| Message ID | 1448037342-18384-4-git-send-email-hch@lst.de (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
Christoph Hellwig <hch@lst.de> writes: > Timer context is not very useful for drivers to perform any meaningful abort > action from. So instead of calling the driver from this useless context > defer it to a workqueue as soon as possible. > > Note that while a delayed_work item would seem the right thing here I didn't > dare to use it due to the magic in blk_add_timer that pokes deep into timer > internals. But maybe this encourages Tejun to add a sensible API for that to > the workqueue API and we'll all be fine in the end :) I don't see where the blk-mq timeout work is ever scheduled. You removed the call to setup_timer for the mq case, so what causes the mq timeout work to run? I must be missing something. -Jeff > > Signed-off-by: Christoph Hellwig <hch@lst.de> > --- > block/blk-core.c | 8 ++++++++ > block/blk-mq.c | 8 +++++--- > block/blk-timeout.c | 5 +++-- > block/blk.h | 2 +- > include/linux/blkdev.h | 1 + > 5 files changed, 18 insertions(+), 6 deletions(-) > > diff --git a/block/blk-core.c b/block/blk-core.c > index 5131993b..1de0974 100644 > --- a/block/blk-core.c > +++ b/block/blk-core.c > @@ -664,6 +664,13 @@ static void blk_queue_usage_counter_release(struct percpu_ref *ref) > wake_up_all(&q->mq_freeze_wq); > } > > +static void blk_rq_timed_out_timer(unsigned long data) > +{ > + struct request_queue *q = (struct request_queue *)data; > + > + kblockd_schedule_work(&q->timeout_work); > +} > + > struct request_queue *blk_alloc_queue_node(gfp_t gfp_mask, int node_id) > { > struct request_queue *q; > @@ -825,6 +832,7 @@ blk_init_allocated_queue(struct request_queue *q, request_fn_proc *rfn, > if (blk_init_rl(&q->root_rl, q, GFP_KERNEL)) > goto fail; > > + INIT_WORK(&q->timeout_work, blk_timeout_work); > q->request_fn = rfn; > q->prep_rq_fn = NULL; > q->unprep_rq_fn = NULL; > diff --git a/block/blk-mq.c b/block/blk-mq.c > index 3ae09de..8354601 100644 > --- a/block/blk-mq.c > +++ b/block/blk-mq.c > @@ -85,6 +85,7 @@ void blk_mq_freeze_queue_start(struct request_queue *q) > freeze_depth = atomic_inc_return(&q->mq_freeze_depth); > if (freeze_depth == 1) { > percpu_ref_kill(&q->q_usage_counter); > + cancel_work_sync(&q->timeout_work); > blk_mq_run_hw_queues(q, false); > } > } > @@ -617,9 +618,10 @@ static void blk_mq_check_expired(struct blk_mq_hw_ctx *hctx, > } > } > > -static void blk_mq_rq_timer(unsigned long priv) > +static void blk_mq_timeout_work(struct work_struct *work) > { > - struct request_queue *q = (struct request_queue *)priv; > + struct request_queue *q = > + container_of(work, struct request_queue, timeout_work); > struct blk_mq_timeout_data data = { > .next = 0, > .next_set = 0, > @@ -2015,7 +2017,7 @@ struct request_queue *blk_mq_init_allocated_queue(struct blk_mq_tag_set *set, > hctxs[i]->queue_num = i; > } > > - setup_timer(&q->timeout, blk_mq_rq_timer, (unsigned long) q); > + INIT_WORK(&q->timeout_work, blk_mq_timeout_work); > blk_queue_rq_timeout(q, set->timeout ? set->timeout : 30 * HZ); > > q->nr_queues = nr_cpu_ids; > diff --git a/block/blk-timeout.c b/block/blk-timeout.c > index aa40aa9..aedd128 100644 > --- a/block/blk-timeout.c > +++ b/block/blk-timeout.c > @@ -127,9 +127,10 @@ static void blk_rq_check_expired(struct request *rq, unsigned long *next_timeout > } > } > > -void blk_rq_timed_out_timer(unsigned long data) > +void blk_timeout_work(struct work_struct *work) > { > - struct request_queue *q = (struct request_queue *) data; > + struct request_queue *q = > + container_of(work, struct request_queue, timeout_work); > unsigned long flags, next = 0; > struct request *rq, *tmp; > int next_set = 0; > diff --git a/block/blk.h b/block/blk.h > index da722eb..37b9165 100644 > --- a/block/blk.h > +++ b/block/blk.h > @@ -95,7 +95,7 @@ static inline void blk_flush_integrity(void) > } > #endif > > -void blk_rq_timed_out_timer(unsigned long data); > +void blk_timeout_work(struct work_struct *work); > unsigned long blk_rq_timeout(unsigned long timeout); > void blk_add_timer(struct request *req); > void blk_delete_timer(struct request *); > diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h > index 3fe27f8..9a8424a 100644 > --- a/include/linux/blkdev.h > +++ b/include/linux/blkdev.h > @@ -407,6 +407,7 @@ struct request_queue { > > unsigned int rq_timeout; > struct timer_list timeout; > + struct work_struct timeout_work; > struct list_head timeout_list; > > struct list_head icq_list; -- To unsubscribe from this list: send the line "unsubscribe linux-block" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Mon, Nov 23, 2015 at 03:31:32PM -0500, Jeff Moyer wrote: > I don't see where the blk-mq timeout work is ever scheduled. You > removed the call to setup_timer for the mq case, so what causes the mq > timeout work to run? I must be missing something. No we use the timer which we initialize in blk_alloc_queue_node instead of overwriting it later in the blk-mq case. The timer gets added in blk_add_timer either way. -- To unsubscribe from this list: send the line "unsubscribe linux-block" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Christoph Hellwig <hch@lst.de> writes: > On Mon, Nov 23, 2015 at 03:31:32PM -0500, Jeff Moyer wrote: >> I don't see where the blk-mq timeout work is ever scheduled. You >> removed the call to setup_timer for the mq case, so what causes the mq >> timeout work to run? I must be missing something. > > No we use the timer which we initialize in blk_alloc_queue_node instead > of overwriting it later in the blk-mq case. The timer gets added in > blk_add_timer either way. Ah, I see. I missed the call to blk_alloc_queue_node from blk_mq_init_queue. Thanks. The patch looks good to me. Everything will still be running in the same context (irqs off). Reviewed-by: Jeff Moyer <jmoyer@redhat.com> -- To unsubscribe from this list: send the line "unsubscribe linux-block" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
diff --git a/block/blk-core.c b/block/blk-core.c index 5131993b..1de0974 100644 --- a/block/blk-core.c +++ b/block/blk-core.c @@ -664,6 +664,13 @@ static void blk_queue_usage_counter_release(struct percpu_ref *ref) wake_up_all(&q->mq_freeze_wq); } +static void blk_rq_timed_out_timer(unsigned long data) +{ + struct request_queue *q = (struct request_queue *)data; + + kblockd_schedule_work(&q->timeout_work); +} + struct request_queue *blk_alloc_queue_node(gfp_t gfp_mask, int node_id) { struct request_queue *q; @@ -825,6 +832,7 @@ blk_init_allocated_queue(struct request_queue *q, request_fn_proc *rfn, if (blk_init_rl(&q->root_rl, q, GFP_KERNEL)) goto fail; + INIT_WORK(&q->timeout_work, blk_timeout_work); q->request_fn = rfn; q->prep_rq_fn = NULL; q->unprep_rq_fn = NULL; diff --git a/block/blk-mq.c b/block/blk-mq.c index 3ae09de..8354601 100644 --- a/block/blk-mq.c +++ b/block/blk-mq.c @@ -85,6 +85,7 @@ void blk_mq_freeze_queue_start(struct request_queue *q) freeze_depth = atomic_inc_return(&q->mq_freeze_depth); if (freeze_depth == 1) { percpu_ref_kill(&q->q_usage_counter); + cancel_work_sync(&q->timeout_work); blk_mq_run_hw_queues(q, false); } } @@ -617,9 +618,10 @@ static void blk_mq_check_expired(struct blk_mq_hw_ctx *hctx, } } -static void blk_mq_rq_timer(unsigned long priv) +static void blk_mq_timeout_work(struct work_struct *work) { - struct request_queue *q = (struct request_queue *)priv; + struct request_queue *q = + container_of(work, struct request_queue, timeout_work); struct blk_mq_timeout_data data = { .next = 0, .next_set = 0, @@ -2015,7 +2017,7 @@ struct request_queue *blk_mq_init_allocated_queue(struct blk_mq_tag_set *set, hctxs[i]->queue_num = i; } - setup_timer(&q->timeout, blk_mq_rq_timer, (unsigned long) q); + INIT_WORK(&q->timeout_work, blk_mq_timeout_work); blk_queue_rq_timeout(q, set->timeout ? set->timeout : 30 * HZ); q->nr_queues = nr_cpu_ids; diff --git a/block/blk-timeout.c b/block/blk-timeout.c index aa40aa9..aedd128 100644 --- a/block/blk-timeout.c +++ b/block/blk-timeout.c @@ -127,9 +127,10 @@ static void blk_rq_check_expired(struct request *rq, unsigned long *next_timeout } } -void blk_rq_timed_out_timer(unsigned long data) +void blk_timeout_work(struct work_struct *work) { - struct request_queue *q = (struct request_queue *) data; + struct request_queue *q = + container_of(work, struct request_queue, timeout_work); unsigned long flags, next = 0; struct request *rq, *tmp; int next_set = 0; diff --git a/block/blk.h b/block/blk.h index da722eb..37b9165 100644 --- a/block/blk.h +++ b/block/blk.h @@ -95,7 +95,7 @@ static inline void blk_flush_integrity(void) } #endif -void blk_rq_timed_out_timer(unsigned long data); +void blk_timeout_work(struct work_struct *work); unsigned long blk_rq_timeout(unsigned long timeout); void blk_add_timer(struct request *req); void blk_delete_timer(struct request *); diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h index 3fe27f8..9a8424a 100644 --- a/include/linux/blkdev.h +++ b/include/linux/blkdev.h @@ -407,6 +407,7 @@ struct request_queue { unsigned int rq_timeout; struct timer_list timeout; + struct work_struct timeout_work; struct list_head timeout_list; struct list_head icq_list;
Timer context is not very useful for drivers to perform any meaningful abort action from. So instead of calling the driver from this useless context defer it to a workqueue as soon as possible. Note that while a delayed_work item would seem the right thing here I didn't dare to use it due to the magic in blk_add_timer that pokes deep into timer internals. But maybe this encourages Tejun to add a sensible API for that to the workqueue API and we'll all be fine in the end :) Signed-off-by: Christoph Hellwig <hch@lst.de> --- block/blk-core.c | 8 ++++++++ block/blk-mq.c | 8 +++++--- block/blk-timeout.c | 5 +++-- block/blk.h | 2 +- include/linux/blkdev.h | 1 + 5 files changed, 18 insertions(+), 6 deletions(-)