blk-mq: insert passthrough request into hctx->dispatch directly
authorMing Lei <ming.lei@redhat.com>
Tue, 25 Feb 2020 01:04:32 +0000 (09:04 +0800)
committerJens Axboe <axboe@kernel.dk>
Tue, 25 Feb 2020 01:50:48 +0000 (18:50 -0700)
commit01e99aeca3979600302913cef3f89076786f32c8
treefd971e355c8d579a32d0689755870b498d3b2b11
parentf8788d86ab28f61f7b46eb6be375f8a726783636
blk-mq: insert passthrough request into hctx->dispatch directly

For some reason, device may be in one situation which can't handle
FS request, so STS_RESOURCE is always returned and the FS request
will be added to hctx->dispatch. However passthrough request may
be required at that time for fixing the problem. If passthrough
request is added to scheduler queue, there isn't any chance for
blk-mq to dispatch it given we prioritize requests in hctx->dispatch.
Then the FS IO request may never be completed, and IO hang is caused.

So passthrough request has to be added to hctx->dispatch directly
for fixing the IO hang.

Fix this issue by inserting passthrough request into hctx->dispatch
directly together withing adding FS request to the tail of
hctx->dispatch in blk_mq_dispatch_rq_list(). Actually we add FS request
to tail of hctx->dispatch at default, see blk_mq_request_bypass_insert().

Then it becomes consistent with original legacy IO request
path, in which passthrough request is always added to q->queue_head.

Cc: Dongli Zhang <dongli.zhang@oracle.com>
Cc: Christoph Hellwig <hch@infradead.org>
Cc: Ewan D. Milne <emilne@redhat.com>
Signed-off-by: Ming Lei <ming.lei@redhat.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
block/blk-flush.c
block/blk-mq-sched.c
block/blk-mq.c
block/blk-mq.h