block/wbt: fix negative inflight counter when remove scsi device
authorLaibin Qiu <qiulaibin@huawei.com>
Sat, 22 Jan 2022 11:10:45 +0000 (19:10 +0800)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Wed, 23 Feb 2022 11:01:04 +0000 (12:01 +0100)
commit598dbaf74b64f51aa09bdbb8d15e957330ff2659
tree12f91fc931f334d688c981227e128463138a1a73
parentdc6faa0ede4d190b25980dd648e2be808fc58ff5
block/wbt: fix negative inflight counter when remove scsi device

commit e92bc4cd34de2ce454bdea8cd198b8067ee4e123 upstream.

Now that we disable wbt by set WBT_STATE_OFF_DEFAULT in
wbt_disable_default() when switch elevator to bfq. And when
we remove scsi device, wbt will be enabled by wbt_enable_default.
If it become false positive between wbt_wait() and wbt_track()
when submit write request.

The following is the scenario that triggered the problem.

T1                          T2                           T3
                            elevator_switch_mq
                            bfq_init_queue
                            wbt_disable_default <= Set
                            rwb->enable_state (OFF)
Submit_bio
blk_mq_make_request
rq_qos_throttle
<= rwb->enable_state (OFF)
                                                         scsi_remove_device
                                                         sd_remove
                                                         del_gendisk
                                                         blk_unregister_queue
                                                         elv_unregister_queue
                                                         wbt_enable_default
                                                         <= Set rwb->enable_state (ON)
q_qos_track
<= rwb->enable_state (ON)
^^^^^^ this request will mark WBT_TRACKED without inflight add and will
lead to drop rqw->inflight to -1 in wbt_done() which will trigger IO hung.

Fix this by move wbt_enable_default() from elv_unregister to
bfq_exit_queue(). Only re-enable wbt when bfq exit.

Fixes: 76a8040817b4b ("blk-wbt: make sure throttle is enabled properly")

Remove oneline stale comment, and kill one oneshot local variable.

Signed-off-by: Ming Lei <ming.lei@rehdat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Link: https://lore.kernel.org/linux-block/20211214133103.551813-1-qiulaibin@huawei.com/
Signed-off-by: Laibin Qiu <qiulaibin@huawei.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
block/bfq-iosched.c
block/elevator.c