block: flush: fix IO hang in case of flood fua req

This patch fixes one issue reported by Kent, which can
be triggered in bcachefs over sata disk. Actually it
is a generic issue in block flush vs. blk-tag.

Cc: Christoph Hellwig <hch@infradead.org>
Reported-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Ming Lei <tom.leiming@gmail.com>
Signed-off-by: Jens Axboe <axboe@fb.com>

diff --git a/block/blk-flush.c b/block/blk-flush.c
index 6a14b68..3c882cb 100644 (file)
--- a/block/blk-flush.c
+++ b/block/blk-flush.c
@@ -342,6 +342,34 @@ static void flush_data_end_io(struct request *rq, int error)
        struct request_queue *q = rq->q;
        struct blk_flush_queue *fq = blk_get_flush_queue(q, NULL);
 
+       /*
+        * Updating q->in_flight[] here for making this tag usable
+        * early. Because in blk_queue_start_tag(),
+        * q->in_flight[BLK_RW_ASYNC] is used to limit async I/O and
+        * reserve tags for sync I/O.
+        *
+        * More importantly this way can avoid the following I/O
+        * deadlock:
+        *
+        * - suppose there are 40 fua requests comming to flush queue
+        *   and queue depth is 31
+        * - 30 rqs are scheduled then blk_queue_start_tag() can't alloc
+        *   tag for async I/O any more
+        * - all the 30 rqs are completed before FLUSH_PENDING_TIMEOUT
+        *   and flush_data_end_io() is called
+        * - the other rqs still can't go ahead if not updating
+        *   q->in_flight[BLK_RW_ASYNC] here, meantime these rqs
+        *   are held in flush data queue and make no progress of
+        *   handling post flush rq
+        * - only after the post flush rq is handled, all these rqs
+        *   can be completed
+        */
+
+       elv_completed_request(q, rq);
+
+       /* for avoiding double accounting */
+       rq->cmd_flags &= ~REQ_STARTED;
+
        /*
         * After populating an empty queue, kick it to avoid stall.  Read
         * the comment in flush_end_io().