IB/srp: Split send and recieve CQs to reduce number of interrupts

We can reduce the number of IB interrupts from two interrupts per
srp_queuecommand() call to one by using separate CQs for send and
receive completions and processing send completions by polling every
time a TX IU is allocated.

Receive completion events still trigger an interrupt.

Signed-off-by: Bart Van Assche <bart.vanassche@gmail.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>

2 files changed, 53 insertions, 25 deletions

diff --git a/drivers/infiniband/ulp/srp/ib_srp.c b/drivers/infiniband/ulp/srp/ib_srp.c
index 54c8fe25c423..441ea7c2e7c4 100644
--- a/drivers/infiniband/ulp/srp/ib_srp.c
+++ b/drivers/infiniband/ulp/srp/ib_srp.c
@@ -80,7 +80,8 @@ MODULE_PARM_DESC(mellanox_workarounds,
 
 static void srp_add_one(struct ib_device *device);
 static void srp_remove_one(struct ib_device *device);
-static void srp_completion(struct ib_cq *cq, void *target_ptr);
+static void srp_recv_completion(struct ib_cq *cq, void *target_ptr);
+static void srp_send_completion(struct ib_cq *cq, void *target_ptr);
 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;
@@ -227,14 +228,22 @@ static int srp_create_target_ib(struct srp_target_port *target)
 	if (!init_attr)
 		return -ENOMEM;
 
-	target->cq = ib_create_cq(target->srp_host->srp_dev->dev,
-				  srp_completion, NULL, target, SRP_CQ_SIZE, 0);
-	if (IS_ERR(target->cq)) {
-		ret = PTR_ERR(target->cq);
+	target->recv_cq = ib_create_cq(target->srp_host->srp_dev->dev,
+				       srp_recv_completion, NULL, target, SRP_RQ_SIZE, 0);
+	if (IS_ERR(target->recv_cq)) {
+		ret = PTR_ERR(target->recv_cq);
 		goto out;
 	}
 
-	ib_req_notify_cq(target->cq, IB_CQ_NEXT_COMP);
+	target->send_cq = ib_create_cq(target->srp_host->srp_dev->dev,
+				       srp_send_completion, NULL, target, SRP_SQ_SIZE, 0);
+	if (IS_ERR(target->send_cq)) {
+		ret = PTR_ERR(target->send_cq);
+		ib_destroy_cq(target->recv_cq);
+		goto out;
+	}
+
+	ib_req_notify_cq(target->recv_cq, IB_CQ_NEXT_COMP);
 
 	init_attr->event_handler       = srp_qp_event;
 	init_attr->cap.max_send_wr     = SRP_SQ_SIZE;
@@ -243,20 +252,22 @@ static int srp_create_target_ib(struct srp_target_port *target)
 	init_attr->cap.max_send_sge    = 1;
 	init_attr->sq_sig_type         = IB_SIGNAL_ALL_WR;
 	init_attr->qp_type             = IB_QPT_RC;
-	init_attr->send_cq             = target->cq;
-	init_attr->recv_cq             = target->cq;
+	init_attr->send_cq             = target->send_cq;
+	init_attr->recv_cq             = target->recv_cq;
 
 	target->qp = ib_create_qp(target->srp_host->srp_dev->pd, init_attr);
 	if (IS_ERR(target->qp)) {
 		ret = PTR_ERR(target->qp);
-		ib_destroy_cq(target->cq);
+		ib_destroy_cq(target->send_cq);
+		ib_destroy_cq(target->recv_cq);
 		goto out;
 	}
 
 	ret = srp_init_qp(target, target->qp);
 	if (ret) {
 		ib_destroy_qp(target->qp);
-		ib_destroy_cq(target->cq);
+		ib_destroy_cq(target->send_cq);
+		ib_destroy_cq(target->recv_cq);
 		goto out;
 	}
 
@@ -270,7 +281,8 @@ static void srp_free_target_ib(struct srp_target_port *target)
 	int i;
 
 	ib_destroy_qp(target->qp);
-	ib_destroy_cq(target->cq);
+	ib_destroy_cq(target->send_cq);
+	ib_destroy_cq(target->recv_cq);
 
 	for (i = 0; i < SRP_RQ_SIZE; ++i)
 		srp_free_iu(target->srp_host, target->rx_ring[i]);
@@ -568,7 +580,9 @@ static int srp_reconnect_target(struct srp_target_port *target)
 	if (ret)
 		goto err;
 
-	while (ib_poll_cq(target->cq, 1, &wc) > 0)
+	while (ib_poll_cq(target->recv_cq, 1, &wc) > 0)
+		; /* nothing */
+	while (ib_poll_cq(target->send_cq, 1, &wc) > 0)
 		; /* nothing */
 
 	spin_lock_irq(target->scsi_host->host_lock);
@@ -851,7 +865,7 @@ static void srp_handle_recv(struct srp_target_port *target, struct ib_wc *wc)
 	struct srp_iu *iu;
 	u8 opcode;
 
-	iu = target->rx_ring[wc->wr_id & ~SRP_OP_RECV];
+	iu = target->rx_ring[wc->wr_id];
 
 	dev = target->srp_host->srp_dev->dev;
 	ib_dma_sync_single_for_cpu(dev, iu->dma, target->max_ti_iu_len,
@@ -898,7 +912,7 @@ static void srp_handle_recv(struct srp_target_port *target, struct ib_wc *wc)
 				      DMA_FROM_DEVICE);
 }
 
-static void srp_completion(struct ib_cq *cq, void *target_ptr)
+static void srp_recv_completion(struct ib_cq *cq, void *target_ptr)
 {
 	struct srp_target_port *target = target_ptr;
 	struct ib_wc wc;
@@ -907,17 +921,31 @@ static void srp_completion(struct ib_cq *cq, void *target_ptr)
 	while (ib_poll_cq(cq, 1, &wc) > 0) {
 		if (wc.status) {
 			shost_printk(KERN_ERR, target->scsi_host,
-				     PFX "failed %s status %d\n",
-				     wc.wr_id & SRP_OP_RECV ? "receive" : "send",
+				     PFX "failed receive status %d\n",
 				     wc.status);
 			target->qp_in_error = 1;
 			break;
 		}
 
-		if (wc.wr_id & SRP_OP_RECV)
-			srp_handle_recv(target, &wc);
-		else
-			++target->tx_tail;
+		srp_handle_recv(target, &wc);
+	}
+}
+
+static void srp_send_completion(struct ib_cq *cq, void *target_ptr)
+{
+	struct srp_target_port *target = target_ptr;
+	struct ib_wc wc;
+
+	while (ib_poll_cq(cq, 1, &wc) > 0) {
+		if (wc.status) {
+			shost_printk(KERN_ERR, target->scsi_host,
+				     PFX "failed send status %d\n",
+				     wc.status);
+			target->qp_in_error = 1;
+			break;
+		}
+
+		++target->tx_tail;
 	}
 }
 
@@ -930,7 +958,7 @@ static int __srp_post_recv(struct srp_target_port *target)
 	int ret;
 
 	next 	 = target->rx_head & (SRP_RQ_SIZE - 1);
-	wr.wr_id = next | SRP_OP_RECV;
+	wr.wr_id = next;
 	iu 	 = target->rx_ring[next];
 
 	list.addr   = iu->dma;
@@ -970,6 +998,8 @@ static struct srp_iu *__srp_get_tx_iu(struct srp_target_port *target,
 {
 	s32 min = (req_type == SRP_REQ_TASK_MGMT) ? 1 : 2;
 
+	srp_send_completion(target->send_cq, target);
+
 	if (target->tx_head - target->tx_tail >= SRP_SQ_SIZE)
 		return NULL;
 
diff --git a/drivers/infiniband/ulp/srp/ib_srp.h b/drivers/infiniband/ulp/srp/ib_srp.h
index e185b907fc12..5a80eac6fdaa 100644
--- a/drivers/infiniband/ulp/srp/ib_srp.h
+++ b/drivers/infiniband/ulp/srp/ib_srp.h
@@ -60,7 +60,6 @@ enum {
 	SRP_RQ_SHIFT    	= 6,
 	SRP_RQ_SIZE		= 1 << SRP_RQ_SHIFT,
 	SRP_SQ_SIZE		= SRP_RQ_SIZE - 1,
-	SRP_CQ_SIZE		= SRP_SQ_SIZE + SRP_RQ_SIZE,
 
 	SRP_TAG_TSK_MGMT	= 1 << (SRP_RQ_SHIFT + 1),
 
@@ -69,8 +68,6 @@ enum {
 	SRP_FMR_DIRTY_SIZE	= SRP_FMR_POOL_SIZE / 4
 };
 
-#define SRP_OP_RECV		(1 << 31)
-
 enum srp_target_state {
 	SRP_TARGET_LIVE,
 	SRP_TARGET_CONNECTING,
@@ -133,7 +130,8 @@ struct srp_target_port {
 	int			path_query_id;
 
 	struct ib_cm_id	       *cm_id;
-	struct ib_cq	       *cq;
+	struct ib_cq	       *recv_cq;
+	struct ib_cq	       *send_cq;
 	struct ib_qp	       *qp;
 
 	int			max_ti_iu_len;