On 4/2/25 10:08 AM, Ming Lei wrote: > Use blk_mq_no_io() to prevent IO from entering queue for avoiding lock > dependency between freeze lock and elevator lock, and we have got many > such reports: > > Reported-by: syzbot+4c7e0f9b94ad65811efb@xxxxxxxxxxxxxxxxxxxxxxxxx > Closes: https://lore.kernel.org/linux-block/67e6b425.050a0220.2f068f.007b.GAE@xxxxxxxxxx/ > Reported-by: Valdis Klētnieks <valdis.kletnieks@xxxxxx> > Closes: https://lore.kernel.org/linux-block/7755.1743228130@turing-police/#t > Signed-off-by: Ming Lei <ming.lei@xxxxxxxxxx> I tested this series on my system and this works well as we cut dependency between ->elevator_lock and ->freeze_lock. However don't we plan to now model blk_mq_enter_no_io and blk_mq_exit_no_io as lock/unlock for supporting lockdep? Maybe we don't. Overall changes looks good to me: Reviewed-by: Nilay Shroff <nilay@xxxxxxxxxxxxx>
