd, OP_EQ,  1, S1_IPV6L, 0, S1_CLNP,  \
+  LD_SUM, 0x015,  1, 0x0, 0x0000}, \
+{ "IPV6 len", 0xf000, 0x6000, OP_EQ,  0, S1_IPV6c, 0, S1_CLNP, \
+  IM_R1,  0x128,  1, 0x0, 0xffff}, \
+{ "IPV6 cont?", 0x0000, 0x0000, OP_EQ,  3, S1_TCP64, 0, S1_CLNP, \
+  LD_FID, 0x484,  1, 0x0, 0xffff}, /* FID IP6&TCP src+dst */ \
+{ "TCP64?", 0xff00, 0x0600, OP_EQ, 18, S1_TCPSQ, 0, S1_CLNP, \
+  LD_LEN, 0x03f,  1, 0x0, 0xffff}
+
+#ifdef USE_HP_IP46TCP4
+static cas_hp_inst_t cas_prog_ip46tcp4tab[] = {
+	CAS_PROG_IP46TCP4_PREAMBLE,
+	{ "TCP seq", /* DADDR should point to dest port */
+	  0x0000, 0x0000, OP_EQ, 0, S1_TCPFG, 4, S1_TCPFG, LD_SEQ,
+	  0x081,  3, 0x0, 0xffff}, /* Load TCP seq # */
+	{ "TCP control flags", 0x0000, 0x0000, OP_EQ,  0, S1_TCPHL, 0,
+	  S1_TCPHL, ST_FLG, 0x045,  3, 0x0, 0x002f}, /* Load TCP flags */
+	{ "TCP length", 0x0000, 0x0000, OP_EQ,  0, S1_TCPHc, 0,
+	  S1_TCPHc, LD_R1,  0x205,  3, 0xB, 0xf000},
+	{ "TCP length cont", 0x0000, 0x0000, OP_EQ,  0, S1_PCKT,  0,
+	  S1_PCKT,  LD_HDR, 0x0ff,  3, 0x0, 0xffff},
+	{ "Cleanup", 0x0000, 0x0000, OP_EQ,  0, S1_CLNP2,  0, S1_CLNP2,
+	  IM_CTL, 0x001,  3, 0x0, 0x0001},
+	{ "Cleanup 2", 0x0000, 0x0000, OP_EQ,  0, S1_PCKT,  0, S1_PCKT,
+	  IM_CTL, 0x000,  0, 0x0, 0x0000},
+	{ "Drop packet", 0x0000, 0x0000, OP_EQ,  0, S1_PCKT,  0, S1_PCKT,
+	  IM_CTL, 0x080,  3, 0x0, 0xffff},
+	{ NULL },
+};
+#ifdef HP_IP46TCP4_DEFAULT
+#define CAS_HP_FIRMWARE               cas_prog_ip46tcp4tab
+#endif
+#endif
+
+/*
+ * Alternate table load which excludes HTTP server traffic from reassembly.
+ * It is substantially similar to the basic table, with one extra state
+ * and a few extra compares. */
+#ifdef USE_HP_IP46TCP4NOHTTP
+static cas_hp_inst_t cas_prog_ip46tcp4nohttptab[] = {
+	CAS_PROG_IP46TCP4_PREAMBLE,
+	{ "TCP seq", /* DADDR should point to dest port */
+	  0xFFFF, 0x0080, OP_EQ,  0, S2_HTTP,  0, S1_TCPFG, LD_SEQ,
+	  0x081,  3, 0x0, 0xffff} , /* Load TCP seq # */
+	{ "TCP control flags", 0xFFFF, 0x8080, OP_EQ,  0, S2_HTTP,  0,
+	  S1_TCPHL, ST_FLG, 0x145,  2, 0x0, 0x002f, }, /* Load TCP flags */
+	{ "TCP length", 0x0000, 0x0000, OP_EQ,  0, S1_TCPHc, 0, S1_TCPHc,
+	  LD_R1,  0x205,  3, 0xB, 0xf000},
+	{ "TCP length cont", 0x0000, 0x0000, OP_EQ,  0, S1_PCKT,  0, S1_PCKT,
+	  LD_HDR, 0x0ff,  3, 0x0, 0xffff},
+	{ "Cleanup", 0x0000, 0x0000, OP_EQ,  0, S1_CLNP2,  0, S1_CLNP2,
+	  IM_CTL, 0x001,  3, 0x0, 0x0001},
+	{ "Cleanup 2", 0x0000, 0x0000, OP_EQ,  0, S1_PCKT,  0, S1_PCKT,
+	  CL_REG, 0x002,  3, 0x0, 0x0000},
+	{ "Drop packet", 0x0000, 0x0000, OP_EQ,  0, S1_PCKT,  0, S1_PCKT,
+	  IM_CTL, 0x080,  3, 0x0, 0xffff},
+	{ "No HTTP", 0x0000, 0x0000, OP_EQ,  0, S1_PCKT,  0, S1_PCKT,
+	  IM_CTL, 0x044,  3, 0x0, 0xffff},
+	{ NULL },
+};
+#ifdef HP_IP46TCP4NOHTTP_DEFAULT
+#define CAS_HP_FIRMWARE               cas_prog_ip46tcp4nohttptab
+#endif
+#endif
+
+/* match step #s for IP4FRAG */
+#define	S3_IPV6c	11
+#define	S3_TCP64	12
+#define	S3_TCPSQ	13
+#define	S3_TCPFG	14
+#define	S3_TCPHL	15
+#define	S3_TCPHc	16
+#define	S3_FRAG		17
+#define	S3_FOFF		18
+#define	S3_CLNP		19
+
+#ifdef USE_HP_IP4FRAG
+static cas_hp_inst_t cas_prog_ip4fragtab[] = {
+	{ "packet arrival?", 0xffff, 0x0000, OP_NP,  6, S1_VLAN,  0, S1_PCKT,
+	  CL_REG, 0x3ff, 1, 0x0, 0x0000},
+	{ "VLAN?", 0xffff, 0x8100, OP_EQ,  1, S1_CFI,   0, S1_8023,
+	  IM_CTL, 0x00a,  3, 0x0, 0xffff},
+	{ "CFI?", 0x1000, 0x1000, OP_EQ,  0, S3_CLNP,  1, S1_8023,
+	  CL_REG, 0x000,  0, 0x0, 0x0000},
+	{ "8023?", 0xffff, 0x0600, OP_LT,  1, S1_LLC,   0, S1_IPV4,
+	  CL_REG, 0x000,  0, 0x0, 0x0000},
+	{ "LLC?", 0xffff, 0xaaaa, OP_EQ,  1, S1_LLCc,  0, S3_CLNP,
+	  CL_REG, 0x000,  0, 0x0, 0x0000},
+	{ "LLCc?",0xff00, 0x0300, OP_EQ,  2, S1_IPV4,  0, S3_CLNP,
+	  CL_REG, 0x000,  0, 0x0, 0x0000},
+	{ "IPV4?", 0xffff, 0x0800, OP_EQ,  1, S1_IPV4c, 0, S1_IPV6,
+	  LD_SAP, 0x100,  3, 0x0, 0xffff},
+	{ "IPV4 cont?", 0xff00, 0x4500, OP_EQ,  3, S1_IPV4F, 0, S3_CLNP,
+	  LD_SUM, 0x00a,  1, 0x0, 0x0000},
+	{ "IPV4 frag?", 0x3fff, 0x0000, OP_EQ,  1, S1_TCP44, 0, S3_FRAG,
+	  LD_LEN, 0x03e,  3, 0x0, 0xffff},
+	{ "TCP44?", 0x00ff, 0x0006, OP_EQ,  7, S3_TCPSQ, 0, S3_CLNP,
+	  LD_FID, 0x182,  3, 0x0, 0xffff}, /* FID IP4&TCP src+dst */
+	{ "IPV6?", 0xffff, 0x86dd, OP_EQ,  1, S3_IPV6c, 0, S3_CLNP,
+	  LD_SUM, 0x015,  1, 0x0, 0x0000},
+	{ "IPV6 cont?", 0xf000, 0x6000, OP_EQ,  3, S3_TCP64, 0, S3_CLNP,
+	  LD_FID, 0x484,  1, 0x0, 0xffff}, /* FID IP6&TCP src+dst */
+	{ "TCP64?", 0xff00, 0x0600, OP_EQ, 18, S3_TCPSQ, 0, S3_CLNP,
+	  LD_LEN, 0x03f,  1, 0x0, 0xffff},
+	{ "TCP seq",	/* DADDR should point to dest port */
+	  0x0000, 0x0000, OP_EQ,  0, S3_TCPFG, 4, S3_TCPFG, LD_SEQ,
+	  0x081,  3, 0x0, 0xffff}, /* Load TCP seq # */
+	{ "TCP control flags", 0x0000, 0x0000, OP_EQ,  0, S3_TCPHL, 0,
+	  S3_TCPHL, ST_FLG, 0x045,  3, 0x0, 0x002f}, /* Load TCP flags */
+	{ "TCP length", 0x0000, 0x0000, OP_EQ,  0, S3_TCPHc, 0, S3_TCPHc,
+	  LD_R1,  0x205,  3, 0xB, 0xf000},
+	{ "TCP length cont", 0x0000, 0x0000, OP_EQ,  0, S1_PCKT,  0, S1_PCKT,
+	  LD_HDR, 0x0ff,  3, 0x0, 0xffff},
+	{ "IP4 Fragment", 0x0000, 0x0000, OP_EQ,  0, S3_FOFF,  0, S3_FOFF,
+	  LD_FID, 0x103,  3, 0x0, 0xffff}, /* FID IP4 src+dst */
+	{ "IP4 frag offset", 0x0000, 0x0000, OP_EQ,  0, S3_FOFF,  0, S3_FOFF,
+	  LD_SEQ, 0x040,  1, 0xD, 0xfff8},
+	{ "Cleanup", 0x0000, 0x0000, OP_EQ,  0, S1_PCKT,  0, S1_PCKT,
+	  IM_CTL, 0x001,  3, 0x0, 0x0001},
+	{ NULL },
+};
+#ifdef HP_IP4FRAG_DEFAULT
+#define CAS_HP_FIRMWARE               cas_prog_ip4fragtab
+#endif
+#endif
+
+/*
+ * Alternate table which does batching without reassembly
+ */
+#ifdef USE_HP_IP46TCP4BATCH
+static cas_hp_inst_t cas_prog_ip46tcp4batchtab[] = {
+	CAS_PROG_IP46TCP4_PREAMBLE,
+	{ "TCP seq",	/* DADDR should point to dest port */
+	  0x0000, 0x0000, OP_EQ,  0, S1_TCPFG, 0, S1_TCPFG, LD_SEQ,
+	  0x081,  3, 0x0, 0xffff}, /* Load TCP seq # */
+	{ "TCP control flags", 0x0000, 0x0000, OP_EQ,  0, S1_TCPHL, 0,
+	  S1_TCPHL, ST_FLG, 0x000,  3, 0x0, 0x0000}, /* Load TCP flags */
+	{ "TCP length", 0x0000, 0x0000, OP_EQ,  0, S1_TCPHc, 0,
+	  S1_TCPHc, LD_R1,  0x205,  3, 0xB, 0xf000},
+	{ "TCP length cont", 0x0000, 0x0000, OP_EQ,  0, S1_PCKT,  0,
+	  S1_PCKT,  IM_CTL, 0x040,  3, 0x0, 0xffff}, /* set batch bit */
+	{ "Cleanup", 0x0000, 0x0000, OP_EQ,  0, S1_PCKT,  0, S1_PCKT,
+	  IM_CTL, 0x001,  3, 0x0, 0x0001},
+	{ "Drop packet", 0x0000, 0x0000, OP_EQ,  0, S1_PCKT,  0,
+	  S1_PCKT,  IM_CTL, 0x080,  3, 0x0, 0xffff},
+	{ NULL },
+};
+#ifdef HP_IP46TCP4BATCH_DEFAULT
+#define CAS_HP_FIRMWARE               cas_prog_ip46tcp4batchtab
+#endif
+#endif
+
+/* Workaround for Cassini rev2 descriptor corruption problem.
+ * Does batching without reassembly, and sets the SAP to a known
+ * data pattern for all packets.
+ */
+#ifdef USE_HP_WORKAROUND
+static cas_hp_inst_t  cas_prog_workaroundtab[] = {
+	{ "packet arrival?", 0xffff, 0x0000, OP_NP,  6, S1_VLAN,  0,
+	  S1_PCKT,  CL_REG, 0x3ff,  1, 0x0, 0x0000} ,
+	{ "VLAN?", 0xffff, 0x8100, OP_EQ,  1, S1_CFI, 0, S1_8023,
+	  IM_CTL, 0x04a,  3, 0x0, 0xffff},
+	{ "CFI?", 0x1000, 0x1000, OP_EQ,  0, S1_CLNP,  1, S1_8023,
+	  CL_REG, 0x000,  0, 0x0, 0x0000},
+	{ "8023?", 0xffff, 0x0600, OP_LT,  1, S1_LLC,   0, S1_IPV4,
+	  CL_REG, 0x000,  0, 0x0, 0x0000},
+	{ "LLC?", 0xffff, 0xaaaa, OP_EQ,  1, S1_LLCc,  0, S1_CLNP,
+	  CL_REG, 0x000,  0, 0x0, 0x0000},
+	{ "LLCc?", 0xff00, 0x0300, OP_EQ,  2, S1_IPV4,  0, S1_CLNP,
+	  CL_REG, 0x000,  0, 0x0, 0x0000},
+	{ "IPV4?", 0xffff, 0x0800, OP_EQ,  1, S1_IPV4c, 0, S1_IPV6,
+	  IM_SAP, 0x6AE,  3, 0x0, 0xffff},
+	{ "IPV4 cont?", 0xff00, 0x4500, OP_EQ,  3, S1_IPV4F, 0, S1_CLNP,
+	  LD_SUM, 0x00a,  1, 0x0, 0x0000},
+	{ "IPV4 frag?", 0x3fff, 0x0000, OP_EQ,  1, S1_TCP44, 0, S1_CLNP,
+	  LD_LEN, 0x03e,  1, 0x0, 0xffff},
+	{ "TCP44?", 0x00ff, 0x0006, OP_EQ,  7, S1_TCPSQ, 0, S1_CLNP,
+	  LD_FID, 0x182,  3, 0x0, 0xffff}, /* FID IP4&TCP src+dst */
+	{ "IPV6?", 0xffff, 0x86dd, OP_EQ,  1, S1_IPV6L, 0, S1_CLNP,
+	  LD_SUM, 0x015,  1, 0x0, 0x0000},
+	{ "IPV6 len", 0xf000, 0x6000, OP_EQ,  0, S1_IPV6c, 0, S1_CLNP,
+	  IM_R1,  0x128,  1, 0x0, 0xffff},
+	{ "IPV6 cont?", 0x0000, 0x0000, OP_EQ,  3, S1_TCP64, 0, S1_CLNP,
+	  LD_FID, 0x484,  1, 0x0, 0xffff}, /* FID IP6&TCP src+dst */
+	{ "TCP64?", 0xff00, 0x0600, OP_EQ, 18, S1_TCPSQ, 0, S1_CLNP,
+	  LD_LEN, 0x03f,  1, 0x0, 0xffff},
+	{ "TCP seq",      /* DADDR should point to dest port */
+	  0x0000, 0x0000, OP_EQ,  0, S1_TCPFG, 4, S1_TCPFG, LD_SEQ,
+	  0x081,  3, 0x0, 0xffff}, /* Load TCP seq # */
+	{ "TCP control flags", 0x0000, 0x0000, OP_EQ,  0, S1_TCPHL, 0,
+	  S1_TCPHL, ST_FLG, 0x045,  3, 0x0, 0x002f}, /* Load TCP flags */
+	{ "TCP length", 0x0000, 0x0000, OP_EQ,  0, S1_TCPHc, 0, S1_TCPHc,
+	  LD_R1,  0x205,  3, 0xB, 0xf000},
+	{ "TCP length cont", 0x0000, 0x0000, OP_EQ,  0, S1_PCKT,  0,
+	  S1_PCKT,  LD_HDR, 0x0ff,  3, 0x0, 0xffff},
+	{ "Cleanup", 0x0000, 0x0000, OP_EQ,  0, S1_CLNP2, 0, S1_CLNP2,
+	  IM_SAP, 0x6AE,  3, 0x0, 0xffff} ,
+	{ "Cleanup 2", 0x0000, 0x0000, OP_EQ,  0, S1_PCKT,  0, S1_PCKT,
+	  IM_CTL, 0x001,  3, 0x0, 0x0001},
+	{ NULL },
+};
+#ifdef HP_WORKAROUND_DEFAULT
+#define CAS_HP_FIRMWARE               cas_prog_workaroundtab
+#endif
+#endif
+
+#ifdef USE_HP_ENCRYPT
+static cas_hp_inst_t  cas_prog_encryptiontab[] = {
+	{ "packet arrival?", 0xffff, 0x0000, OP_NP,  6, S1_VLAN,  0,
+	  S1_PCKT,  CL_REG, 0x3ff,  1, 0x0, 0x0000},
+	{ "VLAN?", 0xffff, 0x8100, OP_EQ,  1, S1_CFI,   0, S1_8023,
+	  IM_CTL, 0x00a,  3, 0x0, 0xffff},
+#if 0
+//"CFI?", /* 02 FIND CFI and If FIND go to S1_DROP */
+//0x1000, 0x1000, OP_EQ,  0, S1_DROP,  1, S1_8023,  CL_REG, 0x000,  0, 0x0, 0x00
+	00,
+#endif
+	{ "CFI?", /* FIND CFI and If FIND go to CleanUP1 (ignore and send to host) */
+	  0x1000, 0x1000, OP_EQ,  0, S1_CLNP,  1, S1_8023,
+	  CL_REG, 0x000,  0, 0x0, 0x0000},
+	{ "8023?", 0xffff, 0x0600, OP_LT,  1, S1_LLC,   0, S1_IPV4,
+	  CL_REG, 0x000,  0, 0x0, 0x0000},
+	{ "LLC?", 0xffff, 0xaaaa, OP_EQ,  1, S1_LLCc,  0, S1_CLNP,
+	  CL_REG, 0x000,  0, 0x0, 0x0000},
+	{ "LLCc?", 0xff00, 0x0300, OP_EQ,  2, S1_IPV4,  0, S1_CLNP,
+	  CL_REG, 0x000,  0, 0x0, 0x0000},
+	{ "IPV4?", 0xffff, 0x0800, OP_EQ,  1, S1_IPV4c, 0, S1_IPV6,
+	  LD_SAP, 0x100,  3, 0x0, 0xffff},
+	{ "IPV4 cont?", 0xff00, 0x4500, OP_EQ,  3, S1_IPV4F, 0, S1_CLNP,
+	  LD_SUM, 0x00a,  1, 0x0, 0x0000},
+	{ "IPV4 frag?", 0x3fff, 0x0000, OP_EQ,  1, S1_TCP44, 0, S1_CLNP,
+	  LD_LEN, 0x03e,  1, 0x0, 0xffff},
+	{ "TCP44?", 0x00ff, 0x0006, OP_EQ,  7, S1_TCPSQ, 0, S1_ESP4,
+	  LD_FID, 0x182,  1, 0x0, 0xffff}, /* FID IP4&TCP src+dst */
+	{ "IPV6?", 0xffff, 0x86dd, OP_EQ,  1, S1_IPV6L, 0, S1_CLNP,
+	  LD_SUM, 0x015,  1, 0x0, 0x0000},
+	{ "IPV6 len", 0xf000, 0x6000, OP_EQ,  0, S1_IPV6c, 0, S1_CLNP,
+	  IM_R1,  0x128,  1, 0x0, 0xffff},
+	{ "IPV6 cont?", 0x0000, 0x0000, OP_EQ,  3, S1_TCP64, 0, S1_CLNP,
+	  LD_FID, 0x484,  1, 0x0, 0xffff}, /*  FID IP6&TCP src+dst */
+	{ "TCP64?",
+#if 0
+//@@@0xff00, 0x0600, OP_EQ, 18, S1_TCPSQ, 0, S1_ESP6,  LD_LEN, 0x03f,  1, 0x0, 0xffff,
+#endif
+	  0xff00, 0x0600, OP_EQ, 12, S1_TCPSQ, 0, S1_ESP6,  LD_LEN,
+	  0x03f,  1, 0x0, 0xffff},
+	{ "TCP seq", /* 14:DADDR should point to dest port */
+	  0xFFFF, 0x0080, OP_EQ,  0, S2_HTTP,  0, S1_TCPFG, LD_SEQ,
+	  0x081,  3, 0x0, 0xffff}, /* Load TCP seq # */
+	{ "TCP control flags", 0xFFFF, 0x8080, OP_EQ,  0, S2_HTTP,  0,
+	  S1_TCPHL, ST_FLG, 0x145,  2, 0x0, 0x002f}, /* Load TCP flags */
+	{ "TCP length", 0x0000, 0x0000, OP_EQ,  0, S1_TCPHc, 0, S1_TCPHc,
+	  LD_R1,  0x205,  3, 0xB, 0xf000} ,
+	{ "TCP length cont", 0x0000, 0x0000, OP_EQ,  0, S1_PCKT,  0,
+	  S1_PCKT,  LD_HDR, 0x0ff,  3, 0x0, 0xffff},
+	{ "Cleanup", 0x0000, 0x0000, OP_EQ,  0, S1_CLNP2,  0, S1_CLNP2,
+	  IM_CTL, 0x001,  3, 0x0, 0x0001},
+	{ "Cleanup 2", 0x0000, 0x0000, OP_EQ,  0, S1_PCKT,  0, S1_PCKT,
+	  CL_REG, 0x002,  3, 0x0, 0x0000},
+	{ "Drop packet", 0x0000, 0x0000, OP_EQ,  0, S1_PCKT,  0, S1_PCKT,
+	  IM_CTL, 0x080,  3, 0x0, 0xffff},
+	{ "No HTTP", 0x0000, 0x0000, OP_EQ,  0, S1_PCKT,  0, S1_PCKT,
+	  IM_CTL, 0x044,  3, 0x0, 0xffff},
+	{ "IPV4 ESP encrypted?",  /* S1_ESP4 */
+	  0x00ff, 0x0032, OP_EQ,  0, S1_CLNP2, 0, S1_AH4, IM_CTL,
+	  0x021, 1,  0x0, 0xffff},
+	{ "IPV4 AH encrypted?",   /* S1_AH4 */
+	  0x00ff, 0x0033, OP_EQ,  0, S1_CLNP2, 0, S1_CLNP, IM_CTL,
+	  0x021, 1,  0x0, 0xffff},
+	{ "IPV6 ESP encrypted?",  /* S1_ESP6 */
+#if 0
+//@@@0x00ff, 0x0032, OP_EQ,  0, S1_CLNP2, 0, S1_AH6, IM_CTL, 0x021, 1,  0x0, 0xffff,
+#endif
+	  0xff00, 0x3200, OP_EQ,  0, S1_CLNP2, 0, S1_AH6, IM_CTL,
+	  0x021, 1,  0x0, 0xffff},
+	{ "IPV6 AH encrypted?",   /* S1_AH6 */
+#if 0
+//@@@0x00ff, 0x0033, OP_EQ,  0, S1_CLNP2, 0, S1_CLNP, IM_CTL, 0x021, 1,  0x0, 0xffff,
+#endif
+	  0xff00, 0x3300, OP_EQ,  0, S1_CLNP2, 0, S1_CLNP, IM_CTL,
+	  0x021, 1,  0x0, 0xffff},
+	{ NULL },
+};
+#ifdef HP_ENCRYPT_DEFAULT
+#define CAS_HP_FIRMWARE               cas_prog_encryptiontab
+#endif
+#endif
+
+static cas_hp_inst_t cas_prog_null[] = { {NULL} };
+#ifdef HP_NULL_DEFAULT
+#define CAS_HP_FIRMWARE               cas_prog_null
+#endif
+
+/* phy types */
+#define   CAS_PHY_UNKNOWN       0x00
+#define   CAS_PHY_SERDES        0x01
+#define   CAS_PHY_MII_MDIO0     0x02
+#define   CAS_PHY_MII_MDIO1     0x04
+#define   CAS_PHY_MII(x)        ((x) & (CAS_PHY_MII_MDIO0 | CAS_PHY_MII_MDIO1))
+
+/* _RING_INDEX is the index for the ring sizes to be used.  _RING_SIZE
+ * is the actual size. the default index for the various rings is
+ * 8. NOTE: there a bunch of alignment constraints for the rings. to
+ * deal with that, i just allocate rings to create the desired
+ * alignment. here are the constraints:
+ *   RX DESC and COMP rings must be 8KB aligned
+ *   TX DESC must be 2KB aligned.
+ * if you change the numbers, be cognizant of how the alignment will change
+ * in INIT_BLOCK as well.
+ */
+
+#define DESC_RING_I_TO_S(x)  (32*(1 << (x)))
+#define COMP_RING_I_TO_S(x)  (128*(1 << (x)))
+#define TX_DESC_RING_INDEX 4  /* 512 = 8k */
+#define RX_DESC_RING_INDEX 4  /* 512 = 8k */
+#define RX_COMP_RING_INDEX 4  /* 2048 = 64k: should be 4x rx ring size */
+
+#if (TX_DESC_RING_INDEX > 8) || (TX_DESC_RING_INDEX < 0)
+#error TX_DESC_RING_INDEX must be between 0 and 8
+#endif
+
+#if (RX_DESC_RING_INDEX > 8) || (RX_DESC_RING_INDEX < 0)
+#error RX_DESC_RING_INDEX must be between 0 and 8
+#endif
+
+#if (RX_COMP_RING_INDEX > 8) || (RX_COMP_RING_INDEX < 0)
+#error RX_COMP_RING_INDEX must be between 0 and 8
+#endif
+
+#define N_TX_RINGS                    MAX_TX_RINGS      /* for QoS */
+#define N_TX_RINGS_MASK               MAX_TX_RINGS_MASK
+#define N_RX_DESC_RINGS               MAX_RX_DESC_RINGS /* 1 for ipsec */
+#define N_RX_COMP_RINGS               0x1 /* for mult. PCI interrupts */
+
+/* number of flows that can go through re-assembly */
+#define N_RX_FLOWS                    64
+
+#define TX_DESC_RING_SIZE  DESC_RING_I_TO_S(TX_DESC_RING_INDEX)
+#define RX_DESC_RING_SIZE  DESC_RING_I_TO_S(RX_DESC_RING_INDEX)
+#define RX_COMP_RING_SIZE  COMP_RING_I_TO_S(RX_COMP_RING_INDEX)
+#define TX_DESC_RINGN_INDEX(x) TX_DESC_RING_INDEX
+#define RX_DESC_RINGN_INDEX(x) RX_DESC_RING_INDEX
+#define RX_COMP_RINGN_INDEX(x) RX_COMP_RING_INDEX
+#define TX_DESC_RINGN_SIZE(x)  TX_DESC_RING_SIZE
+#define RX_DESC_RINGN_SIZE(x)  RX_DESC_RING_SIZE
+#define RX_COMP_RINGN_SIZE(x)  RX_COMP_RING_SIZE
+
+/* convert values */
+#define CAS_BASE(x, y)                (((y) << (x ## _SHIFT)) & (x ## _MASK))
+#define CAS_VAL(x, y)                 (((y) & (x ## _MASK)) >> (x ## _SHIFT))
+#define CAS_TX_RINGN_BASE(y)          ((TX_DESC_RINGN_INDEX(y) << \
+                                        TX_CFG_DESC_RINGN_SHIFT(y)) & \
+                                        TX_CFG_DESC_RINGN_MASK(y))
+
+/* min is 2k, but we can't do jumbo frames unless it's at least 8k */
+#define CAS_MIN_PAGE_SHIFT            11 /* 2048 */
+#define CAS_JUMBO_PAGE_SHIFT          13 /* 8192 */
+#define CAS_MAX_PAGE_SHIFT            14 /* 16384 */
+
+#define TX_DESC_BUFLEN_MASK         0x0000000000003FFFULL /* buffer length in
+							     bytes. 0 - 9256 */
+#define TX_DESC_BUFLEN_SHIFT        0
+#define TX_DESC_CSUM_START_MASK     0x00000000001F8000ULL /* checksum start. #
+							     of bytes to be
+							     skipped before
+							     csum calc begins.
+							     value must be
+							     even */
+#define TX_DESC_CSUM_START_SHIFT    15
+#define TX_DESC_CSUM_STUFF_MASK     0x000000001FE00000ULL /* checksum stuff.
+							     byte offset w/in
+							     the pkt for the
+							     1st csum byte.
+							     must be > 8 */
+#define TX_DESC_CSUM_STUFF_SHIFT    21
+#define TX_DESC_CSUM_EN             0x0000000020000000ULL /* enable checksum */
+#define TX_DESC_EOF                 0x0000000040000000ULL /* end of frame */
+#define TX_DESC_SOF                 0x0000000080000000ULL /* start of frame */
+#define TX_DESC_INTME               0x0000000100000000ULL /* interrupt me */
+#define TX_DESC_NO_CRC              0x0000000200000000ULL /* debugging only.
+							     CRC will not be
+							     inserted into
+							     outgoing frame. */
+struct cas_tx_desc {
+	__le64     control;
+	__le64     buffer;
+};
+
+/* descriptor ring for free buffers contains page-sized buffers. the index
+ * value is not used by the hw in any way. it's just stored and returned in
+ * the completion ring.
+ */
+struct cas_rx_desc {
+	__le64     index;
+	__le64     buffer;
+};
+
+/* received packets are put on the completion ring. */
+/* word 1 */
+#define RX_COMP1_DATA_SIZE_MASK           0x0000000007FFE000ULL
+#define RX_COMP1_DATA_SIZE_SHIFT          13
+#define RX_COMP1_DATA_OFF_MASK            0x000001FFF8000000ULL
+#define RX_COMP1_DATA_OFF_SHIFT           27
+#define RX_COMP1_DATA_INDEX_MASK          0x007FFE0000000000ULL
+#define RX_COMP1_DATA_INDEX_SHIFT         41
+#define RX_COMP1_SKIP_MASK                0x0180000000000000ULL
+#define RX_COMP1_SKIP_SHIFT               55
+#define RX_COMP1_RELEASE_NEXT             0x0200000000000000ULL
+#define RX_COMP1_SPLIT_PKT                0x0400000000000000ULL
+#define RX_COMP1_RELEASE_FLOW             0x0800000000000000ULL
+#define RX_COMP1_RELEASE_DATA             0x1000000000000000ULL
+#define RX_COMP1_RELEASE_HDR              0x2000000000000000ULL
+#define RX_COMP1_TYPE_MASK                0xC000000000000000ULL
+#define RX_COMP1_TYPE_SHIFT               62
+
+/* word 2 */
+#define RX_COMP2_NEXT_INDEX_MASK          0x00000007FFE00000ULL
+#define RX_COMP2_NEXT_INDEX_SHIFT         21
+#define RX_COMP2_HDR_SIZE_MASK            0x00000FF800000000ULL
+#define RX_COMP2_HDR_SIZE_SHIFT           35
+#define RX_COMP2_HDR_OFF_MASK             0x0003F00000000000ULL
+#define RX_COMP2_HDR_OFF_SHIFT            44
+#define RX_COMP2_HDR_INDEX_MASK           0xFFFC000000000000ULL
+#define RX_COMP2_HDR_INDEX_SHIFT          50
+
+/* word 3 */
+#define RX_COMP3_SMALL_PKT                0x0000000000000001ULL
+#define RX_COMP3_JUMBO_PKT                0x0000000000000002ULL
+#define RX_COMP3_JUMBO_HDR_SPLIT_EN       0x0000000000000004ULL
+#define RX_COMP3_CSUM_START_MASK          0x000000000007F000ULL
+#define RX_COMP3_CSUM_START_SHIFT         12
+#define RX_COMP3_FLOWID_MASK              0x0000000001F80000ULL
+#define RX_COMP3_FLOWID_SHIFT             19
+#define RX_COMP3_OPCODE_MASK              0x000000000E000000ULL
+#define RX_COMP3_OPCODE_SHIFT             25
+#define RX_COMP3_FORCE_FLAG               0x0000000010000000ULL
+#define RX_COMP3_NO_ASSIST                0x0000000020000000ULL
+#define RX_COMP3_LOAD_BAL_MASK            0x000001F800000000ULL
+#define RX_COMP3_LOAD_BAL_SHIFT           35
+#define RX_PLUS_COMP3_ENC_PKT             0x0000020000000000ULL /* cas+ */
+#define RX_COMP3_L3_HEAD_OFF_MASK         0x0000FE0000000000ULL /* cas */
+#define RX_COMP3_L3_HEAD_OFF_SHIFT        41
+#define RX_PLUS_COMP_L3_HEAD_OFF_MASK     0x0000FC0000000000ULL /* cas+ */
+#define RX_PLUS_COMP_L3_HEAD_OFF_SHIFT    42
+#define RX_COMP3_SAP_MASK                 0xFFFF000000000000ULL
+#define RX_COMP3_SAP_SHIFT                48
+
+/* word 4 */
+#define RX_COMP4_TCP_CSUM_MASK            0x000000000000FFFFULL
+#define RX_COMP4_TCP_CSUM_SHIFT           0
+#define RX_COMP4_PKT_LEN_MASK             0x000000003FFF0000ULL
+#define RX_COMP4_PKT_LEN_SHIFT            16
+#define RX_COMP4_PERFECT_MATCH_MASK       0x00000003C0000000ULL
+#define RX_COMP4_PERFECT_MATCH_SHIFT      30
+#define RX_COMP4_ZERO                     0x0000080000000000ULL
+#define RX_COMP4_HASH_VAL_MASK            0x0FFFF00000000000ULL
+#define RX_COMP4_HASH_VAL_SHIFT           44
+#define RX_COMP4_HASH_PASS                0x1000000000000000ULL
+#define RX_COMP4_BAD                      0x4000000000000000ULL
+#define RX_COMP4_LEN_MISMATCH             0x8000000000000000ULL
+
+/* we encode the following: ring/index/release. only 14 bits
+ * are usable.
+ * NOTE: the encoding is dependent upon RX_DESC_RING_SIZE and
+ *       MAX_RX_DESC_RINGS. */
+#define RX_INDEX_NUM_MASK                 0x0000000000000FFFULL
+#define RX_INDEX_NUM_SHIFT                0
+#define RX_INDEX_RING_MASK                0x0000000000001000ULL
+#define RX_INDEX_RING_SHIFT               12
+#define RX_INDEX_RELEASE                  0x0000000000002000ULL
+
+struct cas_rx_comp {
+	__le64     word1;
+	__le64     word2;
+	__le64     word3;
+	__le64     word4;
+};
+
+enum link_state {
+	link_down = 0,	/* No link, will retry */
+	link_aneg,	/* Autoneg in progress */
+	link_force_try,	/* Try Forced link speed */
+	link_force_ret,	/* Forced mode worked, retrying autoneg */
+	link_force_ok,	/* Stay in forced mode */
+	link_up		/* Link is up */
+};
+
+typedef struct cas_page {
+	struct list_head list;
+	struct page *buffer;
+	dma_addr_t dma_addr;
+	int used;
+} cas_page_t;
+
+
+/* some alignment constraints:
+ * TX DESC, RX DESC, and RX COMP must each be 8K aligned.
+ * TX COMPWB must be 8-byte aligned.
+ * to accomplish this, here's what we do:
+ *
+ * INIT_BLOCK_RX_COMP  = 64k (already aligned)
+ * INIT_BLOCK_RX_DESC  = 8k
+ * INIT_BLOCK_TX       = 8k
+ * INIT_BLOCK_RX1_DESC = 8k
+ * TX COMPWB
+ */
+#define INIT_BLOCK_TX           (TX_DESC_RING_SIZE)
+#define INIT_BLOCK_RX_DESC      (RX_DESC_RING_SIZE)
+#define INIT_BLOCK_RX_COMP      (RX_COMP_RING_SIZE)
+
+struct cas_init_block {
+	struct cas_rx_comp rxcs[N_RX_COMP_RINGS][INIT_BLOCK_RX_COMP];
+	struct cas_rx_desc rxds[N_RX_DESC_RINGS][INIT_BLOCK_RX_DESC];
+	struct cas_tx_desc txds[N_TX_RINGS][INIT_BLOCK_TX];
+	__le64 tx_compwb;
+};
+
+/* tiny buffers to deal with target abort issue. we allocate a bit
+ * over so that we don't have target abort issues with these buffers
+ * as well.
+ */
+#define TX_TINY_BUF_LEN    0x100
+#define TX_TINY_BUF_BLOCK  ((INIT_BLOCK_TX + 1)*TX_TINY_BUF_LEN)
+
+struct cas_tiny_count {
+	int nbufs;
+	int used;
+};
+
+struct cas {
+	spinlock_t lock; /* for most bits */
+	spinlock_t tx_lock[N_TX_RINGS]; /* tx bits */
+	spinlock_t stat_lock[N_TX_RINGS + 1]; /* for stat gathering */
+	spinlock_t rx_inuse_lock; /* rx inuse list */
+	spinlock_t rx_spare_lock; /* rx spare list */
+
+	void __iomem *regs;
+	int tx_new[N_TX_RINGS], tx_old[N_TX_RINGS];
+	int rx_old[N_RX_DESC_RINGS];
+	int rx_cur[N_RX_COMP_RINGS], rx_new[N_RX_COMP_RINGS];
+	int rx_last[N_RX_DESC_RINGS];
+
+	struct napi_struct napi;
+
+	/* Set when chip is actually in operational state
+	 * (ie. not power managed) */
+	int hw_running;
+	int opened;
+	struct mutex pm_mutex; /* open/close/suspend/resume */
+
+	struct cas_init_block *init_block;
+	struct cas_tx_desc *init_txds[MAX_TX_RINGS];
+	struct cas_rx_desc *init_rxds[MAX_RX_DESC_RINGS];
+	struct cas_rx_comp *init_rxcs[MAX_RX_COMP_RINGS];
+
+	/* we use sk_buffs for tx and pages for rx. the rx skbuffs
+	 * are there for flow re-assembly. */
+	struct sk_buff      *tx_skbs[N_TX_RINGS][TX_DESC_RING_SIZE];
+	struct sk_buff_head  rx_flows[N_RX_FLOWS];
+	cas_page_t          *rx_pages[N_RX_DESC_RINGS][RX_DESC_RING_SIZE];
+	struct list_head     rx_spare_list, rx_inuse_list;
+	int                  rx_spares_needed;
+
+	/* for small packets when copying would be quicker than
+	   mapping */
+	struct cas_tiny_count tx_tiny_use[N_TX_RINGS][TX_DESC_RING_SIZE];
+	u8 *tx_tiny_bufs[N_TX_RINGS];
+
+	u32			msg_enable;
+
+	/* N_TX_RINGS must be >= N_RX_DESC_RINGS */
+	struct net_device_stats net_stats[N_TX_RINGS + 1];
+
+	u32			pci_cfg[64 >> 2];
+	u8                      pci_revision;
+
+	int                     phy_type;
+	int			phy_addr;
+	u32                     phy_id;
+#define CAS_FLAG_1000MB_CAP     0x00000001
+#define CAS_FLAG_REG_PLUS       0x00000002
+#define CAS_FLAG_TARGET_ABORT   0x00000004
+#define CAS_FLAG_SATURN         0x00000008
+#define CAS_FLAG_RXD_POST_MASK  0x000000F0
+#define CAS_FLAG_RXD_POST_SHIFT 4
+#define CAS_FLAG_RXD_POST(x)    ((1 << (CAS_FLAG_RXD_POST_SHIFT + (x))) & \
+                                 CAS_FLAG_RXD_POST_MASK)
+#define CAS_FLAG_ENTROPY_DEV    0x00000100
+#define CAS_FLAG_NO_HW_CSUM     0x00000200
+	u32                     cas_flags;
+	int                     packet_min; /* minimum packet size */
+	int			tx_fifo_size;
+	int			rx_fifo_size;
+	int			rx_pause_off;
+	int			rx_pause_on;
+	int                     crc_size;      /* 4 if half-duplex */
+
+	int                     pci_irq_INTC;
+	int                     min_frame_size; /* for tx fifo workaround */
+
+	/* page size allocation */
+	int                     page_size;
+	int                     page_order;
+	int                     mtu_stride;
+
+	u32			mac_rx_cfg;
+
+	/* Autoneg & PHY control */
+	int			link_cntl;
+	int			link_fcntl;
+	enum link_state		lstate;
+	struct timer_list	link_timer;
+	int			timer_ticks;
+	struct work_struct	reset_task;
+#if 0
+	atomic_t		reset_task_pending;
+#else
+	atomic_t		reset_task_pending;
+	atomic_t		reset_task_pending_mtu;
+	atomic_t		reset_task_pending_spare;
+	atomic_t		reset_task_pending_all;
+#endif
+
+	/* Link-down problem workaround */
+#define LINK_TRANSITION_UNKNOWN 	0
+#define LINK_TRANSITION_ON_FAILURE 	1
+#define LINK_TRANSITION_STILL_FAILED 	2
+#define LINK_TRANSITION_LINK_UP 	3
+#define LINK_TRANSITION_LINK_CONFIG	4
+#define LINK_TRANSITION_LINK_DOWN	5
+#define LINK_TRANSITION_REQUESTED_RESET	6
+	int			link_transition;
+	int 			link_transition_jiffies_valid;
+	unsigned long		link_transition_jiffies;
+
+	/* Tuning */
+	u8 orig_cacheline_size;	/* value when loaded */
+#define CAS_PREF_CACHELINE_SIZE	 0x20	/* Minimum desired */
+
+	/* Diagnostic counters and state. */
+	int 			casreg_len; /* reg-space size for dumping */
+	u64			pause_entered;
+	u16			pause_last_time_recvd;
+
+	dma_addr_t block_dvma, tx_tiny_dvma[N_TX_RINGS];
+	struct pci_dev *pdev;
+	struct net_device *dev;
+#if defined(CONFIG_OF)
+	struct device_node	*of_node;
+#endif
+
+	/* Firmware Info */
+	u16			fw_load_addr;
+	u32			fw_size;
+	u8			*fw_data;
+};
+
+#define TX_DESC_NEXT(r, x)  (((x) + 1) & (TX_DESC_RINGN_SIZE(r) - 1))
+#define RX_DESC_ENTRY(r, x) ((x) & (RX_DESC_RINGN_SIZE(r) - 1))
+#define RX_COMP_ENTRY(r, x) ((x) & (RX_COMP_RINGN_SIZE(r) - 1))
+
+#define TX_BUFF_COUNT(r, x, y)    ((x) <= (y) ? ((y) - (x)) : \
+        (TX_DESC_RINGN_SIZE(r) - (x) + (y)))
+
+#define TX_BUFFS_AVAIL(cp, i)	((cp)->tx_old[(i)] <= (cp)->tx_new[(i)] ? \
+        (cp)->tx_old[(i)] + (TX_DESC_RINGN_SIZE(i) - 1) - (cp)->tx_new[(i)] : \
+        (cp)->tx_old[(i)] - (cp)->tx_new[(i)] - 1)
+
+#define CAS_ALIGN(addr, align) \
+     (((unsigned long) (addr) + ((align) - 1UL)) & ~((align) - 1))
+
+#define RX_FIFO_SIZE                  16384
+#define EXPANSION_ROM_SIZE            65536
+
+#define CAS_MC_EXACT_MATCH_SIZE       15
+#define CAS_MC_HASH_SIZE              256
+#define CAS_MC_HASH_MAX              (CAS_MC_EXACT_MATCH_SIZE + \
+                                      CAS_MC_HASH_SIZE)
+
+#define TX_TARGET_ABORT_LEN           0x20
+#define RX_SWIVEL_OFF_VAL             0x2
+#define RX_AE_FREEN_VAL(x)            (RX_DESC_RINGN_SIZE(x) >> 1)
+#define RX_AE_COMP_VAL                (RX_COMP_RING_SIZE >> 1)
+#define RX_BLANK_INTR_PKT_VAL         0x05
+#define RX_BLANK_INTR_TIME_VAL        0x0F
+#define HP_TCP_THRESH_VAL             1530 /* reduce to enable reassembly */
+
+#define RX_SPARE_COUNT                (RX_DESC_RING_SIZE >> 1)
+#define RX_SPARE_RECOVER_VAL          (RX_SPARE_COUNT >> 2)
+
+#endif /* _CASSINI_H */
diff --git a/drivers/net/ethernet/sun/ldmvsw.c b/drivers/net/ethernet/sun/ldmvsw.c
new file mode 100644
index 000000000..05689e62c
--- /dev/null
+++ b/drivers/net/ethernet/sun/ldmvsw.c
@@ -0,0 +1,478 @@
+// SPDX-License-Identifier: GPL-2.0
+/* ldmvsw.c: Sun4v LDOM Virtual Switch Driver.
+ *
+ * Copyright (C) 2016-2017 Oracle. All rights reserved.
+ */
+
+#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
+
+#include <linux/delay.h>
+#include <linux/etherdevice.h>
+#include <linux/ethtool.h>
+#include <linux/highmem.h>
+#include <linux/if_vlan.h>
+#include <linux/init.h>
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/mutex.h>
+#include <linux/netdevice.h>
+#include <linux/slab.h>
+#include <linux/types.h>
+
+#if defined(CONFIG_IPV6)
+#include <linux/icmpv6.h>
+#endif
+
+#include <net/ip.h>
+#include <net/icmp.h>
+#include <net/route.h>
+
+#include <asm/vio.h>
+#include <asm/ldc.h>
+
+/* This driver makes use of the common code in sunvnet_common.c */
+#include "sunvnet_common.h"
+
+/* Length of time before we decide the hardware is hung,
+ * and dev->tx_timeout() should be called to fix the problem.
+ */
+#define VSW_TX_TIMEOUT			(10 * HZ)
+
+/* Static HW Addr used for the network interfaces representing vsw ports */
+static u8 vsw_port_hwaddr[ETH_ALEN] = {0xFE, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF};
+
+#define DRV_MODULE_NAME		"ldmvsw"
+#define DRV_MODULE_VERSION	"1.2"
+#define DRV_MODULE_RELDATE	"March 4, 2017"
+
+static char version[] =
+	DRV_MODULE_NAME " " DRV_MODULE_VERSION " (" DRV_MODULE_RELDATE ")";
+MODULE_AUTHOR("Oracle");
+MODULE_DESCRIPTION("Sun4v LDOM Virtual Switch Driver");
+MODULE_LICENSE("GPL");
+MODULE_VERSION(DRV_MODULE_VERSION);
+
+/* Ordered from largest major to lowest */
+static struct vio_version vsw_versions[] = {
+	{ .major = 1, .minor = 8 },
+	{ .major = 1, .minor = 7 },
+	{ .major = 1, .minor = 6 },
+	{ .major = 1, .minor = 0 },
+};
+
+static void vsw_get_drvinfo(struct net_device *dev,
+			    struct ethtool_drvinfo *info)
+{
+	strscpy(info->driver, DRV_MODULE_NAME, sizeof(info->driver));
+	strscpy(info->version, DRV_MODULE_VERSION, sizeof(info->version));
+}
+
+static u32 vsw_get_msglevel(struct net_device *dev)
+{
+	struct vnet_port *port = netdev_priv(dev);
+
+	return port->vp->msg_enable;
+}
+
+static void vsw_set_msglevel(struct net_device *dev, u32 value)
+{
+	struct vnet_port *port = netdev_priv(dev);
+
+	port->vp->msg_enable = value;
+}
+
+static const struct ethtool_ops vsw_ethtool_ops = {
+	.get_drvinfo		= vsw_get_drvinfo,
+	.get_msglevel		= vsw_get_msglevel,
+	.set_msglevel		= vsw_set_msglevel,
+	.get_link		= ethtool_op_get_link,
+};
+
+static LIST_HEAD(vnet_list);
+static DEFINE_MUTEX(vnet_list_mutex);
+
+/* func arg to vnet_start_xmit_common() to get the proper tx port */
+static struct vnet_port *vsw_tx_port_find(struct sk_buff *skb,
+					  struct net_device *dev)
+{
+	struct vnet_port *port = netdev_priv(dev);
+
+	return port;
+}
+
+static u16 vsw_select_queue(struct net_device *dev, struct sk_buff *skb,
+			    struct net_device *sb_dev)
+{
+	struct vnet_port *port = netdev_priv(dev);
+
+	if (!port)
+		return 0;
+
+	return port->q_index;
+}
+
+/* Wrappers to common functions */
+static netdev_tx_t vsw_start_xmit(struct sk_buff *skb, struct net_device *dev)
+{
+	return sunvnet_start_xmit_common(skb, dev, vsw_tx_port_find);
+}
+
+static void vsw_set_rx_mode(struct net_device *dev)
+{
+	struct vnet_port *port = netdev_priv(dev);
+
+	return sunvnet_set_rx_mode_common(dev, port->vp);
+}
+
+static int ldmvsw_open(struct net_device *dev)
+{
+	struct vnet_port *port = netdev_priv(dev);
+	struct vio_driver_state *vio = &port->vio;
+
+	/* reset the channel */
+	vio_link_state_change(vio, LDC_EVENT_RESET);
+	vnet_port_reset(port);
+	vio_port_up(vio);
+
+	return 0;
+}
+
+#ifdef CONFIG_NET_POLL_CONTROLLER
+static void vsw_poll_controller(struct net_device *dev)
+{
+	struct vnet_port *port = netdev_priv(dev);
+
+	return sunvnet_poll_controller_common(dev, port->vp);
+}
+#endif
+
+static const struct net_device_ops vsw_ops = {
+	.ndo_open		= ldmvsw_open,
+	.ndo_stop		= sunvnet_close_common,
+	.ndo_set_rx_mode	= vsw_set_rx_mode,
+	.ndo_set_mac_address	= sunvnet_set_mac_addr_common,
+	.ndo_validate_addr	= eth_validate_addr,
+	.ndo_tx_timeout		= sunvnet_tx_timeout_common,
+	.ndo_start_xmit		= vsw_start_xmit,
+	.ndo_select_queue	= vsw_select_queue,
+#ifdef CONFIG_NET_POLL_CONTROLLER
+	.ndo_poll_controller    = vsw_poll_controller,
+#endif
+};
+
+static const char *local_mac_prop = "local-mac-address";
+static const char *cfg_handle_prop = "cfg-handle";
+
+static struct vnet *vsw_get_vnet(struct mdesc_handle *hp,
+				 u64 port_node,
+				 u64 *handle)
+{
+	struct vnet *vp;
+	struct vnet *iter;
+	const u64 *local_mac = NULL;
+	const u64 *cfghandle = NULL;
+	u64 a;
+
+	/* Get the parent virtual-network-switch macaddr and cfghandle */
+	mdesc_for_each_arc(a, hp, port_node, MDESC_ARC_TYPE_BACK) {
+		u64 target = mdesc_arc_target(hp, a);
+		const char *name;
+
+		name = mdesc_get_property(hp, target, "name", NULL);
+		if (!name || strcmp(name, "virtual-network-switch"))
+			continue;
+
+		local_mac = mdesc_get_property(hp, target,
+					       local_mac_prop, NULL);
+		cfghandle = mdesc_get_property(hp, target,
+					       cfg_handle_prop, NULL);
+		break;
+	}
+	if (!local_mac || !cfghandle)
+		return ERR_PTR(-ENODEV);
+
+	/* find or create associated vnet */
+	vp = NULL;
+	mutex_lock(&vnet_list_mutex);
+	list_for_each_entry(iter, &vnet_list, list) {
+		if (iter->local_mac == *local_mac) {
+			vp = iter;
+			break;
+		}
+	}
+
+	if (!vp) {
+		vp = kzalloc_obj(*vp);
+		if (unlikely(!vp)) {
+			mutex_unlock(&vnet_list_mutex);
+			return ERR_PTR(-ENOMEM);
+		}
+
+		spin_lock_init(&vp->lock);
+		INIT_LIST_HEAD(&vp->port_list);
+		INIT_LIST_HEAD(&vp->list);
+		vp->local_mac = *local_mac;
+		list_add(&vp->list, &vnet_list);
+	}
+
+	mutex_unlock(&vnet_list_mutex);
+
+	*handle = (u64)*cfghandle;
+
+	return vp;
+}
+
+static struct net_device *vsw_alloc_netdev(u8 hwaddr[],
+					   struct vio_dev *vdev,
+					   u64 handle,
+					   u64 port_id)
+{
+	struct net_device *dev;
+	struct vnet_port *port;
+
+	dev = alloc_etherdev_mqs(sizeof(*port), VNET_MAX_TXQS, 1);
+	if (!dev)
+		return ERR_PTR(-ENOMEM);
+	dev->needed_headroom = VNET_PACKET_SKIP + 8;
+	dev->needed_tailroom = 8;
+
+	eth_hw_addr_set(dev, hwaddr);
+	ether_addr_copy(dev->perm_addr, dev->dev_addr);
+
+	sprintf(dev->name, "vif%d.%d", (int)handle, (int)port_id);
+
+	dev->netdev_ops = &vsw_ops;
+	dev->ethtool_ops = &vsw_ethtool_ops;
+	dev->watchdog_timeo = VSW_TX_TIMEOUT;
+
+	dev->hw_features = NETIF_F_HW_CSUM | NETIF_F_SG;
+	dev->features = dev->hw_features;
+
+	/* MTU range: 68 - 65535 */
+	dev->min_mtu = ETH_MIN_MTU;
+	dev->max_mtu = VNET_MAX_MTU;
+
+	SET_NETDEV_DEV(dev, &vdev->dev);
+
+	return dev;
+}
+
+static struct ldc_channel_config vsw_ldc_cfg = {
+	.event		= sunvnet_event_common,
+	.mtu		= 64,
+	.mode		= LDC_MODE_UNRELIABLE,
+};
+
+static struct vio_driver_ops vsw_vio_ops = {
+	.send_attr		= sunvnet_send_attr_common,
+	.handle_attr		= sunvnet_handle_attr_common,
+	.handshake_complete	= sunvnet_handshake_complete_common,
+};
+
+static const char *remote_macaddr_prop = "remote-mac-address";
+static const char *id_prop = "id";
+
+static int vsw_port_probe(struct vio_dev *vdev, const struct vio_device_id *id)
+{
+	struct mdesc_handle *hp;
+	struct vnet_port *port;
+	unsigned long flags;
+	struct vnet *vp;
+	struct net_device *dev;
+	const u64 *rmac;
+	int len, i, err;
+	const u64 *port_id;
+	u64 handle;
+
+	hp = mdesc_grab();
+
+	if (!hp)
+		return -ENODEV;
+
+	rmac = mdesc_get_property(hp, vdev->mp, remote_macaddr_prop, &len);
+	err = -ENODEV;
+	if (!rmac) {
+		pr_err("Port lacks %s property\n", remote_macaddr_prop);
+		mdesc_release(hp);
+		return err;
+	}
+
+	port_id = mdesc_get_property(hp, vdev->mp, id_prop, NULL);
+	err = -ENODEV;
+	if (!port_id) {
+		pr_err("Port lacks %s property\n", id_prop);
+		mdesc_release(hp);
+		return err;
+	}
+
+	/* Get (or create) the vnet associated with this port */
+	vp = vsw_get_vnet(hp, vdev->mp, &handle);
+	if (IS_ERR(vp)) {
+		err = PTR_ERR(vp);
+		pr_err("Failed to get vnet for vsw-port\n");
+		mdesc_release(hp);
+		return err;
+	}
+
+	mdesc_release(hp);
+
+	dev = vsw_alloc_netdev(vsw_port_hwaddr, vdev, handle, *port_id);
+	if (IS_ERR(dev)) {
+		err = PTR_ERR(dev);
+		pr_err("Failed to alloc netdev for vsw-port\n");
+		return err;
+	}
+
+	port = netdev_priv(dev);
+
+	INIT_LIST_HEAD(&port->list);
+
+	for (i = 0; i < ETH_ALEN; i++)
+		port->raddr[i] = (*rmac >> (5 - i) * 8) & 0xff;
+
+	port->vp = vp;
+	port->dev = dev;
+	port->switch_port = 1;
+	port->tso = false; /* no tso in vsw, misbehaves in bridge */
+	port->tsolen = 0;
+
+	/* Mark the port as belonging to ldmvsw which directs the
+	 * common code to use the net_device in the vnet_port
+	 * rather than the net_device in the vnet (which is used
+	 * by sunvnet). This bit is used by the VNET_PORT_TO_NET_DEVICE
+	 * macro.
+	 */
+	port->vsw = 1;
+
+	err = vio_driver_init(&port->vio, vdev, VDEV_NETWORK,
+			      vsw_versions, ARRAY_SIZE(vsw_versions),
+			      &vsw_vio_ops, dev->name);
+	if (err)
+		goto err_out_free_dev;
+
+	err = vio_ldc_alloc(&port->vio, &vsw_ldc_cfg, port);
+	if (err)
+		goto err_out_free_dev;
+
+	dev_set_drvdata(&vdev->dev, port);
+
+	netif_napi_add(dev, &port->napi, sunvnet_poll_common);
+
+	spin_lock_irqsave(&vp->lock, flags);
+	list_add_rcu(&port->list, &vp->port_list);
+	spin_unlock_irqrestore(&vp->lock, flags);
+
+	timer_setup(&port->clean_timer, sunvnet_clean_timer_expire_common, 0);
+
+	err = register_netdev(dev);
+	if (err) {
+		pr_err("Cannot register net device, aborting\n");
+		goto err_out_del_timer;
+	}
+
+	spin_lock_irqsave(&vp->lock, flags);
+	sunvnet_port_add_txq_common(port);
+	spin_unlock_irqrestore(&vp->lock, flags);
+
+	napi_enable(&port->napi);
+	vio_port_up(&port->vio);
+
+	/* assure no carrier until we receive an LDC_EVENT_UP,
+	 * even if the vsw config script tries to force us up
+	 */
+	netif_carrier_off(dev);
+
+	netdev_info(dev, "LDOM vsw-port %pM\n", dev->dev_addr);
+
+	pr_info("%s: PORT ( remote-mac %pM%s )\n", dev->name,
+		port->raddr, " switch-port");
+
+	return 0;
+
+err_out_del_timer:
+	timer_delete_sync(&port->clean_timer);
+	list_del_rcu(&port->list);
+	synchronize_rcu();
+	netif_napi_del(&port->napi);
+	dev_set_drvdata(&vdev->dev, NULL);
+	vio_ldc_free(&port->vio);
+
+err_out_free_dev:
+	free_netdev(dev);
+	return err;
+}
+
+static void vsw_port_remove(struct vio_dev *vdev)
+{
+	struct vnet_port *port = dev_get_drvdata(&vdev->dev);
+	unsigned long flags;
+
+	if (port) {
+		timer_delete_sync(&port->vio.timer);
+		timer_delete_sync(&port->clean_timer);
+
+		napi_disable(&port->napi);
+		unregister_netdev(port->dev);
+
+		list_del_rcu(&port->list);
+
+		synchronize_rcu();
+		spin_lock_irqsave(&port->vp->lock, flags);
+		sunvnet_port_rm_txq_common(port);
+		spin_unlock_irqrestore(&port->vp->lock, flags);
+		netif_napi_del(&port->napi);
+		sunvnet_port_free_tx_bufs_common(port);
+		vio_ldc_free(&port->vio);
+
+		dev_set_drvdata(&vdev->dev, NULL);
+
+		free_netdev(port->dev);
+	}
+}
+
+static void vsw_cleanup(void)
+{
+	struct vnet *vp;
+
+	/* just need to free up the vnet list */
+	mutex_lock(&vnet_list_mutex);
+	while (!list_empty(&vnet_list)) {
+		vp = list_first_entry(&vnet_list, struct vnet, list);
+		list_del(&vp->list);
+		/* vio_unregister_driver() should have cleaned up port_list */
+		if (!list_empty(&vp->port_list))
+			pr_err("Ports not removed by VIO subsystem!\n");
+		kfree(vp);
+	}
+	mutex_unlock(&vnet_list_mutex);
+}
+
+static const struct vio_device_id vsw_port_match[] = {
+	{
+		.type = "vsw-port",
+	},
+	{},
+};
+MODULE_DEVICE_TABLE(vio, vsw_port_match);
+
+static struct vio_driver vsw_port_driver = {
+	.id_table	= vsw_port_match,
+	.probe		= vsw_port_probe,
+	.remove		= vsw_port_remove,
+	.name		= "vsw_port",
+};
+
+static int __init vsw_init(void)
+{
+	pr_info("%s\n", version);
+	return vio_register_driver(&vsw_port_driver);
+}
+
+static void __exit vsw_exit(void)
+{
+	vio_unregister_driver(&vsw_port_driver);
+	vsw_cleanup();
+}
+
+module_init(vsw_init);
+module_exit(vsw_exit);
diff --git a/drivers/net/ethernet/sun/niu.c b/drivers/net/ethernet/sun/niu.c
new file mode 100644
index 000000000..54dd72811
--- /dev/null
+++ b/drivers/net/ethernet/sun/niu.c
@@ -0,0 +1,10267 @@
+// SPDX-License-Identifier: GPL-2.0
+/* niu.c: Neptune ethernet driver.
+ *
+ * Copyright (C) 2007, 2008 David S. Miller (davem@davemloft.net)
+ */
+
+#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
+
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/interrupt.h>
+#include <linux/pci.h>
+#include <linux/dma-mapping.h>
+#include <linux/netdevice.h>
+#include <linux/ethtool.h>
+#include <linux/etherdevice.h>
+#include <linux/platform_device.h>
+#include <linux/delay.h>
+#include <linux/bitops.h>
+#include <linux/mii.h>
+#include <linux/if.h>
+#include <linux/if_ether.h>
+#include <linux/if_vlan.h>
+#include <linux/ip.h>
+#include <linux/in.h>
+#include <linux/ipv6.h>
+#include <linux/log2.h>
+#include <linux/jiffies.h>
+#include <linux/crc32.h>
+#include <linux/list.h>
+#include <linux/slab.h>
+
+#include <linux/io.h>
+#include <linux/of.h>
+
+#include "niu.h"
+
+/* This driver wants to store a link to a "next page" within the
+ * page struct itself by overloading the content of the "mapping"
+ * member. This is not expected by the page API, but does currently
+ * work. However, the randstruct plugin gets very bothered by this
+ * case because "mapping" (struct address_space) is randomized, so
+ * casts to/from it trigger warnings. Hide this by way of a union,
+ * to create a typed alias of "mapping", since that's how it is
+ * actually being used here.
+ */
+union niu_page {
+	struct page page;
+	struct {
+		unsigned long __flags;	/* unused alias of "flags" */
+		struct list_head __lru;	/* unused alias of "lru" */
+		struct page *next;	/* alias of "mapping" */
+	};
+};
+#define niu_next_page(p)	container_of(p, union niu_page, page)->next
+
+#define DRV_MODULE_NAME		"niu"
+#define DRV_MODULE_VERSION	"1.1"
+#define DRV_MODULE_RELDATE	"Apr 22, 2010"
+
+static char version[] =
+	DRV_MODULE_NAME ".c:v" DRV_MODULE_VERSION " (" DRV_MODULE_RELDATE ")\n";
+
+MODULE_AUTHOR("David S. Miller <davem@davemloft.net>");
+MODULE_DESCRIPTION("NIU ethernet driver");
+MODULE_LICENSE("GPL");
+MODULE_VERSION(DRV_MODULE_VERSION);
+
+#ifndef readq
+static u64 readq(void __iomem *reg)
+{
+	return ((u64) readl(reg)) | (((u64) readl(reg + 4UL)) << 32);
+}
+
+static void writeq(u64 val, void __iomem *reg)
+{
+	writel(val & 0xffffffff, reg);
+	writel(val >> 32, reg + 0x4UL);
+}
+#endif
+
+static const struct pci_device_id niu_pci_tbl[] = {
+	{PCI_DEVICE(PCI_VENDOR_ID_SUN, 0xabcd)},
+	{}
+};
+
+MODULE_DEVICE_TABLE(pci, niu_pci_tbl);
+
+#define NIU_TX_TIMEOUT			(5 * HZ)
+
+#define nr64(reg)		readq(np->regs + (reg))
+#define nw64(reg, val)		writeq((val), np->regs + (reg))
+
+#define nr64_mac(reg)		readq(np->mac_regs + (reg))
+#define nw64_mac(reg, val)	writeq((val), np->mac_regs + (reg))
+
+#define nr64_ipp(reg)		readq(np->regs + np->ipp_off + (reg))
+#define nw64_ipp(reg, val)	writeq((val), np->regs + np->ipp_off + (reg))
+
+#define nr64_pcs(reg)		readq(np->regs + np->pcs_off + (reg))
+#define nw64_pcs(reg, val)	writeq((val), np->regs + np->pcs_off + (reg))
+
+#define nr64_xpcs(reg)		readq(np->regs + np->xpcs_off + (reg))
+#define nw64_xpcs(reg, val)	writeq((val), np->regs + np->xpcs_off + (reg))
+
+#define NIU_MSG_DEFAULT (NETIF_MSG_DRV | NETIF_MSG_PROBE | NETIF_MSG_LINK)
+
+static int niu_debug;
+static int debug = -1;
+module_param(debug, int, 0);
+MODULE_PARM_DESC(debug, "NIU debug level");
+
+#define niu_lock_parent(np, flags) \
+	spin_lock_irqsave(&np->parent->lock, flags)
+#define niu_unlock_parent(np, flags) \
+	spin_unlock_irqrestore(&np->parent->lock, flags)
+
+static int serdes_init_10g_serdes(struct niu *np);
+
+static int __niu_wait_bits_clear_mac(struct niu *np, unsigned long reg,
+				     u64 bits, int limit, int delay)
+{
+	while (--limit >= 0) {
+		u64 val = nr64_mac(reg);
+
+		if (!(val & bits))
+			break;
+		udelay(delay);
+	}
+	if (limit < 0)
+		return -ENODEV;
+	return 0;
+}
+
+static int __niu_set_and_wait_clear_mac(struct niu *np, unsigned long reg,
+					u64 bits, int limit, int delay,
+					const char *reg_name)
+{
+	int err;
+
+	nw64_mac(reg, bits);
+	err = __niu_wait_bits_clear_mac(np, reg, bits, limit, delay);
+	if (err)
+		netdev_err(np->dev, "bits (%llx) of register %s would not clear, val[%llx]\n",
+			   (unsigned long long)bits, reg_name,
+			   (unsigned long long)nr64_mac(reg));
+	return err;
+}
+
+#define niu_set_and_wait_clear_mac(NP, REG, BITS, LIMIT, DELAY, REG_NAME) \
+({	BUILD_BUG_ON(LIMIT <= 0 || DELAY < 0); \
+	__niu_set_and_wait_clear_mac(NP, REG, BITS, LIMIT, DELAY, REG_NAME); \
+})
+
+static int __niu_wait_bits_clear_ipp(struct niu *np, unsigned long reg,
+				     u64 bits, int limit, int delay)
+{
+	while (--limit >= 0) {
+		u64 val = nr64_ipp(reg);
+
+		if (!(val & bits))
+			break;
+		udelay(delay);
+	}
+	if (limit < 0)
+		return -ENODEV;
+	return 0;
+}
+
+static int __niu_set_and_wait_clear_ipp(struct niu *np, unsigned long reg,
+					u64 bits, int limit, int delay,
+					const char *reg_name)
+{
+	int err;
+	u64 val;
+
+	val = nr64_ipp(reg);
+	val |= bits;
+	nw64_ipp(reg, val);
+
+	err = __niu_wait_bits_clear_ipp(np, reg, bits, limit, delay);
+	if (err)
+		netdev_err(np->dev, "bits (%llx) of register %s would not clear, val[%llx]\n",
+			   (unsigned long long)bits, reg_name,
+			   (unsigned long long)nr64_ipp(reg));
+	return err;
+}
+
+#define niu_set_and_wait_clear_ipp(NP, REG, BITS, LIMIT, DELAY, REG_NAME) \
+({	BUILD_BUG_ON(LIMIT <= 0 || DELAY < 0); \
+	__niu_set_and_wait_clear_ipp(NP, REG, BITS, LIMIT, DELAY, REG_NAME); \
+})
+
+static int __niu_wait_bits_clear(struct niu *np, unsigned long reg,
+				 u64 bits, int limit, int delay)
+{
+	while (--limit >= 0) {
+		u64 val = nr64(reg);
+
+		if (!(val & bits))
+			break;
+		udelay(delay);
+	}
+	if (limit < 0)
+		return -ENODEV;
+	return 0;
+}
+
+#define niu_wait_bits_clear(NP, REG, BITS, LIMIT, DELAY) \
+({	BUILD_BUG_ON(LIMIT <= 0 || DELAY < 0); \
+	__niu_wait_bits_clear(NP, REG, BITS, LIMIT, DELAY); \
+})
+
+static int __niu_set_and_wait_clear(struct niu *np, unsigned long reg,
+				    u64 bits, int limit, int delay,
+				    const char *reg_name)
+{
+	int err;
+
+	nw64(reg, bits);
+	err = __niu_wait_bits_clear(np, reg, bits, limit, delay);
+	if (err)
+		netdev_err(np->dev, "bits (%llx) of register %s would not clear, val[%llx]\n",
+			   (unsigned long long)bits, reg_name,
+			   (unsigned long long)nr64(reg));
+	return err;
+}
+
+#define niu_set_and_wait_clear(NP, REG, BITS, LIMIT, DELAY, REG_NAME) \
+({	BUILD_BUG_ON(LIMIT <= 0 || DELAY < 0); \
+	__niu_set_and_wait_clear(NP, REG, BITS, LIMIT, DELAY, REG_NAME); \
+})
+
+static void niu_ldg_rearm(struct niu *np, struct niu_ldg *lp, int on)
+{
+	u64 val = (u64) lp->timer;
+
+	if (on)
+		val |= LDG_IMGMT_ARM;
+
+	nw64(LDG_IMGMT(lp->ldg_num), val);
+}
+
+static int niu_ldn_irq_enable(struct niu *np, int ldn, int on)
+{
+	unsigned long mask_reg, bits;
+	u64 val;
+
+	if (ldn < 0 || ldn > LDN_MAX)
+		return -EINVAL;
+
+	if (ldn < 64) {
+		mask_reg = LD_IM0(ldn);
+		bits = LD_IM0_MASK;
+	} else {
+		mask_reg = LD_IM1(ldn - 64);
+		bits = LD_IM1_MASK;
+	}
+
+	val = nr64(mask_reg);
+	if (on)
+		val &= ~bits;
+	else
+		val |= bits;
+	nw64(mask_reg, val);
+
+	return 0;
+}
+
+static int niu_enable_ldn_in_ldg(struct niu *np, struct niu_ldg *lp, int on)
+{
+	struct niu_parent *parent = np->parent;
+	int i;
+
+	for (i = 0; i <= LDN_MAX; i++) {
+		int err;
+
+		if (parent->ldg_map[i] != lp->ldg_num)
+			continue;
+
+		err = niu_ldn_irq_enable(np, i, on);
+		if (err)
+			return err;
+	}
+	return 0;
+}
+
+static int niu_enable_interrupts(struct niu *np, int on)
+{
+	int i;
+
+	for (i = 0; i < np->num_ldg; i++) {
+		struct niu_ldg *lp = &np->ldg[i];
+		int err;
+
+		err = niu_enable_ldn_in_ldg(np, lp, on);
+		if (err)
+			return err;
+	}
+	for (i = 0; i < np->num_ldg; i++)
+		niu_ldg_rearm(np, &np->ldg[i], on);
+
+	return 0;
+}
+
+static u32 phy_encode(u32 type, int port)
+{
+	return type << (port * 2);
+}
+
+static u32 phy_decode(u32 val, int port)
+{
+	return (val >> (port * 2)) & PORT_TYPE_MASK;
+}
+
+static int mdio_wait(struct niu *np)
+{
+	int limit = 1000;
+	u64 val;
+
+	while (--limit > 0) {
+		val = nr64(MIF_FRAME_OUTPUT);
+		if ((val >> MIF_FRAME_OUTPUT_TA_SHIFT) & 0x1)
+			return val & MIF_FRAME_OUTPUT_DATA;
+
+		udelay(10);
+	}
+
+	return -ENODEV;
+}
+
+static int mdio_read(struct niu *np, int port, int dev, int reg)
+{
+	int err;
+
+	nw64(MIF_FRAME_OUTPUT, MDIO_ADDR_OP(port, dev, reg));
+	err = mdio_wait(np);
+	if (err < 0)
+		return err;
+
+	nw64(MIF_FRAME_OUTPUT, MDIO_READ_OP(port, dev));
+	return mdio_wait(np);
+}
+
+static int mdio_write(struct niu *np, int port, int dev, int reg, int data)
+{
+	int err;
+
+	nw64(MIF_FRAME_OUTPUT, MDIO_ADDR_OP(port, dev, reg));
+	err = mdio_wait(np);
+	if (err < 0)
+		return err;
+
+	nw64(MIF_FRAME_OUTPUT, MDIO_WRITE_OP(port, dev, data));
+	err = mdio_wait(np);
+	if (err < 0)
+		return err;
+
+	return 0;
+}
+
+static int mii_read(struct niu *np, int port, int reg)
+{
+	nw64(MIF_FRAME_OUTPUT, MII_READ_OP(port, reg));
+	return mdio_wait(np);
+}
+
+static int mii_write(struct niu *np, int port, int reg, int data)
+{
+	int err;
+
+	nw64(MIF_FRAME_OUTPUT, MII_WRITE_OP(port, reg, data));
+	err = mdio_wait(np);
+	if (err < 0)
+		return err;
+
+	return 0;
+}
+
+static int esr2_set_tx_cfg(struct niu *np, unsigned long channel, u32 val)
+{
+	int err;
+
+	err = mdio_write(np, np->port, NIU_ESR2_DEV_ADDR,
+			 ESR2_TI_PLL_TX_CFG_L(channel),
+			 val & 0xffff);
+	if (!err)
+		err = mdio_write(np, np->port, NIU_ESR2_DEV_ADDR,
+				 ESR2_TI_PLL_TX_CFG_H(channel),
+				 val >> 16);
+	return err;
+}
+
+static int esr2_set_rx_cfg(struct niu *np, unsigned long channel, u32 val)
+{
+	int err;
+
+	err = mdio_write(np, np->port, NIU_ESR2_DEV_ADDR,
+			 ESR2_TI_PLL_RX_CFG_L(channel),
+			 val & 0xffff);
+	if (!err)
+		err = mdio_write(np, np->port, NIU_ESR2_DEV_ADDR,
+				 ESR2_TI_PLL_RX_CFG_H(channel),
+				 val >> 16);
+	return err;
+}
+
+/* Mode is always 10G fiber.  */
+static int serdes_init_niu_10g_fiber(struct niu *np)
+{
+	struct niu_link_config *lp = &np->link_config;
+	u32 tx_cfg, rx_cfg;
+	unsigned long i;
+
+	tx_cfg = (PLL_TX_CFG_ENTX | PLL_TX_CFG_SWING_1375MV);
+	rx_cfg = (PLL_RX_CFG_ENRX | PLL_RX_CFG_TERM_0P8VDDT |
+		  PLL_RX_CFG_ALIGN_ENA | PLL_RX_CFG_LOS_LTHRESH |
+		  PLL_RX_CFG_EQ_LP_ADAPTIVE);
+
+	if (lp->loopback_mode == LOOPBACK_PHY) {
+		u16 test_cfg = PLL_TEST_CFG_LOOPBACK_CML_DIS;
+
+		mdio_write(np, np->port, NIU_ESR2_DEV_ADDR,
+			   ESR2_TI_PLL_TEST_CFG_L, test_cfg);
+
+		tx_cfg |= PLL_TX_CFG_ENTEST;
+		rx_cfg |= PLL_RX_CFG_ENTEST;
+	}
+
+	/* Initialize all 4 lanes of the SERDES.  */
+	for (i = 0; i < 4; i++) {
+		int err = esr2_set_tx_cfg(np, i, tx_cfg);
+		if (err)
+			return err;
+	}
+
+	for (i = 0; i < 4; i++) {
+		int err = esr2_set_rx_cfg(np, i, rx_cfg);
+		if (err)
+			return err;
+	}
+
+	return 0;
+}
+
+static int serdes_init_niu_1g_serdes(struct niu *np)
+{
+	struct niu_link_config *lp = &np->link_config;
+	u16 pll_cfg, pll_sts;
+	int max_retry = 100;
+	u64 sig, mask, val;
+	u32 tx_cfg, rx_cfg;
+	unsigned long i;
+	int err;
+
+	tx_cfg = (PLL_TX_CFG_ENTX | PLL_TX_CFG_SWING_1375MV |
+		  PLL_TX_CFG_RATE_HALF);
+	rx_cfg = (PLL_RX_CFG_ENRX | PLL_RX_CFG_TERM_0P8VDDT |
+		  PLL_RX_CFG_ALIGN_ENA | PLL_RX_CFG_LOS_LTHRESH |
+		  PLL_RX_CFG_RATE_HALF);
+
+	if (np->port == 0)
+		rx_cfg |= PLL_RX_CFG_EQ_LP_ADAPTIVE;
+
+	if (lp->loopback_mode == LOOPBACK_PHY) {
+		u16 test_cfg = PLL_TEST_CFG_LOOPBACK_CML_DIS;
+
+		mdio_write(np, np->port, NIU_ESR2_DEV_ADDR,
+			   ESR2_TI_PLL_TEST_CFG_L, test_cfg);
+
+		tx_cfg |= PLL_TX_CFG_ENTEST;
+		rx_cfg |= PLL_RX_CFG_ENTEST;
+	}
+
+	/* Initialize PLL for 1G */
+	pll_cfg = (PLL_CFG_ENPLL | PLL_CFG_MPY_8X);
+
+	err = mdio_write(np, np->port, NIU_ESR2_DEV_ADDR,
+			 ESR2_TI_PLL_CFG_L, pll_cfg);
+	if (err) {
+		netdev_err(np->dev, "NIU Port %d %s() mdio write to ESR2_TI_PLL_CFG_L failed\n",
+			   np->port, __func__);
+		return err;
+	}
+
+	pll_sts = PLL_CFG_ENPLL;
+
+	err = mdio_write(np, np->port, NIU_ESR2_DEV_ADDR,
+			 ESR2_TI_PLL_STS_L, pll_sts);
+	if (err) {
+		netdev_err(np->dev, "NIU Port %d %s() mdio write to ESR2_TI_PLL_STS_L failed\n",
+			   np->port, __func__);
+		return err;
+	}
+
+	udelay(200);
+
+	/* Initialize all 4 lanes of the SERDES.  */
+	for (i = 0; i < 4; i++) {
+		err = esr2_set_tx_cfg(np, i, tx_cfg);
+		if (err)
+			return err;
+	}
+
+	for (i = 0; i < 4; i++) {
+		err = esr2_set_rx_cfg(np, i, rx_cfg);
+		if (err)
+			return err;
+	}
+
+	switch (np->port) {
+	case 0:
+		val = (ESR_INT_SRDY0_P0 | ESR_INT_DET0_P0);
+		mask = val;
+		break;
+
+	case 1:
+		val = (ESR_INT_SRDY0_P1 | ESR_INT_DET0_P1);
+		mask = val;
+		break;
+
+	default:
+		return -EINVAL;
+	}
+
+	while (max_retry--) {
+		sig = nr64(ESR_INT_SIGNALS);
+		if ((sig & mask) == val)
+			break;
+
+		mdelay(500);
+	}
+
+	if ((sig & mask) != val) {
+		netdev_err(np->dev, "Port %u signal bits [%08x] are not [%08x]\n",
+			   np->port, (int)(sig & mask), (int)val);
+		return -ENODEV;
+	}
+
+	return 0;
+}
+
+static int serdes_init_niu_10g_serdes(struct niu *np)
+{
+	struct niu_link_config *lp = &np->link_config;
+	u32 tx_cfg, rx_cfg, pll_cfg, pll_sts;
+	int max_retry = 100;
+	u64 sig, mask, val;
+	unsigned long i;
+	int err;
+
+	tx_cfg = (PLL_TX_CFG_ENTX | PLL_TX_CFG_SWING_1375MV);
+	rx_cfg = (PLL_RX_CFG_ENRX | PLL_RX_CFG_TERM_0P8VDDT |
+		  PLL_RX_CFG_ALIGN_ENA | PLL_RX_CFG_LOS_LTHRESH |
+		  PLL_RX_CFG_EQ_LP_ADAPTIVE);
+
+	if (lp->loopback_mode == LOOPBACK_PHY) {
+		u16 test_cfg = PLL_TEST_CFG_LOOPBACK_CML_DIS;
+
+		mdio_write(np, np->port, NIU_ESR2_DEV_ADDR,
+			   ESR2_TI_PLL_TEST_CFG_L, test_cfg);
+
+		tx_cfg |= PLL_TX_CFG_ENTEST;
+		rx_cfg |= PLL_RX_CFG_ENTEST;
+	}
+
+	/* Initialize PLL for 10G */
+	pll_cfg = (PLL_CFG_ENPLL | PLL_CFG_MPY_10X);
+
+	err = mdio_write(np, np->port, NIU_ESR2_DEV_ADDR,
+			 ESR2_TI_PLL_CFG_L, pll_cfg & 0xffff);
+	if (err) {
+		netdev_err(np->dev, "NIU Port %d %s() mdio write to ESR2_TI_PLL_CFG_L failed\n",
+			   np->port, __func__);
+		return err;
+	}
+
+	pll_sts = PLL_CFG_ENPLL;
+
+	err = mdio_write(np, np->port, NIU_ESR2_DEV_ADDR,
+			 ESR2_TI_PLL_STS_L, pll_sts & 0xffff);
+	if (err) {
+		netdev_err(np->dev, "NIU Port %d %s() mdio write to ESR2_TI_PLL_STS_L failed\n",
+			   np->port, __func__);
+		return err;
+	}
+
+	udelay(200);
+
+	/* Initialize all 4 lanes of the SERDES.  */
+	for (i = 0; i < 4; i++) {
+		err = esr2_set_tx_cfg(np, i, tx_cfg);
+		if (err)
+			return err;
+	}
+
+	for (i = 0; i < 4; i++) {
+		err = esr2_set_rx_cfg(np, i, rx_cfg);
+		if (err)
+			return err;
+	}
+
+	/* check if serdes is ready */
+
+	switch (np->port) {
+	case 0:
+		mask = ESR_INT_SIGNALS_P0_BITS;
+		val = (ESR_INT_SRDY0_P0 |
+		       ESR_INT_DET0_P0 |
+		       ESR_INT_XSRDY_P0 |
+		       ESR_INT_XDP_P0_CH3 |
+		       ESR_INT_XDP_P0_CH2 |
+		       ESR_INT_XDP_P0_CH1 |
+		       ESR_INT_XDP_P0_CH0);
+		break;
+
+	case 1:
+		mask = ESR_INT_SIGNALS_P1_BITS;
+		val = (ESR_INT_SRDY0_P1 |
+		       ESR_INT_DET0_P1 |
+		       ESR_INT_XSRDY_P1 |
+		       ESR_INT_XDP_P1_CH3 |
+		       ESR_INT_XDP_P1_CH2 |
+		       ESR_INT_XDP_P1_CH1 |
+		       ESR_INT_XDP_P1_CH0);
+		break;
+
+	default:
+		return -EINVAL;
+	}
+
+	while (max_retry--) {
+		sig = nr64(ESR_INT_SIGNALS);
+		if ((sig & mask) == val)
+			break;
+
+		mdelay(500);
+	}
+
+	if ((sig & mask) != val) {
+		pr_info("NIU Port %u signal bits [%08x] are not [%08x] for 10G...trying 1G\n",
+			np->port, (int)(sig & mask), (int)val);
+
+		/* 10G failed, try initializing at 1G */
+		err = serdes_init_niu_1g_serdes(np);
+		if (!err) {
+			np->flags &= ~NIU_FLAGS_10G;
+			np->mac_xcvr = MAC_XCVR_PCS;
+		}  else {
+			netdev_err(np->dev, "Port %u 10G/1G SERDES Link Failed\n",
+				   np->port);
+			return -ENODEV;
+		}
+	}
+	return 0;
+}
+
+static int esr_read_rxtx_ctrl(struct niu *np, unsigned long chan, u32 *val)
+{
+	int err;
+
+	err = mdio_read(np, np->port, NIU_ESR_DEV_ADDR, ESR_RXTX_CTRL_L(chan));
+	if (err >= 0) {
+		*val = (err & 0xffff);
+		err = mdio_read(np, np->port, NIU_ESR_DEV_ADDR,
+				ESR_RXTX_CTRL_H(chan));
+		if (err >= 0)
+			*val |= ((err & 0xffff) << 16);
+		err = 0;
+	}
+	return err;
+}
+
+static int esr_read_glue0(struct niu *np, unsigned long chan, u32 *val)
+{
+	int err;
+
+	err = mdio_read(np, np->port, NIU_ESR_DEV_ADDR,
+			ESR_GLUE_CTRL0_L(chan));
+	if (err >= 0) {
+		*val = (err & 0xffff);
+		err = mdio_read(np, np->port, NIU_ESR_DEV_ADDR,
+				ESR_GLUE_CTRL0_H(chan));
+		if (err >= 0) {
+			*val |= ((err & 0xffff) << 16);
+			err = 0;
+		}
+	}
+	return err;
+}
+
+static int esr_read_reset(struct niu *np, u32 *val)
+{
+	int err;
+
+	err = mdio_read(np, np->port, NIU_ESR_DEV_ADDR,
+			ESR_RXTX_RESET_CTRL_L);
+	if (err >= 0) {
+		*val = (err & 0xffff);
+		err = mdio_read(np, np->port, NIU_ESR_DEV_ADDR,
+				ESR_RXTX_RESET_CTRL_H);
+		if (err >= 0) {
+			*val |= ((err & 0xffff) << 16);
+			err = 0;
+		}
+	}
+	return err;
+}
+
+static int esr_write_rxtx_ctrl(struct niu *np, unsigned long chan, u32 val)
+{
+	int err;
+
+	err = mdio_write(np, np->port, NIU_ESR_DEV_ADDR,
+			 ESR_RXTX_CTRL_L(chan), val & 0xffff);
+	if (!err)
+		err = mdio_write(np, np->port, NIU_ESR_DEV_ADDR,
+				 ESR_RXTX_CTRL_H(chan), (val >> 16));
+	return err;
+}
+
+static int esr_write_glue0(struct niu *np, unsigned long chan, u32 val)
+{
+	int err;
+
+	err = mdio_write(np, np->port, NIU_ESR_DEV_ADDR,
+			ESR_GLUE_CTRL0_L(chan), val & 0xffff);
+	if (!err)
+		err = mdio_write(np, np->port, NIU_ESR_DEV_ADDR,
+				 ESR_GLUE_CTRL0_H(chan), (val >> 16));
+	return err;
+}
+
+static int esr_reset(struct niu *np)
+{
+	u32 reset;
+	int err;
+
+	err = mdio_write(np, np->port, NIU_ESR_DEV_ADDR,
+			 ESR_RXTX_RESET_CTRL_L, 0x0000);
+	if (err)
+		return err;
+	err = mdio_write(np, np->port, NIU_ESR_DEV_ADDR,
+			 ESR_RXTX_RESET_CTRL_H, 0xffff);
+	if (err)
+		return err;
+	udelay(200);
+
+	err = mdio_write(np, np->port, NIU_ESR_DEV_ADDR,
+			 ESR_RXTX_RESET_CTRL_L, 0xffff);
+	if (err)
+		return err;
+	udelay(200);
+
+	err = mdio_write(np, np->port, NIU_ESR_DEV_ADDR,
+			 ESR_RXTX_RESET_CTRL_H, 0x0000);
+	if (err)
+		return err;
+	udelay(200);
+
+	err = esr_read_reset(np, &reset);
+	if (err)
+		return err;
+	if (reset != 0) {
+		netdev_err(np->dev, "Port %u ESR_RESET did not clear [%08x]\n",
+			   np->port, reset);
+		return -ENODEV;
+	}
+
+	return 0;
+}
+
+static int serdes_init_10g(struct niu *np)
+{
+	struct niu_link_config *lp = &np->link_config;
+	unsigned long ctrl_reg, test_cfg_reg, i;
+	u64 ctrl_val, test_cfg_val, sig, mask, val;
+	int err;
+
+	switch (np->port) {
+	case 0:
+		ctrl_reg = ENET_SERDES_0_CTRL_CFG;
+		test_cfg_reg = ENET_SERDES_0_TEST_CFG;
+		break;
+	case 1:
+		ctrl_reg = ENET_SERDES_1_CTRL_CFG;
+		test_cfg_reg = ENET_SERDES_1_TEST_CFG;
+		break;
+
+	default:
+		return -EINVAL;
+	}
+	ctrl_val = (ENET_SERDES_CTRL_SDET_0 |
+		    ENET_SERDES_CTRL_SDET_1 |
+		    ENET_SERDES_CTRL_SDET_2 |
+		    ENET_SERDES_CTRL_SDET_3 |
+		    (0x5 << ENET_SERDES_CTRL_EMPH_0_SHIFT) |
+		    (0x5 << ENET_SERDES_CTRL_EMPH_1_SHIFT) |
+		    (0x5 << ENET_SERDES_CTRL_EMPH_2_SHIFT) |
+		    (0x5 << ENET_SERDES_CTRL_EMPH_3_SHIFT) |
+		    (0x1 << ENET_SERDES_CTRL_LADJ_0_SHIFT) |
+		    (0x1 << ENET_SERDES_CTRL_LADJ_1_SHIFT) |
+		    (0x1 << ENET_SERDES_CTRL_LADJ_2_SHIFT) |
+		    (0x1 << ENET_SERDES_CTRL_LADJ_3_SHIFT));
+	test_cfg_val = 0;
+
+	if (lp->loopback_mode == LOOPBACK_PHY) {
+		test_cfg_val |= ((ENET_TEST_MD_PAD_LOOPBACK <<
+				  ENET_SERDES_TEST_MD_0_SHIFT) |
+				 (ENET_TEST_MD_PAD_LOOPBACK <<
+				  ENET_SERDES_TEST_MD_1_SHIFT) |
+				 (ENET_TEST_MD_PAD_LOOPBACK <<
+				  ENET_SERDES_TEST_MD_2_SHIFT) |
+				 (ENET_TEST_MD_PAD_LOOPBACK <<
+				  ENET_SERDES_TEST_MD_3_SHIFT));
+	}
+
+	nw64(ctrl_reg, ctrl_val);
+	nw64(test_cfg_reg, test_cfg_val);
+
+	/* Initialize all 4 lanes of the SERDES.  */
+	for (i = 0; i < 4; i++) {
+		u32 rxtx_ctrl, glue0;
+
+		err = esr_read_rxtx_ctrl(np, i, &rxtx_ctrl);
+		if (err)
+			return err;
+		err = esr_read_glue0(np, i, &glue0);
+		if (err)
+			return err;
+
+		rxtx_ctrl &= ~(ESR_RXTX_CTRL_VMUXLO);
+		rxtx_ctrl |= (ESR_RXTX_CTRL_ENSTRETCH |
+			      (2 << ESR_RXTX_CTRL_VMUXLO_SHIFT));
+
+		glue0 &= ~(ESR_GLUE_CTRL0_SRATE |
+			   ESR_GLUE_CTRL0_THCNT |
+			   ESR_GLUE_CTRL0_BLTIME);
+		glue0 |= (ESR_GLUE_CTRL0_RXLOSENAB |
+			  (0xf << ESR_GLUE_CTRL0_SRATE_SHIFT) |
+			  (0xff << ESR_GLUE_CTRL0_THCNT_SHIFT) |
+			  (BLTIME_300_CYCLES <<
+			   ESR_GLUE_CTRL0_BLTIME_SHIFT));
+
+		err = esr_write_rxtx_ctrl(np, i, rxtx_ctrl);
+		if (err)
+			return err;
+		err = esr_write_glue0(np, i, glue0);
+		if (err)
+			return err;
+	}
+
+	err = esr_reset(np);
+	if (err)
+		return err;
+
+	sig = nr64(ESR_INT_SIGNALS);
+	switch (np->port) {
+	case 0:
+		mask = ESR_INT_SIGNALS_P0_BITS;
+		val = (ESR_INT_SRDY0_P0 |
+		       ESR_INT_DET0_P0 |
+		       ESR_INT_XSRDY_P0 |
+		       ESR_INT_XDP_P0_CH3 |
+		       ESR_INT_XDP_P0_CH2 |
+		       ESR_INT_XDP_P0_CH1 |
+		       ESR_INT_XDP_P0_CH0);
+		break;
+
+	case 1:
+		mask = ESR_INT_SIGNALS_P1_BITS;
+		val = (ESR_INT_SRDY0_P1 |
+		       ESR_INT_DET0_P1 |
+		       ESR_INT_XSRDY_P1 |
+		       ESR_INT_XDP_P1_CH3 |
+		       ESR_INT_XDP_P1_CH2 |
+		       ESR_INT_XDP_P1_CH1 |
+		       ESR_INT_XDP_P1_CH0);
+		break;
+
+	default:
+		return -EINVAL;
+	}
+
+	if ((sig & mask) != val) {
+		if (np->flags & NIU_FLAGS_HOTPLUG_PHY) {
+			np->flags &= ~NIU_FLAGS_HOTPLUG_PHY_PRESENT;
+			return 0;
+		}
+		netdev_err(np->dev, "Port %u signal bits [%08x] are not [%08x]\n",
+			   np->port, (int)(sig & mask), (int)val);
+		return -ENODEV;
+	}
+	if (np->flags & NIU_FLAGS_HOTPLUG_PHY)
+		np->flags |= NIU_FLAGS_HOTPLUG_PHY_PRESENT;
+	return 0;
+}
+
+static int serdes_init_1g(struct niu *np)
+{
+	u64 val;
+
+	val = nr64(ENET_SERDES_1_PLL_CFG);
+	val &= ~ENET_SERDES_PLL_FBDIV2;
+	switch (np->port) {
+	case 0:
+		val |= ENET_SERDES_PLL_HRATE0;
+		break;
+	case 1:
+		val |= ENET_SERDES_PLL_HRATE1;
+		break;
+	case 2:
+		val |= ENET_SERDES_PLL_HRATE2;
+		break;
+	case 3:
+		val |= ENET_SERDES_PLL_HRATE3;
+		break;
+	default:
+		return -EINVAL;
+	}
+	nw64(ENET_SERDES_1_PLL_CFG, val);
+
+	return 0;
+}
+
+static int serdes_init_1g_serdes(struct niu *np)
+{
+	struct niu_link_config *lp = &np->link_config;
+	unsigned long ctrl_reg, test_cfg_reg, pll_cfg, i;
+	u64 ctrl_val, test_cfg_val, sig, mask, val;
+	int err;
+	u64 reset_val, val_rd;
+
+	val = ENET_SERDES_PLL_HRATE0 | ENET_SERDES_PLL_HRATE1 |
+		ENET_SERDES_PLL_HRATE2 | ENET_SERDES_PLL_HRATE3 |
+		ENET_SERDES_PLL_FBDIV0;
+	switch (np->port) {
+	case 0:
+		reset_val =  ENET_SERDES_RESET_0;
+		ctrl_reg = ENET_SERDES_0_CTRL_CFG;
+		test_cfg_reg = ENET_SERDES_0_TEST_CFG;
+		pll_cfg = ENET_SERDES_0_PLL_CFG;
+		break;
+	case 1:
+		reset_val =  ENET_SERDES_RESET_1;
+		ctrl_reg = ENET_SERDES_1_CTRL_CFG;
+		test_cfg_reg = ENET_SERDES_1_TEST_CFG;
+		pll_cfg = ENET_SERDES_1_PLL_CFG;
+		break;
+
+	default:
+		return -EINVAL;
+	}
+	ctrl_val = (ENET_SERDES_CTRL_SDET_0 |
+		    ENET_SERDES_CTRL_SDET_1 |
+		    ENET_SERDES_CTRL_SDET_2 |
+		    ENET_SERDES_CTRL_SDET_3 |
+		    (0x5 << ENET_SERDES_CTRL_EMPH_0_SHIFT) |
+		    (0x5 << ENET_SERDES_CTRL_EMPH_1_SHIFT) |
+		    (0x5 << ENET_SERDES_CTRL_EMPH_2_SHIFT) |
+		    (0x5 << ENET_SERDES_CTRL_EMPH_3_SHIFT) |
+		    (0x1 << ENET_SERDES_CTRL_LADJ_0_SHIFT) |
+		    (0x1 << ENET_SERDES_CTRL_LADJ_1_SHIFT) |
+		    (0x1 << ENET_SERDES_CTRL_LADJ_2_SHIFT) |
+		    (0x1 << ENET_SERDES_CTRL_LADJ_3_SHIFT));
+	test_cfg_val = 0;
+
+	if (lp->loopback_mode == LOOPBACK_PHY) {
+		test_cfg_val |= ((ENET_TEST_MD_PAD_LOOPBACK <<
+				  ENET_SERDES_TEST_MD_0_SHIFT) |
+				 (ENET_TEST_MD_PAD_LOOPBACK <<
+				  ENET_SERDES_TEST_MD_1_SHIFT) |
+				 (ENET_TEST_MD_PAD_LOOPBACK <<
+				  ENET_SERDES_TEST_MD_2_SHIFT) |
+				 (ENET_TEST_MD_PAD_LOOPBACK <<
+				  ENET_SERDES_TEST_MD_3_SHIFT));
+	}
+
+	nw64(ENET_SERDES_RESET, reset_val);
+	mdelay(20);
+	val_rd = nr64(ENET_SERDES_RESET);
+	val_rd &= ~reset_val;
+	nw64(pll_cfg, val);
+	nw64(ctrl_reg, ctrl_val);
+	nw64(test_cfg_reg, test_cfg_val);
+	nw64(ENET_SERDES_RESET, val_rd);
+	mdelay(2000);
+
+	/* Initialize all 4 lanes of the SERDES.  */
+	for (i = 0; i < 4; i++) {
+		u32 rxtx_ctrl, glue0;
+
+		err = esr_read_rxtx_ctrl(np, i, &rxtx_ctrl);
+		if (err)
+			return err;
+		err = esr_read_glue0(np, i, &glue0);
+		if (err)
+			return err;
+
+		rxtx_ctrl &= ~(ESR_RXTX_CTRL_VMUXLO);
+		rxtx_ctrl |= (ESR_RXTX_CTRL_ENSTRETCH |
+			      (2 << ESR_RXTX_CTRL_VMUXLO_SHIFT));
+
+		glue0 &= ~(ESR_GLUE_CTRL0_SRATE |
+			   ESR_GLUE_CTRL0_THCNT |
+			   ESR_GLUE_CTRL0_BLTIME);
+		glue0 |= (ESR_GLUE_CTRL0_RXLOSENAB |
+			  (0xf << ESR_GLUE_CTRL0_SRATE_SHIFT) |
+			  (0xff << ESR_GLUE_CTRL0_THCNT_SHIFT) |
+			  (BLTIME_300_CYCLES <<
+			   ESR_GLUE_CTRL0_BLTIME_SHIFT));
+
+		err = esr_write_rxtx_ctrl(np, i, rxtx_ctrl);
+		if (err)
+			return err;
+		err = esr_write_glue0(np, i, glue0);
+		if (err)
+			return err;
+	}
+
+
+	sig = nr64(ESR_INT_SIGNALS);
+	switch (np->port) {
+	case 0:
+		val = (ESR_INT_SRDY0_P0 | ESR_INT_DET0_P0);
+		mask = val;
+		break;
+
+	case 1:
+		val = (ESR_INT_SRDY0_P1 | ESR_INT_DET0_P1);
+		mask = val;
+		break;
+
+	default:
+		return -EINVAL;
+	}
+
+	if ((sig & mask) != val) {
+		netdev_err(np->dev, "Port %u signal bits [%08x] are not [%08x]\n",
+			   np->port, (int)(sig & mask), (int)val);
+		return -ENODEV;
+	}
+
+	return 0;
+}
+
+static int link_status_1g_serdes(struct niu *np, int *link_up_p)
+{
+	struct niu_link_config *lp = &np->link_config;
+	int link_up;
+	u64 val;
+	u16 current_speed;
+	unsigned long flags;
+	u8 current_duplex;
+
+	link_up = 0;
+	current_speed = SPEED_INVALID;
+	current_duplex = DUPLEX_INVALID;
+
+	spin_lock_irqsave(&np->lock, flags);
+
+	val = nr64_pcs(PCS_MII_STAT);
+
+	if (val & PCS_MII_STAT_LINK_STATUS) {
+		link_up = 1;
+		current_speed = SPEED_1000;
+		current_duplex = DUPLEX_FULL;
+	}
+
+	lp->active_speed = current_speed;
+	lp->active_duplex = current_duplex;
+	spin_unlock_irqrestore(&np->lock, flags);
+
+	*link_up_p = link_up;
+	return 0;
+}
+
+static int link_status_10g_serdes(struct niu *np, int *link_up_p)
+{
+	unsigned long flags;
+	struct niu_link_config *lp = &np->link_config;
+	int link_up = 0;
+	int link_ok = 1;
+	u64 val, val2;
+	u16 current_speed;
+	u8 current_duplex;
+
+	if (!(np->flags & NIU_FLAGS_10G))
+		return link_status_1g_serdes(np, link_up_p);
+
+	current_speed = SPEED_INVALID;
+	current_duplex = DUPLEX_INVALID;
+	spin_lock_irqsave(&np->lock, flags);
+
+	val = nr64_xpcs(XPCS_STATUS(0));
+	val2 = nr64_mac(XMAC_INTER2);
+	if (val2 & 0x01000000)
+		link_ok = 0;
+
+	if ((val & 0x1000ULL) && link_ok) {
+		link_up = 1;
+		current_speed = SPEED_10000;
+		current_duplex = DUPLEX_FULL;
+	}
+	lp->active_speed = current_speed;
+	lp->active_duplex = current_duplex;
+	spin_unlock_irqrestore(&np->lock, flags);
+	*link_up_p = link_up;
+	return 0;
+}
+
+static int link_status_mii(struct niu *np, int *link_up_p)
+{
+	struct niu_link_config *lp = &np->link_config;
+	int err;
+	int bmsr, advert, ctrl1000, stat1000, lpa, bmcr, estatus;
+	int supported, advertising, active_speed, active_duplex;
+
+	err = mii_read(np, np->phy_addr, MII_BMCR);
+	if (unlikely(err < 0))
+		return err;
+	bmcr = err;
+
+	err = mii_read(np, np->phy_addr, MII_BMSR);
+	if (unlikely(err < 0))
+		return err;
+	bmsr = err;
+
+	err = mii_read(np, np->phy_addr, MII_ADVERTISE);
+	if (unlikely(err < 0))
+		return err;
+	advert = err;
+
+	err = mii_read(np, np->phy_addr, MII_LPA);
+	if (unlikely(err < 0))
+		return err;
+	lpa = err;
+
+	if (likely(bmsr & BMSR_ESTATEN)) {
+		err = mii_read(np, np->phy_addr, MII_ESTATUS);
+		if (unlikely(err < 0))
+			return err;
+		estatus = err;
+
+		err = mii_read(np, np->phy_addr, MII_CTRL1000);
+		if (unlikely(err < 0))
+			return err;
+		ctrl1000 = err;
+
+		err = mii_read(np, np->phy_addr, MII_STAT1000);
+		if (unlikely(err < 0))
+			return err;
+		stat1000 = err;
+	} else
+		estatus = ctrl1000 = stat1000 = 0;
+
+	supported = 0;
+	if (bmsr & BMSR_ANEGCAPABLE)
+		supported |= SUPPORTED_Autoneg;
+	if (bmsr & BMSR_10HALF)
+		supported |= SUPPORTED_10baseT_Half;
+	if (bmsr & BMSR_10FULL)
+		supported |= SUPPORTED_10baseT_Full;
+	if (bmsr & BMSR_100HALF)
+		supported |= SUPPORTED_100baseT_Half;
+	if (bmsr & BMSR_100FULL)
+		supported |= SUPPORTED_100baseT_Full;
+	if (estatus & ESTATUS_1000_THALF)
+		supported |= SUPPORTED_1000baseT_Half;
+	if (estatus & ESTATUS_1000_TFULL)
+		supported |= SUPPORTED_1000baseT_Full;
+	lp->supported = supported;
+
+	advertising = mii_adv_to_ethtool_adv_t(advert);
+	advertising |= mii_ctrl1000_to_ethtool_adv_t(ctrl1000);
+
+	if (bmcr & BMCR_ANENABLE) {
+		int neg, neg1000;
+
+		lp->active_autoneg = 1;
+		advertising |= ADVERTISED_Autoneg;
+
+		neg = advert & lpa;
+		neg1000 = (ctrl1000 << 2) & stat1000;
+
+		if (neg1000 & (LPA_1000FULL | LPA_1000HALF))
+			active_speed = SPEED_1000;
+		else if (neg & LPA_100)
+			active_speed = SPEED_100;
+		else if (neg & (LPA_10HALF | LPA_10FULL))
+			active_speed = SPEED_10;
+		else
+			active_speed = SPEED_INVALID;
+
+		if ((neg1000 & LPA_1000FULL) || (neg & LPA_DUPLEX))
+			active_duplex = DUPLEX_FULL;
+		else if (active_speed != SPEED_INVALID)
+			active_duplex = DUPLEX_HALF;
+		else
+			active_duplex = DUPLEX_INVALID;
+	} else {
+		lp->active_autoneg = 0;
+
+		if ((bmcr & BMCR_SPEED1000) && !(bmcr & BMCR_SPEED100))
+			active_speed = SPEED_1000;
+		else if (bmcr & BMCR_SPEED100)
+			active_speed = SPEED_100;
+		else
+			active_speed = SPEED_10;
+
+		if (bmcr & BMCR_FULLDPLX)
+			active_duplex = DUPLEX_FULL;
+		else
+			active_duplex = DUPLEX_HALF;
+	}
+
+	lp->active_advertising = advertising;
+	lp->active_speed = active_speed;
+	lp->active_duplex = active_duplex;
+	*link_up_p = !!(bmsr & BMSR_LSTATUS);
+
+	return 0;
+}
+
+static int link_status_1g_rgmii(struct niu *np, int *link_up_p)
+{
+	struct niu_link_config *lp = &np->link_config;
+	u16 current_speed, bmsr;
+	unsigned long flags;
+	u8 current_duplex;
+	int err, link_up;
+
+	link_up = 0;
+	current_speed = SPEED_INVALID;
+	current_duplex = DUPLEX_INVALID;
+
+	spin_lock_irqsave(&np->lock, flags);
+
+	err = mii_read(np, np->phy_addr, MII_BMSR);
+	if (err < 0)
+		goto out;
+
+	bmsr = err;
+	if (bmsr & BMSR_LSTATUS) {
+		link_up = 1;
+		current_speed = SPEED_1000;
+		current_duplex = DUPLEX_FULL;
+	}
+	lp->active_speed = current_speed;
+	lp->active_duplex = current_duplex;
+	err = 0;
+
+out:
+	spin_unlock_irqrestore(&np->lock, flags);
+
+	*link_up_p = link_up;
+	return err;
+}
+
+static int link_status_1g(struct niu *np, int *link_up_p)
+{
+	struct niu_link_config *lp = &np->link_config;
+	unsigned long flags;
+	int err;
+
+	spin_lock_irqsave(&np->lock, flags);
+
+	err = link_status_mii(np, link_up_p);
+	lp->supported |= SUPPORTED_TP;
+	lp->active_advertising |= ADVERTISED_TP;
+
+	spin_unlock_irqrestore(&np->lock, flags);
+	return err;
+}
+
+static int bcm8704_reset(struct niu *np)
+{
+	int err, limit;
+
+	err = mdio_read(np, np->phy_addr,
+			BCM8704_PHYXS_DEV_ADDR, MII_BMCR);
+	if (err < 0 || err == 0xffff)
+		return err;
+	err |= BMCR_RESET;
+	err = mdio_write(np, np->phy_addr, BCM8704_PHYXS_DEV_ADDR,
+			 MII_BMCR, err);
+	if (err)
+		return err;
+
+	limit = 1000;
+	while (--limit >= 0) {
+		err = mdio_read(np, np->phy_addr,
+				BCM8704_PHYXS_DEV_ADDR, MII_BMCR);
+		if (err < 0)
+			return err;
+		if (!(err & BMCR_RESET))
+			break;
+	}
+	if (limit < 0) {
+		netdev_err(np->dev, "Port %u PHY will not reset (bmcr=%04x)\n",
+			   np->port, (err & 0xffff));
+		return -ENODEV;
+	}
+	return 0;
+}
+
+/* When written, certain PHY registers need to be read back twice
+ * in order for the bits to settle properly.
+ */
+static int bcm8704_user_dev3_readback(struct niu *np, int reg)
+{
+	int err = mdio_read(np, np->phy_addr, BCM8704_USER_DEV3_ADDR, reg);
+	if (err < 0)
+		return err;
+	err = mdio_read(np, np->phy_addr, BCM8704_USER_DEV3_ADDR, reg);
+	if (err < 0)
+		return err;
+	return 0;
+}
+
+static int bcm8706_init_user_dev3(struct niu *np)
+{
+	int err;
+
+
+	err = mdio_read(np, np->phy_addr, BCM8704_USER_DEV3_ADDR,
+			BCM8704_USER_OPT_DIGITAL_CTRL);
+	if (err < 0)
+		return err;
+	err &= ~USER_ODIG_CTRL_GPIOS;
+	err |= (0x3 << USER_ODIG_CTRL_GPIOS_SHIFT);
+	err |=  USER_ODIG_CTRL_RESV2;
+	err = mdio_write(np, np->phy_addr, BCM8704_USER_DEV3_ADDR,
+			 BCM8704_USER_OPT_DIGITAL_CTRL, err);
+	if (err)
+		return err;
+
+	mdelay(1000);
+
+	return 0;
+}
+
+static int bcm8704_init_user_dev3(struct niu *np)
+{
+	int err;
+
+	err = mdio_write(np, np->phy_addr,
+			 BCM8704_USER_DEV3_ADDR, BCM8704_USER_CONTROL,
+			 (USER_CONTROL_OPTXRST_LVL |
+			  USER_CONTROL_OPBIASFLT_LVL |
+			  USER_CONTROL_OBTMPFLT_LVL |
+			  USER_CONTROL_OPPRFLT_LVL |
+			  USER_CONTROL_OPTXFLT_LVL |
+			  USER_CONTROL_OPRXLOS_LVL |
+			  USER_CONTROL_OPRXFLT_LVL |
+			  USER_CONTROL_OPTXON_LVL |
+			  (0x3f << USER_CONTROL_RES1_SHIFT)));
+	if (err)
+		return err;
+
+	err = mdio_write(np, np->phy_addr,
+			 BCM8704_USER_DEV3_ADDR, BCM8704_USER_PMD_TX_CONTROL,
+			 (USER_PMD_TX_CTL_XFP_CLKEN |
+			  (1 << USER_PMD_TX_CTL_TX_DAC_TXD_SH) |
+			  (2 << USER_PMD_TX_CTL_TX_DAC_TXCK_SH) |
+			  USER_PMD_TX_CTL_TSCK_LPWREN));
+	if (err)
+		return err;
+
+	err = bcm8704_user_dev3_readback(np, BCM8704_USER_CONTROL);
+	if (err)
+		return err;
+	err = bcm8704_user_dev3_readback(np, BCM8704_USER_PMD_TX_CONTROL);
+	if (err)
+		return err;
+
+	err = mdio_read(np, np->phy_addr, BCM8704_USER_DEV3_ADDR,
+			BCM8704_USER_OPT_DIGITAL_CTRL);
+	if (err < 0)
+		return err;
+	err &= ~USER_ODIG_CTRL_GPIOS;
+	err |= (0x3 << USER_ODIG_CTRL_GPIOS_SHIFT);
+	err = mdio_write(np, np->phy_addr, BCM8704_USER_DEV3_ADDR,
+			 BCM8704_USER_OPT_DIGITAL_CTRL, err);
+	if (err)
+		return err;
+
+	mdelay(1000);
+
+	return 0;
+}
+
+static int mrvl88x2011_act_led(struct niu *np, int val)
+{
+	int	err;
+
+	err  = mdio_read(np, np->phy_addr, MRVL88X2011_USER_DEV2_ADDR,
+		MRVL88X2011_LED_8_TO_11_CTL);
+	if (err < 0)
+		return err;
+
+	err &= ~MRVL88X2011_LED(MRVL88X2011_LED_ACT,MRVL88X2011_LED_CTL_MASK);
+	err |=  MRVL88X2011_LED(MRVL88X2011_LED_ACT,val);
+
+	return mdio_write(np, np->phy_addr, MRVL88X2011_USER_DEV2_ADDR,
+			  MRVL88X2011_LED_8_TO_11_CTL, err);
+}
+
+static int mrvl88x2011_led_blink_rate(struct niu *np, int rate)
+{
+	int	err;
+
+	err = mdio_read(np, np->phy_addr, MRVL88X2011_USER_DEV2_ADDR,
+			MRVL88X2011_LED_BLINK_CTL);
+	if (err >= 0) {
+		err &= ~MRVL88X2011_LED_BLKRATE_MASK;
+		err |= (rate << 4);
+
+		err = mdio_write(np, np->phy_addr, MRVL88X2011_USER_DEV2_ADDR,
+				 MRVL88X2011_LED_BLINK_CTL, err);
+	}
+
+	return err;
+}
+
+static int xcvr_init_10g_mrvl88x2011(struct niu *np)
+{
+	int	err;
+
+	/* Set LED functions */
+	err = mrvl88x2011_led_blink_rate(np, MRVL88X2011_LED_BLKRATE_134MS);
+	if (err)
+		return err;
+
+	/* led activity */
+	err = mrvl88x2011_act_led(np, MRVL88X2011_LED_CTL_OFF);
+	if (err)
+		return err;
+
+	err = mdio_read(np, np->phy_addr, MRVL88X2011_USER_DEV3_ADDR,
+			MRVL88X2011_GENERAL_CTL);
+	if (err < 0)
+		return err;
+
+	err |= MRVL88X2011_ENA_XFPREFCLK;
+
+	err = mdio_write(np, np->phy_addr, MRVL88X2011_USER_DEV3_ADDR,
+			 MRVL88X2011_GENERAL_CTL, err);
+	if (err < 0)
+		return err;
+
+	err = mdio_read(np, np->phy_addr, MRVL88X2011_USER_DEV1_ADDR,
+			MRVL88X2011_PMA_PMD_CTL_1);
+	if (err < 0)
+		return err;
+
+	if (np->link_config.loopback_mode == LOOPBACK_MAC)
+		err |= MRVL88X2011_LOOPBACK;
+	else
+		err &= ~MRVL88X2011_LOOPBACK;
+
+	err = mdio_write(np, np->phy_addr, MRVL88X2011_USER_DEV1_ADDR,
+			 MRVL88X2011_PMA_PMD_CTL_1, err);
+	if (err < 0)
+		return err;
+
+	/* Enable PMD  */
+	return mdio_write(np, np->phy_addr, MRVL88X2011_USER_DEV1_ADDR,
+			  MRVL88X2011_10G_PMD_TX_DIS, MRVL88X2011_ENA_PMDTX);
+}
+
+
+static int xcvr_diag_bcm870x(struct niu *np)
+{
+	u16 analog_stat0, tx_alarm_status;
+	int err = 0;
+
+#if 1
+	err = mdio_read(np, np->phy_addr, BCM8704_PMA_PMD_DEV_ADDR,
+			MII_STAT1000);
+	if (err < 0)
+		return err;
+	pr_info("Port %u PMA_PMD(MII_STAT1000) [%04x]\n", np->port, err);
+
+	err = mdio_read(np, np->phy_addr, BCM8704_USER_DEV3_ADDR, 0x20);
+	if (err < 0)
+		return err;
+	pr_info("Port %u USER_DEV3(0x20) [%04x]\n", np->port, err);
+
+	err = mdio_read(np, np->phy_addr, BCM8704_PHYXS_DEV_ADDR,
+			MII_NWAYTEST);
+	if (err < 0)
+		return err;
+	pr_info("Port %u PHYXS(MII_NWAYTEST) [%04x]\n", np->port, err);
+#endif
+
+	/* XXX dig this out it might not be so useful XXX */
+	err = mdio_read(np, np->phy_addr, BCM8704_USER_DEV3_ADDR,
+			BCM8704_USER_ANALOG_STATUS0);
+	if (err < 0)
+		return err;
+	err = mdio_read(np, np->phy_addr, BCM8704_USER_DEV3_ADDR,
+			BCM8704_USER_ANALOG_STATUS0);
+	if (err < 0)
+		return err;
+	analog_stat0 = err;
+
+	err = mdio_read(np, np->phy_addr, BCM8704_USER_DEV3_ADDR,
+			BCM8704_USER_TX_ALARM_STATUS);
+	if (err < 0)
+		return err;
+	err = mdio_read(np, np->phy_addr, BCM8704_USER_DEV3_ADDR,
+			BCM8704_USER_TX_ALARM_STATUS);
+	if (err < 0)
+		return err;
+	tx_alarm_status = err;
+
+	if (analog_stat0 != 0x03fc) {
+		if ((analog_stat0 == 0x43bc) && (tx_alarm_status != 0)) {
+			pr_info("Port %u cable not connected or bad cable\n",
+				np->port);
+		} else if (analog_stat0 == 0x639c) {
+			pr_info("Port %u optical module is bad or missing\n",
+				np->port);
+		}
+	}
+
+	return 0;
+}
+
+static int xcvr_10g_set_lb_bcm870x(struct niu *np)
+{
+	struct niu_link_config *lp = &np->link_config;
+	int err;
+
+	err = mdio_read(np, np->phy_addr, BCM8704_PCS_DEV_ADDR,
+			MII_BMCR);
+	if (err < 0)
+		return err;
+
+	err &= ~BMCR_LOOPBACK;
+
+	if (lp->loopback_mode == LOOPBACK_MAC)
+		err |= BMCR_LOOPBACK;
+
+	err = mdio_write(np, np->phy_addr, BCM8704_PCS_DEV_ADDR,
+			 MII_BMCR, err);
+	if (err)
+		return err;
+
+	return 0;
+}
+
+static int xcvr_init_10g_bcm8706(struct niu *np)
+{
+	int err = 0;
+	u64 val;
+
+	if ((np->flags & NIU_FLAGS_HOTPLUG_PHY) &&
+	    (np->flags & NIU_FLAGS_HOTPLUG_PHY_PRESENT) == 0)
+			return err;
+
+	val = nr64_mac(XMAC_CONFIG);
+	val &= ~XMAC_CONFIG_LED_POLARITY;
+	val |= XMAC_CONFIG_FORCE_LED_ON;
+	nw64_mac(XMAC_CONFIG, val);
+
+	val = nr64(MIF_CONFIG);
+	val |= MIF_CONFIG_INDIRECT_MODE;
+	nw64(MIF_CONFIG, val);
+
+	err = bcm8704_reset(np);
+	if (err)
+		return err;
+
+	err = xcvr_10g_set_lb_bcm870x(np);
+	if (err)
+		return err;
+
+	err = bcm8706_init_user_dev3(np);
+	if (err)
+		return err;
+
+	err = xcvr_diag_bcm870x(np);
+	if (err)
+		return err;
+
+	return 0;
+}
+
+static int xcvr_init_10g_bcm8704(struct niu *np)
+{
+	int err;
+
+	err = bcm8704_reset(np);
+	if (err)
+		return err;
+
+	err = bcm8704_init_user_dev3(np);
+	if (err)
+		return err;
+
+	err = xcvr_10g_set_lb_bcm870x(np);
+	if (err)
+		return err;
+
+	err =  xcvr_diag_bcm870x(np);
+	if (err)
+		return err;
+
+	return 0;
+}
+
+static int xcvr_init_10g(struct niu *np)
+{
+	int phy_id, err;
+	u64 val;
+
+	val = nr64_mac(XMAC_CONFIG);
+	val &= ~XMAC_CONFIG_LED_POLARITY;
+	val |= XMAC_CONFIG_FORCE_LED_ON;
+	nw64_mac(XMAC_CONFIG, val);
+
+	/* XXX shared resource, lock parent XXX */
+	val = nr64(MIF_CONFIG);
+	val |= MIF_CONFIG_INDIRECT_MODE;
+	nw64(MIF_CONFIG, val);
+
+	phy_id = phy_decode(np->parent->port_phy, np->port);
+	phy_id = np->parent->phy_probe_info.phy_id[phy_id][np->port];
+
+	/* handle different phy types */
+	switch (phy_id & NIU_PHY_ID_MASK) {
+	case NIU_PHY_ID_MRVL88X2011:
+		err = xcvr_init_10g_mrvl88x2011(np);
+		break;
+
+	default: /* bcom 8704 */
+		err = xcvr_init_10g_bcm8704(np);
+		break;
+	}
+
+	return err;
+}
+
+static int mii_reset(struct niu *np)
+{
+	int limit, err;
+
+	err = mii_write(np, np->phy_addr, MII_BMCR, BMCR_RESET);
+	if (err)
+		return err;
+
+	limit = 1000;
+	while (--limit >= 0) {
+		udelay(500);
+		err = mii_read(np, np->phy_addr, MII_BMCR);
+		if (err < 0)
+			return err;
+		if (!(err & BMCR_RESET))
+			break;
+	}
+	if (limit < 0) {
+		netdev_err(np->dev, "Port %u MII would not reset, bmcr[%04x]\n",
+			   np->port, err);
+		return -ENODEV;
+	}
+
+	return 0;
+}
+
+static int xcvr_init_1g_rgmii(struct niu *np)
+{
+	int err;
+	u64 val;
+	u16 bmcr, bmsr, estat;
+
+	val = nr64(MIF_CONFIG);
+	val &= ~MIF_CONFIG_INDIRECT_MODE;
+	nw64(MIF_CONFIG, val);
+
+	err = mii_reset(np);
+	if (err)
+		return err;
+
+	err = mii_read(np, np->phy_addr, MII_BMSR);
+	if (err < 0)
+		return err;
+	bmsr = err;
+
+	estat = 0;
+	if (bmsr & BMSR_ESTATEN) {
+		err = mii_read(np, np->phy_addr, MII_ESTATUS);
+		if (err < 0)
+			return err;
+		estat = err;
+	}
+
+	bmcr = 0;
+	err = mii_write(np, np->phy_addr, MII_BMCR, bmcr);
+	if (err)
+		return err;
+
+	if (bmsr & BMSR_ESTATEN) {
+		u16 ctrl1000 = 0;
+
+		if (estat & ESTATUS_1000_TFULL)
+			ctrl1000 |= ADVERTISE_1000FULL;
+		err = mii_write(np, np->phy_addr, MII_CTRL1000, ctrl1000);
+		if (err)
+			return err;
+	}
+
+	bmcr = (BMCR_SPEED1000 | BMCR_FULLDPLX);
+
+	err = mii_write(np, np->phy_addr, MII_BMCR, bmcr);
+	if (err)
+		return err;
+
+	err = mii_read(np, np->phy_addr, MII_BMCR);
+	if (err < 0)
+		return err;
+	bmcr = mii_read(np, np->phy_addr, MII_BMCR);
+
+	err = mii_read(np, np->phy_addr, MII_BMSR);
+	if (err < 0)
+		return err;
+
+	return 0;
+}
+
+static int mii_init_common(struct niu *np)
+{
+	struct niu_link_config *lp = &np->link_config;
+	u16 bmcr, bmsr, adv, estat;
+	int err;
+
+	err = mii_reset(np);
+	if (err)
+		return err;
+
+	err = mii_read(np, np->phy_addr, MII_BMSR);
+	if (err < 0)
+		return err;
+	bmsr = err;
+
+	estat = 0;
+	if (bmsr & BMSR_ESTATEN) {
+		err = mii_read(np, np->phy_addr, MII_ESTATUS);
+		if (err < 0)
+			return err;
+		estat = err;
+	}
+
+	bmcr = 0;
+	err = mii_write(np, np->phy_addr, MII_BMCR, bmcr);
+	if (err)
+		return err;
+
+	if (lp->loopback_mode == LOOPBACK_MAC) {
+		bmcr |= BMCR_LOOPBACK;
+		if (lp->active_speed == SPEED_1000)
+			bmcr |= BMCR_SPEED1000;
+		if (lp->active_duplex == DUPLEX_FULL)
+			bmcr |= BMCR_FULLDPLX;
+	}
+
+	if (lp->loopback_mode == LOOPBACK_PHY) {
+		u16 aux;
+
+		aux = (BCM5464R_AUX_CTL_EXT_LB |
+		       BCM5464R_AUX_CTL_WRITE_1);
+		err = mii_write(np, np->phy_addr, BCM5464R_AUX_CTL, aux);
+		if (err)
+			return err;
+	}
+
+	if (lp->autoneg) {
+		u16 ctrl1000;
+
+		adv = ADVERTISE_CSMA | ADVERTISE_PAUSE_CAP;
+		if ((bmsr & BMSR_10HALF) &&
+			(lp->advertising & ADVERTISED_10baseT_Half))
+			adv |= ADVERTISE_10HALF;
+		if ((bmsr & BMSR_10FULL) &&
+			(lp->advertising & ADVERTISED_10baseT_Full))
+			adv |= ADVERTISE_10FULL;
+		if ((bmsr & BMSR_100HALF) &&
+			(lp->advertising & ADVERTISED_100baseT_Half))
+			adv |= ADVERTISE_100HALF;
+		if ((bmsr & BMSR_100FULL) &&
+			(lp->advertising & ADVERTISED_100baseT_Full))
+			adv |= ADVERTISE_100FULL;
+		err = mii_write(np, np->phy_addr, MII_ADVERTISE, adv);
+		if (err)
+			return err;
+
+		if (likely(bmsr & BMSR_ESTATEN)) {
+			ctrl1000 = 0;
+			if ((estat & ESTATUS_1000_THALF) &&
+				(lp->advertising & ADVERTISED_1000baseT_Half))
+				ctrl1000 |= ADVERTISE_1000HALF;
+			if ((estat & ESTATUS_1000_TFULL) &&
+				(lp->advertising & ADVERTISED_1000baseT_Full))
+				ctrl1000 |= ADVERTISE_1000FULL;
+			err = mii_write(np, np->phy_addr,
+					MII_CTRL1000, ctrl1000);
+			if (err)
+				return err;
+		}
+
+		bmcr |= (BMCR_ANENABLE | BMCR_ANRESTART);
+	} else {
+		/* !lp->autoneg */
+		int fulldpx;
+
+		if (lp->duplex == DUPLEX_FULL) {
+			bmcr |= BMCR_FULLDPLX;
+			fulldpx = 1;
+		} else if (lp->duplex == DUPLEX_HALF)
+			fulldpx = 0;
+		else
+			return -EINVAL;
+
+		if (lp->speed == SPEED_1000) {
+			/* if X-full requested while not supported, or
+			   X-half requested while not supported... */
+			if ((fulldpx && !(estat & ESTATUS_1000_TFULL)) ||
+				(!fulldpx && !(estat & ESTATUS_1000_THALF)))
+				return -EINVAL;
+			bmcr |= BMCR_SPEED1000;
+		} else if (lp->speed == SPEED_100) {
+			if ((fulldpx && !(bmsr & BMSR_100FULL)) ||
+				(!fulldpx && !(bmsr & BMSR_100HALF)))
+				return -EINVAL;
+			bmcr |= BMCR_SPEED100;
+		} else if (lp->speed == SPEED_10) {
+			if ((fulldpx && !(bmsr & BMSR_10FULL)) ||
+				(!fulldpx && !(bmsr & BMSR_10HALF)))
+				return -EINVAL;
+		} else
+			return -EINVAL;
+	}
+
+	err = mii_write(np, np->phy_addr, MII_BMCR, bmcr);
+	if (err)
+		return err;
+
+#if 0
+	err = mii_read(np, np->phy_addr, MII_BMCR);
+	if (err < 0)
+		return err;
+	bmcr = err;
+
+	err = mii_read(np, np->phy_addr, MII_BMSR);
+	if (err < 0)
+		return err;
+	bmsr = err;
+
+	pr_info("Port %u after MII init bmcr[%04x] bmsr[%04x]\n",
+		np->port, bmcr, bmsr);
+#endif
+
+	return 0;
+}
+
+static int xcvr_init_1g(struct niu *np)
+{
+	u64 val;
+
+	/* XXX shared resource, lock parent XXX */
+	val = nr64(MIF_CONFIG);
+	val &= ~MIF_CONFIG_INDIRECT_MODE;
+	nw64(MIF_CONFIG, val);
+
+	return mii_init_common(np);
+}
+
+static int niu_xcvr_init(struct niu *np)
+{
+	const struct niu_phy_ops *ops = np->phy_ops;
+	int err;
+
+	err = 0;
+	if (ops->xcvr_init)
+		err = ops->xcvr_init(np);
+
+	return err;
+}
+
+static int niu_serdes_init(struct niu *np)
+{
+	const struct niu_phy_ops *ops = np->phy_ops;
+	int err;
+
+	err = 0;
+	if (ops->serdes_init)
+		err = ops->serdes_init(np);
+
+	return err;
+}
+
+static void niu_init_xif(struct niu *);
+static void niu_handle_led(struct niu *, int status);
+
+static int niu_link_status_common(struct niu *np, int link_up)
+{
+	struct niu_link_config *lp = &np->link_config;
+	struct net_device *dev = np->dev;
+	unsigned long flags;
+
+	if (!netif_carrier_ok(dev) && link_up) {
+		netif_info(np, link, dev, "Link is up at %s, %s duplex\n",
+			   lp->active_speed == SPEED_10000 ? "10Gb/sec" :
+			   lp->active_speed == SPEED_1000 ? "1Gb/sec" :
+			   lp->active_speed == SPEED_100 ? "100Mbit/sec" :
+			   "10Mbit/sec",
+			   lp->active_duplex == DUPLEX_FULL ? "full" : "half");
+
+		spin_lock_irqsave(&np->lock, flags);
+		niu_init_xif(np);
+		niu_handle_led(np, 1);
+		spin_unlock_irqrestore(&np->lock, flags);
+
+		netif_carrier_on(dev);
+	} else if (netif_carrier_ok(dev) && !link_up) {
+		netif_warn(np, link, dev, "Link is down\n");
+		spin_lock_irqsave(&np->lock, flags);
+		niu_handle_led(np, 0);
+		spin_unlock_irqrestore(&np->lock, flags);
+		netif_carrier_off(dev);
+	}
+
+	return 0;
+}
+
+static int link_status_10g_mrvl(struct niu *np, int *link_up_p)
+{
+	int err, link_up, pma_status, pcs_status;
+
+	link_up = 0;
+
+	err = mdio_read(np, np->phy_addr, MRVL88X2011_USER_DEV1_ADDR,
+			MRVL88X2011_10G_PMD_STATUS_2);
+	if (err < 0)
+		goto out;
+
+	/* Check PMA/PMD Register: 1.0001.2 == 1 */
+	err = mdio_read(np, np->phy_addr, MRVL88X2011_USER_DEV1_ADDR,
+			MRVL88X2011_PMA_PMD_STATUS_1);
+	if (err < 0)
+		goto out;
+
+	pma_status = ((err & MRVL88X2011_LNK_STATUS_OK) ? 1 : 0);
+
+        /* Check PMC Register : 3.0001.2 == 1: read twice */
+	err = mdio_read(np, np->phy_addr, MRVL88X2011_USER_DEV3_ADDR,
+			MRVL88X2011_PMA_PMD_STATUS_1);
+	if (err < 0)
+		goto out;
+
+	err = mdio_read(np, np->phy_addr, MRVL88X2011_USER_DEV3_ADDR,
+			MRVL88X2011_PMA_PMD_STATUS_1);
+	if (err < 0)
+		goto out;
+
+	pcs_status = ((err & MRVL88X2011_LNK_STATUS_OK) ? 1 : 0);
+
+        /* Check XGXS Register : 4.0018.[0-3,12] */
+	err = mdio_read(np, np->phy_addr, MRVL88X2011_USER_DEV4_ADDR,
+			MRVL88X2011_10G_XGXS_LANE_STAT);
+	if (err < 0)
+		goto out;
+
+	if (err == (PHYXS_XGXS_LANE_STAT_ALINGED | PHYXS_XGXS_LANE_STAT_LANE3 |
+		    PHYXS_XGXS_LANE_STAT_LANE2 | PHYXS_XGXS_LANE_STAT_LANE1 |
+		    PHYXS_XGXS_LANE_STAT_LANE0 | PHYXS_XGXS_LANE_STAT_MAGIC |
+		    0x800))
+		link_up = (pma_status && pcs_status) ? 1 : 0;
+
+	np->link_config.active_speed = SPEED_10000;
+	np->link_config.active_duplex = DUPLEX_FULL;
+	err = 0;
+out:
+	mrvl88x2011_act_led(np, (link_up ?
+				 MRVL88X2011_LED_CTL_PCS_ACT :
+				 MRVL88X2011_LED_CTL_OFF));
+
+	*link_up_p = link_up;
+	return err;
+}
+
+static int link_status_10g_bcm8706(struct niu *np, int *link_up_p)
+{
+	int err, link_up;
+	link_up = 0;
+
+	err = mdio_read(np, np->phy_addr, BCM8704_PMA_PMD_DEV_ADDR,
+			BCM8704_PMD_RCV_SIGDET);
+	if (err < 0 || err == 0xffff)
+		goto out;
+	if (!(err & PMD_RCV_SIGDET_GLOBAL)) {
+		err = 0;
+		goto out;
+	}
+
+	err = mdio_read(np, np->phy_addr, BCM8704_PCS_DEV_ADDR,
+			BCM8704_PCS_10G_R_STATUS);
+	if (err < 0)
+		goto out;
+
+	if (!(err & PCS_10G_R_STATUS_BLK_LOCK)) {
+		err = 0;
+		goto out;
+	}
+
+	err = mdio_read(np, np->phy_addr, BCM8704_PHYXS_DEV_ADDR,
+			BCM8704_PHYXS_XGXS_LANE_STAT);
+	if (err < 0)
+		goto out;
+	if (err != (PHYXS_XGXS_LANE_STAT_ALINGED |
+		    PHYXS_XGXS_LANE_STAT_MAGIC |
+		    PHYXS_XGXS_LANE_STAT_PATTEST |
+		    PHYXS_XGXS_LANE_STAT_LANE3 |
+		    PHYXS_XGXS_LANE_STAT_LANE2 |
+		    PHYXS_XGXS_LANE_STAT_LANE1 |
+		    PHYXS_XGXS_LANE_STAT_LANE0)) {
+		err = 0;
+		np->link_config.active_speed = SPEED_INVALID;
+		np->link_config.active_duplex = DUPLEX_INVALID;
+		goto out;
+	}
+
+	link_up = 1;
+	np->link_config.active_speed = SPEED_10000;
+	np->link_config.active_duplex = DUPLEX_FULL;
+	err = 0;
+
+out:
+	*link_up_p = link_up;
+	return err;
+}
+
+static int link_status_10g_bcom(struct niu *np, int *link_up_p)
+{
+	int err, link_up;
+
+	link_up = 0;
+
+	err = mdio_read(np, np->phy_addr, BCM8704_PMA_PMD_DEV_ADDR,
+			BCM8704_PMD_RCV_SIGDET);
+	if (err < 0)
+		goto out;
+	if (!(err & PMD_RCV_SIGDET_GLOBAL)) {
+		err = 0;
+		goto out;
+	}
+
+	err = mdio_read(np, np->phy_addr, BCM8704_PCS_DEV_ADDR,
+			BCM8704_PCS_10G_R_STATUS);
+	if (err < 0)
+		goto out;
+	if (!(err & PCS_10G_R_STATUS_BLK_LOCK)) {
+		err = 0;
+		goto out;
+	}
+
+	err = mdio_read(np, np->phy_addr, BCM8704_PHYXS_DEV_ADDR,
+			BCM8704_PHYXS_XGXS_LANE_STAT);
+	if (err < 0)
+		goto out;
+
+	if (err != (PHYXS_XGXS_LANE_STAT_ALINGED |
+		    PHYXS_XGXS_LANE_STAT_MAGIC |
+		    PHYXS_XGXS_LANE_STAT_LANE3 |
+		    PHYXS_XGXS_LANE_STAT_LANE2 |
+		    PHYXS_XGXS_LANE_STAT_LANE1 |
+		    PHYXS_XGXS_LANE_STAT_LANE0)) {
+		err = 0;
+		goto out;
+	}
+
+	link_up = 1;
+	np->link_config.active_speed = SPEED_10000;
+	np->link_config.active_duplex = DUPLEX_FULL;
+	err = 0;
+
+out:
+	*link_up_p = link_up;
+	return err;
+}
+
+static int link_status_10g(struct niu *np, int *link_up_p)
+{
+	unsigned long flags;
+	int err = -EINVAL;
+
+	spin_lock_irqsave(&np->lock, flags);
+
+	if (np->link_config.loopback_mode == LOOPBACK_DISABLED) {
+		int phy_id;
+
+		phy_id = phy_decode(np->parent->port_phy, np->port);
+		phy_id = np->parent->phy_probe_info.phy_id[phy_id][np->port];
+
+		/* handle different phy types */
+		switch (phy_id & NIU_PHY_ID_MASK) {
+		case NIU_PHY_ID_MRVL88X2011:
+			err = link_status_10g_mrvl(np, link_up_p);
+			break;
+
+		default: /* bcom 8704 */
+			err = link_status_10g_bcom(np, link_up_p);
+			break;
+		}
+	}
+
+	spin_unlock_irqrestore(&np->lock, flags);
+
+	return err;
+}
+
+static int niu_10g_phy_present(struct niu *np)
+{
+	u64 sig, mask, val;
+
+	sig = nr64(ESR_INT_SIGNALS);
+	switch (np->port) {
+	case 0:
+		mask = ESR_INT_SIGNALS_P0_BITS;
+		val = (ESR_INT_SRDY0_P0 |
+		       ESR_INT_DET0_P0 |
+		       ESR_INT_XSRDY_P0 |
+		       ESR_INT_XDP_P0_CH3 |
+		       ESR_INT_XDP_P0_CH2 |
+		       ESR_INT_XDP_P0_CH1 |
+		       ESR_INT_XDP_P0_CH0);
+		break;
+
+	case 1:
+		mask = ESR_INT_SIGNALS_P1_BITS;
+		val = (ESR_INT_SRDY0_P1 |
+		       ESR_INT_DET0_P1 |
+		       ESR_INT_XSRDY_P1 |
+		       ESR_INT_XDP_P1_CH3 |
+		       ESR_INT_XDP_P1_CH2 |
+		       ESR_INT_XDP_P1_CH1 |
+		       ESR_INT_XDP_P1_CH0);
+		break;
+
+	default:
+		return 0;
+	}
+
+	if ((sig & mask) != val)
+		return 0;
+	return 1;
+}
+
+static int link_status_10g_hotplug(struct niu *np, int *link_up_p)
+{
+	unsigned long flags;
+	int err = 0;
+	int phy_present;
+	int phy_present_prev;
+
+	spin_lock_irqsave(&np->lock, flags);
+
+	if (np->link_config.loopback_mode == LOOPBACK_DISABLED) {
+		phy_present_prev = (np->flags & NIU_FLAGS_HOTPLUG_PHY_PRESENT) ?
+			1 : 0;
+		phy_present = niu_10g_phy_present(np);
+		if (phy_present != phy_present_prev) {
+			/* state change */
+			if (phy_present) {
+				/* A NEM was just plugged in */
+				np->flags |= NIU_FLAGS_HOTPLUG_PHY_PRESENT;
+				if (np->phy_ops->xcvr_init)
+					err = np->phy_ops->xcvr_init(np);
+				if (err) {
+					err = mdio_read(np, np->phy_addr,
+						BCM8704_PHYXS_DEV_ADDR, MII_BMCR);
+					if (err == 0xffff) {
+						/* No mdio, back-to-back XAUI */
+						goto out;
+					}
+					/* debounce */
+					np->flags &= ~NIU_FLAGS_HOTPLUG_PHY_PRESENT;
+				}
+			} else {
+				np->flags &= ~NIU_FLAGS_HOTPLUG_PHY_PRESENT;
+				*link_up_p = 0;
+				netif_warn(np, link, np->dev,
+					   "Hotplug PHY Removed\n");
+			}
+		}
+out:
+		if (np->flags & NIU_FLAGS_HOTPLUG_PHY_PRESENT) {
+			err = link_status_10g_bcm8706(np, link_up_p);
+			if (err == 0xffff) {
+				/* No mdio, back-to-back XAUI: it is C10NEM */
+				*link_up_p = 1;
+				np->link_config.active_speed = SPEED_10000;
+				np->link_config.active_duplex = DUPLEX_FULL;
+			}
+		}
+	}
+
+	spin_unlock_irqrestore(&np->lock, flags);
+
+	return 0;
+}
+
+static int niu_link_status(struct niu *np, int *link_up_p)
+{
+	const struct niu_phy_ops *ops = np->phy_ops;
+	int err;
+
+	err = 0;
+	if (ops->link_status)
+		err = ops->link_status(np, link_up_p);
+
+	return err;
+}
+
+static void niu_timer(struct timer_list *t)
+{
+	struct niu *np = timer_container_of(np, t, timer);
+	unsigned long off;
+	int err, link_up;
+
+	err = niu_link_status(np, &link_up);
+	if (!err)
+		niu_link_status_common(np, link_up);
+
+	if (netif_carrier_ok(np->dev))
+		off = 5 * HZ;
+	else
+		off = 1 * HZ;
+	np->timer.expires = jiffies + off;
+
+	add_timer(&np->timer);
+}
+
+static const struct niu_phy_ops phy_ops_10g_serdes = {
+	.serdes_init		= serdes_init_10g_serdes,
+	.link_status		= link_status_10g_serdes,
+};
+
+static const struct niu_phy_ops phy_ops_10g_serdes_niu = {
+	.serdes_init		= serdes_init_niu_10g_serdes,
+	.link_status		= link_status_10g_serdes,
+};
+
+static const struct niu_phy_ops phy_ops_1g_serdes_niu = {
+	.serdes_init		= serdes_init_niu_1g_serdes,
+	.link_status		= link_status_1g_serdes,
+};
+
+static const struct niu_phy_ops phy_ops_1g_rgmii = {
+	.xcvr_init		= xcvr_init_1g_rgmii,
+	.link_status		= link_status_1g_rgmii,
+};
+
+static const struct niu_phy_ops phy_ops_10g_fiber_niu = {
+	.serdes_init		= serdes_init_niu_10g_fiber,
+	.xcvr_init		= xcvr_init_10g,
+	.link_status		= link_status_10g,
+};
+
+static const struct niu_phy_ops phy_ops_10g_fiber = {
+	.serdes_init		= serdes_init_10g,
+	.xcvr_init		= xcvr_init_10g,
+	.link_status		= link_status_10g,
+};
+
+static const struct niu_phy_ops phy_ops_10g_fiber_hotplug = {
+	.serdes_init		= serdes_init_10g,
+	.xcvr_init		= xcvr_init_10g_bcm8706,
+	.link_status		= link_status_10g_hotplug,
+};
+
+static const struct niu_phy_ops phy_ops_niu_10g_hotplug = {
+	.serdes_init		= serdes_init_niu_10g_fiber,
+	.xcvr_init		= xcvr_init_10g_bcm8706,
+	.link_status		= link_status_10g_hotplug,
+};
+
+static const struct niu_phy_ops phy_ops_10g_copper = {
+	.serdes_init		= serdes_init_10g,
+	.link_status		= link_status_10g, /* XXX */
+};
+
+static const struct niu_phy_ops phy_ops_1g_fiber = {
+	.serdes_init		= serdes_init_1g,
+	.xcvr_init		= xcvr_init_1g,
+	.link_status		= link_status_1g,
+};
+
+static const struct niu_phy_ops phy_ops_1g_copper = {
+	.xcvr_init		= xcvr_init_1g,
+	.link_status		= link_status_1g,
+};
+
+struct niu_phy_template {
+	const struct niu_phy_ops	*ops;
+	u32				phy_addr_base;
+};
+
+static const struct niu_phy_template phy_template_niu_10g_fiber = {
+	.ops		= &phy_ops_10g_fiber_niu,
+	.phy_addr_base	= 16,
+};
+
+static const struct niu_phy_template phy_template_niu_10g_serdes = {
+	.ops		= &phy_ops_10g_serdes_niu,
+	.phy_addr_base	= 0,
+};
+
+static const struct niu_phy_template phy_template_niu_1g_serdes = {
+	.ops		= &phy_ops_1g_serdes_niu,
+	.phy_addr_base	= 0,
+};
+
+static const struct niu_phy_template phy_template_10g_fiber = {
+	.ops		= &phy_ops_10g_fiber,
+	.phy_addr_base	= 8,
+};
+
+static const struct niu_phy_template phy_template_10g_fiber_hotplug = {
+	.ops		= &phy_ops_10g_fiber_hotplug,
+	.phy_addr_base	= 8,
+};
+
+static const struct niu_phy_template phy_template_niu_10g_hotplug = {
+	.ops		= &phy_ops_niu_10g_hotplug,
+	.phy_addr_base	= 8,
+};
+
+static const struct niu_phy_template phy_template_10g_copper = {
+	.ops		= &phy_ops_10g_copper,
+	.phy_addr_base	= 10,
+};
+
+static const struct niu_phy_template phy_template_1g_fiber = {
+	.ops		= &phy_ops_1g_fiber,
+	.phy_addr_base	= 0,
+};
+
+static const struct niu_phy_template phy_template_1g_copper = {
+	.ops		= &phy_ops_1g_copper,
+	.phy_addr_base	= 0,
+};
+
+static const struct niu_phy_template phy_template_1g_rgmii = {
+	.ops		= &phy_ops_1g_rgmii,
+	.phy_addr_base	= 0,
+};
+
+static const struct niu_phy_template phy_template_10g_serdes = {
+	.ops		= &phy_ops_10g_serdes,
+	.phy_addr_base	= 0,
+};
+
+static int niu_atca_port_num[4] = {
+	0, 0,  11, 10
+};
+
+static int serdes_init_10g_serdes(struct niu *np)
+{
+	struct niu_link_config *lp = &np->link_config;
+	unsigned long ctrl_reg, test_cfg_reg, pll_cfg, i;
+	u64 ctrl_val, test_cfg_val, sig, mask, val;
+
+	switch (np->port) {
+	case 0:
+		ctrl_reg = ENET_SERDES_0_CTRL_CFG;
+		test_cfg_reg = ENET_SERDES_0_TEST_CFG;
+		pll_cfg = ENET_SERDES_0_PLL_CFG;
+		break;
+	case 1:
+		ctrl_reg = ENET_SERDES_1_CTRL_CFG;
+		test_cfg_reg = ENET_SERDES_1_TEST_CFG;
+		pll_cfg = ENET_SERDES_1_PLL_CFG;
+		break;
+
+	default:
+		return -EINVAL;
+	}
+	ctrl_val = (ENET_SERDES_CTRL_SDET_0 |
+		    ENET_SERDES_CTRL_SDET_1 |
+		    ENET_SERDES_CTRL_SDET_2 |
+		    ENET_SERDES_CTRL_SDET_3 |
+		    (0x5 << ENET_SERDES_CTRL_EMPH_0_SHIFT) |
+		    (0x5 << ENET_SERDES_CTRL_EMPH_1_SHIFT) |
+		    (0x5 << ENET_SERDES_CTRL_EMPH_2_SHIFT) |
+		    (0x5 << ENET_SERDES_CTRL_EMPH_3_SHIFT) |
+		    (0x1 << ENET_SERDES_CTRL_LADJ_0_SHIFT) |
+		    (0x1 << ENET_SERDES_CTRL_LADJ_1_SHIFT) |
+		    (0x1 << ENET_SERDES_CTRL_LADJ_2_SHIFT) |
+		    (0x1 << ENET_SERDES_CTRL_LADJ_3_SHIFT));
+	test_cfg_val = 0;
+
+	if (lp->loopback_mode == LOOPBACK_PHY) {
+		test_cfg_val |= ((ENET_TEST_MD_PAD_LOOPBACK <<
+				  ENET_SERDES_TEST_MD_0_SHIFT) |
+				 (ENET_TEST_MD_PAD_LOOPBACK <<
+				  ENET_SERDES_TEST_MD_1_SHIFT) |
+				 (ENET_TEST_MD_PAD_LOOPBACK <<
+				  ENET_SERDES_TEST_MD_2_SHIFT) |
+				 (ENET_TEST_MD_PAD_LOOPBACK <<
+				  ENET_SERDES_TEST_MD_3_SHIFT));
+	}
+
+	esr_reset(np);
+	nw64(pll_cfg, ENET_SERDES_PLL_FBDIV2);
+	nw64(ctrl_reg, ctrl_val);
+	nw64(test_cfg_reg, test_cfg_val);
+
+	/* Initialize all 4 lanes of the SERDES.  */
+	for (i = 0; i < 4; i++) {
+		u32 rxtx_ctrl, glue0;
+		int err;
+
+		err = esr_read_rxtx_ctrl(np, i, &rxtx_ctrl);
+		if (err)
+			return err;
+		err = esr_read_glue0(np, i, &glue0);
+		if (err)
+			return err;
+
+		rxtx_ctrl &= ~(ESR_RXTX_CTRL_VMUXLO);
+		rxtx_ctrl |= (ESR_RXTX_CTRL_ENSTRETCH |
+			      (2 << ESR_RXTX_CTRL_VMUXLO_SHIFT));
+
+		glue0 &= ~(ESR_GLUE_CTRL0_SRATE |
+			   ESR_GLUE_CTRL0_THCNT |
+			   ESR_GLUE_CTRL0_BLTIME);
+		glue0 |= (ESR_GLUE_CTRL0_RXLOSENAB |
+			  (0xf << ESR_GLUE_CTRL0_SRATE_SHIFT) |
+			  (0xff << ESR_GLUE_CTRL0_THCNT_SHIFT) |
+			  (BLTIME_300_CYCLES <<
+			   ESR_GLUE_CTRL0_BLTIME_SHIFT));
+
+		err = esr_write_rxtx_ctrl(np, i, rxtx_ctrl);
+		if (err)
+			return err;
+		err = esr_write_glue0(np, i, glue0);
+		if (err)
+			return err;
+	}
+
+
+	sig = nr64(ESR_INT_SIGNALS);
+	switch (np->port) {
+	case 0:
+		mask = ESR_INT_SIGNALS_P0_BITS;
+		val = (ESR_INT_SRDY0_P0 |
+		       ESR_INT_DET0_P0 |
+		       ESR_INT_XSRDY_P0 |
+		       ESR_INT_XDP_P0_CH3 |
+		       ESR_INT_XDP_P0_CH2 |
+		       ESR_INT_XDP_P0_CH1 |
+		       ESR_INT_XDP_P0_CH0);
+		break;
+
+	case 1:
+		mask = ESR_INT_SIGNALS_P1_BITS;
+		val = (ESR_INT_SRDY0_P1 |
+		       ESR_INT_DET0_P1 |
+		       ESR_INT_XSRDY_P1 |
+		       ESR_INT_XDP_P1_CH3 |
+		       ESR_INT_XDP_P1_CH2 |
+		       ESR_INT_XDP_P1_CH1 |
+		       ESR_INT_XDP_P1_CH0);
+		break;
+
+	default:
+		return -EINVAL;
+	}
+
+	if ((sig & mask) != val) {
+		int err;
+		err = serdes_init_1g_serdes(np);
+		if (!err) {
+			np->flags &= ~NIU_FLAGS_10G;
+			np->mac_xcvr = MAC_XCVR_PCS;
+		}  else {
+			netdev_err(np->dev, "Port %u 10G/1G SERDES Link Failed\n",
+				   np->port);
+			return -ENODEV;
+		}
+	}
+
+	return 0;
+}
+
+static int niu_determine_phy_disposition(struct niu *np)
+{
+	struct niu_parent *parent = np->parent;
+	u8 plat_type = parent->plat_type;
+	const struct niu_phy_template *tp;
+	u32 phy_addr_off = 0;
+
+	if (plat_type == PLAT_TYPE_NIU) {
+		switch (np->flags &
+			(NIU_FLAGS_10G |
+			 NIU_FLAGS_FIBER |
+			 NIU_FLAGS_XCVR_SERDES)) {
+		case NIU_FLAGS_10G | NIU_FLAGS_XCVR_SERDES:
+			/* 10G Serdes */
+			tp = &phy_template_niu_10g_serdes;
+			break;
+		case NIU_FLAGS_XCVR_SERDES:
+			/* 1G Serdes */
+			tp = &phy_template_niu_1g_serdes;
+			break;
+		case NIU_FLAGS_10G | NIU_FLAGS_FIBER:
+			/* 10G Fiber */
+		default:
+			if (np->flags & NIU_FLAGS_HOTPLUG_PHY) {
+				tp = &phy_template_niu_10g_hotplug;
+				if (np->port == 0)
+					phy_addr_off = 8;
+				if (np->port == 1)
+					phy_addr_off = 12;
+			} else {
+				tp = &phy_template_niu_10g_fiber;
+				phy_addr_off += np->port;
+			}
+			break;
+		}
+	} else {
+		switch (np->flags &
+			(NIU_FLAGS_10G |
+			 NIU_FLAGS_FIBER |
+			 NIU_FLAGS_XCVR_SERDES)) {
+		case 0:
+			/* 1G copper */
+			tp = &phy_template_1g_copper;
+			if (plat_type == PLAT_TYPE_VF_P0)
+				phy_addr_off = 10;
+			else if (plat_type == PLAT_TYPE_VF_P1)
+				phy_addr_off = 26;
+
+			phy_addr_off += (np->port ^ 0x3);
+			break;
+
+		case NIU_FLAGS_10G:
+			/* 10G copper */
+			tp = &phy_template_10g_copper;
+			break;
+
+		case NIU_FLAGS_FIBER:
+			/* 1G fiber */
+			tp = &phy_template_1g_fiber;
+			break;
+
+		case NIU_FLAGS_10G | NIU_FLAGS_FIBER:
+			/* 10G fiber */
+			tp = &phy_template_10g_fiber;
+			if (plat_type == PLAT_TYPE_VF_P0 ||
+			    plat_type == PLAT_TYPE_VF_P1)
+				phy_addr_off = 8;
+			phy_addr_off += np->port;
+			if (np->flags & NIU_FLAGS_HOTPLUG_PHY) {
+				tp = &phy_template_10g_fiber_hotplug;
+				if (np->port == 0)
+					phy_addr_off = 8;
+				if (np->port == 1)
+					phy_addr_off = 12;
+			}
+			break;
+
+		case NIU_FLAGS_10G | NIU_FLAGS_XCVR_SERDES:
+		case NIU_FLAGS_XCVR_SERDES | NIU_FLAGS_FIBER:
+		case NIU_FLAGS_XCVR_SERDES:
+			switch(np->port) {
+			case 0:
+			case 1:
+				tp = &phy_template_10g_serdes;
+				break;
+			case 2:
+			case 3:
+				tp = &phy_template_1g_rgmii;
+				break;
+			default:
+				return -EINVAL;
+			}
+			phy_addr_off = niu_atca_port_num[np->port];
+			break;
+
+		default:
+			return -EINVAL;
+		}
+	}
+
+	np->phy_ops = tp->ops;
+	np->phy_addr = tp->phy_addr_base + phy_addr_off;
+
+	return 0;
+}
+
+static int niu_init_link(struct niu *np)
+{
+	struct niu_parent *parent = np->parent;
+	int err, ignore;
+
+	if (parent->plat_type == PLAT_TYPE_NIU) {
+		err = niu_xcvr_init(np);
+		if (err)
+			return err;
+		msleep(200);
+	}
+	err = niu_serdes_init(np);
+	if (err && !(np->flags & NIU_FLAGS_HOTPLUG_PHY))
+		return err;
+	msleep(200);
+	err = niu_xcvr_init(np);
+	if (!err || (np->flags & NIU_FLAGS_HOTPLUG_PHY))
+		niu_link_status(np, &ignore);
+	return 0;
+}
+
+static void niu_set_primary_mac(struct niu *np, const unsigned char *addr)
+{
+	u16 reg0 = addr[4] << 8 | addr[5];
+	u16 reg1 = addr[2] << 8 | addr[3];
+	u16 reg2 = addr[0] << 8 | addr[1];
+
+	if (np->flags & NIU_FLAGS_XMAC) {
+		nw64_mac(XMAC_ADDR0, reg0);
+		nw64_mac(XMAC_ADDR1, reg1);
+		nw64_mac(XMAC_ADDR2, reg2);
+	} else {
+		nw64_mac(BMAC_ADDR0, reg0);
+		nw64_mac(BMAC_ADDR1, reg1);
+		nw64_mac(BMAC_ADDR2, reg2);
+	}
+}
+
+static int niu_num_alt_addr(struct niu *np)
+{
+	if (np->flags & NIU_FLAGS_XMAC)
+		return XMAC_NUM_ALT_ADDR;
+	else
+		return BMAC_NUM_ALT_ADDR;
+}
+
+static int niu_set_alt_mac(struct niu *np, int index, unsigned char *addr)
+{
+	u16 reg0 = addr[4] << 8 | addr[5];
+	u16 reg1 = addr[2] << 8 | addr[3];
+	u16 reg2 = addr[0] << 8 | addr[1];
+
+	if (index >= niu_num_alt_addr(np))
+		return -EINVAL;
+
+	if (np->flags & NIU_FLAGS_XMAC) {
+		nw64_mac(XMAC_ALT_ADDR0(index), reg0);
+		nw64_mac(XMAC_ALT_ADDR1(index), reg1);
+		nw64_mac(XMAC_ALT_ADDR2(index), reg2);
+	} else {
+		nw64_mac(BMAC_ALT_ADDR0(index), reg0);
+		nw64_mac(BMAC_ALT_ADDR1(index), reg1);
+		nw64_mac(BMAC_ALT_ADDR2(index), reg2);
+	}
+
+	return 0;
+}
+
+static int niu_enable_alt_mac(struct niu *np, int index, int on)
+{
+	unsigned long reg;
+	u64 val, mask;
+
+	if (index >= niu_num_alt_addr(np))
+		return -EINVAL;
+
+	if (np->flags & NIU_FLAGS_XMAC) {
+		reg = XMAC_ADDR_CMPEN;
+		mask = 1 << index;
+	} else {
+		reg = BMAC_ADDR_CMPEN;
+		mask = 1 << (index + 1);
+	}
+
+	val = nr64_mac(reg);
+	if (on)
+		val |= mask;
+	else
+		val &= ~mask;
+	nw64_mac(reg, val);
+
+	return 0;
+}
+
+static void __set_rdc_table_num_hw(struct niu *np, unsigned long reg,
+				   int num, int mac_pref)
+{
+	u64 val = nr64_mac(reg);
+	val &= ~(HOST_INFO_MACRDCTBLN | HOST_INFO_MPR);
+	val |= num;
+	if (mac_pref)
+		val |= HOST_INFO_MPR;
+	nw64_mac(reg, val);
+}
+
+static int __set_rdc_table_num(struct niu *np,
+			       int xmac_index, int bmac_index,
+			       int rdc_table_num, int mac_pref)
+{
+	unsigned long reg;
+
+	if (rdc_table_num & ~HOST_INFO_MACRDCTBLN)
+		return -EINVAL;
+	if (np->flags & NIU_FLAGS_XMAC)
+		reg = XMAC_HOST_INFO(xmac_index);
+	else
+		reg = BMAC_HOST_INFO(bmac_index);
+	__set_rdc_table_num_hw(np, reg, rdc_table_num, mac_pref);
+	return 0;
+}
+
+static int niu_set_primary_mac_rdc_table(struct niu *np, int table_num,
+					 int mac_pref)
+{
+	return __set_rdc_table_num(np, 17, 0, table_num, mac_pref);
+}
+
+static int niu_set_multicast_mac_rdc_table(struct niu *np, int table_num,
+					   int mac_pref)
+{
+	return __set_rdc_table_num(np, 16, 8, table_num, mac_pref);
+}
+
+static int niu_set_alt_mac_rdc_table(struct niu *np, int idx,
+				     int table_num, int mac_pref)
+{
+	if (idx >= niu_num_alt_addr(np))
+		return -EINVAL;
+	return __set_rdc_table_num(np, idx, idx + 1, table_num, mac_pref);
+}
+
+static u64 vlan_entry_set_parity(u64 reg_val)
+{
+	u64 port01_mask;
+	u64 port23_mask;
+
+	port01_mask = 0x00ff;
+	port23_mask = 0xff00;
+
+	if (hweight64(reg_val & port01_mask) & 1)
+		reg_val |= ENET_VLAN_TBL_PARITY0;
+	else
+		reg_val &= ~ENET_VLAN_TBL_PARITY0;
+
+	if (hweight64(reg_val & port23_mask) & 1)
+		reg_val |= ENET_VLAN_TBL_PARITY1;
+	else
+		reg_val &= ~ENET_VLAN_TBL_PARITY1;
+
+	return reg_val;
+}
+
+static void vlan_tbl_write(struct niu *np, unsigned long index,
+			   int port, int vpr, int rdc_table)
+{
+	u64 reg_val = nr64(ENET_VLAN_TBL(index));
+
+	reg_val &= ~((ENET_VLAN_TBL_VPR |
+		      ENET_VLAN_TBL_VLANRDCTBLN) <<
+		     ENET_VLAN_TBL_SHIFT(port));
+	if (vpr)
+		reg_val |= (ENET_VLAN_TBL_VPR <<
+			    ENET_VLAN_TBL_SHIFT(port));
+	reg_val |= (rdc_table << ENET_VLAN_TBL_SHIFT(port));
+
+	reg_val = vlan_entry_set_parity(reg_val);
+
+	nw64(ENET_VLAN_TBL(index), reg_val);
+}
+
+static void vlan_tbl_clear(struct niu *np)
+{
+	int i;
+
+	for (i = 0; i < ENET_VLAN_TBL_NUM_ENTRIES; i++)
+		nw64(ENET_VLAN_TBL(i), 0);
+}
+
+static int tcam_wait_bit(struct niu *np, u64 bit)
+{
+	int limit = 1000;
+
+	while (--limit > 0) {
+		if (nr64(TCAM_CTL) & bit)
+			break;
+		udelay(1);
+	}
+	if (limit <= 0)
+		return -ENODEV;
+
+	return 0;
+}
+
+static int tcam_flush(struct niu *np, int index)
+{
+	nw64(TCAM_KEY_0, 0x00);
+	nw64(TCAM_KEY_MASK_0, 0xff);
+	nw64(TCAM_CTL, (TCAM_CTL_RWC_TCAM_WRITE | index));
+
+	return tcam_wait_bit(np, TCAM_CTL_STAT);
+}
+
+#if 0
+static int tcam_read(struct niu *np, int index,
+		     u64 *key, u64 *mask)
+{
+	int err;
+
+	nw64(TCAM_CTL, (TCAM_CTL_RWC_TCAM_READ | index));
+	err = tcam_wait_bit(np, TCAM_CTL_STAT);
+	if (!err) {
+		key[0] = nr64(TCAM_KEY_0);
+		key[1] = nr64(TCAM_KEY_1);
+		key[2] = nr64(TCAM_KEY_2);
+		key[3] = nr64(TCAM_KEY_3);
+		mask[0] = nr64(TCAM_KEY_MASK_0);
+		mask[1] = nr64(TCAM_KEY_MASK_1);
+		mask[2] = nr64(TCAM_KEY_MASK_2);
+		mask[3] = nr64(TCAM_KEY_MASK_3);
+	}
+	return err;
+}
+#endif
+
+static int tcam_write(struct niu *np, int index,
+		      u64 *key, u64 *mask)
+{
+	nw64(TCAM_KEY_0, key[0]);
+	nw64(TCAM_KEY_1, key[1]);
+	nw64(TCAM_KEY_2, key[2]);
+	nw64(TCAM_KEY_3, key[3]);
+	nw64(TCAM_KEY_MASK_0, mask[0]);
+	nw64(TCAM_KEY_MASK_1, mask[1]);
+	nw64(TCAM_KEY_MASK_2, mask[2]);
+	nw64(TCAM_KEY_MASK_3, mask[3]);
+	nw64(TCAM_CTL, (TCAM_CTL_RWC_TCAM_WRITE | index));
+
+	return tcam_wait_bit(np, TCAM_CTL_STAT);
+}
+
+#if 0
+static int tcam_assoc_read(struct niu *np, int index, u64 *data)
+{
+	int err;
+
+	nw64(TCAM_CTL, (TCAM_CTL_RWC_RAM_READ | index));
+	err = tcam_wait_bit(np, TCAM_CTL_STAT);
+	if (!err)
+		*data = nr64(TCAM_KEY_1);
+
+	return err;
+}
+#endif
+
+static int tcam_assoc_write(struct niu *np, int index, u64 assoc_data)
+{
+	nw64(TCAM_KEY_1, assoc_data);
+	nw64(TCAM_CTL, (TCAM_CTL_RWC_RAM_WRITE | index));
+
+	return tcam_wait_bit(np, TCAM_CTL_STAT);
+}
+
+static void tcam_enable(struct niu *np, int on)
+{
+	u64 val = nr64(FFLP_CFG_1);
+
+	if (on)
+		val &= ~FFLP_CFG_1_TCAM_DIS;
+	else
+		val |= FFLP_CFG_1_TCAM_DIS;
+	nw64(FFLP_CFG_1, val);
+}
+
+static void tcam_set_lat_and_ratio(struct niu *np, u64 latency, u64 ratio)
+{
+	u64 val = nr64(FFLP_CFG_1);
+
+	val &= ~(FFLP_CFG_1_FFLPINITDONE |
+		 FFLP_CFG_1_CAMLAT |
+		 FFLP_CFG_1_CAMRATIO);
+	val |= (latency << FFLP_CFG_1_CAMLAT_SHIFT);
+	val |= (ratio << FFLP_CFG_1_CAMRATIO_SHIFT);
+	nw64(FFLP_CFG_1, val);
+
+	val = nr64(FFLP_CFG_1);
+	val |= FFLP_CFG_1_FFLPINITDONE;
+	nw64(FFLP_CFG_1, val);
+}
+
+static int tcam_user_eth_class_enable(struct niu *np, unsigned long class,
+				      int on)
+{
+	unsigned long reg;
+	u64 val;
+
+	if (class < CLASS_CODE_ETHERTYPE1 ||
+	    class > CLASS_CODE_ETHERTYPE2)
+		return -EINVAL;
+
+	reg = L2_CLS(class - CLASS_CODE_ETHERTYPE1);
+	val = nr64(reg);
+	if (on)
+		val |= L2_CLS_VLD;
+	else
+		val &= ~L2_CLS_VLD;
+	nw64(reg, val);
+
+	return 0;
+}
+
+#if 0
+static int tcam_user_eth_class_set(struct niu *np, unsigned long class,
+				   u64 ether_type)
+{
+	unsigned long reg;
+	u64 val;
+
+	if (class < CLASS_CODE_ETHERTYPE1 ||
+	    class > CLASS_CODE_ETHERTYPE2 ||
+	    (ether_type & ~(u64)0xffff) != 0)
+		return -EINVAL;
+
+	reg = L2_CLS(class - CLASS_CODE_ETHERTYPE1);
+	val = nr64(reg);
+	val &= ~L2_CLS_ETYPE;
+	val |= (ether_type << L2_CLS_ETYPE_SHIFT);
+	nw64(reg, val);
+
+	return 0;
+}
+#endif
+
+static int tcam_user_ip_class_enable(struct niu *np, unsigned long class,
+				     int on)
+{
+	unsigned long reg;
+	u64 val;
+
+	if (class < CLASS_CODE_USER_PROG1 ||
+	    class > CLASS_CODE_USER_PROG4)
+		return -EINVAL;
+
+	reg = L3_CLS(class - CLASS_CODE_USER_PROG1);
+	val = nr64(reg);
+	if (on)
+		val |= L3_CLS_VALID;
+	else
+		val &= ~L3_CLS_VALID;
+	nw64(reg, val);
+
+	return 0;
+}
+
+static int tcam_user_ip_class_set(struct niu *np, unsigned long class,
+				  int ipv6, u64 protocol_id,
+				  u64 tos_mask, u64 tos_val)
+{
+	unsigned long reg;
+	u64 val;
+
+	if (class < CLASS_CODE_USER_PROG1 ||
+	    class > CLASS_CODE_USER_PROG4 ||
+	    (protocol_id & ~(u64)0xff) != 0 ||
+	    (tos_mask & ~(u64)0xff) != 0 ||
+	    (tos_val & ~(u64)0xff) != 0)
+		return -EINVAL;
+
+	reg = L3_CLS(class - CLASS_CODE_USER_PROG1);
+	val = nr64(reg);
+	val &= ~(L3_CLS_IPVER | L3_CLS_PID |
+		 L3_CLS_TOSMASK | L3_CLS_TOS);
+	if (ipv6)
+		val |= L3_CLS_IPVER;
+	val |= (protocol_id << L3_CLS_PID_SHIFT);
+	val |= (tos_mask << L3_CLS_TOSMASK_SHIFT);
+	val |= (tos_val << L3_CLS_TOS_SHIFT);
+	nw64(reg, val);
+
+	return 0;
+}
+
+static int tcam_early_init(struct niu *np)
+{
+	unsigned long i;
+	int err;
+
+	tcam_enable(np, 0);
+	tcam_set_lat_and_ratio(np,
+			       DEFAULT_TCAM_LATENCY,
+			       DEFAULT_TCAM_ACCESS_RATIO);
+	for (i = CLASS_CODE_ETHERTYPE1; i <= CLASS_CODE_ETHERTYPE2; i++) {
+		err = tcam_user_eth_class_enable(np, i, 0);
+		if (err)
+			return err;
+	}
+	for (i = CLASS_CODE_USER_PROG1; i <= CLASS_CODE_USER_PROG4; i++) {
+		err = tcam_user_ip_class_enable(np, i, 0);
+		if (err)
+			return err;
+	}
+
+	return 0;
+}
+
+static int tcam_flush_all(struct niu *np)
+{
+	unsigned long i;
+
+	for (i = 0; i < np->parent->tcam_num_entries; i++) {
+		int err = tcam_flush(np, i);
+		if (err)
+			return err;
+	}
+	return 0;
+}
+
+static u64 hash_addr_regval(unsigned long index, unsigned long num_entries)
+{
+	return (u64)index | (num_entries == 1 ? HASH_TBL_ADDR_AUTOINC : 0);
+}
+
+#if 0
+static int hash_read(struct niu *np, unsigned long partition,
+		     unsigned long index, unsigned long num_entries,
+		     u64 *data)
+{
+	u64 val = hash_addr_regval(index, num_entries);
+	unsigned long i;
+
+	if (partition >= FCRAM_NUM_PARTITIONS ||
+	    index + num_entries > FCRAM_SIZE)
+		return -EINVAL;
+
+	nw64(HASH_TBL_ADDR(partition), val);
+	for (i = 0; i < num_entries; i++)
+		data[i] = nr64(HASH_TBL_DATA(partition));
+
+	return 0;
+}
+#endif
+
+static int hash_write(struct niu *np, unsigned long partition,
+		      unsigned long index, unsigned long num_entries,
+		      u64 *data)
+{
+	u64 val = hash_addr_regval(index, num_entries);
+	unsigned long i;
+
+	if (partition >= FCRAM_NUM_PARTITIONS ||
+	    index + (num_entries * 8) > FCRAM_SIZE)
+		return -EINVAL;
+
+	nw64(HASH_TBL_ADDR(partition), val);
+	for (i = 0; i < num_entries; i++)
+		nw64(HASH_TBL_DATA(partition), data[i]);
+
+	return 0;
+}
+
+static void fflp_reset(struct niu *np)
+{
+	u64 val;
+
+	nw64(FFLP_CFG_1, FFLP_CFG_1_PIO_FIO_RST);
+	udelay(10);
+	nw64(FFLP_CFG_1, 0);
+
+	val = FFLP_CFG_1_FCRAMOUTDR_NORMAL | FFLP_CFG_1_FFLPINITDONE;
+	nw64(FFLP_CFG_1, val);
+}
+
+static void fflp_set_timings(struct niu *np)
+{
+	u64 val = nr64(FFLP_CFG_1);
+
+	val &= ~FFLP_CFG_1_FFLPINITDONE;
+	val |= (DEFAULT_FCRAMRATIO << FFLP_CFG_1_FCRAMRATIO_SHIFT);
+	nw64(FFLP_CFG_1, val);
+
+	val = nr64(FFLP_CFG_1);
+	val |= FFLP_CFG_1_FFLPINITDONE;
+	nw64(FFLP_CFG_1, val);
+
+	val = nr64(FCRAM_REF_TMR);
+	val &= ~(FCRAM_REF_TMR_MAX | FCRAM_REF_TMR_MIN);
+	val |= (DEFAULT_FCRAM_REFRESH_MAX << FCRAM_REF_TMR_MAX_SHIFT);
+	val |= (DEFAULT_FCRAM_REFRESH_MIN << FCRAM_REF_TMR_MIN_SHIFT);
+	nw64(FCRAM_REF_TMR, val);
+}
+
+static int fflp_set_partition(struct niu *np, u64 partition,
+			      u64 mask, u64 base, int enable)
+{
+	unsigned long reg;
+	u64 val;
+
+	if (partition >= FCRAM_NUM_PARTITIONS ||
+	    (mask & ~(u64)0x1f) != 0 ||
+	    (base & ~(u64)0x1f) != 0)
+		return -EINVAL;
+
+	reg = FLW_PRT_SEL(partition);
+
+	val = nr64(reg);
+	val &= ~(FLW_PRT_SEL_EXT | FLW_PRT_SEL_MASK | FLW_PRT_SEL_BASE);
+	val |= (mask << FLW_PRT_SEL_MASK_SHIFT);
+	val |= (base << FLW_PRT_SEL_BASE_SHIFT);
+	if (enable)
+		val |= FLW_PRT_SEL_EXT;
+	nw64(reg, val);
+
+	return 0;
+}
+
+static int fflp_disable_all_partitions(struct niu *np)
+{
+	unsigned long i;
+
+	for (i = 0; i < FCRAM_NUM_PARTITIONS; i++) {
+		int err = fflp_set_partition(np, 0, 0, 0, 0);
+		if (err)
+			return err;
+	}
+	return 0;
+}
+
+static void fflp_llcsnap_enable(struct niu *np, int on)
+{
+	u64 val = nr64(FFLP_CFG_1);
+
+	if (on)
+		val |= FFLP_CFG_1_LLCSNAP;
+	else
+		val &= ~FFLP_CFG_1_LLCSNAP;
+	nw64(FFLP_CFG_1, val);
+}
+
+static void fflp_errors_enable(struct niu *np, int on)
+{
+	u64 val = nr64(FFLP_CFG_1);
+
+	if (on)
+		val &= ~FFLP_CFG_1_ERRORDIS;
+	else
+		val |= FFLP_CFG_1_ERRORDIS;
+	nw64(FFLP_CFG_1, val);
+}
+
+static int fflp_hash_clear(struct niu *np)
+{
+	struct fcram_hash_ipv4 ent;
+	unsigned long i;
+
+	/* IPV4 hash entry with valid bit clear, rest is don't care.  */
+	memset(&ent, 0, sizeof(ent));
+	ent.header = HASH_HEADER_EXT;
+
+	for (i = 0; i < FCRAM_SIZE; i += sizeof(ent)) {
+		int err = hash_write(np, 0, i, 1, (u64 *) &ent);
+		if (err)
+			return err;
+	}
+	return 0;
+}
+
+static int fflp_early_init(struct niu *np)
+{
+	struct niu_parent *parent;
+	unsigned long flags;
+	int err;
+
+	niu_lock_parent(np, flags);
+
+	parent = np->parent;
+	err = 0;
+	if (!(parent->flags & PARENT_FLGS_CLS_HWINIT)) {
+		if (np->parent->plat_type != PLAT_TYPE_NIU) {
+			fflp_reset(np);
+			fflp_set_timings(np);
+			err = fflp_disable_all_partitions(np);
+			if (err) {
+				netif_printk(np, probe, KERN_DEBUG, np->dev,
+					     "fflp_disable_all_partitions failed, err=%d\n",
+					     err);
+				goto out;
+			}
+		}
+
+		err = tcam_early_init(np);
+		if (err) {
+			netif_printk(np, probe, KERN_DEBUG, np->dev,
+				     "tcam_early_init failed, err=%d\n", err);
+			goto out;
+		}
+		fflp_llcsnap_enable(np, 1);
+		fflp_errors_enable(np, 0);
+		nw64(H1POLY, 0);
+		nw64(H2POLY, 0);
+
+		err = tcam_flush_all(np);
+		if (err) {
+			netif_printk(np, probe, KERN_DEBUG, np->dev,
+				     "tcam_flush_all failed, err=%d\n", err);
+			goto out;
+		}
+		if (np->parent->plat_type != PLAT_TYPE_NIU) {
+			err = fflp_hash_clear(np);
+			if (err) {
+				netif_printk(np, probe, KERN_DEBUG, np->dev,
+					     "fflp_hash_clear failed, err=%d\n",
+					     err);
+				goto out;
+			}
+		}
+
+		vlan_tbl_clear(np);
+
+		parent->flags |= PARENT_FLGS_CLS_HWINIT;
+	}
+out:
+	niu_unlock_parent(np, flags);
+	return err;
+}
+
+static int niu_set_flow_key(struct niu *np, unsigned long class_code, u64 key)
+{
+	if (class_code < CLASS_CODE_USER_PROG1 ||
+	    class_code > CLASS_CODE_SCTP_IPV6)
+		return -EINVAL;
+
+	nw64(FLOW_KEY(class_code - CLASS_CODE_USER_PROG1), key);
+	return 0;
+}
+
+static int niu_set_tcam_key(struct niu *np, unsigned long class_code, u64 key)
+{
+	if (class_code < CLASS_CODE_USER_PROG1 ||
+	    class_code > CLASS_CODE_SCTP_IPV6)
+		return -EINVAL;
+
+	nw64(TCAM_KEY(class_code - CLASS_CODE_USER_PROG1), key);
+	return 0;
+}
+
+/* Entries for the ports are interleaved in the TCAM */
+static u16 tcam_get_index(struct niu *np, u16 idx)
+{
+	/* One entry reserved for IP fragment rule */
+	if (idx >= (np->clas.tcam_sz - 1))
+		idx = 0;
+	return np->clas.tcam_top + ((idx+1) * np->parent->num_ports);
+}
+
+static u16 tcam_get_size(struct niu *np)
+{
+	/* One entry reserved for IP fragment rule */
+	return np->clas.tcam_sz - 1;
+}
+
+static u16 tcam_get_valid_entry_cnt(struct niu *np)
+{
+	/* One entry reserved for IP fragment rule */
+	return np->clas.tcam_valid_entries - 1;
+}
+
+static void niu_rx_skb_append(struct sk_buff *skb, struct page *page,
+			      u32 offset, u32 size, u32 truesize)
+{
+	skb_fill_page_desc(skb, skb_shinfo(skb)->nr_frags, page, offset, size);
+
+	skb->len += size;
+	skb->data_len += size;
+	skb->truesize += truesize;
+}
+
+static unsigned int niu_hash_rxaddr(struct rx_ring_info *rp, u64 a)
+{
+	a >>= PAGE_SHIFT;
+	a ^= (a >> ilog2(MAX_RBR_RING_SIZE));
+
+	return a & (MAX_RBR_RING_SIZE - 1);
+}
+
+static struct page *niu_find_rxpage(struct rx_ring_info *rp, u64 addr,
+				    struct page ***link)
+{
+	unsigned int h = niu_hash_rxaddr(rp, addr);
+	struct page *p, **pp;
+
+	addr &= PAGE_MASK;
+	pp = &rp->rxhash[h];
+	for (; (p = *pp) != NULL; pp = &niu_next_page(p)) {
+		if (p->private == addr) {
+			*link = pp;
+			goto found;
+		}
+	}
+	BUG();
+
+found:
+	return p;
+}
+
+static void niu_hash_page(struct rx_ring_info *rp, struct page *page, u64 base)
+{
+	unsigned int h = niu_hash_rxaddr(rp, base);
+
+	page->private = base;
+	niu_next_page(page) = rp->rxhash[h];
+	rp->rxhash[h] = page;
+}
+
+static int niu_rbr_add_page(struct niu *np, struct rx_ring_info *rp,
+			    gfp_t mask, int start_index)
+{
+	struct page *page;
+	u64 addr;
+	int i;
+
+	page = alloc_page(mask);
+	if (!page)
+		return -ENOMEM;
+
+	addr = np->ops->map_page(np->device, page, 0,
+				 PAGE_SIZE, DMA_FROM_DEVICE);
+	if (np->ops->mapping_error(np->device, addr)) {
+		__free_page(page);
+		return -ENOMEM;
+	}
+
+	niu_hash_page(rp, page, addr);
+	if (rp->rbr_blocks_per_page > 1)
+		page_ref_add(page, rp->rbr_blocks_per_page - 1);
+
+	for (i = 0; i < rp->rbr_blocks_per_page; i++) {
+		__le32 *rbr = &rp->rbr[start_index + i];
+
+		*rbr = cpu_to_le32(addr >> RBR_DESCR_ADDR_SHIFT);
+		addr += rp->rbr_block_size;
+	}
+
+	return 0;
+}
+
+static void niu_rbr_refill(struct niu *np, struct rx_ring_info *rp, gfp_t mask)
+{
+	int index = rp->rbr_index;
+
+	rp->rbr_pending++;
+	if ((rp->rbr_pending % rp->rbr_blocks_per_page) == 0) {
+		int err = niu_rbr_add_page(np, rp, mask, index);
+
+		if (unlikely(err)) {
+			rp->rbr_pending--;
+			return;
+		}
+
+		rp->rbr_index += rp->rbr_blocks_per_page;
+		BUG_ON(rp->rbr_index > rp->rbr_table_size);
+		if (rp->rbr_index == rp->rbr_table_size)
+			rp->rbr_index = 0;
+
+		if (rp->rbr_pending >= rp->rbr_kick_thresh) {
+			nw64(RBR_KICK(rp->rx_channel), rp->rbr_pending);
+			rp->rbr_pending = 0;
+		}
+	}
+}
+
+static int niu_rx_pkt_ignore(struct niu *np, struct rx_ring_info *rp)
+{
+	unsigned int index = rp->rcr_index;
+	int num_rcr = 0;
+
+	rp->rx_dropped++;
+	while (1) {
+		struct page *page, **link;
+		u64 addr, val;
+		u32 rcr_size;
+
+		num_rcr++;
+
+		val = le64_to_cpup(&rp->rcr[index]);
+		addr = (val & RCR_ENTRY_PKT_BUF_ADDR) <<
+			RCR_ENTRY_PKT_BUF_ADDR_SHIFT;
+		page = niu_find_rxpage(rp, addr, &link);
+
+		rcr_size = rp->rbr_sizes[(val & RCR_ENTRY_PKTBUFSZ) >>
+					 RCR_ENTRY_PKTBUFSZ_SHIFT];
+		if ((page->private + PAGE_SIZE) - rcr_size == addr) {
+			*link = niu_next_page(page);
+			np->ops->unmap_page(np->device, page->private,
+					    PAGE_SIZE, DMA_FROM_DEVICE);
+			page->private = 0;
+			niu_next_page(page) = NULL;
+			__free_page(page);
+			rp->rbr_refill_pending++;
+		}
+
+		index = NEXT_RCR(rp, index);
+		if (!(val & RCR_ENTRY_MULTI))
+			break;
+
+	}
+	rp->rcr_index = index;
+
+	return num_rcr;
+}
+
+static int niu_process_rx_pkt(struct napi_struct *napi, struct niu *np,
+			      struct rx_ring_info *rp)
+{
+	unsigned int index = rp->rcr_index;
+	struct rx_pkt_hdr1 *rh;
+	struct sk_buff *skb;
+	int len, num_rcr;
+
+	skb = netdev_alloc_skb(np->dev, RX_SKB_ALLOC_SIZE);
+	if (unlikely(!skb))
+		return niu_rx_pkt_ignore(np, rp);
+
+	num_rcr = 0;
+	while (1) {
+		struct page *page, **link;
+		u32 rcr_size, append_size;
+		u64 addr, val, off;
+
+		num_rcr++;
+
+		val = le64_to_cpup(&rp->rcr[index]);
+
+		len = (val & RCR_ENTRY_L2_LEN) >>
+			RCR_ENTRY_L2_LEN_SHIFT;
+		append_size = len + ETH_HLEN + ETH_FCS_LEN;
+
+		addr = (val & RCR_ENTRY_PKT_BUF_ADDR) <<
+			RCR_ENTRY_PKT_BUF_ADDR_SHIFT;
+		page = niu_find_rxpage(rp, addr, &link);
+
+		rcr_size = rp->rbr_sizes[(val & RCR_ENTRY_PKTBUFSZ) >>
+					 RCR_ENTRY_PKTBUFSZ_SHIFT];
+
+		off = addr & ~PAGE_MASK;
+		if (num_rcr == 1) {
+			int ptype;
+
+			ptype = (val >> RCR_ENTRY_PKT_TYPE_SHIFT);
+			if ((ptype == RCR_PKT_TYPE_TCP ||
+			     ptype == RCR_PKT_TYPE_UDP) &&
+			    !(val & (RCR_ENTRY_NOPORT |
+				     RCR_ENTRY_ERROR)))
+				skb->ip_summed = CHECKSUM_UNNECESSARY;
+			else
+				skb_checksum_none_assert(skb);
+		} else if (!(val & RCR_ENTRY_MULTI))
+			append_size = append_size - skb->len;
+
+		niu_rx_skb_append(skb, page, off, append_size, rcr_size);
+		if ((page->private + rp->rbr_block_size) - rcr_size == addr) {
+			*link = niu_next_page(page);
+			np->ops->unmap_page(np->device, page->private,
+					    PAGE_SIZE, DMA_FROM_DEVICE);
+			page->private = 0;
+			niu_next_page(page) = NULL;
+			rp->rbr_refill_pending++;
+		} else
+			get_page(page);
+
+		index = NEXT_RCR(rp, index);
+		if (!(val & RCR_ENTRY_MULTI))
+			break;
+
+	}
+	rp->rcr_index = index;
+
+	len += sizeof(*rh);
+	len = min_t(int, len, sizeof(*rh) + VLAN_ETH_HLEN);
+	__pskb_pull_tail(skb, len);
+
+	rh = (struct rx_pkt_hdr1 *) skb->data;
+	if (np->dev->features & NETIF_F_RXHASH)
+		skb_set_hash(skb,
+			     ((u32)rh->hashval2_0 << 24 |
+			      (u32)rh->hashval2_1 << 16 |
+			      (u32)rh->hashval1_1 << 8 |
+			      (u32)rh->hashval1_2 << 0),
+			     PKT_HASH_TYPE_L3);
+	skb_pull(skb, sizeof(*rh));
+
+	rp->rx_packets++;
+	rp->rx_bytes += skb->len;
+
+	skb->protocol = eth_type_trans(skb, np->dev);
+	skb_record_rx_queue(skb, rp->rx_channel);
+	napi_gro_receive(napi, skb);
+
+	return num_rcr;
+}
+
+static int niu_rbr_fill(struct niu *np, struct rx_ring_info *rp, gfp_t mask)
+{
+	int blocks_per_page = rp->rbr_blocks_per_page;
+	int err, index = rp->rbr_index;
+
+	err = 0;
+	while (index < (rp->rbr_table_size - blocks_per_page)) {
+		err = niu_rbr_add_page(np, rp, mask, index);
+		if (unlikely(err))
+			break;
+
+		index += blocks_per_page;
+	}
+
+	rp->rbr_index = index;
+	return err;
+}
+
+static void niu_rbr_free(struct niu *np, struct rx_ring_info *rp)
+{
+	int i;
+
+	for (i = 0; i < MAX_RBR_RING_SIZE; i++) {
+		struct page *page;
+
+		page = rp->rxhash[i];
+		while (page) {
+			struct page *next = niu_next_page(page);
+			u64 base = page->private;
+
+			np->ops->unmap_page(np->device, base, PAGE_SIZE,
+					    DMA_FROM_DEVICE);
+			page->private = 0;
+			niu_next_page(page) = NULL;
+
+			__free_page(page);
+
+			page = next;
+		}
+	}
+
+	for (i = 0; i < rp->rbr_table_size; i++)
+		rp->rbr[i] = cpu_to_le32(0);
+	rp->rbr_index = 0;
+}
+
+static int release_tx_packet(struct niu *np, struct tx_ring_info *rp, int idx)
+{
+	struct tx_buff_info *tb = &rp->tx_buffs[idx];
+	struct sk_buff *skb = tb->skb;
+	struct tx_pkt_hdr *tp;
+	u64 tx_flags;
+	int i, len;
+
+	tp = (struct tx_pkt_hdr *) skb->data;
+	tx_flags = le64_to_cpup(&tp->flags);
+
+	rp->tx_packets++;
+	rp->tx_bytes += (((tx_flags & TXHDR_LEN) >> TXHDR_LEN_SHIFT) -
+			 ((tx_flags & TXHDR_PAD) / 2));
+
+	len = skb_headlen(skb);
+	np->ops->unmap_single(np->device, tb->mapping,
+			      len, DMA_TO_DEVICE);
+
+	if (le64_to_cpu(rp->descr[idx]) & TX_DESC_MARK)
+		rp->mark_pending--;
+
+	tb->skb = NULL;
+	do {
+		idx = NEXT_TX(rp, idx);
+		len -= MAX_TX_DESC_LEN;
+	} while (len > 0);
+
+	for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
+		tb = &rp->tx_buffs[idx];
+		BUG_ON(tb->skb != NULL);
+		np->ops->unmap_page(np->device, tb->mapping,
+				    skb_frag_size(&skb_shinfo(skb)->frags[i]),
+				    DMA_TO_DEVICE);
+		idx = NEXT_TX(rp, idx);
+	}
+
+	dev_kfree_skb(skb);
+
+	return idx;
+}
+
+#define NIU_TX_WAKEUP_THRESH(rp)		((rp)->pending / 4)
+
+static void niu_tx_work(struct niu *np, struct tx_ring_info *rp)
+{
+	struct netdev_queue *txq;
+	u16 pkt_cnt, tmp;
+	int cons, index;
+	u64 cs;
+
+	index = (rp - np->tx_rings);
+	txq = netdev_get_tx_queue(np->dev, index);
+
+	cs = rp->tx_cs;
+	if (unlikely(!(cs & (TX_CS_MK | TX_CS_MMK))))
+		goto out;
+
+	tmp = pkt_cnt = (cs & TX_CS_PKT_CNT) >> TX_CS_PKT_CNT_SHIFT;
+	pkt_cnt = (pkt_cnt - rp->last_pkt_cnt) &
+		(TX_CS_PKT_CNT >> TX_CS_PKT_CNT_SHIFT);
+
+	rp->last_pkt_cnt = tmp;
+
+	cons = rp->cons;
+
+	netif_printk(np, tx_done, KERN_DEBUG, np->dev,
+		     "%s() pkt_cnt[%u] cons[%d]\n", __func__, pkt_cnt, cons);
+
+	while (pkt_cnt--)
+		cons = release_tx_packet(np, rp, cons);
+
+	rp->cons = cons;
+	smp_mb();
+
+out:
+	if (unlikely(netif_tx_queue_stopped(txq) &&
+		     (niu_tx_avail(rp) > NIU_TX_WAKEUP_THRESH(rp)))) {
+		__netif_tx_lock(txq, smp_processor_id());
+		if (netif_tx_queue_stopped(txq) &&
+		    (niu_tx_avail(rp) > NIU_TX_WAKEUP_THRESH(rp)))
+			netif_tx_wake_queue(txq);
+		__netif_tx_unlock(txq);
+	}
+}
+
+static inline void niu_sync_rx_discard_stats(struct niu *np,
+					     struct rx_ring_info *rp,
+					     const int limit)
+{
+	/* This elaborate scheme is needed for reading the RX discard
+	 * counters, as they are only 16-bit and can overflow quickly,
+	 * and because the overflow indication bit is not usable as
+	 * the counter value does not wrap, but remains at max value
+	 * 0xFFFF.
+	 *
+	 * In theory and in practice counters can be lost in between
+	 * reading nr64() and clearing the counter nw64().  For this
+	 * reason, the number of counter clearings nw64() is
+	 * limited/reduced though the limit parameter.
+	 */
+	int rx_channel = rp->rx_channel;
+	u32 misc, wred;
+
+	/* RXMISC (Receive Miscellaneous Discard Count), covers the
+	 * following discard events: IPP (Input Port Process),
+	 * FFLP/TCAM, Full RCR (Receive Completion Ring) RBR (Receive
+	 * Block Ring) prefetch buffer is empty.
+	 */
+	misc = nr64(RXMISC(rx_channel));
+	if (unlikely((misc & RXMISC_COUNT) > limit)) {
+		nw64(RXMISC(rx_channel), 0);
+		rp->rx_errors += misc & RXMISC_COUNT;
+
+		if (unlikely(misc & RXMISC_OFLOW))
+			dev_err(np->device, "rx-%d: Counter overflow RXMISC discard\n",
+				rx_channel);
+
+		netif_printk(np, rx_err, KERN_DEBUG, np->dev,
+			     "rx-%d: MISC drop=%u over=%u\n",
+			     rx_channel, misc, misc-limit);
+	}
+
+	/* WRED (Weighted Random Early Discard) by hardware */
+	wred = nr64(RED_DIS_CNT(rx_channel));
+	if (unlikely((wred & RED_DIS_CNT_COUNT) > limit)) {
+		nw64(RED_DIS_CNT(rx_channel), 0);
+		rp->rx_dropped += wred & RED_DIS_CNT_COUNT;
+
+		if (unlikely(wred & RED_DIS_CNT_OFLOW))
+			dev_err(np->device, "rx-%d: Counter overflow WRED discard\n", rx_channel);
+
+		netif_printk(np, rx_err, KERN_DEBUG, np->dev,
+			     "rx-%d: WRED drop=%u over=%u\n",
+			     rx_channel, wred, wred-limit);
+	}
+}
+
+static int niu_rx_work(struct napi_struct *napi, struct niu *np,
+		       struct rx_ring_info *rp, int budget)
+{
+	int qlen, rcr_done = 0, work_done = 0;
+	struct rxdma_mailbox *mbox = rp->mbox;
+	u64 stat;
+
+#if 1
+	stat = nr64(RX_DMA_CTL_STAT(rp->rx_channel));
+	qlen = nr64(RCRSTAT_A(rp->rx_channel)) & RCRSTAT_A_QLEN;
+#else
+	stat = le64_to_cpup(&mbox->rx_dma_ctl_stat);
+	qlen = (le64_to_cpup(&mbox->rcrstat_a) & RCRSTAT_A_QLEN);
+#endif
+	mbox->rx_dma_ctl_stat = 0;
+	mbox->rcrstat_a = 0;
+
+	netif_printk(np, rx_status, KERN_DEBUG, np->dev,
+		     "%s(chan[%d]), stat[%llx] qlen=%d\n",
+		     __func__, rp->rx_channel, (unsigned long long)stat, qlen);
+
+	rcr_done = work_done = 0;
+	qlen = min(qlen, budget);
+	while (work_done < qlen) {
+		rcr_done += niu_process_rx_pkt(napi, np, rp);
+		work_done++;
+	}
+
+	if (rp->rbr_refill_pending >= rp->rbr_kick_thresh) {
+		unsigned int i;
+
+		for (i = 0; i < rp->rbr_refill_pending; i++)
+			niu_rbr_refill(np, rp, GFP_ATOMIC);
+		rp->rbr_refill_pending = 0;
+	}
+
+	stat = (RX_DMA_CTL_STAT_MEX |
+		((u64)work_done << RX_DMA_CTL_STAT_PKTREAD_SHIFT) |
+		((u64)rcr_done << RX_DMA_CTL_STAT_PTRREAD_SHIFT));
+
+	nw64(RX_DMA_CTL_STAT(rp->rx_channel), stat);
+
+	/* Only sync discards stats when qlen indicate potential for drops */
+	if (qlen > 10)
+		niu_sync_rx_discard_stats(np, rp, 0x7FFF);
+
+	return work_done;
+}
+
+static int niu_poll_core(struct niu *np, struct niu_ldg *lp, int budget)
+{
+	u64 v0 = lp->v0;
+	u32 tx_vec = (v0 >> 32);
+	u32 rx_vec = (v0 & 0xffffffff);
+	int i, work_done = 0;
+
+	netif_printk(np, intr, KERN_DEBUG, np->dev,
+		     "%s() v0[%016llx]\n", __func__, (unsigned long long)v0);
+
+	for (i = 0; i < np->num_tx_rings; i++) {
+		struct tx_ring_info *rp = &np->tx_rings[i];
+		if (tx_vec & (1 << rp->tx_channel))
+			niu_tx_work(np, rp);
+		nw64(LD_IM0(LDN_TXDMA(rp->tx_channel)), 0);
+	}
+
+	for (i = 0; i < np->num_rx_rings; i++) {
+		struct rx_ring_info *rp = &np->rx_rings[i];
+
+		if (rx_vec & (1 << rp->rx_channel)) {
+			int this_work_done;
+
+			this_work_done = niu_rx_work(&lp->napi, np, rp,
+						     budget);
+
+			budget -= this_work_done;
+			work_done += this_work_done;
+		}
+		nw64(LD_IM0(LDN_RXDMA(rp->rx_channel)), 0);
+	}
+
+	return work_done;
+}
+
+static int niu_poll(struct napi_struct *napi, int budget)
+{
+	struct niu_ldg *lp = container_of(napi, struct niu_ldg, napi);
+	struct niu *np = lp->np;
+	int work_done;
+
+	work_done = niu_poll_core(np, lp, budget);
+
+	if (work_done < budget) {
+		napi_complete_done(napi, work_done);
+		niu_ldg_rearm(np, lp, 1);
+	}
+	return work_done;
+}
+
+static void niu_log_rxchan_errors(struct niu *np, struct rx_ring_info *rp,
+				  u64 stat)
+{
+	netdev_err(np->dev, "RX channel %u errors ( ", rp->rx_channel);
+
+	if (stat & RX_DMA_CTL_STAT_RBR_TMOUT)
+		pr_cont("RBR_TMOUT ");
+	if (stat & RX_DMA_CTL_STAT_RSP_CNT_ERR)
+		pr_cont("RSP_CNT ");
+	if (stat & RX_DMA_CTL_STAT_BYTE_EN_BUS)
+		pr_cont("BYTE_EN_BUS ");
+	if (stat & RX_DMA_CTL_STAT_RSP_DAT_ERR)
+		pr_cont("RSP_DAT ");
+	if (stat & RX_DMA_CTL_STAT_RCR_ACK_ERR)
+		pr_cont("RCR_ACK ");
+	if (stat & RX_DMA_CTL_STAT_RCR_SHA_PAR)
+		pr_cont("RCR_SHA_PAR ");
+	if (stat & RX_DMA_CTL_STAT_RBR_PRE_PAR)
+		pr_cont("RBR_PRE_PAR ");
+	if (stat & RX_DMA_CTL_STAT_CONFIG_ERR)
+		pr_cont("CONFIG ");
+	if (stat & RX_DMA_CTL_STAT_RCRINCON)
+		pr_cont("RCRINCON ");
+	if (stat & RX_DMA_CTL_STAT_RCRFULL)
+		pr_cont("RCRFULL ");
+	if (stat & RX_DMA_CTL_STAT_RBRFULL)
+		pr_cont("RBRFULL ");
+	if (stat & RX_DMA_CTL_STAT_RBRLOGPAGE)
+		pr_cont("RBRLOGPAGE ");
+	if (stat & RX_DMA_CTL_STAT_CFIGLOGPAGE)
+		pr_cont("CFIGLOGPAGE ");
+	if (stat & RX_DMA_CTL_STAT_DC_FIFO_ERR)
+		pr_cont("DC_FIDO ");
+
+	pr_cont(")\n");
+}
+
+static int niu_rx_error(struct niu *np, struct rx_ring_info *rp)
+{
+	u64 stat = nr64(RX_DMA_CTL_STAT(rp->rx_channel));
+	int err = 0;
+
+
+	if (stat & (RX_DMA_CTL_STAT_CHAN_FATAL |
+		    RX_DMA_CTL_STAT_PORT_FATAL))
+		err = -EINVAL;
+
+	if (err) {
+		netdev_err(np->dev, "RX channel %u error, stat[%llx]\n",
+			   rp->rx_channel,
+			   (unsigned long long) stat);
+
+		niu_log_rxchan_errors(np, rp, stat);
+	}
+
+	nw64(RX_DMA_CTL_STAT(rp->rx_channel),
+	     stat & RX_DMA_CTL_WRITE_CLEAR_ERRS);
+
+	return err;
+}
+
+static void niu_log_txchan_errors(struct niu *np, struct tx_ring_info *rp,
+				  u64 cs)
+{
+	netdev_err(np->dev, "TX channel %u errors ( ", rp->tx_channel);
+
+	if (cs & TX_CS_MBOX_ERR)
+		pr_cont("MBOX ");
+	if (cs & TX_CS_PKT_SIZE_ERR)
+		pr_cont("PKT_SIZE ");
+	if (cs & TX_CS_TX_RING_OFLOW)
+		pr_cont("TX_RING_OFLOW ");
+	if (cs & TX_CS_PREF_BUF_PAR_ERR)
+		pr_cont("PREF_BUF_PAR ");
+	if (cs & TX_CS_NACK_PREF)
+		pr_cont("NACK_PREF ");
+	if (cs & TX_CS_NACK_PKT_RD)
+		pr_cont("NACK_PKT_RD ");
+	if (cs & TX_CS_CONF_PART_ERR)
+		pr_cont("CONF_PART ");
+	if (cs & TX_CS_PKT_PRT_ERR)
+		pr_cont("PKT_PTR ");
+
+	pr_cont(")\n");
+}
+
+static int niu_tx_error(struct niu *np, struct tx_ring_info *rp)
+{
+	u64 cs, logh, logl;
+
+	cs = nr64(TX_CS(rp->tx_channel));
+	logh = nr64(TX_RNG_ERR_LOGH(rp->tx_channel));
+	logl = nr64(TX_RNG_ERR_LOGL(rp->tx_channel));
+
+	netdev_err(np->dev, "TX channel %u error, cs[%llx] logh[%llx] logl[%llx]\n",
+		   rp->tx_channel,
+		   (unsigned long long)cs,
+		   (unsigned long long)logh,
+		   (unsigned long long)logl);
+
+	niu_log_txchan_errors(np, rp, cs);
+
+	return -ENODEV;
+}
+
+static int niu_mif_interrupt(struct niu *np)
+{
+	u64 mif_status = nr64(MIF_STATUS);
+	int phy_mdint = 0;
+
+	if (np->flags & NIU_FLAGS_XMAC) {
+		u64 xrxmac_stat = nr64_mac(XRXMAC_STATUS);
+
+		if (xrxmac_stat & XRXMAC_STATUS_PHY_MDINT)
+			phy_mdint = 1;
+	}
+
+	netdev_err(np->dev, "MIF interrupt, stat[%llx] phy_mdint(%d)\n",
+		   (unsigned long long)mif_status, phy_mdint);
+
+	return -ENODEV;
+}
+
+static void niu_xmac_interrupt(struct niu *np)
+{
+	struct niu_xmac_stats *mp = &np->mac_stats.xmac;
+	u64 val;
+
+	val = nr64_mac(XTXMAC_STATUS);
+	if (val & XTXMAC_STATUS_FRAME_CNT_EXP)
+		mp->tx_frames += TXMAC_FRM_CNT_COUNT;
+	if (val & XTXMAC_STATUS_BYTE_CNT_EXP)
+		mp->tx_bytes += TXMAC_BYTE_CNT_COUNT;
+	if (val & XTXMAC_STATUS_TXFIFO_XFR_ERR)
+		mp->tx_fifo_errors++;
+	if (val & XTXMAC_STATUS_TXMAC_OFLOW)
+		mp->tx_overflow_errors++;
+	if (val & XTXMAC_STATUS_MAX_PSIZE_ERR)
+		mp->tx_max_pkt_size_errors++;
+	if (val & XTXMAC_STATUS_TXMAC_UFLOW)
+		mp->tx_underflow_errors++;
+
+	val = nr64_mac(XRXMAC_STATUS);
+	if (val & XRXMAC_STATUS_LCL_FLT_STATUS)
+		mp->rx_local_faults++;
+	if (val & XRXMAC_STATUS_RFLT_DET)
+		mp->rx_remote_faults++;
+	if (val & XRXMAC_STATUS_LFLT_CNT_EXP)
+		mp->rx_link_faults += LINK_FAULT_CNT_COUNT;
+	if (val & XRXMAC_STATUS_ALIGNERR_CNT_EXP)
+		mp->rx_align_errors += RXMAC_ALIGN_ERR_CNT_COUNT;
+	if (val & XRXMAC_STATUS_RXFRAG_CNT_EXP)
+		mp->rx_frags += RXMAC_FRAG_CNT_COUNT;
+	if (val & XRXMAC_STATUS_RXMULTF_CNT_EXP)
+		mp->rx_mcasts += RXMAC_MC_FRM_CNT_COUNT;
+	if (val & XRXMAC_STATUS_RXBCAST_CNT_EXP)
+		mp->rx_bcasts += RXMAC_BC_FRM_CNT_COUNT;
+	if (val & XRXMAC_STATUS_RXHIST1_CNT_EXP)
+		mp->rx_hist_cnt1 += RXMAC_HIST_CNT1_COUNT;
+	if (val & XRXMAC_STATUS_RXHIST2_CNT_EXP)
+		mp->rx_hist_cnt2 += RXMAC_HIST_CNT2_COUNT;
+	if (val & XRXMAC_STATUS_RXHIST3_CNT_EXP)
+		mp->rx_hist_cnt3 += RXMAC_HIST_CNT3_COUNT;
+	if (val & XRXMAC_STATUS_RXHIST4_CNT_EXP)
+		mp->rx_hist_cnt4 += RXMAC_HIST_CNT4_COUNT;
+	if (val & XRXMAC_STATUS_RXHIST5_CNT_EXP)
+		mp->rx_hist_cnt5 += RXMAC_HIST_CNT5_COUNT;
+	if (val & XRXMAC_STATUS_RXHIST6_CNT_EXP)
+		mp->rx_hist_cnt6 += RXMAC_HIST_CNT6_COUNT;
+	if (val & XRXMAC_STATUS_RXHIST7_CNT_EXP)
+		mp->rx_hist_cnt7 += RXMAC_HIST_CNT7_COUNT;
+	if (val & XRXMAC_STATUS_RXOCTET_CNT_EXP)
+		mp->rx_octets += RXMAC_BT_CNT_COUNT;
+	if (val & XRXMAC_STATUS_CVIOLERR_CNT_EXP)
+		mp->rx_code_violations += RXMAC_CD_VIO_CNT_COUNT;
+	if (val & XRXMAC_STATUS_LENERR_CNT_EXP)
+		mp->rx_len_errors += RXMAC_MPSZER_CNT_COUNT;
+	if (val & XRXMAC_STATUS_CRCERR_CNT_EXP)
+		mp->rx_crc_errors += RXMAC_CRC_ER_CNT_COUNT;
+	if (val & XRXMAC_STATUS_RXUFLOW)
+		mp->rx_underflows++;
+	if (val & XRXMAC_STATUS_RXOFLOW)
+		mp->rx_overflows++;
+
+	val = nr64_mac(XMAC_FC_STAT);
+	if (val & XMAC_FC_STAT_TX_MAC_NPAUSE)
+		mp->pause_off_state++;
+	if (val & XMAC_FC_STAT_TX_MAC_PAUSE)
+		mp->pause_on_state++;
+	if (val & XMAC_FC_STAT_RX_MAC_RPAUSE)
+		mp->pause_received++;
+}
+
+static void niu_bmac_interrupt(struct niu *np)
+{
+	struct niu_bmac_stats *mp = &np->mac_stats.bmac;
+	u64 val;
+
+	val = nr64_mac(BTXMAC_STATUS);
+	if (val & BTXMAC_STATUS_UNDERRUN)
+		mp->tx_underflow_errors++;
+	if (val & BTXMAC_STATUS_MAX_PKT_ERR)
+		mp->tx_max_pkt_size_errors++;
+	if (val & BTXMAC_STATUS_BYTE_CNT_EXP)
+		mp->tx_bytes += BTXMAC_BYTE_CNT_COUNT;
+	if (val & BTXMAC_STATUS_FRAME_CNT_EXP)
+		mp->tx_frames += BTXMAC_FRM_CNT_COUNT;
+
+	val = nr64_mac(BRXMAC_STATUS);
+	if (val & BRXMAC_STATUS_OVERFLOW)
+		mp->rx_overflows++;
+	if (val & BRXMAC_STATUS_FRAME_CNT_EXP)
+		mp->rx_frames += BRXMAC_FRAME_CNT_COUNT;
+	if (val & BRXMAC_STATUS_ALIGN_ERR_EXP)
+		mp->rx_align_errors += BRXMAC_ALIGN_ERR_CNT_COUNT;
+	if (val & BRXMAC_STATUS_CRC_ERR_EXP)
+		mp->rx_crc_errors += BRXMAC_ALIGN_ERR_CNT_COUNT;
+	if (val & BRXMAC_STATUS_LEN_ERR_EXP)
+		mp->rx_len_errors += BRXMAC_CODE_VIOL_ERR_CNT_COUNT;
+
+	val = nr64_mac(BMAC_CTRL_STATUS);
+	if (val & BMAC_CTRL_STATUS_NOPAUSE)
+		mp->pause_off_state++;
+	if (val & BMAC_CTRL_STATUS_PAUSE)
+		mp->pause_on_state++;
+	if (val & BMAC_CTRL_STATUS_PAUSE_RECV)
+		mp->pause_received++;
+}
+
+static int niu_mac_interrupt(struct niu *np)
+{
+	if (np->flags & NIU_FLAGS_XMAC)
+		niu_xmac_interrupt(np);
+	else
+		niu_bmac_interrupt(np);
+
+	return 0;
+}
+
+static void niu_log_device_error(struct niu *np, u64 stat)
+{
+	netdev_err(np->dev, "Core device errors ( ");
+
+	if (stat & SYS_ERR_MASK_META2)
+		pr_cont("META2 ");
+	if (stat & SYS_ERR_MASK_META1)
+		pr_cont("META1 ");
+	if (stat & SYS_ERR_MASK_PEU)
+		pr_cont("PEU ");
+	if (stat & SYS_ERR_MASK_TXC)
+		pr_cont("TXC ");
+	if (stat & SYS_ERR_MASK_RDMC)
+		pr_cont("RDMC ");
+	if (stat & SYS_ERR_MASK_TDMC)
+		pr_cont("TDMC ");
+	if (stat & SYS_ERR_MASK_ZCP)
+		pr_cont("ZCP ");
+	if (stat & SYS_ERR_MASK_FFLP)
+		pr_cont("FFLP ");
+	if (stat & SYS_ERR_MASK_IPP)
+		pr_cont("IPP ");
+	if (stat & SYS_ERR_MASK_MAC)
+		pr_cont("MAC ");
+	if (stat & SYS_ERR_MASK_SMX)
+		pr_cont("SMX ");
+
+	pr_cont(")\n");
+}
+
+static int niu_device_error(struct niu *np)
+{
+	u64 stat = nr64(SYS_ERR_STAT);
+
+	netdev_err(np->dev, "Core device error, stat[%llx]\n",
+		   (unsigned long long)stat);
+
+	niu_log_device_error(np, stat);
+
+	return -ENODEV;
+}
+
+static int niu_slowpath_interrupt(struct niu *np, struct niu_ldg *lp,
+			      u64 v0, u64 v1, u64 v2)
+{
+
+	int i, err = 0;
+
+	lp->v0 = v0;
+	lp->v1 = v1;
+	lp->v2 = v2;
+
+	if (v1 & 0x00000000ffffffffULL) {
+		u32 rx_vec = (v1 & 0xffffffff);
+
+		for (i = 0; i < np->num_rx_rings; i++) {
+			struct rx_ring_info *rp = &np->rx_rings[i];
+
+			if (rx_vec & (1 << rp->rx_channel)) {
+				int r = niu_rx_error(np, rp);
+				if (r) {
+					err = r;
+				} else {
+					if (!v0)
+						nw64(RX_DMA_CTL_STAT(rp->rx_channel),
+						     RX_DMA_CTL_STAT_MEX);
+				}
+			}
+		}
+	}
+	if (v1 & 0x7fffffff00000000ULL) {
+		u32 tx_vec = (v1 >> 32) & 0x7fffffff;
+
+		for (i = 0; i < np->num_tx_rings; i++) {
+			struct tx_ring_info *rp = &np->tx_rings[i];
+
+			if (tx_vec & (1 << rp->tx_channel)) {
+				int r = niu_tx_error(np, rp);
+				if (r)
+					err = r;
+			}
+		}
+	}
+	if ((v0 | v1) & 0x8000000000000000ULL) {
+		int r = niu_mif_interrupt(np);
+		if (r)
+			err = r;
+	}
+	if (v2) {
+		if (v2 & 0x01ef) {
+			int r = niu_mac_interrupt(np);
+			if (r)
+				err = r;
+		}
+		if (v2 & 0x0210) {
+			int r = niu_device_error(np);
+			if (r)
+				err = r;
+		}
+	}
+
+	if (err)
+		niu_enable_interrupts(np, 0);
+
+	return err;
+}
+
+static void niu_rxchan_intr(struct niu *np, struct rx_ring_info *rp,
+			    int ldn)
+{
+	struct rxdma_mailbox *mbox = rp->mbox;
+	u64 stat_write, stat = le64_to_cpup(&mbox->rx_dma_ctl_stat);
+
+	stat_write = (RX_DMA_CTL_STAT_RCRTHRES |
+		      RX_DMA_CTL_STAT_RCRTO);
+	nw64(RX_DMA_CTL_STAT(rp->rx_channel), stat_write);
+
+	netif_printk(np, intr, KERN_DEBUG, np->dev,
+		     "%s() stat[%llx]\n", __func__, (unsigned long long)stat);
+}
+
+static void niu_txchan_intr(struct niu *np, struct tx_ring_info *rp,
+			    int ldn)
+{
+	rp->tx_cs = nr64(TX_CS(rp->tx_channel));
+
+	netif_printk(np, intr, KERN_DEBUG, np->dev,
+		     "%s() cs[%llx]\n", __func__, (unsigned long long)rp->tx_cs);
+}
+
+static void __niu_fastpath_interrupt(struct niu *np, int ldg, u64 v0)
+{
+	struct niu_parent *parent = np->parent;
+	u32 rx_vec, tx_vec;
+	int i;
+
+	tx_vec = (v0 >> 32);
+	rx_vec = (v0 & 0xffffffff);
+
+	for (i = 0; i < np->num_rx_rings; i++) {
+		struct rx_ring_info *rp = &np->rx_rings[i];
+		int ldn = LDN_RXDMA(rp->rx_channel);
+
+		if (parent->ldg_map[ldn] != ldg)
+			continue;
+
+		nw64(LD_IM0(ldn), LD_IM0_MASK);
+		if (rx_vec & (1 << rp->rx_channel))
+			niu_rxchan_intr(np, rp, ldn);
+	}
+
+	for (i = 0; i < np->num_tx_rings; i++) {
+		struct tx_ring_info *rp = &np->tx_rings[i];
+		int ldn = LDN_TXDMA(rp->tx_channel);
+
+		if (parent->ldg_map[ldn] != ldg)
+			continue;
+
+		nw64(LD_IM0(ldn), LD_IM0_MASK);
+		if (tx_vec & (1 << rp->tx_channel))
+			niu_txchan_intr(np, rp, ldn);
+	}
+}
+
+static void niu_schedule_napi(struct niu *np, struct niu_ldg *lp,
+			      u64 v0, u64 v1, u64 v2)
+{
+	if (likely(napi_schedule_prep(&lp->napi))) {
+		lp->v0 = v0;
+		lp->v1 = v1;
+		lp->v2 = v2;
+		__niu_fastpath_interrupt(np, lp->ldg_num, v0);
+		__napi_schedule(&lp->napi);
+	}
+}
+
+static irqreturn_t niu_interrupt(int irq, void *dev_id)
+{
+	struct niu_ldg *lp = dev_id;
+	struct niu *np = lp->np;
+	int ldg = lp->ldg_num;
+	unsigned long flags;
+	u64 v0, v1, v2;
+
+	if (netif_msg_intr(np))
+		printk(KERN_DEBUG KBUILD_MODNAME ": " "%s() ldg[%p](%d)",
+		       __func__, lp, ldg);
+
+	spin_lock_irqsave(&np->lock, flags);
+
+	v0 = nr64(LDSV0(ldg));
+	v1 = nr64(LDSV1(ldg));
+	v2 = nr64(LDSV2(ldg));
+
+	if (netif_msg_intr(np))
+		pr_cont(" v0[%llx] v1[%llx] v2[%llx]\n",
+		       (unsigned long long) v0,
+		       (unsigned long long) v1,
+		       (unsigned long long) v2);
+
+	if (unlikely(!v0 && !v1 && !v2)) {
+		spin_unlock_irqrestore(&np->lock, flags);
+		return IRQ_NONE;
+	}
+
+	if (unlikely((v0 & ((u64)1 << LDN_MIF)) || v1 || v2)) {
+		int err = niu_slowpath_interrupt(np, lp, v0, v1, v2);
+		if (err)
+			goto out;
+	}
+	if (likely(v0 & ~((u64)1 << LDN_MIF)))
+		niu_schedule_napi(np, lp, v0, v1, v2);
+	else
+		niu_ldg_rearm(np, lp, 1);
+out:
+	spin_unlock_irqrestore(&np->lock, flags);
+
+	return IRQ_HANDLED;
+}
+
+static void niu_free_rx_ring_info(struct niu *np, struct rx_ring_info *rp)
+{
+	if (rp->mbox) {
+		np->ops->free_coherent(np->device,
+				       sizeof(struct rxdma_mailbox),
+				       rp->mbox, rp->mbox_dma);
+		rp->mbox = NULL;
+	}
+	if (rp->rcr) {
+		np->ops->free_coherent(np->device,
+				       MAX_RCR_RING_SIZE * sizeof(__le64),
+				       rp->rcr, rp->rcr_dma);
+		rp->rcr = NULL;
+		rp->rcr_table_size = 0;
+		rp->rcr_index = 0;
+	}
+	if (rp->rbr) {
+		niu_rbr_free(np, rp);
+
+		np->ops->free_coherent(np->device,
+				       MAX_RBR_RING_SIZE * sizeof(__le32),
+				       rp->rbr, rp->rbr_dma);
+		rp->rbr = NULL;
+		rp->rbr_table_size = 0;
+		rp->rbr_index = 0;
+	}
+	kfree(rp->rxhash);
+	rp->rxhash = NULL;
+}
+
+static void niu_free_tx_ring_info(struct niu *np, struct tx_ring_info *rp)
+{
+	if (rp->mbox) {
+		np->ops->free_coherent(np->device,
+				       sizeof(struct txdma_mailbox),
+				       rp->mbox, rp->mbox_dma);
+		rp->mbox = NULL;
+	}
+	if (rp->descr) {
+		int i;
+
+		for (i = 0; i < MAX_TX_RING_SIZE; i++) {
+			if (rp->tx_buffs[i].skb)
+				(void) release_tx_packet(np, rp, i);
+		}
+
+		np->ops->free_coherent(np->device,
+				       MAX_TX_RING_SIZE * sizeof(__le64),
+				       rp->descr, rp->descr_dma);
+		rp->descr = NULL;
+		rp->pending = 0;
+		rp->prod = 0;
+		rp->cons = 0;
+		rp->wrap_bit = 0;
+	}
+}
+
+static void niu_free_channels(struct niu *np)
+{
+	int i;
+
+	if (np->rx_rings) {
+		for (i = 0; i < np->num_rx_rings; i++) {
+			struct rx_ring_info *rp = &np->rx_rings[i];
+
+			niu_free_rx_ring_info(np, rp);
+		}
+		kfree(np->rx_rings);
+		np->rx_rings = NULL;
+		np->num_rx_rings = 0;
+	}
+
+	if (np->tx_rings) {
+		for (i = 0; i < np->num_tx_rings; i++) {
+			struct tx_ring_info *rp = &np->tx_rings[i];
+
+			niu_free_tx_ring_info(np, rp);
+		}
+		kfree(np->tx_rings);
+		np->tx_rings = NULL;
+		np->num_tx_rings = 0;
+	}
+}
+
+static int niu_alloc_rx_ring_info(struct niu *np,
+				  struct rx_ring_info *rp)
+{
+	BUILD_BUG_ON(sizeof(struct rxdma_mailbox) != 64);
+
+	rp->rxhash = kzalloc_objs(struct page *, MAX_RBR_RING_SIZE);
+	if (!rp->rxhash)
+		return -ENOMEM;
+
+	rp->mbox = np->ops->alloc_coherent(np->device,
+					   sizeof(struct rxdma_mailbox),
+					   &rp->mbox_dma, GFP_KERNEL);
+	if (!rp->mbox)
+		return -ENOMEM;
+	if ((unsigned long)rp->mbox & (64UL - 1)) {
+		netdev_err(np->dev, "Coherent alloc gives misaligned RXDMA mailbox %p\n",
+			   rp->mbox);
+		return -EINVAL;
+	}
+
+	rp->rcr = np->ops->alloc_coherent(np->device,
+					  MAX_RCR_RING_SIZE * sizeof(__le64),
+					  &rp->rcr_dma, GFP_KERNEL);
+	if (!rp->rcr)
+		return -ENOMEM;
+	if ((unsigned long)rp->rcr & (64UL - 1)) {
+		netdev_err(np->dev, "Coherent alloc gives misaligned RXDMA RCR table %p\n",
+			   rp->rcr);
+		return -EINVAL;
+	}
+	rp->rcr_table_size = MAX_RCR_RING_SIZE;
+	rp->rcr_index = 0;
+
+	rp->rbr = np->ops->alloc_coherent(np->device,
+					  MAX_RBR_RING_SIZE * sizeof(__le32),
+					  &rp->rbr_dma, GFP_KERNEL);
+	if (!rp->rbr)
+		return -ENOMEM;
+	if ((unsigned long)rp->rbr & (64UL - 1)) {
+		netdev_err(np->dev, "Coherent alloc gives misaligned RXDMA RBR table %p\n",
+			   rp->rbr);
+		return -EINVAL;
+	}
+	rp->rbr_table_size = MAX_RBR_RING_SIZE;
+	rp->rbr_index = 0;
+	rp->rbr_pending = 0;
+
+	return 0;
+}
+
+static void niu_set_max_burst(struct niu *np, struct tx_ring_info *rp)
+{
+	int mtu = np->dev->mtu;
+
+	/* These values are recommended by the HW designers for fair
+	 * utilization of DRR amongst the rings.
+	 */
+	rp->max_burst = mtu + 32;
+	if (rp->max_burst > 4096)
+		rp->max_burst = 4096;
+}
+
+static int niu_alloc_tx_ring_info(struct niu *np,
+				  struct tx_ring_info *rp)
+{
+	BUILD_BUG_ON(sizeof(struct txdma_mailbox) != 64);
+
+	rp->mbox = np->ops->alloc_coherent(np->device,
+					   sizeof(struct txdma_mailbox),
+					   &rp->mbox_dma, GFP_KERNEL);
+	if (!rp->mbox)
+		return -ENOMEM;
+	if ((unsigned long)rp->mbox & (64UL - 1)) {
+		netdev_err(np->dev, "Coherent alloc gives misaligned TXDMA mailbox %p\n",
+			   rp->mbox);
+		return -EINVAL;
+	}
+
+	rp->descr = np->ops->alloc_coherent(np->device,
+					    MAX_TX_RING_SIZE * sizeof(__le64),
+					    &rp->descr_dma, GFP_KERNEL);
+	if (!rp->descr)
+		return -ENOMEM;
+	if ((unsigned long)rp->descr & (64UL - 1)) {
+		netdev_err(np->dev, "Coherent alloc gives misaligned TXDMA descr table %p\n",
+			   rp->descr);
+		return -EINVAL;
+	}
+
+	rp->pending = MAX_TX_RING_SIZE;
+	rp->prod = 0;
+	rp->cons = 0;
+	rp->wrap_bit = 0;
+
+	/* XXX make these configurable... XXX */
+	rp->mark_freq = rp->pending / 4;
+
+	niu_set_max_burst(np, rp);
+
+	return 0;
+}
+
+static void niu_size_rbr(struct niu *np, struct rx_ring_info *rp)
+{
+	u16 bss;
+
+	bss = min(PAGE_SHIFT, 15);
+
+	rp->rbr_block_size = 1 << bss;
+	rp->rbr_blocks_per_page = 1 << (PAGE_SHIFT-bss);
+
+	rp->rbr_sizes[0] = 256;
+	rp->rbr_sizes[1] = 1024;
+	if (np->dev->mtu > ETH_DATA_LEN) {
+		switch (PAGE_SIZE) {
+		case 4 * 1024:
+			rp->rbr_sizes[2] = 4096;
+			break;
+
+		default:
+			rp->rbr_sizes[2] = 8192;
+			break;
+		}
+	} else {
+		rp->rbr_sizes[2] = 2048;
+	}
+	rp->rbr_sizes[3] = rp->rbr_block_size;
+}
+
+static int niu_alloc_channels(struct niu *np)
+{
+	struct niu_parent *parent = np->parent;
+	int first_rx_channel, first_tx_channel;
+	int num_rx_rings, num_tx_rings;
+	struct rx_ring_info *rx_rings;
+	struct tx_ring_info *tx_rings;
+	int i, port, err;
+
+	port = np->port;
+	first_rx_channel = first_tx_channel = 0;
+	for (i = 0; i < port; i++) {
+		first_rx_channel += parent->rxchan_per_port[i];
+		first_tx_channel += parent->txchan_per_port[i];
+	}
+
+	num_rx_rings = parent->rxchan_per_port[port];
+	num_tx_rings = parent->txchan_per_port[port];
+
+	rx_rings = kzalloc_objs(struct rx_ring_info, num_rx_rings);
+	err = -ENOMEM;
+	if (!rx_rings)
+		goto out_err;
+
+	np->num_rx_rings = num_rx_rings;
+	smp_wmb();
+	np->rx_rings = rx_rings;
+
+	netif_set_real_num_rx_queues(np->dev, num_rx_rings);
+
+	for (i = 0; i < np->num_rx_rings; i++) {
+		struct rx_ring_info *rp = &np->rx_rings[i];
+
+		rp->np = np;
+		rp->rx_channel = first_rx_channel + i;
+
+		err = niu_alloc_rx_ring_info(np, rp);
+		if (err)
+			goto out_err;
+
+		niu_size_rbr(np, rp);
+
+		/* XXX better defaults, configurable, etc... XXX */
+		rp->nonsyn_window = 64;
+		rp->nonsyn_threshold = rp->rcr_table_size - 64;
+		rp->syn_window = 64;
+		rp->syn_threshold = rp->rcr_table_size - 64;
+		rp->rcr_pkt_threshold = 16;
+		rp->rcr_timeout = 8;
+		rp->rbr_kick_thresh = RBR_REFILL_MIN;
+		if (rp->rbr_kick_thresh < rp->rbr_blocks_per_page)
+			rp->rbr_kick_thresh = rp->rbr_blocks_per_page;
+
+		err = niu_rbr_fill(np, rp, GFP_KERNEL);
+		if (err)
+			goto out_err;
+	}
+
+	tx_rings = kzalloc_objs(struct tx_ring_info, num_tx_rings);
+	err = -ENOMEM;
+	if (!tx_rings)
+		goto out_err;
+
+	np->num_tx_rings = num_tx_rings;
+	smp_wmb();
+	np->tx_rings = tx_rings;
+
+	netif_set_real_num_tx_queues(np->dev, num_tx_rings);
+
+	for (i = 0; i < np->num_tx_rings; i++) {
+		struct tx_ring_info *rp = &np->tx_rings[i];
+
+		rp->np = np;
+		rp->tx_channel = first_tx_channel + i;
+
+		err = niu_alloc_tx_ring_info(np, rp);
+		if (err)
+			goto out_err;
+	}
+
+	return 0;
+
+out_err:
+	niu_free_channels(np);
+	return err;
+}
+
+static int niu_tx_cs_sng_poll(struct niu *np, int channel)
+{
+	int limit = 1000;
+
+	while (--limit > 0) {
+		u64 val = nr64(TX_CS(channel));
+		if (val & TX_CS_SNG_STATE)
+			return 0;
+	}
+	return -ENODEV;
+}
+
+static int niu_tx_channel_stop(struct niu *np, int channel)
+{
+	u64 val = nr64(TX_CS(channel));
+
+	val |= TX_CS_STOP_N_GO;
+	nw64(TX_CS(channel), val);
+
+	return niu_tx_cs_sng_poll(np, channel);
+}
+
+static int niu_tx_cs_reset_poll(struct niu *np, int channel)
+{
+	int limit = 1000;
+
+	while (--limit > 0) {
+		u64 val = nr64(TX_CS(channel));
+		if (!(val & TX_CS_RST))
+			return 0;
+	}
+	return -ENODEV;
+}
+
+static int niu_tx_channel_reset(struct niu *np, int channel)
+{
+	u64 val = nr64(TX_CS(channel));
+	int err;
+
+	val |= TX_CS_RST;
+	nw64(TX_CS(channel), val);
+
+	err = niu_tx_cs_reset_poll(np, channel);
+	if (!err)
+		nw64(TX_RING_KICK(channel), 0);
+
+	return err;
+}
+
+static int niu_tx_channel_lpage_init(struct niu *np, int channel)
+{
+	u64 val;
+
+	nw64(TX_LOG_MASK1(channel), 0);
+	nw64(TX_LOG_VAL1(channel), 0);
+	nw64(TX_LOG_MASK2(channel), 0);
+	nw64(TX_LOG_VAL2(channel), 0);
+	nw64(TX_LOG_PAGE_RELO1(channel), 0);
+	nw64(TX_LOG_PAGE_RELO2(channel), 0);
+	nw64(TX_LOG_PAGE_HDL(channel), 0);
+
+	val  = (u64)np->port << TX_LOG_PAGE_VLD_FUNC_SHIFT;
+	val |= (TX_LOG_PAGE_VLD_PAGE0 | TX_LOG_PAGE_VLD_PAGE1);
+	nw64(TX_LOG_PAGE_VLD(channel), val);
+
+	/* XXX TXDMA 32bit mode? XXX */
+
+	return 0;
+}
+
+static void niu_txc_enable_port(struct niu *np, int on)
+{
+	unsigned long flags;
+	u64 val, mask;
+
+	niu_lock_parent(np, flags);
+	val = nr64(TXC_CONTROL);
+	mask = (u64)1 << np->port;
+	if (on) {
+		val |= TXC_CONTROL_ENABLE | mask;
+	} else {
+		val &= ~mask;
+		if ((val & ~TXC_CONTROL_ENABLE) == 0)
+			val &= ~TXC_CONTROL_ENABLE;
+	}
+	nw64(TXC_CONTROL, val);
+	niu_unlock_parent(np, flags);
+}
+
+static void niu_txc_set_imask(struct niu *np, u64 imask)
+{
+	unsigned long flags;
+	u64 val;
+
+	niu_lock_parent(np, flags);
+	val = nr64(TXC_INT_MASK);
+	val &= ~TXC_INT_MASK_VAL(np->port);
+	val |= (imask << TXC_INT_MASK_VAL_SHIFT(np->port));
+	niu_unlock_parent(np, flags);
+}
+
+static void niu_txc_port_dma_enable(struct niu *np, int on)
+{
+	u64 val = 0;
+
+	if (on) {
+		int i;
+
+		for (i = 0; i < np->num_tx_rings; i++)
+			val |= (1 << np->tx_rings[i].tx_channel);
+	}
+	nw64(TXC_PORT_DMA(np->port), val);
+}
+
+static int niu_init_one_tx_channel(struct niu *np, struct tx_ring_info *rp)
+{
+	int err, channel = rp->tx_channel;
+	u64 val, ring_len;
+
+	err = niu_tx_channel_stop(np, channel);
+	if (err)
+		return err;
+
+	err = niu_tx_channel_reset(np, channel);
+	if (err)
+		return err;
+
+	err = niu_tx_channel_lpage_init(np, channel);
+	if (err)
+		return err;
+
+	nw64(TXC_DMA_MAX(channel), rp->max_burst);
+	nw64(TX_ENT_MSK(channel), 0);
+
+	if (rp->descr_dma & ~(TX_RNG_CFIG_STADDR_BASE |
+			      TX_RNG_CFIG_STADDR)) {
+		netdev_err(np->dev, "TX ring channel %d DMA addr (%llx) is not aligned\n",
+			   channel, (unsigned long long)rp->descr_dma);
+		return -EINVAL;
+	}
+
+	/* The length field in TX_RNG_CFIG is measured in 64-byte
+	 * blocks.  rp->pending is the number of TX descriptors in
+	 * our ring, 8 bytes each, thus we divide by 8 bytes more
+	 * to get the proper value the chip wants.
+	 */
+	ring_len = (rp->pending / 8);
+
+	val = ((ring_len << TX_RNG_CFIG_LEN_SHIFT) |
+	       rp->descr_dma);
+	nw64(TX_RNG_CFIG(channel), val);
+
+	if (((rp->mbox_dma >> 32) & ~TXDMA_MBH_MBADDR) ||
+	    ((u32)rp->mbox_dma & ~TXDMA_MBL_MBADDR)) {
+		netdev_err(np->dev, "TX ring channel %d MBOX addr (%llx) has invalid bits\n",
+			    channel, (unsigned long long)rp->mbox_dma);
+		return -EINVAL;
+	}
+	nw64(TXDMA_MBH(channel), rp->mbox_dma >> 32);
+	nw64(TXDMA_MBL(channel), rp->mbox_dma & TXDMA_MBL_MBADDR);
+
+	nw64(TX_CS(channel), 0);
+
+	rp->last_pkt_cnt = 0;
+
+	return 0;
+}
+
+static void niu_init_rdc_groups(struct niu *np)
+{
+	struct niu_rdc_tables *tp = &np->parent->rdc_group_cfg[np->port];
+	int i, first_table_num = tp->first_table_num;
+
+	for (i = 0; i < tp->num_tables; i++) {
+		struct rdc_table *tbl = &tp->tables[i];
+		int this_table = first_table_num + i;
+		int slot;
+
+		for (slot = 0; slot < NIU_RDC_TABLE_SLOTS; slot++)
+			nw64(RDC_TBL(this_table, slot),
+			     tbl->rxdma_channel[slot]);
+	}
+
+	nw64(DEF_RDC(np->port), np->parent->rdc_default[np->port]);
+}
+
+static void niu_init_drr_weight(struct niu *np)
+{
+	int type = phy_decode(np->parent->port_phy, np->port);
+	u64 val;
+
+	switch (type) {
+	case PORT_TYPE_10G:
+		val = PT_DRR_WEIGHT_DEFAULT_10G;
+		break;
+
+	case PORT_TYPE_1G:
+	default:
+		val = PT_DRR_WEIGHT_DEFAULT_1G;
+		break;
+	}
+	nw64(PT_DRR_WT(np->port), val);
+}
+
+static int niu_init_hostinfo(struct niu *np)
+{
+	struct niu_parent *parent = np->parent;
+	struct niu_rdc_tables *tp = &parent->rdc_group_cfg[np->port];
+	int i, err, num_alt = niu_num_alt_addr(np);
+	int first_rdc_table = tp->first_table_num;
+
+	err = niu_set_primary_mac_rdc_table(np, first_rdc_table, 1);
+	if (err)
+		return err;
+
+	err = niu_set_multicast_mac_rdc_table(np, first_rdc_table, 1);
+	if (err)
+		return err;
+
+	for (i = 0; i < num_alt; i++) {
+		err = niu_set_alt_mac_rdc_table(np, i, first_rdc_table, 1);
+		if (err)
+			return err;
+	}
+
+	return 0;
+}
+
+static int niu_rx_channel_reset(struct niu *np, int channel)
+{
+	return niu_set_and_wait_clear(np, RXDMA_CFIG1(channel),
+				      RXDMA_CFIG1_RST, 1000, 10,
+				      "RXDMA_CFIG1");
+}
+
+static int niu_rx_channel_lpage_init(struct niu *np, int channel)
+{
+	u64 val;
+
+	nw64(RX_LOG_MASK1(channel), 0);
+	nw64(RX_LOG_VAL1(channel), 0);
+	nw64(RX_LOG_MASK2(channel), 0);
+	nw64(RX_LOG_VAL2(channel), 0);
+	nw64(RX_LOG_PAGE_RELO1(channel), 0);
+	nw64(RX_LOG_PAGE_RELO2(channel), 0);
+	nw64(RX_LOG_PAGE_HDL(channel), 0);
+
+	val  = (u64)np->port << RX_LOG_PAGE_VLD_FUNC_SHIFT;
+	val |= (RX_LOG_PAGE_VLD_PAGE0 | RX_LOG_PAGE_VLD_PAGE1);
+	nw64(RX_LOG_PAGE_VLD(channel), val);
+
+	return 0;
+}
+
+static void niu_rx_channel_wred_init(struct niu *np, struct rx_ring_info *rp)
+{
+	u64 val;
+
+	val = (((u64)rp->nonsyn_window << RDC_RED_PARA_WIN_SHIFT) |
+	       ((u64)rp->nonsyn_threshold << RDC_RED_PARA_THRE_SHIFT) |
+	       ((u64)rp->syn_window << RDC_RED_PARA_WIN_SYN_SHIFT) |
+	       ((u64)rp->syn_threshold << RDC_RED_PARA_THRE_SYN_SHIFT));
+	nw64(RDC_RED_PARA(rp->rx_channel), val);
+}
+
+static int niu_compute_rbr_cfig_b(struct rx_ring_info *rp, u64 *ret)
+{
+	u64 val = 0;
+
+	*ret = 0;
+	switch (rp->rbr_block_size) {
+	case 4 * 1024:
+		val |= (RBR_BLKSIZE_4K << RBR_CFIG_B_BLKSIZE_SHIFT);
+		break;
+	case 8 * 1024:
+		val |= (RBR_BLKSIZE_8K << RBR_CFIG_B_BLKSIZE_SHIFT);
+		break;
+	case 16 * 1024:
+		val |= (RBR_BLKSIZE_16K << RBR_CFIG_B_BLKSIZE_SHIFT);
+		break;
+	case 32 * 1024:
+		val |= (RBR_BLKSIZE_32K << RBR_CFIG_B_BLKSIZE_SHIFT);
+		break;
+	default:
+		return -EINVAL;
+	}
+	val |= RBR_CFIG_B_VLD2;
+	switch (rp->rbr_sizes[2]) {
+	case 2 * 1024:
+		val |= (RBR_BUFSZ2_2K << RBR_CFIG_B_BUFSZ2_SHIFT);
+		break;
+	case 4 * 1024:
+		val |= (RBR_BUFSZ2_4K << RBR_CFIG_B_BUFSZ2_SHIFT);
+		break;
+	case 8 * 1024:
+		val |= (RBR_BUFSZ2_8K << RBR_CFIG_B_BUFSZ2_SHIFT);
+		break;
+	case 16 * 1024:
+		val |= (RBR_BUFSZ2_16K << RBR_CFIG_B_BUFSZ2_SHIFT);
+		break;
+
+	default:
+		return -EINVAL;
+	}
+	val |= RBR_CFIG_B_VLD1;
+	switch (rp->rbr_sizes[1]) {
+	case 1 * 1024:
+		val |= (RBR_BUFSZ1_1K << RBR_CFIG_B_BUFSZ1_SHIFT);
+		break;
+	case 2 * 1024:
+		val |= (RBR_BUFSZ1_2K << RBR_CFIG_B_BUFSZ1_SHIFT);
+		break;
+	case 4 * 1024:
+		val |= (RBR_BUFSZ1_4K << RBR_CFIG_B_BUFSZ1_SHIFT);
+		break;
+	case 8 * 1024:
+		val |= (RBR_BUFSZ1_8K << RBR_CFIG_B_BUFSZ1_SHIFT);
+		break;
+
+	default:
+		return -EINVAL;
+	}
+	val |= RBR_CFIG_B_VLD0;
+	switch (rp->rbr_sizes[0]) {
+	case 256:
+		val |= (RBR_BUFSZ0_256 << RBR_CFIG_B_BUFSZ0_SHIFT);
+		break;
+	case 512:
+		val |= (RBR_BUFSZ0_512 << RBR_CFIG_B_BUFSZ0_SHIFT);
+		break;
+	case 1 * 1024:
+		val |= (RBR_BUFSZ0_1K << RBR_CFIG_B_BUFSZ0_SHIFT);
+		break;
+	case 2 * 1024:
+		val |= (RBR_BUFSZ0_2K << RBR_CFIG_B_BUFSZ0_SHIFT);
+		break;
+
+	default:
+		return -EINVAL;
+	}
+
+	*ret = val;
+	return 0;
+}
+
+static int niu_enable_rx_channel(struct niu *np, int channel, int on)
+{
+	u64 val = nr64(RXDMA_CFIG1(channel));
+	int limit;
+
+	if (on)
+		val |= RXDMA_CFIG1_EN;
+	else
+		val &= ~RXDMA_CFIG1_EN;
+	nw64(RXDMA_CFIG1(channel), val);
+
+	limit = 1000;
+	while (--limit > 0) {
+		if (nr64(RXDMA_CFIG1(channel)) & RXDMA_CFIG1_QST)
+			break;
+		udelay(10);
+	}
+	if (limit <= 0)
+		return -ENODEV;
+	return 0;
+}
+
+static int niu_init_one_rx_channel(struct niu *np, struct rx_ring_info *rp)
+{
+	int err, channel = rp->rx_channel;
+	u64 val;
+
+	err = niu_rx_channel_reset(np, channel);
+	if (err)
+		return err;
+
+	err = niu_rx_channel_lpage_init(np, channel);
+	if (err)
+		return err;
+
+	niu_rx_channel_wred_init(np, rp);
+
+	nw64(RX_DMA_ENT_MSK(channel), RX_DMA_ENT_MSK_RBR_EMPTY);
+	nw64(RX_DMA_CTL_STAT(channel),
+	     (RX_DMA_CTL_STAT_MEX |
+	      RX_DMA_CTL_STAT_RCRTHRES |
+	      RX_DMA_CTL_STAT_RCRTO |
+	      RX_DMA_CTL_STAT_RBR_EMPTY));
+	nw64(RXDMA_CFIG1(channel), rp->mbox_dma >> 32);
+	nw64(RXDMA_CFIG2(channel),
+	     ((rp->mbox_dma & RXDMA_CFIG2_MBADDR_L) |
+	      RXDMA_CFIG2_FULL_HDR));
+	nw64(RBR_CFIG_A(channel),
+	     ((u64)rp->rbr_table_size << RBR_CFIG_A_LEN_SHIFT) |
+	     (rp->rbr_dma & (RBR_CFIG_A_STADDR_BASE | RBR_CFIG_A_STADDR)));
+	err = niu_compute_rbr_cfig_b(rp, &val);
+	if (err)
+		return err;
+	nw64(RBR_CFIG_B(channel), val);
+	nw64(RCRCFIG_A(channel),
+	     ((u64)rp->rcr_table_size << RCRCFIG_A_LEN_SHIFT) |
+	     (rp->rcr_dma & (RCRCFIG_A_STADDR_BASE | RCRCFIG_A_STADDR)));
+	nw64(RCRCFIG_B(channel),
+	     ((u64)rp->rcr_pkt_threshold << RCRCFIG_B_PTHRES_SHIFT) |
+	     RCRCFIG_B_ENTOUT |
+	     ((u64)rp->rcr_timeout << RCRCFIG_B_TIMEOUT_SHIFT));
+
+	err = niu_enable_rx_channel(np, channel, 1);
+	if (err)
+		return err;
+
+	nw64(RBR_KICK(channel), rp->rbr_index);
+
+	val = nr64(RX_DMA_CTL_STAT(channel));
+	val |= RX_DMA_CTL_STAT_RBR_EMPTY;
+	nw64(RX_DMA_CTL_STAT(channel), val);
+
+	return 0;
+}
+
+static int niu_init_rx_channels(struct niu *np)
+{
+	unsigned long flags;
+	u64 seed = jiffies_64;
+	int err, i;
+
+	niu_lock_parent(np, flags);
+	nw64(RX_DMA_CK_DIV, np->parent->rxdma_clock_divider);
+	nw64(RED_RAN_INIT, RED_RAN_INIT_OPMODE | (seed & RED_RAN_INIT_VAL));
+	niu_unlock_parent(np, flags);
+
+	/* XXX RXDMA 32bit mode? XXX */
+
+	niu_init_rdc_groups(np);
+	niu_init_drr_weight(np);
+
+	err = niu_init_hostinfo(np);
+	if (err)
+		return err;
+
+	for (i = 0; i < np->num_rx_rings; i++) {
+		struct rx_ring_info *rp = &np->rx_rings[i];
+
+		err = niu_init_one_rx_channel(np, rp);
+		if (err)
+			return err;
+	}
+
+	return 0;
+}
+
+static int niu_set_ip_frag_rule(struct niu *np)
+{
+	struct niu_parent *parent = np->parent;
+	struct niu_classifier *cp = &np->clas;
+	struct niu_tcam_entry *tp;
+	int index, err;
+
+	index = cp->tcam_top;
+	tp = &parent->tcam[index];
+
+	/* Note that the noport bit is the same in both ipv4 and
+	 * ipv6 format TCAM entries.
+	 */
+	memset(tp, 0, sizeof(*tp));
+	tp->key[1] = TCAM_V4KEY1_NOPORT;
+	tp->key_mask[1] = TCAM_V4KEY1_NOPORT;
+	tp->assoc_data = (TCAM_ASSOCDATA_TRES_USE_OFFSET |
+			  ((u64)0 << TCAM_ASSOCDATA_OFFSET_SHIFT));
+	err = tcam_write(np, index, tp->key, tp->key_mask);
+	if (err)
+		return err;
+	err = tcam_assoc_write(np, index, tp->assoc_data);
+	if (err)
+		return err;
+	tp->valid = 1;
+	cp->tcam_valid_entries++;
+
+	return 0;
+}
+
+static int niu_init_classifier_hw(struct niu *np)
+{
+	struct niu_parent *parent = np->parent;
+	struct niu_classifier *cp = &np->clas;
+	int i, err;
+
+	nw64(H1POLY, cp->h1_init);
+	nw64(H2POLY, cp->h2_init);
+
+	err = niu_init_hostinfo(np);
+	if (err)
+		return err;
+
+	for (i = 0; i < ENET_VLAN_TBL_NUM_ENTRIES; i++) {
+		struct niu_vlan_rdc *vp = &cp->vlan_mappings[i];
+
+		vlan_tbl_write(np, i, np->port,
+			       vp->vlan_pref, vp->rdc_num);
+	}
+
+	for (i = 0; i < cp->num_alt_mac_mappings; i++) {
+		struct niu_altmac_rdc *ap = &cp->alt_mac_mappings[i];
+
+		err = niu_set_alt_mac_rdc_table(np, ap->alt_mac_num,
+						ap->rdc_num, ap->mac_pref);
+		if (err)
+			return err;
+	}
+
+	for (i = CLASS_CODE_USER_PROG1; i <= CLASS_CODE_SCTP_IPV6; i++) {
+		int index = i - CLASS_CODE_USER_PROG1;
+
+		err = niu_set_tcam_key(np, i, parent->tcam_key[index]);
+		if (err)
+			return err;
+		err = niu_set_flow_key(np, i, parent->flow_key[index]);
+		if (err)
+			return err;
+	}
+
+	err = niu_set_ip_frag_rule(np);
+	if (err)
+		return err;
+
+	tcam_enable(np, 1);
+
+	return 0;
+}
+
+static int niu_zcp_write(struct niu *np, int index, u64 *data)
+{
+	nw64(ZCP_RAM_DATA0, data[0]);
+	nw64(ZCP_RAM_DATA1, data[1]);
+	nw64(ZCP_RAM_DATA2, data[2]);
+	nw64(ZCP_RAM_DATA3, data[3]);
+	nw64(ZCP_RAM_DATA4, data[4]);
+	nw64(ZCP_RAM_BE, ZCP_RAM_BE_VAL);
+	nw64(ZCP_RAM_ACC,
+	     (ZCP_RAM_ACC_WRITE |
+	      (0 << ZCP_RAM_ACC_ZFCID_SHIFT) |
+	      (ZCP_RAM_SEL_CFIFO(np->port) << ZCP_RAM_ACC_RAM_SEL_SHIFT)));
+
+	return niu_wait_bits_clear(np, ZCP_RAM_ACC, ZCP_RAM_ACC_BUSY,
+				   1000, 100);
+}
+
+static int niu_zcp_read(struct niu *np, int index, u64 *data)
+{
+	int err;
+
+	err = niu_wait_bits_clear(np, ZCP_RAM_ACC, ZCP_RAM_ACC_BUSY,
+				  1000, 100);
+	if (err) {
+		netdev_err(np->dev, "ZCP read busy won't clear, ZCP_RAM_ACC[%llx]\n",
+			   (unsigned long long)nr64(ZCP_RAM_ACC));
+		return err;
+	}
+
+	nw64(ZCP_RAM_ACC,
+	     (ZCP_RAM_ACC_READ |
+	      (0 << ZCP_RAM_ACC_ZFCID_SHIFT) |
+	      (ZCP_RAM_SEL_CFIFO(np->port) << ZCP_RAM_ACC_RAM_SEL_SHIFT)));
+
+	err = niu_wait_bits_clear(np, ZCP_RAM_ACC, ZCP_RAM_ACC_BUSY,
+				  1000, 100);
+	if (err) {
+		netdev_err(np->dev, "ZCP read busy2 won't clear, ZCP_RAM_ACC[%llx]\n",
+			   (unsigned long long)nr64(ZCP_RAM_ACC));
+		return err;
+	}
+
+	data[0] = nr64(ZCP_RAM_DATA0);
+	data[1] = nr64(ZCP_RAM_DATA1);
+	data[2] = nr64(ZCP_RAM_DATA2);
+	data[3] = nr64(ZCP_RAM_DATA3);
+	data[4] = nr64(ZCP_RAM_DATA4);
+
+	return 0;
+}
+
+static void niu_zcp_cfifo_reset(struct niu *np)
+{
+	u64 val = nr64(RESET_CFIFO);
+
+	val |= RESET_CFIFO_RST(np->port);
+	nw64(RESET_CFIFO, val);
+	udelay(10);
+
+	val &= ~RESET_CFIFO_RST(np->port);
+	nw64(RESET_CFIFO, val);
+}
+
+static int niu_init_zcp(struct niu *np)
+{
+	u64 data[5], rbuf[5];
+	int i, max, err;
+
+	if (np->parent->plat_type != PLAT_TYPE_NIU) {
+		if (np->port == 0 || np->port == 1)
+			max = ATLAS_P0_P1_CFIFO_ENTRIES;
+		else
+			max = ATLAS_P2_P3_CFIFO_ENTRIES;
+	} else
+		max = NIU_CFIFO_ENTRIES;
+
+	data[0] = 0;
+	data[1] = 0;
+	data[2] = 0;
+	data[3] = 0;
+	data[4] = 0;
+
+	for (i = 0; i < max; i++) {
+		err = niu_zcp_write(np, i, data);
+		if (err)
+			return err;
+		err = niu_zcp_read(np, i, rbuf);
+		if (err)
+			return err;
+	}
+
+	niu_zcp_cfifo_reset(np);
+	nw64(CFIFO_ECC(np->port), 0);
+	nw64(ZCP_INT_STAT, ZCP_INT_STAT_ALL);
+	(void) nr64(ZCP_INT_STAT);
+	nw64(ZCP_INT_MASK, ZCP_INT_MASK_ALL);
+
+	return 0;
+}
+
+static void niu_ipp_write(struct niu *np, int index, u64 *data)
+{
+	u64 val = nr64_ipp(IPP_CFIG);
+
+	nw64_ipp(IPP_CFIG, val | IPP_CFIG_DFIFO_PIO_W);
+	nw64_ipp(IPP_DFIFO_WR_PTR, index);
+	nw64_ipp(IPP_DFIFO_WR0, data[0]);
+	nw64_ipp(IPP_DFIFO_WR1, data[1]);
+	nw64_ipp(IPP_DFIFO_WR2, data[2]);
+	nw64_ipp(IPP_DFIFO_WR3, data[3]);
+	nw64_ipp(IPP_DFIFO_WR4, data[4]);
+	nw64_ipp(IPP_CFIG, val & ~IPP_CFIG_DFIFO_PIO_W);
+}
+
+static void niu_ipp_read(struct niu *np, int index, u64 *data)
+{
+	nw64_ipp(IPP_DFIFO_RD_PTR, index);
+	data[0] = nr64_ipp(IPP_DFIFO_RD0);
+	data[1] = nr64_ipp(IPP_DFIFO_RD1);
+	data[2] = nr64_ipp(IPP_DFIFO_RD2);
+	data[3] = nr64_ipp(IPP_DFIFO_RD3);
+	data[4] = nr64_ipp(IPP_DFIFO_RD4);
+}
+
+static int niu_ipp_reset(struct niu *np)
+{
+	return niu_set_and_wait_clear_ipp(np, IPP_CFIG, IPP_CFIG_SOFT_RST,
+					  1000, 100, "IPP_CFIG");
+}
+
+static int niu_init_ipp(struct niu *np)
+{
+	u64 data[5], rbuf[5], val;
+	int i, max, err;
+
+	if (np->parent->plat_type != PLAT_TYPE_NIU) {
+		if (np->port == 0 || np->port == 1)
+			max = ATLAS_P0_P1_DFIFO_ENTRIES;
+		else
+			max = ATLAS_P2_P3_DFIFO_ENTRIES;
+	} else
+		max = NIU_DFIFO_ENTRIES;
+
+	data[0] = 0;
+	data[1] = 0;
+	data[2] = 0;
+	data[3] = 0;
+	data[4] = 0;
+
+	for (i = 0; i < max; i++) {
+		niu_ipp_write(np, i, data);
+		niu_ipp_read(np, i, rbuf);
+	}
+
+	(void) nr64_ipp(IPP_INT_STAT);
+	(void) nr64_ipp(IPP_INT_STAT);
+
+	err = niu_ipp_reset(np);
+	if (err)
+		return err;
+
+	(void) nr64_ipp(IPP_PKT_DIS);
+	(void) nr64_ipp(IPP_BAD_CS_CNT);
+	(void) nr64_ipp(IPP_ECC);
+
+	(void) nr64_ipp(IPP_INT_STAT);
+
+	nw64_ipp(IPP_MSK, ~IPP_MSK_ALL);
+
+	val = nr64_ipp(IPP_CFIG);
+	val &= ~IPP_CFIG_IP_MAX_PKT;
+	val |= (IPP_CFIG_IPP_ENABLE |
+		IPP_CFIG_DFIFO_ECC_EN |
+		IPP_CFIG_DROP_BAD_CRC |
+		IPP_CFIG_CKSUM_EN |
+		(0x1ffff << IPP_CFIG_IP_MAX_PKT_SHIFT));
+	nw64_ipp(IPP_CFIG, val);
+
+	return 0;
+}
+
+static void niu_handle_led(struct niu *np, int status)
+{
+	u64 val;
+	val = nr64_mac(XMAC_CONFIG);
+
+	if ((np->flags & NIU_FLAGS_10G) != 0 &&
+	    (np->flags & NIU_FLAGS_FIBER) != 0) {
+		if (status) {
+			val |= XMAC_CONFIG_LED_POLARITY;
+			val &= ~XMAC_CONFIG_FORCE_LED_ON;
+		} else {
+			val |= XMAC_CONFIG_FORCE_LED_ON;
+			val &= ~XMAC_CONFIG_LED_POLARITY;
+		}
+	}
+
+	nw64_mac(XMAC_CONFIG, val);
+}
+
+static void niu_init_xif_xmac(struct niu *np)
+{
+	struct niu_link_config *lp = &np->link_config;
+	u64 val;
+
+	if (np->flags & NIU_FLAGS_XCVR_SERDES) {
+		val = nr64(MIF_CONFIG);
+		val |= MIF_CONFIG_ATCA_GE;
+		nw64(MIF_CONFIG, val);
+	}
+
+	val = nr64_mac(XMAC_CONFIG);
+	val &= ~XMAC_CONFIG_SEL_POR_CLK_SRC;
+
+	val |= XMAC_CONFIG_TX_OUTPUT_EN;
+
+	if (lp->loopback_mode == LOOPBACK_MAC) {
+		val &= ~XMAC_CONFIG_SEL_POR_CLK_SRC;
+		val |= XMAC_CONFIG_LOOPBACK;
+	} else {
+		val &= ~XMAC_CONFIG_LOOPBACK;
+	}
+
+	if (np->flags & NIU_FLAGS_10G) {
+		val &= ~XMAC_CONFIG_LFS_DISABLE;
+	} else {
+		val |= XMAC_CONFIG_LFS_DISABLE;
+		if (!(np->flags & NIU_FLAGS_FIBER) &&
+		    !(np->flags & NIU_FLAGS_XCVR_SERDES))
+			val |= XMAC_CONFIG_1G_PCS_BYPASS;
+		else
+			val &= ~XMAC_CONFIG_1G_PCS_BYPASS;
+	}
+
+	val &= ~XMAC_CONFIG_10G_XPCS_BYPASS;
+
+	if (lp->active_speed == SPEED_100)
+		val |= XMAC_CONFIG_SEL_CLK_25MHZ;
+	else
+		val &= ~XMAC_CONFIG_SEL_CLK_25MHZ;
+
+	nw64_mac(XMAC_CONFIG, val);
+
+	val = nr64_mac(XMAC_CONFIG);
+	val &= ~XMAC_CONFIG_MODE_MASK;
+	if (np->flags & NIU_FLAGS_10G) {
+		val |= XMAC_CONFIG_MODE_XGMII;
+	} else {
+		if (lp->active_speed == SPEED_1000)
+			val |= XMAC_CONFIG_MODE_GMII;
+		else
+			val |= XMAC_CONFIG_MODE_MII;
+	}
+
+	nw64_mac(XMAC_CONFIG, val);
+}
+
+static void niu_init_xif_bmac(struct niu *np)
+{
+	struct niu_link_config *lp = &np->link_config;
+	u64 val;
+
+	val = BMAC_XIF_CONFIG_TX_OUTPUT_EN;
+
+	if (lp->loopback_mode == LOOPBACK_MAC)
+		val |= BMAC_XIF_CONFIG_MII_LOOPBACK;
+	else
+		val &= ~BMAC_XIF_CONFIG_MII_LOOPBACK;
+
+	if (lp->active_speed == SPEED_1000)
+		val |= BMAC_XIF_CONFIG_GMII_MODE;
+	else
+		val &= ~BMAC_XIF_CONFIG_GMII_MODE;
+
+	val &= ~(BMAC_XIF_CONFIG_LINK_LED |
+		 BMAC_XIF_CONFIG_LED_POLARITY);
+
+	if (!(np->flags & NIU_FLAGS_10G) &&
+	    !(np->flags & NIU_FLAGS_FIBER) &&
+	    lp->active_speed == SPEED_100)
+		val |= BMAC_XIF_CONFIG_25MHZ_CLOCK;
+	else
+		val &= ~BMAC_XIF_CONFIG_25MHZ_CLOCK;
+
+	nw64_mac(BMAC_XIF_CONFIG, val);
+}
+
+static void niu_init_xif(struct niu *np)
+{
+	if (np->flags & NIU_FLAGS_XMAC)
+		niu_init_xif_xmac(np);
+	else
+		niu_init_xif_bmac(np);
+}
+
+static void niu_pcs_mii_reset(struct niu *np)
+{
+	int limit = 1000;
+	u64 val = nr64_pcs(PCS_MII_CTL);
+	val |= PCS_MII_CTL_RST;
+	nw64_pcs(PCS_MII_CTL, val);
+	while ((--limit >= 0) && (val & PCS_MII_CTL_RST)) {
+		udelay(100);
+		val = nr64_pcs(PCS_MII_CTL);
+	}
+}
+
+static void niu_xpcs_reset(struct niu *np)
+{
+	int limit = 1000;
+	u64 val = nr64_xpcs(XPCS_CONTROL1);
+	val |= XPCS_CONTROL1_RESET;
+	nw64_xpcs(XPCS_CONTROL1, val);
+	while ((--limit >= 0) && (val & XPCS_CONTROL1_RESET)) {
+		udelay(100);
+		val = nr64_xpcs(XPCS_CONTROL1);
+	}
+}
+
+static int niu_init_pcs(struct niu *np)
+{
+	struct niu_link_config *lp = &np->link_config;
+	u64 val;
+
+	switch (np->flags & (NIU_FLAGS_10G |
+			     NIU_FLAGS_FIBER |
+			     NIU_FLAGS_XCVR_SERDES)) {
+	case NIU_FLAGS_FIBER:
+		/* 1G fiber */
+		nw64_pcs(PCS_CONF, PCS_CONF_MASK | PCS_CONF_ENABLE);
+		nw64_pcs(PCS_DPATH_MODE, 0);
+		niu_pcs_mii_reset(np);
+		break;
+
+	case NIU_FLAGS_10G:
+	case NIU_FLAGS_10G | NIU_FLAGS_FIBER:
+	case NIU_FLAGS_10G | NIU_FLAGS_XCVR_SERDES:
+		/* 10G SERDES */
+		if (!(np->flags & NIU_FLAGS_XMAC))
+			return -EINVAL;
+
+		/* 10G copper or fiber */
+		val = nr64_mac(XMAC_CONFIG);
+		val &= ~XMAC_CONFIG_10G_XPCS_BYPASS;
+		nw64_mac(XMAC_CONFIG, val);
+
+		niu_xpcs_reset(np);
+
+		val = nr64_xpcs(XPCS_CONTROL1);
+		if (lp->loopback_mode == LOOPBACK_PHY)
+			val |= XPCS_CONTROL1_LOOPBACK;
+		else
+			val &= ~XPCS_CONTROL1_LOOPBACK;
+		nw64_xpcs(XPCS_CONTROL1, val);
+
+		nw64_xpcs(XPCS_DESKEW_ERR_CNT, 0);
+		(void) nr64_xpcs(XPCS_SYMERR_CNT01);
+		(void) nr64_xpcs(XPCS_SYMERR_CNT23);
+		break;
+
+
+	case NIU_FLAGS_XCVR_SERDES:
+		/* 1G SERDES */
+		niu_pcs_mii_reset(np);
+		nw64_pcs(PCS_CONF, PCS_CONF_MASK | PCS_CONF_ENABLE);
+		nw64_pcs(PCS_DPATH_MODE, 0);
+		break;
+
+	case 0:
+		/* 1G copper */
+	case NIU_FLAGS_XCVR_SERDES | NIU_FLAGS_FIBER:
+		/* 1G RGMII FIBER */
+		nw64_pcs(PCS_DPATH_MODE, PCS_DPATH_MODE_MII);
+		niu_pcs_mii_reset(np);
+		break;
+
+	default:
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
+static int niu_reset_tx_xmac(struct niu *np)
+{
+	return niu_set_and_wait_clear_mac(np, XTXMAC_SW_RST,
+					  (XTXMAC_SW_RST_REG_RS |
+					   XTXMAC_SW_RST_SOFT_RST),
+					  1000, 100, "XTXMAC_SW_RST");
+}
+
+static int niu_reset_tx_bmac(struct niu *np)
+{
+	int limit;
+
+	nw64_mac(BTXMAC_SW_RST, BTXMAC_SW_RST_RESET);
+	limit = 1000;
+	while (--limit >= 0) {
+		if (!(nr64_mac(BTXMAC_SW_RST) & BTXMAC_SW_RST_RESET))
+			break;
+		udelay(100);
+	}
+	if (limit < 0) {
+		dev_err(np->device, "Port %u TX BMAC would not reset, BTXMAC_SW_RST[%llx]\n",
+			np->port,
+			(unsigned long long) nr64_mac(BTXMAC_SW_RST));
+		return -ENODEV;
+	}
+
+	return 0;
+}
+
+static int niu_reset_tx_mac(struct niu *np)
+{
+	if (np->flags & NIU_FLAGS_XMAC)
+		return niu_reset_tx_xmac(np);
+	else
+		return niu_reset_tx_bmac(np);
+}
+
+static void niu_init_tx_xmac(struct niu *np, u64 min, u64 max)
+{
+	u64 val;
+
+	val = nr64_mac(XMAC_MIN);
+	val &= ~(XMAC_MIN_TX_MIN_PKT_SIZE |
+		 XMAC_MIN_RX_MIN_PKT_SIZE);
+	val |= (min << XMAC_MIN_RX_MIN_PKT_SIZE_SHFT);
+	val |= (min << XMAC_MIN_TX_MIN_PKT_SIZE_SHFT);
+	nw64_mac(XMAC_MIN, val);
+
+	nw64_mac(XMAC_MAX, max);
+
+	nw64_mac(XTXMAC_STAT_MSK, ~(u64)0);
+
+	val = nr64_mac(XMAC_IPG);
+	if (np->flags & NIU_FLAGS_10G) {
+		val &= ~XMAC_IPG_IPG_XGMII;
+		val |= (IPG_12_15_XGMII << XMAC_IPG_IPG_XGMII_SHIFT);
+	} else {
+		val &= ~XMAC_IPG_IPG_MII_GMII;
+		val |= (IPG_12_MII_GMII << XMAC_IPG_IPG_MII_GMII_SHIFT);
+	}
+	nw64_mac(XMAC_IPG, val);
+
+	val = nr64_mac(XMAC_CONFIG);
+	val &= ~(XMAC_CONFIG_ALWAYS_NO_CRC |
+		 XMAC_CONFIG_STRETCH_MODE |
+		 XMAC_CONFIG_VAR_MIN_IPG_EN |
+		 XMAC_CONFIG_TX_ENABLE);
+	nw64_mac(XMAC_CONFIG, val);
+
+	nw64_mac(TXMAC_FRM_CNT, 0);
+	nw64_mac(TXMAC_BYTE_CNT, 0);
+}
+
+static void niu_init_tx_bmac(struct niu *np, u64 min, u64 max)
+{
+	u64 val;
+
+	nw64_mac(BMAC_MIN_FRAME, min);
+	nw64_mac(BMAC_MAX_FRAME, max);
+
+	nw64_mac(BTXMAC_STATUS_MASK, ~(u64)0);
+	nw64_mac(BMAC_CTRL_TYPE, 0x8808);
+	nw64_mac(BMAC_PREAMBLE_SIZE, 7);
+
+	val = nr64_mac(BTXMAC_CONFIG);
+	val &= ~(BTXMAC_CONFIG_FCS_DISABLE |
+		 BTXMAC_CONFIG_ENABLE);
+	nw64_mac(BTXMAC_CONFIG, val);
+}
+
+static void niu_init_tx_mac(struct niu *np)
+{
+	u64 min, max;
+
+	min = 64;
+	if (np->dev->mtu > ETH_DATA_LEN)
+		max = 9216;
+	else
+		max = 1522;
+
+	/* The XMAC_MIN register only accepts values for TX min which
+	 * have the low 3 bits cleared.
+	 */
+	BUG_ON(min & 0x7);
+
+	if (np->flags & NIU_FLAGS_XMAC)
+		niu_init_tx_xmac(np, min, max);
+	else
+		niu_init_tx_bmac(np, min, max);
+}
+
+static int niu_reset_rx_xmac(struct niu *np)
+{
+	int limit;
+
+	nw64_mac(XRXMAC_SW_RST,
+		 XRXMAC_SW_RST_REG_RS | XRXMAC_SW_RST_SOFT_RST);
+	limit = 1000;
+	while (--limit >= 0) {
+		if (!(nr64_mac(XRXMAC_SW_RST) & (XRXMAC_SW_RST_REG_RS |
+						 XRXMAC_SW_RST_SOFT_RST)))
+			break;
+		udelay(100);
+	}
+	if (limit < 0) {
+		dev_err(np->device, "Port %u RX XMAC would not reset, XRXMAC_SW_RST[%llx]\n",
+			np->port,
+			(unsigned long long) nr64_mac(XRXMAC_SW_RST));
+		return -ENODEV;
+	}
+
+	return 0;
+}
+
+static int niu_reset_rx_bmac(struct niu *np)
+{
+	int limit;
+
+	nw64_mac(BRXMAC_SW_RST, BRXMAC_SW_RST_RESET);
+	limit = 1000;
+	while (--limit >= 0) {
+		if (!(nr64_mac(BRXMAC_SW_RST) & BRXMAC_SW_RST_RESET))
+			break;
+		udelay(100);
+	}
+	if (limit < 0) {
+		dev_err(np->device, "Port %u RX BMAC would not reset, BRXMAC_SW_RST[%llx]\n",
+			np->port,
+			(unsigned long long) nr64_mac(BRXMAC_SW_RST));
+		return -ENODEV;
+	}
+
+	return 0;
+}
+
+static int niu_reset_rx_mac(struct niu *np)
+{
+	if (np->flags & NIU_FLAGS_XMAC)
+		return niu_reset_rx_xmac(np);
+	else
+		return niu_reset_rx_bmac(np);
+}
+
+static void niu_init_rx_xmac(struct niu *np)
+{
+	struct niu_parent *parent = np->parent;
+	struct niu_rdc_tables *tp = &parent->rdc_group_cfg[np->port];
+	int first_rdc_table = tp->first_table_num;
+	unsigned long i;
+	u64 val;
+
+	nw64_mac(XMAC_ADD_FILT0, 0);
+	nw64_mac(XMAC_ADD_FILT1, 0);
+	nw64_mac(XMAC_ADD_FILT2, 0);
+	nw64_mac(XMAC_ADD_FILT12_MASK, 0);
+	nw64_mac(XMAC_ADD_FILT00_MASK, 0);
+	for (i = 0; i < MAC_NUM_HASH; i++)
+		nw64_mac(XMAC_HASH_TBL(i), 0);
+	nw64_mac(XRXMAC_STAT_MSK, ~(u64)0);
+	niu_set_primary_mac_rdc_table(np, first_rdc_table, 1);
+	niu_set_multicast_mac_rdc_table(np, first_rdc_table, 1);
+
+	val = nr64_mac(XMAC_CONFIG);
+	val &= ~(XMAC_CONFIG_RX_MAC_ENABLE |
+		 XMAC_CONFIG_PROMISCUOUS |
+		 XMAC_CONFIG_PROMISC_GROUP |
+		 XMAC_CONFIG_ERR_CHK_DIS |
+		 XMAC_CONFIG_RX_CRC_CHK_DIS |
+		 XMAC_CONFIG_RESERVED_MULTICAST |
+		 XMAC_CONFIG_RX_CODEV_CHK_DIS |
+		 XMAC_CONFIG_ADDR_FILTER_EN |
+		 XMAC_CONFIG_RCV_PAUSE_ENABLE |
+		 XMAC_CONFIG_STRIP_CRC |
+		 XMAC_CONFIG_PASS_FLOW_CTRL |
+		 XMAC_CONFIG_MAC2IPP_PKT_CNT_EN);
+	val |= (XMAC_CONFIG_HASH_FILTER_EN);
+	nw64_mac(XMAC_CONFIG, val);
+
+	nw64_mac(RXMAC_BT_CNT, 0);
+	nw64_mac(RXMAC_BC_FRM_CNT, 0);
+	nw64_mac(RXMAC_MC_FRM_CNT, 0);
+	nw64_mac(RXMAC_FRAG_CNT, 0);
+	nw64_mac(RXMAC_HIST_CNT1, 0);
+	nw64_mac(RXMAC_HIST_CNT2, 0);
+	nw64_mac(RXMAC_HIST_CNT3, 0);
+	nw64_mac(RXMAC_HIST_CNT4, 0);
+	nw64_mac(RXMAC_HIST_CNT5, 0);
+	nw64_mac(RXMAC_HIST_CNT6, 0);
+	nw64_mac(RXMAC_HIST_CNT7, 0);
+	nw64_mac(RXMAC_MPSZER_CNT, 0);
+	nw64_mac(RXMAC_CRC_ER_CNT, 0);
+	nw64_mac(RXMAC_CD_VIO_CNT, 0);
+	nw64_mac(LINK_FAULT_CNT, 0);
+}
+
+static void niu_init_rx_bmac(struct niu *np)
+{
+	struct niu_parent *parent = np->parent;
+	struct niu_rdc_tables *tp = &parent->rdc_group_cfg[np->port];
+	int first_rdc_table = tp->first_table_num;
+	unsigned long i;
+	u64 val;
+
+	nw64_mac(BMAC_ADD_FILT0, 0);
+	nw64_mac(BMAC_ADD_FILT1, 0);
+	nw64_mac(BMAC_ADD_FILT2, 0);
+	nw64_mac(BMAC_ADD_FILT12_MASK, 0);
+	nw64_mac(BMAC_ADD_FILT00_MASK, 0);
+	for (i = 0; i < MAC_NUM_HASH; i++)
+		nw64_mac(BMAC_HASH_TBL(i), 0);
+	niu_set_primary_mac_rdc_table(np, first_rdc_table, 1);
+	niu_set_multicast_mac_rdc_table(np, first_rdc_table, 1);
+	nw64_mac(BRXMAC_STATUS_MASK, ~(u64)0);
+
+	val = nr64_mac(BRXMAC_CONFIG);
+	val &= ~(BRXMAC_CONFIG_ENABLE |
+		 BRXMAC_CONFIG_STRIP_PAD |
+		 BRXMAC_CONFIG_STRIP_FCS |
+		 BRXMAC_CONFIG_PROMISC |
+		 BRXMAC_CONFIG_PROMISC_GRP |
+		 BRXMAC_CONFIG_ADDR_FILT_EN |
+		 BRXMAC_CONFIG_DISCARD_DIS);
+	val |= (BRXMAC_CONFIG_HASH_FILT_EN);
+	nw64_mac(BRXMAC_CONFIG, val);
+
+	val = nr64_mac(BMAC_ADDR_CMPEN);
+	val |= BMAC_ADDR_CMPEN_EN0;
+	nw64_mac(BMAC_ADDR_CMPEN, val);
+}
+
+static void niu_init_rx_mac(struct niu *np)
+{
+	niu_set_primary_mac(np, np->dev->dev_addr);
+
+	if (np->flags & NIU_FLAGS_XMAC)
+		niu_init_rx_xmac(np);
+	else
+		niu_init_rx_bmac(np);
+}
+
+static void niu_enable_tx_xmac(struct niu *np, int on)
+{
+	u64 val = nr64_mac(XMAC_CONFIG);
+
+	if (on)
+		val |= XMAC_CONFIG_TX_ENABLE;
+	else
+		val &= ~XMAC_CONFIG_TX_ENABLE;
+	nw64_mac(XMAC_CONFIG, val);
+}
+
+static void niu_enable_tx_bmac(struct niu *np, int on)
+{
+	u64 val = nr64_mac(BTXMAC_CONFIG);
+
+	if (on)
+		val |= BTXMAC_CONFIG_ENABLE;
+	else
+		val &= ~BTXMAC_CONFIG_ENABLE;
+	nw64_mac(BTXMAC_CONFIG, val);
+}
+
+static void niu_enable_tx_mac(struct niu *np, int on)
+{
+	if (np->flags & NIU_FLAGS_XMAC)
+		niu_enable_tx_xmac(np, on);
+	else
+		niu_enable_tx_bmac(np, on);
+}
+
+static void niu_enable_rx_xmac(struct niu *np, int on)
+{
+	u64 val = nr64_mac(XMAC_CONFIG);
+
+	val &= ~(XMAC_CONFIG_HASH_FILTER_EN |
+		 XMAC_CONFIG_PROMISCUOUS);
+
+	if (np->flags & NIU_FLAGS_MCAST)
+		val |= XMAC_CONFIG_HASH_FILTER_EN;
+	if (np->flags & NIU_FLAGS_PROMISC)
+		val |= XMAC_CONFIG_PROMISCUOUS;
+
+	if (on)
+		val |= XMAC_CONFIG_RX_MAC_ENABLE;
+	else
+		val &= ~XMAC_CONFIG_RX_MAC_ENABLE;
+	nw64_mac(XMAC_CONFIG, val);
+}
+
+static void niu_enable_rx_bmac(struct niu *np, int on)
+{
+	u64 val = nr64_mac(BRXMAC_CONFIG);
+
+	val &= ~(BRXMAC_CONFIG_HASH_FILT_EN |
+		 BRXMAC_CONFIG_PROMISC);
+
+	if (np->flags & NIU_FLAGS_MCAST)
+		val |= BRXMAC_CONFIG_HASH_FILT_EN;
+	if (np->flags & NIU_FLAGS_PROMISC)
+		val |= BRXMAC_CONFIG_PROMISC;
+
+	if (on)
+		val |= BRXMAC_CONFIG_ENABLE;
+	else
+		val &= ~BRXMAC_CONFIG_ENABLE;
+	nw64_mac(BRXMAC_CONFIG, val);
+}
+
+static void niu_enable_rx_mac(struct niu *np, int on)
+{
+	if (np->flags & NIU_FLAGS_XMAC)
+		niu_enable_rx_xmac(np, on);
+	else
+		niu_enable_rx_bmac(np, on);
+}
+
+static int niu_init_mac(struct niu *np)
+{
+	int err;
+
+	niu_init_xif(np);
+	err = niu_init_pcs(np);
+	if (err)
+		return err;
+
+	err = niu_reset_tx_mac(np);
+	if (err)
+		return err;
+	niu_init_tx_mac(np);
+	err = niu_reset_rx_mac(np);
+	if (err)
+		return err;
+	niu_init_rx_mac(np);
+
+	/* This looks hookey but the RX MAC reset we just did will
+	 * undo some of the state we setup in niu_init_tx_mac() so we
+	 * have to call it again.  In particular, the RX MAC reset will
+	 * set the XMAC_MAX register back to its default value.
+	 */
+	niu_init_tx_mac(np);
+	niu_enable_tx_mac(np, 1);
+
+	niu_enable_rx_mac(np, 1);
+
+	return 0;
+}
+
+static void niu_stop_one_tx_channel(struct niu *np, struct tx_ring_info *rp)
+{
+	(void) niu_tx_channel_stop(np, rp->tx_channel);
+}
+
+static void niu_stop_tx_channels(struct niu *np)
+{
+	int i;
+
+	for (i = 0; i < np->num_tx_rings; i++) {
+		struct tx_ring_info *rp = &np->tx_rings[i];
+
+		niu_stop_one_tx_channel(np, rp);
+	}
+}
+
+static void niu_reset_one_tx_channel(struct niu *np, struct tx_ring_info *rp)
+{
+	(void) niu_tx_channel_reset(np, rp->tx_channel);
+}
+
+static void niu_reset_tx_channels(struct niu *np)
+{
+	int i;
+
+	for (i = 0; i < np->num_tx_rings; i++) {
+		struct tx_ring_info *rp = &np->tx_rings[i];
+
+		niu_reset_one_tx_channel(np, rp);
+	}
+}
+
+static void niu_stop_one_rx_channel(struct niu *np, struct rx_ring_info *rp)
+{
+	(void) niu_enable_rx_channel(np, rp->rx_channel, 0);
+}
+
+static void niu_stop_rx_channels(struct niu *np)
+{
+	int i;
+
+	for (i = 0; i < np->num_rx_rings; i++) {
+		struct rx_ring_info *rp = &np->rx_rings[i];
+
+		niu_stop_one_rx_channel(np, rp);
+	}
+}
+
+static void niu_reset_one_rx_channel(struct niu *np, struct rx_ring_info *rp)
+{
+	int channel = rp->rx_channel;
+
+	(void) niu_rx_channel_reset(np, channel);
+	nw64(RX_DMA_ENT_MSK(channel), RX_DMA_ENT_MSK_ALL);
+	nw64(RX_DMA_CTL_STAT(channel), 0);
+	(void) niu_enable_rx_channel(np, channel, 0);
+}
+
+static void niu_reset_rx_channels(struct niu *np)
+{
+	int i;
+
+	for (i = 0; i < np->num_rx_rings; i++) {
+		struct rx_ring_info *rp = &np->rx_rings[i];
+
+		niu_reset_one_rx_channel(np, rp);
+	}
+}
+
+static void niu_disable_ipp(struct niu *np)
+{
+	u64 rd, wr, val;
+	int limit;
+
+	rd = nr64_ipp(IPP_DFIFO_RD_PTR);
+	wr = nr64_ipp(IPP_DFIFO_WR_PTR);
+	limit = 100;
+	while (--limit >= 0 && (rd != wr)) {
+		rd = nr64_ipp(IPP_DFIFO_RD_PTR);
+		wr = nr64_ipp(IPP_DFIFO_WR_PTR);
+	}
+	if (limit < 0 &&
+	    (rd != 0 && wr != 1)) {
+		netdev_err(np->dev, "IPP would not quiesce, rd_ptr[%llx] wr_ptr[%llx]\n",
+			   (unsigned long long)nr64_ipp(IPP_DFIFO_RD_PTR),
+			   (unsigned long long)nr64_ipp(IPP_DFIFO_WR_PTR));
+	}
+
+	val = nr64_ipp(IPP_CFIG);
+	val &= ~(IPP_CFIG_IPP_ENABLE |
+		 IPP_CFIG_DFIFO_ECC_EN |
+		 IPP_CFIG_DROP_BAD_CRC |
+		 IPP_CFIG_CKSUM_EN);
+	nw64_ipp(IPP_CFIG, val);
+
+	(void) niu_ipp_reset(np);
+}
+
+static int niu_init_hw(struct niu *np)
+{
+	int i, err;
+
+	netif_printk(np, ifup, KERN_DEBUG, np->dev, "Initialize TXC\n");
+	niu_txc_enable_port(np, 1);
+	niu_txc_port_dma_enable(np, 1);
+	niu_txc_set_imask(np, 0);
+
+	netif_printk(np, ifup, KERN_DEBUG, np->dev, "Initialize TX channels\n");
+	for (i = 0; i < np->num_tx_rings; i++) {
+		struct tx_ring_info *rp = &np->tx_rings[i];
+
+		err = niu_init_one_tx_channel(np, rp);
+		if (err)
+			return err;
+	}
+
+	netif_printk(np, ifup, KERN_DEBUG, np->dev, "Initialize RX channels\n");
+	err = niu_init_rx_channels(np);
+	if (err)
+		goto out_uninit_tx_channels;
+
+	netif_printk(np, ifup, KERN_DEBUG, np->dev, "Initialize classifier\n");
+	err = niu_init_classifier_hw(np);
+	if (err)
+		goto out_uninit_rx_channels;
+
+	netif_printk(np, ifup, KERN_DEBUG, np->dev, "Initialize ZCP\n");
+	err = niu_init_zcp(np);
+	if (err)
+		goto out_uninit_rx_channels;
+
+	netif_printk(np, ifup, KERN_DEBUG, np->dev, "Initialize IPP\n");
+	err = niu_init_ipp(np);
+	if (err)
+		goto out_uninit_rx_channels;
+
+	netif_printk(np, ifup, KERN_DEBUG, np->dev, "Initialize MAC\n");
+	err = niu_init_mac(np);
+	if (err)
+		goto out_uninit_ipp;
+
+	return 0;
+
+out_uninit_ipp:
+	netif_printk(np, ifup, KERN_DEBUG, np->dev, "Uninit IPP\n");
+	niu_disable_ipp(np);
+
+out_uninit_rx_channels:
+	netif_printk(np, ifup, KERN_DEBUG, np->dev, "Uninit RX channels\n");
+	niu_stop_rx_channels(np);
+	niu_reset_rx_channels(np);
+
+out_uninit_tx_channels:
+	netif_printk(np, ifup, KERN_DEBUG, np->dev, "Uninit TX channels\n");
+	niu_stop_tx_channels(np);
+	niu_reset_tx_channels(np);
+
+	return err;
+}
+
+static void niu_stop_hw(struct niu *np)
+{
+	netif_printk(np, ifdown, KERN_DEBUG, np->dev, "Disable interrupts\n");
+	niu_enable_interrupts(np, 0);
+
+	netif_printk(np, ifdown, KERN_DEBUG, np->dev, "Disable RX MAC\n");
+	niu_enable_rx_mac(np, 0);
+
+	netif_printk(np, ifdown, KERN_DEBUG, np->dev, "Disable IPP\n");
+	niu_disable_ipp(np);
+
+	netif_printk(np, ifdown, KERN_DEBUG, np->dev, "Stop TX channels\n");
+	niu_stop_tx_channels(np);
+
+	netif_printk(np, ifdown, KERN_DEBUG, np->dev, "Stop RX channels\n");
+	niu_stop_rx_channels(np);
+
+	netif_printk(np, ifdown, KERN_DEBUG, np->dev, "Reset TX channels\n");
+	niu_reset_tx_channels(np);
+
+	netif_printk(np, ifdown, KERN_DEBUG, np->dev, "Reset RX channels\n");
+	niu_reset_rx_channels(np);
+}
+
+static void niu_set_irq_name(struct niu *np)
+{
+	int port = np->port;
+	int i, j = 1;
+
+	snprintf(np->irq_name[0], sizeof(np->irq_name[0]), "%s:MAC", np->dev->name);
+
+	if (port == 0) {
+		snprintf(np->irq_name[1], sizeof(np->irq_name[1]), "%s:MIF", np->dev->name);
+		snprintf(np->irq_name[2], sizeof(np->irq_name[2]), "%s:SYSERR", np->dev->name);
+		j = 3;
+	}
+
+	for (i = 0; i < np->num_ldg - j; i++) {
+		if (i < np->num_rx_rings)
+			sprintf(np->irq_name[i+j], "%s-rx-%d",
+				np->dev->name, i);
+		else if (i < np->num_tx_rings + np->num_rx_rings)
+			sprintf(np->irq_name[i+j], "%s-tx-%d", np->dev->name,
+				i - np->num_rx_rings);
+	}
+}
+
+static int niu_request_irq(struct niu *np)
+{
+	int i, j, err;
+
+	niu_set_irq_name(np);
+
+	err = 0;
+	for (i = 0; i < np->num_ldg; i++) {
+		struct niu_ldg *lp = &np->ldg[i];
+
+		err = request_irq(lp->irq, niu_interrupt, IRQF_SHARED,
+				  np->irq_name[i], lp);
+		if (err)
+			goto out_free_irqs;
+
+	}
+
+	return 0;
+
+out_free_irqs:
+	for (j = 0; j < i; j++) {
+		struct niu_ldg *lp = &np->ldg[j];
+
+		free_irq(lp->irq, lp);
+	}
+	return err;
+}
+
+static void niu_free_irq(struct niu *np)
+{
+	int i;
+
+	for (i = 0; i < np->num_ldg; i++) {
+		struct niu_ldg *lp = &np->ldg[i];
+
+		free_irq(lp->irq, lp);
+	}
+}
+
+static void niu_enable_napi(struct niu *np)
+{
+	int i;
+
+	for (i = 0; i < np->num_ldg; i++)
+		napi_enable_locked(&np->ldg[i].napi);
+}
+
+static void niu_disable_napi(struct niu *np)
+{
+	int i;
+
+	for (i = 0; i < np->num_ldg; i++)
+		napi_disable(&np->ldg[i].napi);
+}
+
+static int niu_open(struct net_device *dev)
+{
+	struct niu *np = netdev_priv(dev);
+	int err;
+
+	netif_carrier_off(dev);
+
+	err = niu_alloc_channels(np);
+	if (err)
+		goto out_err;
+
+	err = niu_enable_interrupts(np, 0);
+	if (err)
+		goto out_free_channels;
+
+	err = niu_request_irq(np);
+	if (err)
+		goto out_free_channels;
+
+	netdev_lock(dev);
+	niu_enable_napi(np);
+	netdev_unlock(dev);
+
+	spin_lock_irq(&np->lock);
+
+	err = niu_init_hw(np);
+	if (!err) {
+		timer_setup(&np->timer, niu_timer, 0);
+		np->timer.expires = jiffies + HZ;
+
+		err = niu_enable_interrupts(np, 1);
+		if (err)
+			niu_stop_hw(np);
+	}
+
+	spin_unlock_irq(&np->lock);
+
+	if (err) {
+		niu_disable_napi(np);
+		goto out_free_irq;
+	}
+
+	netif_tx_start_all_queues(dev);
+
+	if (np->link_config.loopback_mode != LOOPBACK_DISABLED)
+		netif_carrier_on(dev);
+
+	add_timer(&np->timer);
+
+	return 0;
+
+out_free_irq:
+	niu_free_irq(np);
+
+out_free_channels:
+	niu_free_channels(np);
+
+out_err:
+	return err;
+}
+
+static void niu_full_shutdown(struct niu *np, struct net_device *dev)
+{
+	cancel_work_sync(&np->reset_task);
+
+	niu_disable_napi(np);
+	netif_tx_stop_all_queues(dev);
+
+	timer_delete_sync(&np->timer);
+
+	spin_lock_irq(&np->lock);
+
+	niu_stop_hw(np);
+
+	spin_unlock_irq(&np->lock);
+}
+
+static int niu_close(struct net_device *dev)
+{
+	struct niu *np = netdev_priv(dev);
+
+	niu_full_shutdown(np, dev);
+
+	niu_free_irq(np);
+
+	niu_free_channels(np);
+
+	niu_handle_led(np, 0);
+
+	return 0;
+}
+
+static void niu_sync_xmac_stats(struct niu *np)
+{
+	struct niu_xmac_stats *mp = &np->mac_stats.xmac;
+
+	mp->tx_frames += nr64_mac(TXMAC_FRM_CNT);
+	mp->tx_bytes += nr64_mac(TXMAC_BYTE_CNT);
+
+	mp->rx_link_faults += nr64_mac(LINK_FAULT_CNT);
+	mp->rx_align_errors += nr64_mac(RXMAC_ALIGN_ERR_CNT);
+	mp->rx_frags += nr64_mac(RXMAC_FRAG_CNT);
+	mp->rx_mcasts += nr64_mac(RXMAC_MC_FRM_CNT);
+	mp->rx_bcasts += nr64_mac(RXMAC_BC_FRM_CNT);
+	mp->rx_hist_cnt1 += nr64_mac(RXMAC_HIST_CNT1);
+	mp->rx_hist_cnt2 += nr64_mac(RXMAC_HIST_CNT2);
+	mp->rx_hist_cnt3 += nr64_mac(RXMAC_HIST_CNT3);
+	mp->rx_hist_cnt4 += nr64_mac(RXMAC_HIST_CNT4);
+	mp->rx_hist_cnt5 += nr64_mac(RXMAC_HIST_CNT5);
+	mp->rx_hist_cnt6 += nr64_mac(RXMAC_HIST_CNT6);
+	mp->rx_hist_cnt7 += nr64_mac(RXMAC_HIST_CNT7);
+	mp->rx_octets += nr64_mac(RXMAC_BT_CNT);
+	mp->rx_code_violations += nr64_mac(RXMAC_CD_VIO_CNT);
+	mp->rx_len_errors += nr64_mac(RXMAC_MPSZER_CNT);
+	mp->rx_crc_errors += nr64_mac(RXMAC_CRC_ER_CNT);
+}
+
+static void niu_sync_bmac_stats(struct niu *np)
+{
+	struct niu_bmac_stats *mp = &np->mac_stats.bmac;
+
+	mp->tx_bytes += nr64_mac(BTXMAC_BYTE_CNT);
+	mp->tx_frames += nr64_mac(BTXMAC_FRM_CNT);
+
+	mp->rx_frames += nr64_mac(BRXMAC_FRAME_CNT);
+	mp->rx_align_errors += nr64_mac(BRXMAC_ALIGN_ERR_CNT);
+	mp->rx_crc_errors += nr64_mac(BRXMAC_ALIGN_ERR_CNT);
+	mp->rx_len_errors += nr64_mac(BRXMAC_CODE_VIOL_ERR_CNT);
+}
+
+static void niu_sync_mac_stats(struct niu *np)
+{
+	if (np->flags & NIU_FLAGS_XMAC)
+		niu_sync_xmac_stats(np);
+	else
+		niu_sync_bmac_stats(np);
+}
+
+static void niu_get_rx_stats(struct niu *np,
+			     struct rtnl_link_stats64 *stats)
+{
+	u64 pkts, dropped, errors, bytes;
+	struct rx_ring_info *rx_rings;
+	int i;
+
+	pkts = dropped = errors = bytes = 0;
+
+	rx_rings = READ_ONCE(np->rx_rings);
+	if (!rx_rings)
+		goto no_rings;
+
+	for (i = 0; i < np->num_rx_rings; i++) {
+		struct rx_ring_info *rp = &rx_rings[i];
+
+		niu_sync_rx_discard_stats(np, rp, 0);
+
+		pkts += rp->rx_packets;
+		bytes += rp->rx_bytes;
+		dropped += rp->rx_dropped;
+		errors += rp->rx_errors;
+	}
+
+no_rings:
+	stats->rx_packets = pkts;
+	stats->rx_bytes = bytes;
+	stats->rx_dropped = dropped;
+	stats->rx_errors = errors;
+}
+
+static void niu_get_tx_stats(struct niu *np,
+			     struct rtnl_link_stats64 *stats)
+{
+	u64 pkts, errors, bytes;
+	struct tx_ring_info *tx_rings;
+	int i;
+
+	pkts = errors = bytes = 0;
+
+	tx_rings = READ_ONCE(np->tx_rings);
+	if (!tx_rings)
+		goto no_rings;
+
+	for (i = 0; i < np->num_tx_rings; i++) {
+		struct tx_ring_info *rp = &tx_rings[i];
+
+		pkts += rp->tx_packets;
+		bytes += rp->tx_bytes;
+		errors += rp->tx_errors;
+	}
+
+no_rings:
+	stats->tx_packets = pkts;
+	stats->tx_bytes = bytes;
+	stats->tx_errors = errors;
+}
+
+static void niu_get_stats(struct net_device *dev,
+			  struct rtnl_link_stats64 *stats)
+{
+	struct niu *np = netdev_priv(dev);
+
+	if (netif_running(dev)) {
+		niu_get_rx_stats(np, stats);
+		niu_get_tx_stats(np, stats);
+	}
+}
+
+static void niu_load_hash_xmac(struct niu *np, u16 *hash)
+{
+	int i;
+
+	for (i = 0; i < 16; i++)
+		nw64_mac(XMAC_HASH_TBL(i), hash[i]);
+}
+
+static void niu_load_hash_bmac(struct niu *np, u16 *hash)
+{
+	int i;
+
+	for (i = 0; i < 16; i++)
+		nw64_mac(BMAC_HASH_TBL(i), hash[i]);
+}
+
+static void niu_load_hash(struct niu *np, u16 *hash)
+{
+	if (np->flags & NIU_FLAGS_XMAC)
+		niu_load_hash_xmac(np, hash);
+	else
+		niu_load_hash_bmac(np, hash);
+}
+
+static void niu_set_rx_mode(struct net_device *dev)
+{
+	struct niu *np = netdev_priv(dev);
+	int i, alt_cnt, err;
+	struct netdev_hw_addr *ha;
+	unsigned long flags;
+	u16 hash[16] = { 0, };
+
+	spin_lock_irqsave(&np->lock, flags);
+	niu_enable_rx_mac(np, 0);
+
+	np->flags &= ~(NIU_FLAGS_MCAST | NIU_FLAGS_PROMISC);
+	if (dev->flags & IFF_PROMISC)
+		np->flags |= NIU_FLAGS_PROMISC;
+	if ((dev->flags & IFF_ALLMULTI) || (!netdev_mc_empty(dev)))
+		np->flags |= NIU_FLAGS_MCAST;
+
+	alt_cnt = netdev_uc_count(dev);
+	if (alt_cnt > niu_num_alt_addr(np)) {
+		alt_cnt = 0;
+		np->flags |= NIU_FLAGS_PROMISC;
+	}
+
+	if (alt_cnt) {
+		int index = 0;
+
+		netdev_for_each_uc_addr(ha, dev) {
+			err = niu_set_alt_mac(np, index, ha->addr);
+			if (err)
+				netdev_warn(dev, "Error %d adding alt mac %d\n",
+					    err, index);
+			err = niu_enable_alt_mac(np, index, 1);
+			if (err)
+				netdev_warn(dev, "Error %d enabling alt mac %d\n",
+					    err, index);
+
+			index++;
+		}
+	} else {
+		int alt_start;
+		if (np->flags & NIU_FLAGS_XMAC)
+			alt_start = 0;
+		else
+			alt_start = 1;
+		for (i = alt_start; i < niu_num_alt_addr(np); i++) {
+			err = niu_enable_alt_mac(np, i, 0);
+			if (err)
+				netdev_warn(dev, "Error %d disabling alt mac %d\n",
+					    err, i);
+		}
+	}
+	if (dev->flags & IFF_ALLMULTI) {
+		for (i = 0; i < 16; i++)
+			hash[i] = 0xffff;
+	} else if (!netdev_mc_empty(dev)) {
+		netdev_for_each_mc_addr(ha, dev) {
+			u32 crc = ether_crc_le(ETH_ALEN, ha->addr);
+
+			crc >>= 24;
+			hash[crc >> 4] |= (1 << (15 - (crc & 0xf)));
+		}
+	}
+
+	if (np->flags & NIU_FLAGS_MCAST)
+		niu_load_hash(np, hash);
+
+	niu_enable_rx_mac(np, 1);
+	spin_unlock_irqrestore(&np->lock, flags);
+}
+
+static int niu_set_mac_addr(struct net_device *dev, void *p)
+{
+	struct niu *np = netdev_priv(dev);
+	struct sockaddr *addr = p;
+	unsigned long flags;
+
+	if (!is_valid_ether_addr(addr->sa_data))
+		return -EADDRNOTAVAIL;
+
+	eth_hw_addr_set(dev, addr->sa_data);
+
+	if (!netif_running(dev))
+		return 0;
+
+	spin_lock_irqsave(&np->lock, flags);
+	niu_enable_rx_mac(np, 0);
+	niu_set_primary_mac(np, dev->dev_addr);
+	niu_enable_rx_mac(np, 1);
+	spin_unlock_irqrestore(&np->lock, flags);
+
+	return 0;
+}
+
+static int niu_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
+{
+	return -EOPNOTSUPP;
+}
+
+static void niu_netif_stop(struct niu *np)
+{
+	netif_trans_update(np->dev);	/* prevent tx timeout */
+
+	niu_disable_napi(np);
+
+	netif_tx_disable(np->dev);
+}
+
+static void niu_netif_start(struct niu *np)
+{
+	/* NOTE: unconditional netif_wake_queue is only appropriate
+	 * so long as all callers are assured to have free tx slots
+	 * (such as after niu_init_hw).
+	 */
+	netif_tx_wake_all_queues(np->dev);
+
+	niu_enable_napi(np);
+
+	niu_enable_interrupts(np, 1);
+}
+
+static void niu_reset_buffers(struct niu *np)
+{
+	int i, j, k, err;
+
+	if (np->rx_rings) {
+		for (i = 0; i < np->num_rx_rings; i++) {
+			struct rx_ring_info *rp = &np->rx_rings[i];
+
+			for (j = 0, k = 0; j < MAX_RBR_RING_SIZE; j++) {
+				struct page *page;
+
+				page = rp->rxhash[j];
+				while (page) {
+					struct page *next = niu_next_page(page);
+					u64 base = page->private;
+					base = base >> RBR_DESCR_ADDR_SHIFT;
+					rp->rbr[k++] = cpu_to_le32(base);
+					page = next;
+				}
+			}
+			for (; k < MAX_RBR_RING_SIZE; k++) {
+				err = niu_rbr_add_page(np, rp, GFP_ATOMIC, k);
+				if (unlikely(err))
+					break;
+			}
+
+			rp->rbr_index = rp->rbr_table_size - 1;
+			rp->rcr_index = 0;
+			rp->rbr_pending = 0;
+			rp->rbr_refill_pending = 0;
+		}
+	}
+	if (np->tx_rings) {
+		for (i = 0; i < np->num_tx_rings; i++) {
+			struct tx_ring_info *rp = &np->tx_rings[i];
+
+			for (j = 0; j < MAX_TX_RING_SIZE; j++) {
+				if (rp->tx_buffs[j].skb)
+					(void) release_tx_packet(np, rp, j);
+			}
+
+			rp->pending = MAX_TX_RING_SIZE;
+			rp->prod = 0;
+			rp->cons = 0;
+			rp->wrap_bit = 0;
+		}
+	}
+}
+
+static void niu_reset_task(struct work_struct *work)
+{
+	struct niu *np = container_of(work, struct niu, reset_task);
+	unsigned long flags;
+	int err;
+
+	spin_lock_irqsave(&np->lock, flags);
+	if (!netif_running(np->dev)) {
+		spin_unlock_irqrestore(&np->lock, flags);
+		return;
+	}
+
+	spin_unlock_irqrestore(&np->lock, flags);
+
+	timer_delete_sync(&np->timer);
+
+	niu_netif_stop(np);
+
+	spin_lock_irqsave(&np->lock, flags);
+
+	niu_stop_hw(np);
+
+	spin_unlock_irqrestore(&np->lock, flags);
+
+	niu_reset_buffers(np);
+
+	netdev_lock(np->dev);
+	spin_lock_irqsave(&np->lock, flags);
+
+	err = niu_init_hw(np);
+	if (!err) {
+		np->timer.expires = jiffies + HZ;
+		add_timer(&np->timer);
+		niu_netif_start(np);
+	}
+
+	spin_unlock_irqrestore(&np->lock, flags);
+	netdev_unlock(np->dev);
+}
+
+static void niu_tx_timeout(struct net_device *dev, unsigned int txqueue)
+{
+	struct niu *np = netdev_priv(dev);
+
+	dev_err(np->device, "%s: Transmit timed out, resetting\n",
+		dev->name);
+
+	schedule_work(&np->reset_task);
+}
+
+static void niu_set_txd(struct tx_ring_info *rp, int index,
+			u64 mapping, u64 len, u64 mark,
+			u64 n_frags)
+{
+	__le64 *desc = &rp->descr[index];
+
+	*desc = cpu_to_le64(mark |
+			    (n_frags << TX_DESC_NUM_PTR_SHIFT) |
+			    (len << TX_DESC_TR_LEN_SHIFT) |
+			    (mapping & TX_DESC_SAD));
+}
+
+static u64 niu_compute_tx_flags(struct sk_buff *skb, struct ethhdr *ehdr,
+				u64 pad_bytes, u64 len)
+{
+	u16 eth_proto, eth_proto_inner;
+	u64 csum_bits, l3off, ihl, ret;
+	u8 ip_proto;
+	int ipv6;
+
+	eth_proto = be16_to_cpu(ehdr->h_proto);
+	eth_proto_inner = eth_proto;
+	if (eth_proto == ETH_P_8021Q) {
+		struct vlan_ethhdr *vp = (struct vlan_ethhdr *) ehdr;
+		__be16 val = vp->h_vlan_encapsulated_proto;
+
+		eth_proto_inner = be16_to_cpu(val);
+	}
+
+	ipv6 = ihl = 0;
+	switch (skb->protocol) {
+	case cpu_to_be16(ETH_P_IP):
+		ip_proto = ip_hdr(skb)->protocol;
+		ihl = ip_hdr(skb)->ihl;
+		break;
+	case cpu_to_be16(ETH_P_IPV6):
+		ip_proto = ipv6_hdr(skb)->nexthdr;
+		ihl = (40 >> 2);
+		ipv6 = 1;
+		break;
+	default:
+		ip_proto = ihl = 0;
+		break;
+	}
+
+	csum_bits = TXHDR_CSUM_NONE;
+	if (skb->ip_summed == CHECKSUM_PARTIAL) {
+		u64 start, stuff;
+
+		csum_bits = (ip_proto == IPPROTO_TCP ?
+			     TXHDR_CSUM_TCP :
+			     (ip_proto == IPPROTO_UDP ?
+			      TXHDR_CSUM_UDP : TXHDR_CSUM_SCTP));
+
+		start = skb_checksum_start_offset(skb) -
+			(pad_bytes + sizeof(struct tx_pkt_hdr));
+		stuff = start + skb->csum_offset;
+
+		csum_bits |= (start / 2) << TXHDR_L4START_SHIFT;
+		csum_bits |= (stuff / 2) << TXHDR_L4STUFF_SHIFT;
+	}
+
+	l3off = skb_network_offset(skb) -
+		(pad_bytes + sizeof(struct tx_pkt_hdr));
+
+	ret = (((pad_bytes / 2) << TXHDR_PAD_SHIFT) |
+	       (len << TXHDR_LEN_SHIFT) |
+	       ((l3off / 2) << TXHDR_L3START_SHIFT) |
+	       (ihl << TXHDR_IHL_SHIFT) |
+	       ((eth_proto_inner < ETH_P_802_3_MIN) ? TXHDR_LLC : 0) |
+	       ((eth_proto == ETH_P_8021Q) ? TXHDR_VLAN : 0) |
+	       (ipv6 ? TXHDR_IP_VER : 0) |
+	       csum_bits);
+
+	return ret;
+}
+
+static netdev_tx_t niu_start_xmit(struct sk_buff *skb,
+				  struct net_device *dev)
+{
+	struct niu *np = netdev_priv(dev);
+	unsigned long align, headroom;
+	struct netdev_queue *txq;
+	struct tx_ring_info *rp;
+	struct tx_pkt_hdr *tp;
+	unsigned int len, nfg;
+	struct ethhdr *ehdr;
+	int prod, i, tlen;
+	u64 mapping, mrk;
+
+	i = skb_get_queue_mapping(skb);
+	rp = &np->tx_rings[i];
+	txq = netdev_get_tx_queue(dev, i);
+
+	if (niu_tx_avail(rp) <= (skb_shinfo(skb)->nr_frags + 1)) {
+		netif_tx_stop_queue(txq);
+		dev_err(np->device, "%s: BUG! Tx ring full when queue awake!\n", dev->name);
+		rp->tx_errors++;
+		return NETDEV_TX_BUSY;
+	}
+
+	if (eth_skb_pad(skb))
+		goto out;
+
+	len = sizeof(struct tx_pkt_hdr) + 15;
+	if (skb_headroom(skb) < len) {
+		struct sk_buff *skb_new;
+
+		skb_new = skb_realloc_headroom(skb, len);
+		if (!skb_new)
+			goto out_drop;
+		kfree_skb(skb);
+		skb = skb_new;
+	} else
+		skb_orphan(skb);
+
+	align = ((unsigned long) skb->data & (16 - 1));
+	headroom = align + sizeof(struct tx_pkt_hdr);
+
+	ehdr = (struct ethhdr *) skb->data;
+	tp = skb_push(skb, headroom);
+
+	len = skb->len - sizeof(struct tx_pkt_hdr);
+	tp->flags = cpu_to_le64(niu_compute_tx_flags(skb, ehdr, align, len));
+	tp->resv = 0;
+
+	len = skb_headlen(skb);
+	mapping = np->ops->map_single(np->device, skb->data,
+				      len, DMA_TO_DEVICE);
+	if (np->ops->mapping_error(np->device, mapping))
+		goto out_drop;
+
+	prod = rp->prod;
+
+	rp->tx_buffs[prod].skb = skb;
+	rp->tx_buffs[prod].mapping = mapping;
+
+	mrk = TX_DESC_SOP;
+	if (++rp->mark_counter == rp->mark_freq) {
+		rp->mark_counter = 0;
+		mrk |= TX_DESC_MARK;
+		rp->mark_pending++;
+	}
+
+	tlen = len;
+	nfg = skb_shinfo(skb)->nr_frags;
+	while (tlen > 0) {
+		tlen -= MAX_TX_DESC_LEN;
+		nfg++;
+	}
+
+	while (len > 0) {
+		unsigned int this_len = len;
+
+		if (this_len > MAX_TX_DESC_LEN)
+			this_len = MAX_TX_DESC_LEN;
+
+		niu_set_txd(rp, prod, mapping, this_len, mrk, nfg);
+		mrk = nfg = 0;
+
+		prod = NEXT_TX(rp, prod);
+		mapping += this_len;
+		len -= this_len;
+	}
+
+	for (i = 0; i <  skb_shinfo(skb)->nr_frags; i++) {
+		const skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
+
+		len = skb_frag_size(frag);
+		mapping = np->ops->map_page(np->device, skb_frag_page(frag),
+					    skb_frag_off(frag), len,
+					    DMA_TO_DEVICE);
+		if (np->ops->mapping_error(np->device, mapping))
+			goto out_unmap;
+
+		rp->tx_buffs[prod].skb = NULL;
+		rp->tx_buffs[prod].mapping = mapping;
+
+		niu_set_txd(rp, prod, mapping, len, 0, 0);
+
+		prod = NEXT_TX(rp, prod);
+	}
+
+	if (prod < rp->prod)
+		rp->wrap_bit ^= TX_RING_KICK_WRAP;
+	rp->prod = prod;
+
+	nw64(TX_RING_KICK(rp->tx_channel), rp->wrap_bit | (prod << 3));
+
+	if (unlikely(niu_tx_avail(rp) <= (MAX_SKB_FRAGS + 1))) {
+		netif_tx_stop_queue(txq);
+		if (niu_tx_avail(rp) > NIU_TX_WAKEUP_THRESH(rp))
+			netif_tx_wake_queue(txq);
+	}
+
+out:
+	return NETDEV_TX_OK;
+
+out_unmap:
+	while (i--) {
+		const skb_frag_t *frag;
+
+		prod = PREVIOUS_TX(rp, prod);
+		frag = &skb_shinfo(skb)->frags[i];
+		np->ops->unmap_page(np->device, rp->tx_buffs[prod].mapping,
+				    skb_frag_size(frag), DMA_TO_DEVICE);
+	}
+
+	np->ops->unmap_single(np->device, rp->tx_buffs[rp->prod].mapping,
+			      skb_headlen(skb), DMA_TO_DEVICE);
+
+out_drop:
+	rp->tx_errors++;
+	kfree_skb(skb);
+	goto out;
+}
+
+static int niu_change_mtu(struct net_device *dev, int new_mtu)
+{
+	struct niu *np = netdev_priv(dev);
+	int err, orig_jumbo, new_jumbo;
+
+	orig_jumbo = (dev->mtu > ETH_DATA_LEN);
+	new_jumbo = (new_mtu > ETH_DATA_LEN);
+
+	WRITE_ONCE(dev->mtu, new_mtu);
+
+	if (!netif_running(dev) ||
+	    (orig_jumbo == new_jumbo))
+		return 0;
+
+	niu_full_shutdown(np, dev);
+
+	niu_free_channels(np);
+
+	netdev_lock(dev);
+	niu_enable_napi(np);
+	netdev_unlock(dev);
+
+	err = niu_alloc_channels(np);
+	if (err)
+		return err;
+
+	spin_lock_irq(&np->lock);
+
+	err = niu_init_hw(np);
+	if (!err) {
+		timer_setup(&np->timer, niu_timer, 0);
+		np->timer.expires = jiffies + HZ;
+
+		err = niu_enable_interrupts(np, 1);
+		if (err)
+			niu_stop_hw(np);
+	}
+
+	spin_unlock_irq(&np->lock);
+
+	if (!err) {
+		netif_tx_start_all_queues(dev);
+		if (np->link_config.loopback_mode != LOOPBACK_DISABLED)
+			netif_carrier_on(dev);
+
+		add_timer(&np->timer);
+	}
+
+	return err;
+}
+
+static void niu_get_drvinfo(struct net_device *dev,
+			    struct ethtool_drvinfo *info)
+{
+	struct niu *np = netdev_priv(dev);
+	struct niu_vpd *vpd = &np->vpd;
+
+	strscpy(info->driver, DRV_MODULE_NAME, sizeof(info->driver));
+	strscpy(info->version, DRV_MODULE_VERSION, sizeof(info->version));
+	snprintf(info->fw_version, sizeof(info->fw_version), "%d.%d",
+		vpd->fcode_major, vpd->fcode_minor);
+	if (np->parent->plat_type != PLAT_TYPE_NIU)
+		strscpy(info->bus_info, pci_name(np->pdev),
+			sizeof(info->bus_info));
+}
+
+static int niu_get_link_ksettings(struct net_device *dev,
+				  struct ethtool_link_ksettings *cmd)
+{
+	struct niu *np = netdev_priv(dev);
+	struct niu_link_config *lp;
+
+	lp = &np->link_config;
+
+	memset(cmd, 0, sizeof(*cmd));
+	cmd->base.phy_address = np->phy_addr;
+	ethtool_convert_legacy_u32_to_link_mode(cmd->link_modes.supported,
+						lp->supported);
+	ethtool_convert_legacy_u32_to_link_mode(cmd->link_modes.advertising,
+						lp->active_advertising);
+	cmd->base.autoneg = lp->active_autoneg;
+	cmd->base.speed = lp->active_speed;
+	cmd->base.duplex = lp->active_duplex;
+	cmd->base.port = (np->flags & NIU_FLAGS_FIBER) ? PORT_FIBRE : PORT_TP;
+
+	return 0;
+}
+
+static int niu_set_link_ksettings(struct net_device *dev,
+				  const struct ethtool_link_ksettings *cmd)
+{
+	struct niu *np = netdev_priv(dev);
+	struct niu_link_config *lp = &np->link_config;
+
+	ethtool_convert_link_mode_to_legacy_u32(&lp->advertising,
+						cmd->link_modes.advertising);
+	lp->speed = cmd->base.speed;
+	lp->duplex = cmd->base.duplex;
+	lp->autoneg = cmd->base.autoneg;
+	return niu_init_link(np);
+}
+
+static u32 niu_get_msglevel(struct net_device *dev)
+{
+	struct niu *np = netdev_priv(dev);
+	return np->msg_enable;
+}
+
+static void niu_set_msglevel(struct net_device *dev, u32 value)
+{
+	struct niu *np = netdev_priv(dev);
+	np->msg_enable = value;
+}
+
+static int niu_nway_reset(struct net_device *dev)
+{
+	struct niu *np = netdev_priv(dev);
+
+	if (np->link_config.autoneg)
+		return niu_init_link(np);
+
+	return 0;
+}
+
+static int niu_get_eeprom_len(struct net_device *dev)
+{
+	struct niu *np = netdev_priv(dev);
+
+	return np->eeprom_len;
+}
+
+static int niu_get_eeprom(struct net_device *dev,
+			  struct ethtool_eeprom *eeprom, u8 *data)
+{
+	struct niu *np = netdev_priv(dev);
+	u32 offset, len, val;
+
+	offset = eeprom->offset;
+	len = eeprom->len;
+
+	if (offset + len < offset)
+		return -EINVAL;
+	if (offset >= np->eeprom_len)
+		return -EINVAL;
+	if (offset + len > np->eeprom_len)
+		len = eeprom->len = np->eeprom_len - offset;
+
+	if (offset & 3) {
+		u32 b_offset, b_count;
+
+		b_offset = offset & 3;
+		b_count = 4 - b_offset;
+		if (b_count > len)
+			b_count = len;
+
+		val = nr64(ESPC_NCR((offset - b_offset) / 4));
+		memcpy(data, ((char *)&val) + b_offset, b_count);
+		data += b_count;
+		len -= b_count;
+		offset += b_count;
+	}
+	while (len >= 4) {
+		val = nr64(ESPC_NCR(offset / 4));
+		memcpy(data, &val, 4);
+		data += 4;
+		len -= 4;
+		offset += 4;
+	}
+	if (len) {
+		val = nr64(ESPC_NCR(offset / 4));
+		memcpy(data, &val, len);
+	}
+	return 0;
+}
+
+static void niu_ethflow_to_l3proto(int flow_type, u8 *pid)
+{
+	switch (flow_type) {
+	case TCP_V4_FLOW:
+	case TCP_V6_FLOW:
+		*pid = IPPROTO_TCP;
+		break;
+	case UDP_V4_FLOW:
+	case UDP_V6_FLOW:
+		*pid = IPPROTO_UDP;
+		break;
+	case SCTP_V4_FLOW:
+	case SCTP_V6_FLOW:
+		*pid = IPPROTO_SCTP;
+		break;
+	case AH_V4_FLOW:
+	case AH_V6_FLOW:
+		*pid = IPPROTO_AH;
+		break;
+	case ESP_V4_FLOW:
+	case ESP_V6_FLOW:
+		*pid = IPPROTO_ESP;
+		break;
+	default:
+		*pid = 0;
+		break;
+	}
+}
+
+static int niu_class_to_ethflow(u64 class, int *flow_type)
+{
+	switch (class) {
+	case CLASS_CODE_TCP_IPV4:
+		*flow_type = TCP_V4_FLOW;
+		break;
+	case CLASS_CODE_UDP_IPV4:
+		*flow_type = UDP_V4_FLOW;
+		break;
+	case CLASS_CODE_AH_ESP_IPV4:
+		*flow_type = AH_V4_FLOW;
+		break;
+	case CLASS_CODE_SCTP_IPV4:
+		*flow_type = SCTP_V4_FLOW;
+		break;
+	case CLASS_CODE_TCP_IPV6:
+		*flow_type = TCP_V6_FLOW;
+		break;
+	case CLASS_CODE_UDP_IPV6:
+		*flow_type = UDP_V6_FLOW;
+		break;
+	case CLASS_CODE_AH_ESP_IPV6:
+		*flow_type = AH_V6_FLOW;
+		break;
+	case CLASS_CODE_SCTP_IPV6:
+		*flow_type = SCTP_V6_FLOW;
+		break;
+	case CLASS_CODE_USER_PROG1:
+	case CLASS_CODE_USER_PROG2:
+	case CLASS_CODE_USER_PROG3:
+	case CLASS_CODE_USER_PROG4:
+		*flow_type = IP_USER_FLOW;
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
+static int niu_ethflow_to_class(int flow_type, u64 *class)
+{
+	switch (flow_type) {
+	case TCP_V4_FLOW:
+		*class = CLASS_CODE_TCP_IPV4;
+		break;
+	case UDP_V4_FLOW:
+		*class = CLASS_CODE_UDP_IPV4;
+		break;
+	case AH_ESP_V4_FLOW:
+	case AH_V4_FLOW:
+	case ESP_V4_FLOW:
+		*class = CLASS_CODE_AH_ESP_IPV4;
+		break;
+	case SCTP_V4_FLOW:
+		*class = CLASS_CODE_SCTP_IPV4;
+		break;
+	case TCP_V6_FLOW:
+		*class = CLASS_CODE_TCP_IPV6;
+		break;
+	case UDP_V6_FLOW:
+		*class = CLASS_CODE_UDP_IPV6;
+		break;
+	case AH_ESP_V6_FLOW:
+	case AH_V6_FLOW:
+	case ESP_V6_FLOW:
+		*class = CLASS_CODE_AH_ESP_IPV6;
+		break;
+	case SCTP_V6_FLOW:
+		*class = CLASS_CODE_SCTP_IPV6;
+		break;
+	default:
+		return 0;
+	}
+
+	return 1;
+}
+
+static u64 niu_flowkey_to_ethflow(u64 flow_key)
+{
+	u64 ethflow = 0;
+
+	if (flow_key & FLOW_KEY_L2DA)
+		ethflow |= RXH_L2DA;
+	if (flow_key & FLOW_KEY_VLAN)
+		ethflow |= RXH_VLAN;
+	if (flow_key & FLOW_KEY_IPSA)
+		ethflow |= RXH_IP_SRC;
+	if (flow_key & FLOW_KEY_IPDA)
+		ethflow |= RXH_IP_DST;
+	if (flow_key & FLOW_KEY_PROTO)
+		ethflow |= RXH_L3_PROTO;
+	if (flow_key & (FLOW_KEY_L4_BYTE12 << FLOW_KEY_L4_0_SHIFT))
+		ethflow |= RXH_L4_B_0_1;
+	if (flow_key & (FLOW_KEY_L4_BYTE12 << FLOW_KEY_L4_1_SHIFT))
+		ethflow |= RXH_L4_B_2_3;
+
+	return ethflow;
+
+}
+
+static int niu_ethflow_to_flowkey(u64 ethflow, u64 *flow_key)
+{
+	u64 key = 0;
+
+	if (ethflow & RXH_L2DA)
+		key |= FLOW_KEY_L2DA;
+	if (ethflow & RXH_VLAN)
+		key |= FLOW_KEY_VLAN;
+	if (ethflow & RXH_IP_SRC)
+		key |= FLOW_KEY_IPSA;
+	if (ethflow & RXH_IP_DST)
+		key |= FLOW_KEY_IPDA;
+	if (ethflow & RXH_L3_PROTO)
+		key |= FLOW_KEY_PROTO;
+	if (ethflow & RXH_L4_B_0_1)
+		key |= (FLOW_KEY_L4_BYTE12 << FLOW_KEY_L4_0_SHIFT);
+	if (ethflow & RXH_L4_B_2_3)
+		key |= (FLOW_KEY_L4_BYTE12 << FLOW_KEY_L4_1_SHIFT);
+
+	*flow_key = key;
+
+	return 1;
+
+}
+
+static int niu_get_rxfh_fields(struct net_device *dev,
+			       struct ethtool_rxfh_fields *nfc)
+{
+	struct niu *np = netdev_priv(dev);
+	u64 class;
+
+	nfc->data = 0;
+
+	if (!niu_ethflow_to_class(nfc->flow_type, &class))
+		return -EINVAL;
+
+	if (np->parent->tcam_key[class - CLASS_CODE_USER_PROG1] &
+	    TCAM_KEY_DISC)
+		nfc->data = RXH_DISCARD;
+	else
+		nfc->data = niu_flowkey_to_ethflow(np->parent->flow_key[class -
+						      CLASS_CODE_USER_PROG1]);
+	return 0;
+}
+
+static void niu_get_ip4fs_from_tcam_key(struct niu_tcam_entry *tp,
+					struct ethtool_rx_flow_spec *fsp)
+{
+	u32 tmp;
+	u16 prt;
+
+	tmp = (tp->key[3] & TCAM_V4KEY3_SADDR) >> TCAM_V4KEY3_SADDR_SHIFT;
+	fsp->h_u.tcp_ip4_spec.ip4src = cpu_to_be32(tmp);
+
+	tmp = (tp->key[3] & TCAM_V4KEY3_DADDR) >> TCAM_V4KEY3_DADDR_SHIFT;
+	fsp->h_u.tcp_ip4_spec.ip4dst = cpu_to_be32(tmp);
+
+	tmp = (tp->key_mask[3] & TCAM_V4KEY3_SADDR) >> TCAM_V4KEY3_SADDR_SHIFT;
+	fsp->m_u.tcp_ip4_spec.ip4src = cpu_to_be32(tmp);
+
+	tmp = (tp->key_mask[3] & TCAM_V4KEY3_DADDR) >> TCAM_V4KEY3_DADDR_SHIFT;
+	fsp->m_u.tcp_ip4_spec.ip4dst = cpu_to_be32(tmp);
+
+	fsp->h_u.tcp_ip4_spec.tos = (tp->key[2] & TCAM_V4KEY2_TOS) >>
+		TCAM_V4KEY2_TOS_SHIFT;
+	fsp->m_u.tcp_ip4_spec.tos = (tp->key_mask[2] & TCAM_V4KEY2_TOS) >>
+		TCAM_V4KEY2_TOS_SHIFT;
+
+	switch (fsp->flow_type) {
+	case TCP_V4_FLOW:
+	case UDP_V4_FLOW:
+	case SCTP_V4_FLOW:
+		prt = ((tp->key[2] & TCAM_V4KEY2_PORT_SPI) >>
+			TCAM_V4KEY2_PORT_SPI_SHIFT) >> 16;
+		fsp->h_u.tcp_ip4_spec.psrc = cpu_to_be16(prt);
+
+		prt = ((tp->key[2] & TCAM_V4KEY2_PORT_SPI) >>
+			TCAM_V4KEY2_PORT_SPI_SHIFT) & 0xffff;
+		fsp->h_u.tcp_ip4_spec.pdst = cpu_to_be16(prt);
+
+		prt = ((tp->key_mask[2] & TCAM_V4KEY2_PORT_SPI) >>
+			TCAM_V4KEY2_PORT_SPI_SHIFT) >> 16;
+		fsp->m_u.tcp_ip4_spec.psrc = cpu_to_be16(prt);
+
+		prt = ((tp->key_mask[2] & TCAM_V4KEY2_PORT_SPI) >>
+			 TCAM_V4KEY2_PORT_SPI_SHIFT) & 0xffff;
+		fsp->m_u.tcp_ip4_spec.pdst = cpu_to_be16(prt);
+		break;
+	case AH_V4_FLOW:
+	case ESP_V4_FLOW:
+		tmp = (tp->key[2] & TCAM_V4KEY2_PORT_SPI) >>
+			TCAM_V4KEY2_PORT_SPI_SHIFT;
+		fsp->h_u.ah_ip4_spec.spi = cpu_to_be32(tmp);
+
+		tmp = (tp->key_mask[2] & TCAM_V4KEY2_PORT_SPI) >>
+			TCAM_V4KEY2_PORT_SPI_SHIFT;
+		fsp->m_u.ah_ip4_spec.spi = cpu_to_be32(tmp);
+		break;
+	case IP_USER_FLOW:
+		tmp = (tp->key[2] & TCAM_V4KEY2_PORT_SPI) >>
+			TCAM_V4KEY2_PORT_SPI_SHIFT;
+		fsp->h_u.usr_ip4_spec.l4_4_bytes = cpu_to_be32(tmp);
+
+		tmp = (tp->key_mask[2] & TCAM_V4KEY2_PORT_SPI) >>
+			TCAM_V4KEY2_PORT_SPI_SHIFT;
+		fsp->m_u.usr_ip4_spec.l4_4_bytes = cpu_to_be32(tmp);
+
+		fsp->h_u.usr_ip4_spec.proto =
+			(tp->key[2] & TCAM_V4KEY2_PROTO) >>
+			TCAM_V4KEY2_PROTO_SHIFT;
+		fsp->m_u.usr_ip4_spec.proto =
+			(tp->key_mask[2] & TCAM_V4KEY2_PROTO) >>
+			TCAM_V4KEY2_PROTO_SHIFT;
+
+		fsp->h_u.usr_ip4_spec.ip_ver = ETH_RX_NFC_IP4;
+		break;
+	default:
+		break;
+	}
+}
+
+static int niu_get_ethtool_tcam_entry(struct niu *np,
+				      struct ethtool_rxnfc *nfc)
+{
+	struct niu_parent *parent = np->parent;
+	struct niu_tcam_entry *tp;
+	struct ethtool_rx_flow_spec *fsp = &nfc->fs;
+	u16 idx;
+	u64 class;
+	int ret = 0;
+
+	idx = tcam_get_index(np, (u16)nfc->fs.location);
+
+	tp = &parent->tcam[idx];
+	if (!tp->valid) {
+		netdev_info(np->dev, "niu%d: entry [%d] invalid for idx[%d]\n",
+			    parent->index, (u16)nfc->fs.location, idx);
+		return -EINVAL;
+	}
+
+	/* fill the flow spec entry */
+	class = (tp->key[0] & TCAM_V4KEY0_CLASS_CODE) >>
+		TCAM_V4KEY0_CLASS_CODE_SHIFT;
+	ret = niu_class_to_ethflow(class, &fsp->flow_type);
+	if (ret < 0) {
+		netdev_info(np->dev, "niu%d: niu_class_to_ethflow failed\n",
+			    parent->index);
+		goto out;
+	}
+
+	if (fsp->flow_type == AH_V4_FLOW || fsp->flow_type == AH_V6_FLOW) {
+		u32 proto = (tp->key[2] & TCAM_V4KEY2_PROTO) >>
+			TCAM_V4KEY2_PROTO_SHIFT;
+		if (proto == IPPROTO_ESP) {
+			if (fsp->flow_type == AH_V4_FLOW)
+				fsp->flow_type = ESP_V4_FLOW;
+			else
+				fsp->flow_type = ESP_V6_FLOW;
+		}
+	}
+
+	switch (fsp->flow_type) {
+	case TCP_V4_FLOW:
+	case UDP_V4_FLOW:
+	case SCTP_V4_FLOW:
+	case AH_V4_FLOW:
+	case ESP_V4_FLOW:
+		niu_get_ip4fs_from_tcam_key(tp, fsp);
+		break;
+	case TCP_V6_FLOW:
+	case UDP_V6_FLOW:
+	case SCTP_V6_FLOW:
+	case AH_V6_FLOW:
+	case ESP_V6_FLOW:
+		/* Not yet implemented */
+		ret = -EINVAL;
+		break;
+	case IP_USER_FLOW:
+		niu_get_ip4fs_from_tcam_key(tp, fsp);
+		break;
+	default:
+		ret = -EINVAL;
+		break;
+	}
+
+	if (ret < 0)
+		goto out;
+
+	if (tp->assoc_data & TCAM_ASSOCDATA_DISC)
+		fsp->ring_cookie = RX_CLS_FLOW_DISC;
+	else
+		fsp->ring_cookie = (tp->assoc_data & TCAM_ASSOCDATA_OFFSET) >>
+			TCAM_ASSOCDATA_OFFSET_SHIFT;
+
+	/* put the tcam size here */
+	nfc->data = tcam_get_size(np);
+out:
+	return ret;
+}
+
+static int niu_get_ethtool_tcam_all(struct niu *np,
+				    struct ethtool_rxnfc *nfc,
+				    u32 *rule_locs)
+{
+	struct niu_parent *parent = np->parent;
+	struct niu_tcam_entry *tp;
+	int i, idx, cnt;
+	unsigned long flags;
+	int ret = 0;
+
+	/* put the tcam size here */
+	nfc->data = tcam_get_size(np);
+
+	niu_lock_parent(np, flags);
+	for (cnt = 0, i = 0; i < nfc->data; i++) {
+		idx = tcam_get_index(np, i);
+		tp = &parent->tcam[idx];
+		if (!tp->valid)
+			continue;
+		if (cnt == nfc->rule_cnt) {
+			ret = -EMSGSIZE;
+			break;
+		}
+		rule_locs[cnt] = i;
+		cnt++;
+	}
+	niu_unlock_parent(np, flags);
+
+	nfc->rule_cnt = cnt;
+
+	return ret;
+}
+
+static u32 niu_get_rx_ring_count(struct net_device *dev)
+{
+	struct niu *np = netdev_priv(dev);
+
+	return np->num_rx_rings;
+}
+
+static int niu_get_nfc(struct net_device *dev, struct ethtool_rxnfc *cmd,
+		       u32 *rule_locs)
+{
+	struct niu *np = netdev_priv(dev);
+	int ret = 0;
+
+	switch (cmd->cmd) {
+	case ETHTOOL_GRXCLSRLCNT:
+		cmd->rule_cnt = tcam_get_valid_entry_cnt(np);
+		break;
+	case ETHTOOL_GRXCLSRULE:
+		ret = niu_get_ethtool_tcam_entry(np, cmd);
+		break;
+	case ETHTOOL_GRXCLSRLALL:
+		ret = niu_get_ethtool_tcam_all(np, cmd, rule_locs);
+		break;
+	default:
+		ret = -EINVAL;
+		break;
+	}
+
+	return ret;
+}
+
+static int niu_set_rxfh_fields(struct net_device *dev,
+			       const struct ethtool_rxfh_fields *nfc,
+			       struct netlink_ext_ack *extack)
+{
+	struct niu *np = netdev_priv(dev);
+	u64 class;
+	u64 flow_key = 0;
+	unsigned long flags;
+
+	if (!niu_ethflow_to_class(nfc->flow_type, &class))
+		return -EINVAL;
+
+	if (class < CLASS_CODE_USER_PROG1 ||
+	    class > CLASS_CODE_SCTP_IPV6)
+		return -EINVAL;
+
+	if (nfc->data & RXH_DISCARD) {
+		niu_lock_parent(np, flags);
+		flow_key = np->parent->tcam_key[class -
+					       CLASS_CODE_USER_PROG1];
+		flow_key |= TCAM_KEY_DISC;
+		nw64(TCAM_KEY(class - CLASS_CODE_USER_PROG1), flow_key);
+		np->parent->tcam_key[class - CLASS_CODE_USER_PROG1] = flow_key;
+		niu_unlock_parent(np, flags);
+		return 0;
+	} else {
+		/* Discard was set before, but is not set now */
+		if (np->parent->tcam_key[class - CLASS_CODE_USER_PROG1] &
+		    TCAM_KEY_DISC) {
+			niu_lock_parent(np, flags);
+			flow_key = np->parent->tcam_key[class -
+					       CLASS_CODE_USER_PROG1];
+			flow_key &= ~TCAM_KEY_DISC;
+			nw64(TCAM_KEY(class - CLASS_CODE_USER_PROG1),
+			     flow_key);
+			np->parent->tcam_key[class - CLASS_CODE_USER_PROG1] =
+				flow_key;
+			niu_unlock_parent(np, flags);
+		}
+	}
+
+	if (!niu_ethflow_to_flowkey(nfc->data, &flow_key))
+		return -EINVAL;
+
+	niu_lock_parent(np, flags);
+	nw64(FLOW_KEY(class - CLASS_CODE_USER_PROG1), flow_key);
+	np->parent->flow_key[class - CLASS_CODE_USER_PROG1] = flow_key;
+	niu_unlock_parent(np, flags);
+
+	return 0;
+}
+
+static void niu_get_tcamkey_from_ip4fs(struct ethtool_rx_flow_spec *fsp,
+				       struct niu_tcam_entry *tp,
+				       int l2_rdc_tab, u64 class)
+{
+	u8 pid = 0;
+	u32 sip, dip, sipm, dipm, spi, spim;
+	u16 sport, dport, spm, dpm;
+
+	sip = be32_to_cpu(fsp->h_u.tcp_ip4_spec.ip4src);
+	sipm = be32_to_cpu(fsp->m_u.tcp_ip4_spec.ip4src);
+	dip = be32_to_cpu(fsp->h_u.tcp_ip4_spec.ip4dst);
+	dipm = be32_to_cpu(fsp->m_u.tcp_ip4_spec.ip4dst);
+
+	tp->key[0] = class << TCAM_V4KEY0_CLASS_CODE_SHIFT;
+	tp->key_mask[0] = TCAM_V4KEY0_CLASS_CODE;
+	tp->key[1] = (u64)l2_rdc_tab << TCAM_V4KEY1_L2RDCNUM_SHIFT;
+	tp->key_mask[1] = TCAM_V4KEY1_L2RDCNUM;
+
+	tp->key[3] = (u64)sip << TCAM_V4KEY3_SADDR_SHIFT;
+	tp->key[3] |= dip;
+
+	tp->key_mask[3] = (u64)sipm << TCAM_V4KEY3_SADDR_SHIFT;
+	tp->key_mask[3] |= dipm;
+
+	tp->key[2] |= ((u64)fsp->h_u.tcp_ip4_spec.tos <<
+		       TCAM_V4KEY2_TOS_SHIFT);
+	tp->key_mask[2] |= ((u64)fsp->m_u.tcp_ip4_spec.tos <<
+			    TCAM_V4KEY2_TOS_SHIFT);
+	switch (fsp->flow_type) {
+	case TCP_V4_FLOW:
+	case UDP_V4_FLOW:
+	case SCTP_V4_FLOW:
+		sport = be16_to_cpu(fsp->h_u.tcp_ip4_spec.psrc);
+		spm = be16_to_cpu(fsp->m_u.tcp_ip4_spec.psrc);
+		dport = be16_to_cpu(fsp->h_u.tcp_ip4_spec.pdst);
+		dpm = be16_to_cpu(fsp->m_u.tcp_ip4_spec.pdst);
+
+		tp->key[2] |= (((u64)sport << 16) | dport);
+		tp->key_mask[2] |= (((u64)spm << 16) | dpm);
+		niu_ethflow_to_l3proto(fsp->flow_type, &pid);
+		break;
+	case AH_V4_FLOW:
+	case ESP_V4_FLOW:
+		spi = be32_to_cpu(fsp->h_u.ah_ip4_spec.spi);
+		spim = be32_to_cpu(fsp->m_u.ah_ip4_spec.spi);
+
+		tp->key[2] |= spi;
+		tp->key_mask[2] |= spim;
+		niu_ethflow_to_l3proto(fsp->flow_type, &pid);
+		break;
+	case IP_USER_FLOW:
+		spi = be32_to_cpu(fsp->h_u.usr_ip4_spec.l4_4_bytes);
+		spim = be32_to_cpu(fsp->m_u.usr_ip4_spec.l4_4_bytes);
+
+		tp->key[2] |= spi;
+		tp->key_mask[2] |= spim;
+		pid = fsp->h_u.usr_ip4_spec.proto;
+		break;
+	default:
+		break;
+	}
+
+	tp->key[2] |= ((u64)pid << TCAM_V4KEY2_PROTO_SHIFT);
+	if (pid) {
+		tp->key_mask[2] |= TCAM_V4KEY2_PROTO;
+	}
+}
+
+static int niu_add_ethtool_tcam_entry(struct niu *np,
+				      struct ethtool_rxnfc *nfc)
+{
+	struct niu_parent *parent = np->parent;
+	struct niu_tcam_entry *tp;
+	struct ethtool_rx_flow_spec *fsp = &nfc->fs;
+	struct niu_rdc_tables *rdc_table = &parent->rdc_group_cfg[np->port];
+	int l2_rdc_table = rdc_table->first_table_num;
+	u16 idx;
+	u64 class;
+	unsigned long flags;
+	int err, ret;
+
+	ret = 0;
+
+	idx = nfc->fs.location;
+	if (idx >= tcam_get_size(np))
+		return -EINVAL;
+
+	if (fsp->flow_type == IP_USER_FLOW) {
+		int i;
+		int add_usr_cls = 0;
+		struct ethtool_usrip4_spec *uspec = &fsp->h_u.usr_ip4_spec;
+		struct ethtool_usrip4_spec *umask = &fsp->m_u.usr_ip4_spec;
+
+		if (uspec->ip_ver != ETH_RX_NFC_IP4)
+			return -EINVAL;
+
+		niu_lock_parent(np, flags);
+
+		for (i = 0; i < NIU_L3_PROG_CLS; i++) {
+			if (parent->l3_cls[i]) {
+				if (uspec->proto == parent->l3_cls_pid[i]) {
+					class = parent->l3_cls[i];
+					parent->l3_cls_refcnt[i]++;
+					add_usr_cls = 1;
+					break;
+				}
+			} else {
+				/* Program new user IP class */
+				switch (i) {
+				case 0:
+					class = CLASS_CODE_USER_PROG1;
+					break;
+				case 1:
+					class = CLASS_CODE_USER_PROG2;
+					break;
+				case 2:
+					class = CLASS_CODE_USER_PROG3;
+					break;
+				case 3:
+					class = CLASS_CODE_USER_PROG4;
+					break;
+				default:
+					class = CLASS_CODE_UNRECOG;
+					break;
+				}
+				ret = tcam_user_ip_class_set(np, class, 0,
+							     uspec->proto,
+							     uspec->tos,
+							     umask->tos);
+				if (ret)
+					goto out;
+
+				ret = tcam_user_ip_class_enable(np, class, 1);
+				if (ret)
+					goto out;
+				parent->l3_cls[i] = class;
+				parent->l3_cls_pid[i] = uspec->proto;
+				parent->l3_cls_refcnt[i]++;
+				add_usr_cls = 1;
+				break;
+			}
+		}
+		if (!add_usr_cls) {
+			netdev_info(np->dev, "niu%d: %s(): Could not find/insert class for pid %d\n",
+				    parent->index, __func__, uspec->proto);
+			ret = -EINVAL;
+			goto out;
+		}
+		niu_unlock_parent(np, flags);
+	} else {
+		if (!niu_ethflow_to_class(fsp->flow_type, &class)) {
+			return -EINVAL;
+		}
+	}
+
+	niu_lock_parent(np, flags);
+
+	idx = tcam_get_index(np, idx);
+	tp = &parent->tcam[idx];
+
+	memset(tp, 0, sizeof(*tp));
+
+	/* fill in the tcam key and mask */
+	switch (fsp->flow_type) {
+	case TCP_V4_FLOW:
+	case UDP_V4_FLOW:
+	case SCTP_V4_FLOW:
+	case AH_V4_FLOW:
+	case ESP_V4_FLOW:
+		niu_get_tcamkey_from_ip4fs(fsp, tp, l2_rdc_table, class);
+		break;
+	case TCP_V6_FLOW:
+	case UDP_V6_FLOW:
+	case SCTP_V6_FLOW:
+	case AH_V6_FLOW:
+	case ESP_V6_FLOW:
+		/* Not yet implemented */
+		netdev_info(np->dev, "niu%d: In %s(): flow %d for IPv6 not implemented\n",
+			    parent->index, __func__, fsp->flow_type);
+		ret = -EINVAL;
+		goto out;
+	case IP_USER_FLOW:
+		niu_get_tcamkey_from_ip4fs(fsp, tp, l2_rdc_table, class);
+		break;
+	default:
+		netdev_info(np->dev, "niu%d: In %s(): Unknown flow type %d\n",
+			    parent->index, __func__, fsp->flow_type);
+		ret = -EINVAL;
+		goto out;
+	}
+
+	/* fill in the assoc data */
+	if (fsp->ring_cookie == RX_CLS_FLOW_DISC) {
+		tp->assoc_data = TCAM_ASSOCDATA_DISC;
+	} else {
+		if (fsp->ring_cookie >= np->num_rx_rings) {
+			netdev_info(np->dev, "niu%d: In %s(): Invalid RX ring %lld\n",
+				    parent->index, __func__,
+				    (long long)fsp->ring_cookie);
+			ret = -EINVAL;
+			goto out;
+		}
+		tp->assoc_data = (TCAM_ASSOCDATA_TRES_USE_OFFSET |
+				  (fsp->ring_cookie <<
+				   TCAM_ASSOCDATA_OFFSET_SHIFT));
+	}
+
+	err = tcam_write(np, idx, tp->key, tp->key_mask);
+	if (err) {
+		ret = -EINVAL;
+		goto out;
+	}
+	err = tcam_assoc_write(np, idx, tp->assoc_data);
+	if (err) {
+		ret = -EINVAL;
+		goto out;
+	}
+
+	/* validate the entry */
+	tp->valid = 1;
+	np->clas.tcam_valid_entries++;
+out:
+	niu_unlock_parent(np, flags);
+
+	return ret;
+}
+
+static int niu_del_ethtool_tcam_entry(struct niu *np, u32 loc)
+{
+	struct niu_parent *parent = np->parent;
+	struct niu_tcam_entry *tp;
+	u16 idx;
+	unsigned long flags;
+	u64 class;
+	int ret = 0;
+
+	if (loc >= tcam_get_size(np))
+		return -EINVAL;
+
+	niu_lock_parent(np, flags);
+
+	idx = tcam_get_index(np, loc);
+	tp = &parent->tcam[idx];
+
+	/* if the entry is of a user defined class, then update*/
+	class = (tp->key[0] & TCAM_V4KEY0_CLASS_CODE) >>
+		TCAM_V4KEY0_CLASS_CODE_SHIFT;
+
+	if (class >= CLASS_CODE_USER_PROG1 && class <= CLASS_CODE_USER_PROG4) {
+		int i;
+		for (i = 0; i < NIU_L3_PROG_CLS; i++) {
+			if (parent->l3_cls[i] == class) {
+				parent->l3_cls_refcnt[i]--;
+				if (!parent->l3_cls_refcnt[i]) {
+					/* disable class */
+					ret = tcam_user_ip_class_enable(np,
+									class,
+									0);
+					if (ret)
+						goto out;
+					parent->l3_cls[i] = 0;
+					parent->l3_cls_pid[i] = 0;
+				}
+				break;
+			}
+		}
+		if (i == NIU_L3_PROG_CLS) {
+			netdev_info(np->dev, "niu%d: In %s(): Usr class 0x%llx not found\n",
+				    parent->index, __func__,
+				    (unsigned long long)class);
+			ret = -EINVAL;
+			goto out;
+		}
+	}
+
+	ret = tcam_flush(np, idx);
+	if (ret)
+		goto out;
+
+	/* invalidate the entry */
+	tp->valid = 0;
+	np->clas.tcam_valid_entries--;
+out:
+	niu_unlock_parent(np, flags);
+
+	return ret;
+}
+
+static int niu_set_nfc(struct net_device *dev, struct ethtool_rxnfc *cmd)
+{
+	struct niu *np = netdev_priv(dev);
+	int ret = 0;
+
+	switch (cmd->cmd) {
+	case ETHTOOL_SRXCLSRLINS:
+		ret = niu_add_ethtool_tcam_entry(np, cmd);
+		break;
+	case ETHTOOL_SRXCLSRLDEL:
+		ret = niu_del_ethtool_tcam_entry(np, cmd->fs.location);
+		break;
+	default:
+		ret = -EINVAL;
+		break;
+	}
+
+	return ret;
+}
+
+static const struct {
+	const char string[ETH_GSTRING_LEN];
+} niu_xmac_stat_keys[] = {
+	{ "tx_frames" },
+	{ "tx_bytes" },
+	{ "tx_fifo_errors" },
+	{ "tx_overflow_errors" },
+	{ "tx_max_pkt_size_errors" },
+	{ "tx_underflow_errors" },
+	{ "rx_local_faults" },
+	{ "rx_remote_faults" },
+	{ "rx_link_faults" },
+	{ "rx_align_errors" },
+	{ "rx_frags" },
+	{ "rx_mcasts" },
+	{ "rx_bcasts" },
+	{ "rx_hist_cnt1" },
+	{ "rx_hist_cnt2" },
+	{ "rx_hist_cnt3" },
+	{ "rx_hist_cnt4" },
+	{ "rx_hist_cnt5" },
+	{ "rx_hist_cnt6" },
+	{ "rx_hist_cnt7" },
+	{ "rx_octets" },
+	{ "rx_code_violations" },
+	{ "rx_len_errors" },
+	{ "rx_crc_errors" },
+	{ "rx_underflows" },
+	{ "rx_overflows" },
+	{ "pause_off_state" },
+	{ "pause_on_state" },
+	{ "pause_received" },
+};
+
+#define NUM_XMAC_STAT_KEYS	ARRAY_SIZE(niu_xmac_stat_keys)
+
+static const struct {
+	const char string[ETH_GSTRING_LEN];
+} niu_bmac_stat_keys[] = {
+	{ "tx_underflow_errors" },
+	{ "tx_max_pkt_size_errors" },
+	{ "tx_bytes" },
+	{ "tx_frames" },
+	{ "rx_overflows" },
+	{ "rx_frames" },
+	{ "rx_align_errors" },
+	{ "rx_crc_errors" },
+	{ "rx_len_errors" },
+	{ "pause_off_state" },
+	{ "pause_on_state" },
+	{ "pause_received" },
+};
+
+#define NUM_BMAC_STAT_KEYS	ARRAY_SIZE(niu_bmac_stat_keys)
+
+static const struct {
+	const char string[ETH_GSTRING_LEN];
+} niu_rxchan_stat_keys[] = {
+	{ "rx_channel" },
+	{ "rx_packets" },
+	{ "rx_bytes" },
+	{ "rx_dropped" },
+	{ "rx_errors" },
+};
+
+#define NUM_RXCHAN_STAT_KEYS	ARRAY_SIZE(niu_rxchan_stat_keys)
+
+static const struct {
+	const char string[ETH_GSTRING_LEN];
+} niu_txchan_stat_keys[] = {
+	{ "tx_channel" },
+	{ "tx_packets" },
+	{ "tx_bytes" },
+	{ "tx_errors" },
+};
+
+#define NUM_TXCHAN_STAT_KEYS	ARRAY_SIZE(niu_txchan_stat_keys)
+
+static void niu_get_strings(struct net_device *dev, u32 stringset, u8 *data)
+{
+	struct niu *np = netdev_priv(dev);
+	int i;
+
+	if (stringset != ETH_SS_STATS)
+		return;
+
+	if (np->flags & NIU_FLAGS_XMAC) {
+		memcpy(data, niu_xmac_stat_keys,
+		       sizeof(niu_xmac_stat_keys));
+		data += sizeof(niu_xmac_stat_keys);
+	} else {
+		memcpy(data, niu_bmac_stat_keys,
+		       sizeof(niu_bmac_stat_keys));
+		data += sizeof(niu_bmac_stat_keys);
+	}
+	for (i = 0; i < np->num_rx_rings; i++) {
+		memcpy(data, niu_rxchan_stat_keys,
+		       sizeof(niu_rxchan_stat_keys));
+		data += sizeof(niu_rxchan_stat_keys);
+	}
+	for (i = 0; i < np->num_tx_rings; i++) {
+		memcpy(data, niu_txchan_stat_keys,
+		       sizeof(niu_txchan_stat_keys));
+		data += sizeof(niu_txchan_stat_keys);
+	}
+}
+
+static int niu_get_sset_count(struct net_device *dev, int stringset)
+{
+	struct niu *np = netdev_priv(dev);
+
+	if (stringset != ETH_SS_STATS)
+		return -EINVAL;
+
+	return (np->flags & NIU_FLAGS_XMAC ?
+		 NUM_XMAC_STAT_KEYS :
+		 NUM_BMAC_STAT_KEYS) +
+		(np->num_rx_rings * NUM_RXCHAN_STAT_KEYS) +
+		(np->num_tx_rings * NUM_TXCHAN_STAT_KEYS);
+}
+
+static void niu_get_ethtool_stats(struct net_device *dev,
+				  struct ethtool_stats *stats, u64 *data)
+{
+	struct niu *np = netdev_priv(dev);
+	int i;
+
+	niu_sync_mac_stats(np);
+	if (np->flags & NIU_FLAGS_XMAC) {
+		memcpy(data, &np->mac_stats.xmac,
+		       sizeof(struct niu_xmac_stats));
+		data += (sizeof(struct niu_xmac_stats) / sizeof(u64));
+	} else {
+		memcpy(data, &np->mac_stats.bmac,
+		       sizeof(struct niu_bmac_stats));
+		data += (sizeof(struct niu_bmac_stats) / sizeof(u64));
+	}
+	for (i = 0; i < np->num_rx_rings; i++) {
+		struct rx_ring_info *rp = &np->rx_rings[i];
+
+		niu_sync_rx_discard_stats(np, rp, 0);
+
+		data[0] = rp->rx_channel;
+		data[1] = rp->rx_packets;
+		data[2] = rp->rx_bytes;
+		data[3] = rp->rx_dropped;
+		data[4] = rp->rx_errors;
+		data += 5;
+	}
+	for (i = 0; i < np->num_tx_rings; i++) {
+		struct tx_ring_info *rp = &np->tx_rings[i];
+
+		data[0] = rp->tx_channel;
+		data[1] = rp->tx_packets;
+		data[2] = rp->tx_bytes;
+		data[3] = rp->tx_errors;
+		data += 4;
+	}
+}
+
+static u64 niu_led_state_save(struct niu *np)
+{
+	if (np->flags & NIU_FLAGS_XMAC)
+		return nr64_mac(XMAC_CONFIG);
+	else
+		return nr64_mac(BMAC_XIF_CONFIG);
+}
+
+static void niu_led_state_restore(struct niu *np, u64 val)
+{
+	if (np->flags & NIU_FLAGS_XMAC)
+		nw64_mac(XMAC_CONFIG, val);
+	else
+		nw64_mac(BMAC_XIF_CONFIG, val);
+}
+
+static void niu_force_led(struct niu *np, int on)
+{
+	u64 val, reg, bit;
+
+	if (np->flags & NIU_FLAGS_XMAC) {
+		reg = XMAC_CONFIG;
+		bit = XMAC_CONFIG_FORCE_LED_ON;
+	} else {
+		reg = BMAC_XIF_CONFIG;
+		bit = BMAC_XIF_CONFIG_LINK_LED;
+	}
+
+	val = nr64_mac(reg);
+	if (on)
+		val |= bit;
+	else
+		val &= ~bit;
+	nw64_mac(reg, val);
+}
+
+static int niu_set_phys_id(struct net_device *dev,
+			   enum ethtool_phys_id_state state)
+
+{
+	struct niu *np = netdev_priv(dev);
+
+	if (!netif_running(dev))
+		return -EAGAIN;
+
+	switch (state) {
+	case ETHTOOL_ID_ACTIVE:
+		np->orig_led_state = niu_led_state_save(np);
+		return 1;	/* cycle on/off once per second */
+
+	case ETHTOOL_ID_ON:
+		niu_force_led(np, 1);
+		break;
+
+	case ETHTOOL_ID_OFF:
+		niu_force_led(np, 0);
+		break;
+
+	case ETHTOOL_ID_INACTIVE:
+		niu_led_state_restore(np, np->orig_led_state);
+	}
+
+	return 0;
+}
+
+static const struct ethtool_ops niu_ethtool_ops = {
+	.get_drvinfo		= niu_get_drvinfo,
+	.get_link		= ethtool_op_get_link,
+	.get_msglevel		= niu_get_msglevel,
+	.set_msglevel		= niu_set_msglevel,
+	.nway_reset		= niu_nway_reset,
+	.get_eeprom_len		= niu_get_eeprom_len,
+	.get_eeprom		= niu_get_eeprom,
+	.get_strings		= niu_get_strings,
+	.get_sset_count		= niu_get_sset_count,
+	.get_ethtool_stats	= niu_get_ethtool_stats,
+	.set_phys_id		= niu_set_phys_id,
+	.get_rxnfc		= niu_get_nfc,
+	.set_rxnfc		= niu_set_nfc,
+	.get_rx_ring_count	= niu_get_rx_ring_count,
+	.get_rxfh_fields	= niu_get_rxfh_fields,
+	.set_rxfh_fields	= niu_set_rxfh_fields,
+	.get_link_ksettings	= niu_get_link_ksettings,
+	.set_link_ksettings	= niu_set_link_ksettings,
+};
+
+static int niu_ldg_assign_ldn(struct niu *np, struct niu_parent *parent,
+			      int ldg, int ldn)
+{
+	if (ldg < NIU_LDG_MIN || ldg > NIU_LDG_MAX)
+		return -EINVAL;
+	if (ldn < 0 || ldn > LDN_MAX)
+		return -EINVAL;
+
+	parent->ldg_map[ldn] = ldg;
+
+	if (np->parent->plat_type == PLAT_TYPE_NIU) {
+		/* On N2 NIU, the ldn-->ldg assignments are setup and fixed by
+		 * the firmware, and we're not supposed to change them.
+		 * Validate the mapping, because if it's wrong we probably
+		 * won't get any interrupts and that's painful to debug.
+		 */
+		if (nr64(LDG_NUM(ldn)) != ldg) {
+			dev_err(np->device, "Port %u, mismatched LDG assignment for ldn %d, should be %d is %llu\n",
+				np->port, ldn, ldg,
+				(unsigned long long) nr64(LDG_NUM(ldn)));
+			return -EINVAL;
+		}
+	} else
+		nw64(LDG_NUM(ldn), ldg);
+
+	return 0;
+}
+
+static int niu_set_ldg_timer_res(struct niu *np, int res)
+{
+	if (res < 0 || res > LDG_TIMER_RES_VAL)
+		return -EINVAL;
+
+
+	nw64(LDG_TIMER_RES, res);
+
+	return 0;
+}
+
+static int niu_set_ldg_sid(struct niu *np, int ldg, int func, int vector)
+{
+	if ((ldg < NIU_LDG_MIN || ldg > NIU_LDG_MAX) ||
+	    (func < 0 || func > 3) ||
+	    (vector < 0 || vector > 0x1f))
+		return -EINVAL;
+
+	nw64(SID(ldg), (func << SID_FUNC_SHIFT) | vector);
+
+	return 0;
+}
+
+static int niu_pci_eeprom_read(struct niu *np, u32 addr)
+{
+	u64 frame, frame_base = (ESPC_PIO_STAT_READ_START |
+				 (addr << ESPC_PIO_STAT_ADDR_SHIFT));
+	int limit;
+
+	if (addr > (ESPC_PIO_STAT_ADDR >> ESPC_PIO_STAT_ADDR_SHIFT))
+		return -EINVAL;
+
+	frame = frame_base;
+	nw64(ESPC_PIO_STAT, frame);
+	limit = 64;
+	do {
+		udelay(5);
+		frame = nr64(ESPC_PIO_STAT);
+		if (frame & ESPC_PIO_STAT_READ_END)
+			break;
+	} while (limit--);
+	if (!(frame & ESPC_PIO_STAT_READ_END)) {
+		dev_err(np->device, "EEPROM read timeout frame[%llx]\n",
+			(unsigned long long) frame);
+		return -ENODEV;
+	}
+
+	frame = frame_base;
+	nw64(ESPC_PIO_STAT, frame);
+	limit = 64;
+	do {
+		udelay(5);
+		frame = nr64(ESPC_PIO_STAT);
+		if (frame & ESPC_PIO_STAT_READ_END)
+			break;
+	} while (limit--);
+	if (!(frame & ESPC_PIO_STAT_READ_END)) {
+		dev_err(np->device, "EEPROM read timeout frame[%llx]\n",
+			(unsigned long long) frame);
+		return -ENODEV;
+	}
+
+	frame = nr64(ESPC_PIO_STAT);
+	return (frame & ESPC_PIO_STAT_DATA) >> ESPC_PIO_STAT_DATA_SHIFT;
+}
+
+static int niu_pci_eeprom_read16(struct niu *np, u32 off)
+{
+	int err = niu_pci_eeprom_read(np, off);
+	u16 val;
+
+	if (err < 0)
+		return err;
+	val = (err << 8);
+	err = niu_pci_eeprom_read(np, off + 1);
+	if (err < 0)
+		return err;
+	val |= (err & 0xff);
+
+	return val;
+}
+
+static int niu_pci_eeprom_read16_swp(struct niu *np, u32 off)
+{
+	int err = niu_pci_eeprom_read(np, off);
+	u16 val;
+
+	if (err < 0)
+		return err;
+
+	val = (err & 0xff);
+	err = niu_pci_eeprom_read(np, off + 1);
+	if (err < 0)
+		return err;
+
+	val |= (err & 0xff) << 8;
+
+	return val;
+}
+
+static int niu_pci_vpd_get_propname(struct niu *np, u32 off, char *namebuf,
+				    int namebuf_len)
+{
+	int i;
+
+	for (i = 0; i < namebuf_len; i++) {
+		int err = niu_pci_eeprom_read(np, off + i);
+		if (err < 0)
+			return err;
+		*namebuf++ = err;
+		if (!err)
+			break;
+	}
+	if (i >= namebuf_len)
+		return -EINVAL;
+
+	return i + 1;
+}
+
+static void niu_vpd_parse_version(struct niu *np)
+{
+	struct niu_vpd *vpd = &np->vpd;
+	int len = strlen(vpd->version) + 1;
+	const char *s = vpd->version;
+	int i;
+
+	for (i = 0; i < len - 5; i++) {
+		if (!strncmp(s + i, "FCode ", 6))
+			break;
+	}
+	if (i >= len - 5)
+		return;
+
+	s += i + 5;
+	sscanf(s, "%d.%d", &vpd->fcode_major, &vpd->fcode_minor);
+
+	netif_printk(np, probe, KERN_DEBUG, np->dev,
+		     "VPD_SCAN: FCODE major(%d) minor(%d)\n",
+		     vpd->fcode_major, vpd->fcode_minor);
+	if (vpd->fcode_major > NIU_VPD_MIN_MAJOR ||
+	    (vpd->fcode_major == NIU_VPD_MIN_MAJOR &&
+	     vpd->fcode_minor >= NIU_VPD_MIN_MINOR))
+		np->flags |= NIU_FLAGS_VPD_VALID;
+}
+
+/* ESPC_PIO_EN_ENABLE must be set */
+static int niu_pci_vpd_scan_props(struct niu *np, u32 start, u32 end)
+{
+	unsigned int found_mask = 0;
+#define FOUND_MASK_MODEL	0x00000001
+#define FOUND_MASK_BMODEL	0x00000002
+#define FOUND_MASK_VERS		0x00000004
+#define FOUND_MASK_MAC		0x00000008
+#define FOUND_MASK_NMAC		0x00000010
+#define FOUND_MASK_PHY		0x00000020
+#define FOUND_MASK_ALL		0x0000003f
+
+	netif_printk(np, probe, KERN_DEBUG, np->dev,
+		     "VPD_SCAN: start[%x] end[%x]\n", start, end);
+	while (start < end) {
+		int len, err, prop_len;
+		char namebuf[64];
+		u8 *prop_buf;
+		int max_len;
+
+		if (found_mask == FOUND_MASK_ALL) {
+			niu_vpd_parse_version(np);
+			return 1;
+		}
+
+		err = niu_pci_eeprom_read(np, start + 2);
+		if (err < 0)
+			return err;
+		len = err;
+		start += 3;
+
+		prop_len = niu_pci_eeprom_read(np, start + 4);
+		if (prop_len < 0)
+			return prop_len;
+		err = niu_pci_vpd_get_propname(np, start + 5, namebuf, 64);
+		if (err < 0)
+			return err;
+
+		prop_buf = NULL;
+		max_len = 0;
+		if (!strcmp(namebuf, "model")) {
+			prop_buf = np->vpd.model;
+			max_len = NIU_VPD_MODEL_MAX;
+			found_mask |= FOUND_MASK_MODEL;
+		} else if (!strcmp(namebuf, "board-model")) {
+			prop_buf = np->vpd.board_model;
+			max_len = NIU_VPD_BD_MODEL_MAX;
+			found_mask |= FOUND_MASK_BMODEL;
+		} else if (!strcmp(namebuf, "version")) {
+			prop_buf = np->vpd.version;
+			max_len = NIU_VPD_VERSION_MAX;
+			found_mask |= FOUND_MASK_VERS;
+		} else if (!strcmp(namebuf, "local-mac-address")) {
+			prop_buf = np->vpd.local_mac;
+			max_len = ETH_ALEN;
+			found_mask |= FOUND_MASK_MAC;
+		} else if (!strcmp(namebuf, "num-mac-addresses")) {
+			prop_buf = &np->vpd.mac_num;
+			max_len = 1;
+			found_mask |= FOUND_MASK_NMAC;
+		} else if (!strcmp(namebuf, "phy-type")) {
+			prop_buf = np->vpd.phy_type;
+			max_len = NIU_VPD_PHY_TYPE_MAX;
+			found_mask |= FOUND_MASK_PHY;
+		}
+
+		if (max_len && prop_len > max_len) {
+			dev_err(np->device, "Property '%s' length (%d) is too long\n", namebuf, prop_len);
+			return -EINVAL;
+		}
+
+		if (prop_buf) {
+			u32 off = start + 5 + err;
+			int i;
+
+			netif_printk(np, probe, KERN_DEBUG, np->dev,
+				     "VPD_SCAN: Reading in property [%s] len[%d]\n",
+				     namebuf, prop_len);
+			for (i = 0; i < prop_len; i++) {
+				err =  niu_pci_eeprom_read(np, off + i);
+				if (err < 0)
+					return err;
+				*prop_buf++ = err;
+			}
+		}
+
+		start += len;
+	}
+
+	return 0;
+}
+
+/* ESPC_PIO_EN_ENABLE must be set */
+static int niu_pci_vpd_fetch(struct niu *np, u32 start)
+{
+	u32 offset;
+	int err;
+
+	err = niu_pci_eeprom_read16_swp(np, start + 1);
+	if (err < 0)
+		return err;
+
+	offset = err + 3;
+
+	while (start + offset < ESPC_EEPROM_SIZE) {
+		u32 here = start + offset;
+		u32 end;
+
+		err = niu_pci_eeprom_read(np, here);
+		if (err < 0)
+			return err;
+		if (err != 0x90)
+			return -EINVAL;
+
+		err = niu_pci_eeprom_read16_swp(np, here + 1);
+		if (err < 0)
+			return err;
+
+		here = start + offset + 3;
+		end = start + offset + err;
+
+		offset += err;
+
+		err = niu_pci_vpd_scan_props(np, here, end);
+		if (err < 0)
+			return err;
+		/* ret == 1 is not an error */
+		if (err == 1)
+			return 0;
+	}
+	return 0;
+}
+
+/* ESPC_PIO_EN_ENABLE must be set */
+static u32 niu_pci_vpd_offset(struct niu *np)
+{
+	u32 start = 0, end = ESPC_EEPROM_SIZE, ret;
+	int err;
+
+	while (start < end) {
+		ret = start;
+
+		/* ROM header signature?  */
+		err = niu_pci_eeprom_read16(np, start +  0);
+		if (err != 0x55aa)
+			return 0;
+
+		/* Apply offset to PCI data structure.  */
+		err = niu_pci_eeprom_read16(np, start + 23);
+		if (err < 0)
+			return 0;
+		start += err;
+
+		/* Check for "PCIR" signature.  */
+		err = niu_pci_eeprom_read16(np, start +  0);
+		if (err != 0x5043)
+			return 0;
+		err = niu_pci_eeprom_read16(np, start +  2);
+		if (err != 0x4952)
+			return 0;
+
+		/* Check for OBP image type.  */
+		err = niu_pci_eeprom_read(np, start + 20);
+		if (err < 0)
+			return 0;
+		if (err != 0x01) {
+			err = niu_pci_eeprom_read(np, ret + 2);
+			if (err < 0)
+				return 0;
+
+			start = ret + (err * 512);
+			continue;
+		}
+
+		err = niu_pci_eeprom_read16_swp(np, start + 8);
+		if (err < 0)
+			return err;
+		ret += err;
+
+		err = niu_pci_eeprom_read(np, ret + 0);
+		if (err != 0x82)
+			return 0;
+
+		return ret;
+	}
+
+	return 0;
+}
+
+static int niu_phy_type_prop_decode(struct niu *np, const char *phy_prop)
+{
+	if (!strcmp(phy_prop, "mif")) {
+		/* 1G copper, MII */
+		np->flags &= ~(NIU_FLAGS_FIBER |
+			       NIU_FLAGS_10G);
+		np->mac_xcvr = MAC_XCVR_MII;
+	} else if (!strcmp(phy_prop, "xgf")) {
+		/* 10G fiber, XPCS */
+		np->flags |= (NIU_FLAGS_10G |
+			      NIU_FLAGS_FIBER);
+		np->mac_xcvr = MAC_XCVR_XPCS;
+	} else if (!strcmp(phy_prop, "pcs")) {
+		/* 1G fiber, PCS */
+		np->flags &= ~NIU_FLAGS_10G;
+		np->flags |= NIU_FLAGS_FIBER;
+		np->mac_xcvr = MAC_XCVR_PCS;
+	} else if (!strcmp(phy_prop, "xgc")) {
+		/* 10G copper, XPCS */
+		np->flags |= NIU_FLAGS_10G;
+		np->flags &= ~NIU_FLAGS_FIBER;
+		np->mac_xcvr = MAC_XCVR_XPCS;
+	} else if (!strcmp(phy_prop, "xgsd") || !strcmp(phy_prop, "gsd")) {
+		/* 10G Serdes or 1G Serdes, default to 10G */
+		np->flags |= NIU_FLAGS_10G;
+		np->flags &= ~NIU_FLAGS_FIBER;
+		np->flags |= NIU_FLAGS_XCVR_SERDES;
+		np->mac_xcvr = MAC_XCVR_XPCS;
+	} else {
+		return -EINVAL;
+	}
+	return 0;
+}
+
+static int niu_pci_vpd_get_nports(struct niu *np)
+{
+	int ports = 0;
+
+	if ((!strcmp(np->vpd.model, NIU_QGC_LP_MDL_STR)) ||
+	    (!strcmp(np->vpd.model, NIU_QGC_PEM_MDL_STR)) ||
+	    (!strcmp(np->vpd.model, NIU_MARAMBA_MDL_STR)) ||
+	    (!strcmp(np->vpd.model, NIU_KIMI_MDL_STR)) ||
+	    (!strcmp(np->vpd.model, NIU_ALONSO_MDL_STR))) {
+		ports = 4;
+	} else if ((!strcmp(np->vpd.model, NIU_2XGF_LP_MDL_STR)) ||
+		   (!strcmp(np->vpd.model, NIU_2XGF_PEM_MDL_STR)) ||
+		   (!strcmp(np->vpd.model, NIU_FOXXY_MDL_STR)) ||
+		   (!strcmp(np->vpd.model, NIU_2XGF_MRVL_MDL_STR))) {
+		ports = 2;
+	}
+
+	return ports;
+}
+
+static void niu_pci_vpd_validate(struct niu *np)
+{
+	struct net_device *dev = np->dev;
+	struct niu_vpd *vpd = &np->vpd;
+	u8 addr[ETH_ALEN];
+	u8 val8;
+
+	if (!is_valid_ether_addr(&vpd->local_mac[0])) {
+		dev_err(np->device, "VPD MAC invalid, falling back to SPROM\n");
+
+		np->flags &= ~NIU_FLAGS_VPD_VALID;
+		return;
+	}
+
+	if (!strcmp(np->vpd.model, NIU_ALONSO_MDL_STR) ||
+	    !strcmp(np->vpd.model, NIU_KIMI_MDL_STR)) {
+		np->flags |= NIU_FLAGS_10G;
+		np->flags &= ~NIU_FLAGS_FIBER;
+		np->flags |= NIU_FLAGS_XCVR_SERDES;
+		np->mac_xcvr = MAC_XCVR_PCS;
+		if (np->port > 1) {
+			np->flags |= NIU_FLAGS_FIBER;
+			np->flags &= ~NIU_FLAGS_10G;
+		}
+		if (np->flags & NIU_FLAGS_10G)
+			np->mac_xcvr = MAC_XCVR_XPCS;
+	} else if (!strcmp(np->vpd.model, NIU_FOXXY_MDL_STR)) {
+		np->flags |= (NIU_FLAGS_10G | NIU_FLAGS_FIBER |
+			      NIU_FLAGS_HOTPLUG_PHY);
+	} else if (niu_phy_type_prop_decode(np, np->vpd.phy_type)) {
+		dev_err(np->device, "Illegal phy string [%s]\n",
+			np->vpd.phy_type);
+		dev_err(np->device, "Falling back to SPROM\n");
+		np->flags &= ~NIU_FLAGS_VPD_VALID;
+		return;
+	}
+
+	ether_addr_copy(addr, vpd->local_mac);
+
+	val8 = addr[5];
+	addr[5] += np->port;
+	if (addr[5] < val8)
+		addr[4]++;
+
+	eth_hw_addr_set(dev, addr);
+}
+
+static int niu_pci_probe_sprom(struct niu *np)
+{
+	struct net_device *dev = np->dev;
+	u8 addr[ETH_ALEN];
+	int len, i;
+	u64 val, sum;
+	u8 val8;
+
+	val = (nr64(ESPC_VER_IMGSZ) & ESPC_VER_IMGSZ_IMGSZ);
+	val >>= ESPC_VER_IMGSZ_IMGSZ_SHIFT;
+	len = val / 4;
+
+	np->eeprom_len = len;
+
+	netif_printk(np, probe, KERN_DEBUG, np->dev,
+		     "SPROM: Image size %llu\n", (unsigned long long)val);
+
+	sum = 0;
+	for (i = 0; i < len; i++) {
+		val = nr64(ESPC_NCR(i));
+		sum += (val >>  0) & 0xff;
+		sum += (val >>  8) & 0xff;
+		sum += (val >> 16) & 0xff;
+		sum += (val >> 24) & 0xff;
+	}
+	netif_printk(np, probe, KERN_DEBUG, np->dev,
+		     "SPROM: Checksum %x\n", (int)(sum & 0xff));
+	if ((sum & 0xff) != 0xab) {
+		dev_err(np->device, "Bad SPROM checksum (%x, should be 0xab)\n", (int)(sum & 0xff));
+		return -EINVAL;
+	}
+
+	val = nr64(ESPC_PHY_TYPE);
+	switch (np->port) {
+	case 0:
+		val8 = (val & ESPC_PHY_TYPE_PORT0) >>
+			ESPC_PHY_TYPE_PORT0_SHIFT;
+		break;
+	case 1:
+		val8 = (val & ESPC_PHY_TYPE_PORT1) >>
+			ESPC_PHY_TYPE_PORT1_SHIFT;
+		break;
+	case 2:
+		val8 = (val & ESPC_PHY_TYPE_PORT2) >>
+			ESPC_PHY_TYPE_PORT2_SHIFT;
+		break;
+	case 3:
+		val8 = (val & ESPC_PHY_TYPE_PORT3) >>
+			ESPC_PHY_TYPE_PORT3_SHIFT;
+		break;
+	default:
+		dev_err(np->device, "Bogus port number %u\n",
+			np->port);
+		return -EINVAL;
+	}
+	netif_printk(np, probe, KERN_DEBUG, np->dev,
+		     "SPROM: PHY type %x\n", val8);
+
+	switch (val8) {
+	case ESPC_PHY_TYPE_1G_COPPER:
+		/* 1G copper, MII */
+		np->flags &= ~(NIU_FLAGS_FIBER |
+			       NIU_FLAGS_10G);
+		np->mac_xcvr = MAC_XCVR_MII;
+		break;
+
+	case ESPC_PHY_TYPE_1G_FIBER:
+		/* 1G fiber, PCS */
+		np->flags &= ~NIU_FLAGS_10G;
+		np->flags |= NIU_FLAGS_FIBER;
+		np->mac_xcvr = MAC_XCVR_PCS;
+		break;
+
+	case ESPC_PHY_TYPE_10G_COPPER:
+		/* 10G copper, XPCS */
+		np->flags |= NIU_FLAGS_10G;
+		np->flags &= ~NIU_FLAGS_FIBER;
+		np->mac_xcvr = MAC_XCVR_XPCS;
+		break;
+
+	case ESPC_PHY_TYPE_10G_FIBER:
+		/* 10G fiber, XPCS */
+		np->flags |= (NIU_FLAGS_10G |
+			      NIU_FLAGS_FIBER);
+		np->mac_xcvr = MAC_XCVR_XPCS;
+		break;
+
+	default:
+		dev_err(np->device, "Bogus SPROM phy type %u\n", val8);
+		return -EINVAL;
+	}
+
+	val = nr64(ESPC_MAC_ADDR0);
+	netif_printk(np, probe, KERN_DEBUG, np->dev,
+		     "SPROM: MAC_ADDR0[%08llx]\n", (unsigned long long)val);
+	addr[0] = (val >>  0) & 0xff;
+	addr[1] = (val >>  8) & 0xff;
+	addr[2] = (val >> 16) & 0xff;
+	addr[3] = (val >> 24) & 0xff;
+
+	val = nr64(ESPC_MAC_ADDR1);
+	netif_printk(np, probe, KERN_DEBUG, np->dev,
+		     "SPROM: MAC_ADDR1[%08llx]\n", (unsigned long long)val);
+	addr[4] = (val >>  0) & 0xff;
+	addr[5] = (val >>  8) & 0xff;
+
+	if (!is_valid_ether_addr(addr)) {
+		dev_err(np->device, "SPROM MAC address invalid [ %pM ]\n",
+			addr);
+		return -EINVAL;
+	}
+
+	val8 = addr[5];
+	addr[5] += np->port;
+	if (addr[5] < val8)
+		addr[4]++;
+
+	eth_hw_addr_set(dev, addr);
+
+	val = nr64(ESPC_MOD_STR_LEN);
+	netif_printk(np, probe, KERN_DEBUG, np->dev,
+		     "SPROM: MOD_STR_LEN[%llu]\n", (unsigned long long)val);
+	if (val >= 8 * 4)
+		return -EINVAL;
+
+	for (i = 0; i < val; i += 4) {
+		u64 tmp = nr64(ESPC_NCR(5 + (i / 4)));
+
+		np->vpd.model[i + 3] = (tmp >>  0) & 0xff;
+		np->vpd.model[i + 2] = (tmp >>  8) & 0xff;
+		np->vpd.model[i + 1] = (tmp >> 16) & 0xff;
+		np->vpd.model[i + 0] = (tmp >> 24) & 0xff;
+	}
+	np->vpd.model[val] = '\0';
+
+	val = nr64(ESPC_BD_MOD_STR_LEN);
+	netif_printk(np, probe, KERN_DEBUG, np->dev,
+		     "SPROM: BD_MOD_STR_LEN[%llu]\n", (unsigned long long)val);
+	if (val >= 4 * 4)
+		return -EINVAL;
+
+	for (i = 0; i < val; i += 4) {
+		u64 tmp = nr64(ESPC_NCR(14 + (i / 4)));
+
+		np->vpd.board_model[i + 3] = (tmp >>  0) & 0xff;
+		np->vpd.board_model[i + 2] = (tmp >>  8) & 0xff;
+		np->vpd.board_model[i + 1] = (tmp >> 16) & 0xff;
+		np->vpd.board_model[i + 0] = (tmp >> 24) & 0xff;
+	}
+	np->vpd.board_model[val] = '\0';
+
+	np->vpd.mac_num =
+		nr64(ESPC_NUM_PORTS_MACS) & ESPC_NUM_PORTS_MACS_VAL;
+	netif_printk(np, probe, KERN_DEBUG, np->dev,
+		     "SPROM: NUM_PORTS_MACS[%d]\n", np->vpd.mac_num);
+
+	return 0;
+}
+
+static int niu_get_and_validate_port(struct niu *np)
+{
+	struct niu_parent *parent = np->parent;
+
+	if (np->port <= 1)
+		np->flags |= NIU_FLAGS_XMAC;
+
+	if (!parent->num_ports) {
+		if (parent->plat_type == PLAT_TYPE_NIU) {
+			parent->num_ports = 2;
+		} else {
+			parent->num_ports = niu_pci_vpd_get_nports(np);
+			if (!parent->num_ports) {
+				/* Fall back to SPROM as last resort.
+				 * This will fail on most cards.
+				 */
+				parent->num_ports = nr64(ESPC_NUM_PORTS_MACS) &
+					ESPC_NUM_PORTS_MACS_VAL;
+
+				/* All of the current probing methods fail on
+				 * Maramba on-board parts.
+				 */
+				if (!parent->num_ports)
+					parent->num_ports = 4;
+			}
+		}
+	}
+
+	if (np->port >= parent->num_ports)
+		return -ENODEV;
+
+	return 0;
+}
+
+static int phy_record(struct niu_parent *parent, struct phy_probe_info *p,
+		      int dev_id_1, int dev_id_2, u8 phy_port, int type)
+{
+	u32 id = (dev_id_1 << 16) | dev_id_2;
+	u8 idx;
+
+	if (dev_id_1 < 0 || dev_id_2 < 0)
+		return 0;
+	if (type == PHY_TYPE_PMA_PMD || type == PHY_TYPE_PCS) {
+		/* Because of the NIU_PHY_ID_MASK being applied, the 8704
+		 * test covers the 8706 as well.
+		 */
+		if (((id & NIU_PHY_ID_MASK) != NIU_PHY_ID_BCM8704) &&
+		    ((id & NIU_PHY_ID_MASK) != NIU_PHY_ID_MRVL88X2011))
+			return 0;
+	} else {
+		if ((id & NIU_PHY_ID_MASK) != NIU_PHY_ID_BCM5464R)
+			return 0;
+	}
+
+	pr_info("niu%d: Found PHY %08x type %s at phy_port %u\n",
+		parent->index, id,
+		type == PHY_TYPE_PMA_PMD ? "PMA/PMD" :
+		type == PHY_TYPE_PCS ? "PCS" : "MII",
+		phy_port);
+
+	if (p->cur[type] >= NIU_MAX_PORTS) {
+		pr_err("Too many PHY ports\n");
+		return -EINVAL;
+	}
+	idx = p->cur[type];
+	p->phy_id[type][idx] = id;
+	p->phy_port[type][idx] = phy_port;
+	p->cur[type] = idx + 1;
+	return 0;
+}
+
+static int port_has_10g(struct phy_probe_info *p, int port)
+{
+	int i;
+
+	for (i = 0; i < p->cur[PHY_TYPE_PMA_PMD]; i++) {
+		if (p->phy_port[PHY_TYPE_PMA_PMD][i] == port)
+			return 1;
+	}
+	for (i = 0; i < p->cur[PHY_TYPE_PCS]; i++) {
+		if (p->phy_port[PHY_TYPE_PCS][i] == port)
+			return 1;
+	}
+
+	return 0;
+}
+
+static int count_10g_ports(struct phy_probe_info *p, int *lowest)
+{
+	int port, cnt;
+
+	cnt = 0;
+	*lowest = 32;
+	for (port = 8; port < 32; port++) {
+		if (port_has_10g(p, port)) {
+			if (!cnt)
+				*lowest = port;
+			cnt++;
+		}
+	}
+
+	return cnt;
+}
+
+static int count_1g_ports(struct phy_probe_info *p, int *lowest)
+{
+	*lowest = 32;
+	if (p->cur[PHY_TYPE_MII])
+		*lowest = p->phy_port[PHY_TYPE_MII][0];
+
+	return p->cur[PHY_TYPE_MII];
+}
+
+static void niu_n2_divide_channels(struct niu_parent *parent)
+{
+	int num_ports = parent->num_ports;
+	int i;
+
+	for (i = 0; i < num_ports; i++) {
+		parent->rxchan_per_port[i] = (16 / num_ports);
+		parent->txchan_per_port[i] = (16 / num_ports);
+
+		pr_info("niu%d: Port %u [%u RX chans] [%u TX chans]\n",
+			parent->index, i,
+			parent->rxchan_per_port[i],
+			parent->txchan_per_port[i]);
+	}
+}
+
+static void niu_divide_channels(struct niu_parent *parent,
+				int num_10g, int num_1g)
+{
+	int num_ports = parent->num_ports;
+	int rx_chans_per_10g, rx_chans_per_1g;
+	int tx_chans_per_10g, tx_chans_per_1g;
+	int i, tot_rx, tot_tx;
+
+	if (!num_10g || !num_1g) {
+		rx_chans_per_10g = rx_chans_per_1g =
+			(NIU_NUM_RXCHAN / num_ports);
+		tx_chans_per_10g = tx_chans_per_1g =
+			(NIU_NUM_TXCHAN / num_ports);
+	} else {
+		rx_chans_per_1g = NIU_NUM_RXCHAN / 8;
+		rx_chans_per_10g = (NIU_NUM_RXCHAN -
+				    (rx_chans_per_1g * num_1g)) /
+			num_10g;
+
+		tx_chans_per_1g = NIU_NUM_TXCHAN / 6;
+		tx_chans_per_10g = (NIU_NUM_TXCHAN -
+				    (tx_chans_per_1g * num_1g)) /
+			num_10g;
+	}
+
+	tot_rx = tot_tx = 0;
+	for (i = 0; i < num_ports; i++) {
+		int type = phy_decode(parent->port_phy, i);
+
+		if (type == PORT_TYPE_10G) {
+			parent->rxchan_per_port[i] = rx_chans_per_10g;
+			parent->txchan_per_port[i] = tx_chans_per_10g;
+		} else {
+			parent->rxchan_per_port[i] = rx_chans_per_1g;
+			parent->txchan_per_port[i] = tx_chans_per_1g;
+		}
+		pr_info("niu%d: Port %u [%u RX chans] [%u TX chans]\n",
+			parent->index, i,
+			parent->rxchan_per_port[i],
+			parent->txchan_per_port[i]);
+		tot_rx += parent->rxchan_per_port[i];
+		tot_tx += parent->txchan_per_port[i];
+	}
+
+	if (tot_rx > NIU_NUM_RXCHAN) {
+		pr_err("niu%d: Too many RX channels (%d), resetting to one per port\n",
+		       parent->index, tot_rx);
+		for (i = 0; i < num_ports; i++)
+			parent->rxchan_per_port[i] = 1;
+	}
+	if (tot_tx > NIU_NUM_TXCHAN) {
+		pr_err("niu%d: Too many TX channels (%d), resetting to one per port\n",
+		       parent->index, tot_tx);
+		for (i = 0; i < num_ports; i++)
+			parent->txchan_per_port[i] = 1;
+	}
+	if (tot_rx < NIU_NUM_RXCHAN || tot_tx < NIU_NUM_TXCHAN) {
+		pr_warn("niu%d: Driver bug, wasted channels, RX[%d] TX[%d]\n",
+			parent->index, tot_rx, tot_tx);
+	}
+}
+
+static void niu_divide_rdc_groups(struct niu_parent *parent,
+				  int num_10g, int num_1g)
+{
+	int i, num_ports = parent->num_ports;
+	int rdc_group, rdc_groups_per_port;
+	int rdc_channel_base;
+
+	rdc_group = 0;
+	rdc_groups_per_port = NIU_NUM_RDC_TABLES / num_ports;
+
+	rdc_channel_base = 0;
+
+	for (i = 0; i < num_ports; i++) {
+		struct niu_rdc_tables *tp = &parent->rdc_group_cfg[i];
+		int grp, num_channels = parent->rxchan_per_port[i];
+		int this_channel_offset;
+
+		tp->first_table_num = rdc_group;
+		tp->num_tables = rdc_groups_per_port;
+		this_channel_offset = 0;
+		for (grp = 0; grp < tp->num_tables; grp++) {
+			struct rdc_table *rt = &tp->tables[grp];
+			int slot;
+
+			pr_info("niu%d: Port %d RDC tbl(%d) [ ",
+				parent->index, i, tp->first_table_num + grp);
+			for (slot = 0; slot < NIU_RDC_TABLE_SLOTS; slot++) {
+				rt->rxdma_channel[slot] =
+					rdc_channel_base + this_channel_offset;
+
+				pr_cont("%d ", rt->rxdma_channel[slot]);
+
+				if (++this_channel_offset == num_channels)
+					this_channel_offset = 0;
+			}
+			pr_cont("]\n");
+		}
+
+		parent->rdc_default[i] = rdc_channel_base;
+
+		rdc_channel_base += num_channels;
+		rdc_group += rdc_groups_per_port;
+	}
+}
+
+static int fill_phy_probe_info(struct niu *np, struct niu_parent *parent,
+			       struct phy_probe_info *info)
+{
+	unsigned long flags;
+	int port, err;
+
+	memset(info, 0, sizeof(*info));
+
+	/* Port 0 to 7 are reserved for onboard Serdes, probe the rest.  */
+	niu_lock_parent(np, flags);
+	err = 0;
+	for (port = 8; port < 32; port++) {
+		int dev_id_1, dev_id_2;
+
+		dev_id_1 = mdio_read(np, port,
+				     NIU_PMA_PMD_DEV_ADDR, MII_PHYSID1);
+		dev_id_2 = mdio_read(np, port,
+				     NIU_PMA_PMD_DEV_ADDR, MII_PHYSID2);
+		err = phy_record(parent, info, dev_id_1, dev_id_2, port,
+				 PHY_TYPE_PMA_PMD);
+		if (err)
+			break;
+		dev_id_1 = mdio_read(np, port,
+				     NIU_PCS_DEV_ADDR, MII_PHYSID1);
+		dev_id_2 = mdio_read(np, port,
+				     NIU_PCS_DEV_ADDR, MII_PHYSID2);
+		err = phy_record(parent, info, dev_id_1, dev_id_2, port,
+				 PHY_TYPE_PCS);
+		if (err)
+			break;
+		dev_id_1 = mii_read(np, port, MII_PHYSID1);
+		dev_id_2 = mii_read(np, port, MII_PHYSID2);
+		err = phy_record(parent, info, dev_id_1, dev_id_2, port,
+				 PHY_TYPE_MII);
+		if (err)
+			break;
+	}
+	niu_unlock_parent(np, flags);
+
+	return err;
+}
+
+static int walk_phys(struct niu *np, struct niu_parent *parent)
+{
+	struct phy_probe_info *info = &parent->phy_probe_info;
+	int lowest_10g, lowest_1g;
+	int num_10g, num_1g;
+	u32 val;
+	int err;
+
+	num_10g = num_1g = 0;
+
+	if (!strcmp(np->vpd.model, NIU_ALONSO_MDL_STR) ||
+	    !strcmp(np->vpd.model, NIU_KIMI_MDL_STR)) {
+		num_10g = 0;
+		num_1g = 2;
+		parent->plat_type = PLAT_TYPE_ATCA_CP3220;
+		parent->num_ports = 4;
+		val = (phy_encode(PORT_TYPE_1G, 0) |
+		       phy_encode(PORT_TYPE_1G, 1) |
+		       phy_encode(PORT_TYPE_1G, 2) |
+		       phy_encode(PORT_TYPE_1G, 3));
+	} else if (!strcmp(np->vpd.model, NIU_FOXXY_MDL_STR)) {
+		num_10g = 2;
+		num_1g = 0;
+		parent->num_ports = 2;
+		val = (phy_encode(PORT_TYPE_10G, 0) |
+		       phy_encode(PORT_TYPE_10G, 1));
+	} else if ((np->flags & NIU_FLAGS_XCVR_SERDES) &&
+		   (parent->plat_type == PLAT_TYPE_NIU)) {
+		/* this is the Monza case */
+		if (np->flags & NIU_FLAGS_10G) {
+			val = (phy_encode(PORT_TYPE_10G, 0) |
+			       phy_encode(PORT_TYPE_10G, 1));
+		} else {
+			val = (phy_encode(PORT_TYPE_1G, 0) |
+			       phy_encode(PORT_TYPE_1G, 1));
+		}
+	} else {
+		err = fill_phy_probe_info(np, parent, info);
+		if (err)
+			return err;
+
+		num_10g = count_10g_ports(info, &lowest_10g);
+		num_1g = count_1g_ports(info, &lowest_1g);
+
+		switch ((num_10g << 4) | num_1g) {
+		case 0x24:
+			if (lowest_1g == 10)
+				parent->plat_type = PLAT_TYPE_VF_P0;
+			else if (lowest_1g == 26)
+				parent->plat_type = PLAT_TYPE_VF_P1;
+			else
+				goto unknown_vg_1g_port;
+
+			fallthrough;
+		case 0x22:
+			val = (phy_encode(PORT_TYPE_10G, 0) |
+			       phy_encode(PORT_TYPE_10G, 1) |
+			       phy_encode(PORT_TYPE_1G, 2) |
+			       phy_encode(PORT_TYPE_1G, 3));
+			break;
+
+		case 0x20:
+			val = (phy_encode(PORT_TYPE_10G, 0) |
+			       phy_encode(PORT_TYPE_10G, 1));
+			break;
+
+		case 0x10:
+			val = phy_encode(PORT_TYPE_10G, np->port);
+			break;
+
+		case 0x14:
+			if (lowest_1g == 10)
+				parent->plat_type = PLAT_TYPE_VF_P0;
+			else if (lowest_1g == 26)
+				parent->plat_type = PLAT_TYPE_VF_P1;
+			else
+				goto unknown_vg_1g_port;
+
+			fallthrough;
+		case 0x13:
+			if ((lowest_10g & 0x7) == 0)
+				val = (phy_encode(PORT_TYPE_10G, 0) |
+				       phy_encode(PORT_TYPE_1G, 1) |
+				       phy_encode(PORT_TYPE_1G, 2) |
+				       phy_encode(PORT_TYPE_1G, 3));
+			else
+				val = (phy_encode(PORT_TYPE_1G, 0) |
+				       phy_encode(PORT_TYPE_10G, 1) |
+				       phy_encode(PORT_TYPE_1G, 2) |
+				       phy_encode(PORT_TYPE_1G, 3));
+			break;
+
+		case 0x04:
+			if (lowest_1g == 10)
+				parent->plat_type = PLAT_TYPE_VF_P0;
+			else if (lowest_1g == 26)
+				parent->plat_type = PLAT_TYPE_VF_P1;
+			else
+				goto unknown_vg_1g_port;
+
+			val = (phy_encode(PORT_TYPE_1G, 0) |
+			       phy_encode(PORT_TYPE_1G, 1) |
+			       phy_encode(PORT_TYPE_1G, 2) |
+			       phy_encode(PORT_TYPE_1G, 3));
+			break;
+
+		default:
+			pr_err("Unsupported port config 10G[%d] 1G[%d]\n",
+			       num_10g, num_1g);
+			return -EINVAL;
+		}
+	}
+
+	parent->port_phy = val;
+
+	if (parent->plat_type == PLAT_TYPE_NIU)
+		niu_n2_divide_channels(parent);
+	else
+		niu_divide_channels(parent, num_10g, num_1g);
+
+	niu_divide_rdc_groups(parent, num_10g, num_1g);
+
+	return 0;
+
+unknown_vg_1g_port:
+	pr_err("Cannot identify platform type, 1gport=%d\n", lowest_1g);
+	return -EINVAL;
+}
+
+static int niu_probe_ports(struct niu *np)
+{
+	struct niu_parent *parent = np->parent;
+	int err, i;
+
+	if (parent->port_phy == PORT_PHY_UNKNOWN) {
+		err = walk_phys(np, parent);
+		if (err)
+			return err;
+
+		niu_set_ldg_timer_res(np, 2);
+		for (i = 0; i <= LDN_MAX; i++)
+			niu_ldn_irq_enable(np, i, 0);
+	}
+
+	if (parent->port_phy == PORT_PHY_INVALID)
+		return -EINVAL;
+
+	return 0;
+}
+
+static int niu_classifier_swstate_init(struct niu *np)
+{
+	struct niu_classifier *cp = &np->clas;
+
+	cp->tcam_top = (u16) np->port;
+	cp->tcam_sz = np->parent->tcam_num_entries / np->parent->num_ports;
+	cp->h1_init = 0xffffffff;
+	cp->h2_init = 0xffff;
+
+	return fflp_early_init(np);
+}
+
+static void niu_link_config_init(struct niu *np)
+{
+	struct niu_link_config *lp = &np->link_config;
+
+	lp->advertising = (ADVERTISED_10baseT_Half |
+			   ADVERTISED_10baseT_Full |
+			   ADVERTISED_100baseT_Half |
+			   ADVERTISED_100baseT_Full |
+			   ADVERTISED_1000baseT_Half |
+			   ADVERTISED_1000baseT_Full |
+			   ADVERTISED_10000baseT_Full |
+			   ADVERTISED_Autoneg);
+	lp->speed = lp->active_speed = SPEED_INVALID;
+	lp->duplex = DUPLEX_FULL;
+	lp->active_duplex = DUPLEX_INVALID;
+	lp->autoneg = 1;
+#if 0
+	lp->loopback_mode = LOOPBACK_MAC;
+	lp->active_speed = SPEED_10000;
+	lp->active_duplex = DUPLEX_FULL;
+#else
+	lp->loopback_mode = LOOPBACK_DISABLED;
+#endif
+}
+
+static int niu_init_mac_ipp_pcs_base(struct niu *np)
+{
+	switch (np->port) {
+	case 0:
+		np->mac_regs = np->regs + XMAC_PORT0_OFF;
+		np->ipp_off  = 0x00000;
+		np->pcs_off  = 0x04000;
+		np->xpcs_off = 0x02000;
+		break;
+
+	case 1:
+		np->mac_regs = np->regs + XMAC_PORT1_OFF;
+		np->ipp_off  = 0x08000;
+		np->pcs_off  = 0x0a000;
+		np->xpcs_off = 0x08000;
+		break;
+
+	case 2:
+		np->mac_regs = np->regs + BMAC_PORT2_OFF;
+		np->ipp_off  = 0x04000;
+		np->pcs_off  = 0x0e000;
+		np->xpcs_off = ~0UL;
+		break;
+
+	case 3:
+		np->mac_regs = np->regs + BMAC_PORT3_OFF;
+		np->ipp_off  = 0x0c000;
+		np->pcs_off  = 0x12000;
+		np->xpcs_off = ~0UL;
+		break;
+
+	default:
+		dev_err(np->device, "Port %u is invalid, cannot compute MAC block offset\n", np->port);
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
+static void niu_try_msix(struct niu *np, u8 *ldg_num_map)
+{
+	struct msix_entry msi_vec[NIU_NUM_LDG];
+	struct niu_parent *parent = np->parent;
+	struct pci_dev *pdev = np->pdev;
+	int i, num_irqs;
+	u8 first_ldg;
+
+	first_ldg = (NIU_NUM_LDG / parent->num_ports) * np->port;
+	for (i = 0; i < (NIU_NUM_LDG / parent->num_ports); i++)
+		ldg_num_map[i] = first_ldg + i;
+
+	num_irqs = (parent->rxchan_per_port[np->port] +
+		    parent->txchan_per_port[np->port] +
+		    (np->port == 0 ? 3 : 1));
+	BUG_ON(num_irqs > (NIU_NUM_LDG / parent->num_ports));
+
+	for (i = 0; i < num_irqs; i++) {
+		msi_vec[i].vector = 0;
+		msi_vec[i].entry = i;
+	}
+
+	pdev->dev_flags |= PCI_DEV_FLAGS_MSIX_TOUCH_ENTRY_DATA_FIRST;
+
+	num_irqs = pci_enable_msix_range(pdev, msi_vec, 1, num_irqs);
+	if (num_irqs < 0) {
+		np->flags &= ~NIU_FLAGS_MSIX;
+		return;
+	}
+
+	np->flags |= NIU_FLAGS_MSIX;
+	for (i = 0; i < num_irqs; i++)
+		np->ldg[i].irq = msi_vec[i].vector;
+	np->num_ldg = num_irqs;
+}
+
+static int niu_n2_irq_init(struct niu *np, u8 *ldg_num_map)
+{
+#ifdef CONFIG_SPARC64
+	struct platform_device *op = np->op;
+	const u32 *int_prop;
+	int i;
+
+	int_prop = of_get_property(op->dev.of_node, "interrupts", NULL);
+	if (!int_prop)
+		return -ENODEV;
+
+	for (i = 0; i < op->archdata.num_irqs; i++) {
+		ldg_num_map[i] = int_prop[i];
+		np->ldg[i].irq = op->archdata.irqs[i];
+	}
+
+	np->num_ldg = op->archdata.num_irqs;
+
+	return 0;
+#else
+	return -EINVAL;
+#endif
+}
+
+static int niu_ldg_init(struct niu *np)
+{
+	struct niu_parent *parent = np->parent;
+	u8 ldg_num_map[NIU_NUM_LDG];
+	int first_chan, num_chan;
+	int i, err, ldg_rotor;
+	u8 port;
+
+	np->num_ldg = 1;
+	np->ldg[0].irq = np->dev->irq;
+	if (parent->plat_type == PLAT_TYPE_NIU) {
+		err = niu_n2_irq_init(np, ldg_num_map);
+		if (err)
+			return err;
+	} else
+		niu_try_msix(np, ldg_num_map);
+
+	port = np->port;
+	for (i = 0; i < np->num_ldg; i++) {
+		struct niu_ldg *lp = &np->ldg[i];
+
+		netif_napi_add(np->dev, &lp->napi, niu_poll);
+
+		lp->np = np;
+		lp->ldg_num = ldg_num_map[i];
+		lp->timer = 2; /* XXX */
+
+		/* On N2 NIU the firmware has setup the SID mappings so they go
+		 * to the correct values that will route the LDG to the proper
+		 * interrupt in the NCU interrupt table.
+		 */
+		if (np->parent->plat_type != PLAT_TYPE_NIU) {
+			err = niu_set_ldg_sid(np, lp->ldg_num, port, i);
+			if (err)
+				return err;
+		}
+	}
+
+	/* We adopt the LDG assignment ordering used by the N2 NIU
+	 * 'interrupt' properties because that simplifies a lot of
+	 * things.  This ordering is:
+	 *
+	 *	MAC
+	 *	MIF	(if port zero)
+	 *	SYSERR	(if port zero)
+	 *	RX channels
+	 *	TX channels
+	 */
+
+	ldg_rotor = 0;
+
+	err = niu_ldg_assign_ldn(np, parent, ldg_num_map[ldg_rotor],
+				  LDN_MAC(port));
+	if (err)
+		return err;
+
+	ldg_rotor++;
+	if (ldg_rotor == np->num_ldg)
+		ldg_rotor = 0;
+
+	if (port == 0) {
+		err = niu_ldg_assign_ldn(np, parent,
+					 ldg_num_map[ldg_rotor],
+					 LDN_MIF);
+		if (err)
+			return err;
+
+		ldg_rotor++;
+		if (ldg_rotor == np->num_ldg)
+			ldg_rotor = 0;
+
+		err = niu_ldg_assign_ldn(np, parent,
+					 ldg_num_map[ldg_rotor],
+					 LDN_DEVICE_ERROR);
+		if (err)
+			return err;
+
+		ldg_rotor++;
+		if (ldg_rotor == np->num_ldg)
+			ldg_rotor = 0;
+
+	}
+
+	first_chan = 0;
+	for (i = 0; i < port; i++)
+		first_chan += parent->rxchan_per_port[i];
+	num_chan = parent->rxchan_per_port[port];
+
+	for (i = first_chan; i < (first_chan + num_chan); i++) {
+		err = niu_ldg_assign_ldn(np, parent,
+					 ldg_num_map[ldg_rotor],
+					 LDN_RXDMA(i));
+		if (err)
+			return err;
+		ldg_rotor++;
+		if (ldg_rotor == np->num_ldg)
+			ldg_rotor = 0;
+	}
+
+	first_chan = 0;
+	for (i = 0; i < port; i++)
+		first_chan += parent->txchan_per_port[i];
+	num_chan = parent->txchan_per_port[port];
+	for (i = first_chan; i < (first_chan + num_chan); i++) {
+		err = niu_ldg_assign_ldn(np, parent,
+					 ldg_num_map[ldg_rotor],
+					 LDN_TXDMA(i));
+		if (err)
+			return err;
+		ldg_rotor++;
+		if (ldg_rotor == np->num_ldg)
+			ldg_rotor = 0;
+	}
+
+	return 0;
+}
+
+static void niu_ldg_free(struct niu *np)
+{
+	if (np->flags & NIU_FLAGS_MSIX)
+		pci_disable_msix(np->pdev);
+}
+
+static int niu_get_of_props(struct niu *np)
+{
+#ifdef CONFIG_SPARC64
+	struct net_device *dev = np->dev;
+	struct device_node *dp;
+	const char *phy_type;
+	const u8 *mac_addr;
+	const char *model;
+	int prop_len;
+
+	if (np->parent->plat_type == PLAT_TYPE_NIU)
+		dp = np->op->dev.of_node;
+	else
+		dp = pci_device_to_OF_node(np->pdev);
+
+	phy_type = of_get_property(dp, "phy-type", NULL);
+	if (!phy_type) {
+		netdev_err(dev, "%pOF: OF node lacks phy-type property\n", dp);
+		return -EINVAL;
+	}
+
+	if (!strcmp(phy_type, "none"))
+		return -ENODEV;
+
+	strcpy(np->vpd.phy_type, phy_type);
+
+	if (niu_phy_type_prop_decode(np, np->vpd.phy_type)) {
+		netdev_err(dev, "%pOF: Illegal phy string [%s]\n",
+			   dp, np->vpd.phy_type);
+		return -EINVAL;
+	}
+
+	mac_addr = of_get_property(dp, "local-mac-address", &prop_len);
+	if (!mac_addr) {
+		netdev_err(dev, "%pOF: OF node lacks local-mac-address property\n",
+			   dp);
+		return -EINVAL;
+	}
+	if (prop_len != dev->addr_len) {
+		netdev_err(dev, "%pOF: OF MAC address prop len (%d) is wrong\n",
+			   dp, prop_len);
+	}
+	eth_hw_addr_set(dev, mac_addr);
+	if (!is_valid_ether_addr(&dev->dev_addr[0])) {
+		netdev_err(dev, "%pOF: OF MAC address is invalid\n", dp);
+		netdev_err(dev, "%pOF: [ %pM ]\n", dp, dev->dev_addr);
+		return -EINVAL;
+	}
+
+	model = of_get_property(dp, "model", NULL);
+
+	if (model)
+		strcpy(np->vpd.model, model);
+
+	if (of_property_read_bool(dp, "hot-swappable-phy")) {
+		np->flags |= (NIU_FLAGS_10G | NIU_FLAGS_FIBER |
+			NIU_FLAGS_HOTPLUG_PHY);
+	}
+
+	return 0;
+#else
+	return -EINVAL;
+#endif
+}
+
+static int niu_get_invariants(struct niu *np)
+{
+	int err, have_props;
+	u32 offset;
+
+	err = niu_get_of_props(np);
+	if (err == -ENODEV)
+		return err;
+
+	have_props = !err;
+
+	err = niu_init_mac_ipp_pcs_base(np);
+	if (err)
+		return err;
+
+	if (have_props) {
+		err = niu_get_and_validate_port(np);
+		if (err)
+			return err;
+
+	} else  {
+		if (np->parent->plat_type == PLAT_TYPE_NIU)
+			return -EINVAL;
+
+		nw64(ESPC_PIO_EN, ESPC_PIO_EN_ENABLE);
+		offset = niu_pci_vpd_offset(np);
+		netif_printk(np, probe, KERN_DEBUG, np->dev,
+			     "%s() VPD offset [%08x]\n", __func__, offset);
+		if (offset) {
+			err = niu_pci_vpd_fetch(np, offset);
+			if (err < 0)
+				return err;
+		}
+		nw64(ESPC_PIO_EN, 0);
+
+		if (np->flags & NIU_FLAGS_VPD_VALID) {
+			niu_pci_vpd_validate(np);
+			err = niu_get_and_validate_port(np);
+			if (err)
+				return err;
+		}
+
+		if (!(np->flags & NIU_FLAGS_VPD_VALID)) {
+			err = niu_get_and_validate_port(np);
+			if (err)
+				return err;
+			err = niu_pci_probe_sprom(np);
+			if (err)
+				return err;
+		}
+	}
+
+	err = niu_probe_ports(np);
+	if (err)
+		return err;
+
+	niu_ldg_init(np);
+
+	niu_classifier_swstate_init(np);
+	niu_link_config_init(np);
+
+	err = niu_determine_phy_disposition(np);
+	if (!err)
+		err = niu_init_link(np);
+
+	return err;
+}
+
+static LIST_HEAD(niu_parent_list);
+static DEFINE_MUTEX(niu_parent_lock);
+static int niu_parent_index;
+
+static ssize_t show_port_phy(struct device *dev,
+			     struct device_attribute *attr, char *buf)
+{
+	struct platform_device *plat_dev = to_platform_device(dev);
+	struct niu_parent *p = dev_get_platdata(&plat_dev->dev);
+	u32 port_phy = p->port_phy;
+	char *orig_buf = buf;
+	int i;
+
+	if (port_phy == PORT_PHY_UNKNOWN ||
+	    port_phy == PORT_PHY_INVALID)
+		return 0;
+
+	for (i = 0; i < p->num_ports; i++) {
+		const char *type_str;
+		int type;
+
+		type = phy_decode(port_phy, i);
+		if (type == PORT_TYPE_10G)
+			type_str = "10G";
+		else
+			type_str = "1G";
+		buf += sprintf(buf,
+			       (i == 0) ? "%s" : " %s",
+			       type_str);
+	}
+	buf += sprintf(buf, "\n");
+	return buf - orig_buf;
+}
+
+static ssize_t show_plat_type(struct device *dev,
+			      struct device_attribute *attr, char *buf)
+{
+	struct platform_device *plat_dev = to_platform_device(dev);
+	struct niu_parent *p = dev_get_platdata(&plat_dev->dev);
+	const char *type_str;
+
+	switch (p->plat_type) {
+	case PLAT_TYPE_ATLAS:
+		type_str = "atlas";
+		break;
+	case PLAT_TYPE_NIU:
+		type_str = "niu";
+		break;
+	case PLAT_TYPE_VF_P0:
+		type_str = "vf_p0";
+		break;
+	case PLAT_TYPE_VF_P1:
+		type_str = "vf_p1";
+		break;
+	default:
+		type_str = "unknown";
+		break;
+	}
+
+	return sprintf(buf, "%s\n", type_str);
+}
+
+static ssize_t __show_chan_per_port(struct device *dev,
+				    struct device_attribute *attr, char *buf,
+				    int rx)
+{
+	struct platform_device *plat_dev = to_platform_device(dev);
+	struct niu_parent *p = dev_get_platdata(&plat_dev->dev);
+	char *orig_buf = buf;
+	u8 *arr;
+	int i;
+
+	arr = (rx ? p->rxchan_per_port : p->txchan_per_port);
+
+	for (i = 0; i < p->num_ports; i++) {
+		buf += sprintf(buf,
+			       (i == 0) ? "%d" : " %d",
+			       arr[i]);
+	}
+	buf += sprintf(buf, "\n");
+
+	return buf - orig_buf;
+}
+
+static ssize_t show_rxchan_per_port(struct device *dev,
+				    struct device_attribute *attr, char *buf)
+{
+	return __show_chan_per_port(dev, attr, buf, 1);
+}
+
+static ssize_t show_txchan_per_port(struct device *dev,
+				    struct device_attribute *attr, char *buf)
+{
+	return __show_chan_per_port(dev, attr, buf, 1);
+}
+
+static ssize_t show_num_ports(struct device *dev,
+			      struct device_attribute *attr, char *buf)
+{
+	struct platform_device *plat_dev = to_platform_device(dev);
+	struct niu_parent *p = dev_get_platdata(&plat_dev->dev);
+
+	return sprintf(buf, "%d\n", p->num_ports);
+}
+
+static struct device_attribute niu_parent_attributes[] = {
+	__ATTR(port_phy, 0444, show_port_phy, NULL),
+	__ATTR(plat_type, 0444, show_plat_type, NULL),
+	__ATTR(rxchan_per_port, 0444, show_rxchan_per_port, NULL),
+	__ATTR(txchan_per_port, 0444, show_txchan_per_port, NULL),
+	__ATTR(num_ports, 0444, show_num_ports, NULL),
+	{}
+};
+
+static struct niu_parent *niu_new_parent(struct niu *np,
+					 union niu_parent_id *id, u8 ptype)
+{
+	struct platform_device *plat_dev;
+	struct niu_parent *p;
+	int i;
+
+	plat_dev = platform_device_register_simple("niu-board", niu_parent_index,
+						   NULL, 0);
+	if (IS_ERR(plat_dev))
+		return NULL;
+
+	for (i = 0; niu_parent_attributes[i].attr.name; i++) {
+		int err = device_create_file(&plat_dev->dev,
+					     &niu_parent_attributes[i]);
+		if (err)
+			goto fail_unregister;
+	}
+
+	p = kzalloc_obj(*p);
+	if (!p)
+		goto fail_unregister;
+
+	p->index = niu_parent_index++;
+
+	plat_dev->dev.platform_data = p;
+	p->plat_dev = plat_dev;
+
+	memcpy(&p->id, id, sizeof(*id));
+	p->plat_type = ptype;
+	INIT_LIST_HEAD(&p->list);
+	atomic_set(&p->refcnt, 0);
+	list_add(&p->list, &niu_parent_list);
+	spin_lock_init(&p->lock);
+
+	p->rxdma_clock_divider = 7500;
+
+	p->tcam_num_entries = NIU_PCI_TCAM_ENTRIES;
+	if (p->plat_type == PLAT_TYPE_NIU)
+		p->tcam_num_entries = NIU_NONPCI_TCAM_ENTRIES;
+
+	for (i = CLASS_CODE_USER_PROG1; i <= CLASS_CODE_SCTP_IPV6; i++) {
+		int index = i - CLASS_CODE_USER_PROG1;
+
+		p->tcam_key[index] = TCAM_KEY_TSEL;
+		p->flow_key[index] = (FLOW_KEY_IPSA |
+				      FLOW_KEY_IPDA |
+				      FLOW_KEY_PROTO |
+				      (FLOW_KEY_L4_BYTE12 <<
+				       FLOW_KEY_L4_0_SHIFT) |
+				      (FLOW_KEY_L4_BYTE12 <<
+				       FLOW_KEY_L4_1_SHIFT));
+	}
+
+	for (i = 0; i < LDN_MAX + 1; i++)
+		p->ldg_map[i] = LDG_INVALID;
+
+	return p;
+
+fail_unregister:
+	platform_device_unregister(plat_dev);
+	return NULL;
+}
+
+static struct niu_parent *niu_get_parent(struct niu *np,
+					 union niu_parent_id *id, u8 ptype)
+{
+	struct niu_parent *p, *tmp;
+	int port = np->port;
+
+	mutex_lock(&niu_parent_lock);
+	p = NULL;
+	list_for_each_entry(tmp, &niu_parent_list, list) {
+		if (!memcmp(id, &tmp->id, sizeof(*id))) {
+			p = tmp;
+			break;
+		}
+	}
+	if (!p)
+		p = niu_new_parent(np, id, ptype);
+
+	if (p) {
+		char port_name[8];
+		int err;
+
+		sprintf(port_name, "port%d", port);
+		err = sysfs_create_link(&p->plat_dev->dev.kobj,
+					&np->device->kobj,
+					port_name);
+		if (!err) {
+			p->ports[port] = np;
+			atomic_inc(&p->refcnt);
+		}
+	}
+	mutex_unlock(&niu_parent_lock);
+
+	return p;
+}
+
+static void niu_put_parent(struct niu *np)
+{
+	struct niu_parent *p = np->parent;
+	u8 port = np->port;
+	char port_name[8];
+
+	BUG_ON(!p || p->ports[port] != np);
+
+	netif_printk(np, probe, KERN_DEBUG, np->dev,
+		     "%s() port[%u]\n", __func__, port);
+
+	sprintf(port_name, "port%d", port);
+
+	mutex_lock(&niu_parent_lock);
+
+	sysfs_remove_link(&p->plat_dev->dev.kobj, port_name);
+
+	p->ports[port] = NULL;
+	np->parent = NULL;
+
+	if (atomic_dec_and_test(&p->refcnt)) {
+		list_del(&p->list);
+		platform_device_unregister(p->plat_dev);
+	}
+
+	mutex_unlock(&niu_parent_lock);
+}
+
+static void *niu_pci_alloc_coherent(struct device *dev, size_t size,
+				    u64 *handle, gfp_t flag)
+{
+	dma_addr_t dh;
+	void *ret;
+
+	ret = dma_alloc_coherent(dev, size, &dh, flag);
+	if (ret)
+		*handle = dh;
+	return ret;
+}
+
+static void niu_pci_free_coherent(struct device *dev, size_t size,
+				  void *cpu_addr, u64 handle)
+{
+	dma_free_coherent(dev, size, cpu_addr, handle);
+}
+
+static u64 niu_pci_map_page(struct device *dev, struct page *page,
+			    unsigned long offset, size_t size,
+			    enum dma_data_direction direction)
+{
+	return dma_map_page(dev, page, offset, size, direction);
+}
+
+static void niu_pci_unmap_page(struct device *dev, u64 dma_address,
+			       size_t size, enum dma_data_direction direction)
+{
+	dma_unmap_page(dev, dma_address, size, direction);
+}
+
+static u64 niu_pci_map_single(struct device *dev, void *cpu_addr,
+			      size_t size,
+			      enum dma_data_direction direction)
+{
+	return dma_map_single(dev, cpu_addr, size, direction);
+}
+
+static void niu_pci_unmap_single(struct device *dev, u64 dma_address,
+				 size_t size,
+				 enum dma_data_direction direction)
+{
+	dma_unmap_single(dev, dma_address, size, direction);
+}
+
+static int niu_pci_mapping_error(struct device *dev, u64 addr)
+{
+	return dma_mapping_error(dev, addr);
+}
+
+static const struct niu_ops niu_pci_ops = {
+	.alloc_coherent	= niu_pci_alloc_coherent,
+	.free_coherent	= niu_pci_free_coherent,
+	.map_page	= niu_pci_map_page,
+	.unmap_page	= niu_pci_unmap_page,
+	.map_single	= niu_pci_map_single,
+	.unmap_single	= niu_pci_unmap_single,
+	.mapping_error	= niu_pci_mapping_error,
+};
+
+static void niu_driver_version(void)
+{
+	static int niu_version_printed;
+
+	if (niu_version_printed++ == 0)
+		pr_info("%s", version);
+}
+
+static struct net_device *niu_alloc_and_init(struct device *gen_dev,
+					     struct pci_dev *pdev,
+					     struct platform_device *op,
+					     const struct niu_ops *ops, u8 port)
+{
+	struct net_device *dev;
+	struct niu *np;
+
+	dev = alloc_etherdev_mq(sizeof(struct niu), NIU_NUM_TXCHAN);
+	if (!dev)
+		return NULL;
+
+	SET_NETDEV_DEV(dev, gen_dev);
+
+	np = netdev_priv(dev);
+	np->dev = dev;
+	np->pdev = pdev;
+	np->op = op;
+	np->device = gen_dev;
+	np->ops = ops;
+
+	np->msg_enable = niu_debug;
+
+	spin_lock_init(&np->lock);
+	INIT_WORK(&np->reset_task, niu_reset_task);
+
+	np->port = port;
+
+	return dev;
+}
+
+static const struct net_device_ops niu_netdev_ops = {
+	.ndo_open		= niu_open,
+	.ndo_stop		= niu_close,
+	.ndo_start_xmit		= niu_start_xmit,
+	.ndo_get_stats64	= niu_get_stats,
+	.ndo_set_rx_mode	= niu_set_rx_mode,
+	.ndo_validate_addr	= eth_validate_addr,
+	.ndo_set_mac_address	= niu_set_mac_addr,
+	.ndo_eth_ioctl		= niu_ioctl,
+	.ndo_tx_timeout		= niu_tx_timeout,
+	.ndo_change_mtu		= niu_change_mtu,
+};
+
+static void niu_assign_netdev_ops(struct net_device *dev)
+{
+	dev->netdev_ops = &niu_netdev_ops;
+	dev->ethtool_ops = &niu_ethtool_ops;
+	dev->watchdog_timeo = NIU_TX_TIMEOUT;
+}
+
+static void niu_device_announce(struct niu *np)
+{
+	struct net_device *dev = np->dev;
+
+	pr_info("%s: NIU Ethernet %pM\n", dev->name, dev->dev_addr);
+
+	if (np->parent->plat_type == PLAT_TYPE_ATCA_CP3220) {
+		pr_info("%s: Port type[%s] mode[%s:%s] XCVR[%s] phy[%s]\n",
+				dev->name,
+				(np->flags & NIU_FLAGS_XMAC ? "XMAC" : "BMAC"),
+				(np->flags & NIU_FLAGS_10G ? "10G" : "1G"),
+				(np->flags & NIU_FLAGS_FIBER ? "RGMII FIBER" : "SERDES"),
+				(np->mac_xcvr == MAC_XCVR_MII ? "MII" :
+				 (np->mac_xcvr == MAC_XCVR_PCS ? "PCS" : "XPCS")),
+				np->vpd.phy_type);
+	} else {
+		pr_info("%s: Port type[%s] mode[%s:%s] XCVR[%s] phy[%s]\n",
+				dev->name,
+				(np->flags & NIU_FLAGS_XMAC ? "XMAC" : "BMAC"),
+				(np->flags & NIU_FLAGS_10G ? "10G" : "1G"),
+				(np->flags & NIU_FLAGS_FIBER ? "FIBER" :
+				 (np->flags & NIU_FLAGS_XCVR_SERDES ? "SERDES" :
+				  "COPPER")),
+				(np->mac_xcvr == MAC_XCVR_MII ? "MII" :
+				 (np->mac_xcvr == MAC_XCVR_PCS ? "PCS" : "XPCS")),
+				np->vpd.phy_type);
+	}
+}
+
+static void niu_set_basic_features(struct net_device *dev)
+{
+	dev->hw_features = NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_RXHASH;
+	dev->features |= dev->hw_features | NETIF_F_RXCSUM;
+}
+
+static int niu_pci_init_one(struct pci_dev *pdev,
+			    const struct pci_device_id *ent)
+{
+	union niu_parent_id parent_id;
+	struct net_device *dev;
+	struct niu *np;
+	int err;
+
+	niu_driver_version();
+
+	err = pci_enable_device(pdev);
+	if (err) {
+		dev_err(&pdev->dev, "Cannot enable PCI device, aborting\n");
+		return err;
+	}
+
+	if (!(pci_resource_flags(pdev, 0) & IORESOURCE_MEM) ||
+	    !(pci_resource_flags(pdev, 2) & IORESOURCE_MEM)) {
+		dev_err(&pdev->dev, "Cannot find proper PCI device base addresses, aborting\n");
+		err = -ENODEV;
+		goto err_out_disable_pdev;
+	}
+
+	err = pci_request_regions(pdev, DRV_MODULE_NAME);
+	if (err) {
+		dev_err(&pdev->dev, "Cannot obtain PCI resources, aborting\n");
+		goto err_out_disable_pdev;
+	}
+
+	if (!pci_is_pcie(pdev)) {
+		dev_err(&pdev->dev, "Cannot find PCI Express capability, aborting\n");
+		err = -ENODEV;
+		goto err_out_free_res;
+	}
+
+	dev = niu_alloc_and_init(&pdev->dev, pdev, NULL,
+				 &niu_pci_ops, PCI_FUNC(pdev->devfn));
+	if (!dev) {
+		err = -ENOMEM;
+		goto err_out_free_res;
+	}
+	np = netdev_priv(dev);
+
+	memset(&parent_id, 0, sizeof(parent_id));
+	parent_id.pci.domain = pci_domain_nr(pdev->bus);
+	parent_id.pci.bus = pdev->bus->number;
+	parent_id.pci.device = PCI_SLOT(pdev->devfn);
+
+	np->parent = niu_get_parent(np, &parent_id,
+				    PLAT_TYPE_ATLAS);
+	if (!np->parent) {
+		err = -ENOMEM;
+		goto err_out_free_dev;
+	}
+
+	pcie_capability_clear_and_set_word(pdev, PCI_EXP_DEVCTL,
+		PCI_EXP_DEVCTL_NOSNOOP_EN,
+		PCI_EXP_DEVCTL_CERE | PCI_EXP_DEVCTL_NFERE |
+		PCI_EXP_DEVCTL_FERE | PCI_EXP_DEVCTL_URRE |
+		PCI_EXP_DEVCTL_RELAX_EN);
+
+	err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(44));
+	if (!err)
+		dev->features |= NETIF_F_HIGHDMA;
+	if (err) {
+		err = dma_set_mask(&pdev->dev, DMA_BIT_MASK(32));
+		if (err) {
+			dev_err(&pdev->dev, "No usable DMA configuration, aborting\n");
+			goto err_out_release_parent;
+		}
+	}
+
+	niu_set_basic_features(dev);
+
+	dev->priv_flags |= IFF_UNICAST_FLT;
+
+	np->regs = pci_ioremap_bar(pdev, 0);
+	if (!np->regs) {
+		dev_err(&pdev->dev, "Cannot map device registers, aborting\n");
+		err = -ENOMEM;
+		goto err_out_release_parent;
+	}
+
+	pci_set_master(pdev);
+	pci_save_state(pdev);
+
+	dev->irq = pdev->irq;
+
+	/* MTU range: 68 - 9216 */
+	dev->min_mtu = ETH_MIN_MTU;
+	dev->max_mtu = NIU_MAX_MTU;
+
+	niu_assign_netdev_ops(dev);
+
+	err = niu_get_invariants(np);
+	if (err) {
+		if (err != -ENODEV)
+			dev_err(&pdev->dev, "Problem fetching invariants of chip, aborting\n");
+		goto err_out_iounmap;
+	}
+
+	err = register_netdev(dev);
+	if (err) {
+		dev_err(&pdev->dev, "Cannot register net device, aborting\n");
+		goto err_out_iounmap;
+	}
+
+	pci_set_drvdata(pdev, dev);
+
+	niu_device_announce(np);
+
+	return 0;
+
+err_out_iounmap:
+	if (np->regs) {
+		iounmap(np->regs);
+		np->regs = NULL;
+	}
+
+err_out_release_parent:
+	niu_put_parent(np);
+
+err_out_free_dev:
+	free_netdev(dev);
+
+err_out_free_res:
+	pci_release_regions(pdev);
+
+err_out_disable_pdev:
+	pci_disable_device(pdev);
+
+	return err;
+}
+
+static void niu_pci_remove_one(struct pci_dev *pdev)
+{
+	struct net_device *dev = pci_get_drvdata(pdev);
+
+	if (dev) {
+		struct niu *np = netdev_priv(dev);
+
+		unregister_netdev(dev);
+		if (np->regs) {
+			iounmap(np->regs);
+			np->regs = NULL;
+		}
+
+		niu_ldg_free(np);
+
+		niu_put_parent(np);
+
+		free_netdev(dev);
+		pci_release_regions(pdev);
+		pci_disable_device(pdev);
+	}
+}
+
+static int __maybe_unused niu_suspend(struct device *dev_d)
+{
+	struct net_device *dev = dev_get_drvdata(dev_d);
+	struct niu *np = netdev_priv(dev);
+	unsigned long flags;
+
+	if (!netif_running(dev))
+		return 0;
+
+	flush_work(&np->reset_task);
+	niu_netif_stop(np);
+
+	timer_delete_sync(&np->timer);
+
+	spin_lock_irqsave(&np->lock, flags);
+	niu_enable_interrupts(np, 0);
+	spin_unlock_irqrestore(&np->lock, flags);
+
+	netif_device_detach(dev);
+
+	spin_lock_irqsave(&np->lock, flags);
+	niu_stop_hw(np);
+	spin_unlock_irqrestore(&np->lock, flags);
+
+	return 0;
+}
+
+static int __maybe_unused niu_resume(struct device *dev_d)
+{
+	struct net_device *dev = dev_get_drvdata(dev_d);
+	struct niu *np = netdev_priv(dev);
+	unsigned long flags;
+	int err;
+
+	if (!netif_running(dev))
+		return 0;
+
+	netif_device_attach(dev);
+
+	spin_lock_irqsave(&np->lock, flags);
+
+	netdev_lock(dev);
+	err = niu_init_hw(np);
+	if (!err) {
+		np->timer.expires = jiffies + HZ;
+		add_timer(&np->timer);
+		niu_netif_start(np);
+	}
+
+	spin_unlock_irqrestore(&np->lock, flags);
+	netdev_unlock(dev);
+
+	return err;
+}
+
+static SIMPLE_DEV_PM_OPS(niu_pm_ops, niu_suspend, niu_resume);
+
+static struct pci_driver niu_pci_driver = {
+	.name		= DRV_MODULE_NAME,
+	.id_table	= niu_pci_tbl,
+	.probe		= niu_pci_init_one,
+	.remove		= niu_pci_remove_one,
+	.driver.pm	= &niu_pm_ops,
+};
+
+#ifdef CONFIG_SPARC64
+static void *niu_phys_alloc_coherent(struct device *dev, size_t size,
+				     u64 *dma_addr, gfp_t flag)
+{
+	unsigned long order = get_order(size);
+	unsigned long page = __get_free_pages(flag, order);
+
+	if (page == 0UL)
+		return NULL;
+	memset((char *)page, 0, PAGE_SIZE << order);
+	*dma_addr = __pa(page);
+
+	return (void *) page;
+}
+
+static void niu_phys_free_coherent(struct device *dev, size_t size,
+				   void *cpu_addr, u64 handle)
+{
+	unsigned long order = get_order(size);
+
+	free_pages((unsigned long) cpu_addr, order);
+}
+
+static u64 niu_phys_map_page(struct device *dev, struct page *page,
+			     unsigned long offset, size_t size,
+			     enum dma_data_direction direction)
+{
+	return page_to_phys(page) + offset;
+}
+
+static void niu_phys_unmap_page(struct device *dev, u64 dma_address,
+				size_t size, enum dma_data_direction direction)
+{
+	/* Nothing to do.  */
+}
+
+static u64 niu_phys_map_single(struct device *dev, void *cpu_addr,
+			       size_t size,
+			       enum dma_data_direction direction)
+{
+	return __pa(cpu_addr);
+}
+
+static void niu_phys_unmap_single(struct device *dev, u64 dma_address,
+				  size_t size,
+				  enum dma_data_direction direction)
+{
+	/* Nothing to do.  */
+}
+
+static int niu_phys_mapping_error(struct device *dev, u64 dma_address)
+{
+	return false;
+}
+
+static const struct niu_ops niu_phys_ops = {
+	.alloc_coherent	= niu_phys_alloc_coherent,
+	.free_coherent	= niu_phys_free_coherent,
+	.map_page	= niu_phys_map_page,
+	.unmap_page	= niu_phys_unmap_page,
+	.map_single	= niu_phys_map_single,
+	.unmap_single	= niu_phys_unmap_single,
+	.mapping_error	= niu_phys_mapping_error,
+};
+
+static int niu_of_probe(struct platform_device *op)
+{
+	union niu_parent_id parent_id;
+	struct net_device *dev;
+	struct niu *np;
+	const u32 *reg;
+	int err;
+
+	niu_driver_version();
+
+	reg = of_get_property(op->dev.of_node, "reg", NULL);
+	if (!reg) {
+		dev_err(&op->dev, "%pOF: No 'reg' property, aborting\n",
+			op->dev.of_node);
+		return -ENODEV;
+	}
+
+	dev = niu_alloc_and_init(&op->dev, NULL, op,
+				 &niu_phys_ops, reg[0] & 0x1);
+	if (!dev) {
+		err = -ENOMEM;
+		goto err_out;
+	}
+	np = netdev_priv(dev);
+
+	memset(&parent_id, 0, sizeof(parent_id));
+	parent_id.of = of_get_parent(op->dev.of_node);
+
+	np->parent = niu_get_parent(np, &parent_id,
+				    PLAT_TYPE_NIU);
+	if (!np->parent) {
+		err = -ENOMEM;
+		goto err_out_free_dev;
+	}
+
+	niu_set_basic_features(dev);
+
+	np->regs = of_ioremap(&op->resource[1], 0,
+			      resource_size(&op->resource[1]),
+			      "niu regs");
+	if (!np->regs) {
+		dev_err(&op->dev, "Cannot map device registers, aborting\n");
+		err = -ENOMEM;
+		goto err_out_release_parent;
+	}
+
+	np->vir_regs_1 = of_ioremap(&op->resource[2], 0,
+				    resource_size(&op->resource[2]),
+				    "niu vregs-1");
+	if (!np->vir_regs_1) {
+		dev_err(&op->dev, "Cannot map device vir registers 1, aborting\n");
+		err = -ENOMEM;
+		goto err_out_iounmap;
+	}
+
+	np->vir_regs_2 = of_ioremap(&op->resource[3], 0,
+				    resource_size(&op->resource[3]),
+				    "niu vregs-2");
+	if (!np->vir_regs_2) {
+		dev_err(&op->dev, "Cannot map device vir registers 2, aborting\n");
+		err = -ENOMEM;
+		goto err_out_iounmap;
+	}
+
+	niu_assign_netdev_ops(dev);
+
+	err = niu_get_invariants(np);
+	if (err) {
+		if (err != -ENODEV)
+			dev_err(&op->dev, "Problem fetching invariants of chip, aborting\n");
+		goto err_out_iounmap;
+	}
+
+	err = register_netdev(dev);
+	if (err) {
+		dev_err(&op->dev, "Cannot register net device, aborting\n");
+		goto err_out_iounmap;
+	}
+
+	platform_set_drvdata(op, dev);
+
+	niu_device_announce(np);
+
+	return 0;
+
+err_out_iounmap:
+	if (np->vir_regs_1) {
+		of_iounmap(&op->resource[2], np->vir_regs_1,
+			   resource_size(&op->resource[2]));
+		np->vir_regs_1 = NULL;
+	}
+
+	if (np->vir_regs_2) {
+		of_iounmap(&op->resource[3], np->vir_regs_2,
+			   resource_size(&op->resource[3]));
+		np->vir_regs_2 = NULL;
+	}
+
+	if (np->regs) {
+		of_iounmap(&op->resource[1], np->regs,
+			   resource_size(&op->resource[1]));
+		np->regs = NULL;
+	}
+
+err_out_release_parent:
+	niu_put_parent(np);
+
+err_out_free_dev:
+	free_netdev(dev);
+
+err_out:
+	return err;
+}
+
+static void niu_of_remove(struct platform_device *op)
+{
+	struct net_device *dev = platform_get_drvdata(op);
+
+	if (dev) {
+		struct niu *np = netdev_priv(dev);
+
+		unregister_netdev(dev);
+
+		if (np->vir_regs_1) {
+			of_iounmap(&op->resource[2], np->vir_regs_1,
+				   resource_size(&op->resource[2]));
+			np->vir_regs_1 = NULL;
+		}
+
+		if (np->vir_regs_2) {
+			of_iounmap(&op->resource[3], np->vir_regs_2,
+				   resource_size(&op->resource[3]));
+			np->vir_regs_2 = NULL;
+		}
+
+		if (np->regs) {
+			of_iounmap(&op->resource[1], np->regs,
+				   resource_size(&op->resource[1]));
+			np->regs = NULL;
+		}
+
+		niu_ldg_free(np);
+
+		niu_put_parent(np);
+
+		free_netdev(dev);
+	}
+}
+
+static const struct of_device_id niu_match[] = {
+	{
+		.name = "network",
+		.compatible = "SUNW,niusl",
+	},
+	{},
+};
+MODULE_DEVICE_TABLE(of, niu_match);
+
+static struct platform_driver niu_of_driver = {
+	.driver = {
+		.name = "niu",
+		.of_match_table = niu_match,
+	},
+	.probe		= niu_of_probe,
+	.remove		= niu_of_remove,
+};
+
+#endif /* CONFIG_SPARC64 */
+
+static int __init niu_init(void)
+{
+	int err = 0;
+
+	BUILD_BUG_ON(PAGE_SIZE < 4 * 1024);
+
+	BUILD_BUG_ON(offsetof(struct page, mapping) !=
+		     offsetof(union niu_page, next));
+
+	niu_debug = netif_msg_init(debug, NIU_MSG_DEFAULT);
+
+#ifdef CONFIG_SPARC64
+	err = platform_driver_register(&niu_of_driver);
+#endif
+
+	if (!err) {
+		err = pci_register_driver(&niu_pci_driver);
+#ifdef CONFIG_SPARC64
+		if (err)
+			platform_driver_unregister(&niu_of_driver);
+#endif
+	}
+
+	return err;
+}
+
+static void __exit niu_exit(void)
+{
+	pci_unregister_driver(&niu_pci_driver);
+#ifdef CONFIG_SPARC64
+	platform_driver_unregister(&niu_of_driver);
+#endif
+}
+
+module_init(niu_init);
+module_exit(niu_exit);
diff --git a/drivers/net/ethernet/sun/niu.h b/drivers/net/ethernet/sun/niu.h
new file mode 100644
index 000000000..676d31499
--- /dev/null
+++ b/drivers/net/ethernet/sun/niu.h
@@ -0,0 +1,3311 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/* niu.h: Definitions for Neptune ethernet driver.
+ *
+ * Copyright (C) 2007 David S. Miller (davem@davemloft.net)
+ */
+
+#ifndef _NIU_H
+#define _NIU_H
+
+#define PIO			0x000000UL
+#define FZC_PIO			0x080000UL
+#define FZC_MAC			0x180000UL
+#define FZC_IPP			0x280000UL
+#define FFLP			0x300000UL
+#define FZC_FFLP		0x380000UL
+#define PIO_VADDR		0x400000UL
+#define ZCP			0x500000UL
+#define FZC_ZCP			0x580000UL
+#define DMC			0x600000UL
+#define FZC_DMC			0x680000UL
+#define TXC			0x700000UL
+#define FZC_TXC			0x780000UL
+#define PIO_LDSV		0x800000UL
+#define PIO_PIO_LDGIM		0x900000UL
+#define PIO_IMASK0		0xa00000UL
+#define PIO_IMASK1		0xb00000UL
+#define FZC_PROM		0xc80000UL
+#define FZC_PIM			0xd80000UL
+
+#define LDSV0(LDG)		(PIO_LDSV + 0x00000UL + (LDG) * 0x2000UL)
+#define LDSV1(LDG)		(PIO_LDSV + 0x00008UL + (LDG) * 0x2000UL)
+#define LDSV2(LDG)		(PIO_LDSV + 0x00010UL + (LDG) * 0x2000UL)
+
+#define LDG_IMGMT(LDG)		(PIO_LDSV + 0x00018UL + (LDG) * 0x2000UL)
+#define  LDG_IMGMT_ARM		0x0000000080000000ULL
+#define  LDG_IMGMT_TIMER	0x000000000000003fULL
+
+#define LD_IM0(IDX)		(PIO_IMASK0 + 0x00000UL + (IDX) * 0x2000UL)
+#define  LD_IM0_MASK		0x0000000000000003ULL
+
+#define LD_IM1(IDX)		(PIO_IMASK1 + 0x00000UL + (IDX) * 0x2000UL)
+#define  LD_IM1_MASK		0x0000000000000003ULL
+
+#define LDG_TIMER_RES		(FZC_PIO + 0x00008UL)
+#define  LDG_TIMER_RES_VAL	0x00000000000fffffULL
+
+#define DIRTY_TID_CTL		(FZC_PIO + 0x00010UL)
+#define  DIRTY_TID_CTL_NPTHRED	0x00000000003f0000ULL
+#define  DIRTY_TID_CTL_RDTHRED	0x00000000000003f0ULL
+#define  DIRTY_TID_CTL_DTIDCLR	0x0000000000000002ULL
+#define  DIRTY_TID_CTL_DTIDENAB	0x0000000000000001ULL
+
+#define DIRTY_TID_STAT		(FZC_PIO + 0x00018UL)
+#define  DIRTY_TID_STAT_NPWSTAT	0x0000000000003f00ULL
+#define  DIRTY_TID_STAT_RDSTAT	0x000000000000003fULL
+
+#define RST_CTL			(FZC_PIO + 0x00038UL)
+#define  RST_CTL_MAC_RST3	0x0000000000400000ULL
+#define  RST_CTL_MAC_RST2	0x0000000000200000ULL
+#define  RST_CTL_MAC_RST1	0x0000000000100000ULL
+#define  RST_CTL_MAC_RST0	0x0000000000080000ULL
+#define  RST_CTL_ACK_TO_EN	0x0000000000000800ULL
+#define  RST_CTL_ACK_TO_VAL	0x00000000000007feULL
+
+#define SMX_CFIG_DAT		(FZC_PIO + 0x00040UL)
+#define  SMX_CFIG_DAT_RAS_DET	0x0000000080000000ULL
+#define  SMX_CFIG_DAT_RAS_INJ	0x0000000040000000ULL
+#define  SMX_CFIG_DAT_XACT_TO	0x000000000fffffffULL
+
+#define SMX_INT_STAT		(FZC_PIO + 0x00048UL)
+#define  SMX_INT_STAT_STAT	0x00000000ffffffffULL
+
+#define SMX_CTL			(FZC_PIO + 0x00050UL)
+#define  SMX_CTL_CTL		0x00000000ffffffffULL
+
+#define SMX_DBG_VEC		(FZC_PIO + 0x00058UL)
+#define  SMX_DBG_VEC_VEC	0x00000000ffffffffULL
+
+#define PIO_DBG_SEL		(FZC_PIO + 0x00060UL)
+#define  PIO_DBG_SEL_SEL	0x000000000000003fULL
+
+#define PIO_TRAIN_VEC		(FZC_PIO + 0x00068UL)
+#define  PIO_TRAIN_VEC_VEC	0x00000000ffffffffULL
+
+#define PIO_ARB_CTL		(FZC_PIO + 0x00070UL)
+#define  PIO_ARB_CTL_CTL	0x00000000ffffffffULL
+
+#define PIO_ARB_DBG_VEC		(FZC_PIO + 0x00078UL)
+#define  PIO_ARB_DBG_VEC_VEC	0x00000000ffffffffULL
+
+#define SYS_ERR_MASK		(FZC_PIO + 0x00090UL)
+#define  SYS_ERR_MASK_META2	0x0000000000000400ULL
+#define  SYS_ERR_MASK_META1	0x0000000000000200ULL
+#define  SYS_ERR_MASK_PEU	0x0000000000000100ULL
+#define  SYS_ERR_MASK_TXC	0x0000000000000080ULL
+#define  SYS_ERR_MASK_RDMC	0x0000000000000040ULL
+#define  SYS_ERR_MASK_TDMC	0x0000000000000020ULL
+#define  SYS_ERR_MASK_ZCP	0x0000000000000010ULL
+#define  SYS_ERR_MASK_FFLP	0x0000000000000008ULL
+#define  SYS_ERR_MASK_IPP	0x0000000000000004ULL
+#define  SYS_ERR_MASK_MAC	0x0000000000000002ULL
+#define  SYS_ERR_MASK_SMX	0x0000000000000001ULL
+
+#define SYS_ERR_STAT			(FZC_PIO + 0x00098UL)
+#define  SYS_ERR_STAT_META2		0x0000000000000400ULL
+#define  SYS_ERR_STAT_META1		0x0000000000000200ULL
+#define  SYS_ERR_STAT_PEU		0x0000000000000100ULL
+#define  SYS_ERR_STAT_TXC		0x0000000000000080ULL
+#define  SYS_ERR_STAT_RDMC		0x0000000000000040ULL
+#define  SYS_ERR_STAT_TDMC		0x0000000000000020ULL
+#define  SYS_ERR_STAT_ZCP		0x0000000000000010ULL
+#define  SYS_ERR_STAT_FFLP		0x0000000000000008ULL
+#define  SYS_ERR_STAT_IPP		0x0000000000000004ULL
+#define  SYS_ERR_STAT_MAC		0x0000000000000002ULL
+#define  SYS_ERR_STAT_SMX		0x0000000000000001ULL
+
+#define SID(LDG)			(FZC_PIO + 0x10200UL + (LDG) * 8UL)
+#define  SID_FUNC			0x0000000000000060ULL
+#define  SID_FUNC_SHIFT			5
+#define  SID_VECTOR			0x000000000000001fULL
+#define  SID_VECTOR_SHIFT		0
+
+#define LDG_NUM(LDN)			(FZC_PIO + 0x20000UL + (LDN) * 8UL)
+
+#define XMAC_PORT0_OFF			(FZC_MAC + 0x000000)
+#define XMAC_PORT1_OFF			(FZC_MAC + 0x006000)
+#define BMAC_PORT2_OFF			(FZC_MAC + 0x00c000)
+#define BMAC_PORT3_OFF			(FZC_MAC + 0x010000)
+
+/* XMAC registers, offset from np->mac_regs  */
+
+#define XTXMAC_SW_RST			0x00000UL
+#define  XTXMAC_SW_RST_REG_RS		0x0000000000000002ULL
+#define  XTXMAC_SW_RST_SOFT_RST		0x0000000000000001ULL
+
+#define XRXMAC_SW_RST			0x00008UL
+#define  XRXMAC_SW_RST_REG_RS		0x0000000000000002ULL
+#define  XRXMAC_SW_RST_SOFT_RST		0x0000000000000001ULL
+
+#define XTXMAC_STATUS			0x00020UL
+#define  XTXMAC_STATUS_FRAME_CNT_EXP	0x0000000000000800ULL
+#define  XTXMAC_STATUS_BYTE_CNT_EXP	0x0000000000000400ULL
+#define  XTXMAC_STATUS_TXFIFO_XFR_ERR	0x0000000000000010ULL
+#define  XTXMAC_STATUS_TXMAC_OFLOW	0x0000000000000008ULL
+#define  XTXMAC_STATUS_MAX_PSIZE_ERR	0x0000000000000004ULL
+#define  XTXMAC_STATUS_TXMAC_UFLOW	0x0000000000000002ULL
+#define  XTXMAC_STATUS_FRAME_XMITED	0x0000000000000001ULL
+
+#define XRXMAC_STATUS			0x00028UL
+#define  XRXMAC_STATUS_RXHIST7_CNT_EXP	0x0000000000100000ULL
+#define  XRXMAC_STATUS_LCL_FLT_STATUS	0x0000000000080000ULL
+#define  XRXMAC_STATUS_RFLT_DET		0x0000000000040000ULL
+#define  XRXMAC_STATUS_LFLT_CNT_EXP	0x0000000000020000ULL
+#define  XRXMAC_STATUS_PHY_MDINT	0x0000000000010000ULL
+#define  XRXMAC_STATUS_ALIGNERR_CNT_EXP	0x0000000000010000ULL
+#define  XRXMAC_STATUS_RXFRAG_CNT_EXP	0x0000000000008000ULL
+#define  XRXMAC_STATUS_RXMULTF_CNT_EXP	0x0000000000004000ULL
+#define  XRXMAC_STATUS_RXBCAST_CNT_EXP	0x0000000000002000ULL
+#define  XRXMAC_STATUS_RXHIST6_CNT_EXP	0x0000000000001000ULL
+#define  XRXMAC_STATUS_RXHIST5_CNT_EXP	0x0000000000000800ULL
+#define  XRXMAC_STATUS_RXHIST4_CNT_EXP	0x0000000000000400ULL
+#define  XRXMAC_STATUS_RXHIST3_CNT_EXP	0x0000000000000200ULL
+#define  XRXMAC_STATUS_RXHIST2_CNT_EXP	0x0000000000000100ULL
+#define  XRXMAC_STATUS_RXHIST1_CNT_EXP	0x0000000000000080ULL
+#define  XRXMAC_STATUS_RXOCTET_CNT_EXP	0x0000000000000040ULL
+#define  XRXMAC_STATUS_CVIOLERR_CNT_EXP	0x0000000000000020ULL
+#define  XRXMAC_STATUS_LENERR_CNT_EXP	0x0000000000000010ULL
+#define  XRXMAC_STATUS_CRCERR_CNT_EXP	0x0000000000000008ULL
+#define  XRXMAC_STATUS_RXUFLOW		0x0000000000000004ULL
+#define  XRXMAC_STATUS_RXOFLOW		0x0000000000000002ULL
+#define  XRXMAC_STATUS_FRAME_RCVD	0x0000000000000001ULL
+
+#define XMAC_FC_STAT			0x00030UL
+#define  XMAC_FC_STAT_RX_RCV_PAUSE_TIME	0x00000000ffff0000ULL
+#define  XMAC_FC_STAT_TX_MAC_NPAUSE	0x0000000000000004ULL
+#define  XMAC_FC_STAT_TX_MAC_PAUSE	0x0000000000000002ULL
+#define  XMAC_FC_STAT_RX_MAC_RPAUSE	0x0000000000000001ULL
+
+#define XTXMAC_STAT_MSK			0x00040UL
+#define  XTXMAC_STAT_MSK_FRAME_CNT_EXP	0x0000000000000800ULL
+#define  XTXMAC_STAT_MSK_BYTE_CNT_EXP	0x0000000000000400ULL
+#define  XTXMAC_STAT_MSK_TXFIFO_XFR_ERR	0x0000000000000010ULL
+#define  XTXMAC_STAT_MSK_TXMAC_OFLOW	0x0000000000000008ULL
+#define  XTXMAC_STAT_MSK_MAX_PSIZE_ERR	0x0000000000000004ULL
+#define  XTXMAC_STAT_MSK_TXMAC_UFLOW	0x0000000000000002ULL
+#define  XTXMAC_STAT_MSK_FRAME_XMITED	0x0000000000000001ULL
+
+#define XRXMAC_STAT_MSK				0x00048UL
+#define  XRXMAC_STAT_MSK_LCL_FLT_STAT_MSK	0x0000000000080000ULL
+#define  XRXMAC_STAT_MSK_RFLT_DET		0x0000000000040000ULL
+#define  XRXMAC_STAT_MSK_LFLT_CNT_EXP		0x0000000000020000ULL
+#define  XRXMAC_STAT_MSK_PHY_MDINT		0x0000000000010000ULL
+#define  XRXMAC_STAT_MSK_RXFRAG_CNT_EXP		0x0000000000008000ULL
+#define  XRXMAC_STAT_MSK_RXMULTF_CNT_EXP	0x0000000000004000ULL
+#define  XRXMAC_STAT_MSK_RXBCAST_CNT_EXP	0x0000000000002000ULL
+#define  XRXMAC_STAT_MSK_RXHIST6_CNT_EXP	0x0000000000001000ULL
+#define  XRXMAC_STAT_MSK_RXHIST5_CNT_EXP	0x0000000000000800ULL
+#define  XRXMAC_STAT_MSK_RXHIST4_CNT_EXP	0x0000000000000400ULL
+#define  XRXMAC_STAT_MSK_RXHIST3_CNT_EXP	0x0000000000000200ULL
+#define  XRXMAC_STAT_MSK_RXHIST2_CNT_EXP	0x0000000000000100ULL
+#define  XRXMAC_STAT_MSK_RXHIST1_CNT_EXP	0x0000000000000080ULL
+#define  XRXMAC_STAT_MSK_RXOCTET_CNT_EXP	0x0000000000000040ULL
+#define  XRXMAC_STAT_MSK_CVIOLERR_CNT_EXP	0x0000000000000020ULL
+#define  XRXMAC_STAT_MSK_LENERR_CNT_EXP		0x0000000000000010ULL
+#define  XRXMAC_STAT_MSK_CRCERR_CNT_EXP		0x0000000000000008ULL
+#define  XRXMAC_STAT_MSK_RXUFLOW_CNT_EXP	0x0000000000000004ULL
+#define  XRXMAC_STAT_MSK_RXOFLOW_CNT_EXP	0x0000000000000002ULL
+#define  XRXMAC_STAT_MSK_FRAME_RCVD		0x0000000000000001ULL
+
+#define XMAC_FC_MSK			0x00050UL
+#define  XMAC_FC_MSK_TX_MAC_NPAUSE	0x0000000000000004ULL
+#define  XMAC_FC_MSK_TX_MAC_PAUSE	0x0000000000000002ULL
+#define  XMAC_FC_MSK_RX_MAC_RPAUSE	0x0000000000000001ULL
+
+#define XMAC_CONFIG			0x00060UL
+#define  XMAC_CONFIG_SEL_CLK_25MHZ	0x0000000080000000ULL
+#define  XMAC_CONFIG_1G_PCS_BYPASS	0x0000000040000000ULL
+#define  XMAC_CONFIG_10G_XPCS_BYPASS	0x0000000020000000ULL
+#define  XMAC_CONFIG_MODE_MASK		0x0000000018000000ULL
+#define  XMAC_CONFIG_MODE_XGMII		0x0000000000000000ULL
+#define  XMAC_CONFIG_MODE_GMII		0x0000000008000000ULL
+#define  XMAC_CONFIG_MODE_MII		0x0000000010000000ULL
+#define  XMAC_CONFIG_LFS_DISABLE	0x0000000004000000ULL
+#define  XMAC_CONFIG_LOOPBACK		0x0000000002000000ULL
+#define  XMAC_CONFIG_TX_OUTPUT_EN	0x0000000001000000ULL
+#define  XMAC_CONFIG_SEL_POR_CLK_SRC	0x0000000000800000ULL
+#define  XMAC_CONFIG_LED_POLARITY	0x0000000000400000ULL
+#define  XMAC_CONFIG_FORCE_LED_ON	0x0000000000200000ULL
+#define  XMAC_CONFIG_PASS_FLOW_CTRL	0x0000000000100000ULL
+#define  XMAC_CONFIG_RCV_PAUSE_ENABLE	0x0000000000080000ULL
+#define  XMAC_CONFIG_MAC2IPP_PKT_CNT_EN	0x0000000000040000ULL
+#define  XMAC_CONFIG_STRIP_CRC		0x0000000000020000ULL
+#define  XMAC_CONFIG_ADDR_FILTER_EN	0x0000000000010000ULL
+#define  XMAC_CONFIG_HASH_FILTER_EN	0x0000000000008000ULL
+#define  XMAC_CONFIG_RX_CODEV_CHK_DIS	0x0000000000004000ULL
+#define  XMAC_CONFIG_RESERVED_MULTICAST	0x0000000000002000ULL
+#define  XMAC_CONFIG_RX_CRC_CHK_DIS	0x0000000000001000ULL
+#define  XMAC_CONFIG_ERR_CHK_DIS	0x0000000000000800ULL
+#define  XMAC_CONFIG_PROMISC_GROUP	0x0000000000000400ULL
+#define  XMAC_CONFIG_PROMISCUOUS	0x0000000000000200ULL
+#define  XMAC_CONFIG_RX_MAC_ENABLE	0x0000000000000100ULL
+#define  XMAC_CONFIG_WARNING_MSG_EN	0x0000000000000080ULL
+#define  XMAC_CONFIG_ALWAYS_NO_CRC	0x0000000000000008ULL
+#define  XMAC_CONFIG_VAR_MIN_IPG_EN	0x0000000000000004ULL
+#define  XMAC_CONFIG_STRETCH_MODE	0x0000000000000002ULL
+#define  XMAC_CONFIG_TX_ENABLE		0x0000000000000001ULL
+
+#define XMAC_IPG			0x00080UL
+#define  XMAC_IPG_STRETCH_CONST		0x0000000000e00000ULL
+#define  XMAC_IPG_STRETCH_CONST_SHIFT	21
+#define  XMAC_IPG_STRETCH_RATIO		0x00000000001f0000ULL
+#define  XMAC_IPG_STRETCH_RATIO_SHIFT	16
+#define  XMAC_IPG_IPG_MII_GMII		0x000000000000ff00ULL
+#define  XMAC_IPG_IPG_MII_GMII_SHIFT	8
+#define  XMAC_IPG_IPG_XGMII		0x0000000000000007ULL
+#define  XMAC_IPG_IPG_XGMII_SHIFT	0
+
+#define IPG_12_15_XGMII			3
+#define IPG_16_19_XGMII			4
+#define IPG_20_23_XGMII			5
+#define IPG_12_MII_GMII			10
+#define IPG_13_MII_GMII			11
+#define IPG_14_MII_GMII			12
+#define IPG_15_MII_GMII			13
+#define IPG_16_MII_GMII			14
+
+#define XMAC_MIN			0x00088UL
+#define  XMAC_MIN_RX_MIN_PKT_SIZE	0x000000003ff00000ULL
+#define  XMAC_MIN_RX_MIN_PKT_SIZE_SHFT	20
+#define  XMAC_MIN_SLOT_TIME		0x000000000003fc00ULL
+#define  XMAC_MIN_SLOT_TIME_SHFT	10
+#define  XMAC_MIN_TX_MIN_PKT_SIZE	0x00000000000003ffULL
+#define  XMAC_MIN_TX_MIN_PKT_SIZE_SHFT	0
+
+#define XMAC_MAX			0x00090UL
+#define  XMAC_MAX_FRAME_SIZE		0x0000000000003fffULL
+#define  XMAC_MAX_FRAME_SIZE_SHFT	0
+
+#define XMAC_ADDR0			0x000a0UL
+#define  XMAC_ADDR0_ADDR0		0x000000000000ffffULL
+
+#define XMAC_ADDR1			0x000a8UL
+#define  XMAC_ADDR1_ADDR1		0x000000000000ffffULL
+
+#define XMAC_ADDR2			0x000b0UL
+#define  XMAC_ADDR2_ADDR2		0x000000000000ffffULL
+
+#define XMAC_ADDR_CMPEN			0x00208UL
+#define  XMAC_ADDR_CMPEN_EN15		0x0000000000008000ULL
+#define  XMAC_ADDR_CMPEN_EN14		0x0000000000004000ULL
+#define  XMAC_ADDR_CMPEN_EN13		0x0000000000002000ULL
+#define  XMAC_ADDR_CMPEN_EN12		0x0000000000001000ULL
+#define  XMAC_ADDR_CMPEN_EN11		0x0000000000000800ULL
+#define  XMAC_ADDR_CMPEN_EN10		0x0000000000000400ULL
+#define  XMAC_ADDR_CMPEN_EN9		0x0000000000000200ULL
+#define  XMAC_ADDR_CMPEN_EN8		0x0000000000000100ULL
+#define  XMAC_ADDR_CMPEN_EN7		0x0000000000000080ULL
+#define  XMAC_ADDR_CMPEN_EN6		0x0000000000000040ULL
+#define  XMAC_ADDR_CMPEN_EN5		0x0000000000000020ULL
+#define  XMAC_ADDR_CMPEN_EN4		0x0000000000000010ULL
+#define  XMAC_ADDR_CMPEN_EN3		0x0000000000000008ULL
+#define  XMAC_ADDR_CMPEN_EN2		0x0000000000000004ULL
+#define  XMAC_ADDR_CMPEN_EN1		0x0000000000000002ULL
+#define  XMAC_ADDR_CMPEN_EN0		0x0000000000000001ULL
+
+#define XMAC_NUM_ALT_ADDR		16
+
+#define XMAC_ALT_ADDR0(NUM)		(0x00218UL + (NUM)*0x18UL)
+#define  XMAC_ALT_ADDR0_ADDR0		0x000000000000ffffULL
+
+#define XMAC_ALT_ADDR1(NUM)		(0x00220UL + (NUM)*0x18UL)
+#define  XMAC_ALT_ADDR1_ADDR1		0x000000000000ffffULL
+
+#define XMAC_ALT_ADDR2(NUM)		(0x00228UL + (NUM)*0x18UL)
+#define  XMAC_ALT_ADDR2_ADDR2		0x000000000000ffffULL
+
+#define XMAC_ADD_FILT0			0x00818UL
+#define  XMAC_ADD_FILT0_FILT0		0x000000000000ffffULL
+
+#define XMAC_ADD_FILT1			0x00820UL
+#define  XMAC_ADD_FILT1_FILT1		0x000000000000ffffULL
+
+#define XMAC_ADD_FILT2			0x00828UL
+#define  XMAC_ADD_FILT2_FILT2		0x000000000000ffffULL
+
+#define XMAC_ADD_FILT12_MASK		0x00830UL
+#define  XMAC_ADD_FILT12_MASK_VAL	0x00000000000000ffULL
+
+#define XMAC_ADD_FILT00_MASK		0x00838UL
+#define  XMAC_ADD_FILT00_MASK_VAL	0x000000000000ffffULL
+
+#define XMAC_HASH_TBL(NUM)		(0x00840UL + (NUM) * 0x8UL)
+#define XMAC_HASH_TBL_VAL		0x000000000000ffffULL
+
+#define XMAC_NUM_HOST_INFO		20
+
+#define XMAC_HOST_INFO(NUM)		(0x00900UL + (NUM) * 0x8UL)
+
+#define XMAC_PA_DATA0			0x00b80UL
+#define XMAC_PA_DATA0_VAL		0x00000000ffffffffULL
+
+#define XMAC_PA_DATA1			0x00b88UL
+#define XMAC_PA_DATA1_VAL		0x00000000ffffffffULL
+
+#define XMAC_DEBUG_SEL			0x00b90UL
+#define  XMAC_DEBUG_SEL_XMAC		0x0000000000000078ULL
+#define  XMAC_DEBUG_SEL_MAC		0x0000000000000007ULL
+
+#define XMAC_TRAIN_VEC			0x00b98UL
+#define  XMAC_TRAIN_VEC_VAL		0x00000000ffffffffULL
+
+#define RXMAC_BT_CNT			0x00100UL
+#define  RXMAC_BT_CNT_COUNT		0x00000000ffffffffULL
+
+#define RXMAC_BC_FRM_CNT		0x00108UL
+#define  RXMAC_BC_FRM_CNT_COUNT		0x00000000001fffffULL
+
+#define RXMAC_MC_FRM_CNT		0x00110UL
+#define  RXMAC_MC_FRM_CNT_COUNT		0x00000000001fffffULL
+
+#define RXMAC_FRAG_CNT			0x00118UL
+#define  RXMAC_FRAG_CNT_COUNT		0x00000000001fffffULL
+
+#define RXMAC_HIST_CNT1			0x00120UL
+#define  RXMAC_HIST_CNT1_COUNT		0x00000000001fffffULL
+
+#define RXMAC_HIST_CNT2			0x00128UL
+#define  RXMAC_HIST_CNT2_COUNT		0x00000000001fffffULL
+
+#define RXMAC_HIST_CNT3			0x00130UL
+#define  RXMAC_HIST_CNT3_COUNT		0x00000000000fffffULL
+
+#define RXMAC_HIST_CNT4			0x00138UL
+#define  RXMAC_HIST_CNT4_COUNT		0x000000000007ffffULL
+
+#define RXMAC_HIST_CNT5			0x00140UL
+#define  RXMAC_HIST_CNT5_COUNT		0x000000000003ffffULL
+
+#define RXMAC_HIST_CNT6			0x00148UL
+#define  RXMAC_HIST_CNT6_COUNT		0x000000000000ffffULL
+
+#define RXMAC_MPSZER_CNT		0x00150UL
+#define  RXMAC_MPSZER_CNT_COUNT		0x00000000000000ffULL
+
+#define RXMAC_CRC_ER_CNT		0x00158UL
+#define  RXMAC_CRC_ER_CNT_COUNT		0x00000000000000ffULL
+
+#define RXMAC_CD_VIO_CNT		0x00160UL
+#define  RXMAC_CD_VIO_CNT_COUNT		0x00000000000000ffULL
+
+#define RXMAC_ALIGN_ERR_CNT		0x00168UL
+#define  RXMAC_ALIGN_ERR_CNT_COUNT	0x00000000000000ffULL
+
+#define TXMAC_FRM_CNT			0x00170UL
+#define  TXMAC_FRM_CNT_COUNT		0x00000000ffffffffULL
+
+#define TXMAC_BYTE_CNT			0x00178UL
+#define  TXMAC_BYTE_CNT_COUNT		0x00000000ffffffffULL
+
+#define LINK_FAULT_CNT			0x00180UL
+#define  LINK_FAULT_CNT_COUNT		0x00000000000000ffULL
+
+#define RXMAC_HIST_CNT7			0x00188UL
+#define  RXMAC_HIST_CNT7_COUNT		0x0000000007ffffffULL
+
+#define XMAC_SM_REG			0x001a8UL
+#define  XMAC_SM_REG_STATE		0x00000000ffffffffULL
+
+#define XMAC_INTER1			0x001b0UL
+#define  XMAC_INTERN1_SIGNALS1		0x00000000ffffffffULL
+
+#define XMAC_INTER2			0x001b8UL
+#define  XMAC_INTERN2_SIGNALS2		0x00000000ffffffffULL
+
+/* BMAC registers, offset from np->mac_regs  */
+
+#define BTXMAC_SW_RST			0x00000UL
+#define  BTXMAC_SW_RST_RESET		0x0000000000000001ULL
+
+#define BRXMAC_SW_RST			0x00008UL
+#define  BRXMAC_SW_RST_RESET		0x0000000000000001ULL
+
+#define BMAC_SEND_PAUSE			0x00010UL
+#define  BMAC_SEND_PAUSE_SEND		0x0000000000010000ULL
+#define  BMAC_SEND_PAUSE_TIME		0x000000000000ffffULL
+
+#define BTXMAC_STATUS			0x00020UL
+#define  BTXMAC_STATUS_XMIT		0x0000000000000001ULL
+#define  BTXMAC_STATUS_UNDERRUN		0x0000000000000002ULL
+#define  BTXMAC_STATUS_MAX_PKT_ERR	0x0000000000000004ULL
+#define  BTXMAC_STATUS_BYTE_CNT_EXP	0x0000000000000400ULL
+#define  BTXMAC_STATUS_FRAME_CNT_EXP	0x0000000000000800ULL
+
+#define BRXMAC_STATUS			0x00028UL
+#define  BRXMAC_STATUS_RX_PKT		0x0000000000000001ULL
+#define  BRXMAC_STATUS_OVERFLOW		0x0000000000000002ULL
+#define  BRXMAC_STATUS_FRAME_CNT_EXP	0x0000000000000004ULL
+#define  BRXMAC_STATUS_ALIGN_ERR_EXP	0x0000000000000008ULL
+#define  BRXMAC_STATUS_CRC_ERR_EXP	0x0000000000000010ULL
+#define  BRXMAC_STATUS_LEN_ERR_EXP	0x0000000000000020ULL
+
+#define BMAC_CTRL_STATUS		0x00030UL
+#define  BMAC_CTRL_STATUS_PAUSE_RECV	0x0000000000000001ULL
+#define  BMAC_CTRL_STATUS_PAUSE		0x0000000000000002ULL
+#define  BMAC_CTRL_STATUS_NOPAUSE	0x0000000000000004ULL
+#define  BMAC_CTRL_STATUS_TIME		0x00000000ffff0000ULL
+#define  BMAC_CTRL_STATUS_TIME_SHIFT	16
+
+#define BTXMAC_STATUS_MASK		0x00040UL
+#define BRXMAC_STATUS_MASK		0x00048UL
+#define BMAC_CTRL_STATUS_MASK		0x00050UL
+
+#define BTXMAC_CONFIG			0x00060UL
+#define  BTXMAC_CONFIG_ENABLE		0x0000000000000001ULL
+#define  BTXMAC_CONFIG_FCS_DISABLE	0x0000000000000002ULL
+
+#define BRXMAC_CONFIG			0x00068UL
+#define  BRXMAC_CONFIG_DISCARD_DIS	0x0000000000000080ULL
+#define  BRXMAC_CONFIG_ADDR_FILT_EN	0x0000000000000040ULL
+#define  BRXMAC_CONFIG_HASH_FILT_EN	0x0000000000000020ULL
+#define  BRXMAC_CONFIG_PROMISC_GRP	0x0000000000000010ULL
+#define  BRXMAC_CONFIG_PROMISC		0x0000000000000008ULL
+#define  BRXMAC_CONFIG_STRIP_FCS	0x0000000000000004ULL
+#define  BRXMAC_CONFIG_STRIP_PAD	0x0000000000000002ULL
+#define  BRXMAC_CONFIG_ENABLE		0x0000000000000001ULL
+
+#define BMAC_CTRL_CONFIG		0x00070UL
+#define  BMAC_CTRL_CONFIG_TX_PAUSE_EN	0x0000000000000001ULL
+#define  BMAC_CTRL_CONFIG_RX_PAUSE_EN	0x0000000000000002ULL
+#define  BMAC_CTRL_CONFIG_PASS_CTRL	0x0000000000000004ULL
+
+#define BMAC_XIF_CONFIG			0x00078UL
+#define  BMAC_XIF_CONFIG_TX_OUTPUT_EN	0x0000000000000001ULL
+#define  BMAC_XIF_CONFIG_MII_LOOPBACK	0x0000000000000002ULL
+#define  BMAC_XIF_CONFIG_GMII_MODE	0x0000000000000008ULL
+#define  BMAC_XIF_CONFIG_LINK_LED	0x0000000000000020ULL
+#define  BMAC_XIF_CONFIG_LED_POLARITY	0x0000000000000040ULL
+#define  BMAC_XIF_CONFIG_25MHZ_CLOCK	0x0000000000000080ULL
+
+#define BMAC_MIN_FRAME			0x000a0UL
+#define  BMAC_MIN_FRAME_VAL		0x00000000000003ffULL
+
+#define BMAC_MAX_FRAME			0x000a8UL
+#define  BMAC_MAX_FRAME_MAX_BURST	0x000000003fff0000ULL
+#define  BMAC_MAX_FRAME_MAX_BURST_SHIFT	16
+#define  BMAC_MAX_FRAME_MAX_FRAME	0x0000000000003fffULL
+#define  BMAC_MAX_FRAME_MAX_FRAME_SHIFT	0
+
+#define BMAC_PREAMBLE_SIZE		0x000b0UL
+#define  BMAC_PREAMBLE_SIZE_VAL		0x00000000000003ffULL
+
+#define BMAC_CTRL_TYPE			0x000c8UL
+
+#define BMAC_ADDR0			0x00100UL
+#define  BMAC_ADDR0_ADDR0		0x000000000000ffffULL
+
+#define BMAC_ADDR1			0x00108UL
+#define  BMAC_ADDR1_ADDR1		0x000000000000ffffULL
+
+#define BMAC_ADDR2			0x00110UL
+#define  BMAC_ADDR2_ADDR2		0x000000000000ffffULL
+
+#define BMAC_NUM_ALT_ADDR		6
+
+#define BMAC_ALT_ADDR0(NUM)		(0x00118UL + (NUM)*0x18UL)
+#define  BMAC_ALT_ADDR0_ADDR0		0x000000000000ffffULL
+
+#define BMAC_ALT_ADDR1(NUM)		(0x00120UL + (NUM)*0x18UL)
+#define  BMAC_ALT_ADDR1_ADDR1		0x000000000000ffffULL
+
+#define BMAC_ALT_ADDR2(NUM)		(0x00128UL + (NUM)*0x18UL)
+#define  BMAC_ALT_ADDR2_ADDR2		0x000000000000ffffULL
+
+#define BMAC_FC_ADDR0			0x00268UL
+#define  BMAC_FC_ADDR0_ADDR0		0x000000000000ffffULL
+
+#define BMAC_FC_ADDR1			0x00270UL
+#define  BMAC_FC_ADDR1_ADDR1		0x000000000000ffffULL
+
+#define BMAC_FC_ADDR2			0x00278UL
+#define  BMAC_FC_ADDR2_ADDR2		0x000000000000ffffULL
+
+#define BMAC_ADD_FILT0			0x00298UL
+#define  BMAC_ADD_FILT0_FILT0		0x000000000000ffffULL
+
+#define BMAC_ADD_FILT1			0x002a0UL
+#define  BMAC_ADD_FILT1_FILT1		0x000000000000ffffULL
+
+#define BMAC_ADD_FILT2			0x002a8UL
+#define  BMAC_ADD_FILT2_FILT2		0x000000000000ffffULL
+
+#define BMAC_ADD_FILT12_MASK		0x002b0UL
+#define  BMAC_ADD_FILT12_MASK_VAL	0x00000000000000ffULL
+
+#define BMAC_ADD_FILT00_MASK		0x002b8UL
+#define  BMAC_ADD_FILT00_MASK_VAL	0x000000000000ffffULL
+
+#define BMAC_HASH_TBL(NUM)		(0x002c0UL + (NUM) * 0x8UL)
+#define BMAC_HASH_TBL_VAL		0x000000000000ffffULL
+
+#define BRXMAC_FRAME_CNT		0x00370
+#define  BRXMAC_FRAME_CNT_COUNT		0x000000000000ffffULL
+
+#define BRXMAC_MAX_LEN_ERR_CNT		0x00378
+
+#define BRXMAC_ALIGN_ERR_CNT		0x00380
+#define  BRXMAC_ALIGN_ERR_CNT_COUNT	0x000000000000ffffULL
+
+#define BRXMAC_CRC_ERR_CNT		0x00388
+#define  BRXMAC_ALIGN_ERR_CNT_COUNT	0x000000000000ffffULL
+
+#define BRXMAC_CODE_VIOL_ERR_CNT	0x00390
+#define  BRXMAC_CODE_VIOL_ERR_CNT_COUNT	0x000000000000ffffULL
+
+#define BMAC_STATE_MACHINE		0x003a0
+
+#define BMAC_ADDR_CMPEN			0x003f8UL
+#define  BMAC_ADDR_CMPEN_EN15		0x0000000000008000ULL
+#define  BMAC_ADDR_CMPEN_EN14		0x0000000000004000ULL
+#define  BMAC_ADDR_CMPEN_EN13		0x0000000000002000ULL
+#define  BMAC_ADDR_CMPEN_EN12		0x0000000000001000ULL
+#define  BMAC_ADDR_CMPEN_EN11		0x0000000000000800ULL
+#define  BMAC_ADDR_CMPEN_EN10		0x0000000000000400ULL
+#define  BMAC_ADDR_CMPEN_EN9		0x0000000000000200ULL
+#define  BMAC_ADDR_CMPEN_EN8		0x0000000000000100ULL
+#define  BMAC_ADDR_CMPEN_EN7		0x0000000000000080ULL
+#define  BMAC_ADDR_CMPEN_EN6		0x0000000000000040ULL
+#define  BMAC_ADDR_CMPEN_EN5		0x0000000000000020ULL
+#define  BMAC_ADDR_CMPEN_EN4		0x0000000000000010ULL
+#define  BMAC_ADDR_CMPEN_EN3		0x0000000000000008ULL
+#define  BMAC_ADDR_CMPEN_EN2		0x0000000000000004ULL
+#define  BMAC_ADDR_CMPEN_EN1		0x0000000000000002ULL
+#define  BMAC_ADDR_CMPEN_EN0		0x0000000000000001ULL
+
+#define BMAC_NUM_HOST_INFO		9
+
+#define BMAC_HOST_INFO(NUM)		(0x00400UL + (NUM) * 0x8UL)
+
+#define BTXMAC_BYTE_CNT			0x00448UL
+#define  BTXMAC_BYTE_CNT_COUNT		0x00000000ffffffffULL
+
+#define BTXMAC_FRM_CNT			0x00450UL
+#define  BTXMAC_FRM_CNT_COUNT		0x00000000ffffffffULL
+
+#define BRXMAC_BYTE_CNT			0x00458UL
+#define  BRXMAC_BYTE_CNT_COUNT		0x00000000ffffffffULL
+
+#define HOST_INFO_MPR			0x0000000000000100ULL
+#define HOST_INFO_MACRDCTBLN		0x0000000000000007ULL
+
+/* XPCS registers, offset from np->regs + np->xpcs_off  */
+
+#define XPCS_CONTROL1			(FZC_MAC + 0x00000UL)
+#define  XPCS_CONTROL1_RESET		0x0000000000008000ULL
+#define  XPCS_CONTROL1_LOOPBACK		0x0000000000004000ULL
+#define  XPCS_CONTROL1_SPEED_SELECT3	0x0000000000002000ULL
+#define  XPCS_CONTROL1_CSR_LOW_PWR	0x0000000000000800ULL
+#define  XPCS_CONTROL1_CSR_SPEED1	0x0000000000000040ULL
+#define  XPCS_CONTROL1_CSR_SPEED0	0x000000000000003cULL
+
+#define XPCS_STATUS1			(FZC_MAC + 0x00008UL)
+#define  XPCS_STATUS1_CSR_FAULT		0x0000000000000080ULL
+#define  XPCS_STATUS1_CSR_RXLNK_STAT	0x0000000000000004ULL
+#define  XPCS_STATUS1_CSR_LPWR_ABLE	0x0000000000000002ULL
+
+#define XPCS_DEVICE_IDENTIFIER		(FZC_MAC + 0x00010UL)
+#define  XPCS_DEVICE_IDENTIFIER_VAL	0x00000000ffffffffULL
+
+#define XPCS_SPEED_ABILITY		(FZC_MAC + 0x00018UL)
+#define  XPCS_SPEED_ABILITY_10GIG	0x0000000000000001ULL
+
+#define XPCS_DEV_IN_PKG			(FZC_MAC + 0x00020UL)
+#define  XPCS_DEV_IN_PKG_CSR_VEND2	0x0000000080000000ULL
+#define  XPCS_DEV_IN_PKG_CSR_VEND1	0x0000000040000000ULL
+#define  XPCS_DEV_IN_PKG_DTE_XS		0x0000000000000020ULL
+#define  XPCS_DEV_IN_PKG_PHY_XS		0x0000000000000010ULL
+#define  XPCS_DEV_IN_PKG_PCS		0x0000000000000008ULL
+#define  XPCS_DEV_IN_PKG_WIS		0x0000000000000004ULL
+#define  XPCS_DEV_IN_PKG_PMD_PMA	0x0000000000000002ULL
+#define  XPCS_DEV_IN_PKG_CLS22		0x0000000000000001ULL
+
+#define XPCS_CONTROL2			(FZC_MAC + 0x00028UL)
+#define  XPCS_CONTROL2_CSR_PSC_SEL	0x0000000000000003ULL
+
+#define XPCS_STATUS2			(FZC_MAC + 0x00030UL)
+#define  XPCS_STATUS2_CSR_DEV_PRES	0x000000000000c000ULL
+#define  XPCS_STATUS2_CSR_TX_FAULT	0x0000000000000800ULL
+#define  XPCS_STATUS2_CSR_RCV_FAULT	0x0000000000000400ULL
+#define  XPCS_STATUS2_TEN_GBASE_W	0x0000000000000004ULL
+#define  XPCS_STATUS2_TEN_GBASE_X	0x0000000000000002ULL
+#define  XPCS_STATUS2_TEN_GBASE_R	0x0000000000000001ULL
+
+#define XPCS_PKG_ID			(FZC_MAC + 0x00038UL)
+#define  XPCS_PKG_ID_VAL		0x00000000ffffffffULL
+
+#define XPCS_STATUS(IDX)		(FZC_MAC + 0x00040UL)
+#define  XPCS_STATUS_CSR_LANE_ALIGN	0x0000000000001000ULL
+#define  XPCS_STATUS_CSR_PATTEST_CAP	0x0000000000000800ULL
+#define  XPCS_STATUS_CSR_LANE3_SYNC	0x0000000000000008ULL
+#define  XPCS_STATUS_CSR_LANE2_SYNC	0x0000000000000004ULL
+#define  XPCS_STATUS_CSR_LANE1_SYNC	0x0000000000000002ULL
+#define  XPCS_STATUS_CSR_LANE0_SYNC	0x0000000000000001ULL
+
+#define XPCS_TEST_CONTROL		(FZC_MAC + 0x00048UL)
+#define  XPCS_TEST_CONTROL_TXTST_EN	0x0000000000000004ULL
+#define  XPCS_TEST_CONTROL_TPAT_SEL	0x0000000000000003ULL
+
+#define XPCS_CFG_VENDOR1		(FZC_MAC + 0x00050UL)
+#define  XPCS_CFG_VENDOR1_DBG_IOTST	0x0000000000000080ULL
+#define  XPCS_CFG_VENDOR1_DBG_SEL	0x0000000000000078ULL
+#define  XPCS_CFG_VENDOR1_BYPASS_DET	0x0000000000000004ULL
+#define  XPCS_CFG_VENDOR1_TXBUF_EN	0x0000000000000002ULL
+#define  XPCS_CFG_VENDOR1_XPCS_EN	0x0000000000000001ULL
+
+#define XPCS_DIAG_VENDOR2		(FZC_MAC + 0x00058UL)
+#define  XPCS_DIAG_VENDOR2_SSM_LANE3	0x0000000001e00000ULL
+#define  XPCS_DIAG_VENDOR2_SSM_LANE2	0x00000000001e0000ULL
+#define  XPCS_DIAG_VENDOR2_SSM_LANE1	0x000000000001e000ULL
+#define  XPCS_DIAG_VENDOR2_SSM_LANE0	0x0000000000001e00ULL
+#define  XPCS_DIAG_VENDOR2_EBUF_SM	0x00000000000001feULL
+#define  XPCS_DIAG_VENDOR2_RCV_SM	0x0000000000000001ULL
+
+#define XPCS_MASK1			(FZC_MAC + 0x00060UL)
+#define  XPCS_MASK1_FAULT_MASK		0x0000000000000080ULL
+#define  XPCS_MASK1_RXALIGN_STAT_MSK	0x0000000000000004ULL
+
+#define XPCS_PKT_COUNT			(FZC_MAC + 0x00068UL)
+#define  XPCS_PKT_COUNT_TX		0x00000000ffff0000ULL
+#define  XPCS_PKT_COUNT_RX		0x000000000000ffffULL
+
+#define XPCS_TX_SM			(FZC_MAC + 0x00070UL)
+#define  XPCS_TX_SM_VAL			0x000000000000000fULL
+
+#define XPCS_DESKEW_ERR_CNT		(FZC_MAC + 0x00078UL)
+#define  XPCS_DESKEW_ERR_CNT_VAL	0x00000000000000ffULL
+
+#define XPCS_SYMERR_CNT01		(FZC_MAC + 0x00080UL)
+#define  XPCS_SYMERR_CNT01_LANE1	0x00000000ffff0000ULL
+#define  XPCS_SYMERR_CNT01_LANE0	0x000000000000ffffULL
+
+#define XPCS_SYMERR_CNT23		(FZC_MAC + 0x00088UL)
+#define  XPCS_SYMERR_CNT23_LANE3	0x00000000ffff0000ULL
+#define  XPCS_SYMERR_CNT23_LANE2	0x000000000000ffffULL
+
+#define XPCS_TRAINING_VECTOR		(FZC_MAC + 0x00090UL)
+#define  XPCS_TRAINING_VECTOR_VAL	0x00000000ffffffffULL
+
+/* PCS registers, offset from np->regs + np->pcs_off  */
+
+#define PCS_MII_CTL			(FZC_MAC + 0x00000UL)
+#define  PCS_MII_CTL_RST		0x0000000000008000ULL
+#define  PCS_MII_CTL_10_100_SPEED	0x0000000000002000ULL
+#define  PCS_MII_AUTONEG_EN		0x0000000000001000ULL
+#define  PCS_MII_PWR_DOWN		0x0000000000000800ULL
+#define  PCS_MII_ISOLATE		0x0000000000000400ULL
+#define  PCS_MII_AUTONEG_RESTART	0x0000000000000200ULL
+#define  PCS_MII_DUPLEX			0x0000000000000100ULL
+#define  PCS_MII_COLL_TEST		0x0000000000000080ULL
+#define  PCS_MII_1000MB_SPEED		0x0000000000000040ULL
+
+#define PCS_MII_STAT			(FZC_MAC + 0x00008UL)
+#define  PCS_MII_STAT_EXT_STATUS	0x0000000000000100ULL
+#define  PCS_MII_STAT_AUTONEG_DONE	0x0000000000000020ULL
+#define  PCS_MII_STAT_REMOTE_FAULT	0x0000000000000010ULL
+#define  PCS_MII_STAT_AUTONEG_ABLE	0x0000000000000008ULL
+#define  PCS_MII_STAT_LINK_STATUS	0x0000000000000004ULL
+#define  PCS_MII_STAT_JABBER_DET	0x0000000000000002ULL
+#define  PCS_MII_STAT_EXT_CAP		0x0000000000000001ULL
+
+#define PCS_MII_ADV			(FZC_MAC + 0x00010UL)
+#define  PCS_MII_ADV_NEXT_PAGE		0x0000000000008000ULL
+#define  PCS_MII_ADV_ACK		0x0000000000004000ULL
+#define  PCS_MII_ADV_REMOTE_FAULT	0x0000000000003000ULL
+#define  PCS_MII_ADV_ASM_DIR		0x0000000000000100ULL
+#define  PCS_MII_ADV_PAUSE		0x0000000000000080ULL
+#define  PCS_MII_ADV_HALF_DUPLEX	0x0000000000000040ULL
+#define  PCS_MII_ADV_FULL_DUPLEX	0x0000000000000020ULL
+
+#define PCS_MII_PARTNER			(FZC_MAC + 0x00018UL)
+#define  PCS_MII_PARTNER_NEXT_PAGE	0x0000000000008000ULL
+#define  PCS_MII_PARTNER_ACK		0x0000000000004000ULL
+#define  PCS_MII_PARTNER_REMOTE_FAULT	0x0000000000002000ULL
+#define  PCS_MII_PARTNER_PAUSE		0x0000000000000180ULL
+#define  PCS_MII_PARTNER_HALF_DUPLEX	0x0000000000000040ULL
+#define  PCS_MII_PARTNER_FULL_DUPLEX	0x0000000000000020ULL
+
+#define PCS_CONF			(FZC_MAC + 0x00020UL)
+#define  PCS_CONF_MASK			0x0000000000000040ULL
+#define  PCS_CONF_10MS_TMR_OVERRIDE	0x0000000000000020ULL
+#define  PCS_CONF_JITTER_STUDY		0x0000000000000018ULL
+#define  PCS_CONF_SIGDET_ACTIVE_LOW	0x0000000000000004ULL
+#define  PCS_CONF_SIGDET_OVERRIDE	0x0000000000000002ULL
+#define  PCS_CONF_ENABLE		0x0000000000000001ULL
+
+#define PCS_STATE			(FZC_MAC + 0x00028UL)
+#define  PCS_STATE_D_PARTNER_FAIL	0x0000000020000000ULL
+#define  PCS_STATE_D_WAIT_C_CODES_ACK	0x0000000010000000ULL
+#define  PCS_STATE_D_SYNC_LOSS		0x0000000008000000ULL
+#define  PCS_STATE_D_NO_GOOD_C_CODES	0x0000000004000000ULL
+#define  PCS_STATE_D_SERDES		0x0000000002000000ULL
+#define  PCS_STATE_D_BREAKLINK_C_CODES	0x0000000001000000ULL
+#define  PCS_STATE_L_SIGDET		0x0000000000400000ULL
+#define  PCS_STATE_L_SYNC_LOSS		0x0000000000200000ULL
+#define  PCS_STATE_L_C_CODES		0x0000000000100000ULL
+#define  PCS_STATE_LINK_CFG_STATE	0x000000000001e000ULL
+#define  PCS_STATE_SEQ_DET_STATE	0x0000000000001800ULL
+#define  PCS_STATE_WORD_SYNC_STATE	0x0000000000000700ULL
+#define  PCS_STATE_NO_IDLE		0x000000000000000fULL
+
+#define PCS_INTERRUPT			(FZC_MAC + 0x00030UL)
+#define  PCS_INTERRUPT_LSTATUS		0x0000000000000004ULL
+
+#define PCS_DPATH_MODE			(FZC_MAC + 0x000a0UL)
+#define  PCS_DPATH_MODE_PCS		0x0000000000000000ULL
+#define  PCS_DPATH_MODE_MII		0x0000000000000002ULL
+#define  PCS_DPATH_MODE_LINKUP_F_ENAB	0x0000000000000001ULL
+
+#define PCS_PKT_CNT			(FZC_MAC + 0x000c0UL)
+#define  PCS_PKT_CNT_RX			0x0000000007ff0000ULL
+#define  PCS_PKT_CNT_TX			0x00000000000007ffULL
+
+#define MIF_BB_MDC			(FZC_MAC + 0x16000UL)
+#define  MIF_BB_MDC_CLK			0x0000000000000001ULL
+
+#define MIF_BB_MDO			(FZC_MAC + 0x16008UL)
+#define  MIF_BB_MDO_DAT			0x0000000000000001ULL
+
+#define MIF_BB_MDO_EN			(FZC_MAC + 0x16010UL)
+#define  MIF_BB_MDO_EN_VAL		0x0000000000000001ULL
+
+#define MIF_FRAME_OUTPUT		(FZC_MAC + 0x16018UL)
+#define  MIF_FRAME_OUTPUT_ST		0x00000000c0000000ULL
+#define  MIF_FRAME_OUTPUT_ST_SHIFT	30
+#define  MIF_FRAME_OUTPUT_OP_ADDR	0x0000000000000000ULL
+#define  MIF_FRAME_OUTPUT_OP_WRITE	0x0000000010000000ULL
+#define  MIF_FRAME_OUTPUT_OP_READ_INC	0x0000000020000000ULL
+#define  MIF_FRAME_OUTPUT_OP_READ	0x0000000030000000ULL
+#define  MIF_FRAME_OUTPUT_OP_SHIFT	28
+#define  MIF_FRAME_OUTPUT_PORT		0x000000000f800000ULL
+#define  MIF_FRAME_OUTPUT_PORT_SHIFT	23
+#define  MIF_FRAME_OUTPUT_REG		0x00000000007c0000ULL
+#define  MIF_FRAME_OUTPUT_REG_SHIFT	18
+#define  MIF_FRAME_OUTPUT_TA		0x0000000000030000ULL
+#define  MIF_FRAME_OUTPUT_TA_SHIFT	16
+#define  MIF_FRAME_OUTPUT_DATA		0x000000000000ffffULL
+#define  MIF_FRAME_OUTPUT_DATA_SHIFT	0
+
+#define MDIO_ADDR_OP(port, dev, reg) \
+	((0 << MIF_FRAME_OUTPUT_ST_SHIFT) | \
+	 MIF_FRAME_OUTPUT_OP_ADDR | \
+	 (port << MIF_FRAME_OUTPUT_PORT_SHIFT) | \
+	 (dev << MIF_FRAME_OUTPUT_REG_SHIFT) | \
+	 (0x2 << MIF_FRAME_OUTPUT_TA_SHIFT) | \
+	 (reg << MIF_FRAME_OUTPUT_DATA_SHIFT))
+
+#define MDIO_READ_OP(port, dev) \
+	((0 << MIF_FRAME_OUTPUT_ST_SHIFT) | \
+	 MIF_FRAME_OUTPUT_OP_READ | \
+	 (port << MIF_FRAME_OUTPUT_PORT_SHIFT) | \
+	 (dev << MIF_FRAME_OUTPUT_REG_SHIFT) | \
+	 (0x2 << MIF_FRAME_OUTPUT_TA_SHIFT))
+
+#define MDIO_WRITE_OP(port, dev, data) \
+	((0 << MIF_FRAME_OUTPUT_ST_SHIFT) | \
+	 MIF_FRAME_OUTPUT_OP_WRITE | \
+	 (port << MIF_FRAME_OUTPUT_PORT_SHIFT) | \
+	 (dev << MIF_FRAME_OUTPUT_REG_SHIFT) | \
+	 (0x2 << MIF_FRAME_OUTPUT_TA_SHIFT) | \
+	 (data << MIF_FRAME_OUTPUT_DATA_SHIFT))
+
+#define MII_READ_OP(port, reg) \
+	((1 << MIF_FRAME_OUTPUT_ST_SHIFT) | \
+	 (2 << MIF_FRAME_OUTPUT_OP_SHIFT) | \
+	 (port << MIF_FRAME_OUTPUT_PORT_SHIFT) | \
+	 (reg << MIF_FRAME_OUTPUT_REG_SHIFT) | \
+	 (0x2 << MIF_FRAME_OUTPUT_TA_SHIFT))
+
+#define MII_WRITE_OP(port, reg, data) \
+	((1 << MIF_FRAME_OUTPUT_ST_SHIFT) | \
+	 (1 << MIF_FRAME_OUTPUT_OP_SHIFT) | \
+	 (port << MIF_FRAME_OUTPUT_PORT_SHIFT) | \
+	 (reg << MIF_FRAME_OUTPUT_REG_SHIFT) | \
+	 (0x2 << MIF_FRAME_OUTPUT_TA_SHIFT) | \
+	 (data << MIF_FRAME_OUTPUT_DATA_SHIFT))
+
+#define MIF_CONFIG			(FZC_MAC + 0x16020UL)
+#define  MIF_CONFIG_ATCA_GE		0x0000000000010000ULL
+#define  MIF_CONFIG_INDIRECT_MODE	0x0000000000008000ULL
+#define  MIF_CONFIG_POLL_PRT_PHYADDR	0x0000000000003c00ULL
+#define  MIF_CONFIG_POLL_DEV_REG_ADDR	0x00000000000003e0ULL
+#define  MIF_CONFIG_BB_MODE		0x0000000000000010ULL
+#define  MIF_CONFIG_POLL_EN		0x0000000000000008ULL
+#define  MIF_CONFIG_BB_SER_SEL		0x0000000000000006ULL
+#define  MIF_CONFIG_MANUAL_MODE		0x0000000000000001ULL
+
+#define MIF_POLL_STATUS			(FZC_MAC + 0x16028UL)
+#define  MIF_POLL_STATUS_DATA		0x00000000ffff0000ULL
+#define  MIF_POLL_STATUS_STAT		0x000000000000ffffULL
+
+#define MIF_POLL_MASK			(FZC_MAC + 0x16030UL)
+#define  MIF_POLL_MASK_VAL		0x000000000000ffffULL
+
+#define MIF_SM				(FZC_MAC + 0x16038UL)
+#define  MIF_SM_PORT_ADDR		0x00000000001f0000ULL
+#define  MIF_SM_MDI_1			0x0000000000004000ULL
+#define  MIF_SM_MDI_0			0x0000000000002400ULL
+#define  MIF_SM_MDCLK			0x0000000000001000ULL
+#define  MIF_SM_MDO_EN			0x0000000000000800ULL
+#define  MIF_SM_MDO			0x0000000000000400ULL
+#define  MIF_SM_MDI			0x0000000000000200ULL
+#define  MIF_SM_CTL			0x00000000000001c0ULL
+#define  MIF_SM_EX			0x000000000000003fULL
+
+#define MIF_STATUS			(FZC_MAC + 0x16040UL)
+#define  MIF_STATUS_MDINT1		0x0000000000000020ULL
+#define  MIF_STATUS_MDINT0		0x0000000000000010ULL
+
+#define MIF_MASK			(FZC_MAC + 0x16048UL)
+#define  MIF_MASK_MDINT1		0x0000000000000020ULL
+#define  MIF_MASK_MDINT0		0x0000000000000010ULL
+#define  MIF_MASK_PEU_ERR		0x0000000000000008ULL
+#define  MIF_MASK_YC			0x0000000000000004ULL
+#define  MIF_MASK_XGE_ERR0		0x0000000000000002ULL
+#define  MIF_MASK_MIF_INIT_DONE		0x0000000000000001ULL
+
+#define ENET_SERDES_RESET		(FZC_MAC + 0x14000UL)
+#define  ENET_SERDES_RESET_1		0x0000000000000002ULL
+#define  ENET_SERDES_RESET_0		0x0000000000000001ULL
+
+#define ENET_SERDES_CFG			(FZC_MAC + 0x14008UL)
+#define  ENET_SERDES_BE_LOOPBACK	0x0000000000000002ULL
+#define  ENET_SERDES_CFG_FORCE_RDY	0x0000000000000001ULL
+
+#define ENET_SERDES_0_PLL_CFG		(FZC_MAC + 0x14010UL)
+#define  ENET_SERDES_PLL_FBDIV0		0x0000000000000001ULL
+#define  ENET_SERDES_PLL_FBDIV1		0x0000000000000002ULL
+#define  ENET_SERDES_PLL_FBDIV2		0x0000000000000004ULL
+#define  ENET_SERDES_PLL_HRATE0		0x0000000000000008ULL
+#define  ENET_SERDES_PLL_HRATE1		0x0000000000000010ULL
+#define  ENET_SERDES_PLL_HRATE2		0x0000000000000020ULL
+#define  ENET_SERDES_PLL_HRATE3		0x0000000000000040ULL
+
+#define ENET_SERDES_0_CTRL_CFG		(FZC_MAC + 0x14018UL)
+#define  ENET_SERDES_CTRL_SDET_0	0x0000000000000001ULL
+#define  ENET_SERDES_CTRL_SDET_1	0x0000000000000002ULL
+#define  ENET_SERDES_CTRL_SDET_2	0x0000000000000004ULL
+#define  ENET_SERDES_CTRL_SDET_3	0x0000000000000008ULL
+#define  ENET_SERDES_CTRL_EMPH_0	0x0000000000000070ULL
+#define  ENET_SERDES_CTRL_EMPH_0_SHIFT	4
+#define  ENET_SERDES_CTRL_EMPH_1	0x0000000000000380ULL
+#define  ENET_SERDES_CTRL_EMPH_1_SHIFT	7
+#define  ENET_SERDES_CTRL_EMPH_2	0x0000000000001c00ULL
+#define  ENET_SERDES_CTRL_EMPH_2_SHIFT	10
+#define  ENET_SERDES_CTRL_EMPH_3	0x000000000000e000ULL
+#define  ENET_SERDES_CTRL_EMPH_3_SHIFT	13
+#define  ENET_SERDES_CTRL_LADJ_0	0x0000000000070000ULL
+#define  ENET_SERDES_CTRL_LADJ_0_SHIFT	16
+#define  ENET_SERDES_CTRL_LADJ_1	0x0000000000380000ULL
+#define  ENET_SERDES_CTRL_LADJ_1_SHIFT	19
+#define  ENET_SERDES_CTRL_LADJ_2	0x0000000001c00000ULL
+#define  ENET_SERDES_CTRL_LADJ_2_SHIFT	22
+#define  ENET_SERDES_CTRL_LADJ_3	0x000000000e000000ULL
+#define  ENET_SERDES_CTRL_LADJ_3_SHIFT	25
+#define  ENET_SERDES_CTRL_RXITERM_0	0x0000000010000000ULL
+#define  ENET_SERDES_CTRL_RXITERM_1	0x0000000020000000ULL
+#define  ENET_SERDES_CTRL_RXITERM_2	0x0000000040000000ULL
+#define  ENET_SERDES_CTRL_RXITERM_3	0x0000000080000000ULL
+
+#define ENET_SERDES_0_TEST_CFG		(FZC_MAC + 0x14020UL)
+#define  ENET_SERDES_TEST_MD_0		0x0000000000000003ULL
+#define  ENET_SERDES_TEST_MD_0_SHIFT	0
+#define  ENET_SERDES_TEST_MD_1		0x000000000000000cULL
+#define  ENET_SERDES_TEST_MD_1_SHIFT	2
+#define  ENET_SERDES_TEST_MD_2		0x0000000000000030ULL
+#define  ENET_SERDES_TEST_MD_2_SHIFT	4
+#define  ENET_SERDES_TEST_MD_3		0x00000000000000c0ULL
+#define  ENET_SERDES_TEST_MD_3_SHIFT	6
+
+#define ENET_TEST_MD_NO_LOOPBACK	0x0
+#define ENET_TEST_MD_EWRAP		0x1
+#define ENET_TEST_MD_PAD_LOOPBACK	0x2
+#define ENET_TEST_MD_REV_LOOPBACK	0x3
+
+#define ENET_SERDES_1_PLL_CFG		(FZC_MAC + 0x14028UL)
+#define ENET_SERDES_1_CTRL_CFG		(FZC_MAC + 0x14030UL)
+#define ENET_SERDES_1_TEST_CFG		(FZC_MAC + 0x14038UL)
+
+#define ENET_RGMII_CFG_REG		(FZC_MAC + 0x14040UL)
+
+#define ESR_INT_SIGNALS			(FZC_MAC + 0x14800UL)
+#define  ESR_INT_SIGNALS_ALL		0x00000000ffffffffULL
+#define  ESR_INT_SIGNALS_P0_BITS	0x0000000033e0000fULL
+#define  ESR_INT_SIGNALS_P1_BITS	0x000000000c1f00f0ULL
+#define  ESR_INT_SRDY0_P0		0x0000000020000000ULL
+#define  ESR_INT_DET0_P0		0x0000000010000000ULL
+#define  ESR_INT_SRDY0_P1		0x0000000008000000ULL
+#define  ESR_INT_DET0_P1		0x0000000004000000ULL
+#define  ESR_INT_XSRDY_P0		0x0000000002000000ULL
+#define  ESR_INT_XDP_P0_CH3		0x0000000001000000ULL
+#define  ESR_INT_XDP_P0_CH2		0x0000000000800000ULL
+#define  ESR_INT_XDP_P0_CH1		0x0000000000400000ULL
+#define  ESR_INT_XDP_P0_CH0		0x0000000000200000ULL
+#define  ESR_INT_XSRDY_P1		0x0000000000100000ULL
+#define  ESR_INT_XDP_P1_CH3		0x0000000000080000ULL
+#define  ESR_INT_XDP_P1_CH2		0x0000000000040000ULL
+#define  ESR_INT_XDP_P1_CH1		0x0000000000020000ULL
+#define  ESR_INT_XDP_P1_CH0		0x0000000000010000ULL
+#define  ESR_INT_SLOSS_P1_CH3		0x0000000000000080ULL
+#define  ESR_INT_SLOSS_P1_CH2		0x0000000000000040ULL
+#define  ESR_INT_SLOSS_P1_CH1		0x0000000000000020ULL
+#define  ESR_INT_SLOSS_P1_CH0		0x0000000000000010ULL
+#define  ESR_INT_SLOSS_P0_CH3		0x0000000000000008ULL
+#define  ESR_INT_SLOSS_P0_CH2		0x0000000000000004ULL
+#define  ESR_INT_SLOSS_P0_CH1		0x0000000000000002ULL
+#define  ESR_INT_SLOSS_P0_CH0		0x0000000000000001ULL
+
+#define ESR_DEBUG_SEL			(FZC_MAC + 0x14808UL)
+#define  ESR_DEBUG_SEL_VAL		0x000000000000003fULL
+
+/* SerDes registers behind MIF */
+#define NIU_ESR_DEV_ADDR		0x1e
+#define ESR_BASE			0x0000
+
+#define ESR_RXTX_COMM_CTRL_L		(ESR_BASE + 0x0000)
+#define ESR_RXTX_COMM_CTRL_H		(ESR_BASE + 0x0001)
+
+#define ESR_RXTX_RESET_CTRL_L		(ESR_BASE + 0x0002)
+#define ESR_RXTX_RESET_CTRL_H		(ESR_BASE + 0x0003)
+
+#define ESR_RX_POWER_CTRL_L		(ESR_BASE + 0x0004)
+#define ESR_RX_POWER_CTRL_H		(ESR_BASE + 0x0005)
+
+#define ESR_TX_POWER_CTRL_L		(ESR_BASE + 0x0006)
+#define ESR_TX_POWER_CTRL_H		(ESR_BASE + 0x0007)
+
+#define ESR_MISC_POWER_CTRL_L		(ESR_BASE + 0x0008)
+#define ESR_MISC_POWER_CTRL_H		(ESR_BASE + 0x0009)
+
+#define ESR_RXTX_CTRL_L(CHAN)		(ESR_BASE + 0x0080 + (CHAN) * 0x10)
+#define ESR_RXTX_CTRL_H(CHAN)		(ESR_BASE + 0x0081 + (CHAN) * 0x10)
+#define  ESR_RXTX_CTRL_BIASCNTL		0x80000000
+#define  ESR_RXTX_CTRL_RESV1		0x7c000000
+#define  ESR_RXTX_CTRL_TDENFIFO		0x02000000
+#define  ESR_RXTX_CTRL_TDWS20		0x01000000
+#define  ESR_RXTX_CTRL_VMUXLO		0x00c00000
+#define  ESR_RXTX_CTRL_VMUXLO_SHIFT	22
+#define  ESR_RXTX_CTRL_VPULSELO		0x00300000
+#define  ESR_RXTX_CTRL_VPULSELO_SHIFT	20
+#define  ESR_RXTX_CTRL_RESV2		0x000f0000
+#define  ESR_RXTX_CTRL_RESV3		0x0000c000
+#define  ESR_RXTX_CTRL_RXPRESWIN	0x00003000
+#define  ESR_RXTX_CTRL_RXPRESWIN_SHIFT	12
+#define  ESR_RXTX_CTRL_RESV4		0x00000800
+#define  ESR_RXTX_CTRL_RISEFALL		0x00000700
+#define  ESR_RXTX_CTRL_RISEFALL_SHIFT	8
+#define  ESR_RXTX_CTRL_RESV5		0x000000fe
+#define  ESR_RXTX_CTRL_ENSTRETCH	0x00000001
+
+#define ESR_RXTX_TUNING_L(CHAN)		(ESR_BASE + 0x0082 + (CHAN) * 0x10)
+#define ESR_RXTX_TUNING_H(CHAN)		(ESR_BASE + 0x0083 + (CHAN) * 0x10)
+
+#define ESR_RX_SYNCCHAR_L(CHAN)		(ESR_BASE + 0x0084 + (CHAN) * 0x10)
+#define ESR_RX_SYNCCHAR_H(CHAN)		(ESR_BASE + 0x0085 + (CHAN) * 0x10)
+
+#define ESR_RXTX_TEST_L(CHAN)		(ESR_BASE + 0x0086 + (CHAN) * 0x10)
+#define ESR_RXTX_TEST_H(CHAN)		(ESR_BASE + 0x0087 + (CHAN) * 0x10)
+
+#define ESR_GLUE_CTRL0_L(CHAN)		(ESR_BASE + 0x0088 + (CHAN) * 0x10)
+#define ESR_GLUE_CTRL0_H(CHAN)		(ESR_BASE + 0x0089 + (CHAN) * 0x10)
+#define  ESR_GLUE_CTRL0_RESV1		0xf8000000
+#define  ESR_GLUE_CTRL0_BLTIME		0x07000000
+#define  ESR_GLUE_CTRL0_BLTIME_SHIFT	24
+#define  ESR_GLUE_CTRL0_RESV2		0x00ff0000
+#define  ESR_GLUE_CTRL0_RXLOS_TEST	0x00008000
+#define  ESR_GLUE_CTRL0_RESV3		0x00004000
+#define  ESR_GLUE_CTRL0_RXLOSENAB	0x00002000
+#define  ESR_GLUE_CTRL0_FASTRESYNC	0x00001000
+#define  ESR_GLUE_CTRL0_SRATE		0x00000f00
+#define  ESR_GLUE_CTRL0_SRATE_SHIFT	8
+#define  ESR_GLUE_CTRL0_THCNT		0x000000ff
+#define  ESR_GLUE_CTRL0_THCNT_SHIFT	0
+
+#define BLTIME_64_CYCLES		0
+#define BLTIME_128_CYCLES		1
+#define BLTIME_256_CYCLES		2
+#define BLTIME_300_CYCLES		3
+#define BLTIME_384_CYCLES		4
+#define BLTIME_512_CYCLES		5
+#define BLTIME_1024_CYCLES		6
+#define BLTIME_2048_CYCLES		7
+
+#define ESR_GLUE_CTRL1_L(CHAN)		(ESR_BASE + 0x008a + (CHAN) * 0x10)
+#define ESR_GLUE_CTRL1_H(CHAN)		(ESR_BASE + 0x008b + (CHAN) * 0x10)
+#define ESR_RXTX_TUNING1_L(CHAN)	(ESR_BASE + 0x00c2 + (CHAN) * 0x10)
+#define ESR_RXTX_TUNING1_H(CHAN)	(ESR_BASE + 0x00c2 + (CHAN) * 0x10)
+#define ESR_RXTX_TUNING2_L(CHAN)	(ESR_BASE + 0x0102 + (CHAN) * 0x10)
+#define ESR_RXTX_TUNING2_H(CHAN)	(ESR_BASE + 0x0102 + (CHAN) * 0x10)
+#define ESR_RXTX_TUNING3_L(CHAN)	(ESR_BASE + 0x0142 + (CHAN) * 0x10)
+#define ESR_RXTX_TUNING3_H(CHAN)	(ESR_BASE + 0x0142 + (CHAN) * 0x10)
+
+#define NIU_ESR2_DEV_ADDR		0x1e
+#define ESR2_BASE			0x8000
+
+#define ESR2_TI_PLL_CFG_L		(ESR2_BASE + 0x000)
+#define ESR2_TI_PLL_CFG_H		(ESR2_BASE + 0x001)
+#define  PLL_CFG_STD			0x00000c00
+#define  PLL_CFG_STD_SHIFT		10
+#define  PLL_CFG_LD			0x00000300
+#define  PLL_CFG_LD_SHIFT		8
+#define  PLL_CFG_MPY			0x0000001e
+#define  PLL_CFG_MPY_SHIFT		1
+#define  PLL_CFG_MPY_4X		0x0
+#define  PLL_CFG_MPY_5X		0x00000002
+#define  PLL_CFG_MPY_6X		0x00000004
+#define  PLL_CFG_MPY_8X		0x00000008
+#define  PLL_CFG_MPY_10X		0x0000000a
+#define  PLL_CFG_MPY_12X		0x0000000c
+#define  PLL_CFG_MPY_12P5X		0x0000000e
+#define  PLL_CFG_ENPLL			0x00000001
+
+#define ESR2_TI_PLL_STS_L		(ESR2_BASE + 0x002)
+#define ESR2_TI_PLL_STS_H		(ESR2_BASE + 0x003)
+#define  PLL_STS_LOCK			0x00000001
+
+#define ESR2_TI_PLL_TEST_CFG_L		(ESR2_BASE + 0x004)
+#define ESR2_TI_PLL_TEST_CFG_H		(ESR2_BASE + 0x005)
+#define  PLL_TEST_INVPATT		0x00004000
+#define  PLL_TEST_RATE			0x00003000
+#define  PLL_TEST_RATE_SHIFT		12
+#define  PLL_TEST_CFG_ENBSAC		0x00000400
+#define  PLL_TEST_CFG_ENBSRX		0x00000200
+#define  PLL_TEST_CFG_ENBSTX		0x00000100
+#define  PLL_TEST_CFG_LOOPBACK_PAD	0x00000040
+#define  PLL_TEST_CFG_LOOPBACK_CML_DIS	0x00000080
+#define  PLL_TEST_CFG_LOOPBACK_CML_EN	0x000000c0
+#define  PLL_TEST_CFG_CLKBYP		0x00000030
+#define  PLL_TEST_CFG_CLKBYP_SHIFT	4
+#define  PLL_TEST_CFG_EN_RXPATT		0x00000008
+#define  PLL_TEST_CFG_EN_TXPATT		0x00000004
+#define  PLL_TEST_CFG_TPATT		0x00000003
+#define  PLL_TEST_CFG_TPATT_SHIFT	0
+
+#define ESR2_TI_PLL_TX_CFG_L(CHAN)	(ESR2_BASE + 0x100 + (CHAN) * 4)
+#define ESR2_TI_PLL_TX_CFG_H(CHAN)	(ESR2_BASE + 0x101 + (CHAN) * 4)
+#define  PLL_TX_CFG_RDTCT		0x00600000
+#define  PLL_TX_CFG_RDTCT_SHIFT		21
+#define  PLL_TX_CFG_ENIDL		0x00100000
+#define  PLL_TX_CFG_BSTX		0x00020000
+#define  PLL_TX_CFG_ENFTP		0x00010000
+#define  PLL_TX_CFG_DE			0x0000f000
+#define  PLL_TX_CFG_DE_SHIFT		12
+#define  PLL_TX_CFG_SWING_125MV		0x00000000
+#define  PLL_TX_CFG_SWING_250MV		0x00000200
+#define  PLL_TX_CFG_SWING_500MV		0x00000400
+#define  PLL_TX_CFG_SWING_625MV		0x00000600
+#define  PLL_TX_CFG_SWING_750MV		0x00000800
+#define  PLL_TX_CFG_SWING_1000MV	0x00000a00
+#define  PLL_TX_CFG_SWING_1250MV	0x00000c00
+#define  PLL_TX_CFG_SWING_1375MV	0x00000e00
+#define  PLL_TX_CFG_CM			0x00000100
+#define  PLL_TX_CFG_INVPAIR		0x00000080
+#define  PLL_TX_CFG_RATE		0x00000060
+#define  PLL_TX_CFG_RATE_SHIFT		5
+#define  PLL_TX_CFG_RATE_FULL		0x0
+#define  PLL_TX_CFG_RATE_HALF		0x20
+#define  PLL_TX_CFG_RATE_QUAD		0x40
+#define  PLL_TX_CFG_BUSWIDTH		0x0000001c
+#define  PLL_TX_CFG_BUSWIDTH_SHIFT	2
+#define  PLL_TX_CFG_ENTEST		0x00000002
+#define  PLL_TX_CFG_ENTX		0x00000001
+
+#define ESR2_TI_PLL_TX_STS_L(CHAN)	(ESR2_BASE + 0x102 + (CHAN) * 4)
+#define ESR2_TI_PLL_TX_STS_H(CHAN)	(ESR2_BASE + 0x103 + (CHAN) * 4)
+#define  PLL_TX_STS_RDTCTIP		0x00000002
+#define  PLL_TX_STS_TESTFAIL		0x00000001
+
+#define ESR2_TI_PLL_RX_CFG_L(CHAN)	(ESR2_BASE + 0x120 + (CHAN) * 4)
+#define ESR2_TI_PLL_RX_CFG_H(CHAN)	(ESR2_BASE + 0x121 + (CHAN) * 4)
+#define  PLL_RX_CFG_BSINRXN		0x02000000
+#define  PLL_RX_CFG_BSINRXP		0x01000000
+#define  PLL_RX_CFG_EQ_MAX_LF		0x00000000
+#define  PLL_RX_CFG_EQ_LP_ADAPTIVE	0x00080000
+#define  PLL_RX_CFG_EQ_LP_1084MHZ	0x00400000
+#define  PLL_RX_CFG_EQ_LP_805MHZ	0x00480000
+#define  PLL_RX_CFG_EQ_LP_573MHZ	0x00500000
+#define  PLL_RX_CFG_EQ_LP_402MHZ	0x00580000
+#define  PLL_RX_CFG_EQ_LP_304MHZ	0x00600000
+#define  PLL_RX_CFG_EQ_LP_216MHZ	0x00680000
+#define  PLL_RX_CFG_EQ_LP_156MHZ	0x00700000
+#define  PLL_RX_CFG_EQ_LP_135MHZ	0x00780000
+#define  PLL_RX_CFG_EQ_SHIFT		19
+#define  PLL_RX_CFG_CDR			0x00070000
+#define  PLL_RX_CFG_CDR_SHIFT		16
+#define  PLL_RX_CFG_LOS_DIS		0x00000000
+#define  PLL_RX_CFG_LOS_HTHRESH		0x00004000
+#define  PLL_RX_CFG_LOS_LTHRESH		0x00008000
+#define  PLL_RX_CFG_ALIGN_DIS		0x00000000
+#define  PLL_RX_CFG_ALIGN_ENA		0x00001000
+#define  PLL_RX_CFG_ALIGN_JOG		0x00002000
+#define  PLL_RX_CFG_TERM_VDDT		0x00000000
+#define  PLL_RX_CFG_TERM_0P8VDDT	0x00000100
+#define  PLL_RX_CFG_TERM_FLOAT		0x00000300
+#define  PLL_RX_CFG_INVPAIR		0x00000080
+#define  PLL_RX_CFG_RATE		0x00000060
+#define  PLL_RX_CFG_RATE_SHIFT		5
+#define  PLL_RX_CFG_RATE_FULL		0x0
+#define  PLL_RX_CFG_RATE_HALF		0x20
+#define  PLL_RX_CFG_RATE_QUAD		0x40
+#define  PLL_RX_CFG_BUSWIDTH		0x0000001c
+#define  PLL_RX_CFG_BUSWIDTH_SHIFT	2
+#define  PLL_RX_CFG_ENTEST		0x00000002
+#define  PLL_RX_CFG_ENRX		0x00000001
+
+#define ESR2_TI_PLL_RX_STS_L(CHAN)	(ESR2_BASE + 0x122 + (CHAN) * 4)
+#define ESR2_TI_PLL_RX_STS_H(CHAN)	(ESR2_BASE + 0x123 + (CHAN) * 4)
+#define  PLL_RX_STS_CRCIDTCT		0x00000200
+#define  PLL_RX_STS_CWDTCT		0x00000100
+#define  PLL_RX_STS_BSRXN		0x00000020
+#define  PLL_RX_STS_BSRXP		0x00000010
+#define  PLL_RX_STS_LOSDTCT		0x00000008
+#define  PLL_RX_STS_ODDCG		0x00000004
+#define  PLL_RX_STS_SYNC		0x00000002
+#define  PLL_RX_STS_TESTFAIL		0x00000001
+
+#define ENET_VLAN_TBL(IDX)		(FZC_FFLP + 0x00000UL + (IDX) * 8UL)
+#define  ENET_VLAN_TBL_PARITY1		0x0000000000020000ULL
+#define  ENET_VLAN_TBL_PARITY0		0x0000000000010000ULL
+#define  ENET_VLAN_TBL_VPR		0x0000000000000008ULL
+#define  ENET_VLAN_TBL_VLANRDCTBLN	0x0000000000000007ULL
+#define  ENET_VLAN_TBL_SHIFT(PORT)	((PORT) * 4)
+
+#define ENET_VLAN_TBL_NUM_ENTRIES	4096
+
+#define FFLP_VLAN_PAR_ERR		(FZC_FFLP + 0x0800UL)
+#define  FFLP_VLAN_PAR_ERR_ERR		0x0000000080000000ULL
+#define  FFLP_VLAN_PAR_ERR_M_ERR	0x0000000040000000ULL
+#define  FFLP_VLAN_PAR_ERR_ADDR		0x000000003ffc0000ULL
+#define  FFLP_VLAN_PAR_ERR_DATA		0x000000000003ffffULL
+
+#define L2_CLS(IDX)			(FZC_FFLP + 0x20000UL + (IDX) * 8UL)
+#define  L2_CLS_VLD			0x0000000000010000ULL
+#define  L2_CLS_ETYPE			0x000000000000ffffULL
+#define  L2_CLS_ETYPE_SHIFT		0
+
+#define L3_CLS(IDX)			(FZC_FFLP + 0x20010UL + (IDX) * 8UL)
+#define  L3_CLS_VALID			0x0000000002000000ULL
+#define  L3_CLS_IPVER			0x0000000001000000ULL
+#define  L3_CLS_PID			0x0000000000ff0000ULL
+#define  L3_CLS_PID_SHIFT		16
+#define  L3_CLS_TOSMASK			0x000000000000ff00ULL
+#define  L3_CLS_TOSMASK_SHIFT		8
+#define  L3_CLS_TOS			0x00000000000000ffULL
+#define  L3_CLS_TOS_SHIFT		0
+
+#define TCAM_KEY(IDX)			(FZC_FFLP + 0x20030UL + (IDX) * 8UL)
+#define  TCAM_KEY_DISC			0x0000000000000008ULL
+#define  TCAM_KEY_TSEL			0x0000000000000004ULL
+#define  TCAM_KEY_IPADDR		0x0000000000000001ULL
+
+#define TCAM_KEY_0			(FZC_FFLP + 0x20090UL)
+#define  TCAM_KEY_0_KEY			0x00000000000000ffULL /* bits 192-199 */
+
+#define TCAM_KEY_1			(FZC_FFLP + 0x20098UL)
+#define  TCAM_KEY_1_KEY			0xffffffffffffffffULL /* bits 128-191 */
+
+#define TCAM_KEY_2			(FZC_FFLP + 0x200a0UL)
+#define  TCAM_KEY_2_KEY			0xffffffffffffffffULL /* bits 64-127 */
+
+#define TCAM_KEY_3			(FZC_FFLP + 0x200a8UL)
+#define  TCAM_KEY_3_KEY			0xffffffffffffffffULL /* bits 0-63 */
+
+#define TCAM_KEY_MASK_0			(FZC_FFLP + 0x200b0UL)
+#define  TCAM_KEY_MASK_0_KEY_SEL	0x00000000000000ffULL /* bits 192-199 */
+
+#define TCAM_KEY_MASK_1			(FZC_FFLP + 0x200b8UL)
+#define  TCAM_KEY_MASK_1_KEY_SEL	0xffffffffffffffffULL /* bits 128-191 */
+
+#define TCAM_KEY_MASK_2			(FZC_FFLP + 0x200c0UL)
+#define  TCAM_KEY_MASK_2_KEY_SEL	0xffffffffffffffffULL /* bits 64-127 */
+
+#define TCAM_KEY_MASK_3			(FZC_FFLP + 0x200c8UL)
+#define  TCAM_KEY_MASK_3_KEY_SEL	0xffffffffffffffffULL /* bits 0-63 */
+
+#define TCAM_CTL			(FZC_FFLP + 0x200d0UL)
+#define  TCAM_CTL_RWC			0x00000000001c0000ULL
+#define  TCAM_CTL_RWC_TCAM_WRITE	0x0000000000000000ULL
+#define  TCAM_CTL_RWC_TCAM_READ		0x0000000000040000ULL
+#define  TCAM_CTL_RWC_TCAM_COMPARE	0x0000000000080000ULL
+#define  TCAM_CTL_RWC_RAM_WRITE		0x0000000000100000ULL
+#define  TCAM_CTL_RWC_RAM_READ		0x0000000000140000ULL
+#define  TCAM_CTL_STAT			0x0000000000020000ULL
+#define  TCAM_CTL_MATCH			0x0000000000010000ULL
+#define  TCAM_CTL_LOC			0x00000000000003ffULL
+
+#define TCAM_ERR			(FZC_FFLP + 0x200d8UL)
+#define  TCAM_ERR_ERR			0x0000000080000000ULL
+#define  TCAM_ERR_P_ECC			0x0000000040000000ULL
+#define  TCAM_ERR_MULT			0x0000000020000000ULL
+#define  TCAM_ERR_ADDR			0x0000000000ff0000ULL
+#define  TCAM_ERR_SYNDROME		0x000000000000ffffULL
+
+#define HASH_LOOKUP_ERR_LOG1		(FZC_FFLP + 0x200e0UL)
+#define  HASH_LOOKUP_ERR_LOG1_ERR	0x0000000000000008ULL
+#define  HASH_LOOKUP_ERR_LOG1_MULT_LK	0x0000000000000004ULL
+#define  HASH_LOOKUP_ERR_LOG1_CU	0x0000000000000002ULL
+#define  HASH_LOOKUP_ERR_LOG1_MULT_BIT	0x0000000000000001ULL
+
+#define HASH_LOOKUP_ERR_LOG2		(FZC_FFLP + 0x200e8UL)
+#define  HASH_LOOKUP_ERR_LOG2_H1	0x000000007ffff800ULL
+#define  HASH_LOOKUP_ERR_LOG2_SUBAREA	0x0000000000000700ULL
+#define  HASH_LOOKUP_ERR_LOG2_SYNDROME	0x00000000000000ffULL
+
+#define FFLP_CFG_1			(FZC_FFLP + 0x20100UL)
+#define  FFLP_CFG_1_TCAM_DIS		0x0000000004000000ULL
+#define  FFLP_CFG_1_PIO_DBG_SEL		0x0000000003800000ULL
+#define  FFLP_CFG_1_PIO_FIO_RST		0x0000000000400000ULL
+#define  FFLP_CFG_1_PIO_FIO_LAT		0x0000000000300000ULL
+#define  FFLP_CFG_1_CAMLAT		0x00000000000f0000ULL
+#define  FFLP_CFG_1_CAMLAT_SHIFT	16
+#define  FFLP_CFG_1_CAMRATIO		0x000000000000f000ULL
+#define  FFLP_CFG_1_CAMRATIO_SHIFT	12
+#define  FFLP_CFG_1_FCRAMRATIO		0x0000000000000f00ULL
+#define  FFLP_CFG_1_FCRAMRATIO_SHIFT	8
+#define  FFLP_CFG_1_FCRAMOUTDR_MASK	0x00000000000000f0ULL
+#define  FFLP_CFG_1_FCRAMOUTDR_NORMAL	0x0000000000000000ULL
+#define  FFLP_CFG_1_FCRAMOUTDR_STRONG	0x0000000000000050ULL
+#define  FFLP_CFG_1_FCRAMOUTDR_WEAK	0x00000000000000a0ULL
+#define  FFLP_CFG_1_FCRAMQS		0x0000000000000008ULL
+#define  FFLP_CFG_1_ERRORDIS		0x0000000000000004ULL
+#define  FFLP_CFG_1_FFLPINITDONE	0x0000000000000002ULL
+#define  FFLP_CFG_1_LLCSNAP		0x0000000000000001ULL
+
+#define DEFAULT_FCRAMRATIO		10
+
+#define DEFAULT_TCAM_LATENCY		4
+#define DEFAULT_TCAM_ACCESS_RATIO	10
+
+#define TCP_CFLAG_MSK			(FZC_FFLP + 0x20108UL)
+#define  TCP_CFLAG_MSK_MASK		0x0000000000000fffULL
+
+#define FCRAM_REF_TMR			(FZC_FFLP + 0x20110UL)
+#define  FCRAM_REF_TMR_MAX		0x00000000ffff0000ULL
+#define  FCRAM_REF_TMR_MAX_SHIFT	16
+#define  FCRAM_REF_TMR_MIN		0x000000000000ffffULL
+#define  FCRAM_REF_TMR_MIN_SHIFT	0
+
+#define DEFAULT_FCRAM_REFRESH_MAX	512
+#define DEFAULT_FCRAM_REFRESH_MIN	512
+
+#define FCRAM_FIO_ADDR			(FZC_FFLP + 0x20118UL)
+#define  FCRAM_FIO_ADDR_ADDR		0x00000000000000ffULL
+
+#define FCRAM_FIO_DAT			(FZC_FFLP + 0x20120UL)
+#define  FCRAM_FIO_DAT_DATA		0x000000000000ffffULL
+
+#define FCRAM_ERR_TST0			(FZC_FFLP + 0x20128UL)
+#define  FCRAM_ERR_TST0_SYND		0x00000000000000ffULL
+
+#define FCRAM_ERR_TST1			(FZC_FFLP + 0x20130UL)
+#define  FCRAM_ERR_TST1_DAT		0x00000000ffffffffULL
+
+#define FCRAM_ERR_TST2			(FZC_FFLP + 0x20138UL)
+#define  FCRAM_ERR_TST2_DAT		0x00000000ffffffffULL
+
+#define FFLP_ERR_MASK			(FZC_FFLP + 0x20140UL)
+#define  FFLP_ERR_MASK_HSH_TBL_DAT	0x00000000000007f8ULL
+#define  FFLP_ERR_MASK_HSH_TBL_LKUP	0x0000000000000004ULL
+#define  FFLP_ERR_MASK_TCAM		0x0000000000000002ULL
+#define  FFLP_ERR_MASK_VLAN		0x0000000000000001ULL
+
+#define FFLP_DBG_TRAIN_VCT		(FZC_FFLP + 0x20148UL)
+#define  FFLP_DBG_TRAIN_VCT_VECTOR	0x00000000ffffffffULL
+
+#define FCRAM_PHY_RD_LAT		(FZC_FFLP + 0x20150UL)
+#define  FCRAM_PHY_RD_LAT_LAT		0x00000000000000ffULL
+
+/* Ethernet TCAM format */
+#define TCAM_ETHKEY0_RESV1		0xffffffffffffff00ULL
+#define TCAM_ETHKEY0_CLASS_CODE		0x00000000000000f8ULL
+#define TCAM_ETHKEY0_CLASS_CODE_SHIFT	3
+#define TCAM_ETHKEY0_RESV2		0x0000000000000007ULL
+#define TCAM_ETHKEY1_FRAME_BYTE0_7(NUM)	(0xff << ((7 - NUM) * 8))
+#define TCAM_ETHKEY2_FRAME_BYTE8	0xff00000000000000ULL
+#define TCAM_ETHKEY2_FRAME_BYTE8_SHIFT	56
+#define TCAM_ETHKEY2_FRAME_BYTE9	0x00ff000000000000ULL
+#define TCAM_ETHKEY2_FRAME_BYTE9_SHIFT	48
+#define TCAM_ETHKEY2_FRAME_BYTE10	0x0000ff0000000000ULL
+#define TCAM_ETHKEY2_FRAME_BYTE10_SHIFT	40
+#define TCAM_ETHKEY2_FRAME_RESV		0x000000ffffffffffULL
+#define TCAM_ETHKEY3_FRAME_RESV		0xffffffffffffffffULL
+
+/* IPV4 TCAM format */
+#define TCAM_V4KEY0_RESV1		0xffffffffffffff00ULL
+#define TCAM_V4KEY0_CLASS_CODE		0x00000000000000f8ULL
+#define TCAM_V4KEY0_CLASS_CODE_SHIFT	3
+#define TCAM_V4KEY0_RESV2		0x0000000000000007ULL
+#define TCAM_V4KEY1_L2RDCNUM		0xf800000000000000ULL
+#define TCAM_V4KEY1_L2RDCNUM_SHIFT	59
+#define TCAM_V4KEY1_NOPORT		0x0400000000000000ULL
+#define TCAM_V4KEY1_RESV		0x03ffffffffffffffULL
+#define TCAM_V4KEY2_RESV		0xffff000000000000ULL
+#define TCAM_V4KEY2_TOS			0x0000ff0000000000ULL
+#define TCAM_V4KEY2_TOS_SHIFT		40
+#define TCAM_V4KEY2_PROTO		0x000000ff00000000ULL
+#define TCAM_V4KEY2_PROTO_SHIFT		32
+#define TCAM_V4KEY2_PORT_SPI		0x00000000ffffffffULL
+#define TCAM_V4KEY2_PORT_SPI_SHIFT	0
+#define TCAM_V4KEY3_SADDR		0xffffffff00000000ULL
+#define TCAM_V4KEY3_SADDR_SHIFT		32
+#define TCAM_V4KEY3_DADDR		0x00000000ffffffffULL
+#define TCAM_V4KEY3_DADDR_SHIFT		0
+
+/* IPV6 TCAM format */
+#define TCAM_V6KEY0_RESV1		0xffffffffffffff00ULL
+#define TCAM_V6KEY0_CLASS_CODE		0x00000000000000f8ULL
+#define TCAM_V6KEY0_CLASS_CODE_SHIFT	3
+#define TCAM_V6KEY0_RESV2		0x0000000000000007ULL
+#define TCAM_V6KEY1_L2RDCNUM		0xf800000000000000ULL
+#define TCAM_V6KEY1_L2RDCNUM_SHIFT	59
+#define TCAM_V6KEY1_NOPORT		0x0400000000000000ULL
+#define TCAM_V6KEY1_RESV		0x03ff000000000000ULL
+#define TCAM_V6KEY1_TOS			0x0000ff0000000000ULL
+#define TCAM_V6KEY1_TOS_SHIFT		40
+#define TCAM_V6KEY1_NEXT_HDR		0x000000ff00000000ULL
+#define TCAM_V6KEY1_NEXT_HDR_SHIFT	32
+#define TCAM_V6KEY1_PORT_SPI		0x00000000ffffffffULL
+#define TCAM_V6KEY1_PORT_SPI_SHIFT	0
+#define TCAM_V6KEY2_ADDR_HIGH		0xffffffffffffffffULL
+#define TCAM_V6KEY3_ADDR_LOW		0xffffffffffffffffULL
+
+#define TCAM_ASSOCDATA_SYNDROME		0x000003fffc000000ULL
+#define TCAM_ASSOCDATA_SYNDROME_SHIFT	26
+#define TCAM_ASSOCDATA_ZFID		0x0000000003ffc000ULL
+#define TCAM_ASSOCDATA_ZFID_SHIFT	14
+#define TCAM_ASSOCDATA_V4_ECC_OK	0x0000000000002000ULL
+#define TCAM_ASSOCDATA_DISC		0x0000000000001000ULL
+#define TCAM_ASSOCDATA_TRES_MASK	0x0000000000000c00ULL
+#define TCAM_ASSOCDATA_TRES_USE_L2RDC	0x0000000000000000ULL
+#define TCAM_ASSOCDATA_TRES_USE_OFFSET	0x0000000000000400ULL
+#define TCAM_ASSOCDATA_TRES_OVR_RDC	0x0000000000000800ULL
+#define TCAM_ASSOCDATA_TRES_OVR_RDC_OFF	0x0000000000000c00ULL
+#define TCAM_ASSOCDATA_RDCTBL		0x0000000000000380ULL
+#define TCAM_ASSOCDATA_RDCTBL_SHIFT	7
+#define TCAM_ASSOCDATA_OFFSET		0x000000000000007cULL
+#define TCAM_ASSOCDATA_OFFSET_SHIFT	2
+#define TCAM_ASSOCDATA_ZFVLD		0x0000000000000002ULL
+#define TCAM_ASSOCDATA_AGE		0x0000000000000001ULL
+
+#define FLOW_KEY(IDX)			(FZC_FFLP + 0x40000UL + (IDX) * 8UL)
+#define  FLOW_KEY_PORT			0x0000000000000200ULL
+#define  FLOW_KEY_L2DA			0x0000000000000100ULL
+#define  FLOW_KEY_VLAN			0x0000000000000080ULL
+#define  FLOW_KEY_IPSA			0x0000000000000040ULL
+#define  FLOW_KEY_IPDA			0x0000000000000020ULL
+#define  FLOW_KEY_PROTO			0x0000000000000010ULL
+#define  FLOW_KEY_L4_0			0x000000000000000cULL
+#define  FLOW_KEY_L4_0_SHIFT		2
+#define  FLOW_KEY_L4_1			0x0000000000000003ULL
+#define  FLOW_KEY_L4_1_SHIFT		0
+
+#define  FLOW_KEY_L4_NONE		0x0
+#define  FLOW_KEY_L4_RESV		0x1
+#define  FLOW_KEY_L4_BYTE12		0x2
+#define  FLOW_KEY_L4_BYTE56		0x3
+
+#define H1POLY				(FZC_FFLP + 0x40060UL)
+#define  H1POLY_INITVAL			0x00000000ffffffffULL
+
+#define H2POLY				(FZC_FFLP + 0x40068UL)
+#define  H2POLY_INITVAL			0x000000000000ffffULL
+
+#define FLW_PRT_SEL(IDX)		(FZC_FFLP + 0x40070UL + (IDX) * 8UL)
+#define  FLW_PRT_SEL_EXT		0x0000000000010000ULL
+#define  FLW_PRT_SEL_MASK		0x0000000000001f00ULL
+#define  FLW_PRT_SEL_MASK_SHIFT		8
+#define  FLW_PRT_SEL_BASE		0x000000000000001fULL
+#define  FLW_PRT_SEL_BASE_SHIFT		0
+
+#define HASH_TBL_ADDR(IDX)		(FFLP + 0x00000UL + (IDX) * 8192UL)
+#define  HASH_TBL_ADDR_AUTOINC		0x0000000000800000ULL
+#define  HASH_TBL_ADDR_ADDR		0x00000000007fffffULL
+
+#define HASH_TBL_DATA(IDX)		(FFLP + 0x00008UL + (IDX) * 8192UL)
+#define  HASH_TBL_DATA_DATA		0xffffffffffffffffULL
+
+/* FCRAM hash table entries are up to 8 64-bit words in size.
+ * The layout of each entry is determined by the settings in the
+ * first word, which is the header.
+ *
+ * The indexing is controllable per partition (there is one partition
+ * per RDC group, thus a total of eight) using the BASE and MASK fields
+ * of FLW_PRT_SEL above.
+ */
+#define FCRAM_SIZE			0x800000
+#define FCRAM_NUM_PARTITIONS		8
+
+/* Generic HASH entry header, used for all non-optimized formats.  */
+#define HASH_HEADER_FMT			0x8000000000000000ULL
+#define HASH_HEADER_EXT			0x4000000000000000ULL
+#define HASH_HEADER_VALID		0x2000000000000000ULL
+#define HASH_HEADER_RESVD		0x1000000000000000ULL
+#define HASH_HEADER_L2_DADDR		0x0ffffffffffff000ULL
+#define HASH_HEADER_L2_DADDR_SHIFT	12
+#define HASH_HEADER_VLAN		0x0000000000000fffULL
+#define HASH_HEADER_VLAN_SHIFT		0
+
+/* Optimized format, just a header with a special layout defined below.
+ * Set FMT and EXT both to zero to indicate this layout is being used.
+ */
+#define HASH_OPT_HEADER_FMT		0x8000000000000000ULL
+#define HASH_OPT_HEADER_EXT		0x4000000000000000ULL
+#define HASH_OPT_HEADER_VALID		0x2000000000000000ULL
+#define HASH_OPT_HEADER_RDCOFF		0x1f00000000000000ULL
+#define HASH_OPT_HEADER_RDCOFF_SHIFT	56
+#define HASH_OPT_HEADER_HASH2		0x00ffff0000000000ULL
+#define HASH_OPT_HEADER_HASH2_SHIFT	40
+#define HASH_OPT_HEADER_RESVD		0x000000ff00000000ULL
+#define HASH_OPT_HEADER_USERINFO	0x00000000ffffffffULL
+#define HASH_OPT_HEADER_USERINFO_SHIFT	0
+
+/* Port and protocol word used for ipv4 and ipv6 layouts.  */
+#define HASH_PORT_DPORT			0xffff000000000000ULL
+#define HASH_PORT_DPORT_SHIFT		48
+#define HASH_PORT_SPORT			0x0000ffff00000000ULL
+#define HASH_PORT_SPORT_SHIFT		32
+#define HASH_PORT_PROTO			0x00000000ff000000ULL
+#define HASH_PORT_PROTO_SHIFT		24
+#define HASH_PORT_PORT_OFF		0x0000000000c00000ULL
+#define HASH_PORT_PORT_OFF_SHIFT	22
+#define HASH_PORT_PORT_RESV		0x00000000003fffffULL
+
+/* Action word used for ipv4 and ipv6 layouts.  */
+#define HASH_ACTION_RESV1		0xe000000000000000ULL
+#define HASH_ACTION_RDCOFF		0x1f00000000000000ULL
+#define HASH_ACTION_RDCOFF_SHIFT	56
+#define HASH_ACTION_ZFVALID		0x0080000000000000ULL
+#define HASH_ACTION_RESV2		0x0070000000000000ULL
+#define HASH_ACTION_ZFID		0x000fff0000000000ULL
+#define HASH_ACTION_ZFID_SHIFT		40
+#define HASH_ACTION_RESV3		0x000000ff00000000ULL
+#define HASH_ACTION_USERINFO		0x00000000ffffffffULL
+#define HASH_ACTION_USERINFO_SHIFT	0
+
+/* IPV4 address word.  Addresses are in network endian. */
+#define HASH_IP4ADDR_SADDR		0xffffffff00000000ULL
+#define HASH_IP4ADDR_SADDR_SHIFT	32
+#define HASH_IP4ADDR_DADDR		0x00000000ffffffffULL
+#define HASH_IP4ADDR_DADDR_SHIFT	0
+
+/* IPV6 address layout is 4 words, first two are saddr, next two
+ * are daddr.  Addresses are in network endian.
+ */
+
+struct fcram_hash_opt {
+	u64	header;
+};
+
+/* EXT=1, FMT=0 */
+struct fcram_hash_ipv4 {
+	u64	header;
+	u64	addrs;
+	u64	ports;
+	u64	action;
+};
+
+/* EXT=1, FMT=1 */
+struct fcram_hash_ipv6 {
+	u64	header;
+	u64	addrs[4];
+	u64	ports;
+	u64	action;
+};
+
+#define HASH_TBL_DATA_LOG(IDX)		(FFLP + 0x00010UL + (IDX) * 8192UL)
+#define  HASH_TBL_DATA_LOG_ERR		0x0000000080000000ULL
+#define  HASH_TBL_DATA_LOG_ADDR		0x000000007fffff00ULL
+#define  HASH_TBL_DATA_LOG_SYNDROME	0x00000000000000ffULL
+
+#define RX_DMA_CK_DIV			(FZC_DMC + 0x00000UL)
+#define  RX_DMA_CK_DIV_CNT		0x000000000000ffffULL
+
+#define DEF_RDC(IDX)			(FZC_DMC + 0x00008UL + (IDX) * 0x8UL)
+#define  DEF_RDC_VAL			0x000000000000001fULL
+
+#define PT_DRR_WT(IDX)			(FZC_DMC + 0x00028UL + (IDX) * 0x8UL)
+#define  PT_DRR_WT_VAL			0x000000000000ffffULL
+
+#define PT_DRR_WEIGHT_DEFAULT_10G	0x0400
+#define PT_DRR_WEIGHT_DEFAULT_1G	0x0066
+
+#define PT_USE(IDX)			(FZC_DMC + 0x00048UL + (IDX) * 0x8UL)
+#define  PT_USE_CNT			0x00000000000fffffULL
+
+#define RED_RAN_INIT			(FZC_DMC + 0x00068UL)
+#define  RED_RAN_INIT_OPMODE		0x0000000000010000ULL
+#define  RED_RAN_INIT_VAL		0x000000000000ffffULL
+
+#define RX_ADDR_MD			(FZC_DMC + 0x00070UL)
+#define  RX_ADDR_MD_DBG_PT_MUX_SEL	0x000000000000000cULL
+#define  RX_ADDR_MD_RAM_ACC		0x0000000000000002ULL
+#define  RX_ADDR_MD_MODE32		0x0000000000000001ULL
+
+#define RDMC_PRE_PAR_ERR		(FZC_DMC + 0x00078UL)
+#define  RDMC_PRE_PAR_ERR_ERR		0x0000000000008000ULL
+#define  RDMC_PRE_PAR_ERR_MERR		0x0000000000004000ULL
+#define  RDMC_PRE_PAR_ERR_ADDR		0x00000000000000ffULL
+
+#define RDMC_SHA_PAR_ERR		(FZC_DMC + 0x00080UL)
+#define  RDMC_SHA_PAR_ERR_ERR		0x0000000000008000ULL
+#define  RDMC_SHA_PAR_ERR_MERR		0x0000000000004000ULL
+#define  RDMC_SHA_PAR_ERR_ADDR		0x00000000000000ffULL
+
+#define RDMC_MEM_ADDR			(FZC_DMC + 0x00088UL)
+#define  RDMC_MEM_ADDR_PRE_SHAD		0x0000000000000100ULL
+#define  RDMC_MEM_ADDR_ADDR		0x00000000000000ffULL
+
+#define RDMC_MEM_DAT0			(FZC_DMC + 0x00090UL)
+#define  RDMC_MEM_DAT0_DATA		0x00000000ffffffffULL /* bits 31:0 */
+
+#define RDMC_MEM_DAT1			(FZC_DMC + 0x00098UL)
+#define  RDMC_MEM_DAT1_DATA		0x00000000ffffffffULL /* bits 63:32 */
+
+#define RDMC_MEM_DAT2			(FZC_DMC + 0x000a0UL)
+#define  RDMC_MEM_DAT2_DATA		0x00000000ffffffffULL /* bits 95:64 */
+
+#define RDMC_MEM_DAT3			(FZC_DMC + 0x000a8UL)
+#define  RDMC_MEM_DAT3_DATA		0x00000000ffffffffULL /* bits 127:96 */
+
+#define RDMC_MEM_DAT4			(FZC_DMC + 0x000b0UL)
+#define  RDMC_MEM_DAT4_DATA		0x00000000000fffffULL /* bits 147:128 */
+
+#define RX_CTL_DAT_FIFO_STAT			(FZC_DMC + 0x000b8UL)
+#define  RX_CTL_DAT_FIFO_STAT_ID_MISMATCH	0x0000000000000100ULL
+#define  RX_CTL_DAT_FIFO_STAT_ZCP_EOP_ERR	0x00000000000000f0ULL
+#define  RX_CTL_DAT_FIFO_STAT_IPP_EOP_ERR	0x000000000000000fULL
+
+#define RX_CTL_DAT_FIFO_MASK			(FZC_DMC + 0x000c0UL)
+#define  RX_CTL_DAT_FIFO_MASK_ID_MISMATCH	0x0000000000000100ULL
+#define  RX_CTL_DAT_FIFO_MASK_ZCP_EOP_ERR	0x00000000000000f0ULL
+#define  RX_CTL_DAT_FIFO_MASK_IPP_EOP_ERR	0x000000000000000fULL
+
+#define RDMC_TRAINING_VECTOR			(FZC_DMC + 0x000c8UL)
+#define  RDMC_TRAINING_VECTOR_TRAINING_VECTOR	0x00000000ffffffffULL
+
+#define RX_CTL_DAT_FIFO_STAT_DBG		(FZC_DMC + 0x000d0UL)
+#define  RX_CTL_DAT_FIFO_STAT_DBG_ID_MISMATCH	0x0000000000000100ULL
+#define  RX_CTL_DAT_FIFO_STAT_DBG_ZCP_EOP_ERR	0x00000000000000f0ULL
+#define  RX_CTL_DAT_FIFO_STAT_DBG_IPP_EOP_ERR	0x000000000000000fULL
+
+#define RDC_TBL(TBL,SLOT)		(FZC_ZCP + 0x10000UL + \
+					 (TBL) * (8UL * 16UL) + \
+					 (SLOT) * 8UL)
+#define  RDC_TBL_RDC			0x000000000000000fULL
+
+#define RX_LOG_PAGE_VLD(IDX)		(FZC_DMC + 0x20000UL + (IDX) * 0x40UL)
+#define  RX_LOG_PAGE_VLD_FUNC		0x000000000000000cULL
+#define  RX_LOG_PAGE_VLD_FUNC_SHIFT	2
+#define  RX_LOG_PAGE_VLD_PAGE1		0x0000000000000002ULL
+#define  RX_LOG_PAGE_VLD_PAGE0		0x0000000000000001ULL
+
+#define RX_LOG_MASK1(IDX)		(FZC_DMC + 0x20008UL + (IDX) * 0x40UL)
+#define  RX_LOG_MASK1_MASK		0x00000000ffffffffULL
+
+#define RX_LOG_VAL1(IDX)		(FZC_DMC + 0x20010UL + (IDX) * 0x40UL)
+#define  RX_LOG_VAL1_VALUE		0x00000000ffffffffULL
+
+#define RX_LOG_MASK2(IDX)		(FZC_DMC + 0x20018UL + (IDX) * 0x40UL)
+#define  RX_LOG_MASK2_MASK		0x00000000ffffffffULL
+
+#define RX_LOG_VAL2(IDX)		(FZC_DMC + 0x20020UL + (IDX) * 0x40UL)
+#define  RX_LOG_VAL2_VALUE		0x00000000ffffffffULL
+
+#define RX_LOG_PAGE_RELO1(IDX)		(FZC_DMC + 0x20028UL + (IDX) * 0x40UL)
+#define  RX_LOG_PAGE_RELO1_RELO		0x00000000ffffffffULL
+
+#define RX_LOG_PAGE_RELO2(IDX)		(FZC_DMC + 0x20030UL + (IDX) * 0x40UL)
+#define  RX_LOG_PAGE_RELO2_RELO		0x00000000ffffffffULL
+
+#define RX_LOG_PAGE_HDL(IDX)		(FZC_DMC + 0x20038UL + (IDX) * 0x40UL)
+#define  RX_LOG_PAGE_HDL_HANDLE		0x00000000000fffffULL
+
+#define TX_LOG_PAGE_VLD(IDX)		(FZC_DMC + 0x40000UL + (IDX) * 0x200UL)
+#define  TX_LOG_PAGE_VLD_FUNC		0x000000000000000cULL
+#define  TX_LOG_PAGE_VLD_FUNC_SHIFT	2
+#define  TX_LOG_PAGE_VLD_PAGE1		0x0000000000000002ULL
+#define  TX_LOG_PAGE_VLD_PAGE0		0x0000000000000001ULL
+
+#define TX_LOG_MASK1(IDX)		(FZC_DMC + 0x40008UL + (IDX) * 0x200UL)
+#define  TX_LOG_MASK1_MASK		0x00000000ffffffffULL
+
+#define TX_LOG_VAL1(IDX)		(FZC_DMC + 0x40010UL + (IDX) * 0x200UL)
+#define  TX_LOG_VAL1_VALUE		0x00000000ffffffffULL
+
+#define TX_LOG_MASK2(IDX)		(FZC_DMC + 0x40018UL + (IDX) * 0x200UL)
+#define  TX_LOG_MASK2_MASK		0x00000000ffffffffULL
+
+#define TX_LOG_VAL2(IDX)		(FZC_DMC + 0x40020UL + (IDX) * 0x200UL)
+#define  TX_LOG_VAL2_VALUE		0x00000000ffffffffULL
+
+#define TX_LOG_PAGE_RELO1(IDX)		(FZC_DMC + 0x40028UL + (IDX) * 0x200UL)
+#define  TX_LOG_PAGE_RELO1_RELO		0x00000000ffffffffULL
+
+#define TX_LOG_PAGE_RELO2(IDX)		(FZC_DMC + 0x40030UL + (IDX) * 0x200UL)
+#define  TX_LOG_PAGE_RELO2_RELO		0x00000000ffffffffULL
+
+#define TX_LOG_PAGE_HDL(IDX)		(FZC_DMC + 0x40038UL + (IDX) * 0x200UL)
+#define  TX_LOG_PAGE_HDL_HANDLE		0x00000000000fffffULL
+
+#define TX_ADDR_MD			(FZC_DMC + 0x45000UL)
+#define  TX_ADDR_MD_MODE32		0x0000000000000001ULL
+
+#define RDC_RED_PARA(IDX)		(FZC_DMC + 0x30000UL + (IDX) * 0x40UL)
+#define  RDC_RED_PARA_THRE_SYN		0x00000000fff00000ULL
+#define  RDC_RED_PARA_THRE_SYN_SHIFT	20
+#define  RDC_RED_PARA_WIN_SYN		0x00000000000f0000ULL
+#define  RDC_RED_PARA_WIN_SYN_SHIFT	16
+#define  RDC_RED_PARA_THRE		0x000000000000fff0ULL
+#define  RDC_RED_PARA_THRE_SHIFT	4
+#define  RDC_RED_PARA_WIN		0x000000000000000fULL
+#define  RDC_RED_PARA_WIN_SHIFT		0
+
+#define RED_DIS_CNT(IDX)		(FZC_DMC + 0x30008UL + (IDX) * 0x40UL)
+#define  RED_DIS_CNT_OFLOW		0x0000000000010000ULL
+#define  RED_DIS_CNT_COUNT		0x000000000000ffffULL
+
+#define IPP_CFIG			(FZC_IPP + 0x00000UL)
+#define  IPP_CFIG_SOFT_RST		0x0000000080000000ULL
+#define  IPP_CFIG_IP_MAX_PKT		0x0000000001ffff00ULL
+#define  IPP_CFIG_IP_MAX_PKT_SHIFT	8
+#define  IPP_CFIG_FFLP_CS_PIO_W		0x0000000000000080ULL
+#define  IPP_CFIG_PFIFO_PIO_W		0x0000000000000040ULL
+#define  IPP_CFIG_DFIFO_PIO_W		0x0000000000000020ULL
+#define  IPP_CFIG_CKSUM_EN		0x0000000000000010ULL
+#define  IPP_CFIG_DROP_BAD_CRC		0x0000000000000008ULL
+#define  IPP_CFIG_DFIFO_ECC_EN		0x0000000000000004ULL
+#define  IPP_CFIG_DEBUG_BUS_OUT_EN	0x0000000000000002ULL
+#define  IPP_CFIG_IPP_ENABLE		0x0000000000000001ULL
+
+#define IPP_PKT_DIS			(FZC_IPP + 0x00020UL)
+#define  IPP_PKT_DIS_COUNT		0x0000000000003fffULL
+
+#define IPP_BAD_CS_CNT			(FZC_IPP + 0x00028UL)
+#define  IPP_BAD_CS_CNT_COUNT		0x0000000000003fffULL
+
+#define IPP_ECC				(FZC_IPP + 0x00030UL)
+#define  IPP_ECC_COUNT			0x00000000000000ffULL
+
+#define IPP_INT_STAT			(FZC_IPP + 0x00040UL)
+#define  IPP_INT_STAT_SOP_MISS		0x0000000080000000ULL
+#define  IPP_INT_STAT_EOP_MISS		0x0000000040000000ULL
+#define  IPP_INT_STAT_DFIFO_UE		0x0000000030000000ULL
+#define  IPP_INT_STAT_DFIFO_CE		0x000000000c000000ULL
+#define  IPP_INT_STAT_DFIFO_ECC		0x0000000003000000ULL
+#define  IPP_INT_STAT_DFIFO_ECC_IDX	0x00000000007ff000ULL
+#define  IPP_INT_STAT_PFIFO_PERR	0x0000000000000800ULL
+#define  IPP_INT_STAT_ECC_ERR_MAX	0x0000000000000400ULL
+#define  IPP_INT_STAT_PFIFO_ERR_IDX	0x00000000000003f0ULL
+#define  IPP_INT_STAT_PFIFO_OVER	0x0000000000000008ULL
+#define  IPP_INT_STAT_PFIFO_UND		0x0000000000000004ULL
+#define  IPP_INT_STAT_BAD_CS_MX		0x0000000000000002ULL
+#define  IPP_INT_STAT_PKT_DIS_MX	0x0000000000000001ULL
+#define  IPP_INT_STAT_ALL		0x00000000ff7fffffULL
+
+#define IPP_MSK				(FZC_IPP + 0x00048UL)
+#define  IPP_MSK_ECC_ERR_MX		0x0000000000000080ULL
+#define  IPP_MSK_DFIFO_EOP_SOP		0x0000000000000040ULL
+#define  IPP_MSK_DFIFO_UC		0x0000000000000020ULL
+#define  IPP_MSK_PFIFO_PAR		0x0000000000000010ULL
+#define  IPP_MSK_PFIFO_OVER		0x0000000000000008ULL
+#define  IPP_MSK_PFIFO_UND		0x0000000000000004ULL
+#define  IPP_MSK_BAD_CS			0x0000000000000002ULL
+#define  IPP_MSK_PKT_DIS_CNT		0x0000000000000001ULL
+#define  IPP_MSK_ALL			0x00000000000000ffULL
+
+#define IPP_PFIFO_RD0			(FZC_IPP + 0x00060UL)
+#define  IPP_PFIFO_RD0_DATA		0x00000000ffffffffULL /* bits 31:0 */
+
+#define IPP_PFIFO_RD1			(FZC_IPP + 0x00068UL)
+#define  IPP_PFIFO_RD1_DATA		0x00000000ffffffffULL /* bits 63:32 */
+
+#define IPP_PFIFO_RD2			(FZC_IPP + 0x00070UL)
+#define  IPP_PFIFO_RD2_DATA		0x00000000ffffffffULL /* bits 95:64 */
+
+#define IPP_PFIFO_RD3			(FZC_IPP + 0x00078UL)
+#define  IPP_PFIFO_RD3_DATA		0x00000000ffffffffULL /* bits 127:96 */
+
+#define IPP_PFIFO_RD4			(FZC_IPP + 0x00080UL)
+#define  IPP_PFIFO_RD4_DATA		0x00000000ffffffffULL /* bits 145:128 */
+
+#define IPP_PFIFO_WR0			(FZC_IPP + 0x00088UL)
+#define  IPP_PFIFO_WR0_DATA		0x00000000ffffffffULL /* bits 31:0 */
+
+#define IPP_PFIFO_WR1			(FZC_IPP + 0x00090UL)
+#define  IPP_PFIFO_WR1_DATA		0x00000000ffffffffULL /* bits 63:32 */
+
+#define IPP_PFIFO_WR2			(FZC_IPP + 0x00098UL)
+#define  IPP_PFIFO_WR2_DATA		0x00000000ffffffffULL /* bits 95:64 */
+
+#define IPP_PFIFO_WR3			(FZC_IPP + 0x000a0UL)
+#define  IPP_PFIFO_WR3_DATA		0x00000000ffffffffULL /* bits 127:96 */
+
+#define IPP_PFIFO_WR4			(FZC_IPP + 0x000a8UL)
+#define  IPP_PFIFO_WR4_DATA		0x00000000ffffffffULL /* bits 145:128 */
+
+#define IPP_PFIFO_RD_PTR		(FZC_IPP + 0x000b0UL)
+#define  IPP_PFIFO_RD_PTR_PTR		0x000000000000003fULL
+
+#define IPP_PFIFO_WR_PTR		(FZC_IPP + 0x000b8UL)
+#define  IPP_PFIFO_WR_PTR_PTR		0x000000000000007fULL
+
+#define IPP_DFIFO_RD0			(FZC_IPP + 0x000c0UL)
+#define  IPP_DFIFO_RD0_DATA		0x00000000ffffffffULL /* bits 31:0 */
+
+#define IPP_DFIFO_RD1			(FZC_IPP + 0x000c8UL)
+#define  IPP_DFIFO_RD1_DATA		0x00000000ffffffffULL /* bits 63:32 */
+
+#define IPP_DFIFO_RD2			(FZC_IPP + 0x000d0UL)
+#define  IPP_DFIFO_RD2_DATA		0x00000000ffffffffULL /* bits 95:64 */
+
+#define IPP_DFIFO_RD3			(FZC_IPP + 0x000d8UL)
+#define  IPP_DFIFO_RD3_DATA		0x00000000ffffffffULL /* bits 127:96 */
+
+#define IPP_DFIFO_RD4			(FZC_IPP + 0x000e0UL)
+#define  IPP_DFIFO_RD4_DATA		0x00000000ffffffffULL /* bits 145:128 */
+
+#define IPP_DFIFO_WR0			(FZC_IPP + 0x000e8UL)
+#define  IPP_DFIFO_WR0_DATA		0x00000000ffffffffULL /* bits 31:0 */
+
+#define IPP_DFIFO_WR1			(FZC_IPP + 0x000f0UL)
+#define  IPP_DFIFO_WR1_DATA		0x00000000ffffffffULL /* bits 63:32 */
+
+#define IPP_DFIFO_WR2			(FZC_IPP + 0x000f8UL)
+#define  IPP_DFIFO_WR2_DATA		0x00000000ffffffffULL /* bits 95:64 */
+
+#define IPP_DFIFO_WR3			(FZC_IPP + 0x00100UL)
+#define  IPP_DFIFO_WR3_DATA		0x00000000ffffffffULL /* bits 127:96 */
+
+#define IPP_DFIFO_WR4			(FZC_IPP + 0x00108UL)
+#define  IPP_DFIFO_WR4_DATA		0x00000000ffffffffULL /* bits 145:128 */
+
+#define IPP_DFIFO_RD_PTR		(FZC_IPP + 0x00110UL)
+#define  IPP_DFIFO_RD_PTR_PTR		0x0000000000000fffULL
+
+#define IPP_DFIFO_WR_PTR		(FZC_IPP + 0x00118UL)
+#define  IPP_DFIFO_WR_PTR_PTR		0x0000000000000fffULL
+
+#define IPP_SM				(FZC_IPP + 0x00120UL)
+#define  IPP_SM_SM			0x00000000ffffffffULL
+
+#define IPP_CS_STAT			(FZC_IPP + 0x00128UL)
+#define  IPP_CS_STAT_BCYC_CNT		0x00000000ff000000ULL
+#define  IPP_CS_STAT_IP_LEN		0x0000000000fff000ULL
+#define  IPP_CS_STAT_CS_FAIL		0x0000000000000800ULL
+#define  IPP_CS_STAT_TERM		0x0000000000000400ULL
+#define  IPP_CS_STAT_BAD_NUM		0x0000000000000200ULL
+#define  IPP_CS_STAT_CS_STATE		0x00000000000001ffULL
+
+#define IPP_FFLP_CS_INFO		(FZC_IPP + 0x00130UL)
+#define  IPP_FFLP_CS_INFO_PKT_ID	0x0000000000003c00ULL
+#define  IPP_FFLP_CS_INFO_L4_PROTO	0x0000000000000300ULL
+#define  IPP_FFLP_CS_INFO_V4_HD_LEN	0x00000000000000f0ULL
+#define  IPP_FFLP_CS_INFO_L3_VER	0x000000000000000cULL
+#define  IPP_FFLP_CS_INFO_L2_OP		0x0000000000000003ULL
+
+#define IPP_DBG_SEL			(FZC_IPP + 0x00138UL)
+#define  IPP_DBG_SEL_SEL		0x000000000000000fULL
+
+#define IPP_DFIFO_ECC_SYND		(FZC_IPP + 0x00140UL)
+#define  IPP_DFIFO_ECC_SYND_SYND	0x000000000000ffffULL
+
+#define IPP_DFIFO_EOP_RD_PTR		(FZC_IPP + 0x00148UL)
+#define  IPP_DFIFO_EOP_RD_PTR_PTR	0x0000000000000fffULL
+
+#define IPP_ECC_CTL			(FZC_IPP + 0x00150UL)
+#define  IPP_ECC_CTL_DIS_DBL		0x0000000080000000ULL
+#define  IPP_ECC_CTL_COR_DBL		0x0000000000020000ULL
+#define  IPP_ECC_CTL_COR_SNG		0x0000000000010000ULL
+#define  IPP_ECC_CTL_COR_ALL		0x0000000000000400ULL
+#define  IPP_ECC_CTL_COR_1		0x0000000000000100ULL
+#define  IPP_ECC_CTL_COR_LST		0x0000000000000004ULL
+#define  IPP_ECC_CTL_COR_SND		0x0000000000000002ULL
+#define  IPP_ECC_CTL_COR_FSR		0x0000000000000001ULL
+
+#define NIU_DFIFO_ENTRIES		1024
+#define ATLAS_P0_P1_DFIFO_ENTRIES	2048
+#define ATLAS_P2_P3_DFIFO_ENTRIES	1024
+
+#define ZCP_CFIG			(FZC_ZCP + 0x00000UL)
+#define  ZCP_CFIG_ZCP_32BIT_MODE	0x0000000001000000ULL
+#define  ZCP_CFIG_ZCP_DEBUG_SEL		0x0000000000ff0000ULL
+#define  ZCP_CFIG_DMA_TH		0x000000000000ffe0ULL
+#define  ZCP_CFIG_ECC_CHK_DIS		0x0000000000000010ULL
+#define  ZCP_CFIG_PAR_CHK_DIS		0x0000000000000008ULL
+#define  ZCP_CFIG_DIS_BUFF_RSP_IF	0x0000000000000004ULL
+#define  ZCP_CFIG_DIS_BUFF_REQ_IF	0x0000000000000002ULL
+#define  ZCP_CFIG_ZC_ENABLE		0x0000000000000001ULL
+
+#define ZCP_INT_STAT			(FZC_ZCP + 0x00008UL)
+#define  ZCP_INT_STAT_RRFIFO_UNDERRUN	0x0000000000008000ULL
+#define  ZCP_INT_STAT_RRFIFO_OVERRUN	0x0000000000004000ULL
+#define  ZCP_INT_STAT_RSPFIFO_UNCOR_ERR	0x0000000000001000ULL
+#define  ZCP_INT_STAT_BUFFER_OVERFLOW	0x0000000000000800ULL
+#define  ZCP_INT_STAT_STAT_TBL_PERR	0x0000000000000400ULL
+#define  ZCP_INT_STAT_DYN_TBL_PERR	0x0000000000000200ULL
+#define  ZCP_INT_STAT_BUF_TBL_PERR	0x0000000000000100ULL
+#define  ZCP_INT_STAT_TT_PROGRAM_ERR	0x0000000000000080ULL
+#define  ZCP_INT_STAT_RSP_TT_INDEX_ERR	0x0000000000000040ULL
+#define  ZCP_INT_STAT_SLV_TT_INDEX_ERR	0x0000000000000020ULL
+#define  ZCP_INT_STAT_ZCP_TT_INDEX_ERR	0x0000000000000010ULL
+#define  ZCP_INT_STAT_CFIFO_ECC3	0x0000000000000008ULL
+#define  ZCP_INT_STAT_CFIFO_ECC2	0x0000000000000004ULL
+#define  ZCP_INT_STAT_CFIFO_ECC1	0x0000000000000002ULL
+#define  ZCP_INT_STAT_CFIFO_ECC0	0x0000000000000001ULL
+#define  ZCP_INT_STAT_ALL		0x000000000000ffffULL
+
+#define ZCP_INT_MASK			(FZC_ZCP + 0x00010UL)
+#define  ZCP_INT_MASK_RRFIFO_UNDERRUN	0x0000000000008000ULL
+#define  ZCP_INT_MASK_RRFIFO_OVERRUN	0x0000000000004000ULL
+#define  ZCP_INT_MASK_LOJ		0x0000000000002000ULL
+#define  ZCP_INT_MASK_RSPFIFO_UNCOR_ERR	0x0000000000001000ULL
+#define  ZCP_INT_MASK_BUFFER_OVERFLOW	0x0000000000000800ULL
+#define  ZCP_INT_MASK_STAT_TBL_PERR	0x0000000000000400ULL
+#define  ZCP_INT_MASK_DYN_TBL_PERR	0x0000000000000200ULL
+#define  ZCP_INT_MASK_BUF_TBL_PERR	0x0000000000000100ULL
+#define  ZCP_INT_MASK_TT_PROGRAM_ERR	0x0000000000000080ULL
+#define  ZCP_INT_MASK_RSP_TT_INDEX_ERR	0x0000000000000040ULL
+#define  ZCP_INT_MASK_SLV_TT_INDEX_ERR	0x0000000000000020ULL
+#define  ZCP_INT_MASK_ZCP_TT_INDEX_ERR	0x0000000000000010ULL
+#define  ZCP_INT_MASK_CFIFO_ECC3	0x0000000000000008ULL
+#define  ZCP_INT_MASK_CFIFO_ECC2	0x0000000000000004ULL
+#define  ZCP_INT_MASK_CFIFO_ECC1	0x0000000000000002ULL
+#define  ZCP_INT_MASK_CFIFO_ECC0	0x0000000000000001ULL
+#define  ZCP_INT_MASK_ALL		0x000000000000ffffULL
+
+#define BAM4BUF				(FZC_ZCP + 0x00018UL)
+#define  BAM4BUF_LOJ			0x0000000080000000ULL
+#define  BAM4BUF_EN_CK			0x0000000040000000ULL
+#define  BAM4BUF_IDX_END0		0x000000003ff00000ULL
+#define  BAM4BUF_IDX_ST0		0x00000000000ffc00ULL
+#define  BAM4BUF_OFFSET0		0x00000000000003ffULL
+
+#define BAM8BUF				(FZC_ZCP + 0x00020UL)
+#define  BAM8BUF_LOJ			0x0000000080000000ULL
+#define  BAM8BUF_EN_CK			0x0000000040000000ULL
+#define  BAM8BUF_IDX_END1		0x000000003ff00000ULL
+#define  BAM8BUF_IDX_ST1		0x00000000000ffc00ULL
+#define  BAM8BUF_OFFSET1		0x00000000000003ffULL
+
+#define BAM16BUF			(FZC_ZCP + 0x00028UL)
+#define  BAM16BUF_LOJ			0x0000000080000000ULL
+#define  BAM16BUF_EN_CK			0x0000000040000000ULL
+#define  BAM16BUF_IDX_END2		0x000000003ff00000ULL
+#define  BAM16BUF_IDX_ST2		0x00000000000ffc00ULL
+#define  BAM16BUF_OFFSET2		0x00000000000003ffULL
+
+#define BAM32BUF			(FZC_ZCP + 0x00030UL)
+#define  BAM32BUF_LOJ			0x0000000080000000ULL
+#define  BAM32BUF_EN_CK			0x0000000040000000ULL
+#define  BAM32BUF_IDX_END3		0x000000003ff00000ULL
+#define  BAM32BUF_IDX_ST3		0x00000000000ffc00ULL
+#define  BAM32BUF_OFFSET3		0x00000000000003ffULL
+
+#define DST4BUF				(FZC_ZCP + 0x00038UL)
+#define  DST4BUF_DS_OFFSET0		0x00000000000003ffULL
+
+#define DST8BUF				(FZC_ZCP + 0x00040UL)
+#define  DST8BUF_DS_OFFSET1		0x00000000000003ffULL
+
+#define DST16BUF			(FZC_ZCP + 0x00048UL)
+#define  DST16BUF_DS_OFFSET2		0x00000000000003ffULL
+
+#define DST32BUF			(FZC_ZCP + 0x00050UL)
+#define  DST32BUF_DS_OFFSET3		0x00000000000003ffULL
+
+#define ZCP_RAM_DATA0			(FZC_ZCP + 0x00058UL)
+#define  ZCP_RAM_DATA0_DAT0		0x00000000ffffffffULL
+
+#define ZCP_RAM_DATA1			(FZC_ZCP + 0x00060UL)
+#define  ZCP_RAM_DAT10_DAT1		0x00000000ffffffffULL
+
+#define ZCP_RAM_DATA2			(FZC_ZCP + 0x00068UL)
+#define  ZCP_RAM_DATA2_DAT2		0x00000000ffffffffULL
+
+#define ZCP_RAM_DATA3			(FZC_ZCP + 0x00070UL)
+#define  ZCP_RAM_DATA3_DAT3		0x00000000ffffffffULL
+
+#define ZCP_RAM_DATA4			(FZC_ZCP + 0x00078UL)
+#define  ZCP_RAM_DATA4_DAT4		0x00000000000000ffULL
+
+#define ZCP_RAM_BE			(FZC_ZCP + 0x00080UL)
+#define  ZCP_RAM_BE_VAL			0x000000000001ffffULL
+
+#define ZCP_RAM_ACC			(FZC_ZCP + 0x00088UL)
+#define  ZCP_RAM_ACC_BUSY		0x0000000080000000ULL
+#define  ZCP_RAM_ACC_READ		0x0000000040000000ULL
+#define  ZCP_RAM_ACC_WRITE		0x0000000000000000ULL
+#define  ZCP_RAM_ACC_LOJ		0x0000000020000000ULL
+#define  ZCP_RAM_ACC_ZFCID		0x000000001ffe0000ULL
+#define  ZCP_RAM_ACC_ZFCID_SHIFT	17
+#define  ZCP_RAM_ACC_RAM_SEL		0x000000000001f000ULL
+#define  ZCP_RAM_ACC_RAM_SEL_SHIFT	12
+#define  ZCP_RAM_ACC_CFIFOADDR		0x0000000000000fffULL
+#define  ZCP_RAM_ACC_CFIFOADDR_SHIFT	0
+
+#define ZCP_RAM_SEL_BAM(INDEX)		(0x00 + (INDEX))
+#define ZCP_RAM_SEL_TT_STATIC		0x08
+#define ZCP_RAM_SEL_TT_DYNAMIC		0x09
+#define ZCP_RAM_SEL_CFIFO(PORT)		(0x10 + (PORT))
+
+#define NIU_CFIFO_ENTRIES		1024
+#define ATLAS_P0_P1_CFIFO_ENTRIES	2048
+#define ATLAS_P2_P3_CFIFO_ENTRIES	1024
+
+#define CHK_BIT_DATA			(FZC_ZCP + 0x00090UL)
+#define  CHK_BIT_DATA_DATA		0x000000000000ffffULL
+
+#define RESET_CFIFO			(FZC_ZCP + 0x00098UL)
+#define  RESET_CFIFO_RST(PORT)		(0x1 << (PORT))
+
+#define CFIFO_ECC(PORT)			(FZC_ZCP + 0x000a0UL + (PORT) * 8UL)
+#define  CFIFO_ECC_DIS_DBLBIT_ERR	0x0000000080000000ULL
+#define  CFIFO_ECC_DBLBIT_ERR		0x0000000000020000ULL
+#define  CFIFO_ECC_SINGLEBIT_ERR	0x0000000000010000ULL
+#define  CFIFO_ECC_ALL_PKT		0x0000000000000400ULL
+#define  CFIFO_ECC_LAST_LINE		0x0000000000000004ULL
+#define  CFIFO_ECC_2ND_LINE		0x0000000000000002ULL
+#define  CFIFO_ECC_1ST_LINE		0x0000000000000001ULL
+
+#define ZCP_TRAINING_VECTOR		(FZC_ZCP + 0x000c0UL)
+#define  ZCP_TRAINING_VECTOR_VECTOR	0x00000000ffffffffULL
+
+#define ZCP_STATE_MACHINE		(FZC_ZCP + 0x000c8UL)
+#define  ZCP_STATE_MACHINE_SM		0x00000000ffffffffULL
+
+/* Same bits as ZCP_INT_STAT */
+#define ZCP_INT_STAT_TEST		(FZC_ZCP + 0x00108UL)
+
+#define RXDMA_CFIG1(IDX)		(DMC + 0x00000UL + (IDX) * 0x200UL)
+#define  RXDMA_CFIG1_EN			0x0000000080000000ULL
+#define  RXDMA_CFIG1_RST		0x0000000040000000ULL
+#define  RXDMA_CFIG1_QST		0x0000000020000000ULL
+#define  RXDMA_CFIG1_MBADDR_H		0x0000000000000fffULL /* mboxaddr 43:32 */
+
+#define RXDMA_CFIG2(IDX)		(DMC + 0x00008UL + (IDX) * 0x200UL)
+#define  RXDMA_CFIG2_MBADDR_L		0x00000000ffffffc0ULL /* mboxaddr 31:6 */
+#define  RXDMA_CFIG2_OFFSET		0x0000000000000006ULL
+#define  RXDMA_CFIG2_OFFSET_SHIFT	1
+#define  RXDMA_CFIG2_FULL_HDR		0x0000000000000001ULL
+
+#define RBR_CFIG_A(IDX)			(DMC + 0x00010UL + (IDX) * 0x200UL)
+#define  RBR_CFIG_A_LEN			0xffff000000000000ULL
+#define  RBR_CFIG_A_LEN_SHIFT		48
+#define  RBR_CFIG_A_STADDR_BASE		0x00000ffffffc0000ULL
+#define  RBR_CFIG_A_STADDR		0x000000000003ffc0ULL
+
+#define RBR_CFIG_B(IDX)			(DMC + 0x00018UL + (IDX) * 0x200UL)
+#define  RBR_CFIG_B_BLKSIZE		0x0000000003000000ULL
+#define  RBR_CFIG_B_BLKSIZE_SHIFT	24
+#define  RBR_CFIG_B_VLD2		0x0000000000800000ULL
+#define  RBR_CFIG_B_BUFSZ2		0x0000000000030000ULL
+#define  RBR_CFIG_B_BUFSZ2_SHIFT	16
+#define  RBR_CFIG_B_VLD1		0x0000000000008000ULL
+#define  RBR_CFIG_B_BUFSZ1		0x0000000000000300ULL
+#define  RBR_CFIG_B_BUFSZ1_SHIFT	8
+#define  RBR_CFIG_B_VLD0		0x0000000000000080ULL
+#define  RBR_CFIG_B_BUFSZ0		0x0000000000000003ULL
+#define  RBR_CFIG_B_BUFSZ0_SHIFT	0
+
+#define RBR_BLKSIZE_4K			0x0
+#define RBR_BLKSIZE_8K			0x1
+#define RBR_BLKSIZE_16K			0x2
+#define RBR_BLKSIZE_32K			0x3
+#define RBR_BUFSZ2_2K			0x0
+#define RBR_BUFSZ2_4K			0x1
+#define RBR_BUFSZ2_8K			0x2
+#define RBR_BUFSZ2_16K			0x3
+#define RBR_BUFSZ1_1K			0x0
+#define RBR_BUFSZ1_2K			0x1
+#define RBR_BUFSZ1_4K			0x2
+#define RBR_BUFSZ1_8K			0x3
+#define RBR_BUFSZ0_256			0x0
+#define RBR_BUFSZ0_512			0x1
+#define RBR_BUFSZ0_1K			0x2
+#define RBR_BUFSZ0_2K			0x3
+
+#define RBR_KICK(IDX)			(DMC + 0x00020UL + (IDX) * 0x200UL)
+#define  RBR_KICK_BKADD			0x000000000000ffffULL
+
+#define RBR_STAT(IDX)			(DMC + 0x00028UL + (IDX) * 0x200UL)
+#define  RBR_STAT_QLEN			0x000000000000ffffULL
+
+#define RBR_HDH(IDX)			(DMC + 0x00030UL + (IDX) * 0x200UL)
+#define  RBR_HDH_HEAD_H			0x0000000000000fffULL
+
+#define RBR_HDL(IDX)			(DMC + 0x00038UL + (IDX) * 0x200UL)
+#define  RBR_HDL_HEAD_L			0x00000000fffffffcULL
+
+#define RCRCFIG_A(IDX)			(DMC + 0x00040UL + (IDX) * 0x200UL)
+#define  RCRCFIG_A_LEN			0xffff000000000000ULL
+#define  RCRCFIG_A_LEN_SHIFT		48
+#define  RCRCFIG_A_STADDR_BASE		0x00000ffffff80000ULL
+#define  RCRCFIG_A_STADDR		0x000000000007ffc0ULL
+
+#define RCRCFIG_B(IDX)			(DMC + 0x00048UL + (IDX) * 0x200UL)
+#define  RCRCFIG_B_PTHRES		0x00000000ffff0000ULL
+#define  RCRCFIG_B_PTHRES_SHIFT		16
+#define  RCRCFIG_B_ENTOUT		0x0000000000008000ULL
+#define  RCRCFIG_B_TIMEOUT		0x000000000000003fULL
+#define  RCRCFIG_B_TIMEOUT_SHIFT	0
+
+#define RCRSTAT_A(IDX)			(DMC + 0x00050UL + (IDX) * 0x200UL)
+#define  RCRSTAT_A_QLEN			0x000000000000ffffULL
+
+#define RCRSTAT_B(IDX)			(DMC + 0x00058UL + (IDX) * 0x200UL)
+#define  RCRSTAT_B_TIPTR_H		0x0000000000000fffULL
+
+#define RCRSTAT_C(IDX)			(DMC + 0x00060UL + (IDX) * 0x200UL)
+#define  RCRSTAT_C_TIPTR_L		0x00000000fffffff8ULL
+
+#define RX_DMA_CTL_STAT(IDX)		(DMC + 0x00070UL + (IDX) * 0x200UL)
+#define  RX_DMA_CTL_STAT_RBR_TMOUT	0x0020000000000000ULL
+#define  RX_DMA_CTL_STAT_RSP_CNT_ERR	0x0010000000000000ULL
+#define  RX_DMA_CTL_STAT_BYTE_EN_BUS	0x0008000000000000ULL
+#define  RX_DMA_CTL_STAT_RSP_DAT_ERR	0x0004000000000000ULL
+#define  RX_DMA_CTL_STAT_RCR_ACK_ERR	0x0002000000000000ULL
+#define  RX_DMA_CTL_STAT_DC_FIFO_ERR	0x0001000000000000ULL
+#define  RX_DMA_CTL_STAT_MEX		0x0000800000000000ULL
+#define  RX_DMA_CTL_STAT_RCRTHRES	0x0000400000000000ULL
+#define  RX_DMA_CTL_STAT_RCRTO		0x0000200000000000ULL
+#define  RX_DMA_CTL_STAT_RCR_SHA_PAR	0x0000100000000000ULL
+#define  RX_DMA_CTL_STAT_RBR_PRE_PAR	0x0000080000000000ULL
+#define  RX_DMA_CTL_STAT_PORT_DROP_PKT	0x0000040000000000ULL
+#define  RX_DMA_CTL_STAT_WRED_DROP	0x0000020000000000ULL
+#define  RX_DMA_CTL_STAT_RBR_PRE_EMTY	0x0000010000000000ULL
+#define  RX_DMA_CTL_STAT_RCRSHADOW_FULL	0x0000008000000000ULL
+#define  RX_DMA_CTL_STAT_CONFIG_ERR	0x0000004000000000ULL
+#define  RX_DMA_CTL_STAT_RCRINCON	0x0000002000000000ULL
+#define  RX_DMA_CTL_STAT_RCRFULL	0x0000001000000000ULL
+#define  RX_DMA_CTL_STAT_RBR_EMPTY	0x0000000800000000ULL
+#define  RX_DMA_CTL_STAT_RBRFULL	0x0000000400000000ULL
+#define  RX_DMA_CTL_STAT_RBRLOGPAGE	0x0000000200000000ULL
+#define  RX_DMA_CTL_STAT_CFIGLOGPAGE	0x0000000100000000ULL
+#define  RX_DMA_CTL_STAT_PTRREAD	0x00000000ffff0000ULL
+#define  RX_DMA_CTL_STAT_PTRREAD_SHIFT	16
+#define  RX_DMA_CTL_STAT_PKTREAD	0x000000000000ffffULL
+#define  RX_DMA_CTL_STAT_PKTREAD_SHIFT	0
+
+#define  RX_DMA_CTL_STAT_CHAN_FATAL	(RX_DMA_CTL_STAT_RBR_TMOUT | \
+					 RX_DMA_CTL_STAT_RSP_CNT_ERR | \
+					 RX_DMA_CTL_STAT_BYTE_EN_BUS | \
+					 RX_DMA_CTL_STAT_RSP_DAT_ERR | \
+					 RX_DMA_CTL_STAT_RCR_ACK_ERR | \
+					 RX_DMA_CTL_STAT_RCR_SHA_PAR | \
+					 RX_DMA_CTL_STAT_RBR_PRE_PAR | \
+					 RX_DMA_CTL_STAT_CONFIG_ERR | \
+					 RX_DMA_CTL_STAT_RCRINCON | \
+					 RX_DMA_CTL_STAT_RCRFULL | \
+					 RX_DMA_CTL_STAT_RBRFULL | \
+					 RX_DMA_CTL_STAT_RBRLOGPAGE | \
+					 RX_DMA_CTL_STAT_CFIGLOGPAGE)
+
+#define RX_DMA_CTL_STAT_PORT_FATAL	(RX_DMA_CTL_STAT_DC_FIFO_ERR)
+
+#define RX_DMA_CTL_WRITE_CLEAR_ERRS	(RX_DMA_CTL_STAT_RBR_EMPTY | \
+					 RX_DMA_CTL_STAT_RCRSHADOW_FULL | \
+					 RX_DMA_CTL_STAT_RBR_PRE_EMTY | \
+					 RX_DMA_CTL_STAT_WRED_DROP | \
+					 RX_DMA_CTL_STAT_PORT_DROP_PKT | \
+					 RX_DMA_CTL_STAT_RCRTO | \
+					 RX_DMA_CTL_STAT_RCRTHRES | \
+					 RX_DMA_CTL_STAT_DC_FIFO_ERR)
+
+#define RCR_FLSH(IDX)			(DMC + 0x00078UL + (IDX) * 0x200UL)
+#define  RCR_FLSH_FLSH			0x0000000000000001ULL
+
+#define RXMISC(IDX)			(DMC + 0x00090UL + (IDX) * 0x200UL)
+#define  RXMISC_OFLOW			0x0000000000010000ULL
+#define  RXMISC_COUNT			0x000000000000ffffULL
+
+#define RX_DMA_CTL_STAT_DBG(IDX)	(DMC + 0x00098UL + (IDX) * 0x200UL)
+#define  RX_DMA_CTL_STAT_DBG_RBR_TMOUT		0x0020000000000000ULL
+#define  RX_DMA_CTL_STAT_DBG_RSP_CNT_ERR	0x0010000000000000ULL
+#define  RX_DMA_CTL_STAT_DBG_BYTE_EN_BUS	0x0008000000000000ULL
+#define  RX_DMA_CTL_STAT_DBG_RSP_DAT_ERR	0x0004000000000000ULL
+#define  RX_DMA_CTL_STAT_DBG_RCR_ACK_ERR	0x0002000000000000ULL
+#define  RX_DMA_CTL_STAT_DBG_DC_FIFO_ERR	0x0001000000000000ULL
+#define  RX_DMA_CTL_STAT_DBG_MEX		0x0000800000000000ULL
+#define  RX_DMA_CTL_STAT_DBG_RCRTHRES		0x0000400000000000ULL
+#define  RX_DMA_CTL_STAT_DBG_RCRTO		0x0000200000000000ULL
+#define  RX_DMA_CTL_STAT_DBG_RCR_SHA_PAR	0x0000100000000000ULL
+#define  RX_DMA_CTL_STAT_DBG_RBR_PRE_PAR	0x0000080000000000ULL
+#define  RX_DMA_CTL_STAT_DBG_PORT_DROP_PKT	0x0000040000000000ULL
+#define  RX_DMA_CTL_STAT_DBG_WRED_DROP		0x0000020000000000ULL
+#define  RX_DMA_CTL_STAT_DBG_RBR_PRE_EMTY	0x0000010000000000ULL
+#define  RX_DMA_CTL_STAT_DBG_RCRSHADOW_FULL	0x0000008000000000ULL
+#define  RX_DMA_CTL_STAT_DBG_CONFIG_ERR		0x0000004000000000ULL
+#define  RX_DMA_CTL_STAT_DBG_RCRINCON		0x0000002000000000ULL
+#define  RX_DMA_CTL_STAT_DBG_RCRFULL		0x0000001000000000ULL
+#define  RX_DMA_CTL_STAT_DBG_RBR_EMPTY		0x0000000800000000ULL
+#define  RX_DMA_CTL_STAT_DBG_RBRFULL		0x0000000400000000ULL
+#define  RX_DMA_CTL_STAT_DBG_RBRLOGPAGE		0x0000000200000000ULL
+#define  RX_DMA_CTL_STAT_DBG_CFIGLOGPAGE	0x0000000100000000ULL
+#define  RX_DMA_CTL_STAT_DBG_PTRREAD		0x00000000ffff0000ULL
+#define  RX_DMA_CTL_STAT_DBG_PKTREAD		0x000000000000ffffULL
+
+#define RX_DMA_ENT_MSK(IDX)		(DMC + 0x00068UL + (IDX) * 0x200UL)
+#define  RX_DMA_ENT_MSK_RBR_TMOUT	0x0000000000200000ULL
+#define  RX_DMA_ENT_MSK_RSP_CNT_ERR	0x0000000000100000ULL
+#define  RX_DMA_ENT_MSK_BYTE_EN_BUS	0x0000000000080000ULL
+#define  RX_DMA_ENT_MSK_RSP_DAT_ERR	0x0000000000040000ULL
+#define  RX_DMA_ENT_MSK_RCR_ACK_ERR	0x0000000000020000ULL
+#define  RX_DMA_ENT_MSK_DC_FIFO_ERR	0x0000000000010000ULL
+#define  RX_DMA_ENT_MSK_RCRTHRES	0x0000000000004000ULL
+#define  RX_DMA_ENT_MSK_RCRTO		0x0000000000002000ULL
+#define  RX_DMA_ENT_MSK_RCR_SHA_PAR	0x0000000000001000ULL
+#define  RX_DMA_ENT_MSK_RBR_PRE_PAR	0x0000000000000800ULL
+#define  RX_DMA_ENT_MSK_PORT_DROP_PKT	0x0000000000000400ULL
+#define  RX_DMA_ENT_MSK_WRED_DROP	0x0000000000000200ULL
+#define  RX_DMA_ENT_MSK_RBR_PRE_EMTY	0x0000000000000100ULL
+#define  RX_DMA_ENT_MSK_RCR_SHADOW_FULL	0x0000000000000080ULL
+#define  RX_DMA_ENT_MSK_CONFIG_ERR	0x0000000000000040ULL
+#define  RX_DMA_ENT_MSK_RCRINCON	0x0000000000000020ULL
+#define  RX_DMA_ENT_MSK_RCRFULL		0x0000000000000010ULL
+#define  RX_DMA_ENT_MSK_RBR_EMPTY	0x0000000000000008ULL
+#define  RX_DMA_ENT_MSK_RBRFULL		0x0000000000000004ULL
+#define  RX_DMA_ENT_MSK_RBRLOGPAGE	0x0000000000000002ULL
+#define  RX_DMA_ENT_MSK_CFIGLOGPAGE	0x0000000000000001ULL
+#define  RX_DMA_ENT_MSK_ALL		0x00000000003f7fffULL
+
+#define TX_RNG_CFIG(IDX)		(DMC + 0x40000UL + (IDX) * 0x200UL)
+#define  TX_RNG_CFIG_LEN		0x1fff000000000000ULL
+#define  TX_RNG_CFIG_LEN_SHIFT		48
+#define  TX_RNG_CFIG_STADDR_BASE	0x00000ffffff80000ULL
+#define  TX_RNG_CFIG_STADDR		0x000000000007ffc0ULL
+
+#define TX_RING_HDL(IDX)		(DMC + 0x40010UL + (IDX) * 0x200UL)
+#define  TX_RING_HDL_WRAP		0x0000000000080000ULL
+#define  TX_RING_HDL_HEAD		0x000000000007fff8ULL
+#define  TX_RING_HDL_HEAD_SHIFT		3
+
+#define TX_RING_KICK(IDX)		(DMC + 0x40018UL + (IDX) * 0x200UL)
+#define  TX_RING_KICK_WRAP		0x0000000000080000ULL
+#define  TX_RING_KICK_TAIL		0x000000000007fff8ULL
+
+#define TX_ENT_MSK(IDX)			(DMC + 0x40020UL + (IDX) * 0x200UL)
+#define  TX_ENT_MSK_MK			0x0000000000008000ULL
+#define  TX_ENT_MSK_MBOX_ERR		0x0000000000000080ULL
+#define  TX_ENT_MSK_PKT_SIZE_ERR	0x0000000000000040ULL
+#define  TX_ENT_MSK_TX_RING_OFLOW	0x0000000000000020ULL
+#define  TX_ENT_MSK_PREF_BUF_ECC_ERR	0x0000000000000010ULL
+#define  TX_ENT_MSK_NACK_PREF		0x0000000000000008ULL
+#define  TX_ENT_MSK_NACK_PKT_RD		0x0000000000000004ULL
+#define  TX_ENT_MSK_CONF_PART_ERR	0x0000000000000002ULL
+#define  TX_ENT_MSK_PKT_PRT_ERR		0x0000000000000001ULL
+
+#define TX_CS(IDX)			(DMC + 0x40028UL + (IDX)*0x200UL)
+#define  TX_CS_PKT_CNT			0x0fff000000000000ULL
+#define  TX_CS_PKT_CNT_SHIFT		48
+#define  TX_CS_LASTMARK			0x00000fff00000000ULL
+#define  TX_CS_LASTMARK_SHIFT		32
+#define  TX_CS_RST			0x0000000080000000ULL
+#define  TX_CS_RST_STATE		0x0000000040000000ULL
+#define  TX_CS_MB			0x0000000020000000ULL
+#define  TX_CS_STOP_N_GO		0x0000000010000000ULL
+#define  TX_CS_SNG_STATE		0x0000000008000000ULL
+#define  TX_CS_MK			0x0000000000008000ULL
+#define  TX_CS_MMK			0x0000000000004000ULL
+#define  TX_CS_MBOX_ERR			0x0000000000000080ULL
+#define  TX_CS_PKT_SIZE_ERR		0x0000000000000040ULL
+#define  TX_CS_TX_RING_OFLOW		0x0000000000000020ULL
+#define  TX_CS_PREF_BUF_PAR_ERR		0x0000000000000010ULL
+#define  TX_CS_NACK_PREF		0x0000000000000008ULL
+#define  TX_CS_NACK_PKT_RD		0x0000000000000004ULL
+#define  TX_CS_CONF_PART_ERR		0x0000000000000002ULL
+#define  TX_CS_PKT_PRT_ERR		0x0000000000000001ULL
+
+#define TXDMA_MBH(IDX)			(DMC + 0x40030UL + (IDX) * 0x200UL)
+#define  TXDMA_MBH_MBADDR		0x0000000000000fffULL
+
+#define TXDMA_MBL(IDX)			(DMC + 0x40038UL + (IDX) * 0x200UL)
+#define  TXDMA_MBL_MBADDR		0x00000000ffffffc0ULL
+
+#define TX_DMA_PRE_ST(IDX)		(DMC + 0x40040UL + (IDX) * 0x200UL)
+#define  TX_DMA_PRE_ST_SHADOW_HD	0x000000000007ffffULL
+
+#define TX_RNG_ERR_LOGH(IDX)		(DMC + 0x40048UL + (IDX) * 0x200UL)
+#define  TX_RNG_ERR_LOGH_ERR		0x0000000080000000ULL
+#define  TX_RNG_ERR_LOGH_MERR		0x0000000040000000ULL
+#define  TX_RNG_ERR_LOGH_ERRCODE	0x0000000038000000ULL
+#define  TX_RNG_ERR_LOGH_ERRADDR	0x0000000000000fffULL
+
+#define TX_RNG_ERR_LOGL(IDX)		(DMC + 0x40050UL + (IDX) * 0x200UL)
+#define  TX_RNG_ERR_LOGL_ERRADDR	0x00000000ffffffffULL
+
+#define TDMC_INTR_DBG(IDX)		(DMC + 0x40060UL + (IDX) * 0x200UL)
+#define  TDMC_INTR_DBG_MK		0x0000000000008000ULL
+#define  TDMC_INTR_DBG_MBOX_ERR		0x0000000000000080ULL
+#define  TDMC_INTR_DBG_PKT_SIZE_ERR	0x0000000000000040ULL
+#define  TDMC_INTR_DBG_TX_RING_OFLOW	0x0000000000000020ULL
+#define  TDMC_INTR_DBG_PREF_BUF_PAR_ERR	0x0000000000000010ULL
+#define  TDMC_INTR_DBG_NACK_PREF	0x0000000000000008ULL
+#define  TDMC_INTR_DBG_NACK_PKT_RD	0x0000000000000004ULL
+#define  TDMC_INTR_DBG_CONF_PART_ERR	0x0000000000000002ULL
+#define  TDMC_INTR_DBG_PKT_PART_ERR	0x0000000000000001ULL
+
+#define TX_CS_DBG(IDX)			(DMC + 0x40068UL + (IDX) * 0x200UL)
+#define  TX_CS_DBG_PKT_CNT		0x0fff000000000000ULL
+
+#define TDMC_INJ_PAR_ERR(IDX)		(DMC + 0x45040UL + (IDX) * 0x200UL)
+#define  TDMC_INJ_PAR_ERR_VAL		0x000000000000ffffULL
+
+#define TDMC_DBG_SEL(IDX)		(DMC + 0x45080UL + (IDX) * 0x200UL)
+#define  TDMC_DBG_SEL_DBG_SEL		0x000000000000003fULL
+
+#define TDMC_TRAINING_VECTOR(IDX)	(DMC + 0x45088UL + (IDX) * 0x200UL)
+#define  TDMC_TRAINING_VECTOR_VEC	0x00000000ffffffffULL
+
+#define TXC_DMA_MAX(CHAN)		(FZC_TXC + 0x00000UL + (CHAN)*0x1000UL)
+#define TXC_DMA_MAX_LEN(CHAN)		(FZC_TXC + 0x00008UL + (CHAN)*0x1000UL)
+
+#define TXC_CONTROL			(FZC_TXC + 0x20000UL)
+#define  TXC_CONTROL_ENABLE		0x0000000000000010ULL
+#define  TXC_CONTROL_PORT_ENABLE(X)	(1 << (X))
+
+#define TXC_TRAINING_VEC		(FZC_TXC + 0x20008UL)
+#define  TXC_TRAINING_VEC_MASK		0x00000000ffffffffULL
+
+#define TXC_DEBUG			(FZC_TXC + 0x20010UL)
+#define  TXC_DEBUG_SELECT		0x000000000000003fULL
+
+#define TXC_MAX_REORDER			(FZC_TXC + 0x20018UL)
+#define  TXC_MAX_REORDER_PORT3		0x000000000f000000ULL
+#define  TXC_MAX_REORDER_PORT2		0x00000000000f0000ULL
+#define  TXC_MAX_REORDER_PORT1		0x0000000000000f00ULL
+#define  TXC_MAX_REORDER_PORT0		0x000000000000000fULL
+
+#define TXC_PORT_CTL(PORT)		(FZC_TXC + 0x20020UL + (PORT)*0x100UL)
+#define  TXC_PORT_CTL_CLR_ALL_STAT	0x0000000000000001ULL
+
+#define TXC_PKT_STUFFED(PORT)		(FZC_TXC + 0x20030UL + (PORT)*0x100UL)
+#define  TXC_PKT_STUFFED_PP_REORDER	0x00000000ffff0000ULL
+#define  TXC_PKT_STUFFED_PP_PACKETASSY	0x000000000000ffffULL
+
+#define TXC_PKT_XMIT(PORT)		(FZC_TXC + 0x20038UL + (PORT)*0x100UL)
+#define  TXC_PKT_XMIT_BYTES		0x00000000ffff0000ULL
+#define  TXC_PKT_XMIT_PKTS		0x000000000000ffffULL
+
+#define TXC_ROECC_CTL(PORT)		(FZC_TXC + 0x20040UL + (PORT)*0x100UL)
+#define  TXC_ROECC_CTL_DISABLE_UE	0x0000000080000000ULL
+#define  TXC_ROECC_CTL_DBL_BIT_ERR	0x0000000000020000ULL
+#define  TXC_ROECC_CTL_SNGL_BIT_ERR	0x0000000000010000ULL
+#define  TXC_ROECC_CTL_ALL_PKTS		0x0000000000000400ULL
+#define  TXC_ROECC_CTL_ALT_PKTS		0x0000000000000200ULL
+#define  TXC_ROECC_CTL_ONE_PKT_ONLY	0x0000000000000100ULL
+#define  TXC_ROECC_CTL_LST_PKT_LINE	0x0000000000000004ULL
+#define  TXC_ROECC_CTL_2ND_PKT_LINE	0x0000000000000002ULL
+#define  TXC_ROECC_CTL_1ST_PKT_LINE	0x0000000000000001ULL
+
+#define TXC_ROECC_ST(PORT)		(FZC_TXC + 0x20048UL + (PORT)*0x100UL)
+#define  TXC_ROECC_CLR_ST		0x0000000080000000ULL
+#define  TXC_ROECC_CE			0x0000000000020000ULL
+#define  TXC_ROECC_UE			0x0000000000010000ULL
+#define  TXC_ROECC_ST_ECC_ADDR		0x00000000000003ffULL
+
+#define TXC_RO_DATA0(PORT)		(FZC_TXC + 0x20050UL + (PORT)*0x100UL)
+#define  TXC_RO_DATA0_DATA0		0x00000000ffffffffULL /* bits 31:0 */
+
+#define TXC_RO_DATA1(PORT)		(FZC_TXC + 0x20058UL + (PORT)*0x100UL)
+#define  TXC_RO_DATA1_DATA1		0x00000000ffffffffULL /* bits 63:32 */
+
+#define TXC_RO_DATA2(PORT)		(FZC_TXC + 0x20060UL + (PORT)*0x100UL)
+#define  TXC_RO_DATA2_DATA2		0x00000000ffffffffULL /* bits 95:64 */
+
+#define TXC_RO_DATA3(PORT)		(FZC_TXC + 0x20068UL + (PORT)*0x100UL)
+#define  TXC_RO_DATA3_DATA3		0x00000000ffffffffULL /* bits 127:96 */
+
+#define TXC_RO_DATA4(PORT)		(FZC_TXC + 0x20070UL + (PORT)*0x100UL)
+#define  TXC_RO_DATA4_DATA4		0x0000000000ffffffULL /* bits 151:128 */
+
+#define TXC_SFECC_CTL(PORT)		(FZC_TXC + 0x20078UL + (PORT)*0x100UL)
+#define  TXC_SFECC_CTL_DISABLE_UE	0x0000000080000000ULL
+#define  TXC_SFECC_CTL_DBL_BIT_ERR	0x0000000000020000ULL
+#define  TXC_SFECC_CTL_SNGL_BIT_ERR	0x0000000000010000ULL
+#define  TXC_SFECC_CTL_ALL_PKTS		0x0000000000000400ULL
+#define  TXC_SFECC_CTL_ALT_PKTS		0x0000000000000200ULL
+#define  TXC_SFECC_CTL_ONE_PKT_ONLY	0x0000000000000100ULL
+#define  TXC_SFECC_CTL_LST_PKT_LINE	0x0000000000000004ULL
+#define  TXC_SFECC_CTL_2ND_PKT_LINE	0x0000000000000002ULL
+#define  TXC_SFECC_CTL_1ST_PKT_LINE	0x0000000000000001ULL
+
+#define TXC_SFECC_ST(PORT)		(FZC_TXC + 0x20080UL + (PORT)*0x100UL)
+#define  TXC_SFECC_ST_CLR_ST		0x0000000080000000ULL
+#define  TXC_SFECC_ST_CE		0x0000000000020000ULL
+#define  TXC_SFECC_ST_UE		0x0000000000010000ULL
+#define  TXC_SFECC_ST_ECC_ADDR		0x00000000000003ffULL
+
+#define TXC_SF_DATA0(PORT)		(FZC_TXC + 0x20088UL + (PORT)*0x100UL)
+#define  TXC_SF_DATA0_DATA0		0x00000000ffffffffULL /* bits 31:0 */
+
+#define TXC_SF_DATA1(PORT)		(FZC_TXC + 0x20090UL + (PORT)*0x100UL)
+#define  TXC_SF_DATA1_DATA1		0x00000000ffffffffULL /* bits 63:32 */
+
+#define TXC_SF_DATA2(PORT)		(FZC_TXC + 0x20098UL + (PORT)*0x100UL)
+#define  TXC_SF_DATA2_DATA2		0x00000000ffffffffULL /* bits 95:64 */
+
+#define TXC_SF_DATA3(PORT)		(FZC_TXC + 0x200a0UL + (PORT)*0x100UL)
+#define  TXC_SF_DATA3_DATA3		0x00000000ffffffffULL /* bits 127:96 */
+
+#define TXC_SF_DATA4(PORT)		(FZC_TXC + 0x200a8UL + (PORT)*0x100UL)
+#define  TXC_SF_DATA4_DATA4		0x0000000000ffffffULL /* bits 151:128 */
+
+#define TXC_RO_TIDS(PORT)		(FZC_TXC + 0x200b0UL + (PORT)*0x100UL)
+#define  TXC_RO_TIDS_IN_USE		0x00000000ffffffffULL
+
+#define TXC_RO_STATE0(PORT)		(FZC_TXC + 0x200b8UL + (PORT)*0x100UL)
+#define  TXC_RO_STATE0_DUPLICATE_TID	0x00000000ffffffffULL
+
+#define TXC_RO_STATE1(PORT)		(FZC_TXC + 0x200c0UL + (PORT)*0x100UL)
+#define  TXC_RO_STATE1_UNUSED_TID	0x00000000ffffffffULL
+
+#define TXC_RO_STATE2(PORT)		(FZC_TXC + 0x200c8UL + (PORT)*0x100UL)
+#define  TXC_RO_STATE2_TRANS_TIMEOUT	0x00000000ffffffffULL
+
+#define TXC_RO_STATE3(PORT)		(FZC_TXC + 0x200d0UL + (PORT)*0x100UL)
+#define  TXC_RO_STATE3_ENAB_SPC_WMARK	0x0000000080000000ULL
+#define  TXC_RO_STATE3_RO_SPC_WMARK	0x000000007fe00000ULL
+#define  TXC_RO_STATE3_ROFIFO_SPC_AVAIL	0x00000000001ff800ULL
+#define  TXC_RO_STATE3_ENAB_RO_WMARK	0x0000000000000100ULL
+#define  TXC_RO_STATE3_HIGH_RO_USED	0x00000000000000f0ULL
+#define  TXC_RO_STATE3_NUM_RO_USED	0x000000000000000fULL
+
+#define TXC_RO_CTL(PORT)		(FZC_TXC + 0x200d8UL + (PORT)*0x100UL)
+#define  TXC_RO_CTL_CLR_FAIL_STATE	0x0000000080000000ULL
+#define  TXC_RO_CTL_RO_ADDR		0x000000000f000000ULL
+#define  TXC_RO_CTL_ADDR_FAILED		0x0000000000400000ULL
+#define  TXC_RO_CTL_DMA_FAILED		0x0000000000200000ULL
+#define  TXC_RO_CTL_LEN_FAILED		0x0000000000100000ULL
+#define  TXC_RO_CTL_CAPT_ADDR_FAILED	0x0000000000040000ULL
+#define  TXC_RO_CTL_CAPT_DMA_FAILED	0x0000000000020000ULL
+#define  TXC_RO_CTL_CAPT_LEN_FAILED	0x0000000000010000ULL
+#define  TXC_RO_CTL_RO_STATE_RD_DONE	0x0000000000000080ULL
+#define  TXC_RO_CTL_RO_STATE_WR_DONE	0x0000000000000040ULL
+#define  TXC_RO_CTL_RO_STATE_RD		0x0000000000000020ULL
+#define  TXC_RO_CTL_RO_STATE_WR		0x0000000000000010ULL
+#define  TXC_RO_CTL_RO_STATE_ADDR	0x000000000000000fULL
+
+#define TXC_RO_ST_DATA0(PORT)		(FZC_TXC + 0x200e0UL + (PORT)*0x100UL)
+#define  TXC_RO_ST_DATA0_DATA0		0x00000000ffffffffULL
+
+#define TXC_RO_ST_DATA1(PORT)		(FZC_TXC + 0x200e8UL + (PORT)*0x100UL)
+#define  TXC_RO_ST_DATA1_DATA1		0x00000000ffffffffULL
+
+#define TXC_RO_ST_DATA2(PORT)		(FZC_TXC + 0x200f0UL + (PORT)*0x100UL)
+#define  TXC_RO_ST_DATA2_DATA2		0x00000000ffffffffULL
+
+#define TXC_RO_ST_DATA3(PORT)		(FZC_TXC + 0x200f8UL + (PORT)*0x100UL)
+#define  TXC_RO_ST_DATA3_DATA3		0x00000000ffffffffULL
+
+#define TXC_PORT_PACKET_REQ(PORT)	(FZC_TXC + 0x20100UL + (PORT)*0x100UL)
+#define  TXC_PORT_PACKET_REQ_GATHER_REQ	0x00000000f0000000ULL
+#define  TXC_PORT_PACKET_REQ_PKT_REQ	0x000000000fff0000ULL
+#define  TXC_PORT_PACKET_REQ_PERR_ABRT	0x000000000000ffffULL
+
+	/* bits are same as TXC_INT_STAT */
+#define TXC_INT_STAT_DBG		(FZC_TXC + 0x20420UL)
+
+#define TXC_INT_STAT			(FZC_TXC + 0x20428UL)
+#define  TXC_INT_STAT_VAL_SHIFT(PORT)	((PORT) * 8)
+#define  TXC_INT_STAT_VAL(PORT)		(0x3f << TXC_INT_STAT_VAL_SHIFT(PORT))
+#define  TXC_INT_STAT_SF_CE(PORT)	(0x01 << TXC_INT_STAT_VAL_SHIFT(PORT))
+#define  TXC_INT_STAT_SF_UE(PORT)	(0x02 << TXC_INT_STAT_VAL_SHIFT(PORT))
+#define  TXC_INT_STAT_RO_CE(PORT)	(0x04 << TXC_INT_STAT_VAL_SHIFT(PORT))
+#define  TXC_INT_STAT_RO_UE(PORT)	(0x08 << TXC_INT_STAT_VAL_SHIFT(PORT))
+#define  TXC_INT_STAT_REORDER_ERR(PORT)	(0x10 << TXC_INT_STAT_VAL_SHIFT(PORT))
+#define  TXC_INT_STAT_PKTASM_DEAD(PORT)	(0x20 << TXC_INT_STAT_VAL_SHIFT(PORT))
+
+#define TXC_INT_MASK			(FZC_TXC + 0x20430UL)
+#define  TXC_INT_MASK_VAL_SHIFT(PORT)	((PORT) * 8)
+#define  TXC_INT_MASK_VAL(PORT)		(0x3f << TXC_INT_STAT_VAL_SHIFT(PORT))
+
+#define TXC_INT_MASK_SF_CE		0x01
+#define TXC_INT_MASK_SF_UE		0x02
+#define TXC_INT_MASK_RO_CE		0x04
+#define TXC_INT_MASK_RO_UE		0x08
+#define TXC_INT_MASK_REORDER_ERR	0x10
+#define TXC_INT_MASK_PKTASM_DEAD	0x20
+#define TXC_INT_MASK_ALL		0x3f
+
+#define TXC_PORT_DMA(IDX)		(FZC_TXC + 0x20028UL + (IDX)*0x100UL)
+
+#define ESPC_PIO_EN			(FZC_PROM + 0x40000UL)
+#define  ESPC_PIO_EN_ENABLE		0x0000000000000001ULL
+
+#define ESPC_PIO_STAT			(FZC_PROM + 0x40008UL)
+#define  ESPC_PIO_STAT_READ_START	0x0000000080000000ULL
+#define  ESPC_PIO_STAT_READ_END		0x0000000040000000ULL
+#define  ESPC_PIO_STAT_WRITE_INIT	0x0000000020000000ULL
+#define  ESPC_PIO_STAT_WRITE_END	0x0000000010000000ULL
+#define  ESPC_PIO_STAT_ADDR		0x0000000003ffff00ULL
+#define  ESPC_PIO_STAT_ADDR_SHIFT	8
+#define  ESPC_PIO_STAT_DATA		0x00000000000000ffULL
+#define  ESPC_PIO_STAT_DATA_SHIFT	0
+
+#define ESPC_NCR(IDX)			(FZC_PROM + 0x40020UL + (IDX)*0x8UL)
+#define  ESPC_NCR_VAL			0x00000000ffffffffULL
+
+#define ESPC_MAC_ADDR0			ESPC_NCR(0)
+#define ESPC_MAC_ADDR1			ESPC_NCR(1)
+#define ESPC_NUM_PORTS_MACS		ESPC_NCR(2)
+#define  ESPC_NUM_PORTS_MACS_VAL	0x00000000000000ffULL
+#define ESPC_MOD_STR_LEN		ESPC_NCR(4)
+#define ESPC_MOD_STR_1			ESPC_NCR(5)
+#define ESPC_MOD_STR_2			ESPC_NCR(6)
+#define ESPC_MOD_STR_3			ESPC_NCR(7)
+#define ESPC_MOD_STR_4			ESPC_NCR(8)
+#define ESPC_MOD_STR_5			ESPC_NCR(9)
+#define ESPC_MOD_STR_6			ESPC_NCR(10)
+#define ESPC_MOD_STR_7			ESPC_NCR(11)
+#define ESPC_MOD_STR_8			ESPC_NCR(12)
+#define ESPC_BD_MOD_STR_LEN		ESPC_NCR(13)
+#define ESPC_BD_MOD_STR_1		ESPC_NCR(14)
+#define ESPC_BD_MOD_STR_2		ESPC_NCR(15)
+#define ESPC_BD_MOD_STR_3		ESPC_NCR(16)
+#define ESPC_BD_MOD_STR_4		ESPC_NCR(17)
+
+#define ESPC_PHY_TYPE			ESPC_NCR(18)
+#define  ESPC_PHY_TYPE_PORT0		0x00000000ff000000ULL
+#define  ESPC_PHY_TYPE_PORT0_SHIFT	24
+#define  ESPC_PHY_TYPE_PORT1		0x0000000000ff0000ULL
+#define  ESPC_PHY_TYPE_PORT1_SHIFT	16
+#define  ESPC_PHY_TYPE_PORT2		0x000000000000ff00ULL
+#define  ESPC_PHY_TYPE_PORT2_SHIFT	8
+#define  ESPC_PHY_TYPE_PORT3		0x00000000000000ffULL
+#define  ESPC_PHY_TYPE_PORT3_SHIFT	0
+
+#define  ESPC_PHY_TYPE_1G_COPPER	3
+#define  ESPC_PHY_TYPE_1G_FIBER		2
+#define  ESPC_PHY_TYPE_10G_COPPER	1
+#define  ESPC_PHY_TYPE_10G_FIBER	0
+
+#define ESPC_MAX_FM_SZ			ESPC_NCR(19)
+
+#define ESPC_INTR_NUM			ESPC_NCR(20)
+#define  ESPC_INTR_NUM_PORT0		0x00000000ff000000ULL
+#define  ESPC_INTR_NUM_PORT1		0x0000000000ff0000ULL
+#define  ESPC_INTR_NUM_PORT2		0x000000000000ff00ULL
+#define  ESPC_INTR_NUM_PORT3		0x00000000000000ffULL
+
+#define ESPC_VER_IMGSZ			ESPC_NCR(21)
+#define  ESPC_VER_IMGSZ_IMGSZ		0x00000000ffff0000ULL
+#define  ESPC_VER_IMGSZ_IMGSZ_SHIFT	16
+#define  ESPC_VER_IMGSZ_VER		0x000000000000ffffULL
+#define  ESPC_VER_IMGSZ_VER_SHIFT	0
+
+#define ESPC_CHKSUM			ESPC_NCR(22)
+#define  ESPC_CHKSUM_SUM		0x00000000000000ffULL
+
+#define ESPC_EEPROM_SIZE		0x100000
+
+#define CLASS_CODE_UNRECOG		0x00
+#define CLASS_CODE_DUMMY1		0x01
+#define CLASS_CODE_ETHERTYPE1		0x02
+#define CLASS_CODE_ETHERTYPE2		0x03
+#define CLASS_CODE_USER_PROG1		0x04
+#define CLASS_CODE_USER_PROG2		0x05
+#define CLASS_CODE_USER_PROG3		0x06
+#define CLASS_CODE_USER_PROG4		0x07
+#define CLASS_CODE_TCP_IPV4		0x08
+#define CLASS_CODE_UDP_IPV4		0x09
+#define CLASS_CODE_AH_ESP_IPV4		0x0a
+#define CLASS_CODE_SCTP_IPV4		0x0b
+#define CLASS_CODE_TCP_IPV6		0x0c
+#define CLASS_CODE_UDP_IPV6		0x0d
+#define CLASS_CODE_AH_ESP_IPV6		0x0e
+#define CLASS_CODE_SCTP_IPV6		0x0f
+#define CLASS_CODE_ARP			0x10
+#define CLASS_CODE_RARP			0x11
+#define CLASS_CODE_DUMMY2		0x12
+#define CLASS_CODE_DUMMY3		0x13
+#define CLASS_CODE_DUMMY4		0x14
+#define CLASS_CODE_DUMMY5		0x15
+#define CLASS_CODE_DUMMY6		0x16
+#define CLASS_CODE_DUMMY7		0x17
+#define CLASS_CODE_DUMMY8		0x18
+#define CLASS_CODE_DUMMY9		0x19
+#define CLASS_CODE_DUMMY10		0x1a
+#define CLASS_CODE_DUMMY11		0x1b
+#define CLASS_CODE_DUMMY12		0x1c
+#define CLASS_CODE_DUMMY13		0x1d
+#define CLASS_CODE_DUMMY14		0x1e
+#define CLASS_CODE_DUMMY15		0x1f
+
+/* Logical devices and device groups */
+#define LDN_RXDMA(CHAN)			(0 + (CHAN))
+#define LDN_RESV1(OFF)			(16 + (OFF))
+#define LDN_TXDMA(CHAN)			(32 + (CHAN))
+#define LDN_RESV2(OFF)			(56 + (OFF))
+#define LDN_MIF				63
+#define LDN_MAC(PORT)			(64 + (PORT))
+#define LDN_DEVICE_ERROR		68
+#define LDN_MAX				LDN_DEVICE_ERROR
+
+#define NIU_LDG_MIN			0
+#define NIU_LDG_MAX			63
+#define NIU_NUM_LDG			64
+#define LDG_INVALID			0xff
+
+/* PHY stuff */
+#define NIU_PMA_PMD_DEV_ADDR		1
+#define NIU_PCS_DEV_ADDR		3
+
+#define NIU_PHY_ID_MASK			0xfffff0f0
+#define NIU_PHY_ID_BCM8704		0x00206030
+#define NIU_PHY_ID_BCM8706		0x00206035
+#define NIU_PHY_ID_BCM5464R		0x002060b0
+#define NIU_PHY_ID_MRVL88X2011		0x01410020
+
+/* MRVL88X2011 register addresses */
+#define MRVL88X2011_USER_DEV1_ADDR	1
+#define MRVL88X2011_USER_DEV2_ADDR	2
+#define MRVL88X2011_USER_DEV3_ADDR	3
+#define MRVL88X2011_USER_DEV4_ADDR	4
+#define MRVL88X2011_PMA_PMD_CTL_1	0x0000
+#define MRVL88X2011_PMA_PMD_STATUS_1	0x0001
+#define MRVL88X2011_10G_PMD_STATUS_2	0x0008
+#define MRVL88X2011_10G_PMD_TX_DIS	0x0009
+#define MRVL88X2011_10G_XGXS_LANE_STAT	0x0018
+#define MRVL88X2011_GENERAL_CTL		0x8300
+#define MRVL88X2011_LED_BLINK_CTL	0x8303
+#define MRVL88X2011_LED_8_TO_11_CTL	0x8306
+
+/* MRVL88X2011 register control */
+#define MRVL88X2011_ENA_XFPREFCLK	0x0001
+#define MRVL88X2011_ENA_PMDTX		0x0000
+#define MRVL88X2011_LOOPBACK            0x1
+#define MRVL88X2011_LED_ACT		0x1
+#define MRVL88X2011_LNK_STATUS_OK	0x4
+#define MRVL88X2011_LED_BLKRATE_MASK	0x70
+#define MRVL88X2011_LED_BLKRATE_034MS	0x0
+#define MRVL88X2011_LED_BLKRATE_067MS	0x1
+#define MRVL88X2011_LED_BLKRATE_134MS	0x2
+#define MRVL88X2011_LED_BLKRATE_269MS	0x3
+#define MRVL88X2011_LED_BLKRATE_538MS	0x4
+#define MRVL88X2011_LED_CTL_OFF		0x0
+#define MRVL88X2011_LED_CTL_PCS_ACT	0x5
+#define MRVL88X2011_LED_CTL_MASK	0x7
+#define MRVL88X2011_LED(n,v)		((v)<<((n)*4))
+#define MRVL88X2011_LED_STAT(n,v)	((v)>>((n)*4))
+
+#define BCM8704_PMA_PMD_DEV_ADDR	1
+#define BCM8704_PCS_DEV_ADDR		2
+#define BCM8704_USER_DEV3_ADDR		3
+#define BCM8704_PHYXS_DEV_ADDR		4
+#define BCM8704_USER_DEV4_ADDR		4
+
+#define BCM8704_PMD_RCV_SIGDET		0x000a
+#define  PMD_RCV_SIGDET_LANE3		0x0010
+#define  PMD_RCV_SIGDET_LANE2		0x0008
+#define  PMD_RCV_SIGDET_LANE1		0x0004
+#define  PMD_RCV_SIGDET_LANE0		0x0002
+#define  PMD_RCV_SIGDET_GLOBAL		0x0001
+
+#define BCM8704_PCS_10G_R_STATUS	0x0020
+#define  PCS_10G_R_STATUS_LINKSTAT	0x1000
+#define  PCS_10G_R_STATUS_PRBS31_ABLE	0x0004
+#define  PCS_10G_R_STATUS_HI_BER	0x0002
+#define  PCS_10G_R_STATUS_BLK_LOCK	0x0001
+
+#define BCM8704_USER_CONTROL		0xc800
+#define  USER_CONTROL_OPTXENB_LVL	0x8000
+#define  USER_CONTROL_OPTXRST_LVL	0x4000
+#define  USER_CONTROL_OPBIASFLT_LVL	0x2000
+#define  USER_CONTROL_OBTMPFLT_LVL	0x1000
+#define  USER_CONTROL_OPPRFLT_LVL	0x0800
+#define  USER_CONTROL_OPTXFLT_LVL	0x0400
+#define  USER_CONTROL_OPRXLOS_LVL	0x0200
+#define  USER_CONTROL_OPRXFLT_LVL	0x0100
+#define  USER_CONTROL_OPTXON_LVL	0x0080
+#define  USER_CONTROL_RES1		0x007f
+#define  USER_CONTROL_RES1_SHIFT	0
+
+#define BCM8704_USER_ANALOG_CLK		0xc801
+#define BCM8704_USER_PMD_RX_CONTROL	0xc802
+
+#define BCM8704_USER_PMD_TX_CONTROL	0xc803
+#define  USER_PMD_TX_CTL_RES1		0xfe00
+#define  USER_PMD_TX_CTL_XFP_CLKEN	0x0100
+#define  USER_PMD_TX_CTL_TX_DAC_TXD	0x00c0
+#define  USER_PMD_TX_CTL_TX_DAC_TXD_SH	6
+#define  USER_PMD_TX_CTL_TX_DAC_TXCK	0x0030
+#define  USER_PMD_TX_CTL_TX_DAC_TXCK_SH	4
+#define  USER_PMD_TX_CTL_TSD_LPWREN	0x0008
+#define  USER_PMD_TX_CTL_TSCK_LPWREN	0x0004
+#define  USER_PMD_TX_CTL_CMU_LPWREN	0x0002
+#define  USER_PMD_TX_CTL_SFIFORST	0x0001
+
+#define BCM8704_USER_ANALOG_STATUS0	0xc804
+#define BCM8704_USER_OPT_DIGITAL_CTRL	0xc808
+#define BCM8704_USER_TX_ALARM_STATUS	0x9004
+
+#define  USER_ODIG_CTRL_FMODE		0x8000
+#define  USER_ODIG_CTRL_TX_PDOWN	0x4000
+#define  USER_ODIG_CTRL_RX_PDOWN	0x2000
+#define  USER_ODIG_CTRL_EFILT_EN	0x1000
+#define  USER_ODIG_CTRL_OPT_RST		0x0800
+#define  USER_ODIG_CTRL_PCS_TIB		0x0400
+#define  USER_ODIG_CTRL_PCS_RI		0x0200
+#define  USER_ODIG_CTRL_RESV1		0x0180
+#define  USER_ODIG_CTRL_GPIOS		0x0060
+#define  USER_ODIG_CTRL_GPIOS_SHIFT	5
+#define  USER_ODIG_CTRL_RESV2		0x0010
+#define  USER_ODIG_CTRL_LB_ERR_DIS	0x0008
+#define  USER_ODIG_CTRL_RESV3		0x0006
+#define  USER_ODIG_CTRL_TXONOFF_PD_DIS	0x0001
+
+#define BCM8704_PHYXS_XGXS_LANE_STAT	0x0018
+#define  PHYXS_XGXS_LANE_STAT_ALINGED	0x1000
+#define  PHYXS_XGXS_LANE_STAT_PATTEST	0x0800
+#define  PHYXS_XGXS_LANE_STAT_MAGIC	0x0400
+#define  PHYXS_XGXS_LANE_STAT_LANE3	0x0008
+#define  PHYXS_XGXS_LANE_STAT_LANE2	0x0004
+#define  PHYXS_XGXS_LANE_STAT_LANE1	0x0002
+#define  PHYXS_XGXS_LANE_STAT_LANE0	0x0001
+
+#define BCM5464R_AUX_CTL		24
+#define  BCM5464R_AUX_CTL_EXT_LB	0x8000
+#define  BCM5464R_AUX_CTL_EXT_PLEN	0x4000
+#define  BCM5464R_AUX_CTL_ER1000	0x3000
+#define  BCM5464R_AUX_CTL_ER1000_SHIFT	12
+#define  BCM5464R_AUX_CTL_RESV1		0x0800
+#define  BCM5464R_AUX_CTL_WRITE_1	0x0400
+#define  BCM5464R_AUX_CTL_RESV2		0x0300
+#define  BCM5464R_AUX_CTL_PRESP_DIS	0x0080
+#define  BCM5464R_AUX_CTL_RESV3		0x0040
+#define  BCM5464R_AUX_CTL_ER100		0x0030
+#define  BCM5464R_AUX_CTL_ER100_SHIFT	4
+#define  BCM5464R_AUX_CTL_DIAG_MODE	0x0008
+#define  BCM5464R_AUX_CTL_SR_SEL	0x0007
+#define  BCM5464R_AUX_CTL_SR_SEL_SHIFT	0
+
+#define  BCM5464R_CTRL1000_AS_MASTER		0x0800
+#define  BCM5464R_CTRL1000_ENABLE_AS_MASTER	0x1000
+
+#define RCR_ENTRY_MULTI			0x8000000000000000ULL
+#define RCR_ENTRY_PKT_TYPE		0x6000000000000000ULL
+#define RCR_ENTRY_PKT_TYPE_SHIFT	61
+#define RCR_ENTRY_ZERO_COPY		0x1000000000000000ULL
+#define RCR_ENTRY_NOPORT		0x0800000000000000ULL
+#define RCR_ENTRY_PROMISC		0x0400000000000000ULL
+#define RCR_ENTRY_ERROR			0x0380000000000000ULL
+#define RCR_ENTRY_DCF_ERR		0x0040000000000000ULL
+#define RCR_ENTRY_L2_LEN		0x003fff0000000000ULL
+#define RCR_ENTRY_L2_LEN_SHIFT		40
+#define RCR_ENTRY_PKTBUFSZ		0x000000c000000000ULL
+#define RCR_ENTRY_PKTBUFSZ_SHIFT	38
+#define RCR_ENTRY_PKT_BUF_ADDR		0x0000003fffffffffULL /* bits 43:6 */
+#define RCR_ENTRY_PKT_BUF_ADDR_SHIFT	6
+
+#define RCR_PKT_TYPE_OTHER		0x0
+#define RCR_PKT_TYPE_TCP		0x1
+#define RCR_PKT_TYPE_UDP		0x2
+#define RCR_PKT_TYPE_SCTP		0x3
+
+#define NIU_RXPULL_MAX			ETH_HLEN
+
+struct rx_pkt_hdr0 {
+#if defined(__LITTLE_ENDIAN_BITFIELD)
+	u8	inputport:2,
+		maccheck:1,
+		class:5;
+	u8	vlan:1,
+		llcsnap:1,
+		noport:1,
+		badip:1,
+		tcamhit:1,
+		tres:2,
+		tzfvld:1;
+#elif defined(__BIG_ENDIAN_BITFIELD)
+	u8	class:5,
+		maccheck:1,
+		inputport:2;
+	u8	tzfvld:1,
+		tres:2,
+		tcamhit:1,
+		badip:1,
+		noport:1,
+		llcsnap:1,
+		vlan:1;
+#endif
+};
+
+struct rx_pkt_hdr1 {
+	u8	hwrsvd1;
+	u8	tcammatch;
+#if defined(__LITTLE_ENDIAN_BITFIELD)
+	u8	hwrsvd2:2,
+		hashit:1,
+		exact:1,
+		hzfvld:1,
+		hashsidx:3;
+#elif defined(__BIG_ENDIAN_BITFIELD)
+	u8	hashsidx:3,
+		hzfvld:1,
+		exact:1,
+		hashit:1,
+		hwrsvd2:2;
+#endif
+	u8	zcrsvd;
+
+	/* Bits 11:8 of zero copy flow ID.  */
+#if defined(__LITTLE_ENDIAN_BITFIELD)
+	u8	hwrsvd3:4, zflowid0:4;
+#elif defined(__BIG_ENDIAN_BITFIELD)
+	u8	zflowid0:4, hwrsvd3:4;
+#endif
+
+	/* Bits 7:0 of zero copy flow ID.  */
+	u8	zflowid1;
+
+	/* Bits 15:8 of hash value, H2.  */
+	u8	hashval2_0;
+
+	/* Bits 7:0 of hash value, H2.  */
+	u8	hashval2_1;
+
+	/* Bits 19:16 of hash value, H1.  */
+#if defined(__LITTLE_ENDIAN_BITFIELD)
+	u8	hwrsvd4:4, hashval1_0:4;
+#elif defined(__BIG_ENDIAN_BITFIELD)
+	u8	hashval1_0:4, hwrsvd4:4;
+#endif
+
+	/* Bits 15:8 of hash value, H1.  */
+	u8	hashval1_1;
+
+	/* Bits 7:0 of hash value, H1.  */
+	u8	hashval1_2;
+
+	u8	hwrsvd5;
+	u8	hwrsvd6;
+
+	u8	usrdata_0;	/* Bits 39:32 of user data.  */
+	u8	usrdata_1;	/* Bits 31:24 of user data.  */
+	u8	usrdata_2;	/* Bits 23:16 of user data.  */
+	u8	usrdata_3;	/* Bits 15:8 of user data.  */
+	u8	usrdata_4;	/* Bits 7:0 of user data.  */
+};
+
+struct tx_dma_mbox {
+	u64	tx_dma_pre_st;
+	u64	tx_cs;
+	u64	tx_ring_kick;
+	u64	tx_ring_hdl;
+	u64	resv1;
+	u32	tx_rng_err_logl;
+	u32	tx_rng_err_logh;
+	u64	resv2;
+	u64	resv3;
+};
+
+struct tx_pkt_hdr {
+	__le64	flags;
+#define TXHDR_PAD		0x0000000000000007ULL
+#define  TXHDR_PAD_SHIFT	0
+#define TXHDR_LEN		0x000000003fff0000ULL
+#define  TXHDR_LEN_SHIFT	16
+#define TXHDR_L4STUFF		0x0000003f00000000ULL
+#define  TXHDR_L4STUFF_SHIFT	32
+#define TXHDR_L4START		0x00003f0000000000ULL
+#define  TXHDR_L4START_SHIFT	40
+#define TXHDR_L3START		0x000f000000000000ULL
+#define  TXHDR_L3START_SHIFT	48
+#define TXHDR_IHL		0x00f0000000000000ULL
+#define  TXHDR_IHL_SHIFT	52
+#define TXHDR_VLAN		0x0100000000000000ULL
+#define TXHDR_LLC		0x0200000000000000ULL
+#define TXHDR_IP_VER		0x2000000000000000ULL
+#define TXHDR_CSUM_NONE		0x0000000000000000ULL
+#define TXHDR_CSUM_TCP		0x4000000000000000ULL
+#define TXHDR_CSUM_UDP		0x8000000000000000ULL
+#define TXHDR_CSUM_SCTP		0xc000000000000000ULL
+	__le64	resv;
+};
+
+#define TX_DESC_SOP		0x8000000000000000ULL
+#define TX_DESC_MARK		0x4000000000000000ULL
+#define TX_DESC_NUM_PTR		0x3c00000000000000ULL
+#define TX_DESC_NUM_PTR_SHIFT	58
+#define TX_DESC_TR_LEN		0x01fff00000000000ULL
+#define TX_DESC_TR_LEN_SHIFT	44
+#define TX_DESC_SAD		0x00000fffffffffffULL
+#define TX_DESC_SAD_SHIFT	0
+
+struct tx_buff_info {
+	struct sk_buff *skb;
+	u64 mapping;
+};
+
+struct txdma_mailbox {
+	__le64	tx_dma_pre_st;
+	__le64	tx_cs;
+	__le64	tx_ring_kick;
+	__le64	tx_ring_hdl;
+	__le64	resv1;
+	__le32	tx_rng_err_logl;
+	__le32	tx_rng_err_logh;
+	__le64	resv2[2];
+} __attribute__((aligned(64)));
+
+#define MAX_TX_RING_SIZE	256
+#define MAX_TX_DESC_LEN		4076
+
+struct tx_ring_info {
+	struct tx_buff_info	tx_buffs[MAX_TX_RING_SIZE];
+	struct niu		*np;
+	u64			tx_cs;
+	int			pending;
+	int			prod;
+	int			cons;
+	int			wrap_bit;
+	u16			last_pkt_cnt;
+	u16			tx_channel;
+	u16			mark_counter;
+	u16			mark_freq;
+	u16			mark_pending;
+	u16			__pad;
+	struct txdma_mailbox	*mbox;
+	__le64			*descr;
+
+	u64			tx_packets;
+	u64			tx_bytes;
+	u64			tx_errors;
+
+	u64			mbox_dma;
+	u64			descr_dma;
+	int			max_burst;
+};
+
+#define NEXT_TX(tp, index) \
+	(((index) + 1) < (tp)->pending ? ((index) + 1) : 0)
+
+#define PREVIOUS_TX(tp, index) \
+	(((index) - 1) >= 0 ? ((index) - 1) : (((tp)->pending) - 1))
+
+static inline u32 niu_tx_avail(struct tx_ring_info *tp)
+{
+	return (tp->pending -
+		((tp->prod - tp->cons) & (MAX_TX_RING_SIZE - 1)));
+}
+
+struct rxdma_mailbox {
+	__le64	rx_dma_ctl_stat;
+	__le64	rbr_stat;
+	__le32	rbr_hdl;
+	__le32	rbr_hdh;
+	__le64	resv1;
+	__le32	rcrstat_c;
+	__le32	rcrstat_b;
+	__le64	rcrstat_a;
+	__le64	resv2[2];
+} __attribute__((aligned(64)));
+
+#define MAX_RBR_RING_SIZE	128
+#define MAX_RCR_RING_SIZE	(MAX_RBR_RING_SIZE * 2)
+
+#define RBR_REFILL_MIN		16
+
+#define RX_SKB_ALLOC_SIZE	128 + NET_IP_ALIGN
+
+struct rx_ring_info {
+	struct niu		*np;
+	int			rx_channel;
+	u16			rbr_block_size;
+	u16			rbr_blocks_per_page;
+	u16			rbr_sizes[4];
+	unsigned int		rcr_index;
+	unsigned int		rcr_table_size;
+	unsigned int		rbr_index;
+	unsigned int		rbr_pending;
+	unsigned int		rbr_refill_pending;
+	unsigned int		rbr_kick_thresh;
+	unsigned int		rbr_table_size;
+	struct page		**rxhash;
+	struct rxdma_mailbox	*mbox;
+	__le64			*rcr;
+	__le32			*rbr;
+#define RBR_DESCR_ADDR_SHIFT	12
+
+	u64			rx_packets;
+	u64			rx_bytes;
+	u64			rx_dropped;
+	u64			rx_errors;
+
+	u64			mbox_dma;
+	u64			rcr_dma;
+	u64			rbr_dma;
+
+	/* WRED */
+	int			nonsyn_window;
+	int			nonsyn_threshold;
+	int			syn_window;
+	int			syn_threshold;
+
+	/* interrupt mitigation */
+	int			rcr_pkt_threshold;
+	int			rcr_timeout;
+};
+
+#define NEXT_RCR(rp, index) \
+	(((index) + 1) < (rp)->rcr_table_size ? ((index) + 1) : 0)
+#define NEXT_RBR(rp, index) \
+	(((index) + 1) < (rp)->rbr_table_size ? ((index) + 1) : 0)
+
+#define NIU_MAX_PORTS		4
+#define NIU_NUM_RXCHAN		16
+#define NIU_NUM_TXCHAN		24
+#define MAC_NUM_HASH		16
+
+#define NIU_MAX_MTU		9216
+
+/* VPD strings */
+#define	NIU_QGC_LP_BM_STR	"501-7606"
+#define	NIU_2XGF_LP_BM_STR	"501-7283"
+#define	NIU_QGC_PEM_BM_STR	"501-7765"
+#define	NIU_2XGF_PEM_BM_STR	"501-7626"
+#define	NIU_ALONSO_BM_STR	"373-0202"
+#define	NIU_FOXXY_BM_STR	"501-7961"
+#define	NIU_2XGF_MRVL_BM_STR	"SK-6E82"
+#define	NIU_QGC_LP_MDL_STR	"SUNW,pcie-qgc"
+#define	NIU_2XGF_LP_MDL_STR	"SUNW,pcie-2xgf"
+#define	NIU_QGC_PEM_MDL_STR	"SUNW,pcie-qgc-pem"
+#define	NIU_2XGF_PEM_MDL_STR	"SUNW,pcie-2xgf-pem"
+#define	NIU_ALONSO_MDL_STR	"SUNW,CP3220"
+#define	NIU_KIMI_MDL_STR	"SUNW,CP3260"
+#define	NIU_MARAMBA_MDL_STR	"SUNW,pcie-neptune"
+#define	NIU_FOXXY_MDL_STR	"SUNW,pcie-rfem"
+#define	NIU_2XGF_MRVL_MDL_STR	"SysKonnect,pcie-2xgf"
+
+#define NIU_VPD_MIN_MAJOR	3
+#define NIU_VPD_MIN_MINOR	4
+
+#define NIU_VPD_MODEL_MAX	32
+#define NIU_VPD_BD_MODEL_MAX	16
+#define NIU_VPD_VERSION_MAX	64
+#define NIU_VPD_PHY_TYPE_MAX	8
+
+struct niu_vpd {
+	char			model[NIU_VPD_MODEL_MAX];
+	char			board_model[NIU_VPD_BD_MODEL_MAX];
+	char			version[NIU_VPD_VERSION_MAX];
+	char			phy_type[NIU_VPD_PHY_TYPE_MAX];
+	u8			mac_num;
+	u8			__pad;
+	u8			local_mac[6];
+	int			fcode_major;
+	int			fcode_minor;
+};
+
+struct niu_altmac_rdc {
+	u8			alt_mac_num;
+	u8			rdc_num;
+	u8			mac_pref;
+};
+
+struct niu_vlan_rdc {
+	u8			rdc_num;
+	u8			vlan_pref;
+};
+
+struct niu_classifier {
+	struct niu_altmac_rdc	alt_mac_mappings[16];
+	struct niu_vlan_rdc	vlan_mappings[ENET_VLAN_TBL_NUM_ENTRIES];
+
+	u16			tcam_top;
+	u16			tcam_sz;
+	u16			tcam_valid_entries;
+	u16			num_alt_mac_mappings;
+
+	u32			h1_init;
+	u16			h2_init;
+};
+
+#define NIU_NUM_RDC_TABLES	8
+#define NIU_RDC_TABLE_SLOTS	16
+
+struct rdc_table {
+	u8			rxdma_channel[NIU_RDC_TABLE_SLOTS];
+};
+
+struct niu_rdc_tables {
+	struct rdc_table	tables[NIU_NUM_RDC_TABLES];
+	int			first_table_num;
+	int			num_tables;
+};
+
+#define PHY_TYPE_PMA_PMD	0
+#define PHY_TYPE_PCS		1
+#define PHY_TYPE_MII		2
+#define PHY_TYPE_MAX		3
+
+struct phy_probe_info {
+	u32	phy_id[PHY_TYPE_MAX][NIU_MAX_PORTS];
+	u8	phy_port[PHY_TYPE_MAX][NIU_MAX_PORTS];
+	u8	cur[PHY_TYPE_MAX];
+
+	struct device_attribute	phy_port_attrs[PHY_TYPE_MAX * NIU_MAX_PORTS];
+	struct device_attribute	phy_type_attrs[PHY_TYPE_MAX * NIU_MAX_PORTS];
+	struct device_attribute	phy_id_attrs[PHY_TYPE_MAX * NIU_MAX_PORTS];
+};
+
+struct niu_tcam_entry {
+	u8			valid;
+	u64			key[4];
+	u64			key_mask[4];
+	u64			assoc_data;
+};
+
+struct device_node;
+union niu_parent_id {
+	struct {
+		int		domain;
+		int		bus;
+		int		device;
+	} pci;
+	struct device_node	*of;
+};
+
+struct niu;
+struct niu_parent {
+	struct platform_device	*plat_dev;
+	int			index;
+
+	union niu_parent_id	id;
+
+	struct niu		*ports[NIU_MAX_PORTS];
+
+	atomic_t		refcnt;
+	struct list_head	list;
+
+	spinlock_t		lock;
+
+	u32			flags;
+#define PARENT_FLGS_CLS_HWINIT	0x00000001
+
+	u32			port_phy;
+#define PORT_PHY_UNKNOWN	0x00000000
+#define PORT_PHY_INVALID	0xffffffff
+#define PORT_TYPE_10G		0x01
+#define PORT_TYPE_1G		0x02
+#define PORT_TYPE_MASK		0x03
+
+	u8			rxchan_per_port[NIU_MAX_PORTS];
+	u8			txchan_per_port[NIU_MAX_PORTS];
+
+	struct niu_rdc_tables	rdc_group_cfg[NIU_MAX_PORTS];
+	u8			rdc_default[NIU_MAX_PORTS];
+
+	u8			ldg_map[LDN_MAX + 1];
+
+	u8			plat_type;
+#define PLAT_TYPE_INVALID	0x00
+#define PLAT_TYPE_ATLAS		0x01
+#define PLAT_TYPE_NIU		0x02
+#define PLAT_TYPE_VF_P0		0x03
+#define PLAT_TYPE_VF_P1		0x04
+#define PLAT_TYPE_ATCA_CP3220	0x08
+
+	u8			num_ports;
+
+	u16			tcam_num_entries;
+#define NIU_PCI_TCAM_ENTRIES	256
+#define NIU_NONPCI_TCAM_ENTRIES	128
+#define NIU_TCAM_ENTRIES_MAX	256
+
+	int			rxdma_clock_divider;
+
+	struct phy_probe_info	phy_probe_info;
+
+	struct niu_tcam_entry	tcam[NIU_TCAM_ENTRIES_MAX];
+
+#define	NIU_L2_PROG_CLS		2
+#define	NIU_L3_PROG_CLS		4
+	u64			l2_cls[NIU_L2_PROG_CLS];
+	u64			l3_cls[NIU_L3_PROG_CLS];
+	u64			tcam_key[12];
+	u64			flow_key[12];
+	u16			l3_cls_refcnt[NIU_L3_PROG_CLS];
+	u8			l3_cls_pid[NIU_L3_PROG_CLS];
+};
+
+struct niu_ops {
+	void *(*alloc_coherent)(struct device *dev, size_t size,
+				u64 *handle, gfp_t flag);
+	void (*free_coherent)(struct device *dev, size_t size,
+			      void *cpu_addr, u64 handle);
+	u64 (*map_page)(struct device *dev, struct page *page,
+			unsigned long offset, size_t size,
+			enum dma_data_direction direction);
+	void (*unmap_page)(struct device *dev, u64 dma_address,
+			   size_t size, enum dma_data_direction direction);
+	u64 (*map_single)(struct device *dev, void *cpu_addr,
+			  size_t size,
+			  enum dma_data_direction direction);
+	void (*unmap_single)(struct device *dev, u64 dma_address,
+			     size_t size, enum dma_data_direction direction);
+	int (*mapping_error)(struct device *dev, u64 dma_address);
+};
+
+struct niu_link_config {
+	u32				supported;
+
+	/* Describes what we're trying to get. */
+	u32				advertising;
+	u16				speed;
+	u8				duplex;
+	u8				autoneg;
+
+	/* Describes what we actually have. */
+	u32				active_advertising;
+	u16				active_speed;
+	u8				active_duplex;
+	u8				active_autoneg;
+#define SPEED_INVALID		0xffff
+#define DUPLEX_INVALID		0xff
+#define AUTONEG_INVALID		0xff
+
+	u8				loopback_mode;
+#define LOOPBACK_DISABLED	0x00
+#define LOOPBACK_PHY		0x01
+#define LOOPBACK_MAC		0x02
+};
+
+struct niu_ldg {
+	struct napi_struct	napi;
+	struct niu	*np;
+	u8		ldg_num;
+	u8		timer;
+	u64		v0, v1, v2;
+	unsigned int	irq;
+};
+
+struct niu_xmac_stats {
+	u64	tx_frames;
+	u64	tx_bytes;
+	u64	tx_fifo_errors;
+	u64	tx_overflow_errors;
+	u64	tx_max_pkt_size_errors;
+	u64	tx_underflow_errors;
+
+	u64	rx_local_faults;
+	u64	rx_remote_faults;
+	u64	rx_link_faults;
+	u64	rx_align_errors;
+	u64	rx_frags;
+	u64	rx_mcasts;
+	u64	rx_bcasts;
+	u64	rx_hist_cnt1;
+	u64	rx_hist_cnt2;
+	u64	rx_hist_cnt3;
+	u64	rx_hist_cnt4;
+	u64	rx_hist_cnt5;
+	u64	rx_hist_cnt6;
+	u64	rx_hist_cnt7;
+	u64	rx_octets;
+	u64	rx_code_violations;
+	u64	rx_len_errors;
+	u64	rx_crc_errors;
+	u64	rx_underflows;
+	u64	rx_overflows;
+
+	u64	pause_off_state;
+	u64	pause_on_state;
+	u64	pause_received;
+};
+
+struct niu_bmac_stats {
+	u64	tx_underflow_errors;
+	u64	tx_max_pkt_size_errors;
+	u64	tx_bytes;
+	u64	tx_frames;
+
+	u64	rx_overflows;
+	u64	rx_frames;
+	u64	rx_align_errors;
+	u64	rx_crc_errors;
+	u64	rx_len_errors;
+
+	u64	pause_off_state;
+	u64	pause_on_state;
+	u64	pause_received;
+};
+
+union niu_mac_stats {
+	struct niu_xmac_stats	xmac;
+	struct niu_bmac_stats	bmac;
+};
+
+struct niu_phy_ops {
+	int (*serdes_init)(struct niu *np);
+	int (*xcvr_init)(struct niu *np);
+	int (*link_status)(struct niu *np, int *);
+};
+
+struct platform_device;
+struct niu {
+	void __iomem			*regs;
+	struct net_device		*dev;
+	struct pci_dev			*pdev;
+	struct device			*device;
+	struct niu_parent		*parent;
+
+	u32				flags;
+#define NIU_FLAGS_HOTPLUG_PHY_PRESENT	0x02000000 /* Removable PHY detected*/
+#define NIU_FLAGS_HOTPLUG_PHY		0x01000000 /* Removable PHY */
+#define NIU_FLAGS_VPD_VALID		0x00800000 /* VPD has valid version */
+#define NIU_FLAGS_MSIX			0x00400000 /* MSI-X in use */
+#define NIU_FLAGS_MCAST			0x00200000 /* multicast filter enabled */
+#define NIU_FLAGS_PROMISC		0x00100000 /* PROMISC enabled */
+#define NIU_FLAGS_XCVR_SERDES		0x00080000 /* 0=PHY 1=SERDES */
+#define NIU_FLAGS_10G			0x00040000 /* 0=1G 1=10G */
+#define NIU_FLAGS_FIBER			0x00020000 /* 0=COPPER 1=FIBER */
+#define NIU_FLAGS_XMAC			0x00010000 /* 0=BMAC 1=XMAC */
+
+	u32				msg_enable;
+	char                            irq_name[NIU_NUM_RXCHAN + NIU_NUM_TXCHAN + 3][IFNAMSIZ + 7];
+
+	/* Protects hw programming, and ring state.  */
+	spinlock_t			lock;
+
+	const struct niu_ops		*ops;
+	union niu_mac_stats		mac_stats;
+
+	struct rx_ring_info		*rx_rings;
+	struct tx_ring_info		*tx_rings;
+	int				num_rx_rings;
+	int				num_tx_rings;
+
+	struct niu_ldg			ldg[NIU_NUM_LDG];
+	int				num_ldg;
+
+	void __iomem			*mac_regs;
+	unsigned long			ipp_off;
+	unsigned long			pcs_off;
+	unsigned long			xpcs_off;
+
+	struct timer_list		timer;
+	u64				orig_led_state;
+	const struct niu_phy_ops	*phy_ops;
+	int				phy_addr;
+
+	struct niu_link_config		link_config;
+
+	struct work_struct		reset_task;
+
+	u8				port;
+	u8				mac_xcvr;
+#define MAC_XCVR_MII			1
+#define MAC_XCVR_PCS			2
+#define MAC_XCVR_XPCS			3
+
+	struct niu_classifier		clas;
+
+	struct niu_vpd			vpd;
+	u32				eeprom_len;
+
+	struct platform_device		*op;
+	void __iomem			*vir_regs_1;
+	void __iomem			*vir_regs_2;
+};
+
+#endif /* _NIU_H */
diff --git a/drivers/net/ethernet/sun/sunbmac.c b/drivers/net/ethernet/sun/sunbmac.c
new file mode 100644
index 000000000..edb2fd3a6
--- /dev/null
+++ b/drivers/net/ethernet/sun/sunbmac.c
@@ -0,0 +1,1278 @@
+// SPDX-License-Identifier: GPL-2.0
+/* sunbmac.c: Driver for Sparc BigMAC 100baseT ethernet adapters.
+ *
+ * Copyright (C) 1997, 1998, 1999, 2003, 2008 David S. Miller (davem@davemloft.net)
+ */
+
+#include <linux/module.h>
+#include <linux/pgtable.h>
+
+#include <linux/kernel.h>
+#include <linux/types.h>
+#include <linux/fcntl.h>
+#include <linux/interrupt.h>
+#include <linux/ioport.h>
+#include <linux/in.h>
+#include <linux/string.h>
+#include <linux/delay.h>
+#include <linux/crc32.h>
+#include <linux/errno.h>
+#include <linux/ethtool.h>
+#include <linux/mii.h>
+#include <linux/netdevice.h>
+#include <linux/etherdevice.h>
+#include <linux/skbuff.h>
+#include <linux/bitops.h>
+#include <linux/dma-mapping.h>
+#include <linux/of.h>
+#include <linux/platform_device.h>
+#include <linux/gfp.h>
+
+#include <asm/auxio.h>
+#include <asm/byteorder.h>
+#include <asm/dma.h>
+#include <asm/idprom.h>
+#include <asm/io.h>
+#include <asm/openprom.h>
+#include <asm/oplib.h>
+
+#include "sunbmac.h"
+
+#define DRV_NAME	"sunbmac"
+#define DRV_VERSION	"2.1"
+#define DRV_RELDATE	"August 26, 2008"
+#define DRV_AUTHOR	"David S. Miller (davem@davemloft.net)"
+
+static char version[] =
+	DRV_NAME ".c:v" DRV_VERSION " " DRV_RELDATE " " DRV_AUTHOR "\n";
+
+MODULE_VERSION(DRV_VERSION);
+MODULE_AUTHOR(DRV_AUTHOR);
+MODULE_DESCRIPTION("Sun BigMAC 100baseT ethernet driver");
+MODULE_LICENSE("GPL");
+
+#undef DEBUG_PROBE
+#undef DEBUG_TX
+#undef DEBUG_IRQ
+
+#ifdef DEBUG_PROBE
+#define DP(x)  printk x
+#else
+#define DP(x)
+#endif
+
+#ifdef DEBUG_TX
+#define DTX(x)  printk x
+#else
+#define DTX(x)
+#endif
+
+#ifdef DEBUG_IRQ
+#define DIRQ(x)  printk x
+#else
+#define DIRQ(x)
+#endif
+
+#define DEFAULT_JAMSIZE    4 /* Toe jam */
+
+#define QEC_RESET_TRIES 200
+
+static int qec_global_reset(void __iomem *gregs)
+{
+	int tries = QEC_RESET_TRIES;
+
+	sbus_writel(GLOB_CTRL_RESET, gregs + GLOB_CTRL);
+	while (--tries) {
+		if (sbus_readl(gregs + GLOB_CTRL) & GLOB_CTRL_RESET) {
+			udelay(20);
+			continue;
+		}
+		break;
+	}
+	if (tries)
+		return 0;
+	printk(KERN_ERR "BigMAC: Cannot reset the QEC.\n");
+	return -1;
+}
+
+static void qec_init(struct bigmac *bp)
+{
+	struct platform_device *qec_op = bp->qec_op;
+	void __iomem *gregs = bp->gregs;
+	u8 bsizes = bp->bigmac_bursts;
+	u32 regval;
+
+	/* 64byte bursts do not work at the moment, do
+	 * not even try to enable them.  -DaveM
+	 */
+	if (bsizes & DMA_BURST32)
+		regval = GLOB_CTRL_B32;
+	else
+		regval = GLOB_CTRL_B16;
+	sbus_writel(regval | GLOB_CTRL_BMODE, gregs + GLOB_CTRL);
+	sbus_writel(GLOB_PSIZE_2048, gregs + GLOB_PSIZE);
+
+	/* All of memsize is given to bigmac. */
+	sbus_writel(resource_size(&qec_op->resource[1]),
+		    gregs + GLOB_MSIZE);
+
+	/* Half to the transmitter, half to the receiver. */
+	sbus_writel(resource_size(&qec_op->resource[1]) >> 1,
+		    gregs + GLOB_TSIZE);
+	sbus_writel(resource_size(&qec_op->resource[1]) >> 1,
+		    gregs + GLOB_RSIZE);
+}
+
+#define TX_RESET_TRIES     32
+#define RX_RESET_TRIES     32
+
+static void bigmac_tx_reset(void __iomem *bregs)
+{
+	int tries = TX_RESET_TRIES;
+
+	sbus_writel(0, bregs + BMAC_TXCFG);
+
+	/* The fifo threshold bit is read-only and does
+	 * not clear.  -DaveM
+	 */
+	while ((sbus_readl(bregs + BMAC_TXCFG) & ~(BIGMAC_TXCFG_FIFO)) != 0 &&
+	       --tries != 0)
+		udelay(20);
+
+	if (!tries) {
+		printk(KERN_ERR "BIGMAC: Transmitter will not reset.\n");
+		printk(KERN_ERR "BIGMAC: tx_cfg is %08x\n",
+		       sbus_readl(bregs + BMAC_TXCFG));
+	}
+}
+
+static void bigmac_rx_reset(void __iomem *bregs)
+{
+	int tries = RX_RESET_TRIES;
+
+	sbus_writel(0, bregs + BMAC_RXCFG);
+	while (sbus_readl(bregs + BMAC_RXCFG) && --tries)
+		udelay(20);
+
+	if (!tries) {
+		printk(KERN_ERR "BIGMAC: Receiver will not reset.\n");
+		printk(KERN_ERR "BIGMAC: rx_cfg is %08x\n",
+		       sbus_readl(bregs + BMAC_RXCFG));
+	}
+}
+
+/* Reset the transmitter and receiver. */
+static void bigmac_stop(struct bigmac *bp)
+{
+	bigmac_tx_reset(bp->bregs);
+	bigmac_rx_reset(bp->bregs);
+}
+
+static void bigmac_get_counters(struct bigmac *bp, void __iomem *bregs)
+{
+	struct net_device_stats *stats = &bp->dev->stats;
+
+	stats->rx_crc_errors += sbus_readl(bregs + BMAC_RCRCECTR);
+	sbus_writel(0, bregs + BMAC_RCRCECTR);
+
+	stats->rx_frame_errors += sbus_readl(bregs + BMAC_UNALECTR);
+	sbus_writel(0, bregs + BMAC_UNALECTR);
+
+	stats->rx_length_errors += sbus_readl(bregs + BMAC_GLECTR);
+	sbus_writel(0, bregs + BMAC_GLECTR);
+
+	stats->tx_aborted_errors += sbus_readl(bregs + BMAC_EXCTR);
+
+	stats->collisions +=
+		(sbus_readl(bregs + BMAC_EXCTR) +
+		 sbus_readl(bregs + BMAC_LTCTR));
+	sbus_writel(0, bregs + BMAC_EXCTR);
+	sbus_writel(0, bregs + BMAC_LTCTR);
+}
+
+static void bigmac_clean_rings(struct bigmac *bp)
+{
+	int i;
+
+	for (i = 0; i < RX_RING_SIZE; i++) {
+		if (bp->rx_skbs[i] != NULL) {
+			dev_kfree_skb_any(bp->rx_skbs[i]);
+			bp->rx_skbs[i] = NULL;
+		}
+	}
+
+	for (i = 0; i < TX_RING_SIZE; i++) {
+		if (bp->tx_skbs[i] != NULL) {
+			dev_kfree_skb_any(bp->tx_skbs[i]);
+			bp->tx_skbs[i] = NULL;
+		}
+	}
+}
+
+static void bigmac_init_rings(struct bigmac *bp, bool non_blocking)
+{
+	struct bmac_init_block *bb = bp->bmac_block;
+	int i;
+	gfp_t gfp_flags = GFP_KERNEL;
+
+	if (non_blocking)
+		gfp_flags = GFP_ATOMIC;
+
+	bp->rx_new = bp->rx_old = bp->tx_new = bp->tx_old = 0;
+
+	/* Free any skippy bufs left around in the rings. */
+	bigmac_clean_rings(bp);
+
+	/* Now get new skbufs for the receive ring. */
+	for (i = 0; i < RX_RING_SIZE; i++) {
+		struct sk_buff *skb;
+
+		skb = big_mac_alloc_skb(RX_BUF_ALLOC_SIZE, gfp_flags);
+		if (!skb)
+			continue;
+
+		bp->rx_skbs[i] = skb;
+
+		/* Because we reserve afterwards. */
+		skb_put(skb, ETH_FRAME_LEN);
+		skb_reserve(skb, 34);
+
+		bb->be_rxd[i].rx_addr =
+			dma_map_single(&bp->bigmac_op->dev,
+				       skb->data,
+				       RX_BUF_ALLOC_SIZE - 34,
+				       DMA_FROM_DEVICE);
+		bb->be_rxd[i].rx_flags =
+			(RXD_OWN | ((RX_BUF_ALLOC_SIZE - 34) & RXD_LENGTH));
+	}
+
+	for (i = 0; i < TX_RING_SIZE; i++)
+		bb->be_txd[i].tx_flags = bb->be_txd[i].tx_addr = 0;
+}
+
+#define MGMT_CLKON  (MGMT_PAL_INT_MDIO|MGMT_PAL_EXT_MDIO|MGMT_PAL_OENAB|MGMT_PAL_DCLOCK)
+#define MGMT_CLKOFF (MGMT_PAL_INT_MDIO|MGMT_PAL_EXT_MDIO|MGMT_PAL_OENAB)
+
+static void idle_transceiver(void __iomem *tregs)
+{
+	int i = 20;
+
+	while (i--) {
+		sbus_writel(MGMT_CLKOFF, tregs + TCVR_MPAL);
+		sbus_readl(tregs + TCVR_MPAL);
+		sbus_writel(MGMT_CLKON, tregs + TCVR_MPAL);
+		sbus_readl(tregs + TCVR_MPAL);
+	}
+}
+
+static void write_tcvr_bit(struct bigmac *bp, void __iomem *tregs, int bit)
+{
+	if (bp->tcvr_type == internal) {
+		bit = (bit & 1) << 3;
+		sbus_writel(bit | (MGMT_PAL_OENAB | MGMT_PAL_EXT_MDIO),
+			    tregs + TCVR_MPAL);
+		sbus_readl(tregs + TCVR_MPAL);
+		sbus_writel(bit | MGMT_PAL_OENAB | MGMT_PAL_EXT_MDIO | MGMT_PAL_DCLOCK,
+			    tregs + TCVR_MPAL);
+		sbus_readl(tregs + TCVR_MPAL);
+	} else if (bp->tcvr_type == external) {
+		bit = (bit & 1) << 2;
+		sbus_writel(bit | MGMT_PAL_INT_MDIO | MGMT_PAL_OENAB,
+			    tregs + TCVR_MPAL);
+		sbus_readl(tregs + TCVR_MPAL);
+		sbus_writel(bit | MGMT_PAL_INT_MDIO | MGMT_PAL_OENAB | MGMT_PAL_DCLOCK,
+			    tregs + TCVR_MPAL);
+		sbus_readl(tregs + TCVR_MPAL);
+	} else {
+		printk(KERN_ERR "write_tcvr_bit: No transceiver type known!\n");
+	}
+}
+
+static int read_tcvr_bit(struct bigmac *bp, void __iomem *tregs)
+{
+	int retval = 0;
+
+	if (bp->tcvr_type == internal) {
+		sbus_writel(MGMT_PAL_EXT_MDIO, tregs + TCVR_MPAL);
+		sbus_readl(tregs + TCVR_MPAL);
+		sbus_writel(MGMT_PAL_EXT_MDIO | MGMT_PAL_DCLOCK,
+			    tregs + TCVR_MPAL);
+		sbus_readl(tregs + TCVR_MPAL);
+		retval = (sbus_readl(tregs + TCVR_MPAL) & MGMT_PAL_INT_MDIO) >> 3;
+	} else if (bp->tcvr_type == external) {
+		sbus_writel(MGMT_PAL_INT_MDIO, tregs + TCVR_MPAL);
+		sbus_readl(tregs + TCVR_MPAL);
+		sbus_writel(MGMT_PAL_INT_MDIO | MGMT_PAL_DCLOCK, tregs + TCVR_MPAL);
+		sbus_readl(tregs + TCVR_MPAL);
+		retval = (sbus_readl(tregs + TCVR_MPAL) & MGMT_PAL_EXT_MDIO) >> 2;
+	} else {
+		printk(KERN_ERR "read_tcvr_bit: No transceiver type known!\n");
+	}
+	return retval;
+}
+
+static int read_tcvr_bit2(struct bigmac *bp, void __iomem *tregs)
+{
+	int retval = 0;
+
+	if (bp->tcvr_type == internal) {
+		sbus_writel(MGMT_PAL_EXT_MDIO, tregs + TCVR_MPAL);
+		sbus_readl(tregs + TCVR_MPAL);
+		retval = (sbus_readl(tregs + TCVR_MPAL) & MGMT_PAL_INT_MDIO) >> 3;
+		sbus_writel(MGMT_PAL_EXT_MDIO | MGMT_PAL_DCLOCK, tregs + TCVR_MPAL);
+		sbus_readl(tregs + TCVR_MPAL);
+	} else if (bp->tcvr_type == external) {
+		sbus_writel(MGMT_PAL_INT_MDIO, tregs + TCVR_MPAL);
+		sbus_readl(tregs + TCVR_MPAL);
+		retval = (sbus_readl(tregs + TCVR_MPAL) & MGMT_PAL_EXT_MDIO) >> 2;
+		sbus_writel(MGMT_PAL_INT_MDIO | MGMT_PAL_DCLOCK, tregs + TCVR_MPAL);
+		sbus_readl(tregs + TCVR_MPAL);
+	} else {
+		printk(KERN_ERR "read_tcvr_bit2: No transceiver type known!\n");
+	}
+	return retval;
+}
+
+static void put_tcvr_byte(struct bigmac *bp,
+			  void __iomem *tregs,
+			  unsigned int byte)
+{
+	int shift = 4;
+
+	do {
+		write_tcvr_bit(bp, tregs, ((byte >> shift) & 1));
+		shift -= 1;
+	} while (shift >= 0);
+}
+
+static void bigmac_tcvr_write(struct bigmac *bp, void __iomem *tregs,
+			      int reg, unsigned short val)
+{
+	int shift;
+
+	reg &= 0xff;
+	val &= 0xffff;
+	switch(bp->tcvr_type) {
+	case internal:
+	case external:
+		break;
+
+	default:
+		printk(KERN_ERR "bigmac_tcvr_read: Whoops, no known transceiver type.\n");
+		return;
+	}
+
+	idle_transceiver(tregs);
+	write_tcvr_bit(bp, tregs, 0);
+	write_tcvr_bit(bp, tregs, 1);
+	write_tcvr_bit(bp, tregs, 0);
+	write_tcvr_bit(bp, tregs, 1);
+
+	put_tcvr_byte(bp, tregs,
+		      ((bp->tcvr_type == internal) ?
+		       BIGMAC_PHY_INTERNAL : BIGMAC_PHY_EXTERNAL));
+
+	put_tcvr_byte(bp, tregs, reg);
+
+	write_tcvr_bit(bp, tregs, 1);
+	write_tcvr_bit(bp, tregs, 0);
+
+	shift = 15;
+	do {
+		write_tcvr_bit(bp, tregs, (val >> shift) & 1);
+		shift -= 1;
+	} while (shift >= 0);
+}
+
+static unsigned short bigmac_tcvr_read(struct bigmac *bp,
+				       void __iomem *tregs,
+				       int reg)
+{
+	unsigned short retval = 0;
+
+	reg &= 0xff;
+	switch(bp->tcvr_type) {
+	case internal:
+	case external:
+		break;
+
+	default:
+		printk(KERN_ERR "bigmac_tcvr_read: Whoops, no known transceiver type.\n");
+		return 0xffff;
+	}
+
+	idle_transceiver(tregs);
+	write_tcvr_bit(bp, tregs, 0);
+	write_tcvr_bit(bp, tregs, 1);
+	write_tcvr_bit(bp, tregs, 1);
+	write_tcvr_bit(bp, tregs, 0);
+
+	put_tcvr_byte(bp, tregs,
+		      ((bp->tcvr_type == internal) ?
+		       BIGMAC_PHY_INTERNAL : BIGMAC_PHY_EXTERNAL));
+
+	put_tcvr_byte(bp, tregs, reg);
+
+	if (bp->tcvr_type == external) {
+		int shift = 15;
+
+		(void) read_tcvr_bit2(bp, tregs);
+		(void) read_tcvr_bit2(bp, tregs);
+
+		do {
+			int tmp;
+
+			tmp = read_tcvr_bit2(bp, tregs);
+			retval |= ((tmp & 1) << shift);
+			shift -= 1;
+		} while (shift >= 0);
+
+		(void) read_tcvr_bit2(bp, tregs);
+		(void) read_tcvr_bit2(bp, tregs);
+		(void) read_tcvr_bit2(bp, tregs);
+	} else {
+		int shift = 15;
+
+		(void) read_tcvr_bit(bp, tregs);
+		(void) read_tcvr_bit(bp, tregs);
+
+		do {
+			int tmp;
+
+			tmp = read_tcvr_bit(bp, tregs);
+			retval |= ((tmp & 1) << shift);
+			shift -= 1;
+		} while (shift >= 0);
+
+		(void) read_tcvr_bit(bp, tregs);
+		(void) read_tcvr_bit(bp, tregs);
+		(void) read_tcvr_bit(bp, tregs);
+	}
+	return retval;
+}
+
+static void bigmac_tcvr_init(struct bigmac *bp)
+{
+	void __iomem *tregs = bp->tregs;
+	u32 mpal;
+
+	idle_transceiver(tregs);
+	sbus_writel(MGMT_PAL_INT_MDIO | MGMT_PAL_EXT_MDIO | MGMT_PAL_DCLOCK,
+		    tregs + TCVR_MPAL);
+	sbus_readl(tregs + TCVR_MPAL);
+
+	/* Only the bit for the present transceiver (internal or
+	 * external) will stick, set them both and see what stays.
+	 */
+	sbus_writel(MGMT_PAL_INT_MDIO | MGMT_PAL_EXT_MDIO, tregs + TCVR_MPAL);
+	sbus_readl(tregs + TCVR_MPAL);
+	udelay(20);
+
+	mpal = sbus_readl(tregs + TCVR_MPAL);
+	if (mpal & MGMT_PAL_EXT_MDIO) {
+		bp->tcvr_type = external;
+		sbus_writel(~(TCVR_PAL_EXTLBACK | TCVR_PAL_MSENSE | TCVR_PAL_LTENABLE),
+			    tregs + TCVR_TPAL);
+		sbus_readl(tregs + TCVR_TPAL);
+	} else if (mpal & MGMT_PAL_INT_MDIO) {
+		bp->tcvr_type = internal;
+		sbus_writel(~(TCVR_PAL_SERIAL | TCVR_PAL_EXTLBACK |
+			      TCVR_PAL_MSENSE | TCVR_PAL_LTENABLE),
+			    tregs + TCVR_TPAL);
+		sbus_readl(tregs + TCVR_TPAL);
+	} else {
+		printk(KERN_ERR "BIGMAC: AIEEE, neither internal nor "
+		       "external MDIO available!\n");
+		printk(KERN_ERR "BIGMAC: mgmt_pal[%08x] tcvr_pal[%08x]\n",
+		       sbus_readl(tregs + TCVR_MPAL),
+		       sbus_readl(tregs + TCVR_TPAL));
+	}
+}
+
+static int bigmac_init_hw(struct bigmac *, bool);
+
+static int try_next_permutation(struct bigmac *bp, void __iomem *tregs)
+{
+	if (bp->sw_bmcr & BMCR_SPEED100) {
+		int timeout;
+
+		/* Reset the PHY. */
+		bp->sw_bmcr	= (BMCR_ISOLATE | BMCR_PDOWN | BMCR_LOOPBACK);
+		bigmac_tcvr_write(bp, tregs, MII_BMCR, bp->sw_bmcr);
+		bp->sw_bmcr	= (BMCR_RESET);
+		bigmac_tcvr_write(bp, tregs, MII_BMCR, bp->sw_bmcr);
+
+		timeout = 64;
+		while (--timeout) {
+			bp->sw_bmcr = bigmac_tcvr_read(bp, tregs, MII_BMCR);
+			if ((bp->sw_bmcr & BMCR_RESET) == 0)
+				break;
+			udelay(20);
+		}
+		if (timeout == 0)
+			printk(KERN_ERR "%s: PHY reset failed.\n", bp->dev->name);
+
+		bp->sw_bmcr = bigmac_tcvr_read(bp, tregs, MII_BMCR);
+
+		/* Now we try 10baseT. */
+		bp->sw_bmcr &= ~(BMCR_SPEED100);
+		bigmac_tcvr_write(bp, tregs, MII_BMCR, bp->sw_bmcr);
+		return 0;
+	}
+
+	/* We've tried them all. */
+	return -1;
+}
+
+static void bigmac_timer(struct timer_list *t)
+{
+	struct bigmac *bp = timer_container_of(bp, t, bigmac_timer);
+	void __iomem *tregs = bp->tregs;
+	int restart_timer = 0;
+
+	bp->timer_ticks++;
+	if (bp->timer_state == ltrywait) {
+		bp->sw_bmsr = bigmac_tcvr_read(bp, tregs, MII_BMSR);
+		bp->sw_bmcr = bigmac_tcvr_read(bp, tregs, MII_BMCR);
+		if (bp->sw_bmsr & BMSR_LSTATUS) {
+			printk(KERN_INFO "%s: Link is now up at %s.\n",
+			       bp->dev->name,
+			       (bp->sw_bmcr & BMCR_SPEED100) ?
+			       "100baseT" : "10baseT");
+			bp->timer_state = asleep;
+			restart_timer = 0;
+		} else {
+			if (bp->timer_ticks >= 4) {
+				int ret;
+
+				ret = try_next_permutation(bp, tregs);
+				if (ret == -1) {
+					printk(KERN_ERR "%s: Link down, cable problem?\n",
+					       bp->dev->name);
+					ret = bigmac_init_hw(bp, true);
+					if (ret) {
+						printk(KERN_ERR "%s: Error, cannot re-init the "
+						       "BigMAC.\n", bp->dev->name);
+					}
+					return;
+				}
+				bp->timer_ticks = 0;
+				restart_timer = 1;
+			} else {
+				restart_timer = 1;
+			}
+		}
+	} else {
+		/* Can't happens.... */
+		printk(KERN_ERR "%s: Aieee, link timer is asleep but we got one anyways!\n",
+		       bp->dev->name);
+		restart_timer = 0;
+		bp->timer_ticks = 0;
+		bp->timer_state = asleep; /* foo on you */
+	}
+
+	if (restart_timer != 0) {
+		bp->bigmac_timer.expires = jiffies + ((12 * HZ)/10); /* 1.2 sec. */
+		add_timer(&bp->bigmac_timer);
+	}
+}
+
+/* Well, really we just force the chip into 100baseT then
+ * 10baseT, each time checking for a link status.
+ */
+static void bigmac_begin_auto_negotiation(struct bigmac *bp)
+{
+	void __iomem *tregs = bp->tregs;
+	int timeout;
+
+	/* Grab new software copies of PHY registers. */
+	bp->sw_bmsr	= bigmac_tcvr_read(bp, tregs, MII_BMSR);
+	bp->sw_bmcr	= bigmac_tcvr_read(bp, tregs, MII_BMCR);
+
+	/* Reset the PHY. */
+	bp->sw_bmcr	= (BMCR_ISOLATE | BMCR_PDOWN | BMCR_LOOPBACK);
+	bigmac_tcvr_write(bp, tregs, MII_BMCR, bp->sw_bmcr);
+	bp->sw_bmcr	= (BMCR_RESET);
+	bigmac_tcvr_write(bp, tregs, MII_BMCR, bp->sw_bmcr);
+
+	timeout = 64;
+	while (--timeout) {
+		bp->sw_bmcr = bigmac_tcvr_read(bp, tregs, MII_BMCR);
+		if ((bp->sw_bmcr & BMCR_RESET) == 0)
+			break;
+		udelay(20);
+	}
+	if (timeout == 0)
+		printk(KERN_ERR "%s: PHY reset failed.\n", bp->dev->name);
+
+	bp->sw_bmcr = bigmac_tcvr_read(bp, tregs, MII_BMCR);
+
+	/* First we try 100baseT. */
+	bp->sw_bmcr |= BMCR_SPEED100;
+	bigmac_tcvr_write(bp, tregs, MII_BMCR, bp->sw_bmcr);
+
+	bp->timer_state = ltrywait;
+	bp->timer_ticks = 0;
+	bp->bigmac_timer.expires = jiffies + (12 * HZ) / 10;
+	add_timer(&bp->bigmac_timer);
+}
+
+static int bigmac_init_hw(struct bigmac *bp, bool non_blocking)
+{
+	void __iomem *gregs        = bp->gregs;
+	void __iomem *cregs        = bp->creg;
+	void __iomem *bregs        = bp->bregs;
+	__u32 bblk_dvma = (__u32)bp->bblock_dvma;
+	const unsigned char *e = &bp->dev->dev_addr[0];
+
+	/* Latch current counters into statistics. */
+	bigmac_get_counters(bp, bregs);
+
+	/* Reset QEC. */
+	qec_global_reset(gregs);
+
+	/* Init QEC. */
+	qec_init(bp);
+
+	/* Alloc and reset the tx/rx descriptor chains. */
+	bigmac_init_rings(bp, non_blocking);
+
+	/* Initialize the PHY. */
+	bigmac_tcvr_init(bp);
+
+	/* Stop transmitter and receiver. */
+	bigmac_stop(bp);
+
+	/* Set hardware ethernet address. */
+	sbus_writel(((e[4] << 8) | e[5]), bregs + BMAC_MACADDR2);
+	sbus_writel(((e[2] << 8) | e[3]), bregs + BMAC_MACADDR1);
+	sbus_writel(((e[0] << 8) | e[1]), bregs + BMAC_MACADDR0);
+
+	/* Clear the hash table until mc upload occurs. */
+	sbus_writel(0, bregs + BMAC_HTABLE3);
+	sbus_writel(0, bregs + BMAC_HTABLE2);
+	sbus_writel(0, bregs + BMAC_HTABLE1);
+	sbus_writel(0, bregs + BMAC_HTABLE0);
+
+	/* Enable Big Mac hash table filter. */
+	sbus_writel(BIGMAC_RXCFG_HENABLE | BIGMAC_RXCFG_FIFO,
+		    bregs + BMAC_RXCFG);
+	udelay(20);
+
+	/* Ok, configure the Big Mac transmitter. */
+	sbus_writel(BIGMAC_TXCFG_FIFO, bregs + BMAC_TXCFG);
+
+	/* The HME docs recommend to use the 10LSB of our MAC here. */
+	sbus_writel(((e[5] | e[4] << 8) & 0x3ff),
+		    bregs + BMAC_RSEED);
+
+	/* Enable the output drivers no matter what. */
+	sbus_writel(BIGMAC_XCFG_ODENABLE | BIGMAC_XCFG_RESV,
+		    bregs + BMAC_XIFCFG);
+
+	/* Tell the QEC where the ring descriptors are. */
+	sbus_writel(bblk_dvma + bib_offset(be_rxd, 0),
+		    cregs + CREG_RXDS);
+	sbus_writel(bblk_dvma + bib_offset(be_txd, 0),
+		    cregs + CREG_TXDS);
+
+	/* Setup the FIFO pointers into QEC local memory. */
+	sbus_writel(0, cregs + CREG_RXRBUFPTR);
+	sbus_writel(0, cregs + CREG_RXWBUFPTR);
+	sbus_writel(sbus_readl(gregs + GLOB_RSIZE),
+		    cregs + CREG_TXRBUFPTR);
+	sbus_writel(sbus_readl(gregs + GLOB_RSIZE),
+		    cregs + CREG_TXWBUFPTR);
+
+	/* Tell bigmac what interrupts we don't want to hear about. */
+	sbus_writel(BIGMAC_IMASK_GOTFRAME | BIGMAC_IMASK_SENTFRAME,
+		    bregs + BMAC_IMASK);
+
+	/* Enable the various other irq's. */
+	sbus_writel(0, cregs + CREG_RIMASK);
+	sbus_writel(0, cregs + CREG_TIMASK);
+	sbus_writel(0, cregs + CREG_QMASK);
+	sbus_writel(0, cregs + CREG_BMASK);
+
+	/* Set jam size to a reasonable default. */
+	sbus_writel(DEFAULT_JAMSIZE, bregs + BMAC_JSIZE);
+
+	/* Clear collision counter. */
+	sbus_writel(0, cregs + CREG_CCNT);
+
+	/* Enable transmitter and receiver. */
+	sbus_writel(sbus_readl(bregs + BMAC_TXCFG) | BIGMAC_TXCFG_ENABLE,
+		    bregs + BMAC_TXCFG);
+	sbus_writel(sbus_readl(bregs + BMAC_RXCFG) | BIGMAC_RXCFG_ENABLE,
+		    bregs + BMAC_RXCFG);
+
+	/* Ok, start detecting link speed/duplex. */
+	bigmac_begin_auto_negotiation(bp);
+
+	/* Success. */
+	return 0;
+}
+
+/* Error interrupts get sent here. */
+static void bigmac_is_medium_rare(struct bigmac *bp, u32 qec_status, u32 bmac_status)
+{
+	printk(KERN_ERR "bigmac_is_medium_rare: ");
+	if (qec_status & (GLOB_STAT_ER | GLOB_STAT_BM)) {
+		if (qec_status & GLOB_STAT_ER)
+			printk("QEC_ERROR, ");
+		if (qec_status & GLOB_STAT_BM)
+			printk("QEC_BMAC_ERROR, ");
+	}
+	if (bmac_status & CREG_STAT_ERRORS) {
+		if (bmac_status & CREG_STAT_BERROR)
+			printk("BMAC_ERROR, ");
+		if (bmac_status & CREG_STAT_TXDERROR)
+			printk("TXD_ERROR, ");
+		if (bmac_status & CREG_STAT_TXLERR)
+			printk("TX_LATE_ERROR, ");
+		if (bmac_status & CREG_STAT_TXPERR)
+			printk("TX_PARITY_ERROR, ");
+		if (bmac_status & CREG_STAT_TXSERR)
+			printk("TX_SBUS_ERROR, ");
+
+		if (bmac_status & CREG_STAT_RXDROP)
+			printk("RX_DROP_ERROR, ");
+
+		if (bmac_status & CREG_STAT_RXSMALL)
+			printk("RX_SMALL_ERROR, ");
+		if (bmac_status & CREG_STAT_RXLERR)
+			printk("RX_LATE_ERROR, ");
+		if (bmac_status & CREG_STAT_RXPERR)
+			printk("RX_PARITY_ERROR, ");
+		if (bmac_status & CREG_STAT_RXSERR)
+			printk("RX_SBUS_ERROR, ");
+	}
+
+	printk(" RESET\n");
+	bigmac_init_hw(bp, true);
+}
+
+/* BigMAC transmit complete service routines. */
+static void bigmac_tx(struct bigmac *bp)
+{
+	struct be_txd *txbase = &bp->bmac_block->be_txd[0];
+	struct net_device *dev = bp->dev;
+	int elem;
+
+	spin_lock(&bp->lock);
+
+	elem = bp->tx_old;
+	DTX(("bigmac_tx: tx_old[%d] ", elem));
+	while (elem != bp->tx_new) {
+		struct sk_buff *skb;
+		struct be_txd *this = &txbase[elem];
+
+		DTX(("this(%p) [flags(%08x)addr(%08x)]",
+		     this, this->tx_flags, this->tx_addr));
+
+		if (this->tx_flags & TXD_OWN)
+			break;
+		skb = bp->tx_skbs[elem];
+		dev->stats.tx_packets++;
+		dev->stats.tx_bytes += skb->len;
+		dma_unmap_single(&bp->bigmac_op->dev,
+				 this->tx_addr, skb->len,
+				 DMA_TO_DEVICE);
+
+		DTX(("skb(%p) ", skb));
+		bp->tx_skbs[elem] = NULL;
+		dev_consume_skb_irq(skb);
+
+		elem = NEXT_TX(elem);
+	}
+	DTX((" DONE, tx_old=%d\n", elem));
+	bp->tx_old = elem;
+
+	if (netif_queue_stopped(dev) &&
+	    TX_BUFFS_AVAIL(bp) > 0)
+		netif_wake_queue(bp->dev);
+
+	spin_unlock(&bp->lock);
+}
+
+/* BigMAC receive complete service routines. */
+static void bigmac_rx(struct bigmac *bp)
+{
+	struct be_rxd *rxbase = &bp->bmac_block->be_rxd[0];
+	struct be_rxd *this;
+	int elem = bp->rx_new, drops = 0;
+	u32 flags;
+
+	this = &rxbase[elem];
+	while (!((flags = this->rx_flags) & RXD_OWN)) {
+		struct sk_buff *skb;
+		int len = (flags & RXD_LENGTH); /* FCS not included */
+
+		/* Check for errors. */
+		if (len < ETH_ZLEN) {
+			bp->dev->stats.rx_errors++;
+			bp->dev->stats.rx_length_errors++;
+
+	drop_it:
+			/* Return it to the BigMAC. */
+			bp->dev->stats.rx_dropped++;
+			this->rx_flags =
+				(RXD_OWN | ((RX_BUF_ALLOC_SIZE - 34) & RXD_LENGTH));
+			goto next;
+		}
+		skb = bp->rx_skbs[elem];
+		if (len > RX_COPY_THRESHOLD) {
+			struct sk_buff *new_skb;
+
+			/* Now refill the entry, if we can. */
+			new_skb = big_mac_alloc_skb(RX_BUF_ALLOC_SIZE, GFP_ATOMIC);
+			if (new_skb == NULL) {
+				drops++;
+				goto drop_it;
+			}
+			dma_unmap_single(&bp->bigmac_op->dev,
+					 this->rx_addr,
+					 RX_BUF_ALLOC_SIZE - 34,
+					 DMA_FROM_DEVICE);
+			bp->rx_skbs[elem] = new_skb;
+			skb_put(new_skb, ETH_FRAME_LEN);
+			skb_reserve(new_skb, 34);
+			this->rx_addr =
+				dma_map_single(&bp->bigmac_op->dev,
+					       new_skb->data,
+					       RX_BUF_ALLOC_SIZE - 34,
+					       DMA_FROM_DEVICE);
+			this->rx_flags =
+				(RXD_OWN | ((RX_BUF_ALLOC_SIZE - 34) & RXD_LENGTH));
+
+			/* Trim the original skb for the netif. */
+			skb_trim(skb, len);
+		} else {
+			struct sk_buff *copy_skb = netdev_alloc_skb(bp->dev, len + 2);
+
+			if (copy_skb == NULL) {
+				drops++;
+				goto drop_it;
+			}
+			skb_reserve(copy_skb, 2);
+			skb_put(copy_skb, len);
+			dma_sync_single_for_cpu(&bp->bigmac_op->dev,
+						this->rx_addr, len,
+						DMA_FROM_DEVICE);
+			skb_copy_to_linear_data(copy_skb, (unsigned char *)skb->data, len);
+			dma_sync_single_for_device(&bp->bigmac_op->dev,
+						   this->rx_addr, len,
+						   DMA_FROM_DEVICE);
+
+			/* Reuse original ring buffer. */
+			this->rx_flags =
+				(RXD_OWN | ((RX_BUF_ALLOC_SIZE - 34) & RXD_LENGTH));
+
+			skb = copy_skb;
+		}
+
+		/* No checksums done by the BigMAC ;-( */
+		skb->protocol = eth_type_trans(skb, bp->dev);
+		netif_rx(skb);
+		bp->dev->stats.rx_packets++;
+		bp->dev->stats.rx_bytes += len;
+	next:
+		elem = NEXT_RX(elem);
+		this = &rxbase[elem];
+	}
+	bp->rx_new = elem;
+	if (drops)
+		printk(KERN_NOTICE "%s: Memory squeeze, deferring packet.\n", bp->dev->name);
+}
+
+static irqreturn_t bigmac_interrupt(int irq, void *dev_id)
+{
+	struct bigmac *bp = (struct bigmac *) dev_id;
+	u32 qec_status, bmac_status;
+
+	DIRQ(("bigmac_interrupt: "));
+
+	/* Latch status registers now. */
+	bmac_status = sbus_readl(bp->creg + CREG_STAT);
+	qec_status = sbus_readl(bp->gregs + GLOB_STAT);
+
+	DIRQ(("qec_status=%08x bmac_status=%08x\n", qec_status, bmac_status));
+	if ((qec_status & (GLOB_STAT_ER | GLOB_STAT_BM)) ||
+	   (bmac_status & CREG_STAT_ERRORS))
+		bigmac_is_medium_rare(bp, qec_status, bmac_status);
+
+	if (bmac_status & CREG_STAT_TXIRQ)
+		bigmac_tx(bp);
+
+	if (bmac_status & CREG_STAT_RXIRQ)
+		bigmac_rx(bp);
+
+	return IRQ_HANDLED;
+}
+
+static int bigmac_open(struct net_device *dev)
+{
+	struct bigmac *bp = netdev_priv(dev);
+	int ret;
+
+	ret = request_irq(dev->irq, bigmac_interrupt, IRQF_SHARED, dev->name, bp);
+	if (ret) {
+		printk(KERN_ERR "BIGMAC: Can't order irq %d to go.\n", dev->irq);
+		return ret;
+	}
+	timer_setup(&bp->bigmac_timer, bigmac_timer, 0);
+	ret = bigmac_init_hw(bp, false);
+	if (ret)
+		free_irq(dev->irq, bp);
+	return ret;
+}
+
+static int bigmac_close(struct net_device *dev)
+{
+	struct bigmac *bp = netdev_priv(dev);
+
+	timer_delete(&bp->bigmac_timer);
+	bp->timer_state = asleep;
+	bp->timer_ticks = 0;
+
+	bigmac_stop(bp);
+	bigmac_clean_rings(bp);
+	free_irq(dev->irq, bp);
+	return 0;
+}
+
+static void bigmac_tx_timeout(struct net_device *dev, unsigned int txqueue)
+{
+	struct bigmac *bp = netdev_priv(dev);
+
+	bigmac_init_hw(bp, true);
+	netif_wake_queue(dev);
+}
+
+/* Put a packet on the wire. */
+static netdev_tx_t
+bigmac_start_xmit(struct sk_buff *skb, struct net_device *dev)
+{
+	struct bigmac *bp = netdev_priv(dev);
+	int len, entry;
+	u32 mapping;
+
+	len = skb->len;
+	mapping = dma_map_single(&bp->bigmac_op->dev, skb->data,
+				 len, DMA_TO_DEVICE);
+
+	/* Avoid a race... */
+	spin_lock_irq(&bp->lock);
+	entry = bp->tx_new;
+	DTX(("bigmac_start_xmit: len(%d) entry(%d)\n", len, entry));
+	bp->bmac_block->be_txd[entry].tx_flags = TXD_UPDATE;
+	bp->tx_skbs[entry] = skb;
+	bp->bmac_block->be_txd[entry].tx_addr = mapping;
+	bp->bmac_block->be_txd[entry].tx_flags =
+		(TXD_OWN | TXD_SOP | TXD_EOP | (len & TXD_LENGTH));
+	bp->tx_new = NEXT_TX(entry);
+	if (TX_BUFFS_AVAIL(bp) <= 0)
+		netif_stop_queue(dev);
+	spin_unlock_irq(&bp->lock);
+
+	/* Get it going. */
+	sbus_writel(CREG_CTRL_TWAKEUP, bp->creg + CREG_CTRL);
+
+
+	return NETDEV_TX_OK;
+}
+
+static struct net_device_stats *bigmac_get_stats(struct net_device *dev)
+{
+	struct bigmac *bp = netdev_priv(dev);
+
+	bigmac_get_counters(bp, bp->bregs);
+	return &dev->stats;
+}
+
+static void bigmac_set_multicast(struct net_device *dev)
+{
+	struct bigmac *bp = netdev_priv(dev);
+	void __iomem *bregs = bp->bregs;
+	struct netdev_hw_addr *ha;
+	u32 tmp, crc;
+
+	/* Disable the receiver.  The bit self-clears when
+	 * the operation is complete.
+	 */
+	tmp = sbus_readl(bregs + BMAC_RXCFG);
+	tmp &= ~(BIGMAC_RXCFG_ENABLE);
+	sbus_writel(tmp, bregs + BMAC_RXCFG);
+	while ((sbus_readl(bregs + BMAC_RXCFG) & BIGMAC_RXCFG_ENABLE) != 0)
+		udelay(20);
+
+	if ((dev->flags & IFF_ALLMULTI) || (netdev_mc_count(dev) > 64)) {
+		sbus_writel(0xffff, bregs + BMAC_HTABLE0);
+		sbus_writel(0xffff, bregs + BMAC_HTABLE1);
+		sbus_writel(0xffff, bregs + BMAC_HTABLE2);
+		sbus_writel(0xffff, bregs + BMAC_HTABLE3);
+	} else if (dev->flags & IFF_PROMISC) {
+		tmp = sbus_readl(bregs + BMAC_RXCFG);
+		tmp |= BIGMAC_RXCFG_PMISC;
+		sbus_writel(tmp, bregs + BMAC_RXCFG);
+	} else {
+		u16 hash_table[4] = { 0 };
+
+		netdev_for_each_mc_addr(ha, dev) {
+			crc = ether_crc_le(6, ha->addr);
+			crc >>= 26;
+			hash_table[crc >> 4] |= 1 << (crc & 0xf);
+		}
+		sbus_writel(hash_table[0], bregs + BMAC_HTABLE0);
+		sbus_writel(hash_table[1], bregs + BMAC_HTABLE1);
+		sbus_writel(hash_table[2], bregs + BMAC_HTABLE2);
+		sbus_writel(hash_table[3], bregs + BMAC_HTABLE3);
+	}
+
+	/* Re-enable the receiver. */
+	tmp = sbus_readl(bregs + BMAC_RXCFG);
+	tmp |= BIGMAC_RXCFG_ENABLE;
+	sbus_writel(tmp, bregs + BMAC_RXCFG);
+}
+
+/* Ethtool support... */
+static void bigmac_get_drvinfo(struct net_device *dev, struct ethtool_drvinfo *info)
+{
+	strscpy(info->driver, "sunbmac", sizeof(info->driver));
+	strscpy(info->version, "2.0", sizeof(info->version));
+}
+
+static u32 bigmac_get_link(struct net_device *dev)
+{
+	struct bigmac *bp = netdev_priv(dev);
+
+	spin_lock_irq(&bp->lock);
+	bp->sw_bmsr = bigmac_tcvr_read(bp, bp->tregs, MII_BMSR);
+	spin_unlock_irq(&bp->lock);
+
+	return (bp->sw_bmsr & BMSR_LSTATUS);
+}
+
+static const struct ethtool_ops bigmac_ethtool_ops = {
+	.get_drvinfo		= bigmac_get_drvinfo,
+	.get_link		= bigmac_get_link,
+};
+
+static const struct net_device_ops bigmac_ops = {
+	.ndo_open		= bigmac_open,
+	.ndo_stop		= bigmac_close,
+	.ndo_start_xmit		= bigmac_start_xmit,
+	.ndo_get_stats		= bigmac_get_stats,
+	.ndo_set_rx_mode	= bigmac_set_multicast,
+	.ndo_tx_timeout		= bigmac_tx_timeout,
+	.ndo_set_mac_address	= eth_mac_addr,
+	.ndo_validate_addr	= eth_validate_addr,
+};
+
+static int bigmac_ether_init(struct platform_device *op,
+			     struct platform_device *qec_op)
+{
+	static int version_printed;
+	struct net_device *dev;
+	u8 bsizes, bsizes_more;
+	struct bigmac *bp;
+
+	/* Get a new device struct for this interface. */
+	dev = alloc_etherdev(sizeof(struct bigmac));
+	if (!dev)
+		return -ENOMEM;
+
+	if (version_printed++ == 0)
+		printk(KERN_INFO "%s", version);
+
+	eth_hw_addr_set(dev, idprom->id_ethaddr);
+
+	/* Setup softc, with backpointers to QEC and BigMAC SBUS device structs. */
+	bp = netdev_priv(dev);
+	bp->qec_op = qec_op;
+	bp->bigmac_op = op;
+
+	SET_NETDEV_DEV(dev, &op->dev);
+
+	spin_lock_init(&bp->lock);
+
+	/* Map in QEC global control registers. */
+	bp->gregs = of_ioremap(&qec_op->resource[0], 0,
+			       GLOB_REG_SIZE, "BigMAC QEC GLobal Regs");
+	if (!bp->gregs) {
+		printk(KERN_ERR "BIGMAC: Cannot map QEC global registers.\n");
+		goto fail_and_cleanup;
+	}
+
+	/* Make sure QEC is in BigMAC mode. */
+	if ((sbus_readl(bp->gregs + GLOB_CTRL) & 0xf0000000) != GLOB_CTRL_BMODE) {
+		printk(KERN_ERR "BigMAC: AIEEE, QEC is not in BigMAC mode!\n");
+		goto fail_and_cleanup;
+	}
+
+	/* Reset the QEC. */
+	if (qec_global_reset(bp->gregs))
+		goto fail_and_cleanup;
+
+	/* Get supported SBUS burst sizes. */
+	bsizes = of_getintprop_default(qec_op->dev.of_node, "burst-sizes", 0xff);
+	bsizes_more = of_getintprop_default(qec_op->dev.of_node, "burst-sizes", 0xff);
+
+	bsizes &= 0xff;
+	if (bsizes_more != 0xff)
+		bsizes &= bsizes_more;
+	if (bsizes == 0xff || (bsizes & DMA_BURST16) == 0 ||
+	    (bsizes & DMA_BURST32) == 0)
+		bsizes = (DMA_BURST32 - 1);
+	bp->bigmac_bursts = bsizes;
+
+	/* Perform QEC initialization. */
+	qec_init(bp);
+
+	/* Map in the BigMAC channel registers. */
+	bp->creg = of_ioremap(&op->resource[0], 0,
+			      CREG_REG_SIZE, "BigMAC QEC Channel Regs");
+	if (!bp->creg) {
+		printk(KERN_ERR "BIGMAC: Cannot map QEC channel registers.\n");
+		goto fail_and_cleanup;
+	}
+
+	/* Map in the BigMAC control registers. */
+	bp->bregs = of_ioremap(&op->resource[1], 0,
+			       BMAC_REG_SIZE, "BigMAC Primary Regs");
+	if (!bp->bregs) {
+		printk(KERN_ERR "BIGMAC: Cannot map BigMAC primary registers.\n");
+		goto fail_and_cleanup;
+	}
+
+	/* Map in the BigMAC transceiver registers, this is how you poke at
+	 * the BigMAC's PHY.
+	 */
+	bp->tregs = of_ioremap(&op->resource[2], 0,
+			       TCVR_REG_SIZE, "BigMAC Transceiver Regs");
+	if (!bp->tregs) {
+		printk(KERN_ERR "BIGMAC: Cannot map BigMAC transceiver registers.\n");
+		goto fail_and_cleanup;
+	}
+
+	/* Stop the BigMAC. */
+	bigmac_stop(bp);
+
+	/* Allocate transmit/receive descriptor DVMA block. */
+	bp->bmac_block = dma_alloc_coherent(&bp->bigmac_op->dev,
+					    PAGE_SIZE,
+					    &bp->bblock_dvma, GFP_ATOMIC);
+	if (bp->bmac_block == NULL || bp->bblock_dvma == 0)
+		goto fail_and_cleanup;
+
+	/* Get the board revision of this BigMAC. */
+	bp->board_rev = of_getintprop_default(bp->bigmac_op->dev.of_node,
+					      "board-version", 1);
+
+	/* Init auto-negotiation timer state. */
+	timer_setup(&bp->bigmac_timer, bigmac_timer, 0);
+	bp->timer_state = asleep;
+	bp->timer_ticks = 0;
+
+	/* Backlink to generic net device struct. */
+	bp->dev = dev;
+
+	/* Set links to our BigMAC open and close routines. */
+	dev->ethtool_ops = &bigmac_ethtool_ops;
+	dev->netdev_ops = &bigmac_ops;
+	dev->watchdog_timeo = 5*HZ;
+
+	/* Finish net device registration. */
+	dev->irq = bp->bigmac_op->archdata.irqs[0];
+	dev->dma = 0;
+
+	if (register_netdev(dev)) {
+		printk(KERN_ERR "BIGMAC: Cannot register device.\n");
+		goto fail_and_cleanup;
+	}
+
+	dev_set_drvdata(&bp->bigmac_op->dev, bp);
+
+	printk(KERN_INFO "%s: BigMAC 100baseT Ethernet %pM\n",
+	       dev->name, dev->dev_addr);
+
+	return 0;
+
+fail_and_cleanup:
+	/* Something went wrong, undo whatever we did so far. */
+	/* Free register mappings if any. */
+	if (bp->gregs)
+		of_iounmap(&qec_op->resource[0], bp->gregs, GLOB_REG_SIZE);
+	if (bp->creg)
+		of_iounmap(&op->resource[0], bp->creg, CREG_REG_SIZE);
+	if (bp->bregs)
+		of_iounmap(&op->resource[1], bp->bregs, BMAC_REG_SIZE);
+	if (bp->tregs)
+		of_iounmap(&op->resource[2], bp->tregs, TCVR_REG_SIZE);
+
+	if (bp->bmac_block)
+		dma_free_coherent(&bp->bigmac_op->dev,
+				  PAGE_SIZE,
+				  bp->bmac_block,
+				  bp->bblock_dvma);
+
+	/* This also frees the co-located private data */
+	free_netdev(dev);
+	return -ENODEV;
+}
+
+/* QEC can be the parent of either QuadEthernet or a BigMAC.  We want
+ * the latter.
+ */
+static int bigmac_sbus_probe(struct platform_device *op)
+{
+	struct device *parent = op->dev.parent;
+	struct platform_device *qec_op;
+
+	qec_op = to_platform_device(parent);
+
+	return bigmac_ether_init(op, qec_op);
+}
+
+static void bigmac_sbus_remove(struct platform_device *op)
+{
+	struct bigmac *bp = platform_get_drvdata(op);
+	struct device *parent = op->dev.parent;
+	struct net_device *net_dev = bp->dev;
+	struct platform_device *qec_op;
+
+	qec_op = to_platform_device(parent);
+
+	unregister_netdev(net_dev);
+
+	of_iounmap(&qec_op->resource[0], bp->gregs, GLOB_REG_SIZE);
+	of_iounmap(&op->resource[0], bp->creg, CREG_REG_SIZE);
+	of_iounmap(&op->resource[1], bp->bregs, BMAC_REG_SIZE);
+	of_iounmap(&op->resource[2], bp->tregs, TCVR_REG_SIZE);
+	dma_free_coherent(&op->dev,
+			  PAGE_SIZE,
+			  bp->bmac_block,
+			  bp->bblock_dvma);
+
+	free_netdev(net_dev);
+}
+
+static const struct of_device_id bigmac_sbus_match[] = {
+	{
+		.name = "be",
+	},
+	{},
+};
+
+MODULE_DEVICE_TABLE(of, bigmac_sbus_match);
+
+static struct platform_driver bigmac_sbus_driver = {
+	.driver = {
+		.name = "sunbmac",
+		.of_match_table = bigmac_sbus_match,
+	},
+	.probe		= bigmac_sbus_probe,
+	.remove		= bigmac_sbus_remove,
+};
+
+module_platform_driver(bigmac_sbus_driver);
diff --git a/drivers/net/ethernet/sun/sunbmac.h b/drivers/net/ethernet/sun/sunbmac.h
new file mode 100644
index 000000000..85778edcf
--- /dev/null
+++ b/drivers/net/ethernet/sun/sunbmac.h
@@ -0,0 +1,338 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/* $Id: sunbmac.h,v 1.7 2000/07/11 22:35:22 davem Exp $
+ * sunbmac.h: Defines for the Sun "Big MAC" 100baseT ethernet cards.
+ *
+ * Copyright (C) 1997 David S. Miller (davem@caip.rutgers.edu)
+ */
+
+#ifndef _SUNBMAC_H
+#define _SUNBMAC_H
+
+/* QEC global registers. */
+#define GLOB_CTRL	0x00UL	/* Control                  */
+#define GLOB_STAT	0x04UL	/* Status                   */
+#define GLOB_PSIZE	0x08UL	/* Packet Size              */
+#define GLOB_MSIZE	0x0cUL	/* Local-mem size (64K)     */
+#define GLOB_RSIZE	0x10UL	/* Receive partition size   */
+#define GLOB_TSIZE	0x14UL	/* Transmit partition size  */
+#define GLOB_REG_SIZE	0x18UL
+
+#define GLOB_CTRL_MMODE       0x40000000 /* MACE qec mode            */
+#define GLOB_CTRL_BMODE       0x10000000 /* BigMAC qec mode          */
+#define GLOB_CTRL_EPAR        0x00000020 /* Enable parity            */
+#define GLOB_CTRL_ACNTRL      0x00000018 /* SBUS arbitration control */
+#define GLOB_CTRL_B64         0x00000004 /* 64 byte dvma bursts      */
+#define GLOB_CTRL_B32         0x00000002 /* 32 byte dvma bursts      */
+#define GLOB_CTRL_B16         0x00000000 /* 16 byte dvma bursts      */
+#define GLOB_CTRL_RESET       0x00000001 /* Reset the QEC            */
+
+#define GLOB_STAT_TX          0x00000008 /* BigMAC Transmit IRQ      */
+#define GLOB_STAT_RX          0x00000004 /* BigMAC Receive IRQ       */
+#define GLOB_STAT_BM          0x00000002 /* BigMAC Global IRQ        */
+#define GLOB_STAT_ER          0x00000001 /* BigMAC Error IRQ         */
+
+#define GLOB_PSIZE_2048       0x00       /* 2k packet size           */
+#define GLOB_PSIZE_4096       0x01       /* 4k packet size           */
+#define GLOB_PSIZE_6144       0x10       /* 6k packet size           */
+#define GLOB_PSIZE_8192       0x11       /* 8k packet size           */
+
+/* QEC BigMAC channel registers. */
+#define CREG_CTRL	0x00UL	/* Control                   */
+#define CREG_STAT	0x04UL	/* Status                    */
+#define CREG_RXDS	0x08UL	/* RX descriptor ring ptr    */
+#define CREG_TXDS	0x0cUL	/* TX descriptor ring ptr    */
+#define CREG_RIMASK	0x10UL	/* RX Interrupt Mask         */
+#define CREG_TIMASK	0x14UL	/* TX Interrupt Mask         */
+#define CREG_QMASK	0x18UL	/* QEC Error Interrupt Mask  */
+#define CREG_BMASK	0x1cUL	/* BigMAC Error Interrupt Mask*/
+#define CREG_RXWBUFPTR	0x20UL	/* Local memory rx write ptr */
+#define CREG_RXRBUFPTR	0x24UL	/* Local memory rx read ptr  */
+#define CREG_TXWBUFPTR	0x28UL	/* Local memory tx write ptr */
+#define CREG_TXRBUFPTR	0x2cUL	/* Local memory tx read ptr  */
+#define CREG_CCNT	0x30UL	/* Collision Counter         */
+#define CREG_REG_SIZE	0x34UL
+
+#define CREG_CTRL_TWAKEUP     0x00000001  /* Transmitter Wakeup, 'go'. */
+
+#define CREG_STAT_BERROR      0x80000000  /* BigMAC error              */
+#define CREG_STAT_TXIRQ       0x00200000  /* Transmit Interrupt        */
+#define CREG_STAT_TXDERROR    0x00080000  /* TX Descriptor is bogus    */
+#define CREG_STAT_TXLERR      0x00040000  /* Late Transmit Error       */
+#define CREG_STAT_TXPERR      0x00020000  /* Transmit Parity Error     */
+#define CREG_STAT_TXSERR      0x00010000  /* Transmit SBUS error ack   */
+#define CREG_STAT_RXIRQ       0x00000020  /* Receive Interrupt         */
+#define CREG_STAT_RXDROP      0x00000010  /* Dropped a RX'd packet     */
+#define CREG_STAT_RXSMALL     0x00000008  /* Receive buffer too small  */
+#define CREG_STAT_RXLERR      0x00000004  /* Receive Late Error        */
+#define CREG_STAT_RXPERR      0x00000002  /* Receive Parity Error      */
+#define CREG_STAT_RXSERR      0x00000001  /* Receive SBUS Error ACK    */
+
+#define CREG_STAT_ERRORS      (CREG_STAT_BERROR|CREG_STAT_TXDERROR|CREG_STAT_TXLERR|   \
+                               CREG_STAT_TXPERR|CREG_STAT_TXSERR|CREG_STAT_RXDROP|     \
+                               CREG_STAT_RXSMALL|CREG_STAT_RXLERR|CREG_STAT_RXPERR|    \
+                               CREG_STAT_RXSERR)
+
+#define CREG_QMASK_TXDERROR   0x00080000  /* TXD error                 */
+#define CREG_QMASK_TXLERR     0x00040000  /* TX late error             */
+#define CREG_QMASK_TXPERR     0x00020000  /* TX parity error           */
+#define CREG_QMASK_TXSERR     0x00010000  /* TX sbus error ack         */
+#define CREG_QMASK_RXDROP     0x00000010  /* RX drop                   */
+#define CREG_QMASK_RXBERROR   0x00000008  /* RX buffer error           */
+#define CREG_QMASK_RXLEERR    0x00000004  /* RX late error             */
+#define CREG_QMASK_RXPERR     0x00000002  /* RX parity error           */
+#define CREG_QMASK_RXSERR     0x00000001  /* RX sbus error ack         */
+
+/* BIGMAC core registers */
+#define BMAC_XIFCFG	0x000UL	/* XIF config register                */
+	/* 0x004-->0x0fc, reserved */
+#define BMAC_STATUS	0x100UL	/* Status register, clear on read     */
+#define BMAC_IMASK	0x104UL	/* Interrupt mask register            */
+	/* 0x108-->0x204, reserved */
+#define BMAC_TXSWRESET	0x208UL	/* Transmitter software reset         */
+#define BMAC_TXCFG	0x20cUL	/* Transmitter config register        */
+#define BMAC_IGAP1	0x210UL	/* Inter-packet gap 1                 */
+#define BMAC_IGAP2	0x214UL	/* Inter-packet gap 2                 */
+#define BMAC_ALIMIT	0x218UL	/* Transmit attempt limit             */
+#define BMAC_STIME	0x21cUL	/* Transmit slot time                 */
+#define BMAC_PLEN	0x220UL	/* Size of transmit preamble          */
+#define BMAC_PPAT	0x224UL	/* Pattern for transmit preamble      */
+#define BMAC_TXDELIM	0x228UL	/* Transmit delimiter                 */
+#define BMAC_JSIZE	0x22cUL	/* Toe jam...                         */
+#define BMAC_TXPMAX	0x230UL	/* Transmit max pkt size              */
+#define BMAC_TXPMIN	0x234UL	/* Transmit min pkt size              */
+#define BMAC_PATTEMPT	0x238UL	/* Count of transmit peak attempts    */
+#define BMAC_DTCTR	0x23cUL	/* Transmit defer timer               */
+#define BMAC_NCCTR	0x240UL	/* Transmit normal-collision counter  */
+#define BMAC_FCCTR	0x244UL	/* Transmit first-collision counter   */
+#define BMAC_EXCTR	0x248UL	/* Transmit excess-collision counter  */
+#define BMAC_LTCTR	0x24cUL	/* Transmit late-collision counter    */
+#define BMAC_RSEED	0x250UL	/* Transmit random number seed        */
+#define BMAC_TXSMACHINE	0x254UL /* Transmit state machine             */
+	/* 0x258-->0x304, reserved */
+#define BMAC_RXSWRESET	0x308UL	/* Receiver software reset            */
+#define BMAC_RXCFG	0x30cUL	/* Receiver config register           */
+#define BMAC_RXPMAX	0x310UL	/* Receive max pkt size               */
+#define BMAC_RXPMIN	0x314UL	/* Receive min pkt size               */
+#define BMAC_MACADDR2	0x318UL	/* Ether address register 2           */
+#define BMAC_MACADDR1	0x31cUL	/* Ether address register 1           */
+#define BMAC_MACADDR0	0x320UL	/* Ether address register 0           */
+#define BMAC_FRCTR	0x324UL	/* Receive frame receive counter      */
+#define BMAC_GLECTR	0x328UL	/* Receive giant-length error counter */
+#define BMAC_UNALECTR	0x32cUL	/* Receive unaligned error counter    */
+#define BMAC_RCRCECTR	0x330UL	/* Receive CRC error counter          */
+#define BMAC_RXSMACHINE	0x334UL	/* Receiver state machine             */
+#define BMAC_RXCVALID	0x338UL	/* Receiver code violation            */
+	/* 0x33c, reserved */
+#define BMAC_HTABLE3	0x340UL	/* Hash table 3                       */
+#define BMAC_HTABLE2	0x344UL	/* Hash table 2                       */
+#define BMAC_HTABLE1	0x348UL	/* Hash table 1                       */
+#define BMAC_HTABLE0	0x34cUL	/* Hash table 0                       */
+#define BMAC_AFILTER2	0x350UL	/* Address filter 2                   */
+#define BMAC_AFILTER1	0x354UL	/* Address filter 1                   */
+#define BMAC_AFILTER0	0x358UL	/* Address filter 0                   */
+#define BMAC_AFMASK	0x35cUL	/* Address filter mask                */
+#define BMAC_REG_SIZE	0x360UL
+
+/* BigMac XIF config register. */
+#define BIGMAC_XCFG_ODENABLE   0x00000001 /* Output driver enable                     */
+#define BIGMAC_XCFG_RESV       0x00000002 /* Reserved, write always as 1              */
+#define BIGMAC_XCFG_MLBACK     0x00000004 /* Loopback-mode MII enable                 */
+#define BIGMAC_XCFG_SMODE      0x00000008 /* Enable serial mode                       */
+
+/* BigMAC status register. */
+#define BIGMAC_STAT_GOTFRAME   0x00000001 /* Received a frame                         */
+#define BIGMAC_STAT_RCNTEXP    0x00000002 /* Receive frame counter expired            */
+#define BIGMAC_STAT_ACNTEXP    0x00000004 /* Align-error counter expired              */
+#define BIGMAC_STAT_CCNTEXP    0x00000008 /* CRC-error counter expired                */
+#define BIGMAC_STAT_LCNTEXP    0x00000010 /* Length-error counter expired             */
+#define BIGMAC_STAT_RFIFOVF    0x00000020 /* Receive FIFO overflow                    */
+#define BIGMAC_STAT_CVCNTEXP   0x00000040 /* Code-violation counter expired           */
+#define BIGMAC_STAT_SENTFRAME  0x00000100 /* Transmitted a frame                      */
+#define BIGMAC_STAT_TFIFO_UND  0x00000200 /* Transmit FIFO underrun                   */
+#define BIGMAC_STAT_MAXPKTERR  0x00000400 /* Max-packet size error                    */
+#define BIGMAC_STAT_NCNTEXP    0x00000800 /* Normal-collision counter expired         */
+#define BIGMAC_STAT_ECNTEXP    0x00001000 /* Excess-collision counter expired         */
+#define BIGMAC_STAT_LCCNTEXP   0x00002000 /* Late-collision counter expired           */
+#define BIGMAC_STAT_FCNTEXP    0x00004000 /* First-collision counter expired          */
+#define BIGMAC_STAT_DTIMEXP    0x00008000 /* Defer-timer expired                      */
+
+/* BigMAC interrupt mask register. */
+#define BIGMAC_IMASK_GOTFRAME  0x00000001 /* Received a frame                         */
+#define BIGMAC_IMASK_RCNTEXP   0x00000002 /* Receive frame counter expired            */
+#define BIGMAC_IMASK_ACNTEXP   0x00000004 /* Align-error counter expired              */
+#define BIGMAC_IMASK_CCNTEXP   0x00000008 /* CRC-error counter expired                */
+#define BIGMAC_IMASK_LCNTEXP   0x00000010 /* Length-error counter expired             */
+#define BIGMAC_IMASK_RFIFOVF   0x00000020 /* Receive FIFO overflow                    */
+#define BIGMAC_IMASK_CVCNTEXP  0x00000040 /* Code-violation counter expired           */
+#define BIGMAC_IMASK_SENTFRAME 0x00000100 /* Transmitted a frame                      */
+#define BIGMAC_IMASK_TFIFO_UND 0x00000200 /* Transmit FIFO underrun                   */
+#define BIGMAC_IMASK_MAXPKTERR 0x00000400 /* Max-packet size error                    */
+#define BIGMAC_IMASK_NCNTEXP   0x00000800 /* Normal-collision counter expired         */
+#define BIGMAC_IMASK_ECNTEXP   0x00001000 /* Excess-collision counter expired         */
+#define BIGMAC_IMASK_LCCNTEXP  0x00002000 /* Late-collision counter expired           */
+#define BIGMAC_IMASK_FCNTEXP   0x00004000 /* First-collision counter expired          */
+#define BIGMAC_IMASK_DTIMEXP   0x00008000 /* Defer-timer expired                      */
+
+/* BigMac transmit config register. */
+#define BIGMAC_TXCFG_ENABLE    0x00000001 /* Enable the transmitter                   */
+#define BIGMAC_TXCFG_FIFO      0x00000010 /* Default tx fthresh...                    */
+#define BIGMAC_TXCFG_SMODE     0x00000020 /* Enable slow transmit mode                */
+#define BIGMAC_TXCFG_CIGN      0x00000040 /* Ignore transmit collisions               */
+#define BIGMAC_TXCFG_FCSOFF    0x00000080 /* Do not emit FCS                          */
+#define BIGMAC_TXCFG_DBACKOFF  0x00000100 /* Disable backoff                          */
+#define BIGMAC_TXCFG_FULLDPLX  0x00000200 /* Enable full-duplex                       */
+
+/* BigMac receive config register. */
+#define BIGMAC_RXCFG_ENABLE    0x00000001 /* Enable the receiver                      */
+#define BIGMAC_RXCFG_FIFO      0x0000000e /* Default rx fthresh...                    */
+#define BIGMAC_RXCFG_PSTRIP    0x00000020 /* Pad byte strip enable                    */
+#define BIGMAC_RXCFG_PMISC     0x00000040 /* Enable promiscuous mode                   */
+#define BIGMAC_RXCFG_DERR      0x00000080 /* Disable error checking                   */
+#define BIGMAC_RXCFG_DCRCS     0x00000100 /* Disable CRC stripping                    */
+#define BIGMAC_RXCFG_ME        0x00000200 /* Receive packets addressed to me          */
+#define BIGMAC_RXCFG_PGRP      0x00000400 /* Enable promisc group mode                */
+#define BIGMAC_RXCFG_HENABLE   0x00000800 /* Enable the hash filter                   */
+#define BIGMAC_RXCFG_AENABLE   0x00001000 /* Enable the address filter                */
+
+/* The BigMAC PHY transceiver.  Not nearly as sophisticated as the happy meal
+ * one.  But it does have the "bit banger", oh baby.
+ */
+#define TCVR_TPAL	0x00UL
+#define TCVR_MPAL	0x04UL
+#define TCVR_REG_SIZE	0x08UL
+
+/* Frame commands. */
+#define FRAME_WRITE           0x50020000
+#define FRAME_READ            0x60020000
+
+/* Transceiver registers. */
+#define TCVR_PAL_SERIAL       0x00000001 /* Enable serial mode              */
+#define TCVR_PAL_EXTLBACK     0x00000002 /* Enable external loopback        */
+#define TCVR_PAL_MSENSE       0x00000004 /* Media sense                     */
+#define TCVR_PAL_LTENABLE     0x00000008 /* Link test enable                */
+#define TCVR_PAL_LTSTATUS     0x00000010 /* Link test status  (P1 only)     */
+
+/* Management PAL. */
+#define MGMT_PAL_DCLOCK       0x00000001 /* Data clock                      */
+#define MGMT_PAL_OENAB        0x00000002 /* Output enabler                  */
+#define MGMT_PAL_MDIO         0x00000004 /* MDIO Data/attached              */
+#define MGMT_PAL_TIMEO        0x00000008 /* Transmit enable timeout error   */
+#define MGMT_PAL_EXT_MDIO     MGMT_PAL_MDIO
+#define MGMT_PAL_INT_MDIO     MGMT_PAL_TIMEO
+
+/* Here are some PHY addresses. */
+#define BIGMAC_PHY_EXTERNAL   0 /* External transceiver */
+#define BIGMAC_PHY_INTERNAL   1 /* Internal transceiver */
+
+/* Ring descriptors and such, same as Quad Ethernet. */
+struct be_rxd {
+	u32 rx_flags;
+	u32 rx_addr;
+};
+
+#define RXD_OWN      0x80000000 /* Ownership.      */
+#define RXD_UPDATE   0x10000000 /* Being Updated?  */
+#define RXD_LENGTH   0x000007ff /* Packet Length.  */
+
+struct be_txd {
+	u32 tx_flags;
+	u32 tx_addr;
+};
+
+#define TXD_OWN      0x80000000 /* Ownership.      */
+#define TXD_SOP      0x40000000 /* Start Of Packet */
+#define TXD_EOP      0x20000000 /* End Of Packet   */
+#define TXD_UPDATE   0x10000000 /* Being Updated?  */
+#define TXD_LENGTH   0x000007ff /* Packet Length.  */
+
+#define TX_RING_MAXSIZE   256
+#define RX_RING_MAXSIZE   256
+
+#define TX_RING_SIZE      256
+#define RX_RING_SIZE      256
+
+#define NEXT_RX(num)       (((num) + 1) & (RX_RING_SIZE - 1))
+#define NEXT_TX(num)       (((num) + 1) & (TX_RING_SIZE - 1))
+#define PREV_RX(num)       (((num) - 1) & (RX_RING_SIZE - 1))
+#define PREV_TX(num)       (((num) - 1) & (TX_RING_SIZE - 1))
+
+#define TX_BUFFS_AVAIL(bp)                                    \
+        (((bp)->tx_old <= (bp)->tx_new) ?                     \
+	  (bp)->tx_old + (TX_RING_SIZE - 1) - (bp)->tx_new :  \
+			    (bp)->tx_old - (bp)->tx_new - 1)
+
+
+#define RX_COPY_THRESHOLD  256
+#define RX_BUF_ALLOC_SIZE  (ETH_FRAME_LEN + (64 * 3))
+
+struct bmac_init_block {
+	struct be_rxd be_rxd[RX_RING_MAXSIZE];
+	struct be_txd be_txd[TX_RING_MAXSIZE];
+};
+
+#define bib_offset(mem, elem) \
+((__u32)((unsigned long)(&(((struct bmac_init_block *)0)->mem[elem]))))
+
+/* Now software state stuff. */
+enum bigmac_transceiver {
+	external = 0,
+	internal = 1,
+	none     = 2,
+};
+
+/* Timer state engine. */
+enum bigmac_timer_state {
+	ltrywait = 1,  /* Forcing try of all modes, from fastest to slowest. */
+	asleep   = 2,  /* Timer inactive.                                    */
+};
+
+struct bigmac {
+	void __iomem	*gregs;	/* QEC Global Registers               */
+	void __iomem	*creg;	/* QEC BigMAC Channel Registers       */
+	void __iomem	*bregs;	/* BigMAC Registers                   */
+	void __iomem	*tregs;	/* BigMAC Transceiver                 */
+	struct bmac_init_block	*bmac_block;	/* RX and TX descriptors */
+	dma_addr_t		bblock_dvma;	/* RX and TX descriptors */
+
+	spinlock_t		lock;
+
+	struct sk_buff		*rx_skbs[RX_RING_SIZE];
+	struct sk_buff		*tx_skbs[TX_RING_SIZE];
+
+	int rx_new, tx_new, rx_old, tx_old;
+
+	int board_rev;				/* BigMAC board revision.             */
+
+	enum bigmac_transceiver	tcvr_type;
+	unsigned int		bigmac_bursts;
+	unsigned int		paddr;
+	unsigned short		sw_bmsr;         /* SW copy of PHY BMSR               */
+	unsigned short		sw_bmcr;         /* SW copy of PHY BMCR               */
+	struct timer_list	bigmac_timer;
+	enum bigmac_timer_state	timer_state;
+	unsigned int		timer_ticks;
+
+	struct platform_device	*qec_op;
+	struct platform_device	*bigmac_op;
+	struct net_device	*dev;
+};
+
+/* We use this to acquire receive skb's that we can DMA directly into. */
+#define ALIGNED_RX_SKB_ADDR(addr) \
+        ((((unsigned long)(addr) + (64 - 1)) & ~(64 - 1)) - (unsigned long)(addr))
+
+static inline struct sk_buff *big_mac_alloc_skb(unsigned int length, gfp_t gfp_flags)
+{
+	struct sk_buff *skb;
+
+	skb = alloc_skb(length + 64, gfp_flags);
+	if(skb) {
+		int offset = ALIGNED_RX_SKB_ADDR(skb->data);
+
+		if(offset)
+			skb_reserve(skb, offset);
+	}
+	return skb;
+}
+
+#endif /* !(_SUNBMAC_H) */
diff --git a/drivers/net/ethernet/sun/sungem.c b/drivers/net/ethernet/sun/sungem.c
new file mode 100644
index 000000000..dc5638d10
--- /dev/null
+++ b/drivers/net/ethernet/sun/sungem.c
@@ -0,0 +1,3030 @@
+// SPDX-License-Identifier: GPL-2.0
+/* $Id: sungem.c,v 1.44.2.22 2002/03/13 01:18:12 davem Exp $
+ * sungem.c: Sun GEM ethernet driver.
+ *
+ * Copyright (C) 2000, 2001, 2002, 2003 David S. Miller (davem@redhat.com)
+ *
+ * Support for Apple GMAC and assorted PHYs, WOL, Power Management
+ * (C) 2001,2002,2003 Benjamin Herrenscmidt (benh@kernel.crashing.org)
+ * (C) 2004,2005 Benjamin Herrenscmidt, IBM Corp.
+ *
+ * NAPI and NETPOLL support
+ * (C) 2004 by Eric Lemoine (eric.lemoine@gmail.com)
+ *
+ */
+
+#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
+
+#include <linux/module.h>
+#include <linux/kernel.h>
+#include <linux/types.h>
+#include <linux/fcntl.h>
+#include <linux/interrupt.h>
+#include <linux/ioport.h>
+#include <linux/in.h>
+#include <linux/sched.h>
+#include <linux/string.h>
+#include <linux/delay.h>
+#include <linux/errno.h>
+#include <linux/pci.h>
+#include <linux/dma-mapping.h>
+#include <linux/netdevice.h>
+#include <linux/etherdevice.h>
+#include <linux/skbuff.h>
+#include <linux/mii.h>
+#include <linux/ethtool.h>
+#include <linux/crc32.h>
+#include <linux/random.h>
+#include <linux/workqueue.h>
+#include <linux/if_vlan.h>
+#include <linux/bitops.h>
+#include <linux/mm.h>
+#include <linux/gfp.h>
+#include <linux/of.h>
+
+#include <asm/io.h>
+#include <asm/byteorder.h>
+#include <linux/uaccess.h>
+#include <asm/irq.h>
+
+#ifdef CONFIG_SPARC
+#include <asm/idprom.h>
+#include <asm/prom.h>
+#endif
+
+#ifdef CONFIG_PPC_PMAC
+#include <asm/machdep.h>
+#include <asm/pmac_feature.h>
+#endif
+
+#include <linux/sungem_phy.h>
+#include "sungem.h"
+
+#define STRIP_FCS
+
+#define DEFAULT_MSG	(NETIF_MSG_DRV		| \
+			 NETIF_MSG_PROBE	| \
+			 NETIF_MSG_LINK)
+
+#define ADVERTISE_MASK	(SUPPORTED_10baseT_Half | SUPPORTED_10baseT_Full | \
+			 SUPPORTED_100baseT_Half | SUPPORTED_100baseT_Full | \
+			 SUPPORTED_1000baseT_Half | SUPPORTED_1000baseT_Full | \
+			 SUPPORTED_Pause | SUPPORTED_Autoneg)
+
+#define DRV_NAME	"sungem"
+#define DRV_VERSION	"1.0"
+#define DRV_AUTHOR	"David S. Miller <davem@redhat.com>"
+
+static char version[] =
+        DRV_NAME ".c:v" DRV_VERSION " " DRV_AUTHOR "\n";
+
+MODULE_AUTHOR(DRV_AUTHOR);
+MODULE_DESCRIPTION("Sun GEM Gbit ethernet driver");
+MODULE_LICENSE("GPL");
+
+#define GEM_MODULE_NAME	"gem"
+
+static const struct pci_device_id gem_pci_tbl[] = {
+	{ PCI_VDEVICE(SUN, PCI_DEVICE_ID_SUN_GEM) },
+
+	/* These models only differ from the original GEM in
+	 * that their tx/rx fifos are of a different size and
+	 * they only support 10/100 speeds. -DaveM
+	 *
+	 * Apple's GMAC does support gigabit on machines with
+	 * the BCM54xx PHYs. -BenH
+	 */
+	{ PCI_VDEVICE(SUN, PCI_DEVICE_ID_SUN_RIO_GEM) },
+	{ PCI_VDEVICE(APPLE, PCI_DEVICE_ID_APPLE_UNI_N_GMAC) },
+	{ PCI_VDEVICE(APPLE, PCI_DEVICE_ID_APPLE_UNI_N_GMACP) },
+	{ PCI_VDEVICE(APPLE, PCI_DEVICE_ID_APPLE_UNI_N_GMAC2) },
+	{ PCI_VDEVICE(APPLE, PCI_DEVICE_ID_APPLE_K2_GMAC) },
+	{ PCI_VDEVICE(APPLE, PCI_DEVICE_ID_APPLE_SH_SUNGEM) },
+	{ PCI_VDEVICE(APPLE, PCI_DEVICE_ID_APPLE_IPID2_GMAC) },
+	{ }
+};
+
+MODULE_DEVICE_TABLE(pci, gem_pci_tbl);
+
+static u16 __sungem_phy_read(struct gem *gp, int phy_addr, int reg)
+{
+	u32 cmd;
+	int limit = 10000;
+
+	cmd  = (1 << 30);
+	cmd |= (2 << 28);
+	cmd |= (phy_addr << 23) & MIF_FRAME_PHYAD;
+	cmd |= (reg << 18) & MIF_FRAME_REGAD;
+	cmd |= (MIF_FRAME_TAMSB);
+	writel(cmd, gp->regs + MIF_FRAME);
+
+	while (--limit) {
+		cmd = readl(gp->regs + MIF_FRAME);
+		if (cmd & MIF_FRAME_TALSB)
+			break;
+
+		udelay(10);
+	}
+
+	if (!limit)
+		cmd = 0xffff;
+
+	return cmd & MIF_FRAME_DATA;
+}
+
+static inline int _sungem_phy_read(struct net_device *dev, int mii_id, int reg)
+{
+	struct gem *gp = netdev_priv(dev);
+	return __sungem_phy_read(gp, mii_id, reg);
+}
+
+static inline u16 sungem_phy_read(struct gem *gp, int reg)
+{
+	return __sungem_phy_read(gp, gp->mii_phy_addr, reg);
+}
+
+static void __sungem_phy_write(struct gem *gp, int phy_addr, int reg, u16 val)
+{
+	u32 cmd;
+	int limit = 10000;
+
+	cmd  = (1 << 30);
+	cmd |= (1 << 28);
+	cmd |= (phy_addr << 23) & MIF_FRAME_PHYAD;
+	cmd |= (reg << 18) & MIF_FRAME_REGAD;
+	cmd |= (MIF_FRAME_TAMSB);
+	cmd |= (val & MIF_FRAME_DATA);
+	writel(cmd, gp->regs + MIF_FRAME);
+
+	while (limit--) {
+		cmd = readl(gp->regs + MIF_FRAME);
+		if (cmd & MIF_FRAME_TALSB)
+			break;
+
+		udelay(10);
+	}
+}
+
+static inline void _sungem_phy_write(struct net_device *dev, int mii_id, int reg, int val)
+{
+	struct gem *gp = netdev_priv(dev);
+	__sungem_phy_write(gp, mii_id, reg, val & 0xffff);
+}
+
+static inline void sungem_phy_write(struct gem *gp, int reg, u16 val)
+{
+	__sungem_phy_write(gp, gp->mii_phy_addr, reg, val);
+}
+
+static inline void gem_enable_ints(struct gem *gp)
+{
+	/* Enable all interrupts but TXDONE */
+	writel(GREG_STAT_TXDONE, gp->regs + GREG_IMASK);
+}
+
+static inline void gem_disable_ints(struct gem *gp)
+{
+	/* Disable all interrupts, including TXDONE */
+	writel(GREG_STAT_NAPI | GREG_STAT_TXDONE, gp->regs + GREG_IMASK);
+	(void)readl(gp->regs + GREG_IMASK); /* write posting */
+}
+
+static void gem_get_cell(struct gem *gp)
+{
+	BUG_ON(gp->cell_enabled < 0);
+	gp->cell_enabled++;
+#ifdef CONFIG_PPC_PMAC
+	if (gp->cell_enabled == 1) {
+		mb();
+		pmac_call_feature(PMAC_FTR_GMAC_ENABLE, gp->of_node, 0, 1);
+		udelay(10);
+	}
+#endif /* CONFIG_PPC_PMAC */
+}
+
+/* Turn off the chip's clock */
+static void gem_put_cell(struct gem *gp)
+{
+	BUG_ON(gp->cell_enabled <= 0);
+	gp->cell_enabled--;
+#ifdef CONFIG_PPC_PMAC
+	if (gp->cell_enabled == 0) {
+		mb();
+		pmac_call_feature(PMAC_FTR_GMAC_ENABLE, gp->of_node, 0, 0);
+		udelay(10);
+	}
+#endif /* CONFIG_PPC_PMAC */
+}
+
+static inline void gem_netif_stop(struct gem *gp)
+{
+	netif_trans_update(gp->dev);	/* prevent tx timeout */
+	napi_disable(&gp->napi);
+	netif_tx_disable(gp->dev);
+}
+
+static inline void gem_netif_start(struct gem *gp)
+{
+	/* NOTE: unconditional netif_wake_queue is only
+	 * appropriate so long as all callers are assured to
+	 * have free tx slots.
+	 */
+	netif_wake_queue(gp->dev);
+	napi_enable(&gp->napi);
+}
+
+static void gem_schedule_reset(struct gem *gp)
+{
+	gp->reset_task_pending = 1;
+	schedule_work(&gp->reset_task);
+}
+
+static void gem_handle_mif_event(struct gem *gp, u32 reg_val, u32 changed_bits)
+{
+	if (netif_msg_intr(gp))
+		printk(KERN_DEBUG "%s: mif interrupt\n", gp->dev->name);
+}
+
+static int gem_pcs_interrupt(struct net_device *dev, struct gem *gp, u32 gem_status)
+{
+	u32 pcs_istat = readl(gp->regs + PCS_ISTAT);
+	u32 pcs_miistat;
+
+	if (netif_msg_intr(gp))
+		printk(KERN_DEBUG "%s: pcs interrupt, pcs_istat: 0x%x\n",
+			gp->dev->name, pcs_istat);
+
+	if (!(pcs_istat & PCS_ISTAT_LSC)) {
+		netdev_err(dev, "PCS irq but no link status change???\n");
+		return 0;
+	}
+
+	/* The link status bit latches on zero, so you must
+	 * read it twice in such a case to see a transition
+	 * to the link being up.
+	 */
+	pcs_miistat = readl(gp->regs + PCS_MIISTAT);
+	if (!(pcs_miistat & PCS_MIISTAT_LS))
+		pcs_miistat |=
+			(readl(gp->regs + PCS_MIISTAT) &
+			 PCS_MIISTAT_LS);
+
+	if (pcs_miistat & PCS_MIISTAT_ANC) {
+		/* The remote-fault indication is only valid
+		 * when autoneg has completed.
+		 */
+		if (pcs_miistat & PCS_MIISTAT_RF)
+			netdev_info(dev, "PCS AutoNEG complete, RemoteFault\n");
+		else
+			netdev_info(dev, "PCS AutoNEG complete\n");
+	}
+
+	if (pcs_miistat & PCS_MIISTAT_LS) {
+		netdev_info(dev, "PCS link is now up\n");
+		netif_carrier_on(gp->dev);
+	} else {
+		netdev_info(dev, "PCS link is now down\n");
+		netif_carrier_off(gp->dev);
+		/* If this happens and the link timer is not running,
+		 * reset so we re-negotiate.
+		 */
+		if (!timer_pending(&gp->link_timer))
+			return 1;
+	}
+
+	return 0;
+}
+
+static int gem_txmac_interrupt(struct net_device *dev, struct gem *gp, u32 gem_status)
+{
+	u32 txmac_stat = readl(gp->regs + MAC_TXSTAT);
+
+	if (netif_msg_intr(gp))
+		printk(KERN_DEBUG "%s: txmac interrupt, txmac_stat: 0x%x\n",
+			gp->dev->name, txmac_stat);
+
+	/* Defer timer expiration is quite normal,
+	 * don't even log the event.
+	 */
+	if ((txmac_stat & MAC_TXSTAT_DTE) &&
+	    !(txmac_stat & ~MAC_TXSTAT_DTE))
+		return 0;
+
+	if (txmac_stat & MAC_TXSTAT_URUN) {
+		netdev_err(dev, "TX MAC xmit underrun\n");
+		dev->stats.tx_fifo_errors++;
+	}
+
+	if (txmac_stat & MAC_TXSTAT_MPE) {
+		netdev_err(dev, "TX MAC max packet size error\n");
+		dev->stats.tx_errors++;
+	}
+
+	/* The rest are all cases of one of the 16-bit TX
+	 * counters expiring.
+	 */
+	if (txmac_stat & MAC_TXSTAT_NCE)
+		dev->stats.collisions += 0x10000;
+
+	if (txmac_stat & MAC_TXSTAT_ECE) {
+		dev->stats.tx_aborted_errors += 0x10000;
+		dev->stats.collisions += 0x10000;
+	}
+
+	if (txmac_stat & MAC_TXSTAT_LCE) {
+		dev->stats.tx_aborted_errors += 0x10000;
+		dev->stats.collisions += 0x10000;
+	}
+
+	/* We do not keep track of MAC_TXSTAT_FCE and
+	 * MAC_TXSTAT_PCE events.
+	 */
+	return 0;
+}
+
+/* When we get a RX fifo overflow, the RX unit in GEM is probably hung
+ * so we do the following.
+ *
+ * If any part of the reset goes wrong, we return 1 and that causes the
+ * whole chip to be reset.
+ */
+static int gem_rxmac_reset(struct gem *gp)
+{
+	struct net_device *dev = gp->dev;
+	int limit, i;
+	u64 desc_dma;
+	u32 val;
+
+	/* First, reset & disable MAC RX. */
+	writel(MAC_RXRST_CMD, gp->regs + MAC_RXRST);
+	for (limit = 0; limit < 5000; limit++) {
+		if (!(readl(gp->regs + MAC_RXRST) & MAC_RXRST_CMD))
+			break;
+		udelay(10);
+	}
+	if (limit == 5000) {
+		netdev_err(dev, "RX MAC will not reset, resetting whole chip\n");
+		return 1;
+	}
+
+	writel(gp->mac_rx_cfg & ~MAC_RXCFG_ENAB,
+	       gp->regs + MAC_RXCFG);
+	for (limit = 0; limit < 5000; limit++) {
+		if (!(readl(gp->regs + MAC_RXCFG) & MAC_RXCFG_ENAB))
+			break;
+		udelay(10);
+	}
+	if (limit == 5000) {
+		netdev_err(dev, "RX MAC will not disable, resetting whole chip\n");
+		return 1;
+	}
+
+	/* Second, disable RX DMA. */
+	writel(0, gp->regs + RXDMA_CFG);
+	for (limit = 0; limit < 5000; limit++) {
+		if (!(readl(gp->regs + RXDMA_CFG) & RXDMA_CFG_ENABLE))
+			break;
+		udelay(10);
+	}
+	if (limit == 5000) {
+		netdev_err(dev, "RX DMA will not disable, resetting whole chip\n");
+		return 1;
+	}
+
+	mdelay(5);
+
+	/* Execute RX reset command. */
+	writel(gp->swrst_base | GREG_SWRST_RXRST,
+	       gp->regs + GREG_SWRST);
+	for (limit = 0; limit < 5000; limit++) {
+		if (!(readl(gp->regs + GREG_SWRST) & GREG_SWRST_RXRST))
+			break;
+		udelay(10);
+	}
+	if (limit == 5000) {
+		netdev_err(dev, "RX reset command will not execute, resetting whole chip\n");
+		return 1;
+	}
+
+	/* Refresh the RX ring. */
+	for (i = 0; i < RX_RING_SIZE; i++) {
+		struct gem_rxd *rxd = &gp->init_block->rxd[i];
+
+		if (gp->rx_skbs[i] == NULL) {
+			netdev_err(dev, "Parts of RX ring empty, resetting whole chip\n");
+			return 1;
+		}
+
+		rxd->status_word = cpu_to_le64(RXDCTRL_FRESH(gp));
+	}
+	gp->rx_new = gp->rx_old = 0;
+
+	/* Now we must reprogram the rest of RX unit. */
+	desc_dma = (u64) gp->gblock_dvma;
+	desc_dma += (INIT_BLOCK_TX_RING_SIZE * sizeof(struct gem_txd));
+	writel(desc_dma >> 32, gp->regs + RXDMA_DBHI);
+	writel(desc_dma & 0xffffffff, gp->regs + RXDMA_DBLOW);
+	writel(RX_RING_SIZE - 4, gp->regs + RXDMA_KICK);
+	val = (RXDMA_CFG_BASE | (RX_OFFSET << 10) |
+	       (ETH_HLEN << 13) | RXDMA_CFG_FTHRESH_128);
+	writel(val, gp->regs + RXDMA_CFG);
+	if (readl(gp->regs + GREG_BIFCFG) & GREG_BIFCFG_M66EN)
+		writel(((5 & RXDMA_BLANK_IPKTS) |
+			((8 << 12) & RXDMA_BLANK_ITIME)),
+		       gp->regs + RXDMA_BLANK);
+	else
+		writel(((5 & RXDMA_BLANK_IPKTS) |
+			((4 << 12) & RXDMA_BLANK_ITIME)),
+		       gp->regs + RXDMA_BLANK);
+	val  = (((gp->rx_pause_off / 64) << 0) & RXDMA_PTHRESH_OFF);
+	val |= (((gp->rx_pause_on / 64) << 12) & RXDMA_PTHRESH_ON);
+	writel(val, gp->regs + RXDMA_PTHRESH);
+	val = readl(gp->regs + RXDMA_CFG);
+	writel(val | RXDMA_CFG_ENABLE, gp->regs + RXDMA_CFG);
+	writel(MAC_RXSTAT_RCV, gp->regs + MAC_RXMASK);
+	val = readl(gp->regs + MAC_RXCFG);
+	writel(val | MAC_RXCFG_ENAB, gp->regs + MAC_RXCFG);
+
+	return 0;
+}
+
+static int gem_rxmac_interrupt(struct net_device *dev, struct gem *gp, u32 gem_status)
+{
+	u32 rxmac_stat = readl(gp->regs + MAC_RXSTAT);
+	int ret = 0;
+
+	if (netif_msg_intr(gp))
+		printk(KERN_DEBUG "%s: rxmac interrupt, rxmac_stat: 0x%x\n",
+			gp->dev->name, rxmac_stat);
+
+	if (rxmac_stat & MAC_RXSTAT_OFLW) {
+		u32 smac = readl(gp->regs + MAC_SMACHINE);
+
+		netdev_err(dev, "RX MAC fifo overflow smac[%08x]\n", smac);
+		dev->stats.rx_over_errors++;
+		dev->stats.rx_fifo_errors++;
+
+		ret = gem_rxmac_reset(gp);
+	}
+
+	if (rxmac_stat & MAC_RXSTAT_ACE)
+		dev->stats.rx_frame_errors += 0x10000;
+
+	if (rxmac_stat & MAC_RXSTAT_CCE)
+		dev->stats.rx_crc_errors += 0x10000;
+
+	if (rxmac_stat & MAC_RXSTAT_LCE)
+		dev->stats.rx_length_errors += 0x10000;
+
+	/* We do not track MAC_RXSTAT_FCE and MAC_RXSTAT_VCE
+	 * events.
+	 */
+	return ret;
+}
+
+static int gem_mac_interrupt(struct net_device *dev, struct gem *gp, u32 gem_status)
+{
+	u32 mac_cstat = readl(gp->regs + MAC_CSTAT);
+
+	if (netif_msg_intr(gp))
+		printk(KERN_DEBUG "%s: mac interrupt, mac_cstat: 0x%x\n",
+			gp->dev->name, mac_cstat);
+
+	/* This interrupt is just for pause frame and pause
+	 * tracking.  It is useful for diagnostics and debug
+	 * but probably by default we will mask these events.
+	 */
+	if (mac_cstat & MAC_CSTAT_PS)
+		gp->pause_entered++;
+
+	if (mac_cstat & MAC_CSTAT_PRCV)
+		gp->pause_last_time_recvd = (mac_cstat >> 16);
+
+	return 0;
+}
+
+static int gem_mif_interrupt(struct net_device *dev, struct gem *gp, u32 gem_status)
+{
+	u32 mif_status = readl(gp->regs + MIF_STATUS);
+	u32 reg_val, changed_bits;
+
+	reg_val = (mif_status & MIF_STATUS_DATA) >> 16;
+	changed_bits = (mif_status & MIF_STATUS_STAT);
+
+	gem_handle_mif_event(gp, reg_val, changed_bits);
+
+	return 0;
+}
+
+static int gem_pci_interrupt(struct net_device *dev, struct gem *gp, u32 gem_status)
+{
+	u32 pci_estat = readl(gp->regs + GREG_PCIESTAT);
+
+	if (gp->pdev->vendor == PCI_VENDOR_ID_SUN &&
+	    gp->pdev->device == PCI_DEVICE_ID_SUN_GEM) {
+		netdev_err(dev, "PCI error [%04x]", pci_estat);
+
+		if (pci_estat & GREG_PCIESTAT_BADACK)
+			pr_cont(" <No ACK64# during ABS64 cycle>");
+		if (pci_estat & GREG_PCIESTAT_DTRTO)
+			pr_cont(" <Delayed transaction timeout>");
+		if (pci_estat & GREG_PCIESTAT_OTHER)
+			pr_cont(" <other>");
+		pr_cont("\n");
+	} else {
+		pci_estat |= GREG_PCIESTAT_OTHER;
+		netdev_err(dev, "PCI error\n");
+	}
+
+	if (pci_estat & GREG_PCIESTAT_OTHER) {
+		int pci_errs;
+
+		/* Interrogate PCI config space for the
+		 * true cause.
+		 */
+		pci_errs = pci_status_get_and_clear_errors(gp->pdev);
+		netdev_err(dev, "PCI status errors[%04x]\n", pci_errs);
+		if (pci_errs & PCI_STATUS_PARITY)
+			netdev_err(dev, "PCI parity error detected\n");
+		if (pci_errs & PCI_STATUS_SIG_TARGET_ABORT)
+			netdev_err(dev, "PCI target abort\n");
+		if (pci_errs & PCI_STATUS_REC_TARGET_ABORT)
+			netdev_err(dev, "PCI master acks target abort\n");
+		if (pci_errs & PCI_STATUS_REC_MASTER_ABORT)
+			netdev_err(dev, "PCI master abort\n");
+		if (pci_errs & PCI_STATUS_SIG_SYSTEM_ERROR)
+			netdev_err(dev, "PCI system error SERR#\n");
+		if (pci_errs & PCI_STATUS_DETECTED_PARITY)
+			netdev_err(dev, "PCI parity error\n");
+	}
+
+	/* For all PCI errors, we should reset the chip. */
+	return 1;
+}
+
+/* All non-normal interrupt conditions get serviced here.
+ * Returns non-zero if we should just exit the interrupt
+ * handler right now (ie. if we reset the card which invalidates
+ * all of the other original irq status bits).
+ */
+static int gem_abnormal_irq(struct net_device *dev, struct gem *gp, u32 gem_status)
+{
+	if (gem_status & GREG_STAT_RXNOBUF) {
+		/* Frame arrived, no free RX buffers available. */
+		if (netif_msg_rx_err(gp))
+			printk(KERN_DEBUG "%s: no buffer for rx frame\n",
+				gp->dev->name);
+		dev->stats.rx_dropped++;
+	}
+
+	if (gem_status & GREG_STAT_RXTAGERR) {
+		/* corrupt RX tag framing */
+		if (netif_msg_rx_err(gp))
+			printk(KERN_DEBUG "%s: corrupt rx tag framing\n",
+				gp->dev->name);
+		dev->stats.rx_errors++;
+
+		return 1;
+	}
+
+	if (gem_status & GREG_STAT_PCS) {
+		if (gem_pcs_interrupt(dev, gp, gem_status))
+			return 1;
+	}
+
+	if (gem_status & GREG_STAT_TXMAC) {
+		if (gem_txmac_interrupt(dev, gp, gem_status))
+			return 1;
+	}
+
+	if (gem_status & GREG_STAT_RXMAC) {
+		if (gem_rxmac_interrupt(dev, gp, gem_status))
+			return 1;
+	}
+
+	if (gem_status & GREG_STAT_MAC) {
+		if (gem_mac_interrupt(dev, gp, gem_status))
+			return 1;
+	}
+
+	if (gem_status & GREG_STAT_MIF) {
+		if (gem_mif_interrupt(dev, gp, gem_status))
+			return 1;
+	}
+
+	if (gem_status & GREG_STAT_PCIERR) {
+		if (gem_pci_interrupt(dev, gp, gem_status))
+			return 1;
+	}
+
+	return 0;
+}
+
+static __inline__ void gem_tx(struct net_device *dev, struct gem *gp, u32 gem_status)
+{
+	int entry, limit;
+
+	entry = gp->tx_old;
+	limit = ((gem_status & GREG_STAT_TXNR) >> GREG_STAT_TXNR_SHIFT);
+	while (entry != limit) {
+		struct sk_buff *skb;
+		struct gem_txd *txd;
+		dma_addr_t dma_addr;
+		u32 dma_len;
+		int frag;
+
+		if (netif_msg_tx_done(gp))
+			printk(KERN_DEBUG "%s: tx done, slot %d\n",
+				gp->dev->name, entry);
+		skb = gp->tx_skbs[entry];
+		if (skb_shinfo(skb)->nr_frags) {
+			int last = entry + skb_shinfo(skb)->nr_frags;
+			int walk = entry;
+			int incomplete = 0;
+
+			last &= (TX_RING_SIZE - 1);
+			for (;;) {
+				walk = NEXT_TX(walk);
+				if (walk == limit)
+					incomplete = 1;
+				if (walk == last)
+					break;
+			}
+			if (incomplete)
+				break;
+		}
+		gp->tx_skbs[entry] = NULL;
+		dev->stats.tx_bytes += skb->len;
+
+		for (frag = 0; frag <= skb_shinfo(skb)->nr_frags; frag++) {
+			txd = &gp->init_block->txd[entry];
+
+			dma_addr = le64_to_cpu(txd->buffer);
+			dma_len = le64_to_cpu(txd->control_word) & TXDCTRL_BUFSZ;
+
+			dma_unmap_page(&gp->pdev->dev, dma_addr, dma_len,
+				       DMA_TO_DEVICE);
+			entry = NEXT_TX(entry);
+		}
+
+		dev->stats.tx_packets++;
+		dev_consume_skb_any(skb);
+	}
+	gp->tx_old = entry;
+
+	/* Need to make the tx_old update visible to gem_start_xmit()
+	 * before checking for netif_queue_stopped().  Without the
+	 * memory barrier, there is a small possibility that gem_start_xmit()
+	 * will miss it and cause the queue to be stopped forever.
+	 */
+	smp_mb();
+
+	if (unlikely(netif_queue_stopped(dev) &&
+		     TX_BUFFS_AVAIL(gp) > (MAX_SKB_FRAGS + 1))) {
+		struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
+
+		__netif_tx_lock(txq, smp_processor_id());
+		if (netif_queue_stopped(dev) &&
+		    TX_BUFFS_AVAIL(gp) > (MAX_SKB_FRAGS + 1))
+			netif_wake_queue(dev);
+		__netif_tx_unlock(txq);
+	}
+}
+
+static __inline__ void gem_post_rxds(struct gem *gp, int limit)
+{
+	int cluster_start, curr, count, kick;
+
+	cluster_start = curr = (gp->rx_new & ~(4 - 1));
+	count = 0;
+	kick = -1;
+	dma_wmb();
+	while (curr != limit) {
+		curr = NEXT_RX(curr);
+		if (++count == 4) {
+			struct gem_rxd *rxd =
+				&gp->init_block->rxd[cluster_start];
+			for (;;) {
+				rxd->status_word = cpu_to_le64(RXDCTRL_FRESH(gp));
+				rxd++;
+				cluster_start = NEXT_RX(cluster_start);
+				if (cluster_start == curr)
+					break;
+			}
+			kick = curr;
+			count = 0;
+		}
+	}
+	if (kick >= 0) {
+		mb();
+		writel(kick, gp->regs + RXDMA_KICK);
+	}
+}
+
+#define ALIGNED_RX_SKB_ADDR(addr) \
+        ((((unsigned long)(addr) + (64UL - 1UL)) & ~(64UL - 1UL)) - (unsigned long)(addr))
+static __inline__ struct sk_buff *gem_alloc_skb(struct net_device *dev, int size,
+						gfp_t gfp_flags)
+{
+	struct sk_buff *skb = alloc_skb(size + 64, gfp_flags);
+
+	if (likely(skb)) {
+		unsigned long offset = ALIGNED_RX_SKB_ADDR(skb->data);
+		skb_reserve(skb, offset);
+	}
+	return skb;
+}
+
+static int gem_rx(struct gem *gp, int work_to_do)
+{
+	struct net_device *dev = gp->dev;
+	int entry, drops, work_done = 0;
+	u32 done;
+
+	if (netif_msg_rx_status(gp))
+		printk(KERN_DEBUG "%s: rx interrupt, done: %d, rx_new: %d\n",
+			gp->dev->name, readl(gp->regs + RXDMA_DONE), gp->rx_new);
+
+	entry = gp->rx_new;
+	drops = 0;
+	done = readl(gp->regs + RXDMA_DONE);
+	for (;;) {
+		struct gem_rxd *rxd = &gp->init_block->rxd[entry];
+		struct sk_buff *skb;
+		u64 status = le64_to_cpu(rxd->status_word);
+		dma_addr_t dma_addr;
+		int len;
+
+		if ((status & RXDCTRL_OWN) != 0)
+			break;
+
+		if (work_done >= RX_RING_SIZE || work_done >= work_to_do)
+			break;
+
+		/* When writing back RX descriptor, GEM writes status
+		 * then buffer address, possibly in separate transactions.
+		 * If we don't wait for the chip to write both, we could
+		 * post a new buffer to this descriptor then have GEM spam
+		 * on the buffer address.  We sync on the RX completion
+		 * register to prevent this from happening.
+		 */
+		if (entry == done) {
+			done = readl(gp->regs + RXDMA_DONE);
+			if (entry == done)
+				break;
+		}
+
+		/* We can now account for the work we're about to do */
+		work_done++;
+
+		skb = gp->rx_skbs[entry];
+
+		len = (status & RXDCTRL_BUFSZ) >> 16;
+		if ((len < ETH_ZLEN) || (status & RXDCTRL_BAD)) {
+			dev->stats.rx_errors++;
+			if (len < ETH_ZLEN)
+				dev->stats.rx_length_errors++;
+			if (len & RXDCTRL_BAD)
+				dev->stats.rx_crc_errors++;
+
+			/* We'll just return it to GEM. */
+		drop_it:
+			dev->stats.rx_dropped++;
+			goto next;
+		}
+
+		dma_addr = le64_to_cpu(rxd->buffer);
+		if (len > RX_COPY_THRESHOLD) {
+			struct sk_buff *new_skb;
+
+			new_skb = gem_alloc_skb(dev, RX_BUF_ALLOC_SIZE(gp), GFP_ATOMIC);
+			if (new_skb == NULL) {
+				drops++;
+				goto drop_it;
+			}
+			dma_unmap_page(&gp->pdev->dev, dma_addr,
+				       RX_BUF_ALLOC_SIZE(gp), DMA_FROM_DEVICE);
+			gp->rx_skbs[entry] = new_skb;
+			skb_put(new_skb, (gp->rx_buf_sz + RX_OFFSET));
+			rxd->buffer = cpu_to_le64(dma_map_page(&gp->pdev->dev,
+							       virt_to_page(new_skb->data),
+							       offset_in_page(new_skb->data),
+							       RX_BUF_ALLOC_SIZE(gp),
+							       DMA_FROM_DEVICE));
+			skb_reserve(new_skb, RX_OFFSET);
+
+			/* Trim the original skb for the netif. */
+			skb_trim(skb, len);
+		} else {
+			struct sk_buff *copy_skb = netdev_alloc_skb(dev, len + 2);
+
+			if (copy_skb == NULL) {
+				drops++;
+				goto drop_it;
+			}
+
+			skb_reserve(copy_skb, 2);
+			skb_put(copy_skb, len);
+			dma_sync_single_for_cpu(&gp->pdev->dev, dma_addr, len,
+						DMA_FROM_DEVICE);
+			skb_copy_from_linear_data(skb, copy_skb->data, len);
+			dma_sync_single_for_device(&gp->pdev->dev, dma_addr,
+						   len, DMA_FROM_DEVICE);
+
+			/* We'll reuse the original ring buffer. */
+			skb = copy_skb;
+		}
+
+		if (likely(dev->features & NETIF_F_RXCSUM)) {
+			__sum16 csum;
+
+			csum = (__force __sum16)htons((status & RXDCTRL_TCPCSUM) ^ 0xffff);
+			skb->csum = csum_unfold(csum);
+			skb->ip_summed = CHECKSUM_COMPLETE;
+		}
+		skb->protocol = eth_type_trans(skb, gp->dev);
+
+		napi_gro_receive(&gp->napi, skb);
+
+		dev->stats.rx_packets++;
+		dev->stats.rx_bytes += len;
+
+	next:
+		entry = NEXT_RX(entry);
+	}
+
+	gem_post_rxds(gp, entry);
+
+	gp->rx_new = entry;
+
+	if (drops)
+		netdev_info(gp->dev, "Memory squeeze, deferring packet\n");
+
+	return work_done;
+}
+
+static int gem_poll(struct napi_struct *napi, int budget)
+{
+	struct gem *gp = container_of(napi, struct gem, napi);
+	struct net_device *dev = gp->dev;
+	int work_done;
+
+	work_done = 0;
+	do {
+		/* Handle anomalies */
+		if (unlikely(gp->status & GREG_STAT_ABNORMAL)) {
+			struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
+			int reset;
+
+			/* We run the abnormal interrupt handling code with
+			 * the Tx lock. It only resets the Rx portion of the
+			 * chip, but we need to guard it against DMA being
+			 * restarted by the link poll timer
+			 */
+			__netif_tx_lock(txq, smp_processor_id());
+			reset = gem_abnormal_irq(dev, gp, gp->status);
+			__netif_tx_unlock(txq);
+			if (reset) {
+				gem_schedule_reset(gp);
+				napi_complete(napi);
+				return work_done;
+			}
+		}
+
+		/* Run TX completion thread */
+		gem_tx(dev, gp, gp->status);
+
+		/* Run RX thread. We don't use any locking here,
+		 * code willing to do bad things - like cleaning the
+		 * rx ring - must call napi_disable(), which
+		 * schedule_timeout()'s if polling is already disabled.
+		 */
+		work_done += gem_rx(gp, budget - work_done);
+
+		if (work_done >= budget)
+			return work_done;
+
+		gp->status = readl(gp->regs + GREG_STAT);
+	} while (gp->status & GREG_STAT_NAPI);
+
+	napi_complete_done(napi, work_done);
+	gem_enable_ints(gp);
+
+	return work_done;
+}
+
+static irqreturn_t gem_interrupt(int irq, void *dev_id)
+{
+	struct net_device *dev = dev_id;
+	struct gem *gp = netdev_priv(dev);
+
+	if (napi_schedule_prep(&gp->napi)) {
+		u32 gem_status = readl(gp->regs + GREG_STAT);
+
+		if (unlikely(gem_status == 0)) {
+			napi_enable(&gp->napi);
+			return IRQ_NONE;
+		}
+		if (netif_msg_intr(gp))
+			printk(KERN_DEBUG "%s: gem_interrupt() gem_status: 0x%x\n",
+			       gp->dev->name, gem_status);
+
+		gp->status = gem_status;
+		gem_disable_ints(gp);
+		__napi_schedule(&gp->napi);
+	}
+
+	/* If polling was disabled at the time we received that
+	 * interrupt, we may return IRQ_HANDLED here while we
+	 * should return IRQ_NONE. No big deal...
+	 */
+	return IRQ_HANDLED;
+}
+
+static void gem_tx_timeout(struct net_device *dev, unsigned int txqueue)
+{
+	struct gem *gp = netdev_priv(dev);
+
+	netdev_err(dev, "transmit timed out, resetting\n");
+
+	netdev_err(dev, "TX_STATE[%08x:%08x:%08x]\n",
+		   readl(gp->regs + TXDMA_CFG),
+		   readl(gp->regs + MAC_TXSTAT),
+		   readl(gp->regs + MAC_TXCFG));
+	netdev_err(dev, "RX_STATE[%08x:%08x:%08x]\n",
+		   readl(gp->regs + RXDMA_CFG),
+		   readl(gp->regs + MAC_RXSTAT),
+		   readl(gp->regs + MAC_RXCFG));
+
+	gem_schedule_reset(gp);
+}
+
+static __inline__ int gem_intme(int entry)
+{
+	/* Algorithm: IRQ every 1/2 of descriptors. */
+	if (!(entry & ((TX_RING_SIZE>>1)-1)))
+		return 1;
+
+	return 0;
+}
+
+static netdev_tx_t gem_start_xmit(struct sk_buff *skb,
+				  struct net_device *dev)
+{
+	struct gem *gp = netdev_priv(dev);
+	int entry;
+	u64 ctrl;
+
+	ctrl = 0;
+	if (skb->ip_summed == CHECKSUM_PARTIAL) {
+		const u64 csum_start_off = skb_checksum_start_offset(skb);
+		const u64 csum_stuff_off = csum_start_off + skb->csum_offset;
+
+		ctrl = (TXDCTRL_CENAB |
+			(csum_start_off << 15) |
+			(csum_stuff_off << 21));
+	}
+
+	if (unlikely(TX_BUFFS_AVAIL(gp) <= (skb_shinfo(skb)->nr_frags + 1))) {
+		/* This is a hard error, log it. */
+		if (!netif_queue_stopped(dev)) {
+			netif_stop_queue(dev);
+			netdev_err(dev, "BUG! Tx Ring full when queue awake!\n");
+		}
+		return NETDEV_TX_BUSY;
+	}
+
+	entry = gp->tx_new;
+	gp->tx_skbs[entry] = skb;
+
+	if (skb_shinfo(skb)->nr_frags == 0) {
+		struct gem_txd *txd = &gp->init_block->txd[entry];
+		dma_addr_t mapping;
+		u32 len;
+
+		len = skb->len;
+		mapping = dma_map_page(&gp->pdev->dev,
+				       virt_to_page(skb->data),
+				       offset_in_page(skb->data),
+				       len, DMA_TO_DEVICE);
+		ctrl |= TXDCTRL_SOF | TXDCTRL_EOF | len;
+		if (gem_intme(entry))
+			ctrl |= TXDCTRL_INTME;
+		txd->buffer = cpu_to_le64(mapping);
+		dma_wmb();
+		txd->control_word = cpu_to_le64(ctrl);
+		entry = NEXT_TX(entry);
+	} else {
+		struct gem_txd *txd;
+		u32 first_len;
+		u64 intme;
+		dma_addr_t first_mapping;
+		int frag, first_entry = entry;
+
+		intme = 0;
+		if (gem_intme(entry))
+			intme |= TXDCTRL_INTME;
+
+		/* We must give this initial chunk to the device last.
+		 * Otherwise we could race with the device.
+		 */
+		first_len = skb_headlen(skb);
+		first_mapping = dma_map_page(&gp->pdev->dev,
+					     virt_to_page(skb->data),
+					     offset_in_page(skb->data),
+					     first_len, DMA_TO_DEVICE);
+		entry = NEXT_TX(entry);
+
+		for (frag = 0; frag < skb_shinfo(skb)->nr_frags; frag++) {
+			const skb_frag_t *this_frag = &skb_shinfo(skb)->frags[frag];
+			u32 len;
+			dma_addr_t mapping;
+			u64 this_ctrl;
+
+			len = skb_frag_size(this_frag);
+			mapping = skb_frag_dma_map(&gp->pdev->dev, this_frag,
+						   0, len, DMA_TO_DEVICE);
+			this_ctrl = ctrl;
+			if (frag == skb_shinfo(skb)->nr_frags - 1)
+				this_ctrl |= TXDCTRL_EOF;
+
+			txd = &gp->init_block->txd[entry];
+			txd->buffer = cpu_to_le64(mapping);
+			dma_wmb();
+			txd->control_word = cpu_to_le64(this_ctrl | len);
+
+			if (gem_intme(entry))
+				intme |= TXDCTRL_INTME;
+
+			entry = NEXT_TX(entry);
+		}
+		txd = &gp->init_block->txd[first_entry];
+		txd->buffer = cpu_to_le64(first_mapping);
+		dma_wmb();
+		txd->control_word =
+			cpu_to_le64(ctrl | TXDCTRL_SOF | intme | first_len);
+	}
+
+	gp->tx_new = entry;
+	if (unlikely(TX_BUFFS_AVAIL(gp) <= (MAX_SKB_FRAGS + 1))) {
+		netif_stop_queue(dev);
+
+		/* netif_stop_queue() must be done before checking
+		 * tx index in TX_BUFFS_AVAIL() below, because
+		 * in gem_tx(), we update tx_old before checking for
+		 * netif_queue_stopped().
+		 */
+		smp_mb();
+		if (TX_BUFFS_AVAIL(gp) > (MAX_SKB_FRAGS + 1))
+			netif_wake_queue(dev);
+	}
+	if (netif_msg_tx_queued(gp))
+		printk(KERN_DEBUG "%s: tx queued, slot %d, skblen %d\n",
+		       dev->name, entry, skb->len);
+	mb();
+	writel(gp->tx_new, gp->regs + TXDMA_KICK);
+
+	return NETDEV_TX_OK;
+}
+
+static void gem_pcs_reset(struct gem *gp)
+{
+	int limit;
+	u32 val;
+
+	/* Reset PCS unit. */
+	val = readl(gp->regs + PCS_MIICTRL);
+	val |= PCS_MIICTRL_RST;
+	writel(val, gp->regs + PCS_MIICTRL);
+
+	limit = 32;
+	while (readl(gp->regs + PCS_MIICTRL) & PCS_MIICTRL_RST) {
+		udelay(100);
+		if (limit-- <= 0)
+			break;
+	}
+	if (limit < 0)
+		netdev_warn(gp->dev, "PCS reset bit would not clear\n");
+}
+
+static void gem_pcs_reinit_adv(struct gem *gp)
+{
+	u32 val;
+
+	/* Make sure PCS is disabled while changing advertisement
+	 * configuration.
+	 */
+	val = readl(gp->regs + PCS_CFG);
+	val &= ~(PCS_CFG_ENABLE | PCS_CFG_TO);
+	writel(val, gp->regs + PCS_CFG);
+
+	/* Advertise all capabilities except asymmetric
+	 * pause.
+	 */
+	val = readl(gp->regs + PCS_MIIADV);
+	val |= (PCS_MIIADV_FD | PCS_MIIADV_HD |
+		PCS_MIIADV_SP | PCS_MIIADV_AP);
+	writel(val, gp->regs + PCS_MIIADV);
+
+	/* Enable and restart auto-negotiation, disable wrapback/loopback,
+	 * and re-enable PCS.
+	 */
+	val = readl(gp->regs + PCS_MIICTRL);
+	val |= (PCS_MIICTRL_RAN | PCS_MIICTRL_ANE);
+	val &= ~PCS_MIICTRL_WB;
+	writel(val, gp->regs + PCS_MIICTRL);
+
+	val = readl(gp->regs + PCS_CFG);
+	val |= PCS_CFG_ENABLE;
+	writel(val, gp->regs + PCS_CFG);
+
+	/* Make sure serialink loopback is off.  The meaning
+	 * of this bit is logically inverted based upon whether
+	 * you are in Serialink or SERDES mode.
+	 */
+	val = readl(gp->regs + PCS_SCTRL);
+	if (gp->phy_type == phy_serialink)
+		val &= ~PCS_SCTRL_LOOP;
+	else
+		val |= PCS_SCTRL_LOOP;
+	writel(val, gp->regs + PCS_SCTRL);
+}
+
+#define STOP_TRIES 32
+
+static void gem_reset(struct gem *gp)
+{
+	int limit;
+	u32 val;
+
+	/* Make sure we won't get any more interrupts */
+	writel(0xffffffff, gp->regs + GREG_IMASK);
+
+	/* Reset the chip */
+	writel(gp->swrst_base | GREG_SWRST_TXRST | GREG_SWRST_RXRST,
+	       gp->regs + GREG_SWRST);
+
+	limit = STOP_TRIES;
+
+	do {
+		udelay(20);
+		val = readl(gp->regs + GREG_SWRST);
+		if (limit-- <= 0)
+			break;
+	} while (val & (GREG_SWRST_TXRST | GREG_SWRST_RXRST));
+
+	if (limit < 0)
+		netdev_err(gp->dev, "SW reset is ghetto\n");
+
+	if (gp->phy_type == phy_serialink || gp->phy_type == phy_serdes)
+		gem_pcs_reinit_adv(gp);
+}
+
+static void gem_start_dma(struct gem *gp)
+{
+	u32 val;
+
+	/* We are ready to rock, turn everything on. */
+	val = readl(gp->regs + TXDMA_CFG);
+	writel(val | TXDMA_CFG_ENABLE, gp->regs + TXDMA_CFG);
+	val = readl(gp->regs + RXDMA_CFG);
+	writel(val | RXDMA_CFG_ENABLE, gp->regs + RXDMA_CFG);
+	val = readl(gp->regs + MAC_TXCFG);
+	writel(val | MAC_TXCFG_ENAB, gp->regs + MAC_TXCFG);
+	val = readl(gp->regs + MAC_RXCFG);
+	writel(val | MAC_RXCFG_ENAB, gp->regs + MAC_RXCFG);
+
+	(void) readl(gp->regs + MAC_RXCFG);
+	udelay(100);
+
+	gem_enable_ints(gp);
+
+	writel(RX_RING_SIZE - 4, gp->regs + RXDMA_KICK);
+}
+
+/* DMA won't be actually stopped before about 4ms tho ...
+ */
+static void gem_stop_dma(struct gem *gp)
+{
+	u32 val;
+
+	/* We are done rocking, turn everything off. */
+	val = readl(gp->regs + TXDMA_CFG);
+	writel(val & ~TXDMA_CFG_ENABLE, gp->regs + TXDMA_CFG);
+	val = readl(gp->regs + RXDMA_CFG);
+	writel(val & ~RXDMA_CFG_ENABLE, gp->regs + RXDMA_CFG);
+	val = readl(gp->regs + MAC_TXCFG);
+	writel(val & ~MAC_TXCFG_ENAB, gp->regs + MAC_TXCFG);
+	val = readl(gp->regs + MAC_RXCFG);
+	writel(val & ~MAC_RXCFG_ENAB, gp->regs + MAC_RXCFG);
+
+	(void) readl(gp->regs + MAC_RXCFG);
+
+	/* Need to wait a bit ... done by the caller */
+}
+
+
+// XXX dbl check what that function should do when called on PCS PHY
+static void gem_begin_auto_negotiation(struct gem *gp,
+				       const struct ethtool_link_ksettings *ep)
+{
+	u32 advertise, features;
+	int autoneg;
+	int speed;
+	int duplex;
+	u32 advertising;
+
+	if (ep)
+		ethtool_convert_link_mode_to_legacy_u32(
+			&advertising, ep->link_modes.advertising);
+
+	if (gp->phy_type != phy_mii_mdio0 &&
+	    gp->phy_type != phy_mii_mdio1)
+		goto non_mii;
+
+	/* Setup advertise */
+	if (found_mii_phy(gp))
+		features = gp->phy_mii.def->features;
+	else
+		features = 0;
+
+	advertise = features & ADVERTISE_MASK;
+	if (gp->phy_mii.advertising != 0)
+		advertise &= gp->phy_mii.advertising;
+
+	autoneg = gp->want_autoneg;
+	speed = gp->phy_mii.speed;
+	duplex = gp->phy_mii.duplex;
+
+	/* Setup link parameters */
+	if (!ep)
+		goto start_aneg;
+	if (ep->base.autoneg == AUTONEG_ENABLE) {
+		advertise = advertising;
+		autoneg = 1;
+	} else {
+		autoneg = 0;
+		speed = ep->base.speed;
+		duplex = ep->base.duplex;
+	}
+
+start_aneg:
+	/* Sanitize settings based on PHY capabilities */
+	if ((features & SUPPORTED_Autoneg) == 0)
+		autoneg = 0;
+	if (speed == SPEED_1000 &&
+	    !(features & (SUPPORTED_1000baseT_Half | SUPPORTED_1000baseT_Full)))
+		speed = SPEED_100;
+	if (speed == SPEED_100 &&
+	    !(features & (SUPPORTED_100baseT_Half | SUPPORTED_100baseT_Full)))
+		speed = SPEED_10;
+	if (duplex == DUPLEX_FULL &&
+	    !(features & (SUPPORTED_1000baseT_Full |
+	    		  SUPPORTED_100baseT_Full |
+	    		  SUPPORTED_10baseT_Full)))
+	    	duplex = DUPLEX_HALF;
+	if (speed == 0)
+		speed = SPEED_10;
+
+	/* If we are asleep, we don't try to actually setup the PHY, we
+	 * just store the settings
+	 */
+	if (!netif_device_present(gp->dev)) {
+		gp->phy_mii.autoneg = gp->want_autoneg = autoneg;
+		gp->phy_mii.speed = speed;
+		gp->phy_mii.duplex = duplex;
+		return;
+	}
+
+	/* Configure PHY & start aneg */
+	gp->want_autoneg = autoneg;
+	if (autoneg) {
+		if (found_mii_phy(gp))
+			gp->phy_mii.def->ops->setup_aneg(&gp->phy_mii, advertise);
+		gp->lstate = link_aneg;
+	} else {
+		if (found_mii_phy(gp))
+			gp->phy_mii.def->ops->setup_forced(&gp->phy_mii, speed, duplex);
+		gp->lstate = link_force_ok;
+	}
+
+non_mii:
+	gp->timer_ticks = 0;
+	mod_timer(&gp->link_timer, jiffies + ((12 * HZ) / 10));
+}
+
+/* A link-up condition has occurred, initialize and enable the
+ * rest of the chip.
+ */
+static int gem_set_link_modes(struct gem *gp)
+{
+	struct netdev_queue *txq = netdev_get_tx_queue(gp->dev, 0);
+	int full_duplex, speed, pause;
+	u32 val;
+
+	full_duplex = 0;
+	speed = SPEED_10;
+	pause = 0;
+
+	if (found_mii_phy(gp)) {
+	    	if (gp->phy_mii.def->ops->read_link(&gp->phy_mii))
+	    		return 1;
+		full_duplex = (gp->phy_mii.duplex == DUPLEX_FULL);
+		speed = gp->phy_mii.speed;
+		pause = gp->phy_mii.pause;
+	} else if (gp->phy_type == phy_serialink ||
+	    	   gp->phy_type == phy_serdes) {
+		u32 pcs_lpa = readl(gp->regs + PCS_MIILP);
+
+		if ((pcs_lpa & PCS_MIIADV_FD) || gp->phy_type == phy_serdes)
+			full_duplex = 1;
+		speed = SPEED_1000;
+	}
+
+	netif_info(gp, link, gp->dev, "Link is up at %d Mbps, %s-duplex\n",
+		   speed, (full_duplex ? "full" : "half"));
+
+
+	/* We take the tx queue lock to avoid collisions between
+	 * this code, the tx path and the NAPI-driven error path
+	 */
+	__netif_tx_lock(txq, smp_processor_id());
+
+	val = (MAC_TXCFG_EIPG0 | MAC_TXCFG_NGU);
+	if (full_duplex) {
+		val |= (MAC_TXCFG_ICS | MAC_TXCFG_ICOLL);
+	} else {
+		/* MAC_TXCFG_NBO must be zero. */
+	}
+	writel(val, gp->regs + MAC_TXCFG);
+
+	val = (MAC_XIFCFG_OE | MAC_XIFCFG_LLED);
+	if (!full_duplex &&
+	    (gp->phy_type == phy_mii_mdio0 ||
+	     gp->phy_type == phy_mii_mdio1)) {
+		val |= MAC_XIFCFG_DISE;
+	} else if (full_duplex) {
+		val |= MAC_XIFCFG_FLED;
+	}
+
+	if (speed == SPEED_1000)
+		val |= (MAC_XIFCFG_GMII);
+
+	writel(val, gp->regs + MAC_XIFCFG);
+
+	/* If gigabit and half-duplex, enable carrier extension
+	 * mode.  Else, disable it.
+	 */
+	if (speed == SPEED_1000 && !full_duplex) {
+		val = readl(gp->regs + MAC_TXCFG);
+		writel(val | MAC_TXCFG_TCE, gp->regs + MAC_TXCFG);
+
+		val = readl(gp->regs + MAC_RXCFG);
+		writel(val | MAC_RXCFG_RCE, gp->regs + MAC_RXCFG);
+	} else {
+		val = readl(gp->regs + MAC_TXCFG);
+		writel(val & ~MAC_TXCFG_TCE, gp->regs + MAC_TXCFG);
+
+		val = readl(gp->regs + MAC_RXCFG);
+		writel(val & ~MAC_RXCFG_RCE, gp->regs + MAC_RXCFG);
+	}
+
+	if (gp->phy_type == phy_serialink ||
+	    gp->phy_type == phy_serdes) {
+		u32 pcs_lpa = readl(gp->regs + PCS_MIILP);
+
+		if (pcs_lpa & (PCS_MIIADV_SP | PCS_MIIADV_AP))
+			pause = 1;
+	}
+
+	if (!full_duplex)
+		writel(512, gp->regs + MAC_STIME);
+	else
+		writel(64, gp->regs + MAC_STIME);
+	val = readl(gp->regs + MAC_MCCFG);
+	if (pause)
+		val |= (MAC_MCCFG_SPE | MAC_MCCFG_RPE);
+	else
+		val &= ~(MAC_MCCFG_SPE | MAC_MCCFG_RPE);
+	writel(val, gp->regs + MAC_MCCFG);
+
+	gem_start_dma(gp);
+
+	__netif_tx_unlock(txq);
+
+	if (netif_msg_link(gp)) {
+		if (pause) {
+			netdev_info(gp->dev,
+				    "Pause is enabled (rxfifo: %d off: %d on: %d)\n",
+				    gp->rx_fifo_sz,
+				    gp->rx_pause_off,
+				    gp->rx_pause_on);
+		} else {
+			netdev_info(gp->dev, "Pause is disabled\n");
+		}
+	}
+
+	return 0;
+}
+
+static int gem_mdio_link_not_up(struct gem *gp)
+{
+	switch (gp->lstate) {
+	case link_force_ret:
+		netif_info(gp, link, gp->dev,
+			   "Autoneg failed again, keeping forced mode\n");
+		gp->phy_mii.def->ops->setup_forced(&gp->phy_mii,
+			gp->last_forced_speed, DUPLEX_HALF);
+		gp->timer_ticks = 5;
+		gp->lstate = link_force_ok;
+		return 0;
+	case link_aneg:
+		/* We try forced modes after a failed aneg only on PHYs that don't
+		 * have "magic_aneg" bit set, which means they internally do the
+		 * while forced-mode thingy. On these, we just restart aneg
+		 */
+		if (gp->phy_mii.def->magic_aneg)
+			return 1;
+		netif_info(gp, link, gp->dev, "switching to forced 100bt\n");
+		/* Try forced modes. */
+		gp->phy_mii.def->ops->setup_forced(&gp->phy_mii, SPEED_100,
+			DUPLEX_HALF);
+		gp->timer_ticks = 5;
+		gp->lstate = link_force_try;
+		return 0;
+	case link_force_try:
+		/* Downgrade from 100 to 10 Mbps if necessary.
+		 * If already at 10Mbps, warn user about the
+		 * situation every 10 ticks.
+		 */
+		if (gp->phy_mii.speed == SPEED_100) {
+			gp->phy_mii.def->ops->setup_forced(&gp->phy_mii, SPEED_10,
+				DUPLEX_HALF);
+			gp->timer_ticks = 5;
+			netif_info(gp, link, gp->dev,
+				   "switching to forced 10bt\n");
+			return 0;
+		} else
+			return 1;
+	default:
+		return 0;
+	}
+}
+
+static void gem_link_timer(struct timer_list *t)
+{
+	struct gem *gp = timer_container_of(gp, t, link_timer);
+	struct net_device *dev = gp->dev;
+	int restart_aneg = 0;
+
+	/* There's no point doing anything if we're going to be reset */
+	if (gp->reset_task_pending)
+		return;
+
+	if (gp->phy_type == phy_serialink ||
+	    gp->phy_type == phy_serdes) {
+		u32 val = readl(gp->regs + PCS_MIISTAT);
+
+		if (!(val & PCS_MIISTAT_LS))
+			val = readl(gp->regs + PCS_MIISTAT);
+
+		if ((val & PCS_MIISTAT_LS) != 0) {
+			if (gp->lstate == link_up)
+				goto restart;
+
+			gp->lstate = link_up;
+			netif_carrier_on(dev);
+			(void)gem_set_link_modes(gp);
+		}
+		goto restart;
+	}
+	if (found_mii_phy(gp) && gp->phy_mii.def->ops->poll_link(&gp->phy_mii)) {
+		/* Ok, here we got a link. If we had it due to a forced
+		 * fallback, and we were configured for autoneg, we do
+		 * retry a short autoneg pass. If you know your hub is
+		 * broken, use ethtool ;)
+		 */
+		if (gp->lstate == link_force_try && gp->want_autoneg) {
+			gp->lstate = link_force_ret;
+			gp->last_forced_speed = gp->phy_mii.speed;
+			gp->timer_ticks = 5;
+			if (netif_msg_link(gp))
+				netdev_info(dev,
+					    "Got link after fallback, retrying autoneg once...\n");
+			gp->phy_mii.def->ops->setup_aneg(&gp->phy_mii, gp->phy_mii.advertising);
+		} else if (gp->lstate != link_up) {
+			gp->lstate = link_up;
+			netif_carrier_on(dev);
+			if (gem_set_link_modes(gp))
+				restart_aneg = 1;
+		}
+	} else {
+		/* If the link was previously up, we restart the
+		 * whole process
+		 */
+		if (gp->lstate == link_up) {
+			gp->lstate = link_down;
+			netif_info(gp, link, dev, "Link down\n");
+			netif_carrier_off(dev);
+			gem_schedule_reset(gp);
+			/* The reset task will restart the timer */
+			return;
+		} else if (++gp->timer_ticks > 10) {
+			if (found_mii_phy(gp))
+				restart_aneg = gem_mdio_link_not_up(gp);
+			else
+				restart_aneg = 1;
+		}
+	}
+	if (restart_aneg) {
+		gem_begin_auto_negotiation(gp, NULL);
+		return;
+	}
+restart:
+	mod_timer(&gp->link_timer, jiffies + ((12 * HZ) / 10));
+}
+
+static void gem_clean_rings(struct gem *gp)
+{
+	struct gem_init_block *gb = gp->init_block;
+	struct sk_buff *skb;
+	int i;
+	dma_addr_t dma_addr;
+
+	for (i = 0; i < RX_RING_SIZE; i++) {
+		struct gem_rxd *rxd;
+
+		rxd = &gb->rxd[i];
+		if (gp->rx_skbs[i] != NULL) {
+			skb = gp->rx_skbs[i];
+			dma_addr = le64_to_cpu(rxd->buffer);
+			dma_unmap_page(&gp->pdev->dev, dma_addr,
+				       RX_BUF_ALLOC_SIZE(gp),
+				       DMA_FROM_DEVICE);
+			dev_kfree_skb_any(skb);
+			gp->rx_skbs[i] = NULL;
+		}
+		rxd->status_word = 0;
+		dma_wmb();
+		rxd->buffer = 0;
+	}
+
+	for (i = 0; i < TX_RING_SIZE; i++) {
+		if (gp->tx_skbs[i] != NULL) {
+			struct gem_txd *txd;
+			int frag;
+
+			skb = gp->tx_skbs[i];
+			gp->tx_skbs[i] = NULL;
+
+			for (frag = 0; frag <= skb_shinfo(skb)->nr_frags; frag++) {
+				int ent = i & (TX_RING_SIZE - 1);
+
+				txd = &gb->txd[ent];
+				dma_addr = le64_to_cpu(txd->buffer);
+				dma_unmap_page(&gp->pdev->dev, dma_addr,
+					       le64_to_cpu(txd->control_word) &
+					       TXDCTRL_BUFSZ, DMA_TO_DEVICE);
+
+				if (frag != skb_shinfo(skb)->nr_frags)
+					i++;
+			}
+			dev_kfree_skb_any(skb);
+		}
+	}
+}
+
+static void gem_init_rings(struct gem *gp)
+{
+	struct gem_init_block *gb = gp->init_block;
+	struct net_device *dev = gp->dev;
+	int i;
+	dma_addr_t dma_addr;
+
+	gp->rx_new = gp->rx_old = gp->tx_new = gp->tx_old = 0;
+
+	gem_clean_rings(gp);
+
+	gp->rx_buf_sz = max(dev->mtu + ETH_HLEN + VLAN_HLEN,
+			    (unsigned)VLAN_ETH_FRAME_LEN);
+
+	for (i = 0; i < RX_RING_SIZE; i++) {
+		struct sk_buff *skb;
+		struct gem_rxd *rxd = &gb->rxd[i];
+
+		skb = gem_alloc_skb(dev, RX_BUF_ALLOC_SIZE(gp), GFP_KERNEL);
+		if (!skb) {
+			rxd->buffer = 0;
+			rxd->status_word = 0;
+			continue;
+		}
+
+		gp->rx_skbs[i] = skb;
+		skb_put(skb, (gp->rx_buf_sz + RX_OFFSET));
+		dma_addr = dma_map_page(&gp->pdev->dev,
+					virt_to_page(skb->data),
+					offset_in_page(skb->data),
+					RX_BUF_ALLOC_SIZE(gp),
+					DMA_FROM_DEVICE);
+		rxd->buffer = cpu_to_le64(dma_addr);
+		dma_wmb();
+		rxd->status_word = cpu_to_le64(RXDCTRL_FRESH(gp));
+		skb_reserve(skb, RX_OFFSET);
+	}
+
+	for (i = 0; i < TX_RING_SIZE; i++) {
+		struct gem_txd *txd = &gb->txd[i];
+
+		txd->control_word = 0;
+		dma_wmb();
+		txd->buffer = 0;
+	}
+	wmb();
+}
+
+/* Init PHY interface and start link poll state machine */
+static void gem_init_phy(struct gem *gp)
+{
+	u32 mifcfg;
+
+	/* Revert MIF CFG setting done on stop_phy */
+	mifcfg = readl(gp->regs + MIF_CFG);
+	mifcfg &= ~MIF_CFG_BBMODE;
+	writel(mifcfg, gp->regs + MIF_CFG);
+
+	if (gp->pdev->vendor == PCI_VENDOR_ID_APPLE) {
+		int i;
+
+		/* Those delays sucks, the HW seems to love them though, I'll
+		 * seriously consider breaking some locks here to be able
+		 * to schedule instead
+		 */
+		for (i = 0; i < 3; i++) {
+#ifdef CONFIG_PPC_PMAC
+			pmac_call_feature(PMAC_FTR_GMAC_PHY_RESET, gp->of_node, 0, 0);
+			msleep(20);
+#endif
+			/* Some PHYs used by apple have problem getting back to us,
+			 * we do an additional reset here
+			 */
+			sungem_phy_write(gp, MII_BMCR, BMCR_RESET);
+			msleep(20);
+			if (sungem_phy_read(gp, MII_BMCR) != 0xffff)
+				break;
+			if (i == 2)
+				netdev_warn(gp->dev, "GMAC PHY not responding !\n");
+		}
+	}
+
+	if (gp->pdev->vendor == PCI_VENDOR_ID_SUN &&
+	    gp->pdev->device == PCI_DEVICE_ID_SUN_GEM) {
+		u32 val;
+
+		/* Init datapath mode register. */
+		if (gp->phy_type == phy_mii_mdio0 ||
+		    gp->phy_type == phy_mii_mdio1) {
+			val = PCS_DMODE_MGM;
+		} else if (gp->phy_type == phy_serialink) {
+			val = PCS_DMODE_SM | PCS_DMODE_GMOE;
+		} else {
+			val = PCS_DMODE_ESM;
+		}
+
+		writel(val, gp->regs + PCS_DMODE);
+	}
+
+	if (gp->phy_type == phy_mii_mdio0 ||
+	    gp->phy_type == phy_mii_mdio1) {
+		/* Reset and detect MII PHY */
+		sungem_phy_probe(&gp->phy_mii, gp->mii_phy_addr);
+
+		/* Init PHY */
+		if (gp->phy_mii.def && gp->phy_mii.def->ops->init)
+			gp->phy_mii.def->ops->init(&gp->phy_mii);
+	} else {
+		gem_pcs_reset(gp);
+		gem_pcs_reinit_adv(gp);
+	}
+
+	/* Default aneg parameters */
+	gp->timer_ticks = 0;
+	gp->lstate = link_down;
+	netif_carrier_off(gp->dev);
+
+	/* Print things out */
+	if (gp->phy_type == phy_mii_mdio0 ||
+	    gp->phy_type == phy_mii_mdio1)
+		netdev_info(gp->dev, "Found %s PHY\n",
+			    gp->phy_mii.def ? gp->phy_mii.def->name : "no");
+
+	gem_begin_auto_negotiation(gp, NULL);
+}
+
+static void gem_init_dma(struct gem *gp)
+{
+	u64 desc_dma = (u64) gp->gblock_dvma;
+	u32 val;
+
+	val = (TXDMA_CFG_BASE | (0x7ff << 10) | TXDMA_CFG_PMODE);
+	writel(val, gp->regs + TXDMA_CFG);
+
+	writel(desc_dma >> 32, gp->regs + TXDMA_DBHI);
+	writel(desc_dma & 0xffffffff, gp->regs + TXDMA_DBLOW);
+	desc_dma += (INIT_BLOCK_TX_RING_SIZE * sizeof(struct gem_txd));
+
+	writel(0, gp->regs + TXDMA_KICK);
+
+	val = (RXDMA_CFG_BASE | (RX_OFFSET << 10) |
+	       (ETH_HLEN << 13) | RXDMA_CFG_FTHRESH_128);
+	writel(val, gp->regs + RXDMA_CFG);
+
+	writel(desc_dma >> 32, gp->regs + RXDMA_DBHI);
+	writel(desc_dma & 0xffffffff, gp->regs + RXDMA_DBLOW);
+
+	writel(RX_RING_SIZE - 4, gp->regs + RXDMA_KICK);
+
+	val  = (((gp->rx_pause_off / 64) << 0) & RXDMA_PTHRESH_OFF);
+	val |= (((gp->rx_pause_on / 64) << 12) & RXDMA_PTHRESH_ON);
+	writel(val, gp->regs + RXDMA_PTHRESH);
+
+	if (readl(gp->regs + GREG_BIFCFG) & GREG_BIFCFG_M66EN)
+		writel(((5 & RXDMA_BLANK_IPKTS) |
+			((8 << 12) & RXDMA_BLANK_ITIME)),
+		       gp->regs + RXDMA_BLANK);
+	else
+		writel(((5 & RXDMA_BLANK_IPKTS) |
+			((4 << 12) & RXDMA_BLANK_ITIME)),
+		       gp->regs + RXDMA_BLANK);
+}
+
+static u32 gem_setup_multicast(struct gem *gp)
+{
+	u32 rxcfg = 0;
+	int i;
+
+	if ((gp->dev->flags & IFF_ALLMULTI) ||
+	    (netdev_mc_count(gp->dev) > 256)) {
+	    	for (i=0; i<16; i++)
+			writel(0xffff, gp->regs + MAC_HASH0 + (i << 2));
+		rxcfg |= MAC_RXCFG_HFE;
+	} else if (gp->dev->flags & IFF_PROMISC) {
+		rxcfg |= MAC_RXCFG_PROM;
+	} else {
+		u16 hash_table[16];
+		u32 crc;
+		struct netdev_hw_addr *ha;
+		int i;
+
+		memset(hash_table, 0, sizeof(hash_table));
+		netdev_for_each_mc_addr(ha, gp->dev) {
+			crc = ether_crc_le(6, ha->addr);
+			crc >>= 24;
+			hash_table[crc >> 4] |= 1 << (15 - (crc & 0xf));
+		}
+	    	for (i=0; i<16; i++)
+			writel(hash_table[i], gp->regs + MAC_HASH0 + (i << 2));
+		rxcfg |= MAC_RXCFG_HFE;
+	}
+
+	return rxcfg;
+}
+
+static void gem_init_mac(struct gem *gp)
+{
+	const unsigned char *e = &gp->dev->dev_addr[0];
+
+	writel(0x1bf0, gp->regs + MAC_SNDPAUSE);
+
+	writel(0x00, gp->regs + MAC_IPG0);
+	writel(0x08, gp->regs + MAC_IPG1);
+	writel(0x04, gp->regs + MAC_IPG2);
+	writel(0x40, gp->regs + MAC_STIME);
+	writel(0x40, gp->regs + MAC_MINFSZ);
+
+	/* Ethernet payload + header + FCS + optional VLAN tag. */
+	writel(0x20000000 | (gp->rx_buf_sz + 4), gp->regs + MAC_MAXFSZ);
+
+	writel(0x07, gp->regs + MAC_PASIZE);
+	writel(0x04, gp->regs + MAC_JAMSIZE);
+	writel(0x10, gp->regs + MAC_ATTLIM);
+	writel(0x8808, gp->regs + MAC_MCTYPE);
+
+	writel((e[5] | (e[4] << 8)) & 0x3ff, gp->regs + MAC_RANDSEED);
+
+	writel((e[4] << 8) | e[5], gp->regs + MAC_ADDR0);
+	writel((e[2] << 8) | e[3], gp->regs + MAC_ADDR1);
+	writel((e[0] << 8) | e[1], gp->regs + MAC_ADDR2);
+
+	writel(0, gp->regs + MAC_ADDR3);
+	writel(0, gp->regs + MAC_ADDR4);
+	writel(0, gp->regs + MAC_ADDR5);
+
+	writel(0x0001, gp->regs + MAC_ADDR6);
+	writel(0xc200, gp->regs + MAC_ADDR7);
+	writel(0x0180, gp->regs + MAC_ADDR8);
+
+	writel(0, gp->regs + MAC_AFILT0);
+	writel(0, gp->regs + MAC_AFILT1);
+	writel(0, gp->regs + MAC_AFILT2);
+	writel(0, gp->regs + MAC_AF21MSK);
+	writel(0, gp->regs + MAC_AF0MSK);
+
+	gp->mac_rx_cfg = gem_setup_multicast(gp);
+#ifdef STRIP_FCS
+	gp->mac_rx_cfg |= MAC_RXCFG_SFCS;
+#endif
+	writel(0, gp->regs + MAC_NCOLL);
+	writel(0, gp->regs + MAC_FASUCC);
+	writel(0, gp->regs + MAC_ECOLL);
+	writel(0, gp->regs + MAC_LCOLL);
+	writel(0, gp->regs + MAC_DTIMER);
+	writel(0, gp->regs + MAC_PATMPS);
+	writel(0, gp->regs + MAC_RFCTR);
+	writel(0, gp->regs + MAC_LERR);
+	writel(0, gp->regs + MAC_AERR);
+	writel(0, gp->regs + MAC_FCSERR);
+	writel(0, gp->regs + MAC_RXCVERR);
+
+	/* Clear RX/TX/MAC/XIF config, we will set these up and enable
+	 * them once a link is established.
+	 */
+	writel(0, gp->regs + MAC_TXCFG);
+	writel(gp->mac_rx_cfg, gp->regs + MAC_RXCFG);
+	writel(0, gp->regs + MAC_MCCFG);
+	writel(0, gp->regs + MAC_XIFCFG);
+
+	/* Setup MAC interrupts.  We want to get all of the interesting
+	 * counter expiration events, but we do not want to hear about
+	 * normal rx/tx as the DMA engine tells us that.
+	 */
+	writel(MAC_TXSTAT_XMIT, gp->regs + MAC_TXMASK);
+	writel(MAC_RXSTAT_RCV, gp->regs + MAC_RXMASK);
+
+	/* Don't enable even the PAUSE interrupts for now, we
+	 * make no use of those events other than to record them.
+	 */
+	writel(0xffffffff, gp->regs + MAC_MCMASK);
+
+	/* Don't enable GEM's WOL in normal operations
+	 */
+	if (gp->has_wol)
+		writel(0, gp->regs + WOL_WAKECSR);
+}
+
+static void gem_init_pause_thresholds(struct gem *gp)
+{
+	u32 cfg;
+
+	/* Calculate pause thresholds.  Setting the OFF threshold to the
+	 * full RX fifo size effectively disables PAUSE generation which
+	 * is what we do for 10/100 only GEMs which have FIFOs too small
+	 * to make real gains from PAUSE.
+	 */
+	if (gp->rx_fifo_sz <= (2 * 1024)) {
+		gp->rx_pause_off = gp->rx_pause_on = gp->rx_fifo_sz;
+	} else {
+		int max_frame = (gp->rx_buf_sz + 4 + 64) & ~63;
+		int off = (gp->rx_fifo_sz - (max_frame * 2));
+		int on = off - max_frame;
+
+		gp->rx_pause_off = off;
+		gp->rx_pause_on = on;
+	}
+
+
+	/* Configure the chip "burst" DMA mode & enable some
+	 * HW bug fixes on Apple version
+	 */
+	cfg  = 0;
+	if (gp->pdev->vendor == PCI_VENDOR_ID_APPLE)
+		cfg |= GREG_CFG_RONPAULBIT | GREG_CFG_ENBUG2FIX;
+#if !defined(CONFIG_SPARC64) && !defined(CONFIG_ALPHA)
+	cfg |= GREG_CFG_IBURST;
+#endif
+	cfg |= ((31 << 1) & GREG_CFG_TXDMALIM);
+	cfg |= ((31 << 6) & GREG_CFG_RXDMALIM);
+	writel(cfg, gp->regs + GREG_CFG);
+
+	/* If Infinite Burst didn't stick, then use different
+	 * thresholds (and Apple bug fixes don't exist)
+	 */
+	if (!(readl(gp->regs + GREG_CFG) & GREG_CFG_IBURST)) {
+		cfg = ((2 << 1) & GREG_CFG_TXDMALIM);
+		cfg |= ((8 << 6) & GREG_CFG_RXDMALIM);
+		writel(cfg, gp->regs + GREG_CFG);
+	}
+}
+
+static int gem_check_invariants(struct gem *gp)
+{
+	struct pci_dev *pdev = gp->pdev;
+	u32 mif_cfg;
+
+	/* On Apple's sungem, we can't rely on registers as the chip
+	 * was been powered down by the firmware. The PHY is looked
+	 * up later on.
+	 */
+	if (pdev->vendor == PCI_VENDOR_ID_APPLE) {
+		gp->phy_type = phy_mii_mdio0;
+		gp->tx_fifo_sz = readl(gp->regs + TXDMA_FSZ) * 64;
+		gp->rx_fifo_sz = readl(gp->regs + RXDMA_FSZ) * 64;
+		gp->swrst_base = 0;
+
+		mif_cfg = readl(gp->regs + MIF_CFG);
+		mif_cfg &= ~(MIF_CFG_PSELECT|MIF_CFG_POLL|MIF_CFG_BBMODE|MIF_CFG_MDI1);
+		mif_cfg |= MIF_CFG_MDI0;
+		writel(mif_cfg, gp->regs + MIF_CFG);
+		writel(PCS_DMODE_MGM, gp->regs + PCS_DMODE);
+		writel(MAC_XIFCFG_OE, gp->regs + MAC_XIFCFG);
+
+		/* We hard-code the PHY address so we can properly bring it out of
+		 * reset later on, we can't really probe it at this point, though
+		 * that isn't an issue.
+		 */
+		if (gp->pdev->device == PCI_DEVICE_ID_APPLE_K2_GMAC)
+			gp->mii_phy_addr = 1;
+		else
+			gp->mii_phy_addr = 0;
+
+		return 0;
+	}
+
+	mif_cfg = readl(gp->regs + MIF_CFG);
+
+	if (pdev->vendor == PCI_VENDOR_ID_SUN &&
+	    pdev->device == PCI_DEVICE_ID_SUN_RIO_GEM) {
+		/* One of the MII PHYs _must_ be present
+		 * as this chip has no gigabit PHY.
+		 */
+		if ((mif_cfg & (MIF_CFG_MDI0 | MIF_CFG_MDI1)) == 0) {
+			pr_err("RIO GEM lacks MII phy, mif_cfg[%08x]\n",
+			       mif_cfg);
+			return -1;
+		}
+	}
+
+	/* Determine initial PHY interface type guess.  MDIO1 is the
+	 * external PHY and thus takes precedence over MDIO0.
+	 */
+
+	if (mif_cfg & MIF_CFG_MDI1) {
+		gp->phy_type = phy_mii_mdio1;
+		mif_cfg |= MIF_CFG_PSELECT;
+		writel(mif_cfg, gp->regs + MIF_CFG);
+	} else if (mif_cfg & MIF_CFG_MDI0) {
+		gp->phy_type = phy_mii_mdio0;
+		mif_cfg &= ~MIF_CFG_PSELECT;
+		writel(mif_cfg, gp->regs + MIF_CFG);
+	} else {
+#ifdef CONFIG_SPARC
+		const char *p;
+
+		p = of_get_property(gp->of_node, "shared-pins", NULL);
+		if (p && !strcmp(p, "serdes"))
+			gp->phy_type = phy_serdes;
+		else
+#endif
+			gp->phy_type = phy_serialink;
+	}
+	if (gp->phy_type == phy_mii_mdio1 ||
+	    gp->phy_type == phy_mii_mdio0) {
+		int i;
+
+		for (i = 0; i < 32; i++) {
+			gp->mii_phy_addr = i;
+			if (sungem_phy_read(gp, MII_BMCR) != 0xffff)
+				break;
+		}
+		if (i == 32) {
+			if (pdev->device != PCI_DEVICE_ID_SUN_GEM) {
+				pr_err("RIO MII phy will not respond\n");
+				return -1;
+			}
+			gp->phy_type = phy_serdes;
+		}
+	}
+
+	/* Fetch the FIFO configurations now too. */
+	gp->tx_fifo_sz = readl(gp->regs + TXDMA_FSZ) * 64;
+	gp->rx_fifo_sz = readl(gp->regs + RXDMA_FSZ) * 64;
+
+	if (pdev->vendor == PCI_VENDOR_ID_SUN) {
+		if (pdev->device == PCI_DEVICE_ID_SUN_GEM) {
+			if (gp->tx_fifo_sz != (9 * 1024) ||
+			    gp->rx_fifo_sz != (20 * 1024)) {
+				pr_err("GEM has bogus fifo sizes tx(%d) rx(%d)\n",
+				       gp->tx_fifo_sz, gp->rx_fifo_sz);
+				return -1;
+			}
+			gp->swrst_base = 0;
+		} else {
+			if (gp->tx_fifo_sz != (2 * 1024) ||
+			    gp->rx_fifo_sz != (2 * 1024)) {
+				pr_err("RIO GEM has bogus fifo sizes tx(%d) rx(%d)\n",
+				       gp->tx_fifo_sz, gp->rx_fifo_sz);
+				return -1;
+			}
+			gp->swrst_base = (64 / 4) << GREG_SWRST_CACHE_SHIFT;
+		}
+	}
+
+	return 0;
+}
+
+static void gem_reinit_chip(struct gem *gp)
+{
+	/* Reset the chip */
+	gem_reset(gp);
+
+	/* Make sure ints are disabled */
+	gem_disable_ints(gp);
+
+	/* Allocate & setup ring buffers */
+	gem_init_rings(gp);
+
+	/* Configure pause thresholds */
+	gem_init_pause_thresholds(gp);
+
+	/* Init DMA & MAC engines */
+	gem_init_dma(gp);
+	gem_init_mac(gp);
+}
+
+
+static void gem_stop_phy(struct gem *gp, int wol)
+{
+	u32 mifcfg;
+
+	/* Let the chip settle down a bit, it seems that helps
+	 * for sleep mode on some models
+	 */
+	msleep(10);
+
+	/* Make sure we aren't polling PHY status change. We
+	 * don't currently use that feature though
+	 */
+	mifcfg = readl(gp->regs + MIF_CFG);
+	mifcfg &= ~MIF_CFG_POLL;
+	writel(mifcfg, gp->regs + MIF_CFG);
+
+	if (wol && gp->has_wol) {
+		const unsigned char *e = &gp->dev->dev_addr[0];
+		u32 csr;
+
+		/* Setup wake-on-lan for MAGIC packet */
+		writel(MAC_RXCFG_HFE | MAC_RXCFG_SFCS | MAC_RXCFG_ENAB,
+		       gp->regs + MAC_RXCFG);
+		writel((e[4] << 8) | e[5], gp->regs + WOL_MATCH0);
+		writel((e[2] << 8) | e[3], gp->regs + WOL_MATCH1);
+		writel((e[0] << 8) | e[1], gp->regs + WOL_MATCH2);
+
+		writel(WOL_MCOUNT_N | WOL_MCOUNT_M, gp->regs + WOL_MCOUNT);
+		csr = WOL_WAKECSR_ENABLE;
+		if ((readl(gp->regs + MAC_XIFCFG) & MAC_XIFCFG_GMII) == 0)
+			csr |= WOL_WAKECSR_MII;
+		writel(csr, gp->regs + WOL_WAKECSR);
+	} else {
+		writel(0, gp->regs + MAC_RXCFG);
+		(void)readl(gp->regs + MAC_RXCFG);
+		/* Machine sleep will die in strange ways if we
+		 * dont wait a bit here, looks like the chip takes
+		 * some time to really shut down
+		 */
+		msleep(10);
+	}
+
+	writel(0, gp->regs + MAC_TXCFG);
+	writel(0, gp->regs + MAC_XIFCFG);
+	writel(0, gp->regs + TXDMA_CFG);
+	writel(0, gp->regs + RXDMA_CFG);
+
+	if (!wol) {
+		gem_reset(gp);
+		writel(MAC_TXRST_CMD, gp->regs + MAC_TXRST);
+		writel(MAC_RXRST_CMD, gp->regs + MAC_RXRST);
+
+		if (found_mii_phy(gp) && gp->phy_mii.def->ops->suspend)
+			gp->phy_mii.def->ops->suspend(&gp->phy_mii);
+
+		/* According to Apple, we must set the MDIO pins to this begnign
+		 * state or we may 1) eat more current, 2) damage some PHYs
+		 */
+		writel(mifcfg | MIF_CFG_BBMODE, gp->regs + MIF_CFG);
+		writel(0, gp->regs + MIF_BBCLK);
+		writel(0, gp->regs + MIF_BBDATA);
+		writel(0, gp->regs + MIF_BBOENAB);
+		writel(MAC_XIFCFG_GMII | MAC_XIFCFG_LBCK, gp->regs + MAC_XIFCFG);
+		(void) readl(gp->regs + MAC_XIFCFG);
+	}
+}
+
+static int gem_do_start(struct net_device *dev)
+{
+	struct gem *gp = netdev_priv(dev);
+	int rc;
+
+	pci_set_master(gp->pdev);
+
+	/* Init & setup chip hardware */
+	gem_reinit_chip(gp);
+
+	/* An interrupt might come in handy */
+	rc = request_irq(gp->pdev->irq, gem_interrupt,
+			 IRQF_SHARED, dev->name, (void *)dev);
+	if (rc) {
+		netdev_err(dev, "failed to request irq !\n");
+
+		gem_reset(gp);
+		gem_clean_rings(gp);
+		gem_put_cell(gp);
+		return rc;
+	}
+
+	/* Mark us as attached again if we come from resume(), this has
+	 * no effect if we weren't detached and needs to be done now.
+	 */
+	netif_device_attach(dev);
+
+	/* Restart NAPI & queues */
+	gem_netif_start(gp);
+
+	/* Detect & init PHY, start autoneg etc... this will
+	 * eventually result in starting DMA operations when
+	 * the link is up
+	 */
+	gem_init_phy(gp);
+
+	return 0;
+}
+
+static void gem_do_stop(struct net_device *dev, int wol)
+{
+	struct gem *gp = netdev_priv(dev);
+
+	/* Stop NAPI and stop tx queue */
+	gem_netif_stop(gp);
+
+	/* Make sure ints are disabled. We don't care about
+	 * synchronizing as NAPI is disabled, thus a stray
+	 * interrupt will do nothing bad (our irq handler
+	 * just schedules NAPI)
+	 */
+	gem_disable_ints(gp);
+
+	/* Stop the link timer */
+	timer_delete_sync(&gp->link_timer);
+
+	/* We cannot cancel the reset task while holding the
+	 * rtnl lock, we'd get an A->B / B->A deadlock stituation
+	 * if we did. This is not an issue however as the reset
+	 * task is synchronized vs. us (rtnl_lock) and will do
+	 * nothing if the device is down or suspended. We do
+	 * still clear reset_task_pending to avoid a spurious
+	 * reset later on in case we do resume before it gets
+	 * scheduled.
+	 */
+	gp->reset_task_pending = 0;
+
+	/* If we are going to sleep with WOL */
+	gem_stop_dma(gp);
+	msleep(10);
+	if (!wol)
+		gem_reset(gp);
+	msleep(10);
+
+	/* Get rid of rings */
+	gem_clean_rings(gp);
+
+	/* No irq needed anymore */
+	free_irq(gp->pdev->irq, (void *) dev);
+
+	/* Shut the PHY down eventually and setup WOL */
+	gem_stop_phy(gp, wol);
+}
+
+static void gem_reset_task(struct work_struct *work)
+{
+	struct gem *gp = container_of(work, struct gem, reset_task);
+
+	/* Lock out the network stack (essentially shield ourselves
+	 * against a racing open, close, control call, or suspend
+	 */
+	rtnl_lock();
+
+	/* Skip the reset task if suspended or closed, or if it's
+	 * been cancelled by gem_do_stop (see comment there)
+	 */
+	if (!netif_device_present(gp->dev) ||
+	    !netif_running(gp->dev) ||
+	    !gp->reset_task_pending) {
+		rtnl_unlock();
+		return;
+	}
+
+	/* Stop the link timer */
+	timer_delete_sync(&gp->link_timer);
+
+	/* Stop NAPI and tx */
+	gem_netif_stop(gp);
+
+	/* Reset the chip & rings */
+	gem_reinit_chip(gp);
+	if (gp->lstate == link_up)
+		gem_set_link_modes(gp);
+
+	/* Restart NAPI and Tx */
+	gem_netif_start(gp);
+
+	/* We are back ! */
+	gp->reset_task_pending = 0;
+
+	/* If the link is not up, restart autoneg, else restart the
+	 * polling timer
+	 */
+	if (gp->lstate != link_up)
+		gem_begin_auto_negotiation(gp, NULL);
+	else
+		mod_timer(&gp->link_timer, jiffies + ((12 * HZ) / 10));
+
+	rtnl_unlock();
+}
+
+static int gem_open(struct net_device *dev)
+{
+	struct gem *gp = netdev_priv(dev);
+	int rc;
+
+	/* We allow open while suspended, we just do nothing,
+	 * the chip will be initialized in resume()
+	 */
+	if (netif_device_present(dev)) {
+		/* Enable the cell */
+		gem_get_cell(gp);
+
+		/* Make sure PCI access and bus master are enabled */
+		rc = pci_enable_device(gp->pdev);
+		if (rc) {
+			netdev_err(dev, "Failed to enable chip on PCI bus !\n");
+
+			/* Put cell and forget it for now, it will be considered
+			 *as still asleep, a new sleep cycle may bring it back
+			 */
+			gem_put_cell(gp);
+			return -ENXIO;
+		}
+		return gem_do_start(dev);
+	}
+
+	return 0;
+}
+
+static int gem_close(struct net_device *dev)
+{
+	struct gem *gp = netdev_priv(dev);
+
+	if (netif_device_present(dev)) {
+		gem_do_stop(dev, 0);
+
+		/* Make sure bus master is disabled */
+		pci_disable_device(gp->pdev);
+
+		/* Cell not needed neither if no WOL */
+		if (!gp->asleep_wol)
+			gem_put_cell(gp);
+	}
+	return 0;
+}
+
+static int __maybe_unused gem_suspend(struct device *dev_d)
+{
+	struct net_device *dev = dev_get_drvdata(dev_d);
+	struct gem *gp = netdev_priv(dev);
+
+	/* Lock the network stack first to avoid racing with open/close,
+	 * reset task and setting calls
+	 */
+	rtnl_lock();
+
+	/* Not running, mark ourselves non-present, no need for
+	 * a lock here
+	 */
+	if (!netif_running(dev)) {
+		netif_device_detach(dev);
+		rtnl_unlock();
+		return 0;
+	}
+	netdev_info(dev, "suspending, WakeOnLan %s\n",
+		    (gp->wake_on_lan && netif_running(dev)) ?
+		    "enabled" : "disabled");
+
+	/* Tell the network stack we're gone. gem_do_stop() below will
+	 * synchronize with TX, stop NAPI etc...
+	 */
+	netif_device_detach(dev);
+
+	/* Switch off chip, remember WOL setting */
+	gp->asleep_wol = !!gp->wake_on_lan;
+	gem_do_stop(dev, gp->asleep_wol);
+
+	/* Cell not needed neither if no WOL */
+	if (!gp->asleep_wol)
+		gem_put_cell(gp);
+
+	/* Unlock the network stack */
+	rtnl_unlock();
+
+	return 0;
+}
+
+static int __maybe_unused gem_resume(struct device *dev_d)
+{
+	struct net_device *dev = dev_get_drvdata(dev_d);
+	struct gem *gp = netdev_priv(dev);
+
+	/* See locking comment in gem_suspend */
+	rtnl_lock();
+
+	/* Not running, mark ourselves present, no need for
+	 * a lock here
+	 */
+	if (!netif_running(dev)) {
+		netif_device_attach(dev);
+		rtnl_unlock();
+		return 0;
+	}
+
+	/* Enable the cell */
+	gem_get_cell(gp);
+
+	/* Restart chip. If that fails there isn't much we can do, we
+	 * leave things stopped.
+	 */
+	gem_do_start(dev);
+
+	/* If we had WOL enabled, the cell clock was never turned off during
+	 * sleep, so we end up being unbalanced. Fix that here
+	 */
+	if (gp->asleep_wol)
+		gem_put_cell(gp);
+
+	/* Unlock the network stack */
+	rtnl_unlock();
+
+	return 0;
+}
+
+static struct net_device_stats *gem_get_stats(struct net_device *dev)
+{
+	struct gem *gp = netdev_priv(dev);
+
+	/* I have seen this being called while the PM was in progress,
+	 * so we shield against this. Let's also not poke at registers
+	 * while the reset task is going on.
+	 *
+	 * TODO: Move stats collection elsewhere (link timer ?) and
+	 * make this a nop to avoid all those synchro issues
+	 */
+	if (!netif_device_present(dev) || !netif_running(dev))
+		goto bail;
+
+	/* Better safe than sorry... */
+	if (WARN_ON(!gp->cell_enabled))
+		goto bail;
+
+	dev->stats.rx_crc_errors += readl(gp->regs + MAC_FCSERR);
+	writel(0, gp->regs + MAC_FCSERR);
+
+	dev->stats.rx_frame_errors += readl(gp->regs + MAC_AERR);
+	writel(0, gp->regs + MAC_AERR);
+
+	dev->stats.rx_length_errors += readl(gp->regs + MAC_LERR);
+	writel(0, gp->regs + MAC_LERR);
+
+	dev->stats.tx_aborted_errors += readl(gp->regs + MAC_ECOLL);
+	dev->stats.collisions +=
+		(readl(gp->regs + MAC_ECOLL) + readl(gp->regs + MAC_LCOLL));
+	writel(0, gp->regs + MAC_ECOLL);
+	writel(0, gp->regs + MAC_LCOLL);
+ bail:
+	return &dev->stats;
+}
+
+static int gem_set_mac_address(struct net_device *dev, void *addr)
+{
+	struct sockaddr *macaddr = (struct sockaddr *) addr;
+	const unsigned char *e = &dev->dev_addr[0];
+	struct gem *gp = netdev_priv(dev);
+
+	if (!is_valid_ether_addr(macaddr->sa_data))
+		return -EADDRNOTAVAIL;
+
+	eth_hw_addr_set(dev, macaddr->sa_data);
+
+	/* We'll just catch it later when the device is up'd or resumed */
+	if (!netif_running(dev) || !netif_device_present(dev))
+		return 0;
+
+	/* Better safe than sorry... */
+	if (WARN_ON(!gp->cell_enabled))
+		return 0;
+
+	writel((e[4] << 8) | e[5], gp->regs + MAC_ADDR0);
+	writel((e[2] << 8) | e[3], gp->regs + MAC_ADDR1);
+	writel((e[0] << 8) | e[1], gp->regs + MAC_ADDR2);
+
+	return 0;
+}
+
+static void gem_set_multicast(struct net_device *dev)
+{
+	struct gem *gp = netdev_priv(dev);
+	u32 rxcfg, rxcfg_new;
+	int limit = 10000;
+
+	if (!netif_running(dev) || !netif_device_present(dev))
+		return;
+
+	/* Better safe than sorry... */
+	if (gp->reset_task_pending || WARN_ON(!gp->cell_enabled))
+		return;
+
+	rxcfg = readl(gp->regs + MAC_RXCFG);
+	rxcfg_new = gem_setup_multicast(gp);
+#ifdef STRIP_FCS
+	rxcfg_new |= MAC_RXCFG_SFCS;
+#endif
+	gp->mac_rx_cfg = rxcfg_new;
+
+	writel(rxcfg & ~MAC_RXCFG_ENAB, gp->regs + MAC_RXCFG);
+	while (readl(gp->regs + MAC_RXCFG) & MAC_RXCFG_ENAB) {
+		if (!limit--)
+			break;
+		udelay(10);
+	}
+
+	rxcfg &= ~(MAC_RXCFG_PROM | MAC_RXCFG_HFE);
+	rxcfg |= rxcfg_new;
+
+	writel(rxcfg, gp->regs + MAC_RXCFG);
+}
+
+/* Jumbo-grams don't seem to work :-( */
+#define GEM_MIN_MTU	ETH_MIN_MTU
+#if 1
+#define GEM_MAX_MTU	ETH_DATA_LEN
+#else
+#define GEM_MAX_MTU	9000
+#endif
+
+static int gem_change_mtu(struct net_device *dev, int new_mtu)
+{
+	struct gem *gp = netdev_priv(dev);
+
+	WRITE_ONCE(dev->mtu, new_mtu);
+
+	/* We'll just catch it later when the device is up'd or resumed */
+	if (!netif_running(dev) || !netif_device_present(dev))
+		return 0;
+
+	/* Better safe than sorry... */
+	if (WARN_ON(!gp->cell_enabled))
+		return 0;
+
+	gem_netif_stop(gp);
+	gem_reinit_chip(gp);
+	if (gp->lstate == link_up)
+		gem_set_link_modes(gp);
+	gem_netif_start(gp);
+
+	return 0;
+}
+
+static void gem_get_drvinfo(struct net_device *dev, struct ethtool_drvinfo *info)
+{
+	struct gem *gp = netdev_priv(dev);
+
+	strscpy(info->driver, DRV_NAME, sizeof(info->driver));
+	strscpy(info->version, DRV_VERSION, sizeof(info->version));
+	strscpy(info->bus_info, pci_name(gp->pdev), sizeof(info->bus_info));
+}
+
+static int gem_get_link_ksettings(struct net_device *dev,
+				  struct ethtool_link_ksettings *cmd)
+{
+	struct gem *gp = netdev_priv(dev);
+	u32 supported, advertising;
+
+	if (gp->phy_type == phy_mii_mdio0 ||
+	    gp->phy_type == phy_mii_mdio1) {
+		if (gp->phy_mii.def)
+			supported = gp->phy_mii.def->features;
+		else
+			supported = (SUPPORTED_10baseT_Half |
+					  SUPPORTED_10baseT_Full);
+
+		/* XXX hardcoded stuff for now */
+		cmd->base.port = PORT_MII;
+		cmd->base.phy_address = 0; /* XXX fixed PHYAD */
+
+		/* Return current PHY settings */
+		cmd->base.autoneg = gp->want_autoneg;
+		cmd->base.speed = gp->phy_mii.speed;
+		cmd->base.duplex = gp->phy_mii.duplex;
+		advertising = gp->phy_mii.advertising;
+
+		/* If we started with a forced mode, we don't have a default
+		 * advertise set, we need to return something sensible so
+		 * userland can re-enable autoneg properly.
+		 */
+		if (advertising == 0)
+			advertising = supported;
+	} else { // XXX PCS ?
+		supported =
+			(SUPPORTED_10baseT_Half | SUPPORTED_10baseT_Full |
+			 SUPPORTED_100baseT_Half | SUPPORTED_100baseT_Full |
+			 SUPPORTED_Autoneg);
+		advertising = supported;
+		cmd->base.speed = 0;
+		cmd->base.duplex = 0;
+		cmd->base.port = 0;
+		cmd->base.phy_address = 0;
+		cmd->base.autoneg = 0;
+
+		/* serdes means usually a Fibre connector, with most fixed */
+		if (gp->phy_type == phy_serdes) {
+			cmd->base.port = PORT_FIBRE;
+			supported = (SUPPORTED_1000baseT_Half |
+				SUPPORTED_1000baseT_Full |
+				SUPPORTED_FIBRE | SUPPORTED_Autoneg |
+				SUPPORTED_Pause | SUPPORTED_Asym_Pause);
+			advertising = supported;
+			if (gp->lstate == link_up)
+				cmd->base.speed = SPEED_1000;
+			cmd->base.duplex = DUPLEX_FULL;
+			cmd->base.autoneg = 1;
+		}
+	}
+
+	ethtool_convert_legacy_u32_to_link_mode(cmd->link_modes.supported,
+						supported);
+	ethtool_convert_legacy_u32_to_link_mode(cmd->link_modes.advertising,
+						advertising);
+
+	return 0;
+}
+
+static int gem_set_link_ksettings(struct net_device *dev,
+				  const struct ethtool_link_ksettings *cmd)
+{
+	struct gem *gp = netdev_priv(dev);
+	u32 speed = cmd->base.speed;
+	u32 advertising;
+
+	ethtool_convert_link_mode_to_legacy_u32(&advertising,
+						cmd->link_modes.advertising);
+
+	/* Verify the settings we care about. */
+	if (cmd->base.autoneg != AUTONEG_ENABLE &&
+	    cmd->base.autoneg != AUTONEG_DISABLE)
+		return -EINVAL;
+
+	if (cmd->base.autoneg == AUTONEG_ENABLE &&
+	    advertising == 0)
+		return -EINVAL;
+
+	if (cmd->base.autoneg == AUTONEG_DISABLE &&
+	    ((speed != SPEED_1000 &&
+	      speed != SPEED_100 &&
+	      speed != SPEED_10) ||
+	     (cmd->base.duplex != DUPLEX_HALF &&
+	      cmd->base.duplex != DUPLEX_FULL)))
+		return -EINVAL;
+
+	/* Apply settings and restart link process. */
+	if (netif_device_present(gp->dev)) {
+		timer_delete_sync(&gp->link_timer);
+		gem_begin_auto_negotiation(gp, cmd);
+	}
+
+	return 0;
+}
+
+static int gem_nway_reset(struct net_device *dev)
+{
+	struct gem *gp = netdev_priv(dev);
+
+	if (!gp->want_autoneg)
+		return -EINVAL;
+
+	/* Restart link process  */
+	if (netif_device_present(gp->dev)) {
+		timer_delete_sync(&gp->link_timer);
+		gem_begin_auto_negotiation(gp, NULL);
+	}
+
+	return 0;
+}
+
+static u32 gem_get_msglevel(struct net_device *dev)
+{
+	struct gem *gp = netdev_priv(dev);
+	return gp->msg_enable;
+}
+
+static void gem_set_msglevel(struct net_device *dev, u32 value)
+{
+	struct gem *gp = netdev_priv(dev);
+	gp->msg_enable = value;
+}
+
+
+/* Add more when I understand how to program the chip */
+/* like WAKE_UCAST | WAKE_MCAST | WAKE_BCAST */
+
+#define WOL_SUPPORTED_MASK	(WAKE_MAGIC)
+
+static void gem_get_wol(struct net_device *dev, struct ethtool_wolinfo *wol)
+{
+	struct gem *gp = netdev_priv(dev);
+
+	/* Add more when I understand how to program the chip */
+	if (gp->has_wol) {
+		wol->supported = WOL_SUPPORTED_MASK;
+		wol->wolopts = gp->wake_on_lan;
+	} else {
+		wol->supported = 0;
+		wol->wolopts = 0;
+	}
+}
+
+static int gem_set_wol(struct net_device *dev, struct ethtool_wolinfo *wol)
+{
+	struct gem *gp = netdev_priv(dev);
+
+	if (!gp->has_wol)
+		return -EOPNOTSUPP;
+	gp->wake_on_lan = wol->wolopts & WOL_SUPPORTED_MASK;
+	return 0;
+}
+
+static const struct ethtool_ops gem_ethtool_ops = {
+	.get_drvinfo		= gem_get_drvinfo,
+	.get_link		= ethtool_op_get_link,
+	.nway_reset		= gem_nway_reset,
+	.get_msglevel		= gem_get_msglevel,
+	.set_msglevel		= gem_set_msglevel,
+	.get_wol		= gem_get_wol,
+	.set_wol		= gem_set_wol,
+	.get_link_ksettings	= gem_get_link_ksettings,
+	.set_link_ksettings	= gem_set_link_ksettings,
+};
+
+static int gem_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
+{
+	struct gem *gp = netdev_priv(dev);
+	struct mii_ioctl_data *data = if_mii(ifr);
+	int rc = -EOPNOTSUPP;
+
+	/* For SIOCGMIIREG and SIOCSMIIREG the core checks for us that
+	 * netif_device_present() is true and holds rtnl_lock for us
+	 * so we have nothing to worry about
+	 */
+
+	switch (cmd) {
+	case SIOCGMIIPHY:		/* Get address of MII PHY in use. */
+		data->phy_id = gp->mii_phy_addr;
+		fallthrough;
+
+	case SIOCGMIIREG:		/* Read MII PHY register. */
+		data->val_out = __sungem_phy_read(gp, data->phy_id & 0x1f,
+					   data->reg_num & 0x1f);
+		rc = 0;
+		break;
+
+	case SIOCSMIIREG:		/* Write MII PHY register. */
+		__sungem_phy_write(gp, data->phy_id & 0x1f, data->reg_num & 0x1f,
+			    data->val_in);
+		rc = 0;
+		break;
+	}
+	return rc;
+}
+
+#if (!defined(CONFIG_SPARC) && !defined(CONFIG_PPC_PMAC))
+/* Fetch MAC address from vital product data of PCI ROM. */
+static int find_eth_addr_in_vpd(void __iomem *rom_base, int len, unsigned char *dev_addr)
+{
+	int this_offset;
+
+	for (this_offset = 0x20; this_offset < len; this_offset++) {
+		void __iomem *p = rom_base + this_offset;
+		int i;
+
+		if (readb(p + 0) != 0x90 ||
+		    readb(p + 1) != 0x00 ||
+		    readb(p + 2) != 0x09 ||
+		    readb(p + 3) != 0x4e ||
+		    readb(p + 4) != 0x41 ||
+		    readb(p + 5) != 0x06)
+			continue;
+
+		this_offset += 6;
+		p += 6;
+
+		for (i = 0; i < 6; i++)
+			dev_addr[i] = readb(p + i);
+		return 1;
+	}
+	return 0;
+}
+
+static void get_gem_mac_nonobp(struct pci_dev *pdev, unsigned char *dev_addr)
+{
+	size_t size;
+	void __iomem *p = pci_map_rom(pdev, &size);
+
+	if (p) {
+		int found;
+
+		found = readb(p) == 0x55 &&
+			readb(p + 1) == 0xaa &&
+			find_eth_addr_in_vpd(p, (64 * 1024), dev_addr);
+		pci_unmap_rom(pdev, p);
+		if (found)
+			return;
+	}
+
+	/* Sun MAC prefix then 3 random bytes. */
+	dev_addr[0] = 0x08;
+	dev_addr[1] = 0x00;
+	dev_addr[2] = 0x20;
+	get_random_bytes(dev_addr + 3, 3);
+}
+#endif /* not Sparc and not PPC */
+
+static int gem_get_device_address(struct gem *gp)
+{
+#if defined(CONFIG_SPARC) || defined(CONFIG_PPC_PMAC)
+	struct net_device *dev = gp->dev;
+	const unsigned char *addr;
+
+	addr = of_get_property(gp->of_node, "local-mac-address", NULL);
+	if (addr == NULL) {
+#ifdef CONFIG_SPARC
+		addr = idprom->id_ethaddr;
+#else
+		printk("\n");
+		pr_err("%s: can't get mac-address\n", dev->name);
+		return -1;
+#endif
+	}
+	eth_hw_addr_set(dev, addr);
+#else
+	u8 addr[ETH_ALEN];
+
+	get_gem_mac_nonobp(gp->pdev, addr);
+	eth_hw_addr_set(gp->dev, addr);
+#endif
+	return 0;
+}
+
+static void gem_remove_one(struct pci_dev *pdev)
+{
+	struct net_device *dev = pci_get_drvdata(pdev);
+
+	if (dev) {
+		struct gem *gp = netdev_priv(dev);
+
+		unregister_netdev(dev);
+
+		/* Ensure reset task is truly gone */
+		cancel_work_sync(&gp->reset_task);
+
+		/* Free resources */
+		dma_free_coherent(&pdev->dev, sizeof(struct gem_init_block),
+				  gp->init_block, gp->gblock_dvma);
+		iounmap(gp->regs);
+		pci_release_regions(pdev);
+		free_netdev(dev);
+	}
+}
+
+static const struct net_device_ops gem_netdev_ops = {
+	.ndo_open		= gem_open,
+	.ndo_stop		= gem_close,
+	.ndo_start_xmit		= gem_start_xmit,
+	.ndo_get_stats		= gem_get_stats,
+	.ndo_set_rx_mode	= gem_set_multicast,
+	.ndo_eth_ioctl		= gem_ioctl,
+	.ndo_tx_timeout		= gem_tx_timeout,
+	.ndo_change_mtu		= gem_change_mtu,
+	.ndo_validate_addr	= eth_validate_addr,
+	.ndo_set_mac_address    = gem_set_mac_address,
+};
+
+static int gem_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
+{
+	unsigned long gemreg_base, gemreg_len;
+	struct net_device *dev;
+	struct gem *gp;
+	int err, pci_using_dac;
+
+	printk_once(KERN_INFO "%s", version);
+
+	/* Apple gmac note: during probe, the chip is powered up by
+	 * the arch code to allow the code below to work (and to let
+	 * the chip be probed on the config space. It won't stay powered
+	 * up until the interface is brought up however, so we can't rely
+	 * on register configuration done at this point.
+	 */
+	err = pci_enable_device(pdev);
+	if (err) {
+		pr_err("Cannot enable MMIO operation, aborting\n");
+		return err;
+	}
+	pci_set_master(pdev);
+
+	/* Configure DMA attributes. */
+
+	/* All of the GEM documentation states that 64-bit DMA addressing
+	 * is fully supported and should work just fine.  However the
+	 * front end for RIO based GEMs is different and only supports
+	 * 32-bit addressing.
+	 *
+	 * For now we assume the various PPC GEMs are 32-bit only as well.
+	 */
+	if (pdev->vendor == PCI_VENDOR_ID_SUN &&
+	    pdev->device == PCI_DEVICE_ID_SUN_GEM &&
+	    !dma_set_mask(&pdev->dev, DMA_BIT_MASK(64))) {
+		pci_using_dac = 1;
+	} else {
+		err = dma_set_mask(&pdev->dev, DMA_BIT_MASK(32));
+		if (err) {
+			pr_err("No usable DMA configuration, aborting\n");
+			goto err_disable_device;
+		}
+		pci_using_dac = 0;
+	}
+
+	gemreg_base = pci_resource_start(pdev, 0);
+	gemreg_len = pci_resource_len(pdev, 0);
+
+	if ((pci_resource_flags(pdev, 0) & IORESOURCE_IO) != 0) {
+		pr_err("Cannot find proper PCI device base address, aborting\n");
+		err = -ENODEV;
+		goto err_disable_device;
+	}
+
+	dev = alloc_etherdev(sizeof(*gp));
+	if (!dev) {
+		err = -ENOMEM;
+		goto err_disable_device;
+	}
+	SET_NETDEV_DEV(dev, &pdev->dev);
+
+	gp = netdev_priv(dev);
+
+	err = pci_request_regions(pdev, DRV_NAME);
+	if (err) {
+		pr_err("Cannot obtain PCI resources, aborting\n");
+		goto err_out_free_netdev;
+	}
+
+	gp->pdev = pdev;
+	gp->dev = dev;
+
+	gp->msg_enable = DEFAULT_MSG;
+
+	timer_setup(&gp->link_timer, gem_link_timer, 0);
+
+	INIT_WORK(&gp->reset_task, gem_reset_task);
+
+	gp->lstate = link_down;
+	gp->timer_ticks = 0;
+	netif_carrier_off(dev);
+
+	gp->regs = ioremap(gemreg_base, gemreg_len);
+	if (!gp->regs) {
+		pr_err("Cannot map device registers, aborting\n");
+		err = -EIO;
+		goto err_out_free_res;
+	}
+
+	/* On Apple, we want a reference to the Open Firmware device-tree
+	 * node. We use it for clock control.
+	 */
+#if defined(CONFIG_PPC_PMAC) || defined(CONFIG_SPARC)
+	gp->of_node = pci_device_to_OF_node(pdev);
+#endif
+
+	/* Only Apple version supports WOL afaik */
+	if (pdev->vendor == PCI_VENDOR_ID_APPLE)
+		gp->has_wol = 1;
+
+	/* Make sure cell is enabled */
+	gem_get_cell(gp);
+
+	/* Make sure everything is stopped and in init state */
+	gem_reset(gp);
+
+	/* Fill up the mii_phy structure (even if we won't use it) */
+	gp->phy_mii.dev = dev;
+	gp->phy_mii.mdio_read = _sungem_phy_read;
+	gp->phy_mii.mdio_write = _sungem_phy_write;
+#ifdef CONFIG_PPC_PMAC
+	gp->phy_mii.platform_data = gp->of_node;
+#endif
+	/* By default, we start with autoneg */
+	gp->want_autoneg = 1;
+
+	/* Check fifo sizes, PHY type, etc... */
+	if (gem_check_invariants(gp)) {
+		err = -ENODEV;
+		goto err_out_iounmap;
+	}
+
+	/* It is guaranteed that the returned buffer will be at least
+	 * PAGE_SIZE aligned.
+	 */
+	gp->init_block = dma_alloc_coherent(&pdev->dev, sizeof(struct gem_init_block),
+					    &gp->gblock_dvma, GFP_KERNEL);
+	if (!gp->init_block) {
+		pr_err("Cannot allocate init block, aborting\n");
+		err = -ENOMEM;
+		goto err_out_iounmap;
+	}
+
+	err = gem_get_device_address(gp);
+	if (err)
+		goto err_out_free_consistent;
+
+	dev->netdev_ops = &gem_netdev_ops;
+	netif_napi_add(dev, &gp->napi, gem_poll);
+	dev->ethtool_ops = &gem_ethtool_ops;
+	dev->watchdog_timeo = 5 * HZ;
+	dev->dma = 0;
+
+	/* Set that now, in case PM kicks in now */
+	pci_set_drvdata(pdev, dev);
+
+	/* We can do scatter/gather and HW checksum */
+	dev->hw_features = NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_RXCSUM;
+	dev->features = dev->hw_features;
+	if (pci_using_dac)
+		dev->features |= NETIF_F_HIGHDMA;
+
+	/* MTU range: 68 - 1500 (Jumbo mode is broken) */
+	dev->min_mtu = GEM_MIN_MTU;
+	dev->max_mtu = GEM_MAX_MTU;
+
+	/* Register with kernel */
+	err = register_netdev(dev);
+	if (err) {
+		pr_err("Cannot register net device, aborting\n");
+		goto err_out_clear_drvdata;
+	}
+
+	/* Undo the get_cell with appropriate locking (we could use
+	 * ndo_init/uninit but that would be even more clumsy imho)
+	 */
+	rtnl_lock();
+	gem_put_cell(gp);
+	rtnl_unlock();
+
+	netdev_info(dev, "Sun GEM (PCI) 10/100/1000BaseT Ethernet %pM\n",
+		    dev->dev_addr);
+	return 0;
+
+err_out_clear_drvdata:
+	pci_set_drvdata(pdev, NULL);
+	netif_napi_del(&gp->napi);
+
+err_out_free_consistent:
+	dma_free_coherent(&pdev->dev, sizeof(struct gem_init_block),
+			  gp->init_block, gp->gblock_dvma);
+err_out_iounmap:
+	gem_put_cell(gp);
+	iounmap(gp->regs);
+
+err_out_free_res:
+	pci_release_regions(pdev);
+
+err_out_free_netdev:
+	free_netdev(dev);
+err_disable_device:
+	pci_disable_device(pdev);
+	return err;
+
+}
+
+static SIMPLE_DEV_PM_OPS(gem_pm_ops, gem_suspend, gem_resume);
+
+static struct pci_driver gem_driver = {
+	.name		= GEM_MODULE_NAME,
+	.id_table	= gem_pci_tbl,
+	.probe		= gem_init_one,
+	.remove		= gem_remove_one,
+	.driver.pm	= &gem_pm_ops,
+};
+
+module_pci_driver(gem_driver);
diff --git a/drivers/net/ethernet/sun/sungem.h b/drivers/net/ethernet/sun/sungem.h
new file mode 100644
index 000000000..b921d3074
--- /dev/null
+++ b/drivers/net/ethernet/sun/sungem.h
@@ -0,0 +1,1028 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/* $Id: sungem.h,v 1.10.2.4 2002/03/11 08:54:48 davem Exp $
+ * sungem.h: Definitions for Sun GEM ethernet driver.
+ *
+ * Copyright (C) 2000 David S. Miller (davem@redhat.com)
+ */
+
+#ifndef _SUNGEM_H
+#define _SUNGEM_H
+
+/* Global Registers */
+#define GREG_SEBSTATE	0x0000UL	/* SEB State Register		*/
+#define GREG_CFG	0x0004UL	/* Configuration Register	*/
+#define GREG_STAT	0x000CUL	/* Status Register		*/
+#define GREG_IMASK	0x0010UL	/* Interrupt Mask Register	*/
+#define GREG_IACK	0x0014UL	/* Interrupt ACK Register	*/
+#define GREG_STAT2	0x001CUL	/* Alias of GREG_STAT		*/
+#define GREG_PCIESTAT	0x1000UL	/* PCI Error Status Register	*/
+#define GREG_PCIEMASK	0x1004UL	/* PCI Error Mask Register	*/
+#define GREG_BIFCFG	0x1008UL	/* BIF Configuration Register	*/
+#define GREG_BIFDIAG	0x100CUL	/* BIF Diagnostics Register	*/
+#define GREG_SWRST	0x1010UL	/* Software Reset Register	*/
+
+/* Global SEB State Register */
+#define GREG_SEBSTATE_ARB	0x00000003	/* State of Arbiter		*/
+#define GREG_SEBSTATE_RXWON	0x00000004	/* RX won internal arbitration	*/
+
+/* Global Configuration Register */
+#define GREG_CFG_IBURST		0x00000001	/* Infinite Burst		*/
+#define GREG_CFG_TXDMALIM	0x0000003e	/* TX DMA grant limit		*/
+#define GREG_CFG_RXDMALIM	0x000007c0	/* RX DMA grant limit		*/
+#define GREG_CFG_RONPAULBIT	0x00000800	/* Use mem read multiple for PCI read
+						 * after infinite burst (Apple) */
+#define GREG_CFG_ENBUG2FIX	0x00001000	/* Fix Rx hang after overflow */
+
+/* Global Interrupt Status Register.
+ *
+ * Reading this register automatically clears bits 0 through 6.
+ * This auto-clearing does not occur when the alias at GREG_STAT2
+ * is read instead.  The rest of the interrupt bits only clear when
+ * the secondary interrupt status register corresponding to that
+ * bit is read (ie. if GREG_STAT_PCS is set, it will be cleared by
+ * reading PCS_ISTAT).
+ */
+#define GREG_STAT_TXINTME	0x00000001	/* TX INTME frame transferred	*/
+#define GREG_STAT_TXALL		0x00000002	/* All TX frames transferred	*/
+#define GREG_STAT_TXDONE	0x00000004	/* One TX frame transferred	*/
+#define GREG_STAT_RXDONE	0x00000010	/* One RX frame arrived		*/
+#define GREG_STAT_RXNOBUF	0x00000020	/* No free RX buffers available	*/
+#define GREG_STAT_RXTAGERR	0x00000040	/* RX tag framing is corrupt	*/
+#define GREG_STAT_PCS		0x00002000	/* PCS signalled interrupt	*/
+#define GREG_STAT_TXMAC		0x00004000	/* TX MAC signalled interrupt	*/
+#define GREG_STAT_RXMAC		0x00008000	/* RX MAC signalled interrupt	*/
+#define GREG_STAT_MAC		0x00010000	/* MAC Control signalled irq	*/
+#define GREG_STAT_MIF		0x00020000	/* MIF signalled interrupt	*/
+#define GREG_STAT_PCIERR	0x00040000	/* PCI Error interrupt		*/
+#define GREG_STAT_TXNR		0xfff80000	/* == TXDMA_TXDONE reg val	*/
+#define GREG_STAT_TXNR_SHIFT	19
+
+#define GREG_STAT_ABNORMAL	(GREG_STAT_RXNOBUF | GREG_STAT_RXTAGERR | \
+				 GREG_STAT_PCS | GREG_STAT_TXMAC | GREG_STAT_RXMAC | \
+				 GREG_STAT_MAC | GREG_STAT_MIF | GREG_STAT_PCIERR)
+
+#define GREG_STAT_NAPI		(GREG_STAT_TXALL  | GREG_STAT_TXINTME | \
+				 GREG_STAT_RXDONE | GREG_STAT_ABNORMAL)
+
+/* The layout of GREG_IMASK and GREG_IACK is identical to GREG_STAT.
+ * Bits set in GREG_IMASK will prevent that interrupt type from being
+ * signalled to the cpu.  GREG_IACK can be used to clear specific top-level
+ * interrupt conditions in GREG_STAT, ie. it only works for bits 0 through 6.
+ * Setting the bit will clear that interrupt, clear bits will have no effect
+ * on GREG_STAT.
+ */
+
+/* Global PCI Error Status Register */
+#define GREG_PCIESTAT_BADACK	0x00000001	/* No ACK64# during ABS64 cycle	*/
+#define GREG_PCIESTAT_DTRTO	0x00000002	/* Delayed transaction timeout	*/
+#define GREG_PCIESTAT_OTHER	0x00000004	/* Other PCI error, check cfg space */
+
+/* The layout of the GREG_PCIEMASK is identical to that of GREG_PCIESTAT.
+ * Bits set in GREG_PCIEMASK will prevent that interrupt type from being
+ * signalled to the cpu.
+ */
+
+/* Global BIF Configuration Register */
+#define GREG_BIFCFG_SLOWCLK	0x00000001	/* Set if PCI runs < 25Mhz	*/
+#define GREG_BIFCFG_B64DIS	0x00000002	/* Disable 64bit wide data cycle*/
+#define GREG_BIFCFG_M66EN	0x00000004	/* Set if on 66Mhz PCI segment	*/
+
+/* Global BIF Diagnostics Register */
+#define GREG_BIFDIAG_BURSTSM	0x007f0000	/* PCI Burst state machine	*/
+#define GREG_BIFDIAG_BIFSM	0xff000000	/* BIF state machine		*/
+
+/* Global Software Reset Register.
+ *
+ * This register is used to perform a global reset of the RX and TX portions
+ * of the GEM asic.  Setting the RX or TX reset bit will start the reset.
+ * The driver _MUST_ poll these bits until they clear.  One may not attempt
+ * to program any other part of GEM until the bits clear.
+ */
+#define GREG_SWRST_TXRST	0x00000001	/* TX Software Reset		*/
+#define GREG_SWRST_RXRST	0x00000002	/* RX Software Reset		*/
+#define GREG_SWRST_RSTOUT	0x00000004	/* Force RST# pin active	*/
+#define GREG_SWRST_CACHESIZE	0x00ff0000	/* RIO only: cache line size	*/
+#define GREG_SWRST_CACHE_SHIFT	16
+
+/* TX DMA Registers */
+#define TXDMA_KICK	0x2000UL	/* TX Kick Register		*/
+#define TXDMA_CFG	0x2004UL	/* TX Configuration Register	*/
+#define TXDMA_DBLOW	0x2008UL	/* TX Desc. Base Low		*/
+#define TXDMA_DBHI	0x200CUL	/* TX Desc. Base High		*/
+#define TXDMA_FWPTR	0x2014UL	/* TX FIFO Write Pointer	*/
+#define TXDMA_FSWPTR	0x2018UL	/* TX FIFO Shadow Write Pointer	*/
+#define TXDMA_FRPTR	0x201CUL	/* TX FIFO Read Pointer		*/
+#define TXDMA_FSRPTR	0x2020UL	/* TX FIFO Shadow Read Pointer	*/
+#define TXDMA_PCNT	0x2024UL	/* TX FIFO Packet Counter	*/
+#define TXDMA_SMACHINE	0x2028UL	/* TX State Machine Register	*/
+#define TXDMA_DPLOW	0x2030UL	/* TX Data Pointer Low		*/
+#define TXDMA_DPHI	0x2034UL	/* TX Data Pointer High		*/
+#define TXDMA_TXDONE	0x2100UL	/* TX Completion Register	*/
+#define TXDMA_FADDR	0x2104UL	/* TX FIFO Address		*/
+#define TXDMA_FTAG	0x2108UL	/* TX FIFO Tag			*/
+#define TXDMA_DLOW	0x210CUL	/* TX FIFO Data Low		*/
+#define TXDMA_DHIT1	0x2110UL	/* TX FIFO Data HighT1		*/
+#define TXDMA_DHIT0	0x2114UL	/* TX FIFO Data HighT0		*/
+#define TXDMA_FSZ	0x2118UL	/* TX FIFO Size			*/
+
+/* TX Kick Register.
+ *
+ * This 13-bit register is programmed by the driver to hold the descriptor
+ * entry index which follows the last valid transmit descriptor.
+ */
+
+/* TX Completion Register.
+ *
+ * This 13-bit register is updated by GEM to hold to descriptor entry index
+ * which follows the last descriptor already processed by GEM.  Note that
+ * this value is mirrored in GREG_STAT which eliminates the need to even
+ * access this register in the driver during interrupt processing.
+ */
+
+/* TX Configuration Register.
+ *
+ * Note that TXDMA_CFG_FTHRESH, the TX FIFO Threshold, is an obsolete feature
+ * that was meant to be used with jumbo packets.  It should be set to the
+ * maximum value of 0x4ff, else one risks getting TX MAC Underrun errors.
+ */
+#define TXDMA_CFG_ENABLE	0x00000001	/* Enable TX DMA channel	*/
+#define TXDMA_CFG_RINGSZ	0x0000001e	/* TX descriptor ring size	*/
+#define TXDMA_CFG_RINGSZ_32	0x00000000	/* 32 TX descriptors		*/
+#define TXDMA_CFG_RINGSZ_64	0x00000002	/* 64 TX descriptors		*/
+#define TXDMA_CFG_RINGSZ_128	0x00000004	/* 128 TX descriptors		*/
+#define TXDMA_CFG_RINGSZ_256	0x00000006	/* 256 TX descriptors		*/
+#define TXDMA_CFG_RINGSZ_512	0x00000008	/* 512 TX descriptors		*/
+#define TXDMA_CFG_RINGSZ_1K	0x0000000a	/* 1024 TX descriptors		*/
+#define TXDMA_CFG_RINGSZ_2K	0x0000000c	/* 2048 TX descriptors		*/
+#define TXDMA_CFG_RINGSZ_4K	0x0000000e	/* 4096 TX descriptors		*/
+#define TXDMA_CFG_RINGSZ_8K	0x00000010	/* 8192 TX descriptors		*/
+#define TXDMA_CFG_PIOSEL	0x00000020	/* Enable TX FIFO PIO from cpu	*/
+#define TXDMA_CFG_FTHRESH	0x001ffc00	/* TX FIFO Threshold, obsolete	*/
+#define TXDMA_CFG_PMODE		0x00200000	/* TXALL irq means TX FIFO empty*/
+
+/* TX Descriptor Base Low/High.
+ *
+ * These two registers store the 53 most significant bits of the base address
+ * of the TX descriptor table.  The 11 least significant bits are always
+ * zero.  As a result, the TX descriptor table must be 2K aligned.
+ */
+
+/* The rest of the TXDMA_* registers are for diagnostics and debug, I will document
+ * them later. -DaveM
+ */
+
+/* WakeOnLan Registers	*/
+#define WOL_MATCH0	0x3000UL
+#define WOL_MATCH1	0x3004UL
+#define WOL_MATCH2	0x3008UL
+#define WOL_MCOUNT	0x300CUL
+#define WOL_WAKECSR	0x3010UL
+
+/* WOL Match count register
+ */
+#define WOL_MCOUNT_N		0x00000010
+#define WOL_MCOUNT_M		0x00000000 /* 0 << 8 */
+
+#define WOL_WAKECSR_ENABLE	0x00000001
+#define WOL_WAKECSR_MII		0x00000002
+#define WOL_WAKECSR_SEEN	0x00000004
+#define WOL_WAKECSR_FILT_UCAST	0x00000008
+#define WOL_WAKECSR_FILT_MCAST	0x00000010
+#define WOL_WAKECSR_FILT_BCAST	0x00000020
+#define WOL_WAKECSR_FILT_SEEN	0x00000040
+
+
+/* Receive DMA Registers */
+#define RXDMA_CFG	0x4000UL	/* RX Configuration Register	*/
+#define RXDMA_DBLOW	0x4004UL	/* RX Descriptor Base Low	*/
+#define RXDMA_DBHI	0x4008UL	/* RX Descriptor Base High	*/
+#define RXDMA_FWPTR	0x400CUL	/* RX FIFO Write Pointer	*/
+#define RXDMA_FSWPTR	0x4010UL	/* RX FIFO Shadow Write Pointer	*/
+#define RXDMA_FRPTR	0x4014UL	/* RX FIFO Read Pointer		*/
+#define RXDMA_PCNT	0x4018UL	/* RX FIFO Packet Counter	*/
+#define RXDMA_SMACHINE	0x401CUL	/* RX State Machine Register	*/
+#define RXDMA_PTHRESH	0x4020UL	/* Pause Thresholds		*/
+#define RXDMA_DPLOW	0x4024UL	/* RX Data Pointer Low		*/
+#define RXDMA_DPHI	0x4028UL	/* RX Data Pointer High		*/
+#define RXDMA_KICK	0x4100UL	/* RX Kick Register		*/
+#define RXDMA_DONE	0x4104UL	/* RX Completion Register	*/
+#define RXDMA_BLANK	0x4108UL	/* RX Blanking Register		*/
+#define RXDMA_FADDR	0x410CUL	/* RX FIFO Address		*/
+#define RXDMA_FTAG	0x4110UL	/* RX FIFO Tag			*/
+#define RXDMA_DLOW	0x4114UL	/* RX FIFO Data Low		*/
+#define RXDMA_DHIT1	0x4118UL	/* RX FIFO Data HighT0		*/
+#define RXDMA_DHIT0	0x411CUL	/* RX FIFO Data HighT1		*/
+#define RXDMA_FSZ	0x4120UL	/* RX FIFO Size			*/
+
+/* RX Configuration Register. */
+#define RXDMA_CFG_ENABLE	0x00000001	/* Enable RX DMA channel	*/
+#define RXDMA_CFG_RINGSZ	0x0000001e	/* RX descriptor ring size	*/
+#define RXDMA_CFG_RINGSZ_32	0x00000000	/* - 32   entries		*/
+#define RXDMA_CFG_RINGSZ_64	0x00000002	/* - 64   entries		*/
+#define RXDMA_CFG_RINGSZ_128	0x00000004	/* - 128  entries		*/
+#define RXDMA_CFG_RINGSZ_256	0x00000006	/* - 256  entries		*/
+#define RXDMA_CFG_RINGSZ_512	0x00000008	/* - 512  entries		*/
+#define RXDMA_CFG_RINGSZ_1K	0x0000000a	/* - 1024 entries		*/
+#define RXDMA_CFG_RINGSZ_2K	0x0000000c	/* - 2048 entries		*/
+#define RXDMA_CFG_RINGSZ_4K	0x0000000e	/* - 4096 entries		*/
+#define RXDMA_CFG_RINGSZ_8K	0x00000010	/* - 8192 entries		*/
+#define RXDMA_CFG_RINGSZ_BDISAB	0x00000020	/* Disable RX desc batching	*/
+#define RXDMA_CFG_FBOFF		0x00001c00	/* Offset of first data byte	*/
+#define RXDMA_CFG_CSUMOFF	0x000fe000	/* Skip bytes before csum calc	*/
+#define RXDMA_CFG_FTHRESH	0x07000000	/* RX FIFO dma start threshold	*/
+#define RXDMA_CFG_FTHRESH_64	0x00000000	/* - 64   bytes			*/
+#define RXDMA_CFG_FTHRESH_128	0x01000000	/* - 128  bytes			*/
+#define RXDMA_CFG_FTHRESH_256	0x02000000	/* - 256  bytes			*/
+#define RXDMA_CFG_FTHRESH_512	0x03000000	/* - 512  bytes			*/
+#define RXDMA_CFG_FTHRESH_1K	0x04000000	/* - 1024 bytes			*/
+#define RXDMA_CFG_FTHRESH_2K	0x05000000	/* - 2048 bytes			*/
+
+/* RX Descriptor Base Low/High.
+ *
+ * These two registers store the 53 most significant bits of the base address
+ * of the RX descriptor table.  The 11 least significant bits are always
+ * zero.  As a result, the RX descriptor table must be 2K aligned.
+ */
+
+/* RX PAUSE Thresholds.
+ *
+ * These values determine when XOFF and XON PAUSE frames are emitted by
+ * GEM.  The thresholds measure RX FIFO occupancy in units of 64 bytes.
+ */
+#define RXDMA_PTHRESH_OFF	0x000001ff	/* XOFF emitted w/FIFO > this	*/
+#define RXDMA_PTHRESH_ON	0x001ff000	/* XON emitted w/FIFO < this	*/
+
+/* RX Kick Register.
+ *
+ * This 13-bit register is written by the host CPU and holds the last
+ * valid RX descriptor number plus one.  This is, if 'N' is written to
+ * this register, it means that all RX descriptors up to but excluding
+ * 'N' are valid.
+ *
+ * The hardware requires that RX descriptors are posted in increments
+ * of 4.  This means 'N' must be a multiple of four.  For the best
+ * performance, the first new descriptor being posted should be (PCI)
+ * cache line aligned.
+ */
+
+/* RX Completion Register.
+ *
+ * This 13-bit register is updated by GEM to indicate which RX descriptors
+ * have already been used for receive frames.  All descriptors up to but
+ * excluding the value in this register are ready to be processed.  GEM
+ * updates this register value after the RX FIFO empties completely into
+ * the RX descriptor's buffer, but before the RX_DONE bit is set in the
+ * interrupt status register.
+ */
+
+/* RX Blanking Register. */
+#define RXDMA_BLANK_IPKTS	0x000001ff	/* RX_DONE asserted after this
+						 * many packets received since
+						 * previous RX_DONE.
+						 */
+#define RXDMA_BLANK_ITIME	0x000ff000	/* RX_DONE asserted after this
+						 * many clocks (measured in 2048
+						 * PCI clocks) were counted since
+						 * the previous RX_DONE.
+						 */
+
+/* RX FIFO Size.
+ *
+ * This 11-bit read-only register indicates how large, in units of 64-bytes,
+ * the RX FIFO is.  The driver uses this to properly configure the RX PAUSE
+ * thresholds.
+ */
+
+/* The rest of the RXDMA_* registers are for diagnostics and debug, I will document
+ * them later. -DaveM
+ */
+
+/* MAC Registers */
+#define MAC_TXRST	0x6000UL	/* TX MAC Software Reset Command*/
+#define MAC_RXRST	0x6004UL	/* RX MAC Software Reset Command*/
+#define MAC_SNDPAUSE	0x6008UL	/* Send Pause Command Register	*/
+#define MAC_TXSTAT	0x6010UL	/* TX MAC Status Register	*/
+#define MAC_RXSTAT	0x6014UL	/* RX MAC Status Register	*/
+#define MAC_CSTAT	0x6018UL	/* MAC Control Status Register	*/
+#define MAC_TXMASK	0x6020UL	/* TX MAC Mask Register		*/
+#define MAC_RXMASK	0x6024UL	/* RX MAC Mask Register		*/
+#define MAC_MCMASK	0x6028UL	/* MAC Control Mask Register	*/
+#define MAC_TXCFG	0x6030UL	/* TX MAC Configuration Register*/
+#define MAC_RXCFG	0x6034UL	/* RX MAC Configuration Register*/
+#define MAC_MCCFG	0x6038UL	/* MAC Control Config Register	*/
+#define MAC_XIFCFG	0x603CUL	/* XIF Configuration Register	*/
+#define MAC_IPG0	0x6040UL	/* InterPacketGap0 Register	*/
+#define MAC_IPG1	0x6044UL	/* InterPacketGap1 Register	*/
+#define MAC_IPG2	0x6048UL	/* InterPacketGap2 Register	*/
+#define MAC_STIME	0x604CUL	/* SlotTime Register		*/
+#define MAC_MINFSZ	0x6050UL	/* MinFrameSize Register	*/
+#define MAC_MAXFSZ	0x6054UL	/* MaxFrameSize Register	*/
+#define MAC_PASIZE	0x6058UL	/* PA Size Register		*/
+#define MAC_JAMSIZE	0x605CUL	/* JamSize Register		*/
+#define MAC_ATTLIM	0x6060UL	/* Attempt Limit Register	*/
+#define MAC_MCTYPE	0x6064UL	/* MAC Control Type Register	*/
+#define MAC_ADDR0	0x6080UL	/* MAC Address 0 Register	*/
+#define MAC_ADDR1	0x6084UL	/* MAC Address 1 Register	*/
+#define MAC_ADDR2	0x6088UL	/* MAC Address 2 Register	*/
+#define MAC_ADDR3	0x608CUL	/* MAC Address 3 Register	*/
+#define MAC_ADDR4	0x6090UL	/* MAC Address 4 Register	*/
+#define MAC_ADDR5	0x6094UL	/* MAC Address 5 Register	*/
+#define MAC_ADDR6	0x6098UL	/* MAC Address 6 Register	*/
+#define MAC_ADDR7	0x609CUL	/* MAC Address 7 Register	*/
+#define MAC_ADDR8	0x60A0UL	/* MAC Address 8 Register	*/
+#define MAC_AFILT0	0x60A4UL	/* Address Filter 0 Register	*/
+#define MAC_AFILT1	0x60A8UL	/* Address Filter 1 Register	*/
+#define MAC_AFILT2	0x60ACUL	/* Address Filter 2 Register	*/
+#define MAC_AF21MSK	0x60B0UL	/* Address Filter 2&1 Mask Reg	*/
+#define MAC_AF0MSK	0x60B4UL	/* Address Filter 0 Mask Reg	*/
+#define MAC_HASH0	0x60C0UL	/* Hash Table 0 Register	*/
+#define MAC_HASH1	0x60C4UL	/* Hash Table 1 Register	*/
+#define MAC_HASH2	0x60C8UL	/* Hash Table 2 Register	*/
+#define MAC_HASH3	0x60CCUL	/* Hash Table 3 Register	*/
+#define MAC_HASH4	0x60D0UL	/* Hash Table 4 Register	*/
+#define MAC_HASH5	0x60D4UL	/* Hash Table 5 Register	*/
+#define MAC_HASH6	0x60D8UL	/* Hash Table 6 Register	*/
+#define MAC_HASH7	0x60DCUL	/* Hash Table 7 Register	*/
+#define MAC_HASH8	0x60E0UL	/* Hash Table 8 Register	*/
+#define MAC_HASH9	0x60E4UL	/* Hash Table 9 Register	*/
+#define MAC_HASH10	0x60E8UL	/* Hash Table 10 Register	*/
+#define MAC_HASH11	0x60ECUL	/* Hash Table 11 Register	*/
+#define MAC_HASH12	0x60F0UL	/* Hash Table 12 Register	*/
+#define MAC_HASH13	0x60F4UL	/* Hash Table 13 Register	*/
+#define MAC_HASH14	0x60F8UL	/* Hash Table 14 Register	*/
+#define MAC_HASH15	0x60FCUL	/* Hash Table 15 Register	*/
+#define MAC_NCOLL	0x6100UL	/* Normal Collision Counter	*/
+#define MAC_FASUCC	0x6104UL	/* First Attempt. Succ Coll Ctr.*/
+#define MAC_ECOLL	0x6108UL	/* Excessive Collision Counter	*/
+#define MAC_LCOLL	0x610CUL	/* Late Collision Counter	*/
+#define MAC_DTIMER	0x6110UL	/* Defer Timer			*/
+#define MAC_PATMPS	0x6114UL	/* Peak Attempts Register	*/
+#define MAC_RFCTR	0x6118UL	/* Receive Frame Counter	*/
+#define MAC_LERR	0x611CUL	/* Length Error Counter		*/
+#define MAC_AERR	0x6120UL	/* Alignment Error Counter	*/
+#define MAC_FCSERR	0x6124UL	/* FCS Error Counter		*/
+#define MAC_RXCVERR	0x6128UL	/* RX code Violation Error Ctr	*/
+#define MAC_RANDSEED	0x6130UL	/* Random Number Seed Register	*/
+#define MAC_SMACHINE	0x6134UL	/* State Machine Register	*/
+
+/* TX MAC Software Reset Command. */
+#define MAC_TXRST_CMD	0x00000001	/* Start sw reset, self-clears	*/
+
+/* RX MAC Software Reset Command. */
+#define MAC_RXRST_CMD	0x00000001	/* Start sw reset, self-clears	*/
+
+/* Send Pause Command. */
+#define MAC_SNDPAUSE_TS	0x0000ffff	/* The pause_time operand used in
+					 * Send_Pause and flow-control
+					 * handshakes.
+					 */
+#define MAC_SNDPAUSE_SP	0x00010000	/* Setting this bit instructs the MAC
+					 * to send a Pause Flow Control
+					 * frame onto the network.
+					 */
+
+/* TX MAC Status Register. */
+#define MAC_TXSTAT_XMIT	0x00000001	/* Frame Transmitted		*/
+#define MAC_TXSTAT_URUN	0x00000002	/* TX Underrun			*/
+#define MAC_TXSTAT_MPE	0x00000004	/* Max Packet Size Error	*/
+#define MAC_TXSTAT_NCE	0x00000008	/* Normal Collision Cntr Expire	*/
+#define MAC_TXSTAT_ECE	0x00000010	/* Excess Collision Cntr Expire	*/
+#define MAC_TXSTAT_LCE	0x00000020	/* Late Collision Cntr Expire	*/
+#define MAC_TXSTAT_FCE	0x00000040	/* First Collision Cntr Expire	*/
+#define MAC_TXSTAT_DTE	0x00000080	/* Defer Timer Expire		*/
+#define MAC_TXSTAT_PCE	0x00000100	/* Peak Attempts Cntr Expire	*/
+
+/* RX MAC Status Register. */
+#define MAC_RXSTAT_RCV	0x00000001	/* Frame Received		*/
+#define MAC_RXSTAT_OFLW	0x00000002	/* Receive Overflow		*/
+#define MAC_RXSTAT_FCE	0x00000004	/* Frame Cntr Expire		*/
+#define MAC_RXSTAT_ACE	0x00000008	/* Align Error Cntr Expire	*/
+#define MAC_RXSTAT_CCE	0x00000010	/* CRC Error Cntr Expire	*/
+#define MAC_RXSTAT_LCE	0x00000020	/* Length Error Cntr Expire	*/
+#define MAC_RXSTAT_VCE	0x00000040	/* Code Violation Cntr Expire	*/
+
+/* MAC Control Status Register. */
+#define MAC_CSTAT_PRCV	0x00000001	/* Pause Received		*/
+#define MAC_CSTAT_PS	0x00000002	/* Paused State			*/
+#define MAC_CSTAT_NPS	0x00000004	/* Not Paused State		*/
+#define MAC_CSTAT_PTR	0xffff0000	/* Pause Time Received		*/
+
+/* The layout of the MAC_{TX,RX,C}MASK registers is identical to that
+ * of MAC_{TX,RX,C}STAT.  Bits set in MAC_{TX,RX,C}MASK will prevent
+ * that interrupt type from being signalled to front end of GEM.  For
+ * the interrupt to actually get sent to the cpu, it is necessary to
+ * properly set the appropriate GREG_IMASK_{TX,RX,}MAC bits as well.
+ */
+
+/* TX MAC Configuration Register.
+ *
+ * NOTE: The TX MAC Enable bit must be cleared and polled until
+ *	 zero before any other bits in this register are changed.
+ *
+ *	 Also, enabling the Carrier Extension feature of GEM is
+ *	 a 3 step process 1) Set TX Carrier Extension 2) Set
+ *	 RX Carrier Extension 3) Set Slot Time to 0x200.  This
+ *	 mode must be enabled when in half-duplex at 1Gbps, else
+ *	 it must be disabled.
+ */
+#define MAC_TXCFG_ENAB	0x00000001	/* TX MAC Enable		*/
+#define MAC_TXCFG_ICS	0x00000002	/* Ignore Carrier Sense		*/
+#define MAC_TXCFG_ICOLL	0x00000004	/* Ignore Collisions		*/
+#define MAC_TXCFG_EIPG0	0x00000008	/* Enable IPG0			*/
+#define MAC_TXCFG_NGU	0x00000010	/* Never Give Up		*/
+#define MAC_TXCFG_NGUL	0x00000020	/* Never Give Up Limit		*/
+#define MAC_TXCFG_NBO	0x00000040	/* No Backoff			*/
+#define MAC_TXCFG_SD	0x00000080	/* Slow Down			*/
+#define MAC_TXCFG_NFCS	0x00000100	/* No FCS			*/
+#define MAC_TXCFG_TCE	0x00000200	/* TX Carrier Extension		*/
+
+/* RX MAC Configuration Register.
+ *
+ * NOTE: The RX MAC Enable bit must be cleared and polled until
+ *	 zero before any other bits in this register are changed.
+ *
+ *	 Similar rules apply to the Hash Filter Enable bit when
+ *	 programming the hash table registers, and the Address Filter
+ *	 Enable bit when programming the address filter registers.
+ */
+#define MAC_RXCFG_ENAB	0x00000001	/* RX MAC Enable		*/
+#define MAC_RXCFG_SPAD	0x00000002	/* Strip Pad			*/
+#define MAC_RXCFG_SFCS	0x00000004	/* Strip FCS			*/
+#define MAC_RXCFG_PROM	0x00000008	/* Promiscuous Mode		*/
+#define MAC_RXCFG_PGRP	0x00000010	/* Promiscuous Group		*/
+#define MAC_RXCFG_HFE	0x00000020	/* Hash Filter Enable		*/
+#define MAC_RXCFG_AFE	0x00000040	/* Address Filter Enable	*/
+#define MAC_RXCFG_DDE	0x00000080	/* Disable Discard on Error	*/
+#define MAC_RXCFG_RCE	0x00000100	/* RX Carrier Extension		*/
+
+/* MAC Control Config Register. */
+#define MAC_MCCFG_SPE	0x00000001	/* Send Pause Enable		*/
+#define MAC_MCCFG_RPE	0x00000002	/* Receive Pause Enable		*/
+#define MAC_MCCFG_PMC	0x00000004	/* Pass MAC Control		*/
+
+/* XIF Configuration Register.
+ *
+ * NOTE: When leaving or entering loopback mode, a global hardware
+ *       init of GEM should be performed.
+ */
+#define MAC_XIFCFG_OE	0x00000001	/* MII TX Output Driver Enable	*/
+#define MAC_XIFCFG_LBCK	0x00000002	/* Loopback TX to RX		*/
+#define MAC_XIFCFG_DISE	0x00000004	/* Disable RX path during TX	*/
+#define MAC_XIFCFG_GMII	0x00000008	/* Use GMII clocks + datapath	*/
+#define MAC_XIFCFG_MBOE	0x00000010	/* Controls MII_BUF_EN pin	*/
+#define MAC_XIFCFG_LLED	0x00000020	/* Force LINKLED# active (low)	*/
+#define MAC_XIFCFG_FLED	0x00000040	/* Force FDPLXLED# active (low)	*/
+
+/* InterPacketGap0 Register.  This 8-bit value is used as an extension
+ * to the InterPacketGap1 Register.  Specifically it contributes to the
+ * timing of the RX-to-TX IPG.  This value is ignored and presumed to
+ * be zero for TX-to-TX IPG calculations and/or when the Enable IPG0 bit
+ * is cleared in the TX MAC Configuration Register.
+ *
+ * This value in this register in terms of media byte time.
+ *
+ * Recommended value: 0x00
+ */
+
+/* InterPacketGap1 Register.  This 8-bit value defines the first 2/3
+ * portion of the Inter Packet Gap.
+ *
+ * This value in this register in terms of media byte time.
+ *
+ * Recommended value: 0x08
+ */
+
+/* InterPacketGap2 Register.  This 8-bit value defines the second 1/3
+ * portion of the Inter Packet Gap.
+ *
+ * This value in this register in terms of media byte time.
+ *
+ * Recommended value: 0x04
+ */
+
+/* Slot Time Register.  This 10-bit value specifies the slot time
+ * parameter in units of media byte time.  It determines the physical
+ * span of the network.
+ *
+ * Recommended value: 0x40
+ */
+
+/* Minimum Frame Size Register.  This 10-bit register specifies the
+ * smallest sized frame the TXMAC will send onto the medium, and the
+ * RXMAC will receive from the medium.
+ *
+ * Recommended value: 0x40
+ */
+
+/* Maximum Frame and Burst Size Register.
+ *
+ * This register specifies two things.  First it specifies the maximum
+ * sized frame the TXMAC will send and the RXMAC will recognize as
+ * valid.  Second, it specifies the maximum run length of a burst of
+ * packets sent in half-duplex gigabit modes.
+ *
+ * Recommended value: 0x200005ee
+ */
+#define MAC_MAXFSZ_MFS	0x00007fff	/* Max Frame Size		*/
+#define MAC_MAXFSZ_MBS	0x7fff0000	/* Max Burst Size		*/
+
+/* PA Size Register.  This 10-bit register specifies the number of preamble
+ * bytes which will be transmitted at the beginning of each frame.  A
+ * value of two or greater should be programmed here.
+ *
+ * Recommended value: 0x07
+ */
+
+/* Jam Size Register.  This 4-bit register specifies the duration of
+ * the jam in units of media byte time.
+ *
+ * Recommended value: 0x04
+ */
+
+/* Attempts Limit Register.  This 8-bit register specifies the number
+ * of attempts that the TXMAC will make to transmit a frame, before it
+ * resets its Attempts Counter.  After reaching the Attempts Limit the
+ * TXMAC may or may not drop the frame, as determined by the NGU
+ * (Never Give Up) and NGUL (Never Give Up Limit) bits in the TXMAC
+ * Configuration Register.
+ *
+ * Recommended value: 0x10
+ */
+
+/* MAX Control Type Register.  This 16-bit register specifies the
+ * "type" field of a MAC Control frame.  The TXMAC uses this field to
+ * encapsulate the MAC Control frame for transmission, and the RXMAC
+ * uses it for decoding valid MAC Control frames received from the
+ * network.
+ *
+ * Recommended value: 0x8808
+ */
+
+/* MAC Address Registers.  Each of these registers specify the
+ * ethernet MAC of the interface, 16-bits at a time.  Register
+ * 0 specifies bits [47:32], register 1 bits [31:16], and register
+ * 2 bits [15:0].
+ *
+ * Registers 3 through and including 5 specify an alternate
+ * MAC address for the interface.
+ *
+ * Registers 6 through and including 8 specify the MAC Control
+ * Address, which must be the reserved multicast address for MAC
+ * Control frames.
+ *
+ * Example: To program primary station address a:b:c:d:e:f into
+ *	    the chip.
+ *		MAC_Address_2 = (a << 8) | b
+ *		MAC_Address_1 = (c << 8) | d
+ *		MAC_Address_0 = (e << 8) | f
+ */
+
+/* Address Filter Registers.  Registers 0 through 2 specify bit
+ * fields [47:32] through [15:0], respectively, of the address
+ * filter.  The Address Filter 2&1 Mask Register denotes the 8-bit
+ * nibble mask for Address Filter Registers 2 and 1.  The Address
+ * Filter 0 Mask Register denotes the 16-bit mask for the Address
+ * Filter Register 0.
+ */
+
+/* Hash Table Registers.  Registers 0 through 15 specify bit fields
+ * [255:240] through [15:0], respectively, of the hash table.
+ */
+
+/* Statistics Registers.  All of these registers are 16-bits and
+ * track occurrences of a specific event.  GEM can be configured
+ * to interrupt the host cpu when any of these counters overflow.
+ * They should all be explicitly initialized to zero when the interface
+ * is brought up.
+ */
+
+/* Random Number Seed Register.  This 10-bit value is used as the
+ * RNG seed inside GEM for the CSMA/CD backoff algorithm.  It is
+ * recommended to program this register to the 10 LSB of the
+ * interfaces MAC address.
+ */
+
+/* Pause Timer, read-only.  This 16-bit timer is used to time the pause
+ * interval as indicated by a received pause flow control frame.
+ * A non-zero value in this timer indicates that the MAC is currently in
+ * the paused state.
+ */
+
+/* MIF Registers */
+#define MIF_BBCLK	0x6200UL	/* MIF Bit-Bang Clock		*/
+#define MIF_BBDATA	0x6204UL	/* MIF Bit-Band Data		*/
+#define MIF_BBOENAB	0x6208UL	/* MIF Bit-Bang Output Enable	*/
+#define MIF_FRAME	0x620CUL	/* MIF Frame/Output Register	*/
+#define MIF_CFG		0x6210UL	/* MIF Configuration Register	*/
+#define MIF_MASK	0x6214UL	/* MIF Mask Register		*/
+#define MIF_STATUS	0x6218UL	/* MIF Status Register		*/
+#define MIF_SMACHINE	0x621CUL	/* MIF State Machine Register	*/
+
+/* MIF Bit-Bang Clock.  This 1-bit register is used to generate the
+ * MDC clock waveform on the MII Management Interface when the MIF is
+ * programmed in the "Bit-Bang" mode.  Writing a '1' after a '0' into
+ * this register will create a rising edge on the MDC, while writing
+ * a '0' after a '1' will create a falling edge.  For every bit that
+ * is transferred on the management interface, both edges have to be
+ * generated.
+ */
+
+/* MIF Bit-Bang Data.  This 1-bit register is used to generate the
+ * outgoing data (MDO) on the MII Management Interface when the MIF
+ * is programmed in the "Bit-Bang" mode.  The daa will be steered to the
+ * appropriate MDIO based on the state of the PHY_Select bit in the MIF
+ * Configuration Register.
+ */
+
+/* MIF Big-Band Output Enable.  THis 1-bit register is used to enable
+ * ('1') or disable ('0') the I-directional driver on the MII when the
+ * MIF is programmed in the "Bit-Bang" mode.  The MDIO should be enabled
+ * when data bits are transferred from the MIF to the transceiver, and it
+ * should be disabled when the interface is idle or when data bits are
+ * transferred from the transceiver to the MIF (data portion of a read
+ * instruction).  Only one MDIO will be enabled at a given time, depending
+ * on the state of the PHY_Select bit in the MIF Configuration Register.
+ */
+
+/* MIF Configuration Register.  This 15-bit register controls the operation
+ * of the MIF.
+ */
+#define MIF_CFG_PSELECT	0x00000001	/* Xcvr slct: 0=mdio0 1=mdio1	*/
+#define MIF_CFG_POLL	0x00000002	/* Enable polling mechanism	*/
+#define MIF_CFG_BBMODE	0x00000004	/* 1=bit-bang 0=frame mode	*/
+#define MIF_CFG_PRADDR	0x000000f8	/* Xcvr poll register address	*/
+#define MIF_CFG_MDI0	0x00000100	/* MDIO_0 present or read-bit	*/
+#define MIF_CFG_MDI1	0x00000200	/* MDIO_1 present or read-bit	*/
+#define MIF_CFG_PPADDR	0x00007c00	/* Xcvr poll PHY address	*/
+
+/* MIF Frame/Output Register.  This 32-bit register allows the host to
+ * communicate with a transceiver in frame mode (as opposed to big-bang
+ * mode).  Writes by the host specify an instruction.  After being issued
+ * the host must poll this register for completion.  Also, after
+ * completion this register holds the data returned by the transceiver
+ * if applicable.
+ */
+#define MIF_FRAME_ST	0xc0000000	/* STart of frame		*/
+#define MIF_FRAME_OP	0x30000000	/* OPcode			*/
+#define MIF_FRAME_PHYAD	0x0f800000	/* PHY ADdress			*/
+#define MIF_FRAME_REGAD	0x007c0000	/* REGister ADdress		*/
+#define MIF_FRAME_TAMSB	0x00020000	/* Turn Around MSB		*/
+#define MIF_FRAME_TALSB	0x00010000	/* Turn Around LSB		*/
+#define MIF_FRAME_DATA	0x0000ffff	/* Instruction Payload		*/
+
+/* MIF Status Register.  This register reports status when the MIF is
+ * operating in the poll mode.  The poll status field is auto-clearing
+ * on read.
+ */
+#define MIF_STATUS_DATA	0xffff0000	/* Live image of XCVR reg	*/
+#define MIF_STATUS_STAT	0x0000ffff	/* Which bits have changed	*/
+
+/* MIF Mask Register.  This 16-bit register is used when in poll mode
+ * to say which bits of the polled register will cause an interrupt
+ * when changed.
+ */
+
+/* PCS/Serialink Registers */
+#define PCS_MIICTRL	0x9000UL	/* PCS MII Control Register	*/
+#define PCS_MIISTAT	0x9004UL	/* PCS MII Status Register	*/
+#define PCS_MIIADV	0x9008UL	/* PCS MII Advertisement Reg	*/
+#define PCS_MIILP	0x900CUL	/* PCS MII Link Partner Ability	*/
+#define PCS_CFG		0x9010UL	/* PCS Configuration Register	*/
+#define PCS_SMACHINE	0x9014UL	/* PCS State Machine Register	*/
+#define PCS_ISTAT	0x9018UL	/* PCS Interrupt Status Reg	*/
+#define PCS_DMODE	0x9050UL	/* Datapath Mode Register	*/
+#define PCS_SCTRL	0x9054UL	/* Serialink Control Register	*/
+#define PCS_SOS		0x9058UL	/* Shared Output Select Reg	*/
+#define PCS_SSTATE	0x905CUL	/* Serialink State Register	*/
+
+/* PCD MII Control Register. */
+#define PCS_MIICTRL_SPD	0x00000040	/* Read as one, writes ignored	*/
+#define PCS_MIICTRL_CT	0x00000080	/* Force COL signal active	*/
+#define PCS_MIICTRL_DM	0x00000100	/* Duplex mode, forced low	*/
+#define PCS_MIICTRL_RAN	0x00000200	/* Restart auto-neg, self clear	*/
+#define PCS_MIICTRL_ISO	0x00000400	/* Read as zero, writes ignored	*/
+#define PCS_MIICTRL_PD	0x00000800	/* Read as zero, writes ignored	*/
+#define PCS_MIICTRL_ANE	0x00001000	/* Auto-neg enable		*/
+#define PCS_MIICTRL_SS	0x00002000	/* Read as zero, writes ignored	*/
+#define PCS_MIICTRL_WB	0x00004000	/* Wrapback, loopback at 10-bit
+					 * input side of Serialink
+					 */
+#define PCS_MIICTRL_RST	0x00008000	/* Resets PCS, self clearing	*/
+
+/* PCS MII Status Register. */
+#define PCS_MIISTAT_EC	0x00000001	/* Ext Capability: Read as zero	*/
+#define PCS_MIISTAT_JD	0x00000002	/* Jabber Detect: Read as zero	*/
+#define PCS_MIISTAT_LS	0x00000004	/* Link Status: 1=up 0=down	*/
+#define PCS_MIISTAT_ANA	0x00000008	/* Auto-neg Ability, always 1	*/
+#define PCS_MIISTAT_RF	0x00000010	/* Remote Fault			*/
+#define PCS_MIISTAT_ANC	0x00000020	/* Auto-neg complete		*/
+#define PCS_MIISTAT_ES	0x00000100	/* Extended Status, always 1	*/
+
+/* PCS MII Advertisement Register. */
+#define PCS_MIIADV_FD	0x00000020	/* Advertise Full Duplex	*/
+#define PCS_MIIADV_HD	0x00000040	/* Advertise Half Duplex	*/
+#define PCS_MIIADV_SP	0x00000080	/* Advertise Symmetric Pause	*/
+#define PCS_MIIADV_AP	0x00000100	/* Advertise Asymmetric Pause	*/
+#define PCS_MIIADV_RF	0x00003000	/* Remote Fault			*/
+#define PCS_MIIADV_ACK	0x00004000	/* Read-only			*/
+#define PCS_MIIADV_NP	0x00008000	/* Next-page, forced low	*/
+
+/* PCS MII Link Partner Ability Register.   This register is equivalent
+ * to the Link Partnet Ability Register of the standard MII register set.
+ * It's layout corresponds to the PCS MII Advertisement Register.
+ */
+
+/* PCS Configuration Register. */
+#define PCS_CFG_ENABLE	0x00000001	/* Must be zero while changing
+					 * PCS MII advertisement reg.
+					 */
+#define PCS_CFG_SDO	0x00000002	/* Signal detect override	*/
+#define PCS_CFG_SDL	0x00000004	/* Signal detect active low	*/
+#define PCS_CFG_JS	0x00000018	/* Jitter-study:
+					 * 0 = normal operation
+					 * 1 = high-frequency test pattern
+					 * 2 = low-frequency test pattern
+					 * 3 = reserved
+					 */
+#define PCS_CFG_TO	0x00000020	/* 10ms auto-neg timer override	*/
+
+/* PCS Interrupt Status Register.  This register is self-clearing
+ * when read.
+ */
+#define PCS_ISTAT_LSC	0x00000004	/* Link Status Change		*/
+
+/* Datapath Mode Register. */
+#define PCS_DMODE_SM	0x00000001	/* 1 = use internal Serialink	*/
+#define PCS_DMODE_ESM	0x00000002	/* External SERDES mode		*/
+#define PCS_DMODE_MGM	0x00000004	/* MII/GMII mode		*/
+#define PCS_DMODE_GMOE	0x00000008	/* GMII Output Enable		*/
+
+/* Serialink Control Register.
+ *
+ * NOTE: When in SERDES mode, the loopback bit has inverse logic.
+ */
+#define PCS_SCTRL_LOOP	0x00000001	/* Loopback enable		*/
+#define PCS_SCTRL_ESCD	0x00000002	/* Enable sync char detection	*/
+#define PCS_SCTRL_LOCK	0x00000004	/* Lock to reference clock	*/
+#define PCS_SCTRL_EMP	0x00000018	/* Output driver emphasis	*/
+#define PCS_SCTRL_STEST	0x000001c0	/* Self test patterns		*/
+#define PCS_SCTRL_PDWN	0x00000200	/* Software power-down		*/
+#define PCS_SCTRL_RXZ	0x00000c00	/* PLL input to Serialink	*/
+#define PCS_SCTRL_RXP	0x00003000	/* PLL input to Serialink	*/
+#define PCS_SCTRL_TXZ	0x0000c000	/* PLL input to Serialink	*/
+#define PCS_SCTRL_TXP	0x00030000	/* PLL input to Serialink	*/
+
+/* Shared Output Select Register.  For test and debug, allows multiplexing
+ * test outputs into the PROM address pins.  Set to zero for normal
+ * operation.
+ */
+#define PCS_SOS_PADDR	0x00000003	/* PROM Address			*/
+
+/* PROM Image Space */
+#define PROM_START	0x100000UL	/* Expansion ROM run time access*/
+#define PROM_SIZE	0x0fffffUL	/* Size of ROM			*/
+#define PROM_END	0x200000UL	/* End of ROM			*/
+
+/* MII definitions missing from mii.h */
+
+#define BMCR_SPD2	0x0040		/* Gigabit enable? (bcm5411)	*/
+#define LPA_PAUSE	0x0400
+
+/* More PHY registers (specific to Broadcom models) */
+
+/* MII BCM5201 MULTIPHY interrupt register */
+#define MII_BCM5201_INTERRUPT			0x1A
+#define MII_BCM5201_INTERRUPT_INTENABLE		0x4000
+
+#define MII_BCM5201_AUXMODE2			0x1B
+#define MII_BCM5201_AUXMODE2_LOWPOWER		0x0008
+
+#define MII_BCM5201_MULTIPHY                    0x1E
+
+/* MII BCM5201 MULTIPHY register bits */
+#define MII_BCM5201_MULTIPHY_SERIALMODE         0x0002
+#define MII_BCM5201_MULTIPHY_SUPERISOLATE       0x0008
+
+/* MII BCM5400 1000-BASET Control register */
+#define MII_BCM5400_GB_CONTROL			0x09
+#define MII_BCM5400_GB_CONTROL_FULLDUPLEXCAP	0x0200
+
+/* MII BCM5400 AUXCONTROL register */
+#define MII_BCM5400_AUXCONTROL                  0x18
+#define MII_BCM5400_AUXCONTROL_PWR10BASET       0x0004
+
+/* MII BCM5400 AUXSTATUS register */
+#define MII_BCM5400_AUXSTATUS                   0x19
+#define MII_BCM5400_AUXSTATUS_LINKMODE_MASK     0x0700
+#define MII_BCM5400_AUXSTATUS_LINKMODE_SHIFT    8
+
+/* When it can, GEM internally caches 4 aligned TX descriptors
+ * at a time, so that it can use full cacheline DMA reads.
+ *
+ * Note that unlike HME, there is no ownership bit in the descriptor
+ * control word.  The same functionality is obtained via the TX-Kick
+ * and TX-Complete registers.  As a result, GEM need not write back
+ * updated values to the TX descriptor ring, it only performs reads.
+ *
+ * Since TX descriptors are never modified by GEM, the driver can
+ * use the buffer DMA address as a place to keep track of allocated
+ * DMA mappings for a transmitted packet.
+ */
+struct gem_txd {
+	__le64	control_word;
+	__le64	buffer;
+};
+
+#define TXDCTRL_BUFSZ	0x0000000000007fffULL	/* Buffer Size		*/
+#define TXDCTRL_CSTART	0x00000000001f8000ULL	/* CSUM Start Offset	*/
+#define TXDCTRL_COFF	0x000000001fe00000ULL	/* CSUM Stuff Offset	*/
+#define TXDCTRL_CENAB	0x0000000020000000ULL	/* CSUM Enable		*/
+#define TXDCTRL_EOF	0x0000000040000000ULL	/* End of Frame		*/
+#define TXDCTRL_SOF	0x0000000080000000ULL	/* Start of Frame	*/
+#define TXDCTRL_INTME	0x0000000100000000ULL	/* "Interrupt Me"	*/
+#define TXDCTRL_NOCRC	0x0000000200000000ULL	/* No CRC Present	*/
+
+/* GEM requires that RX descriptors are provided four at a time,
+ * aligned.  Also, the RX ring may not wrap around.  This means that
+ * there will be at least 4 unused descriptor entries in the middle
+ * of the RX ring at all times.
+ *
+ * Similar to HME, GEM assumes that it can write garbage bytes before
+ * the beginning of the buffer and right after the end in order to DMA
+ * whole cachelines.
+ *
+ * Unlike for TX, GEM does update the status word in the RX descriptors
+ * when packets arrive.  Therefore an ownership bit does exist in the
+ * RX descriptors.  It is advisory, GEM clears it but does not check
+ * it in any way.  So when buffers are posted to the RX ring (via the
+ * RX Kick register) by the driver it must make sure the buffers are
+ * truly ready and that the ownership bits are set properly.
+ *
+ * Even though GEM modifies the RX descriptors, it guarantees that the
+ * buffer DMA address field will stay the same when it performs these
+ * updates.  Therefore it can be used to keep track of DMA mappings
+ * by the host driver just as in the TX descriptor case above.
+ */
+struct gem_rxd {
+	__le64	status_word;
+	__le64	buffer;
+};
+
+#define RXDCTRL_TCPCSUM	0x000000000000ffffULL	/* TCP Pseudo-CSUM	*/
+#define RXDCTRL_BUFSZ	0x000000007fff0000ULL	/* Buffer Size		*/
+#define RXDCTRL_OWN	0x0000000080000000ULL	/* GEM owns this entry	*/
+#define RXDCTRL_HASHVAL	0x0ffff00000000000ULL	/* Hash Value		*/
+#define RXDCTRL_HPASS	0x1000000000000000ULL	/* Passed Hash Filter	*/
+#define RXDCTRL_ALTMAC	0x2000000000000000ULL	/* Matched ALT MAC	*/
+#define RXDCTRL_BAD	0x4000000000000000ULL	/* Frame has bad CRC	*/
+
+#define RXDCTRL_FRESH(gp)	\
+	((((RX_BUF_ALLOC_SIZE(gp) - RX_OFFSET) << 16) & RXDCTRL_BUFSZ) | \
+	 RXDCTRL_OWN)
+
+#define TX_RING_SIZE 128
+#define RX_RING_SIZE 128
+
+#if TX_RING_SIZE == 32
+#define TXDMA_CFG_BASE	TXDMA_CFG_RINGSZ_32
+#elif TX_RING_SIZE == 64
+#define TXDMA_CFG_BASE	TXDMA_CFG_RINGSZ_64
+#elif TX_RING_SIZE == 128
+#define TXDMA_CFG_BASE	TXDMA_CFG_RINGSZ_128
+#elif TX_RING_SIZE == 256
+#define TXDMA_CFG_BASE	TXDMA_CFG_RINGSZ_256
+#elif TX_RING_SIZE == 512
+#define TXDMA_CFG_BASE	TXDMA_CFG_RINGSZ_512
+#elif TX_RING_SIZE == 1024
+#define TXDMA_CFG_BASE	TXDMA_CFG_RINGSZ_1K
+#elif TX_RING_SIZE == 2048
+#define TXDMA_CFG_BASE	TXDMA_CFG_RINGSZ_2K
+#elif TX_RING_SIZE == 4096
+#define TXDMA_CFG_BASE	TXDMA_CFG_RINGSZ_4K
+#elif TX_RING_SIZE == 8192
+#define TXDMA_CFG_BASE	TXDMA_CFG_RINGSZ_8K
+#else
+#error TX_RING_SIZE value is illegal...
+#endif
+
+#if RX_RING_SIZE == 32
+#define RXDMA_CFG_BASE	RXDMA_CFG_RINGSZ_32
+#elif RX_RING_SIZE == 64
+#define RXDMA_CFG_BASE	RXDMA_CFG_RINGSZ_64
+#elif RX_RING_SIZE == 128
+#define RXDMA_CFG_BASE	RXDMA_CFG_RINGSZ_128
+#elif RX_RING_SIZE == 256
+#define RXDMA_CFG_BASE	RXDMA_CFG_RINGSZ_256
+#elif RX_RING_SIZE == 512
+#define RXDMA_CFG_BASE	RXDMA_CFG_RINGSZ_512
+#elif RX_RING_SIZE == 1024
+#define RXDMA_CFG_BASE	RXDMA_CFG_RINGSZ_1K
+#elif RX_RING_SIZE == 2048
+#define RXDMA_CFG_BASE	RXDMA_CFG_RINGSZ_2K
+#elif RX_RING_SIZE == 4096
+#define RXDMA_CFG_BASE	RXDMA_CFG_RINGSZ_4K
+#elif RX_RING_SIZE == 8192
+#define RXDMA_CFG_BASE	RXDMA_CFG_RINGSZ_8K
+#else
+#error RX_RING_SIZE is illegal...
+#endif
+
+#define NEXT_TX(N)	(((N) + 1) & (TX_RING_SIZE - 1))
+#define NEXT_RX(N)	(((N) + 1) & (RX_RING_SIZE - 1))
+
+#define TX_BUFFS_AVAIL(GP)					\
+	(((GP)->tx_old <= (GP)->tx_new) ?			\
+	  (GP)->tx_old + (TX_RING_SIZE - 1) - (GP)->tx_new :	\
+	  (GP)->tx_old - (GP)->tx_new - 1)
+
+#define RX_OFFSET          2
+#define RX_BUF_ALLOC_SIZE(gp)	((gp)->rx_buf_sz + 28 + RX_OFFSET + 64)
+
+#define RX_COPY_THRESHOLD  256
+
+#if TX_RING_SIZE < 128
+#define INIT_BLOCK_TX_RING_SIZE		128
+#else
+#define INIT_BLOCK_TX_RING_SIZE		TX_RING_SIZE
+#endif
+
+#if RX_RING_SIZE < 128
+#define INIT_BLOCK_RX_RING_SIZE		128
+#else
+#define INIT_BLOCK_RX_RING_SIZE		RX_RING_SIZE
+#endif
+
+struct gem_init_block {
+	struct gem_txd	txd[INIT_BLOCK_TX_RING_SIZE];
+	struct gem_rxd	rxd[INIT_BLOCK_RX_RING_SIZE];
+};
+
+enum gem_phy_type {
+	phy_mii_mdio0,
+	phy_mii_mdio1,
+	phy_serialink,
+	phy_serdes,
+};
+
+enum link_state {
+	link_down = 0,	/* No link, will retry */
+	link_aneg,	/* Autoneg in progress */
+	link_force_try,	/* Try Forced link speed */
+	link_force_ret,	/* Forced mode worked, retrying autoneg */
+	link_force_ok,	/* Stay in forced mode */
+	link_up		/* Link is up */
+};
+
+struct gem {
+	void __iomem		*regs;
+	int			rx_new, rx_old;
+	int			tx_new, tx_old;
+
+	unsigned int has_wol : 1;	/* chip supports wake-on-lan */
+	unsigned int asleep_wol : 1;	/* was asleep with WOL enabled */
+
+	int			cell_enabled;
+	u32			msg_enable;
+	u32			status;
+
+	struct napi_struct	napi;
+
+	int			tx_fifo_sz;
+	int			rx_fifo_sz;
+	int			rx_pause_off;
+	int			rx_pause_on;
+	int			rx_buf_sz;
+	u64			pause_entered;
+	u16			pause_last_time_recvd;
+	u32			mac_rx_cfg;
+	u32			swrst_base;
+
+	int			want_autoneg;
+	int			last_forced_speed;
+	enum link_state		lstate;
+	struct timer_list	link_timer;
+	int			timer_ticks;
+	int			wake_on_lan;
+	struct work_struct	reset_task;
+	volatile int		reset_task_pending;
+
+	enum gem_phy_type	phy_type;
+	struct mii_phy		phy_mii;
+	int			mii_phy_addr;
+
+	struct gem_init_block	*init_block;
+	struct sk_buff		*rx_skbs[RX_RING_SIZE];
+	struct sk_buff		*tx_skbs[TX_RING_SIZE];
+	dma_addr_t		gblock_dvma;
+
+	struct pci_dev		*pdev;
+	struct net_device	*dev;
+#if defined(CONFIG_PPC_PMAC) || defined(CONFIG_SPARC)
+	struct device_node	*of_node;
+#endif
+};
+
+#define found_mii_phy(gp) ((gp->phy_type == phy_mii_mdio0 || gp->phy_type == phy_mii_mdio1) && \
+			   gp->phy_mii.def && gp->phy_mii.def->ops)
+
+#endif /* _SUNGEM_H */
diff --git a/drivers/net/ethernet/sun/sunhme.c b/drivers/net/ethernet/sun/sunhme.c
new file mode 100644
index 000000000..a70f704eb
--- /dev/null
+++ b/drivers/net/ethernet/sun/sunhme.c
@@ -0,0 +1,2901 @@
+// SPDX-License-Identifier: GPL-2.0
+/* sunhme.c: Sparc HME/BigMac 10/100baseT half/full duplex auto switching,
+ *           auto carrier detecting ethernet driver.  Also known as the
+ *           "Happy Meal Ethernet" found on SunSwift SBUS cards.
+ *
+ * Copyright (C) 1996, 1998, 1999, 2002, 2003,
+ *		2006, 2008 David S. Miller (davem@davemloft.net)
+ *
+ * Changes :
+ * 2000/11/11 Willy Tarreau <willy AT meta-x.org>
+ *   - port to non-sparc architectures. Tested only on x86 and
+ *     only currently works with QFE PCI cards.
+ *   - ability to specify the MAC address at module load time by passing this
+ *     argument : macaddr=0x00,0x10,0x20,0x30,0x40,0x50
+ */
+
+#include <linux/bitops.h>
+#include <linux/crc32.h>
+#include <linux/delay.h>
+#include <linux/dma-mapping.h>
+#include <linux/errno.h>
+#include <linux/etherdevice.h>
+#include <linux/ethtool.h>
+#include <linux/fcntl.h>
+#include <linux/in.h>
+#include <linux/init.h>
+#include <linux/interrupt.h>
+#include <linux/io.h>
+#include <linux/ioport.h>
+#include <linux/kernel.h>
+#include <linux/mii.h>
+#include <linux/mm.h>
+#include <linux/module.h>
+#include <linux/netdevice.h>
+#include <linux/of.h>
+#include <linux/of_device.h>
+#include <linux/pci.h>
+#include <linux/platform_device.h>
+#include <linux/random.h>
+#include <linux/skbuff.h>
+#include <linux/slab.h>
+#include <linux/string.h>
+#include <linux/types.h>
+#include <linux/uaccess.h>
+
+#include <asm/byteorder.h>
+#include <asm/dma.h>
+#include <asm/irq.h>
+
+#ifdef CONFIG_SPARC
+#include <asm/auxio.h>
+#include <asm/idprom.h>
+#include <asm/openprom.h>
+#include <asm/oplib.h>
+#include <asm/prom.h>
+#endif
+
+#include "sunhme.h"
+
+#define DRV_NAME	"sunhme"
+
+MODULE_AUTHOR("David S. Miller <davem@davemloft.net>");
+MODULE_DESCRIPTION("Sun HappyMealEthernet(HME) 10/100baseT ethernet driver");
+MODULE_LICENSE("GPL");
+
+static int macaddr[6];
+
+/* accept MAC address of the form macaddr=0x08,0x00,0x20,0x30,0x40,0x50 */
+module_param_array(macaddr, int, NULL, 0);
+MODULE_PARM_DESC(macaddr, "Happy Meal MAC address to set");
+
+#ifdef CONFIG_SBUS
+static struct quattro *qfe_sbus_list;
+#endif
+
+#ifdef CONFIG_PCI
+static struct quattro *qfe_pci_list;
+#endif
+
+#define hme_debug(fmt, ...) pr_debug("%s: " fmt, __func__, ##__VA_ARGS__)
+#define HMD hme_debug
+
+/* "Auto Switch Debug" aka phy debug */
+#if 1
+#define ASD hme_debug
+#else
+#define ASD(...)
+#endif
+
+#if 0
+struct hme_tx_logent {
+	unsigned int tstamp;
+	int tx_new, tx_old;
+	unsigned int action;
+#define TXLOG_ACTION_IRQ	0x01
+#define TXLOG_ACTION_TXMIT	0x02
+#define TXLOG_ACTION_TBUSY	0x04
+#define TXLOG_ACTION_NBUFS	0x08
+	unsigned int status;
+};
+#define TX_LOG_LEN	128
+static struct hme_tx_logent tx_log[TX_LOG_LEN];
+static int txlog_cur_entry;
+static __inline__ void tx_add_log(struct happy_meal *hp, unsigned int a, unsigned int s)
+{
+	struct hme_tx_logent *tlp;
+	unsigned long flags;
+
+	local_irq_save(flags);
+	tlp = &tx_log[txlog_cur_entry];
+	tlp->tstamp = (unsigned int)jiffies;
+	tlp->tx_new = hp->tx_new;
+	tlp->tx_old = hp->tx_old;
+	tlp->action = a;
+	tlp->status = s;
+	txlog_cur_entry = (txlog_cur_entry + 1) & (TX_LOG_LEN - 1);
+	local_irq_restore(flags);
+}
+static __inline__ void tx_dump_log(void)
+{
+	int i, this;
+
+	this = txlog_cur_entry;
+	for (i = 0; i < TX_LOG_LEN; i++) {
+		pr_err("TXLOG[%d]: j[%08x] tx[N(%d)O(%d)] action[%08x] stat[%08x]\n", i,
+		       tx_log[this].tstamp,
+		       tx_log[this].tx_new, tx_log[this].tx_old,
+		       tx_log[this].action, tx_log[this].status);
+		this = (this + 1) & (TX_LOG_LEN - 1);
+	}
+}
+#else
+#define tx_add_log(hp, a, s)
+#define tx_dump_log()
+#endif
+
+#define DEFAULT_IPG0      16 /* For lance-mode only */
+#define DEFAULT_IPG1       8 /* For all modes */
+#define DEFAULT_IPG2       4 /* For all modes */
+#define DEFAULT_JAMSIZE    4 /* Toe jam */
+
+/* NOTE: In the descriptor writes one _must_ write the address
+ *	 member _first_.  The card must not be allowed to see
+ *	 the updated descriptor flags until the address is
+ *	 correct.  I've added a write memory barrier between
+ *	 the two stores so that I can sleep well at night... -DaveM
+ */
+
+#if defined(CONFIG_SBUS) && defined(CONFIG_PCI)
+static void sbus_hme_write32(void __iomem *reg, u32 val)
+{
+	sbus_writel(val, reg);
+}
+
+static u32 sbus_hme_read32(void __iomem *reg)
+{
+	return sbus_readl(reg);
+}
+
+static void sbus_hme_write_rxd(struct happy_meal_rxd *rxd, u32 flags, u32 addr)
+{
+	rxd->rx_addr = (__force hme32)addr;
+	dma_wmb();
+	rxd->rx_flags = (__force hme32)flags;
+}
+
+static void sbus_hme_write_txd(struct happy_meal_txd *txd, u32 flags, u32 addr)
+{
+	txd->tx_addr = (__force hme32)addr;
+	dma_wmb();
+	txd->tx_flags = (__force hme32)flags;
+}
+
+static u32 sbus_hme_read_desc32(hme32 *p)
+{
+	return (__force u32)*p;
+}
+
+static void pci_hme_write32(void __iomem *reg, u32 val)
+{
+	writel(val, reg);
+}
+
+static u32 pci_hme_read32(void __iomem *reg)
+{
+	return readl(reg);
+}
+
+static void pci_hme_write_rxd(struct happy_meal_rxd *rxd, u32 flags, u32 addr)
+{
+	rxd->rx_addr = (__force hme32)cpu_to_le32(addr);
+	dma_wmb();
+	rxd->rx_flags = (__force hme32)cpu_to_le32(flags);
+}
+
+static void pci_hme_write_txd(struct happy_meal_txd *txd, u32 flags, u32 addr)
+{
+	txd->tx_addr = (__force hme32)cpu_to_le32(addr);
+	dma_wmb();
+	txd->tx_flags = (__force hme32)cpu_to_le32(flags);
+}
+
+static u32 pci_hme_read_desc32(hme32 *p)
+{
+	return le32_to_cpup((__le32 *)p);
+}
+
+#define hme_write32(__hp, __reg, __val) \
+	((__hp)->write32((__reg), (__val)))
+#define hme_read32(__hp, __reg) \
+	((__hp)->read32(__reg))
+#define hme_write_rxd(__hp, __rxd, __flags, __addr) \
+	((__hp)->write_rxd((__rxd), (__flags), (__addr)))
+#define hme_write_txd(__hp, __txd, __flags, __addr) \
+	((__hp)->write_txd((__txd), (__flags), (__addr)))
+#define hme_read_desc32(__hp, __p) \
+	((__hp)->read_desc32(__p))
+#else
+#ifdef CONFIG_SBUS
+/* SBUS only compilation */
+#define hme_write32(__hp, __reg, __val) \
+	sbus_writel((__val), (__reg))
+#define hme_read32(__hp, __reg) \
+	sbus_readl(__reg)
+#define hme_write_rxd(__hp, __rxd, __flags, __addr) \
+do {	(__rxd)->rx_addr = (__force hme32)(u32)(__addr); \
+	dma_wmb(); \
+	(__rxd)->rx_flags = (__force hme32)(u32)(__flags); \
+} while(0)
+#define hme_write_txd(__hp, __txd, __flags, __addr) \
+do {	(__txd)->tx_addr = (__force hme32)(u32)(__addr); \
+	dma_wmb(); \
+	(__txd)->tx_flags = (__force hme32)(u32)(__flags); \
+} while(0)
+#define hme_read_desc32(__hp, __p)	((__force u32)(hme32)*(__p))
+#else
+/* PCI only compilation */
+#define hme_write32(__hp, __reg, __val) \
+	writel((__val), (__reg))
+#define hme_read32(__hp, __reg) \
+	readl(__reg)
+#define hme_write_rxd(__hp, __rxd, __flags, __addr) \
+do {	(__rxd)->rx_addr = (__force hme32)cpu_to_le32(__addr); \
+	dma_wmb(); \
+	(__rxd)->rx_flags = (__force hme32)cpu_to_le32(__flags); \
+} while(0)
+#define hme_write_txd(__hp, __txd, __flags, __addr) \
+do {	(__txd)->tx_addr = (__force hme32)cpu_to_le32(__addr); \
+	dma_wmb(); \
+	(__txd)->tx_flags = (__force hme32)cpu_to_le32(__flags); \
+} while(0)
+static inline u32 hme_read_desc32(struct happy_meal *hp, hme32 *p)
+{
+	return le32_to_cpup((__le32 *)p);
+}
+#endif
+#endif
+
+
+/* Oh yes, the MIF BitBang is mighty fun to program.  BitBucket is more like it. */
+static void BB_PUT_BIT(struct happy_meal *hp, void __iomem *tregs, int bit)
+{
+	hme_write32(hp, tregs + TCVR_BBDATA, bit);
+	hme_write32(hp, tregs + TCVR_BBCLOCK, 0);
+	hme_write32(hp, tregs + TCVR_BBCLOCK, 1);
+}
+
+#if 0
+static u32 BB_GET_BIT(struct happy_meal *hp, void __iomem *tregs, int internal)
+{
+	u32 ret;
+
+	hme_write32(hp, tregs + TCVR_BBCLOCK, 0);
+	hme_write32(hp, tregs + TCVR_BBCLOCK, 1);
+	ret = hme_read32(hp, tregs + TCVR_CFG);
+	if (internal)
+		ret &= TCV_CFG_MDIO0;
+	else
+		ret &= TCV_CFG_MDIO1;
+
+	return ret;
+}
+#endif
+
+static u32 BB_GET_BIT2(struct happy_meal *hp, void __iomem *tregs, int internal)
+{
+	u32 retval;
+
+	hme_write32(hp, tregs + TCVR_BBCLOCK, 0);
+	udelay(1);
+	retval = hme_read32(hp, tregs + TCVR_CFG);
+	if (internal)
+		retval &= TCV_CFG_MDIO0;
+	else
+		retval &= TCV_CFG_MDIO1;
+	hme_write32(hp, tregs + TCVR_BBCLOCK, 1);
+
+	return retval;
+}
+
+#define TCVR_FAILURE      0x80000000     /* Impossible MIF read value */
+
+static int happy_meal_bb_read(struct happy_meal *hp,
+			      void __iomem *tregs, int reg)
+{
+	u32 tmp;
+	int retval = 0;
+	int i;
+
+	/* Enable the MIF BitBang outputs. */
+	hme_write32(hp, tregs + TCVR_BBOENAB, 1);
+
+	/* Force BitBang into the idle state. */
+	for (i = 0; i < 32; i++)
+		BB_PUT_BIT(hp, tregs, 1);
+
+	/* Give it the read sequence. */
+	BB_PUT_BIT(hp, tregs, 0);
+	BB_PUT_BIT(hp, tregs, 1);
+	BB_PUT_BIT(hp, tregs, 1);
+	BB_PUT_BIT(hp, tregs, 0);
+
+	/* Give it the PHY address. */
+	tmp = hp->paddr & 0xff;
+	for (i = 4; i >= 0; i--)
+		BB_PUT_BIT(hp, tregs, ((tmp >> i) & 1));
+
+	/* Tell it what register we want to read. */
+	tmp = (reg & 0xff);
+	for (i = 4; i >= 0; i--)
+		BB_PUT_BIT(hp, tregs, ((tmp >> i) & 1));
+
+	/* Close down the MIF BitBang outputs. */
+	hme_write32(hp, tregs + TCVR_BBOENAB, 0);
+
+	/* Now read in the value. */
+	(void) BB_GET_BIT2(hp, tregs, (hp->tcvr_type == internal));
+	for (i = 15; i >= 0; i--)
+		retval |= BB_GET_BIT2(hp, tregs, (hp->tcvr_type == internal));
+	(void) BB_GET_BIT2(hp, tregs, (hp->tcvr_type == internal));
+	(void) BB_GET_BIT2(hp, tregs, (hp->tcvr_type == internal));
+	(void) BB_GET_BIT2(hp, tregs, (hp->tcvr_type == internal));
+	ASD("reg=%d value=%x\n", reg, retval);
+	return retval;
+}
+
+static void happy_meal_bb_write(struct happy_meal *hp,
+				void __iomem *tregs, int reg,
+				unsigned short value)
+{
+	u32 tmp;
+	int i;
+
+	ASD("reg=%d value=%x\n", reg, value);
+
+	/* Enable the MIF BitBang outputs. */
+	hme_write32(hp, tregs + TCVR_BBOENAB, 1);
+
+	/* Force BitBang into the idle state. */
+	for (i = 0; i < 32; i++)
+		BB_PUT_BIT(hp, tregs, 1);
+
+	/* Give it write sequence. */
+	BB_PUT_BIT(hp, tregs, 0);
+	BB_PUT_BIT(hp, tregs, 1);
+	BB_PUT_BIT(hp, tregs, 0);
+	BB_PUT_BIT(hp, tregs, 1);
+
+	/* Give it the PHY address. */
+	tmp = (hp->paddr & 0xff);
+	for (i = 4; i >= 0; i--)
+		BB_PUT_BIT(hp, tregs, ((tmp >> i) & 1));
+
+	/* Tell it what register we will be writing. */
+	tmp = (reg & 0xff);
+	for (i = 4; i >= 0; i--)
+		BB_PUT_BIT(hp, tregs, ((tmp >> i) & 1));
+
+	/* Tell it to become ready for the bits. */
+	BB_PUT_BIT(hp, tregs, 1);
+	BB_PUT_BIT(hp, tregs, 0);
+
+	for (i = 15; i >= 0; i--)
+		BB_PUT_BIT(hp, tregs, ((value >> i) & 1));
+
+	/* Close down the MIF BitBang outputs. */
+	hme_write32(hp, tregs + TCVR_BBOENAB, 0);
+}
+
+#define TCVR_READ_TRIES   16
+
+static int happy_meal_tcvr_read(struct happy_meal *hp,
+				void __iomem *tregs, int reg)
+{
+	int tries = TCVR_READ_TRIES;
+	int retval;
+
+	if (hp->tcvr_type == none) {
+		ASD("no transceiver, value=TCVR_FAILURE\n");
+		return TCVR_FAILURE;
+	}
+
+	if (!(hp->happy_flags & HFLAG_FENABLE)) {
+		ASD("doing bit bang\n");
+		return happy_meal_bb_read(hp, tregs, reg);
+	}
+
+	hme_write32(hp, tregs + TCVR_FRAME,
+		    (FRAME_READ | (hp->paddr << 23) | ((reg & 0xff) << 18)));
+	while (!(hme_read32(hp, tregs + TCVR_FRAME) & 0x10000) && --tries)
+		udelay(20);
+	if (!tries) {
+		netdev_err(hp->dev, "Aieee, transceiver MIF read bolixed\n");
+		return TCVR_FAILURE;
+	}
+	retval = hme_read32(hp, tregs + TCVR_FRAME) & 0xffff;
+	ASD("reg=0x%02x value=%04x\n", reg, retval);
+	return retval;
+}
+
+#define TCVR_WRITE_TRIES  16
+
+static void happy_meal_tcvr_write(struct happy_meal *hp,
+				  void __iomem *tregs, int reg,
+				  unsigned short value)
+{
+	int tries = TCVR_WRITE_TRIES;
+
+	ASD("reg=0x%02x value=%04x\n", reg, value);
+
+	/* Welcome to Sun Microsystems, can I take your order please? */
+	if (!(hp->happy_flags & HFLAG_FENABLE)) {
+		happy_meal_bb_write(hp, tregs, reg, value);
+		return;
+	}
+
+	/* Would you like fries with that? */
+	hme_write32(hp, tregs + TCVR_FRAME,
+		    (FRAME_WRITE | (hp->paddr << 23) |
+		     ((reg & 0xff) << 18) | (value & 0xffff)));
+	while (!(hme_read32(hp, tregs + TCVR_FRAME) & 0x10000) && --tries)
+		udelay(20);
+
+	/* Anything else? */
+	if (!tries)
+		netdev_err(hp->dev, "Aieee, transceiver MIF write bolixed\n");
+
+	/* Fifty-two cents is your change, have a nice day. */
+}
+
+/* Auto negotiation.  The scheme is very simple.  We have a timer routine
+ * that keeps watching the auto negotiation process as it progresses.
+ * The DP83840 is first told to start doing it's thing, we set up the time
+ * and place the timer state machine in its initial state.
+ *
+ * Here the timer peeks at the DP83840 status registers at each click to see
+ * if the auto negotiation has completed, we assume here that the DP83840 PHY
+ * will time out at some point and just tell us what (didn't) happen.  For
+ * complete coverage we only allow so many of the ticks at this level to run,
+ * when this has expired we print a warning message and try another strategy.
+ * This "other" strategy is to force the interface into various speed/duplex
+ * configurations and we stop when we see a link-up condition before the
+ * maximum number of "peek" ticks have occurred.
+ *
+ * Once a valid link status has been detected we configure the BigMAC and
+ * the rest of the Happy Meal to speak the most efficient protocol we could
+ * get a clean link for.  The priority for link configurations, highest first
+ * is:
+ *                 100 Base-T Full Duplex
+ *                 100 Base-T Half Duplex
+ *                 10 Base-T Full Duplex
+ *                 10 Base-T Half Duplex
+ *
+ * We start a new timer now, after a successful auto negotiation status has
+ * been detected.  This timer just waits for the link-up bit to get set in
+ * the BMCR of the DP83840.  When this occurs we print a kernel log message
+ * describing the link type in use and the fact that it is up.
+ *
+ * If a fatal error of some sort is signalled and detected in the interrupt
+ * service routine, and the chip is reset, or the link is ifconfig'd down
+ * and then back up, this entire process repeats itself all over again.
+ */
+static int try_next_permutation(struct happy_meal *hp, void __iomem *tregs)
+{
+	hp->sw_bmcr = happy_meal_tcvr_read(hp, tregs, MII_BMCR);
+
+	/* Downgrade from full to half duplex.  Only possible
+	 * via ethtool.
+	 */
+	if (hp->sw_bmcr & BMCR_FULLDPLX) {
+		hp->sw_bmcr &= ~(BMCR_FULLDPLX);
+		happy_meal_tcvr_write(hp, tregs, MII_BMCR, hp->sw_bmcr);
+		return 0;
+	}
+
+	/* Downgrade from 100 to 10. */
+	if (hp->sw_bmcr & BMCR_SPEED100) {
+		hp->sw_bmcr &= ~(BMCR_SPEED100);
+		happy_meal_tcvr_write(hp, tregs, MII_BMCR, hp->sw_bmcr);
+		return 0;
+	}
+
+	/* We've tried everything. */
+	return -1;
+}
+
+static void display_link_mode(struct happy_meal *hp, void __iomem *tregs)
+{
+	hp->sw_lpa = happy_meal_tcvr_read(hp, tregs, MII_LPA);
+
+	netdev_info(hp->dev,
+		    "Link is up using %s transceiver at %dMb/s, %s Duplex.\n",
+		    hp->tcvr_type == external ? "external" : "internal",
+		    hp->sw_lpa & (LPA_100HALF | LPA_100FULL) ? 100 : 10,
+		    hp->sw_lpa & (LPA_100FULL | LPA_10FULL) ? "Full" : "Half");
+}
+
+static void display_forced_link_mode(struct happy_meal *hp, void __iomem *tregs)
+{
+	hp->sw_bmcr = happy_meal_tcvr_read(hp, tregs, MII_BMCR);
+
+	netdev_info(hp->dev,
+		    "Link has been forced up using %s transceiver at %dMb/s, %s Duplex.\n",
+		    hp->tcvr_type == external ? "external" : "internal",
+		    hp->sw_bmcr & BMCR_SPEED100 ? 100 : 10,
+		    hp->sw_bmcr & BMCR_FULLDPLX ? "Full" : "Half");
+}
+
+static int set_happy_link_modes(struct happy_meal *hp, void __iomem *tregs)
+{
+	int full;
+
+	/* All we care about is making sure the bigmac tx_cfg has a
+	 * proper duplex setting.
+	 */
+	if (hp->timer_state == arbwait) {
+		hp->sw_lpa = happy_meal_tcvr_read(hp, tregs, MII_LPA);
+		if (!(hp->sw_lpa & (LPA_10HALF | LPA_10FULL | LPA_100HALF | LPA_100FULL)))
+			goto no_response;
+		if (hp->sw_lpa & LPA_100FULL)
+			full = 1;
+		else if (hp->sw_lpa & LPA_100HALF)
+			full = 0;
+		else if (hp->sw_lpa & LPA_10FULL)
+			full = 1;
+		else
+			full = 0;
+	} else {
+		/* Forcing a link mode. */
+		hp->sw_bmcr = happy_meal_tcvr_read(hp, tregs, MII_BMCR);
+		if (hp->sw_bmcr & BMCR_FULLDPLX)
+			full = 1;
+		else
+			full = 0;
+	}
+
+	/* Before changing other bits in the tx_cfg register, and in
+	 * general any of other the TX config registers too, you
+	 * must:
+	 * 1) Clear Enable
+	 * 2) Poll with reads until that bit reads back as zero
+	 * 3) Make TX configuration changes
+	 * 4) Set Enable once more
+	 */
+	hme_write32(hp, hp->bigmacregs + BMAC_TXCFG,
+		    hme_read32(hp, hp->bigmacregs + BMAC_TXCFG) &
+		    ~(BIGMAC_TXCFG_ENABLE));
+	while (hme_read32(hp, hp->bigmacregs + BMAC_TXCFG) & BIGMAC_TXCFG_ENABLE)
+		barrier();
+	if (full) {
+		hp->happy_flags |= HFLAG_FULL;
+		hme_write32(hp, hp->bigmacregs + BMAC_TXCFG,
+			    hme_read32(hp, hp->bigmacregs + BMAC_TXCFG) |
+			    BIGMAC_TXCFG_FULLDPLX);
+	} else {
+		hp->happy_flags &= ~(HFLAG_FULL);
+		hme_write32(hp, hp->bigmacregs + BMAC_TXCFG,
+			    hme_read32(hp, hp->bigmacregs + BMAC_TXCFG) &
+			    ~(BIGMAC_TXCFG_FULLDPLX));
+	}
+	hme_write32(hp, hp->bigmacregs + BMAC_TXCFG,
+		    hme_read32(hp, hp->bigmacregs + BMAC_TXCFG) |
+		    BIGMAC_TXCFG_ENABLE);
+	return 0;
+no_response:
+	return 1;
+}
+
+static int is_lucent_phy(struct happy_meal *hp)
+{
+	void __iomem *tregs = hp->tcvregs;
+	unsigned short mr2, mr3;
+	int ret = 0;
+
+	mr2 = happy_meal_tcvr_read(hp, tregs, 2);
+	mr3 = happy_meal_tcvr_read(hp, tregs, 3);
+	if ((mr2 & 0xffff) == 0x0180 &&
+	    ((mr3 & 0xffff) >> 10) == 0x1d)
+		ret = 1;
+
+	return ret;
+}
+
+/* hp->happy_lock must be held */
+static void
+happy_meal_begin_auto_negotiation(struct happy_meal *hp,
+				  void __iomem *tregs,
+				  const struct ethtool_link_ksettings *ep)
+{
+	int timeout;
+
+	/* Read all of the registers we are interested in now. */
+	hp->sw_bmsr      = happy_meal_tcvr_read(hp, tregs, MII_BMSR);
+	hp->sw_bmcr      = happy_meal_tcvr_read(hp, tregs, MII_BMCR);
+	hp->sw_physid1   = happy_meal_tcvr_read(hp, tregs, MII_PHYSID1);
+	hp->sw_physid2   = happy_meal_tcvr_read(hp, tregs, MII_PHYSID2);
+
+	/* XXX Check BMSR_ANEGCAPABLE, should not be necessary though. */
+
+	hp->sw_advertise = happy_meal_tcvr_read(hp, tregs, MII_ADVERTISE);
+	if (!ep || ep->base.autoneg == AUTONEG_ENABLE) {
+		/* Advertise everything we can support. */
+		if (hp->sw_bmsr & BMSR_10HALF)
+			hp->sw_advertise |= (ADVERTISE_10HALF);
+		else
+			hp->sw_advertise &= ~(ADVERTISE_10HALF);
+
+		if (hp->sw_bmsr & BMSR_10FULL)
+			hp->sw_advertise |= (ADVERTISE_10FULL);
+		else
+			hp->sw_advertise &= ~(ADVERTISE_10FULL);
+		if (hp->sw_bmsr & BMSR_100HALF)
+			hp->sw_advertise |= (ADVERTISE_100HALF);
+		else
+			hp->sw_advertise &= ~(ADVERTISE_100HALF);
+		if (hp->sw_bmsr & BMSR_100FULL)
+			hp->sw_advertise |= (ADVERTISE_100FULL);
+		else
+			hp->sw_advertise &= ~(ADVERTISE_100FULL);
+		happy_meal_tcvr_write(hp, tregs, MII_ADVERTISE, hp->sw_advertise);
+
+		/* XXX Currently no Happy Meal cards I know off support 100BaseT4,
+		 * XXX and this is because the DP83840 does not support it, changes
+		 * XXX would need to be made to the tx/rx logic in the driver as well
+		 * XXX so I completely skip checking for it in the BMSR for now.
+		 */
+
+		ASD("Advertising [ %s%s%s%s]\n",
+		    hp->sw_advertise & ADVERTISE_10HALF ? "10H " : "",
+		    hp->sw_advertise & ADVERTISE_10FULL ? "10F " : "",
+		    hp->sw_advertise & ADVERTISE_100HALF ? "100H " : "",
+		    hp->sw_advertise & ADVERTISE_100FULL ? "100F " : "");
+
+		/* Enable Auto-Negotiation, this is usually on already... */
+		hp->sw_bmcr |= BMCR_ANENABLE;
+		happy_meal_tcvr_write(hp, tregs, MII_BMCR, hp->sw_bmcr);
+
+		/* Restart it to make sure it is going. */
+		hp->sw_bmcr |= BMCR_ANRESTART;
+		happy_meal_tcvr_write(hp, tregs, MII_BMCR, hp->sw_bmcr);
+
+		/* BMCR_ANRESTART self clears when the process has begun. */
+
+		timeout = 64;  /* More than enough. */
+		while (--timeout) {
+			hp->sw_bmcr = happy_meal_tcvr_read(hp, tregs, MII_BMCR);
+			if (!(hp->sw_bmcr & BMCR_ANRESTART))
+				break; /* got it. */
+			udelay(10);
+		}
+		if (!timeout) {
+			netdev_err(hp->dev,
+				   "Happy Meal would not start auto negotiation BMCR=0x%04x\n",
+				   hp->sw_bmcr);
+			netdev_notice(hp->dev,
+				      "Performing force link detection.\n");
+			goto force_link;
+		} else {
+			hp->timer_state = arbwait;
+		}
+	} else {
+force_link:
+		/* Force the link up, trying first a particular mode.
+		 * Either we are here at the request of ethtool or
+		 * because the Happy Meal would not start to autoneg.
+		 */
+
+		/* Disable auto-negotiation in BMCR, enable the duplex and
+		 * speed setting, init the timer state machine, and fire it off.
+		 */
+		if (!ep || ep->base.autoneg == AUTONEG_ENABLE) {
+			hp->sw_bmcr = BMCR_SPEED100;
+		} else {
+			if (ep->base.speed == SPEED_100)
+				hp->sw_bmcr = BMCR_SPEED100;
+			else
+				hp->sw_bmcr = 0;
+			if (ep->base.duplex == DUPLEX_FULL)
+				hp->sw_bmcr |= BMCR_FULLDPLX;
+		}
+		happy_meal_tcvr_write(hp, tregs, MII_BMCR, hp->sw_bmcr);
+
+		if (!is_lucent_phy(hp)) {
+			/* OK, seems we need do disable the transceiver for the first
+			 * tick to make sure we get an accurate link state at the
+			 * second tick.
+			 */
+			hp->sw_csconfig = happy_meal_tcvr_read(hp, tregs,
+							       DP83840_CSCONFIG);
+			hp->sw_csconfig &= ~(CSCONFIG_TCVDISAB);
+			happy_meal_tcvr_write(hp, tregs, DP83840_CSCONFIG,
+					      hp->sw_csconfig);
+		}
+		hp->timer_state = ltrywait;
+	}
+
+	hp->timer_ticks = 0;
+	hp->happy_timer.expires = jiffies + (12 * HZ)/10;  /* 1.2 sec. */
+	add_timer(&hp->happy_timer);
+}
+
+static void happy_meal_timer(struct timer_list *t)
+{
+	struct happy_meal *hp = timer_container_of(hp, t, happy_timer);
+	void __iomem *tregs = hp->tcvregs;
+	int restart_timer = 0;
+
+	spin_lock_irq(&hp->happy_lock);
+
+	hp->timer_ticks++;
+	switch(hp->timer_state) {
+	case arbwait:
+		/* Only allow for 5 ticks, thats 10 seconds and much too
+		 * long to wait for arbitration to complete.
+		 */
+		if (hp->timer_ticks >= 10) {
+			/* Enter force mode. */
+	do_force_mode:
+			hp->sw_bmcr = happy_meal_tcvr_read(hp, tregs, MII_BMCR);
+			netdev_notice(hp->dev,
+				      "Auto-Negotiation unsuccessful, trying force link mode\n");
+			hp->sw_bmcr = BMCR_SPEED100;
+			happy_meal_tcvr_write(hp, tregs, MII_BMCR, hp->sw_bmcr);
+
+			if (!is_lucent_phy(hp)) {
+				/* OK, seems we need do disable the transceiver for the first
+				 * tick to make sure we get an accurate link state at the
+				 * second tick.
+				 */
+				hp->sw_csconfig = happy_meal_tcvr_read(hp, tregs, DP83840_CSCONFIG);
+				hp->sw_csconfig &= ~(CSCONFIG_TCVDISAB);
+				happy_meal_tcvr_write(hp, tregs, DP83840_CSCONFIG, hp->sw_csconfig);
+			}
+			hp->timer_state = ltrywait;
+			hp->timer_ticks = 0;
+			restart_timer = 1;
+		} else {
+			/* Anything interesting happen? */
+			hp->sw_bmsr = happy_meal_tcvr_read(hp, tregs, MII_BMSR);
+			if (hp->sw_bmsr & BMSR_ANEGCOMPLETE) {
+				int ret;
+
+				/* Just what we've been waiting for... */
+				ret = set_happy_link_modes(hp, tregs);
+				if (ret) {
+					/* Ooops, something bad happened, go to force
+					 * mode.
+					 *
+					 * XXX Broken hubs which don't support 802.3u
+					 * XXX auto-negotiation make this happen as well.
+					 */
+					goto do_force_mode;
+				}
+
+				/* Success, at least so far, advance our state engine. */
+				hp->timer_state = lupwait;
+				restart_timer = 1;
+			} else {
+				restart_timer = 1;
+			}
+		}
+		break;
+
+	case lupwait:
+		/* Auto negotiation was successful and we are awaiting a
+		 * link up status.  I have decided to let this timer run
+		 * forever until some sort of error is signalled, reporting
+		 * a message to the user at 10 second intervals.
+		 */
+		hp->sw_bmsr = happy_meal_tcvr_read(hp, tregs, MII_BMSR);
+		if (hp->sw_bmsr & BMSR_LSTATUS) {
+			/* Wheee, it's up, display the link mode in use and put
+			 * the timer to sleep.
+			 */
+			display_link_mode(hp, tregs);
+			hp->timer_state = asleep;
+			restart_timer = 0;
+		} else {
+			if (hp->timer_ticks >= 10) {
+				netdev_notice(hp->dev,
+					      "Auto negotiation successful, link still not completely up.\n");
+				hp->timer_ticks = 0;
+				restart_timer = 1;
+			} else {
+				restart_timer = 1;
+			}
+		}
+		break;
+
+	case ltrywait:
+		/* Making the timeout here too long can make it take
+		 * annoyingly long to attempt all of the link mode
+		 * permutations, but then again this is essentially
+		 * error recovery code for the most part.
+		 */
+		hp->sw_bmsr = happy_meal_tcvr_read(hp, tregs, MII_BMSR);
+		hp->sw_csconfig = happy_meal_tcvr_read(hp, tregs, DP83840_CSCONFIG);
+		if (hp->timer_ticks == 1) {
+			if (!is_lucent_phy(hp)) {
+				/* Re-enable transceiver, we'll re-enable the transceiver next
+				 * tick, then check link state on the following tick.
+				 */
+				hp->sw_csconfig |= CSCONFIG_TCVDISAB;
+				happy_meal_tcvr_write(hp, tregs,
+						      DP83840_CSCONFIG, hp->sw_csconfig);
+			}
+			restart_timer = 1;
+			break;
+		}
+		if (hp->timer_ticks == 2) {
+			if (!is_lucent_phy(hp)) {
+				hp->sw_csconfig &= ~(CSCONFIG_TCVDISAB);
+				happy_meal_tcvr_write(hp, tregs,
+						      DP83840_CSCONFIG, hp->sw_csconfig);
+			}
+			restart_timer = 1;
+			break;
+		}
+		if (hp->sw_bmsr & BMSR_LSTATUS) {
+			/* Force mode selection success. */
+			display_forced_link_mode(hp, tregs);
+			set_happy_link_modes(hp, tregs); /* XXX error? then what? */
+			hp->timer_state = asleep;
+			restart_timer = 0;
+		} else {
+			if (hp->timer_ticks >= 4) { /* 6 seconds or so... */
+				int ret;
+
+				ret = try_next_permutation(hp, tregs);
+				if (ret == -1) {
+					/* Aieee, tried them all, reset the
+					 * chip and try all over again.
+					 */
+
+					/* Let the user know... */
+					netdev_notice(hp->dev,
+						      "Link down, cable problem?\n");
+
+					happy_meal_begin_auto_negotiation(hp, tregs, NULL);
+					goto out;
+				}
+				if (!is_lucent_phy(hp)) {
+					hp->sw_csconfig = happy_meal_tcvr_read(hp, tregs,
+									       DP83840_CSCONFIG);
+					hp->sw_csconfig |= CSCONFIG_TCVDISAB;
+					happy_meal_tcvr_write(hp, tregs,
+							      DP83840_CSCONFIG, hp->sw_csconfig);
+				}
+				hp->timer_ticks = 0;
+				restart_timer = 1;
+			} else {
+				restart_timer = 1;
+			}
+		}
+		break;
+
+	case asleep:
+	default:
+		/* Can't happens.... */
+		netdev_err(hp->dev,
+			   "Aieee, link timer is asleep but we got one anyways!\n");
+		restart_timer = 0;
+		hp->timer_ticks = 0;
+		hp->timer_state = asleep; /* foo on you */
+		break;
+	}
+
+	if (restart_timer) {
+		hp->happy_timer.expires = jiffies + ((12 * HZ)/10); /* 1.2 sec. */
+		add_timer(&hp->happy_timer);
+	}
+
+out:
+	spin_unlock_irq(&hp->happy_lock);
+}
+
+#define TX_RESET_TRIES     32
+#define RX_RESET_TRIES     32
+
+/* hp->happy_lock must be held */
+static void happy_meal_tx_reset(struct happy_meal *hp, void __iomem *bregs)
+{
+	int tries = TX_RESET_TRIES;
+
+	HMD("reset...\n");
+
+	/* Would you like to try our SMCC Delux? */
+	hme_write32(hp, bregs + BMAC_TXSWRESET, 0);
+	while ((hme_read32(hp, bregs + BMAC_TXSWRESET) & 1) && --tries)
+		udelay(20);
+
+	/* Lettuce, tomato, buggy hardware (no extra charge)? */
+	if (!tries)
+		netdev_err(hp->dev, "Transceiver BigMac ATTACK!");
+
+	/* Take care. */
+	HMD("done\n");
+}
+
+/* hp->happy_lock must be held */
+static void happy_meal_rx_reset(struct happy_meal *hp, void __iomem *bregs)
+{
+	int tries = RX_RESET_TRIES;
+
+	HMD("reset...\n");
+
+	/* We have a special on GNU/Viking hardware bugs today. */
+	hme_write32(hp, bregs + BMAC_RXSWRESET, 0);
+	while ((hme_read32(hp, bregs + BMAC_RXSWRESET) & 1) && --tries)
+		udelay(20);
+
+	/* Will that be all? */
+	if (!tries)
+		netdev_err(hp->dev, "Receiver BigMac ATTACK!\n");
+
+	/* Don't forget your vik_1137125_wa.  Have a nice day. */
+	HMD("done\n");
+}
+
+#define STOP_TRIES         16
+
+/* hp->happy_lock must be held */
+static void happy_meal_stop(struct happy_meal *hp, void __iomem *gregs)
+{
+	int tries = STOP_TRIES;
+
+	HMD("reset...\n");
+
+	/* We're consolidating our STB products, it's your lucky day. */
+	hme_write32(hp, gregs + GREG_SWRESET, GREG_RESET_ALL);
+	while (hme_read32(hp, gregs + GREG_SWRESET) && --tries)
+		udelay(20);
+
+	/* Come back next week when we are "Sun Microelectronics". */
+	if (!tries)
+		netdev_err(hp->dev, "Fry guys.\n");
+
+	/* Remember: "Different name, same old buggy as shit hardware." */
+	HMD("done\n");
+}
+
+/* hp->happy_lock must be held */
+static void happy_meal_get_counters(struct happy_meal *hp, void __iomem *bregs)
+{
+	struct net_device_stats *stats = &hp->dev->stats;
+
+	stats->rx_crc_errors += hme_read32(hp, bregs + BMAC_RCRCECTR);
+	hme_write32(hp, bregs + BMAC_RCRCECTR, 0);
+
+	stats->rx_frame_errors += hme_read32(hp, bregs + BMAC_UNALECTR);
+	hme_write32(hp, bregs + BMAC_UNALECTR, 0);
+
+	stats->rx_length_errors += hme_read32(hp, bregs + BMAC_GLECTR);
+	hme_write32(hp, bregs + BMAC_GLECTR, 0);
+
+	stats->tx_aborted_errors += hme_read32(hp, bregs + BMAC_EXCTR);
+
+	stats->collisions +=
+		(hme_read32(hp, bregs + BMAC_EXCTR) +
+		 hme_read32(hp, bregs + BMAC_LTCTR));
+	hme_write32(hp, bregs + BMAC_EXCTR, 0);
+	hme_write32(hp, bregs + BMAC_LTCTR, 0);
+}
+
+/* Only Sun can take such nice parts and fuck up the programming interface
+ * like this.  Good job guys...
+ */
+#define TCVR_RESET_TRIES       16 /* It should reset quickly        */
+#define TCVR_UNISOLATE_TRIES   32 /* Dis-isolation can take longer. */
+
+/* hp->happy_lock must be held */
+static int happy_meal_tcvr_reset(struct happy_meal *hp, void __iomem *tregs)
+{
+	u32 tconfig;
+	int result, tries = TCVR_RESET_TRIES;
+
+	tconfig = hme_read32(hp, tregs + TCVR_CFG);
+	ASD("tcfg=%08x\n", tconfig);
+	if (hp->tcvr_type == external) {
+		hme_write32(hp, tregs + TCVR_CFG, tconfig & ~(TCV_CFG_PSELECT));
+		hp->tcvr_type = internal;
+		hp->paddr = TCV_PADDR_ITX;
+		happy_meal_tcvr_write(hp, tregs, MII_BMCR,
+				      (BMCR_LOOPBACK|BMCR_PDOWN|BMCR_ISOLATE));
+		result = happy_meal_tcvr_read(hp, tregs, MII_BMCR);
+		if (result == TCVR_FAILURE) {
+			ASD("phyread_fail\n");
+			return -1;
+		}
+		ASD("external: ISOLATE, phyread_ok, PSELECT\n");
+		hme_write32(hp, tregs + TCVR_CFG, tconfig | TCV_CFG_PSELECT);
+		hp->tcvr_type = external;
+		hp->paddr = TCV_PADDR_ETX;
+	} else {
+		if (tconfig & TCV_CFG_MDIO1) {
+			hme_write32(hp, tregs + TCVR_CFG, (tconfig | TCV_CFG_PSELECT));
+			happy_meal_tcvr_write(hp, tregs, MII_BMCR,
+					      (BMCR_LOOPBACK|BMCR_PDOWN|BMCR_ISOLATE));
+			result = happy_meal_tcvr_read(hp, tregs, MII_BMCR);
+			if (result == TCVR_FAILURE) {
+				ASD("phyread_fail>\n");
+				return -1;
+			}
+			ASD("internal: PSELECT, ISOLATE, phyread_ok, ~PSELECT\n");
+			hme_write32(hp, tregs + TCVR_CFG, (tconfig & ~(TCV_CFG_PSELECT)));
+			hp->tcvr_type = internal;
+			hp->paddr = TCV_PADDR_ITX;
+		}
+	}
+
+	ASD("BMCR_RESET...\n");
+	happy_meal_tcvr_write(hp, tregs, MII_BMCR, BMCR_RESET);
+
+	while (--tries) {
+		result = happy_meal_tcvr_read(hp, tregs, MII_BMCR);
+		if (result == TCVR_FAILURE)
+			return -1;
+		hp->sw_bmcr = result;
+		if (!(result & BMCR_RESET))
+			break;
+		udelay(20);
+	}
+	if (!tries) {
+		ASD("BMCR RESET FAILED!\n");
+		return -1;
+	}
+	ASD("RESET_OK\n");
+
+	/* Get fresh copies of the PHY registers. */
+	hp->sw_bmsr      = happy_meal_tcvr_read(hp, tregs, MII_BMSR);
+	hp->sw_physid1   = happy_meal_tcvr_read(hp, tregs, MII_PHYSID1);
+	hp->sw_physid2   = happy_meal_tcvr_read(hp, tregs, MII_PHYSID2);
+	hp->sw_advertise = happy_meal_tcvr_read(hp, tregs, MII_ADVERTISE);
+
+	ASD("UNISOLATE...\n");
+	hp->sw_bmcr &= ~(BMCR_ISOLATE);
+	happy_meal_tcvr_write(hp, tregs, MII_BMCR, hp->sw_bmcr);
+
+	tries = TCVR_UNISOLATE_TRIES;
+	while (--tries) {
+		result = happy_meal_tcvr_read(hp, tregs, MII_BMCR);
+		if (result == TCVR_FAILURE)
+			return -1;
+		if (!(result & BMCR_ISOLATE))
+			break;
+		udelay(20);
+	}
+	if (!tries) {
+		ASD("UNISOLATE FAILED!\n");
+		return -1;
+	}
+	ASD("SUCCESS and CSCONFIG_DFBYPASS\n");
+	if (!is_lucent_phy(hp)) {
+		result = happy_meal_tcvr_read(hp, tregs,
+					      DP83840_CSCONFIG);
+		happy_meal_tcvr_write(hp, tregs,
+				      DP83840_CSCONFIG, (result | CSCONFIG_DFBYPASS));
+	}
+	return 0;
+}
+
+/* Figure out whether we have an internal or external transceiver.
+ *
+ * hp->happy_lock must be held
+ */
+static void happy_meal_transceiver_check(struct happy_meal *hp, void __iomem *tregs)
+{
+	unsigned long tconfig = hme_read32(hp, tregs + TCVR_CFG);
+	u32 reread = hme_read32(hp, tregs + TCVR_CFG);
+
+	ASD("tcfg=%08lx\n", tconfig);
+	if (reread & TCV_CFG_MDIO1) {
+		hme_write32(hp, tregs + TCVR_CFG, tconfig | TCV_CFG_PSELECT);
+		hp->paddr = TCV_PADDR_ETX;
+		hp->tcvr_type = external;
+		ASD("not polling, external\n");
+	} else {
+		if (reread & TCV_CFG_MDIO0) {
+			hme_write32(hp, tregs + TCVR_CFG,
+				    tconfig & ~(TCV_CFG_PSELECT));
+			hp->paddr = TCV_PADDR_ITX;
+			hp->tcvr_type = internal;
+			ASD("not polling, internal\n");
+		} else {
+			netdev_err(hp->dev,
+				   "Transceiver and a coke please.");
+			hp->tcvr_type = none; /* Grrr... */
+			ASD("not polling, none\n");
+		}
+	}
+}
+
+/* The receive ring buffers are a bit tricky to get right.  Here goes...
+ *
+ * The buffers we dma into must be 64 byte aligned.  So we use a special
+ * alloc_skb() routine for the happy meal to allocate 64 bytes more than
+ * we really need.
+ *
+ * We use skb_reserve() to align the data block we get in the skb.  We
+ * also program the etxregs->cfg register to use an offset of 2.  This
+ * empirical constant plus the ethernet header size will always leave
+ * us with a nicely aligned ip header once we pass things up to the
+ * protocol layers.
+ *
+ * The numbers work out to:
+ *
+ *         Max ethernet frame size         1518
+ *         Ethernet header size              14
+ *         Happy Meal base offset             2
+ *
+ * Say a skb data area is at 0xf001b010, and its size alloced is
+ * (ETH_FRAME_LEN + 64 + 2) = (1514 + 64 + 2) = 1580 bytes.
+ *
+ * First our alloc_skb() routine aligns the data base to a 64 byte
+ * boundary.  We now have 0xf001b040 as our skb data address.  We
+ * plug this into the receive descriptor address.
+ *
+ * Next, we skb_reserve() 2 bytes to account for the Happy Meal offset.
+ * So now the data we will end up looking at starts at 0xf001b042.  When
+ * the packet arrives, we will check out the size received and subtract
+ * this from the skb->length.  Then we just pass the packet up to the
+ * protocols as is, and allocate a new skb to replace this slot we have
+ * just received from.
+ *
+ * The ethernet layer will strip the ether header from the front of the
+ * skb we just sent to it, this leaves us with the ip header sitting
+ * nicely aligned at 0xf001b050.  Also, for tcp and udp packets the
+ * Happy Meal has even checksummed the tcp/udp data for us.  The 16
+ * bit checksum is obtained from the low bits of the receive descriptor
+ * flags, thus:
+ *
+ * 	skb->csum = rxd->rx_flags & 0xffff;
+ * 	skb->ip_summed = CHECKSUM_COMPLETE;
+ *
+ * before sending off the skb to the protocols, and we are good as gold.
+ */
+static void happy_meal_clean_rings(struct happy_meal *hp)
+{
+	int i;
+
+	for (i = 0; i < RX_RING_SIZE; i++) {
+		if (hp->rx_skbs[i] != NULL) {
+			struct sk_buff *skb = hp->rx_skbs[i];
+			struct happy_meal_rxd *rxd;
+			u32 dma_addr;
+
+			rxd = &hp->happy_block->happy_meal_rxd[i];
+			dma_addr = hme_read_desc32(hp, &rxd->rx_addr);
+			dma_unmap_single(hp->dma_dev, dma_addr,
+					 RX_BUF_ALLOC_SIZE, DMA_FROM_DEVICE);
+			dev_kfree_skb_any(skb);
+			hp->rx_skbs[i] = NULL;
+		}
+	}
+
+	for (i = 0; i < TX_RING_SIZE; i++) {
+		if (hp->tx_skbs[i] != NULL) {
+			struct sk_buff *skb = hp->tx_skbs[i];
+			struct happy_meal_txd *txd;
+			u32 dma_addr;
+			int frag;
+
+			hp->tx_skbs[i] = NULL;
+
+			for (frag = 0; frag <= skb_shinfo(skb)->nr_frags; frag++) {
+				txd = &hp->happy_block->happy_meal_txd[i];
+				dma_addr = hme_read_desc32(hp, &txd->tx_addr);
+				if (!frag)
+					dma_unmap_single(hp->dma_dev, dma_addr,
+							 (hme_read_desc32(hp, &txd->tx_flags)
+							  & TXFLAG_SIZE),
+							 DMA_TO_DEVICE);
+				else
+					dma_unmap_page(hp->dma_dev, dma_addr,
+							 (hme_read_desc32(hp, &txd->tx_flags)
+							  & TXFLAG_SIZE),
+							 DMA_TO_DEVICE);
+
+				if (frag != skb_shinfo(skb)->nr_frags)
+					i++;
+			}
+
+			dev_kfree_skb_any(skb);
+		}
+	}
+}
+
+/* hp->happy_lock must be held */
+static void happy_meal_init_rings(struct happy_meal *hp)
+{
+	struct hmeal_init_block *hb = hp->happy_block;
+	int i;
+
+	HMD("counters to zero\n");
+	hp->rx_new = hp->rx_old = hp->tx_new = hp->tx_old = 0;
+
+	/* Free any skippy bufs left around in the rings. */
+	happy_meal_clean_rings(hp);
+
+	/* Now get new skippy bufs for the receive ring. */
+	HMD("init rxring\n");
+	for (i = 0; i < RX_RING_SIZE; i++) {
+		struct sk_buff *skb;
+		u32 mapping;
+
+		skb = happy_meal_alloc_skb(RX_BUF_ALLOC_SIZE, GFP_ATOMIC);
+		if (!skb) {
+			hme_write_rxd(hp, &hb->happy_meal_rxd[i], 0, 0);
+			continue;
+		}
+		hp->rx_skbs[i] = skb;
+
+		/* Because we reserve afterwards. */
+		skb_put(skb, (ETH_FRAME_LEN + RX_OFFSET + 4));
+		mapping = dma_map_single(hp->dma_dev, skb->data, RX_BUF_ALLOC_SIZE,
+					 DMA_FROM_DEVICE);
+		if (dma_mapping_error(hp->dma_dev, mapping)) {
+			dev_kfree_skb_any(skb);
+			hme_write_rxd(hp, &hb->happy_meal_rxd[i], 0, 0);
+			continue;
+		}
+		hme_write_rxd(hp, &hb->happy_meal_rxd[i],
+			      (RXFLAG_OWN | ((RX_BUF_ALLOC_SIZE - RX_OFFSET) << 16)),
+			      mapping);
+		skb_reserve(skb, RX_OFFSET);
+	}
+
+	HMD("init txring\n");
+	for (i = 0; i < TX_RING_SIZE; i++)
+		hme_write_txd(hp, &hb->happy_meal_txd[i], 0, 0);
+
+	HMD("done\n");
+}
+
+/* hp->happy_lock must be held */
+static int happy_meal_init(struct happy_meal *hp)
+{
+	const unsigned char *e = &hp->dev->dev_addr[0];
+	void __iomem *gregs        = hp->gregs;
+	void __iomem *etxregs      = hp->etxregs;
+	void __iomem *erxregs      = hp->erxregs;
+	void __iomem *bregs        = hp->bigmacregs;
+	void __iomem *tregs        = hp->tcvregs;
+	const char *bursts = "64";
+	u32 regtmp, rxcfg;
+
+	/* If auto-negotiation timer is running, kill it. */
+	timer_delete(&hp->happy_timer);
+
+	HMD("happy_flags[%08x]\n", hp->happy_flags);
+	if (!(hp->happy_flags & HFLAG_INIT)) {
+		HMD("set HFLAG_INIT\n");
+		hp->happy_flags |= HFLAG_INIT;
+		happy_meal_get_counters(hp, bregs);
+	}
+
+	/* Stop transmitter and receiver. */
+	HMD("to happy_meal_stop\n");
+	happy_meal_stop(hp, gregs);
+
+	/* Alloc and reset the tx/rx descriptor chains. */
+	HMD("to happy_meal_init_rings\n");
+	happy_meal_init_rings(hp);
+
+	/* See if we can enable the MIF frame on this card to speak to the DP83840. */
+	if (hp->happy_flags & HFLAG_FENABLE) {
+		HMD("use frame old[%08x]\n",
+		    hme_read32(hp, tregs + TCVR_CFG));
+		hme_write32(hp, tregs + TCVR_CFG,
+			    hme_read32(hp, tregs + TCVR_CFG) & ~(TCV_CFG_BENABLE));
+	} else {
+		HMD("use bitbang old[%08x]\n",
+		    hme_read32(hp, tregs + TCVR_CFG));
+		hme_write32(hp, tregs + TCVR_CFG,
+			    hme_read32(hp, tregs + TCVR_CFG) | TCV_CFG_BENABLE);
+	}
+
+	/* Check the state of the transceiver. */
+	HMD("to happy_meal_transceiver_check\n");
+	happy_meal_transceiver_check(hp, tregs);
+
+	/* Put the Big Mac into a sane state. */
+	switch(hp->tcvr_type) {
+	case none:
+		/* Cannot operate if we don't know the transceiver type! */
+		HMD("AAIEEE no transceiver type, EAGAIN\n");
+		return -EAGAIN;
+
+	case internal:
+		/* Using the MII buffers. */
+		HMD("internal, using MII\n");
+		hme_write32(hp, bregs + BMAC_XIFCFG, 0);
+		break;
+
+	case external:
+		/* Not using the MII, disable it. */
+		HMD("external, disable MII\n");
+		hme_write32(hp, bregs + BMAC_XIFCFG, BIGMAC_XCFG_MIIDISAB);
+		break;
+	}
+
+	if (happy_meal_tcvr_reset(hp, tregs))
+		return -EAGAIN;
+
+	/* Reset the Happy Meal Big Mac transceiver and the receiver. */
+	HMD("tx/rx reset\n");
+	happy_meal_tx_reset(hp, bregs);
+	happy_meal_rx_reset(hp, bregs);
+
+	/* Set jam size and inter-packet gaps to reasonable defaults. */
+	hme_write32(hp, bregs + BMAC_JSIZE, DEFAULT_JAMSIZE);
+	hme_write32(hp, bregs + BMAC_IGAP1, DEFAULT_IPG1);
+	hme_write32(hp, bregs + BMAC_IGAP2, DEFAULT_IPG2);
+
+	/* Load up the MAC address and random seed. */
+
+	/* The docs recommend to use the 10LSB of our MAC here. */
+	hme_write32(hp, bregs + BMAC_RSEED, ((e[5] | e[4]<<8)&0x3ff));
+
+	hme_write32(hp, bregs + BMAC_MACADDR2, ((e[4] << 8) | e[5]));
+	hme_write32(hp, bregs + BMAC_MACADDR1, ((e[2] << 8) | e[3]));
+	hme_write32(hp, bregs + BMAC_MACADDR0, ((e[0] << 8) | e[1]));
+
+	if ((hp->dev->flags & IFF_ALLMULTI) ||
+	    (netdev_mc_count(hp->dev) > 64)) {
+		hme_write32(hp, bregs + BMAC_HTABLE0, 0xffff);
+		hme_write32(hp, bregs + BMAC_HTABLE1, 0xffff);
+		hme_write32(hp, bregs + BMAC_HTABLE2, 0xffff);
+		hme_write32(hp, bregs + BMAC_HTABLE3, 0xffff);
+	} else if ((hp->dev->flags & IFF_PROMISC) == 0) {
+		u16 hash_table[4];
+		struct netdev_hw_addr *ha;
+		u32 crc;
+
+		memset(hash_table, 0, sizeof(hash_table));
+		netdev_for_each_mc_addr(ha, hp->dev) {
+			crc = ether_crc_le(6, ha->addr);
+			crc >>= 26;
+			hash_table[crc >> 4] |= 1 << (crc & 0xf);
+		}
+		hme_write32(hp, bregs + BMAC_HTABLE0, hash_table[0]);
+		hme_write32(hp, bregs + BMAC_HTABLE1, hash_table[1]);
+		hme_write32(hp, bregs + BMAC_HTABLE2, hash_table[2]);
+		hme_write32(hp, bregs + BMAC_HTABLE3, hash_table[3]);
+	} else {
+		hme_write32(hp, bregs + BMAC_HTABLE3, 0);
+		hme_write32(hp, bregs + BMAC_HTABLE2, 0);
+		hme_write32(hp, bregs + BMAC_HTABLE1, 0);
+		hme_write32(hp, bregs + BMAC_HTABLE0, 0);
+	}
+
+	/* Set the RX and TX ring ptrs. */
+	HMD("ring ptrs rxr[%08x] txr[%08x]\n",
+	    ((__u32)hp->hblock_dvma + hblock_offset(happy_meal_rxd, 0)),
+	    ((__u32)hp->hblock_dvma + hblock_offset(happy_meal_txd, 0)));
+	hme_write32(hp, erxregs + ERX_RING,
+		    ((__u32)hp->hblock_dvma + hblock_offset(happy_meal_rxd, 0)));
+	hme_write32(hp, etxregs + ETX_RING,
+		    ((__u32)hp->hblock_dvma + hblock_offset(happy_meal_txd, 0)));
+
+	/* Parity issues in the ERX unit of some HME revisions can cause some
+	 * registers to not be written unless their parity is even.  Detect such
+	 * lost writes and simply rewrite with a low bit set (which will be ignored
+	 * since the rxring needs to be 2K aligned).
+	 */
+	if (hme_read32(hp, erxregs + ERX_RING) !=
+	    ((__u32)hp->hblock_dvma + hblock_offset(happy_meal_rxd, 0)))
+		hme_write32(hp, erxregs + ERX_RING,
+			    ((__u32)hp->hblock_dvma + hblock_offset(happy_meal_rxd, 0))
+			    | 0x4);
+
+	/* Set the supported burst sizes. */
+#ifndef CONFIG_SPARC
+	/* It is always PCI and can handle 64byte bursts. */
+	hme_write32(hp, gregs + GREG_CFG, GREG_CFG_BURST64);
+#else
+	if ((hp->happy_bursts & DMA_BURST64) &&
+	    ((hp->happy_flags & HFLAG_PCI) != 0
+#ifdef CONFIG_SBUS
+	     || sbus_can_burst64()
+#endif
+	     || 0)) {
+		u32 gcfg = GREG_CFG_BURST64;
+
+		/* I have no idea if I should set the extended
+		 * transfer mode bit for Cheerio, so for now I
+		 * do not.  -DaveM
+		 */
+#ifdef CONFIG_SBUS
+		if ((hp->happy_flags & HFLAG_PCI) == 0) {
+			struct platform_device *op = hp->happy_dev;
+			if (sbus_can_dma_64bit()) {
+				sbus_set_sbus64(&op->dev,
+						hp->happy_bursts);
+				gcfg |= GREG_CFG_64BIT;
+			}
+		}
+#endif
+
+		bursts = "64";
+		hme_write32(hp, gregs + GREG_CFG, gcfg);
+	} else if (hp->happy_bursts & DMA_BURST32) {
+		bursts = "32";
+		hme_write32(hp, gregs + GREG_CFG, GREG_CFG_BURST32);
+	} else if (hp->happy_bursts & DMA_BURST16) {
+		bursts = "16";
+		hme_write32(hp, gregs + GREG_CFG, GREG_CFG_BURST16);
+	} else {
+		bursts = "XXX";
+		hme_write32(hp, gregs + GREG_CFG, 0);
+	}
+#endif /* CONFIG_SPARC */
+
+	HMD("old[%08x] bursts<%s>\n",
+	    hme_read32(hp, gregs + GREG_CFG), bursts);
+
+	/* Turn off interrupts we do not want to hear. */
+	hme_write32(hp, gregs + GREG_IMASK,
+		    (GREG_IMASK_GOTFRAME | GREG_IMASK_RCNTEXP |
+		     GREG_IMASK_SENTFRAME | GREG_IMASK_TXPERR));
+
+	/* Set the transmit ring buffer size. */
+	HMD("tx rsize=%d oreg[%08x]\n", (int)TX_RING_SIZE,
+	    hme_read32(hp, etxregs + ETX_RSIZE));
+	hme_write32(hp, etxregs + ETX_RSIZE, (TX_RING_SIZE >> ETX_RSIZE_SHIFT) - 1);
+
+	/* Enable transmitter DVMA. */
+	HMD("tx dma enable old[%08x]\n", hme_read32(hp, etxregs + ETX_CFG));
+	hme_write32(hp, etxregs + ETX_CFG,
+		    hme_read32(hp, etxregs + ETX_CFG) | ETX_CFG_DMAENABLE);
+
+	/* This chip really rots, for the receiver sometimes when you
+	 * write to its control registers not all the bits get there
+	 * properly.  I cannot think of a sane way to provide complete
+	 * coverage for this hardware bug yet.
+	 */
+	HMD("erx regs bug old[%08x]\n",
+	    hme_read32(hp, erxregs + ERX_CFG));
+	hme_write32(hp, erxregs + ERX_CFG, ERX_CFG_DEFAULT(RX_OFFSET));
+	regtmp = hme_read32(hp, erxregs + ERX_CFG);
+	hme_write32(hp, erxregs + ERX_CFG, ERX_CFG_DEFAULT(RX_OFFSET));
+	if (hme_read32(hp, erxregs + ERX_CFG) != ERX_CFG_DEFAULT(RX_OFFSET)) {
+		netdev_err(hp->dev,
+			   "Eieee, rx config register gets greasy fries.\n");
+		netdev_err(hp->dev,
+			   "Trying to set %08x, reread gives %08x\n",
+			   ERX_CFG_DEFAULT(RX_OFFSET), regtmp);
+		/* XXX Should return failure here... */
+	}
+
+	/* Enable Big Mac hash table filter. */
+	HMD("enable hash rx_cfg_old[%08x]\n",
+	    hme_read32(hp, bregs + BMAC_RXCFG));
+	rxcfg = BIGMAC_RXCFG_HENABLE | BIGMAC_RXCFG_REJME;
+	if (hp->dev->flags & IFF_PROMISC)
+		rxcfg |= BIGMAC_RXCFG_PMISC;
+	hme_write32(hp, bregs + BMAC_RXCFG, rxcfg);
+
+	/* Let the bits settle in the chip. */
+	udelay(10);
+
+	/* Ok, configure the Big Mac transmitter. */
+	HMD("BIGMAC init\n");
+	regtmp = 0;
+	if (hp->happy_flags & HFLAG_FULL)
+		regtmp |= BIGMAC_TXCFG_FULLDPLX;
+
+	/* Don't turn on the "don't give up" bit for now.  It could cause hme
+	 * to deadlock with the PHY if a Jabber occurs.
+	 */
+	hme_write32(hp, bregs + BMAC_TXCFG, regtmp /*| BIGMAC_TXCFG_DGIVEUP*/);
+
+	/* Give up after 16 TX attempts. */
+	hme_write32(hp, bregs + BMAC_ALIMIT, 16);
+
+	/* Enable the output drivers no matter what. */
+	regtmp = BIGMAC_XCFG_ODENABLE;
+
+	/* If card can do lance mode, enable it. */
+	if (hp->happy_flags & HFLAG_LANCE)
+		regtmp |= (DEFAULT_IPG0 << 5) | BIGMAC_XCFG_LANCE;
+
+	/* Disable the MII buffers if using external transceiver. */
+	if (hp->tcvr_type == external)
+		regtmp |= BIGMAC_XCFG_MIIDISAB;
+
+	HMD("XIF config old[%08x]\n", hme_read32(hp, bregs + BMAC_XIFCFG));
+	hme_write32(hp, bregs + BMAC_XIFCFG, regtmp);
+
+	/* Start things up. */
+	HMD("tx old[%08x] and rx [%08x] ON!\n",
+	    hme_read32(hp, bregs + BMAC_TXCFG),
+	    hme_read32(hp, bregs + BMAC_RXCFG));
+
+	/* Set larger TX/RX size to allow for 802.1q */
+	hme_write32(hp, bregs + BMAC_TXMAX, ETH_FRAME_LEN + 8);
+	hme_write32(hp, bregs + BMAC_RXMAX, ETH_FRAME_LEN + 8);
+
+	hme_write32(hp, bregs + BMAC_TXCFG,
+		    hme_read32(hp, bregs + BMAC_TXCFG) | BIGMAC_TXCFG_ENABLE);
+	hme_write32(hp, bregs + BMAC_RXCFG,
+		    hme_read32(hp, bregs + BMAC_RXCFG) | BIGMAC_RXCFG_ENABLE);
+
+	/* Get the autonegotiation started, and the watch timer ticking. */
+	happy_meal_begin_auto_negotiation(hp, tregs, NULL);
+
+	/* Success. */
+	return 0;
+}
+
+/* hp->happy_lock must be held */
+static void happy_meal_set_initial_advertisement(struct happy_meal *hp)
+{
+	void __iomem *tregs	= hp->tcvregs;
+	void __iomem *bregs	= hp->bigmacregs;
+	void __iomem *gregs	= hp->gregs;
+
+	happy_meal_stop(hp, gregs);
+	if (hp->happy_flags & HFLAG_FENABLE)
+		hme_write32(hp, tregs + TCVR_CFG,
+			    hme_read32(hp, tregs + TCVR_CFG) & ~(TCV_CFG_BENABLE));
+	else
+		hme_write32(hp, tregs + TCVR_CFG,
+			    hme_read32(hp, tregs + TCVR_CFG) | TCV_CFG_BENABLE);
+	happy_meal_transceiver_check(hp, tregs);
+	switch(hp->tcvr_type) {
+	case none:
+		return;
+	case internal:
+		hme_write32(hp, bregs + BMAC_XIFCFG, 0);
+		break;
+	case external:
+		hme_write32(hp, bregs + BMAC_XIFCFG, BIGMAC_XCFG_MIIDISAB);
+		break;
+	}
+	if (happy_meal_tcvr_reset(hp, tregs))
+		return;
+
+	/* Latch PHY registers as of now. */
+	hp->sw_bmsr      = happy_meal_tcvr_read(hp, tregs, MII_BMSR);
+	hp->sw_advertise = happy_meal_tcvr_read(hp, tregs, MII_ADVERTISE);
+
+	/* Advertise everything we can support. */
+	if (hp->sw_bmsr & BMSR_10HALF)
+		hp->sw_advertise |= (ADVERTISE_10HALF);
+	else
+		hp->sw_advertise &= ~(ADVERTISE_10HALF);
+
+	if (hp->sw_bmsr & BMSR_10FULL)
+		hp->sw_advertise |= (ADVERTISE_10FULL);
+	else
+		hp->sw_advertise &= ~(ADVERTISE_10FULL);
+	if (hp->sw_bmsr & BMSR_100HALF)
+		hp->sw_advertise |= (ADVERTISE_100HALF);
+	else
+		hp->sw_advertise &= ~(ADVERTISE_100HALF);
+	if (hp->sw_bmsr & BMSR_100FULL)
+		hp->sw_advertise |= (ADVERTISE_100FULL);
+	else
+		hp->sw_advertise &= ~(ADVERTISE_100FULL);
+
+	/* Update the PHY advertisement register. */
+	happy_meal_tcvr_write(hp, tregs, MII_ADVERTISE, hp->sw_advertise);
+}
+
+/* Once status is latched (by happy_meal_interrupt) it is cleared by
+ * the hardware, so we cannot re-read it and get a correct value.
+ *
+ * hp->happy_lock must be held
+ */
+static int happy_meal_is_not_so_happy(struct happy_meal *hp, u32 status)
+{
+	int reset = 0;
+
+	/* Only print messages for non-counter related interrupts. */
+	if (status & (GREG_STAT_STSTERR | GREG_STAT_TFIFO_UND |
+		      GREG_STAT_MAXPKTERR | GREG_STAT_RXERR |
+		      GREG_STAT_RXPERR | GREG_STAT_RXTERR | GREG_STAT_EOPERR |
+		      GREG_STAT_MIFIRQ | GREG_STAT_TXEACK | GREG_STAT_TXLERR |
+		      GREG_STAT_TXPERR | GREG_STAT_TXTERR | GREG_STAT_SLVERR |
+		      GREG_STAT_SLVPERR))
+		netdev_err(hp->dev,
+			   "Error interrupt for happy meal, status = %08x\n",
+			   status);
+
+	if (status & GREG_STAT_RFIFOVF) {
+		/* Receive FIFO overflow is harmless and the hardware will take
+		   care of it, just some packets are lost. Who cares. */
+		netdev_dbg(hp->dev, "Happy Meal receive FIFO overflow.\n");
+	}
+
+	if (status & GREG_STAT_STSTERR) {
+		/* BigMAC SQE link test failed. */
+		netdev_err(hp->dev, "Happy Meal BigMAC SQE test failed.\n");
+		reset = 1;
+	}
+
+	if (status & GREG_STAT_TFIFO_UND) {
+		/* Transmit FIFO underrun, again DMA error likely. */
+		netdev_err(hp->dev,
+			   "Happy Meal transmitter FIFO underrun, DMA error.\n");
+		reset = 1;
+	}
+
+	if (status & GREG_STAT_MAXPKTERR) {
+		/* Driver error, tried to transmit something larger
+		 * than ethernet max mtu.
+		 */
+		netdev_err(hp->dev, "Happy Meal MAX Packet size error.\n");
+		reset = 1;
+	}
+
+	if (status & GREG_STAT_NORXD) {
+		/* This is harmless, it just means the system is
+		 * quite loaded and the incoming packet rate was
+		 * faster than the interrupt handler could keep up
+		 * with.
+		 */
+		netdev_info(hp->dev,
+			    "Happy Meal out of receive descriptors, packet dropped.\n");
+	}
+
+	if (status & (GREG_STAT_RXERR|GREG_STAT_RXPERR|GREG_STAT_RXTERR)) {
+		/* All sorts of DMA receive errors. */
+		netdev_err(hp->dev, "Happy Meal rx DMA errors [ %s%s%s]\n",
+			   status & GREG_STAT_RXERR ? "GenericError " : "",
+			   status & GREG_STAT_RXPERR ? "ParityError " : "",
+			   status & GREG_STAT_RXTERR ? "RxTagBotch " : "");
+		reset = 1;
+	}
+
+	if (status & GREG_STAT_EOPERR) {
+		/* Driver bug, didn't set EOP bit in tx descriptor given
+		 * to the happy meal.
+		 */
+		netdev_err(hp->dev,
+			   "EOP not set in happy meal transmit descriptor!\n");
+		reset = 1;
+	}
+
+	if (status & GREG_STAT_MIFIRQ) {
+		/* MIF signalled an interrupt, were we polling it? */
+		netdev_err(hp->dev, "Happy Meal MIF interrupt.\n");
+	}
+
+	if (status &
+	    (GREG_STAT_TXEACK|GREG_STAT_TXLERR|GREG_STAT_TXPERR|GREG_STAT_TXTERR)) {
+		/* All sorts of transmit DMA errors. */
+		netdev_err(hp->dev, "Happy Meal tx DMA errors [ %s%s%s%s]\n",
+			   status & GREG_STAT_TXEACK ? "GenericError " : "",
+			   status & GREG_STAT_TXLERR ? "LateError " : "",
+			   status & GREG_STAT_TXPERR ? "ParityError " : "",
+			   status & GREG_STAT_TXTERR ? "TagBotch " : "");
+		reset = 1;
+	}
+
+	if (status & (GREG_STAT_SLVERR|GREG_STAT_SLVPERR)) {
+		/* Bus or parity error when cpu accessed happy meal registers
+		 * or it's internal FIFO's.  Should never see this.
+		 */
+		netdev_err(hp->dev,
+			   "Happy Meal register access SBUS slave (%s) error.\n",
+			   (status & GREG_STAT_SLVPERR) ? "parity" : "generic");
+		reset = 1;
+	}
+
+	if (reset) {
+		netdev_notice(hp->dev, "Resetting...\n");
+		happy_meal_init(hp);
+		return 1;
+	}
+	return 0;
+}
+
+/* hp->happy_lock must be held */
+static void happy_meal_tx(struct happy_meal *hp)
+{
+	struct happy_meal_txd *txbase = &hp->happy_block->happy_meal_txd[0];
+	struct happy_meal_txd *this;
+	struct net_device *dev = hp->dev;
+	int elem;
+
+	elem = hp->tx_old;
+	while (elem != hp->tx_new) {
+		struct sk_buff *skb;
+		u32 flags, dma_addr, dma_len;
+		int frag;
+
+		netdev_vdbg(hp->dev, "TX[%d]\n", elem);
+		this = &txbase[elem];
+		flags = hme_read_desc32(hp, &this->tx_flags);
+		if (flags & TXFLAG_OWN)
+			break;
+		skb = hp->tx_skbs[elem];
+		if (skb_shinfo(skb)->nr_frags) {
+			int last;
+
+			last = elem + skb_shinfo(skb)->nr_frags;
+			last &= (TX_RING_SIZE - 1);
+			flags = hme_read_desc32(hp, &txbase[last].tx_flags);
+			if (flags & TXFLAG_OWN)
+				break;
+		}
+		hp->tx_skbs[elem] = NULL;
+		dev->stats.tx_bytes += skb->len;
+
+		for (frag = 0; frag <= skb_shinfo(skb)->nr_frags; frag++) {
+			dma_addr = hme_read_desc32(hp, &this->tx_addr);
+			dma_len = hme_read_desc32(hp, &this->tx_flags);
+
+			dma_len &= TXFLAG_SIZE;
+			if (!frag)
+				dma_unmap_single(hp->dma_dev, dma_addr, dma_len, DMA_TO_DEVICE);
+			else
+				dma_unmap_page(hp->dma_dev, dma_addr, dma_len, DMA_TO_DEVICE);
+
+			elem = NEXT_TX(elem);
+			this = &txbase[elem];
+		}
+
+		dev_consume_skb_irq(skb);
+		dev->stats.tx_packets++;
+	}
+	hp->tx_old = elem;
+
+	if (netif_queue_stopped(dev) &&
+	    TX_BUFFS_AVAIL(hp) > (MAX_SKB_FRAGS + 1))
+		netif_wake_queue(dev);
+}
+
+/* Originally I used to handle the allocation failure by just giving back just
+ * that one ring buffer to the happy meal.  Problem is that usually when that
+ * condition is triggered, the happy meal expects you to do something reasonable
+ * with all of the packets it has DMA'd in.  So now I just drop the entire
+ * ring when we cannot get a new skb and give them all back to the happy meal,
+ * maybe things will be "happier" now.
+ *
+ * hp->happy_lock must be held
+ */
+static void happy_meal_rx(struct happy_meal *hp, struct net_device *dev)
+{
+	struct happy_meal_rxd *rxbase = &hp->happy_block->happy_meal_rxd[0];
+	struct happy_meal_rxd *this;
+	int elem = hp->rx_new, drops = 0;
+	u32 flags;
+
+	this = &rxbase[elem];
+	while (!((flags = hme_read_desc32(hp, &this->rx_flags)) & RXFLAG_OWN)) {
+		struct sk_buff *skb;
+		int len = flags >> 16;
+		u16 csum = flags & RXFLAG_CSUM;
+		u32 dma_addr = hme_read_desc32(hp, &this->rx_addr);
+
+		/* Check for errors. */
+		if ((len < ETH_ZLEN) || (flags & RXFLAG_OVERFLOW)) {
+			netdev_vdbg(dev, "RX[%d ERR(%08x)]", elem, flags);
+			dev->stats.rx_errors++;
+			if (len < ETH_ZLEN)
+				dev->stats.rx_length_errors++;
+			if (len & (RXFLAG_OVERFLOW >> 16)) {
+				dev->stats.rx_over_errors++;
+				dev->stats.rx_fifo_errors++;
+			}
+
+			/* Return it to the Happy meal. */
+	drop_it:
+			dev->stats.rx_dropped++;
+			hme_write_rxd(hp, this,
+				      (RXFLAG_OWN|((RX_BUF_ALLOC_SIZE-RX_OFFSET)<<16)),
+				      dma_addr);
+			goto next;
+		}
+		skb = hp->rx_skbs[elem];
+		if (len > RX_COPY_THRESHOLD) {
+			struct sk_buff *new_skb;
+			u32 mapping;
+
+			/* Now refill the entry, if we can. */
+			new_skb = happy_meal_alloc_skb(RX_BUF_ALLOC_SIZE, GFP_ATOMIC);
+			if (new_skb == NULL) {
+				drops++;
+				goto drop_it;
+			}
+			skb_put(new_skb, (ETH_FRAME_LEN + RX_OFFSET + 4));
+			mapping = dma_map_single(hp->dma_dev, new_skb->data,
+						 RX_BUF_ALLOC_SIZE,
+						 DMA_FROM_DEVICE);
+			if (unlikely(dma_mapping_error(hp->dma_dev, mapping))) {
+				dev_kfree_skb_any(new_skb);
+				drops++;
+				goto drop_it;
+			}
+
+			dma_unmap_single(hp->dma_dev, dma_addr, RX_BUF_ALLOC_SIZE, DMA_FROM_DEVICE);
+			hp->rx_skbs[elem] = new_skb;
+			hme_write_rxd(hp, this,
+				      (RXFLAG_OWN|((RX_BUF_ALLOC_SIZE-RX_OFFSET)<<16)),
+				      mapping);
+			skb_reserve(new_skb, RX_OFFSET);
+
+			/* Trim the original skb for the netif. */
+			skb_trim(skb, len);
+		} else {
+			struct sk_buff *copy_skb = netdev_alloc_skb(dev, len + 2);
+
+			if (copy_skb == NULL) {
+				drops++;
+				goto drop_it;
+			}
+
+			skb_reserve(copy_skb, 2);
+			skb_put(copy_skb, len);
+			dma_sync_single_for_cpu(hp->dma_dev, dma_addr, len + 2, DMA_FROM_DEVICE);
+			skb_copy_from_linear_data(skb, copy_skb->data, len);
+			dma_sync_single_for_device(hp->dma_dev, dma_addr, len + 2, DMA_FROM_DEVICE);
+			/* Reuse original ring buffer. */
+			hme_write_rxd(hp, this,
+				      (RXFLAG_OWN|((RX_BUF_ALLOC_SIZE-RX_OFFSET)<<16)),
+				      dma_addr);
+
+			skb = copy_skb;
+		}
+
+		/* This card is _fucking_ hot... */
+		skb->csum = csum_unfold(~(__force __sum16)htons(csum));
+		skb->ip_summed = CHECKSUM_COMPLETE;
+
+		netdev_vdbg(dev, "RX[%d len=%d csum=%4x]", elem, len, csum);
+		skb->protocol = eth_type_trans(skb, dev);
+		netif_rx(skb);
+
+		dev->stats.rx_packets++;
+		dev->stats.rx_bytes += len;
+	next:
+		elem = NEXT_RX(elem);
+		this = &rxbase[elem];
+	}
+	hp->rx_new = elem;
+	if (drops)
+		netdev_info(hp->dev, "Memory squeeze, deferring packet.\n");
+}
+
+static irqreturn_t happy_meal_interrupt(int irq, void *dev_id)
+{
+	struct net_device *dev = dev_id;
+	struct happy_meal *hp  = netdev_priv(dev);
+	u32 happy_status       = hme_read32(hp, hp->gregs + GREG_STAT);
+
+	HMD("status=%08x\n", happy_status);
+	if (!happy_status)
+		return IRQ_NONE;
+
+	spin_lock(&hp->happy_lock);
+
+	if (happy_status & GREG_STAT_ERRORS) {
+		if (happy_meal_is_not_so_happy(hp, /* un- */ happy_status))
+			goto out;
+	}
+
+	if (happy_status & GREG_STAT_TXALL)
+		happy_meal_tx(hp);
+
+	if (happy_status & GREG_STAT_RXTOHOST)
+		happy_meal_rx(hp, dev);
+
+	HMD("done\n");
+out:
+	spin_unlock(&hp->happy_lock);
+
+	return IRQ_HANDLED;
+}
+
+static int happy_meal_open(struct net_device *dev)
+{
+	struct happy_meal *hp = netdev_priv(dev);
+	int res;
+
+	res = request_irq(hp->irq, happy_meal_interrupt, IRQF_SHARED,
+			  dev->name, dev);
+	if (res) {
+		netdev_err(dev, "Can't order irq %d to go.\n", hp->irq);
+		return res;
+	}
+
+	HMD("to happy_meal_init\n");
+
+	spin_lock_irq(&hp->happy_lock);
+	res = happy_meal_init(hp);
+	spin_unlock_irq(&hp->happy_lock);
+
+	if (res)
+		free_irq(hp->irq, dev);
+	return res;
+}
+
+static int happy_meal_close(struct net_device *dev)
+{
+	struct happy_meal *hp = netdev_priv(dev);
+
+	spin_lock_irq(&hp->happy_lock);
+	happy_meal_stop(hp, hp->gregs);
+	happy_meal_clean_rings(hp);
+
+	/* If auto-negotiation timer is running, kill it. */
+	timer_delete(&hp->happy_timer);
+
+	spin_unlock_irq(&hp->happy_lock);
+
+	free_irq(hp->irq, dev);
+
+	return 0;
+}
+
+static void happy_meal_tx_timeout(struct net_device *dev, unsigned int txqueue)
+{
+	struct happy_meal *hp = netdev_priv(dev);
+
+	netdev_err(dev, "transmit timed out, resetting\n");
+	tx_dump_log();
+	netdev_err(dev, "Happy Status %08x TX[%08x:%08x]\n",
+		   hme_read32(hp, hp->gregs + GREG_STAT),
+		   hme_read32(hp, hp->etxregs + ETX_CFG),
+		   hme_read32(hp, hp->bigmacregs + BMAC_TXCFG));
+
+	spin_lock_irq(&hp->happy_lock);
+	happy_meal_init(hp);
+	spin_unlock_irq(&hp->happy_lock);
+
+	netif_wake_queue(dev);
+}
+
+static void unmap_partial_tx_skb(struct happy_meal *hp, u32 first_mapping,
+				 u32 first_len, u32 first_entry, u32 entry)
+{
+	struct happy_meal_txd *txbase = &hp->happy_block->happy_meal_txd[0];
+
+	dma_unmap_single(hp->dma_dev, first_mapping, first_len, DMA_TO_DEVICE);
+
+	first_entry = NEXT_TX(first_entry);
+	while (first_entry != entry) {
+		struct happy_meal_txd *this = &txbase[first_entry];
+		u32 addr, len;
+
+		addr = hme_read_desc32(hp, &this->tx_addr);
+		len = hme_read_desc32(hp, &this->tx_flags);
+		len &= TXFLAG_SIZE;
+		dma_unmap_page(hp->dma_dev, addr, len, DMA_TO_DEVICE);
+	}
+}
+
+static netdev_tx_t happy_meal_start_xmit(struct sk_buff *skb,
+					 struct net_device *dev)
+{
+	struct happy_meal *hp = netdev_priv(dev);
+	int entry;
+	u32 tx_flags;
+
+	tx_flags = TXFLAG_OWN;
+	if (skb->ip_summed == CHECKSUM_PARTIAL) {
+		const u32 csum_start_off = skb_checksum_start_offset(skb);
+		const u32 csum_stuff_off = csum_start_off + skb->csum_offset;
+
+		tx_flags = (TXFLAG_OWN | TXFLAG_CSENABLE |
+			    ((csum_start_off << 14) & TXFLAG_CSBUFBEGIN) |
+			    ((csum_stuff_off << 20) & TXFLAG_CSLOCATION));
+	}
+
+	spin_lock_irq(&hp->happy_lock);
+
+	if (TX_BUFFS_AVAIL(hp) <= (skb_shinfo(skb)->nr_frags + 1)) {
+		netif_stop_queue(dev);
+		spin_unlock_irq(&hp->happy_lock);
+		netdev_err(dev, "BUG! Tx Ring full when queue awake!\n");
+		return NETDEV_TX_BUSY;
+	}
+
+	entry = hp->tx_new;
+	netdev_vdbg(dev, "SX<l[%d]e[%d]>\n", skb->len, entry);
+	hp->tx_skbs[entry] = skb;
+
+	if (skb_shinfo(skb)->nr_frags == 0) {
+		u32 mapping, len;
+
+		len = skb->len;
+		mapping = dma_map_single(hp->dma_dev, skb->data, len, DMA_TO_DEVICE);
+		if (unlikely(dma_mapping_error(hp->dma_dev, mapping)))
+			goto out_dma_error;
+		tx_flags |= (TXFLAG_SOP | TXFLAG_EOP);
+		hme_write_txd(hp, &hp->happy_block->happy_meal_txd[entry],
+			      (tx_flags | (len & TXFLAG_SIZE)),
+			      mapping);
+		entry = NEXT_TX(entry);
+	} else {
+		u32 first_len, first_mapping;
+		int frag, first_entry = entry;
+
+		/* We must give this initial chunk to the device last.
+		 * Otherwise we could race with the device.
+		 */
+		first_len = skb_headlen(skb);
+		first_mapping = dma_map_single(hp->dma_dev, skb->data, first_len,
+					       DMA_TO_DEVICE);
+		if (unlikely(dma_mapping_error(hp->dma_dev, first_mapping)))
+			goto out_dma_error;
+		entry = NEXT_TX(entry);
+
+		for (frag = 0; frag < skb_shinfo(skb)->nr_frags; frag++) {
+			const skb_frag_t *this_frag = &skb_shinfo(skb)->frags[frag];
+			u32 len, mapping, this_txflags;
+
+			len = skb_frag_size(this_frag);
+			mapping = skb_frag_dma_map(hp->dma_dev, this_frag,
+						   0, len, DMA_TO_DEVICE);
+			if (unlikely(dma_mapping_error(hp->dma_dev, mapping))) {
+				unmap_partial_tx_skb(hp, first_mapping, first_len,
+						     first_entry, entry);
+				goto out_dma_error;
+			}
+			this_txflags = tx_flags;
+			if (frag == skb_shinfo(skb)->nr_frags - 1)
+				this_txflags |= TXFLAG_EOP;
+			hme_write_txd(hp, &hp->happy_block->happy_meal_txd[entry],
+				      (this_txflags | (len & TXFLAG_SIZE)),
+				      mapping);
+			entry = NEXT_TX(entry);
+		}
+		hme_write_txd(hp, &hp->happy_block->happy_meal_txd[first_entry],
+			      (tx_flags | TXFLAG_SOP | (first_len & TXFLAG_SIZE)),
+			      first_mapping);
+	}
+
+	hp->tx_new = entry;
+
+	if (TX_BUFFS_AVAIL(hp) <= (MAX_SKB_FRAGS + 1))
+		netif_stop_queue(dev);
+
+	/* Get it going. */
+	hme_write32(hp, hp->etxregs + ETX_PENDING, ETX_TP_DMAWAKEUP);
+
+	spin_unlock_irq(&hp->happy_lock);
+
+	tx_add_log(hp, TXLOG_ACTION_TXMIT, 0);
+	return NETDEV_TX_OK;
+
+out_dma_error:
+	hp->tx_skbs[hp->tx_new] = NULL;
+	spin_unlock_irq(&hp->happy_lock);
+
+	dev_kfree_skb_any(skb);
+	dev->stats.tx_dropped++;
+	return NETDEV_TX_OK;
+}
+
+static struct net_device_stats *happy_meal_get_stats(struct net_device *dev)
+{
+	struct happy_meal *hp = netdev_priv(dev);
+
+	spin_lock_irq(&hp->happy_lock);
+	happy_meal_get_counters(hp, hp->bigmacregs);
+	spin_unlock_irq(&hp->happy_lock);
+
+	return &dev->stats;
+}
+
+static void happy_meal_set_multicast(struct net_device *dev)
+{
+	struct happy_meal *hp = netdev_priv(dev);
+	void __iomem *bregs = hp->bigmacregs;
+	struct netdev_hw_addr *ha;
+	u32 crc;
+
+	spin_lock_irq(&hp->happy_lock);
+
+	if ((dev->flags & IFF_ALLMULTI) || (netdev_mc_count(dev) > 64)) {
+		hme_write32(hp, bregs + BMAC_HTABLE0, 0xffff);
+		hme_write32(hp, bregs + BMAC_HTABLE1, 0xffff);
+		hme_write32(hp, bregs + BMAC_HTABLE2, 0xffff);
+		hme_write32(hp, bregs + BMAC_HTABLE3, 0xffff);
+	} else if (dev->flags & IFF_PROMISC) {
+		hme_write32(hp, bregs + BMAC_RXCFG,
+			    hme_read32(hp, bregs + BMAC_RXCFG) | BIGMAC_RXCFG_PMISC);
+	} else {
+		u16 hash_table[4];
+
+		memset(hash_table, 0, sizeof(hash_table));
+		netdev_for_each_mc_addr(ha, dev) {
+			crc = ether_crc_le(6, ha->addr);
+			crc >>= 26;
+			hash_table[crc >> 4] |= 1 << (crc & 0xf);
+		}
+		hme_write32(hp, bregs + BMAC_HTABLE0, hash_table[0]);
+		hme_write32(hp, bregs + BMAC_HTABLE1, hash_table[1]);
+		hme_write32(hp, bregs + BMAC_HTABLE2, hash_table[2]);
+		hme_write32(hp, bregs + BMAC_HTABLE3, hash_table[3]);
+	}
+
+	spin_unlock_irq(&hp->happy_lock);
+}
+
+/* Ethtool support... */
+static int hme_get_link_ksettings(struct net_device *dev,
+				  struct ethtool_link_ksettings *cmd)
+{
+	struct happy_meal *hp = netdev_priv(dev);
+	u32 speed;
+	u32 supported;
+
+	supported =
+		(SUPPORTED_10baseT_Half | SUPPORTED_10baseT_Full |
+		 SUPPORTED_100baseT_Half | SUPPORTED_100baseT_Full |
+		 SUPPORTED_Autoneg | SUPPORTED_TP | SUPPORTED_MII);
+
+	/* XXX hardcoded stuff for now */
+	cmd->base.port = PORT_TP; /* XXX no MII support */
+	cmd->base.phy_address = 0; /* XXX fixed PHYAD */
+
+	/* Record PHY settings. */
+	spin_lock_irq(&hp->happy_lock);
+	hp->sw_bmcr = happy_meal_tcvr_read(hp, hp->tcvregs, MII_BMCR);
+	hp->sw_lpa = happy_meal_tcvr_read(hp, hp->tcvregs, MII_LPA);
+	spin_unlock_irq(&hp->happy_lock);
+
+	if (hp->sw_bmcr & BMCR_ANENABLE) {
+		cmd->base.autoneg = AUTONEG_ENABLE;
+		speed = ((hp->sw_lpa & (LPA_100HALF | LPA_100FULL)) ?
+			 SPEED_100 : SPEED_10);
+		if (speed == SPEED_100)
+			cmd->base.duplex =
+				(hp->sw_lpa & (LPA_100FULL)) ?
+				DUPLEX_FULL : DUPLEX_HALF;
+		else
+			cmd->base.duplex =
+				(hp->sw_lpa & (LPA_10FULL)) ?
+				DUPLEX_FULL : DUPLEX_HALF;
+	} else {
+		cmd->base.autoneg = AUTONEG_DISABLE;
+		speed = (hp->sw_bmcr & BMCR_SPEED100) ? SPEED_100 : SPEED_10;
+		cmd->base.duplex =
+			(hp->sw_bmcr & BMCR_FULLDPLX) ?
+			DUPLEX_FULL : DUPLEX_HALF;
+	}
+	cmd->base.speed = speed;
+	ethtool_convert_legacy_u32_to_link_mode(cmd->link_modes.supported,
+						supported);
+
+	return 0;
+}
+
+static int hme_set_link_ksettings(struct net_device *dev,
+				  const struct ethtool_link_ksettings *cmd)
+{
+	struct happy_meal *hp = netdev_priv(dev);
+
+	/* Verify the settings we care about. */
+	if (cmd->base.autoneg != AUTONEG_ENABLE &&
+	    cmd->base.autoneg != AUTONEG_DISABLE)
+		return -EINVAL;
+	if (cmd->base.autoneg == AUTONEG_DISABLE &&
+	    ((cmd->base.speed != SPEED_100 &&
+	      cmd->base.speed != SPEED_10) ||
+	     (cmd->base.duplex != DUPLEX_HALF &&
+	      cmd->base.duplex != DUPLEX_FULL)))
+		return -EINVAL;
+
+	/* Ok, do it to it. */
+	spin_lock_irq(&hp->happy_lock);
+	timer_delete(&hp->happy_timer);
+	happy_meal_begin_auto_negotiation(hp, hp->tcvregs, cmd);
+	spin_unlock_irq(&hp->happy_lock);
+
+	return 0;
+}
+
+static void hme_get_drvinfo(struct net_device *dev, struct ethtool_drvinfo *info)
+{
+	struct happy_meal *hp = netdev_priv(dev);
+
+	strscpy(info->driver, DRV_NAME, sizeof(info->driver));
+	if (hp->happy_flags & HFLAG_PCI) {
+		struct pci_dev *pdev = hp->happy_dev;
+		strscpy(info->bus_info, pci_name(pdev), sizeof(info->bus_info));
+	}
+#ifdef CONFIG_SBUS
+	else {
+		const struct linux_prom_registers *regs;
+		struct platform_device *op = hp->happy_dev;
+		regs = of_get_property(op->dev.of_node, "regs", NULL);
+		if (regs)
+			snprintf(info->bus_info, sizeof(info->bus_info),
+				"SBUS:%d",
+				regs->which_io);
+	}
+#endif
+}
+
+static u32 hme_get_link(struct net_device *dev)
+{
+	struct happy_meal *hp = netdev_priv(dev);
+
+	spin_lock_irq(&hp->happy_lock);
+	hp->sw_bmcr = happy_meal_tcvr_read(hp, hp->tcvregs, MII_BMCR);
+	spin_unlock_irq(&hp->happy_lock);
+
+	return hp->sw_bmsr & BMSR_LSTATUS;
+}
+
+static const struct ethtool_ops hme_ethtool_ops = {
+	.get_drvinfo		= hme_get_drvinfo,
+	.get_link		= hme_get_link,
+	.get_link_ksettings	= hme_get_link_ksettings,
+	.set_link_ksettings	= hme_set_link_ksettings,
+};
+
+#ifdef CONFIG_SBUS
+/* Given a happy meal sbus device, find it's quattro parent.
+ * If none exist, allocate and return a new one.
+ *
+ * Return NULL on failure.
+ */
+static struct quattro *quattro_sbus_find(struct platform_device *child)
+{
+	struct device *parent = child->dev.parent;
+	struct platform_device *op;
+	struct quattro *qp;
+
+	op = to_platform_device(parent);
+	qp = platform_get_drvdata(op);
+	if (qp)
+		return qp;
+
+	qp = kzalloc_obj(*qp);
+	if (!qp)
+		return NULL;
+
+	qp->quattro_dev = child;
+	qp->next = qfe_sbus_list;
+	qfe_sbus_list = qp;
+
+	platform_set_drvdata(op, qp);
+	return qp;
+}
+#endif /* CONFIG_SBUS */
+
+#ifdef CONFIG_PCI
+static struct quattro *quattro_pci_find(struct pci_dev *pdev)
+{
+	int i;
+	struct pci_dev *bdev = pdev->bus->self;
+	struct quattro *qp;
+
+	if (!bdev)
+		return ERR_PTR(-ENODEV);
+
+	for (qp = qfe_pci_list; qp != NULL; qp = qp->next) {
+		struct pci_dev *qpdev = qp->quattro_dev;
+
+		if (qpdev == bdev)
+			return qp;
+	}
+
+	qp = kmalloc_obj(struct quattro);
+	if (!qp)
+		return ERR_PTR(-ENOMEM);
+
+	for (i = 0; i < 4; i++)
+		qp->happy_meals[i] = NULL;
+
+	qp->quattro_dev = bdev;
+	qp->next = qfe_pci_list;
+	qfe_pci_list = qp;
+
+	/* No range tricks necessary on PCI. */
+	qp->nranges = 0;
+	return qp;
+}
+#endif /* CONFIG_PCI */
+
+static const struct net_device_ops hme_netdev_ops = {
+	.ndo_open		= happy_meal_open,
+	.ndo_stop		= happy_meal_close,
+	.ndo_start_xmit		= happy_meal_start_xmit,
+	.ndo_tx_timeout		= happy_meal_tx_timeout,
+	.ndo_get_stats		= happy_meal_get_stats,
+	.ndo_set_rx_mode	= happy_meal_set_multicast,
+	.ndo_set_mac_address 	= eth_mac_addr,
+	.ndo_validate_addr	= eth_validate_addr,
+};
+
+#ifdef CONFIG_PCI
+static int is_quattro_p(struct pci_dev *pdev)
+{
+	struct pci_dev *busdev = pdev->bus->self;
+	struct pci_dev *this_pdev;
+	int n_hmes;
+
+	if (!busdev || busdev->vendor != PCI_VENDOR_ID_DEC ||
+	    busdev->device != PCI_DEVICE_ID_DEC_21153)
+		return 0;
+
+	n_hmes = 0;
+	list_for_each_entry(this_pdev, &pdev->bus->devices, bus_list) {
+		if (this_pdev->vendor == PCI_VENDOR_ID_SUN &&
+		    this_pdev->device == PCI_DEVICE_ID_SUN_HAPPYMEAL)
+			n_hmes++;
+	}
+
+	if (n_hmes != 4)
+		return 0;
+
+	return 1;
+}
+
+/* Fetch MAC address from vital product data of PCI ROM. */
+static int find_eth_addr_in_vpd(void __iomem *rom_base, int len, int index, unsigned char *dev_addr)
+{
+	int this_offset;
+
+	for (this_offset = 0x20; this_offset < len; this_offset++) {
+		void __iomem *p = rom_base + this_offset;
+
+		if (readb(p + 0) != 0x90 ||
+		    readb(p + 1) != 0x00 ||
+		    readb(p + 2) != 0x09 ||
+		    readb(p + 3) != 0x4e ||
+		    readb(p + 4) != 0x41 ||
+		    readb(p + 5) != 0x06)
+			continue;
+
+		this_offset += 6;
+		p += 6;
+
+		if (index == 0) {
+			for (int i = 0; i < 6; i++)
+				dev_addr[i] = readb(p + i);
+			return 1;
+		}
+		index--;
+	}
+	return 0;
+}
+
+static void __maybe_unused get_hme_mac_nonsparc(struct pci_dev *pdev,
+						unsigned char *dev_addr)
+{
+	void __iomem *p;
+	size_t size;
+
+	p = pci_map_rom(pdev, &size);
+	if (p) {
+		int index = 0;
+		int found;
+
+		if (is_quattro_p(pdev))
+			index = PCI_SLOT(pdev->devfn);
+
+		found = readb(p) == 0x55 &&
+			readb(p + 1) == 0xaa &&
+			find_eth_addr_in_vpd(p, (64 * 1024), index, dev_addr);
+		pci_unmap_rom(pdev, p);
+		if (found)
+			return;
+	}
+
+	/* Sun MAC prefix then 3 random bytes. */
+	dev_addr[0] = 0x08;
+	dev_addr[1] = 0x00;
+	dev_addr[2] = 0x20;
+	get_random_bytes(&dev_addr[3], 3);
+}
+#endif
+
+static void happy_meal_addr_init(struct happy_meal *hp,
+				 struct device_node *dp, int qfe_slot)
+{
+	int i;
+
+	for (i = 0; i < 6; i++) {
+		if (macaddr[i] != 0)
+			break;
+	}
+
+	if (i < 6) { /* a mac address was given */
+		u8 addr[ETH_ALEN];
+
+		for (i = 0; i < 6; i++)
+			addr[i] = macaddr[i];
+		eth_hw_addr_set(hp->dev, addr);
+		macaddr[5]++;
+	} else {
+#ifdef CONFIG_SPARC
+		const unsigned char *addr;
+		int len;
+
+		/* If user did not specify a MAC address specifically, use
+		 * the Quattro local-mac-address property...
+		 */
+		if (qfe_slot != -1) {
+			addr = of_get_property(dp, "local-mac-address", &len);
+			if (addr && len == 6) {
+				eth_hw_addr_set(hp->dev, addr);
+				return;
+			}
+		}
+
+		eth_hw_addr_set(hp->dev, idprom->id_ethaddr);
+#else
+		u8 addr[ETH_ALEN];
+
+		get_hme_mac_nonsparc(hp->happy_dev, addr);
+		eth_hw_addr_set(hp->dev, addr);
+#endif
+	}
+}
+
+static int happy_meal_common_probe(struct happy_meal *hp,
+				   struct device_node *dp)
+{
+	struct net_device *dev = hp->dev;
+	int err;
+
+#ifdef CONFIG_SPARC
+	hp->hm_revision = of_getintprop_default(dp, "hm-rev", hp->hm_revision);
+#endif
+
+	/* Now enable the feature flags we can. */
+	if (hp->hm_revision == 0x20 || hp->hm_revision == 0x21)
+		hp->happy_flags |= HFLAG_20_21;
+	else if (hp->hm_revision != 0xa0)
+		hp->happy_flags |= HFLAG_NOT_A0;
+
+	hp->happy_block = dmam_alloc_coherent(hp->dma_dev, PAGE_SIZE,
+					      &hp->hblock_dvma, GFP_KERNEL);
+	if (!hp->happy_block)
+		return -ENOMEM;
+
+	/* Force check of the link first time we are brought up. */
+	hp->linkcheck = 0;
+
+	/* Force timer state to 'asleep' with count of zero. */
+	hp->timer_state = asleep;
+	hp->timer_ticks = 0;
+
+	timer_setup(&hp->happy_timer, happy_meal_timer, 0);
+
+	dev->netdev_ops = &hme_netdev_ops;
+	dev->watchdog_timeo = 5 * HZ;
+	dev->ethtool_ops = &hme_ethtool_ops;
+
+	/* Happy Meal can do it all... */
+	dev->hw_features = NETIF_F_SG | NETIF_F_HW_CSUM;
+	dev->features |= dev->hw_features | NETIF_F_RXCSUM;
+
+
+	/* Grrr, Happy Meal comes up by default not advertising
+	 * full duplex 100baseT capabilities, fix this.
+	 */
+	spin_lock_irq(&hp->happy_lock);
+	happy_meal_set_initial_advertisement(hp);
+	spin_unlock_irq(&hp->happy_lock);
+
+	err = devm_register_netdev(hp->dma_dev, dev);
+	if (err)
+		dev_err(hp->dma_dev, "Cannot register net device, aborting.\n");
+	return err;
+}
+
+#ifdef CONFIG_SBUS
+static int happy_meal_sbus_probe_one(struct platform_device *op, int is_qfe)
+{
+	struct device_node *dp = op->dev.of_node, *sbus_dp;
+	struct quattro *qp = NULL;
+	struct happy_meal *hp;
+	struct net_device *dev;
+	int qfe_slot = -1;
+	int err;
+
+	sbus_dp = op->dev.parent->of_node;
+
+	/* We can match PCI devices too, do not accept those here. */
+	if (!of_node_name_eq(sbus_dp, "sbus") && !of_node_name_eq(sbus_dp, "sbi"))
+		return -ENODEV;
+
+	if (is_qfe) {
+		qp = quattro_sbus_find(op);
+		if (qp == NULL)
+			return -ENODEV;
+		for (qfe_slot = 0; qfe_slot < 4; qfe_slot++)
+			if (qp->happy_meals[qfe_slot] == NULL)
+				break;
+		if (qfe_slot == 4)
+			return -ENODEV;
+	}
+
+	dev = devm_alloc_etherdev(&op->dev, sizeof(struct happy_meal));
+	if (!dev)
+		return -ENOMEM;
+	SET_NETDEV_DEV(dev, &op->dev);
+
+	hp = netdev_priv(dev);
+	hp->dev = dev;
+	hp->happy_dev = op;
+	hp->dma_dev = &op->dev;
+	happy_meal_addr_init(hp, dp, qfe_slot);
+
+	spin_lock_init(&hp->happy_lock);
+
+	if (qp != NULL) {
+		hp->qfe_parent = qp;
+		hp->qfe_ent = qfe_slot;
+		qp->happy_meals[qfe_slot] = dev;
+	}
+
+	hp->gregs = devm_platform_ioremap_resource(op, 0);
+	if (IS_ERR(hp->gregs)) {
+		dev_err(&op->dev, "Cannot map global registers.\n");
+		err = PTR_ERR(hp->gregs);
+		goto err_out_clear_quattro;
+	}
+
+	hp->etxregs = devm_platform_ioremap_resource(op, 1);
+	if (IS_ERR(hp->etxregs)) {
+		dev_err(&op->dev, "Cannot map MAC TX registers.\n");
+		err = PTR_ERR(hp->etxregs);
+		goto err_out_clear_quattro;
+	}
+
+	hp->erxregs = devm_platform_ioremap_resource(op, 2);
+	if (IS_ERR(hp->erxregs)) {
+		dev_err(&op->dev, "Cannot map MAC RX registers.\n");
+		err = PTR_ERR(hp->erxregs);
+		goto err_out_clear_quattro;
+	}
+
+	/* BIGMAC may have bogus sizes */
+	if ((op->resource[3].end - op->resource[3].start) >= BMAC_REG_SIZE)
+		op->resource[3].end = op->resource[3].start + BMAC_REG_SIZE - 1;
+	hp->bigmacregs = devm_platform_ioremap_resource(op, 3);
+	if (IS_ERR(hp->bigmacregs)) {
+		dev_err(&op->dev, "Cannot map BIGMAC registers.\n");
+		err = PTR_ERR(hp->bigmacregs);
+		goto err_out_clear_quattro;
+	}
+
+	hp->tcvregs = devm_platform_ioremap_resource(op, 4);
+	if (IS_ERR(hp->tcvregs)) {
+		dev_err(&op->dev, "Cannot map TCVR registers.\n");
+		err = PTR_ERR(hp->tcvregs);
+		goto err_out_clear_quattro;
+	}
+
+	hp->hm_revision = 0xa0;
+
+	if (qp != NULL)
+		hp->happy_flags |= HFLAG_QUATTRO;
+
+	hp->irq = op->archdata.irqs[0];
+
+	/* Get the supported DVMA burst sizes from our Happy SBUS. */
+	hp->happy_bursts = of_getintprop_default(sbus_dp,
+						 "burst-sizes", 0x00);
+
+#ifdef CONFIG_PCI
+	/* Hook up SBUS register/descriptor accessors. */
+	hp->read_desc32 = sbus_hme_read_desc32;
+	hp->write_txd = sbus_hme_write_txd;
+	hp->write_rxd = sbus_hme_write_rxd;
+	hp->read32 = sbus_hme_read32;
+	hp->write32 = sbus_hme_write32;
+#endif
+
+	err = happy_meal_common_probe(hp, dp);
+	if (err)
+		goto err_out_clear_quattro;
+
+	platform_set_drvdata(op, hp);
+
+	if (qfe_slot != -1)
+		netdev_info(dev,
+			    "Quattro HME slot %d (SBUS) 10/100baseT Ethernet %pM\n",
+			    qfe_slot, dev->dev_addr);
+	else
+		netdev_info(dev, "HAPPY MEAL (SBUS) 10/100baseT Ethernet %pM\n",
+			    dev->dev_addr);
+
+	return 0;
+
+err_out_clear_quattro:
+	if (qp)
+		qp->happy_meals[qfe_slot] = NULL;
+	return err;
+}
+#endif
+
+#ifdef CONFIG_PCI
+static int happy_meal_pci_probe(struct pci_dev *pdev,
+				const struct pci_device_id *ent)
+{
+	struct device_node *dp = NULL;
+	struct quattro *qp = NULL;
+	struct happy_meal *hp;
+	struct net_device *dev;
+	void __iomem *hpreg_base;
+	struct resource *hpreg_res;
+	char prom_name[64];
+	int qfe_slot = -1;
+	int err = -ENODEV;
+
+	/* Now make sure pci_dev cookie is there. */
+#ifdef CONFIG_SPARC
+	dp = pci_device_to_OF_node(pdev);
+	snprintf(prom_name, sizeof(prom_name), "%pOFn", dp);
+#else
+	if (is_quattro_p(pdev))
+		strcpy(prom_name, "SUNW,qfe");
+	else
+		strcpy(prom_name, "SUNW,hme");
+#endif
+
+	err = pcim_enable_device(pdev);
+	if (err)
+		return err;
+	pci_set_master(pdev);
+
+	if (!strcmp(prom_name, "SUNW,qfe") || !strcmp(prom_name, "qfe")) {
+		qp = quattro_pci_find(pdev);
+		if (IS_ERR(qp))
+			return PTR_ERR(qp);
+
+		for (qfe_slot = 0; qfe_slot < 4; qfe_slot++)
+			if (!qp->happy_meals[qfe_slot])
+				break;
+
+		if (qfe_slot == 4)
+			return -ENODEV;
+	}
+
+	dev = devm_alloc_etherdev(&pdev->dev, sizeof(struct happy_meal));
+	if (!dev)
+		return -ENOMEM;
+	SET_NETDEV_DEV(dev, &pdev->dev);
+
+	hp = netdev_priv(dev);
+	hp->dev = dev;
+	hp->happy_dev = pdev;
+	hp->dma_dev = &pdev->dev;
+
+	spin_lock_init(&hp->happy_lock);
+
+	if (qp != NULL) {
+		hp->qfe_parent = qp;
+		hp->qfe_ent = qfe_slot;
+		qp->happy_meals[qfe_slot] = dev;
+	}
+
+	err = -EINVAL;
+	if ((pci_resource_flags(pdev, 0) & IORESOURCE_IO) != 0) {
+		dev_err(&pdev->dev,
+			"Cannot find proper PCI device base address.\n");
+		goto err_out_clear_quattro;
+	}
+
+	hpreg_res = devm_request_mem_region(&pdev->dev,
+					    pci_resource_start(pdev, 0),
+					    pci_resource_len(pdev, 0),
+					    DRV_NAME);
+	if (!hpreg_res) {
+		err = -EBUSY;
+		dev_err(&pdev->dev, "Cannot obtain PCI resources, aborting.\n");
+		goto err_out_clear_quattro;
+	}
+
+	hpreg_base = pcim_iomap(pdev, 0, 0x8000);
+	if (!hpreg_base) {
+		err = -ENOMEM;
+		dev_err(&pdev->dev, "Unable to remap card memory.\n");
+		goto err_out_clear_quattro;
+	}
+
+	happy_meal_addr_init(hp, dp, qfe_slot);
+
+	/* Layout registers. */
+	hp->gregs      = (hpreg_base + 0x0000UL);
+	hp->etxregs    = (hpreg_base + 0x2000UL);
+	hp->erxregs    = (hpreg_base + 0x4000UL);
+	hp->bigmacregs = (hpreg_base + 0x6000UL);
+	hp->tcvregs    = (hpreg_base + 0x7000UL);
+
+	if (IS_ENABLED(CONFIG_SPARC))
+		hp->hm_revision = 0xc0 | (pdev->revision & 0x0f);
+	else
+		hp->hm_revision = 0x20;
+
+	if (qp != NULL)
+		hp->happy_flags |= HFLAG_QUATTRO;
+
+	/* And of course, indicate this is PCI. */
+	hp->happy_flags |= HFLAG_PCI;
+
+#ifdef CONFIG_SPARC
+	/* Assume PCI happy meals can handle all burst sizes. */
+	hp->happy_bursts = DMA_BURSTBITS;
+#endif
+	hp->irq = pdev->irq;
+
+#ifdef CONFIG_SBUS
+	/* Hook up PCI register/descriptor accessors. */
+	hp->read_desc32 = pci_hme_read_desc32;
+	hp->write_txd = pci_hme_write_txd;
+	hp->write_rxd = pci_hme_write_rxd;
+	hp->read32 = pci_hme_read32;
+	hp->write32 = pci_hme_write32;
+#endif
+
+	err = happy_meal_common_probe(hp, dp);
+	if (err)
+		goto err_out_clear_quattro;
+
+	pci_set_drvdata(pdev, hp);
+
+	if (!qfe_slot) {
+		struct pci_dev *qpdev = qp->quattro_dev;
+
+		prom_name[0] = 0;
+		if (!strncmp(dev->name, "eth", 3)) {
+			int i = simple_strtoul(dev->name + 3, NULL, 10);
+			sprintf(prom_name, "-%d", i + 3);
+		}
+		netdev_info(dev,
+			    "%s: Quattro HME (PCI/CheerIO) 10/100baseT Ethernet bridge %04x.%04x\n",
+			    prom_name, qpdev->vendor, qpdev->device);
+	}
+
+	if (qfe_slot != -1)
+		netdev_info(dev,
+			    "Quattro HME slot %d (PCI/CheerIO) 10/100baseT Ethernet %pM\n",
+			    qfe_slot, dev->dev_addr);
+	else
+		netdev_info(dev,
+			    "HAPPY MEAL (PCI/CheerIO) 10/100BaseT Ethernet %pM\n",
+			    dev->dev_addr);
+
+	return 0;
+
+err_out_clear_quattro:
+	if (qp != NULL)
+		qp->happy_meals[qfe_slot] = NULL;
+	return err;
+}
+
+static const struct pci_device_id happymeal_pci_ids[] = {
+	{ PCI_DEVICE(PCI_VENDOR_ID_SUN, PCI_DEVICE_ID_SUN_HAPPYMEAL) },
+	{ }			/* Terminating entry */
+};
+
+MODULE_DEVICE_TABLE(pci, happymeal_pci_ids);
+
+static struct pci_driver hme_pci_driver = {
+	.name		= "hme",
+	.id_table	= happymeal_pci_ids,
+	.probe		= happy_meal_pci_probe,
+};
+
+static int __init happy_meal_pci_init(void)
+{
+	return pci_register_driver(&hme_pci_driver);
+}
+
+static void happy_meal_pci_exit(void)
+{
+	pci_unregister_driver(&hme_pci_driver);
+
+	while (qfe_pci_list) {
+		struct quattro *qfe = qfe_pci_list;
+		struct quattro *next = qfe->next;
+
+		kfree(qfe);
+
+		qfe_pci_list = next;
+	}
+}
+
+#endif
+
+#ifdef CONFIG_SBUS
+static const struct of_device_id hme_sbus_match[];
+static int hme_sbus_probe(struct platform_device *op)
+{
+	const struct of_device_id *match;
+	struct device_node *dp = op->dev.of_node;
+	const char *model = of_get_property(dp, "model", NULL);
+	int is_qfe;
+
+	match = of_match_device(hme_sbus_match, &op->dev);
+	if (!match)
+		return -EINVAL;
+	is_qfe = (match->data != NULL);
+
+	if (!is_qfe && model && !strcmp(model, "SUNW,sbus-qfe"))
+		is_qfe = 1;
+
+	return happy_meal_sbus_probe_one(op, is_qfe);
+}
+
+static const struct of_device_id hme_sbus_match[] = {
+	{
+		.name = "SUNW,hme",
+	},
+	{
+		.name = "SUNW,qfe",
+		.data = (void *) 1,
+	},
+	{
+		.name = "qfe",
+		.data = (void *) 1,
+	},
+	{},
+};
+
+MODULE_DEVICE_TABLE(of, hme_sbus_match);
+
+static struct platform_driver hme_sbus_driver = {
+	.driver = {
+		.name = "hme",
+		.of_match_table = hme_sbus_match,
+	},
+	.probe		= hme_sbus_probe,
+};
+
+static int __init happy_meal_sbus_init(void)
+{
+	return platform_driver_register(&hme_sbus_driver);
+}
+
+static void happy_meal_sbus_exit(void)
+{
+	platform_driver_unregister(&hme_sbus_driver);
+
+	while (qfe_sbus_list) {
+		struct quattro *qfe = qfe_sbus_list;
+		struct quattro *next = qfe->next;
+
+		kfree(qfe);
+
+		qfe_sbus_list = next;
+	}
+}
+#endif
+
+static int __init happy_meal_probe(void)
+{
+	int err = 0;
+
+#ifdef CONFIG_SBUS
+	err = happy_meal_sbus_init();
+#endif
+#ifdef CONFIG_PCI
+	if (!err) {
+		err = happy_meal_pci_init();
+#ifdef CONFIG_SBUS
+		if (err)
+			happy_meal_sbus_exit();
+#endif
+	}
+#endif
+
+	return err;
+}
+
+
+static void __exit happy_meal_exit(void)
+{
+#ifdef CONFIG_SBUS
+	happy_meal_sbus_exit();
+#endif
+#ifdef CONFIG_PCI
+	happy_meal_pci_exit();
+#endif
+}
+
+module_init(happy_meal_probe);
+module_exit(happy_meal_exit);
diff --git a/drivers/net/ethernet/sun/sunhme.h b/drivers/net/ethernet/sun/sunhme.h
new file mode 100644
index 000000000..258b4c7fe
--- /dev/null
+++ b/drivers/net/ethernet/sun/sunhme.h
@@ -0,0 +1,510 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/* $Id: sunhme.h,v 1.33 2001/08/03 06:23:04 davem Exp $
+ * sunhme.h: Definitions for Sparc HME/BigMac 10/100baseT ethernet driver.
+ *           Also known as the "Happy Meal".
+ *
+ * Copyright (C) 1996, 1999 David S. Miller (davem@redhat.com)
+ */
+
+#ifndef _SUNHME_H
+#define _SUNHME_H
+
+#include <linux/pci.h>
+
+/* Happy Meal global registers. */
+#define GREG_SWRESET	0x000UL	/* Software Reset  */
+#define GREG_CFG	0x004UL	/* Config Register */
+#define GREG_STAT	0x100UL	/* Status          */
+#define GREG_IMASK	0x104UL	/* Interrupt Mask  */
+#define GREG_REG_SIZE	0x108UL
+
+/* Global reset register. */
+#define GREG_RESET_ETX         0x01
+#define GREG_RESET_ERX         0x02
+#define GREG_RESET_ALL         0x03
+
+/* Global config register. */
+#define GREG_CFG_BURSTMSK      0x03
+#define GREG_CFG_BURST16       0x00
+#define GREG_CFG_BURST32       0x01
+#define GREG_CFG_BURST64       0x02
+#define GREG_CFG_64BIT         0x04
+#define GREG_CFG_PARITY        0x08
+#define GREG_CFG_RESV          0x10
+
+/* Global status register. */
+#define GREG_STAT_GOTFRAME     0x00000001 /* Received a frame                         */
+#define GREG_STAT_RCNTEXP      0x00000002 /* Receive frame counter expired            */
+#define GREG_STAT_ACNTEXP      0x00000004 /* Align-error counter expired              */
+#define GREG_STAT_CCNTEXP      0x00000008 /* CRC-error counter expired                */
+#define GREG_STAT_LCNTEXP      0x00000010 /* Length-error counter expired             */
+#define GREG_STAT_RFIFOVF      0x00000020 /* Receive FIFO overflow                    */
+#define GREG_STAT_CVCNTEXP     0x00000040 /* Code-violation counter expired           */
+#define GREG_STAT_STSTERR      0x00000080 /* Test error in XIF for SQE                */
+#define GREG_STAT_SENTFRAME    0x00000100 /* Transmitted a frame                      */
+#define GREG_STAT_TFIFO_UND    0x00000200 /* Transmit FIFO underrun                   */
+#define GREG_STAT_MAXPKTERR    0x00000400 /* Max-packet size error                    */
+#define GREG_STAT_NCNTEXP      0x00000800 /* Normal-collision counter expired         */
+#define GREG_STAT_ECNTEXP      0x00001000 /* Excess-collision counter expired         */
+#define GREG_STAT_LCCNTEXP     0x00002000 /* Late-collision counter expired           */
+#define GREG_STAT_FCNTEXP      0x00004000 /* First-collision counter expired          */
+#define GREG_STAT_DTIMEXP      0x00008000 /* Defer-timer expired                      */
+#define GREG_STAT_RXTOHOST     0x00010000 /* Moved from receive-FIFO to host memory   */
+#define GREG_STAT_NORXD        0x00020000 /* No more receive descriptors              */
+#define GREG_STAT_RXERR        0x00040000 /* Error during receive dma                 */
+#define GREG_STAT_RXLATERR     0x00080000 /* Late error during receive dma            */
+#define GREG_STAT_RXPERR       0x00100000 /* Parity error during receive dma          */
+#define GREG_STAT_RXTERR       0x00200000 /* Tag error during receive dma             */
+#define GREG_STAT_EOPERR       0x00400000 /* Transmit descriptor did not have EOP set */
+#define GREG_STAT_MIFIRQ       0x00800000 /* MIF is signaling an interrupt condition  */
+#define GREG_STAT_HOSTTOTX     0x01000000 /* Moved from host memory to transmit-FIFO  */
+#define GREG_STAT_TXALL        0x02000000 /* Transmitted all packets in the tx-fifo   */
+#define GREG_STAT_TXEACK       0x04000000 /* Error during transmit dma                */
+#define GREG_STAT_TXLERR       0x08000000 /* Late error during transmit dma           */
+#define GREG_STAT_TXPERR       0x10000000 /* Parity error during transmit dma         */
+#define GREG_STAT_TXTERR       0x20000000 /* Tag error during transmit dma            */
+#define GREG_STAT_SLVERR       0x40000000 /* PIO access got an error                  */
+#define GREG_STAT_SLVPERR      0x80000000 /* PIO access got a parity error            */
+
+/* All interesting error conditions. */
+#define GREG_STAT_ERRORS       0xfc7efefc
+
+/* Global interrupt mask register. */
+#define GREG_IMASK_GOTFRAME    0x00000001 /* Received a frame                         */
+#define GREG_IMASK_RCNTEXP     0x00000002 /* Receive frame counter expired            */
+#define GREG_IMASK_ACNTEXP     0x00000004 /* Align-error counter expired              */
+#define GREG_IMASK_CCNTEXP     0x00000008 /* CRC-error counter expired                */
+#define GREG_IMASK_LCNTEXP     0x00000010 /* Length-error counter expired             */
+#define GREG_IMASK_RFIFOVF     0x00000020 /* Receive FIFO overflow                    */
+#define GREG_IMASK_CVCNTEXP    0x00000040 /* Code-violation counter expired           */
+#define GREG_IMASK_STSTERR     0x00000080 /* Test error in XIF for SQE                */
+#define GREG_IMASK_SENTFRAME   0x00000100 /* Transmitted a frame                      */
+#define GREG_IMASK_TFIFO_UND   0x00000200 /* Transmit FIFO underrun                   */
+#define GREG_IMASK_MAXPKTERR   0x00000400 /* Max-packet size error                    */
+#define GREG_IMASK_NCNTEXP     0x00000800 /* Normal-collision counter expired         */
+#define GREG_IMASK_ECNTEXP     0x00001000 /* Excess-collision counter expired         */
+#define GREG_IMASK_LCCNTEXP    0x00002000 /* Late-collision counter expired           */
+#define GREG_IMASK_FCNTEXP     0x00004000 /* First-collision counter expired          */
+#define GREG_IMASK_DTIMEXP     0x00008000 /* Defer-timer expired                      */
+#define GREG_IMASK_RXTOHOST    0x00010000 /* Moved from receive-FIFO to host memory   */
+#define GREG_IMASK_NORXD       0x00020000 /* No more receive descriptors              */
+#define GREG_IMASK_RXERR       0x00040000 /* Error during receive dma                 */
+#define GREG_IMASK_RXLATERR    0x00080000 /* Late error during receive dma            */
+#define GREG_IMASK_RXPERR      0x00100000 /* Parity error during receive dma          */
+#define GREG_IMASK_RXTERR      0x00200000 /* Tag error during receive dma             */
+#define GREG_IMASK_EOPERR      0x00400000 /* Transmit descriptor did not have EOP set */
+#define GREG_IMASK_MIFIRQ      0x00800000 /* MIF is signaling an interrupt condition  */
+#define GREG_IMASK_HOSTTOTX    0x01000000 /* Moved from host memory to transmit-FIFO  */
+#define GREG_IMASK_TXALL       0x02000000 /* Transmitted all packets in the tx-fifo   */
+#define GREG_IMASK_TXEACK      0x04000000 /* Error during transmit dma                */
+#define GREG_IMASK_TXLERR      0x08000000 /* Late error during transmit dma           */
+#define GREG_IMASK_TXPERR      0x10000000 /* Parity error during transmit dma         */
+#define GREG_IMASK_TXTERR      0x20000000 /* Tag error during transmit dma            */
+#define GREG_IMASK_SLVERR      0x40000000 /* PIO access got an error                  */
+#define GREG_IMASK_SLVPERR     0x80000000 /* PIO access got a parity error            */
+
+/* Happy Meal external transmitter registers. */
+#define ETX_PENDING	0x00UL	/* Transmit pending/wakeup register */
+#define ETX_CFG		0x04UL	/* Transmit config register         */
+#define ETX_RING	0x08UL	/* Transmit ring pointer            */
+#define ETX_BBASE	0x0cUL	/* Transmit buffer base             */
+#define ETX_BDISP	0x10UL	/* Transmit buffer displacement     */
+#define ETX_FIFOWPTR	0x14UL	/* FIFO write ptr                   */
+#define ETX_FIFOSWPTR	0x18UL	/* FIFO write ptr (shadow register) */
+#define ETX_FIFORPTR	0x1cUL	/* FIFO read ptr                    */
+#define ETX_FIFOSRPTR	0x20UL	/* FIFO read ptr (shadow register)  */
+#define ETX_FIFOPCNT	0x24UL	/* FIFO packet counter              */
+#define ETX_SMACHINE	0x28UL	/* Transmitter state machine        */
+#define ETX_RSIZE	0x2cUL	/* Ring descriptor size             */
+#define ETX_BPTR	0x30UL	/* Transmit data buffer ptr         */
+#define ETX_REG_SIZE	0x34UL
+
+/* ETX transmit pending register. */
+#define ETX_TP_DMAWAKEUP         0x00000001 /* Restart transmit dma             */
+
+/* ETX config register. */
+#define ETX_CFG_DMAENABLE        0x00000001 /* Enable transmit dma              */
+#define ETX_CFG_FIFOTHRESH       0x000003fe /* Transmit FIFO threshold          */
+#define ETX_CFG_IRQDAFTER        0x00000400 /* Interrupt after TX-FIFO drained  */
+#define ETX_CFG_IRQDBEFORE       0x00000000 /* Interrupt before TX-FIFO drained */
+
+#define ETX_RSIZE_SHIFT          4
+
+/* Happy Meal external receiver registers. */
+#define ERX_CFG		0x00UL	/* Receiver config register         */
+#define ERX_RING	0x04UL	/* Receiver ring ptr                */
+#define ERX_BPTR	0x08UL	/* Receiver buffer ptr              */
+#define ERX_FIFOWPTR	0x0cUL	/* FIFO write ptr                   */
+#define ERX_FIFOSWPTR	0x10UL	/* FIFO write ptr (shadow register) */
+#define ERX_FIFORPTR	0x14UL	/* FIFO read ptr                    */
+#define ERX_FIFOSRPTR	0x18UL	/* FIFO read ptr (shadow register)  */
+#define ERX_SMACHINE	0x1cUL	/* Receiver state machine           */
+#define ERX_REG_SIZE	0x20UL
+
+/* ERX config register. */
+#define ERX_CFG_DMAENABLE    0x00000001 /* Enable receive DMA        */
+#define ERX_CFG_RESV1        0x00000006 /* Unused...                 */
+#define ERX_CFG_BYTEOFFSET   0x00000038 /* Receive first byte offset */
+#define ERX_CFG_RESV2        0x000001c0 /* Unused...                 */
+#define ERX_CFG_SIZE32       0x00000000 /* Receive ring size == 32   */
+#define ERX_CFG_SIZE64       0x00000200 /* Receive ring size == 64   */
+#define ERX_CFG_SIZE128      0x00000400 /* Receive ring size == 128  */
+#define ERX_CFG_SIZE256      0x00000600 /* Receive ring size == 256  */
+#define ERX_CFG_RESV3        0x0000f800 /* Unused...                 */
+#define ERX_CFG_CSUMSTART    0x007f0000 /* Offset of checksum start,
+					 * in halfwords. */
+
+/* I'd like a Big Mac, small fries, small coke, and SparcLinux please. */
+#define BMAC_XIFCFG	0x0000UL	/* XIF config register                */
+	/* 0x4-->0x204, reserved */
+#define BMAC_TXSWRESET	0x208UL	/* Transmitter software reset         */
+#define BMAC_TXCFG	0x20cUL	/* Transmitter config register        */
+#define BMAC_IGAP1	0x210UL	/* Inter-packet gap 1                 */
+#define BMAC_IGAP2	0x214UL	/* Inter-packet gap 2                 */
+#define BMAC_ALIMIT	0x218UL	/* Transmit attempt limit             */
+#define BMAC_STIME	0x21cUL	/* Transmit slot time                 */
+#define BMAC_PLEN	0x220UL	/* Size of transmit preamble          */
+#define BMAC_PPAT	0x224UL	/* Pattern for transmit preamble      */
+#define BMAC_TXSDELIM	0x228UL	/* Transmit delimiter                 */
+#define BMAC_JSIZE	0x22cUL	/* Jam size                           */
+#define BMAC_TXMAX	0x230UL	/* Transmit max pkt size              */
+#define BMAC_TXMIN	0x234UL	/* Transmit min pkt size              */
+#define BMAC_PATTEMPT	0x238UL	/* Count of transmit peak attempts    */
+#define BMAC_DTCTR	0x23cUL	/* Transmit defer timer               */
+#define BMAC_NCCTR	0x240UL	/* Transmit normal-collision counter  */
+#define BMAC_FCCTR	0x244UL	/* Transmit first-collision counter   */
+#define BMAC_EXCTR	0x248UL	/* Transmit excess-collision counter  */
+#define BMAC_LTCTR	0x24cUL	/* Transmit late-collision counter    */
+#define BMAC_RSEED	0x250UL	/* Transmit random number seed        */
+#define BMAC_TXSMACHINE	0x254UL	/* Transmit state machine             */
+	/* 0x258-->0x304, reserved */
+#define BMAC_RXSWRESET	0x308UL	/* Receiver software reset            */
+#define BMAC_RXCFG	0x30cUL	/* Receiver config register           */
+#define BMAC_RXMAX	0x310UL	/* Receive max pkt size               */
+#define BMAC_RXMIN	0x314UL	/* Receive min pkt size               */
+#define BMAC_MACADDR2	0x318UL	/* Ether address register 2           */
+#define BMAC_MACADDR1	0x31cUL	/* Ether address register 1           */
+#define BMAC_MACADDR0	0x320UL	/* Ether address register 0           */
+#define BMAC_FRCTR	0x324UL	/* Receive frame receive counter      */
+#define BMAC_GLECTR	0x328UL	/* Receive giant-length error counter */
+#define BMAC_UNALECTR	0x32cUL	/* Receive unaligned error counter    */
+#define BMAC_RCRCECTR	0x330UL	/* Receive CRC error counter          */
+#define BMAC_RXSMACHINE	0x334UL	/* Receiver state machine             */
+#define BMAC_RXCVALID	0x338UL	/* Receiver code violation            */
+	/* 0x33c, reserved */
+#define BMAC_HTABLE3	0x340UL	/* Hash table 3                       */
+#define BMAC_HTABLE2	0x344UL	/* Hash table 2                       */
+#define BMAC_HTABLE1	0x348UL	/* Hash table 1                       */
+#define BMAC_HTABLE0	0x34cUL	/* Hash table 0                       */
+#define BMAC_AFILTER2	0x350UL	/* Address filter 2                   */
+#define BMAC_AFILTER1	0x354UL	/* Address filter 1                   */
+#define BMAC_AFILTER0	0x358UL	/* Address filter 0                   */
+#define BMAC_AFMASK	0x35cUL	/* Address filter mask                */
+#define BMAC_REG_SIZE	0x360UL
+
+/* BigMac XIF config register. */
+#define BIGMAC_XCFG_ODENABLE  0x00000001 /* Output driver enable         */
+#define BIGMAC_XCFG_XLBACK    0x00000002 /* Loopback-mode XIF enable     */
+#define BIGMAC_XCFG_MLBACK    0x00000004 /* Loopback-mode MII enable     */
+#define BIGMAC_XCFG_MIIDISAB  0x00000008 /* MII receive buffer disable   */
+#define BIGMAC_XCFG_SQENABLE  0x00000010 /* SQE test enable              */
+#define BIGMAC_XCFG_SQETWIN   0x000003e0 /* SQE time window              */
+#define BIGMAC_XCFG_LANCE     0x00000010 /* Lance mode enable            */
+#define BIGMAC_XCFG_LIPG0     0x000003e0 /* Lance mode IPG0              */
+
+/* BigMac transmit config register. */
+#define BIGMAC_TXCFG_ENABLE   0x00000001 /* Enable the transmitter       */
+#define BIGMAC_TXCFG_SMODE    0x00000020 /* Enable slow transmit mode    */
+#define BIGMAC_TXCFG_CIGN     0x00000040 /* Ignore transmit collisions   */
+#define BIGMAC_TXCFG_FCSOFF   0x00000080 /* Do not emit FCS              */
+#define BIGMAC_TXCFG_DBACKOFF 0x00000100 /* Disable backoff              */
+#define BIGMAC_TXCFG_FULLDPLX 0x00000200 /* Enable full-duplex           */
+#define BIGMAC_TXCFG_DGIVEUP  0x00000400 /* Don't give up on transmits   */
+
+/* BigMac receive config register. */
+#define BIGMAC_RXCFG_ENABLE   0x00000001 /* Enable the receiver             */
+#define BIGMAC_RXCFG_PSTRIP   0x00000020 /* Pad byte strip enable           */
+#define BIGMAC_RXCFG_PMISC    0x00000040 /* Enable promiscuous mode          */
+#define BIGMAC_RXCFG_DERR     0x00000080 /* Disable error checking          */
+#define BIGMAC_RXCFG_DCRCS    0x00000100 /* Disable CRC stripping           */
+#define BIGMAC_RXCFG_REJME    0x00000200 /* Reject packets addressed to me  */
+#define BIGMAC_RXCFG_PGRP     0x00000400 /* Enable promisc group mode       */
+#define BIGMAC_RXCFG_HENABLE  0x00000800 /* Enable the hash filter          */
+#define BIGMAC_RXCFG_AENABLE  0x00001000 /* Enable the address filter       */
+
+/* These are the "Management Interface" (ie. MIF) registers of the transceiver. */
+#define TCVR_BBCLOCK	0x00UL	/* Bit bang clock register          */
+#define TCVR_BBDATA	0x04UL	/* Bit bang data register           */
+#define TCVR_BBOENAB	0x08UL	/* Bit bang output enable           */
+#define TCVR_FRAME	0x0cUL	/* Frame control/data register      */
+#define TCVR_CFG	0x10UL	/* MIF config register              */
+#define TCVR_IMASK	0x14UL	/* MIF interrupt mask               */
+#define TCVR_STATUS	0x18UL	/* MIF status                       */
+#define TCVR_SMACHINE	0x1cUL	/* MIF state machine                */
+#define TCVR_REG_SIZE	0x20UL
+
+/* Frame commands. */
+#define FRAME_WRITE           0x50020000
+#define FRAME_READ            0x60020000
+
+/* Transceiver config register */
+#define TCV_CFG_PSELECT       0x00000001 /* Select PHY                      */
+#define TCV_CFG_PENABLE       0x00000002 /* Enable MIF polling              */
+#define TCV_CFG_BENABLE       0x00000004 /* Enable the "bit banger" oh baby */
+#define TCV_CFG_PREGADDR      0x000000f8 /* Address of poll register        */
+#define TCV_CFG_MDIO0         0x00000100 /* MDIO zero, data/attached        */
+#define TCV_CFG_MDIO1         0x00000200 /* MDIO one,  data/attached        */
+#define TCV_CFG_PDADDR        0x00007c00 /* Device PHY address polling      */
+
+/* Here are some PHY addresses. */
+#define TCV_PADDR_ETX         0          /* Internal transceiver            */
+#define TCV_PADDR_ITX         1          /* External transceiver            */
+
+/* Transceiver status register */
+#define TCV_STAT_BASIC        0xffff0000 /* The "basic" part                */
+#define TCV_STAT_NORMAL       0x0000ffff /* The "non-basic" part            */
+
+/* Inside the Happy Meal transceiver is the physical layer, they use an
+ * implementations for National Semiconductor, part number DP83840VCE.
+ * You can retrieve the data sheets and programming docs for this beast
+ * from http://www.national.com/
+ *
+ * The DP83840 is capable of both 10 and 100Mbps ethernet, in both
+ * half and full duplex mode.  It also supports auto negotiation.
+ *
+ * But.... THIS THING IS A PAIN IN THE ASS TO PROGRAM!
+ * Debugging eeprom burnt code is more fun than programming this chip!
+ */
+
+/* Generic MII registers defined in linux/mii.h, these below
+ * are DP83840 specific.
+ */
+#define DP83840_CSCONFIG        0x17        /* CS configuration            */
+
+/* The Carrier Sense config register. */
+#define CSCONFIG_RESV1          0x0001  /* Unused...                   */
+#define CSCONFIG_LED4           0x0002  /* Pin for full-dplx LED4      */
+#define CSCONFIG_LED1           0x0004  /* Pin for conn-status LED1    */
+#define CSCONFIG_RESV2          0x0008  /* Unused...                   */
+#define CSCONFIG_TCVDISAB       0x0010  /* Turns off the transceiver   */
+#define CSCONFIG_DFBYPASS       0x0020  /* Bypass disconnect function  */
+#define CSCONFIG_GLFORCE        0x0040  /* Good link force for 100mbps */
+#define CSCONFIG_CLKTRISTATE    0x0080  /* Tristate 25m clock          */
+#define CSCONFIG_RESV3          0x0700  /* Unused...                   */
+#define CSCONFIG_ENCODE         0x0800  /* 1=MLT-3, 0=binary           */
+#define CSCONFIG_RENABLE        0x1000  /* Repeater mode enable        */
+#define CSCONFIG_TCDISABLE      0x2000  /* Disable timeout counter     */
+#define CSCONFIG_RESV4          0x4000  /* Unused...                   */
+#define CSCONFIG_NDISABLE       0x8000  /* Disable NRZI                */
+
+/* Happy Meal descriptor rings and such.
+ * All descriptor rings must be aligned on a 2K boundary.
+ * All receive buffers must be 64 byte aligned.
+ * Always write the address first before setting the ownership
+ * bits to avoid races with the hardware scanning the ring.
+ */
+typedef u32 __bitwise hme32;
+
+struct happy_meal_rxd {
+	hme32 rx_flags;
+	hme32 rx_addr;
+};
+
+#define RXFLAG_OWN         0x80000000 /* 1 = hardware, 0 = software */
+#define RXFLAG_OVERFLOW    0x40000000 /* 1 = buffer overflow        */
+#define RXFLAG_SIZE        0x3fff0000 /* Size of the buffer         */
+#define RXFLAG_CSUM        0x0000ffff /* HW computed checksum       */
+
+struct happy_meal_txd {
+	hme32 tx_flags;
+	hme32 tx_addr;
+};
+
+#define TXFLAG_OWN         0x80000000 /* 1 = hardware, 0 = software */
+#define TXFLAG_SOP         0x40000000 /* 1 = start of packet        */
+#define TXFLAG_EOP         0x20000000 /* 1 = end of packet          */
+#define TXFLAG_CSENABLE    0x10000000 /* 1 = enable hw-checksums    */
+#define TXFLAG_CSLOCATION  0x0ff00000 /* Where to stick the csum    */
+#define TXFLAG_CSBUFBEGIN  0x000fc000 /* Where to begin checksum    */
+#define TXFLAG_SIZE        0x00003fff /* Size of the packet         */
+
+#define TX_RING_SIZE       32         /* Must be >16 and <255, multiple of 16  */
+#define RX_RING_SIZE       32         /* see ERX_CFG_SIZE* for possible values */
+
+#if (TX_RING_SIZE < 16 || TX_RING_SIZE > 256 || (TX_RING_SIZE % 16) != 0)
+#error TX_RING_SIZE holds illegal value
+#endif
+
+#define TX_RING_MAXSIZE    256
+#define RX_RING_MAXSIZE    256
+
+/* We use a 14 byte offset for checksum computation. */
+#if (RX_RING_SIZE == 32)
+#define ERX_CFG_DEFAULT(off) (ERX_CFG_DMAENABLE|((off)<<3)|ERX_CFG_SIZE32|((14/2)<<16))
+#else
+#if (RX_RING_SIZE == 64)
+#define ERX_CFG_DEFAULT(off) (ERX_CFG_DMAENABLE|((off)<<3)|ERX_CFG_SIZE64|((14/2)<<16))
+#else
+#if (RX_RING_SIZE == 128)
+#define ERX_CFG_DEFAULT(off) (ERX_CFG_DMAENABLE|((off)<<3)|ERX_CFG_SIZE128|((14/2)<<16))
+#else
+#if (RX_RING_SIZE == 256)
+#define ERX_CFG_DEFAULT(off) (ERX_CFG_DMAENABLE|((off)<<3)|ERX_CFG_SIZE256|((14/2)<<16))
+#else
+#error RX_RING_SIZE holds illegal value
+#endif
+#endif
+#endif
+#endif
+
+#define NEXT_RX(num)       (((num) + 1) & (RX_RING_SIZE - 1))
+#define NEXT_TX(num)       (((num) + 1) & (TX_RING_SIZE - 1))
+#define PREV_RX(num)       (((num) - 1) & (RX_RING_SIZE - 1))
+#define PREV_TX(num)       (((num) - 1) & (TX_RING_SIZE - 1))
+
+#define TX_BUFFS_AVAIL(hp)                                    \
+        (((hp)->tx_old <= (hp)->tx_new) ?                     \
+	  (hp)->tx_old + (TX_RING_SIZE - 1) - (hp)->tx_new :  \
+			    (hp)->tx_old - (hp)->tx_new - 1)
+
+#define RX_OFFSET          2
+#define RX_BUF_ALLOC_SIZE  (1546 + RX_OFFSET + 64)
+
+#define RX_COPY_THRESHOLD  256
+
+struct hmeal_init_block {
+	struct happy_meal_rxd happy_meal_rxd[RX_RING_MAXSIZE];
+	struct happy_meal_txd happy_meal_txd[TX_RING_MAXSIZE];
+};
+
+#define hblock_offset(mem, elem) \
+((__u32)((unsigned long)(&(((struct hmeal_init_block *)0)->mem[elem]))))
+
+/* Now software state stuff. */
+enum happy_transceiver {
+	external = 0,
+	internal = 1,
+	none     = 2,
+};
+
+/* Timer state engine. */
+enum happy_timer_state {
+	arbwait  = 0,  /* Waiting for auto negotiation to complete.          */
+	lupwait  = 1,  /* Auto-neg complete, awaiting link-up status.        */
+	ltrywait = 2,  /* Forcing try of all modes, from fastest to slowest. */
+	asleep   = 3,  /* Time inactive.                                     */
+};
+
+struct quattro;
+
+/* Happy happy, joy joy! */
+struct happy_meal {
+	void __iomem	*gregs;			/* Happy meal global registers       */
+	struct hmeal_init_block  *happy_block;	/* RX and TX descriptors (CPU addr)  */
+
+#if defined(CONFIG_SBUS) && defined(CONFIG_PCI)
+	u32 (*read_desc32)(hme32 *);
+	void (*write_txd)(struct happy_meal_txd *, u32, u32);
+	void (*write_rxd)(struct happy_meal_rxd *, u32, u32);
+#endif
+
+	/* This is either an platform_device or a pci_dev. */
+	void			  *happy_dev;
+	struct device		  *dma_dev;
+
+	spinlock_t		  happy_lock;
+
+	struct sk_buff           *rx_skbs[RX_RING_SIZE];
+	struct sk_buff           *tx_skbs[TX_RING_SIZE];
+
+	int rx_new, tx_new, rx_old, tx_old;
+
+#if defined(CONFIG_SBUS) && defined(CONFIG_PCI)
+	u32 (*read32)(void __iomem *);
+	void (*write32)(void __iomem *, u32);
+#endif
+
+	void __iomem	*etxregs;        /* External transmitter regs        */
+	void __iomem	*erxregs;        /* External receiver regs           */
+	void __iomem	*bigmacregs;     /* BIGMAC core regs		     */
+	void __iomem	*tcvregs;        /* MIF transceiver regs             */
+
+	dma_addr_t                hblock_dvma;    /* DVMA visible address happy block  */
+	unsigned int              happy_flags;    /* Driver state flags                */
+	int                       irq;
+	enum happy_transceiver    tcvr_type;      /* Kind of transceiver in use        */
+	unsigned int              happy_bursts;   /* Get your mind out of the gutter   */
+	unsigned int              paddr;          /* PHY address for transceiver       */
+	unsigned short            hm_revision;    /* Happy meal revision               */
+	unsigned short            sw_bmcr;        /* SW copy of BMCR                   */
+	unsigned short            sw_bmsr;        /* SW copy of BMSR                   */
+	unsigned short            sw_physid1;     /* SW copy of PHYSID1                */
+	unsigned short            sw_physid2;     /* SW copy of PHYSID2                */
+	unsigned short            sw_advertise;   /* SW copy of ADVERTISE              */
+	unsigned short            sw_lpa;         /* SW copy of LPA                    */
+	unsigned short            sw_expansion;   /* SW copy of EXPANSION              */
+	unsigned short            sw_csconfig;    /* SW copy of CSCONFIG               */
+	unsigned int              auto_speed;     /* Auto-nego link speed              */
+        unsigned int              forced_speed;   /* Force mode link speed             */
+	unsigned int              poll_data;      /* MIF poll data                     */
+	unsigned int              poll_flag;      /* MIF poll flag                     */
+	unsigned int              linkcheck;      /* Have we checked the link yet?     */
+	unsigned int              lnkup;          /* Is the link up as far as we know? */
+	unsigned int              lnkdown;        /* Trying to force the link down?    */
+	unsigned int              lnkcnt;         /* Counter for link-up attempts.     */
+	struct timer_list         happy_timer;    /* To watch the link when coming up. */
+	enum happy_timer_state    timer_state;    /* State of the auto-neg timer.      */
+	unsigned int              timer_ticks;    /* Number of clicks at each state.   */
+
+	struct net_device	 *dev;		/* Backpointer                       */
+	struct quattro		 *qfe_parent;	/* For Quattro cards                 */
+	int			  qfe_ent;	/* Which instance on quattro         */
+};
+
+/* Here are the happy flags. */
+#define HFLAG_FENABLE             0x00000002      /* The MII frame is enabled          */
+#define HFLAG_LANCE               0x00000004      /* We are using lance-mode           */
+#define HFLAG_RXENABLE            0x00000008      /* Receiver is enabled               */
+#define HFLAG_AUTO                0x00000010      /* Using auto-negotiation, 0 = force */
+#define HFLAG_FULL                0x00000020      /* Full duplex enable                */
+#define HFLAG_MACFULL             0x00000040      /* Using full duplex in the MAC      */
+#define HFLAG_RXCV                0x00000100      /* XXX RXCV ENABLE                   */
+#define HFLAG_INIT                0x00000200      /* Init called at least once         */
+#define HFLAG_LINKUP              0x00000400      /* 1 = Link is up                    */
+#define HFLAG_PCI                 0x00000800      /* PCI based Happy Meal              */
+#define HFLAG_QUATTRO		  0x00001000      /* On QFE/Quattro card	       */
+
+#define HFLAG_20_21  HFLAG_FENABLE
+#define HFLAG_NOT_A0 (HFLAG_FENABLE | HFLAG_LANCE | HFLAG_RXCV)
+
+/* Support for QFE/Quattro cards. */
+struct quattro {
+	struct net_device	*happy_meals[4];
+
+	/* This is either a sbus_dev or a pci_dev. */
+	void			*quattro_dev;
+
+	struct quattro		*next;
+
+	/* PROM ranges, if any. */
+#ifdef CONFIG_SBUS
+	struct linux_prom_ranges  ranges[8];
+#endif
+	int			  nranges;
+};
+
+/* We use this to acquire receive skb's that we can DMA directly into. */
+#define ALIGNED_RX_SKB_ADDR(addr) \
+        ((((unsigned long)(addr) + (64UL - 1UL)) & ~(64UL - 1UL)) - (unsigned long)(addr))
+#define happy_meal_alloc_skb(__length, __gfp_flags) \
+({	struct sk_buff *__skb; \
+	__skb = alloc_skb((__length) + 64, (__gfp_flags)); \
+	if(__skb) { \
+		int __offset = (int) ALIGNED_RX_SKB_ADDR(__skb->data); \
+		if(__offset) \
+			skb_reserve(__skb, __offset); \
+	} \
+	__skb; \
+})
+
+#endif /* !(_SUNHME_H) */
diff --git a/drivers/net/ethernet/sun/sunqe.c b/drivers/net/ethernet/sun/sunqe.c
new file mode 100644
index 000000000..acef22cb9
--- /dev/null
+++ b/drivers/net/ethernet/sun/sunqe.c
@@ -0,0 +1,994 @@
+// SPDX-License-Identifier: GPL-2.0
+/* sunqe.c: Sparc QuadEthernet 10baseT SBUS card driver.
+ *          Once again I am out to prove that every ethernet
+ *          controller out there can be most efficiently programmed
+ *          if you make it look like a LANCE.
+ *
+ * Copyright (C) 1996, 1999, 2003, 2006, 2008 David S. Miller (davem@davemloft.net)
+ */
+
+#include <linux/module.h>
+#include <linux/kernel.h>
+#include <linux/types.h>
+#include <linux/errno.h>
+#include <linux/fcntl.h>
+#include <linux/interrupt.h>
+#include <linux/ioport.h>
+#include <linux/in.h>
+#include <linux/slab.h>
+#include <linux/string.h>
+#include <linux/delay.h>
+#include <linux/init.h>
+#include <linux/crc32.h>
+#include <linux/netdevice.h>
+#include <linux/etherdevice.h>
+#include <linux/skbuff.h>
+#include <linux/ethtool.h>
+#include <linux/bitops.h>
+#include <linux/dma-mapping.h>
+#include <linux/of.h>
+#include <linux/pgtable.h>
+#include <linux/platform_device.h>
+
+#include <asm/io.h>
+#include <asm/dma.h>
+#include <asm/byteorder.h>
+#include <asm/idprom.h>
+#include <asm/openprom.h>
+#include <asm/oplib.h>
+#include <asm/auxio.h>
+#include <asm/irq.h>
+
+#include "sunqe.h"
+
+#define DRV_NAME	"sunqe"
+#define DRV_VERSION	"4.1"
+#define DRV_RELDATE	"August 27, 2008"
+#define DRV_AUTHOR	"David S. Miller (davem@davemloft.net)"
+
+static char version[] =
+	DRV_NAME ".c:v" DRV_VERSION " " DRV_RELDATE " " DRV_AUTHOR "\n";
+
+MODULE_VERSION(DRV_VERSION);
+MODULE_AUTHOR(DRV_AUTHOR);
+MODULE_DESCRIPTION("Sun QuadEthernet 10baseT SBUS card driver");
+MODULE_LICENSE("GPL");
+
+static struct sunqec *root_qec_dev;
+
+static void qe_set_multicast(struct net_device *dev);
+
+#define QEC_RESET_TRIES 200
+
+static inline int qec_global_reset(void __iomem *gregs)
+{
+	int tries = QEC_RESET_TRIES;
+
+	sbus_writel(GLOB_CTRL_RESET, gregs + GLOB_CTRL);
+	while (--tries) {
+		u32 tmp = sbus_readl(gregs + GLOB_CTRL);
+		if (tmp & GLOB_CTRL_RESET) {
+			udelay(20);
+			continue;
+		}
+		break;
+	}
+	if (tries)
+		return 0;
+	printk(KERN_ERR "QuadEther: AIEEE cannot reset the QEC!\n");
+	return -1;
+}
+
+#define MACE_RESET_RETRIES 200
+#define QE_RESET_RETRIES   200
+
+static inline int qe_stop(struct sunqe *qep)
+{
+	void __iomem *cregs = qep->qcregs;
+	void __iomem *mregs = qep->mregs;
+	int tries;
+
+	/* Reset the MACE, then the QEC channel. */
+	sbus_writeb(MREGS_BCONFIG_RESET, mregs + MREGS_BCONFIG);
+	tries = MACE_RESET_RETRIES;
+	while (--tries) {
+		u8 tmp = sbus_readb(mregs + MREGS_BCONFIG);
+		if (tmp & MREGS_BCONFIG_RESET) {
+			udelay(20);
+			continue;
+		}
+		break;
+	}
+	if (!tries) {
+		printk(KERN_ERR "QuadEther: AIEEE cannot reset the MACE!\n");
+		return -1;
+	}
+
+	sbus_writel(CREG_CTRL_RESET, cregs + CREG_CTRL);
+	tries = QE_RESET_RETRIES;
+	while (--tries) {
+		u32 tmp = sbus_readl(cregs + CREG_CTRL);
+		if (tmp & CREG_CTRL_RESET) {
+			udelay(20);
+			continue;
+		}
+		break;
+	}
+	if (!tries) {
+		printk(KERN_ERR "QuadEther: Cannot reset QE channel!\n");
+		return -1;
+	}
+	return 0;
+}
+
+static void qe_init_rings(struct sunqe *qep)
+{
+	struct qe_init_block *qb = qep->qe_block;
+	struct sunqe_buffers *qbufs = qep->buffers;
+	__u32 qbufs_dvma = (__u32)qep->buffers_dvma;
+	int i;
+
+	qep->rx_new = qep->rx_old = qep->tx_new = qep->tx_old = 0;
+	memset(qb, 0, sizeof(struct qe_init_block));
+	memset(qbufs, 0, sizeof(struct sunqe_buffers));
+	for (i = 0; i < RX_RING_SIZE; i++) {
+		qb->qe_rxd[i].rx_addr = qbufs_dvma + qebuf_offset(rx_buf, i);
+		qb->qe_rxd[i].rx_flags =
+			(RXD_OWN | ((RXD_PKT_SZ) & RXD_LENGTH));
+	}
+}
+
+static int qe_init(struct sunqe *qep, int from_irq)
+{
+	struct sunqec *qecp = qep->parent;
+	void __iomem *cregs = qep->qcregs;
+	void __iomem *mregs = qep->mregs;
+	void __iomem *gregs = qecp->gregs;
+	const unsigned char *e = &qep->dev->dev_addr[0];
+	__u32 qblk_dvma = (__u32)qep->qblock_dvma;
+	u32 tmp;
+	int i;
+
+	/* Shut it up. */
+	if (qe_stop(qep))
+		return -EAGAIN;
+
+	/* Setup initial rx/tx init block pointers. */
+	sbus_writel(qblk_dvma + qib_offset(qe_rxd, 0), cregs + CREG_RXDS);
+	sbus_writel(qblk_dvma + qib_offset(qe_txd, 0), cregs + CREG_TXDS);
+
+	/* Enable/mask the various irq's. */
+	sbus_writel(0, cregs + CREG_RIMASK);
+	sbus_writel(1, cregs + CREG_TIMASK);
+
+	sbus_writel(0, cregs + CREG_QMASK);
+	sbus_writel(CREG_MMASK_RXCOLL, cregs + CREG_MMASK);
+
+	/* Setup the FIFO pointers into QEC local memory. */
+	tmp = qep->channel * sbus_readl(gregs + GLOB_MSIZE);
+	sbus_writel(tmp, cregs + CREG_RXRBUFPTR);
+	sbus_writel(tmp, cregs + CREG_RXWBUFPTR);
+
+	tmp = sbus_readl(cregs + CREG_RXRBUFPTR) +
+		sbus_readl(gregs + GLOB_RSIZE);
+	sbus_writel(tmp, cregs + CREG_TXRBUFPTR);
+	sbus_writel(tmp, cregs + CREG_TXWBUFPTR);
+
+	/* Clear the channel collision counter. */
+	sbus_writel(0, cregs + CREG_CCNT);
+
+	/* For 10baseT, inter frame space nor throttle seems to be necessary. */
+	sbus_writel(0, cregs + CREG_PIPG);
+
+	/* Now dork with the AMD MACE. */
+	sbus_writeb(MREGS_PHYCONFIG_AUTO, mregs + MREGS_PHYCONFIG);
+	sbus_writeb(MREGS_TXFCNTL_AUTOPAD, mregs + MREGS_TXFCNTL);
+	sbus_writeb(0, mregs + MREGS_RXFCNTL);
+
+	/* The QEC dma's the rx'd packets from local memory out to main memory,
+	 * and therefore it interrupts when the packet reception is "complete".
+	 * So don't listen for the MACE talking about it.
+	 */
+	sbus_writeb(MREGS_IMASK_COLL | MREGS_IMASK_RXIRQ, mregs + MREGS_IMASK);
+	sbus_writeb(MREGS_BCONFIG_BSWAP | MREGS_BCONFIG_64TS, mregs + MREGS_BCONFIG);
+	sbus_writeb((MREGS_FCONFIG_TXF16 | MREGS_FCONFIG_RXF32 |
+		     MREGS_FCONFIG_RFWU | MREGS_FCONFIG_TFWU),
+		    mregs + MREGS_FCONFIG);
+
+	/* Only usable interface on QuadEther is twisted pair. */
+	sbus_writeb(MREGS_PLSCONFIG_TP, mregs + MREGS_PLSCONFIG);
+
+	/* Tell MACE we are changing the ether address. */
+	sbus_writeb(MREGS_IACONFIG_ACHNGE | MREGS_IACONFIG_PARESET,
+		    mregs + MREGS_IACONFIG);
+	while ((sbus_readb(mregs + MREGS_IACONFIG) & MREGS_IACONFIG_ACHNGE) != 0)
+		barrier();
+	sbus_writeb(e[0], mregs + MREGS_ETHADDR);
+	sbus_writeb(e[1], mregs + MREGS_ETHADDR);
+	sbus_writeb(e[2], mregs + MREGS_ETHADDR);
+	sbus_writeb(e[3], mregs + MREGS_ETHADDR);
+	sbus_writeb(e[4], mregs + MREGS_ETHADDR);
+	sbus_writeb(e[5], mregs + MREGS_ETHADDR);
+
+	/* Clear out the address filter. */
+	sbus_writeb(MREGS_IACONFIG_ACHNGE | MREGS_IACONFIG_LARESET,
+		    mregs + MREGS_IACONFIG);
+	while ((sbus_readb(mregs + MREGS_IACONFIG) & MREGS_IACONFIG_ACHNGE) != 0)
+		barrier();
+	for (i = 0; i < 8; i++)
+		sbus_writeb(0, mregs + MREGS_FILTER);
+
+	/* Address changes are now complete. */
+	sbus_writeb(0, mregs + MREGS_IACONFIG);
+
+	qe_init_rings(qep);
+
+	/* Wait a little bit for the link to come up... */
+	mdelay(5);
+	if (!(sbus_readb(mregs + MREGS_PHYCONFIG) & MREGS_PHYCONFIG_LTESTDIS)) {
+		int tries = 50;
+
+		while (--tries) {
+			u8 tmp;
+
+			mdelay(5);
+			barrier();
+			tmp = sbus_readb(mregs + MREGS_PHYCONFIG);
+			if ((tmp & MREGS_PHYCONFIG_LSTAT) != 0)
+				break;
+		}
+		if (tries == 0)
+			printk(KERN_NOTICE "%s: Warning, link state is down.\n", qep->dev->name);
+	}
+
+	/* Missed packet counter is cleared on a read. */
+	sbus_readb(mregs + MREGS_MPCNT);
+
+	/* Reload multicast information, this will enable the receiver
+	 * and transmitter.
+	 */
+	qe_set_multicast(qep->dev);
+
+	/* QEC should now start to show interrupts. */
+	return 0;
+}
+
+/* Grrr, certain error conditions completely lock up the AMD MACE,
+ * so when we get these we _must_ reset the chip.
+ */
+static int qe_is_bolixed(struct sunqe *qep, u32 qe_status)
+{
+	struct net_device *dev = qep->dev;
+	int mace_hwbug_workaround = 0;
+
+	if (qe_status & CREG_STAT_EDEFER) {
+		printk(KERN_ERR "%s: Excessive transmit defers.\n", dev->name);
+		dev->stats.tx_errors++;
+	}
+
+	if (qe_status & CREG_STAT_CLOSS) {
+		printk(KERN_ERR "%s: Carrier lost, link down?\n", dev->name);
+		dev->stats.tx_errors++;
+		dev->stats.tx_carrier_errors++;
+	}
+
+	if (qe_status & CREG_STAT_ERETRIES) {
+		printk(KERN_ERR "%s: Excessive transmit retries (more than 16).\n", dev->name);
+		dev->stats.tx_errors++;
+		mace_hwbug_workaround = 1;
+	}
+
+	if (qe_status & CREG_STAT_LCOLL) {
+		printk(KERN_ERR "%s: Late transmit collision.\n", dev->name);
+		dev->stats.tx_errors++;
+		dev->stats.collisions++;
+		mace_hwbug_workaround = 1;
+	}
+
+	if (qe_status & CREG_STAT_FUFLOW) {
+		printk(KERN_ERR "%s: Transmit fifo underflow, driver bug.\n", dev->name);
+		dev->stats.tx_errors++;
+		mace_hwbug_workaround = 1;
+	}
+
+	if (qe_status & CREG_STAT_JERROR) {
+		printk(KERN_ERR "%s: Jabber error.\n", dev->name);
+	}
+
+	if (qe_status & CREG_STAT_BERROR) {
+		printk(KERN_ERR "%s: Babble error.\n", dev->name);
+	}
+
+	if (qe_status & CREG_STAT_CCOFLOW) {
+		dev->stats.tx_errors += 256;
+		dev->stats.collisions += 256;
+	}
+
+	if (qe_status & CREG_STAT_TXDERROR) {
+		printk(KERN_ERR "%s: Transmit descriptor is bogus, driver bug.\n", dev->name);
+		dev->stats.tx_errors++;
+		dev->stats.tx_aborted_errors++;
+		mace_hwbug_workaround = 1;
+	}
+
+	if (qe_status & CREG_STAT_TXLERR) {
+		printk(KERN_ERR "%s: Transmit late error.\n", dev->name);
+		dev->stats.tx_errors++;
+		mace_hwbug_workaround = 1;
+	}
+
+	if (qe_status & CREG_STAT_TXPERR) {
+		printk(KERN_ERR "%s: Transmit DMA parity error.\n", dev->name);
+		dev->stats.tx_errors++;
+		dev->stats.tx_aborted_errors++;
+		mace_hwbug_workaround = 1;
+	}
+
+	if (qe_status & CREG_STAT_TXSERR) {
+		printk(KERN_ERR "%s: Transmit DMA sbus error ack.\n", dev->name);
+		dev->stats.tx_errors++;
+		dev->stats.tx_aborted_errors++;
+		mace_hwbug_workaround = 1;
+	}
+
+	if (qe_status & CREG_STAT_RCCOFLOW) {
+		dev->stats.rx_errors += 256;
+		dev->stats.collisions += 256;
+	}
+
+	if (qe_status & CREG_STAT_RUOFLOW) {
+		dev->stats.rx_errors += 256;
+		dev->stats.rx_over_errors += 256;
+	}
+
+	if (qe_status & CREG_STAT_MCOFLOW) {
+		dev->stats.rx_errors += 256;
+		dev->stats.rx_missed_errors += 256;
+	}
+
+	if (qe_status & CREG_STAT_RXFOFLOW) {
+		printk(KERN_ERR "%s: Receive fifo overflow.\n", dev->name);
+		dev->stats.rx_errors++;
+		dev->stats.rx_over_errors++;
+	}
+
+	if (qe_status & CREG_STAT_RLCOLL) {
+		printk(KERN_ERR "%s: Late receive collision.\n", dev->name);
+		dev->stats.rx_errors++;
+		dev->stats.collisions++;
+	}
+
+	if (qe_status & CREG_STAT_FCOFLOW) {
+		dev->stats.rx_errors += 256;
+		dev->stats.rx_frame_errors += 256;
+	}
+
+	if (qe_status & CREG_STAT_CECOFLOW) {
+		dev->stats.rx_errors += 256;
+		dev->stats.rx_crc_errors += 256;
+	}
+
+	if (qe_status & CREG_STAT_RXDROP) {
+		printk(KERN_ERR "%s: Receive packet dropped.\n", dev->name);
+		dev->stats.rx_errors++;
+		dev->stats.rx_dropped++;
+		dev->stats.rx_missed_errors++;
+	}
+
+	if (qe_status & CREG_STAT_RXSMALL) {
+		printk(KERN_ERR "%s: Receive buffer too small, driver bug.\n", dev->name);
+		dev->stats.rx_errors++;
+		dev->stats.rx_length_errors++;
+	}
+
+	if (qe_status & CREG_STAT_RXLERR) {
+		printk(KERN_ERR "%s: Receive late error.\n", dev->name);
+		dev->stats.rx_errors++;
+		mace_hwbug_workaround = 1;
+	}
+
+	if (qe_status & CREG_STAT_RXPERR) {
+		printk(KERN_ERR "%s: Receive DMA parity error.\n", dev->name);
+		dev->stats.rx_errors++;
+		dev->stats.rx_missed_errors++;
+		mace_hwbug_workaround = 1;
+	}
+
+	if (qe_status & CREG_STAT_RXSERR) {
+		printk(KERN_ERR "%s: Receive DMA sbus error ack.\n", dev->name);
+		dev->stats.rx_errors++;
+		dev->stats.rx_missed_errors++;
+		mace_hwbug_workaround = 1;
+	}
+
+	if (mace_hwbug_workaround)
+		qe_init(qep, 1);
+	return mace_hwbug_workaround;
+}
+
+/* Per-QE receive interrupt service routine.  Just like on the happy meal
+ * we receive directly into skb's with a small packet copy water mark.
+ */
+static void qe_rx(struct sunqe *qep)
+{
+	struct qe_rxd *rxbase = &qep->qe_block->qe_rxd[0];
+	struct net_device *dev = qep->dev;
+	struct qe_rxd *this;
+	struct sunqe_buffers *qbufs = qep->buffers;
+	__u32 qbufs_dvma = (__u32)qep->buffers_dvma;
+	int elem = qep->rx_new;
+	u32 flags;
+
+	this = &rxbase[elem];
+	while (!((flags = this->rx_flags) & RXD_OWN)) {
+		struct sk_buff *skb;
+		unsigned char *this_qbuf =
+			&qbufs->rx_buf[elem & (RX_RING_SIZE - 1)][0];
+		__u32 this_qbuf_dvma = qbufs_dvma +
+			qebuf_offset(rx_buf, (elem & (RX_RING_SIZE - 1)));
+		struct qe_rxd *end_rxd =
+			&rxbase[(elem+RX_RING_SIZE)&(RX_RING_MAXSIZE-1)];
+		int len = (flags & RXD_LENGTH) - 4;  /* QE adds ether FCS size to len */
+
+		/* Check for errors. */
+		if (len < ETH_ZLEN) {
+			dev->stats.rx_errors++;
+			dev->stats.rx_length_errors++;
+			dev->stats.rx_dropped++;
+		} else {
+			skb = netdev_alloc_skb(dev, len + 2);
+			if (skb == NULL) {
+				dev->stats.rx_dropped++;
+			} else {
+				skb_reserve(skb, 2);
+				skb_put(skb, len);
+				skb_copy_to_linear_data(skb, this_qbuf,
+						 len);
+				skb->protocol = eth_type_trans(skb, qep->dev);
+				netif_rx(skb);
+				dev->stats.rx_packets++;
+				dev->stats.rx_bytes += len;
+			}
+		}
+		end_rxd->rx_addr = this_qbuf_dvma;
+		end_rxd->rx_flags = (RXD_OWN | ((RXD_PKT_SZ) & RXD_LENGTH));
+
+		elem = NEXT_RX(elem);
+		this = &rxbase[elem];
+	}
+	qep->rx_new = elem;
+}
+
+static void qe_tx_reclaim(struct sunqe *qep);
+
+/* Interrupts for all QE's get filtered out via the QEC master controller,
+ * so we just run through each qe and check to see who is signaling
+ * and thus needs to be serviced.
+ */
+static irqreturn_t qec_interrupt(int irq, void *dev_id)
+{
+	struct sunqec *qecp = dev_id;
+	u32 qec_status;
+	int channel = 0;
+
+	/* Latch the status now. */
+	qec_status = sbus_readl(qecp->gregs + GLOB_STAT);
+	while (channel < 4) {
+		if (qec_status & 0xf) {
+			struct sunqe *qep = qecp->qes[channel];
+			u32 qe_status;
+
+			qe_status = sbus_readl(qep->qcregs + CREG_STAT);
+			if (qe_status & CREG_STAT_ERRORS) {
+				if (qe_is_bolixed(qep, qe_status))
+					goto next;
+			}
+			if (qe_status & CREG_STAT_RXIRQ)
+				qe_rx(qep);
+			if (netif_queue_stopped(qep->dev) &&
+			    (qe_status & CREG_STAT_TXIRQ)) {
+				spin_lock(&qep->lock);
+				qe_tx_reclaim(qep);
+				if (TX_BUFFS_AVAIL(qep) > 0) {
+					/* Wake net queue and return to
+					 * lazy tx reclaim.
+					 */
+					netif_wake_queue(qep->dev);
+					sbus_writel(1, qep->qcregs + CREG_TIMASK);
+				}
+				spin_unlock(&qep->lock);
+			}
+	next:
+			;
+		}
+		qec_status >>= 4;
+		channel++;
+	}
+
+	return IRQ_HANDLED;
+}
+
+static int qe_open(struct net_device *dev)
+{
+	struct sunqe *qep = netdev_priv(dev);
+
+	qep->mconfig = (MREGS_MCONFIG_TXENAB |
+			MREGS_MCONFIG_RXENAB |
+			MREGS_MCONFIG_MBAENAB);
+	return qe_init(qep, 0);
+}
+
+static int qe_close(struct net_device *dev)
+{
+	struct sunqe *qep = netdev_priv(dev);
+
+	qe_stop(qep);
+	return 0;
+}
+
+/* Reclaim TX'd frames from the ring.  This must always run under
+ * the IRQ protected qep->lock.
+ */
+static void qe_tx_reclaim(struct sunqe *qep)
+{
+	struct qe_txd *txbase = &qep->qe_block->qe_txd[0];
+	int elem = qep->tx_old;
+
+	while (elem != qep->tx_new) {
+		u32 flags = txbase[elem].tx_flags;
+
+		if (flags & TXD_OWN)
+			break;
+		elem = NEXT_TX(elem);
+	}
+	qep->tx_old = elem;
+}
+
+static void qe_tx_timeout(struct net_device *dev, unsigned int txqueue)
+{
+	struct sunqe *qep = netdev_priv(dev);
+	int tx_full;
+
+	spin_lock_irq(&qep->lock);
+
+	/* Try to reclaim, if that frees up some tx
+	 * entries, we're fine.
+	 */
+	qe_tx_reclaim(qep);
+	tx_full = TX_BUFFS_AVAIL(qep) <= 0;
+
+	spin_unlock_irq(&qep->lock);
+
+	if (! tx_full)
+		goto out;
+
+	printk(KERN_ERR "%s: transmit timed out, resetting\n", dev->name);
+	qe_init(qep, 1);
+
+out:
+	netif_wake_queue(dev);
+}
+
+/* Get a packet queued to go onto the wire. */
+static netdev_tx_t qe_start_xmit(struct sk_buff *skb, struct net_device *dev)
+{
+	struct sunqe *qep = netdev_priv(dev);
+	struct sunqe_buffers *qbufs = qep->buffers;
+	__u32 txbuf_dvma, qbufs_dvma = (__u32)qep->buffers_dvma;
+	unsigned char *txbuf;
+	int len, entry;
+
+	spin_lock_irq(&qep->lock);
+
+	qe_tx_reclaim(qep);
+
+	len = skb->len;
+	entry = qep->tx_new;
+
+	txbuf = &qbufs->tx_buf[entry & (TX_RING_SIZE - 1)][0];
+	txbuf_dvma = qbufs_dvma +
+		qebuf_offset(tx_buf, (entry & (TX_RING_SIZE - 1)));
+
+	/* Avoid a race... */
+	qep->qe_block->qe_txd[entry].tx_flags = TXD_UPDATE;
+
+	skb_copy_from_linear_data(skb, txbuf, len);
+
+	qep->qe_block->qe_txd[entry].tx_addr = txbuf_dvma;
+	qep->qe_block->qe_txd[entry].tx_flags =
+		(TXD_OWN | TXD_SOP | TXD_EOP | (len & TXD_LENGTH));
+	qep->tx_new = NEXT_TX(entry);
+
+	/* Get it going. */
+	sbus_writel(CREG_CTRL_TWAKEUP, qep->qcregs + CREG_CTRL);
+
+	dev->stats.tx_packets++;
+	dev->stats.tx_bytes += len;
+
+	if (TX_BUFFS_AVAIL(qep) <= 0) {
+		/* Halt the net queue and enable tx interrupts.
+		 * When the tx queue empties the tx irq handler
+		 * will wake up the queue and return us back to
+		 * the lazy tx reclaim scheme.
+		 */
+		netif_stop_queue(dev);
+		sbus_writel(0, qep->qcregs + CREG_TIMASK);
+	}
+	spin_unlock_irq(&qep->lock);
+
+	dev_kfree_skb(skb);
+
+	return NETDEV_TX_OK;
+}
+
+static void qe_set_multicast(struct net_device *dev)
+{
+	struct sunqe *qep = netdev_priv(dev);
+	struct netdev_hw_addr *ha;
+	u8 new_mconfig = qep->mconfig;
+	int i;
+	u32 crc;
+
+	/* Lock out others. */
+	netif_stop_queue(dev);
+
+	if ((dev->flags & IFF_ALLMULTI) || (netdev_mc_count(dev) > 64)) {
+		sbus_writeb(MREGS_IACONFIG_ACHNGE | MREGS_IACONFIG_LARESET,
+			    qep->mregs + MREGS_IACONFIG);
+		while ((sbus_readb(qep->mregs + MREGS_IACONFIG) & MREGS_IACONFIG_ACHNGE) != 0)
+			barrier();
+		for (i = 0; i < 8; i++)
+			sbus_writeb(0xff, qep->mregs + MREGS_FILTER);
+		sbus_writeb(0, qep->mregs + MREGS_IACONFIG);
+	} else if (dev->flags & IFF_PROMISC) {
+		new_mconfig |= MREGS_MCONFIG_PROMISC;
+	} else {
+		u16 hash_table[4];
+		u8 *hbytes = (unsigned char *) &hash_table[0];
+
+		memset(hash_table, 0, sizeof(hash_table));
+		netdev_for_each_mc_addr(ha, dev) {
+			crc = ether_crc_le(6, ha->addr);
+			crc >>= 26;
+			hash_table[crc >> 4] |= 1 << (crc & 0xf);
+		}
+		/* Program the qe with the new filter value. */
+		sbus_writeb(MREGS_IACONFIG_ACHNGE | MREGS_IACONFIG_LARESET,
+			    qep->mregs + MREGS_IACONFIG);
+		while ((sbus_readb(qep->mregs + MREGS_IACONFIG) & MREGS_IACONFIG_ACHNGE) != 0)
+			barrier();
+		for (i = 0; i < 8; i++) {
+			u8 tmp = *hbytes++;
+			sbus_writeb(tmp, qep->mregs + MREGS_FILTER);
+		}
+		sbus_writeb(0, qep->mregs + MREGS_IACONFIG);
+	}
+
+	/* Any change of the logical address filter, the physical address,
+	 * or enabling/disabling promiscuous mode causes the MACE to disable
+	 * the receiver.  So we must re-enable them here or else the MACE
+	 * refuses to listen to anything on the network.  Sheesh, took
+	 * me a day or two to find this bug.
+	 */
+	qep->mconfig = new_mconfig;
+	sbus_writeb(qep->mconfig, qep->mregs + MREGS_MCONFIG);
+
+	/* Let us get going again. */
+	netif_wake_queue(dev);
+}
+
+/* Ethtool support... */
+static void qe_get_drvinfo(struct net_device *dev, struct ethtool_drvinfo *info)
+{
+	const struct linux_prom_registers *regs;
+	struct sunqe *qep = netdev_priv(dev);
+	struct platform_device *op;
+
+	strscpy(info->driver, "sunqe", sizeof(info->driver));
+	strscpy(info->version, "3.0", sizeof(info->version));
+
+	op = qep->op;
+	regs = of_get_property(op->dev.of_node, "reg", NULL);
+	if (regs)
+		snprintf(info->bus_info, sizeof(info->bus_info), "SBUS:%d",
+			 regs->which_io);
+
+}
+
+static u32 qe_get_link(struct net_device *dev)
+{
+	struct sunqe *qep = netdev_priv(dev);
+	void __iomem *mregs = qep->mregs;
+	u8 phyconfig;
+
+	spin_lock_irq(&qep->lock);
+	phyconfig = sbus_readb(mregs + MREGS_PHYCONFIG);
+	spin_unlock_irq(&qep->lock);
+
+	return phyconfig & MREGS_PHYCONFIG_LSTAT;
+}
+
+static const struct ethtool_ops qe_ethtool_ops = {
+	.get_drvinfo		= qe_get_drvinfo,
+	.get_link		= qe_get_link,
+};
+
+/* This is only called once at boot time for each card probed. */
+static void qec_init_once(struct sunqec *qecp, struct platform_device *op)
+{
+	u8 bsizes = qecp->qec_bursts;
+
+	if (sbus_can_burst64() && (bsizes & DMA_BURST64)) {
+		sbus_writel(GLOB_CTRL_B64, qecp->gregs + GLOB_CTRL);
+	} else if (bsizes & DMA_BURST32) {
+		sbus_writel(GLOB_CTRL_B32, qecp->gregs + GLOB_CTRL);
+	} else {
+		sbus_writel(GLOB_CTRL_B16, qecp->gregs + GLOB_CTRL);
+	}
+
+	/* Packetsize only used in 100baseT BigMAC configurations,
+	 * set it to zero just to be on the safe side.
+	 */
+	sbus_writel(GLOB_PSIZE_2048, qecp->gregs + GLOB_PSIZE);
+
+	/* Set the local memsize register, divided up to one piece per QE channel. */
+	sbus_writel((resource_size(&op->resource[1]) >> 2),
+		    qecp->gregs + GLOB_MSIZE);
+
+	/* Divide up the local QEC memory amongst the 4 QE receiver and
+	 * transmitter FIFOs.  Basically it is (total / 2 / num_channels).
+	 */
+	sbus_writel((resource_size(&op->resource[1]) >> 2) >> 1,
+		    qecp->gregs + GLOB_TSIZE);
+	sbus_writel((resource_size(&op->resource[1]) >> 2) >> 1,
+		    qecp->gregs + GLOB_RSIZE);
+}
+
+static u8 qec_get_burst(struct device_node *dp)
+{
+	u8 bsizes, bsizes_more;
+
+	/* Find and set the burst sizes for the QEC, since it
+	 * does the actual dma for all 4 channels.
+	 */
+	bsizes = of_getintprop_default(dp, "burst-sizes", 0xff);
+	bsizes &= 0xff;
+	bsizes_more = of_getintprop_default(dp->parent, "burst-sizes", 0xff);
+
+	if (bsizes_more != 0xff)
+		bsizes &= bsizes_more;
+	if (bsizes == 0xff || (bsizes & DMA_BURST16) == 0 ||
+	    (bsizes & DMA_BURST32)==0)
+		bsizes = (DMA_BURST32 - 1);
+
+	return bsizes;
+}
+
+static struct sunqec *get_qec(struct platform_device *child)
+{
+	struct platform_device *op = to_platform_device(child->dev.parent);
+	struct sunqec *qecp;
+
+	qecp = platform_get_drvdata(op);
+	if (!qecp) {
+		qecp = kzalloc_obj(struct sunqec);
+		if (qecp) {
+			u32 ctrl;
+
+			qecp->op = op;
+			qecp->gregs = of_ioremap(&op->resource[0], 0,
+						 GLOB_REG_SIZE,
+						 "QEC Global Registers");
+			if (!qecp->gregs)
+				goto fail;
+
+			/* Make sure the QEC is in MACE mode. */
+			ctrl = sbus_readl(qecp->gregs + GLOB_CTRL);
+			ctrl &= 0xf0000000;
+			if (ctrl != GLOB_CTRL_MMODE) {
+				printk(KERN_ERR "qec: Not in MACE mode!\n");
+				goto fail;
+			}
+
+			if (qec_global_reset(qecp->gregs))
+				goto fail;
+
+			qecp->qec_bursts = qec_get_burst(op->dev.of_node);
+
+			qec_init_once(qecp, op);
+
+			if (request_irq(op->archdata.irqs[0], qec_interrupt,
+					IRQF_SHARED, "qec", (void *) qecp)) {
+				printk(KERN_ERR "qec: Can't register irq.\n");
+				goto fail;
+			}
+
+			platform_set_drvdata(op, qecp);
+
+			qecp->next_module = root_qec_dev;
+			root_qec_dev = qecp;
+		}
+	}
+
+	return qecp;
+
+fail:
+	if (qecp->gregs)
+		of_iounmap(&op->resource[0], qecp->gregs, GLOB_REG_SIZE);
+	kfree(qecp);
+	return NULL;
+}
+
+static const struct net_device_ops qec_ops = {
+	.ndo_open		= qe_open,
+	.ndo_stop		= qe_close,
+	.ndo_start_xmit		= qe_start_xmit,
+	.ndo_set_rx_mode	= qe_set_multicast,
+	.ndo_tx_timeout		= qe_tx_timeout,
+	.ndo_set_mac_address	= eth_mac_addr,
+	.ndo_validate_addr	= eth_validate_addr,
+};
+
+static int qec_ether_init(struct platform_device *op)
+{
+	static unsigned version_printed;
+	struct net_device *dev;
+	struct sunqec *qecp;
+	struct sunqe *qe;
+	int i, res;
+
+	if (version_printed++ == 0)
+		printk(KERN_INFO "%s", version);
+
+	dev = alloc_etherdev(sizeof(struct sunqe));
+	if (!dev)
+		return -ENOMEM;
+
+	eth_hw_addr_set(dev, idprom->id_ethaddr);
+
+	qe = netdev_priv(dev);
+
+	res = -ENODEV;
+
+	i = of_getintprop_default(op->dev.of_node, "channel#", -1);
+	if (i == -1)
+		goto fail;
+	qe->channel = i;
+	spin_lock_init(&qe->lock);
+
+	qecp = get_qec(op);
+	if (!qecp)
+		goto fail;
+
+	qecp->qes[qe->channel] = qe;
+	qe->dev = dev;
+	qe->parent = qecp;
+	qe->op = op;
+
+	res = -ENOMEM;
+	qe->qcregs = of_ioremap(&op->resource[0], 0,
+				CREG_REG_SIZE, "QEC Channel Registers");
+	if (!qe->qcregs) {
+		printk(KERN_ERR "qe: Cannot map channel registers.\n");
+		goto fail;
+	}
+
+	qe->mregs = of_ioremap(&op->resource[1], 0,
+			       MREGS_REG_SIZE, "QE MACE Registers");
+	if (!qe->mregs) {
+		printk(KERN_ERR "qe: Cannot map MACE registers.\n");
+		goto fail;
+	}
+
+	qe->qe_block = dma_alloc_coherent(&op->dev, PAGE_SIZE,
+					  &qe->qblock_dvma, GFP_ATOMIC);
+	qe->buffers = dma_alloc_coherent(&op->dev, sizeof(struct sunqe_buffers),
+					 &qe->buffers_dvma, GFP_ATOMIC);
+	if (qe->qe_block == NULL || qe->qblock_dvma == 0 ||
+	    qe->buffers == NULL || qe->buffers_dvma == 0)
+		goto fail;
+
+	/* Stop this QE. */
+	qe_stop(qe);
+
+	SET_NETDEV_DEV(dev, &op->dev);
+
+	dev->watchdog_timeo = 5*HZ;
+	dev->irq = op->archdata.irqs[0];
+	dev->dma = 0;
+	dev->ethtool_ops = &qe_ethtool_ops;
+	dev->netdev_ops = &qec_ops;
+
+	res = register_netdev(dev);
+	if (res)
+		goto fail;
+
+	platform_set_drvdata(op, qe);
+
+	printk(KERN_INFO "%s: qe channel[%d] %pM\n", dev->name, qe->channel,
+	       dev->dev_addr);
+	return 0;
+
+fail:
+	if (qe->qcregs)
+		of_iounmap(&op->resource[0], qe->qcregs, CREG_REG_SIZE);
+	if (qe->mregs)
+		of_iounmap(&op->resource[1], qe->mregs, MREGS_REG_SIZE);
+	if (qe->qe_block)
+		dma_free_coherent(&op->dev, PAGE_SIZE,
+				  qe->qe_block, qe->qblock_dvma);
+	if (qe->buffers)
+		dma_free_coherent(&op->dev,
+				  sizeof(struct sunqe_buffers),
+				  qe->buffers,
+				  qe->buffers_dvma);
+
+	free_netdev(dev);
+
+	return res;
+}
+
+static int qec_sbus_probe(struct platform_device *op)
+{
+	return qec_ether_init(op);
+}
+
+static void qec_sbus_remove(struct platform_device *op)
+{
+	struct sunqe *qp = platform_get_drvdata(op);
+	struct net_device *net_dev = qp->dev;
+
+	unregister_netdev(net_dev);
+
+	of_iounmap(&op->resource[0], qp->qcregs, CREG_REG_SIZE);
+	of_iounmap(&op->resource[1], qp->mregs, MREGS_REG_SIZE);
+	dma_free_coherent(&op->dev, PAGE_SIZE,
+			  qp->qe_block, qp->qblock_dvma);
+	dma_free_coherent(&op->dev, sizeof(struct sunqe_buffers),
+			  qp->buffers, qp->buffers_dvma);
+
+	free_netdev(net_dev);
+}
+
+static const struct of_device_id qec_sbus_match[] = {
+	{
+		.name = "qe",
+	},
+	{},
+};
+
+MODULE_DEVICE_TABLE(of, qec_sbus_match);
+
+static struct platform_driver qec_sbus_driver = {
+	.driver = {
+		.name = "qec",
+		.of_match_table = qec_sbus_match,
+	},
+	.probe		= qec_sbus_probe,
+	.remove		= qec_sbus_remove,
+};
+
+static int __init qec_init(void)
+{
+	return platform_driver_register(&qec_sbus_driver);
+}
+
+static void __exit qec_exit(void)
+{
+	platform_driver_unregister(&qec_sbus_driver);
+
+	while (root_qec_dev) {
+		struct sunqec *next = root_qec_dev->next_module;
+		struct platform_device *op = root_qec_dev->op;
+
+		free_irq(op->archdata.irqs[0], (void *) root_qec_dev);
+		of_iounmap(&op->resource[0], root_qec_dev->gregs,
+			   GLOB_REG_SIZE);
+		kfree(root_qec_dev);
+
+		root_qec_dev = next;
+	}
+}
+
+module_init(qec_init);
+module_exit(qec_exit);
diff --git a/drivers/net/ethernet/sun/sunqe.h b/drivers/net/ethernet/sun/sunqe.h
new file mode 100644
index 000000000..300631e8a
--- /dev/null
+++ b/drivers/net/ethernet/sun/sunqe.h
@@ -0,0 +1,351 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/* $Id: sunqe.h,v 1.13 2000/02/09 11:15:42 davem Exp $
+ * sunqe.h: Definitions for the Sun QuadEthernet driver.
+ *
+ * Copyright (C) 1996 David S. Miller (davem@caip.rutgers.edu)
+ */
+
+#ifndef _SUNQE_H
+#define _SUNQE_H
+
+/* QEC global registers. */
+#define GLOB_CTRL	0x00UL		/* Control			*/
+#define GLOB_STAT	0x04UL		/* Status			*/
+#define GLOB_PSIZE	0x08UL		/* Packet Size			*/
+#define GLOB_MSIZE	0x0cUL		/* Local-memory Size		*/
+#define GLOB_RSIZE	0x10UL		/* Receive partition size	*/
+#define GLOB_TSIZE	0x14UL		/* Transmit partition size	*/
+#define GLOB_REG_SIZE	0x18UL
+
+#define GLOB_CTRL_MMODE       0x40000000 /* MACE qec mode            */
+#define GLOB_CTRL_BMODE       0x10000000 /* BigMAC qec mode          */
+#define GLOB_CTRL_EPAR        0x00000020 /* Enable parity            */
+#define GLOB_CTRL_ACNTRL      0x00000018 /* SBUS arbitration control */
+#define GLOB_CTRL_B64         0x00000004 /* 64 byte dvma bursts      */
+#define GLOB_CTRL_B32         0x00000002 /* 32 byte dvma bursts      */
+#define GLOB_CTRL_B16         0x00000000 /* 16 byte dvma bursts      */
+#define GLOB_CTRL_RESET       0x00000001 /* Reset the QEC            */
+
+#define GLOB_STAT_TX          0x00000008 /* BigMAC Transmit IRQ      */
+#define GLOB_STAT_RX          0x00000004 /* BigMAC Receive IRQ       */
+#define GLOB_STAT_BM          0x00000002 /* BigMAC Global IRQ        */
+#define GLOB_STAT_ER          0x00000001 /* BigMAC Error IRQ         */
+
+#define GLOB_PSIZE_2048       0x00       /* 2k packet size           */
+#define GLOB_PSIZE_4096       0x01       /* 4k packet size           */
+#define GLOB_PSIZE_6144       0x10       /* 6k packet size           */
+#define GLOB_PSIZE_8192       0x11       /* 8k packet size           */
+
+/* In MACE mode, there are four qe channels.  Each channel has its own
+ * status bits in the QEC status register.  This macro picks out the
+ * ones you want.
+ */
+#define GLOB_STAT_PER_QE(status, channel) (((status) >> ((channel) * 4)) & 0xf)
+
+/* The following registers are for per-qe channel information/status. */
+#define CREG_CTRL	0x00UL	/* Control                   */
+#define CREG_STAT	0x04UL	/* Status                    */
+#define CREG_RXDS	0x08UL	/* RX descriptor ring ptr    */
+#define CREG_TXDS	0x0cUL	/* TX descriptor ring ptr    */
+#define CREG_RIMASK	0x10UL	/* RX Interrupt Mask         */
+#define CREG_TIMASK	0x14UL	/* TX Interrupt Mask         */
+#define CREG_QMASK	0x18UL	/* QEC Error Interrupt Mask  */
+#define CREG_MMASK	0x1cUL	/* MACE Error Interrupt Mask */
+#define CREG_RXWBUFPTR	0x20UL	/* Local memory rx write ptr */
+#define CREG_RXRBUFPTR	0x24UL	/* Local memory rx read ptr  */
+#define CREG_TXWBUFPTR	0x28UL	/* Local memory tx write ptr */
+#define CREG_TXRBUFPTR	0x2cUL	/* Local memory tx read ptr  */
+#define CREG_CCNT	0x30UL	/* Collision Counter         */
+#define CREG_PIPG	0x34UL	/* Inter-Frame Gap           */
+#define CREG_REG_SIZE	0x38UL
+
+#define CREG_CTRL_RXOFF       0x00000004  /* Disable this qe's receiver*/
+#define CREG_CTRL_RESET       0x00000002  /* Reset this qe channel     */
+#define CREG_CTRL_TWAKEUP     0x00000001  /* Transmitter Wakeup, 'go'. */
+
+#define CREG_STAT_EDEFER      0x10000000  /* Excessive Defers          */
+#define CREG_STAT_CLOSS       0x08000000  /* Carrier Loss              */
+#define CREG_STAT_ERETRIES    0x04000000  /* More than 16 retries      */
+#define CREG_STAT_LCOLL       0x02000000  /* Late TX Collision         */
+#define CREG_STAT_FUFLOW      0x01000000  /* FIFO Underflow            */
+#define CREG_STAT_JERROR      0x00800000  /* Jabber Error              */
+#define CREG_STAT_BERROR      0x00400000  /* Babble Error              */
+#define CREG_STAT_TXIRQ       0x00200000  /* Transmit Interrupt        */
+#define CREG_STAT_CCOFLOW     0x00100000  /* TX Coll-counter Overflow  */
+#define CREG_STAT_TXDERROR    0x00080000  /* TX Descriptor is bogus    */
+#define CREG_STAT_TXLERR      0x00040000  /* Late Transmit Error       */
+#define CREG_STAT_TXPERR      0x00020000  /* Transmit Parity Error     */
+#define CREG_STAT_TXSERR      0x00010000  /* Transmit SBUS error ack   */
+#define CREG_STAT_RCCOFLOW    0x00001000  /* RX Coll-counter Overflow  */
+#define CREG_STAT_RUOFLOW     0x00000800  /* Runt Counter Overflow     */
+#define CREG_STAT_MCOFLOW     0x00000400  /* Missed Counter Overflow   */
+#define CREG_STAT_RXFOFLOW    0x00000200  /* RX FIFO Overflow          */
+#define CREG_STAT_RLCOLL      0x00000100  /* RX Late Collision         */
+#define CREG_STAT_FCOFLOW     0x00000080  /* Frame Counter Overflow    */
+#define CREG_STAT_CECOFLOW    0x00000040  /* CRC Error-counter Overflow*/
+#define CREG_STAT_RXIRQ       0x00000020  /* Receive Interrupt         */
+#define CREG_STAT_RXDROP      0x00000010  /* Dropped a RX'd packet     */
+#define CREG_STAT_RXSMALL     0x00000008  /* Receive buffer too small  */
+#define CREG_STAT_RXLERR      0x00000004  /* Receive Late Error        */
+#define CREG_STAT_RXPERR      0x00000002  /* Receive Parity Error      */
+#define CREG_STAT_RXSERR      0x00000001  /* Receive SBUS Error ACK    */
+
+#define CREG_STAT_ERRORS      (CREG_STAT_EDEFER|CREG_STAT_CLOSS|CREG_STAT_ERETRIES|     \
+			       CREG_STAT_LCOLL|CREG_STAT_FUFLOW|CREG_STAT_JERROR|       \
+			       CREG_STAT_BERROR|CREG_STAT_CCOFLOW|CREG_STAT_TXDERROR|   \
+			       CREG_STAT_TXLERR|CREG_STAT_TXPERR|CREG_STAT_TXSERR|      \
+			       CREG_STAT_RCCOFLOW|CREG_STAT_RUOFLOW|CREG_STAT_MCOFLOW| \
+			       CREG_STAT_RXFOFLOW|CREG_STAT_RLCOLL|CREG_STAT_FCOFLOW|   \
+			       CREG_STAT_CECOFLOW|CREG_STAT_RXDROP|CREG_STAT_RXSMALL|   \
+			       CREG_STAT_RXLERR|CREG_STAT_RXPERR|CREG_STAT_RXSERR)
+
+#define CREG_QMASK_COFLOW     0x00100000  /* CollCntr overflow         */
+#define CREG_QMASK_TXDERROR   0x00080000  /* TXD error                 */
+#define CREG_QMASK_TXLERR     0x00040000  /* TX late error             */
+#define CREG_QMASK_TXPERR     0x00020000  /* TX parity error           */
+#define CREG_QMASK_TXSERR     0x00010000  /* TX sbus error ack         */
+#define CREG_QMASK_RXDROP     0x00000010  /* RX drop                   */
+#define CREG_QMASK_RXBERROR   0x00000008  /* RX buffer error           */
+#define CREG_QMASK_RXLEERR    0x00000004  /* RX late error             */
+#define CREG_QMASK_RXPERR     0x00000002  /* RX parity error           */
+#define CREG_QMASK_RXSERR     0x00000001  /* RX sbus error ack         */
+
+#define CREG_MMASK_EDEFER     0x10000000  /* Excess defer              */
+#define CREG_MMASK_CLOSS      0x08000000  /* Carrier loss              */
+#define CREG_MMASK_ERETRY     0x04000000  /* Excess retry              */
+#define CREG_MMASK_LCOLL      0x02000000  /* Late collision error      */
+#define CREG_MMASK_UFLOW      0x01000000  /* Underflow                 */
+#define CREG_MMASK_JABBER     0x00800000  /* Jabber error              */
+#define CREG_MMASK_BABBLE     0x00400000  /* Babble error              */
+#define CREG_MMASK_OFLOW      0x00000800  /* Overflow                  */
+#define CREG_MMASK_RXCOLL     0x00000400  /* RX Coll-Cntr overflow     */
+#define CREG_MMASK_RPKT       0x00000200  /* Runt pkt overflow         */
+#define CREG_MMASK_MPKT       0x00000100  /* Missed pkt overflow       */
+
+#define CREG_PIPG_TENAB       0x00000020  /* Enable Throttle           */
+#define CREG_PIPG_MMODE       0x00000010  /* Manual Mode               */
+#define CREG_PIPG_WMASK       0x0000000f  /* SBUS Wait Mask            */
+
+/* Per-channel AMD 79C940 MACE registers. */
+#define MREGS_RXFIFO	0x00UL	/* Receive FIFO                   */
+#define MREGS_TXFIFO	0x01UL	/* Transmit FIFO                  */
+#define MREGS_TXFCNTL	0x02UL	/* Transmit Frame Control         */
+#define MREGS_TXFSTAT	0x03UL	/* Transmit Frame Status          */
+#define MREGS_TXRCNT	0x04UL	/* Transmit Retry Count           */
+#define MREGS_RXFCNTL	0x05UL	/* Receive Frame Control          */
+#define MREGS_RXFSTAT	0x06UL	/* Receive Frame Status           */
+#define MREGS_FFCNT	0x07UL	/* FIFO Frame Count               */
+#define MREGS_IREG	0x08UL	/* Interrupt Register             */
+#define MREGS_IMASK	0x09UL	/* Interrupt Mask                 */
+#define MREGS_POLL	0x0aUL	/* POLL Register                  */
+#define MREGS_BCONFIG	0x0bUL	/* BIU Config                     */
+#define MREGS_FCONFIG	0x0cUL	/* FIFO Config                    */
+#define MREGS_MCONFIG	0x0dUL	/* MAC Config                     */
+#define MREGS_PLSCONFIG	0x0eUL	/* PLS Config                     */
+#define MREGS_PHYCONFIG	0x0fUL	/* PHY Config                     */
+#define MREGS_CHIPID1	0x10UL	/* Chip-ID, low bits              */
+#define MREGS_CHIPID2	0x11UL	/* Chip-ID, high bits             */
+#define MREGS_IACONFIG	0x12UL	/* Internal Address Config        */
+	/* 0x13UL, reserved */
+#define MREGS_FILTER	0x14UL	/* Logical Address Filter         */
+#define MREGS_ETHADDR	0x15UL	/* Our Ethernet Address           */
+	/* 0x16UL, reserved */
+	/* 0x17UL, reserved */
+#define MREGS_MPCNT	0x18UL	/* Missed Packet Count            */
+	/* 0x19UL, reserved */
+#define MREGS_RPCNT	0x1aUL	/* Runt Packet Count              */
+#define MREGS_RCCNT	0x1bUL	/* RX Collision Count             */
+	/* 0x1cUL, reserved */
+#define MREGS_UTEST	0x1dUL	/* User Test                      */
+#define MREGS_RTEST1	0x1eUL	/* Reserved Test 1                */
+#define MREGS_RTEST2	0x1fUL	/* Reserved Test 2                */
+#define MREGS_REG_SIZE	0x20UL
+
+#define MREGS_TXFCNTL_DRETRY        0x80 /* Retry disable                  */
+#define MREGS_TXFCNTL_DFCS          0x08 /* Disable TX FCS                 */
+#define MREGS_TXFCNTL_AUTOPAD       0x01 /* TX auto pad                    */
+
+#define MREGS_TXFSTAT_VALID         0x80 /* TX valid                       */
+#define MREGS_TXFSTAT_UNDERFLOW     0x40 /* TX underflow                   */
+#define MREGS_TXFSTAT_LCOLL         0x20 /* TX late collision              */
+#define MREGS_TXFSTAT_MRETRY        0x10 /* TX > 1 retries                 */
+#define MREGS_TXFSTAT_ORETRY        0x08 /* TX 1 retry                     */
+#define MREGS_TXFSTAT_PDEFER        0x04 /* TX pkt deferred                */
+#define MREGS_TXFSTAT_CLOSS         0x02 /* TX carrier lost                */
+#define MREGS_TXFSTAT_RERROR        0x01 /* TX retry error                 */
+
+#define MREGS_TXRCNT_EDEFER         0x80 /* TX Excess defers               */
+#define MREGS_TXRCNT_CMASK          0x0f /* TX retry count                 */
+
+#define MREGS_RXFCNTL_LOWLAT        0x08 /* RX low latency                 */
+#define MREGS_RXFCNTL_AREJECT       0x04 /* RX addr match rej              */
+#define MREGS_RXFCNTL_AUTOSTRIP     0x01 /* RX auto strip                  */
+
+#define MREGS_RXFSTAT_OVERFLOW      0x80 /* RX overflow                    */
+#define MREGS_RXFSTAT_LCOLL         0x40 /* RX late collision              */
+#define MREGS_RXFSTAT_FERROR        0x20 /* RX framing error               */
+#define MREGS_RXFSTAT_FCSERROR      0x10 /* RX FCS error                   */
+#define MREGS_RXFSTAT_RBCNT         0x0f /* RX msg byte count              */
+
+#define MREGS_FFCNT_RX              0xf0 /* RX FIFO frame cnt              */
+#define MREGS_FFCNT_TX              0x0f /* TX FIFO frame cnt              */
+
+#define MREGS_IREG_JABBER           0x80 /* IRQ Jabber error               */
+#define MREGS_IREG_BABBLE           0x40 /* IRQ Babble error               */
+#define MREGS_IREG_COLL             0x20 /* IRQ Collision error            */
+#define MREGS_IREG_RCCO             0x10 /* IRQ Collision cnt overflow     */
+#define MREGS_IREG_RPKTCO           0x08 /* IRQ Runt packet count overflow */
+#define MREGS_IREG_MPKTCO           0x04 /* IRQ missed packet cnt overflow */
+#define MREGS_IREG_RXIRQ            0x02 /* IRQ RX'd a packet              */
+#define MREGS_IREG_TXIRQ            0x01 /* IRQ TX'd a packet              */
+
+#define MREGS_IMASK_BABBLE          0x40 /* IMASK Babble errors            */
+#define MREGS_IMASK_COLL            0x20 /* IMASK Collision errors         */
+#define MREGS_IMASK_MPKTCO          0x04 /* IMASK Missed pkt cnt overflow  */
+#define MREGS_IMASK_RXIRQ           0x02 /* IMASK RX interrupts            */
+#define MREGS_IMASK_TXIRQ           0x01 /* IMASK TX interrupts            */
+
+#define MREGS_POLL_TXVALID          0x80 /* TX is valid                    */
+#define MREGS_POLL_TDTR             0x40 /* TX data transfer request       */
+#define MREGS_POLL_RDTR             0x20 /* RX data transfer request       */
+
+#define MREGS_BCONFIG_BSWAP         0x40 /* Byte Swap                      */
+#define MREGS_BCONFIG_4TS           0x00 /* 4byte transmit start point     */
+#define MREGS_BCONFIG_16TS          0x10 /* 16byte transmit start point    */
+#define MREGS_BCONFIG_64TS          0x20 /* 64byte transmit start point    */
+#define MREGS_BCONFIG_112TS         0x30 /* 112byte transmit start point   */
+#define MREGS_BCONFIG_RESET         0x01 /* SW-Reset the MACE              */
+
+#define MREGS_FCONFIG_TXF8          0x00 /* TX fifo 8 write cycles         */
+#define MREGS_FCONFIG_TXF32         0x80 /* TX fifo 32 write cycles        */
+#define MREGS_FCONFIG_TXF16         0x40 /* TX fifo 16 write cycles        */
+#define MREGS_FCONFIG_RXF64         0x20 /* RX fifo 64 write cycles        */
+#define MREGS_FCONFIG_RXF32         0x10 /* RX fifo 32 write cycles        */
+#define MREGS_FCONFIG_RXF16         0x00 /* RX fifo 16 write cycles        */
+#define MREGS_FCONFIG_TFWU          0x08 /* TX fifo watermark update       */
+#define MREGS_FCONFIG_RFWU          0x04 /* RX fifo watermark update       */
+#define MREGS_FCONFIG_TBENAB        0x02 /* TX burst enable                */
+#define MREGS_FCONFIG_RBENAB        0x01 /* RX burst enable                */
+
+#define MREGS_MCONFIG_PROMISC       0x80 /* Promiscuous mode enable        */
+#define MREGS_MCONFIG_TPDDISAB      0x40 /* TX 2part deferral enable       */
+#define MREGS_MCONFIG_MBAENAB       0x20 /* Modified backoff enable        */
+#define MREGS_MCONFIG_RPADISAB      0x08 /* RX physical addr disable       */
+#define MREGS_MCONFIG_RBDISAB       0x04 /* RX broadcast disable           */
+#define MREGS_MCONFIG_TXENAB        0x02 /* Enable transmitter             */
+#define MREGS_MCONFIG_RXENAB        0x01 /* Enable receiver                */
+
+#define MREGS_PLSCONFIG_TXMS        0x08 /* TX mode select                 */
+#define MREGS_PLSCONFIG_GPSI        0x06 /* Use GPSI connector             */
+#define MREGS_PLSCONFIG_DAI         0x04 /* Use DAI connector              */
+#define MREGS_PLSCONFIG_TP          0x02 /* Use TwistedPair connector      */
+#define MREGS_PLSCONFIG_AUI         0x00 /* Use AUI connector              */
+#define MREGS_PLSCONFIG_IOENAB      0x01 /* PLS I/O enable                 */
+
+#define MREGS_PHYCONFIG_LSTAT       0x80 /* Link status                    */
+#define MREGS_PHYCONFIG_LTESTDIS    0x40 /* Disable link test logic        */
+#define MREGS_PHYCONFIG_RXPOLARITY  0x20 /* RX polarity                    */
+#define MREGS_PHYCONFIG_APCDISAB    0x10 /* AutoPolarityCorrect disab      */
+#define MREGS_PHYCONFIG_LTENAB      0x08 /* Select low threshold           */
+#define MREGS_PHYCONFIG_AUTO        0x04 /* Connector port auto-sel        */
+#define MREGS_PHYCONFIG_RWU         0x02 /* Remote WakeUp                  */
+#define MREGS_PHYCONFIG_AW          0x01 /* Auto Wakeup                    */
+
+#define MREGS_IACONFIG_ACHNGE       0x80 /* Do address change              */
+#define MREGS_IACONFIG_PARESET      0x04 /* Physical address reset         */
+#define MREGS_IACONFIG_LARESET      0x02 /* Logical address reset          */
+
+#define MREGS_UTEST_RTRENAB         0x80 /* Enable resv test register      */
+#define MREGS_UTEST_RTRDISAB        0x40 /* Disab resv test register       */
+#define MREGS_UTEST_RPACCEPT        0x20 /* Accept runt packets            */
+#define MREGS_UTEST_FCOLL           0x10 /* Force collision status         */
+#define MREGS_UTEST_FCSENAB         0x08 /* Enable FCS on RX               */
+#define MREGS_UTEST_INTLOOPM        0x06 /* Intern lpback w/MENDEC         */
+#define MREGS_UTEST_INTLOOP         0x04 /* Intern lpback                  */
+#define MREGS_UTEST_EXTLOOP         0x02 /* Extern lpback                  */
+#define MREGS_UTEST_NOLOOP          0x00 /* No loopback                    */
+
+struct qe_rxd {
+	u32 rx_flags;
+	u32 rx_addr;
+};
+
+#define RXD_OWN      0x80000000 /* Ownership.      */
+#define RXD_UPDATE   0x10000000 /* Being Updated?  */
+#define RXD_LENGTH   0x000007ff /* Packet Length.  */
+
+struct qe_txd {
+	u32 tx_flags;
+	u32 tx_addr;
+};
+
+#define TXD_OWN      0x80000000 /* Ownership.      */
+#define TXD_SOP      0x40000000 /* Start Of Packet */
+#define TXD_EOP      0x20000000 /* End Of Packet   */
+#define TXD_UPDATE   0x10000000 /* Being Updated?  */
+#define TXD_LENGTH   0x000007ff /* Packet Length.  */
+
+#define TX_RING_MAXSIZE   256
+#define RX_RING_MAXSIZE   256
+
+#define TX_RING_SIZE      16
+#define RX_RING_SIZE      16
+
+#define NEXT_RX(num)       (((num) + 1) & (RX_RING_MAXSIZE - 1))
+#define NEXT_TX(num)       (((num) + 1) & (TX_RING_MAXSIZE - 1))
+#define PREV_RX(num)       (((num) - 1) & (RX_RING_MAXSIZE - 1))
+#define PREV_TX(num)       (((num) - 1) & (TX_RING_MAXSIZE - 1))
+
+#define TX_BUFFS_AVAIL(qp)                                    \
+        (((qp)->tx_old <= (qp)->tx_new) ?                     \
+	  (qp)->tx_old + (TX_RING_SIZE - 1) - (qp)->tx_new :  \
+			    (qp)->tx_old - (qp)->tx_new - 1)
+
+struct qe_init_block {
+	struct qe_rxd qe_rxd[RX_RING_MAXSIZE];
+	struct qe_txd qe_txd[TX_RING_MAXSIZE];
+};
+
+#define qib_offset(mem, elem) \
+((__u32)((unsigned long)(&(((struct qe_init_block *)0)->mem[elem]))))
+
+struct sunqe;
+
+struct sunqec {
+	void __iomem		*gregs;		/* QEC Global Registers         */
+	struct sunqe		*qes[4];	/* Each child MACE              */
+	unsigned int            qec_bursts;	/* Support burst sizes          */
+	struct platform_device	*op;		/* QEC's OF device              */
+	struct sunqec		*next_module;	/* List of all QECs in system   */
+};
+
+#define PKT_BUF_SZ	1664
+#define RXD_PKT_SZ	1664
+
+struct sunqe_buffers {
+	u8	tx_buf[TX_RING_SIZE][PKT_BUF_SZ];
+	u8	__pad[2];
+	u8	rx_buf[RX_RING_SIZE][PKT_BUF_SZ];
+};
+
+#define qebuf_offset(mem, elem) \
+((__u32)((unsigned long)(&(((struct sunqe_buffers *)0)->mem[elem][0]))))
+
+struct sunqe {
+	void __iomem			*qcregs;		/* QEC per-channel Registers   */
+	void __iomem			*mregs;		/* Per-channel MACE Registers  */
+	struct qe_init_block      	*qe_block;	/* RX and TX descriptors       */
+	dma_addr_t			qblock_dvma;	/* RX and TX descriptors       */
+	spinlock_t			lock;		/* Protects txfull state       */
+	int                        	rx_new, rx_old;	/* RX ring extents	       */
+	int			   	tx_new, tx_old;	/* TX ring extents	       */
+	struct sunqe_buffers		*buffers;	/* CPU visible address.        */
+	dma_addr_t			buffers_dvma;	/* DVMA visible address.       */
+	struct sunqec			*parent;
+	u8				mconfig;	/* Base MACE mconfig value     */
+	struct platform_device		*op;		/* QE's OF device struct       */
+	struct net_device		*dev;		/* QE's netdevice struct       */
+	int				channel;	/* Who am I?                   */
+};
+
+#endif /* !(_SUNQE_H) */
diff --git a/drivers/net/ethernet/sun/sunvnet.c b/drivers/net/ethernet/sun/sunvnet.c
new file mode 100644
index 000000000..6d477401f
--- /dev/null
+++ b/drivers/net/ethernet/sun/sunvnet.c
@@ -0,0 +1,556 @@
+// SPDX-License-Identifier: GPL-2.0
+/* sunvnet.c: Sun LDOM Virtual Network Driver.
+ *
+ * Copyright (C) 2007, 2008 David S. Miller <davem@davemloft.net>
+ * Copyright (C) 2016-2017 Oracle. All rights reserved.
+ */
+
+#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
+
+#include <linux/module.h>
+#include <linux/kernel.h>
+#include <linux/types.h>
+#include <linux/slab.h>
+#include <linux/delay.h>
+#include <linux/init.h>
+#include <linux/netdevice.h>
+#include <linux/ethtool.h>
+#include <linux/etherdevice.h>
+#include <linux/mutex.h>
+#include <linux/highmem.h>
+#include <linux/if_vlan.h>
+
+#if IS_ENABLED(CONFIG_IPV6)
+#include <linux/icmpv6.h>
+#endif
+
+#include <net/ip.h>
+#include <net/icmp.h>
+#include <net/route.h>
+
+#include <asm/vio.h>
+#include <asm/ldc.h>
+
+#include "sunvnet_common.h"
+
+/* length of time before we decide the hardware is borked,
+ * and dev->tx_timeout() should be called to fix the problem
+ */
+#define VNET_TX_TIMEOUT			(5 * HZ)
+
+#define DRV_MODULE_NAME		"sunvnet"
+#define DRV_MODULE_VERSION	"2.0"
+#define DRV_MODULE_RELDATE	"February 3, 2017"
+
+static char version[] =
+	DRV_MODULE_NAME " " DRV_MODULE_VERSION " (" DRV_MODULE_RELDATE ")";
+MODULE_AUTHOR("David S. Miller <davem@davemloft.net>");
+MODULE_DESCRIPTION("Sun LDOM virtual network driver");
+MODULE_LICENSE("GPL");
+MODULE_VERSION(DRV_MODULE_VERSION);
+
+/* Ordered from largest major to lowest */
+static struct vio_version vnet_versions[] = {
+	{ .major = 1, .minor = 8 },
+	{ .major = 1, .minor = 7 },
+	{ .major = 1, .minor = 6 },
+	{ .major = 1, .minor = 0 },
+};
+
+static void vnet_get_drvinfo(struct net_device *dev,
+			     struct ethtool_drvinfo *info)
+{
+	strscpy(info->driver, DRV_MODULE_NAME, sizeof(info->driver));
+	strscpy(info->version, DRV_MODULE_VERSION, sizeof(info->version));
+}
+
+static u32 vnet_get_msglevel(struct net_device *dev)
+{
+	struct vnet *vp = netdev_priv(dev);
+
+	return vp->msg_enable;
+}
+
+static void vnet_set_msglevel(struct net_device *dev, u32 value)
+{
+	struct vnet *vp = netdev_priv(dev);
+
+	vp->msg_enable = value;
+}
+
+static const struct {
+	const char string[ETH_GSTRING_LEN];
+} ethtool_stats_keys[] = {
+	{ "rx_packets" },
+	{ "tx_packets" },
+	{ "rx_bytes" },
+	{ "tx_bytes" },
+	{ "rx_errors" },
+	{ "tx_errors" },
+	{ "rx_dropped" },
+	{ "tx_dropped" },
+	{ "multicast" },
+	{ "rx_length_errors" },
+	{ "rx_frame_errors" },
+	{ "rx_missed_errors" },
+	{ "tx_carrier_errors" },
+	{ "nports" },
+};
+
+static int vnet_get_sset_count(struct net_device *dev, int sset)
+{
+	struct vnet *vp = (struct vnet *)netdev_priv(dev);
+
+	switch (sset) {
+	case ETH_SS_STATS:
+		return ARRAY_SIZE(ethtool_stats_keys)
+			+ (NUM_VNET_PORT_STATS * vp->nports);
+	default:
+		return -EOPNOTSUPP;
+	}
+}
+
+static void vnet_get_strings(struct net_device *dev, u32 stringset, u8 *buf)
+{
+	struct vnet *vp = (struct vnet *)netdev_priv(dev);
+	struct vnet_port *port;
+
+	switch (stringset) {
+	case ETH_SS_STATS:
+		memcpy(buf, &ethtool_stats_keys, sizeof(ethtool_stats_keys));
+		buf += sizeof(ethtool_stats_keys);
+
+		rcu_read_lock();
+		list_for_each_entry_rcu(port, &vp->port_list, list) {
+			ethtool_sprintf(&buf, "p%u.%s-%pM", port->q_index,
+					port->switch_port ? "s" : "q",
+					port->raddr);
+			ethtool_sprintf(&buf, "p%u.rx_packets", port->q_index);
+			ethtool_sprintf(&buf, "p%u.tx_packets", port->q_index);
+			ethtool_sprintf(&buf, "p%u.rx_bytes", port->q_index);
+			ethtool_sprintf(&buf, "p%u.tx_bytes", port->q_index);
+			ethtool_sprintf(&buf, "p%u.event_up", port->q_index);
+			ethtool_sprintf(&buf, "p%u.event_reset", port->q_index);
+		}
+		rcu_read_unlock();
+		break;
+	default:
+		WARN_ON(1);
+		break;
+	}
+}
+
+static void vnet_get_ethtool_stats(struct net_device *dev,
+				   struct ethtool_stats *estats, u64 *data)
+{
+	struct vnet *vp = (struct vnet *)netdev_priv(dev);
+	struct vnet_port *port;
+	int i = 0;
+
+	data[i++] = dev->stats.rx_packets;
+	data[i++] = dev->stats.tx_packets;
+	data[i++] = dev->stats.rx_bytes;
+	data[i++] = dev->stats.tx_bytes;
+	data[i++] = dev->stats.rx_errors;
+	data[i++] = dev->stats.tx_errors;
+	data[i++] = dev->stats.rx_dropped;
+	data[i++] = dev->stats.tx_dropped;
+	data[i++] = dev->stats.multicast;
+	data[i++] = dev->stats.rx_length_errors;
+	data[i++] = dev->stats.rx_frame_errors;
+	data[i++] = dev->stats.rx_missed_errors;
+	data[i++] = dev->stats.tx_carrier_errors;
+	data[i++] = vp->nports;
+
+	rcu_read_lock();
+	list_for_each_entry_rcu(port, &vp->port_list, list) {
+		data[i++] = port->q_index;
+		data[i++] = port->stats.rx_packets;
+		data[i++] = port->stats.tx_packets;
+		data[i++] = port->stats.rx_bytes;
+		data[i++] = port->stats.tx_bytes;
+		data[i++] = port->stats.event_up;
+		data[i++] = port->stats.event_reset;
+	}
+	rcu_read_unlock();
+}
+
+static const struct ethtool_ops vnet_ethtool_ops = {
+	.get_drvinfo		= vnet_get_drvinfo,
+	.get_msglevel		= vnet_get_msglevel,
+	.set_msglevel		= vnet_set_msglevel,
+	.get_link		= ethtool_op_get_link,
+	.get_sset_count		= vnet_get_sset_count,
+	.get_strings		= vnet_get_strings,
+	.get_ethtool_stats	= vnet_get_ethtool_stats,
+};
+
+static LIST_HEAD(vnet_list);
+static DEFINE_MUTEX(vnet_list_mutex);
+
+static struct vnet_port *__tx_port_find(struct vnet *vp, struct sk_buff *skb)
+{
+	unsigned int hash = vnet_hashfn(skb->data);
+	struct hlist_head *hp = &vp->port_hash[hash];
+	struct vnet_port *port;
+
+	hlist_for_each_entry_rcu(port, hp, hash) {
+		if (!sunvnet_port_is_up_common(port))
+			continue;
+		if (ether_addr_equal(port->raddr, skb->data))
+			return port;
+	}
+	list_for_each_entry_rcu(port, &vp->port_list, list) {
+		if (!port->switch_port)
+			continue;
+		if (!sunvnet_port_is_up_common(port))
+			continue;
+		return port;
+	}
+	return NULL;
+}
+
+/* func arg to vnet_start_xmit_common() to get the proper tx port */
+static struct vnet_port *vnet_tx_port_find(struct sk_buff *skb,
+					   struct net_device *dev)
+{
+	struct vnet *vp = netdev_priv(dev);
+
+	return __tx_port_find(vp, skb);
+}
+
+static u16 vnet_select_queue(struct net_device *dev, struct sk_buff *skb,
+			     struct net_device *sb_dev)
+{
+	struct vnet *vp = netdev_priv(dev);
+	struct vnet_port *port = __tx_port_find(vp, skb);
+
+	if (!port)
+		return 0;
+
+	return port->q_index;
+}
+
+/* Wrappers to common functions */
+static netdev_tx_t vnet_start_xmit(struct sk_buff *skb, struct net_device *dev)
+{
+	return sunvnet_start_xmit_common(skb, dev, vnet_tx_port_find);
+}
+
+static void vnet_set_rx_mode(struct net_device *dev)
+{
+	struct vnet *vp = netdev_priv(dev);
+
+	return sunvnet_set_rx_mode_common(dev, vp);
+}
+
+#ifdef CONFIG_NET_POLL_CONTROLLER
+static void vnet_poll_controller(struct net_device *dev)
+{
+	struct vnet *vp = netdev_priv(dev);
+
+	return sunvnet_poll_controller_common(dev, vp);
+}
+#endif
+
+static const struct net_device_ops vnet_ops = {
+	.ndo_open		= sunvnet_open_common,
+	.ndo_stop		= sunvnet_close_common,
+	.ndo_set_rx_mode	= vnet_set_rx_mode,
+	.ndo_set_mac_address	= sunvnet_set_mac_addr_common,
+	.ndo_validate_addr	= eth_validate_addr,
+	.ndo_tx_timeout		= sunvnet_tx_timeout_common,
+	.ndo_start_xmit		= vnet_start_xmit,
+	.ndo_select_queue	= vnet_select_queue,
+#ifdef CONFIG_NET_POLL_CONTROLLER
+	.ndo_poll_controller	= vnet_poll_controller,
+#endif
+};
+
+static struct vnet *vnet_new(const u64 *local_mac,
+			     struct vio_dev *vdev)
+{
+	struct net_device *dev;
+	u8 addr[ETH_ALEN];
+	struct vnet *vp;
+	int err, i;
+
+	dev = alloc_etherdev_mqs(sizeof(*vp), VNET_MAX_TXQS, 1);
+	if (!dev)
+		return ERR_PTR(-ENOMEM);
+	dev->needed_headroom = VNET_PACKET_SKIP + 8;
+	dev->needed_tailroom = 8;
+
+	for (i = 0; i < ETH_ALEN; i++)
+		addr[i] = (*local_mac >> (5 - i) * 8) & 0xff;
+	eth_hw_addr_set(dev, addr);
+
+	vp = netdev_priv(dev);
+
+	spin_lock_init(&vp->lock);
+	vp->dev = dev;
+
+	INIT_LIST_HEAD(&vp->port_list);
+	for (i = 0; i < VNET_PORT_HASH_SIZE; i++)
+		INIT_HLIST_HEAD(&vp->port_hash[i]);
+	INIT_LIST_HEAD(&vp->list);
+	vp->local_mac = *local_mac;
+
+	dev->netdev_ops = &vnet_ops;
+	dev->ethtool_ops = &vnet_ethtool_ops;
+	dev->watchdog_timeo = VNET_TX_TIMEOUT;
+
+	dev->hw_features = NETIF_F_TSO | NETIF_F_GSO | NETIF_F_ALL_TSO |
+			   NETIF_F_HW_CSUM | NETIF_F_SG;
+	dev->features = dev->hw_features;
+
+	/* MTU range: 68 - 65535 */
+	dev->min_mtu = ETH_MIN_MTU;
+	dev->max_mtu = VNET_MAX_MTU;
+
+	SET_NETDEV_DEV(dev, &vdev->dev);
+
+	err = register_netdev(dev);
+	if (err) {
+		pr_err("Cannot register net device, aborting\n");
+		goto err_out_free_dev;
+	}
+
+	netdev_info(dev, "Sun LDOM vnet %pM\n", dev->dev_addr);
+
+	list_add(&vp->list, &vnet_list);
+
+	return vp;
+
+err_out_free_dev:
+	free_netdev(dev);
+
+	return ERR_PTR(err);
+}
+
+static struct vnet *vnet_find_or_create(const u64 *local_mac,
+					struct vio_dev *vdev)
+{
+	struct vnet *iter, *vp;
+
+	mutex_lock(&vnet_list_mutex);
+	vp = NULL;
+	list_for_each_entry(iter, &vnet_list, list) {
+		if (iter->local_mac == *local_mac) {
+			vp = iter;
+			break;
+		}
+	}
+	if (!vp)
+		vp = vnet_new(local_mac, vdev);
+	mutex_unlock(&vnet_list_mutex);
+
+	return vp;
+}
+
+static void vnet_cleanup(void)
+{
+	struct vnet *vp;
+	struct net_device *dev;
+
+	mutex_lock(&vnet_list_mutex);
+	while (!list_empty(&vnet_list)) {
+		vp = list_first_entry(&vnet_list, struct vnet, list);
+		list_del(&vp->list);
+		dev = vp->dev;
+		/* vio_unregister_driver() should have cleaned up port_list */
+		BUG_ON(!list_empty(&vp->port_list));
+		unregister_netdev(dev);
+		free_netdev(dev);
+	}
+	mutex_unlock(&vnet_list_mutex);
+}
+
+static const char *local_mac_prop = "local-mac-address";
+
+static struct vnet *vnet_find_parent(struct mdesc_handle *hp,
+				     u64 port_node,
+				     struct vio_dev *vdev)
+{
+	const u64 *local_mac = NULL;
+	u64 a;
+
+	mdesc_for_each_arc(a, hp, port_node, MDESC_ARC_TYPE_BACK) {
+		u64 target = mdesc_arc_target(hp, a);
+		const char *name;
+
+		name = mdesc_get_property(hp, target, "name", NULL);
+		if (!name || strcmp(name, "network"))
+			continue;
+
+		local_mac = mdesc_get_property(hp, target,
+					       local_mac_prop, NULL);
+		if (local_mac)
+			break;
+	}
+	if (!local_mac)
+		return ERR_PTR(-ENODEV);
+
+	return vnet_find_or_create(local_mac, vdev);
+}
+
+static struct ldc_channel_config vnet_ldc_cfg = {
+	.event		= sunvnet_event_common,
+	.mtu		= 64,
+	.mode		= LDC_MODE_UNRELIABLE,
+};
+
+static struct vio_driver_ops vnet_vio_ops = {
+	.send_attr		= sunvnet_send_attr_common,
+	.handle_attr		= sunvnet_handle_attr_common,
+	.handshake_complete	= sunvnet_handshake_complete_common,
+};
+
+const char *remote_macaddr_prop = "remote-mac-address";
+
+static int vnet_port_probe(struct vio_dev *vdev, const struct vio_device_id *id)
+{
+	struct mdesc_handle *hp;
+	struct vnet_port *port;
+	unsigned long flags;
+	struct vnet *vp;
+	const u64 *rmac;
+	int len, i, err, switch_port;
+
+	hp = mdesc_grab();
+
+	if (!hp)
+		return -ENODEV;
+
+	vp = vnet_find_parent(hp, vdev->mp, vdev);
+	if (IS_ERR(vp)) {
+		pr_err("Cannot find port parent vnet\n");
+		err = PTR_ERR(vp);
+		goto err_out_put_mdesc;
+	}
+
+	rmac = mdesc_get_property(hp, vdev->mp, remote_macaddr_prop, &len);
+	err = -ENODEV;
+	if (!rmac) {
+		pr_err("Port lacks %s property\n", remote_macaddr_prop);
+		goto err_out_put_mdesc;
+	}
+
+	port = kzalloc_obj(*port);
+	err = -ENOMEM;
+	if (!port)
+		goto err_out_put_mdesc;
+
+	for (i = 0; i < ETH_ALEN; i++)
+		port->raddr[i] = (*rmac >> (5 - i) * 8) & 0xff;
+
+	port->vp = vp;
+
+	err = vio_driver_init(&port->vio, vdev, VDEV_NETWORK,
+			      vnet_versions, ARRAY_SIZE(vnet_versions),
+			      &vnet_vio_ops, vp->dev->name);
+	if (err)
+		goto err_out_free_port;
+
+	err = vio_ldc_alloc(&port->vio, &vnet_ldc_cfg, port);
+	if (err)
+		goto err_out_free_port;
+
+	netif_napi_add(port->vp->dev, &port->napi, sunvnet_poll_common);
+
+	INIT_HLIST_NODE(&port->hash);
+	INIT_LIST_HEAD(&port->list);
+
+	switch_port = 0;
+	if (mdesc_get_property(hp, vdev->mp, "switch-port", NULL))
+		switch_port = 1;
+	port->switch_port = switch_port;
+	port->tso = true;
+	port->tsolen = 0;
+
+	spin_lock_irqsave(&vp->lock, flags);
+	if (switch_port)
+		list_add_rcu(&port->list, &vp->port_list);
+	else
+		list_add_tail_rcu(&port->list, &vp->port_list);
+	hlist_add_head_rcu(&port->hash,
+			   &vp->port_hash[vnet_hashfn(port->raddr)]);
+	sunvnet_port_add_txq_common(port);
+	spin_unlock_irqrestore(&vp->lock, flags);
+
+	dev_set_drvdata(&vdev->dev, port);
+
+	pr_info("%s: PORT ( remote-mac %pM%s )\n",
+		vp->dev->name, port->raddr, switch_port ? " switch-port" : "");
+
+	timer_setup(&port->clean_timer, sunvnet_clean_timer_expire_common, 0);
+
+	napi_enable(&port->napi);
+	vio_port_up(&port->vio);
+
+	mdesc_release(hp);
+
+	return 0;
+
+err_out_free_port:
+	kfree(port);
+
+err_out_put_mdesc:
+	mdesc_release(hp);
+	return err;
+}
+
+static void vnet_port_remove(struct vio_dev *vdev)
+{
+	struct vnet_port *port = dev_get_drvdata(&vdev->dev);
+
+	if (port) {
+		timer_delete_sync(&port->vio.timer);
+
+		napi_disable(&port->napi);
+
+		list_del_rcu(&port->list);
+		hlist_del_rcu(&port->hash);
+
+		synchronize_rcu();
+		timer_shutdown_sync(&port->clean_timer);
+		sunvnet_port_rm_txq_common(port);
+		netif_napi_del(&port->napi);
+		sunvnet_port_free_tx_bufs_common(port);
+		vio_ldc_free(&port->vio);
+
+		dev_set_drvdata(&vdev->dev, NULL);
+
+		kfree(port);
+	}
+}
+
+static const struct vio_device_id vnet_port_match[] = {
+	{
+		.type = "vnet-port",
+	},
+	{},
+};
+MODULE_DEVICE_TABLE(vio, vnet_port_match);
+
+static struct vio_driver vnet_port_driver = {
+	.id_table	= vnet_port_match,
+	.probe		= vnet_port_probe,
+	.remove		= vnet_port_remove,
+	.name		= "vnet_port",
+};
+
+static int __init vnet_init(void)
+{
+	pr_info("%s\n", version);
+	return vio_register_driver(&vnet_port_driver);
+}
+
+static void __exit vnet_exit(void)
+{
+	vio_unregister_driver(&vnet_port_driver);
+	vnet_cleanup();
+}
+
+module_init(vnet_init);
+module_exit(vnet_exit);
diff --git a/drivers/net/ethernet/sun/sunvnet_common.c b/drivers/net/ethernet/sun/sunvnet_common.c
new file mode 100644
index 000000000..5cca95869
--- /dev/null
+++ b/drivers/net/ethernet/sun/sunvnet_common.c
@@ -0,0 +1,1814 @@
+// SPDX-License-Identifier: GPL-2.0
+/* sunvnet.c: Sun LDOM Virtual Network Driver.
+ *
+ * Copyright (C) 2007, 2008 David S. Miller <davem@davemloft.net>
+ * Copyright (C) 2016-2017 Oracle. All rights reserved.
+ */
+
+#include <linux/module.h>
+#include <linux/kernel.h>
+#include <linux/types.h>
+#include <linux/slab.h>
+#include <linux/delay.h>
+#include <linux/init.h>
+#include <linux/netdevice.h>
+#include <linux/ethtool.h>
+#include <linux/etherdevice.h>
+#include <linux/mutex.h>
+#include <linux/highmem.h>
+#include <linux/if_vlan.h>
+#define CREATE_TRACE_POINTS
+#include <trace/events/sunvnet.h>
+
+#if IS_ENABLED(CONFIG_IPV6)
+#include <linux/icmpv6.h>
+#endif
+
+#include <net/ip.h>
+#include <net/gso.h>
+#include <net/icmp.h>
+#include <net/route.h>
+
+#include <asm/vio.h>
+#include <asm/ldc.h>
+
+#include "sunvnet_common.h"
+
+/* Heuristic for the number of times to exponentially backoff and
+ * retry sending an LDC trigger when EAGAIN is encountered
+ */
+#define	VNET_MAX_RETRIES	10
+
+MODULE_AUTHOR("David S. Miller <davem@davemloft.net>");
+MODULE_DESCRIPTION("Sun LDOM virtual network support library");
+MODULE_LICENSE("GPL");
+MODULE_VERSION("1.1");
+
+static int __vnet_tx_trigger(struct vnet_port *port, u32 start);
+
+static inline u32 vnet_tx_dring_avail(struct vio_dring_state *dr)
+{
+	return vio_dring_avail(dr, VNET_TX_RING_SIZE);
+}
+
+static int vnet_handle_unknown(struct vnet_port *port, void *arg)
+{
+	struct vio_msg_tag *pkt = arg;
+
+	pr_err("Received unknown msg [%02x:%02x:%04x:%08x]\n",
+	       pkt->type, pkt->stype, pkt->stype_env, pkt->sid);
+	pr_err("Resetting connection\n");
+
+	ldc_disconnect(port->vio.lp);
+
+	return -ECONNRESET;
+}
+
+static int vnet_port_alloc_tx_ring(struct vnet_port *port);
+
+int sunvnet_send_attr_common(struct vio_driver_state *vio)
+{
+	struct vnet_port *port = to_vnet_port(vio);
+	struct net_device *dev = VNET_PORT_TO_NET_DEVICE(port);
+	struct vio_net_attr_info pkt;
+	int framelen = ETH_FRAME_LEN;
+	int i, err;
+
+	err = vnet_port_alloc_tx_ring(to_vnet_port(vio));
+	if (err)
+		return err;
+
+	memset(&pkt, 0, sizeof(pkt));
+	pkt.tag.type = VIO_TYPE_CTRL;
+	pkt.tag.stype = VIO_SUBTYPE_INFO;
+	pkt.tag.stype_env = VIO_ATTR_INFO;
+	pkt.tag.sid = vio_send_sid(vio);
+	if (vio_version_before(vio, 1, 2))
+		pkt.xfer_mode = VIO_DRING_MODE;
+	else
+		pkt.xfer_mode = VIO_NEW_DRING_MODE;
+	pkt.addr_type = VNET_ADDR_ETHERMAC;
+	pkt.ack_freq = 0;
+	for (i = 0; i < 6; i++)
+		pkt.addr |= (u64)dev->dev_addr[i] << ((5 - i) * 8);
+	if (vio_version_after(vio, 1, 3)) {
+		if (port->rmtu) {
+			port->rmtu = min(VNET_MAXPACKET, port->rmtu);
+			pkt.mtu = port->rmtu;
+		} else {
+			port->rmtu = VNET_MAXPACKET;
+			pkt.mtu = port->rmtu;
+		}
+		if (vio_version_after_eq(vio, 1, 6))
+			pkt.options = VIO_TX_DRING;
+	} else if (vio_version_before(vio, 1, 3)) {
+		pkt.mtu = framelen;
+	} else { /* v1.3 */
+		pkt.mtu = framelen + VLAN_HLEN;
+	}
+
+	pkt.cflags = 0;
+	if (vio_version_after_eq(vio, 1, 7) && port->tso) {
+		pkt.cflags |= VNET_LSO_IPV4_CAPAB;
+		if (!port->tsolen)
+			port->tsolen = VNET_MAXTSO;
+		pkt.ipv4_lso_maxlen = port->tsolen;
+	}
+
+	pkt.plnk_updt = PHYSLINK_UPDATE_NONE;
+
+	viodbg(HS, "SEND NET ATTR xmode[0x%x] atype[0x%x] addr[%llx] "
+	       "ackfreq[%u] plnk_updt[0x%02x] opts[0x%02x] mtu[%llu] "
+	       "cflags[0x%04x] lso_max[%u]\n",
+	       pkt.xfer_mode, pkt.addr_type,
+	       (unsigned long long)pkt.addr,
+	       pkt.ack_freq, pkt.plnk_updt, pkt.options,
+	       (unsigned long long)pkt.mtu, pkt.cflags, pkt.ipv4_lso_maxlen);
+
+	return vio_ldc_send(vio, &pkt, sizeof(pkt));
+}
+EXPORT_SYMBOL_GPL(sunvnet_send_attr_common);
+
+static int handle_attr_info(struct vio_driver_state *vio,
+			    struct vio_net_attr_info *pkt)
+{
+	struct vnet_port *port = to_vnet_port(vio);
+	u64	localmtu;
+	u8	xfer_mode;
+
+	viodbg(HS, "GOT NET ATTR xmode[0x%x] atype[0x%x] addr[%llx] "
+	       "ackfreq[%u] plnk_updt[0x%02x] opts[0x%02x] mtu[%llu] "
+	       " (rmtu[%llu]) cflags[0x%04x] lso_max[%u]\n",
+	       pkt->xfer_mode, pkt->addr_type,
+	       (unsigned long long)pkt->addr,
+	       pkt->ack_freq, pkt->plnk_updt, pkt->options,
+	       (unsigned long long)pkt->mtu, port->rmtu, pkt->cflags,
+	       pkt->ipv4_lso_maxlen);
+
+	pkt->tag.sid = vio_send_sid(vio);
+
+	xfer_mode = pkt->xfer_mode;
+	/* for version < 1.2, VIO_DRING_MODE = 0x3 and no bitmask */
+	if (vio_version_before(vio, 1, 2) && xfer_mode == VIO_DRING_MODE)
+		xfer_mode = VIO_NEW_DRING_MODE;
+
+	/* MTU negotiation:
+	 *	< v1.3 - ETH_FRAME_LEN exactly
+	 *	> v1.3 - MIN(pkt.mtu, VNET_MAXPACKET, port->rmtu) and change
+	 *			pkt->mtu for ACK
+	 *	= v1.3 - ETH_FRAME_LEN + VLAN_HLEN exactly
+	 */
+	if (vio_version_before(vio, 1, 3)) {
+		localmtu = ETH_FRAME_LEN;
+	} else if (vio_version_after(vio, 1, 3)) {
+		localmtu = port->rmtu ? port->rmtu : VNET_MAXPACKET;
+		localmtu = min(pkt->mtu, localmtu);
+		pkt->mtu = localmtu;
+	} else { /* v1.3 */
+		localmtu = ETH_FRAME_LEN + VLAN_HLEN;
+	}
+	port->rmtu = localmtu;
+
+	/* LSO negotiation */
+	if (vio_version_after_eq(vio, 1, 7))
+		port->tso &= !!(pkt->cflags & VNET_LSO_IPV4_CAPAB);
+	else
+		port->tso = false;
+	if (port->tso) {
+		if (!port->tsolen)
+			port->tsolen = VNET_MAXTSO;
+		port->tsolen = min(port->tsolen, pkt->ipv4_lso_maxlen);
+		if (port->tsolen < VNET_MINTSO) {
+			port->tso = false;
+			port->tsolen = 0;
+			pkt->cflags &= ~VNET_LSO_IPV4_CAPAB;
+		}
+		pkt->ipv4_lso_maxlen = port->tsolen;
+	} else {
+		pkt->cflags &= ~VNET_LSO_IPV4_CAPAB;
+		pkt->ipv4_lso_maxlen = 0;
+		port->tsolen = 0;
+	}
+
+	/* for version >= 1.6, ACK packet mode we support */
+	if (vio_version_after_eq(vio, 1, 6)) {
+		pkt->xfer_mode = VIO_NEW_DRING_MODE;
+		pkt->options = VIO_TX_DRING;
+	}
+
+	if (!(xfer_mode | VIO_NEW_DRING_MODE) ||
+	    pkt->addr_type != VNET_ADDR_ETHERMAC ||
+	    pkt->mtu != localmtu) {
+		viodbg(HS, "SEND NET ATTR NACK\n");
+
+		pkt->tag.stype = VIO_SUBTYPE_NACK;
+
+		(void)vio_ldc_send(vio, pkt, sizeof(*pkt));
+
+		return -ECONNRESET;
+	}
+
+	viodbg(HS, "SEND NET ATTR ACK xmode[0x%x] atype[0x%x] "
+	       "addr[%llx] ackfreq[%u] plnk_updt[0x%02x] opts[0x%02x] "
+	       "mtu[%llu] (rmtu[%llu]) cflags[0x%04x] lso_max[%u]\n",
+	       pkt->xfer_mode, pkt->addr_type,
+	       (unsigned long long)pkt->addr,
+	       pkt->ack_freq, pkt->plnk_updt, pkt->options,
+	       (unsigned long long)pkt->mtu, port->rmtu, pkt->cflags,
+	       pkt->ipv4_lso_maxlen);
+
+	pkt->tag.stype = VIO_SUBTYPE_ACK;
+
+	return vio_ldc_send(vio, pkt, sizeof(*pkt));
+}
+
+static int handle_attr_ack(struct vio_driver_state *vio,
+			   struct vio_net_attr_info *pkt)
+{
+	viodbg(HS, "GOT NET ATTR ACK\n");
+
+	return 0;
+}
+
+static int handle_attr_nack(struct vio_driver_state *vio,
+			    struct vio_net_attr_info *pkt)
+{
+	viodbg(HS, "GOT NET ATTR NACK\n");
+
+	return -ECONNRESET;
+}
+
+int sunvnet_handle_attr_common(struct vio_driver_state *vio, void *arg)
+{
+	struct vio_net_attr_info *pkt = arg;
+
+	switch (pkt->tag.stype) {
+	case VIO_SUBTYPE_INFO:
+		return handle_attr_info(vio, pkt);
+
+	case VIO_SUBTYPE_ACK:
+		return handle_attr_ack(vio, pkt);
+
+	case VIO_SUBTYPE_NACK:
+		return handle_attr_nack(vio, pkt);
+
+	default:
+		return -ECONNRESET;
+	}
+}
+EXPORT_SYMBOL_GPL(sunvnet_handle_attr_common);
+
+void sunvnet_handshake_complete_common(struct vio_driver_state *vio)
+{
+	struct vio_dring_state *dr;
+
+	dr = &vio->drings[VIO_DRIVER_RX_RING];
+	dr->rcv_nxt = 1;
+	dr->snd_nxt = 1;
+
+	dr = &vio->drings[VIO_DRIVER_TX_RING];
+	dr->rcv_nxt = 1;
+	dr->snd_nxt = 1;
+}
+EXPORT_SYMBOL_GPL(sunvnet_handshake_complete_common);
+
+/* The hypervisor interface that implements copying to/from imported
+ * memory from another domain requires that copies are done to 8-byte
+ * aligned buffers, and that the lengths of such copies are also 8-byte
+ * multiples.
+ *
+ * So we align skb->data to an 8-byte multiple and pad-out the data
+ * area so we can round the copy length up to the next multiple of
+ * 8 for the copy.
+ *
+ * The transmitter puts the actual start of the packet 6 bytes into
+ * the buffer it sends over, so that the IP headers after the ethernet
+ * header are aligned properly.  These 6 bytes are not in the descriptor
+ * length, they are simply implied.  This offset is represented using
+ * the VNET_PACKET_SKIP macro.
+ */
+static struct sk_buff *alloc_and_align_skb(struct net_device *dev,
+					   unsigned int len)
+{
+	struct sk_buff *skb;
+	unsigned long addr, off;
+
+	skb = netdev_alloc_skb(dev, len + VNET_PACKET_SKIP + 8 + 8);
+	if (unlikely(!skb))
+		return NULL;
+
+	addr = (unsigned long)skb->data;
+	off = ((addr + 7UL) & ~7UL) - addr;
+	if (off)
+		skb_reserve(skb, off);
+
+	return skb;
+}
+
+static inline void vnet_fullcsum_ipv4(struct sk_buff *skb)
+{
+	struct iphdr *iph = ip_hdr(skb);
+	int offset = skb_transport_offset(skb);
+
+	if (skb->protocol != htons(ETH_P_IP))
+		return;
+	if (iph->protocol != IPPROTO_TCP &&
+	    iph->protocol != IPPROTO_UDP)
+		return;
+	skb->ip_summed = CHECKSUM_NONE;
+	skb->csum_level = 1;
+	skb->csum = 0;
+	if (iph->protocol == IPPROTO_TCP) {
+		struct tcphdr *ptcp = tcp_hdr(skb);
+
+		ptcp->check = 0;
+		skb->csum = skb_checksum(skb, offset, skb->len - offset, 0);
+		ptcp->check = csum_tcpudp_magic(iph->saddr, iph->daddr,
+						skb->len - offset, IPPROTO_TCP,
+						skb->csum);
+	} else if (iph->protocol == IPPROTO_UDP) {
+		struct udphdr *pudp = udp_hdr(skb);
+
+		pudp->check = 0;
+		skb->csum = skb_checksum(skb, offset, skb->len - offset, 0);
+		pudp->check = csum_tcpudp_magic(iph->saddr, iph->daddr,
+						skb->len - offset, IPPROTO_UDP,
+						skb->csum);
+	}
+}
+
+#if IS_ENABLED(CONFIG_IPV6)
+static inline void vnet_fullcsum_ipv6(struct sk_buff *skb)
+{
+	struct ipv6hdr *ip6h = ipv6_hdr(skb);
+	int offset = skb_transport_offset(skb);
+
+	if (skb->protocol != htons(ETH_P_IPV6))
+		return;
+	if (ip6h->nexthdr != IPPROTO_TCP &&
+	    ip6h->nexthdr != IPPROTO_UDP)
+		return;
+	skb->ip_summed = CHECKSUM_NONE;
+	skb->csum_level = 1;
+	skb->csum = 0;
+	if (ip6h->nexthdr == IPPROTO_TCP) {
+		struct tcphdr *ptcp = tcp_hdr(skb);
+
+		ptcp->check = 0;
+		skb->csum = skb_checksum(skb, offset, skb->len - offset, 0);
+		ptcp->check = csum_ipv6_magic(&ip6h->saddr, &ip6h->daddr,
+					      skb->len - offset, IPPROTO_TCP,
+					      skb->csum);
+	} else if (ip6h->nexthdr == IPPROTO_UDP) {
+		struct udphdr *pudp = udp_hdr(skb);
+
+		pudp->check = 0;
+		skb->csum = skb_checksum(skb, offset, skb->len - offset, 0);
+		pudp->check = csum_ipv6_magic(&ip6h->saddr, &ip6h->daddr,
+					      skb->len - offset, IPPROTO_UDP,
+					      skb->csum);
+	}
+}
+#endif
+
+static int vnet_rx_one(struct vnet_port *port, struct vio_net_desc *desc)
+{
+	struct net_device *dev = VNET_PORT_TO_NET_DEVICE(port);
+	unsigned int len = desc->size;
+	unsigned int copy_len;
+	struct sk_buff *skb;
+	int maxlen;
+	int err;
+
+	err = -EMSGSIZE;
+	if (port->tso && port->tsolen > port->rmtu)
+		maxlen = port->tsolen;
+	else
+		maxlen = port->rmtu;
+	if (unlikely(len < ETH_ZLEN || len > maxlen)) {
+		dev->stats.rx_length_errors++;
+		goto out_dropped;
+	}
+
+	skb = alloc_and_align_skb(dev, len);
+	err = -ENOMEM;
+	if (unlikely(!skb)) {
+		dev->stats.rx_missed_errors++;
+		goto out_dropped;
+	}
+
+	copy_len = (len + VNET_PACKET_SKIP + 7U) & ~7U;
+	skb_put(skb, copy_len);
+	err = ldc_copy(port->vio.lp, LDC_COPY_IN,
+		       skb->data, copy_len, 0,
+		       desc->cookies, desc->ncookies);
+	if (unlikely(err < 0)) {
+		dev->stats.rx_frame_errors++;
+		goto out_free_skb;
+	}
+
+	skb_pull(skb, VNET_PACKET_SKIP);
+	skb_trim(skb, len);
+	skb->protocol = eth_type_trans(skb, dev);
+
+	if (vio_version_after_eq(&port->vio, 1, 8)) {
+		struct vio_net_dext *dext = vio_net_ext(desc);
+
+		skb_reset_network_header(skb);
+
+		if (dext->flags & VNET_PKT_HCK_IPV4_HDRCKSUM) {
+			if (skb->protocol == ETH_P_IP) {
+				struct iphdr *iph = ip_hdr(skb);
+
+				iph->check = 0;
+				ip_send_check(iph);
+			}
+		}
+		if ((dext->flags & VNET_PKT_HCK_FULLCKSUM) &&
+		    skb->ip_summed == CHECKSUM_NONE) {
+			if (skb->protocol == htons(ETH_P_IP)) {
+				struct iphdr *iph = ip_hdr(skb);
+				int ihl = iph->ihl * 4;
+
+				skb_set_transport_header(skb, ihl);
+				vnet_fullcsum_ipv4(skb);
+#if IS_ENABLED(CONFIG_IPV6)
+			} else if (skb->protocol == htons(ETH_P_IPV6)) {
+				skb_set_transport_header(skb,
+							 sizeof(struct ipv6hdr));
+				vnet_fullcsum_ipv6(skb);
+#endif
+			}
+		}
+		if (dext->flags & VNET_PKT_HCK_IPV4_HDRCKSUM_OK) {
+			skb->ip_summed = CHECKSUM_PARTIAL;
+			skb->csum_level = 0;
+			if (dext->flags & VNET_PKT_HCK_FULLCKSUM_OK)
+				skb->csum_level = 1;
+		}
+	}
+
+	skb->ip_summed = port->switch_port ? CHECKSUM_NONE : CHECKSUM_PARTIAL;
+
+	if (unlikely(is_multicast_ether_addr(eth_hdr(skb)->h_dest)))
+		dev->stats.multicast++;
+	dev->stats.rx_packets++;
+	dev->stats.rx_bytes += len;
+	port->stats.rx_packets++;
+	port->stats.rx_bytes += len;
+	napi_gro_receive(&port->napi, skb);
+	return 0;
+
+out_free_skb:
+	kfree_skb(skb);
+
+out_dropped:
+	dev->stats.rx_dropped++;
+	return err;
+}
+
+static int vnet_send_ack(struct vnet_port *port, struct vio_dring_state *dr,
+			 u32 start, u32 end, u8 vio_dring_state)
+{
+	struct vio_dring_data hdr = {
+		.tag = {
+			.type		= VIO_TYPE_DATA,
+			.stype		= VIO_SUBTYPE_ACK,
+			.stype_env	= VIO_DRING_DATA,
+			.sid		= vio_send_sid(&port->vio),
+		},
+		.dring_ident		= dr->ident,
+		.start_idx		= start,
+		.end_idx		= end,
+		.state			= vio_dring_state,
+	};
+	int err, delay;
+	int retries = 0;
+
+	hdr.seq = dr->snd_nxt;
+	delay = 1;
+	do {
+		err = vio_ldc_send(&port->vio, &hdr, sizeof(hdr));
+		if (err > 0) {
+			dr->snd_nxt++;
+			break;
+		}
+		udelay(delay);
+		if ((delay <<= 1) > 128)
+			delay = 128;
+		if (retries++ > VNET_MAX_RETRIES) {
+			pr_info("ECONNRESET %x:%x:%x:%x:%x:%x\n",
+				port->raddr[0], port->raddr[1],
+				port->raddr[2], port->raddr[3],
+				port->raddr[4], port->raddr[5]);
+			break;
+		}
+	} while (err == -EAGAIN);
+
+	if (err <= 0 && vio_dring_state == VIO_DRING_STOPPED) {
+		port->stop_rx_idx = end;
+		port->stop_rx = true;
+	} else {
+		port->stop_rx_idx = 0;
+		port->stop_rx = false;
+	}
+
+	return err;
+}
+
+static struct vio_net_desc *get_rx_desc(struct vnet_port *port,
+					struct vio_dring_state *dr,
+					u32 index)
+{
+	struct vio_net_desc *desc = port->vio.desc_buf;
+	int err;
+
+	err = ldc_get_dring_entry(port->vio.lp, desc, dr->entry_size,
+				  (index * dr->entry_size),
+				  dr->cookies, dr->ncookies);
+	if (err < 0)
+		return ERR_PTR(err);
+
+	return desc;
+}
+
+static int put_rx_desc(struct vnet_port *port,
+		       struct vio_dring_state *dr,
+		       struct vio_net_desc *desc,
+		       u32 index)
+{
+	int err;
+
+	err = ldc_put_dring_entry(port->vio.lp, desc, dr->entry_size,
+				  (index * dr->entry_size),
+				  dr->cookies, dr->ncookies);
+	if (err < 0)
+		return err;
+
+	return 0;
+}
+
+static int vnet_walk_rx_one(struct vnet_port *port,
+			    struct vio_dring_state *dr,
+			    u32 index, int *needs_ack)
+{
+	struct vio_net_desc *desc = get_rx_desc(port, dr, index);
+	struct vio_driver_state *vio = &port->vio;
+	int err;
+
+	BUG_ON(!desc);
+	if (IS_ERR(desc))
+		return PTR_ERR(desc);
+
+	if (desc->hdr.state != VIO_DESC_READY)
+		return 1;
+
+	dma_rmb();
+
+	viodbg(DATA, "vio_walk_rx_one desc[%02x:%02x:%08x:%08x:%llx:%llx]\n",
+	       desc->hdr.state, desc->hdr.ack,
+	       desc->size, desc->ncookies,
+	       desc->cookies[0].cookie_addr,
+	       desc->cookies[0].cookie_size);
+
+	err = vnet_rx_one(port, desc);
+	if (err == -ECONNRESET)
+		return err;
+	trace_vnet_rx_one(port->vio._local_sid, port->vio._peer_sid,
+			  index, desc->hdr.ack);
+	desc->hdr.state = VIO_DESC_DONE;
+	err = put_rx_desc(port, dr, desc, index);
+	if (err < 0)
+		return err;
+	*needs_ack = desc->hdr.ack;
+	return 0;
+}
+
+static int vnet_walk_rx(struct vnet_port *port, struct vio_dring_state *dr,
+			u32 start, u32 end, int *npkts, int budget)
+{
+	struct vio_driver_state *vio = &port->vio;
+	int ack_start = -1, ack_end = -1;
+	bool send_ack = true;
+
+	end = (end == (u32)-1) ? vio_dring_prev(dr, start)
+			       : vio_dring_next(dr, end);
+
+	viodbg(DATA, "vnet_walk_rx start[%08x] end[%08x]\n", start, end);
+
+	while (start != end) {
+		int ack = 0, err = vnet_walk_rx_one(port, dr, start, &ack);
+
+		if (err == -ECONNRESET)
+			return err;
+		if (err != 0)
+			break;
+		(*npkts)++;
+		if (ack_start == -1)
+			ack_start = start;
+		ack_end = start;
+		start = vio_dring_next(dr, start);
+		if (ack && start != end) {
+			err = vnet_send_ack(port, dr, ack_start, ack_end,
+					    VIO_DRING_ACTIVE);
+			if (err == -ECONNRESET)
+				return err;
+			ack_start = -1;
+		}
+		if ((*npkts) >= budget) {
+			send_ack = false;
+			break;
+		}
+	}
+	if (unlikely(ack_start == -1)) {
+		ack_end = vio_dring_prev(dr, start);
+		ack_start = ack_end;
+	}
+	if (send_ack) {
+		port->napi_resume = false;
+		trace_vnet_tx_send_stopped_ack(port->vio._local_sid,
+					       port->vio._peer_sid,
+					       ack_end, *npkts);
+		return vnet_send_ack(port, dr, ack_start, ack_end,
+				     VIO_DRING_STOPPED);
+	} else  {
+		trace_vnet_tx_defer_stopped_ack(port->vio._local_sid,
+						port->vio._peer_sid,
+						ack_end, *npkts);
+		port->napi_resume = true;
+		port->napi_stop_idx = ack_end;
+		return 1;
+	}
+}
+
+static int vnet_rx(struct vnet_port *port, void *msgbuf, int *npkts,
+		   int budget)
+{
+	struct vio_dring_data *pkt = msgbuf;
+	struct vio_dring_state *dr = &port->vio.drings[VIO_DRIVER_RX_RING];
+	struct vio_driver_state *vio = &port->vio;
+
+	viodbg(DATA, "vnet_rx stype_env[%04x] seq[%016llx] rcv_nxt[%016llx]\n",
+	       pkt->tag.stype_env, pkt->seq, dr->rcv_nxt);
+
+	if (unlikely(pkt->tag.stype_env != VIO_DRING_DATA))
+		return 0;
+	if (unlikely(pkt->seq != dr->rcv_nxt)) {
+		pr_err("RX out of sequence seq[0x%llx] rcv_nxt[0x%llx]\n",
+		       pkt->seq, dr->rcv_nxt);
+		return 0;
+	}
+
+	if (!port->napi_resume)
+		dr->rcv_nxt++;
+
+	/* XXX Validate pkt->start_idx and pkt->end_idx XXX */
+
+	return vnet_walk_rx(port, dr, pkt->start_idx, pkt->end_idx,
+			    npkts, budget);
+}
+
+static int idx_is_pending(struct vio_dring_state *dr, u32 end)
+{
+	u32 idx = dr->cons;
+	int found = 0;
+
+	while (idx != dr->prod) {
+		if (idx == end) {
+			found = 1;
+			break;
+		}
+		idx = vio_dring_next(dr, idx);
+	}
+	return found;
+}
+
+static int vnet_ack(struct vnet_port *port, void *msgbuf)
+{
+	struct vio_dring_state *dr = &port->vio.drings[VIO_DRIVER_TX_RING];
+	struct vio_dring_data *pkt = msgbuf;
+	struct net_device *dev;
+	u32 end;
+	struct vio_net_desc *desc;
+	struct netdev_queue *txq;
+
+	if (unlikely(pkt->tag.stype_env != VIO_DRING_DATA))
+		return 0;
+
+	end = pkt->end_idx;
+	dev = VNET_PORT_TO_NET_DEVICE(port);
+	netif_tx_lock(dev);
+	if (unlikely(!idx_is_pending(dr, end))) {
+		netif_tx_unlock(dev);
+		return 0;
+	}
+
+	/* sync for race conditions with vnet_start_xmit() and tell xmit it
+	 * is time to send a trigger.
+	 */
+	trace_vnet_rx_stopped_ack(port->vio._local_sid,
+				  port->vio._peer_sid, end);
+	dr->cons = vio_dring_next(dr, end);
+	desc = vio_dring_entry(dr, dr->cons);
+	if (desc->hdr.state == VIO_DESC_READY && !port->start_cons) {
+		/* vnet_start_xmit() just populated this dring but missed
+		 * sending the "start" LDC message to the consumer.
+		 * Send a "start" trigger on its behalf.
+		 */
+		if (__vnet_tx_trigger(port, dr->cons) > 0)
+			port->start_cons = false;
+		else
+			port->start_cons = true;
+	} else {
+		port->start_cons = true;
+	}
+	netif_tx_unlock(dev);
+
+	txq = netdev_get_tx_queue(dev, port->q_index);
+	if (unlikely(netif_tx_queue_stopped(txq) &&
+		     vnet_tx_dring_avail(dr) >= VNET_TX_WAKEUP_THRESH(dr)))
+		return 1;
+
+	return 0;
+}
+
+static int vnet_nack(struct vnet_port *port, void *msgbuf)
+{
+	/* XXX just reset or similar XXX */
+	return 0;
+}
+
+static int handle_mcast(struct vnet_port *port, void *msgbuf)
+{
+	struct vio_net_mcast_info *pkt = msgbuf;
+	struct net_device *dev = VNET_PORT_TO_NET_DEVICE(port);
+
+	if (pkt->tag.stype != VIO_SUBTYPE_ACK)
+		pr_err("%s: Got unexpected MCAST reply [%02x:%02x:%04x:%08x]\n",
+		       dev->name,
+		       pkt->tag.type,
+		       pkt->tag.stype,
+		       pkt->tag.stype_env,
+		       pkt->tag.sid);
+
+	return 0;
+}
+
+/* If the queue is stopped, wake it up so that we'll
+ * send out another START message at the next TX.
+ */
+static void maybe_tx_wakeup(struct vnet_port *port)
+{
+	struct netdev_queue *txq;
+
+	txq = netdev_get_tx_queue(VNET_PORT_TO_NET_DEVICE(port),
+				  port->q_index);
+	__netif_tx_lock(txq, smp_processor_id());
+	if (likely(netif_tx_queue_stopped(txq)))
+		netif_tx_wake_queue(txq);
+	__netif_tx_unlock(txq);
+}
+
+bool sunvnet_port_is_up_common(struct vnet_port *vnet)
+{
+	struct vio_driver_state *vio = &vnet->vio;
+
+	return !!(vio->hs_state & VIO_HS_COMPLETE);
+}
+EXPORT_SYMBOL_GPL(sunvnet_port_is_up_common);
+
+static int vnet_event_napi(struct vnet_port *port, int budget)
+{
+	struct net_device *dev = VNET_PORT_TO_NET_DEVICE(port);
+	struct vio_driver_state *vio = &port->vio;
+	int tx_wakeup, err;
+	int npkts = 0;
+
+	/* we don't expect any other bits */
+	BUG_ON(port->rx_event & ~(LDC_EVENT_DATA_READY |
+				  LDC_EVENT_RESET |
+				  LDC_EVENT_UP));
+
+	/* RESET takes precedent over any other event */
+	if (port->rx_event & LDC_EVENT_RESET) {
+		/* a link went down */
+
+		if (port->vsw == 1) {
+			netif_tx_stop_all_queues(dev);
+			netif_carrier_off(dev);
+		}
+
+		vio_link_state_change(vio, LDC_EVENT_RESET);
+		vnet_port_reset(port);
+		vio_port_up(vio);
+
+		/* If the device is running but its tx queue was
+		 * stopped (due to flow control), restart it.
+		 * This is necessary since vnet_port_reset()
+		 * clears the tx drings and thus we may never get
+		 * back a VIO_TYPE_DATA ACK packet - which is
+		 * the normal mechanism to restart the tx queue.
+		 */
+		if (netif_running(dev))
+			maybe_tx_wakeup(port);
+
+		port->rx_event = 0;
+		port->stats.event_reset++;
+		return 0;
+	}
+
+	if (port->rx_event & LDC_EVENT_UP) {
+		/* a link came up */
+
+		if (port->vsw == 1) {
+			netif_carrier_on(port->dev);
+			netif_tx_start_all_queues(port->dev);
+		}
+
+		vio_link_state_change(vio, LDC_EVENT_UP);
+		port->rx_event = 0;
+		port->stats.event_up++;
+		return 0;
+	}
+
+	err = 0;
+	tx_wakeup = 0;
+	while (1) {
+		union {
+			struct vio_msg_tag tag;
+			u64 raw[8];
+		} msgbuf;
+
+		if (port->napi_resume) {
+			struct vio_dring_data *pkt =
+				(struct vio_dring_data *)&msgbuf;
+			struct vio_dring_state *dr =
+				&port->vio.drings[VIO_DRIVER_RX_RING];
+
+			pkt->tag.type = VIO_TYPE_DATA;
+			pkt->tag.stype = VIO_SUBTYPE_INFO;
+			pkt->tag.stype_env = VIO_DRING_DATA;
+			pkt->seq = dr->rcv_nxt;
+			pkt->start_idx = vio_dring_next(dr,
+							port->napi_stop_idx);
+			pkt->end_idx = -1;
+		} else {
+			err = ldc_read(vio->lp, &msgbuf, sizeof(msgbuf));
+			if (unlikely(err < 0)) {
+				if (err == -ECONNRESET)
+					vio_conn_reset(vio);
+				break;
+			}
+			if (err == 0)
+				break;
+			viodbg(DATA, "TAG [%02x:%02x:%04x:%08x]\n",
+			       msgbuf.tag.type,
+			       msgbuf.tag.stype,
+			       msgbuf.tag.stype_env,
+			       msgbuf.tag.sid);
+			err = vio_validate_sid(vio, &msgbuf.tag);
+			if (err < 0)
+				break;
+		}
+
+		if (likely(msgbuf.tag.type == VIO_TYPE_DATA)) {
+			if (msgbuf.tag.stype == VIO_SUBTYPE_INFO) {
+				if (!sunvnet_port_is_up_common(port)) {
+					/* failures like handshake_failure()
+					 * may have cleaned up dring, but
+					 * NAPI polling may bring us here.
+					 */
+					err = -ECONNRESET;
+					break;
+				}
+				err = vnet_rx(port, &msgbuf, &npkts, budget);
+				if (npkts >= budget)
+					break;
+				if (npkts == 0)
+					break;
+			} else if (msgbuf.tag.stype == VIO_SUBTYPE_ACK) {
+				err = vnet_ack(port, &msgbuf);
+				if (err > 0)
+					tx_wakeup |= err;
+			} else if (msgbuf.tag.stype == VIO_SUBTYPE_NACK) {
+				err = vnet_nack(port, &msgbuf);
+			}
+		} else if (msgbuf.tag.type == VIO_TYPE_CTRL) {
+			if (msgbuf.tag.stype_env == VNET_MCAST_INFO)
+				err = handle_mcast(port, &msgbuf);
+			else
+				err = vio_control_pkt_engine(vio, &msgbuf);
+			if (err)
+				break;
+		} else {
+			err = vnet_handle_unknown(port, &msgbuf);
+		}
+		if (err == -ECONNRESET)
+			break;
+	}
+	if (unlikely(tx_wakeup && err != -ECONNRESET))
+		maybe_tx_wakeup(port);
+	return npkts;
+}
+
+int sunvnet_poll_common(struct napi_struct *napi, int budget)
+{
+	struct vnet_port *port = container_of(napi, struct vnet_port, napi);
+	struct vio_driver_state *vio = &port->vio;
+	int processed = vnet_event_napi(port, budget);
+
+	if (processed < budget) {
+		napi_complete_done(napi, processed);
+		port->rx_event &= ~LDC_EVENT_DATA_READY;
+		vio_set_intr(vio->vdev->rx_ino, HV_INTR_ENABLED);
+	}
+	return processed;
+}
+EXPORT_SYMBOL_GPL(sunvnet_poll_common);
+
+void sunvnet_event_common(void *arg, int event)
+{
+	struct vnet_port *port = arg;
+	struct vio_driver_state *vio = &port->vio;
+
+	port->rx_event |= event;
+	vio_set_intr(vio->vdev->rx_ino, HV_INTR_DISABLED);
+	napi_schedule(&port->napi);
+}
+EXPORT_SYMBOL_GPL(sunvnet_event_common);
+
+static int __vnet_tx_trigger(struct vnet_port *port, u32 start)
+{
+	struct vio_dring_state *dr = &port->vio.drings[VIO_DRIVER_TX_RING];
+	struct vio_dring_data hdr = {
+		.tag = {
+			.type		= VIO_TYPE_DATA,
+			.stype		= VIO_SUBTYPE_INFO,
+			.stype_env	= VIO_DRING_DATA,
+			.sid		= vio_send_sid(&port->vio),
+		},
+		.dring_ident		= dr->ident,
+		.start_idx		= start,
+		.end_idx		= (u32)-1,
+	};
+	int err, delay;
+	int retries = 0;
+
+	if (port->stop_rx) {
+		trace_vnet_tx_pending_stopped_ack(port->vio._local_sid,
+						  port->vio._peer_sid,
+						  port->stop_rx_idx, -1);
+		err = vnet_send_ack(port,
+				    &port->vio.drings[VIO_DRIVER_RX_RING],
+				    port->stop_rx_idx, -1,
+				    VIO_DRING_STOPPED);
+		if (err <= 0)
+			return err;
+	}
+
+	hdr.seq = dr->snd_nxt;
+	delay = 1;
+	do {
+		err = vio_ldc_send(&port->vio, &hdr, sizeof(hdr));
+		if (err > 0) {
+			dr->snd_nxt++;
+			break;
+		}
+		udelay(delay);
+		if ((delay <<= 1) > 128)
+			delay = 128;
+		if (retries++ > VNET_MAX_RETRIES)
+			break;
+	} while (err == -EAGAIN);
+	trace_vnet_tx_trigger(port->vio._local_sid,
+			      port->vio._peer_sid, start, err);
+
+	return err;
+}
+
+static struct sk_buff *vnet_clean_tx_ring(struct vnet_port *port,
+					  unsigned *pending)
+{
+	struct vio_dring_state *dr = &port->vio.drings[VIO_DRIVER_TX_RING];
+	struct sk_buff *skb = NULL;
+	int i, txi;
+
+	*pending = 0;
+
+	txi = dr->prod;
+	for (i = 0; i < VNET_TX_RING_SIZE; ++i) {
+		struct vio_net_desc *d;
+
+		--txi;
+		if (txi < 0)
+			txi = VNET_TX_RING_SIZE - 1;
+
+		d = vio_dring_entry(dr, txi);
+
+		if (d->hdr.state == VIO_DESC_READY) {
+			(*pending)++;
+			continue;
+		}
+		if (port->tx_bufs[txi].skb) {
+			if (d->hdr.state != VIO_DESC_DONE)
+				pr_notice("invalid ring buffer state %d\n",
+					  d->hdr.state);
+			BUG_ON(port->tx_bufs[txi].skb->next);
+
+			port->tx_bufs[txi].skb->next = skb;
+			skb = port->tx_bufs[txi].skb;
+			port->tx_bufs[txi].skb = NULL;
+
+			ldc_unmap(port->vio.lp,
+				  port->tx_bufs[txi].cookies,
+				  port->tx_bufs[txi].ncookies);
+		} else if (d->hdr.state == VIO_DESC_FREE) {
+			break;
+		}
+		d->hdr.state = VIO_DESC_FREE;
+	}
+	return skb;
+}
+
+static inline void vnet_free_skbs(struct sk_buff *skb)
+{
+	struct sk_buff *next;
+
+	while (skb) {
+		next = skb->next;
+		skb->next = NULL;
+		dev_kfree_skb(skb);
+		skb = next;
+	}
+}
+
+void sunvnet_clean_timer_expire_common(struct timer_list *t)
+{
+	struct vnet_port *port = timer_container_of(port, t, clean_timer);
+	struct sk_buff *freeskbs;
+	unsigned pending;
+
+	netif_tx_lock(VNET_PORT_TO_NET_DEVICE(port));
+	freeskbs = vnet_clean_tx_ring(port, &pending);
+	netif_tx_unlock(VNET_PORT_TO_NET_DEVICE(port));
+
+	vnet_free_skbs(freeskbs);
+
+	if (pending)
+		(void)mod_timer(&port->clean_timer,
+				jiffies + VNET_CLEAN_TIMEOUT);
+	 else
+		timer_delete(&port->clean_timer);
+}
+EXPORT_SYMBOL_GPL(sunvnet_clean_timer_expire_common);
+
+static inline int vnet_skb_map(struct ldc_channel *lp, struct sk_buff *skb,
+			       struct ldc_trans_cookie *cookies, int ncookies,
+			       unsigned int map_perm)
+{
+	int i, nc, err, blen;
+
+	/* header */
+	blen = skb_headlen(skb);
+	if (blen < ETH_ZLEN)
+		blen = ETH_ZLEN;
+	blen += VNET_PACKET_SKIP;
+	blen += 8 - (blen & 7);
+
+	err = ldc_map_single(lp, skb->data - VNET_PACKET_SKIP, blen, cookies,
+			     ncookies, map_perm);
+	if (err < 0)
+		return err;
+	nc = err;
+
+	for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
+		skb_frag_t *f = &skb_shinfo(skb)->frags[i];
+		u8 *vaddr;
+
+		if (nc < ncookies) {
+			vaddr = kmap_local_page(skb_frag_page(f));
+			blen = skb_frag_size(f);
+			blen += 8 - (blen & 7);
+			err = ldc_map_single(lp, vaddr + skb_frag_off(f),
+					     blen, cookies + nc, ncookies - nc,
+					     map_perm);
+			kunmap_local(vaddr);
+		} else {
+			err = -EMSGSIZE;
+		}
+
+		if (err < 0) {
+			ldc_unmap(lp, cookies, nc);
+			return err;
+		}
+		nc += err;
+	}
+	return nc;
+}
+
+static inline struct sk_buff *vnet_skb_shape(struct sk_buff *skb, int ncookies)
+{
+	struct sk_buff *nskb;
+	int i, len, pad, docopy;
+
+	len = skb->len;
+	pad = 0;
+	if (len < ETH_ZLEN) {
+		pad += ETH_ZLEN - skb->len;
+		len += pad;
+	}
+	len += VNET_PACKET_SKIP;
+	pad += 8 - (len & 7);
+
+	/* make sure we have enough cookies and alignment in every frag */
+	docopy = skb_shinfo(skb)->nr_frags >= ncookies;
+	for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
+		skb_frag_t *f = &skb_shinfo(skb)->frags[i];
+
+		docopy |= skb_frag_off(f) & 7;
+	}
+	if (((unsigned long)skb->data & 7) != VNET_PACKET_SKIP ||
+	    skb_tailroom(skb) < pad ||
+	    skb_headroom(skb) < VNET_PACKET_SKIP || docopy) {
+		int start = 0, offset;
+		__wsum csum;
+
+		len = skb->len > ETH_ZLEN ? skb->len : ETH_ZLEN;
+		nskb = alloc_and_align_skb(skb->dev, len);
+		if (!nskb) {
+			dev_kfree_skb(skb);
+			return NULL;
+		}
+		skb_reserve(nskb, VNET_PACKET_SKIP);
+
+		nskb->protocol = skb->protocol;
+		offset = skb_mac_header(skb) - skb->data;
+		skb_set_mac_header(nskb, offset);
+		offset = skb_network_offset(skb);
+		skb_set_network_header(nskb, offset);
+		offset = skb_transport_offset(skb);
+		skb_set_transport_header(nskb, offset);
+
+		offset = 0;
+		nskb->csum_offset = skb->csum_offset;
+		nskb->ip_summed = skb->ip_summed;
+
+		if (skb->ip_summed == CHECKSUM_PARTIAL)
+			start = skb_checksum_start_offset(skb);
+		if (start) {
+			int offset = start + nskb->csum_offset;
+
+			/* copy the headers, no csum here */
+			if (skb_copy_bits(skb, 0, nskb->data, start)) {
+				dev_kfree_skb(nskb);
+				dev_kfree_skb(skb);
+				return NULL;
+			}
+
+			/* copy the rest, with csum calculation */
+			*(__sum16 *)(skb->data + offset) = 0;
+			csum = skb_copy_and_csum_bits(skb, start,
+						      nskb->data + start,
+						      skb->len - start);
+
+			/* add in the header checksums */
+			if (skb->protocol == htons(ETH_P_IP)) {
+				struct iphdr *iph = ip_hdr(nskb);
+
+				if (iph->protocol == IPPROTO_TCP ||
+				    iph->protocol == IPPROTO_UDP) {
+					csum = csum_tcpudp_magic(iph->saddr,
+								 iph->daddr,
+								 skb->len - start,
+								 iph->protocol,
+								 csum);
+				}
+			} else if (skb->protocol == htons(ETH_P_IPV6)) {
+				struct ipv6hdr *ip6h = ipv6_hdr(nskb);
+
+				if (ip6h->nexthdr == IPPROTO_TCP ||
+				    ip6h->nexthdr == IPPROTO_UDP) {
+					csum = csum_ipv6_magic(&ip6h->saddr,
+							       &ip6h->daddr,
+							       skb->len - start,
+							       ip6h->nexthdr,
+							       csum);
+				}
+			}
+
+			/* save the final result */
+			*(__sum16 *)(nskb->data + offset) = csum;
+
+			nskb->ip_summed = CHECKSUM_NONE;
+		} else if (skb_copy_bits(skb, 0, nskb->data, skb->len)) {
+			dev_kfree_skb(nskb);
+			dev_kfree_skb(skb);
+			return NULL;
+		}
+		(void)skb_put(nskb, skb->len);
+		if (skb_is_gso(skb)) {
+			skb_shinfo(nskb)->gso_size = skb_shinfo(skb)->gso_size;
+			skb_shinfo(nskb)->gso_type = skb_shinfo(skb)->gso_type;
+		}
+		nskb->queue_mapping = skb->queue_mapping;
+		dev_kfree_skb(skb);
+		skb = nskb;
+	}
+	return skb;
+}
+
+static netdev_tx_t
+vnet_handle_offloads(struct vnet_port *port, struct sk_buff *skb,
+		     struct vnet_port *(*vnet_tx_port)
+		     (struct sk_buff *, struct net_device *))
+{
+	struct net_device *dev = VNET_PORT_TO_NET_DEVICE(port);
+	struct vio_dring_state *dr = &port->vio.drings[VIO_DRIVER_TX_RING];
+	struct sk_buff *segs, *curr, *next;
+	int maclen, datalen;
+	int status;
+	int gso_size, gso_type, gso_segs;
+	int hlen = skb_transport_header(skb) - skb_mac_header(skb);
+	int proto = IPPROTO_IP;
+
+	if (skb->protocol == htons(ETH_P_IP))
+		proto = ip_hdr(skb)->protocol;
+	else if (skb->protocol == htons(ETH_P_IPV6))
+		proto = ipv6_hdr(skb)->nexthdr;
+
+	if (proto == IPPROTO_TCP) {
+		hlen += tcp_hdr(skb)->doff * 4;
+	} else if (proto == IPPROTO_UDP) {
+		hlen += sizeof(struct udphdr);
+	} else {
+		pr_err("vnet_handle_offloads GSO with unknown transport "
+		       "protocol %d tproto %d\n", skb->protocol, proto);
+		hlen = 128; /* XXX */
+	}
+	datalen = port->tsolen - hlen;
+
+	gso_size = skb_shinfo(skb)->gso_size;
+	gso_type = skb_shinfo(skb)->gso_type;
+	gso_segs = skb_shinfo(skb)->gso_segs;
+
+	if (port->tso && gso_size < datalen)
+		gso_segs = DIV_ROUND_UP(skb->len - hlen, datalen);
+
+	if (unlikely(vnet_tx_dring_avail(dr) < gso_segs)) {
+		struct netdev_queue *txq;
+
+		txq  = netdev_get_tx_queue(dev, port->q_index);
+		netif_tx_stop_queue(txq);
+		if (vnet_tx_dring_avail(dr) < skb_shinfo(skb)->gso_segs)
+			return NETDEV_TX_BUSY;
+		netif_tx_wake_queue(txq);
+	}
+
+	maclen = skb_network_header(skb) - skb_mac_header(skb);
+	skb_pull(skb, maclen);
+
+	if (port->tso && gso_size < datalen) {
+		if (skb_unclone(skb, GFP_ATOMIC))
+			goto out_dropped;
+
+		/* segment to TSO size */
+		skb_shinfo(skb)->gso_size = datalen;
+		skb_shinfo(skb)->gso_segs = gso_segs;
+	}
+	segs = skb_gso_segment(skb, dev->features & ~NETIF_F_TSO);
+	if (IS_ERR(segs))
+		goto out_dropped;
+
+	skb_push(skb, maclen);
+	skb_reset_mac_header(skb);
+
+	status = 0;
+	skb_list_walk_safe(segs, curr, next) {
+		skb_mark_not_on_list(curr);
+		if (port->tso && curr->len > dev->mtu) {
+			skb_shinfo(curr)->gso_size = gso_size;
+			skb_shinfo(curr)->gso_type = gso_type;
+			skb_shinfo(curr)->gso_segs =
+				DIV_ROUND_UP(curr->len - hlen, gso_size);
+		} else {
+			skb_shinfo(curr)->gso_size = 0;
+		}
+
+		skb_push(curr, maclen);
+		skb_reset_mac_header(curr);
+		memcpy(skb_mac_header(curr), skb_mac_header(skb),
+		       maclen);
+		curr->csum_start = skb_transport_header(curr) - curr->head;
+		if (ip_hdr(curr)->protocol == IPPROTO_TCP)
+			curr->csum_offset = offsetof(struct tcphdr, check);
+		else if (ip_hdr(curr)->protocol == IPPROTO_UDP)
+			curr->csum_offset = offsetof(struct udphdr, check);
+
+		if (!(status & NETDEV_TX_MASK))
+			status = sunvnet_start_xmit_common(curr, dev,
+							   vnet_tx_port);
+		if (status & NETDEV_TX_MASK)
+			dev_kfree_skb_any(curr);
+	}
+
+	if (!(status & NETDEV_TX_MASK))
+		dev_kfree_skb_any(skb);
+	return status;
+out_dropped:
+	dev->stats.tx_dropped++;
+	dev_kfree_skb_any(skb);
+	return NETDEV_TX_OK;
+}
+
+netdev_tx_t
+sunvnet_start_xmit_common(struct sk_buff *skb, struct net_device *dev,
+			  struct vnet_port *(*vnet_tx_port)
+			  (struct sk_buff *, struct net_device *))
+{
+	struct vnet_port *port = NULL;
+	struct vio_dring_state *dr;
+	struct vio_net_desc *d;
+	unsigned int len;
+	struct sk_buff *freeskbs = NULL;
+	int i, err, txi;
+	unsigned pending = 0;
+	struct netdev_queue *txq;
+
+	rcu_read_lock();
+	port = vnet_tx_port(skb, dev);
+	if (unlikely(!port))
+		goto out_dropped;
+
+	if (skb_is_gso(skb) && skb->len > port->tsolen) {
+		err = vnet_handle_offloads(port, skb, vnet_tx_port);
+		rcu_read_unlock();
+		return err;
+	}
+
+	if (!skb_is_gso(skb) && skb->len > port->rmtu) {
+		unsigned long localmtu = port->rmtu - ETH_HLEN;
+
+		if (vio_version_after_eq(&port->vio, 1, 3))
+			localmtu -= VLAN_HLEN;
+
+		if (skb->protocol == htons(ETH_P_IP))
+			icmp_ndo_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
+				      htonl(localmtu));
+#if IS_ENABLED(CONFIG_IPV6)
+		else if (skb->protocol == htons(ETH_P_IPV6))
+			icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, localmtu);
+#endif
+		goto out_dropped;
+	}
+
+	skb = vnet_skb_shape(skb, 2);
+
+	if (unlikely(!skb))
+		goto out_dropped;
+
+	if (skb->ip_summed == CHECKSUM_PARTIAL) {
+		if (skb->protocol == htons(ETH_P_IP))
+			vnet_fullcsum_ipv4(skb);
+#if IS_ENABLED(CONFIG_IPV6)
+		else if (skb->protocol == htons(ETH_P_IPV6))
+			vnet_fullcsum_ipv6(skb);
+#endif
+	}
+
+	dr = &port->vio.drings[VIO_DRIVER_TX_RING];
+	i = skb_get_queue_mapping(skb);
+	txq = netdev_get_tx_queue(dev, i);
+	if (unlikely(vnet_tx_dring_avail(dr) < 1)) {
+		if (!netif_tx_queue_stopped(txq)) {
+			netif_tx_stop_queue(txq);
+
+			/* This is a hard error, log it. */
+			netdev_err(dev, "BUG! Tx Ring full when queue awake!\n");
+			dev->stats.tx_errors++;
+		}
+		rcu_read_unlock();
+		return NETDEV_TX_BUSY;
+	}
+
+	d = vio_dring_cur(dr);
+
+	txi = dr->prod;
+
+	freeskbs = vnet_clean_tx_ring(port, &pending);
+
+	BUG_ON(port->tx_bufs[txi].skb);
+
+	len = skb->len;
+	if (len < ETH_ZLEN)
+		len = ETH_ZLEN;
+
+	err = vnet_skb_map(port->vio.lp, skb, port->tx_bufs[txi].cookies, 2,
+			   (LDC_MAP_SHADOW | LDC_MAP_DIRECT | LDC_MAP_RW));
+	if (err < 0) {
+		netdev_info(dev, "tx buffer map error %d\n", err);
+		goto out_dropped;
+	}
+
+	port->tx_bufs[txi].skb = skb;
+	skb = NULL;
+	port->tx_bufs[txi].ncookies = err;
+
+	/* We don't rely on the ACKs to free the skb in vnet_start_xmit(),
+	 * thus it is safe to not set VIO_ACK_ENABLE for each transmission:
+	 * the protocol itself does not require it as long as the peer
+	 * sends a VIO_SUBTYPE_ACK for VIO_DRING_STOPPED.
+	 *
+	 * An ACK for every packet in the ring is expensive as the
+	 * sending of LDC messages is slow and affects performance.
+	 */
+	d->hdr.ack = VIO_ACK_DISABLE;
+	d->size = len;
+	d->ncookies = port->tx_bufs[txi].ncookies;
+	for (i = 0; i < d->ncookies; i++)
+		d->cookies[i] = port->tx_bufs[txi].cookies[i];
+	if (vio_version_after_eq(&port->vio, 1, 7)) {
+		struct vio_net_dext *dext = vio_net_ext(d);
+
+		memset(dext, 0, sizeof(*dext));
+		if (skb_is_gso(port->tx_bufs[txi].skb)) {
+			dext->ipv4_lso_mss = skb_shinfo(port->tx_bufs[txi].skb)
+					     ->gso_size;
+			dext->flags |= VNET_PKT_IPV4_LSO;
+		}
+		if (vio_version_after_eq(&port->vio, 1, 8) &&
+		    !port->switch_port) {
+			dext->flags |= VNET_PKT_HCK_IPV4_HDRCKSUM_OK;
+			dext->flags |= VNET_PKT_HCK_FULLCKSUM_OK;
+		}
+	}
+
+	/* This has to be a non-SMP write barrier because we are writing
+	 * to memory which is shared with the peer LDOM.
+	 */
+	dma_wmb();
+
+	d->hdr.state = VIO_DESC_READY;
+
+	/* Exactly one ldc "start" trigger (for dr->cons) needs to be sent
+	 * to notify the consumer that some descriptors are READY.
+	 * After that "start" trigger, no additional triggers are needed until
+	 * a DRING_STOPPED is received from the consumer. The dr->cons field
+	 * (set up by vnet_ack()) has the value of the next dring index
+	 * that has not yet been ack-ed. We send a "start" trigger here
+	 * if, and only if, start_cons is true (reset it afterward). Conversely,
+	 * vnet_ack() should check if the dring corresponding to cons
+	 * is marked READY, but start_cons was false.
+	 * If so, vnet_ack() should send out the missed "start" trigger.
+	 *
+	 * Note that the dma_wmb() above makes sure the cookies et al. are
+	 * not globally visible before the VIO_DESC_READY, and that the
+	 * stores are ordered correctly by the compiler. The consumer will
+	 * not proceed until the VIO_DESC_READY is visible assuring that
+	 * the consumer does not observe anything related to descriptors
+	 * out of order. The HV trap from the LDC start trigger is the
+	 * producer to consumer announcement that work is available to the
+	 * consumer
+	 */
+	if (!port->start_cons) { /* previous trigger suffices */
+		trace_vnet_skip_tx_trigger(port->vio._local_sid,
+					   port->vio._peer_sid, dr->cons);
+		goto ldc_start_done;
+	}
+
+	err = __vnet_tx_trigger(port, dr->cons);
+	if (unlikely(err < 0)) {
+		netdev_info(dev, "TX trigger error %d\n", err);
+		d->hdr.state = VIO_DESC_FREE;
+		skb = port->tx_bufs[txi].skb;
+		port->tx_bufs[txi].skb = NULL;
+		dev->stats.tx_carrier_errors++;
+		goto out_dropped;
+	}
+
+ldc_start_done:
+	port->start_cons = false;
+
+	dev->stats.tx_packets++;
+	dev->stats.tx_bytes += port->tx_bufs[txi].skb->len;
+	port->stats.tx_packets++;
+	port->stats.tx_bytes += port->tx_bufs[txi].skb->len;
+
+	dr->prod = (dr->prod + 1) & (VNET_TX_RING_SIZE - 1);
+	if (unlikely(vnet_tx_dring_avail(dr) < 1)) {
+		netif_tx_stop_queue(txq);
+		smp_rmb();
+		if (vnet_tx_dring_avail(dr) > VNET_TX_WAKEUP_THRESH(dr))
+			netif_tx_wake_queue(txq);
+	}
+
+	(void)mod_timer(&port->clean_timer, jiffies + VNET_CLEAN_TIMEOUT);
+	rcu_read_unlock();
+
+	vnet_free_skbs(freeskbs);
+
+	return NETDEV_TX_OK;
+
+out_dropped:
+	if (pending)
+		(void)mod_timer(&port->clean_timer,
+				jiffies + VNET_CLEAN_TIMEOUT);
+	else if (port)
+		timer_delete(&port->clean_timer);
+	rcu_read_unlock();
+	dev_kfree_skb(skb);
+	vnet_free_skbs(freeskbs);
+	dev->stats.tx_dropped++;
+	return NETDEV_TX_OK;
+}
+EXPORT_SYMBOL_GPL(sunvnet_start_xmit_common);
+
+void sunvnet_tx_timeout_common(struct net_device *dev, unsigned int txqueue)
+{
+	/* XXX Implement me XXX */
+}
+EXPORT_SYMBOL_GPL(sunvnet_tx_timeout_common);
+
+int sunvnet_open_common(struct net_device *dev)
+{
+	netif_carrier_on(dev);
+	netif_tx_start_all_queues(dev);
+
+	return 0;
+}
+EXPORT_SYMBOL_GPL(sunvnet_open_common);
+
+int sunvnet_close_common(struct net_device *dev)
+{
+	netif_tx_stop_all_queues(dev);
+	netif_carrier_off(dev);
+
+	return 0;
+}
+EXPORT_SYMBOL_GPL(sunvnet_close_common);
+
+static struct vnet_mcast_entry *__vnet_mc_find(struct vnet *vp, u8 *addr)
+{
+	struct vnet_mcast_entry *m;
+
+	for (m = vp->mcast_list; m; m = m->next) {
+		if (ether_addr_equal(m->addr, addr))
+			return m;
+	}
+	return NULL;
+}
+
+static void __update_mc_list(struct vnet *vp, struct net_device *dev)
+{
+	struct netdev_hw_addr *ha;
+
+	netdev_for_each_mc_addr(ha, dev) {
+		struct vnet_mcast_entry *m;
+
+		m = __vnet_mc_find(vp, ha->addr);
+		if (m) {
+			m->hit = 1;
+			continue;
+		}
+
+		if (!m) {
+			m = kzalloc_obj(*m, GFP_ATOMIC);
+			if (!m)
+				continue;
+			memcpy(m->addr, ha->addr, ETH_ALEN);
+			m->hit = 1;
+
+			m->next = vp->mcast_list;
+			vp->mcast_list = m;
+		}
+	}
+}
+
+static void __send_mc_list(struct vnet *vp, struct vnet_port *port)
+{
+	struct vio_net_mcast_info info;
+	struct vnet_mcast_entry *m, **pp;
+	int n_addrs;
+
+	memset(&info, 0, sizeof(info));
+
+	info.tag.type = VIO_TYPE_CTRL;
+	info.tag.stype = VIO_SUBTYPE_INFO;
+	info.tag.stype_env = VNET_MCAST_INFO;
+	info.tag.sid = vio_send_sid(&port->vio);
+	info.set = 1;
+
+	n_addrs = 0;
+	for (m = vp->mcast_list; m; m = m->next) {
+		if (m->sent)
+			continue;
+		m->sent = 1;
+		memcpy(&info.mcast_addr[n_addrs * ETH_ALEN],
+		       m->addr, ETH_ALEN);
+		if (++n_addrs == VNET_NUM_MCAST) {
+			info.count = n_addrs;
+
+			(void)vio_ldc_send(&port->vio, &info,
+					   sizeof(info));
+			n_addrs = 0;
+		}
+	}
+	if (n_addrs) {
+		info.count = n_addrs;
+		(void)vio_ldc_send(&port->vio, &info, sizeof(info));
+	}
+
+	info.set = 0;
+
+	n_addrs = 0;
+	pp = &vp->mcast_list;
+	while ((m = *pp) != NULL) {
+		if (m->hit) {
+			m->hit = 0;
+			pp = &m->next;
+			continue;
+		}
+
+		memcpy(&info.mcast_addr[n_addrs * ETH_ALEN],
+		       m->addr, ETH_ALEN);
+		if (++n_addrs == VNET_NUM_MCAST) {
+			info.count = n_addrs;
+			(void)vio_ldc_send(&port->vio, &info,
+					   sizeof(info));
+			n_addrs = 0;
+		}
+
+		*pp = m->next;
+		kfree(m);
+	}
+	if (n_addrs) {
+		info.count = n_addrs;
+		(void)vio_ldc_send(&port->vio, &info, sizeof(info));
+	}
+}
+
+void sunvnet_set_rx_mode_common(struct net_device *dev, struct vnet *vp)
+{
+	struct vnet_port *port;
+
+	rcu_read_lock();
+	list_for_each_entry_rcu(port, &vp->port_list, list) {
+		if (port->switch_port) {
+			__update_mc_list(vp, dev);
+			__send_mc_list(vp, port);
+			break;
+		}
+	}
+	rcu_read_unlock();
+}
+EXPORT_SYMBOL_GPL(sunvnet_set_rx_mode_common);
+
+int sunvnet_set_mac_addr_common(struct net_device *dev, void *p)
+{
+	return -EINVAL;
+}
+EXPORT_SYMBOL_GPL(sunvnet_set_mac_addr_common);
+
+void sunvnet_port_free_tx_bufs_common(struct vnet_port *port)
+{
+	struct vio_dring_state *dr;
+	int i;
+
+	dr = &port->vio.drings[VIO_DRIVER_TX_RING];
+
+	if (!dr->base)
+		return;
+
+	for (i = 0; i < VNET_TX_RING_SIZE; i++) {
+		struct vio_net_desc *d;
+		void *skb = port->tx_bufs[i].skb;
+
+		if (!skb)
+			continue;
+
+		d = vio_dring_entry(dr, i);
+
+		ldc_unmap(port->vio.lp,
+			  port->tx_bufs[i].cookies,
+			  port->tx_bufs[i].ncookies);
+		dev_kfree_skb(skb);
+		port->tx_bufs[i].skb = NULL;
+		d->hdr.state = VIO_DESC_FREE;
+	}
+	ldc_free_exp_dring(port->vio.lp, dr->base,
+			   (dr->entry_size * dr->num_entries),
+			   dr->cookies, dr->ncookies);
+	dr->base = NULL;
+	dr->entry_size = 0;
+	dr->num_entries = 0;
+	dr->pending = 0;
+	dr->ncookies = 0;
+}
+EXPORT_SYMBOL_GPL(sunvnet_port_free_tx_bufs_common);
+
+void vnet_port_reset(struct vnet_port *port)
+{
+	timer_delete(&port->clean_timer);
+	sunvnet_port_free_tx_bufs_common(port);
+	port->rmtu = 0;
+	port->tso = (port->vsw == 0);  /* no tso in vsw, misbehaves in bridge */
+	port->tsolen = 0;
+}
+EXPORT_SYMBOL_GPL(vnet_port_reset);
+
+static int vnet_port_alloc_tx_ring(struct vnet_port *port)
+{
+	struct vio_dring_state *dr;
+	unsigned long len, elen;
+	int i, err, ncookies;
+	void *dring;
+
+	dr = &port->vio.drings[VIO_DRIVER_TX_RING];
+
+	elen = sizeof(struct vio_net_desc) +
+	       sizeof(struct ldc_trans_cookie) * 2;
+	if (vio_version_after_eq(&port->vio, 1, 7))
+		elen += sizeof(struct vio_net_dext);
+	len = VNET_TX_RING_SIZE * elen;
+
+	ncookies = VIO_MAX_RING_COOKIES;
+	dring = ldc_alloc_exp_dring(port->vio.lp, len,
+				    dr->cookies, &ncookies,
+				    (LDC_MAP_SHADOW |
+				     LDC_MAP_DIRECT |
+				     LDC_MAP_RW));
+	if (IS_ERR(dring)) {
+		err = PTR_ERR(dring);
+		goto err_out;
+	}
+
+	dr->base = dring;
+	dr->entry_size = elen;
+	dr->num_entries = VNET_TX_RING_SIZE;
+	dr->prod = 0;
+	dr->cons = 0;
+	port->start_cons  = true; /* need an initial trigger */
+	dr->pending = VNET_TX_RING_SIZE;
+	dr->ncookies = ncookies;
+
+	for (i = 0; i < VNET_TX_RING_SIZE; ++i) {
+		struct vio_net_desc *d;
+
+		d = vio_dring_entry(dr, i);
+		d->hdr.state = VIO_DESC_FREE;
+	}
+	return 0;
+
+err_out:
+	sunvnet_port_free_tx_bufs_common(port);
+
+	return err;
+}
+
+#ifdef CONFIG_NET_POLL_CONTROLLER
+void sunvnet_poll_controller_common(struct net_device *dev, struct vnet *vp)
+{
+	struct vnet_port *port;
+	unsigned long flags;
+
+	spin_lock_irqsave(&vp->lock, flags);
+	if (!list_empty(&vp->port_list)) {
+		port = list_entry(vp->port_list.next, struct vnet_port, list);
+		napi_schedule(&port->napi);
+	}
+	spin_unlock_irqrestore(&vp->lock, flags);
+}
+EXPORT_SYMBOL_GPL(sunvnet_poll_controller_common);
+#endif
+
+void sunvnet_port_add_txq_common(struct vnet_port *port)
+{
+	struct vnet *vp = port->vp;
+	int smallest = 0;
+	int i;
+
+	/* find the first least-used q
+	 * When there are more ldoms than q's, we start to
+	 * double up on ports per queue.
+	 */
+	for (i = 0; i < VNET_MAX_TXQS; i++) {
+		if (vp->q_used[i] == 0) {
+			smallest = i;
+			break;
+		}
+		if (vp->q_used[i] < vp->q_used[smallest])
+			smallest = i;
+	}
+
+	vp->nports++;
+	vp->q_used[smallest]++;
+	port->q_index = smallest;
+}
+EXPORT_SYMBOL_GPL(sunvnet_port_add_txq_common);
+
+void sunvnet_port_rm_txq_common(struct vnet_port *port)
+{
+	port->vp->nports--;
+	port->vp->q_used[port->q_index]--;
+	port->q_index = 0;
+}
+EXPORT_SYMBOL_GPL(sunvnet_port_rm_txq_common);
diff --git a/drivers/net/ethernet/sun/sunvnet_common.h b/drivers/net/ethernet/sun/sunvnet_common.h
new file mode 100644
index 000000000..5416a3cb9
--- /dev/null
+++ b/drivers/net/ethernet/sun/sunvnet_common.h
@@ -0,0 +1,157 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+#ifndef _SUNVNETCOMMON_H
+#define _SUNVNETCOMMON_H
+
+#include <linux/interrupt.h>
+
+/* length of time (or less) we expect pending descriptors to be marked
+ * as VIO_DESC_DONE and skbs ready to be freed
+ */
+#define	VNET_CLEAN_TIMEOUT		((HZ / 100) + 1)
+
+#define VNET_MAXPACKET			(65535ULL + ETH_HLEN + VLAN_HLEN)
+#define VNET_TX_RING_SIZE		512
+#define VNET_TX_WAKEUP_THRESH(dr)	((dr)->pending / 4)
+
+#define	VNET_MINTSO	 2048	/* VIO protocol's minimum TSO len */
+#define	VNET_MAXTSO	65535	/* VIO protocol's maximum TSO len */
+
+#define VNET_MAX_MTU	65535
+
+/* VNET packets are sent in buffers with the first 6 bytes skipped
+ * so that after the ethernet header the IPv4/IPv6 headers are aligned
+ * properly.
+ */
+#define VNET_PACKET_SKIP		6
+
+#define	VNET_MAXCOOKIES			(VNET_MAXPACKET / PAGE_SIZE + 1)
+
+#define	VNET_MAX_TXQS		16
+
+struct vnet_tx_entry {
+	struct sk_buff		*skb;
+	unsigned int		ncookies;
+	struct ldc_trans_cookie	cookies[VNET_MAXCOOKIES];
+};
+
+struct vnet;
+
+struct vnet_port_stats {
+	/* keep them all the same size */
+	u32 rx_bytes;
+	u32 tx_bytes;
+	u32 rx_packets;
+	u32 tx_packets;
+	u32 event_up;
+	u32 event_reset;
+	u32 q_placeholder;
+};
+
+#define NUM_VNET_PORT_STATS  (sizeof(struct vnet_port_stats) / sizeof(u32))
+
+/* Structure to describe a vnet-port or vsw-port in the MD.
+ * If the vsw bit is set, this structure represents a vswitch
+ * port, and the net_device can be found from ->dev. If the
+ * vsw bit is not set, the net_device is available from ->vp->dev.
+ * See the VNET_PORT_TO_NET_DEVICE macro below.
+ */
+struct vnet_port {
+	struct vio_driver_state	vio;
+
+	struct vnet_port_stats stats;
+
+	struct hlist_node	hash;
+	u8			raddr[ETH_ALEN];
+	unsigned		switch_port:1;
+	unsigned		tso:1;
+	unsigned		vsw:1;
+	unsigned		__pad:13;
+
+	struct vnet		*vp;
+	struct net_device	*dev;
+
+	struct vnet_tx_entry	tx_bufs[VNET_TX_RING_SIZE];
+
+	struct list_head	list;
+
+	u32			stop_rx_idx;
+	bool			stop_rx;
+	bool			start_cons;
+
+	struct timer_list	clean_timer;
+
+	u64			rmtu;
+	u16			tsolen;
+
+	struct napi_struct	napi;
+	u32			napi_stop_idx;
+	bool			napi_resume;
+	int			rx_event;
+	u16			q_index;
+};
+
+static inline struct vnet_port *to_vnet_port(struct vio_driver_state *vio)
+{
+	return container_of(vio, struct vnet_port, vio);
+}
+
+#define VNET_PORT_HASH_SIZE	16
+#define VNET_PORT_HASH_MASK	(VNET_PORT_HASH_SIZE - 1)
+
+static inline unsigned int vnet_hashfn(u8 *mac)
+{
+	unsigned int val = mac[4] ^ mac[5];
+
+	return val & (VNET_PORT_HASH_MASK);
+}
+
+struct vnet_mcast_entry {
+	u8			addr[ETH_ALEN];
+	u8			sent;
+	u8			hit;
+	struct vnet_mcast_entry	*next;
+};
+
+struct vnet {
+	spinlock_t		lock; /* Protects port_list and port_hash.  */
+	struct net_device	*dev;
+	u32			msg_enable;
+	u8			q_used[VNET_MAX_TXQS];
+	struct list_head	port_list;
+	struct hlist_head	port_hash[VNET_PORT_HASH_SIZE];
+	struct vnet_mcast_entry	*mcast_list;
+	struct list_head	list;
+	u64			local_mac;
+	int			nports;
+};
+
+/* Def used by common code to get the net_device from the proper location */
+#define VNET_PORT_TO_NET_DEVICE(__port) \
+	((__port)->vsw ? (__port)->dev : (__port)->vp->dev)
+
+/* Common funcs */
+void sunvnet_clean_timer_expire_common(struct timer_list *t);
+int sunvnet_open_common(struct net_device *dev);
+int sunvnet_close_common(struct net_device *dev);
+void sunvnet_set_rx_mode_common(struct net_device *dev, struct vnet *vp);
+int sunvnet_set_mac_addr_common(struct net_device *dev, void *p);
+void sunvnet_tx_timeout_common(struct net_device *dev, unsigned int txqueue);
+netdev_tx_t
+sunvnet_start_xmit_common(struct sk_buff *skb, struct net_device *dev,
+			  struct vnet_port *(*vnet_tx_port)
+			  (struct sk_buff *, struct net_device *));
+#ifdef CONFIG_NET_POLL_CONTROLLER
+void sunvnet_poll_controller_common(struct net_device *dev, struct vnet *vp);
+#endif
+void sunvnet_event_common(void *arg, int event);
+int sunvnet_send_attr_common(struct vio_driver_state *vio);
+int sunvnet_handle_attr_common(struct vio_driver_state *vio, void *arg);
+void sunvnet_handshake_complete_common(struct vio_driver_state *vio);
+int sunvnet_poll_common(struct napi_struct *napi, int budget);
+void sunvnet_port_free_tx_bufs_common(struct vnet_port *port);
+void vnet_port_reset(struct vnet_port *port);
+bool sunvnet_port_is_up_common(struct vnet_port *vnet);
+void sunvnet_port_add_txq_common(struct vnet_port *port);
+void sunvnet_port_rm_txq_common(struct vnet_port *port);
+
+#endif /* _SUNVNETCOMMON_H */