Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.34 as permitted sender) client-ip=23.128.96.34;
From:   Chengfeng Ye <dg573847474@gmail.com>
To:     hch@lst.de, sagi@grimberg.me, kch@nvidia.com
Cc:     linux-nvme@lists.infradead.org, linux-kernel@vger.kernel.org,
        Chengfeng Ye <dg573847474@gmail.com>
Subject: [PATCH] nvmet-rdma: use spin_lock_bh() on rsp_wr_wait_lock
Date:   Tue, 26 Sep 2023 17:22:08 +0000
Message-Id: <20230926172208.55478-1-dg573847474@gmail.com>
Precedence: bulk

It seems to me that read_cqe.done could be executed under softirq
context, as done callbacks always do, and it acquires rsp_wr_wait_lock
along the following call chain.

   nvmet_rdma_read_data_done()
   --> nvmet_rdma_release_rsp()
   --> spin_lock(&queue->rsp_wr_wait_lock)

So it seems more reasonable to use spin_lock_bh() on it, otherwise
there could be following potential deadlocks.

nvmet_rdma_queue_response()
--> nvmet_rdma_release_rsp()
--> spin_lock(&queue->rsp_wr_wait_lock)
<interrupt>
   --> nvmet_rdma_read_data_done()
   --> nvmet_rdma_release_rsp()
   --> spin_lock(&queue->rsp_wr_wait_lock)

nvmet_rdma_cm_handler()
--> nvmet_rdma_handle_command()
--> spin_lock(&queue->rsp_wr_wait_lock)
<interrupt>
   --> nvmet_rdma_read_data_done()
   --> nvmet_rdma_release_rsp()
   --> spin_lock(&queue->rsp_wr_wait_lock)

Signed-off-by: Chengfeng Ye <dg573847474@gmail.com>
---
 drivers/nvme/target/rdma.c | 12 ++++++------
 1 file changed, 6 insertions(+), 6 deletions(-)

diff --git a/drivers/nvme/target/rdma.c b/drivers/nvme/target/rdma.c
index 4597bca43a6d..a01ed29fbd8a 100644
--- a/drivers/nvme/target/rdma.c
+++ b/drivers/nvme/target/rdma.c
@@ -520,7 +520,7 @@ static int nvmet_rdma_post_recv(struct nvmet_rdma_device *ndev,
 
 static void nvmet_rdma_process_wr_wait_list(struct nvmet_rdma_queue *queue)
 {
-	spin_lock(&queue->rsp_wr_wait_lock);
+	spin_lock_bh(&queue->rsp_wr_wait_lock);
 	while (!list_empty(&queue->rsp_wr_wait_list)) {
 		struct nvmet_rdma_rsp *rsp;
 		bool ret;
@@ -529,16 +529,16 @@ static void nvmet_rdma_process_wr_wait_list(struct nvmet_rdma_queue *queue)
 				struct nvmet_rdma_rsp, wait_list);
 		list_del(&rsp->wait_list);
 
-		spin_unlock(&queue->rsp_wr_wait_lock);
+		spin_unlock_bh(&queue->rsp_wr_wait_lock);
 		ret = nvmet_rdma_execute_command(rsp);
-		spin_lock(&queue->rsp_wr_wait_lock);
+		spin_lock_bh(&queue->rsp_wr_wait_lock);
 
 		if (!ret) {
 			list_add(&rsp->wait_list, &queue->rsp_wr_wait_list);
 			break;
 		}
 	}
-	spin_unlock(&queue->rsp_wr_wait_lock);
+	spin_unlock_bh(&queue->rsp_wr_wait_lock);
 }
 
 static u16 nvmet_rdma_check_pi_status(struct ib_mr *sig_mr)
@@ -994,9 +994,9 @@ static void nvmet_rdma_handle_command(struct nvmet_rdma_queue *queue,
 		goto out_err;
 
 	if (unlikely(!nvmet_rdma_execute_command(cmd))) {
-		spin_lock(&queue->rsp_wr_wait_lock);
+		spin_lock_bh(&queue->rsp_wr_wait_lock);
 		list_add_tail(&cmd->wait_list, &queue->rsp_wr_wait_list);
-		spin_unlock(&queue->rsp_wr_wait_lock);
+		spin_unlock_bh(&queue->rsp_wr_wait_lock);
 	}
 
 	return;
-- 
2.17.1