Merge tag 'trace-v7.3-2' of git://git.kernel.org/pub/scm/linux/kernel/git/trace/linux-tracegrafted

Pull tracing fixes from Steven Rostedt:

 - Fix error output of boot instance creation failure

   Currently if a boot instance creation fails, instead of printing out
   the name of the instance that failed, it prints "(null)". That is
   because it prints "cur_str" that had already been processed by
   strsep(). Print the saved name instead.

   While at it, print the error code of the failure.

 - Fix use-after-free for same named historgrams

   Histograms can be named so that they can be used in multiple events.
   But if the named histogram has a variable attached, the second event
   that uses the named histogram which duplicates it and needs to free
   the original after duplication leaves the old variable in place and
   still visible. If another histogram uses than variable, it will use
   the stale one which will try to reference the freed duplicate
   histogram and crash the kernel.

   Free the duplicate variables along with the duplicated histogram
   data.

 - Check return value of kthread_run() in event self test

   The events self tests uses a kthread for testing but does not check
   if it succeeded in creating a kthread. If the kthread creation were
   to fail, the code will still try to call kthread_stop() on the error
   returned.

 - Fix race between reading trace_pipe and updating subbuffer size

   If a user is reading the trace_pipe file at the same time they update
   the ring buffer sub-buffer size, can cause the trace_pipe read to
   read stale data. Add trace_access_lock() around updating the ring
   buffer sub-buffer size.

 - Fix eventfs_inode on failure path in creation of the events directory

   In the creation of the "events" directory, if after allocating the
   eventfs_inode a failure is detected, it calls cleanup_ei() which
   calls free_ei(). The free_ei() will test if eventfs_inode being freed
   has no children. It is a bug if it does. But on the failure case of
   the creation of the "events" directory, the children lists have not
   yet been initialized and the free will trigger a warning because
   list_empty() on an uninitialized list returns false.

   Move the initialization into init_ei() where it makes more sense and
   makes sure that a created eventfs_inode has its lists initialized
   upon creation.

 - Check return value of kthread_run() in ftrace direct sample code

   The sample code that shows how to use the ftrace direct calls does
   not test the return of kthread_run() to see if it succeeds. Return a
   failure if the kthread_run() doesn't succeed.

 - Clear user events state on fork in case of alloc failure

   On fork, the child gets a pointer to the parent's user events state.
   It makes a copy of it then updates the child's pointer to it. But if
   the allocation fails, the duplication function leaves the child with
   a pointer to its parent's descriptor. When the child cleans up its
   data, it will free the parent's descriptor while the parent is still
   using it.

   In the duplication function, set the child's user_event_mm to NULL
   before testing if the allocation succeeded, and when it exits it will
   not free the parent's descriptor.

 - Fix retry exhaustion in simple ring buffer reader swap

   simple_ring_buffer_swap_reader_page() starts with retry set to 8 and
   post-decrements it only after a failed link replacement. On the final
   attempt, a successful replacement leaves retry at zero, while a
   failed replacement leaves it at -1.

   But the check for success expects the retry value to be non-zero and
   exits with an error on zero. This is the opposite result. Fix it.

 - Fail nicely when the remote swap_reader_page() returns an error

   Currently, if the swap_reader_page() of a remote buffer fails, it
   triggers a WARN_ON_ONCE() and continues normally. Instead, have it
   exit with an error and a pr_warn() print instead of a full WARNING.

* tag 'trace-v7.3-2' of git://git.kernel.org/pub/scm/linux/kernel/git/trace/linux-trace:
  ring-buffer: Stop remote reader update when page swap fails
  tracing: Fix retry exhaustion in simple ring buffer reader swap
  tracing/user_events: Clear copied tracing state before fork duplication
  samples/ftrace: Fix kthread_stop() on ERR_PTR in ftrace-direct-multi-modify
  samples/ftrace: Fix kthread_stop() on ERR_PTR in ftrace-direct-modify
  eventfs: Initialize ei->children and ei->list in init_ei()
  tracing: Fix use-after-free in trace_pipe read on sub-buffer order change
  tracing: Fix crash passing ERR_PTR to kthread_stop()
  tracing: Fix use-after-free with same-name named triggers
  tracing: Fix logged instance name on creation failure

1 files changed, 1573 insertions, 0 deletions

diff --git a/drivers/net/fddi/defza.c b/drivers/net/fddi/defza.c
new file mode 100644
index 000000000..9bfecc87d
--- /dev/null
+++ b/drivers/net/fddi/defza.c
@@ -0,0 +1,1573 @@
+// SPDX-License-Identifier: GPL-2.0+
+/*	FDDI network adapter driver for DEC FDDIcontroller 700/700-C devices.
+ *
+ *	Copyright (c) 2018  Maciej W. Rozycki
+ *
+ *	This program is free software; you can redistribute it and/or
+ *	modify it under the terms of the GNU General Public License
+ *	as published by the Free Software Foundation; either version
+ *	2 of the License, or (at your option) any later version.
+ *
+ *	References:
+ *
+ *	Dave Sawyer & Phil Weeks & Frank Itkowsky,
+ *	"DEC FDDIcontroller 700 Port Specification",
+ *	Revision 1.1, Digital Equipment Corporation
+ */
+
+/* ------------------------------------------------------------------------- */
+/* FZA configurable parameters.                                              */
+
+/* The number of transmit ring descriptors; either 0 for 512 or 1 for 1024.  */
+#define FZA_RING_TX_MODE 0
+
+/* The number of receive ring descriptors; from 2 up to 256.  */
+#define FZA_RING_RX_SIZE 256
+
+/* End of FZA configurable parameters.  No need to change anything below.    */
+/* ------------------------------------------------------------------------- */
+
+#include <linux/delay.h>
+#include <linux/device.h>
+#include <linux/dma-mapping.h>
+#include <linux/init.h>
+#include <linux/interrupt.h>
+#include <linux/io.h>
+#include <linux/io-64-nonatomic-lo-hi.h>
+#include <linux/ioport.h>
+#include <linux/kernel.h>
+#include <linux/list.h>
+#include <linux/module.h>
+#include <linux/netdevice.h>
+#include <linux/fddidevice.h>
+#include <linux/sched.h>
+#include <linux/skbuff.h>
+#include <linux/spinlock.h>
+#include <linux/stat.h>
+#include <linux/tc.h>
+#include <linux/timer.h>
+#include <linux/types.h>
+#include <linux/wait.h>
+
+#include <asm/barrier.h>
+
+#include "defza.h"
+
+#define DRV_NAME "defza"
+#define DRV_VERSION "v.1.1.4"
+#define DRV_RELDATE "Oct  6 2018"
+
+static const char version[] =
+	DRV_NAME ": " DRV_VERSION "  " DRV_RELDATE "  Maciej W. Rozycki\n";
+
+MODULE_AUTHOR("Maciej W. Rozycki <macro@orcam.me.uk>");
+MODULE_DESCRIPTION("DEC FDDIcontroller 700 (DEFZA-xx) driver");
+MODULE_LICENSE("GPL");
+
+static int loopback;
+module_param(loopback, int, 0644);
+
+/* Ring Purger Multicast */
+static u8 hw_addr_purger[8] = { 0x09, 0x00, 0x2b, 0x02, 0x01, 0x05 };
+/* Directed Beacon Multicast */
+static u8 hw_addr_beacon[8] = { 0x01, 0x80, 0xc2, 0x00, 0x01, 0x00 };
+
+/* Shorthands for MMIO accesses that we require to be strongly ordered
+ * WRT preceding MMIO accesses.
+ */
+#define readw_o readw_relaxed
+#define readl_o readl_relaxed
+
+#define writew_o writew_relaxed
+#define writel_o writel_relaxed
+
+/* Shorthands for MMIO accesses that we are happy with being weakly ordered
+ * WRT preceding MMIO accesses.
+ */
+#define readw_u readw_relaxed
+#define readl_u readl_relaxed
+#define readq_u readq_relaxed
+
+#define writew_u writew_relaxed
+#define writel_u writel_relaxed
+#define writeq_u writeq_relaxed
+
+static inline struct sk_buff *fza_alloc_skb_irq(struct net_device *dev,
+						unsigned int length)
+{
+	return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
+}
+
+static inline struct sk_buff *fza_alloc_skb(struct net_device *dev,
+					    unsigned int length)
+{
+	return __netdev_alloc_skb(dev, length, GFP_KERNEL);
+}
+
+static inline void fza_skb_align(struct sk_buff *skb, unsigned int v)
+{
+	unsigned long x, y;
+
+	x = (unsigned long)skb->data;
+	y = ALIGN(x, v);
+
+	skb_reserve(skb, y - x);
+}
+
+static inline void fza_reads(const void __iomem *from, void *to,
+			     unsigned long size)
+{
+	if (sizeof(unsigned long) == 8) {
+		const u64 __iomem *src = from;
+		const u32 __iomem *src_trail;
+		u64 *dst = to;
+		u32 *dst_trail;
+
+		for (size = (size + 3) / 4; size > 1; size -= 2)
+			*dst++ = readq_u(src++);
+		if (size) {
+			src_trail = (u32 __iomem *)src;
+			dst_trail = (u32 *)dst;
+			*dst_trail = readl_u(src_trail);
+		}
+	} else {
+		const u32 __iomem *src = from;
+		u32 *dst = to;
+
+		for (size = (size + 3) / 4; size; size--)
+			*dst++ = readl_u(src++);
+	}
+}
+
+static inline void fza_writes(const void *from, void __iomem *to,
+			      unsigned long size)
+{
+	if (sizeof(unsigned long) == 8) {
+		const u64 *src = from;
+		const u32 *src_trail;
+		u64 __iomem *dst = to;
+		u32 __iomem *dst_trail;
+
+		for (size = (size + 3) / 4; size > 1; size -= 2)
+			writeq_u(*src++, dst++);
+		if (size) {
+			src_trail = (u32 *)src;
+			dst_trail = (u32 __iomem *)dst;
+			writel_u(*src_trail, dst_trail);
+		}
+	} else {
+		const u32 *src = from;
+		u32 __iomem *dst = to;
+
+		for (size = (size + 3) / 4; size; size--)
+			writel_u(*src++, dst++);
+	}
+}
+
+static inline void fza_moves(const void __iomem *from, void __iomem *to,
+			     unsigned long size)
+{
+	if (sizeof(unsigned long) == 8) {
+		const u64 __iomem *src = from;
+		const u32 __iomem *src_trail;
+		u64 __iomem *dst = to;
+		u32 __iomem *dst_trail;
+
+		for (size = (size + 3) / 4; size > 1; size -= 2)
+			writeq_u(readq_u(src++), dst++);
+		if (size) {
+			src_trail = (u32 __iomem *)src;
+			dst_trail = (u32 __iomem *)dst;
+			writel_u(readl_u(src_trail), dst_trail);
+		}
+	} else {
+		const u32 __iomem *src = from;
+		u32 __iomem *dst = to;
+
+		for (size = (size + 3) / 4; size; size--)
+			writel_u(readl_u(src++), dst++);
+	}
+}
+
+static inline void fza_zeros(void __iomem *to, unsigned long size)
+{
+	if (sizeof(unsigned long) == 8) {
+		u64 __iomem *dst = to;
+		u32 __iomem *dst_trail;
+
+		for (size = (size + 3) / 4; size > 1; size -= 2)
+			writeq_u(0, dst++);
+		if (size) {
+			dst_trail = (u32 __iomem *)dst;
+			writel_u(0, dst_trail);
+		}
+	} else {
+		u32 __iomem *dst = to;
+
+		for (size = (size + 3) / 4; size; size--)
+			writel_u(0, dst++);
+	}
+}
+
+static inline void fza_regs_dump(struct fza_private *fp)
+{
+	pr_debug("%s: iomem registers:\n", fp->name);
+	pr_debug(" reset:           0x%04x\n", readw_o(&fp->regs->reset));
+	pr_debug(" interrupt event: 0x%04x\n", readw_u(&fp->regs->int_event));
+	pr_debug(" status:          0x%04x\n", readw_u(&fp->regs->status));
+	pr_debug(" interrupt mask:  0x%04x\n", readw_u(&fp->regs->int_mask));
+	pr_debug(" control A:       0x%04x\n", readw_u(&fp->regs->control_a));
+	pr_debug(" control B:       0x%04x\n", readw_u(&fp->regs->control_b));
+}
+
+static inline void fza_do_reset(struct fza_private *fp)
+{
+	/* Reset the board. */
+	writew_o(FZA_RESET_INIT, &fp->regs->reset);
+	readw_o(&fp->regs->reset);	/* Synchronize. */
+	readw_o(&fp->regs->reset);	/* Read it back for a small delay. */
+	writew_o(FZA_RESET_CLR, &fp->regs->reset);
+
+	/* Enable all interrupt events we handle. */
+	writew_o(fp->int_mask, &fp->regs->int_mask);
+	readw_o(&fp->regs->int_mask);	/* Synchronize. */
+}
+
+static inline void fza_do_shutdown(struct fza_private *fp)
+{
+	/* Disable the driver mode. */
+	writew_o(FZA_CONTROL_B_IDLE, &fp->regs->control_b);
+
+	/* And reset the board. */
+	writew_o(FZA_RESET_INIT, &fp->regs->reset);
+	readw_o(&fp->regs->reset);	/* Synchronize. */
+	writew_o(FZA_RESET_CLR, &fp->regs->reset);
+	readw_o(&fp->regs->reset);	/* Synchronize. */
+}
+
+static int fza_reset(struct fza_private *fp)
+{
+	unsigned long flags;
+	uint status, state;
+	long t;
+
+	pr_info("%s: resetting the board...\n", fp->name);
+
+	spin_lock_irqsave(&fp->lock, flags);
+	fp->state_chg_flag = 0;
+	fza_do_reset(fp);
+	spin_unlock_irqrestore(&fp->lock, flags);
+
+	/* DEC says RESET needs up to 30 seconds to complete.  My DEFZA-AA
+	 * rev. C03 happily finishes in 9.7 seconds. :-)  But we need to
+	 * be on the safe side...
+	 */
+	t = wait_event_timeout(fp->state_chg_wait, fp->state_chg_flag,
+			       45 * HZ);
+	status = readw_u(&fp->regs->status);
+	state = FZA_STATUS_GET_STATE(status);
+	if (fp->state_chg_flag == 0) {
+		pr_err("%s: RESET timed out!, state %x\n", fp->name, state);
+		return -EIO;
+	}
+	if (state != FZA_STATE_UNINITIALIZED) {
+		pr_err("%s: RESET failed!, state %x, failure ID %x\n",
+		       fp->name, state, FZA_STATUS_GET_TEST(status));
+		return -EIO;
+	}
+	pr_info("%s: OK\n", fp->name);
+	pr_debug("%s: RESET: %lums elapsed\n", fp->name,
+		 (45 * HZ - t) * 1000 / HZ);
+
+	return 0;
+}
+
+static struct fza_ring_cmd __iomem *fza_cmd_send(struct net_device *dev,
+						 int command)
+{
+	struct fza_private *fp = netdev_priv(dev);
+	struct fza_ring_cmd __iomem *ring = fp->ring_cmd + fp->ring_cmd_index;
+	unsigned int old_mask, new_mask;
+	union fza_cmd_buf __iomem *buf;
+	struct netdev_hw_addr *ha;
+	int i;
+
+	old_mask = fp->int_mask;
+	new_mask = old_mask & ~FZA_MASK_STATE_CHG;
+	writew_u(new_mask, &fp->regs->int_mask);
+	readw_o(&fp->regs->int_mask);			/* Synchronize. */
+	fp->int_mask = new_mask;
+
+	buf = fp->mmio + readl_u(&ring->buffer);
+
+	if ((readl_u(&ring->cmd_own) & FZA_RING_OWN_MASK) !=
+	    FZA_RING_OWN_HOST) {
+		pr_warn("%s: command buffer full, command: %u!\n", fp->name,
+			command);
+		return NULL;
+	}
+
+	switch (command) {
+	case FZA_RING_CMD_INIT:
+		writel_u(FZA_RING_TX_MODE, &buf->init.tx_mode);
+		writel_u(FZA_RING_RX_SIZE, &buf->init.hst_rx_size);
+		fza_zeros(&buf->init.counters, sizeof(buf->init.counters));
+		break;
+
+	case FZA_RING_CMD_MODCAM:
+		i = 0;
+		fza_writes(&hw_addr_purger, &buf->cam.hw_addr[i++],
+			   sizeof(*buf->cam.hw_addr));
+		fza_writes(&hw_addr_beacon, &buf->cam.hw_addr[i++],
+			   sizeof(*buf->cam.hw_addr));
+		netdev_for_each_mc_addr(ha, dev) {
+			if (i >= FZA_CMD_CAM_SIZE)
+				break;
+			fza_writes(ha->addr, &buf->cam.hw_addr[i++],
+				   sizeof(*buf->cam.hw_addr));
+		}
+		while (i < FZA_CMD_CAM_SIZE)
+			fza_zeros(&buf->cam.hw_addr[i++],
+				  sizeof(*buf->cam.hw_addr));
+		break;
+
+	case FZA_RING_CMD_PARAM:
+		writel_u(loopback, &buf->param.loop_mode);
+		writel_u(fp->t_max, &buf->param.t_max);
+		writel_u(fp->t_req, &buf->param.t_req);
+		writel_u(fp->tvx, &buf->param.tvx);
+		writel_u(fp->lem_threshold, &buf->param.lem_threshold);
+		fza_writes(&fp->station_id, &buf->param.station_id,
+			   sizeof(buf->param.station_id));
+		/* Convert to milliseconds due to buggy firmware. */
+		writel_u(fp->rtoken_timeout / 12500,
+			 &buf->param.rtoken_timeout);
+		writel_u(fp->ring_purger, &buf->param.ring_purger);
+		break;
+
+	case FZA_RING_CMD_MODPROM:
+		if (dev->flags & IFF_PROMISC) {
+			writel_u(1, &buf->modprom.llc_prom);
+			writel_u(1, &buf->modprom.smt_prom);
+		} else {
+			writel_u(0, &buf->modprom.llc_prom);
+			writel_u(0, &buf->modprom.smt_prom);
+		}
+		if (dev->flags & IFF_ALLMULTI ||
+		    netdev_mc_count(dev) > FZA_CMD_CAM_SIZE - 2)
+			writel_u(1, &buf->modprom.llc_multi);
+		else
+			writel_u(0, &buf->modprom.llc_multi);
+		writel_u(1, &buf->modprom.llc_bcast);
+		break;
+	}
+
+	/* Trigger the command. */
+	writel_u(FZA_RING_OWN_FZA | command, &ring->cmd_own);
+	writew_o(FZA_CONTROL_A_CMD_POLL, &fp->regs->control_a);
+
+	fp->ring_cmd_index = (fp->ring_cmd_index + 1) % FZA_RING_CMD_SIZE;
+
+	fp->int_mask = old_mask;
+	writew_u(fp->int_mask, &fp->regs->int_mask);
+
+	return ring;
+}
+
+static int fza_init_send(struct net_device *dev,
+			 struct fza_cmd_init *__iomem *init)
+{
+	struct fza_private *fp = netdev_priv(dev);
+	struct fza_ring_cmd __iomem *ring;
+	unsigned long flags;
+	u32 stat;
+	long t;
+
+	spin_lock_irqsave(&fp->lock, flags);
+	fp->cmd_done_flag = 0;
+	ring = fza_cmd_send(dev, FZA_RING_CMD_INIT);
+	spin_unlock_irqrestore(&fp->lock, flags);
+	if (!ring)
+		/* This should never happen in the uninitialized state,
+		 * so do not try to recover and just consider it fatal.
+		 */
+		return -ENOBUFS;
+
+	/* INIT may take quite a long time (160ms for my C03). */
+	t = wait_event_timeout(fp->cmd_done_wait, fp->cmd_done_flag, 3 * HZ);
+	if (fp->cmd_done_flag == 0) {
+		pr_err("%s: INIT command timed out!, state %x\n", fp->name,
+		       FZA_STATUS_GET_STATE(readw_u(&fp->regs->status)));
+		return -EIO;
+	}
+	stat = readl_u(&ring->stat);
+	if (stat != FZA_RING_STAT_SUCCESS) {
+		pr_err("%s: INIT command failed!, status %02x, state %x\n",
+		       fp->name, stat,
+		       FZA_STATUS_GET_STATE(readw_u(&fp->regs->status)));
+		return -EIO;
+	}
+	pr_debug("%s: INIT: %lums elapsed\n", fp->name,
+		 (3 * HZ - t) * 1000 / HZ);
+
+	if (init)
+		*init = fp->mmio + readl_u(&ring->buffer);
+	return 0;
+}
+
+static void fza_rx_init(struct fza_private *fp)
+{
+	int i;
+
+	/* Fill the host receive descriptor ring. */
+	for (i = 0; i < FZA_RING_RX_SIZE; i++) {
+		writel_o(0, &fp->ring_hst_rx[i].rmc);
+		writel_o((fp->rx_dma[i] + 0x1000) >> 9,
+			 &fp->ring_hst_rx[i].buffer1);
+		writel_o(fp->rx_dma[i] >> 9 | FZA_RING_OWN_FZA,
+			 &fp->ring_hst_rx[i].buf0_own);
+	}
+}
+
+static void fza_set_rx_mode(struct net_device *dev)
+{
+	fza_cmd_send(dev, FZA_RING_CMD_MODCAM);
+	fza_cmd_send(dev, FZA_RING_CMD_MODPROM);
+}
+
+union fza_buffer_txp {
+	struct fza_buffer_tx *data_ptr;
+	struct fza_buffer_tx __iomem *mmio_ptr;
+};
+
+static int fza_do_xmit(union fza_buffer_txp ub, int len,
+		       struct net_device *dev, int smt)
+{
+	struct fza_private *fp = netdev_priv(dev);
+	struct fza_buffer_tx __iomem *rmc_tx_ptr;
+	int i, first, frag_len, left_len;
+	u32 own, rmc;
+
+	if (((((fp->ring_rmc_txd_index - 1 + fp->ring_rmc_tx_size) -
+	       fp->ring_rmc_tx_index) % fp->ring_rmc_tx_size) *
+	     FZA_TX_BUFFER_SIZE) < len)
+		return 1;
+
+	first = fp->ring_rmc_tx_index;
+
+	left_len = len;
+	frag_len = FZA_TX_BUFFER_SIZE;
+	/* First descriptor is relinquished last. */
+	own = FZA_RING_TX_OWN_HOST;
+	/* First descriptor carries frame length; we don't use cut-through. */
+	rmc = FZA_RING_TX_SOP | FZA_RING_TX_VBC | len;
+	do {
+		i = fp->ring_rmc_tx_index;
+		rmc_tx_ptr = &fp->buffer_tx[i];
+
+		if (left_len < FZA_TX_BUFFER_SIZE)
+			frag_len = left_len;
+		left_len -= frag_len;
+
+		/* Length must be a multiple of 4 as only word writes are
+		 * permitted!
+		 */
+		frag_len = (frag_len + 3) & ~3;
+		if (smt)
+			fza_moves(ub.mmio_ptr, rmc_tx_ptr, frag_len);
+		else
+			fza_writes(ub.data_ptr, rmc_tx_ptr, frag_len);
+
+		if (left_len == 0)
+			rmc |= FZA_RING_TX_EOP;		/* Mark last frag. */
+
+		writel_o(rmc, &fp->ring_rmc_tx[i].rmc);
+		writel_o(own, &fp->ring_rmc_tx[i].own);
+
+		ub.data_ptr++;
+		fp->ring_rmc_tx_index = (fp->ring_rmc_tx_index + 1) %
+					fp->ring_rmc_tx_size;
+
+		/* Settings for intermediate frags. */
+		own = FZA_RING_TX_OWN_RMC;
+		rmc = 0;
+	} while (left_len > 0);
+
+	if (((((fp->ring_rmc_txd_index - 1 + fp->ring_rmc_tx_size) -
+	       fp->ring_rmc_tx_index) % fp->ring_rmc_tx_size) *
+	     FZA_TX_BUFFER_SIZE) < dev->mtu + dev->hard_header_len) {
+		netif_stop_queue(dev);
+		pr_debug("%s: queue stopped\n", fp->name);
+	}
+
+	writel_o(FZA_RING_TX_OWN_RMC, &fp->ring_rmc_tx[first].own);
+
+	/* Go, go, go! */
+	writew_o(FZA_CONTROL_A_TX_POLL, &fp->regs->control_a);
+
+	return 0;
+}
+
+static int fza_do_recv_smt(struct fza_buffer_tx *data_ptr, int len,
+			   u32 rmc, struct net_device *dev)
+{
+	struct fza_private *fp = netdev_priv(dev);
+	struct fza_buffer_tx __iomem *smt_rx_ptr;
+	u32 own;
+	int i;
+
+	i = fp->ring_smt_rx_index;
+	own = readl_o(&fp->ring_smt_rx[i].own);
+	if ((own & FZA_RING_OWN_MASK) == FZA_RING_OWN_FZA)
+		return 1;
+
+	smt_rx_ptr = fp->mmio + readl_u(&fp->ring_smt_rx[i].buffer);
+
+	/* Length must be a multiple of 4 as only word writes are permitted! */
+	fza_writes(data_ptr, smt_rx_ptr, (len + 3) & ~3);
+
+	writel_o(rmc, &fp->ring_smt_rx[i].rmc);
+	writel_o(FZA_RING_OWN_FZA, &fp->ring_smt_rx[i].own);
+
+	fp->ring_smt_rx_index =
+		(fp->ring_smt_rx_index + 1) % fp->ring_smt_rx_size;
+
+	/* Grab it! */
+	writew_o(FZA_CONTROL_A_SMT_RX_POLL, &fp->regs->control_a);
+
+	return 0;
+}
+
+static void fza_tx(struct net_device *dev)
+{
+	struct fza_private *fp = netdev_priv(dev);
+	u32 own, rmc;
+	int i;
+
+	while (1) {
+		i = fp->ring_rmc_txd_index;
+		if (i == fp->ring_rmc_tx_index)
+			break;
+		own = readl_o(&fp->ring_rmc_tx[i].own);
+		if ((own & FZA_RING_OWN_MASK) == FZA_RING_TX_OWN_RMC)
+			break;
+
+		rmc = readl_u(&fp->ring_rmc_tx[i].rmc);
+		/* Only process the first descriptor. */
+		if ((rmc & FZA_RING_TX_SOP) != 0) {
+			if ((rmc & FZA_RING_TX_DCC_MASK) ==
+			    FZA_RING_TX_DCC_SUCCESS) {
+				int pkt_len = (rmc & FZA_RING_PBC_MASK) - 3;
+								/* Omit PRH. */
+
+				fp->stats.tx_packets++;
+				fp->stats.tx_bytes += pkt_len;
+			} else {
+				fp->stats.tx_errors++;
+				switch (rmc & FZA_RING_TX_DCC_MASK) {
+				case FZA_RING_TX_DCC_DTP_SOP:
+				case FZA_RING_TX_DCC_DTP:
+				case FZA_RING_TX_DCC_ABORT:
+					fp->stats.tx_aborted_errors++;
+					break;
+				case FZA_RING_TX_DCC_UNDRRUN:
+					fp->stats.tx_fifo_errors++;
+					break;
+				case FZA_RING_TX_DCC_PARITY:
+				default:
+					break;
+				}
+			}
+		}
+
+		fp->ring_rmc_txd_index = (fp->ring_rmc_txd_index + 1) %
+					 fp->ring_rmc_tx_size;
+	}
+
+	if (((((fp->ring_rmc_txd_index - 1 + fp->ring_rmc_tx_size) -
+	       fp->ring_rmc_tx_index) % fp->ring_rmc_tx_size) *
+	     FZA_TX_BUFFER_SIZE) >= dev->mtu + dev->hard_header_len) {
+		if (fp->queue_active) {
+			netif_wake_queue(dev);
+			pr_debug("%s: queue woken\n", fp->name);
+		}
+	}
+}
+
+static inline int fza_rx_err(struct fza_private *fp,
+			     const u32 rmc, const u8 fc)
+{
+	int len, min_len, max_len;
+
+	len = rmc & FZA_RING_PBC_MASK;
+
+	if (unlikely((rmc & FZA_RING_RX_BAD) != 0)) {
+		fp->stats.rx_errors++;
+
+		/* Check special status codes. */
+		if ((rmc & (FZA_RING_RX_CRC | FZA_RING_RX_RRR_MASK |
+			    FZA_RING_RX_DA_MASK | FZA_RING_RX_SA_MASK)) ==
+		     (FZA_RING_RX_CRC | FZA_RING_RX_RRR_DADDR |
+		      FZA_RING_RX_DA_CAM | FZA_RING_RX_SA_ALIAS)) {
+			if (len >= 8190)
+				fp->stats.rx_length_errors++;
+			return 1;
+		}
+		if ((rmc & (FZA_RING_RX_CRC | FZA_RING_RX_RRR_MASK |
+			    FZA_RING_RX_DA_MASK | FZA_RING_RX_SA_MASK)) ==
+		     (FZA_RING_RX_CRC | FZA_RING_RX_RRR_DADDR |
+		      FZA_RING_RX_DA_CAM | FZA_RING_RX_SA_CAM)) {
+			/* Halt the interface to trigger a reset. */
+			writew_o(FZA_CONTROL_A_HALT, &fp->regs->control_a);
+			readw_o(&fp->regs->control_a);	/* Synchronize. */
+			return 1;
+		}
+
+		/* Check the MAC status. */
+		switch (rmc & FZA_RING_RX_RRR_MASK) {
+		case FZA_RING_RX_RRR_OK:
+			if ((rmc & FZA_RING_RX_CRC) != 0)
+				fp->stats.rx_crc_errors++;
+			else if ((rmc & FZA_RING_RX_FSC_MASK) == 0 ||
+				 (rmc & FZA_RING_RX_FSB_ERR) != 0)
+				fp->stats.rx_frame_errors++;
+			return 1;
+		case FZA_RING_RX_RRR_SADDR:
+		case FZA_RING_RX_RRR_DADDR:
+		case FZA_RING_RX_RRR_ABORT:
+			/* Halt the interface to trigger a reset. */
+			writew_o(FZA_CONTROL_A_HALT, &fp->regs->control_a);
+			readw_o(&fp->regs->control_a);	/* Synchronize. */
+			return 1;
+		case FZA_RING_RX_RRR_LENGTH:
+			fp->stats.rx_frame_errors++;
+			return 1;
+		default:
+			return 1;
+		}
+	}
+
+	/* Packet received successfully; validate the length. */
+	switch (fc & FDDI_FC_K_FORMAT_MASK) {
+	case FDDI_FC_K_FORMAT_MANAGEMENT:
+		if ((fc & FDDI_FC_K_CLASS_MASK) == FDDI_FC_K_CLASS_ASYNC)
+			min_len = 37;
+		else
+			min_len = 17;
+		break;
+	case FDDI_FC_K_FORMAT_LLC:
+		min_len = 20;
+		break;
+	default:
+		min_len = 17;
+		break;
+	}
+	max_len = 4495;
+	if (len < min_len || len > max_len) {
+		fp->stats.rx_errors++;
+		fp->stats.rx_length_errors++;
+		return 1;
+	}
+
+	return 0;
+}
+
+static void fza_rx(struct net_device *dev)
+{
+	struct fza_private *fp = netdev_priv(dev);
+	struct sk_buff *skb, *newskb;
+	struct fza_fddihdr *frame;
+	dma_addr_t dma, newdma;
+	u32 own, rmc, buf;
+	int i, len;
+	u8 fc;
+
+	while (1) {
+		i = fp->ring_hst_rx_index;
+		own = readl_o(&fp->ring_hst_rx[i].buf0_own);
+		if ((own & FZA_RING_OWN_MASK) == FZA_RING_OWN_FZA)
+			break;
+
+		rmc = readl_u(&fp->ring_hst_rx[i].rmc);
+		skb = fp->rx_skbuff[i];
+		dma = fp->rx_dma[i];
+
+		/* The RMC doesn't count the preamble and the starting
+		 * delimiter.  We fix it up here for a total of 3 octets.
+		 */
+		dma_rmb();
+		len = (rmc & FZA_RING_PBC_MASK) + 3;
+		frame = (struct fza_fddihdr *)skb->data;
+
+		/* We need to get at real FC. */
+		dma_sync_single_for_cpu(fp->bdev,
+					dma +
+					((u8 *)&frame->hdr.fc - (u8 *)frame),
+					sizeof(frame->hdr.fc),
+					DMA_FROM_DEVICE);
+		fc = frame->hdr.fc;
+
+		if (fza_rx_err(fp, rmc, fc))
+			goto err_rx;
+
+		/* We have to 512-byte-align RX buffers... */
+		newskb = fza_alloc_skb_irq(dev, FZA_RX_BUFFER_SIZE + 511);
+		if (newskb) {
+			fza_skb_align(newskb, 512);
+			newdma = dma_map_single(fp->bdev, newskb->data,
+						FZA_RX_BUFFER_SIZE,
+						DMA_FROM_DEVICE);
+			if (dma_mapping_error(fp->bdev, newdma)) {
+				dev_kfree_skb_irq(newskb);
+				newskb = NULL;
+			}
+		}
+		if (newskb) {
+			int pkt_len = len - 7;	/* Omit P, SD and FCS. */
+			int is_multi;
+			int rx_stat;
+
+			dma_unmap_single(fp->bdev, dma, FZA_RX_BUFFER_SIZE,
+					 DMA_FROM_DEVICE);
+
+			/* Queue SMT frames to the SMT receive ring. */
+			if ((fc & (FDDI_FC_K_CLASS_MASK |
+				   FDDI_FC_K_FORMAT_MASK)) ==
+			     (FDDI_FC_K_CLASS_ASYNC |
+			      FDDI_FC_K_FORMAT_MANAGEMENT) &&
+			    (rmc & FZA_RING_RX_DA_MASK) !=
+			     FZA_RING_RX_DA_PROM) {
+				if (fza_do_recv_smt((struct fza_buffer_tx *)
+						    skb->data, len, rmc,
+						    dev)) {
+					writel_o(FZA_CONTROL_A_SMT_RX_OVFL,
+						 &fp->regs->control_a);
+				}
+			}
+
+			is_multi = ((frame->hdr.daddr[0] & 0x01) != 0);
+
+			skb_reserve(skb, 3);	/* Skip over P and SD. */
+			skb_put(skb, pkt_len);	/* And cut off FCS. */
+			skb->protocol = fddi_type_trans(skb, dev);
+
+			rx_stat = netif_rx(skb);
+			if (rx_stat != NET_RX_DROP) {
+				fp->stats.rx_packets++;
+				fp->stats.rx_bytes += pkt_len;
+				if (is_multi)
+					fp->stats.multicast++;
+			} else {
+				fp->stats.rx_dropped++;
+			}
+
+			skb = newskb;
+			dma = newdma;
+			fp->rx_skbuff[i] = skb;
+			fp->rx_dma[i] = dma;
+		} else {
+			fp->stats.rx_dropped++;
+			pr_notice_ratelimited(
+				"%s: memory squeeze, dropping packet\n",
+				fp->name);
+		}
+
+err_rx:
+		writel_o(0, &fp->ring_hst_rx[i].rmc);
+		buf = (dma + 0x1000) >> 9;
+		writel_o(buf, &fp->ring_hst_rx[i].buffer1);
+		buf = dma >> 9 | FZA_RING_OWN_FZA;
+		writel_o(buf, &fp->ring_hst_rx[i].buf0_own);
+		fp->ring_hst_rx_index =
+			(fp->ring_hst_rx_index + 1) % fp->ring_hst_rx_size;
+	}
+}
+
+static void fza_tx_smt(struct net_device *dev)
+{
+	struct fza_private *fp = netdev_priv(dev);
+	struct fza_buffer_tx __iomem *smt_tx_ptr;
+	int i, len;
+	u32 own;
+
+	while (1) {
+		i = fp->ring_smt_tx_index;
+		own = readl_o(&fp->ring_smt_tx[i].own);
+		if ((own & FZA_RING_OWN_MASK) == FZA_RING_OWN_FZA)
+			break;
+
+		smt_tx_ptr = fp->mmio + readl_u(&fp->ring_smt_tx[i].buffer);
+		len = readl_u(&fp->ring_smt_tx[i].rmc) & FZA_RING_PBC_MASK;
+
+		if (!netif_queue_stopped(dev)) {
+			if (dev_nit_active(dev)) {
+				struct fza_buffer_tx *skb_data_ptr;
+				struct sk_buff *skb;
+
+				/* Length must be a multiple of 4 as only word
+				 * reads are permitted!
+				 */
+				skb = fza_alloc_skb_irq(dev, (len + 3) & ~3);
+				if (!skb)
+					goto err_no_skb;	/* Drop. */
+
+				skb_data_ptr = (struct fza_buffer_tx *)
+					       skb->data;
+
+				fza_reads(smt_tx_ptr, skb_data_ptr,
+					  (len + 3) & ~3);
+				skb->dev = dev;
+				skb_reserve(skb, 3);	/* Skip over PRH. */
+				skb_put(skb, len - 3);
+				skb_reset_network_header(skb);
+
+				dev_queue_xmit_nit(skb, dev);
+
+				dev_kfree_skb_irq(skb);
+
+err_no_skb:
+				;
+			}
+
+			/* Queue the frame to the RMC transmit ring. */
+			fza_do_xmit((union fza_buffer_txp)
+				    { .mmio_ptr = smt_tx_ptr },
+				    len, dev, 1);
+		}
+
+		writel_o(FZA_RING_OWN_FZA, &fp->ring_smt_tx[i].own);
+		fp->ring_smt_tx_index =
+			(fp->ring_smt_tx_index + 1) % fp->ring_smt_tx_size;
+	}
+}
+
+static void fza_uns(struct net_device *dev)
+{
+	struct fza_private *fp = netdev_priv(dev);
+	u32 own;
+	int i;
+
+	while (1) {
+		i = fp->ring_uns_index;
+		own = readl_o(&fp->ring_uns[i].own);
+		if ((own & FZA_RING_OWN_MASK) == FZA_RING_OWN_FZA)
+			break;
+
+		if (readl_u(&fp->ring_uns[i].id) == FZA_RING_UNS_RX_OVER) {
+			fp->stats.rx_errors++;
+			fp->stats.rx_over_errors++;
+		}
+
+		writel_o(FZA_RING_OWN_FZA, &fp->ring_uns[i].own);
+		fp->ring_uns_index =
+			(fp->ring_uns_index + 1) % FZA_RING_UNS_SIZE;
+	}
+}
+
+static void fza_tx_flush(struct net_device *dev)
+{
+	struct fza_private *fp = netdev_priv(dev);
+	u32 own;
+	int i;
+
+	/* Clean up the SMT TX ring. */
+	i = fp->ring_smt_tx_index;
+	do {
+		writel_o(FZA_RING_OWN_FZA, &fp->ring_smt_tx[i].own);
+		fp->ring_smt_tx_index =
+			(fp->ring_smt_tx_index + 1) % fp->ring_smt_tx_size;
+
+	} while (i != fp->ring_smt_tx_index);
+
+	/* Clean up the RMC TX ring. */
+	i = fp->ring_rmc_tx_index;
+	do {
+		own = readl_o(&fp->ring_rmc_tx[i].own);
+		if ((own & FZA_RING_OWN_MASK) == FZA_RING_TX_OWN_RMC) {
+			u32 rmc = readl_u(&fp->ring_rmc_tx[i].rmc);
+
+			writel_u(rmc | FZA_RING_TX_DTP,
+				 &fp->ring_rmc_tx[i].rmc);
+		}
+		fp->ring_rmc_tx_index =
+			(fp->ring_rmc_tx_index + 1) % fp->ring_rmc_tx_size;
+
+	} while (i != fp->ring_rmc_tx_index);
+
+	/* Done. */
+	writew_o(FZA_CONTROL_A_FLUSH_DONE, &fp->regs->control_a);
+}
+
+static irqreturn_t fza_interrupt(int irq, void *dev_id)
+{
+	struct net_device *dev = dev_id;
+	struct fza_private *fp = netdev_priv(dev);
+	uint int_event;
+
+	/* Get interrupt events. */
+	int_event = readw_o(&fp->regs->int_event) & fp->int_mask;
+	if (int_event == 0)
+		return IRQ_NONE;
+
+	/* Clear the events. */
+	writew_u(int_event, &fp->regs->int_event);
+
+	/* Now handle the events.  The order matters. */
+
+	/* Command finished interrupt. */
+	if ((int_event & FZA_EVENT_CMD_DONE) != 0) {
+		fp->irq_count_cmd_done++;
+
+		spin_lock(&fp->lock);
+		fp->cmd_done_flag = 1;
+		wake_up(&fp->cmd_done_wait);
+		spin_unlock(&fp->lock);
+	}
+
+	/* Transmit finished interrupt. */
+	if ((int_event & FZA_EVENT_TX_DONE) != 0) {
+		fp->irq_count_tx_done++;
+		fza_tx(dev);
+	}
+
+	/* Host receive interrupt. */
+	if ((int_event & FZA_EVENT_RX_POLL) != 0) {
+		fp->irq_count_rx_poll++;
+		fza_rx(dev);
+	}
+
+	/* SMT transmit interrupt. */
+	if ((int_event & FZA_EVENT_SMT_TX_POLL) != 0) {
+		fp->irq_count_smt_tx_poll++;
+		fza_tx_smt(dev);
+	}
+
+	/* Transmit ring flush request. */
+	if ((int_event & FZA_EVENT_FLUSH_TX) != 0) {
+		fp->irq_count_flush_tx++;
+		fza_tx_flush(dev);
+	}
+
+	/* Link status change interrupt. */
+	if ((int_event & FZA_EVENT_LINK_ST_CHG) != 0) {
+		uint status;
+
+		fp->irq_count_link_st_chg++;
+		status = readw_u(&fp->regs->status);
+		if (FZA_STATUS_GET_LINK(status) == FZA_LINK_ON) {
+			netif_carrier_on(dev);
+			pr_info("%s: link available\n", fp->name);
+		} else {
+			netif_carrier_off(dev);
+			pr_info("%s: link unavailable\n", fp->name);
+		}
+	}
+
+	/* Unsolicited event interrupt. */
+	if ((int_event & FZA_EVENT_UNS_POLL) != 0) {
+		fp->irq_count_uns_poll++;
+		fza_uns(dev);
+	}
+
+	/* State change interrupt. */
+	if ((int_event & FZA_EVENT_STATE_CHG) != 0) {
+		uint status, state;
+
+		fp->irq_count_state_chg++;
+
+		status = readw_u(&fp->regs->status);
+		state = FZA_STATUS_GET_STATE(status);
+		pr_debug("%s: state change: %x\n", fp->name, state);
+		switch (state) {
+		case FZA_STATE_RESET:
+			break;
+
+		case FZA_STATE_UNINITIALIZED:
+			netif_carrier_off(dev);
+			timer_delete_sync_try(&fp->reset_timer);
+			fp->ring_cmd_index = 0;
+			fp->ring_uns_index = 0;
+			fp->ring_rmc_tx_index = 0;
+			fp->ring_rmc_txd_index = 0;
+			fp->ring_hst_rx_index = 0;
+			fp->ring_smt_tx_index = 0;
+			fp->ring_smt_rx_index = 0;
+			if (fp->state > state) {
+				pr_info("%s: OK\n", fp->name);
+				fza_cmd_send(dev, FZA_RING_CMD_INIT);
+			}
+			break;
+
+		case FZA_STATE_INITIALIZED:
+			if (fp->state > state) {
+				fza_set_rx_mode(dev);
+				fza_cmd_send(dev, FZA_RING_CMD_PARAM);
+			}
+			break;
+
+		case FZA_STATE_RUNNING:
+		case FZA_STATE_MAINTENANCE:
+			fp->state = state;
+			fza_rx_init(fp);
+			fp->queue_active = 1;
+			netif_wake_queue(dev);
+			pr_debug("%s: queue woken\n", fp->name);
+			break;
+
+		case FZA_STATE_HALTED:
+			fp->queue_active = 0;
+			netif_stop_queue(dev);
+			pr_debug("%s: queue stopped\n", fp->name);
+
+			spin_lock(&fp->lock);
+			timer_delete(&fp->reset_timer);
+			pr_warn("%s: halted, reason: %x\n", fp->name,
+				FZA_STATUS_GET_HALT(status));
+			fza_regs_dump(fp);
+			pr_info("%s: resetting the board...\n", fp->name);
+			fza_do_reset(fp);
+			fp->timer_state = 0;
+			fp->reset_timer.expires = jiffies + 45 * HZ;
+			add_timer(&fp->reset_timer);
+			spin_unlock(&fp->lock);
+
+			break;
+
+		default:
+			pr_warn("%s: undefined state: %x\n", fp->name, state);
+			break;
+		}
+
+		spin_lock(&fp->lock);
+		fp->state_chg_flag = 1;
+		wake_up(&fp->state_chg_wait);
+		spin_unlock(&fp->lock);
+	}
+
+	return IRQ_HANDLED;
+}
+
+static void fza_reset_timer(struct timer_list *t)
+{
+	struct fza_private *fp = timer_container_of(fp, t, reset_timer);
+	unsigned long flags;
+
+	spin_lock_irqsave(&fp->lock, flags);
+	if (!fp->timer_state) {
+		pr_err("%s: RESET timed out!\n", fp->name);
+		pr_info("%s: trying harder...\n", fp->name);
+
+		/* Assert the board reset. */
+		writew_o(FZA_RESET_INIT, &fp->regs->reset);
+		readw_o(&fp->regs->reset);		/* Synchronize. */
+
+		fp->timer_state = 1;
+		fp->reset_timer.expires = jiffies + HZ;
+	} else {
+		/* Clear the board reset. */
+		writew_u(FZA_RESET_CLR, &fp->regs->reset);
+
+		/* Enable all interrupt events we handle. */
+		writew_o(fp->int_mask, &fp->regs->int_mask);
+		readw_o(&fp->regs->int_mask);		/* Synchronize. */
+
+		fp->timer_state = 0;
+		fp->reset_timer.expires = jiffies + 45 * HZ;
+	}
+	add_timer(&fp->reset_timer);
+	spin_unlock_irqrestore(&fp->lock, flags);
+}
+
+static int fza_set_mac_address(struct net_device *dev, void *addr)
+{
+	return -EOPNOTSUPP;
+}
+
+static netdev_tx_t fza_start_xmit(struct sk_buff *skb, struct net_device *dev)
+{
+	struct fza_private *fp = netdev_priv(dev);
+	unsigned int old_mask, new_mask;
+	int ret;
+	u8 fc;
+
+	skb_push(skb, 3);			/* Make room for PRH. */
+
+	/* Decode FC to set PRH. */
+	fc = skb->data[3];
+	skb->data[0] = 0;
+	skb->data[1] = 0;
+	skb->data[2] = FZA_PRH2_NORMAL;
+	if ((fc & FDDI_FC_K_CLASS_MASK) == FDDI_FC_K_CLASS_SYNC)
+		skb->data[0] |= FZA_PRH0_FRAME_SYNC;
+	switch (fc & FDDI_FC_K_FORMAT_MASK) {
+	case FDDI_FC_K_FORMAT_MANAGEMENT:
+		if ((fc & FDDI_FC_K_CONTROL_MASK) == 0) {
+			/* Token. */
+			skb->data[0] |= FZA_PRH0_TKN_TYPE_IMM;
+			skb->data[1] |= FZA_PRH1_TKN_SEND_NONE;
+		} else {
+			/* SMT or MAC. */
+			skb->data[0] |= FZA_PRH0_TKN_TYPE_UNR;
+			skb->data[1] |= FZA_PRH1_TKN_SEND_UNR;
+		}
+		skb->data[1] |= FZA_PRH1_CRC_NORMAL;
+		break;
+	case FDDI_FC_K_FORMAT_LLC:
+	case FDDI_FC_K_FORMAT_FUTURE:
+		skb->data[0] |= FZA_PRH0_TKN_TYPE_UNR;
+		skb->data[1] |= FZA_PRH1_CRC_NORMAL | FZA_PRH1_TKN_SEND_UNR;
+		break;
+	case FDDI_FC_K_FORMAT_IMPLEMENTOR:
+		skb->data[0] |= FZA_PRH0_TKN_TYPE_UNR;
+		skb->data[1] |= FZA_PRH1_TKN_SEND_ORIG;
+		break;
+	}
+
+	/* SMT transmit interrupts may sneak frames into the RMC
+	 * transmit ring.  We disable them while queueing a frame
+	 * to maintain consistency.
+	 */
+	old_mask = fp->int_mask;
+	new_mask = old_mask & ~FZA_MASK_SMT_TX_POLL;
+	writew_u(new_mask, &fp->regs->int_mask);
+	readw_o(&fp->regs->int_mask);			/* Synchronize. */
+	fp->int_mask = new_mask;
+	ret = fza_do_xmit((union fza_buffer_txp)
+			  { .data_ptr = (struct fza_buffer_tx *)skb->data },
+			  skb->len, dev, 0);
+	fp->int_mask = old_mask;
+	writew_u(fp->int_mask, &fp->regs->int_mask);
+
+	if (ret) {
+		/* Probably an SMT packet filled the remaining space,
+		 * so just stop the queue, but don't report it as an error.
+		 */
+		netif_stop_queue(dev);
+		pr_debug("%s: queue stopped\n", fp->name);
+		fp->stats.tx_dropped++;
+	}
+
+	dev_kfree_skb(skb);
+
+	return ret;
+}
+
+static int fza_open(struct net_device *dev)
+{
+	struct fza_private *fp = netdev_priv(dev);
+	struct fza_ring_cmd __iomem *ring;
+	struct sk_buff *skb;
+	unsigned long flags;
+	dma_addr_t dma;
+	int ret, i;
+	u32 stat;
+	long t;
+
+	for (i = 0; i < FZA_RING_RX_SIZE; i++) {
+		/* We have to 512-byte-align RX buffers... */
+		skb = fza_alloc_skb(dev, FZA_RX_BUFFER_SIZE + 511);
+		if (skb) {
+			fza_skb_align(skb, 512);
+			dma = dma_map_single(fp->bdev, skb->data,
+					     FZA_RX_BUFFER_SIZE,
+					     DMA_FROM_DEVICE);
+			if (dma_mapping_error(fp->bdev, dma)) {
+				dev_kfree_skb(skb);
+				skb = NULL;
+			}
+		}
+		if (!skb) {
+			for (--i; i >= 0; i--) {
+				dma_unmap_single(fp->bdev, fp->rx_dma[i],
+						 FZA_RX_BUFFER_SIZE,
+						 DMA_FROM_DEVICE);
+				dev_kfree_skb(fp->rx_skbuff[i]);
+				fp->rx_dma[i] = 0;
+				fp->rx_skbuff[i] = NULL;
+			}
+			return -ENOMEM;
+		}
+		fp->rx_skbuff[i] = skb;
+		fp->rx_dma[i] = dma;
+	}
+
+	ret = fza_init_send(dev, NULL);
+	if (ret != 0)
+		return ret;
+
+	/* Purger and Beacon multicasts need to be supplied before PARAM. */
+	fza_set_rx_mode(dev);
+
+	spin_lock_irqsave(&fp->lock, flags);
+	fp->cmd_done_flag = 0;
+	ring = fza_cmd_send(dev, FZA_RING_CMD_PARAM);
+	spin_unlock_irqrestore(&fp->lock, flags);
+	if (!ring)
+		return -ENOBUFS;
+
+	t = wait_event_timeout(fp->cmd_done_wait, fp->cmd_done_flag, 3 * HZ);
+	if (fp->cmd_done_flag == 0) {
+		pr_err("%s: PARAM command timed out!, state %x\n", fp->name,
+		       FZA_STATUS_GET_STATE(readw_u(&fp->regs->status)));
+		return -EIO;
+	}
+	stat = readl_u(&ring->stat);
+	if (stat != FZA_RING_STAT_SUCCESS) {
+		pr_err("%s: PARAM command failed!, status %02x, state %x\n",
+		       fp->name, stat,
+		       FZA_STATUS_GET_STATE(readw_u(&fp->regs->status)));
+		return -EIO;
+	}
+	pr_debug("%s: PARAM: %lums elapsed\n", fp->name,
+		 (3 * HZ - t) * 1000 / HZ);
+
+	return 0;
+}
+
+static int fza_close(struct net_device *dev)
+{
+	struct fza_private *fp = netdev_priv(dev);
+	unsigned long flags;
+	uint state;
+	long t;
+	int i;
+
+	netif_stop_queue(dev);
+	pr_debug("%s: queue stopped\n", fp->name);
+
+	timer_delete_sync(&fp->reset_timer);
+	spin_lock_irqsave(&fp->lock, flags);
+	fp->state = FZA_STATE_UNINITIALIZED;
+	fp->state_chg_flag = 0;
+	/* Shut the interface down. */
+	writew_o(FZA_CONTROL_A_SHUT, &fp->regs->control_a);
+	readw_o(&fp->regs->control_a);			/* Synchronize. */
+	spin_unlock_irqrestore(&fp->lock, flags);
+
+	/* DEC says SHUT needs up to 10 seconds to complete. */
+	t = wait_event_timeout(fp->state_chg_wait, fp->state_chg_flag,
+			       15 * HZ);
+	state = FZA_STATUS_GET_STATE(readw_o(&fp->regs->status));
+	if (fp->state_chg_flag == 0) {
+		pr_err("%s: SHUT timed out!, state %x\n", fp->name, state);
+		return -EIO;
+	}
+	if (state != FZA_STATE_UNINITIALIZED) {
+		pr_err("%s: SHUT failed!, state %x\n", fp->name, state);
+		return -EIO;
+	}
+	pr_debug("%s: SHUT: %lums elapsed\n", fp->name,
+		 (15 * HZ - t) * 1000 / HZ);
+
+	for (i = 0; i < FZA_RING_RX_SIZE; i++)
+		if (fp->rx_skbuff[i]) {
+			dma_unmap_single(fp->bdev, fp->rx_dma[i],
+					 FZA_RX_BUFFER_SIZE, DMA_FROM_DEVICE);
+			dev_kfree_skb(fp->rx_skbuff[i]);
+			fp->rx_dma[i] = 0;
+			fp->rx_skbuff[i] = NULL;
+		}
+
+	return 0;
+}
+
+static struct net_device_stats *fza_get_stats(struct net_device *dev)
+{
+	struct fza_private *fp = netdev_priv(dev);
+
+	return &fp->stats;
+}
+
+static int fza_probe(struct device *bdev)
+{
+	static const struct net_device_ops netdev_ops = {
+		.ndo_open = fza_open,
+		.ndo_stop = fza_close,
+		.ndo_start_xmit = fza_start_xmit,
+		.ndo_set_rx_mode = fza_set_rx_mode,
+		.ndo_set_mac_address = fza_set_mac_address,
+		.ndo_get_stats = fza_get_stats,
+	};
+	static int version_printed;
+	char rom_rev[4], fw_rev[4], rmc_rev[4];
+	struct tc_dev *tdev = to_tc_dev(bdev);
+	struct fza_cmd_init __iomem *init;
+	resource_size_t start, len;
+	struct net_device *dev;
+	struct fza_private *fp;
+	uint smt_ver, pmd_type;
+	void __iomem *mmio;
+	uint hw_addr[2];
+	int ret, i;
+
+	if (!version_printed) {
+		pr_info("%s", version);
+		version_printed = 1;
+	}
+
+	dev = alloc_fddidev(sizeof(*fp));
+	if (!dev)
+		return -ENOMEM;
+	SET_NETDEV_DEV(dev, bdev);
+
+	fp = netdev_priv(dev);
+	dev_set_drvdata(bdev, dev);
+
+	fp->bdev = bdev;
+	fp->name = dev_name(bdev);
+
+	/* Request the I/O MEM resource. */
+	start = tdev->resource.start;
+	len = tdev->resource.end - start + 1;
+	if (!request_mem_region(start, len, dev_name(bdev))) {
+		pr_err("%s: cannot reserve MMIO region\n", fp->name);
+		ret = -EBUSY;
+		goto err_out_kfree;
+	}
+
+	/* MMIO mapping setup. */
+	mmio = ioremap(start, len);
+	if (!mmio) {
+		pr_err("%s: cannot map MMIO\n", fp->name);
+		ret = -ENOMEM;
+		goto err_out_resource;
+	}
+
+	/* Initialize the new device structure. */
+	switch (loopback) {
+	case FZA_LOOP_NORMAL:
+	case FZA_LOOP_INTERN:
+	case FZA_LOOP_EXTERN:
+		break;
+	default:
+		loopback = FZA_LOOP_NORMAL;
+	}
+
+	fp->mmio = mmio;
+	dev->irq = tdev->interrupt;
+
+	pr_info("%s: DEC FDDIcontroller 700 or 700-C at 0x%08llx, irq %d\n",
+		fp->name, (long long)tdev->resource.start, dev->irq);
+	pr_debug("%s: mapped at: 0x%p\n", fp->name, mmio);
+
+	fp->regs = mmio + FZA_REG_BASE;
+	fp->ring_cmd = mmio + FZA_RING_CMD;
+	fp->ring_uns = mmio + FZA_RING_UNS;
+
+	init_waitqueue_head(&fp->state_chg_wait);
+	init_waitqueue_head(&fp->cmd_done_wait);
+	spin_lock_init(&fp->lock);
+	fp->int_mask = FZA_MASK_NORMAL;
+
+	timer_setup(&fp->reset_timer, fza_reset_timer, 0);
+
+	/* Sanitize the board. */
+	fza_regs_dump(fp);
+	fza_do_shutdown(fp);
+
+	ret = request_irq(dev->irq, fza_interrupt, IRQF_SHARED, fp->name, dev);
+	if (ret != 0) {
+		pr_err("%s: unable to get IRQ %d!\n", fp->name, dev->irq);
+		goto err_out_map;
+	}
+
+	/* Enable the driver mode. */
+	writew_o(FZA_CONTROL_B_DRIVER, &fp->regs->control_b);
+
+	/* For some reason transmit done interrupts can trigger during
+	 * reset.  This avoids a division error in the handler.
+	 */
+	fp->ring_rmc_tx_size = FZA_RING_TX_SIZE;
+
+	ret = fza_reset(fp);
+	if (ret != 0)
+		goto err_out_irq;
+
+	ret = fza_init_send(dev, &init);
+	if (ret != 0)
+		goto err_out_irq;
+
+	fza_reads(&init->hw_addr, &hw_addr, sizeof(hw_addr));
+	dev_addr_set(dev, (u8 *)&hw_addr);
+
+	fza_reads(&init->rom_rev, &rom_rev, sizeof(rom_rev));
+	fza_reads(&init->fw_rev, &fw_rev, sizeof(fw_rev));
+	fza_reads(&init->rmc_rev, &rmc_rev, sizeof(rmc_rev));
+	for (i = 3; i >= 0 && rom_rev[i] == ' '; i--)
+		rom_rev[i] = 0;
+	for (i = 3; i >= 0 && fw_rev[i] == ' '; i--)
+		fw_rev[i] = 0;
+	for (i = 3; i >= 0 && rmc_rev[i] == ' '; i--)
+		rmc_rev[i] = 0;
+
+	fp->ring_rmc_tx = mmio + readl_u(&init->rmc_tx);
+	fp->ring_rmc_tx_size = readl_u(&init->rmc_tx_size);
+	fp->ring_hst_rx = mmio + readl_u(&init->hst_rx);
+	fp->ring_hst_rx_size = readl_u(&init->hst_rx_size);
+	fp->ring_smt_tx = mmio + readl_u(&init->smt_tx);
+	fp->ring_smt_tx_size = readl_u(&init->smt_tx_size);
+	fp->ring_smt_rx = mmio + readl_u(&init->smt_rx);
+	fp->ring_smt_rx_size = readl_u(&init->smt_rx_size);
+
+	fp->buffer_tx = mmio + FZA_TX_BUFFER_ADDR(readl_u(&init->rmc_tx));
+
+	fp->t_max = readl_u(&init->def_t_max);
+	fp->t_req = readl_u(&init->def_t_req);
+	fp->tvx = readl_u(&init->def_tvx);
+	fp->lem_threshold = readl_u(&init->lem_threshold);
+	fza_reads(&init->def_station_id, &fp->station_id,
+		  sizeof(fp->station_id));
+	fp->rtoken_timeout = readl_u(&init->rtoken_timeout);
+	fp->ring_purger = readl_u(&init->ring_purger);
+
+	smt_ver = readl_u(&init->smt_ver);
+	pmd_type = readl_u(&init->pmd_type);
+
+	pr_debug("%s: INIT parameters:\n", fp->name);
+	pr_debug("        tx_mode: %u\n", readl_u(&init->tx_mode));
+	pr_debug("    hst_rx_size: %u\n", readl_u(&init->hst_rx_size));
+	pr_debug("        rmc_rev: %.4s\n", rmc_rev);
+	pr_debug("        rom_rev: %.4s\n", rom_rev);
+	pr_debug("         fw_rev: %.4s\n", fw_rev);
+	pr_debug("       mop_type: %u\n", readl_u(&init->mop_type));
+	pr_debug("         hst_rx: 0x%08x\n", readl_u(&init->hst_rx));
+	pr_debug("         rmc_tx: 0x%08x\n", readl_u(&init->rmc_tx));
+	pr_debug("    rmc_tx_size: %u\n", readl_u(&init->rmc_tx_size));
+	pr_debug("         smt_tx: 0x%08x\n", readl_u(&init->smt_tx));
+	pr_debug("    smt_tx_size: %u\n", readl_u(&init->smt_tx_size));
+	pr_debug("         smt_rx: 0x%08x\n", readl_u(&init->smt_rx));
+	pr_debug("    smt_rx_size: %u\n", readl_u(&init->smt_rx_size));
+	/* TC systems are always LE, so don't bother swapping. */
+	pr_debug("        hw_addr: 0x%02x%02x%02x%02x%02x%02x%02x%02x\n",
+		 (readl_u(&init->hw_addr[0]) >> 0) & 0xff,
+		 (readl_u(&init->hw_addr[0]) >> 8) & 0xff,
+		 (readl_u(&init->hw_addr[0]) >> 16) & 0xff,
+		 (readl_u(&init->hw_addr[0]) >> 24) & 0xff,
+		 (readl_u(&init->hw_addr[1]) >> 0) & 0xff,
+		 (readl_u(&init->hw_addr[1]) >> 8) & 0xff,
+		 (readl_u(&init->hw_addr[1]) >> 16) & 0xff,
+		 (readl_u(&init->hw_addr[1]) >> 24) & 0xff);
+	pr_debug("      def_t_req: %u\n", readl_u(&init->def_t_req));
+	pr_debug("        def_tvx: %u\n", readl_u(&init->def_tvx));
+	pr_debug("      def_t_max: %u\n", readl_u(&init->def_t_max));
+	pr_debug("  lem_threshold: %u\n", readl_u(&init->lem_threshold));
+	/* Don't bother swapping, see above. */
+	pr_debug(" def_station_id: 0x%02x%02x%02x%02x%02x%02x%02x%02x\n",
+		 (readl_u(&init->def_station_id[0]) >> 0) & 0xff,
+		 (readl_u(&init->def_station_id[0]) >> 8) & 0xff,
+		 (readl_u(&init->def_station_id[0]) >> 16) & 0xff,
+		 (readl_u(&init->def_station_id[0]) >> 24) & 0xff,
+		 (readl_u(&init->def_station_id[1]) >> 0) & 0xff,
+		 (readl_u(&init->def_station_id[1]) >> 8) & 0xff,
+		 (readl_u(&init->def_station_id[1]) >> 16) & 0xff,
+		 (readl_u(&init->def_station_id[1]) >> 24) & 0xff);
+	pr_debug("   pmd_type_alt: %u\n", readl_u(&init->pmd_type_alt));
+	pr_debug("        smt_ver: %u\n", readl_u(&init->smt_ver));
+	pr_debug(" rtoken_timeout: %u\n", readl_u(&init->rtoken_timeout));
+	pr_debug("    ring_purger: %u\n", readl_u(&init->ring_purger));
+	pr_debug("    smt_ver_max: %u\n", readl_u(&init->smt_ver_max));
+	pr_debug("    smt_ver_min: %u\n", readl_u(&init->smt_ver_min));
+	pr_debug("       pmd_type: %u\n", readl_u(&init->pmd_type));
+
+	pr_info("%s: model %s, address %pMF\n",
+		fp->name,
+		pmd_type == FZA_PMD_TYPE_TW ?
+			"700-C (DEFZA-CA), ThinWire PMD selected" :
+			pmd_type == FZA_PMD_TYPE_STP ?
+				"700-C (DEFZA-CA), STP PMD selected" :
+				"700 (DEFZA-AA), MMF PMD",
+		dev->dev_addr);
+	pr_info("%s: ROM rev. %.4s, firmware rev. %.4s, RMC rev. %.4s, "
+		"SMT ver. %u\n", fp->name, rom_rev, fw_rev, rmc_rev, smt_ver);
+
+	/* Now that we fetched initial parameters just shut the interface
+	 * until opened.
+	 */
+	ret = fza_close(dev);
+	if (ret != 0)
+		goto err_out_irq;
+
+	/* The FZA-specific entries in the device structure. */
+	dev->netdev_ops = &netdev_ops;
+
+	ret = register_netdev(dev);
+	if (ret != 0)
+		goto err_out_irq;
+
+	pr_info("%s: registered as %s\n", fp->name, dev->name);
+	fp->name = (const char *)dev->name;
+
+	get_device(bdev);
+	return 0;
+
+err_out_irq:
+	timer_delete_sync(&fp->reset_timer);
+	fza_do_shutdown(fp);
+	free_irq(dev->irq, dev);
+
+err_out_map:
+	iounmap(mmio);
+
+err_out_resource:
+	release_mem_region(start, len);
+
+err_out_kfree:
+	pr_err("%s: initialization failure, aborting!\n", fp->name);
+	free_netdev(dev);
+	return ret;
+}
+
+static int fza_remove(struct device *bdev)
+{
+	struct net_device *dev = dev_get_drvdata(bdev);
+	struct fza_private *fp = netdev_priv(dev);
+	struct tc_dev *tdev = to_tc_dev(bdev);
+	resource_size_t start, len;
+
+	put_device(bdev);
+
+	unregister_netdev(dev);
+
+	timer_delete_sync(&fp->reset_timer);
+	fza_do_shutdown(fp);
+	free_irq(dev->irq, dev);
+
+	iounmap(fp->mmio);
+
+	start = tdev->resource.start;
+	len = tdev->resource.end - start + 1;
+	release_mem_region(start, len);
+
+	free_netdev(dev);
+
+	return 0;
+}
+
+static struct tc_device_id const fza_tc_table[] = {
+	{ "DEC     ", "PMAF-AA " },
+	{ }
+};
+MODULE_DEVICE_TABLE(tc, fza_tc_table);
+
+static struct tc_driver fza_driver = {
+	.id_table	= fza_tc_table,
+	.driver		= {
+		.name	= "defza",
+		.bus	= &tc_bus_type,
+		.probe	= fza_probe,
+		.remove	= fza_remove,
+	},
+};
+
+static int fza_init(void)
+{
+	return tc_register_driver(&fza_driver);
+}
+
+static void fza_exit(void)
+{
+	tc_unregister_driver(&fza_driver);
+}
+
+module_init(fza_init);
+module_exit(fza_exit);