block: queue data commands from the flush state machine at the head

9f87fc4d · Christoph Hellwig <[email protected]> · 2023-07-14 16:30:14 +02:00

We used to insert the data commands following a pre-flush to the head
of the queue until commit 1e82fadfc6b ("blk-mq: do not do head insertions
post-pre-flush commands").  Not doing this seems to cause hangs of
such commands on NFS workloads when exported from file systems with
SATA SSDs.  I have no idea why this would starve these workloads,
but doing a semantic revert of this patch (which looks quite different
due to various other changes) fixes the hangs.

Fixes: 1e82fadfc6b ("blk-mq: do not do head insertions post-pre-flush commands")
Reported-by: Chuck Lever <[email protected]>
Signed-off-by: Christoph Hellwig <[email protected]>
Tested-by: Chuck Lever <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Jens Axboe <[email protected]>

Commit: 9f87fc4d72f52b26ac3e19df5e4584227fe6740c
Parents: 90b46229

Changes

1 files changed · +1 −1

File	Status	Add	Del	Content
block/blk-flush.c	modified	+1	−1	View file

block/blk-flush.c

View file

diff --git a/block/blk-flush.c b/block/blk-flush.c
index dba392c..8220517 100644
--- a/block/blk-flush.c
+++ b/block/blk-flush.c
@@ -189,7 +189,7 @@ static void blk_flush_complete_seq(struct request *rq,
 	case REQ_FSEQ_DATA:
 		list_move_tail(&rq->flush.list, &fq->flush_data_in_flight);
 		spin_lock(&q->requeue_lock);
-		list_add_tail(&rq->queuelist, &q->flush_list);
+		list_add(&rq->queuelist, &q->requeue_list);
 		spin_unlock(&q->requeue_lock);
 		blk_mq_kick_requeue_list(q);
 		break;