| Message ID | 20180613143558.GA1163@lst.de (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
Hi Christoph On 06/13/2018 10:35 PM, hch@lst.de wrote: > On Wed, Jun 13, 2018 at 02:08:12PM +0000, Bart Van Assche wrote: >> __blk_mq_complete_request() is already called today by blk_mq_complete_request(). >> However, it's not clear to me why that function is exported by Jianchao's patch. > > True. I had missed that the patch also started calling the new > mark_rq_complete function from the error handler. > >> The SCSI error handler already waits until all pending requests have finished >> before it starts handling timed out commands. This e-mail thread started with a >> report of a crash in the SCSI error handler, which is a regression introduced in >> the v4.18 merge window. > > ut-requests-again-that-are-in-the.patch > Yeah. I've read back a bit. If your theory of a double invocation of > the timeout handler is correct something like the patch below should sort > it out, right? Yes, it will work for this issue. But we should also clear the RQF_TIMED_OUT in __blk_mq_requeue_request and blk_mq_rq_ctx_init. Thanks Jianchao > > --- >>From d408928360f087c0ad24e31d1d25533c698b8b35 Mon Sep 17 00:00:00 2001 > From: Christoph Hellwig <hch@lst.de> > Date: Wed, 13 Jun 2018 16:25:40 +0200 > Subject: blk-mq: don't time out requests again that are in the timeout handler > > Signed-off-by: Christoph Hellwig <hch@lst.de> > --- > block/blk-mq.c | 4 ++++ > include/linux/blkdev.h | 2 ++ > 2 files changed, 6 insertions(+) > > diff --git a/block/blk-mq.c b/block/blk-mq.c > index e9da5e6a8526..8a2895fed078 100644 > --- a/block/blk-mq.c > +++ b/block/blk-mq.c > @@ -770,6 +770,7 @@ EXPORT_SYMBOL(blk_mq_tag_to_rq); > > static void blk_mq_rq_timed_out(struct request *req, bool reserved) > { > + req->rq_flags |= RQF_TIMED_OUT; > if (req->q->mq_ops->timeout) { > enum blk_eh_timer_return ret; > > @@ -779,6 +780,7 @@ static void blk_mq_rq_timed_out(struct request *req, bool reserved) > WARN_ON_ONCE(ret != BLK_EH_RESET_TIMER); > } > > + req->rq_flags &= ~RQF_TIMED_OUT; > blk_add_timer(req); > } > > @@ -788,6 +790,8 @@ static bool blk_mq_req_expired(struct request *rq, unsigned long *next) > > if (blk_mq_rq_state(rq) != MQ_RQ_IN_FLIGHT) > return false; > + if (rq->rq_flags & RQF_TIMED_OUT) > + return false; > > deadline = blk_rq_deadline(rq); > if (time_after_eq(jiffies, deadline)) > diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h > index bca3a92eb55f..fa6f11751430 100644 > --- a/include/linux/blkdev.h > +++ b/include/linux/blkdev.h > @@ -127,6 +127,8 @@ typedef __u32 __bitwise req_flags_t; > #define RQF_ZONE_WRITE_LOCKED ((__force req_flags_t)(1 << 19)) > /* already slept for hybrid poll */ > #define RQF_MQ_POLL_SLEPT ((__force req_flags_t)(1 << 20)) > +/* ->timeout has been called, don't expire again */ > +#define RQF_TIMED_OUT ((__force req_flags_t)(1 << 21)) > > /* flags that prevent us from merging requests: */ > #define RQF_NOMERGE_FLAGS \ >
On Thu, Jun 14, 2018 at 03:49:28PM +0800, jianchao.wang wrote: > Yes, it will work for this issue. > But we should also clear the RQF_TIMED_OUT in __blk_mq_requeue_request and blk_mq_rq_ctx_init. blk_mq_rq_ctx_init always clears rq_flags, so we don't need to explcitly clear it. __blk_mq_requeue_request probablys needs it so that we handle the case of requeuing from the error handling indeed.
diff --git a/block/blk-mq.c b/block/blk-mq.c index e9da5e6a8526..8a2895fed078 100644 --- a/block/blk-mq.c +++ b/block/blk-mq.c @@ -770,6 +770,7 @@ EXPORT_SYMBOL(blk_mq_tag_to_rq); static void blk_mq_rq_timed_out(struct request *req, bool reserved) { + req->rq_flags |= RQF_TIMED_OUT; if (req->q->mq_ops->timeout) { enum blk_eh_timer_return ret; @@ -779,6 +780,7 @@ static void blk_mq_rq_timed_out(struct request *req, bool reserved) WARN_ON_ONCE(ret != BLK_EH_RESET_TIMER); } + req->rq_flags &= ~RQF_TIMED_OUT; blk_add_timer(req); } @@ -788,6 +790,8 @@ static bool blk_mq_req_expired(struct request *rq, unsigned long *next) if (blk_mq_rq_state(rq) != MQ_RQ_IN_FLIGHT) return false; + if (rq->rq_flags & RQF_TIMED_OUT) + return false; deadline = blk_rq_deadline(rq); if (time_after_eq(jiffies, deadline)) diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h index bca3a92eb55f..fa6f11751430 100644 --- a/include/linux/blkdev.h +++ b/include/linux/blkdev.h @@ -127,6 +127,8 @@ typedef __u32 __bitwise req_flags_t; #define RQF_ZONE_WRITE_LOCKED ((__force req_flags_t)(1 << 19)) /* already slept for hybrid poll */ #define RQF_MQ_POLL_SLEPT ((__force req_flags_t)(1 << 20)) +/* ->timeout has been called, don't expire again */ +#define RQF_TIMED_OUT ((__force req_flags_t)(1 << 21)) /* flags that prevent us from merging requests: */ #define RQF_NOMERGE_FLAGS \