[PATCH] blk-mq: don't issue request directly in case that current is to be blocked

When flushing plug list in case that current will be blocked, we can't
issue request directly because ->queue_rq() may sleep, otherwise scheduler
may complain.

Fixes: dc5fc361d891 ("block: attempt direct issue of plug list")
Signed-off-by: Ming Lei <redacted>
---
 block/blk-mq.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/block/blk-mq.c b/block/blk-mq.c
index c19dfa8ea65e..9840b15f505b 100644
--- a/block/blk-mq.c
+++ b/block/blk-mq.c

@@ -2223,7 +2223,7 @@ void blk_mq_flush_plug_list(struct blk_plug *plug, bool from_schedule)
 		return;
 	plug->rq_count = 0;
 
-	if (!plug->multiple_queues && !plug->has_elevator) {
+	if (!plug->multiple_queues && !plug->has_elevator && !from_schedule) {
 		blk_mq_plug_issue_direct(plug, from_schedule);
 		if (rq_list_empty(plug->mq_list))
 			return;

-- 
2.31.1

On Tue, 26 Oct 2021 16:22:57 +0800, Ming Lei wrote:

When flushing plug list in case that current will be blocked, we can't
issue request directly because ->queue_rq() may sleep, otherwise scheduler
may complain.

Applied, thanks!

[1/1] blk-mq: don't issue request directly in case that current is to be blocked
      commit: ff1552232b3612edff43a95746a4e78e231ef3d4

Best regards,
-- 
Jens Axboe

On 10/26/21 2:22 AM, Ming Lei wrote:

When flushing plug list in case that current will be blocked, we can't
issue request directly because ->queue_rq() may sleep, otherwise scheduler
may complain.

Oops indeed, thanks Ming.

-- 
Jens Axboe