Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/roland/infiniband

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/roland/infiniband: (48 commits)
  IB/srp: Clean up error path in srp_create_target_ib()
  IB/srp: Split send and recieve CQs to reduce number of interrupts
  RDMA/nes: Add support for KR device id 0x0110
  IB/uverbs: Use anon_inodes instead of private infinibandeventfs
  IB/core: Fix and clean up ib_ud_header_init()
  RDMA/cxgb3: Mark RDMA device with CXIO_ERROR_FATAL when removing
  RDMA/cxgb3: Don't allocate the SW queue for user mode CQs
  RDMA/cxgb3: Increase the max CQ depth
  RDMA/cxgb3: Doorbell overflow avoidance and recovery
  IB/core: Pack struct ib_device a little tighter
  IB/ucm: Clean whitespace errors
  IB/ucm: Increase maximum devices supported
  IB/ucm: Use stack variable 'base' in ib_ucm_add_one
  IB/ucm: Use stack variable 'devnum' in ib_ucm_add_one
  IB/umad: Clean whitespace
  IB/umad: Increase maximum devices supported
  IB/umad: Use stack variable 'base' in ib_umad_init_port
  IB/umad: Use stack variable 'devnum' in ib_umad_init_port
  IB/umad: Remove port_table[]
  IB/umad: Convert *cdev to cdev in struct ib_umad_port
  ...

1 files changed, 163 insertions, 343 deletions

diff --git a/drivers/infiniband/ulp/iser/iser_initiator.c b/drivers/infiniband/ulp/iser/iser_initiator.c
index 9de640200ad3..0b9ef0716588 100644
--- a/drivers/infiniband/ulp/iser/iser_initiator.c
+++ b/drivers/infiniband/ulp/iser/iser_initiator.c
@@ -39,29 +39,6 @@
 
 #include "iscsi_iser.h"
 
-/* Constant PDU lengths calculations */
-#define ISER_TOTAL_HEADERS_LEN  (sizeof (struct iser_hdr) + \
-				 sizeof (struct iscsi_hdr))
-
-/* iser_dto_add_regd_buff - increments the reference count for *
- * the registered buffer & adds it to the DTO object           */
-static void iser_dto_add_regd_buff(struct iser_dto *dto,
-				   struct iser_regd_buf *regd_buf,
-				   unsigned long use_offset,
-				   unsigned long use_size)
-{
-	int add_idx;
-
-	atomic_inc(&regd_buf->ref_count);
-
-	add_idx = dto->regd_vector_len;
-	dto->regd[add_idx] = regd_buf;
-	dto->used_sz[add_idx] = use_size;
-	dto->offset[add_idx] = use_offset;
-
-	dto->regd_vector_len++;
-}
-
 /* Register user buffer memory and initialize passive rdma
  *  dto descriptor. Total data size is stored in
  *  iser_task->data[ISER_DIR_IN].data_len
@@ -122,9 +99,9 @@ iser_prepare_write_cmd(struct iscsi_task *task,
 	struct iscsi_iser_task *iser_task = task->dd_data;
 	struct iser_regd_buf *regd_buf;
 	int err;
-	struct iser_dto *send_dto = &iser_task->desc.dto;
 	struct iser_hdr *hdr = &iser_task->desc.iser_header;
 	struct iser_data_buf *buf_out = &iser_task->data[ISER_DIR_OUT];
+	struct ib_sge *tx_dsg = &iser_task->desc.tx_sg[1];
 
 	err = iser_dma_map_task_data(iser_task,
 				     buf_out,
@@ -163,135 +140,100 @@ iser_prepare_write_cmd(struct iscsi_task *task,
 	if (imm_sz > 0) {
 		iser_dbg("Cmd itt:%d, WRITE, adding imm.data sz: %d\n",
 			 task->itt, imm_sz);
-		iser_dto_add_regd_buff(send_dto,
-				       regd_buf,
-				       0,
-				       imm_sz);
+		tx_dsg->addr   = regd_buf->reg.va;
+		tx_dsg->length = imm_sz;
+		tx_dsg->lkey   = regd_buf->reg.lkey;
+		iser_task->desc.num_sge = 2;
 	}
 
 	return 0;
 }
 
-/**
- * iser_post_receive_control - allocates, initializes and posts receive DTO.
- */
-static int iser_post_receive_control(struct iscsi_conn *conn)
+/* creates a new tx descriptor and adds header regd buffer */
+static void iser_create_send_desc(struct iser_conn	*ib_conn,
+				  struct iser_tx_desc	*tx_desc)
 {
-	struct iscsi_iser_conn *iser_conn = conn->dd_data;
-	struct iser_desc     *rx_desc;
-	struct iser_regd_buf *regd_hdr;
-	struct iser_regd_buf *regd_data;
-	struct iser_dto      *recv_dto = NULL;
-	struct iser_device  *device = iser_conn->ib_conn->device;
-	int rx_data_size, err;
-	int posts, outstanding_unexp_pdus;
-
-	/* for the login sequence we must support rx of upto 8K; login is done
-	 * after conn create/bind (connect) and conn stop/bind (reconnect),
-	 * what's common for both schemes is that the connection is not started
-	 */
-	if (conn->c_stage != ISCSI_CONN_STARTED)
-		rx_data_size = ISCSI_DEF_MAX_RECV_SEG_LEN;
-	else /* FIXME till user space sets conn->max_recv_dlength correctly */
-		rx_data_size = 128;
-
-	outstanding_unexp_pdus =
-		atomic_xchg(&iser_conn->ib_conn->unexpected_pdu_count, 0);
-
-	/*
-	 * in addition to the response buffer, replace those consumed by
-	 * unexpected pdus.
-	 */
-	for (posts = 0; posts < 1 + outstanding_unexp_pdus; posts++) {
-		rx_desc = kmem_cache_alloc(ig.desc_cache, GFP_NOIO);
-		if (rx_desc == NULL) {
-			iser_err("Failed to alloc desc for post recv %d\n",
-				 posts);
-			err = -ENOMEM;
-			goto post_rx_cache_alloc_failure;
-		}
-		rx_desc->type = ISCSI_RX;
-		rx_desc->data = kmalloc(rx_data_size, GFP_NOIO);
-		if (rx_desc->data == NULL) {
-			iser_err("Failed to alloc data buf for post recv %d\n",
-				 posts);
-			err = -ENOMEM;
-			goto post_rx_kmalloc_failure;
-		}
-
-		recv_dto = &rx_desc->dto;
-		recv_dto->ib_conn = iser_conn->ib_conn;
-		recv_dto->regd_vector_len = 0;
+	struct iser_device *device = ib_conn->device;
 
-		regd_hdr = &rx_desc->hdr_regd_buf;
-		memset(regd_hdr, 0, sizeof(struct iser_regd_buf));
-		regd_hdr->device  = device;
-		regd_hdr->virt_addr  = rx_desc; /* == &rx_desc->iser_header */
-		regd_hdr->data_size  = ISER_TOTAL_HEADERS_LEN;
+	ib_dma_sync_single_for_cpu(device->ib_device,
+		tx_desc->dma_addr, ISER_HEADERS_LEN, DMA_TO_DEVICE);
 
-		iser_reg_single(device, regd_hdr, DMA_FROM_DEVICE);
-
-		iser_dto_add_regd_buff(recv_dto, regd_hdr, 0, 0);
+	memset(&tx_desc->iser_header, 0, sizeof(struct iser_hdr));
+	tx_desc->iser_header.flags = ISER_VER;
 
-		regd_data = &rx_desc->data_regd_buf;
-		memset(regd_data, 0, sizeof(struct iser_regd_buf));
-		regd_data->device  = device;
-		regd_data->virt_addr  = rx_desc->data;
-		regd_data->data_size  = rx_data_size;
+	tx_desc->num_sge = 1;
 
-		iser_reg_single(device, regd_data, DMA_FROM_DEVICE);
+	if (tx_desc->tx_sg[0].lkey != device->mr->lkey) {
+		tx_desc->tx_sg[0].lkey = device->mr->lkey;
+		iser_dbg("sdesc %p lkey mismatch, fixing\n", tx_desc);
+	}
+}
 
-		iser_dto_add_regd_buff(recv_dto, regd_data, 0, 0);
 
-		err = iser_post_recv(rx_desc);
-		if (err) {
-			iser_err("Failed iser_post_recv for post %d\n", posts);
-			goto post_rx_post_recv_failure;
-		}
+int iser_alloc_rx_descriptors(struct iser_conn *ib_conn)
+{
+	int i, j;
+	u64 dma_addr;
+	struct iser_rx_desc *rx_desc;
+	struct ib_sge       *rx_sg;
+	struct iser_device  *device = ib_conn->device;
+
+	ib_conn->rx_descs = kmalloc(ISER_QP_MAX_RECV_DTOS *
+				sizeof(struct iser_rx_desc), GFP_KERNEL);
+	if (!ib_conn->rx_descs)
+		goto rx_desc_alloc_fail;
+
+	rx_desc = ib_conn->rx_descs;
+
+	for (i = 0; i < ISER_QP_MAX_RECV_DTOS; i++, rx_desc++)  {
+		dma_addr = ib_dma_map_single(device->ib_device, (void *)rx_desc,
+					ISER_RX_PAYLOAD_SIZE, DMA_FROM_DEVICE);
+		if (ib_dma_mapping_error(device->ib_device, dma_addr))
+			goto rx_desc_dma_map_failed;
+
+		rx_desc->dma_addr = dma_addr;
+
+		rx_sg = &rx_desc->rx_sg;
+		rx_sg->addr   = rx_desc->dma_addr;
+		rx_sg->length = ISER_RX_PAYLOAD_SIZE;
+		rx_sg->lkey   = device->mr->lkey;
 	}
-	/* all posts successful */
-	return 0;
 
-post_rx_post_recv_failure:
-	iser_dto_buffs_release(recv_dto);
-	kfree(rx_desc->data);
-post_rx_kmalloc_failure:
-	kmem_cache_free(ig.desc_cache, rx_desc);
-post_rx_cache_alloc_failure:
-	if (posts > 0) {
-		/*
-		 * response buffer posted, but did not replace all unexpected
-		 * pdu recv bufs. Ignore error, retry occurs next send
-		 */
-		outstanding_unexp_pdus -= (posts - 1);
-		err = 0;
-	}
-	atomic_add(outstanding_unexp_pdus,
-		   &iser_conn->ib_conn->unexpected_pdu_count);
+	ib_conn->rx_desc_head = 0;
+	return 0;
 
-	return err;
+rx_desc_dma_map_failed:
+	rx_desc = ib_conn->rx_descs;
+	for (j = 0; j < i; j++, rx_desc++)
+		ib_dma_unmap_single(device->ib_device, rx_desc->dma_addr,
+			ISER_RX_PAYLOAD_SIZE, DMA_FROM_DEVICE);
+	kfree(ib_conn->rx_descs);
+	ib_conn->rx_descs = NULL;
+rx_desc_alloc_fail:
+	iser_err("failed allocating rx descriptors / data buffers\n");
+	return -ENOMEM;
 }
 
-/* creates a new tx descriptor and adds header regd buffer */
-static void iser_create_send_desc(struct iscsi_iser_conn *iser_conn,
-				  struct iser_desc       *tx_desc)
+void iser_free_rx_descriptors(struct iser_conn *ib_conn)
 {
-	struct iser_regd_buf *regd_hdr = &tx_desc->hdr_regd_buf;
-	struct iser_dto      *send_dto = &tx_desc->dto;
+	int i;
+	struct iser_rx_desc *rx_desc;
+	struct iser_device *device = ib_conn->device;
 
-	memset(regd_hdr, 0, sizeof(struct iser_regd_buf));
-	regd_hdr->device  = iser_conn->ib_conn->device;
-	regd_hdr->virt_addr  = tx_desc; /* == &tx_desc->iser_header */
-	regd_hdr->data_size  = ISER_TOTAL_HEADERS_LEN;
+	if (ib_conn->login_buf) {
+		ib_dma_unmap_single(device->ib_device, ib_conn->login_dma,
+			ISER_RX_LOGIN_SIZE, DMA_FROM_DEVICE);
+		kfree(ib_conn->login_buf);
+	}
 
-	send_dto->ib_conn         = iser_conn->ib_conn;
-	send_dto->notify_enable   = 1;
-	send_dto->regd_vector_len = 0;
+	if (!ib_conn->rx_descs)
+		return;
 
-	memset(&tx_desc->iser_header, 0, sizeof(struct iser_hdr));
-	tx_desc->iser_header.flags = ISER_VER;
-
-	iser_dto_add_regd_buff(send_dto, regd_hdr, 0, 0);
+	rx_desc = ib_conn->rx_descs;
+	for (i = 0; i < ISER_QP_MAX_RECV_DTOS; i++, rx_desc++)
+		ib_dma_unmap_single(device->ib_device, rx_desc->dma_addr,
+			ISER_RX_PAYLOAD_SIZE, DMA_FROM_DEVICE);
+	kfree(ib_conn->rx_descs);
 }
 
 /**
@@ -301,46 +243,23 @@ int iser_conn_set_full_featured_mode(struct iscsi_conn *conn)
 {
 	struct iscsi_iser_conn *iser_conn = conn->dd_data;
 
-	int i;
-	/*
-	 * FIXME this value should be declared to the target during login with
-	 * the MaxOutstandingUnexpectedPDUs key when supported
-	 */
-	int initial_post_recv_bufs_num = ISER_MAX_RX_MISC_PDUS;
-
-	iser_dbg("Initially post: %d\n", initial_post_recv_bufs_num);
+	iser_dbg("Initially post: %d\n", ISER_MIN_POSTED_RX);
 
 	/* Check that there is no posted recv or send buffers left - */
 	/* they must be consumed during the login phase */
-	BUG_ON(atomic_read(&iser_conn->ib_conn->post_recv_buf_count) != 0);
+	BUG_ON(iser_conn->ib_conn->post_recv_buf_count != 0);
 	BUG_ON(atomic_read(&iser_conn->ib_conn->post_send_buf_count) != 0);
 
-	/* Initial post receive buffers */
-	for (i = 0; i < initial_post_recv_bufs_num; i++) {
-		if (iser_post_receive_control(conn) != 0) {
-			iser_err("Failed to post recv bufs at:%d conn:0x%p\n",
-				 i, conn);
-			return -ENOMEM;
-		}
-	}
-	iser_dbg("Posted %d post recv bufs, conn:0x%p\n", i, conn);
-	return 0;
-}
+	if (iser_alloc_rx_descriptors(iser_conn->ib_conn))
+		return -ENOMEM;
 
-static int
-iser_check_xmit(struct iscsi_conn *conn, void *task)
-{
-	struct iscsi_iser_conn *iser_conn = conn->dd_data;
+	/* Initial post receive buffers */
+	if (iser_post_recvm(iser_conn->ib_conn, ISER_MIN_POSTED_RX))
+		return -ENOMEM;
 
-	if (atomic_read(&iser_conn->ib_conn->post_send_buf_count) ==
-	    ISER_QP_MAX_REQ_DTOS) {
-		iser_dbg("%ld can't xmit task %p\n",jiffies,task);
-		return -ENOBUFS;
-	}
 	return 0;
 }
 
-
 /**
  * iser_send_command - send command PDU
  */
@@ -349,27 +268,18 @@ int iser_send_command(struct iscsi_conn *conn,
 {
 	struct iscsi_iser_conn *iser_conn = conn->dd_data;
 	struct iscsi_iser_task *iser_task = task->dd_data;
-	struct iser_dto *send_dto = NULL;
 	unsigned long edtl;
-	int err = 0;
+	int err;
 	struct iser_data_buf *data_buf;
 	struct iscsi_cmd *hdr =  (struct iscsi_cmd *)task->hdr;
 	struct scsi_cmnd *sc  =  task->sc;
-
-	if (!iser_conn_state_comp(iser_conn->ib_conn, ISER_CONN_UP)) {
-		iser_err("Failed to send, conn: 0x%p is not up\n", iser_conn->ib_conn);
-		return -EPERM;
-	}
-	if (iser_check_xmit(conn, task))
-		return -ENOBUFS;
+	struct iser_tx_desc *tx_desc = &iser_task->desc;
 
 	edtl = ntohl(hdr->data_length);
 
 	/* build the tx desc regd header and add it to the tx desc dto */
-	iser_task->desc.type = ISCSI_TX_SCSI_COMMAND;
-	send_dto = &iser_task->desc.dto;
-	send_dto->task = iser_task;
-	iser_create_send_desc(iser_conn, &iser_task->desc);
+	tx_desc->type = ISCSI_TX_SCSI_COMMAND;
+	iser_create_send_desc(iser_conn->ib_conn, tx_desc);
 
 	if (hdr->flags & ISCSI_FLAG_CMD_READ)
 		data_buf = &iser_task->data[ISER_DIR_IN];
@@ -398,23 +308,13 @@ int iser_send_command(struct iscsi_conn *conn,
 			goto send_command_error;
 	}
 
-	iser_reg_single(iser_conn->ib_conn->device,
-			send_dto->regd[0], DMA_TO_DEVICE);
-
-	if (iser_post_receive_control(conn) != 0) {
-		iser_err("post_recv failed!\n");
-		err = -ENOMEM;
-		goto send_command_error;
-	}
-
 	iser_task->status = ISER_TASK_STATUS_STARTED;
 
-	err = iser_post_send(&iser_task->desc);
+	err = iser_post_send(iser_conn->ib_conn, tx_desc);
 	if (!err)
 		return 0;
 
 send_command_error:
-	iser_dto_buffs_release(send_dto);
 	iser_err("conn %p failed task->itt %d err %d\n",conn, task->itt, err);
 	return err;
 }
@@ -428,20 +328,13 @@ int iser_send_data_out(struct iscsi_conn *conn,
 {
 	struct iscsi_iser_conn *iser_conn = conn->dd_data;
 	struct iscsi_iser_task *iser_task = task->dd_data;
-	struct iser_desc *tx_desc = NULL;
-	struct iser_dto *send_dto = NULL;
+	struct iser_tx_desc *tx_desc = NULL;
+	struct iser_regd_buf *regd_buf;
 	unsigned long buf_offset;
 	unsigned long data_seg_len;
 	uint32_t itt;
 	int err = 0;
-
-	if (!iser_conn_state_comp(iser_conn->ib_conn, ISER_CONN_UP)) {
-		iser_err("Failed to send, conn: 0x%p is not up\n", iser_conn->ib_conn);
-		return -EPERM;
-	}
-
-	if (iser_check_xmit(conn, task))
-		return -ENOBUFS;
+	struct ib_sge *tx_dsg;
 
 	itt = (__force uint32_t)hdr->itt;
 	data_seg_len = ntoh24(hdr->dlength);
@@ -450,28 +343,25 @@ int iser_send_data_out(struct iscsi_conn *conn,
 	iser_dbg("%s itt %d dseg_len %d offset %d\n",
 		 __func__,(int)itt,(int)data_seg_len,(int)buf_offset);
 
-	tx_desc = kmem_cache_alloc(ig.desc_cache, GFP_NOIO);
+	tx_desc = kmem_cache_zalloc(ig.desc_cache, GFP_ATOMIC);
 	if (tx_desc == NULL) {
 		iser_err("Failed to alloc desc for post dataout\n");
 		return -ENOMEM;
 	}
 
 	tx_desc->type = ISCSI_TX_DATAOUT;
+	tx_desc->iser_header.flags = ISER_VER;
 	memcpy(&tx_desc->iscsi_header, hdr, sizeof(struct iscsi_hdr));
 
-	/* build the tx desc regd header and add it to the tx desc dto */
-	send_dto = &tx_desc->dto;
-	send_dto->task = iser_task;
-	iser_create_send_desc(iser_conn, tx_desc);
-
-	iser_reg_single(iser_conn->ib_conn->device,
-			send_dto->regd[0], DMA_TO_DEVICE);
+	/* build the tx desc */
+	iser_initialize_task_headers(task, tx_desc);
 
-	/* all data was registered for RDMA, we can use the lkey */
-	iser_dto_add_regd_buff(send_dto,
-			       &iser_task->rdma_regd[ISER_DIR_OUT],
-			       buf_offset,
-			       data_seg_len);
+	regd_buf = &iser_task->rdma_regd[ISER_DIR_OUT];
+	tx_dsg = &tx_desc->tx_sg[1];
+	tx_dsg->addr    = regd_buf->reg.va + buf_offset;
+	tx_dsg->length  = data_seg_len;
+	tx_dsg->lkey    = regd_buf->reg.lkey;
+	tx_desc->num_sge = 2;
 
 	if (buf_offset + data_seg_len > iser_task->data[ISER_DIR_OUT].data_len) {
 		iser_err("Offset:%ld & DSL:%ld in Data-Out "
@@ -485,12 +375,11 @@ int iser_send_data_out(struct iscsi_conn *conn,
 		 itt, buf_offset, data_seg_len);
 
 
-	err = iser_post_send(tx_desc);
+	err = iser_post_send(iser_conn->ib_conn, tx_desc);
 	if (!err)
 		return 0;
 
 send_data_out_error:
-	iser_dto_buffs_release(send_dto);
 	kmem_cache_free(ig.desc_cache, tx_desc);
 	iser_err("conn %p failed err %d\n",conn, err);
 	return err;
@@ -501,64 +390,44 @@ int iser_send_control(struct iscsi_conn *conn,
 {
 	struct iscsi_iser_conn *iser_conn = conn->dd_data;
 	struct iscsi_iser_task *iser_task = task->dd_data;
-	struct iser_desc *mdesc = &iser_task->desc;
-	struct iser_dto *send_dto = NULL;
+	struct iser_tx_desc *mdesc = &iser_task->desc;
 	unsigned long data_seg_len;
 	int err = 0;
-	struct iser_regd_buf *regd_buf;
 	struct iser_device *device;
-	unsigned char opcode;
-
-	if (!iser_conn_state_comp(iser_conn->ib_conn, ISER_CONN_UP)) {
-		iser_err("Failed to send, conn: 0x%p is not up\n", iser_conn->ib_conn);
-		return -EPERM;
-	}
-
-	if (iser_check_xmit(conn, task))
-		return -ENOBUFS;
 
 	/* build the tx desc regd header and add it to the tx desc dto */
 	mdesc->type = ISCSI_TX_CONTROL;
-	send_dto = &mdesc->dto;
-	send_dto->task = NULL;
-	iser_create_send_desc(iser_conn, mdesc);
+	iser_create_send_desc(iser_conn->ib_conn, mdesc);
 
 	device = iser_conn->ib_conn->device;
 
-	iser_reg_single(device, send_dto->regd[0], DMA_TO_DEVICE);
-
 	data_seg_len = ntoh24(task->hdr->dlength);
 
 	if (data_seg_len > 0) {
-		regd_buf = &mdesc->data_regd_buf;
-		memset(regd_buf, 0, sizeof(struct iser_regd_buf));
-		regd_buf->device = device;
-		regd_buf->virt_addr = task->data;
-		regd_buf->data_size = task->data_count;
-		iser_reg_single(device, regd_buf,
-				DMA_TO_DEVICE);
-		iser_dto_add_regd_buff(send_dto, regd_buf,
-				       0,
-				       data_seg_len);
+		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;
+		}
+		memcpy(iser_conn->ib_conn->login_buf, task->data,
+							task->data_count);
+		tx_dsg->addr    = iser_conn->ib_conn->login_dma;
+		tx_dsg->length  = data_seg_len;
+		tx_dsg->lkey    = device->mr->lkey;
+		mdesc->num_sge = 2;
 	}
 
-	opcode = task->hdr->opcode & ISCSI_OPCODE_MASK;
-
-	/* post recv buffer for response if one is expected */
-	if (!(opcode == ISCSI_OP_NOOP_OUT && task->hdr->itt == RESERVED_ITT)) {
-		if (iser_post_receive_control(conn) != 0) {
-			iser_err("post_rcv_buff failed!\n");
-			err = -ENOMEM;
+	if (task == conn->login_task) {
+		err = iser_post_recvl(iser_conn->ib_conn);
+		if (err)
 			goto send_control_error;
-		}
 	}
 
-	err = iser_post_send(mdesc);
+	err = iser_post_send(iser_conn->ib_conn, mdesc);
 	if (!err)
 		return 0;
 
 send_control_error:
-	iser_dto_buffs_release(send_dto);
 	iser_err("conn %p failed err %d\n",conn, err);
 	return err;
 }
@@ -566,104 +435,71 @@ send_control_error:
 /**
  * iser_rcv_dto_completion - recv DTO completion
  */
-void iser_rcv_completion(struct iser_desc *rx_desc,
-			 unsigned long dto_xfer_len)
+void iser_rcv_completion(struct iser_rx_desc *rx_desc,
+			 unsigned long rx_xfer_len,
+			 struct iser_conn *ib_conn)
 {
-	struct iser_dto *dto = &rx_desc->dto;
-	struct iscsi_iser_conn *conn = dto->ib_conn->iser_conn;
-	struct iscsi_task *task;
-	struct iscsi_iser_task *iser_task;
+	struct iscsi_iser_conn *conn = ib_conn->iser_conn;
 	struct iscsi_hdr *hdr;
-	char   *rx_data = NULL;
-	int     rx_data_len = 0;
-	unsigned char opcode;
-
-	hdr = &rx_desc->iscsi_header;
+	u64 rx_dma;
+	int rx_buflen, outstanding, count, err;
+
+	/* differentiate between login to all other PDUs */
+	if ((char *)rx_desc == ib_conn->login_buf) {
+		rx_dma = ib_conn->login_dma;
+		rx_buflen = ISER_RX_LOGIN_SIZE;
+	} else {
+		rx_dma = rx_desc->dma_addr;
+		rx_buflen = ISER_RX_PAYLOAD_SIZE;
+	}
 
-	iser_dbg("op 0x%x itt 0x%x\n", hdr->opcode,hdr->itt);
+	ib_dma_sync_single_for_cpu(ib_conn->device->ib_device, rx_dma,
+			rx_buflen, DMA_FROM_DEVICE);
 
-	if (dto_xfer_len > ISER_TOTAL_HEADERS_LEN) { /* we have data */
-		rx_data_len = dto_xfer_len - ISER_TOTAL_HEADERS_LEN;
-		rx_data     = dto->regd[1]->virt_addr;
-		rx_data    += dto->offset[1];
-	}
+	hdr = &rx_desc->iscsi_header;
 
-	opcode = hdr->opcode & ISCSI_OPCODE_MASK;
-
-	if (opcode == ISCSI_OP_SCSI_CMD_RSP) {
-		spin_lock(&conn->iscsi_conn->session->lock);
-		task = iscsi_itt_to_ctask(conn->iscsi_conn, hdr->itt);
-		if (task)
-			__iscsi_get_task(task);
-		spin_unlock(&conn->iscsi_conn->session->lock);
-
-		if (!task)
-			iser_err("itt can't be matched to task!!! "
-				 "conn %p opcode %d itt %d\n",
-				 conn->iscsi_conn, opcode, hdr->itt);
-		else {
-			iser_task = task->dd_data;
-			iser_dbg("itt %d task %p\n",hdr->itt, task);
-			iser_task->status = ISER_TASK_STATUS_COMPLETED;
-			iser_task_rdma_finalize(iser_task);
-			iscsi_put_task(task);
-		}
-	}
-	iser_dto_buffs_release(dto);
+	iser_dbg("op 0x%x itt 0x%x dlen %d\n", hdr->opcode,
+			hdr->itt, (int)(rx_xfer_len - ISER_HEADERS_LEN));
 
-	iscsi_iser_recv(conn->iscsi_conn, hdr, rx_data, rx_data_len);
+	iscsi_iser_recv(conn->iscsi_conn, hdr,
+		rx_desc->data, rx_xfer_len - ISER_HEADERS_LEN);
 
-	kfree(rx_desc->data);
-	kmem_cache_free(ig.desc_cache, rx_desc);
+	ib_dma_sync_single_for_device(ib_conn->device->ib_device, rx_dma,
+			rx_buflen, 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   *
 	 * parallel to the execution of iser_conn_term. So the code that waits *
 	 * for the posted rx bufs refcount to become zero handles everything   */
-	atomic_dec(&conn->ib_conn->post_recv_buf_count);
+	conn->ib_conn->post_recv_buf_count--;
 
-	/*
-	 * if an unexpected PDU was received then the recv wr consumed must
-	 * be replaced, this is done in the next send of a control-type PDU
-	 */
-	if (opcode == ISCSI_OP_NOOP_IN && hdr->itt == RESERVED_ITT) {
-		/* nop-in with itt = 0xffffffff */
-		atomic_inc(&conn->ib_conn->unexpected_pdu_count);
-	}
-	else if (opcode == ISCSI_OP_ASYNC_EVENT) {
-		/* asyncronous message */
-		atomic_inc(&conn->ib_conn->unexpected_pdu_count);
+	if (rx_dma == ib_conn->login_dma)
+		return;
+
+	outstanding = ib_conn->post_recv_buf_count;
+	if (outstanding + ISER_MIN_POSTED_RX <= ISER_QP_MAX_RECV_DTOS) {
+		count = min(ISER_QP_MAX_RECV_DTOS - outstanding,
+						ISER_MIN_POSTED_RX);
+		err = iser_post_recvm(ib_conn, count);
+		if (err)
+			iser_err("posting %d rx bufs err %d\n", count, err);
 	}
-	/* a reject PDU consumes the recv buf posted for the response */
 }
 
-void iser_snd_completion(struct iser_desc *tx_desc)
+void iser_snd_completion(struct iser_tx_desc *tx_desc,
+			struct iser_conn *ib_conn)
 {
-	struct iser_dto        *dto = &tx_desc->dto;
-	struct iser_conn       *ib_conn = dto->ib_conn;
-	struct iscsi_iser_conn *iser_conn = ib_conn->iser_conn;
-	struct iscsi_conn      *conn = iser_conn->iscsi_conn;
 	struct iscsi_task *task;
-	int resume_tx = 0;
-
-	iser_dbg("Initiator, Data sent dto=0x%p\n", dto);
-
-	iser_dto_buffs_release(dto);
+	struct iser_device *device = ib_conn->device;
 
-	if (tx_desc->type == ISCSI_TX_DATAOUT)
+	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);
-
-	if (atomic_read(&iser_conn->ib_conn->post_send_buf_count) ==
-	    ISER_QP_MAX_REQ_DTOS)
-		resume_tx = 1;
+	}
 
 	atomic_dec(&ib_conn->post_send_buf_count);
 
-	if (resume_tx) {
-		iser_dbg("%ld resuming tx\n",jiffies);
-		iscsi_conn_queue_work(conn);
-	}
-
 	if (tx_desc->type == ISCSI_TX_CONTROL) {
 		/* this arithmetic is legal by libiscsi dd_data allocation */
 		task = (void *) ((long)(void *)tx_desc -
@@ -692,7 +528,6 @@ void iser_task_rdma_init(struct iscsi_iser_task *iser_task)
 
 void iser_task_rdma_finalize(struct iscsi_iser_task *iser_task)
 {
-	int deferred;
 	int is_rdma_aligned = 1;
 	struct iser_regd_buf *regd;
 
@@ -710,32 +545,17 @@ void iser_task_rdma_finalize(struct iscsi_iser_task *iser_task)
 
 	if (iser_task->dir[ISER_DIR_IN]) {
 		regd = &iser_task->rdma_regd[ISER_DIR_IN];
-		deferred = iser_regd_buff_release(regd);
-		if (deferred) {
-			iser_err("%d references remain for BUF-IN rdma reg\n",
-				 atomic_read(&regd->ref_count));
-		}
+		if (regd->reg.is_fmr)
+			iser_unreg_mem(&regd->reg);
 	}
 
 	if (iser_task->dir[ISER_DIR_OUT]) {
 		regd = &iser_task->rdma_regd[ISER_DIR_OUT];
-		deferred = iser_regd_buff_release(regd);
-		if (deferred) {
-			iser_err("%d references remain for BUF-OUT rdma reg\n",
-				 atomic_read(&regd->ref_count));
-		}
+		if (regd->reg.is_fmr)
+			iser_unreg_mem(&regd->reg);
 	}
 
        /* if the data was unaligned, it was already unmapped and then copied */
        if (is_rdma_aligned)
 		iser_dma_unmap_task_data(iser_task);
 }
-
-void iser_dto_buffs_release(struct iser_dto *dto)
-{
-	int i;
-
-	for (i = 0; i < dto->regd_vector_len; i++)
-		iser_regd_buff_release(dto->regd[i]);
-}
-