If removal is from breaking error recovery, queues may be frozen, and there may be pending IOs in bio_queue_enter(), and the following del_gendisk() may wait for these IOs, especially from writeback. Similar IO hang exists in flushing scan work too if there are pending IO in scan work context. Fix the kind of issue by unfreezing queues before removing namespace, so that all pending IOs can be handled. Reported-by: Chunguang Xu <brookxu.cn@xxxxxxxxx> https://lore.kernel.org/linux-nvme/cover.1685350577.git.chunguang.xu@xxxxxxxxxx/ Reported-by: Yi Zhang <yi.zhang@xxxxxxxxxx> Signed-off-by: Ming Lei <ming.lei@xxxxxxxxxx> --- drivers/nvme/host/core.c | 3 +++ 1 file changed, 3 insertions(+) diff --git a/drivers/nvme/host/core.c b/drivers/nvme/host/core.c index 6b3f12368196..7d8ff58660ee 100644 --- a/drivers/nvme/host/core.c +++ b/drivers/nvme/host/core.c @@ -4002,6 +4002,9 @@ void nvme_remove_namespaces(struct nvme_ctrl *ctrl) */ nvme_mpath_clear_ctrl_paths(ctrl); + /* unfreeze queues which may be frozen from error recovery */ + nvme_unfreeze_force(ctrl); + /* prevent racing with ns scanning */ flush_work(&ctrl->scan_work); -- 2.40.1
