1 files changed, 55 insertions, 62 deletions

diff --git a/drivers/nvme/host/tcp.c b/drivers/nvme/host/tcp.c
index c15a92163c1f..1843110ec34f 100644
--- a/drivers/nvme/host/tcp.c
+++ b/drivers/nvme/host/tcp.c
@@ -60,6 +60,7 @@ struct nvme_tcp_request {
 enum nvme_tcp_queue_flags {
 	NVME_TCP_Q_ALLOCATED	= 0,
 	NVME_TCP_Q_LIVE		= 1,
+	NVME_TCP_Q_POLLING	= 2,
 };
 
 enum nvme_tcp_recv_state {
@@ -75,6 +76,7 @@ struct nvme_tcp_queue {
 	int			io_cpu;
 
 	spinlock_t		lock;
+	struct mutex		send_mutex;
 	struct list_head	send_list;
 
 	/* recv state */
@@ -131,6 +133,7 @@ static DEFINE_MUTEX(nvme_tcp_ctrl_mutex);
 static struct workqueue_struct *nvme_tcp_wq;
 static struct blk_mq_ops nvme_tcp_mq_ops;
 static struct blk_mq_ops nvme_tcp_admin_mq_ops;
+static int nvme_tcp_try_send(struct nvme_tcp_queue *queue);
 
 static inline struct nvme_tcp_ctrl *to_tcp_ctrl(struct nvme_ctrl *ctrl)
 {
@@ -257,15 +260,29 @@ static inline void nvme_tcp_advance_req(struct nvme_tcp_request *req,
 	}
 }
 
-static inline void nvme_tcp_queue_request(struct nvme_tcp_request *req)
+static inline void nvme_tcp_queue_request(struct nvme_tcp_request *req,
+		bool sync)
 {
 	struct nvme_tcp_queue *queue = req->queue;
+	bool empty;
 
 	spin_lock(&queue->lock);
+	empty = list_empty(&queue->send_list) && !queue->request;
 	list_add_tail(&req->entry, &queue->send_list);
 	spin_unlock(&queue->lock);
 
-	queue_work_on(queue->io_cpu, nvme_tcp_wq, &queue->io_work);
+	/*
+	 * if we're the first on the send_list and we can try to send
+	 * directly, otherwise queue io_work. Also, only do that if we
+	 * are on the same cpu, so we don't introduce contention.
+	 */
+	if (queue->io_cpu == smp_processor_id() &&
+	    sync && empty && mutex_trylock(&queue->send_mutex)) {
+		nvme_tcp_try_send(queue);
+		mutex_unlock(&queue->send_mutex);
+	} else {
+		queue_work_on(queue->io_cpu, nvme_tcp_wq, &queue->io_work);
+	}
 }
 
 static inline struct nvme_tcp_request *
@@ -578,7 +595,7 @@ static int nvme_tcp_handle_r2t(struct nvme_tcp_queue *queue,
 	req->state = NVME_TCP_SEND_H2C_PDU;
 	req->offset = 0;
 
-	nvme_tcp_queue_request(req);
+	nvme_tcp_queue_request(req, false);
 
 	return 0;
 }
@@ -794,11 +811,12 @@ static void nvme_tcp_data_ready(struct sock *sk)
 {
 	struct nvme_tcp_queue *queue;
 
-	read_lock(&sk->sk_callback_lock);
+	read_lock_bh(&sk->sk_callback_lock);
 	queue = sk->sk_user_data;
-	if (likely(queue && queue->rd_enabled))
+	if (likely(queue && queue->rd_enabled) &&
+	    !test_bit(NVME_TCP_Q_POLLING, &queue->flags))
 		queue_work_on(queue->io_cpu, nvme_tcp_wq, &queue->io_work);
-	read_unlock(&sk->sk_callback_lock);
+	read_unlock_bh(&sk->sk_callback_lock);
 }
 
 static void nvme_tcp_write_space(struct sock *sk)
@@ -867,7 +885,7 @@ static int nvme_tcp_try_send_data(struct nvme_tcp_request *req)
 		if (last && !queue->data_digest)
 			flags |= MSG_EOR;
 		else
-			flags |= MSG_MORE;
+			flags |= MSG_MORE | MSG_SENDPAGE_NOTLAST;
 
 		/* can't zcopy slab pages */
 		if (unlikely(PageSlab(page))) {
@@ -906,11 +924,16 @@ static int nvme_tcp_try_send_cmd_pdu(struct nvme_tcp_request *req)
 	struct nvme_tcp_queue *queue = req->queue;
 	struct nvme_tcp_cmd_pdu *pdu = req->pdu;
 	bool inline_data = nvme_tcp_has_inline_data(req);
-	int flags = MSG_DONTWAIT | (inline_data ? MSG_MORE : MSG_EOR);
 	u8 hdgst = nvme_tcp_hdgst_len(queue);
 	int len = sizeof(*pdu) + hdgst - req->offset;
+	int flags = MSG_DONTWAIT;
 	int ret;
 
+	if (inline_data)
+		flags |= MSG_MORE | MSG_SENDPAGE_NOTLAST;
+	else
+		flags |= MSG_EOR;
+
 	if (queue->hdr_digest && !req->offset)
 		nvme_tcp_hdgst(queue->snd_hash, pdu, sizeof(*pdu));
 
@@ -949,7 +972,7 @@ static int nvme_tcp_try_send_data_pdu(struct nvme_tcp_request *req)
 
 	ret = kernel_sendpage(queue->sock, virt_to_page(pdu),
 			offset_in_page(pdu) + req->offset, len,
-			MSG_DONTWAIT | MSG_MORE);
+			MSG_DONTWAIT | MSG_MORE | MSG_SENDPAGE_NOTLAST);
 	if (unlikely(ret <= 0))
 		return ret;
 
@@ -1063,11 +1086,14 @@ static void nvme_tcp_io_work(struct work_struct *w)
 		bool pending = false;
 		int result;
 
-		result = nvme_tcp_try_send(queue);
-		if (result > 0)
-			pending = true;
-		else if (unlikely(result < 0))
-			break;
+		if (mutex_trylock(&queue->send_mutex)) {
+			result = nvme_tcp_try_send(queue);
+			mutex_unlock(&queue->send_mutex);
+			if (result > 0)
+				pending = true;
+			else if (unlikely(result < 0))
+				break;
+		}
 
 		result = nvme_tcp_try_recv(queue);
 		if (result > 0)
@@ -1313,12 +1339,12 @@ static int nvme_tcp_alloc_queue(struct nvme_ctrl *nctrl,
 {
 	struct nvme_tcp_ctrl *ctrl = to_tcp_ctrl(nctrl);
 	struct nvme_tcp_queue *queue = &ctrl->queues[qid];
-	struct linger sol = { .l_onoff = 1, .l_linger = 0 };
-	int ret, opt, rcv_pdu_size;
+	int ret, rcv_pdu_size;
 
 	queue->ctrl = ctrl;
 	INIT_LIST_HEAD(&queue->send_list);
 	spin_lock_init(&queue->lock);
+	mutex_init(&queue->send_mutex);
 	INIT_WORK(&queue->io_work, nvme_tcp_io_work);
 	queue->queue_size = queue_size;
 
@@ -1337,60 +1363,24 @@ static int nvme_tcp_alloc_queue(struct nvme_ctrl *nctrl,
 	}
 
 	/* Single syn retry */
-	opt = 1;
-	ret = kernel_setsockopt(queue->sock, IPPROTO_TCP, TCP_SYNCNT,
-			(char *)&opt, sizeof(opt));
-	if (ret) {
-		dev_err(nctrl->device,
-			"failed to set TCP_SYNCNT sock opt %d\n", ret);
-		goto err_sock;
-	}
+	tcp_sock_set_syncnt(queue->sock->sk, 1);
 
 	/* Set TCP no delay */
-	opt = 1;
-	ret = kernel_setsockopt(queue->sock, IPPROTO_TCP,
-			TCP_NODELAY, (char *)&opt, sizeof(opt));
-	if (ret) {
-		dev_err(nctrl->device,
-			"failed to set TCP_NODELAY sock opt %d\n", ret);
-		goto err_sock;
-	}
+	tcp_sock_set_nodelay(queue->sock->sk);
 
 	/*
 	 * Cleanup whatever is sitting in the TCP transmit queue on socket
 	 * close. This is done to prevent stale data from being sent should
 	 * the network connection be restored before TCP times out.
 	 */
-	ret = kernel_setsockopt(queue->sock, SOL_SOCKET, SO_LINGER,
-			(char *)&sol, sizeof(sol));
-	if (ret) {
-		dev_err(nctrl->device,
-			"failed to set SO_LINGER sock opt %d\n", ret);
-		goto err_sock;
-	}
+	sock_no_linger(queue->sock->sk);
 
-	if (so_priority > 0) {
-		ret = kernel_setsockopt(queue->sock, SOL_SOCKET, SO_PRIORITY,
-				(char *)&so_priority, sizeof(so_priority));
-		if (ret) {
-			dev_err(ctrl->ctrl.device,
-				"failed to set SO_PRIORITY sock opt, ret %d\n",
-				ret);
-			goto err_sock;
-		}
-	}
+	if (so_priority > 0)
+		sock_set_priority(queue->sock->sk, so_priority);
 
 	/* Set socket type of service */
-	if (nctrl->opts->tos >= 0) {
-		opt = nctrl->opts->tos;
-		ret = kernel_setsockopt(queue->sock, SOL_IP, IP_TOS,
-				(char *)&opt, sizeof(opt));
-		if (ret) {
-			dev_err(nctrl->device,
-				"failed to set IP_TOS sock opt %d\n", ret);
-			goto err_sock;
-		}
-	}
+	if (nctrl->opts->tos >= 0)
+		ip_sock_set_tos(queue->sock->sk, nctrl->opts->tos);
 
 	queue->sock->sk->sk_allocation = GFP_ATOMIC;
 	nvme_tcp_set_queue_io_cpu(queue);
@@ -1543,6 +1533,7 @@ static struct blk_mq_tag_set *nvme_tcp_alloc_tagset(struct nvme_ctrl *nctrl,
 		set->queue_depth = NVME_AQ_MQ_TAG_DEPTH;
 		set->reserved_tags = 2; /* connect + keep-alive */
 		set->numa_node = NUMA_NO_NODE;
+		set->flags = BLK_MQ_F_BLOCKING;
 		set->cmd_size = sizeof(struct nvme_tcp_request);
 		set->driver_data = ctrl;
 		set->nr_hw_queues = 1;
@@ -1554,7 +1545,7 @@ static struct blk_mq_tag_set *nvme_tcp_alloc_tagset(struct nvme_ctrl *nctrl,
 		set->queue_depth = nctrl->sqsize + 1;
 		set->reserved_tags = 1; /* fabric connect */
 		set->numa_node = NUMA_NO_NODE;
-		set->flags = BLK_MQ_F_SHOULD_MERGE;
+		set->flags = BLK_MQ_F_SHOULD_MERGE | BLK_MQ_F_BLOCKING;
 		set->cmd_size = sizeof(struct nvme_tcp_request);
 		set->driver_data = ctrl;
 		set->nr_hw_queues = nctrl->queue_count - 1;
@@ -2113,7 +2104,7 @@ static void nvme_tcp_submit_async_event(struct nvme_ctrl *arg)
 	ctrl->async_req.curr_bio = NULL;
 	ctrl->async_req.data_len = 0;
 
-	nvme_tcp_queue_request(&ctrl->async_req);
+	nvme_tcp_queue_request(&ctrl->async_req, true);
 }
 
 static enum blk_eh_timer_return
@@ -2244,7 +2235,7 @@ static blk_status_t nvme_tcp_queue_rq(struct blk_mq_hw_ctx *hctx,
 
 	blk_mq_start_request(rq);
 
-	nvme_tcp_queue_request(req);
+	nvme_tcp_queue_request(req, true);
 
 	return BLK_STS_OK;
 }
@@ -2302,9 +2293,11 @@ static int nvme_tcp_poll(struct blk_mq_hw_ctx *hctx)
 	if (!test_bit(NVME_TCP_Q_LIVE, &queue->flags))
 		return 0;
 
+	set_bit(NVME_TCP_Q_POLLING, &queue->flags);
 	if (sk_can_busy_loop(sk) && skb_queue_empty_lockless(&sk->sk_receive_queue))
 		sk_busy_loop(sk, true);
 	nvme_tcp_try_recv(queue);
+	clear_bit(NVME_TCP_Q_POLLING, &queue->flags);
 	return queue->nr_cqe;
 }