diff mbox

[16/28] ibnbd_clt: add main functionality of ibnbd_client

Message ID 1490352343-20075-17-git-send-email-jinpu.wangl@profitbricks.com (mailing list archive)
State RFC
Headers show

Commit Message

Jinpu Wang March 24, 2017, 10:45 a.m. UTC
From: Jack Wang <jinpu.wang@profitbricks.com>

It provides interface to map remote device as local block devices
(/dev/ibnbdx) and prepare IO for the transfer.

It supports both request mode and multiqueue mode.

Signed-off-by: Jack Wang <jinpu.wang@profitbricks.com>
Signed-off-by: Kleber Souza <kleber.souza@profitbricks.com>
Signed-off-by: Danil Kipnis <danil.kipnis@profitbricks.com>
Signed-off-by: Roman Pen <roman.penyaev@profitbricks.com>
---
 drivers/block/ibnbd_client/ibnbd_clt.c | 2007 ++++++++++++++++++++++++++++++++
 1 file changed, 2007 insertions(+)
 create mode 100644 drivers/block/ibnbd_client/ibnbd_clt.c
diff mbox

Patch

diff --git a/drivers/block/ibnbd_client/ibnbd_clt.c b/drivers/block/ibnbd_client/ibnbd_clt.c
new file mode 100644
index 0000000..945c8df
--- /dev/null
+++ b/drivers/block/ibnbd_client/ibnbd_clt.c
@@ -0,0 +1,2007 @@ 
+/*
+ * InfiniBand Network Block Driver
+ *
+ * Copyright (c) 2014 - 2017 ProfitBricks GmbH. All rights reserved.
+ * Authors: Fabian Holler < mail@fholler.de>
+ *          Jack Wang <jinpu.wang@profitbricks.com>
+ *   	    Kleber Souza <kleber.souza@profitbricks.com>
+ * 	    Danil Kipnis <danil.kipnis@profitbricks.com>
+ *   	    Roman Pen <roman.penyaev@profitbricks.com>
+ *          Milind Dumbare <Milind.dumbare@gmail.com>
+ *
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ * 1. Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions, and the following disclaimer,
+ *    without modification.
+ * 2. Redistributions in binary form must reproduce at minimum a disclaimer
+ *    substantially similar to the "NO WARRANTY" disclaimer below
+ *    ("Disclaimer") and any redistribution must be conditioned upon
+ *    including a substantially similar Disclaimer requirement for further
+ *    binary redistribution.
+ * 3. Neither the names of the above-listed copyright holders nor the names
+ *    of any contributors may be used to endorse or promote products derived
+ *    from this software without specific prior written permission.
+ *
+ * Alternatively, this software may be distributed under the terms of the
+ * GNU General Public License ("GPL") version 2 as published by the Free
+ * Software Foundation.
+ *
+ * NO WARRANTY
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTIBILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * HOLDERS OR CONTRIBUTORS BE LIABLE FOR SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+ * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
+ * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
+ * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
+ * STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING
+ * IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGES.
+ *
+ */
+
+#include <linux/module.h>
+#include <linux/blkdev.h>
+#include <linux/hdreg.h>		/* for hd_geometry */
+#include <linux/scatterlist.h>
+#include <linux/idr.h>
+#include <linux/fs.h>
+#include <linux/uaccess.h>
+#include <rdma/ib.h>
+#include <uapi/linux/in6.h>
+
+#include "ibnbd_clt.h"
+#include "ibnbd_clt_sysfs.h"
+#include "../ibnbd_inc/ibnbd.h"
+#include <rdma/ibtrs.h>
+
+MODULE_AUTHOR("ibnbd@profitbricks.com");
+MODULE_DESCRIPTION("InfiniBand Network Block Device Client");
+MODULE_VERSION(__stringify(IBNBD_VER));
+MODULE_LICENSE("GPL");
+
+static int ibnbd_client_major;
+static DEFINE_IDR(g_index_idr);
+static DEFINE_RWLOCK(g_index_lock);
+static DEFINE_SPINLOCK(sess_lock);
+static DEFINE_SPINLOCK(dev_lock);
+static LIST_HEAD(session_list);
+static LIST_HEAD(devs_list);
+static DECLARE_WAIT_QUEUE_HEAD(sess_list_waitq);
+static struct ibtrs_clt_ops ops;
+
+static bool softirq_enable;
+module_param(softirq_enable, bool, 0444);
+MODULE_PARM_DESC(softirq_enable, "finish request in softirq_fn."
+		 " (default: 0)");
+/*
+ * Maximum number of partitions an instance can have.
+ * 6 bits = 64 minors = 63 partitions (one minor is used for the device itself)
+ */
+#define IBNBD_PART_BITS		6
+#define KERNEL_SECTOR_SIZE      512
+
+inline bool ibnbd_clt_dev_is_open(struct ibnbd_dev *dev)
+{
+	return dev->dev_state == DEV_STATE_OPEN;
+}
+
+static void ibnbd_clt_put_dev(struct ibnbd_dev *dev)
+{
+	if (!atomic_dec_if_positive(&dev->refcount)) {
+		write_lock(&g_index_lock);
+		idr_remove(&g_index_idr, dev->clt_device_id);
+		write_unlock(&g_index_lock);
+		kfree(dev->hw_queues);
+		kfree(dev->close_compl);
+		ibnbd_clt_put_sess(dev->sess);
+		kfree(dev);
+	}
+}
+
+static int ibnbd_clt_get_dev(struct ibnbd_dev *dev)
+{
+	return atomic_inc_not_zero(&dev->refcount);
+}
+
+static struct ibnbd_dev *g_get_dev(int dev_id)
+{
+	struct ibnbd_dev *dev;
+
+	read_lock(&g_index_lock);
+	dev = idr_find(&g_index_idr, dev_id);
+	if (!dev)
+		dev = ERR_PTR(-ENXIO);
+	read_unlock(&g_index_lock);
+
+	return dev;
+}
+
+static void ibnbd_clt_set_dev_attr(struct ibnbd_dev *dev,
+				   const struct ibnbd_msg_open_rsp *rsp)
+{
+	dev->device_id			= rsp->device_id;
+	dev->nsectors			= rsp->nsectors;
+	dev->logical_block_size		= rsp->logical_block_size;
+	dev->physical_block_size	= rsp->physical_block_size;
+	dev->max_write_same_sectors	= rsp->max_write_same_sectors;
+	dev->max_discard_sectors	= rsp->max_discard_sectors;
+	dev->discard_zeroes_data	= rsp->discard_zeroes_data;
+	dev->discard_granularity	= rsp->discard_granularity;
+	dev->discard_alignment		= rsp->discard_alignment;
+	dev->secure_discard		= rsp->secure_discard;
+	dev->rotational			= rsp->rotational;
+	dev->remote_io_mode		= rsp->io_mode;
+
+	if (dev->remote_io_mode == IBNBD_FILEIO) {
+		dev->max_hw_sectors = dev->sess->max_io_size /
+			rsp->logical_block_size;
+		dev->max_segments = BMAX_SEGMENTS;
+	} else {
+		dev->max_hw_sectors = dev->sess->max_io_size /
+			rsp->logical_block_size <
+			rsp->max_hw_sectors ?
+			dev->sess->max_io_size /
+			rsp->logical_block_size : rsp->max_hw_sectors;
+		dev->max_segments = rsp->max_segments > BMAX_SEGMENTS ?
+				    BMAX_SEGMENTS : rsp->max_segments;
+	}
+}
+
+static void ibnbd_clt_revalidate_disk(struct ibnbd_dev *dev,
+				      size_t new_nsectors)
+{
+	int err = 0;
+
+	INFO(dev, "Device size changed from %zu to %zu sectors\n",
+	     dev->nsectors, new_nsectors);
+	dev->nsectors = new_nsectors;
+	set_capacity(dev->gd,
+		     dev->nsectors * (dev->logical_block_size /
+				      KERNEL_SECTOR_SIZE));
+	err = revalidate_disk(dev->gd);
+	if (err)
+		ERR(dev, "Failed to change device size from"
+		    " %zu to %zu, errno: %d\n", dev->nsectors,
+		     new_nsectors, err);
+}
+
+static void process_msg_sess_info_rsp(struct ibnbd_session *sess,
+				      struct ibnbd_msg_sess_info_rsp *msg)
+{
+	sess->ver = min_t(u8, msg->ver, IBNBD_VERSION);
+	DEB("Session to %s (%s) using protocol version %d (client version: %d,"
+	    " server version: %d)\n", sess->str_addr, sess->hostname, sess->ver,
+	    IBNBD_VERSION, msg->ver);
+}
+
+static int process_msg_open_rsp(struct ibnbd_session *sess,
+				struct ibnbd_msg_open_rsp *rsp)
+{
+	struct ibnbd_dev *dev;
+	int err = 0;
+
+	dev = g_get_dev(rsp->clt_device_id);
+	if (IS_ERR(dev)) {
+		ERR_NP("Open-Response message received from session %s"
+		       " for unknown device (id: %d)\n", sess->str_addr,
+		       rsp->clt_device_id);
+		return -ENOENT;
+	}
+
+	if (!ibnbd_clt_get_dev(dev)) {
+		ERR_NP("Failed to process Open-Response message from session"
+		       " %s, unable to get reference to device (id: %d)",
+		       sess->str_addr, rsp->clt_device_id);
+		return -ENOENT;
+	}
+
+	mutex_lock(&dev->lock);
+
+	if (dev->dev_state == DEV_STATE_UNMAPPED) {
+		INFO(dev, "Ignoring Open-Response message from server for "
+		     " unmapped device\n");
+		err = -ENOENT;
+		goto out;
+	}
+
+	if (rsp->result) {
+		ERR(dev, "Server failed to open device for mapping, errno:"
+		    " %d\n", rsp->result);
+		dev->open_errno = rsp->result;
+		if (dev->open_compl)
+			complete(dev->open_compl);
+		goto out;
+	}
+
+	if (dev->dev_state == DEV_STATE_CLOSED) {
+		/* if the device was remapped and the size changed in the
+		 * meantime we need to revalidate it
+		 */
+		if (dev->nsectors != rsp->nsectors)
+			ibnbd_clt_revalidate_disk(dev, (size_t)rsp->nsectors);
+		INFO(dev, "Device online, device remapped successfully\n");
+	}
+
+	ibnbd_clt_set_dev_attr(dev, rsp);
+
+	dev->dev_state = DEV_STATE_OPEN;
+	if (dev->open_compl)
+		complete(dev->open_compl);
+
+out:
+	mutex_unlock(&dev->lock);
+	ibnbd_clt_put_dev(dev);
+
+	return err;
+}
+
+static void process_msg_revalidate(struct ibnbd_session *sess,
+				   struct ibnbd_msg_revalidate *msg)
+{
+	struct ibnbd_dev *dev;
+
+	dev = g_get_dev(msg->clt_device_id);
+	if (IS_ERR(dev)) {
+		ERR_NP("Received device revalidation message from session %s"
+		       " for non-existent device (id %d)\n", sess->str_addr,
+		       msg->clt_device_id);
+		return;
+	}
+
+	if (!ibnbd_clt_get_dev(dev)) {
+		ERR_NP("Failed to process device revalidation message from"
+		       " session %s, unable to get reference to device"
+		       " (id: %d)", sess->str_addr, msg->clt_device_id);
+		return;
+	}
+
+	mutex_lock(&dev->lock);
+
+	if (dev->dev_state == DEV_STATE_UNMAPPED) {
+		ERR(dev, "Received device revalidation message"
+		    " for unmapped device\n");
+		goto out;
+	}
+
+	if (dev->nsectors != msg->nsectors &&
+	    dev->dev_state == DEV_STATE_OPEN) {
+		ibnbd_clt_revalidate_disk(dev, (size_t)msg->nsectors);
+	} else {
+		INFO(dev, "Ignoring device revalidate message, "
+		     "current device size is the same as in the "
+		     "revalidate message, %llu sectors\n", msg->nsectors);
+	}
+
+out:
+	mutex_unlock(&dev->lock);
+	ibnbd_clt_put_dev(dev);
+}
+
+static int send_msg_close(struct ibtrs_session *sess, u32 device_id)
+{
+	struct ibnbd_msg_close msg;
+	struct kvec vec = {
+		.iov_base = &msg,
+		.iov_len  = sizeof(msg)
+	};
+
+	msg.hdr.type	= IBNBD_MSG_CLOSE;
+	msg.device_id	= device_id;
+
+	return ibtrs_clt_send(sess, &vec, 1);
+}
+
+static void ibnbd_clt_recv(void *priv, const void *msg, size_t len)
+{
+	const struct ibnbd_msg_hdr *hdr = msg;
+	struct ibnbd_session *sess = priv;
+
+	if (unlikely(WARN_ON(!hdr) || ibnbd_validate_message(msg, len)))
+		return;
+
+	print_hex_dump_debug("", DUMP_PREFIX_OFFSET, 8, 1, msg, len, true);
+
+	switch (hdr->type) {
+	case IBNBD_MSG_SESS_INFO_RSP: {
+		struct ibnbd_msg_sess_info_rsp *rsp =
+			(struct ibnbd_msg_sess_info_rsp *)msg;
+
+		process_msg_sess_info_rsp(sess, rsp);
+		if (sess->sess_info_compl)
+			complete(sess->sess_info_compl);
+		break;
+	}
+	case IBNBD_MSG_OPEN_RSP: {
+		int err;
+		struct ibnbd_msg_open_rsp *rsp =
+			(struct ibnbd_msg_open_rsp *)msg;
+
+		if (process_msg_open_rsp(sess, rsp) && !rsp->result) {
+			ERR_NP("Failed to process open response message from"
+			       " server, sending close message for dev id:"
+			       " %u\n", rsp->device_id);
+
+			err = send_msg_close(sess->sess, rsp->device_id);
+			if (err)
+				ERR_NP("Failed to send close msg for device"
+				       " with id: %u, errno: %d\n",
+				       rsp->device_id, err);
+		}
+
+		break;
+	}
+	case IBNBD_MSG_CLOSE_RSP: {
+		struct ibnbd_dev *dev;
+		struct ibnbd_msg_close_rsp *rsp =
+			(struct ibnbd_msg_close_rsp *)msg;
+
+		dev = g_get_dev(rsp->clt_device_id);
+		if (IS_ERR(dev)) {
+			ERR_NP("Close-Response message received from session %s"
+			       " for unknown device (id: %u)\n", sess->str_addr,
+			       rsp->clt_device_id);
+			break;
+		}
+		if (dev->close_compl && dev->dev_state == DEV_STATE_UNMAPPED)
+			complete(dev->close_compl);
+
+		break;
+	}
+	case IBNBD_MSG_REVAL:
+		process_msg_revalidate(sess,
+				       (struct ibnbd_msg_revalidate *)msg);
+		break;
+	default:
+		ERR_NP("IBNBD message with unknown type %d received from"
+		       " session %s\n", hdr->type, sess->str_addr);
+		break;
+	}
+}
+
+static void ibnbd_blk_delay_work(struct work_struct *work)
+{
+	struct ibnbd_dev *dev;
+
+	dev = container_of(work, struct ibnbd_dev, rq_delay_work.work);
+	spin_lock_irq(dev->queue->queue_lock);
+	blk_start_queue(dev->queue);
+	spin_unlock_irq(dev->queue->queue_lock);
+}
+
+/**
+ * What is the difference between this and original blk_delay_queue() ?
+ * Here the stop queue flag is cleared, so we are like MQ.
+ */
+static void ibnbd_blk_delay_queue(struct ibnbd_dev *dev, unsigned long msecs)
+{
+	int cpu = get_cpu();
+
+	kblockd_schedule_delayed_work_on(cpu, &dev->rq_delay_work,
+					 msecs_to_jiffies(msecs));
+	put_cpu();
+}
+
+static inline void ibnbd_dev_requeue(struct ibnbd_queue *q)
+{
+	struct ibnbd_dev *dev = q->dev;
+
+	if (dev->queue_mode == BLK_MQ) {
+		if (WARN_ON(!q->hctx))
+			return;
+		blk_mq_delay_queue(q->hctx, 0);
+	} else if (dev->queue_mode == BLK_RQ) {
+		ibnbd_blk_delay_queue(q->dev, 0);
+	} else {
+		WARN(1, "We support requeueing only for RQ or MQ");
+	}
+}
+
+enum {
+	IBNBD_DELAY_10ms   = 10,
+	IBNBD_DELAY_IFBUSY = -1,
+};
+
+/**
+ * ibnbd_get_cpu_qlist() - finds a list with HW queues to be requeued
+ *
+ * Description:
+ *     Each CPU has a list of HW queues, which needs to be requeed.  If a list
+ *     is not empty - it is marked with a bit.  This function finds first
+ *     set bit in a bitmap and returns corresponding CPU list.
+ */
+static struct ibnbd_cpu_qlist *ibnbd_get_cpu_qlist(struct ibnbd_session *sess,
+						   int cpu)
+{
+	int bit;
+
+	/* First half */
+	bit = find_next_bit(sess->cpu_queues_bm, nr_cpu_ids, cpu);
+	if (bit < nr_cpu_ids) {
+		return per_cpu_ptr(sess->cpu_queues, bit);
+	} else if (cpu != 0) {
+		/* Second half */
+		bit = find_next_bit(sess->cpu_queues_bm, cpu, 0);
+		if (bit < cpu)
+			return per_cpu_ptr(sess->cpu_queues, bit);
+	}
+
+	return NULL;
+}
+
+static inline int nxt_cpu(int cpu)
+{
+	return (cpu + 1) % NR_CPUS;
+}
+
+/**
+ * get_cpu_rr_var() - returns pointer to percpu var containing last cpu requeued
+ *
+ * It also sets the var to the current cpu if the var was never set before
+ * (== -1).
+ */
+#define get_cpu_rr_var(percpu)				\
+({							\
+	int *cpup;					\
+							\
+	cpup = &get_cpu_var(*percpu);			\
+	if (unlikely(*cpup < 0))			\
+		*cpup = smp_processor_id();		\
+	cpup;						\
+})
+
+/**
+ * ibnbd_requeue_if_needed() - requeue if CPU queue is marked as non empty
+ *
+ * Description:
+ *     Each CPU has it's own list of HW queues, which should be requeued.
+ *     Function finds such list with HW queues, takes a list lock, picks up
+ *     the first HW queue out of the list and requeues it.
+ *
+ * Return:
+ *     True if the queue was requeued, false otherwise.
+ *
+ * Context:
+ *     Does not matter.
+ */
+static inline bool ibnbd_requeue_if_needed(struct ibnbd_session *sess)
+{
+	struct ibnbd_queue *q = NULL;
+	struct ibnbd_cpu_qlist *cpu_q;
+	unsigned long flags;
+	int cpuv;
+
+	int *uninitialized_var(cpup);
+
+	/*
+	 * To keep fairness and not to let other queues starve we always
+	 * try to wake up someone else in round-robin manner.  That of course
+	 * increases latency but queues always have a chance to be executed.
+	 */
+	cpup = get_cpu_rr_var(sess->cpu_rr);
+	cpuv = (*cpup + 1) % num_online_cpus();
+	for (cpu_q = ibnbd_get_cpu_qlist(sess, cpuv); cpu_q;
+	     cpu_q = ibnbd_get_cpu_qlist(sess, nxt_cpu(cpu_q->cpu))) {
+		if (!spin_trylock_irqsave(&cpu_q->requeue_lock, flags))
+			continue;
+		if (likely(test_bit(cpu_q->cpu, sess->cpu_queues_bm))) {
+			q = list_first_entry_or_null(&cpu_q->requeue_list,
+						     typeof(*q), requeue_list);
+			if (WARN_ON(!q))
+				goto clear_bit;
+			list_del_init(&q->requeue_list);
+			clear_bit_unlock(0, &q->in_list);
+
+			if (list_empty(&cpu_q->requeue_list)) {
+				/* Clear bit if nothing is left */
+clear_bit:
+				clear_bit(cpu_q->cpu, sess->cpu_queues_bm);
+			}
+		}
+		spin_unlock_irqrestore(&cpu_q->requeue_lock, flags);
+
+		if (q)
+			break;
+	}
+
+	/**
+	 * Saves the CPU that is going to be requeued on the per-cpu var. Just
+	 * incrementing it doesn't work because ibnbd_get_cpu_qlist() will
+	 * always return the first CPU with something on the queue list when the
+	 * value stored on the var is greater than the last CPU with something
+	 * on the list.
+	 */
+	if (cpu_q)
+		*cpup = cpu_q->cpu;
+	put_cpu_var(sess->cpu_rr);
+
+	if (q)
+		ibnbd_dev_requeue(q);
+
+	return !!q;
+}
+
+/**
+ * ibnbd_requeue_all_if_idle() - requeue all queues left in the list if
+ *     session is idling (there are no requests in-flight).
+ *
+ * Description:
+ *     This function tries to rerun all stopped queues if there are no
+ *     requests in-flight anymore.  This function tries to solve an obvious
+ *     problem, when number of tags < than number of queues (hctx), which
+ *     are stopped and put to sleep.  If last tag, which has been just put,
+ *     does not wake up all left queues (hctxs), IO requests hang forever.
+ *
+ *     That can happen when all number of tags, say N, have been exhausted
+ *     from one CPU, and we have many block devices per session, say M.
+ *     Each block device has it's own queue (hctx) for each CPU, so eventually
+ *     we can put that number of queues (hctxs) to sleep: M x NR_CPUS.
+ *     If number of tags N < M x NR_CPUS finally we will get an IO hang.
+ *
+ *     To avoid this hang last caller of ibnbd_put_tag() (last caller is the
+ *     one who observes sess->busy == 0) must wake up all remaining queues.
+ *
+ * Context:
+ *     Does not matter.
+ */
+static inline void ibnbd_requeue_all_if_idle(struct ibnbd_session *sess)
+{
+	bool requeued;
+
+	do {
+		requeued = ibnbd_requeue_if_needed(sess);
+	} while (atomic_read(&sess->busy) == 0 && requeued);
+}
+
+static struct ibtrs_tag *ibnbd_get_tag(struct ibnbd_session *sess, int cpu,
+				       size_t tag_bytes, int wait)
+{
+	struct ibtrs_tag *tag;
+
+	tag = ibtrs_get_tag(sess->sess, cpu, tag_bytes,
+			    wait ? IBTRS_TAG_WAIT : IBTRS_TAG_NOWAIT);
+	if (likely(tag))
+		/* We have a subtle rare case here, when all tags can be
+		 * consumed before busy counter increased.  This is safe,
+		 * because loser will get NULL as a tag, observe 0 busy
+		 * counter and immediately restart the queue himself.
+		 */
+		atomic_inc(&sess->busy);
+
+	return tag;
+}
+
+static void ibnbd_put_tag(struct ibnbd_session *sess, struct ibtrs_tag *tag)
+{
+	ibtrs_put_tag(sess->sess, tag);
+	atomic_dec(&sess->busy);
+	/* Paired with ibnbd_dev_add_to_requeue().  Decrement first
+	 * and then check queue bits.
+	 */
+	smp_mb__after_atomic();
+	ibnbd_requeue_all_if_idle(sess);
+}
+
+static struct ibnbd_iu *ibnbd_get_iu(struct ibnbd_session *sess,
+				     size_t tag_bytes, int wait)
+{
+	struct ibnbd_iu *iu;
+	struct ibtrs_tag *tag;
+
+	tag = ibnbd_get_tag(sess, -1, tag_bytes,
+			    wait ? IBTRS_TAG_WAIT : IBTRS_TAG_NOWAIT);
+	if (unlikely(!tag))
+		return NULL;
+	iu = ibtrs_tag_to_pdu(tag);
+	iu->tag = tag; /* yes, ibtrs_tag_from_pdu() can be nice here,
+			* but also we have to think about MQ mode
+			*/
+
+	return iu;
+}
+
+static void ibnbd_put_iu(struct ibnbd_session *sess, struct ibnbd_iu *iu)
+{
+	ibnbd_put_tag(sess, iu->tag);
+}
+
+static void ibnbd_softirq_done_fn(struct request *rq)
+{
+	struct ibnbd_dev *dev		= rq->rq_disk->private_data;
+	struct ibnbd_session *sess	= dev->sess;
+	struct ibnbd_iu *iu;
+
+	switch (dev->queue_mode) {
+	case BLK_MQ:
+		iu = blk_mq_rq_to_pdu(rq);
+		ibnbd_put_tag(sess, iu->tag);
+		blk_mq_end_request(rq, iu->errno);
+		if (rq->rq_flags & RQF_SPECIAL_PAYLOAD)
+			__free_page(rq->special_vec.bv_page);
+		break;
+	case BLK_RQ:
+		iu = rq->special;
+		blk_end_request_all(rq, iu->errno);
+		if (rq->rq_flags & RQF_SPECIAL_PAYLOAD)
+			__free_page(rq->special_vec.bv_page);
+		break;
+	default:
+		WARN(true, "dev->queue_mode , contains unexpected"
+		     " value: %d. Memory Corruption? Inflight I/O stalled!\n",
+		     dev->queue_mode);
+		return;
+	}
+}
+
+static void ibnbd_clt_rdma_ev(void *priv, enum ibtrs_clt_rdma_ev ev, int errno)
+{
+	struct ibnbd_iu *iu		= (struct ibnbd_iu *)priv;
+	struct ibnbd_dev *dev		= iu->dev;
+	struct request *rq;
+	const int flags = iu->msg.rw;
+	bool is_read;
+
+	switch (dev->queue_mode) {
+	case BLK_MQ:
+		rq = iu->rq;
+		is_read = req_op(rq) == READ;
+		iu->errno = errno;
+		if (softirq_enable) {
+			blk_mq_complete_request(rq, errno);
+		} else {
+			ibnbd_put_tag(dev->sess, iu->tag);
+			blk_mq_end_request(rq, errno);
+
+			if (rq->rq_flags & RQF_SPECIAL_PAYLOAD)
+				__free_page(rq->special_vec.bv_page);
+
+		}
+		break;
+	case BLK_RQ:
+		rq = iu->rq;
+		is_read = req_op(rq) == READ;
+		iu->errno = errno;
+		if (softirq_enable) {
+			blk_complete_request(rq);
+		} else {
+			blk_end_request_all(rq, errno);
+			if (rq->rq_flags & RQF_SPECIAL_PAYLOAD)
+				__free_page(rq->special_vec.bv_page);
+		}
+		break;
+	default:
+		WARN(true, "dev->queue_mode , contains unexpected"
+		     " value: %d. Memory Corruption? Inflight I/O stalled!\n",
+		     dev->queue_mode);
+		return;
+	}
+
+	if (errno)
+		INFO_RL(dev, "%s I/O failed with status: %d, flags: 0x%x\n",
+			is_read ? "read" : "write", errno, flags);
+}
+
+static int send_msg_open(struct ibnbd_dev *dev)
+{
+	int err;
+	struct ibnbd_msg_open msg;
+	struct kvec vec = {
+		.iov_base = &msg,
+		.iov_len  = sizeof(msg)
+	};
+
+	msg.hdr.type		= IBNBD_MSG_OPEN;
+	msg.clt_device_id	= dev->clt_device_id;
+	msg.access_mode		= dev->access_mode;
+	msg.io_mode		= dev->io_mode;
+	strlcpy(msg.dev_name, dev->pathname, sizeof(msg.dev_name));
+
+	err = ibtrs_clt_send(dev->sess->sess, &vec, 1);
+
+	return err;
+}
+
+static int send_msg_sess_info(struct ibnbd_session *sess)
+{
+	struct ibnbd_msg_sess_info msg;
+	struct kvec vec = {
+		.iov_base = &msg,
+		.iov_len  = sizeof(msg)
+	};
+
+	msg.hdr.type	= IBNBD_MSG_SESS_INFO;
+	msg.ver		= IBNBD_VERSION;
+
+	return ibtrs_clt_send(sess->sess, &vec, 1);
+}
+
+int open_remote_device(struct ibnbd_dev *dev)
+{
+	int err;
+
+	err = send_msg_open(dev);
+	if (unlikely(err)) {
+		ERR(dev, "Failed to send open msg, err: %d\n", err);
+		return err;
+	}
+	return 0;
+}
+
+static int find_dev_cb(int id, void *ptr, void *data)
+{
+	struct ibnbd_dev *dev = ptr;
+	struct ibnbd_session *sess = data;
+
+	if (dev->sess == sess && dev->dev_state == DEV_STATE_INIT &&
+	    dev->open_compl) {
+		dev->dev_state = DEV_STATE_INIT_CLOSED;
+		dev->open_errno = -ECOMM;
+		complete(dev->open_compl);
+		ERR(dev, "Device offline, session disconnected.\n");
+	} else if (dev->sess == sess && dev->dev_state == DEV_STATE_UNMAPPED &&
+	    dev->close_compl) {
+		complete(dev->close_compl);
+		ERR(dev, "Device closed, session disconnected.\n");
+	}
+
+	return 0;
+}
+
+static void __set_dev_states_closed(struct ibnbd_session *sess)
+{
+	struct ibnbd_dev *dev;
+
+	list_for_each_entry(dev, &sess->devs_list, list) {
+		mutex_lock(&dev->lock);
+		dev->dev_state = DEV_STATE_CLOSED;
+		dev->open_errno = -ECOMM;
+		if (dev->open_compl)
+			complete(dev->open_compl);
+		ERR(dev, "Device offline, session disconnected.\n");
+		mutex_unlock(&dev->lock);
+	}
+	read_lock(&g_index_lock);
+	idr_for_each(&g_index_idr, find_dev_cb, sess);
+	read_unlock(&g_index_lock);
+}
+
+static int update_sess_info(struct ibnbd_session *sess)
+{
+	int err;
+
+	sess->sess_info_compl = kmalloc(sizeof(*sess->sess_info_compl),
+					GFP_KERNEL);
+	if (!sess->sess_info_compl) {
+		ERR_NP("Failed to allocate memory for completion for session"
+		       " %s (%s)\n", sess->str_addr, sess->hostname);
+		return -ENOMEM;
+	}
+
+	init_completion(sess->sess_info_compl);
+
+	err = send_msg_sess_info(sess);
+	if (unlikely(err)) {
+		ERR_NP("Failed to send SESS_INFO message for session %s (%s)\n",
+		       sess->str_addr, sess->hostname);
+		goto out;
+	}
+
+	/* wait for IBNBD_MSG_SESS_INFO_RSP from server */
+	wait_for_completion(sess->sess_info_compl);
+out:
+	kfree(sess->sess_info_compl);
+	sess->sess_info_compl = NULL;
+
+	return err;
+}
+
+static void reopen_worker(struct work_struct *work)
+{
+	struct ibnbd_work *w;
+	struct ibnbd_session *sess;
+	struct ibnbd_dev *dev;
+	int err;
+
+	w = container_of(work, struct ibnbd_work, work);
+	sess = w->sess;
+	kfree(w);
+
+	mutex_lock(&sess->lock);
+	if (sess->state == SESS_STATE_DESTROYED) {
+		mutex_unlock(&sess->lock);
+		return;
+	}
+	err = update_sess_info(sess);
+	if (unlikely(err))
+		goto out;
+	list_for_each_entry(dev, &sess->devs_list, list) {
+		INFO(dev, "session reconnected, remapping device\n");
+		open_remote_device(dev);
+	}
+out:
+	mutex_unlock(&sess->lock);
+
+	ibnbd_clt_put_sess(sess);
+}
+
+static int ibnbd_schedule_reopen(struct ibnbd_session *sess)
+{
+	struct ibnbd_work *w;
+
+	w = kmalloc(sizeof(*w), GFP_KERNEL);
+	if (!w) {
+		ERR_NP("Failed to allocate memory to schedule reopen of"
+		       " devices for session %s\n", sess->str_addr);
+		return -ENOMEM;
+	}
+
+	if (WARN_ON(!ibnbd_clt_get_sess(sess))) {
+		kfree(w);
+		return -ENOENT;
+	}
+
+	w->sess = sess;
+	INIT_WORK(&w->work, reopen_worker);
+	schedule_work(&w->work);
+
+	return 0;
+}
+
+static void ibnbd_clt_sess_ev(void *priv, enum ibtrs_clt_sess_ev ev, int errno)
+{
+	struct ibnbd_session *sess = priv;
+	struct ibtrs_attrs attrs;
+
+	switch (ev) {
+	case IBTRS_CLT_SESS_EV_DISCONNECTED:
+		if (sess->sess_info_compl)
+			complete(sess->sess_info_compl);
+		mutex_lock(&sess->lock);
+		if (sess->state == SESS_STATE_DESTROYED) {
+			mutex_unlock(&sess->lock);
+			return;
+		}
+		sess->state = SESS_STATE_DISCONNECTED;
+		__set_dev_states_closed(sess);
+		mutex_unlock(&sess->lock);
+		break;
+	case IBTRS_CLT_SESS_EV_RECONNECT:
+		mutex_lock(&sess->lock);
+		if (sess->state == SESS_STATE_DESTROYED) {
+			/* This may happen if the session started to be closed
+			 * before the reconnect event arrived. In this case, we
+			 * just return and the session will be closed later
+			 */
+			mutex_unlock(&sess->lock);
+			return;
+		}
+		sess->state = SESS_STATE_READY;
+
+		mutex_unlock(&sess->lock);
+		memset(&attrs, 0, sizeof(attrs));
+		ibtrs_clt_query(sess->sess, &attrs);
+		strlcpy(sess->hostname, attrs.hostname, sizeof(sess->hostname));
+		sess->max_io_size = attrs.max_io_size;
+		ibnbd_schedule_reopen(sess);
+		break;
+	case IBTRS_CLT_SESS_EV_MAX_RECONN_EXCEEDED:
+		INFO_NP("Reconnect attempts exceeded for session %s\n",
+			sess->str_addr);
+		break;
+	default:
+		ERR_NP("Unknown session event received (%d), session: (%s)\n",
+		       ev, sess->str_addr);
+	}
+}
+
+static int ibnbd_cmp_sock_addr(const struct sockaddr_storage *a,
+			       const struct sockaddr_storage *b)
+{
+	if (a->ss_family == b->ss_family) {
+		switch (a->ss_family) {
+		case AF_INET:
+			return memcmp(&((struct sockaddr_in *)a)->sin_addr,
+				      &((struct sockaddr_in *)b)->sin_addr,
+				      sizeof(struct in_addr));
+		case AF_INET6:
+			return memcmp(&((struct sockaddr_in6 *)a)->sin6_addr,
+				      &((struct sockaddr_in6 *)b)->sin6_addr,
+				      sizeof(struct in6_addr));
+		case AF_IB:
+			return memcmp(&((struct sockaddr_ib *)a)->sib_addr,
+				      &((struct sockaddr_ib *)b)->sib_addr,
+				      sizeof(struct ib_addr));
+		default:
+			ERR_NP("Unknown address family: %d\n", a->ss_family);
+			return -EINVAL;
+		}
+	} else {
+		return -1;
+	}
+}
+
+struct ibnbd_session *ibnbd_clt_find_sess(const struct sockaddr_storage *addr)
+{
+	struct ibnbd_session *sess;
+
+	spin_lock(&sess_lock);
+	list_for_each_entry(sess, &session_list, list)
+		if (!ibnbd_cmp_sock_addr(&sess->addr, addr)) {
+			spin_unlock(&sess_lock);
+			return sess;
+		}
+	spin_unlock(&sess_lock);
+
+	return NULL;
+}
+
+static void ibnbd_init_cpu_qlists(struct ibnbd_cpu_qlist __percpu *cpu_queues)
+{
+	unsigned int cpu;
+	struct ibnbd_cpu_qlist *cpu_q;
+
+	for_each_online_cpu(cpu) {
+		cpu_q = per_cpu_ptr(cpu_queues, cpu);
+
+		cpu_q->cpu = cpu;
+		INIT_LIST_HEAD(&cpu_q->requeue_list);
+		spin_lock_init(&cpu_q->requeue_lock);
+	}
+}
+
+static struct blk_mq_ops ibnbd_mq_ops;
+static int setup_mq_tags(struct ibnbd_session *sess)
+{
+	struct blk_mq_tag_set *tags = &sess->tag_set;
+
+	memset(tags, 0, sizeof(*tags));
+	tags->ops		= &ibnbd_mq_ops;
+	tags->queue_depth	= sess->queue_depth;
+	tags->numa_node		= NUMA_NO_NODE;
+	tags->flags		= BLK_MQ_F_SHOULD_MERGE |
+				  BLK_MQ_F_SG_MERGE     |
+				  BLK_MQ_F_TAG_SHARED;
+	tags->cmd_size		= sizeof(struct ibnbd_iu);
+	tags->nr_hw_queues	= num_online_cpus();
+
+	return blk_mq_alloc_tag_set(tags);
+}
+
+static void destroy_mq_tags(struct ibnbd_session *sess)
+{
+	blk_mq_free_tag_set(&sess->tag_set);
+}
+
+struct ibnbd_session *ibnbd_create_session(const struct sockaddr_storage *addr)
+{
+	struct ibnbd_session *sess;
+	struct ibtrs_attrs attrs;
+	char str_addr[IBTRS_ADDRLEN];
+	int err;
+	int cpu;
+
+	err = ibtrs_addr_to_str(addr, str_addr, sizeof(str_addr));
+	if (err < 0) {
+		ERR_NP("Can't create session, invalid address\n");
+		return ERR_PTR(err);
+	}
+
+	DEB("Establishing session to %s\n", str_addr);
+
+	if (ibnbd_clt_find_sess(addr)) {
+		ERR_NP("Can't create session, session to %s already exists\n",
+		       str_addr);
+		return ERR_PTR(-EEXIST);
+	}
+
+	sess = kzalloc_node(sizeof(*sess), GFP_KERNEL, NUMA_NO_NODE);
+	if (unlikely(!sess)) {
+		ERR_NP("Failed to create session to %s,"
+		       " allocating session struct failed\n", str_addr);
+		return ERR_PTR(-ENOMEM);
+	}
+	sess->cpu_queues = alloc_percpu(struct ibnbd_cpu_qlist);
+	if (unlikely(!sess->cpu_queues)) {
+		ERR_NP("Failed to create session to %s,"
+		       " alloc of percpu var (cpu_queues) failed\n", str_addr);
+		kvfree(sess);
+		return ERR_PTR(-ENOMEM);
+	}
+	ibnbd_init_cpu_qlists(sess->cpu_queues);
+
+	/**
+	 * That is simple percpu variable which stores cpu indeces, which are
+	 * incremented on each access.  We need that for the sake of fairness
+	 * to wake up queues in a round-robin manner.
+	 */
+	sess->cpu_rr = alloc_percpu(int);
+	if (unlikely(!sess->cpu_rr)) {
+		ERR_NP("Failed to create session to %s,"
+		       " alloc of percpu var (cpu_rr) failed\n", str_addr);
+		free_percpu(sess->cpu_queues);
+		kfree(sess);
+		return ERR_PTR(-ENOMEM);
+	}
+	for_each_online_cpu(cpu) {
+		*per_cpu_ptr(sess->cpu_rr, cpu) = -1;
+	}
+
+	memset(&attrs, 0, sizeof(attrs));
+	memcpy(&sess->addr, addr, sizeof(sess->addr));
+	strlcpy(sess->str_addr, str_addr, sizeof(sess->str_addr));
+
+	spin_lock(&sess_lock);
+	list_add(&sess->list, &session_list);
+	spin_unlock(&sess_lock);
+
+	atomic_set(&sess->busy, 0);
+	mutex_init(&sess->lock);
+	INIT_LIST_HEAD(&sess->devs_list);
+	bitmap_zero(sess->cpu_queues_bm, NR_CPUS);
+	kref_init(&sess->refcount);
+	sess->state = SESS_STATE_DISCONNECTED;
+
+	sess->sess = ibtrs_clt_open(addr, sizeof(struct ibnbd_iu), sess,
+				    RECONNECT_DELAY, BMAX_SEGMENTS,
+				    MAX_RECONNECTS);
+	if (!IS_ERR(sess->sess)) {
+		mutex_lock(&sess->lock);
+		sess->state = SESS_STATE_READY;
+		mutex_unlock(&sess->lock);
+	} else {
+		err = PTR_ERR(sess->sess);
+		goto out_free;
+	}
+
+	ibtrs_clt_query(sess->sess, &attrs);
+	strlcpy(sess->hostname, attrs.hostname, sizeof(sess->hostname));
+	sess->max_io_size = attrs.max_io_size;
+	sess->queue_depth = attrs.queue_depth;
+
+	err = setup_mq_tags(sess);
+	if (unlikely(err))
+		goto close_sess;
+
+	err = update_sess_info(sess);
+	if (unlikely(err))
+		goto destroy_tags;
+
+	return sess;
+
+destroy_tags:
+	destroy_mq_tags(sess);
+close_sess:
+	ibtrs_clt_close(sess->sess);
+out_free:
+	spin_lock(&sess_lock);
+	list_del(&sess->list);
+	spin_unlock(&sess_lock);
+	free_percpu(sess->cpu_queues);
+	free_percpu(sess->cpu_rr);
+	kfree(sess);
+	return ERR_PTR(err);
+}
+
+static void ibnbd_clt_destroy_session(struct ibnbd_session *sess)
+{
+	mutex_lock(&sess->lock);
+	sess->state = SESS_STATE_DESTROYED;
+
+	if (!list_empty(&sess->devs_list)) {
+		mutex_unlock(&sess->lock);
+		WRN_NP("Device list is not empty,"
+		       " closing session to %s failed\n", sess->str_addr);
+		return;
+	}
+	mutex_unlock(&sess->lock);
+	ibtrs_clt_close(sess->sess);
+
+	destroy_mq_tags(sess);
+	spin_lock(&sess_lock);
+	list_del(&sess->list);
+	spin_unlock(&sess_lock);
+	wake_up(&sess_list_waitq);
+
+	free_percpu(sess->cpu_queues);
+	free_percpu(sess->cpu_rr);
+	kfree(sess);
+}
+
+void ibnbd_clt_sess_release(struct kref *ref)
+{
+	struct ibnbd_session *sess = container_of(ref, struct ibnbd_session,
+						  refcount);
+
+	ibnbd_clt_destroy_session(sess);
+}
+
+static int ibnbd_client_open(struct block_device *block_device, fmode_t mode)
+{
+	struct ibnbd_dev *dev = block_device->bd_disk->private_data;
+
+	if (dev->read_only && (mode & FMODE_WRITE))
+		return -EPERM;
+
+	if (dev->dev_state == DEV_STATE_UNMAPPED ||
+	    !ibnbd_clt_get_dev(dev))
+		return -EIO;
+
+	DEB("OPEN, name=%s, open_cnt=%d\n", dev->gd->disk_name,
+	    atomic_read(&dev->refcount) - 1);
+
+	return 0;
+}
+
+static void ibnbd_client_release(struct gendisk *gen, fmode_t mode)
+{
+	struct ibnbd_dev *dev = gen->private_data;
+
+	DEB("RELEASE, name=%s, open_cnt %d\n", dev->gd->disk_name,
+	    atomic_read(&dev->refcount) - 1);
+
+	ibnbd_clt_put_dev(dev);
+}
+
+static int ibnbd_client_getgeo(struct block_device *block_device,
+			       struct hd_geometry *geo)
+{
+	u64 size;
+	struct ibnbd_dev *dev;
+
+	dev = block_device->bd_disk->private_data;
+	size = dev->size * (dev->logical_block_size / KERNEL_SECTOR_SIZE);
+	geo->cylinders	= (size & ~0x3f) >> 6;	/* size/64 */
+	geo->heads	= 4;
+	geo->sectors	= 16;
+	geo->start	= 0;
+
+	return 0;
+}
+
+static const struct block_device_operations ibnbd_client_ops = {
+	.owner		= THIS_MODULE,
+	.open		= ibnbd_client_open,
+	.release	= ibnbd_client_release,
+	.getgeo		= ibnbd_client_getgeo
+};
+
+static size_t ibnbd_clt_get_sg_size(struct scatterlist *sglist, u32 len)
+{
+	struct scatterlist *sg;
+	size_t tsize = 0;
+	int i;
+
+	for_each_sg(sglist, sg, len, i)
+		tsize += sg->length;
+	return tsize;
+}
+
+static inline int ibnbd_clt_setup_discard(struct request *rq)
+{
+	struct page *page;
+
+	page = alloc_page(GFP_ATOMIC | __GFP_ZERO);
+	if (!page)
+		return -ENOMEM;
+	rq->special_vec.bv_page = page;
+	rq->special_vec.bv_offset = 0;
+	rq->special_vec.bv_len = PAGE_SIZE;
+	rq->rq_flags |= RQF_SPECIAL_PAYLOAD;
+	return 0;
+}
+
+static int ibnbd_client_xfer_request(struct ibnbd_dev *dev, struct request *rq,
+				     struct ibnbd_iu *iu)
+{
+	int err;
+
+	unsigned int sg_cnt;
+	size_t size;
+	struct kvec vec;
+	struct ibtrs_session *sess = dev->sess->sess;
+	struct ibtrs_tag *tag = iu->tag;
+	struct scatterlist *sg = iu->sglist;
+
+	if (req_op(rq) == REQ_OP_DISCARD) {
+		err = ibnbd_clt_setup_discard(rq);
+		if (err)
+			return err;
+	}
+	sg_cnt = blk_rq_nr_phys_segments(rq);
+
+	if (sg_cnt == 0)
+		sg_mark_end(&sg[0]);
+	else
+		sg_mark_end(&sg[sg_cnt - 1]);
+
+	iu->rq		= rq;
+	iu->dev		= dev;
+	iu->msg.sector	= blk_rq_pos(rq);
+	iu->msg.bi_size = blk_rq_bytes(rq);
+	iu->msg.rw	= rq_cmd_to_ibnbd_io_flags(rq);
+
+	sg_cnt = blk_rq_map_sg(dev->queue, rq, sg);
+
+	iu->msg.hdr.type	= IBNBD_MSG_IO;
+	iu->msg.device_id	= dev->device_id;
+
+	size = ibnbd_clt_get_sg_size(sg, sg_cnt);
+	vec = (struct kvec) {
+		.iov_base = &iu->msg,
+		.iov_len  = sizeof(iu->msg)
+	};
+
+	if (req_op(rq) == READ)
+		err = ibtrs_clt_request_rdma_write(sess, tag, iu, &vec, 1, size,
+						   sg, sg_cnt);
+	else
+		err = ibtrs_clt_rdma_write(sess, tag, iu, &vec, 1, size, sg,
+					   sg_cnt);
+	if (unlikely(err)) {
+		ERR_RL(dev, "IBTRS failed to transfer IO, errno: %d\n", err);
+		return err;
+	}
+
+	return 0;
+}
+
+/**
+ * ibnbd_dev_add_to_requeue() - add device to requeue if session is busy
+ *
+ * Description:
+ *     If session is busy, that means someone will requeue us when resources
+ *     are freed.  If session is not doing anything - device is not added to
+ *     the list and @false is returned.
+ */
+static inline bool ibnbd_dev_add_to_requeue(struct ibnbd_dev *dev,
+					    struct ibnbd_queue *q)
+{
+	struct ibnbd_session *sess = dev->sess;
+	struct ibnbd_cpu_qlist *cpu_q;
+	unsigned long flags;
+	bool added = true;
+	bool need_set;
+
+	cpu_q = get_cpu_ptr(sess->cpu_queues);
+	spin_lock_irqsave(&cpu_q->requeue_lock, flags);
+
+	if (likely(!test_and_set_bit_lock(0, &q->in_list))) {
+		if (WARN_ON(!list_empty(&q->requeue_list)))
+			goto unlock;
+
+		need_set = !test_bit(cpu_q->cpu, sess->cpu_queues_bm);
+		if (need_set) {
+			set_bit(cpu_q->cpu, sess->cpu_queues_bm);
+			/* Paired with ibnbd_put_tag().	 Set a bit first
+			 * and then observe the busy counter.
+			 */
+			smp_mb__before_atomic();
+		}
+		if (likely(atomic_read(&sess->busy))) {
+			list_add_tail(&q->requeue_list, &cpu_q->requeue_list);
+		} else {
+			/* Very unlikely, but possible: busy counter was
+			 * observed as zero.  Drop all bits and return
+			 * false to restart the queue by ourselves.
+			 */
+			if (need_set)
+				clear_bit(cpu_q->cpu, sess->cpu_queues_bm);
+			clear_bit_unlock(0, &q->in_list);
+			added = false;
+		}
+	}
+unlock:
+	spin_unlock_irqrestore(&cpu_q->requeue_lock, flags);
+	put_cpu_ptr(sess->cpu_queues);
+
+	return added;
+}
+
+static void ibnbd_dev_kick_mq_queue(struct ibnbd_dev *dev,
+				    struct blk_mq_hw_ctx *hctx,
+				    int delay)
+{
+	struct ibnbd_queue *q = hctx->driver_data;
+
+	if (WARN_ON(dev->queue_mode != BLK_MQ))
+		return;
+	blk_mq_stop_hw_queue(hctx);
+
+	if (delay != IBNBD_DELAY_IFBUSY)
+		blk_mq_delay_queue(hctx, delay);
+	else if (unlikely(!ibnbd_dev_add_to_requeue(dev, q)))
+		/* If session is not busy we have to restart
+		 * the queue ourselves.
+		 */
+		blk_mq_delay_queue(hctx, IBNBD_DELAY_10ms);
+}
+
+static void ibnbd_dev_kick_queue(struct ibnbd_dev *dev, int delay)
+{
+	if (WARN_ON(dev->queue_mode != BLK_RQ))
+		return;
+	blk_stop_queue(dev->queue);
+
+	if (delay != IBNBD_DELAY_IFBUSY)
+		ibnbd_blk_delay_queue(dev, delay);
+	else if (unlikely(!ibnbd_dev_add_to_requeue(dev, dev->hw_queues)))
+		/* If session is not busy we have to restart
+		 * the queue ourselves.
+		 */
+		ibnbd_blk_delay_queue(dev, IBNBD_DELAY_10ms);
+}
+
+static int ibnbd_queue_rq(struct blk_mq_hw_ctx *hctx,
+			  const struct blk_mq_queue_data *bd)
+{
+	struct request *rq = bd->rq;
+	struct ibnbd_dev *dev = rq->rq_disk->private_data;
+	struct ibnbd_iu *iu = blk_mq_rq_to_pdu(rq);
+	int err;
+
+	if (unlikely(!ibnbd_clt_dev_is_open(dev)))
+		return BLK_MQ_RQ_QUEUE_ERROR;
+
+	iu->tag = ibnbd_get_tag(dev->sess, hctx->next_cpu, blk_rq_bytes(rq),
+				IBTRS_TAG_NOWAIT);
+	if (unlikely(!iu->tag)) {
+		ibnbd_dev_kick_mq_queue(dev, hctx, IBNBD_DELAY_IFBUSY);
+		return BLK_MQ_RQ_QUEUE_BUSY;
+	}
+
+	blk_mq_start_request(rq);
+	err = ibnbd_client_xfer_request(dev, rq, iu);
+	if (likely(err == 0))
+		return BLK_MQ_RQ_QUEUE_OK;
+	if (unlikely(err == -EAGAIN || err == -ENOMEM)) {
+		ibnbd_dev_kick_mq_queue(dev, hctx, IBNBD_DELAY_10ms);
+		ibnbd_put_tag(dev->sess, iu->tag);
+		return BLK_MQ_RQ_QUEUE_BUSY;
+	}
+
+	ibnbd_put_tag(dev->sess, iu->tag);
+	return BLK_MQ_RQ_QUEUE_ERROR;
+}
+
+static int ibnbd_init_request(void *data, struct request *rq,
+			      unsigned int hctx_idx, unsigned int request_idx,
+			      unsigned int numa_node)
+{
+	struct ibnbd_iu *iu = blk_mq_rq_to_pdu(rq);
+
+	sg_init_table(iu->sglist, BMAX_SEGMENTS);
+	return 0;
+}
+
+static inline void ibnbd_init_hw_queue(struct ibnbd_dev *dev,
+				       struct ibnbd_queue *q,
+				       struct blk_mq_hw_ctx *hctx)
+{
+	INIT_LIST_HEAD(&q->requeue_list);
+	q->dev  = dev;
+	q->hctx = hctx;
+}
+
+static void ibnbd_init_mq_hw_queues(struct ibnbd_dev *dev)
+{
+	int i;
+	struct blk_mq_hw_ctx *hctx;
+	struct ibnbd_queue *q;
+
+	queue_for_each_hw_ctx(dev->queue, hctx, i) {
+		q = &dev->hw_queues[i];
+		ibnbd_init_hw_queue(dev, q, hctx);
+		hctx->driver_data = q;
+	}
+}
+
+static struct blk_mq_ops ibnbd_mq_ops = {
+	.queue_rq	= ibnbd_queue_rq,
+	.init_request	= ibnbd_init_request,
+	.complete	= ibnbd_softirq_done_fn,
+};
+
+static int index_to_minor(int index)
+{
+	return index << IBNBD_PART_BITS;
+}
+
+static int minor_to_index(int minor)
+{
+	return minor >> IBNBD_PART_BITS;
+}
+
+static int ibnbd_rq_prep_fn(struct request_queue *q, struct request *rq)
+{
+	struct ibnbd_dev *dev = q->queuedata;
+	struct ibnbd_iu *iu;
+
+	iu = ibnbd_get_iu(dev->sess, blk_rq_bytes(rq), IBTRS_TAG_NOWAIT);
+	if (likely(iu)) {
+		rq->special = iu;
+		rq->rq_flags |= RQF_DONTPREP;
+
+		return BLKPREP_OK;
+	}
+
+	ibnbd_dev_kick_queue(dev, IBNBD_DELAY_IFBUSY);
+	return BLKPREP_DEFER;
+}
+
+static void ibnbd_rq_unprep_fn(struct request_queue *q, struct request *rq)
+{
+	struct ibnbd_dev *dev = q->queuedata;
+
+	if (WARN_ON(!rq->special))
+		return;
+	ibnbd_put_iu(dev->sess, rq->special);
+	rq->special = NULL;
+	rq->rq_flags &= ~RQF_DONTPREP;
+}
+
+static void ibnbd_clt_request(struct request_queue *q)
+__must_hold(q->queue_lock)
+{
+	int err;
+	struct request *req;
+	struct ibnbd_iu *iu;
+	struct ibnbd_dev *dev = q->queuedata;
+
+	while ((req = blk_fetch_request(q)) != NULL) {
+		spin_unlock_irq(q->queue_lock);
+
+		if (unlikely(!ibnbd_clt_dev_is_open(dev))) {
+			err = -EIO;
+			goto next;
+		}
+
+		iu = req->special;
+		if (WARN_ON(!iu)) {
+			err = -EIO;
+			goto next;
+		}
+
+		sg_init_table(iu->sglist, dev->max_segments);
+		err = ibnbd_client_xfer_request(dev, req, iu);
+next:
+		if (unlikely(err == -EAGAIN || err == -ENOMEM)) {
+			ibnbd_rq_unprep_fn(q, req);
+			spin_lock_irq(q->queue_lock);
+			blk_requeue_request(q, req);
+			ibnbd_dev_kick_queue(dev, IBNBD_DELAY_10ms);
+			break;
+		} else if (err) {
+			blk_end_request_all(req, err);
+		}
+
+		spin_lock_irq(q->queue_lock);
+	}
+}
+
+static int setup_mq_dev(struct ibnbd_dev *dev)
+{
+	dev->queue = blk_mq_init_queue(&dev->sess->tag_set);
+	if (IS_ERR(dev->queue)) {
+		ERR(dev, "Initializing multiqueue queue failed, errno: %ld\n",
+		    PTR_ERR(dev->queue));
+		return PTR_ERR(dev->queue);
+	}
+	ibnbd_init_mq_hw_queues(dev);
+	return 0;
+}
+
+static int setup_rq_dev(struct ibnbd_dev *dev)
+{
+	dev->queue = blk_init_queue(ibnbd_clt_request, NULL);
+	if (IS_ERR_OR_NULL(dev->queue)) {
+		if (IS_ERR(dev->queue)) {
+			ERR(dev, "Initializing request queue failed, "
+			    "errno: %ld\n", PTR_ERR(dev->queue));
+			return PTR_ERR(dev->queue);
+		}
+		ERR(dev, "Initializing request queue failed\n");
+		return -ENOMEM;
+	}
+
+	blk_queue_prep_rq(dev->queue, ibnbd_rq_prep_fn);
+	blk_queue_softirq_done(dev->queue, ibnbd_softirq_done_fn);
+	blk_queue_unprep_rq(dev->queue, ibnbd_rq_unprep_fn);
+
+	return 0;
+}
+
+static void setup_request_queue(struct ibnbd_dev *dev)
+{
+	blk_queue_logical_block_size(dev->queue, dev->logical_block_size);
+	blk_queue_physical_block_size(dev->queue, dev->physical_block_size);
+	blk_queue_max_hw_sectors(dev->queue, dev->max_hw_sectors);
+	blk_queue_max_write_same_sectors(dev->queue,
+					 dev->max_write_same_sectors);
+
+	blk_queue_max_discard_sectors(dev->queue, dev->max_discard_sectors);
+	dev->queue->limits.discard_zeroes_data	= dev->discard_zeroes_data;
+	dev->queue->limits.discard_granularity	= dev->discard_granularity;
+	dev->queue->limits.discard_alignment	= dev->discard_alignment;
+	if (dev->max_discard_sectors)
+		queue_flag_set_unlocked(QUEUE_FLAG_DISCARD, dev->queue);
+	if (dev->secure_discard)
+		queue_flag_set_unlocked(QUEUE_FLAG_SECERASE, dev->queue);
+
+	queue_flag_set_unlocked(QUEUE_FLAG_SAME_COMP, dev->queue);
+	queue_flag_set_unlocked(QUEUE_FLAG_SAME_FORCE, dev->queue);
+	/* our hca only support 32 sg cnt, proto use one, so 31 left */
+	blk_queue_max_segments(dev->queue, dev->max_segments);
+	blk_queue_io_opt(dev->queue, dev->sess->max_io_size);
+	blk_queue_write_cache(dev->queue, true, true);
+	dev->queue->queuedata = dev;
+}
+
+static void ibnbd_clt_setup_gen_disk(struct ibnbd_dev *dev, int idx)
+{
+	dev->gd->major		= ibnbd_client_major;
+	dev->gd->first_minor	= index_to_minor(idx);
+	dev->gd->fops		= &ibnbd_client_ops;
+	dev->gd->queue		= dev->queue;
+	dev->gd->private_data	= dev;
+	snprintf(dev->gd->disk_name, sizeof(dev->gd->disk_name), "ibnbd%d",
+		 idx);
+	DEB("disk_name=%s, capacity=%zu, queue_mode=%s\n", dev->gd->disk_name,
+	    dev->nsectors * (dev->logical_block_size / KERNEL_SECTOR_SIZE),
+	    ibnbd_queue_mode_str(dev->queue_mode));
+
+	set_capacity(dev->gd, dev->nsectors * (dev->logical_block_size /
+					       KERNEL_SECTOR_SIZE));
+
+	if (dev->access_mode == IBNBD_ACCESS_RO) {
+		dev->read_only = true;
+		set_disk_ro(dev->gd, true);
+	} else {
+		dev->read_only = false;
+	}
+
+	if (!dev->rotational)
+		queue_flag_set_unlocked(QUEUE_FLAG_NONROT, dev->queue);
+}
+
+static void ibnbd_clt_add_gen_disk(struct ibnbd_dev *dev)
+{
+	add_disk(dev->gd);
+}
+
+static int ibnbd_client_setup_device(struct ibnbd_session *sess,
+				     struct ibnbd_dev *dev, int idx)
+{
+	int err;
+
+	dev->size = dev->nsectors * dev->logical_block_size;
+
+	switch (dev->queue_mode) {
+	case BLK_MQ:
+		err = setup_mq_dev(dev);
+		break;
+	case BLK_RQ:
+		err = setup_rq_dev(dev);
+		break;
+	default:
+		err = -EINVAL;
+	}
+
+	if (err)
+		return err;
+
+	setup_request_queue(dev);
+
+	dev->gd = alloc_disk_node(1 << IBNBD_PART_BITS,	NUMA_NO_NODE);
+	if (!dev->gd) {
+		ERR(dev, "Failed to allocate disk node\n");
+		blk_cleanup_queue(dev->queue);
+		return -ENOMEM;
+	}
+
+	ibnbd_clt_setup_gen_disk(dev, idx);
+
+	return 0;
+}
+
+static struct ibnbd_dev *init_dev(struct ibnbd_session *sess,
+				  enum ibnbd_access_mode access_mode,
+				  enum ibnbd_queue_mode queue_mode,
+				  const char *pathname)
+{
+	int ret;
+	struct ibnbd_dev *dev;
+	size_t nr;
+
+	dev = kzalloc_node(sizeof(*dev), GFP_KERNEL, NUMA_NO_NODE);
+	if (!dev) {
+		ERR_NP("Failed to initialize device '%s' from session %s,"
+		       " allocating device structure failed\n", pathname,
+		       sess->str_addr);
+		return ERR_PTR(-ENOMEM);
+	}
+
+	nr = (queue_mode == BLK_MQ ? num_online_cpus() :
+	      queue_mode == BLK_RQ ? 1 : 0);
+	if (nr) {
+		dev->hw_queues = kcalloc(nr, sizeof(*dev->hw_queues),
+					 GFP_KERNEL);
+		if (unlikely(!dev->hw_queues)) {
+			ERR_NP("Failed to initialize device '%s' from session"
+			       " %s, allocating hw_queues failed.", pathname,
+			       sess->str_addr);
+			ret = -ENOMEM;
+			goto out_alloc;
+		}
+		/* for MQ mode we will init all hw queues after the
+		 * request queue is created
+		 */
+		if (queue_mode == BLK_RQ)
+			ibnbd_init_hw_queue(dev, dev->hw_queues, NULL);
+	}
+
+	idr_preload(GFP_KERNEL);
+	write_lock(&g_index_lock);
+	ret = idr_alloc(&g_index_idr, dev, 0, minor_to_index(1 << MINORBITS),
+			GFP_ATOMIC);
+	write_unlock(&g_index_lock);
+	idr_preload_end();
+	if (ret < 0) {
+		ERR_NP("Failed to initialize device '%s' from session %s,"
+		       " allocating idr failed, errno: %d\n", pathname,
+		       sess->str_addr, ret);
+		goto out_queues;
+	}
+
+	dev->clt_device_id	= ret;
+	dev->close_compl	= kmalloc(sizeof(*dev->close_compl),
+					  GFP_KERNEL);
+	if (!dev->close_compl) {
+		ERR_NP("Failed to initialize device '%s' from session %s,"
+		       " allocating close completion failed, errno: %d\n",
+		       pathname, sess->str_addr, ret);
+		ret = -ENOMEM;
+		goto out_idr;
+	}
+	init_completion(dev->close_compl);
+	dev->sess		= sess;
+	dev->access_mode	= access_mode;
+	dev->queue_mode		= queue_mode;
+
+	strlcpy(dev->pathname, pathname, sizeof(dev->pathname));
+
+	INIT_DELAYED_WORK(&dev->rq_delay_work, ibnbd_blk_delay_work);
+	mutex_init(&dev->lock);
+	atomic_set(&dev->refcount, 1);
+	dev->dev_state = DEV_STATE_INIT;
+
+	return dev;
+
+out_idr:
+	write_lock(&g_index_lock);
+	idr_remove(&g_index_idr, dev->clt_device_id);
+	write_unlock(&g_index_lock);
+out_queues:
+	kfree(dev->hw_queues);
+out_alloc:
+	kfree(dev);
+	return ERR_PTR(ret);
+}
+
+bool ibnbd_clt_dev_is_mapped(const char *pathname)
+{
+	struct ibnbd_dev *dev;
+
+	spin_lock(&dev_lock);
+	list_for_each_entry(dev, &devs_list, g_list)
+		if (!strncmp(dev->pathname, pathname, sizeof(dev->pathname))) {
+			spin_unlock(&dev_lock);
+			return true;
+		}
+	spin_unlock(&dev_lock);
+
+	return false;
+}
+
+static struct ibnbd_dev *__find_sess_dev(const struct ibnbd_session *sess,
+					 const char *pathname)
+{
+	struct ibnbd_dev *dev;
+
+	list_for_each_entry(dev, &sess->devs_list, list)
+		if (!strncmp(dev->pathname, pathname, sizeof(dev->pathname)))
+			return dev;
+
+	return NULL;
+}
+
+struct ibnbd_dev *ibnbd_client_add_device(struct ibnbd_session *sess,
+					  const char *pathname,
+					  enum ibnbd_access_mode access_mode,
+					  enum ibnbd_queue_mode queue_mode,
+					  enum ibnbd_io_mode io_mode)
+{
+	int ret;
+	struct ibnbd_dev *dev;
+	struct completion *open_compl;
+
+	DEB("Add remote device: server=%s, path='%s', access_mode=%d,"
+	    " queue_mode=%d\n", sess->str_addr, pathname, access_mode,
+	    queue_mode);
+
+	mutex_lock(&sess->lock);
+
+	if (sess->state != SESS_STATE_READY) {
+		mutex_unlock(&sess->lock);
+		ERR_NP("map_device: failed to map device '%s' from session %s,"
+		       " session is not connected\n", pathname, sess->str_addr);
+		return ERR_PTR(-ENOENT);
+	}
+
+	if (__find_sess_dev(sess, pathname)) {
+		mutex_unlock(&sess->lock);
+		ERR_NP("map_device: failed to map device '%s' from session %s,"
+		       " device with same path is already mapped\n", pathname,
+		       sess->str_addr);
+		return ERR_PTR(-EEXIST);
+	}
+
+	mutex_unlock(&sess->lock);
+	dev = init_dev(sess, access_mode, queue_mode, pathname);
+	if (IS_ERR(dev)) {
+		ERR_NP("map_device: failed to map device '%s' from session %s,"
+		       " can't initialize device, errno: %ld\n", pathname,
+		       sess->str_addr, PTR_ERR(dev));
+		return dev;
+	}
+
+	ibnbd_clt_get_sess(sess);
+
+	open_compl = kmalloc(sizeof(*open_compl), GFP_KERNEL);
+	if (!open_compl) {
+		ERR(dev, "map_device: failed, Can't allocate memory for"
+		    " completion\n");
+		ret = -ENOMEM;
+		goto out;
+	}
+	init_completion(open_compl);
+	dev->open_compl = open_compl;
+	dev->io_mode = io_mode;
+
+	ret = open_remote_device(dev);
+	if (ret) {
+		ERR(dev, "map_device: failed, can't open remote device,"
+		    " errno: %d\n", ret);
+		kfree(open_compl);
+		dev->open_compl = NULL;
+		ret = -EINVAL;
+		goto out;
+	}
+	wait_for_completion(dev->open_compl);
+	mutex_lock(&dev->lock);
+
+	kfree(open_compl);
+	dev->open_compl = NULL;
+
+	if (!ibnbd_clt_dev_is_open(dev)) {
+		mutex_unlock(&dev->lock);
+		ret = dev->open_errno;
+		ERR(dev, "map_device: failed errno: %d\n", ret);
+		goto out;
+	}
+
+	mutex_lock(&sess->lock);
+	list_add(&dev->list, &sess->devs_list);
+	mutex_unlock(&sess->lock);
+
+	spin_lock(&dev_lock);
+	list_add(&dev->g_list, &devs_list);
+	spin_unlock(&dev_lock);
+
+	DEB("Opened remote device: server=%s, path='%s'\n", sess->str_addr,
+	    pathname);
+	ret = ibnbd_client_setup_device(sess, dev, dev->clt_device_id);
+	if (ret) {
+		ERR(dev, "map_device: Failed to configure device, errno: %d\n",
+		    ret);
+		mutex_unlock(&dev->lock);
+		ret = -EINVAL;
+		goto out_close;
+	}
+
+	INFO(dev, "map_device: Device mapped as %s (nsectors: %zu,"
+	     " logical_block_size: %d, physical_block_size: %d,"
+	     " max_write_same_sectors: %d, max_discard_sectors: %d,"
+	     " discard_zeroes_data: %d, discard_granularity: %d,"
+	     " discard_alignment: %d, secure_discard: %d, max_segments: %d,"
+	     " max_hw_sectors: %d, rotational: %d)\n", dev->gd->disk_name,
+	     dev->nsectors, dev->logical_block_size, dev->physical_block_size,
+	     dev->max_write_same_sectors, dev->max_discard_sectors,
+	     dev->discard_zeroes_data, dev->discard_granularity,
+	     dev->discard_alignment, dev->secure_discard,
+	     dev->max_segments, dev->max_hw_sectors, dev->rotational);
+
+	mutex_unlock(&dev->lock);
+
+	ibnbd_clt_add_gen_disk(dev);
+
+	return dev;
+
+out_close:
+	if (!WARN_ON(ibnbd_close_device(dev, true)))
+		wait_for_completion(dev->close_compl);
+out:
+	ibnbd_clt_put_dev(dev);
+	return ERR_PTR(ret);
+}
+
+void ibnbd_destroy_gen_disk(struct ibnbd_dev *dev)
+{
+	del_gendisk(dev->gd);
+	/*
+	 * Before marking queue as dying (blk_cleanup_queue() does that)
+	 * we have to be sure that everything in-flight has gone.
+	 * Blink with freeze/unfreeze.
+	 */
+	blk_mq_freeze_queue(dev->queue);
+	blk_mq_unfreeze_queue(dev->queue);
+	blk_cleanup_queue(dev->queue);
+	put_disk(dev->gd);
+
+	ibnbd_clt_put_dev(dev);
+}
+
+static int __close_device(struct ibnbd_dev *dev, bool force)
+__must_hold(&dev->sess->lock)
+{
+	enum ibnbd_dev_state prev_state;
+	int refcount, ret = 0;
+
+	mutex_lock(&dev->lock);
+
+	if (dev->dev_state == DEV_STATE_UNMAPPED) {
+		INFO(dev, "Device is already being unmapped\n");
+		ret = -EALREADY;
+		goto out;
+	}
+
+	refcount = atomic_read(&dev->refcount);
+	if (!force && refcount > 1) {
+		ERR(dev, "Closing device failed, device is in use,"
+		    " (%d device users)\n", refcount - 1);
+		ret = -EBUSY;
+		goto out;
+	}
+
+	prev_state = dev->dev_state;
+	dev->dev_state = DEV_STATE_UNMAPPED;
+
+	list_del(&dev->list);
+
+	spin_lock(&dev_lock);
+	list_del(&dev->g_list);
+	spin_unlock(&dev_lock);
+
+	ibnbd_clt_remove_dev_symlink(dev);
+	mutex_unlock(&dev->lock);
+
+	mutex_unlock(&dev->sess->lock);
+	if (prev_state == DEV_STATE_OPEN && dev->sess->sess) {
+		if (send_msg_close(dev->sess->sess, dev->device_id))
+			complete(dev->close_compl);
+	} else {
+		complete(dev->close_compl);
+	}
+
+	mutex_lock(&dev->sess->lock);
+	if (dev->gd)
+		INFO(dev, "Device is unmapped\n");
+	else
+		INFO(dev, "Device is unmapped\n");
+	return 0;
+out:
+	mutex_unlock(&dev->lock);
+	return ret;
+}
+
+int ibnbd_close_device(struct ibnbd_dev *dev, bool force)
+{
+	int ret;
+
+	mutex_lock(&dev->sess->lock);
+	ret = __close_device(dev, force);
+	mutex_unlock(&dev->sess->lock);
+
+	return ret;
+}
+
+static void ibnbd_destroy_sessions(void)
+{
+	struct ibnbd_session *sess, *sn;
+	struct ibnbd_dev *dev, *tn;
+	int ret;
+
+	list_for_each_entry_safe(sess, sn, &session_list, list) {
+		if (!ibnbd_clt_get_sess(sess))
+			continue;
+		mutex_lock(&sess->lock);
+		sess->state = SESS_STATE_DESTROYED;
+		list_for_each_entry_safe(dev, tn, &sess->devs_list, list) {
+			if (!kobject_get(&dev->kobj))
+				continue;
+			ret = __close_device(dev, true);
+			if (ret)
+				WRN(dev, "Closing device failed, errno: %d\n",
+				    ret);
+			else
+				wait_for_completion(dev->close_compl);
+			ibnbd_clt_schedule_dev_destroy(dev);
+			kobject_put(&dev->kobj);
+		}
+		mutex_unlock(&sess->lock);
+		ibnbd_clt_put_sess(sess);
+	}
+}
+
+static int __init ibnbd_client_init(void)
+{
+	int err;
+
+	INFO_NP("Loading module ibnbd_client, version: "
+		__stringify(IBNBD_VER) " (softirq_enable: %d)\n",
+		softirq_enable);
+
+	ibnbd_client_major = register_blkdev(ibnbd_client_major, "ibnbd");
+	if (ibnbd_client_major <= 0) {
+		ERR_NP("Failed to load module,"
+		       " block device registration failed\n");
+		err = -EBUSY;
+		goto out;
+	}
+
+	ops.owner	= THIS_MODULE;
+	ops.recv	= ibnbd_clt_recv;
+	ops.rdma_ev	= ibnbd_clt_rdma_ev;
+	ops.sess_ev	= ibnbd_clt_sess_ev;
+	err = ibtrs_clt_register(&ops);
+	if (err) {
+		ERR_NP("Failed to load module, IBTRS registration failed,"
+		       " errno: %d\n", err);
+		goto out_unregister_blk;
+	}
+	err = ibnbd_clt_create_sysfs_files();
+	if (err) {
+		ERR_NP("Failed to load module,"
+		       " creating sysfs device files failed, error: %d\n", err);
+		goto out_unregister;
+	}
+
+	return 0;
+
+out_unregister:
+	ibtrs_clt_unregister(&ops);
+out_unregister_blk:
+	unregister_blkdev(ibnbd_client_major, "ibnbd");
+out:
+	return err;
+}
+
+static void __exit ibnbd_client_exit(void)
+{
+	INFO_NP("Unloading module\n");
+	ibnbd_clt_destroy_default_group();
+	flush_scheduled_work();
+	ibnbd_destroy_sessions();
+	wait_event(sess_list_waitq, list_empty(&session_list));
+	ibnbd_clt_destroy_sysfs_files();
+	ibtrs_clt_unregister(&ops);
+	unregister_blkdev(ibnbd_client_major, "ibnbd");
+	idr_destroy(&g_index_idr);
+	INFO_NP("Module unloaded\n");
+}
+
+module_init(ibnbd_client_init);
+module_exit(ibnbd_client_exit);