nvme-loop: reset queue count to 1 in nvme_loop_destroy_io_queues()
authorHannes Reinecke <hare@suse.de>
Wed, 26 May 2021 15:23:15 +0000 (17:23 +0200)
committerSasha Levin <sashal@kernel.org>
Wed, 30 Jun 2021 12:48:46 +0000 (08:48 -0400)
[ Upstream commit a6c144f3d2e230f2b3ac5ed8c51e0f0391556197 ]

The queue count is increased in nvme_loop_init_io_queues(), so we
need to reset it to 1 at the end of nvme_loop_destroy_io_queues().
Otherwise the function is not re-entrant safe, and crash will happen
during concurrent reset and remove calls.

Signed-off-by: Hannes Reinecke <hare@suse.de>
Reviewed-by: Chaitanya Kulkarni <chaitanya.kulkarni@wdc.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Sasha Levin <sashal@kernel.org>
drivers/nvme/target/loop.c

index 3388d2788fe055554292965e762320470335d560..5f33c3a9469baead9aac50cc05c5b14f4c37133d 100644 (file)
@@ -322,6 +322,7 @@ static void nvme_loop_destroy_io_queues(struct nvme_loop_ctrl *ctrl)
                clear_bit(NVME_LOOP_Q_LIVE, &ctrl->queues[i].flags);
                nvmet_sq_destroy(&ctrl->queues[i].nvme_sq);
        }
+       ctrl->ctrl.queue_count = 1;
 }
 
 static int nvme_loop_init_io_queues(struct nvme_loop_ctrl *ctrl)