Merge tag 'trace-v7.3-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/trace/linux-tracegrafted

Pull tracing fixes from Steven Rostedt:

 - Don't destroy user event fields when removal fails

   User event fields are destroyed before the event is removed from
   visibility. But that can fail leaving the still visible event with no
   fields. Move the destroying of the fields to after the event is
   successfully removed from visibility.

 - Initialize function graph state is fork before calling
   copy_exec_state()

   For non-CLONE_VM forks, copy_exec_state() allocates a new
   task_exec_state. If that allocation fails, ftrace_graph_exit_task()
   will free the tasks ret_stack pointer. Since that pointer is still
   using the parent's ret_stack, it mistakenly frees the parent's
   pointer too.

   Call ftrace_graph_init() on the task first which will NULL out the
   new tasks's ret_stack and if the copy fails, it will not free
   anything.

 - Remove FGRAPH_MAX_INDEX

   The macro FGRAPH_MAX_INDEX was added but never used. Remove it.

 - Save ent_size in function graph printing of nested functions

   The function graph tracer needs to look at the next event to see if
   the next event is the return of the current function entry. If it is,
   it prints a single line:

	ktime_get();

   Otherwise it prints it like a nested function:

	tick_nohz_irq_exit() {
	    ktime_get();
	    kcpustat_irq_exit();
	}

   In order to look at the next event, it must save the current event so
   that it has the information to print from it. It saves the event in
   the iterator descriptor called "ent". What it doesn't save is the
   ent_size of the event which is now used to know if the function graph
   arguments are to be printed. The peek doesn't save the size so the
   size used happens to be that of the size of the last event that was
   seen.

   Save the entry event size in the iterator descriptor so that the
   correct size is used.

 - Fix several errors with freeing data in the histogram code

   The histogram code had a lot of leaked or or incorrect accounting
   when failures happen. Correct them.

 - Fix histogram regression of .percent and .graph modifiers

   Up until 6.3 histogram values could have "percent" or "graph"
   modifiers that changed how they were printed. But a change that added
   restricting histograms values from being strings, stack traces and
   other modifiers inadvertently prevented them from using the percent
   and graph modifiers, which were legal use cases for values.

   Put back the percent and graph modifiers.

 - Fix various typos in the comments

 - Set the trace_clock before initializing a histogram with clock
   argument

   The histogram API allows the user to specific which trace clock to
   use via a "clock=" string. The histogram is set up first before the
   clock is checked. If the passed in clock is not valid, it exits
   without fully fixing up the histogram leaving it on the list and a
   use-after-free can trigger.

   Update the clock argument first and if it fails then exit gracefully
   before the histogram trigger is placed on any lists.

 - Restore :mod: trailer after parsing in ftrace_set_clr_event

   The function ftrace_set_clr_event() modifies the parse string and
   needs to put it back to what was passed in. It searches for ":mod:"
   via a strsep() but fails to put back the first ':' in the string.

   Add back the ':' in the passed in string.

 - Take trace_array reference when opening a tracer options file

   The options files are dynamically created and some tracers add their
   own options. When a tracer adds their own list of options, the
   trace_array holding them has an array to hold the list of options for
   each tracer. This array increases in size via a krealloc(), and the
   new entry gets a newly allocated array to hold the options of the new
   tracer being added.

   The element in each entry of the tracer's option array holds a
   pointer back to the trace_array, a pointer to the tracer it is
   associated to, a pointer to the flags of the option.

   The issue is that these arrays are freed when the trace_array is
   freed when its instance it represents is removed from the instances
   directory. There's a race that an open of one of these options files
   can happen when the instance is being removed.

   Add a new helper function to be called by the open function of the
   options file to iterate all existing trace_arrays under a lock and
   find the one that has the given option element in one of it's tracer
   arrays. If found, then update the associated trace_array's reference
   counter to keep it from being freed. If not found, have the open call
   return -ENODEV.

 - Disable interrupts when acquiring the lock in rb_wake_up_waiters()

   The function rb_wake_up_waiters() assumes it will be called in
   interrupt context and does not disable irqs when taking
   cpu_buffer->reader_lock, which can be called in hard interrupt
   context. The issue is in PREEMPT_RT, this function is called in
   thread context leaving this lock open to a deadlock.

   Take the lock with interrupts disabled.

 - Use rcu_assign_pointer() for tmp_ops filter hash

   The tmp_ops used in update_ftrace_direct_mod() assigns its
   filter_hash field directly, but that field is annotated as __rcu and
   sparse complains. Assign it with rcu_assign_pointer()

 - Fix use-after-free in enable_trigger_private_data_free()

   The trace_event_call is accessed through the event_trigger_data's
   trace_event_file pointer to put the trace_event_call on freeing. The
   issue is that the trace_event_file data may have been freed already
   causing a use-after-free. Add a field to the event_trigger_data that
   points directly to the trace_event_call so that it can decrement its
   reference directly without needing to go through the
   trace_event_file.

 - Fix accounting of buffer data remote headers

   trace_buffer_desc_size() and trace_remote_alloc_buffer() undercount
   the number of pages is needed for the asked for size as it doesn't
   take into account the meta data on each page. Add a helper function
   to do the calculation properly and use that in these functions.

 - Catch nr_page_va overflow in ring_buffer_desc sizing

   The number of pages per remote ring buffer is capped by
   ring_buffer_desc::nr_page_va (32 bits). A buffer_size large enough to
   overflow that field would silently allocate a descriptor smaller than
   what was asked for.

 - Do not resize the subbuf order if any per_cpu buffer is disabled

   The mmapping of ring buffers disables resizing the subbuffers, but it
   is done per-cpu whereas the subbuf size change is done for all the
   per_cpu buffers under the buffer->mutex. It could change the size of
   some while the mapping is happening on others. Have the resize of the
   subbuf order check all the per_cpu buffers under the lock to see if
   any of them is disabled before starting and causing an inconsistency
   between buffers that are being mapped.

* tag 'trace-v7.3-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/trace/linux-trace: (25 commits)
  ring-buffer: Check resize_disabled before publishing the new subbuf order
  tracing/remotes: Catch nr_page_va overflow in ring_buffer_desc sizing
  tracing/remotes: Account for ring buffer page header in size calculation
  tracing: Don't dereference trace_event_file in deferred trigger free
  ftrace: Use rcu_assign_pointer() for tmp_ops filter hash
  ring-buffer: Acquire the lock with irqsave in rb_wake_up_waiters()
  tracing: Take trace_array reference when opening a tracer options file
  tracing: Fix ring_buffer_read_page_size() kernel-doc
  tracing: Restore :mod: trailer after parsing in ftrace_set_clr_event()
  tracing: Fix memory corruption from a "STACKTRACE" histogram key
  tracing: Fix memory corruption from the histogram stacktrace modifier
  tracing: Undo the registration when enabling the histogram trigger fails
  tracing: Take the reference before publishing the named histogram trigger
  tracing: Set the trace clock before registering the histogram trigger
  tracing: Fix typo "preceeded" in comment
  tracing: Fix typo "availabe" in comment
  tracing: Let histogram values keep the percent and graph modifiers
  tracing: Keep the entry count when the histogram stats allocation fails
  tracing: Free histogram the field rejected for a bad modifier
  tracing: Free histogram the var ref when its initialization fails
  ...

7 files changed, 3758 insertions, 0 deletions

diff --git a/drivers/infiniband/ulp/iser/Kconfig b/drivers/infiniband/ulp/iser/Kconfig
new file mode 100644
index 000000000..3016a0c9a
--- /dev/null
+++ b/drivers/infiniband/ulp/iser/Kconfig
@@ -0,0 +1,13 @@
+# SPDX-License-Identifier: GPL-2.0-only
+config INFINIBAND_ISER
+	tristate "iSCSI Extensions for RDMA (iSER)"
+	depends on SCSI && INET && INFINIBAND_ADDR_TRANS
+	select SCSI_ISCSI_ATTRS
+	help
+	  Support for the iSCSI Extensions for RDMA (iSER) Protocol
+	  over InfiniBand. This allows you to access storage devices
+	  that speak iSCSI over iSER over InfiniBand.
+
+	  The iSER protocol is defined by IETF.
+	  See <http://www.ietf.org/rfc/rfc5046.txt>
+	  and <http://members.infinibandta.org/kwspub/spec/Annex_iSER.PDF>
diff --git a/drivers/infiniband/ulp/iser/Makefile b/drivers/infiniband/ulp/iser/Makefile
new file mode 100644
index 000000000..2f3e78863
--- /dev/null
+++ b/drivers/infiniband/ulp/iser/Makefile
@@ -0,0 +1,5 @@
+# SPDX-License-Identifier: GPL-2.0-only
+obj-$(CONFIG_INFINIBAND_ISER)	+= ib_iser.o
+
+ib_iser-y			:= iser_verbs.o iser_initiator.o iser_memory.o \
+				   iscsi_iser.o
diff --git a/drivers/infiniband/ulp/iser/iscsi_iser.c b/drivers/infiniband/ulp/iser/iscsi_iser.c
new file mode 100644
index 000000000..7df441685
--- /dev/null
+++ b/drivers/infiniband/ulp/iser/iscsi_iser.c
@@ -0,0 +1,1083 @@
+/*
+ * iSCSI Initiator over iSER Data-Path
+ *
+ * Copyright (C) 2004 Dmitry Yusupov
+ * Copyright (C) 2004 Alex Aizman
+ * Copyright (C) 2005 Mike Christie
+ * Copyright (c) 2005, 2006 Voltaire, Inc. All rights reserved.
+ * Copyright (c) 2013-2014 Mellanox Technologies. All rights reserved.
+ * maintained by openib-general@openib.org
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *	- Redistributions of source code must retain the above
+ *	  copyright notice, this list of conditions and the following
+ *	  disclaimer.
+ *
+ *	- Redistributions in binary form must reproduce the above
+ *	  copyright notice, this list of conditions and the following
+ *	  disclaimer in the documentation and/or other materials
+ *	  provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ *
+ * Credits:
+ *	Christoph Hellwig
+ *	FUJITA Tomonori
+ *	Arne Redlich
+ *	Zhenyu Wang
+ * Modified by:
+ *      Erez Zilber
+ */
+
+#include <linux/types.h>
+#include <linux/list.h>
+#include <linux/hardirq.h>
+#include <linux/kfifo.h>
+#include <linux/blkdev.h>
+#include <linux/init.h>
+#include <linux/ioctl.h>
+#include <linux/cdev.h>
+#include <linux/in.h>
+#include <linux/net.h>
+#include <linux/scatterlist.h>
+#include <linux/delay.h>
+#include <linux/slab.h>
+#include <linux/module.h>
+
+#include <net/sock.h>
+
+#include <linux/uaccess.h>
+
+#include <scsi/scsi_cmnd.h>
+#include <scsi/scsi_device.h>
+#include <scsi/scsi_eh.h>
+#include <scsi/scsi_tcq.h>
+#include <scsi/scsi_host.h>
+#include <scsi/scsi.h>
+#include <scsi/scsi_transport_iscsi.h>
+
+#include "iscsi_iser.h"
+
+MODULE_DESCRIPTION("iSER (iSCSI Extensions for RDMA) Datamover");
+MODULE_LICENSE("Dual BSD/GPL");
+MODULE_AUTHOR("Alex Nezhinsky, Dan Bar Dov, Or Gerlitz");
+
+static const struct scsi_host_template iscsi_iser_sht;
+static struct iscsi_transport iscsi_iser_transport;
+static struct scsi_transport_template *iscsi_iser_scsi_transport;
+static struct workqueue_struct *release_wq;
+static DEFINE_MUTEX(unbind_iser_conn_mutex);
+struct iser_global ig;
+
+int iser_debug_level = 0;
+module_param_named(debug_level, iser_debug_level, int, S_IRUGO | S_IWUSR);
+MODULE_PARM_DESC(debug_level, "Enable debug tracing if > 0 (default:disabled)");
+
+static int iscsi_iser_set(const char *val, const struct kernel_param *kp);
+static const struct kernel_param_ops iscsi_iser_size_ops = {
+	.set = iscsi_iser_set,
+	.get = param_get_uint,
+};
+
+static unsigned int iscsi_max_lun = 512;
+module_param_cb(max_lun, &iscsi_iser_size_ops, &iscsi_max_lun, S_IRUGO);
+MODULE_PARM_DESC(max_lun, "Max LUNs to allow per session, should > 0 (default:512)");
+
+unsigned int iser_max_sectors = ISER_DEF_MAX_SECTORS;
+module_param_cb(max_sectors, &iscsi_iser_size_ops, &iser_max_sectors,
+		S_IRUGO | S_IWUSR);
+MODULE_PARM_DESC(max_sectors, "Max number of sectors in a single scsi command, should > 0 (default:1024)");
+
+bool iser_always_reg = true;
+module_param_named(always_register, iser_always_reg, bool, S_IRUGO);
+MODULE_PARM_DESC(always_register,
+		 "Always register memory, even for continuous memory regions (default:true)");
+
+bool iser_pi_enable = false;
+module_param_named(pi_enable, iser_pi_enable, bool, S_IRUGO);
+MODULE_PARM_DESC(pi_enable, "Enable T10-PI offload support (default:disabled)");
+
+static int iscsi_iser_set(const char *val, const struct kernel_param *kp)
+{
+	int ret;
+	unsigned int n = 0;
+
+	ret = kstrtouint(val, 10, &n);
+	if (ret != 0 || n == 0)
+		return -EINVAL;
+
+	return param_set_uint(val, kp);
+}
+
+/*
+ * iscsi_iser_recv() - Process a successful recv completion
+ * @conn:         iscsi connection
+ * @hdr:          iscsi header
+ * @rx_data:      buffer containing receive data payload
+ * @rx_data_len:  length of rx_data
+ *
+ * Notes: In case of data length errors or iscsi PDU completion failures
+ *        this routine will signal iscsi layer of connection failure.
+ */
+void iscsi_iser_recv(struct iscsi_conn *conn, struct iscsi_hdr *hdr,
+		     char *rx_data, int rx_data_len)
+{
+	int rc = 0;
+	int datalen;
+
+	/* verify PDU length */
+	datalen = ntoh24(hdr->dlength);
+	if (datalen > rx_data_len || (datalen + 4) < rx_data_len) {
+		iser_err("wrong datalen %d (hdr), %d (IB)\n",
+			datalen, rx_data_len);
+		rc = ISCSI_ERR_DATALEN;
+		goto error;
+	}
+
+	if (datalen != rx_data_len)
+		iser_dbg("aligned datalen (%d) hdr, %d (IB)\n",
+			datalen, rx_data_len);
+
+	rc = iscsi_complete_pdu(conn, hdr, rx_data, rx_data_len);
+	if (rc && rc != ISCSI_ERR_NO_SCSI_CMD)
+		goto error;
+
+	return;
+error:
+	iscsi_conn_failure(conn, rc);
+}
+
+/**
+ * iscsi_iser_pdu_alloc() - allocate an iscsi-iser PDU
+ * @task:     iscsi task
+ * @opcode:   iscsi command opcode
+ *
+ * Netes: This routine can't fail, just assign iscsi task
+ *        hdr and max hdr size.
+ */
+static int iscsi_iser_pdu_alloc(struct iscsi_task *task, uint8_t opcode)
+{
+	struct iscsi_iser_task *iser_task = task->dd_data;
+
+	task->hdr = (struct iscsi_hdr *)&iser_task->desc.iscsi_header;
+	task->hdr_max = sizeof(iser_task->desc.iscsi_header);
+
+	return 0;
+}
+
+/**
+ * iser_initialize_task_headers() - Initialize task headers
+ * @task:       iscsi task
+ * @tx_desc:    iser tx descriptor
+ *
+ * Notes:
+ * This routine may race with iser teardown flow for scsi
+ * error handling TMFs. So for TMF we should acquire the
+ * state mutex to avoid dereferencing the IB device which
+ * may have already been terminated.
+ */
+int iser_initialize_task_headers(struct iscsi_task *task,
+				 struct iser_tx_desc *tx_desc)
+{
+	struct iser_conn *iser_conn = task->conn->dd_data;
+	struct iser_device *device = iser_conn->ib_conn.device;
+	struct iscsi_iser_task *iser_task = task->dd_data;
+	u64 dma_addr;
+
+	if (unlikely(iser_conn->state != ISER_CONN_UP))
+		return -ENODEV;
+
+	dma_addr = ib_dma_map_single(device->ib_device, (void *)tx_desc,
+				ISER_HEADERS_LEN, DMA_TO_DEVICE);
+	if (ib_dma_mapping_error(device->ib_device, dma_addr))
+		return -ENOMEM;
+
+	tx_desc->inv_wr.next = NULL;
+	tx_desc->reg_wr.wr.next = NULL;
+	tx_desc->mapped = true;
+	tx_desc->dma_addr = dma_addr;
+	tx_desc->tx_sg[0].addr   = tx_desc->dma_addr;
+	tx_desc->tx_sg[0].length = ISER_HEADERS_LEN;
+	tx_desc->tx_sg[0].lkey   = device->pd->local_dma_lkey;
+
+	iser_task->iser_conn = iser_conn;
+
+	return 0;
+}
+
+/**
+ * iscsi_iser_task_init() - Initialize iscsi-iser task
+ * @task: iscsi task
+ *
+ * Initialize the task for the scsi command or mgmt command.
+ *
+ * Return: Returns zero on success or -ENOMEM when failing
+ *         to init task headers (dma mapping error).
+ */
+static int iscsi_iser_task_init(struct iscsi_task *task)
+{
+	struct iscsi_iser_task *iser_task = task->dd_data;
+	int ret;
+
+	ret = iser_initialize_task_headers(task, &iser_task->desc);
+	if (ret) {
+		iser_err("Failed to init task %p, err = %d\n",
+			 iser_task, ret);
+		return ret;
+	}
+
+	/* mgmt task */
+	if (!task->sc)
+		return 0;
+
+	iser_task->command_sent = 0;
+	iser_task_rdma_init(iser_task);
+	iser_task->sc = task->sc;
+
+	return 0;
+}
+
+/**
+ * iscsi_iser_mtask_xmit() - xmit management (immediate) task
+ * @conn: iscsi connection
+ * @task: task management task
+ *
+ * Notes:
+ *	The function can return -EAGAIN in which case caller must
+ *	call it again later, or recover. '0' return code means successful
+ *	xmit.
+ *
+ **/
+static int iscsi_iser_mtask_xmit(struct iscsi_conn *conn,
+				 struct iscsi_task *task)
+{
+	iser_dbg("mtask xmit [cid %d itt 0x%x]\n", conn->id, task->itt);
+
+	/* since iser xmits control with zero copy, tasks can not be recycled
+	 * right after sending them.
+	 * The recycling scheme is based on whether a response is expected
+	 * - if yes, the task is recycled at iscsi_complete_pdu
+	 * - if no,  the task is recycled at iser_snd_completion
+	 */
+	return iser_send_control(conn, task);
+}
+
+static int iscsi_iser_task_xmit_unsol_data(struct iscsi_conn *conn,
+					   struct iscsi_task *task)
+{
+	struct iscsi_r2t_info *r2t = &task->unsol_r2t;
+	struct iscsi_data hdr;
+	int error = 0;
+
+	/* Send data-out PDUs while there's still unsolicited data to send */
+	while (iscsi_task_has_unsol_data(task)) {
+		iscsi_prep_data_out_pdu(task, r2t, &hdr);
+		iser_dbg("Sending data-out: itt 0x%x, data count %d\n",
+			   hdr.itt, r2t->data_count);
+
+		/* the buffer description has been passed with the command */
+		/* Send the command */
+		error = iser_send_data_out(conn, task, &hdr);
+		if (error) {
+			r2t->datasn--;
+			goto iscsi_iser_task_xmit_unsol_data_exit;
+		}
+		r2t->sent += r2t->data_count;
+		iser_dbg("Need to send %d more as data-out PDUs\n",
+			   r2t->data_length - r2t->sent);
+	}
+
+iscsi_iser_task_xmit_unsol_data_exit:
+	return error;
+}
+
+/**
+ * iscsi_iser_task_xmit() - xmit iscsi-iser task
+ * @task: iscsi task
+ *
+ * Return: zero on success or escalates $error on failure.
+ */
+static int iscsi_iser_task_xmit(struct iscsi_task *task)
+{
+	struct iscsi_conn *conn = task->conn;
+	struct iscsi_iser_task *iser_task = task->dd_data;
+	int error = 0;
+
+	if (!task->sc)
+		return iscsi_iser_mtask_xmit(conn, task);
+
+	if (task->sc->sc_data_direction == DMA_TO_DEVICE) {
+		BUG_ON(scsi_bufflen(task->sc) == 0);
+
+		iser_dbg("cmd [itt %x total %d imm %d unsol_data %d\n",
+			   task->itt, scsi_bufflen(task->sc),
+			   task->imm_count, task->unsol_r2t.data_length);
+	}
+
+	iser_dbg("ctask xmit [cid %d itt 0x%x]\n",
+		   conn->id, task->itt);
+
+	/* Send the cmd PDU */
+	if (!iser_task->command_sent) {
+		error = iser_send_command(conn, task);
+		if (error)
+			goto iscsi_iser_task_xmit_exit;
+		iser_task->command_sent = 1;
+	}
+
+	/* Send unsolicited data-out PDU(s) if necessary */
+	if (iscsi_task_has_unsol_data(task))
+		error = iscsi_iser_task_xmit_unsol_data(conn, task);
+
+ iscsi_iser_task_xmit_exit:
+	return error;
+}
+
+/**
+ * iscsi_iser_cleanup_task() - cleanup an iscsi-iser task
+ * @task: iscsi task
+ *
+ * Notes: In case the RDMA device is already NULL (might have
+ *        been removed in DEVICE_REMOVAL CM event it will bail-out
+ *        without doing dma unmapping.
+ */
+static void iscsi_iser_cleanup_task(struct iscsi_task *task)
+{
+	struct iscsi_iser_task *iser_task = task->dd_data;
+	struct iser_tx_desc *tx_desc = &iser_task->desc;
+	struct iser_conn *iser_conn = task->conn->dd_data;
+	struct iser_device *device = iser_conn->ib_conn.device;
+
+	/* DEVICE_REMOVAL event might have already released the device */
+	if (!device)
+		return;
+
+	if (likely(tx_desc->mapped)) {
+		ib_dma_unmap_single(device->ib_device, tx_desc->dma_addr,
+				    ISER_HEADERS_LEN, DMA_TO_DEVICE);
+		tx_desc->mapped = false;
+	}
+
+	/* mgmt tasks do not need special cleanup */
+	if (!task->sc)
+		return;
+
+	if (iser_task->status == ISER_TASK_STATUS_STARTED) {
+		iser_task->status = ISER_TASK_STATUS_COMPLETED;
+		iser_task_rdma_finalize(iser_task);
+	}
+}
+
+/**
+ * iscsi_iser_check_protection() - check protection information status of task.
+ * @task:     iscsi task
+ * @sector:   error sector if exsists (output)
+ *
+ * Return: zero if no data-integrity errors have occurred
+ *         0x1: data-integrity error occurred in the guard-block
+ *         0x2: data-integrity error occurred in the reference tag
+ *         0x3: data-integrity error occurred in the application tag
+ *
+ *         In addition the error sector is marked.
+ */
+static u8 iscsi_iser_check_protection(struct iscsi_task *task, sector_t *sector)
+{
+	struct iscsi_iser_task *iser_task = task->dd_data;
+	enum iser_data_dir dir = iser_task->dir[ISER_DIR_IN] ?
+					ISER_DIR_IN : ISER_DIR_OUT;
+
+	return iser_check_task_pi_status(iser_task, dir, sector);
+}
+
+/**
+ * iscsi_iser_conn_create() - create a new iscsi-iser connection
+ * @cls_session: iscsi class connection
+ * @conn_idx:    connection index within the session (for MCS)
+ *
+ * Return: iscsi_cls_conn when iscsi_conn_setup succeeds or NULL
+ *         otherwise.
+ */
+static struct iscsi_cls_conn *
+iscsi_iser_conn_create(struct iscsi_cls_session *cls_session,
+		       uint32_t conn_idx)
+{
+	struct iscsi_conn *conn;
+	struct iscsi_cls_conn *cls_conn;
+
+	cls_conn = iscsi_conn_setup(cls_session, 0, conn_idx);
+	if (!cls_conn)
+		return NULL;
+	conn = cls_conn->dd_data;
+
+	/*
+	 * due to issues with the login code re iser sematics
+	 * this not set in iscsi_conn_setup - FIXME
+	 */
+	conn->max_recv_dlength = ISER_RECV_DATA_SEG_LEN;
+
+	return cls_conn;
+}
+
+/**
+ * iscsi_iser_conn_bind() - bind iscsi and iser connection structures
+ * @cls_session:     iscsi class session
+ * @cls_conn:        iscsi class connection
+ * @transport_eph:   transport end-point handle
+ * @is_leading:      indicate if this is the session leading connection (MCS)
+ *
+ * Return: zero on success, $error if iscsi_conn_bind fails and
+ *         -EINVAL in case end-point doesn't exists anymore or iser connection
+ *         state is not UP (teardown already started).
+ */
+static int iscsi_iser_conn_bind(struct iscsi_cls_session *cls_session,
+				struct iscsi_cls_conn *cls_conn,
+				uint64_t transport_eph, int is_leading)
+{
+	struct iscsi_conn *conn = cls_conn->dd_data;
+	struct iser_conn *iser_conn;
+	struct iscsi_endpoint *ep;
+	int error;
+
+	error = iscsi_conn_bind(cls_session, cls_conn, is_leading);
+	if (error)
+		return error;
+
+	/* the transport ep handle comes from user space so it must be
+	 * verified against the global ib connections list */
+	ep = iscsi_lookup_endpoint(transport_eph);
+	if (!ep) {
+		iser_err("can't bind eph %llx\n",
+			 (unsigned long long)transport_eph);
+		return -EINVAL;
+	}
+	iser_conn = ep->dd_data;
+
+	mutex_lock(&iser_conn->state_mutex);
+	if (iser_conn->state != ISER_CONN_UP) {
+		error = -EINVAL;
+		iser_err("iser_conn %p state is %d, teardown started\n",
+			 iser_conn, iser_conn->state);
+		goto out;
+	}
+
+	error = iser_alloc_rx_descriptors(iser_conn, conn->session);
+	if (error)
+		goto out;
+
+	/* binds the iSER connection retrieved from the previously
+	 * connected ep_handle to the iSCSI layer connection. exchanges
+	 * connection pointers */
+	iser_info("binding iscsi conn %p to iser_conn %p\n", conn, iser_conn);
+
+	conn->dd_data = iser_conn;
+	iser_conn->iscsi_conn = conn;
+
+out:
+	iscsi_put_endpoint(ep);
+	mutex_unlock(&iser_conn->state_mutex);
+	return error;
+}
+
+/**
+ * iscsi_iser_conn_start() - start iscsi-iser connection
+ * @cls_conn: iscsi class connection
+ *
+ * Notes: Here iser intialize (or re-initialize) stop_completion as
+ *        from this point iscsi must call conn_stop in session/connection
+ *        teardown so iser transport must wait for it.
+ */
+static int iscsi_iser_conn_start(struct iscsi_cls_conn *cls_conn)
+{
+	struct iscsi_conn *iscsi_conn;
+	struct iser_conn *iser_conn;
+
+	iscsi_conn = cls_conn->dd_data;
+	iser_conn = iscsi_conn->dd_data;
+	reinit_completion(&iser_conn->stop_completion);
+
+	return iscsi_conn_start(cls_conn);
+}
+
+/**
+ * iscsi_iser_conn_stop() - stop iscsi-iser connection
+ * @cls_conn:  iscsi class connection
+ * @flag:      indicate if recover or terminate (passed as is)
+ *
+ * Notes: Calling iscsi_conn_stop might theoretically race with
+ *        DEVICE_REMOVAL event and dereference a previously freed RDMA device
+ *        handle, so we call it under iser the state lock to protect against
+ *        this kind of race.
+ */
+static void iscsi_iser_conn_stop(struct iscsi_cls_conn *cls_conn, int flag)
+{
+	struct iscsi_conn *conn = cls_conn->dd_data;
+	struct iser_conn *iser_conn = conn->dd_data;
+
+	iser_info("stopping iscsi_conn: %p, iser_conn: %p\n", conn, iser_conn);
+
+	/*
+	 * Userspace may have goofed up and not bound the connection or
+	 * might have only partially setup the connection.
+	 */
+	if (iser_conn) {
+		mutex_lock(&iser_conn->state_mutex);
+		mutex_lock(&unbind_iser_conn_mutex);
+		iser_conn_terminate(iser_conn);
+		iscsi_conn_stop(cls_conn, flag);
+
+		/* unbind */
+		iser_conn->iscsi_conn = NULL;
+		conn->dd_data = NULL;
+		mutex_unlock(&unbind_iser_conn_mutex);
+
+		complete(&iser_conn->stop_completion);
+		mutex_unlock(&iser_conn->state_mutex);
+	} else {
+		iscsi_conn_stop(cls_conn, flag);
+	}
+}
+
+/**
+ * iscsi_iser_session_destroy() - destroy iscsi-iser session
+ * @cls_session: iscsi class session
+ *
+ * Removes and free iscsi host.
+ */
+static void iscsi_iser_session_destroy(struct iscsi_cls_session *cls_session)
+{
+	struct Scsi_Host *shost = iscsi_session_to_shost(cls_session);
+
+	iscsi_session_teardown(cls_session);
+	iscsi_host_remove(shost, false);
+	iscsi_host_free(shost);
+}
+
+static inline unsigned int iser_dif_prot_caps(int prot_caps)
+{
+	int ret = 0;
+
+	if (prot_caps & IB_PROT_T10DIF_TYPE_1)
+		ret |= SHOST_DIF_TYPE1_PROTECTION |
+		       SHOST_DIX_TYPE0_PROTECTION |
+		       SHOST_DIX_TYPE1_PROTECTION;
+	if (prot_caps & IB_PROT_T10DIF_TYPE_2)
+		ret |= SHOST_DIF_TYPE2_PROTECTION |
+		       SHOST_DIX_TYPE2_PROTECTION;
+	if (prot_caps & IB_PROT_T10DIF_TYPE_3)
+		ret |= SHOST_DIF_TYPE3_PROTECTION |
+		       SHOST_DIX_TYPE3_PROTECTION;
+
+	return ret;
+}
+
+/**
+ * iscsi_iser_session_create() - create an iscsi-iser session
+ * @ep:             iscsi end-point handle
+ * @cmds_max:       maximum commands in this session
+ * @qdepth:         session command queue depth
+ * @initial_cmdsn:  initiator command sequnce number
+ *
+ * Allocates and adds a scsi host, expose DIF supprot if
+ * exists, and sets up an iscsi session.
+ */
+static struct iscsi_cls_session *
+iscsi_iser_session_create(struct iscsi_endpoint *ep,
+			  uint16_t cmds_max, uint16_t qdepth,
+			  uint32_t initial_cmdsn)
+{
+	struct iscsi_cls_session *cls_session;
+	struct Scsi_Host *shost;
+	struct iser_conn *iser_conn = NULL;
+	struct ib_conn *ib_conn;
+	struct ib_device *ib_dev;
+	u32 max_fr_sectors;
+
+	shost = iscsi_host_alloc(&iscsi_iser_sht, 0, 0);
+	if (!shost)
+		return NULL;
+	shost->transportt = iscsi_iser_scsi_transport;
+	shost->cmd_per_lun = qdepth;
+	shost->max_lun = iscsi_max_lun;
+	shost->max_id = 0;
+	shost->max_channel = 0;
+	shost->max_cmd_len = 16;
+
+	/*
+	 * older userspace tools (before 2.0-870) did not pass us
+	 * the leading conn's ep so this will be NULL;
+	 */
+	if (ep) {
+		iser_conn = ep->dd_data;
+		shost->sg_tablesize = iser_conn->scsi_sg_tablesize;
+		shost->can_queue = min_t(u16, cmds_max, iser_conn->max_cmds);
+
+		mutex_lock(&iser_conn->state_mutex);
+		if (iser_conn->state != ISER_CONN_UP) {
+			iser_err("iser conn %p already started teardown\n",
+				 iser_conn);
+			mutex_unlock(&iser_conn->state_mutex);
+			goto free_host;
+		}
+
+		ib_conn = &iser_conn->ib_conn;
+		ib_dev = ib_conn->device->ib_device;
+		if (ib_conn->pi_support) {
+			u32 sig_caps = ib_dev->attrs.sig_prot_cap;
+
+			shost->sg_prot_tablesize = shost->sg_tablesize;
+			scsi_host_set_prot(shost, iser_dif_prot_caps(sig_caps));
+			scsi_host_set_guard(shost, SHOST_DIX_GUARD_IP |
+						   SHOST_DIX_GUARD_CRC);
+		}
+
+		if (!(ib_dev->attrs.kernel_cap_flags & IBK_SG_GAPS_REG))
+			shost->virt_boundary_mask = SZ_4K - 1;
+
+		if (iscsi_host_add(shost, ib_dev->dev.parent)) {
+			mutex_unlock(&iser_conn->state_mutex);
+			goto free_host;
+		}
+		mutex_unlock(&iser_conn->state_mutex);
+	} else {
+		shost->can_queue = min_t(u16, cmds_max, ISER_DEF_XMIT_CMDS_MAX);
+		if (iscsi_host_add(shost, NULL))
+			goto free_host;
+	}
+
+	max_fr_sectors = (shost->sg_tablesize * PAGE_SIZE) >> 9;
+	shost->max_sectors = min(iser_max_sectors, max_fr_sectors);
+
+	iser_dbg("iser_conn %p, sg_tablesize %u, max_sectors %u\n",
+		 iser_conn, shost->sg_tablesize,
+		 shost->max_sectors);
+
+	if (shost->max_sectors < iser_max_sectors)
+		iser_warn("max_sectors was reduced from %u to %u\n",
+			  iser_max_sectors, shost->max_sectors);
+
+	cls_session = iscsi_session_setup(&iscsi_iser_transport, shost,
+					  shost->can_queue, 0,
+					  sizeof(struct iscsi_iser_task),
+					  initial_cmdsn, 0);
+	if (!cls_session)
+		goto remove_host;
+
+	return cls_session;
+
+remove_host:
+	iscsi_host_remove(shost, false);
+free_host:
+	iscsi_host_free(shost);
+	return NULL;
+}
+
+static int iscsi_iser_set_param(struct iscsi_cls_conn *cls_conn,
+				enum iscsi_param param, char *buf, int buflen)
+{
+	int value;
+
+	switch (param) {
+	case ISCSI_PARAM_MAX_RECV_DLENGTH:
+		/* TBD */
+		break;
+	case ISCSI_PARAM_HDRDGST_EN:
+		sscanf(buf, "%d", &value);
+		if (value) {
+			iser_err("DataDigest wasn't negotiated to None\n");
+			return -EPROTO;
+		}
+		break;
+	case ISCSI_PARAM_DATADGST_EN:
+		sscanf(buf, "%d", &value);
+		if (value) {
+			iser_err("DataDigest wasn't negotiated to None\n");
+			return -EPROTO;
+		}
+		break;
+	case ISCSI_PARAM_IFMARKER_EN:
+		sscanf(buf, "%d", &value);
+		if (value) {
+			iser_err("IFMarker wasn't negotiated to No\n");
+			return -EPROTO;
+		}
+		break;
+	case ISCSI_PARAM_OFMARKER_EN:
+		sscanf(buf, "%d", &value);
+		if (value) {
+			iser_err("OFMarker wasn't negotiated to No\n");
+			return -EPROTO;
+		}
+		break;
+	default:
+		return iscsi_set_param(cls_conn, param, buf, buflen);
+	}
+
+	return 0;
+}
+
+/**
+ * iscsi_iser_conn_get_stats() - get iscsi connection statistics
+ * @cls_conn:    iscsi class connection
+ * @stats:       iscsi stats to output
+ *
+ * Output connection statistics.
+ */
+static void iscsi_iser_conn_get_stats(struct iscsi_cls_conn *cls_conn,
+				      struct iscsi_stats *stats)
+{
+	struct iscsi_conn *conn = cls_conn->dd_data;
+
+	stats->txdata_octets = conn->txdata_octets;
+	stats->rxdata_octets = conn->rxdata_octets;
+	stats->scsicmd_pdus = conn->scsicmd_pdus_cnt;
+	stats->dataout_pdus = conn->dataout_pdus_cnt;
+	stats->scsirsp_pdus = conn->scsirsp_pdus_cnt;
+	stats->datain_pdus = conn->datain_pdus_cnt; /* always 0 */
+	stats->r2t_pdus = conn->r2t_pdus_cnt; /* always 0 */
+	stats->tmfcmd_pdus = conn->tmfcmd_pdus_cnt;
+	stats->tmfrsp_pdus = conn->tmfrsp_pdus_cnt;
+	stats->custom_length = 0;
+}
+
+static int iscsi_iser_get_ep_param(struct iscsi_endpoint *ep,
+				   enum iscsi_param param, char *buf)
+{
+	struct iser_conn *iser_conn = ep->dd_data;
+
+	switch (param) {
+	case ISCSI_PARAM_CONN_PORT:
+	case ISCSI_PARAM_CONN_ADDRESS:
+		if (!iser_conn || !iser_conn->ib_conn.cma_id)
+			return -ENOTCONN;
+
+		return iscsi_conn_get_addr_param((struct sockaddr_storage *)
+				&iser_conn->ib_conn.cma_id->route.addr.dst_addr,
+				param, buf);
+	default:
+		break;
+	}
+	return -ENOSYS;
+}
+
+/**
+ * iscsi_iser_ep_connect() - Initiate iSER connection establishment
+ * @shost:          scsi_host
+ * @dst_addr:       destination address
+ * @non_blocking:   indicate if routine can block
+ *
+ * Allocate an iscsi endpoint, an iser_conn structure and bind them.
+ * After that start RDMA connection establishment via rdma_cm. We
+ * don't allocate iser_conn embedded in iscsi_endpoint since in teardown
+ * the endpoint will be destroyed at ep_disconnect while iser_conn will
+ * cleanup its resources asynchronuously.
+ *
+ * Return: iscsi_endpoint created by iscsi layer or ERR_PTR(error)
+ *         if fails.
+ */
+static struct iscsi_endpoint *iscsi_iser_ep_connect(struct Scsi_Host *shost,
+						    struct sockaddr *dst_addr,
+						    int non_blocking)
+{
+	int err;
+	struct iser_conn *iser_conn;
+	struct iscsi_endpoint *ep;
+
+	ep = iscsi_create_endpoint(0);
+	if (!ep)
+		return ERR_PTR(-ENOMEM);
+
+	iser_conn = kzalloc_obj(*iser_conn);
+	if (!iser_conn) {
+		err = -ENOMEM;
+		goto failure;
+	}
+
+	ep->dd_data = iser_conn;
+	iser_conn->ep = ep;
+	iser_conn_init(iser_conn);
+
+	err = iser_connect(iser_conn, NULL, dst_addr, non_blocking);
+	if (err)
+		goto failure;
+
+	return ep;
+failure:
+	iscsi_destroy_endpoint(ep);
+	return ERR_PTR(err);
+}
+
+/**
+ * iscsi_iser_ep_poll() - poll for iser connection establishment to complete
+ * @ep:            iscsi endpoint (created at ep_connect)
+ * @timeout_ms:    polling timeout allowed in ms.
+ *
+ * This routine boils down to waiting for up_completion signaling
+ * that cma_id got CONNECTED event.
+ *
+ * Return: 1 if succeeded in connection establishment, 0 if timeout expired
+ *         (libiscsi will retry will kick in) or -1 if interrupted by signal
+ *         or more likely iser connection state transitioned to TEMINATING or
+ *         DOWN during the wait period.
+ */
+static int iscsi_iser_ep_poll(struct iscsi_endpoint *ep, int timeout_ms)
+{
+	struct iser_conn *iser_conn = ep->dd_data;
+	int rc;
+
+	rc = wait_for_completion_interruptible_timeout(&iser_conn->up_completion,
+						       msecs_to_jiffies(timeout_ms));
+	/* if conn establishment failed, return error code to iscsi */
+	if (rc == 0) {
+		mutex_lock(&iser_conn->state_mutex);
+		if (iser_conn->state == ISER_CONN_TERMINATING ||
+		    iser_conn->state == ISER_CONN_DOWN)
+			rc = -1;
+		mutex_unlock(&iser_conn->state_mutex);
+	}
+
+	iser_info("iser conn %p rc = %d\n", iser_conn, rc);
+
+	if (rc > 0)
+		return 1; /* success, this is the equivalent of EPOLLOUT */
+	else if (!rc)
+		return 0; /* timeout */
+	else
+		return rc; /* signal */
+}
+
+/**
+ * iscsi_iser_ep_disconnect() - Initiate connection teardown process
+ * @ep:    iscsi endpoint handle
+ *
+ * This routine is not blocked by iser and RDMA termination process
+ * completion as we queue a deffered work for iser/RDMA destruction
+ * and cleanup or actually call it immediately in case we didn't pass
+ * iscsi conn bind/start stage, thus it is safe.
+ */
+static void iscsi_iser_ep_disconnect(struct iscsi_endpoint *ep)
+{
+	struct iser_conn *iser_conn = ep->dd_data;
+
+	iser_info("ep %p iser conn %p\n", ep, iser_conn);
+
+	mutex_lock(&iser_conn->state_mutex);
+	iser_conn_terminate(iser_conn);
+
+	/*
+	 * if iser_conn and iscsi_conn are bound, we must wait for
+	 * iscsi_conn_stop and flush errors completion before freeing
+	 * the iser resources. Otherwise we are safe to free resources
+	 * immediately.
+	 */
+	if (iser_conn->iscsi_conn) {
+		INIT_WORK(&iser_conn->release_work, iser_release_work);
+		queue_work(release_wq, &iser_conn->release_work);
+		mutex_unlock(&iser_conn->state_mutex);
+	} else {
+		iser_conn->state = ISER_CONN_DOWN;
+		mutex_unlock(&iser_conn->state_mutex);
+		iser_conn_release(iser_conn);
+	}
+
+	iscsi_destroy_endpoint(ep);
+}
+
+static umode_t iser_attr_is_visible(int param_type, int param)
+{
+	switch (param_type) {
+	case ISCSI_HOST_PARAM:
+		switch (param) {
+		case ISCSI_HOST_PARAM_NETDEV_NAME:
+		case ISCSI_HOST_PARAM_HWADDRESS:
+		case ISCSI_HOST_PARAM_INITIATOR_NAME:
+			return S_IRUGO;
+		default:
+			return 0;
+		}
+	case ISCSI_PARAM:
+		switch (param) {
+		case ISCSI_PARAM_MAX_RECV_DLENGTH:
+		case ISCSI_PARAM_MAX_XMIT_DLENGTH:
+		case ISCSI_PARAM_HDRDGST_EN:
+		case ISCSI_PARAM_DATADGST_EN:
+		case ISCSI_PARAM_CONN_ADDRESS:
+		case ISCSI_PARAM_CONN_PORT:
+		case ISCSI_PARAM_EXP_STATSN:
+		case ISCSI_PARAM_PERSISTENT_ADDRESS:
+		case ISCSI_PARAM_PERSISTENT_PORT:
+		case ISCSI_PARAM_PING_TMO:
+		case ISCSI_PARAM_RECV_TMO:
+		case ISCSI_PARAM_INITIAL_R2T_EN:
+		case ISCSI_PARAM_MAX_R2T:
+		case ISCSI_PARAM_IMM_DATA_EN:
+		case ISCSI_PARAM_FIRST_BURST:
+		case ISCSI_PARAM_MAX_BURST:
+		case ISCSI_PARAM_PDU_INORDER_EN:
+		case ISCSI_PARAM_DATASEQ_INORDER_EN:
+		case ISCSI_PARAM_TARGET_NAME:
+		case ISCSI_PARAM_TPGT:
+		case ISCSI_PARAM_USERNAME:
+		case ISCSI_PARAM_PASSWORD:
+		case ISCSI_PARAM_USERNAME_IN:
+		case ISCSI_PARAM_PASSWORD_IN:
+		case ISCSI_PARAM_FAST_ABORT:
+		case ISCSI_PARAM_ABORT_TMO:
+		case ISCSI_PARAM_LU_RESET_TMO:
+		case ISCSI_PARAM_TGT_RESET_TMO:
+		case ISCSI_PARAM_IFACE_NAME:
+		case ISCSI_PARAM_INITIATOR_NAME:
+		case ISCSI_PARAM_DISCOVERY_SESS:
+			return S_IRUGO;
+		default:
+			return 0;
+		}
+	}
+
+	return 0;
+}
+
+static const struct scsi_host_template iscsi_iser_sht = {
+	.module                 = THIS_MODULE,
+	.name                   = "iSCSI Initiator over iSER",
+	.queuecommand           = iscsi_queuecommand,
+	.change_queue_depth	= scsi_change_queue_depth,
+	.sg_tablesize           = ISCSI_ISER_DEF_SG_TABLESIZE,
+	.cmd_per_lun            = ISER_DEF_CMD_PER_LUN,
+	.eh_timed_out		= iscsi_eh_cmd_timed_out,
+	.eh_abort_handler       = iscsi_eh_abort,
+	.eh_device_reset_handler= iscsi_eh_device_reset,
+	.eh_target_reset_handler = iscsi_eh_recover_target,
+	.target_alloc		= iscsi_target_alloc,
+	.proc_name              = "iscsi_iser",
+	.this_id                = -1,
+	.track_queue_depth	= 1,
+	.cmd_size		= sizeof(struct iscsi_cmd),
+};
+
+static struct iscsi_transport iscsi_iser_transport = {
+	.owner                  = THIS_MODULE,
+	.name                   = "iser",
+	.caps                   = CAP_RECOVERY_L0 | CAP_MULTI_R2T | CAP_TEXT_NEGO,
+	/* session management */
+	.create_session         = iscsi_iser_session_create,
+	.destroy_session        = iscsi_iser_session_destroy,
+	/* connection management */
+	.create_conn            = iscsi_iser_conn_create,
+	.bind_conn              = iscsi_iser_conn_bind,
+	.unbind_conn		= iscsi_conn_unbind,
+	.destroy_conn           = iscsi_conn_teardown,
+	.attr_is_visible	= iser_attr_is_visible,
+	.set_param              = iscsi_iser_set_param,
+	.get_conn_param		= iscsi_conn_get_param,
+	.get_ep_param		= iscsi_iser_get_ep_param,
+	.get_session_param	= iscsi_session_get_param,
+	.start_conn             = iscsi_iser_conn_start,
+	.stop_conn              = iscsi_iser_conn_stop,
+	/* iscsi host params */
+	.get_host_param		= iscsi_host_get_param,
+	.set_host_param		= iscsi_host_set_param,
+	/* IO */
+	.send_pdu		= iscsi_conn_send_pdu,
+	.get_stats		= iscsi_iser_conn_get_stats,
+	.init_task		= iscsi_iser_task_init,
+	.xmit_task		= iscsi_iser_task_xmit,
+	.cleanup_task		= iscsi_iser_cleanup_task,
+	.alloc_pdu		= iscsi_iser_pdu_alloc,
+	.check_protection	= iscsi_iser_check_protection,
+	/* recovery */
+	.session_recovery_timedout = iscsi_session_recovery_timedout,
+
+	.ep_connect             = iscsi_iser_ep_connect,
+	.ep_poll                = iscsi_iser_ep_poll,
+	.ep_disconnect          = iscsi_iser_ep_disconnect
+};
+
+static int __init iser_init(void)
+{
+	int err;
+
+	iser_dbg("Starting iSER datamover...\n");
+
+	memset(&ig, 0, sizeof(struct iser_global));
+
+	ig.desc_cache = kmem_cache_create("iser_descriptors",
+					  sizeof(struct iser_tx_desc),
+					  0, SLAB_HWCACHE_ALIGN,
+					  NULL);
+	if (ig.desc_cache == NULL)
+		return -ENOMEM;
+
+	/* device init is called only after the first addr resolution */
+	mutex_init(&ig.device_list_mutex);
+	INIT_LIST_HEAD(&ig.device_list);
+	mutex_init(&ig.connlist_mutex);
+	INIT_LIST_HEAD(&ig.connlist);
+
+	release_wq = alloc_workqueue("release workqueue", WQ_PERCPU, 0);
+	if (!release_wq) {
+		iser_err("failed to allocate release workqueue\n");
+		err = -ENOMEM;
+		goto err_alloc_wq;
+	}
+
+	iscsi_iser_scsi_transport = iscsi_register_transport(
+							&iscsi_iser_transport);
+	if (!iscsi_iser_scsi_transport) {
+		iser_err("iscsi_register_transport failed\n");
+		err = -EINVAL;
+		goto err_reg;
+	}
+
+	return 0;
+
+err_reg:
+	destroy_workqueue(release_wq);
+err_alloc_wq:
+	kmem_cache_destroy(ig.desc_cache);
+
+	return err;
+}
+
+static void __exit iser_exit(void)
+{
+	struct iser_conn *iser_conn, *n;
+	int connlist_empty;
+
+	iser_dbg("Removing iSER datamover...\n");
+	destroy_workqueue(release_wq);
+
+	mutex_lock(&ig.connlist_mutex);
+	connlist_empty = list_empty(&ig.connlist);
+	mutex_unlock(&ig.connlist_mutex);
+
+	if (!connlist_empty) {
+		iser_err("Error cleanup stage completed but we still have iser "
+			 "connections, destroying them anyway\n");
+		list_for_each_entry_safe(iser_conn, n, &ig.connlist,
+					 conn_list) {
+			iser_conn_release(iser_conn);
+		}
+	}
+
+	iscsi_unregister_transport(&iscsi_iser_transport);
+	kmem_cache_destroy(ig.desc_cache);
+}
+
+module_init(iser_init);
+module_exit(iser_exit);
diff --git a/drivers/infiniband/ulp/iser/iscsi_iser.h b/drivers/infiniband/ulp/iser/iscsi_iser.h
new file mode 100644
index 000000000..1d7ac24c4
--- /dev/null
+++ b/drivers/infiniband/ulp/iser/iscsi_iser.h
@@ -0,0 +1,569 @@
+/*
+ * iSER transport for the Open iSCSI Initiator & iSER transport internals
+ *
+ * Copyright (C) 2004 Dmitry Yusupov
+ * Copyright (C) 2004 Alex Aizman
+ * Copyright (C) 2005 Mike Christie
+ * based on code maintained by open-iscsi@googlegroups.com
+ *
+ * Copyright (c) 2004, 2005, 2006 Voltaire, Inc. All rights reserved.
+ * Copyright (c) 2005, 2006 Cisco Systems.  All rights reserved.
+ * Copyright (c) 2013-2014 Mellanox Technologies. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *	- Redistributions of source code must retain the above
+ *	  copyright notice, this list of conditions and the following
+ *	  disclaimer.
+ *
+ *	- Redistributions in binary form must reproduce the above
+ *	  copyright notice, this list of conditions and the following
+ *	  disclaimer in the documentation and/or other materials
+ *	  provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+#ifndef __ISCSI_ISER_H__
+#define __ISCSI_ISER_H__
+
+#include <linux/types.h>
+#include <linux/net.h>
+#include <linux/printk.h>
+#include <scsi/libiscsi.h>
+#include <scsi/scsi_transport_iscsi.h>
+#include <scsi/scsi_cmnd.h>
+#include <scsi/scsi_device.h>
+#include <scsi/iser.h>
+
+#include <linux/interrupt.h>
+#include <linux/wait.h>
+#include <linux/sched.h>
+#include <linux/list.h>
+#include <linux/slab.h>
+#include <linux/dma-mapping.h>
+#include <linux/mutex.h>
+#include <linux/mempool.h>
+#include <linux/uio.h>
+
+#include <linux/socket.h>
+#include <linux/in.h>
+#include <linux/in6.h>
+
+#include <rdma/ib_verbs.h>
+#include <rdma/rdma_cm.h>
+
+#define DRV_NAME	"iser"
+#define PFX		DRV_NAME ": "
+#define DRV_VER		"1.6"
+
+#define iser_dbg(fmt, arg...)				 \
+	do {						 \
+		if (unlikely(iser_debug_level > 2))	 \
+			printk(KERN_DEBUG PFX "%s: " fmt,\
+				__func__ , ## arg);	 \
+	} while (0)
+
+#define iser_warn(fmt, arg...)				\
+	do {						\
+		if (unlikely(iser_debug_level > 0))	\
+			pr_warn(PFX "%s: " fmt,		\
+				__func__ , ## arg);	\
+	} while (0)
+
+#define iser_info(fmt, arg...)				\
+	do {						\
+		if (unlikely(iser_debug_level > 1))	\
+			pr_info(PFX "%s: " fmt,		\
+				__func__ , ## arg);	\
+	} while (0)
+
+#define iser_err(fmt, arg...) \
+	pr_err(PFX "%s: " fmt, __func__ , ## arg)
+
+/* Default support is 512KB I/O size */
+#define ISER_DEF_MAX_SECTORS		1024
+#define ISCSI_ISER_DEF_SG_TABLESIZE                                            \
+	((ISER_DEF_MAX_SECTORS * SECTOR_SIZE) >> ilog2(SZ_4K))
+/* Maximum support is 16MB I/O size */
+#define ISCSI_ISER_MAX_SG_TABLESIZE ((32768 * SECTOR_SIZE) >> ilog2(SZ_4K))
+
+#define ISER_DEF_XMIT_CMDS_DEFAULT		512
+#if ISCSI_DEF_XMIT_CMDS_MAX > ISER_DEF_XMIT_CMDS_DEFAULT
+	#define ISER_DEF_XMIT_CMDS_MAX		ISCSI_DEF_XMIT_CMDS_MAX
+#else
+	#define ISER_DEF_XMIT_CMDS_MAX		ISER_DEF_XMIT_CMDS_DEFAULT
+#endif
+#define ISER_DEF_CMD_PER_LUN		ISER_DEF_XMIT_CMDS_MAX
+
+/* QP settings */
+/* Maximal bounds on received asynchronous PDUs */
+#define ISER_MAX_RX_MISC_PDUS		4 /* NOOP_IN(2) , ASYNC_EVENT(2)   */
+
+#define ISER_MAX_TX_MISC_PDUS		6 /* NOOP_OUT(2), TEXT(1),         *
+					   * SCSI_TMFUNC(2), LOGOUT(1) */
+
+#define ISER_QP_MAX_RECV_DTOS		(ISER_DEF_XMIT_CMDS_MAX)
+
+/* the max TX (send) WR supported by the iSER QP is defined by                 *
+ * max_send_wr = T * (1 + D) + C ; D is how many inflight dataouts we expect   *
+ * to have at max for SCSI command. The tx posting & completion handling code  *
+ * supports -EAGAIN scheme where tx is suspended till the QP has room for more *
+ * send WR. D=8 comes from 64K/8K                                              */
+
+#define ISER_INFLIGHT_DATAOUTS		8
+
+#define ISER_QP_MAX_REQ_DTOS		(ISER_DEF_XMIT_CMDS_MAX *    \
+					(1 + ISER_INFLIGHT_DATAOUTS) + \
+					ISER_MAX_TX_MISC_PDUS        + \
+					ISER_MAX_RX_MISC_PDUS)
+
+/* Max registration work requests per command */
+#define ISER_MAX_REG_WR_PER_CMD		5
+
+/* For Signature we don't support DATAOUTs so no need to make room for them */
+#define ISER_QP_SIG_MAX_REQ_DTOS	(ISER_DEF_XMIT_CMDS_MAX	*       \
+					(1 + ISER_MAX_REG_WR_PER_CMD) + \
+					ISER_MAX_TX_MISC_PDUS         + \
+					ISER_MAX_RX_MISC_PDUS)
+
+#define ISER_GET_MAX_XMIT_CMDS(send_wr) ((send_wr			\
+					 - ISER_MAX_TX_MISC_PDUS	\
+					 - ISER_MAX_RX_MISC_PDUS) /	\
+					 (1 + ISER_INFLIGHT_DATAOUTS))
+
+/* Constant PDU lengths calculations */
+#define ISER_HEADERS_LEN	(sizeof(struct iser_ctrl) + sizeof(struct iscsi_hdr))
+
+#define ISER_RECV_DATA_SEG_LEN	128
+#define ISER_RX_PAYLOAD_SIZE	(ISER_HEADERS_LEN + ISER_RECV_DATA_SEG_LEN)
+#define ISER_RX_LOGIN_SIZE	(ISER_HEADERS_LEN + ISCSI_DEF_MAX_RECV_SEG_LEN)
+
+/* Length of an object name string */
+#define ISER_OBJECT_NAME_SIZE		    64
+
+enum iser_conn_state {
+	ISER_CONN_INIT,		   /* descriptor allocd, no conn          */
+	ISER_CONN_PENDING,	   /* in the process of being established */
+	ISER_CONN_UP,		   /* up and running                      */
+	ISER_CONN_TERMINATING,	   /* in the process of being terminated  */
+	ISER_CONN_DOWN,		   /* shut down                           */
+	ISER_CONN_STATES_NUM
+};
+
+enum iser_task_status {
+	ISER_TASK_STATUS_INIT = 0,
+	ISER_TASK_STATUS_STARTED,
+	ISER_TASK_STATUS_COMPLETED
+};
+
+enum iser_data_dir {
+	ISER_DIR_IN = 0,	   /* to initiator */
+	ISER_DIR_OUT,		   /* from initiator */
+	ISER_DIRS_NUM
+};
+
+/**
+ * struct iser_data_buf - iSER data buffer
+ *
+ * @sg:           pointer to the sg list
+ * @size:         num entries of this sg
+ * @data_len:     total buffer byte len
+ * @dma_nents:    returned by dma_map_sg
+ */
+struct iser_data_buf {
+	struct scatterlist *sg;
+	int                size;
+	unsigned long      data_len;
+	int                dma_nents;
+};
+
+/* fwd declarations */
+struct iser_device;
+struct iscsi_iser_task;
+struct iscsi_endpoint;
+struct iser_reg_resources;
+
+/**
+ * struct iser_mem_reg - iSER memory registration info
+ *
+ * @sge:          memory region sg element
+ * @rkey:         memory region remote key
+ * @desc:         pointer to fast registration context
+ */
+struct iser_mem_reg {
+	struct ib_sge sge;
+	u32 rkey;
+	struct iser_fr_desc *desc;
+};
+
+enum iser_desc_type {
+	ISCSI_TX_CONTROL ,
+	ISCSI_TX_SCSI_COMMAND,
+	ISCSI_TX_DATAOUT
+};
+
+/**
+ * struct iser_tx_desc - iSER TX descriptor
+ *
+ * @iser_header:   iser header
+ * @iscsi_header:  iscsi header
+ * @type:          command/control/dataout
+ * @dma_addr:      header buffer dma_address
+ * @tx_sg:         sg[0] points to iser/iscsi headers
+ *                 sg[1] optionally points to either of immediate data
+ *                 unsolicited data-out or control
+ * @num_sge:       number sges used on this TX task
+ * @cqe:           completion handler
+ * @mapped:        Is the task header mapped
+ * @reg_wr:        registration WR
+ * @send_wr:       send WR
+ * @inv_wr:        invalidate WR
+ */
+struct iser_tx_desc {
+	struct iser_ctrl             iser_header;
+	struct iscsi_hdr             iscsi_header;
+	enum   iser_desc_type        type;
+	u64		             dma_addr;
+	struct ib_sge		     tx_sg[2];
+	int                          num_sge;
+	struct ib_cqe		     cqe;
+	bool			     mapped;
+	struct ib_reg_wr	     reg_wr;
+	struct ib_send_wr	     send_wr;
+	struct ib_send_wr	     inv_wr;
+};
+
+#define ISER_RX_PAD_SIZE	(256 - (ISER_RX_PAYLOAD_SIZE + \
+				 sizeof(u64) + sizeof(struct ib_sge) + \
+				 sizeof(struct ib_cqe)))
+/**
+ * struct iser_rx_desc - iSER RX descriptor
+ *
+ * @iser_header:   iser header
+ * @iscsi_header:  iscsi header
+ * @data:          received data segment
+ * @dma_addr:      receive buffer dma address
+ * @rx_sg:         ib_sge of receive buffer
+ * @cqe:           completion handler
+ * @pad:           for sense data TODO: Modify to maximum sense length supported
+ */
+struct iser_rx_desc {
+	struct iser_ctrl             iser_header;
+	struct iscsi_hdr             iscsi_header;
+	char		             data[ISER_RECV_DATA_SEG_LEN];
+	u64		             dma_addr;
+	struct ib_sge		     rx_sg;
+	struct ib_cqe		     cqe;
+	char		             pad[ISER_RX_PAD_SIZE];
+} __packed;
+
+/**
+ * struct iser_login_desc - iSER login descriptor
+ *
+ * @req:           pointer to login request buffer
+ * @rsp:           pointer to login response buffer
+ * @req_dma:       DMA address of login request buffer
+ * @rsp_dma:       DMA address of login response buffer
+ * @sge:           IB sge for login post recv
+ * @cqe:           completion handler
+ */
+struct iser_login_desc {
+	void                         *req;
+	void                         *rsp;
+	u64                          req_dma;
+	u64                          rsp_dma;
+	struct ib_sge                sge;
+	struct ib_cqe		     cqe;
+} __packed;
+
+struct iser_conn;
+struct ib_conn;
+
+/**
+ * struct iser_device - iSER device handle
+ *
+ * @ib_device:     RDMA device
+ * @pd:            Protection Domain for this device
+ * @event_handler: IB events handle routine
+ * @ig_list:	   entry in devices list
+ * @refcount:      Reference counter, dominated by open iser connections
+ */
+struct iser_device {
+	struct ib_device             *ib_device;
+	struct ib_pd	             *pd;
+	struct ib_event_handler      event_handler;
+	struct list_head             ig_list;
+	int                          refcount;
+};
+
+/**
+ * struct iser_reg_resources - Fast registration resources
+ *
+ * @mr:         memory region
+ * @sig_mr:     signature memory region
+ */
+struct iser_reg_resources {
+	struct ib_mr                     *mr;
+	struct ib_mr                     *sig_mr;
+};
+
+/**
+ * struct iser_fr_desc - Fast registration descriptor
+ *
+ * @list:           entry in connection fastreg pool
+ * @rsc:            data buffer registration resources
+ * @sig_protected:  is region protected indicator
+ * @all_list:       first and last list members
+ */
+struct iser_fr_desc {
+	struct list_head		  list;
+	struct iser_reg_resources	  rsc;
+	bool				  sig_protected;
+	struct list_head                  all_list;
+};
+
+/**
+ * struct iser_fr_pool - connection fast registration pool
+ *
+ * @list:                list of fastreg descriptors
+ * @lock:                protects fastreg pool
+ * @size:                size of the pool
+ * @all_list:            first and last list members
+ */
+struct iser_fr_pool {
+	struct list_head        list;
+	spinlock_t              lock;
+	int                     size;
+	struct list_head        all_list;
+};
+
+/**
+ * struct ib_conn - Infiniband related objects
+ *
+ * @cma_id:              rdma_cm connection maneger handle
+ * @qp:                  Connection Queue-pair
+ * @cq:                  Connection completion queue
+ * @cq_size:             The number of max outstanding completions
+ * @device:              reference to iser device
+ * @fr_pool:             connection fast registration pool
+ * @pi_support:          Indicate device T10-PI support
+ * @reg_cqe:             completion handler
+ */
+struct ib_conn {
+	struct rdma_cm_id           *cma_id;
+	struct ib_qp	            *qp;
+	struct ib_cq		    *cq;
+	u32			    cq_size;
+	struct iser_device          *device;
+	struct iser_fr_pool          fr_pool;
+	bool			     pi_support;
+	struct ib_cqe		     reg_cqe;
+};
+
+/**
+ * struct iser_conn - iSER connection context
+ *
+ * @ib_conn:          connection RDMA resources
+ * @iscsi_conn:       link to matching iscsi connection
+ * @ep:               transport handle
+ * @state:            connection logical state
+ * @qp_max_recv_dtos: maximum number of data outs, corresponds
+ *                    to max number of post recvs
+ * @max_cmds:         maximum cmds allowed for this connection
+ * @name:             connection peer portal
+ * @release_work:     deferred work for release job
+ * @state_mutex:      protects iser onnection state
+ * @stop_completion:  conn_stop completion
+ * @ib_completion:    RDMA cleanup completion
+ * @up_completion:    connection establishment completed
+ *                    (state is ISER_CONN_UP)
+ * @conn_list:        entry in ig conn list
+ * @login_desc:       login descriptor
+ * @rx_descs:         rx buffers array (cyclic buffer)
+ * @num_rx_descs:     number of rx descriptors
+ * @scsi_sg_tablesize: scsi host sg_tablesize
+ * @pages_per_mr:     maximum pages available for registration
+ * @snd_w_inv:        connection uses remote invalidation
+ */
+struct iser_conn {
+	struct ib_conn		     ib_conn;
+	struct iscsi_conn	     *iscsi_conn;
+	struct iscsi_endpoint	     *ep;
+	enum iser_conn_state	     state;
+	unsigned		     qp_max_recv_dtos;
+	u16                          max_cmds;
+	char 			     name[ISER_OBJECT_NAME_SIZE];
+	struct work_struct	     release_work;
+	struct mutex		     state_mutex;
+	struct completion	     stop_completion;
+	struct completion	     ib_completion;
+	struct completion	     up_completion;
+	struct list_head	     conn_list;
+	struct iser_login_desc       login_desc;
+	struct iser_rx_desc	     *rx_descs;
+	u32                          num_rx_descs;
+	unsigned short               scsi_sg_tablesize;
+	unsigned short               pages_per_mr;
+	bool			     snd_w_inv;
+};
+
+/**
+ * struct iscsi_iser_task - iser task context
+ *
+ * @desc:     TX descriptor
+ * @iser_conn:        link to iser connection
+ * @status:           current task status
+ * @sc:               link to scsi command
+ * @command_sent:     indicate if command was sent
+ * @dir:              iser data direction
+ * @rdma_reg:         task rdma registration desc
+ * @data:             iser data buffer desc
+ * @prot:             iser protection buffer desc
+ */
+struct iscsi_iser_task {
+	struct iser_tx_desc          desc;
+	struct iser_conn	     *iser_conn;
+	enum iser_task_status 	     status;
+	struct scsi_cmnd	     *sc;
+	int                          command_sent;
+	int                          dir[ISER_DIRS_NUM];
+	struct iser_mem_reg          rdma_reg[ISER_DIRS_NUM];
+	struct iser_data_buf         data[ISER_DIRS_NUM];
+	struct iser_data_buf         prot[ISER_DIRS_NUM];
+};
+
+/**
+ * struct iser_global - iSER global context
+ *
+ * @device_list_mutex:    protects device_list
+ * @device_list:          iser devices global list
+ * @connlist_mutex:       protects connlist
+ * @connlist:             iser connections global list
+ * @desc_cache:           kmem cache for tx dataout
+ */
+struct iser_global {
+	struct mutex      device_list_mutex;
+	struct list_head  device_list;
+	struct mutex      connlist_mutex;
+	struct list_head  connlist;
+	struct kmem_cache *desc_cache;
+};
+
+extern struct iser_global ig;
+extern int iser_debug_level;
+extern bool iser_pi_enable;
+extern unsigned int iser_max_sectors;
+extern bool iser_always_reg;
+
+int iser_send_control(struct iscsi_conn *conn,
+		      struct iscsi_task *task);
+
+int iser_send_command(struct iscsi_conn *conn,
+		      struct iscsi_task *task);
+
+int iser_send_data_out(struct iscsi_conn *conn,
+		       struct iscsi_task *task,
+		       struct iscsi_data *hdr);
+
+void iscsi_iser_recv(struct iscsi_conn *conn,
+		     struct iscsi_hdr *hdr,
+		     char *rx_data,
+		     int rx_data_len);
+
+void iser_conn_init(struct iser_conn *iser_conn);
+
+void iser_conn_release(struct iser_conn *iser_conn);
+
+int iser_conn_terminate(struct iser_conn *iser_conn);
+
+void iser_release_work(struct work_struct *work);
+
+void iser_err_comp(struct ib_wc *wc, const char *type);
+void iser_login_rsp(struct ib_cq *cq, struct ib_wc *wc);
+void iser_task_rsp(struct ib_cq *cq, struct ib_wc *wc);
+void iser_cmd_comp(struct ib_cq *cq, struct ib_wc *wc);
+void iser_ctrl_comp(struct ib_cq *cq, struct ib_wc *wc);
+void iser_dataout_comp(struct ib_cq *cq, struct ib_wc *wc);
+void iser_reg_comp(struct ib_cq *cq, struct ib_wc *wc);
+
+void iser_task_rdma_init(struct iscsi_iser_task *task);
+
+void iser_task_rdma_finalize(struct iscsi_iser_task *task);
+
+void iser_free_rx_descriptors(struct iser_conn *iser_conn);
+
+int iser_reg_mem_fastreg(struct iscsi_iser_task *task,
+			 enum iser_data_dir dir,
+			 bool all_imm);
+void iser_unreg_mem_fastreg(struct iscsi_iser_task *task,
+			    enum iser_data_dir dir);
+
+int  iser_connect(struct iser_conn *iser_conn,
+		  struct sockaddr *src_addr,
+		  struct sockaddr *dst_addr,
+		  int non_blocking);
+
+int  iser_post_recvl(struct iser_conn *iser_conn);
+int  iser_post_recvm(struct iser_conn *iser_conn,
+		     struct iser_rx_desc *rx_desc);
+int  iser_post_send(struct ib_conn *ib_conn, struct iser_tx_desc *tx_desc);
+
+int iser_dma_map_task_data(struct iscsi_iser_task *iser_task,
+			   enum iser_data_dir iser_dir,
+			   enum dma_data_direction dma_dir);
+
+void iser_dma_unmap_task_data(struct iscsi_iser_task *iser_task,
+			      enum iser_data_dir iser_dir,
+			      enum dma_data_direction dma_dir);
+
+int  iser_initialize_task_headers(struct iscsi_task *task,
+			struct iser_tx_desc *tx_desc);
+int iser_alloc_rx_descriptors(struct iser_conn *iser_conn,
+			      struct iscsi_session *session);
+int iser_alloc_fastreg_pool(struct ib_conn *ib_conn,
+			    unsigned cmds_max,
+			    unsigned int size);
+void iser_free_fastreg_pool(struct ib_conn *ib_conn);
+u8 iser_check_task_pi_status(struct iscsi_iser_task *iser_task,
+			     enum iser_data_dir cmd_dir, sector_t *sector);
+
+static inline struct iser_conn *
+to_iser_conn(struct ib_conn *ib_conn)
+{
+	return container_of(ib_conn, struct iser_conn, ib_conn);
+}
+
+static inline struct iser_rx_desc *
+iser_rx(struct ib_cqe *cqe)
+{
+	return container_of(cqe, struct iser_rx_desc, cqe);
+}
+
+static inline struct iser_tx_desc *
+iser_tx(struct ib_cqe *cqe)
+{
+	return container_of(cqe, struct iser_tx_desc, cqe);
+}
+
+static inline struct iser_login_desc *
+iser_login(struct ib_cqe *cqe)
+{
+	return container_of(cqe, struct iser_login_desc, cqe);
+}
+
+#endif
diff --git a/drivers/infiniband/ulp/iser/iser_initiator.c b/drivers/infiniband/ulp/iser/iser_initiator.c
new file mode 100644
index 000000000..12a2d12fe
--- /dev/null
+++ b/drivers/infiniband/ulp/iser/iser_initiator.c
@@ -0,0 +1,746 @@
+/*
+ * Copyright (c) 2004, 2005, 2006 Voltaire, Inc. All rights reserved.
+ * Copyright (c) 2013-2014 Mellanox Technologies. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *	- Redistributions of source code must retain the above
+ *	  copyright notice, this list of conditions and the following
+ *	  disclaimer.
+ *
+ *	- Redistributions in binary form must reproduce the above
+ *	  copyright notice, this list of conditions and the following
+ *	  disclaimer in the documentation and/or other materials
+ *	  provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+#include <linux/kernel.h>
+#include <linux/slab.h>
+#include <linux/mm.h>
+#include <linux/scatterlist.h>
+#include <linux/kfifo.h>
+#include <scsi/scsi_cmnd.h>
+#include <scsi/scsi_host.h>
+
+#include "iscsi_iser.h"
+
+/* Register user buffer memory and initialize passive rdma
+ *  dto descriptor. Data size is stored in
+ *  task->data[ISER_DIR_IN].data_len, Protection size
+ *  os stored in task->prot[ISER_DIR_IN].data_len
+ */
+static int iser_prepare_read_cmd(struct iscsi_task *task)
+
+{
+	struct iscsi_iser_task *iser_task = task->dd_data;
+	struct iser_mem_reg *mem_reg;
+	int err;
+	struct iser_ctrl *hdr = &iser_task->desc.iser_header;
+
+	err = iser_dma_map_task_data(iser_task,
+				     ISER_DIR_IN,
+				     DMA_FROM_DEVICE);
+	if (err)
+		return err;
+
+	err = iser_reg_mem_fastreg(iser_task, ISER_DIR_IN, false);
+	if (err) {
+		iser_err("Failed to set up Data-IN RDMA\n");
+		goto out_err;
+	}
+	mem_reg = &iser_task->rdma_reg[ISER_DIR_IN];
+
+	hdr->flags    |= ISER_RSV;
+	hdr->read_stag = cpu_to_be32(mem_reg->rkey);
+	hdr->read_va   = cpu_to_be64(mem_reg->sge.addr);
+
+	iser_dbg("Cmd itt:%d READ tags RKEY:%#.4X VA:%#llX\n",
+		 task->itt, mem_reg->rkey,
+		 (unsigned long long)mem_reg->sge.addr);
+
+	return 0;
+
+out_err:
+	iser_dma_unmap_task_data(iser_task, ISER_DIR_IN, DMA_FROM_DEVICE);
+	return err;
+}
+
+/* Register user buffer memory and initialize passive rdma
+ *  dto descriptor. Data size is stored in
+ *  task->data[ISER_DIR_OUT].data_len, Protection size
+ *  is stored at task->prot[ISER_DIR_OUT].data_len
+ */
+static int iser_prepare_write_cmd(struct iscsi_task *task, unsigned int imm_sz,
+				  unsigned int unsol_sz, unsigned int edtl)
+{
+	struct iscsi_iser_task *iser_task = task->dd_data;
+	struct iser_mem_reg *mem_reg;
+	int err;
+	struct iser_ctrl *hdr = &iser_task->desc.iser_header;
+	struct iser_data_buf *buf_out = &iser_task->data[ISER_DIR_OUT];
+	struct ib_sge *tx_dsg = &iser_task->desc.tx_sg[1];
+
+	err = iser_dma_map_task_data(iser_task,
+				     ISER_DIR_OUT,
+				     DMA_TO_DEVICE);
+	if (err)
+		return err;
+
+	err = iser_reg_mem_fastreg(iser_task, ISER_DIR_OUT,
+				   buf_out->data_len == imm_sz);
+	if (err) {
+		iser_err("Failed to register write cmd RDMA mem\n");
+		goto out_err;
+	}
+
+	mem_reg = &iser_task->rdma_reg[ISER_DIR_OUT];
+
+	if (unsol_sz < edtl) {
+		hdr->flags     |= ISER_WSV;
+		if (buf_out->data_len > imm_sz) {
+			hdr->write_stag = cpu_to_be32(mem_reg->rkey);
+			hdr->write_va = cpu_to_be64(mem_reg->sge.addr + unsol_sz);
+		}
+
+		iser_dbg("Cmd itt:%d, WRITE tags, RKEY:%#.4X VA:%#llX + unsol:%d\n",
+			 task->itt, mem_reg->rkey,
+			 (unsigned long long)mem_reg->sge.addr, unsol_sz);
+	}
+
+	if (imm_sz > 0) {
+		iser_dbg("Cmd itt:%d, WRITE, adding imm.data sz: %d\n",
+			 task->itt, imm_sz);
+		tx_dsg->addr = mem_reg->sge.addr;
+		tx_dsg->length = imm_sz;
+		tx_dsg->lkey = mem_reg->sge.lkey;
+		iser_task->desc.num_sge = 2;
+	}
+
+	return 0;
+
+out_err:
+	iser_dma_unmap_task_data(iser_task, ISER_DIR_OUT, DMA_TO_DEVICE);
+	return err;
+}
+
+/* creates a new tx descriptor and adds header regd buffer */
+static void iser_create_send_desc(struct iser_conn *iser_conn,
+		struct iser_tx_desc *tx_desc, enum iser_desc_type type,
+		void (*done)(struct ib_cq *cq, struct ib_wc *wc))
+{
+	struct iser_device *device = iser_conn->ib_conn.device;
+
+	tx_desc->type = type;
+	tx_desc->cqe.done = done;
+
+	ib_dma_sync_single_for_cpu(device->ib_device,
+		tx_desc->dma_addr, ISER_HEADERS_LEN, DMA_TO_DEVICE);
+
+	memset(&tx_desc->iser_header, 0, sizeof(struct iser_ctrl));
+	tx_desc->iser_header.flags = ISER_VER;
+	tx_desc->num_sge = 1;
+}
+
+static void iser_free_login_buf(struct iser_conn *iser_conn)
+{
+	struct iser_device *device = iser_conn->ib_conn.device;
+	struct iser_login_desc *desc = &iser_conn->login_desc;
+
+	if (!desc->req)
+		return;
+
+	ib_dma_unmap_single(device->ib_device, desc->req_dma,
+			    ISCSI_DEF_MAX_RECV_SEG_LEN, DMA_TO_DEVICE);
+
+	ib_dma_unmap_single(device->ib_device, desc->rsp_dma,
+			    ISER_RX_LOGIN_SIZE, DMA_FROM_DEVICE);
+
+	kfree(desc->req);
+	kfree(desc->rsp);
+
+	/* make sure we never redo any unmapping */
+	desc->req = NULL;
+	desc->rsp = NULL;
+}
+
+static int iser_alloc_login_buf(struct iser_conn *iser_conn)
+{
+	struct iser_device *device = iser_conn->ib_conn.device;
+	struct iser_login_desc *desc = &iser_conn->login_desc;
+
+	desc->req = kmalloc(ISCSI_DEF_MAX_RECV_SEG_LEN, GFP_KERNEL);
+	if (!desc->req)
+		return -ENOMEM;
+
+	desc->req_dma = ib_dma_map_single(device->ib_device, desc->req,
+					  ISCSI_DEF_MAX_RECV_SEG_LEN,
+					  DMA_TO_DEVICE);
+	if (ib_dma_mapping_error(device->ib_device,
+				desc->req_dma))
+		goto free_req;
+
+	desc->rsp = kmalloc(ISER_RX_LOGIN_SIZE, GFP_KERNEL);
+	if (!desc->rsp)
+		goto unmap_req;
+
+	desc->rsp_dma = ib_dma_map_single(device->ib_device, desc->rsp,
+					   ISER_RX_LOGIN_SIZE,
+					   DMA_FROM_DEVICE);
+	if (ib_dma_mapping_error(device->ib_device,
+				desc->rsp_dma))
+		goto free_rsp;
+
+	return 0;
+
+free_rsp:
+	kfree(desc->rsp);
+unmap_req:
+	ib_dma_unmap_single(device->ib_device, desc->req_dma,
+			    ISCSI_DEF_MAX_RECV_SEG_LEN,
+			    DMA_TO_DEVICE);
+free_req:
+	kfree(desc->req);
+
+	return -ENOMEM;
+}
+
+int iser_alloc_rx_descriptors(struct iser_conn *iser_conn,
+			      struct iscsi_session *session)
+{
+	int i, j;
+	u64 dma_addr;
+	struct iser_rx_desc *rx_desc;
+	struct ib_sge       *rx_sg;
+	struct ib_conn *ib_conn = &iser_conn->ib_conn;
+	struct iser_device *device = ib_conn->device;
+
+	iser_conn->qp_max_recv_dtos = session->cmds_max;
+
+	if (iser_alloc_fastreg_pool(ib_conn, session->scsi_cmds_max,
+				    iser_conn->pages_per_mr))
+		goto create_rdma_reg_res_failed;
+
+	if (iser_alloc_login_buf(iser_conn))
+		goto alloc_login_buf_fail;
+
+	iser_conn->num_rx_descs = session->cmds_max;
+	iser_conn->rx_descs = kmalloc_objs(struct iser_rx_desc,
+					   iser_conn->num_rx_descs);
+	if (!iser_conn->rx_descs)
+		goto rx_desc_alloc_fail;
+
+	rx_desc = iser_conn->rx_descs;
+
+	for (i = 0; i < iser_conn->qp_max_recv_dtos; i++, rx_desc++)  {
+		dma_addr = ib_dma_map_single(device->ib_device, (void *)rx_desc,
+					ISER_RX_PAYLOAD_SIZE, DMA_FROM_DEVICE);
+		if (ib_dma_mapping_error(device->ib_device, dma_addr))
+			goto rx_desc_dma_map_failed;
+
+		rx_desc->dma_addr = dma_addr;
+		rx_desc->cqe.done = iser_task_rsp;
+		rx_sg = &rx_desc->rx_sg;
+		rx_sg->addr = rx_desc->dma_addr;
+		rx_sg->length = ISER_RX_PAYLOAD_SIZE;
+		rx_sg->lkey = device->pd->local_dma_lkey;
+	}
+
+	return 0;
+
+rx_desc_dma_map_failed:
+	rx_desc = iser_conn->rx_descs;
+	for (j = 0; j < i; j++, rx_desc++)
+		ib_dma_unmap_single(device->ib_device, rx_desc->dma_addr,
+				    ISER_RX_PAYLOAD_SIZE, DMA_FROM_DEVICE);
+	kfree(iser_conn->rx_descs);
+	iser_conn->rx_descs = NULL;
+rx_desc_alloc_fail:
+	iser_free_login_buf(iser_conn);
+alloc_login_buf_fail:
+	iser_free_fastreg_pool(ib_conn);
+create_rdma_reg_res_failed:
+	iser_err("failed allocating rx descriptors / data buffers\n");
+	return -ENOMEM;
+}
+
+void iser_free_rx_descriptors(struct iser_conn *iser_conn)
+{
+	int i;
+	struct iser_rx_desc *rx_desc;
+	struct ib_conn *ib_conn = &iser_conn->ib_conn;
+	struct iser_device *device = ib_conn->device;
+
+	iser_free_fastreg_pool(ib_conn);
+
+	rx_desc = iser_conn->rx_descs;
+	for (i = 0; i < iser_conn->qp_max_recv_dtos; i++, rx_desc++)
+		ib_dma_unmap_single(device->ib_device, rx_desc->dma_addr,
+				    ISER_RX_PAYLOAD_SIZE, DMA_FROM_DEVICE);
+	kfree(iser_conn->rx_descs);
+	/* make sure we never redo any unmapping */
+	iser_conn->rx_descs = NULL;
+
+	iser_free_login_buf(iser_conn);
+}
+
+static int iser_post_rx_bufs(struct iscsi_conn *conn, struct iscsi_hdr *req)
+{
+	struct iser_conn *iser_conn = conn->dd_data;
+	struct iscsi_session *session = conn->session;
+	int err = 0;
+	int i;
+
+	iser_dbg("req op %x flags %x\n", req->opcode, req->flags);
+	/* check if this is the last login - going to full feature phase */
+	if ((req->flags & ISCSI_FULL_FEATURE_PHASE) != ISCSI_FULL_FEATURE_PHASE)
+		goto out;
+
+	if (session->discovery_sess) {
+		iser_info("Discovery session, re-using login RX buffer\n");
+		goto out;
+	}
+
+	iser_info("Normal session, posting batch of RX %d buffers\n",
+		  iser_conn->qp_max_recv_dtos - 1);
+
+	/*
+	 * Initial post receive buffers.
+	 * There is one already posted recv buffer (for the last login
+	 * response). Therefore, the first recv buffer is skipped here.
+	 */
+	for (i = 1; i < iser_conn->qp_max_recv_dtos; i++) {
+		err = iser_post_recvm(iser_conn, &iser_conn->rx_descs[i]);
+		if (err)
+			goto out;
+	}
+out:
+	return err;
+}
+
+/**
+ * iser_send_command - send command PDU
+ * @conn: link to matching iscsi connection
+ * @task: SCSI command task
+ */
+int iser_send_command(struct iscsi_conn *conn, struct iscsi_task *task)
+{
+	struct iser_conn *iser_conn = conn->dd_data;
+	struct iscsi_iser_task *iser_task = task->dd_data;
+	unsigned long edtl;
+	int err;
+	struct iser_data_buf *data_buf, *prot_buf;
+	struct iscsi_scsi_req *hdr = (struct iscsi_scsi_req *)task->hdr;
+	struct scsi_cmnd *sc  =  task->sc;
+	struct iser_tx_desc *tx_desc = &iser_task->desc;
+
+	edtl = ntohl(hdr->data_length);
+
+	/* build the tx desc regd header and add it to the tx desc dto */
+	iser_create_send_desc(iser_conn, tx_desc, ISCSI_TX_SCSI_COMMAND,
+			      iser_cmd_comp);
+
+	if (hdr->flags & ISCSI_FLAG_CMD_READ) {
+		data_buf = &iser_task->data[ISER_DIR_IN];
+		prot_buf = &iser_task->prot[ISER_DIR_IN];
+	} else {
+		data_buf = &iser_task->data[ISER_DIR_OUT];
+		prot_buf = &iser_task->prot[ISER_DIR_OUT];
+	}
+
+	if (scsi_sg_count(sc)) { /* using a scatter list */
+		data_buf->sg = scsi_sglist(sc);
+		data_buf->size = scsi_sg_count(sc);
+	}
+	data_buf->data_len = scsi_bufflen(sc);
+
+	if (scsi_prot_sg_count(sc)) {
+		prot_buf->sg  = scsi_prot_sglist(sc);
+		prot_buf->size = scsi_prot_sg_count(sc);
+		prot_buf->data_len = (data_buf->data_len >>
+				     ilog2(sc->device->sector_size)) * 8;
+	}
+
+	if (hdr->flags & ISCSI_FLAG_CMD_READ) {
+		err = iser_prepare_read_cmd(task);
+		if (err)
+			goto send_command_error;
+	}
+	if (hdr->flags & ISCSI_FLAG_CMD_WRITE) {
+		err = iser_prepare_write_cmd(task,
+					     task->imm_count,
+				             task->imm_count +
+					     task->unsol_r2t.data_length,
+					     edtl);
+		if (err)
+			goto send_command_error;
+	}
+
+	iser_task->status = ISER_TASK_STATUS_STARTED;
+
+	err = iser_post_send(&iser_conn->ib_conn, tx_desc);
+	if (!err)
+		return 0;
+
+send_command_error:
+	iser_err("conn %p failed task->itt %d err %d\n",conn, task->itt, err);
+	return err;
+}
+
+/**
+ * iser_send_data_out - send data out PDU
+ * @conn: link to matching iscsi connection
+ * @task: SCSI command task
+ * @hdr: pointer to the LLD's iSCSI message header
+ */
+int iser_send_data_out(struct iscsi_conn *conn, struct iscsi_task *task,
+		       struct iscsi_data *hdr)
+{
+	struct iser_conn *iser_conn = conn->dd_data;
+	struct iscsi_iser_task *iser_task = task->dd_data;
+	struct iser_tx_desc *tx_desc;
+	struct iser_mem_reg *mem_reg;
+	unsigned long buf_offset;
+	unsigned long data_seg_len;
+	uint32_t itt;
+	int err;
+	struct ib_sge *tx_dsg;
+
+	itt = (__force uint32_t)hdr->itt;
+	data_seg_len = ntoh24(hdr->dlength);
+	buf_offset   = ntohl(hdr->offset);
+
+	iser_dbg("%s itt %d dseg_len %d offset %d\n",
+		 __func__,(int)itt,(int)data_seg_len,(int)buf_offset);
+
+	tx_desc = kmem_cache_zalloc(ig.desc_cache, GFP_ATOMIC);
+	if (!tx_desc)
+		return -ENOMEM;
+
+	tx_desc->type = ISCSI_TX_DATAOUT;
+	tx_desc->cqe.done = iser_dataout_comp;
+	tx_desc->iser_header.flags = ISER_VER;
+	memcpy(&tx_desc->iscsi_header, hdr, sizeof(struct iscsi_hdr));
+
+	/* build the tx desc */
+	err = iser_initialize_task_headers(task, tx_desc);
+	if (err)
+		goto send_data_out_error;
+
+	mem_reg = &iser_task->rdma_reg[ISER_DIR_OUT];
+	tx_dsg = &tx_desc->tx_sg[1];
+	tx_dsg->addr = mem_reg->sge.addr + buf_offset;
+	tx_dsg->length = data_seg_len;
+	tx_dsg->lkey = mem_reg->sge.lkey;
+	tx_desc->num_sge = 2;
+
+	if (buf_offset + data_seg_len > iser_task->data[ISER_DIR_OUT].data_len) {
+		iser_err("Offset:%ld & DSL:%ld in Data-Out inconsistent with total len:%ld, itt:%d\n",
+			 buf_offset, data_seg_len,
+			 iser_task->data[ISER_DIR_OUT].data_len, itt);
+		err = -EINVAL;
+		goto send_data_out_error;
+	}
+	iser_dbg("data-out itt: %d, offset: %ld, sz: %ld\n",
+		 itt, buf_offset, data_seg_len);
+
+	err = iser_post_send(&iser_conn->ib_conn, tx_desc);
+	if (!err)
+		return 0;
+
+send_data_out_error:
+	kmem_cache_free(ig.desc_cache, tx_desc);
+	iser_err("conn %p failed err %d\n", conn, err);
+	return err;
+}
+
+int iser_send_control(struct iscsi_conn *conn, struct iscsi_task *task)
+{
+	struct iser_conn *iser_conn = conn->dd_data;
+	struct iscsi_iser_task *iser_task = task->dd_data;
+	struct iser_tx_desc *mdesc = &iser_task->desc;
+	unsigned long data_seg_len;
+	int err = 0;
+	struct iser_device *device;
+
+	/* build the tx desc regd header and add it to the tx desc dto */
+	iser_create_send_desc(iser_conn, mdesc, ISCSI_TX_CONTROL,
+			      iser_ctrl_comp);
+
+	device = iser_conn->ib_conn.device;
+
+	data_seg_len = ntoh24(task->hdr->dlength);
+
+	if (data_seg_len > 0) {
+		struct iser_login_desc *desc = &iser_conn->login_desc;
+		struct ib_sge *tx_dsg = &mdesc->tx_sg[1];
+
+		if (task != conn->login_task) {
+			iser_err("data present on non login task!!!\n");
+			goto send_control_error;
+		}
+
+		ib_dma_sync_single_for_cpu(device->ib_device, desc->req_dma,
+					   task->data_count, DMA_TO_DEVICE);
+
+		memcpy(desc->req, task->data, task->data_count);
+
+		ib_dma_sync_single_for_device(device->ib_device, desc->req_dma,
+					      task->data_count, DMA_TO_DEVICE);
+
+		tx_dsg->addr = desc->req_dma;
+		tx_dsg->length = task->data_count;
+		tx_dsg->lkey = device->pd->local_dma_lkey;
+		mdesc->num_sge = 2;
+	}
+
+	if (task == conn->login_task) {
+		iser_dbg("op %x dsl %lx, posting login rx buffer\n",
+			 task->hdr->opcode, data_seg_len);
+		err = iser_post_recvl(iser_conn);
+		if (err)
+			goto send_control_error;
+		err = iser_post_rx_bufs(conn, task->hdr);
+		if (err)
+			goto send_control_error;
+	}
+
+	err = iser_post_send(&iser_conn->ib_conn, mdesc);
+	if (!err)
+		return 0;
+
+send_control_error:
+	iser_err("conn %p failed err %d\n",conn, err);
+	return err;
+}
+
+void iser_login_rsp(struct ib_cq *cq, struct ib_wc *wc)
+{
+	struct ib_conn *ib_conn = wc->qp->qp_context;
+	struct iser_conn *iser_conn = to_iser_conn(ib_conn);
+	struct iser_login_desc *desc = iser_login(wc->wr_cqe);
+	struct iscsi_hdr *hdr;
+	char *data;
+	int length;
+	bool full_feature_phase;
+
+	if (unlikely(wc->status != IB_WC_SUCCESS)) {
+		iser_err_comp(wc, "login_rsp");
+		return;
+	}
+
+	ib_dma_sync_single_for_cpu(ib_conn->device->ib_device,
+				   desc->rsp_dma, ISER_RX_LOGIN_SIZE,
+				   DMA_FROM_DEVICE);
+
+	hdr = desc->rsp + sizeof(struct iser_ctrl);
+	data = desc->rsp + ISER_HEADERS_LEN;
+	length = wc->byte_len - ISER_HEADERS_LEN;
+	full_feature_phase = ((hdr->flags & ISCSI_FULL_FEATURE_PHASE) ==
+			      ISCSI_FULL_FEATURE_PHASE) &&
+			     (hdr->flags & ISCSI_FLAG_CMD_FINAL);
+
+	iser_dbg("op 0x%x itt 0x%x dlen %d\n", hdr->opcode,
+		 hdr->itt, length);
+
+	iscsi_iser_recv(iser_conn->iscsi_conn, hdr, data, length);
+
+	ib_dma_sync_single_for_device(ib_conn->device->ib_device,
+				      desc->rsp_dma, ISER_RX_LOGIN_SIZE,
+				      DMA_FROM_DEVICE);
+
+	if (!full_feature_phase ||
+	    iser_conn->iscsi_conn->session->discovery_sess)
+		return;
+
+	/* Post the first RX buffer that is skipped in iser_post_rx_bufs() */
+	iser_post_recvm(iser_conn, iser_conn->rx_descs);
+}
+
+static inline int iser_inv_desc(struct iser_fr_desc *desc, u32 rkey)
+{
+	if (unlikely((!desc->sig_protected && rkey != desc->rsc.mr->rkey) ||
+		     (desc->sig_protected && rkey != desc->rsc.sig_mr->rkey))) {
+		iser_err("Bogus remote invalidation for rkey %#x\n", rkey);
+		return -EINVAL;
+	}
+
+	if (desc->sig_protected)
+		desc->rsc.sig_mr->need_inval = false;
+	else
+		desc->rsc.mr->need_inval = false;
+
+	return 0;
+}
+
+static int iser_check_remote_inv(struct iser_conn *iser_conn, struct ib_wc *wc,
+				 struct iscsi_hdr *hdr)
+{
+	if (wc->wc_flags & IB_WC_WITH_INVALIDATE) {
+		struct iscsi_task *task;
+		u32 rkey = wc->ex.invalidate_rkey;
+
+		iser_dbg("conn %p: remote invalidation for rkey %#x\n",
+			 iser_conn, rkey);
+
+		if (unlikely(!iser_conn->snd_w_inv)) {
+			iser_err("conn %p: unexpected remote invalidation, terminating connection\n",
+				 iser_conn);
+			return -EPROTO;
+		}
+
+		task = iscsi_itt_to_ctask(iser_conn->iscsi_conn, hdr->itt);
+		if (likely(task)) {
+			struct iscsi_iser_task *iser_task = task->dd_data;
+			struct iser_fr_desc *desc;
+
+			if (iser_task->dir[ISER_DIR_IN]) {
+				desc = iser_task->rdma_reg[ISER_DIR_IN].desc;
+				if (unlikely(iser_inv_desc(desc, rkey)))
+					return -EINVAL;
+			}
+
+			if (iser_task->dir[ISER_DIR_OUT]) {
+				desc = iser_task->rdma_reg[ISER_DIR_OUT].desc;
+				if (unlikely(iser_inv_desc(desc, rkey)))
+					return -EINVAL;
+			}
+		} else {
+			iser_err("failed to get task for itt=%d\n", hdr->itt);
+			return -EINVAL;
+		}
+	}
+
+	return 0;
+}
+
+
+void iser_task_rsp(struct ib_cq *cq, struct ib_wc *wc)
+{
+	struct ib_conn *ib_conn = wc->qp->qp_context;
+	struct iser_conn *iser_conn = to_iser_conn(ib_conn);
+	struct iser_rx_desc *desc = iser_rx(wc->wr_cqe);
+	struct iscsi_hdr *hdr;
+	int length, err;
+
+	if (unlikely(wc->status != IB_WC_SUCCESS)) {
+		iser_err_comp(wc, "task_rsp");
+		return;
+	}
+
+	ib_dma_sync_single_for_cpu(ib_conn->device->ib_device,
+				   desc->dma_addr, ISER_RX_PAYLOAD_SIZE,
+				   DMA_FROM_DEVICE);
+
+	hdr = &desc->iscsi_header;
+	length = wc->byte_len - ISER_HEADERS_LEN;
+
+	iser_dbg("op 0x%x itt 0x%x dlen %d\n", hdr->opcode,
+		 hdr->itt, length);
+
+	if (iser_check_remote_inv(iser_conn, wc, hdr)) {
+		iscsi_conn_failure(iser_conn->iscsi_conn,
+				   ISCSI_ERR_CONN_FAILED);
+		return;
+	}
+
+	iscsi_iser_recv(iser_conn->iscsi_conn, hdr, desc->data, length);
+
+	ib_dma_sync_single_for_device(ib_conn->device->ib_device,
+				      desc->dma_addr, ISER_RX_PAYLOAD_SIZE,
+				      DMA_FROM_DEVICE);
+
+	err = iser_post_recvm(iser_conn, desc);
+	if (err)
+		iser_err("posting rx buffer err %d\n", err);
+}
+
+void iser_cmd_comp(struct ib_cq *cq, struct ib_wc *wc)
+{
+	if (unlikely(wc->status != IB_WC_SUCCESS))
+		iser_err_comp(wc, "command");
+}
+
+void iser_ctrl_comp(struct ib_cq *cq, struct ib_wc *wc)
+{
+	struct iser_tx_desc *desc = iser_tx(wc->wr_cqe);
+	struct iscsi_task *task;
+
+	if (unlikely(wc->status != IB_WC_SUCCESS)) {
+		iser_err_comp(wc, "control");
+		return;
+	}
+
+	/* this arithmetic is legal by libiscsi dd_data allocation */
+	task = (void *)desc - sizeof(struct iscsi_task);
+	if (task->hdr->itt == RESERVED_ITT)
+		iscsi_put_task(task);
+}
+
+void iser_dataout_comp(struct ib_cq *cq, struct ib_wc *wc)
+{
+	struct iser_tx_desc *desc = iser_tx(wc->wr_cqe);
+	struct ib_conn *ib_conn = wc->qp->qp_context;
+	struct iser_device *device = ib_conn->device;
+
+	if (unlikely(wc->status != IB_WC_SUCCESS))
+		iser_err_comp(wc, "dataout");
+
+	ib_dma_unmap_single(device->ib_device, desc->dma_addr,
+			    ISER_HEADERS_LEN, DMA_TO_DEVICE);
+	kmem_cache_free(ig.desc_cache, desc);
+}
+
+void iser_task_rdma_init(struct iscsi_iser_task *iser_task)
+
+{
+	iser_task->status = ISER_TASK_STATUS_INIT;
+
+	iser_task->dir[ISER_DIR_IN] = 0;
+	iser_task->dir[ISER_DIR_OUT] = 0;
+
+	iser_task->data[ISER_DIR_IN].data_len  = 0;
+	iser_task->data[ISER_DIR_OUT].data_len = 0;
+
+	iser_task->prot[ISER_DIR_IN].data_len  = 0;
+	iser_task->prot[ISER_DIR_OUT].data_len = 0;
+
+	iser_task->prot[ISER_DIR_IN].dma_nents = 0;
+	iser_task->prot[ISER_DIR_OUT].dma_nents = 0;
+
+	memset(&iser_task->rdma_reg[ISER_DIR_IN], 0,
+	       sizeof(struct iser_mem_reg));
+	memset(&iser_task->rdma_reg[ISER_DIR_OUT], 0,
+	       sizeof(struct iser_mem_reg));
+}
+
+void iser_task_rdma_finalize(struct iscsi_iser_task *iser_task)
+{
+
+	if (iser_task->dir[ISER_DIR_IN]) {
+		iser_unreg_mem_fastreg(iser_task, ISER_DIR_IN);
+		iser_dma_unmap_task_data(iser_task, ISER_DIR_IN,
+					 DMA_FROM_DEVICE);
+	}
+
+	if (iser_task->dir[ISER_DIR_OUT]) {
+		iser_unreg_mem_fastreg(iser_task, ISER_DIR_OUT);
+		iser_dma_unmap_task_data(iser_task, ISER_DIR_OUT,
+					 DMA_TO_DEVICE);
+	}
+}
diff --git a/drivers/infiniband/ulp/iser/iser_memory.c b/drivers/infiniband/ulp/iser/iser_memory.c
new file mode 100644
index 000000000..6efcb79c8
--- /dev/null
+++ b/drivers/infiniband/ulp/iser/iser_memory.c
@@ -0,0 +1,391 @@
+/*
+ * Copyright (c) 2004, 2005, 2006 Voltaire, Inc. All rights reserved.
+ * Copyright (c) 2013-2014 Mellanox Technologies. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *	- Redistributions of source code must retain the above
+ *	  copyright notice, this list of conditions and the following
+ *	  disclaimer.
+ *
+ *	- Redistributions in binary form must reproduce the above
+ *	  copyright notice, this list of conditions and the following
+ *	  disclaimer in the documentation and/or other materials
+ *	  provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+#include <linux/kernel.h>
+#include <linux/slab.h>
+#include <linux/mm.h>
+#include <linux/highmem.h>
+#include <linux/scatterlist.h>
+
+#include "iscsi_iser.h"
+
+void iser_reg_comp(struct ib_cq *cq, struct ib_wc *wc)
+{
+	iser_err_comp(wc, "memreg");
+}
+
+static struct iser_fr_desc *iser_reg_desc_get_fr(struct ib_conn *ib_conn)
+{
+	struct iser_fr_pool *fr_pool = &ib_conn->fr_pool;
+	struct iser_fr_desc *desc;
+	unsigned long flags;
+
+	spin_lock_irqsave(&fr_pool->lock, flags);
+	desc = list_first_entry(&fr_pool->list,
+				struct iser_fr_desc, list);
+	list_del(&desc->list);
+	spin_unlock_irqrestore(&fr_pool->lock, flags);
+
+	return desc;
+}
+
+static void iser_reg_desc_put_fr(struct ib_conn *ib_conn,
+				 struct iser_fr_desc *desc)
+{
+	struct iser_fr_pool *fr_pool = &ib_conn->fr_pool;
+	unsigned long flags;
+
+	spin_lock_irqsave(&fr_pool->lock, flags);
+	list_add(&desc->list, &fr_pool->list);
+	spin_unlock_irqrestore(&fr_pool->lock, flags);
+}
+
+int iser_dma_map_task_data(struct iscsi_iser_task *iser_task,
+			   enum iser_data_dir iser_dir,
+			   enum dma_data_direction dma_dir)
+{
+	struct iser_data_buf *data = &iser_task->data[iser_dir];
+	struct ib_device *dev;
+
+	iser_task->dir[iser_dir] = 1;
+	dev = iser_task->iser_conn->ib_conn.device->ib_device;
+
+	data->dma_nents = ib_dma_map_sg(dev, data->sg, data->size, dma_dir);
+	if (unlikely(data->dma_nents == 0)) {
+		iser_err("dma_map_sg failed!!!\n");
+		return -EINVAL;
+	}
+
+	if (scsi_prot_sg_count(iser_task->sc)) {
+		struct iser_data_buf *pdata = &iser_task->prot[iser_dir];
+
+		pdata->dma_nents = ib_dma_map_sg(dev, pdata->sg, pdata->size, dma_dir);
+		if (unlikely(pdata->dma_nents == 0)) {
+			iser_err("protection dma_map_sg failed!!!\n");
+			goto out_unmap;
+		}
+	}
+
+	return 0;
+
+out_unmap:
+	ib_dma_unmap_sg(dev, data->sg, data->size, dma_dir);
+	return -EINVAL;
+}
+
+
+void iser_dma_unmap_task_data(struct iscsi_iser_task *iser_task,
+			      enum iser_data_dir iser_dir,
+			      enum dma_data_direction dma_dir)
+{
+	struct iser_data_buf *data = &iser_task->data[iser_dir];
+	struct ib_device *dev;
+
+	dev = iser_task->iser_conn->ib_conn.device->ib_device;
+	ib_dma_unmap_sg(dev, data->sg, data->size, dma_dir);
+
+	if (scsi_prot_sg_count(iser_task->sc)) {
+		struct iser_data_buf *pdata = &iser_task->prot[iser_dir];
+
+		ib_dma_unmap_sg(dev, pdata->sg, pdata->size, dma_dir);
+	}
+}
+
+static int iser_reg_dma(struct iser_device *device, struct iser_data_buf *mem,
+			struct iser_mem_reg *reg)
+{
+	struct scatterlist *sg = mem->sg;
+
+	reg->sge.lkey = device->pd->local_dma_lkey;
+	/*
+	 * FIXME: rework the registration code path to differentiate
+	 * rkey/lkey use cases
+	 */
+
+	if (device->pd->flags & IB_PD_UNSAFE_GLOBAL_RKEY)
+		reg->rkey = device->pd->unsafe_global_rkey;
+	else
+		reg->rkey = 0;
+	reg->sge.addr = sg_dma_address(&sg[0]);
+	reg->sge.length = sg_dma_len(&sg[0]);
+
+	iser_dbg("Single DMA entry: lkey=0x%x, rkey=0x%x, addr=0x%llx,"
+		 " length=0x%x\n", reg->sge.lkey, reg->rkey,
+		 reg->sge.addr, reg->sge.length);
+
+	return 0;
+}
+
+void iser_unreg_mem_fastreg(struct iscsi_iser_task *iser_task,
+			    enum iser_data_dir cmd_dir)
+{
+	struct iser_mem_reg *reg = &iser_task->rdma_reg[cmd_dir];
+	struct iser_fr_desc *desc;
+	struct ib_mr_status mr_status;
+
+	desc = reg->desc;
+	if (!desc)
+		return;
+
+	/*
+	 * The signature MR cannot be invalidated and reused without checking.
+	 * libiscsi calls the check_protection transport handler only if
+	 * SCSI-Response is received. And the signature MR is not checked if
+	 * the task is completed for some other reason like a timeout or error
+	 * handling. That's why we must check the signature MR here before
+	 * putting it to the free pool.
+	 */
+	if (unlikely(desc->sig_protected)) {
+		desc->sig_protected = false;
+		ib_check_mr_status(desc->rsc.sig_mr, IB_MR_CHECK_SIG_STATUS,
+				   &mr_status);
+	}
+	iser_reg_desc_put_fr(&iser_task->iser_conn->ib_conn, reg->desc);
+	reg->desc = NULL;
+}
+
+static void iser_set_dif_domain(struct scsi_cmnd *sc,
+				struct ib_sig_domain *domain)
+{
+	domain->sig_type = IB_SIG_TYPE_T10_DIF;
+	domain->sig.dif.pi_interval = scsi_prot_interval(sc);
+	domain->sig.dif.ref_tag = t10_pi_ref_tag(scsi_cmd_to_rq(sc));
+	/*
+	 * At the moment we hard code those, but in the future
+	 * we will take them from sc.
+	 */
+	domain->sig.dif.apptag_check_mask = 0xffff;
+	domain->sig.dif.app_escape = true;
+	domain->sig.dif.ref_escape = true;
+	if (sc->prot_flags & SCSI_PROT_REF_INCREMENT)
+		domain->sig.dif.ref_remap = true;
+}
+
+static int iser_set_sig_attrs(struct scsi_cmnd *sc,
+			      struct ib_sig_attrs *sig_attrs)
+{
+	switch (scsi_get_prot_op(sc)) {
+	case SCSI_PROT_WRITE_INSERT:
+	case SCSI_PROT_READ_STRIP:
+		sig_attrs->mem.sig_type = IB_SIG_TYPE_NONE;
+		iser_set_dif_domain(sc, &sig_attrs->wire);
+		sig_attrs->wire.sig.dif.bg_type = IB_T10DIF_CRC;
+		break;
+	case SCSI_PROT_READ_INSERT:
+	case SCSI_PROT_WRITE_STRIP:
+		sig_attrs->wire.sig_type = IB_SIG_TYPE_NONE;
+		iser_set_dif_domain(sc, &sig_attrs->mem);
+		sig_attrs->mem.sig.dif.bg_type = sc->prot_flags & SCSI_PROT_IP_CHECKSUM ?
+						IB_T10DIF_CSUM : IB_T10DIF_CRC;
+		break;
+	case SCSI_PROT_READ_PASS:
+	case SCSI_PROT_WRITE_PASS:
+		iser_set_dif_domain(sc, &sig_attrs->wire);
+		sig_attrs->wire.sig.dif.bg_type = IB_T10DIF_CRC;
+		iser_set_dif_domain(sc, &sig_attrs->mem);
+		sig_attrs->mem.sig.dif.bg_type = sc->prot_flags & SCSI_PROT_IP_CHECKSUM ?
+						IB_T10DIF_CSUM : IB_T10DIF_CRC;
+		break;
+	default:
+		iser_err("Unsupported PI operation %d\n",
+			 scsi_get_prot_op(sc));
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
+static inline void iser_set_prot_checks(struct scsi_cmnd *sc, u8 *mask)
+{
+	*mask = 0;
+	if (sc->prot_flags & SCSI_PROT_REF_CHECK)
+		*mask |= IB_SIG_CHECK_REFTAG;
+	if (sc->prot_flags & SCSI_PROT_GUARD_CHECK)
+		*mask |= IB_SIG_CHECK_GUARD;
+}
+
+static inline void iser_inv_rkey(struct ib_send_wr *inv_wr, struct ib_mr *mr,
+				 struct ib_cqe *cqe, struct ib_send_wr *next_wr)
+{
+	inv_wr->opcode = IB_WR_LOCAL_INV;
+	inv_wr->wr_cqe = cqe;
+	inv_wr->ex.invalidate_rkey = mr->rkey;
+	inv_wr->send_flags = 0;
+	inv_wr->num_sge = 0;
+	inv_wr->next = next_wr;
+}
+
+static int iser_reg_sig_mr(struct iscsi_iser_task *iser_task,
+			   struct iser_data_buf *mem,
+			   struct iser_data_buf *sig_mem,
+			   struct iser_reg_resources *rsc,
+			   struct iser_mem_reg *sig_reg)
+{
+	struct iser_tx_desc *tx_desc = &iser_task->desc;
+	struct ib_cqe *cqe = &iser_task->iser_conn->ib_conn.reg_cqe;
+	struct ib_mr *mr = rsc->sig_mr;
+	struct ib_sig_attrs *sig_attrs = mr->sig_attrs;
+	struct ib_reg_wr *wr = &tx_desc->reg_wr;
+	int ret;
+
+	memset(sig_attrs, 0, sizeof(*sig_attrs));
+	ret = iser_set_sig_attrs(iser_task->sc, sig_attrs);
+	if (ret)
+		goto err;
+
+	iser_set_prot_checks(iser_task->sc, &sig_attrs->check_mask);
+
+	if (rsc->sig_mr->need_inval)
+		iser_inv_rkey(&tx_desc->inv_wr, mr, cqe, &wr->wr);
+
+	ib_update_fast_reg_key(mr, ib_inc_rkey(mr->rkey));
+
+	ret = ib_map_mr_sg_pi(mr, mem->sg, mem->dma_nents, NULL,
+			      sig_mem->sg, sig_mem->dma_nents, NULL, SZ_4K);
+	if (unlikely(ret)) {
+		iser_err("failed to map PI sg (%d)\n",
+			 mem->dma_nents + sig_mem->dma_nents);
+		goto err;
+	}
+
+	memset(wr, 0, sizeof(*wr));
+	wr->wr.next = &tx_desc->send_wr;
+	wr->wr.opcode = IB_WR_REG_MR_INTEGRITY;
+	wr->wr.wr_cqe = cqe;
+	wr->wr.num_sge = 0;
+	wr->wr.send_flags = 0;
+	wr->mr = mr;
+	wr->key = mr->rkey;
+	wr->access = IB_ACCESS_LOCAL_WRITE |
+		     IB_ACCESS_REMOTE_READ |
+		     IB_ACCESS_REMOTE_WRITE;
+	rsc->sig_mr->need_inval = true;
+
+	sig_reg->sge.lkey = mr->lkey;
+	sig_reg->rkey = mr->rkey;
+	sig_reg->sge.addr = mr->iova;
+	sig_reg->sge.length = mr->length;
+
+	iser_dbg("lkey=0x%x rkey=0x%x addr=0x%llx length=%u\n",
+		 sig_reg->sge.lkey, sig_reg->rkey, sig_reg->sge.addr,
+		 sig_reg->sge.length);
+err:
+	return ret;
+}
+
+static int iser_fast_reg_mr(struct iscsi_iser_task *iser_task,
+			    struct iser_data_buf *mem,
+			    struct iser_reg_resources *rsc,
+			    struct iser_mem_reg *reg)
+{
+	struct iser_tx_desc *tx_desc = &iser_task->desc;
+	struct ib_cqe *cqe = &iser_task->iser_conn->ib_conn.reg_cqe;
+	struct ib_mr *mr = rsc->mr;
+	struct ib_reg_wr *wr = &tx_desc->reg_wr;
+	int n;
+
+	if (rsc->mr->need_inval)
+		iser_inv_rkey(&tx_desc->inv_wr, mr, cqe, &wr->wr);
+
+	ib_update_fast_reg_key(mr, ib_inc_rkey(mr->rkey));
+
+	n = ib_map_mr_sg(mr, mem->sg, mem->dma_nents, NULL, SZ_4K);
+	if (unlikely(n != mem->dma_nents)) {
+		iser_err("failed to map sg (%d/%d)\n",
+			 n, mem->dma_nents);
+		return n < 0 ? n : -EINVAL;
+	}
+
+	wr->wr.next = &tx_desc->send_wr;
+	wr->wr.opcode = IB_WR_REG_MR;
+	wr->wr.wr_cqe = cqe;
+	wr->wr.send_flags = 0;
+	wr->wr.num_sge = 0;
+	wr->mr = mr;
+	wr->key = mr->rkey;
+	wr->access = IB_ACCESS_LOCAL_WRITE  |
+		     IB_ACCESS_REMOTE_WRITE |
+		     IB_ACCESS_REMOTE_READ;
+
+	rsc->mr->need_inval = true;
+
+	reg->sge.lkey = mr->lkey;
+	reg->rkey = mr->rkey;
+	reg->sge.addr = mr->iova;
+	reg->sge.length = mr->length;
+
+	iser_dbg("lkey=0x%x rkey=0x%x addr=0x%llx length=0x%x\n",
+		 reg->sge.lkey, reg->rkey, reg->sge.addr, reg->sge.length);
+
+	return 0;
+}
+
+int iser_reg_mem_fastreg(struct iscsi_iser_task *task,
+			 enum iser_data_dir dir,
+			 bool all_imm)
+{
+	struct ib_conn *ib_conn = &task->iser_conn->ib_conn;
+	struct iser_device *device = ib_conn->device;
+	struct iser_data_buf *mem = &task->data[dir];
+	struct iser_mem_reg *reg = &task->rdma_reg[dir];
+	struct iser_fr_desc *desc;
+	bool use_dma_key;
+	int err;
+
+	use_dma_key = mem->dma_nents == 1 && (all_imm || !iser_always_reg) &&
+		      scsi_get_prot_op(task->sc) == SCSI_PROT_NORMAL;
+	if (use_dma_key)
+		return iser_reg_dma(device, mem, reg);
+
+	desc = iser_reg_desc_get_fr(ib_conn);
+	if (scsi_get_prot_op(task->sc) == SCSI_PROT_NORMAL) {
+		err = iser_fast_reg_mr(task, mem, &desc->rsc, reg);
+		if (unlikely(err))
+			goto err_reg;
+	} else {
+		err = iser_reg_sig_mr(task, mem, &task->prot[dir],
+				      &desc->rsc, reg);
+		if (unlikely(err))
+			goto err_reg;
+
+		desc->sig_protected = true;
+	}
+
+	reg->desc = desc;
+
+	return 0;
+
+err_reg:
+	iser_reg_desc_put_fr(ib_conn, desc);
+
+	return err;
+}
diff --git a/drivers/infiniband/ulp/iser/iser_verbs.c b/drivers/infiniband/ulp/iser/iser_verbs.c
new file mode 100644
index 000000000..55fe68e5b
--- /dev/null
+++ b/drivers/infiniband/ulp/iser/iser_verbs.c
@@ -0,0 +1,951 @@
+/*
+ * Copyright (c) 2004, 2005, 2006 Voltaire, Inc. All rights reserved.
+ * Copyright (c) 2005, 2006 Cisco Systems.  All rights reserved.
+ * Copyright (c) 2013-2014 Mellanox Technologies. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *	- Redistributions of source code must retain the above
+ *	  copyright notice, this list of conditions and the following
+ *	  disclaimer.
+ *
+ *	- Redistributions in binary form must reproduce the above
+ *	  copyright notice, this list of conditions and the following
+ *	  disclaimer in the documentation and/or other materials
+ *	  provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+#include <linux/kernel.h>
+#include <linux/slab.h>
+#include <linux/delay.h>
+
+#include "iscsi_iser.h"
+
+static void iser_qp_event_callback(struct ib_event *cause, void *context)
+{
+	iser_err("qp event %s (%d)\n",
+		 ib_event_msg(cause->event), cause->event);
+}
+
+static void iser_event_handler(struct ib_event_handler *handler,
+				struct ib_event *event)
+{
+	iser_err("async event %s (%d) on device %s port %d\n",
+		 ib_event_msg(event->event), event->event,
+		dev_name(&event->device->dev), event->element.port_num);
+}
+
+/*
+ * iser_create_device_ib_res - creates Protection Domain (PD), Completion
+ * Queue (CQ), DMA Memory Region (DMA MR) with the device associated with
+ * the adaptor.
+ *
+ * Return: 0 on success, -1 on failure
+ */
+static int iser_create_device_ib_res(struct iser_device *device)
+{
+	struct ib_device *ib_dev = device->ib_device;
+
+	if (!(ib_dev->attrs.device_cap_flags & IB_DEVICE_MEM_MGT_EXTENSIONS)) {
+		iser_err("IB device does not support memory registrations\n");
+		return -1;
+	}
+
+	device->pd = ib_alloc_pd(ib_dev,
+		iser_always_reg ? 0 : IB_PD_UNSAFE_GLOBAL_RKEY);
+	if (IS_ERR(device->pd))
+		goto pd_err;
+
+	INIT_IB_EVENT_HANDLER(&device->event_handler, ib_dev,
+			      iser_event_handler);
+	ib_register_event_handler(&device->event_handler);
+	return 0;
+
+pd_err:
+	iser_err("failed to allocate an IB resource\n");
+	return -1;
+}
+
+/*
+ * iser_free_device_ib_res - destroy/dealloc/dereg the DMA MR,
+ * CQ and PD created with the device associated with the adaptor.
+ */
+static void iser_free_device_ib_res(struct iser_device *device)
+{
+	ib_unregister_event_handler(&device->event_handler);
+	ib_dealloc_pd(device->pd);
+
+	device->pd = NULL;
+}
+
+static struct iser_fr_desc *
+iser_create_fastreg_desc(struct iser_device *device,
+			 struct ib_pd *pd,
+			 bool pi_enable,
+			 unsigned int size)
+{
+	struct iser_fr_desc *desc;
+	struct ib_device *ib_dev = device->ib_device;
+	enum ib_mr_type mr_type;
+	int ret;
+
+	desc = kzalloc_obj(*desc);
+	if (!desc)
+		return ERR_PTR(-ENOMEM);
+
+	if (ib_dev->attrs.kernel_cap_flags & IBK_SG_GAPS_REG)
+		mr_type = IB_MR_TYPE_SG_GAPS;
+	else
+		mr_type = IB_MR_TYPE_MEM_REG;
+
+	desc->rsc.mr = ib_alloc_mr(pd, mr_type, size);
+	if (IS_ERR(desc->rsc.mr)) {
+		ret = PTR_ERR(desc->rsc.mr);
+		iser_err("Failed to allocate ib_fast_reg_mr err=%d\n", ret);
+		goto err_alloc_mr;
+	}
+
+	if (pi_enable) {
+		desc->rsc.sig_mr = ib_alloc_mr_integrity(pd, size, size);
+		if (IS_ERR(desc->rsc.sig_mr)) {
+			ret = PTR_ERR(desc->rsc.sig_mr);
+			iser_err("Failed to allocate sig_mr err=%d\n", ret);
+			goto err_alloc_mr_integrity;
+		}
+	}
+
+	return desc;
+
+err_alloc_mr_integrity:
+	ib_dereg_mr(desc->rsc.mr);
+err_alloc_mr:
+	kfree(desc);
+
+	return ERR_PTR(ret);
+}
+
+static void iser_destroy_fastreg_desc(struct iser_fr_desc *desc)
+{
+	struct iser_reg_resources *res = &desc->rsc;
+
+	ib_dereg_mr(res->mr);
+	if (res->sig_mr) {
+		ib_dereg_mr(res->sig_mr);
+		res->sig_mr = NULL;
+	}
+	kfree(desc);
+}
+
+/**
+ * iser_alloc_fastreg_pool - Creates pool of fast_reg descriptors
+ * for fast registration work requests.
+ * @ib_conn: connection RDMA resources
+ * @cmds_max: max number of SCSI commands for this connection
+ * @size: max number of pages per map request
+ *
+ * Return: 0 on success, or errno code on failure
+ */
+int iser_alloc_fastreg_pool(struct ib_conn *ib_conn,
+			    unsigned cmds_max,
+			    unsigned int size)
+{
+	struct iser_device *device = ib_conn->device;
+	struct iser_fr_pool *fr_pool = &ib_conn->fr_pool;
+	struct iser_fr_desc *desc;
+	int i, ret;
+
+	INIT_LIST_HEAD(&fr_pool->list);
+	INIT_LIST_HEAD(&fr_pool->all_list);
+	spin_lock_init(&fr_pool->lock);
+	fr_pool->size = 0;
+	for (i = 0; i < cmds_max; i++) {
+		desc = iser_create_fastreg_desc(device, device->pd,
+						ib_conn->pi_support, size);
+		if (IS_ERR(desc)) {
+			ret = PTR_ERR(desc);
+			goto err;
+		}
+
+		list_add_tail(&desc->list, &fr_pool->list);
+		list_add_tail(&desc->all_list, &fr_pool->all_list);
+		fr_pool->size++;
+	}
+
+	return 0;
+
+err:
+	iser_free_fastreg_pool(ib_conn);
+	return ret;
+}
+
+/**
+ * iser_free_fastreg_pool - releases the pool of fast_reg descriptors
+ * @ib_conn: connection RDMA resources
+ */
+void iser_free_fastreg_pool(struct ib_conn *ib_conn)
+{
+	struct iser_fr_pool *fr_pool = &ib_conn->fr_pool;
+	struct iser_fr_desc *desc, *tmp;
+	int i = 0;
+
+	if (list_empty(&fr_pool->all_list))
+		return;
+
+	iser_info("freeing conn %p fr pool\n", ib_conn);
+
+	list_for_each_entry_safe(desc, tmp, &fr_pool->all_list, all_list) {
+		list_del(&desc->all_list);
+		iser_destroy_fastreg_desc(desc);
+		++i;
+	}
+
+	if (i < fr_pool->size)
+		iser_warn("pool still has %d regions registered\n",
+			  fr_pool->size - i);
+}
+
+/*
+ * iser_create_ib_conn_res - Queue-Pair (QP)
+ *
+ * Return: 0 on success, -1 on failure
+ */
+static int iser_create_ib_conn_res(struct ib_conn *ib_conn)
+{
+	struct iser_conn *iser_conn = to_iser_conn(ib_conn);
+	struct iser_device	*device;
+	struct ib_device	*ib_dev;
+	struct ib_qp_init_attr	init_attr;
+	int			ret = -ENOMEM;
+	unsigned int max_send_wr, cq_size;
+
+	BUG_ON(ib_conn->device == NULL);
+
+	device = ib_conn->device;
+	ib_dev = device->ib_device;
+
+	/* +1 for drain */
+	if (ib_conn->pi_support)
+		max_send_wr = ISER_QP_SIG_MAX_REQ_DTOS + 1;
+	else
+		max_send_wr = ISER_QP_MAX_REQ_DTOS + 1;
+	max_send_wr = min(max_send_wr, ib_dev->attrs.max_qp_wr);
+
+	cq_size = max_send_wr + ISER_QP_MAX_RECV_DTOS;
+	ib_conn->cq = ib_cq_pool_get(ib_dev, cq_size, -1, IB_POLL_SOFTIRQ);
+	if (IS_ERR(ib_conn->cq)) {
+		ret = PTR_ERR(ib_conn->cq);
+		goto cq_err;
+	}
+	ib_conn->cq_size = cq_size;
+
+	memset(&init_attr, 0, sizeof(init_attr));
+
+	init_attr.event_handler = iser_qp_event_callback;
+	init_attr.qp_context = (void *)ib_conn;
+	init_attr.send_cq = ib_conn->cq;
+	init_attr.recv_cq = ib_conn->cq;
+	/* +1 for drain */
+	init_attr.cap.max_recv_wr = ISER_QP_MAX_RECV_DTOS + 1;
+	init_attr.cap.max_send_sge = 2;
+	init_attr.cap.max_recv_sge = 1;
+	init_attr.sq_sig_type = IB_SIGNAL_REQ_WR;
+	init_attr.qp_type = IB_QPT_RC;
+	init_attr.cap.max_send_wr = max_send_wr;
+	if (ib_conn->pi_support)
+		init_attr.create_flags |= IB_QP_CREATE_INTEGRITY_EN;
+	iser_conn->max_cmds = ISER_GET_MAX_XMIT_CMDS(max_send_wr - 1);
+
+	ret = rdma_create_qp(ib_conn->cma_id, device->pd, &init_attr);
+	if (ret)
+		goto out_err;
+
+	ib_conn->qp = ib_conn->cma_id->qp;
+	iser_info("setting conn %p cma_id %p qp %p max_send_wr %d\n", ib_conn,
+		  ib_conn->cma_id, ib_conn->cma_id->qp, max_send_wr);
+	return ret;
+
+out_err:
+	ib_cq_pool_put(ib_conn->cq, ib_conn->cq_size);
+cq_err:
+	iser_err("unable to alloc mem or create resource, err %d\n", ret);
+
+	return ret;
+}
+
+/*
+ * based on the resolved device node GUID see if there already allocated
+ * device for this device. If there's no such, create one.
+ */
+static
+struct iser_device *iser_device_find_by_ib_device(struct rdma_cm_id *cma_id)
+{
+	struct iser_device *device;
+
+	mutex_lock(&ig.device_list_mutex);
+
+	list_for_each_entry(device, &ig.device_list, ig_list)
+		/* find if there's a match using the node GUID */
+		if (device->ib_device->node_guid == cma_id->device->node_guid)
+			goto inc_refcnt;
+
+	device = kzalloc_obj(*device);
+	if (!device)
+		goto out;
+
+	/* assign this device to the device */
+	device->ib_device = cma_id->device;
+	/* init the device and link it into ig device list */
+	if (iser_create_device_ib_res(device)) {
+		kfree(device);
+		device = NULL;
+		goto out;
+	}
+	list_add(&device->ig_list, &ig.device_list);
+
+inc_refcnt:
+	device->refcount++;
+out:
+	mutex_unlock(&ig.device_list_mutex);
+	return device;
+}
+
+/* if there's no demand for this device, release it */
+static void iser_device_try_release(struct iser_device *device)
+{
+	mutex_lock(&ig.device_list_mutex);
+	device->refcount--;
+	iser_info("device %p refcount %d\n", device, device->refcount);
+	if (!device->refcount) {
+		iser_free_device_ib_res(device);
+		list_del(&device->ig_list);
+		kfree(device);
+	}
+	mutex_unlock(&ig.device_list_mutex);
+}
+
+void iser_release_work(struct work_struct *work)
+{
+	struct iser_conn *iser_conn;
+
+	iser_conn = container_of(work, struct iser_conn, release_work);
+
+	/* Wait for conn_stop to complete */
+	wait_for_completion(&iser_conn->stop_completion);
+	/* Wait for IB resouces cleanup to complete */
+	wait_for_completion(&iser_conn->ib_completion);
+
+	mutex_lock(&iser_conn->state_mutex);
+	iser_conn->state = ISER_CONN_DOWN;
+	mutex_unlock(&iser_conn->state_mutex);
+
+	iser_conn_release(iser_conn);
+}
+
+/**
+ * iser_free_ib_conn_res - release IB related resources
+ * @iser_conn: iser connection struct
+ * @destroy: indicator if we need to try to release the
+ *     iser device and memory regoins pool (only iscsi
+ *     shutdown and DEVICE_REMOVAL will use this).
+ *
+ * This routine is called with the iser state mutex held
+ * so the cm_id removal is out of here. It is Safe to
+ * be invoked multiple times.
+ */
+static void iser_free_ib_conn_res(struct iser_conn *iser_conn, bool destroy)
+{
+	struct ib_conn *ib_conn = &iser_conn->ib_conn;
+	struct iser_device *device = ib_conn->device;
+
+	iser_info("freeing conn %p cma_id %p qp %p\n",
+		  iser_conn, ib_conn->cma_id, ib_conn->qp);
+
+	if (ib_conn->qp) {
+		rdma_destroy_qp(ib_conn->cma_id);
+		ib_cq_pool_put(ib_conn->cq, ib_conn->cq_size);
+		ib_conn->qp = NULL;
+	}
+
+	if (destroy) {
+		if (iser_conn->rx_descs)
+			iser_free_rx_descriptors(iser_conn);
+
+		if (device) {
+			iser_device_try_release(device);
+			ib_conn->device = NULL;
+		}
+	}
+}
+
+/**
+ * iser_conn_release - Frees all conn objects and deallocs conn descriptor
+ * @iser_conn: iSER connection context
+ */
+void iser_conn_release(struct iser_conn *iser_conn)
+{
+	struct ib_conn *ib_conn = &iser_conn->ib_conn;
+
+	mutex_lock(&ig.connlist_mutex);
+	list_del(&iser_conn->conn_list);
+	mutex_unlock(&ig.connlist_mutex);
+
+	mutex_lock(&iser_conn->state_mutex);
+	/* In case we endup here without ep_disconnect being invoked. */
+	if (iser_conn->state != ISER_CONN_DOWN) {
+		iser_warn("iser conn %p state %d, expected state down.\n",
+			  iser_conn, iser_conn->state);
+		iscsi_destroy_endpoint(iser_conn->ep);
+		iser_conn->state = ISER_CONN_DOWN;
+	}
+	/*
+	 * In case we never got to bind stage, we still need to
+	 * release IB resources (which is safe to call more than once).
+	 */
+	iser_free_ib_conn_res(iser_conn, true);
+	mutex_unlock(&iser_conn->state_mutex);
+
+	if (ib_conn->cma_id) {
+		rdma_destroy_id(ib_conn->cma_id);
+		ib_conn->cma_id = NULL;
+	}
+
+	kfree(iser_conn);
+}
+
+/**
+ * iser_conn_terminate - triggers start of the disconnect procedures and
+ * waits for them to be done
+ * @iser_conn: iSER connection context
+ *
+ * Called with state mutex held
+ */
+int iser_conn_terminate(struct iser_conn *iser_conn)
+{
+	struct ib_conn *ib_conn = &iser_conn->ib_conn;
+	int err = 0;
+
+	lockdep_assert_held(&iser_conn->state_mutex);
+
+	/* terminate the iser conn only if the conn state is UP */
+	if (iser_conn->state != ISER_CONN_UP)
+		return 0;
+
+	iser_conn->state = ISER_CONN_TERMINATING;
+	iser_info("iser_conn %p state %d\n", iser_conn, iser_conn->state);
+
+	/* suspend queuing of new iscsi commands */
+	if (iser_conn->iscsi_conn)
+		iscsi_suspend_queue(iser_conn->iscsi_conn);
+
+	/*
+	 * In case we didn't already clean up the cma_id (peer initiated
+	 * a disconnection), we need to Cause the CMA to change the QP
+	 * state to ERROR.
+	 */
+	if (ib_conn->cma_id) {
+		err = rdma_disconnect(ib_conn->cma_id);
+		if (err)
+			iser_err("Failed to disconnect, conn: 0x%p err %d\n",
+				 iser_conn, err);
+
+		/* block until all flush errors are consumed */
+		ib_drain_qp(ib_conn->qp);
+	}
+
+	return 1;
+}
+
+/*
+ * Called with state mutex held
+ */
+static void iser_connect_error(struct rdma_cm_id *cma_id)
+{
+	struct iser_conn *iser_conn = cma_id->context;
+
+	lockdep_assert_held(&iser_conn->state_mutex);
+
+	iser_conn->state = ISER_CONN_TERMINATING;
+}
+
+static void iser_calc_scsi_params(struct iser_conn *iser_conn,
+				  unsigned int max_sectors)
+{
+	struct iser_device *device = iser_conn->ib_conn.device;
+	struct ib_device_attr *attr = &device->ib_device->attrs;
+	unsigned short sg_tablesize, sup_sg_tablesize;
+	unsigned short reserved_mr_pages;
+	u32 max_num_sg;
+
+	/*
+	 * FRs without SG_GAPS can only map up to a (device) page per entry,
+	 * but if the first entry is misaligned we'll end up using two entries
+	 * (head and tail) for a single page worth data, so one additional
+	 * entry is required.
+	 */
+	if (attr->kernel_cap_flags & IBK_SG_GAPS_REG)
+		reserved_mr_pages = 0;
+	else
+		reserved_mr_pages = 1;
+
+	if (iser_conn->ib_conn.pi_support)
+		max_num_sg = attr->max_pi_fast_reg_page_list_len;
+	else
+		max_num_sg = attr->max_fast_reg_page_list_len;
+
+	sg_tablesize = DIV_ROUND_UP(max_sectors * SECTOR_SIZE, SZ_4K);
+	sup_sg_tablesize = min_t(uint, ISCSI_ISER_MAX_SG_TABLESIZE,
+				 max_num_sg - reserved_mr_pages);
+	iser_conn->scsi_sg_tablesize = min(sg_tablesize, sup_sg_tablesize);
+	iser_conn->pages_per_mr =
+		iser_conn->scsi_sg_tablesize + reserved_mr_pages;
+}
+
+/*
+ * Called with state mutex held
+ */
+static void iser_addr_handler(struct rdma_cm_id *cma_id)
+{
+	struct iser_conn *iser_conn = cma_id->context;
+	struct iser_device *device;
+	struct ib_conn *ib_conn;
+	int    ret;
+
+	lockdep_assert_held(&iser_conn->state_mutex);
+
+	if (iser_conn->state != ISER_CONN_PENDING)
+		/* bailout */
+		return;
+
+	ib_conn = &iser_conn->ib_conn;
+	device = iser_device_find_by_ib_device(cma_id);
+	if (!device) {
+		iser_err("device lookup/creation failed\n");
+		iser_connect_error(cma_id);
+		return;
+	}
+
+	ib_conn->device = device;
+
+	/* connection T10-PI support */
+	if (iser_pi_enable) {
+		if (!(device->ib_device->attrs.kernel_cap_flags &
+		      IBK_INTEGRITY_HANDOVER)) {
+			iser_warn("T10-PI requested but not supported on %s, "
+				  "continue without T10-PI\n",
+				  dev_name(&ib_conn->device->ib_device->dev));
+			ib_conn->pi_support = false;
+		} else {
+			ib_conn->pi_support = true;
+		}
+	}
+
+	iser_calc_scsi_params(iser_conn, iser_max_sectors);
+
+	ret = rdma_resolve_route(cma_id, 1000);
+	if (ret) {
+		iser_err("resolve route failed: %d\n", ret);
+		iser_connect_error(cma_id);
+		return;
+	}
+}
+
+/*
+ * Called with state mutex held
+ */
+static void iser_route_handler(struct rdma_cm_id *cma_id)
+{
+	struct rdma_conn_param conn_param;
+	int ret;
+	struct iser_cm_hdr req_hdr;
+	struct iser_conn *iser_conn = cma_id->context;
+	struct ib_conn *ib_conn = &iser_conn->ib_conn;
+	struct ib_device *ib_dev = ib_conn->device->ib_device;
+
+	lockdep_assert_held(&iser_conn->state_mutex);
+
+	if (iser_conn->state != ISER_CONN_PENDING)
+		/* bailout */
+		return;
+
+	ret = iser_create_ib_conn_res(ib_conn);
+	if (ret)
+		goto failure;
+
+	memset(&conn_param, 0, sizeof conn_param);
+	conn_param.responder_resources = min(ib_dev->attrs.max_qp_rd_atom, U8_MAX);
+	conn_param.initiator_depth = 1;
+	conn_param.retry_count = 7;
+	conn_param.rnr_retry_count = 6;
+
+	memset(&req_hdr, 0, sizeof(req_hdr));
+	req_hdr.flags = ISER_ZBVA_NOT_SUP;
+	if (!iser_always_reg)
+		req_hdr.flags |= ISER_SEND_W_INV_NOT_SUP;
+	conn_param.private_data	= (void *)&req_hdr;
+	conn_param.private_data_len = sizeof(struct iser_cm_hdr);
+
+	ret = rdma_connect_locked(cma_id, &conn_param);
+	if (ret) {
+		iser_err("failure connecting: %d\n", ret);
+		goto failure;
+	}
+
+	return;
+failure:
+	iser_connect_error(cma_id);
+}
+
+/*
+ * Called with state mutex held
+ */
+static void iser_connected_handler(struct rdma_cm_id *cma_id,
+				   const void *private_data)
+{
+	struct iser_conn *iser_conn = cma_id->context;
+	struct ib_qp_attr attr;
+	struct ib_qp_init_attr init_attr;
+
+	lockdep_assert_held(&iser_conn->state_mutex);
+
+	if (iser_conn->state != ISER_CONN_PENDING)
+		/* bailout */
+		return;
+
+	(void)ib_query_qp(cma_id->qp, &attr, ~0, &init_attr);
+	iser_info("remote qpn:%x my qpn:%x\n", attr.dest_qp_num, cma_id->qp->qp_num);
+
+	if (private_data) {
+		u8 flags = *(u8 *)private_data;
+
+		iser_conn->snd_w_inv = !(flags & ISER_SEND_W_INV_NOT_SUP);
+	}
+
+	iser_info("conn %p: negotiated %s invalidation\n",
+		  iser_conn, iser_conn->snd_w_inv ? "remote" : "local");
+
+	iser_conn->state = ISER_CONN_UP;
+	complete(&iser_conn->up_completion);
+}
+
+/*
+ * Called with state mutex held
+ */
+static void iser_cleanup_handler(struct rdma_cm_id *cma_id,
+				 bool destroy)
+{
+	struct iser_conn *iser_conn = cma_id->context;
+
+	lockdep_assert_held(&iser_conn->state_mutex);
+	/*
+	 * We are not guaranteed that we visited disconnected_handler
+	 * by now, call it here to be safe that we handle CM drep
+	 * and flush errors.
+	 */
+	if (iser_conn_terminate(iser_conn)) {
+		if (iser_conn->iscsi_conn)
+			iscsi_conn_failure(iser_conn->iscsi_conn,
+					   ISCSI_ERR_CONN_FAILED);
+		else
+			iser_err("iscsi_iser connection isn't bound\n");
+	}
+	iser_free_ib_conn_res(iser_conn, destroy);
+	complete(&iser_conn->ib_completion);
+}
+
+static int iser_cma_handler(struct rdma_cm_id *cma_id,
+			    struct rdma_cm_event *event)
+{
+	struct iser_conn *iser_conn;
+	int ret = 0;
+
+	iser_conn = cma_id->context;
+	iser_info("%s (%d): status %d conn %p id %p\n",
+		  rdma_event_msg(event->event), event->event,
+		  event->status, cma_id->context, cma_id);
+
+	mutex_lock(&iser_conn->state_mutex);
+	switch (event->event) {
+	case RDMA_CM_EVENT_ADDR_RESOLVED:
+		iser_addr_handler(cma_id);
+		break;
+	case RDMA_CM_EVENT_ROUTE_RESOLVED:
+		iser_route_handler(cma_id);
+		break;
+	case RDMA_CM_EVENT_ESTABLISHED:
+		iser_connected_handler(cma_id, event->param.conn.private_data);
+		break;
+	case RDMA_CM_EVENT_REJECTED:
+		iser_info("Connection rejected: %s\n",
+			 rdma_reject_msg(cma_id, event->status));
+		fallthrough;
+	case RDMA_CM_EVENT_ADDR_ERROR:
+	case RDMA_CM_EVENT_ROUTE_ERROR:
+	case RDMA_CM_EVENT_CONNECT_ERROR:
+	case RDMA_CM_EVENT_UNREACHABLE:
+		iser_connect_error(cma_id);
+		break;
+	case RDMA_CM_EVENT_DISCONNECTED:
+	case RDMA_CM_EVENT_ADDR_CHANGE:
+	case RDMA_CM_EVENT_TIMEWAIT_EXIT:
+		iser_cleanup_handler(cma_id, false);
+		break;
+	case RDMA_CM_EVENT_DEVICE_REMOVAL:
+		/*
+		 * we *must* destroy the device as we cannot rely
+		 * on iscsid to be around to initiate error handling.
+		 * also if we are not in state DOWN implicitly destroy
+		 * the cma_id.
+		 */
+		iser_cleanup_handler(cma_id, true);
+		if (iser_conn->state != ISER_CONN_DOWN) {
+			iser_conn->ib_conn.cma_id = NULL;
+			ret = 1;
+		}
+		break;
+	default:
+		iser_err("Unexpected RDMA CM event: %s (%d)\n",
+			 rdma_event_msg(event->event), event->event);
+		break;
+	}
+	mutex_unlock(&iser_conn->state_mutex);
+
+	return ret;
+}
+
+void iser_conn_init(struct iser_conn *iser_conn)
+{
+	struct ib_conn *ib_conn = &iser_conn->ib_conn;
+
+	iser_conn->state = ISER_CONN_INIT;
+	init_completion(&iser_conn->stop_completion);
+	init_completion(&iser_conn->ib_completion);
+	init_completion(&iser_conn->up_completion);
+	INIT_LIST_HEAD(&iser_conn->conn_list);
+	mutex_init(&iser_conn->state_mutex);
+
+	ib_conn->reg_cqe.done = iser_reg_comp;
+}
+
+/*
+ * starts the process of connecting to the target
+ * sleeps until the connection is established or rejected
+ */
+int iser_connect(struct iser_conn *iser_conn, struct sockaddr *src_addr,
+		 struct sockaddr *dst_addr, int non_blocking)
+{
+	struct ib_conn *ib_conn = &iser_conn->ib_conn;
+	int err = 0;
+
+	mutex_lock(&iser_conn->state_mutex);
+
+	sprintf(iser_conn->name, "%pISp", dst_addr);
+
+	iser_info("connecting to: %s\n", iser_conn->name);
+
+	/* the device is known only --after-- address resolution */
+	ib_conn->device = NULL;
+
+	iser_conn->state = ISER_CONN_PENDING;
+
+	ib_conn->cma_id = rdma_create_id(&init_net, iser_cma_handler,
+					 iser_conn, RDMA_PS_TCP, IB_QPT_RC);
+	if (IS_ERR(ib_conn->cma_id)) {
+		err = PTR_ERR(ib_conn->cma_id);
+		iser_err("rdma_create_id failed: %d\n", err);
+		goto id_failure;
+	}
+
+	err = rdma_resolve_addr(ib_conn->cma_id, src_addr, dst_addr, 1000);
+	if (err) {
+		iser_err("rdma_resolve_addr failed: %d\n", err);
+		goto addr_failure;
+	}
+
+	if (!non_blocking) {
+		wait_for_completion_interruptible(&iser_conn->up_completion);
+
+		if (iser_conn->state != ISER_CONN_UP) {
+			err =  -EIO;
+			goto connect_failure;
+		}
+	}
+	mutex_unlock(&iser_conn->state_mutex);
+
+	mutex_lock(&ig.connlist_mutex);
+	list_add(&iser_conn->conn_list, &ig.connlist);
+	mutex_unlock(&ig.connlist_mutex);
+	return 0;
+
+id_failure:
+	ib_conn->cma_id = NULL;
+addr_failure:
+	iser_conn->state = ISER_CONN_DOWN;
+connect_failure:
+	mutex_unlock(&iser_conn->state_mutex);
+	iser_conn_release(iser_conn);
+	return err;
+}
+
+int iser_post_recvl(struct iser_conn *iser_conn)
+{
+	struct ib_conn *ib_conn = &iser_conn->ib_conn;
+	struct iser_login_desc *desc = &iser_conn->login_desc;
+	struct ib_recv_wr wr;
+	int ret;
+
+	desc->sge.addr = desc->rsp_dma;
+	desc->sge.length = ISER_RX_LOGIN_SIZE;
+	desc->sge.lkey = ib_conn->device->pd->local_dma_lkey;
+
+	desc->cqe.done = iser_login_rsp;
+	wr.wr_cqe = &desc->cqe;
+	wr.sg_list = &desc->sge;
+	wr.num_sge = 1;
+	wr.next = NULL;
+
+	ret = ib_post_recv(ib_conn->qp, &wr, NULL);
+	if (unlikely(ret))
+		iser_err("ib_post_recv login failed ret=%d\n", ret);
+
+	return ret;
+}
+
+int iser_post_recvm(struct iser_conn *iser_conn, struct iser_rx_desc *rx_desc)
+{
+	struct ib_conn *ib_conn = &iser_conn->ib_conn;
+	struct ib_recv_wr wr;
+	int ret;
+
+	rx_desc->cqe.done = iser_task_rsp;
+	wr.wr_cqe = &rx_desc->cqe;
+	wr.sg_list = &rx_desc->rx_sg;
+	wr.num_sge = 1;
+	wr.next = NULL;
+
+	ret = ib_post_recv(ib_conn->qp, &wr, NULL);
+	if (unlikely(ret))
+		iser_err("ib_post_recv failed ret=%d\n", ret);
+
+	return ret;
+}
+
+
+/**
+ * iser_post_send - Initiate a Send DTO operation
+ * @ib_conn: connection RDMA resources
+ * @tx_desc: iSER TX descriptor
+ *
+ * Return: 0 on success, -1 on failure
+ */
+int iser_post_send(struct ib_conn *ib_conn, struct iser_tx_desc *tx_desc)
+{
+	struct ib_send_wr *wr = &tx_desc->send_wr;
+	struct ib_send_wr *first_wr;
+	int ret;
+
+	ib_dma_sync_single_for_device(ib_conn->device->ib_device,
+				      tx_desc->dma_addr, ISER_HEADERS_LEN,
+				      DMA_TO_DEVICE);
+
+	wr->next = NULL;
+	wr->wr_cqe = &tx_desc->cqe;
+	wr->sg_list = tx_desc->tx_sg;
+	wr->num_sge = tx_desc->num_sge;
+	wr->opcode = IB_WR_SEND;
+	wr->send_flags = IB_SEND_SIGNALED;
+
+	if (tx_desc->inv_wr.next)
+		first_wr = &tx_desc->inv_wr;
+	else if (tx_desc->reg_wr.wr.next)
+		first_wr = &tx_desc->reg_wr.wr;
+	else
+		first_wr = wr;
+
+	ret = ib_post_send(ib_conn->qp, first_wr, NULL);
+	if (unlikely(ret))
+		iser_err("ib_post_send failed, ret:%d opcode:%d\n",
+			 ret, wr->opcode);
+
+	return ret;
+}
+
+u8 iser_check_task_pi_status(struct iscsi_iser_task *iser_task,
+			     enum iser_data_dir cmd_dir, sector_t *sector)
+{
+	struct iser_mem_reg *reg = &iser_task->rdma_reg[cmd_dir];
+	struct iser_fr_desc *desc = reg->desc;
+	unsigned long sector_size = iser_task->sc->device->sector_size;
+	struct ib_mr_status mr_status;
+	int ret;
+
+	if (desc && desc->sig_protected) {
+		desc->sig_protected = false;
+		ret = ib_check_mr_status(desc->rsc.sig_mr,
+					 IB_MR_CHECK_SIG_STATUS, &mr_status);
+		if (ret) {
+			iser_err("ib_check_mr_status failed, ret %d\n", ret);
+			/* Not a lot we can do, return ambiguous guard error */
+			*sector = 0;
+			return 0x1;
+		}
+
+		if (mr_status.fail_status & IB_MR_CHECK_SIG_STATUS) {
+			sector_t sector_off = mr_status.sig_err.sig_err_offset;
+
+			sector_div(sector_off, sector_size + 8);
+			*sector = scsi_get_sector(iser_task->sc) + sector_off;
+
+			iser_err("PI error found type %d at sector %llx "
+			       "expected %x vs actual %x\n",
+			       mr_status.sig_err.err_type,
+			       (unsigned long long)*sector,
+			       mr_status.sig_err.expected,
+			       mr_status.sig_err.actual);
+
+			switch (mr_status.sig_err.err_type) {
+			case IB_SIG_BAD_GUARD:
+				return 0x1;
+			case IB_SIG_BAD_REFTAG:
+				return 0x3;
+			case IB_SIG_BAD_APPTAG:
+				return 0x2;
+			}
+		}
+	}
+
+	return 0;
+}
+
+void iser_err_comp(struct ib_wc *wc, const char *type)
+{
+	if (wc->status != IB_WC_WR_FLUSH_ERR) {
+		struct iser_conn *iser_conn = to_iser_conn(wc->qp->qp_context);
+
+		iser_err("%s failure: %s (%d) vend_err %#x\n", type,
+			 ib_wc_status_msg(wc->status), wc->status,
+			 wc->vendor_err);
+
+		if (iser_conn->iscsi_conn)
+			iscsi_conn_failure(iser_conn->iscsi_conn,
+					   ISCSI_ERR_CONN_FAILED);
+	} else {
+		iser_dbg("%s failure: %s (%d)\n", type,
+			 ib_wc_status_msg(wc->status), wc->status);
+	}
+}