nvmet-rdma: recheck queue state is LIVE in state lock in recv done

[ Upstream commit 3988ac1c67e6e84d2feb987d7b36d5791174b3da ]

The queue state checking in nvmet_rdma_recv_done is not in queue state
lock.Queue state can transfer to LIVE in cm establish handler between
state checking and state lock here, cause a silent drop of nvme connect
cmd.
Recheck queue state whether in LIVE state in state lock to prevent this
issue.

Signed-off-by: Ruozhu Li <david.li@jaguarmicro.com>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Keith Busch <kbusch@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

1 files changed, 23 insertions, 10 deletions

diff --git a/drivers/nvme/target/rdma.c b/drivers/nvme/target/rdma.c
index 1afd93026f9b..2a4536ef6184 100644
--- a/drivers/nvme/target/rdma.c
+++ b/drivers/nvme/target/rdma.c
@@ -996,6 +996,27 @@ out_err:
 	nvmet_req_complete(&cmd->req, status);
 }
 
+static bool nvmet_rdma_recv_not_live(struct nvmet_rdma_queue *queue,
+		struct nvmet_rdma_rsp *rsp)
+{
+	unsigned long flags;
+	bool ret = true;
+
+	spin_lock_irqsave(&queue->state_lock, flags);
+	/*
+	 * recheck queue state is not live to prevent a race condition
+	 * with RDMA_CM_EVENT_ESTABLISHED handler.
+	 */
+	if (queue->state == NVMET_RDMA_Q_LIVE)
+		ret = false;
+	else if (queue->state == NVMET_RDMA_Q_CONNECTING)
+		list_add_tail(&rsp->wait_list, &queue->rsp_wait_list);
+	else
+		nvmet_rdma_put_rsp(rsp);
+	spin_unlock_irqrestore(&queue->state_lock, flags);
+	return ret;
+}
+
 static void nvmet_rdma_recv_done(struct ib_cq *cq, struct ib_wc *wc)
 {
 	struct nvmet_rdma_cmd *cmd =
@@ -1038,17 +1059,9 @@ static void nvmet_rdma_recv_done(struct ib_cq *cq, struct ib_wc *wc)
 	rsp->n_rdma = 0;
 	rsp->invalidate_rkey = 0;
 
-	if (unlikely(queue->state != NVMET_RDMA_Q_LIVE)) {
-		unsigned long flags;
-
-		spin_lock_irqsave(&queue->state_lock, flags);
-		if (queue->state == NVMET_RDMA_Q_CONNECTING)
-			list_add_tail(&rsp->wait_list, &queue->rsp_wait_list);
-		else
-			nvmet_rdma_put_rsp(rsp);
-		spin_unlock_irqrestore(&queue->state_lock, flags);
+	if (unlikely(queue->state != NVMET_RDMA_Q_LIVE) &&
+	    nvmet_rdma_recv_not_live(queue, rsp))
 		return;
-	}
 
 	nvmet_rdma_handle_command(queue, rsp);
 }