nvme/tcp: fix bug on double requeue when send fails
authorAnton Eidelman <anton@lightbitslabs.com>
Mon, 10 Feb 2020 18:37:18 +0000 (10:37 -0800)
committerJens Axboe <axboe@kernel.dk>
Fri, 14 Feb 2020 17:12:04 +0000 (10:12 -0700)
commit2d570a7c0251c594489a2c16b82b14ae30345c03
tree846d1827b352e9d0f3dffd294f507125f863e756
parent4ec31cb6241d95879aac337cc6b50c45dd10cfcb
nvme/tcp: fix bug on double requeue when send fails

When nvme_tcp_io_work() fails to send to socket due to
connection close/reset, error_recovery work is triggered
from nvme_tcp_state_change() socket callback.
This cancels all the active requests in the tagset,
which requeues them.

The failed request, however, was ended and thus requeued
individually as well unless send returned -EPIPE.
Another return code to be treated the same way is -ECONNRESET.

Double requeue caused BUG_ON(blk_queued_rq(rq))
in blk_mq_requeue_request() from either the individual requeue
of the failed request or the bulk requeue from
blk_mq_tagset_busy_iter(, nvme_cancel_request, );

Signed-off-by: Anton Eidelman <anton@lightbitslabs.com>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Keith Busch <kbusch@kernel.org>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
drivers/nvme/host/tcp.c