Merge branch 'rdma-cq.2' of git://git.infradead.org/users/hch/rdma into 4.5/rdma-cq

Signed-off-by: Doug Ledford <dledford@redhat.com>

Conflicts:
	drivers/infiniband/ulp/srp/ib_srp.c - Conflicts with changes in
	ib_srp.c introduced during 4.4-rc updates
diff --git a/drivers/infiniband/Kconfig b/drivers/infiniband/Kconfig
index aa26f3c..282ec0b 100644
--- a/drivers/infiniband/Kconfig
+++ b/drivers/infiniband/Kconfig
@@ -5,6 +5,7 @@
 	depends on NET
 	depends on INET
 	depends on m || IPV6 != m
+	select IRQ_POLL
 	---help---
 	  Core support for InfiniBand (IB).  Make sure to also select
 	  any protocols you wish to use as well as drivers for your
diff --git a/drivers/infiniband/core/Makefile b/drivers/infiniband/core/Makefile
index d43a899..ae48d8740 100644
--- a/drivers/infiniband/core/Makefile
+++ b/drivers/infiniband/core/Makefile
@@ -8,7 +8,7 @@
 obj-$(CONFIG_INFINIBAND_USER_ACCESS) +=	ib_uverbs.o ib_ucm.o \
 					$(user_access-y)
 
-ib_core-y :=			packer.o ud_header.o verbs.o sysfs.o \
+ib_core-y :=			packer.o ud_header.o verbs.o cq.o sysfs.o \
 				device.o fmr_pool.o cache.o netlink.o \
 				roce_gid_mgmt.o
 ib_core-$(CONFIG_INFINIBAND_USER_MEM) += umem.o
diff --git a/drivers/infiniband/core/cq.c b/drivers/infiniband/core/cq.c
new file mode 100644
index 0000000..a754fc7
--- /dev/null
+++ b/drivers/infiniband/core/cq.c
@@ -0,0 +1,209 @@
+/*
+ * Copyright (c) 2015 HGST, a Western Digital Company.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ */
+#include <linux/module.h>
+#include <linux/err.h>
+#include <linux/slab.h>
+#include <rdma/ib_verbs.h>
+
+/* # of WCs to poll for with a single call to ib_poll_cq */
+#define IB_POLL_BATCH			16
+
+/* # of WCs to iterate over before yielding */
+#define IB_POLL_BUDGET_IRQ		256
+#define IB_POLL_BUDGET_WORKQUEUE	65536
+
+#define IB_POLL_FLAGS \
+	(IB_CQ_NEXT_COMP | IB_CQ_REPORT_MISSED_EVENTS)
+
+static int __ib_process_cq(struct ib_cq *cq, int budget)
+{
+	int i, n, completed = 0;
+
+	while ((n = ib_poll_cq(cq, IB_POLL_BATCH, cq->wc)) > 0) {
+		for (i = 0; i < n; i++) {
+			struct ib_wc *wc = &cq->wc[i];
+
+			if (wc->wr_cqe)
+				wc->wr_cqe->done(cq, wc);
+			else
+				WARN_ON_ONCE(wc->status == IB_WC_SUCCESS);
+		}
+
+		completed += n;
+
+		if (n != IB_POLL_BATCH ||
+		    (budget != -1 && completed >= budget))
+			break;
+	}
+
+	return completed;
+}
+
+/**
+ * ib_process_direct_cq - process a CQ in caller context
+ * @cq:		CQ to process
+ * @budget:	number of CQEs to poll for
+ *
+ * This function is used to process all outstanding CQ entries on a
+ * %IB_POLL_DIRECT CQ.  It does not offload CQ processing to a different
+ * context and does not ask for completion interrupts from the HCA.
+ *
+ * Note: for compatibility reasons -1 can be passed in %budget for unlimited
+ * polling.  Do not use this feature in new code, it will be removed soon.
+ */
+int ib_process_cq_direct(struct ib_cq *cq, int budget)
+{
+	WARN_ON_ONCE(cq->poll_ctx != IB_POLL_DIRECT);
+
+	return __ib_process_cq(cq, budget);
+}
+EXPORT_SYMBOL(ib_process_cq_direct);
+
+static void ib_cq_completion_direct(struct ib_cq *cq, void *private)
+{
+	WARN_ONCE(1, "got unsolicited completion for CQ 0x%p\n", cq);
+}
+
+static int ib_poll_handler(struct irq_poll *iop, int budget)
+{
+	struct ib_cq *cq = container_of(iop, struct ib_cq, iop);
+	int completed;
+
+	completed = __ib_process_cq(cq, budget);
+	if (completed < budget) {
+		irq_poll_complete(&cq->iop);
+		if (ib_req_notify_cq(cq, IB_POLL_FLAGS) > 0)
+			irq_poll_sched(&cq->iop);
+	}
+
+	return completed;
+}
+
+static void ib_cq_completion_softirq(struct ib_cq *cq, void *private)
+{
+	irq_poll_sched(&cq->iop);
+}
+
+static void ib_cq_poll_work(struct work_struct *work)
+{
+	struct ib_cq *cq = container_of(work, struct ib_cq, work);
+	int completed;
+
+	completed = __ib_process_cq(cq, IB_POLL_BUDGET_WORKQUEUE);
+	if (completed >= IB_POLL_BUDGET_WORKQUEUE ||
+	    ib_req_notify_cq(cq, IB_POLL_FLAGS) > 0)
+		queue_work(ib_comp_wq, &cq->work);
+}
+
+static void ib_cq_completion_workqueue(struct ib_cq *cq, void *private)
+{
+	queue_work(ib_comp_wq, &cq->work);
+}
+
+/**
+ * ib_alloc_cq - allocate a completion queue
+ * @dev:		device to allocate the CQ for
+ * @private:		driver private data, accessible from cq->cq_context
+ * @nr_cqe:		number of CQEs to allocate
+ * @comp_vector:	HCA completion vectors for this CQ
+ * @poll_ctx:		context to poll the CQ from.
+ *
+ * This is the proper interface to allocate a CQ for in-kernel users. A
+ * CQ allocated with this interface will automatically be polled from the
+ * specified context.  The ULP needs must use wr->wr_cqe instead of wr->wr_id
+ * to use this CQ abstraction.
+ */
+struct ib_cq *ib_alloc_cq(struct ib_device *dev, void *private,
+		int nr_cqe, int comp_vector, enum ib_poll_context poll_ctx)
+{
+	struct ib_cq_init_attr cq_attr = {
+		.cqe		= nr_cqe,
+		.comp_vector	= comp_vector,
+	};
+	struct ib_cq *cq;
+	int ret = -ENOMEM;
+
+	cq = dev->create_cq(dev, &cq_attr, NULL, NULL);
+	if (IS_ERR(cq))
+		return cq;
+
+	cq->device = dev;
+	cq->uobject = NULL;
+	cq->event_handler = NULL;
+	cq->cq_context = private;
+	cq->poll_ctx = poll_ctx;
+	atomic_set(&cq->usecnt, 0);
+
+	cq->wc = kmalloc_array(IB_POLL_BATCH, sizeof(*cq->wc), GFP_KERNEL);
+	if (!cq->wc)
+		goto out_destroy_cq;
+
+	switch (cq->poll_ctx) {
+	case IB_POLL_DIRECT:
+		cq->comp_handler = ib_cq_completion_direct;
+		break;
+	case IB_POLL_SOFTIRQ:
+		cq->comp_handler = ib_cq_completion_softirq;
+
+		irq_poll_init(&cq->iop, IB_POLL_BUDGET_IRQ, ib_poll_handler);
+		ib_req_notify_cq(cq, IB_CQ_NEXT_COMP);
+		break;
+	case IB_POLL_WORKQUEUE:
+		cq->comp_handler = ib_cq_completion_workqueue;
+		INIT_WORK(&cq->work, ib_cq_poll_work);
+		ib_req_notify_cq(cq, IB_CQ_NEXT_COMP);
+		break;
+	default:
+		ret = -EINVAL;
+		goto out_free_wc;
+	}
+
+	return cq;
+
+out_free_wc:
+	kfree(cq->wc);
+out_destroy_cq:
+	cq->device->destroy_cq(cq);
+	return ERR_PTR(ret);
+}
+EXPORT_SYMBOL(ib_alloc_cq);
+
+/**
+ * ib_free_cq - free a completion queue
+ * @cq:		completion queue to free.
+ */
+void ib_free_cq(struct ib_cq *cq)
+{
+	int ret;
+
+	if (WARN_ON_ONCE(atomic_read(&cq->usecnt)))
+		return;
+
+	switch (cq->poll_ctx) {
+	case IB_POLL_DIRECT:
+		break;
+	case IB_POLL_SOFTIRQ:
+		irq_poll_disable(&cq->iop);
+		break;
+	case IB_POLL_WORKQUEUE:
+		flush_work(&cq->work);
+		break;
+	default:
+		WARN_ON_ONCE(1);
+	}
+
+	kfree(cq->wc);
+	ret = cq->device->destroy_cq(cq);
+	WARN_ON_ONCE(ret);
+}
+EXPORT_SYMBOL(ib_free_cq);
diff --git a/drivers/infiniband/core/device.c b/drivers/infiniband/core/device.c
index 179e813..6421d23 100644
--- a/drivers/infiniband/core/device.c
+++ b/drivers/infiniband/core/device.c
@@ -58,6 +58,7 @@
 	bool		  going_down;
 };
 
+struct workqueue_struct *ib_comp_wq;
 struct workqueue_struct *ib_wq;
 EXPORT_SYMBOL_GPL(ib_wq);
 
@@ -954,10 +955,18 @@
 	if (!ib_wq)
 		return -ENOMEM;
 
+	ib_comp_wq = alloc_workqueue("ib-comp-wq",
+			WQ_UNBOUND | WQ_HIGHPRI | WQ_MEM_RECLAIM,
+			WQ_UNBOUND_MAX_ACTIVE);
+	if (!ib_comp_wq) {
+		ret = -ENOMEM;
+		goto err;
+	}
+
 	ret = class_register(&ib_class);
 	if (ret) {
 		printk(KERN_WARNING "Couldn't create InfiniBand device class\n");
-		goto err;
+		goto err_comp;
 	}
 
 	ret = ibnl_init();
@@ -972,7 +981,8 @@
 
 err_sysfs:
 	class_unregister(&ib_class);
-
+err_comp:
+	destroy_workqueue(ib_comp_wq);
 err:
 	destroy_workqueue(ib_wq);
 	return ret;
@@ -983,6 +993,7 @@
 	ib_cache_cleanup();
 	ibnl_cleanup();
 	class_unregister(&ib_class);
+	destroy_workqueue(ib_comp_wq);
 	/* Make sure that any pending umem accounting work is done. */
 	destroy_workqueue(ib_wq);
 }
diff --git a/drivers/infiniband/ulp/ipoib/ipoib_cm.c b/drivers/infiniband/ulp/ipoib/ipoib_cm.c
index 3ae9726..9b014f1 100644
--- a/drivers/infiniband/ulp/ipoib/ipoib_cm.c
+++ b/drivers/infiniband/ulp/ipoib/ipoib_cm.c
@@ -70,7 +70,6 @@
 #define IPOIB_CM_RX_DRAIN_WRID 0xffffffff
 
 static struct ib_send_wr ipoib_cm_rx_drain_wr = {
-	.wr_id = IPOIB_CM_RX_DRAIN_WRID,
 	.opcode = IB_WR_SEND,
 };
 
@@ -223,6 +222,7 @@
 	 * error" WC will be immediately generated for each WR we post.
 	 */
 	p = list_entry(priv->cm.rx_flush_list.next, typeof(*p), list);
+	ipoib_cm_rx_drain_wr.wr_id = IPOIB_CM_RX_DRAIN_WRID;
 	if (ib_post_send(p->qp, &ipoib_cm_rx_drain_wr, &bad_wr))
 		ipoib_warn(priv, "failed to post drain wr\n");
 
diff --git a/drivers/infiniband/ulp/iser/iscsi_iser.h b/drivers/infiniband/ulp/iser/iscsi_iser.h
index 8a5998e..2411680 100644
--- a/drivers/infiniband/ulp/iser/iscsi_iser.h
+++ b/drivers/infiniband/ulp/iser/iscsi_iser.h
@@ -151,16 +151,12 @@
 					 - ISER_MAX_RX_MISC_PDUS) /	\
 					 (1 + ISER_INFLIGHT_DATAOUTS))
 
-#define ISER_WC_BATCH_COUNT   16
 #define ISER_SIGNAL_CMD_COUNT 32
 
 #define ISER_VER			0x10
 #define ISER_WSV			0x08
 #define ISER_RSV			0x04
 
-#define ISER_FASTREG_LI_WRID		0xffffffffffffffffULL
-#define ISER_BEACON_WRID		0xfffffffffffffffeULL
-
 /**
  * struct iser_hdr - iSER header
  *
@@ -269,7 +265,7 @@
 #define ISER_MAX_WRS 7
 
 /**
- * struct iser_tx_desc - iSER TX descriptor (for send wr_id)
+ * struct iser_tx_desc - iSER TX descriptor
  *
  * @iser_header:   iser header
  * @iscsi_header:  iscsi header
@@ -293,6 +289,7 @@
 	u64		             dma_addr;
 	struct ib_sge		     tx_sg[2];
 	int                          num_sge;
+	struct ib_cqe		     cqe;
 	bool			     mapped;
 	u8                           wr_idx;
 	union iser_wr {
@@ -306,9 +303,10 @@
 };
 
 #define ISER_RX_PAD_SIZE	(256 - (ISER_RX_PAYLOAD_SIZE + \
-					sizeof(u64) + sizeof(struct ib_sge)))
+				 sizeof(u64) + sizeof(struct ib_sge) + \
+				 sizeof(struct ib_cqe)))
 /**
- * struct iser_rx_desc - iSER RX descriptor (for recv wr_id)
+ * struct iser_rx_desc - iSER RX descriptor
  *
  * @iser_header:   iser header
  * @iscsi_header:  iscsi header
@@ -323,7 +321,27 @@
 	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
+ * @resp:          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;
 } __attribute__((packed));
 
 struct iser_conn;
@@ -333,18 +351,12 @@
 /**
  * struct iser_comp - iSER completion context
  *
- * @device:     pointer to device handle
  * @cq:         completion queue
- * @wcs:        work completion array
- * @tasklet:    Tasklet handle
  * @active_qps: Number of active QPs attached
  *              to completion context
  */
 struct iser_comp {
-	struct iser_device      *device;
 	struct ib_cq		*cq;
-	struct ib_wc		 wcs[ISER_WC_BATCH_COUNT];
-	struct tasklet_struct	 tasklet;
 	int                      active_qps;
 };
 
@@ -475,10 +487,11 @@
  * @rx_wr:               receive work request for batch posts
  * @device:              reference to iser device
  * @comp:                iser completion context
- * @pi_support:          Indicate device T10-PI support
- * @beacon:              beacon send wr to signal all flush errors were drained
- * @flush_comp:          completes when all connection completions consumed
  * @fr_pool:             connection fast registration poool
+ * @pi_support:          Indicate device T10-PI support
+ * @last:                last send wr to signal all flush errors were drained
+ * @last_cqe:            cqe handler for last wr
+ * @last_comp:           completes when all connection completions consumed
  */
 struct ib_conn {
 	struct rdma_cm_id           *cma_id;
@@ -488,10 +501,12 @@
 	struct ib_recv_wr	     rx_wr[ISER_MIN_POSTED_RX];
 	struct iser_device          *device;
 	struct iser_comp	    *comp;
-	bool			     pi_support;
-	struct ib_send_wr	     beacon;
-	struct completion	     flush_comp;
 	struct iser_fr_pool          fr_pool;
+	bool			     pi_support;
+	struct ib_send_wr	     last;
+	struct ib_cqe		     last_cqe;
+	struct ib_cqe		     reg_cqe;
+	struct completion	     last_comp;
 };
 
 /**
@@ -514,11 +529,7 @@
  * @up_completion:    connection establishment completed
  *                    (state is ISER_CONN_UP)
  * @conn_list:        entry in ig conn list
- * @login_buf:        login data buffer (stores login parameters)
- * @login_req_buf:    login request buffer
- * @login_req_dma:    login request buffer dma address
- * @login_resp_buf:   login response buffer
- * @login_resp_dma:   login response buffer dma address
+ * @login_desc:       login descriptor
  * @rx_desc_head:     head of rx_descs cyclic buffer
  * @rx_descs:         rx buffers array (cyclic buffer)
  * @num_rx_descs:     number of rx descriptors
@@ -541,10 +552,7 @@
 	struct completion	     ib_completion;
 	struct completion	     up_completion;
 	struct list_head	     conn_list;
-
-	char  			     *login_buf;
-	char			     *login_req_buf, *login_resp_buf;
-	u64			     login_req_dma, login_resp_dma;
+	struct iser_login_desc       login_desc;
 	unsigned int 		     rx_desc_head;
 	struct iser_rx_desc	     *rx_descs;
 	u32                          num_rx_descs;
@@ -633,12 +641,14 @@
 
 void iser_release_work(struct work_struct *work);
 
-void iser_rcv_completion(struct iser_rx_desc *desc,
-			 unsigned long dto_xfer_len,
-			 struct ib_conn *ib_conn);
-
-void iser_snd_completion(struct iser_tx_desc *desc,
-			 struct ib_conn *ib_conn);
+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_last_comp(struct ib_cq *cq, struct ib_wc *wc);
 
 void iser_task_rdma_init(struct iscsi_iser_task *task);
 
@@ -719,4 +729,28 @@
 	return cur_wr;
 }
 
+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
index ffd00c4..44ecb89 100644
--- a/drivers/infiniband/ulp/iser/iser_initiator.c
+++ b/drivers/infiniband/ulp/iser/iser_initiator.c
@@ -174,73 +174,63 @@
 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 (!iser_conn->login_buf)
+	if (!desc->req)
 		return;
 
-	if (iser_conn->login_req_dma)
-		ib_dma_unmap_single(device->ib_device,
-				    iser_conn->login_req_dma,
-				    ISCSI_DEF_MAX_RECV_SEG_LEN, DMA_TO_DEVICE);
+	ib_dma_unmap_single(device->ib_device, desc->req_dma,
+			    ISCSI_DEF_MAX_RECV_SEG_LEN, DMA_TO_DEVICE);
 
-	if (iser_conn->login_resp_dma)
-		ib_dma_unmap_single(device->ib_device,
-				    iser_conn->login_resp_dma,
-				    ISER_RX_LOGIN_SIZE, DMA_FROM_DEVICE);
+	ib_dma_unmap_single(device->ib_device, desc->rsp_dma,
+			    ISER_RX_LOGIN_SIZE, DMA_FROM_DEVICE);
 
-	kfree(iser_conn->login_buf);
+	kfree(desc->req);
+	kfree(desc->rsp);
 
 	/* make sure we never redo any unmapping */
-	iser_conn->login_req_dma = 0;
-	iser_conn->login_resp_dma = 0;
-	iser_conn->login_buf = NULL;
+	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;
-	int			req_err, resp_err;
+	struct iser_login_desc *desc = &iser_conn->login_desc;
 
-	BUG_ON(device == NULL);
+	desc->req = kmalloc(ISCSI_DEF_MAX_RECV_SEG_LEN, GFP_KERNEL);
+	if (!desc->req)
+		return -ENOMEM;
 
-	iser_conn->login_buf = kmalloc(ISCSI_DEF_MAX_RECV_SEG_LEN +
-				     ISER_RX_LOGIN_SIZE, GFP_KERNEL);
-	if (!iser_conn->login_buf)
-		goto out_err;
+	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;
 
-	iser_conn->login_req_buf  = iser_conn->login_buf;
-	iser_conn->login_resp_buf = iser_conn->login_buf +
-						ISCSI_DEF_MAX_RECV_SEG_LEN;
+	desc->rsp = kmalloc(ISER_RX_LOGIN_SIZE, GFP_KERNEL);
+	if (!desc->rsp)
+		goto unmap_req;
 
-	iser_conn->login_req_dma = ib_dma_map_single(device->ib_device,
-						     iser_conn->login_req_buf,
-						     ISCSI_DEF_MAX_RECV_SEG_LEN,
-						     DMA_TO_DEVICE);
+	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;
 
-	iser_conn->login_resp_dma = ib_dma_map_single(device->ib_device,
-						      iser_conn->login_resp_buf,
-						      ISER_RX_LOGIN_SIZE,
-						      DMA_FROM_DEVICE);
-
-	req_err  = ib_dma_mapping_error(device->ib_device,
-					iser_conn->login_req_dma);
-	resp_err = ib_dma_mapping_error(device->ib_device,
-					iser_conn->login_resp_dma);
-
-	if (req_err || resp_err) {
-		if (req_err)
-			iser_conn->login_req_dma = 0;
-		if (resp_err)
-			iser_conn->login_resp_dma = 0;
-		goto free_login_buf;
-	}
 	return 0;
 
-free_login_buf:
-	iser_free_login_buf(iser_conn);
+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);
 
-out_err:
-	iser_err("unable to alloc or map login buf\n");
 	return -ENOMEM;
 }
 
@@ -280,11 +270,11 @@
 			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->addr = rx_desc->dma_addr;
 		rx_sg->length = ISER_RX_PAYLOAD_SIZE;
-		rx_sg->lkey   = device->pd->local_dma_lkey;
+		rx_sg->lkey = device->pd->local_dma_lkey;
 	}
 
 	iser_conn->rx_desc_head = 0;
@@ -383,6 +373,7 @@
 
 	/* build the tx desc regd header and add it to the tx desc dto */
 	tx_desc->type = ISCSI_TX_SCSI_COMMAND;
+	tx_desc->cqe.done = iser_cmd_comp;
 	iser_create_send_desc(iser_conn, tx_desc);
 
 	if (hdr->flags & ISCSI_FLAG_CMD_READ) {
@@ -464,6 +455,7 @@
 	}
 
 	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));
 
@@ -513,6 +505,7 @@
 
 	/* build the tx desc regd header and add it to the tx desc dto */
 	mdesc->type = ISCSI_TX_CONTROL;
+	mdesc->cqe.done = iser_ctrl_comp;
 	iser_create_send_desc(iser_conn, mdesc);
 
 	device = iser_conn->ib_conn.device;
@@ -520,25 +513,25 @@
 	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,
-			iser_conn->login_req_dma, task->data_count,
-			DMA_TO_DEVICE);
+		ib_dma_sync_single_for_cpu(device->ib_device, desc->req_dma,
+					   task->data_count, DMA_TO_DEVICE);
 
-		memcpy(iser_conn->login_req_buf, task->data, task->data_count);
+		memcpy(desc->req, task->data, task->data_count);
 
-		ib_dma_sync_single_for_device(device->ib_device,
-			iser_conn->login_req_dma, task->data_count,
-			DMA_TO_DEVICE);
+		ib_dma_sync_single_for_device(device->ib_device, desc->req_dma,
+					      task->data_count, DMA_TO_DEVICE);
 
-		tx_dsg->addr    = iser_conn->login_req_dma;
-		tx_dsg->length  = task->data_count;
-		tx_dsg->lkey    = device->pd->local_dma_lkey;
+		tx_dsg->addr = desc->req_dma;
+		tx_dsg->length = task->data_count;
+		tx_dsg->lkey = device->pd->local_dma_lkey;
 		mdesc->num_sge = 2;
 	}
 
@@ -562,41 +555,69 @@
 	return err;
 }
 
-/**
- * iser_rcv_dto_completion - recv DTO completion
- */
-void iser_rcv_completion(struct iser_rx_desc *rx_desc,
-			 unsigned long rx_xfer_len,
-			 struct ib_conn *ib_conn)
+void iser_login_rsp(struct ib_cq *cq, struct ib_wc *wc)
 {
-	struct iser_conn *iser_conn = container_of(ib_conn, struct iser_conn,
-						   ib_conn);
+	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;
-	u64 rx_dma;
-	int rx_buflen, outstanding, count, err;
+	char *data;
+	int length;
 
-	/* differentiate between login to all other PDUs */
-	if ((char *)rx_desc == iser_conn->login_resp_buf) {
-		rx_dma = iser_conn->login_resp_dma;
-		rx_buflen = ISER_RX_LOGIN_SIZE;
-	} else {
-		rx_dma = rx_desc->dma_addr;
-		rx_buflen = ISER_RX_PAYLOAD_SIZE;
+	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, rx_dma,
-				   rx_buflen, DMA_FROM_DEVICE);
+	ib_dma_sync_single_for_cpu(ib_conn->device->ib_device,
+				   desc->rsp_dma, ISER_RX_LOGIN_SIZE,
+				   DMA_FROM_DEVICE);
 
-	hdr = &rx_desc->iscsi_header;
+	hdr = desc->rsp + sizeof(struct iser_hdr);
+	data = desc->rsp + ISER_HEADERS_LEN;
+	length = wc->byte_len - ISER_HEADERS_LEN;
 
 	iser_dbg("op 0x%x itt 0x%x dlen %d\n", hdr->opcode,
-			hdr->itt, (int)(rx_xfer_len - ISER_HEADERS_LEN));
+		 hdr->itt, length);
 
-	iscsi_iser_recv(iser_conn->iscsi_conn, hdr, rx_desc->data,
-			rx_xfer_len - ISER_HEADERS_LEN);
+	iscsi_iser_recv(iser_conn->iscsi_conn, hdr, data, length);
 
-	ib_dma_sync_single_for_device(ib_conn->device->ib_device, rx_dma,
-				      rx_buflen, DMA_FROM_DEVICE);
+	ib_dma_sync_single_for_device(ib_conn->device->ib_device,
+				      desc->rsp_dma, ISER_RX_LOGIN_SIZE,
+				      DMA_FROM_DEVICE);
+
+	ib_conn->post_recv_buf_count--;
+}
+
+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;
+	int outstanding, count, 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);
+
+	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);
 
 	/* decrementing conn->post_recv_buf_count only --after-- freeing the   *
 	 * task eliminates the need to worry on tasks which are completed in   *
@@ -604,9 +625,6 @@
 	 * for the posted rx bufs refcount to become zero handles everything   */
 	ib_conn->post_recv_buf_count--;
 
-	if (rx_dma == iser_conn->login_resp_dma)
-		return;
-
 	outstanding = ib_conn->post_recv_buf_count;
 	if (outstanding + iser_conn->min_posted_rx <= iser_conn->qp_max_recv_dtos) {
 		count = min(iser_conn->qp_max_recv_dtos - outstanding,
@@ -617,26 +635,47 @@
 	}
 }
 
-void iser_snd_completion(struct iser_tx_desc *tx_desc,
-			struct ib_conn *ib_conn)
+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 (tx_desc->type == ISCSI_TX_DATAOUT) {
-		ib_dma_unmap_single(device->ib_device, tx_desc->dma_addr,
-					ISER_HEADERS_LEN, DMA_TO_DEVICE);
-		kmem_cache_free(ig.desc_cache, tx_desc);
-		tx_desc = NULL;
-	}
+	if (unlikely(wc->status != IB_WC_SUCCESS))
+		iser_err_comp(wc, "dataout");
 
-	if (tx_desc && tx_desc->type == ISCSI_TX_CONTROL) {
-		/* this arithmetic is legal by libiscsi dd_data allocation */
-		task = (void *) ((long)(void *)tx_desc -
-				  sizeof(struct iscsi_task));
-		if (task->hdr->itt == RESERVED_ITT)
-			iscsi_put_task(task);
-	}
+	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_last_comp(struct ib_cq *cq, struct ib_wc *wc)
+{
+	struct ib_conn *ib_conn = wc->qp->qp_context;
+
+	complete(&ib_conn->last_comp);
 }
 
 void iser_task_rdma_init(struct iscsi_iser_task *iser_task)
diff --git a/drivers/infiniband/ulp/iser/iser_memory.c b/drivers/infiniband/ulp/iser/iser_memory.c
index ea765fb..76ca09b 100644
--- a/drivers/infiniband/ulp/iser/iser_memory.c
+++ b/drivers/infiniband/ulp/iser/iser_memory.c
@@ -67,6 +67,11 @@
 	.reg_desc_put	= iser_reg_desc_put_fmr,
 };
 
+void iser_reg_comp(struct ib_cq *cq, struct ib_wc *wc)
+{
+	iser_err_comp(wc, "memreg");
+}
+
 int iser_assign_reg_ops(struct iser_device *device)
 {
 	struct ib_device_attr *dev_attr = &device->dev_attr;
@@ -414,12 +419,14 @@
 }
 
 static void
-iser_inv_rkey(struct ib_send_wr *inv_wr, struct ib_mr *mr)
+iser_inv_rkey(struct ib_send_wr *inv_wr,
+	      struct ib_mr *mr,
+	      struct ib_cqe *cqe)
 {
 	u32 rkey;
 
 	inv_wr->opcode = IB_WR_LOCAL_INV;
-	inv_wr->wr_id = ISER_FASTREG_LI_WRID;
+	inv_wr->wr_cqe = cqe;
 	inv_wr->ex.invalidate_rkey = mr->rkey;
 	inv_wr->send_flags = 0;
 	inv_wr->num_sge = 0;
@@ -437,6 +444,7 @@
 {
 	struct iser_tx_desc *tx_desc = &iser_task->desc;
 	struct ib_sig_attrs *sig_attrs = &tx_desc->sig_attrs;
+	struct ib_cqe *cqe = &iser_task->iser_conn->ib_conn.reg_cqe;
 	struct ib_sig_handover_wr *wr;
 	int ret;
 
@@ -448,11 +456,11 @@
 	iser_set_prot_checks(iser_task->sc, &sig_attrs->check_mask);
 
 	if (!pi_ctx->sig_mr_valid)
-		iser_inv_rkey(iser_tx_next_wr(tx_desc), pi_ctx->sig_mr);
+		iser_inv_rkey(iser_tx_next_wr(tx_desc), pi_ctx->sig_mr, cqe);
 
 	wr = sig_handover_wr(iser_tx_next_wr(tx_desc));
 	wr->wr.opcode = IB_WR_REG_SIG_MR;
-	wr->wr.wr_id = ISER_FASTREG_LI_WRID;
+	wr->wr.wr_cqe = cqe;
 	wr->wr.sg_list = &data_reg->sge;
 	wr->wr.num_sge = 1;
 	wr->wr.send_flags = 0;
@@ -485,12 +493,13 @@
 			    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;
 	int n;
 
 	if (!rsc->mr_valid)
-		iser_inv_rkey(iser_tx_next_wr(tx_desc), mr);
+		iser_inv_rkey(iser_tx_next_wr(tx_desc), mr, cqe);
 
 	n = ib_map_mr_sg(mr, mem->sg, mem->size, SIZE_4K);
 	if (unlikely(n != mem->size)) {
@@ -501,7 +510,7 @@
 
 	wr = reg_wr(iser_tx_next_wr(tx_desc));
 	wr->wr.opcode = IB_WR_REG_MR;
-	wr->wr.wr_id = ISER_FASTREG_LI_WRID;
+	wr->wr.wr_cqe = cqe;
 	wr->wr.send_flags = 0;
 	wr->wr.num_sge = 0;
 	wr->mr = mr;
diff --git a/drivers/infiniband/ulp/iser/iser_verbs.c b/drivers/infiniband/ulp/iser/iser_verbs.c
index 42f4da6..2f2fc42 100644
--- a/drivers/infiniband/ulp/iser/iser_verbs.c
+++ b/drivers/infiniband/ulp/iser/iser_verbs.c
@@ -44,17 +44,6 @@
 #define ISER_MAX_CQ_LEN		(ISER_MAX_RX_LEN + ISER_MAX_TX_LEN + \
 				 ISCSI_ISER_MAX_CONN)
 
-static int iser_cq_poll_limit = 512;
-
-static void iser_cq_tasklet_fn(unsigned long data);
-static void iser_cq_callback(struct ib_cq *cq, void *cq_context);
-
-static void iser_cq_event_callback(struct ib_event *cause, void *context)
-{
-	iser_err("cq event %s (%d)\n",
-		 ib_event_msg(cause->event), cause->event);
-}
-
 static void iser_qp_event_callback(struct ib_event *cause, void *context)
 {
 	iser_err("qp event %s (%d)\n",
@@ -110,27 +99,14 @@
 		goto pd_err;
 
 	for (i = 0; i < device->comps_used; i++) {
-		struct ib_cq_init_attr cq_attr = {};
 		struct iser_comp *comp = &device->comps[i];
 
-		comp->device = device;
-		cq_attr.cqe = max_cqe;
-		cq_attr.comp_vector = i;
-		comp->cq = ib_create_cq(device->ib_device,
-					iser_cq_callback,
-					iser_cq_event_callback,
-					(void *)comp,
-					&cq_attr);
+		comp->cq = ib_alloc_cq(device->ib_device, comp, max_cqe, i,
+				IB_POLL_SOFTIRQ);
 		if (IS_ERR(comp->cq)) {
 			comp->cq = NULL;
 			goto cq_err;
 		}
-
-		if (ib_req_notify_cq(comp->cq, IB_CQ_NEXT_COMP))
-			goto cq_err;
-
-		tasklet_init(&comp->tasklet, iser_cq_tasklet_fn,
-			     (unsigned long)comp);
 	}
 
 	if (!iser_always_reg) {
@@ -140,7 +116,7 @@
 
 		device->mr = ib_get_dma_mr(device->pd, access);
 		if (IS_ERR(device->mr))
-			goto dma_mr_err;
+			goto cq_err;
 	}
 
 	INIT_IB_EVENT_HANDLER(&device->event_handler, device->ib_device,
@@ -153,15 +129,12 @@
 handler_err:
 	if (device->mr)
 		ib_dereg_mr(device->mr);
-dma_mr_err:
-	for (i = 0; i < device->comps_used; i++)
-		tasklet_kill(&device->comps[i].tasklet);
 cq_err:
 	for (i = 0; i < device->comps_used; i++) {
 		struct iser_comp *comp = &device->comps[i];
 
 		if (comp->cq)
-			ib_destroy_cq(comp->cq);
+			ib_free_cq(comp->cq);
 	}
 	ib_dealloc_pd(device->pd);
 pd_err:
@@ -182,8 +155,7 @@
 	for (i = 0; i < device->comps_used; i++) {
 		struct iser_comp *comp = &device->comps[i];
 
-		tasklet_kill(&comp->tasklet);
-		ib_destroy_cq(comp->cq);
+		ib_free_cq(comp->cq);
 		comp->cq = NULL;
 	}
 
@@ -461,8 +433,7 @@
  */
 static int iser_create_ib_conn_res(struct ib_conn *ib_conn)
 {
-	struct iser_conn *iser_conn = container_of(ib_conn, struct iser_conn,
-						   ib_conn);
+	struct iser_conn *iser_conn = to_iser_conn(ib_conn);
 	struct iser_device	*device;
 	struct ib_device_attr *dev_attr;
 	struct ib_qp_init_attr	init_attr;
@@ -724,13 +695,13 @@
 				 iser_conn, err);
 
 		/* post an indication that all flush errors were consumed */
-		err = ib_post_send(ib_conn->qp, &ib_conn->beacon, &bad_wr);
+		err = ib_post_send(ib_conn->qp, &ib_conn->last, &bad_wr);
 		if (err) {
-			iser_err("conn %p failed to post beacon", ib_conn);
+			iser_err("conn %p failed to post last wr", ib_conn);
 			return 1;
 		}
 
-		wait_for_completion(&ib_conn->flush_comp);
+		wait_for_completion(&ib_conn->last_comp);
 	}
 
 	return 1;
@@ -967,14 +938,21 @@
 
 void iser_conn_init(struct iser_conn *iser_conn)
 {
+	struct ib_conn *ib_conn = &iser_conn->ib_conn;
+
 	iser_conn->state = ISER_CONN_INIT;
-	iser_conn->ib_conn.post_recv_buf_count = 0;
-	init_completion(&iser_conn->ib_conn.flush_comp);
 	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->post_recv_buf_count = 0;
+	ib_conn->reg_cqe.done = iser_reg_comp;
+	ib_conn->last_cqe.done = iser_last_comp;
+	ib_conn->last.wr_cqe = &ib_conn->last_cqe;
+	ib_conn->last.opcode = IB_WR_SEND;
+	init_completion(&ib_conn->last_comp);
 }
 
  /**
@@ -1000,9 +978,6 @@
 
 	iser_conn->state = ISER_CONN_PENDING;
 
-	ib_conn->beacon.wr_id = ISER_BEACON_WRID;
-	ib_conn->beacon.opcode = IB_WR_SEND;
-
 	ib_conn->cma_id = rdma_create_id(&init_net, iser_cma_handler,
 					 (void *)iser_conn,
 					 RDMA_PS_TCP, IB_QPT_RC);
@@ -1045,56 +1020,60 @@
 
 int iser_post_recvl(struct iser_conn *iser_conn)
 {
-	struct ib_recv_wr rx_wr, *rx_wr_failed;
 	struct ib_conn *ib_conn = &iser_conn->ib_conn;
-	struct ib_sge	  sge;
+	struct iser_login_desc *desc = &iser_conn->login_desc;
+	struct ib_recv_wr wr, *wr_failed;
 	int ib_ret;
 
-	sge.addr   = iser_conn->login_resp_dma;
-	sge.length = ISER_RX_LOGIN_SIZE;
-	sge.lkey   = ib_conn->device->pd->local_dma_lkey;
+	desc->sge.addr = desc->rsp_dma;
+	desc->sge.length = ISER_RX_LOGIN_SIZE;
+	desc->sge.lkey = ib_conn->device->pd->local_dma_lkey;
 
-	rx_wr.wr_id   = (uintptr_t)iser_conn->login_resp_buf;
-	rx_wr.sg_list = &sge;
-	rx_wr.num_sge = 1;
-	rx_wr.next    = NULL;
+	desc->cqe.done = iser_login_rsp;
+	wr.wr_cqe = &desc->cqe;
+	wr.sg_list = &desc->sge;
+	wr.num_sge = 1;
+	wr.next = NULL;
 
 	ib_conn->post_recv_buf_count++;
-	ib_ret	= ib_post_recv(ib_conn->qp, &rx_wr, &rx_wr_failed);
+	ib_ret = ib_post_recv(ib_conn->qp, &wr, &wr_failed);
 	if (ib_ret) {
 		iser_err("ib_post_recv failed ret=%d\n", ib_ret);
 		ib_conn->post_recv_buf_count--;
 	}
+
 	return ib_ret;
 }
 
 int iser_post_recvm(struct iser_conn *iser_conn, int count)
 {
-	struct ib_recv_wr *rx_wr, *rx_wr_failed;
-	int i, ib_ret;
 	struct ib_conn *ib_conn = &iser_conn->ib_conn;
 	unsigned int my_rx_head = iser_conn->rx_desc_head;
 	struct iser_rx_desc *rx_desc;
+	struct ib_recv_wr *wr, *wr_failed;
+	int i, ib_ret;
 
-	for (rx_wr = ib_conn->rx_wr, i = 0; i < count; i++, rx_wr++) {
-		rx_desc		= &iser_conn->rx_descs[my_rx_head];
-		rx_wr->wr_id	= (uintptr_t)rx_desc;
-		rx_wr->sg_list	= &rx_desc->rx_sg;
-		rx_wr->num_sge	= 1;
-		rx_wr->next	= rx_wr + 1;
+	for (wr = ib_conn->rx_wr, i = 0; i < count; i++, wr++) {
+		rx_desc = &iser_conn->rx_descs[my_rx_head];
+		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 = wr + 1;
 		my_rx_head = (my_rx_head + 1) & iser_conn->qp_max_recv_dtos_mask;
 	}
 
-	rx_wr--;
-	rx_wr->next = NULL; /* mark end of work requests list */
+	wr--;
+	wr->next = NULL; /* mark end of work requests list */
 
 	ib_conn->post_recv_buf_count += count;
-	ib_ret	= ib_post_recv(ib_conn->qp, ib_conn->rx_wr, &rx_wr_failed);
+	ib_ret = ib_post_recv(ib_conn->qp, ib_conn->rx_wr, &wr_failed);
 	if (ib_ret) {
 		iser_err("ib_post_recv failed ret=%d\n", ib_ret);
 		ib_conn->post_recv_buf_count -= count;
 	} else
 		iser_conn->rx_desc_head = my_rx_head;
+
 	return ib_ret;
 }
 
@@ -1115,7 +1094,7 @@
 				      DMA_TO_DEVICE);
 
 	wr->next = NULL;
-	wr->wr_id = (uintptr_t)tx_desc;
+	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;
@@ -1129,149 +1108,6 @@
 	return ib_ret;
 }
 
-/**
- * is_iser_tx_desc - Indicate if the completion wr_id
- *     is a TX descriptor or not.
- * @iser_conn: iser connection
- * @wr_id: completion WR identifier
- *
- * Since we cannot rely on wc opcode in FLUSH errors
- * we must work around it by checking if the wr_id address
- * falls in the iser connection rx_descs buffer. If so
- * it is an RX descriptor, otherwize it is a TX.
- */
-static inline bool
-is_iser_tx_desc(struct iser_conn *iser_conn, void *wr_id)
-{
-	void *start = iser_conn->rx_descs;
-	int len = iser_conn->num_rx_descs * sizeof(*iser_conn->rx_descs);
-
-	if (wr_id >= start && wr_id < start + len)
-		return false;
-
-	return true;
-}
-
-/**
- * iser_handle_comp_error() - Handle error completion
- * @ib_conn:   connection RDMA resources
- * @wc:        work completion
- *
- * Notes: We may handle a FLUSH error completion and in this case
- *        we only cleanup in case TX type was DATAOUT. For non-FLUSH
- *        error completion we should also notify iscsi layer that
- *        connection is failed (in case we passed bind stage).
- */
-static void
-iser_handle_comp_error(struct ib_conn *ib_conn,
-		       struct ib_wc *wc)
-{
-	void *wr_id = (void *)(uintptr_t)wc->wr_id;
-	struct iser_conn *iser_conn = container_of(ib_conn, struct iser_conn,
-						   ib_conn);
-
-	if (wc->status != IB_WC_WR_FLUSH_ERR)
-		if (iser_conn->iscsi_conn)
-			iscsi_conn_failure(iser_conn->iscsi_conn,
-					   ISCSI_ERR_CONN_FAILED);
-
-	if (wc->wr_id == ISER_FASTREG_LI_WRID)
-		return;
-
-	if (is_iser_tx_desc(iser_conn, wr_id)) {
-		struct iser_tx_desc *desc = wr_id;
-
-		if (desc->type == ISCSI_TX_DATAOUT)
-			kmem_cache_free(ig.desc_cache, desc);
-	} else {
-		ib_conn->post_recv_buf_count--;
-	}
-}
-
-/**
- * iser_handle_wc - handle a single work completion
- * @wc: work completion
- *
- * Soft-IRQ context, work completion can be either
- * SEND or RECV, and can turn out successful or
- * with error (or flush error).
- */
-static void iser_handle_wc(struct ib_wc *wc)
-{
-	struct ib_conn *ib_conn;
-	struct iser_tx_desc *tx_desc;
-	struct iser_rx_desc *rx_desc;
-
-	ib_conn = wc->qp->qp_context;
-	if (likely(wc->status == IB_WC_SUCCESS)) {
-		if (wc->opcode == IB_WC_RECV) {
-			rx_desc = (struct iser_rx_desc *)(uintptr_t)wc->wr_id;
-			iser_rcv_completion(rx_desc, wc->byte_len,
-					    ib_conn);
-		} else
-		if (wc->opcode == IB_WC_SEND) {
-			tx_desc = (struct iser_tx_desc *)(uintptr_t)wc->wr_id;
-			iser_snd_completion(tx_desc, ib_conn);
-		} else {
-			iser_err("Unknown wc opcode %d\n", wc->opcode);
-		}
-	} else {
-		if (wc->status != IB_WC_WR_FLUSH_ERR)
-			iser_err("%s (%d): wr id %llx vend_err %x\n",
-				 ib_wc_status_msg(wc->status), wc->status,
-				 wc->wr_id, wc->vendor_err);
-		else
-			iser_dbg("%s (%d): wr id %llx\n",
-				 ib_wc_status_msg(wc->status), wc->status,
-				 wc->wr_id);
-
-		if (wc->wr_id == ISER_BEACON_WRID)
-			/* all flush errors were consumed */
-			complete(&ib_conn->flush_comp);
-		else
-			iser_handle_comp_error(ib_conn, wc);
-	}
-}
-
-/**
- * iser_cq_tasklet_fn - iSER completion polling loop
- * @data: iSER completion context
- *
- * Soft-IRQ context, polling connection CQ until
- * either CQ was empty or we exausted polling budget
- */
-static void iser_cq_tasklet_fn(unsigned long data)
-{
-	struct iser_comp *comp = (struct iser_comp *)data;
-	struct ib_cq *cq = comp->cq;
-	struct ib_wc *const wcs = comp->wcs;
-	int i, n, completed = 0;
-
-	while ((n = ib_poll_cq(cq, ARRAY_SIZE(comp->wcs), wcs)) > 0) {
-		for (i = 0; i < n; i++)
-			iser_handle_wc(&wcs[i]);
-
-		completed += n;
-		if (completed >= iser_cq_poll_limit)
-			break;
-	}
-
-	/*
-	 * It is assumed here that arming CQ only once its empty
-	 * would not cause interrupts to be missed.
-	 */
-	ib_req_notify_cq(cq, IB_CQ_NEXT_COMP);
-
-	iser_dbg("got %d completions\n", completed);
-}
-
-static void iser_cq_callback(struct ib_cq *cq, void *cq_context)
-{
-	struct iser_comp *comp = cq_context;
-
-	tasklet_schedule(&comp->tasklet);
-}
-
 u8 iser_check_task_pi_status(struct iscsi_iser_task *iser_task,
 			     enum iser_data_dir cmd_dir, sector_t *sector)
 {
@@ -1319,3 +1155,21 @@
 	/* Not alot we can do here, return ambiguous guard error */
 	return 0x1;
 }
+
+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);
+	}
+}
diff --git a/drivers/infiniband/ulp/srp/ib_srp.c b/drivers/infiniband/ulp/srp/ib_srp.c
index 3db9a65..20fa332 100644
--- a/drivers/infiniband/ulp/srp/ib_srp.c
+++ b/drivers/infiniband/ulp/srp/ib_srp.c
@@ -132,8 +132,9 @@
 
 static void srp_add_one(struct ib_device *device);
 static void srp_remove_one(struct ib_device *device, void *client_data);
-static void srp_recv_completion(struct ib_cq *cq, void *ch_ptr);
-static void srp_send_completion(struct ib_cq *cq, void *ch_ptr);
+static void srp_recv_done(struct ib_cq *cq, struct ib_wc *wc);
+static void srp_handle_qp_err(struct ib_cq *cq, struct ib_wc *wc,
+		const char *opname);
 static int srp_cm_handler(struct ib_cm_id *cm_id, struct ib_cm_event *event);
 
 static struct scsi_transport_template *ib_srp_transport_template;
@@ -445,6 +446,17 @@
 				  dev->max_pages_per_mr);
 }
 
+static void srp_drain_done(struct ib_cq *cq, struct ib_wc *wc)
+{
+	struct srp_rdma_ch *ch = cq->cq_context;
+
+	complete(&ch->done);
+}
+
+static struct ib_cqe srp_drain_cqe = {
+	.done		= srp_drain_done,
+};
+
 /**
  * srp_destroy_qp() - destroy an RDMA queue pair
  * @ch: SRP RDMA channel.
@@ -457,10 +469,11 @@
 static void srp_destroy_qp(struct srp_rdma_ch *ch)
 {
 	static struct ib_qp_attr attr = { .qp_state = IB_QPS_ERR };
-	static struct ib_recv_wr wr = { .wr_id = SRP_LAST_WR_ID };
+	static struct ib_recv_wr wr = { 0 };
 	struct ib_recv_wr *bad_wr;
 	int ret;
 
+	wr.wr_cqe = &srp_drain_cqe;
 	/* Destroying a QP and reusing ch->done is only safe if not connected */
 	WARN_ON_ONCE(ch->connected);
 
@@ -489,34 +502,27 @@
 	struct ib_fmr_pool *fmr_pool = NULL;
 	struct srp_fr_pool *fr_pool = NULL;
 	const int m = dev->use_fast_reg ? 3 : 1;
-	struct ib_cq_init_attr cq_attr = {};
 	int ret;
 
 	init_attr = kzalloc(sizeof *init_attr, GFP_KERNEL);
 	if (!init_attr)
 		return -ENOMEM;
 
-	/* + 1 for SRP_LAST_WR_ID */
-	cq_attr.cqe = target->queue_size + 1;
-	cq_attr.comp_vector = ch->comp_vector;
-	recv_cq = ib_create_cq(dev->dev, srp_recv_completion, NULL, ch,
-			       &cq_attr);
+	/* queue_size + 1 for ib_drain_qp */
+	recv_cq = ib_alloc_cq(dev->dev, ch, target->queue_size + 1,
+				ch->comp_vector, IB_POLL_SOFTIRQ);
 	if (IS_ERR(recv_cq)) {
 		ret = PTR_ERR(recv_cq);
 		goto err;
 	}
 
-	cq_attr.cqe = m * target->queue_size;
-	cq_attr.comp_vector = ch->comp_vector;
-	send_cq = ib_create_cq(dev->dev, srp_send_completion, NULL, ch,
-			       &cq_attr);
+	send_cq = ib_alloc_cq(dev->dev, ch, m * target->queue_size,
+				ch->comp_vector, IB_POLL_DIRECT);
 	if (IS_ERR(send_cq)) {
 		ret = PTR_ERR(send_cq);
 		goto err_recv_cq;
 	}
 
-	ib_req_notify_cq(recv_cq, IB_CQ_NEXT_COMP);
-
 	init_attr->event_handler       = srp_qp_event;
 	init_attr->cap.max_send_wr     = m * target->queue_size;
 	init_attr->cap.max_recv_wr     = target->queue_size + 1;
@@ -558,9 +564,9 @@
 	if (ch->qp)
 		srp_destroy_qp(ch);
 	if (ch->recv_cq)
-		ib_destroy_cq(ch->recv_cq);
+		ib_free_cq(ch->recv_cq);
 	if (ch->send_cq)
-		ib_destroy_cq(ch->send_cq);
+		ib_free_cq(ch->send_cq);
 
 	ch->qp = qp;
 	ch->recv_cq = recv_cq;
@@ -580,13 +586,13 @@
 	return 0;
 
 err_qp:
-	ib_destroy_qp(qp);
+	srp_destroy_qp(ch);
 
 err_send_cq:
-	ib_destroy_cq(send_cq);
+	ib_free_cq(send_cq);
 
 err_recv_cq:
-	ib_destroy_cq(recv_cq);
+	ib_free_cq(recv_cq);
 
 err:
 	kfree(init_attr);
@@ -622,9 +628,10 @@
 		if (ch->fmr_pool)
 			ib_destroy_fmr_pool(ch->fmr_pool);
 	}
+
 	srp_destroy_qp(ch);
-	ib_destroy_cq(ch->send_cq);
-	ib_destroy_cq(ch->recv_cq);
+	ib_free_cq(ch->send_cq);
+	ib_free_cq(ch->recv_cq);
 
 	/*
 	 * Avoid that the SCSI error handler tries to use this channel after
@@ -1041,18 +1048,25 @@
 	return ret <= 0 ? ret : -ENODEV;
 }
 
-static int srp_inv_rkey(struct srp_rdma_ch *ch, u32 rkey)
+static void srp_inv_rkey_err_done(struct ib_cq *cq, struct ib_wc *wc)
+{
+	srp_handle_qp_err(cq, wc, "INV RKEY");
+}
+
+static int srp_inv_rkey(struct srp_request *req, struct srp_rdma_ch *ch,
+		u32 rkey)
 {
 	struct ib_send_wr *bad_wr;
 	struct ib_send_wr wr = {
 		.opcode		    = IB_WR_LOCAL_INV,
-		.wr_id		    = LOCAL_INV_WR_ID_MASK,
 		.next		    = NULL,
 		.num_sge	    = 0,
 		.send_flags	    = 0,
 		.ex.invalidate_rkey = rkey,
 	};
 
+	wr.wr_cqe = &req->reg_cqe;
+	req->reg_cqe.done = srp_inv_rkey_err_done;
 	return ib_post_send(ch->qp, &wr, &bad_wr);
 }
 
@@ -1074,7 +1088,7 @@
 		struct srp_fr_desc **pfr;
 
 		for (i = req->nmdesc, pfr = req->fr_list; i > 0; i--, pfr++) {
-			res = srp_inv_rkey(ch, (*pfr)->mr->rkey);
+			res = srp_inv_rkey(req, ch, (*pfr)->mr->rkey);
 			if (res < 0) {
 				shost_printk(KERN_ERR, target->scsi_host, PFX
 				  "Queueing INV WR for rkey %#x failed (%d)\n",
@@ -1312,7 +1326,13 @@
 	return 0;
 }
 
+static void srp_reg_mr_err_done(struct ib_cq *cq, struct ib_wc *wc)
+{
+	srp_handle_qp_err(cq, wc, "FAST REG");
+}
+
 static int srp_map_finish_fr(struct srp_map_state *state,
+			     struct srp_request *req,
 			     struct srp_rdma_ch *ch, int sg_nents)
 {
 	struct srp_target_port *target = ch->target;
@@ -1349,9 +1369,11 @@
 	if (unlikely(n < 0))
 		return n;
 
+	req->reg_cqe.done = srp_reg_mr_err_done;
+
 	wr.wr.next = NULL;
 	wr.wr.opcode = IB_WR_REG_MR;
-	wr.wr.wr_id = FAST_REG_WR_ID_MASK;
+	wr.wr.wr_cqe = &req->reg_cqe;
 	wr.wr.num_sge = 0;
 	wr.wr.send_flags = 0;
 	wr.mr = desc->mr;
@@ -1455,7 +1477,7 @@
 	while (count) {
 		int i, n;
 
-		n = srp_map_finish_fr(state, ch, count);
+		n = srp_map_finish_fr(state, req, ch, count);
 		if (unlikely(n < 0))
 			return n;
 
@@ -1524,7 +1546,7 @@
 #ifdef CONFIG_NEED_SG_DMA_LENGTH
 		idb_sg->dma_length = idb_sg->length;	      /* hack^2 */
 #endif
-		ret = srp_map_finish_fr(&state, ch, 1);
+		ret = srp_map_finish_fr(&state, req, ch, 1);
 		if (ret < 0)
 			return ret;
 	} else if (dev->use_fmr) {
@@ -1719,7 +1741,7 @@
 	s32 rsv = (iu_type == SRP_IU_TSK_MGMT) ? 0 : SRP_TSK_MGMT_SQ_SIZE;
 	struct srp_iu *iu;
 
-	srp_send_completion(ch->send_cq, ch);
+	ib_process_cq_direct(ch->send_cq, -1);
 
 	if (list_empty(&ch->free_tx))
 		return NULL;
@@ -1739,6 +1761,19 @@
 	return iu;
 }
 
+static void srp_send_done(struct ib_cq *cq, struct ib_wc *wc)
+{
+	struct srp_iu *iu = container_of(wc->wr_cqe, struct srp_iu, cqe);
+	struct srp_rdma_ch *ch = cq->cq_context;
+
+	if (unlikely(wc->status != IB_WC_SUCCESS)) {
+		srp_handle_qp_err(cq, wc, "SEND");
+		return;
+	}
+
+	list_add(&iu->list, &ch->free_tx);
+}
+
 static int srp_post_send(struct srp_rdma_ch *ch, struct srp_iu *iu, int len)
 {
 	struct srp_target_port *target = ch->target;
@@ -1749,8 +1784,10 @@
 	list.length = len;
 	list.lkey   = target->lkey;
 
+	iu->cqe.done = srp_send_done;
+
 	wr.next       = NULL;
-	wr.wr_id      = (uintptr_t) iu;
+	wr.wr_cqe     = &iu->cqe;
 	wr.sg_list    = &list;
 	wr.num_sge    = 1;
 	wr.opcode     = IB_WR_SEND;
@@ -1769,8 +1806,10 @@
 	list.length = iu->size;
 	list.lkey   = target->lkey;
 
+	iu->cqe.done = srp_recv_done;
+
 	wr.next     = NULL;
-	wr.wr_id    = (uintptr_t) iu;
+	wr.wr_cqe   = &iu->cqe;
 	wr.sg_list  = &list;
 	wr.num_sge  = 1;
 
@@ -1902,14 +1941,20 @@
 			     "problems processing SRP_AER_REQ\n");
 }
 
-static void srp_handle_recv(struct srp_rdma_ch *ch, struct ib_wc *wc)
+static void srp_recv_done(struct ib_cq *cq, struct ib_wc *wc)
 {
+	struct srp_iu *iu = container_of(wc->wr_cqe, struct srp_iu, cqe);
+	struct srp_rdma_ch *ch = cq->cq_context;
 	struct srp_target_port *target = ch->target;
 	struct ib_device *dev = target->srp_host->srp_dev->dev;
-	struct srp_iu *iu = (struct srp_iu *) (uintptr_t) wc->wr_id;
 	int res;
 	u8 opcode;
 
+	if (unlikely(wc->status != IB_WC_SUCCESS)) {
+		srp_handle_qp_err(cq, wc, "RECV");
+		return;
+	}
+
 	ib_dma_sync_single_for_cpu(dev, iu->dma, ch->max_ti_iu_len,
 				   DMA_FROM_DEVICE);
 
@@ -1972,68 +2017,22 @@
 		srp_start_tl_fail_timers(target->rport);
 }
 
-static void srp_handle_qp_err(u64 wr_id, enum ib_wc_status wc_status,
-			      bool send_err, struct srp_rdma_ch *ch)
+static void srp_handle_qp_err(struct ib_cq *cq, struct ib_wc *wc,
+		const char *opname)
 {
+	struct srp_rdma_ch *ch = cq->cq_context;
 	struct srp_target_port *target = ch->target;
 
-	if (wr_id == SRP_LAST_WR_ID) {
-		complete(&ch->done);
-		return;
-	}
-
 	if (ch->connected && !target->qp_in_error) {
-		if (wr_id & LOCAL_INV_WR_ID_MASK) {
-			shost_printk(KERN_ERR, target->scsi_host, PFX
-				     "LOCAL_INV failed with status %s (%d)\n",
-				     ib_wc_status_msg(wc_status), wc_status);
-		} else if (wr_id & FAST_REG_WR_ID_MASK) {
-			shost_printk(KERN_ERR, target->scsi_host, PFX
-				     "FAST_REG_MR failed status %s (%d)\n",
-				     ib_wc_status_msg(wc_status), wc_status);
-		} else {
-			shost_printk(KERN_ERR, target->scsi_host,
-				     PFX "failed %s status %s (%d) for iu %p\n",
-				     send_err ? "send" : "receive",
-				     ib_wc_status_msg(wc_status), wc_status,
-				     (void *)(uintptr_t)wr_id);
-		}
+		shost_printk(KERN_ERR, target->scsi_host,
+			     PFX "failed %s status %s (%d) for CQE %p\n",
+			     opname, ib_wc_status_msg(wc->status), wc->status,
+			     wc->wr_cqe);
 		queue_work(system_long_wq, &target->tl_err_work);
 	}
 	target->qp_in_error = true;
 }
 
-static void srp_recv_completion(struct ib_cq *cq, void *ch_ptr)
-{
-	struct srp_rdma_ch *ch = ch_ptr;
-	struct ib_wc wc;
-
-	ib_req_notify_cq(cq, IB_CQ_NEXT_COMP);
-	while (ib_poll_cq(cq, 1, &wc) > 0) {
-		if (likely(wc.status == IB_WC_SUCCESS)) {
-			srp_handle_recv(ch, &wc);
-		} else {
-			srp_handle_qp_err(wc.wr_id, wc.status, false, ch);
-		}
-	}
-}
-
-static void srp_send_completion(struct ib_cq *cq, void *ch_ptr)
-{
-	struct srp_rdma_ch *ch = ch_ptr;
-	struct ib_wc wc;
-	struct srp_iu *iu;
-
-	while (ib_poll_cq(cq, 1, &wc) > 0) {
-		if (likely(wc.status == IB_WC_SUCCESS)) {
-			iu = (struct srp_iu *) (uintptr_t) wc.wr_id;
-			list_add(&iu->list, &ch->free_tx);
-		} else {
-			srp_handle_qp_err(wc.wr_id, wc.status, true, ch);
-		}
-	}
-}
-
 static int srp_queuecommand(struct Scsi_Host *shost, struct scsi_cmnd *scmnd)
 {
 	struct srp_target_port *target = host_to_target(shost);
@@ -3587,8 +3586,6 @@
 {
 	int ret;
 
-	BUILD_BUG_ON(FIELD_SIZEOF(struct ib_wc, wr_id) < sizeof(void *));
-
 	if (srp_sg_tablesize) {
 		pr_warn("srp_sg_tablesize is deprecated, please use cmd_sg_entries\n");
 		if (!cmd_sg_entries)
diff --git a/drivers/infiniband/ulp/srp/ib_srp.h b/drivers/infiniband/ulp/srp/ib_srp.h
index f6af531..9e05ce4 100644
--- a/drivers/infiniband/ulp/srp/ib_srp.h
+++ b/drivers/infiniband/ulp/srp/ib_srp.h
@@ -66,11 +66,6 @@
 	SRP_TAG_TSK_MGMT	= 1U << 31,
 
 	SRP_MAX_PAGES_PER_MR	= 512,
-
-	LOCAL_INV_WR_ID_MASK	= 1,
-	FAST_REG_WR_ID_MASK	= 2,
-
-	SRP_LAST_WR_ID		= 0xfffffffcU,
 };
 
 enum srp_target_state {
@@ -128,6 +123,7 @@
 	struct srp_direct_buf  *indirect_desc;
 	dma_addr_t		indirect_dma_addr;
 	short			nmdesc;
+	struct ib_cqe		reg_cqe;
 };
 
 /**
@@ -231,6 +227,7 @@
 	void		       *buf;
 	size_t			size;
 	enum dma_data_direction	direction;
+	struct ib_cqe		cqe;
 };
 
 /**
diff --git a/drivers/infiniband/ulp/srpt/ib_srpt.c b/drivers/infiniband/ulp/srpt/ib_srpt.c
index 2e2fe81..8068aff 100644
--- a/drivers/infiniband/ulp/srpt/ib_srpt.c
+++ b/drivers/infiniband/ulp/srpt/ib_srpt.c
@@ -93,6 +93,8 @@
 static struct ib_client srpt_client;
 static void srpt_release_channel(struct srpt_rdma_ch *ch);
 static int srpt_queue_status(struct se_cmd *cmd);
+static void srpt_recv_done(struct ib_cq *cq, struct ib_wc *wc);
+static void srpt_send_done(struct ib_cq *cq, struct ib_wc *wc);
 
 /**
  * opposite_dma_dir() - Swap DMA_TO_DEVICE and DMA_FROM_DEVICE.
@@ -778,12 +780,12 @@
 	struct ib_recv_wr wr, *bad_wr;
 
 	BUG_ON(!sdev);
-	wr.wr_id = encode_wr_id(SRPT_RECV, ioctx->ioctx.index);
-
 	list.addr = ioctx->ioctx.dma;
 	list.length = srp_max_req_size;
 	list.lkey = sdev->pd->local_dma_lkey;
 
+	ioctx->ioctx.cqe.done = srpt_recv_done;
+	wr.wr_cqe = &ioctx->ioctx.cqe;
 	wr.next = NULL;
 	wr.sg_list = &list;
 	wr.num_sge = 1;
@@ -819,8 +821,9 @@
 	list.length = len;
 	list.lkey = sdev->pd->local_dma_lkey;
 
+	ioctx->ioctx.cqe.done = srpt_send_done;
 	wr.next = NULL;
-	wr.wr_id = encode_wr_id(SRPT_SEND, ioctx->ioctx.index);
+	wr.wr_cqe = &ioctx->ioctx.cqe;
 	wr.sg_list = &list;
 	wr.num_sge = 1;
 	wr.opcode = IB_WR_SEND;
@@ -1052,13 +1055,13 @@
 
 	BUG_ON(!ch);
 	BUG_ON(!ioctx);
-	BUG_ON(ioctx->n_rdma && !ioctx->rdma_ius);
+	BUG_ON(ioctx->n_rdma && !ioctx->rdma_wrs);
 
 	while (ioctx->n_rdma)
-		kfree(ioctx->rdma_ius[--ioctx->n_rdma].sge);
+		kfree(ioctx->rdma_wrs[--ioctx->n_rdma].wr.sg_list);
 
-	kfree(ioctx->rdma_ius);
-	ioctx->rdma_ius = NULL;
+	kfree(ioctx->rdma_wrs);
+	ioctx->rdma_wrs = NULL;
 
 	if (ioctx->mapped_sg_count) {
 		sg = ioctx->sg;
@@ -1082,7 +1085,7 @@
 	struct scatterlist *sg, *sg_orig;
 	int sg_cnt;
 	enum dma_data_direction dir;
-	struct rdma_iu *riu;
+	struct ib_rdma_wr *riu;
 	struct srp_direct_buf *db;
 	dma_addr_t dma_addr;
 	struct ib_sge *sge;
@@ -1109,23 +1112,24 @@
 
 	ioctx->mapped_sg_count = count;
 
-	if (ioctx->rdma_ius && ioctx->n_rdma_ius)
-		nrdma = ioctx->n_rdma_ius;
+	if (ioctx->rdma_wrs && ioctx->n_rdma_wrs)
+		nrdma = ioctx->n_rdma_wrs;
 	else {
 		nrdma = (count + SRPT_DEF_SG_PER_WQE - 1) / SRPT_DEF_SG_PER_WQE
 			+ ioctx->n_rbuf;
 
-		ioctx->rdma_ius = kzalloc(nrdma * sizeof *riu, GFP_KERNEL);
-		if (!ioctx->rdma_ius)
+		ioctx->rdma_wrs = kcalloc(nrdma, sizeof(*ioctx->rdma_wrs),
+				GFP_KERNEL);
+		if (!ioctx->rdma_wrs)
 			goto free_mem;
 
-		ioctx->n_rdma_ius = nrdma;
+		ioctx->n_rdma_wrs = nrdma;
 	}
 
 	db = ioctx->rbufs;
 	tsize = cmd->data_length;
 	dma_len = ib_sg_dma_len(dev, &sg[0]);
-	riu = ioctx->rdma_ius;
+	riu = ioctx->rdma_wrs;
 
 	/*
 	 * For each remote desc - calculate the #ib_sge.
@@ -1139,9 +1143,9 @@
 	     j < count && i < ioctx->n_rbuf && tsize > 0; ++i, ++riu, ++db) {
 		rsize = be32_to_cpu(db->len);
 		raddr = be64_to_cpu(db->va);
-		riu->raddr = raddr;
+		riu->remote_addr = raddr;
 		riu->rkey = be32_to_cpu(db->key);
-		riu->sge_cnt = 0;
+		riu->wr.num_sge = 0;
 
 		/* calculate how many sge required for this remote_buf */
 		while (rsize > 0 && tsize > 0) {
@@ -1165,33 +1169,35 @@
 				rsize = 0;
 			}
 
-			++riu->sge_cnt;
+			++riu->wr.num_sge;
 
-			if (rsize > 0 && riu->sge_cnt == SRPT_DEF_SG_PER_WQE) {
+			if (rsize > 0 &&
+			    riu->wr.num_sge == SRPT_DEF_SG_PER_WQE) {
 				++ioctx->n_rdma;
-				riu->sge =
-				    kmalloc(riu->sge_cnt * sizeof *riu->sge,
-					    GFP_KERNEL);
-				if (!riu->sge)
+				riu->wr.sg_list = kmalloc_array(riu->wr.num_sge,
+						sizeof(*riu->wr.sg_list),
+						GFP_KERNEL);
+				if (!riu->wr.sg_list)
 					goto free_mem;
 
 				++riu;
-				riu->sge_cnt = 0;
-				riu->raddr = raddr;
+				riu->wr.num_sge = 0;
+				riu->remote_addr = raddr;
 				riu->rkey = be32_to_cpu(db->key);
 			}
 		}
 
 		++ioctx->n_rdma;
-		riu->sge = kmalloc(riu->sge_cnt * sizeof *riu->sge,
-				   GFP_KERNEL);
-		if (!riu->sge)
+		riu->wr.sg_list = kmalloc_array(riu->wr.num_sge,
+					sizeof(*riu->wr.sg_list),
+					GFP_KERNEL);
+		if (!riu->wr.sg_list)
 			goto free_mem;
 	}
 
 	db = ioctx->rbufs;
 	tsize = cmd->data_length;
-	riu = ioctx->rdma_ius;
+	riu = ioctx->rdma_wrs;
 	sg = sg_orig;
 	dma_len = ib_sg_dma_len(dev, &sg[0]);
 	dma_addr = ib_sg_dma_address(dev, &sg[0]);
@@ -1200,7 +1206,7 @@
 	for (i = 0, j = 0;
 	     j < count && i < ioctx->n_rbuf && tsize > 0; ++i, ++riu, ++db) {
 		rsize = be32_to_cpu(db->len);
-		sge = riu->sge;
+		sge = riu->wr.sg_list;
 		k = 0;
 
 		while (rsize > 0 && tsize > 0) {
@@ -1232,9 +1238,9 @@
 			}
 
 			++k;
-			if (k == riu->sge_cnt && rsize > 0 && tsize > 0) {
+			if (k == riu->wr.num_sge && rsize > 0 && tsize > 0) {
 				++riu;
-				sge = riu->sge;
+				sge = riu->wr.sg_list;
 				k = 0;
 			} else if (rsize > 0 && tsize > 0)
 				++sge;
@@ -1277,8 +1283,8 @@
 	ioctx->n_rbuf = 0;
 	ioctx->rbufs = NULL;
 	ioctx->n_rdma = 0;
-	ioctx->n_rdma_ius = 0;
-	ioctx->rdma_ius = NULL;
+	ioctx->n_rdma_wrs = 0;
+	ioctx->rdma_wrs = NULL;
 	ioctx->mapped_sg_count = 0;
 	init_completion(&ioctx->tx_done);
 	ioctx->queue_status_only = false;
@@ -1380,118 +1386,44 @@
 }
 
 /**
- * srpt_handle_send_err_comp() - Process an IB_WC_SEND error completion.
- */
-static void srpt_handle_send_err_comp(struct srpt_rdma_ch *ch, u64 wr_id)
-{
-	struct srpt_send_ioctx *ioctx;
-	enum srpt_command_state state;
-	u32 index;
-
-	atomic_inc(&ch->sq_wr_avail);
-
-	index = idx_from_wr_id(wr_id);
-	ioctx = ch->ioctx_ring[index];
-	state = srpt_get_cmd_state(ioctx);
-
-	WARN_ON(state != SRPT_STATE_CMD_RSP_SENT
-		&& state != SRPT_STATE_MGMT_RSP_SENT
-		&& state != SRPT_STATE_NEED_DATA
-		&& state != SRPT_STATE_DONE);
-
-	/* If SRP_RSP sending failed, undo the ch->req_lim change. */
-	if (state == SRPT_STATE_CMD_RSP_SENT
-	    || state == SRPT_STATE_MGMT_RSP_SENT)
-		atomic_dec(&ch->req_lim);
-
-	srpt_abort_cmd(ioctx);
-}
-
-/**
- * srpt_handle_send_comp() - Process an IB send completion notification.
- */
-static void srpt_handle_send_comp(struct srpt_rdma_ch *ch,
-				  struct srpt_send_ioctx *ioctx)
-{
-	enum srpt_command_state state;
-
-	atomic_inc(&ch->sq_wr_avail);
-
-	state = srpt_set_cmd_state(ioctx, SRPT_STATE_DONE);
-
-	if (WARN_ON(state != SRPT_STATE_CMD_RSP_SENT
-		    && state != SRPT_STATE_MGMT_RSP_SENT
-		    && state != SRPT_STATE_DONE))
-		pr_debug("state = %d\n", state);
-
-	if (state != SRPT_STATE_DONE) {
-		srpt_unmap_sg_to_ib_sge(ch, ioctx);
-		transport_generic_free_cmd(&ioctx->cmd, 0);
-	} else {
-		pr_err("IB completion has been received too late for"
-		       " wr_id = %u.\n", ioctx->ioctx.index);
-	}
-}
-
-/**
- * srpt_handle_rdma_comp() - Process an IB RDMA completion notification.
- *
  * XXX: what is now target_execute_cmd used to be asynchronous, and unmapping
  * the data that has been transferred via IB RDMA had to be postponed until the
  * check_stop_free() callback.  None of this is necessary anymore and needs to
  * be cleaned up.
  */
-static void srpt_handle_rdma_comp(struct srpt_rdma_ch *ch,
-				  struct srpt_send_ioctx *ioctx,
-				  enum srpt_opcode opcode)
+static void srpt_rdma_read_done(struct ib_cq *cq, struct ib_wc *wc)
 {
+	struct srpt_rdma_ch *ch = cq->cq_context;
+	struct srpt_send_ioctx *ioctx =
+		container_of(wc->wr_cqe, struct srpt_send_ioctx, ioctx.cqe);
+
 	WARN_ON(ioctx->n_rdma <= 0);
 	atomic_add(ioctx->n_rdma, &ch->sq_wr_avail);
 
-	if (opcode == SRPT_RDMA_READ_LAST) {
-		if (srpt_test_and_set_cmd_state(ioctx, SRPT_STATE_NEED_DATA,
-						SRPT_STATE_DATA_IN))
-			target_execute_cmd(&ioctx->cmd);
-		else
-			pr_err("%s[%d]: wrong state = %d\n", __func__,
-			       __LINE__, srpt_get_cmd_state(ioctx));
-	} else if (opcode == SRPT_RDMA_ABORT) {
-		ioctx->rdma_aborted = true;
-	} else {
-		WARN(true, "unexpected opcode %d\n", opcode);
+	if (unlikely(wc->status != IB_WC_SUCCESS)) {
+		pr_info("RDMA_READ for ioctx 0x%p failed with status %d\n",
+			ioctx, wc->status);
+		srpt_abort_cmd(ioctx);
+		return;
 	}
+
+	if (srpt_test_and_set_cmd_state(ioctx, SRPT_STATE_NEED_DATA,
+					SRPT_STATE_DATA_IN))
+		target_execute_cmd(&ioctx->cmd);
+	else
+		pr_err("%s[%d]: wrong state = %d\n", __func__,
+		       __LINE__, srpt_get_cmd_state(ioctx));
 }
 
-/**
- * srpt_handle_rdma_err_comp() - Process an IB RDMA error completion.
- */
-static void srpt_handle_rdma_err_comp(struct srpt_rdma_ch *ch,
-				      struct srpt_send_ioctx *ioctx,
-				      enum srpt_opcode opcode)
+static void srpt_rdma_write_done(struct ib_cq *cq, struct ib_wc *wc)
 {
-	enum srpt_command_state state;
+	struct srpt_send_ioctx *ioctx =
+		container_of(wc->wr_cqe, struct srpt_send_ioctx, ioctx.cqe);
 
-	state = srpt_get_cmd_state(ioctx);
-	switch (opcode) {
-	case SRPT_RDMA_READ_LAST:
-		if (ioctx->n_rdma <= 0) {
-			pr_err("Received invalid RDMA read"
-			       " error completion with idx %d\n",
-			       ioctx->ioctx.index);
-			break;
-		}
-		atomic_add(ioctx->n_rdma, &ch->sq_wr_avail);
-		if (state == SRPT_STATE_NEED_DATA)
-			srpt_abort_cmd(ioctx);
-		else
-			pr_err("%s[%d]: wrong state = %d\n",
-			       __func__, __LINE__, state);
-		break;
-	case SRPT_RDMA_WRITE_LAST:
-		break;
-	default:
-		pr_err("%s[%d]: opcode = %u\n", __func__, __LINE__, opcode);
-		break;
+	if (unlikely(wc->status != IB_WC_SUCCESS)) {
+		pr_info("RDMA_WRITE for ioctx 0x%p failed with status %d\n",
+			ioctx, wc->status);
+		srpt_abort_cmd(ioctx);
 	}
 }
 
@@ -1926,32 +1858,26 @@
 	return;
 }
 
-static void srpt_process_rcv_completion(struct ib_cq *cq,
-					struct srpt_rdma_ch *ch,
-					struct ib_wc *wc)
+static void srpt_recv_done(struct ib_cq *cq, struct ib_wc *wc)
 {
-	struct srpt_device *sdev = ch->sport->sdev;
-	struct srpt_recv_ioctx *ioctx;
-	u32 index;
+	struct srpt_rdma_ch *ch = cq->cq_context;
+	struct srpt_recv_ioctx *ioctx =
+		container_of(wc->wr_cqe, struct srpt_recv_ioctx, ioctx.cqe);
 
-	index = idx_from_wr_id(wc->wr_id);
 	if (wc->status == IB_WC_SUCCESS) {
 		int req_lim;
 
 		req_lim = atomic_dec_return(&ch->req_lim);
 		if (unlikely(req_lim < 0))
 			pr_err("req_lim = %d < 0\n", req_lim);
-		ioctx = sdev->ioctx_ring[index];
 		srpt_handle_new_iu(ch, ioctx, NULL);
 	} else {
-		pr_info("receiving failed for idx %u with status %d\n",
-			index, wc->status);
+		pr_info("receiving failed for ioctx %p with status %d\n",
+			ioctx, wc->status);
 	}
 }
 
 /**
- * srpt_process_send_completion() - Process an IB send completion.
- *
  * Note: Although this has not yet been observed during tests, at least in
  * theory it is possible that the srpt_get_send_ioctx() call invoked by
  * srpt_handle_new_iu() fails. This is possible because the req_lim_delta
@@ -1964,108 +1890,51 @@
  * are queued on cmd_wait_list. The code below processes these delayed
  * requests one at a time.
  */
-static void srpt_process_send_completion(struct ib_cq *cq,
-					 struct srpt_rdma_ch *ch,
-					 struct ib_wc *wc)
+static void srpt_send_done(struct ib_cq *cq, struct ib_wc *wc)
 {
-	struct srpt_send_ioctx *send_ioctx;
-	uint32_t index;
-	enum srpt_opcode opcode;
+	struct srpt_rdma_ch *ch = cq->cq_context;
+	struct srpt_send_ioctx *ioctx =
+		container_of(wc->wr_cqe, struct srpt_send_ioctx, ioctx.cqe);
+	enum srpt_command_state state;
 
-	index = idx_from_wr_id(wc->wr_id);
-	opcode = opcode_from_wr_id(wc->wr_id);
-	send_ioctx = ch->ioctx_ring[index];
-	if (wc->status == IB_WC_SUCCESS) {
-		if (opcode == SRPT_SEND)
-			srpt_handle_send_comp(ch, send_ioctx);
-		else {
-			WARN_ON(opcode != SRPT_RDMA_ABORT &&
-				wc->opcode != IB_WC_RDMA_READ);
-			srpt_handle_rdma_comp(ch, send_ioctx, opcode);
-		}
-	} else {
-		if (opcode == SRPT_SEND) {
-			pr_info("sending response for idx %u failed"
-				" with status %d\n", index, wc->status);
-			srpt_handle_send_err_comp(ch, wc->wr_id);
-		} else if (opcode != SRPT_RDMA_MID) {
-			pr_info("RDMA t %d for idx %u failed with"
-				" status %d\n", opcode, index, wc->status);
-			srpt_handle_rdma_err_comp(ch, send_ioctx, opcode);
-		}
+	state = srpt_set_cmd_state(ioctx, SRPT_STATE_DONE);
+
+	WARN_ON(state != SRPT_STATE_CMD_RSP_SENT &&
+		state != SRPT_STATE_MGMT_RSP_SENT);
+
+	atomic_inc(&ch->sq_wr_avail);
+
+	if (wc->status != IB_WC_SUCCESS) {
+		pr_info("sending response for ioctx 0x%p failed"
+			" with status %d\n", ioctx, wc->status);
+
+		atomic_dec(&ch->req_lim);
+		srpt_abort_cmd(ioctx);
+		goto out;
 	}
 
-	while (unlikely(opcode == SRPT_SEND
-			&& !list_empty(&ch->cmd_wait_list)
-			&& srpt_get_ch_state(ch) == CH_LIVE
-			&& (send_ioctx = srpt_get_send_ioctx(ch)) != NULL)) {
+	if (state != SRPT_STATE_DONE) {
+		srpt_unmap_sg_to_ib_sge(ch, ioctx);
+		transport_generic_free_cmd(&ioctx->cmd, 0);
+	} else {
+		pr_err("IB completion has been received too late for"
+		       " wr_id = %u.\n", ioctx->ioctx.index);
+	}
+
+out:
+	while (!list_empty(&ch->cmd_wait_list) &&
+	       srpt_get_ch_state(ch) == CH_LIVE &&
+	       (ioctx = srpt_get_send_ioctx(ch)) != NULL) {
 		struct srpt_recv_ioctx *recv_ioctx;
 
 		recv_ioctx = list_first_entry(&ch->cmd_wait_list,
 					      struct srpt_recv_ioctx,
 					      wait_list);
 		list_del(&recv_ioctx->wait_list);
-		srpt_handle_new_iu(ch, recv_ioctx, send_ioctx);
+		srpt_handle_new_iu(ch, recv_ioctx, ioctx);
 	}
 }
 
-static void srpt_process_completion(struct ib_cq *cq, struct srpt_rdma_ch *ch)
-{
-	struct ib_wc *const wc = ch->wc;
-	int i, n;
-
-	WARN_ON(cq != ch->cq);
-
-	ib_req_notify_cq(cq, IB_CQ_NEXT_COMP);
-	while ((n = ib_poll_cq(cq, ARRAY_SIZE(ch->wc), wc)) > 0) {
-		for (i = 0; i < n; i++) {
-			if (opcode_from_wr_id(wc[i].wr_id) == SRPT_RECV)
-				srpt_process_rcv_completion(cq, ch, &wc[i]);
-			else
-				srpt_process_send_completion(cq, ch, &wc[i]);
-		}
-	}
-}
-
-/**
- * srpt_completion() - IB completion queue callback function.
- *
- * Notes:
- * - It is guaranteed that a completion handler will never be invoked
- *   concurrently on two different CPUs for the same completion queue. See also
- *   Documentation/infiniband/core_locking.txt and the implementation of
- *   handle_edge_irq() in kernel/irq/chip.c.
- * - When threaded IRQs are enabled, completion handlers are invoked in thread
- *   context instead of interrupt context.
- */
-static void srpt_completion(struct ib_cq *cq, void *ctx)
-{
-	struct srpt_rdma_ch *ch = ctx;
-
-	wake_up_interruptible(&ch->wait_queue);
-}
-
-static int srpt_compl_thread(void *arg)
-{
-	struct srpt_rdma_ch *ch;
-
-	/* Hibernation / freezing of the SRPT kernel thread is not supported. */
-	current->flags |= PF_NOFREEZE;
-
-	ch = arg;
-	BUG_ON(!ch);
-	pr_info("Session %s: kernel thread %s (PID %d) started\n",
-		ch->sess_name, ch->thread->comm, current->pid);
-	while (!kthread_should_stop()) {
-		wait_event_interruptible(ch->wait_queue,
-			(srpt_process_completion(ch->cq, ch),
-			 kthread_should_stop()));
-	}
-	pr_info("Session %s: kernel thread %s (PID %d) stopped\n",
-		ch->sess_name, ch->thread->comm, current->pid);
-	return 0;
-}
-
 /**
  * srpt_create_ch_ib() - Create receive and send completion queues.
  */
@@ -2075,7 +1944,6 @@
 	struct srpt_port *sport = ch->sport;
 	struct srpt_device *sdev = sport->sdev;
 	u32 srp_sq_size = sport->port_attrib.srp_sq_size;
-	struct ib_cq_init_attr cq_attr = {};
 	int ret;
 
 	WARN_ON(ch->rq_size < 1);
@@ -2086,9 +1954,8 @@
 		goto out;
 
 retry:
-	cq_attr.cqe = ch->rq_size + srp_sq_size;
-	ch->cq = ib_create_cq(sdev->device, srpt_completion, NULL, ch,
-			      &cq_attr);
+	ch->cq = ib_alloc_cq(sdev->device, ch, ch->rq_size + srp_sq_size,
+			0 /* XXX: spread CQs */, IB_POLL_WORKQUEUE);
 	if (IS_ERR(ch->cq)) {
 		ret = PTR_ERR(ch->cq);
 		pr_err("failed to create CQ cqe= %d ret= %d\n",
@@ -2131,18 +1998,6 @@
 	if (ret)
 		goto err_destroy_qp;
 
-	init_waitqueue_head(&ch->wait_queue);
-
-	pr_debug("creating thread for session %s\n", ch->sess_name);
-
-	ch->thread = kthread_run(srpt_compl_thread, ch, "ib_srpt_compl");
-	if (IS_ERR(ch->thread)) {
-		pr_err("failed to create kernel thread %ld\n",
-		       PTR_ERR(ch->thread));
-		ch->thread = NULL;
-		goto err_destroy_qp;
-	}
-
 out:
 	kfree(qp_init);
 	return ret;
@@ -2150,17 +2005,14 @@
 err_destroy_qp:
 	ib_destroy_qp(ch->qp);
 err_destroy_cq:
-	ib_destroy_cq(ch->cq);
+	ib_free_cq(ch->cq);
 	goto out;
 }
 
 static void srpt_destroy_ch_ib(struct srpt_rdma_ch *ch)
 {
-	if (ch->thread)
-		kthread_stop(ch->thread);
-
 	ib_destroy_qp(ch->qp);
-	ib_destroy_cq(ch->cq);
+	ib_free_cq(ch->cq);
 }
 
 /**
@@ -2821,12 +2673,8 @@
 static int srpt_perform_rdmas(struct srpt_rdma_ch *ch,
 			      struct srpt_send_ioctx *ioctx)
 {
-	struct ib_rdma_wr wr;
 	struct ib_send_wr *bad_wr;
-	struct rdma_iu *riu;
-	int i;
-	int ret;
-	int sq_wr_avail;
+	int sq_wr_avail, ret, i;
 	enum dma_data_direction dir;
 	const int n_rdma = ioctx->n_rdma;
 
@@ -2842,59 +2690,32 @@
 		}
 	}
 
-	ioctx->rdma_aborted = false;
-	ret = 0;
-	riu = ioctx->rdma_ius;
-	memset(&wr, 0, sizeof wr);
+	for (i = 0; i < n_rdma; i++) {
+		struct ib_send_wr *wr = &ioctx->rdma_wrs[i].wr;
 
-	for (i = 0; i < n_rdma; ++i, ++riu) {
-		if (dir == DMA_FROM_DEVICE) {
-			wr.wr.opcode = IB_WR_RDMA_WRITE;
-			wr.wr.wr_id = encode_wr_id(i == n_rdma - 1 ?
-						SRPT_RDMA_WRITE_LAST :
-						SRPT_RDMA_MID,
-						ioctx->ioctx.index);
+		wr->opcode = (dir == DMA_FROM_DEVICE) ?
+				IB_WR_RDMA_WRITE : IB_WR_RDMA_READ;
+
+		if (i == n_rdma - 1) {
+			/* only get completion event for the last rdma read */
+			if (dir == DMA_TO_DEVICE) {
+				wr->send_flags = IB_SEND_SIGNALED;
+				ioctx->rdma_cqe.done = srpt_rdma_read_done;
+			} else {
+				ioctx->rdma_cqe.done = srpt_rdma_write_done;
+			}
+			wr->wr_cqe = &ioctx->rdma_cqe;
+			wr->next = NULL;
 		} else {
-			wr.wr.opcode = IB_WR_RDMA_READ;
-			wr.wr.wr_id = encode_wr_id(i == n_rdma - 1 ?
-						SRPT_RDMA_READ_LAST :
-						SRPT_RDMA_MID,
-						ioctx->ioctx.index);
+			wr->wr_cqe = NULL;
+			wr->next = &ioctx->rdma_wrs[i + 1].wr;
 		}
-		wr.wr.next = NULL;
-		wr.remote_addr = riu->raddr;
-		wr.rkey = riu->rkey;
-		wr.wr.num_sge = riu->sge_cnt;
-		wr.wr.sg_list = riu->sge;
-
-		/* only get completion event for the last rdma write */
-		if (i == (n_rdma - 1) && dir == DMA_TO_DEVICE)
-			wr.wr.send_flags = IB_SEND_SIGNALED;
-
-		ret = ib_post_send(ch->qp, &wr.wr, &bad_wr);
-		if (ret)
-			break;
 	}
 
+	ret = ib_post_send(ch->qp, &ioctx->rdma_wrs->wr, &bad_wr);
 	if (ret)
 		pr_err("%s[%d]: ib_post_send() returned %d for %d/%d\n",
 				 __func__, __LINE__, ret, i, n_rdma);
-	if (ret && i > 0) {
-		wr.wr.num_sge = 0;
-		wr.wr.wr_id = encode_wr_id(SRPT_RDMA_ABORT, ioctx->ioctx.index);
-		wr.wr.send_flags = IB_SEND_SIGNALED;
-		while (ch->state == CH_LIVE &&
-			ib_post_send(ch->qp, &wr.wr, &bad_wr) != 0) {
-			pr_info("Trying to abort failed RDMA transfer [%d]\n",
-				ioctx->ioctx.index);
-			msleep(1000);
-		}
-		while (ch->state != CH_RELEASING && !ioctx->rdma_aborted) {
-			pr_info("Waiting until RDMA abort finished [%d]\n",
-				ioctx->ioctx.index);
-			msleep(1000);
-		}
-	}
 out:
 	if (unlikely(dir == DMA_TO_DEVICE && ret < 0))
 		atomic_add(n_rdma, &ch->sq_wr_avail);
diff --git a/drivers/infiniband/ulp/srpt/ib_srpt.h b/drivers/infiniband/ulp/srpt/ib_srpt.h
index 5faad8ac..a98b86b 100644
--- a/drivers/infiniband/ulp/srpt/ib_srpt.h
+++ b/drivers/infiniband/ulp/srpt/ib_srpt.h
@@ -128,36 +128,6 @@
 	DEFAULT_MAX_RDMA_SIZE = 65536,
 };
 
-enum srpt_opcode {
-	SRPT_RECV,
-	SRPT_SEND,
-	SRPT_RDMA_MID,
-	SRPT_RDMA_ABORT,
-	SRPT_RDMA_READ_LAST,
-	SRPT_RDMA_WRITE_LAST,
-};
-
-static inline u64 encode_wr_id(u8 opcode, u32 idx)
-{
-	return ((u64)opcode << 32) | idx;
-}
-static inline enum srpt_opcode opcode_from_wr_id(u64 wr_id)
-{
-	return wr_id >> 32;
-}
-static inline u32 idx_from_wr_id(u64 wr_id)
-{
-	return (u32)wr_id;
-}
-
-struct rdma_iu {
-	u64		raddr;
-	u32		rkey;
-	struct ib_sge	*sge;
-	u32		sge_cnt;
-	int		mem_id;
-};
-
 /**
  * enum srpt_command_state - SCSI command state managed by SRPT.
  * @SRPT_STATE_NEW:           New command arrived and is being processed.
@@ -189,6 +159,7 @@
  * @index: Index of the I/O context in its ioctx_ring array.
  */
 struct srpt_ioctx {
+	struct ib_cqe		cqe;
 	void			*buf;
 	dma_addr_t		dma;
 	uint32_t		index;
@@ -215,32 +186,30 @@
  * @sg:          Pointer to sg-list associated with this I/O context.
  * @sg_cnt:      SG-list size.
  * @mapped_sg_count: ib_dma_map_sg() return value.
- * @n_rdma_ius:  Number of elements in the rdma_ius array.
- * @rdma_ius:    Array with information about the RDMA mapping.
+ * @n_rdma_wrs:  Number of elements in the rdma_wrs array.
+ * @rdma_wrs:    Array with information about the RDMA mapping.
  * @tag:         Tag of the received SRP information unit.
  * @spinlock:    Protects 'state'.
  * @state:       I/O context state.
- * @rdma_aborted: If initiating a multipart RDMA transfer failed, whether
- * 		 the already initiated transfers have finished.
  * @cmd:         Target core command data structure.
  * @sense_data:  SCSI sense data.
  */
 struct srpt_send_ioctx {
 	struct srpt_ioctx	ioctx;
 	struct srpt_rdma_ch	*ch;
-	struct rdma_iu		*rdma_ius;
+	struct ib_rdma_wr	*rdma_wrs;
+	struct ib_cqe		rdma_cqe;
 	struct srp_direct_buf	*rbufs;
 	struct srp_direct_buf	single_rbuf;
 	struct scatterlist	*sg;
 	struct list_head	free_list;
 	spinlock_t		spinlock;
 	enum srpt_command_state	state;
-	bool			rdma_aborted;
 	struct se_cmd		cmd;
 	struct completion	tx_done;
 	int			sg_cnt;
 	int			mapped_sg_count;
-	u16			n_rdma_ius;
+	u16			n_rdma_wrs;
 	u8			n_rdma;
 	u8			n_rbuf;
 	bool			queue_status_only;
@@ -267,9 +236,6 @@
 
 /**
  * struct srpt_rdma_ch - RDMA channel.
- * @wait_queue:    Allows the kernel thread to wait for more work.
- * @thread:        Kernel thread that processes the IB queues associated with
- *                 the channel.
  * @cm_id:         IB CM ID associated with the channel.
  * @qp:            IB queue pair used for communicating over this channel.
  * @cq:            IB completion queue for this channel.
@@ -299,8 +265,6 @@
  * @release_done:  Enables waiting for srpt_release_channel() completion.
  */
 struct srpt_rdma_ch {
-	wait_queue_head_t	wait_queue;
-	struct task_struct	*thread;
 	struct ib_cm_id		*cm_id;
 	struct ib_qp		*qp;
 	struct ib_cq		*cq;
diff --git a/drivers/scsi/Kconfig b/drivers/scsi/Kconfig
index 64eed87..9900e61 100644
--- a/drivers/scsi/Kconfig
+++ b/drivers/scsi/Kconfig
@@ -1104,6 +1104,7 @@
 	tristate "IBM Power Linux RAID adapter support"
 	depends on PCI && SCSI && ATA
 	select FW_LOADER
+	select IRQ_POLL
 	---help---
 	  This driver supports the IBM Power Linux family RAID adapters.
 	  This includes IBM pSeries 5712, 5703, 5709, and 570A, as well
diff --git a/drivers/scsi/be2iscsi/Kconfig b/drivers/scsi/be2iscsi/Kconfig
index 4e7cad2..bad5f32 100644
--- a/drivers/scsi/be2iscsi/Kconfig
+++ b/drivers/scsi/be2iscsi/Kconfig
@@ -3,6 +3,7 @@
 	depends on PCI && SCSI && NET
 	select SCSI_ISCSI_ATTRS
 	select ISCSI_BOOT_SYSFS
+	select IRQ_POLL
 
 	help
 	This driver implements the iSCSI functionality for Emulex
diff --git a/drivers/scsi/be2iscsi/be.h b/drivers/scsi/be2iscsi/be.h
index 77f992e..a41c643 100644
--- a/drivers/scsi/be2iscsi/be.h
+++ b/drivers/scsi/be2iscsi/be.h
@@ -20,7 +20,7 @@
 
 #include <linux/pci.h>
 #include <linux/if_vlan.h>
-#include <linux/blk-iopoll.h>
+#include <linux/irq_poll.h>
 #define FW_VER_LEN	32
 #define MCC_Q_LEN	128
 #define MCC_CQ_LEN	256
@@ -101,7 +101,7 @@
 	struct beiscsi_hba *phba;
 	struct be_queue_info *cq;
 	struct work_struct work_cqs; /* Work Item */
-	struct blk_iopoll	iopoll;
+	struct irq_poll	iopoll;
 };
 
 struct be_mcc_obj {
diff --git a/drivers/scsi/be2iscsi/be_iscsi.c b/drivers/scsi/be2iscsi/be_iscsi.c
index b7087ba..022e87b 100644
--- a/drivers/scsi/be2iscsi/be_iscsi.c
+++ b/drivers/scsi/be2iscsi/be_iscsi.c
@@ -1292,9 +1292,9 @@
 
 	for (i = 0; i < phba->num_cpus; i++) {
 		pbe_eq = &phwi_context->be_eq[i];
-		blk_iopoll_disable(&pbe_eq->iopoll);
+		irq_poll_disable(&pbe_eq->iopoll);
 		beiscsi_process_cq(pbe_eq);
-		blk_iopoll_enable(&pbe_eq->iopoll);
+		irq_poll_enable(&pbe_eq->iopoll);
 	}
 }
 
diff --git a/drivers/scsi/be2iscsi/be_main.c b/drivers/scsi/be2iscsi/be_main.c
index fe0c514..cb9072a 100644
--- a/drivers/scsi/be2iscsi/be_main.c
+++ b/drivers/scsi/be2iscsi/be_main.c
@@ -910,8 +910,7 @@
 	num_eq_processed = 0;
 	while (eqe->dw[offsetof(struct amap_eq_entry, valid) / 32]
 				& EQE_VALID_MASK) {
-		if (!blk_iopoll_sched_prep(&pbe_eq->iopoll))
-			blk_iopoll_sched(&pbe_eq->iopoll);
+		irq_poll_sched(&pbe_eq->iopoll);
 
 		AMAP_SET_BITS(struct amap_eq_entry, valid, eqe, 0);
 		queue_tail_inc(eq);
@@ -972,8 +971,7 @@
 			spin_unlock_irqrestore(&phba->isr_lock, flags);
 			num_mcceq_processed++;
 		} else {
-			if (!blk_iopoll_sched_prep(&pbe_eq->iopoll))
-				blk_iopoll_sched(&pbe_eq->iopoll);
+			irq_poll_sched(&pbe_eq->iopoll);
 			num_ioeq_processed++;
 		}
 		AMAP_SET_BITS(struct amap_eq_entry, valid, eqe, 0);
@@ -2295,7 +2293,7 @@
 	hwi_ring_eq_db(phba, pbe_eq->q.id, 0, 0, 1, 1);
 }
 
-static int be_iopoll(struct blk_iopoll *iop, int budget)
+static int be_iopoll(struct irq_poll *iop, int budget)
 {
 	unsigned int ret;
 	struct beiscsi_hba *phba;
@@ -2306,7 +2304,7 @@
 	pbe_eq->cq_count += ret;
 	if (ret < budget) {
 		phba = pbe_eq->phba;
-		blk_iopoll_complete(iop);
+		irq_poll_complete(iop);
 		beiscsi_log(phba, KERN_INFO,
 			    BEISCSI_LOG_CONFIG | BEISCSI_LOG_IO,
 			    "BM_%d : rearm pbe_eq->q.id =%d\n",
@@ -5293,7 +5291,7 @@
 
 	for (i = 0; i < phba->num_cpus; i++) {
 		pbe_eq = &phwi_context->be_eq[i];
-		blk_iopoll_disable(&pbe_eq->iopoll);
+		irq_poll_disable(&pbe_eq->iopoll);
 	}
 
 	if (unload_state == BEISCSI_CLEAN_UNLOAD) {
@@ -5579,9 +5577,8 @@
 
 	for (i = 0; i < phba->num_cpus; i++) {
 		pbe_eq = &phwi_context->be_eq[i];
-		blk_iopoll_init(&pbe_eq->iopoll, be_iopoll_budget,
+		irq_poll_init(&pbe_eq->iopoll, be_iopoll_budget,
 				be_iopoll);
-		blk_iopoll_enable(&pbe_eq->iopoll);
 	}
 
 	i = (phba->msix_enabled) ? i : 0;
@@ -5752,9 +5749,8 @@
 
 	for (i = 0; i < phba->num_cpus; i++) {
 		pbe_eq = &phwi_context->be_eq[i];
-		blk_iopoll_init(&pbe_eq->iopoll, be_iopoll_budget,
+		irq_poll_init(&pbe_eq->iopoll, be_iopoll_budget,
 				be_iopoll);
-		blk_iopoll_enable(&pbe_eq->iopoll);
 	}
 
 	i = (phba->msix_enabled) ? i : 0;
@@ -5795,7 +5791,7 @@
 	destroy_workqueue(phba->wq);
 	for (i = 0; i < phba->num_cpus; i++) {
 		pbe_eq = &phwi_context->be_eq[i];
-		blk_iopoll_disable(&pbe_eq->iopoll);
+		irq_poll_disable(&pbe_eq->iopoll);
 	}
 free_twq:
 	beiscsi_clean_port(phba);
diff --git a/drivers/scsi/ipr.c b/drivers/scsi/ipr.c
index 536cd5a..82031e0 100644
--- a/drivers/scsi/ipr.c
+++ b/drivers/scsi/ipr.c
@@ -3638,7 +3638,7 @@
 	.store = ipr_store_reset_adapter
 };
 
-static int ipr_iopoll(struct blk_iopoll *iop, int budget);
+static int ipr_iopoll(struct irq_poll *iop, int budget);
  /**
  * ipr_show_iopoll_weight - Show ipr polling mode
  * @dev:	class device struct
@@ -3681,34 +3681,33 @@
 	int i;
 
 	if (!ioa_cfg->sis64) {
-		dev_info(&ioa_cfg->pdev->dev, "blk-iopoll not supported on this adapter\n");
+		dev_info(&ioa_cfg->pdev->dev, "irq_poll not supported on this adapter\n");
 		return -EINVAL;
 	}
 	if (kstrtoul(buf, 10, &user_iopoll_weight))
 		return -EINVAL;
 
 	if (user_iopoll_weight > 256) {
-		dev_info(&ioa_cfg->pdev->dev, "Invalid blk-iopoll weight. It must be less than 256\n");
+		dev_info(&ioa_cfg->pdev->dev, "Invalid irq_poll weight. It must be less than 256\n");
 		return -EINVAL;
 	}
 
 	if (user_iopoll_weight == ioa_cfg->iopoll_weight) {
-		dev_info(&ioa_cfg->pdev->dev, "Current blk-iopoll weight has the same weight\n");
+		dev_info(&ioa_cfg->pdev->dev, "Current irq_poll weight has the same weight\n");
 		return strlen(buf);
 	}
 
 	if (ioa_cfg->iopoll_weight && ioa_cfg->sis64 && ioa_cfg->nvectors > 1) {
 		for (i = 1; i < ioa_cfg->hrrq_num; i++)
-			blk_iopoll_disable(&ioa_cfg->hrrq[i].iopoll);
+			irq_poll_disable(&ioa_cfg->hrrq[i].iopoll);
 	}
 
 	spin_lock_irqsave(shost->host_lock, lock_flags);
 	ioa_cfg->iopoll_weight = user_iopoll_weight;
 	if (ioa_cfg->iopoll_weight && ioa_cfg->sis64 && ioa_cfg->nvectors > 1) {
 		for (i = 1; i < ioa_cfg->hrrq_num; i++) {
-			blk_iopoll_init(&ioa_cfg->hrrq[i].iopoll,
+			irq_poll_init(&ioa_cfg->hrrq[i].iopoll,
 					ioa_cfg->iopoll_weight, ipr_iopoll);
-			blk_iopoll_enable(&ioa_cfg->hrrq[i].iopoll);
 		}
 	}
 	spin_unlock_irqrestore(shost->host_lock, lock_flags);
@@ -5569,7 +5568,7 @@
 	return num_hrrq;
 }
 
-static int ipr_iopoll(struct blk_iopoll *iop, int budget)
+static int ipr_iopoll(struct irq_poll *iop, int budget)
 {
 	struct ipr_ioa_cfg *ioa_cfg;
 	struct ipr_hrr_queue *hrrq;
@@ -5585,7 +5584,7 @@
 	completed_ops = ipr_process_hrrq(hrrq, budget, &doneq);
 
 	if (completed_ops < budget)
-		blk_iopoll_complete(iop);
+		irq_poll_complete(iop);
 	spin_unlock_irqrestore(hrrq->lock, hrrq_flags);
 
 	list_for_each_entry_safe(ipr_cmd, temp, &doneq, queue) {
@@ -5693,8 +5692,7 @@
 	if (ioa_cfg->iopoll_weight && ioa_cfg->sis64 && ioa_cfg->nvectors > 1) {
 		if ((be32_to_cpu(*hrrq->hrrq_curr) & IPR_HRRQ_TOGGLE_BIT) ==
 		       hrrq->toggle_bit) {
-			if (!blk_iopoll_sched_prep(&hrrq->iopoll))
-				blk_iopoll_sched(&hrrq->iopoll);
+			irq_poll_sched(&hrrq->iopoll);
 			spin_unlock_irqrestore(hrrq->lock, hrrq_flags);
 			return IRQ_HANDLED;
 		}
@@ -10405,9 +10403,8 @@
 
 	if (ioa_cfg->iopoll_weight && ioa_cfg->sis64 && ioa_cfg->nvectors > 1) {
 		for (i = 1; i < ioa_cfg->hrrq_num; i++) {
-			blk_iopoll_init(&ioa_cfg->hrrq[i].iopoll,
+			irq_poll_init(&ioa_cfg->hrrq[i].iopoll,
 					ioa_cfg->iopoll_weight, ipr_iopoll);
-			blk_iopoll_enable(&ioa_cfg->hrrq[i].iopoll);
 		}
 	}
 
@@ -10436,7 +10433,7 @@
 	if (ioa_cfg->iopoll_weight && ioa_cfg->sis64 && ioa_cfg->nvectors > 1) {
 		ioa_cfg->iopoll_weight = 0;
 		for (i = 1; i < ioa_cfg->hrrq_num; i++)
-			blk_iopoll_disable(&ioa_cfg->hrrq[i].iopoll);
+			irq_poll_disable(&ioa_cfg->hrrq[i].iopoll);
 	}
 
 	while (ioa_cfg->in_reset_reload) {
diff --git a/drivers/scsi/ipr.h b/drivers/scsi/ipr.h
index a34c7a5..56c5706 100644
--- a/drivers/scsi/ipr.h
+++ b/drivers/scsi/ipr.h
@@ -32,7 +32,7 @@
 #include <linux/libata.h>
 #include <linux/list.h>
 #include <linux/kref.h>
-#include <linux/blk-iopoll.h>
+#include <linux/irq_poll.h>
 #include <scsi/scsi.h>
 #include <scsi/scsi_cmnd.h>
 
@@ -517,7 +517,7 @@
 	u8 allow_cmds:1;
 	u8 removing_ioa:1;
 
-	struct blk_iopoll iopoll;
+	struct irq_poll iopoll;
 };
 
 /* Command packet structure */