From: Tetsuo Handa Subject: Re: INFO: task hung in blk_queue_enter Date: Wed, 16 May 2018 22:05:06 +0900 Message-ID: References: <0000000000009b212b056ae6dbad@google.com> <343bbbf6-64eb-879e-d19e-96aebb037d47@I-love.SAKURA.ne.jp> Mime-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Transfer-Encoding: 7bit Cc: linux-kernel@vger.kernel.org, linux-ext4@vger.kernel.org, dvyukov@google.com, Alan Jenkins , Bart Van Assche , Christoph Hellwig , Hannes Reinecke , Johannes Thumshirn , Keith Busch , "Martin K. Petersen" , Martin Steigerwald , Ming Lei , Oleksandr Natalenko , Ross Zwisler To: syzbot , linux-block@vger.kernel.org, syzkaller-bugs@googlegroups.com, Dan Williams , Jens Axboe Return-path: In-Reply-To: <343bbbf6-64eb-879e-d19e-96aebb037d47@I-love.SAKURA.ne.jp> Content-Language: en-US Sender: linux-kernel-owner@vger.kernel.org List-Id: linux-ext4.vger.kernel.org Tetsuo Handa wrote: > I couldn't check whether freeze_depth in blk_freeze_queue_start() was 1, > but presumably q->mq_freeze_depth > 0 because syz-executor7(PID=5010) is > stuck at wait_event() in blk_queue_enter(). > > Since flags == 0, preempt == false. Since stuck at wait_event(), success == false. > Thus, atomic_read(&q->mq_freeze_depth) > 0 if blk_queue_dying(q) == false. And I > guess blk_queue_dying(q) == false because we are just trying to freeze/unfreeze. > I was able to reproduce the hung up using modified reproducer, and got values using below debug printk() patch. --- a/block/blk-core.c +++ b/block/blk-core.c @@ -950,10 +950,12 @@ int blk_queue_enter(struct request_queue *q, blk_mq_req_flags_t flags) */ smp_rmb(); - wait_event(q->mq_freeze_wq, - (atomic_read(&q->mq_freeze_depth) == 0 && - (preempt || !blk_queue_preempt_only(q))) || - blk_queue_dying(q)); + while (wait_event_timeout(q->mq_freeze_wq, + (atomic_read(&q->mq_freeze_depth) == 0 && + (preempt || !blk_queue_preempt_only(q))) || + blk_queue_dying(q), 10 * HZ) == 0) + printk("%s(%u): q->mq_freeze_depth=%d preempt=%d blk_queue_preempt_only(q)=%d blk_queue_dying(q)=%d\n", + current->comm, current->pid, atomic_read(&q->mq_freeze_depth), preempt, blk_queue_preempt_only(q), blk_queue_dying(q)); if (blk_queue_dying(q)) return -ENODEV; } [ 75.869126] print_req_error: I/O error, dev loop0, sector 0 [ 85.983146] a.out(8838): q->mq_freeze_depth=1 preempt=0 blk_queue_preempt_only(q)=0 blk_queue_dying(q)=0 [ 96.222884] a.out(8838): q->mq_freeze_depth=1 preempt=0 blk_queue_preempt_only(q)=0 blk_queue_dying(q)=0 [ 106.463322] a.out(8838): q->mq_freeze_depth=1 preempt=0 blk_queue_preempt_only(q)=0 blk_queue_dying(q)=0 [ 116.702912] a.out(8838): q->mq_freeze_depth=1 preempt=0 blk_queue_preempt_only(q)=0 blk_queue_dying(q)=0 One ore more threads are waiting for q->mq_freeze_depth to become 0. But the thread who incremented q->mq_freeze_depth at blk_freeze_queue_start(q) from blk_freeze_queue() is waiting at blk_mq_freeze_queue_wait(). Therefore, atomic_read(&q->mq_freeze_depth) == 0 condition for wait_event() in blk_queue_enter() will never be satisfied. But what does that wait_event() want to do? Isn't "start freezing" a sort of blk_queue_dying(q) == true? Since percpu_ref_tryget_live(&q->q_usage_counter) failed and the queue is about to be frozen, shouldn't we treat atomic_read(&q->mq_freeze_depth) != 0 as if blk_queue_dying(q) == true? That is, something like below: diff --git a/block/blk-core.c b/block/blk-core.c index 85909b4..59e2496 100644 --- a/block/blk-core.c +++ b/block/blk-core.c @@ -951,10 +951,10 @@ int blk_queue_enter(struct request_queue *q, blk_mq_req_flags_t flags) smp_rmb(); wait_event(q->mq_freeze_wq, - (atomic_read(&q->mq_freeze_depth) == 0 && - (preempt || !blk_queue_preempt_only(q))) || + atomic_read(&q->mq_freeze_depth) || + (preempt || !blk_queue_preempt_only(q)) || blk_queue_dying(q)); - if (blk_queue_dying(q)) + if (atomic_read(&q->mq_freeze_depth) || blk_queue_dying(q)) return -ENODEV; } }