| Message ID | 20180108185011.30018-3-bart.vanassche@wdc.com (mailing list archive) |
|---|---|
| State | Not Applicable |
| Headers | show |
On 01/08/2018 07:50 PM, Bart Van Assche wrote: > Introduce functions that allow block drivers to wait while a request > queue is in the quiesced state (blk-mq) or in the stopped state (legacy > block layer). The next patch will add calls to these functions in the > SCSI core. > > Signed-off-by: Bart Van Assche <bart.vanassche@wdc.com> > Cc: Martin K. Petersen <martin.petersen@oracle.com> > Cc: Christoph Hellwig <hch@lst.de> > Cc: Hannes Reinecke <hare@suse.de> > Cc: Johannes Thumshirn <jthumshirn@suse.de> > Cc: Ming Lei <ming.lei@redhat.com> > --- > block/blk-core.c | 1 + > block/blk-mq.c | 61 ++++++++++++++++++++++++++++++++++++++++++++++++++ > include/linux/blk-mq.h | 2 ++ > 3 files changed, 64 insertions(+) > > diff --git a/block/blk-core.c b/block/blk-core.c > index 605599a2ab3b..d70ff53e6505 100644 > --- a/block/blk-core.c > +++ b/block/blk-core.c > @@ -285,6 +285,7 @@ void blk_start_queue(struct request_queue *q) > WARN_ON_ONCE(q->mq_ops); > > queue_flag_clear(QUEUE_FLAG_STOPPED, q); > + wake_up_all(&q->mq_wq); > __blk_run_queue(q); > } > EXPORT_SYMBOL(blk_start_queue); > diff --git a/block/blk-mq.c b/block/blk-mq.c > index 8118890fb66f..c79b102680fe 100644 > --- a/block/blk-mq.c > +++ b/block/blk-mq.c > @@ -248,11 +248,72 @@ void blk_mq_unquiesce_queue(struct request_queue *q) > queue_flag_clear(QUEUE_FLAG_QUIESCED, q); > spin_unlock_irqrestore(q->queue_lock, flags); > > + wake_up_all(&q->mq_wq); > + > /* dispatch requests which are inserted during quiescing */ > blk_mq_run_hw_queues(q, true); > } > EXPORT_SYMBOL_GPL(blk_mq_unquiesce_queue); > > +/** > + * blk_wait_if_quiesced() - wait if a queue is quiesced (blk-mq) or stopped (legacy block layer) > + * @q: Request queue pointer. > + * > + * Some block drivers, e.g. the SCSI core, can bypass the block layer core > + * request submission mechanism. Surround such code with blk_wait_if_quiesced() > + * / blk_finish_wait_if_quiesced() to avoid that request submission can happen > + * while a queue is quiesced or stopped. > + * > + * Returns with the RCU read lock held (blk-mq) or with q->queue_lock held > + * (legacy block layer). > + * > + * Note: this function does not support block drivers whose .queue_rq() > + * implementation can sleep (BLK_MQ_F_BLOCKING). > + */ > +int blk_wait_if_quiesced(struct request_queue *q) > +{ > + struct blk_mq_hw_ctx *hctx; > + unsigned int i; > + > + might_sleep(); > + > + if (q->mq_ops) { > + queue_for_each_hw_ctx(q, hctx, i) > + WARN_ON(hctx->flags & BLK_MQ_F_BLOCKING); > + > + rcu_read_lock(); > + while (!blk_queue_dying(q) && blk_queue_quiesced(q)) { > + rcu_read_unlock(); > + wait_event(q->mq_wq, blk_queue_dying(q) || > + !blk_queue_quiesced(q)); > + rcu_read_lock(); > + } > + } else { > + spin_lock_irq(q->queue_lock); > + wait_event_lock_irq(q->mq_wq, > + blk_queue_dying(q) || !blk_queue_stopped(q), > + *q->queue_lock); > + q->request_fn_active++; > + } > + return blk_queue_dying(q) ? -ENODEV : 0; > +} > +EXPORT_SYMBOL(blk_wait_if_quiesced); > + > +/** > + * blk_finish_wait_if_quiesced() - counterpart of blk_wait_if_quiesced() > + * @q: Request queue pointer. > + */ > +void blk_finish_wait_if_quiesced(struct request_queue *q) > +{ > + if (q->mq_ops) { > + rcu_read_unlock(); > + } else { > + q->request_fn_active--; > + spin_unlock_irq(q->queue_lock); > + } > +} > +EXPORT_SYMBOL(blk_finish_wait_if_quiesced); > + > void blk_mq_wake_waiters(struct request_queue *q) > { > struct blk_mq_hw_ctx *hctx; > diff --git a/include/linux/blk-mq.h b/include/linux/blk-mq.h > index 95c9a5c862e2..f6b787bd244e 100644 > --- a/include/linux/blk-mq.h > +++ b/include/linux/blk-mq.h > @@ -266,6 +266,8 @@ void blk_mq_start_stopped_hw_queue(struct blk_mq_hw_ctx *hctx, bool async); > void blk_mq_start_stopped_hw_queues(struct request_queue *q, bool async); > void blk_mq_quiesce_queue(struct request_queue *q); > void blk_mq_unquiesce_queue(struct request_queue *q); > +int blk_wait_if_quiesced(struct request_queue *q); > +void blk_finish_wait_if_quiesced(struct request_queue *q); > void blk_mq_delay_run_hw_queue(struct blk_mq_hw_ctx *hctx, unsigned long msecs); > bool blk_mq_run_hw_queue(struct blk_mq_hw_ctx *hctx, bool async); > void blk_mq_run_hw_queues(struct request_queue *q, bool async); > I'm always a bit cautious when having rcu_read_lock() and rcu_read_unlock() in two separate functions. Can we make this dependency more explicit by renaming the first function to blk_start_wait_if_quiesced() and updating the comment to the effect that both functions must be used in tandem? Cheers, Hannes
On Tue, 2018-01-09 at 07:41 +0100, Hannes Reinecke wrote: > I'm always a bit cautious when having rcu_read_lock() and > rcu_read_unlock() in two separate functions. > Can we make this dependency more explicit by renaming the first function > to blk_start_wait_if_quiesced() and updating the comment to the effect > that both functions must be used in tandem? Hello Hannes, That sounds like a good idea to me. I will make these changes and repost this patch series. Bart.
diff --git a/block/blk-core.c b/block/blk-core.c index 605599a2ab3b..d70ff53e6505 100644 --- a/block/blk-core.c +++ b/block/blk-core.c @@ -285,6 +285,7 @@ void blk_start_queue(struct request_queue *q) WARN_ON_ONCE(q->mq_ops); queue_flag_clear(QUEUE_FLAG_STOPPED, q); + wake_up_all(&q->mq_wq); __blk_run_queue(q); } EXPORT_SYMBOL(blk_start_queue); diff --git a/block/blk-mq.c b/block/blk-mq.c index 8118890fb66f..c79b102680fe 100644 --- a/block/blk-mq.c +++ b/block/blk-mq.c @@ -248,11 +248,72 @@ void blk_mq_unquiesce_queue(struct request_queue *q) queue_flag_clear(QUEUE_FLAG_QUIESCED, q); spin_unlock_irqrestore(q->queue_lock, flags); + wake_up_all(&q->mq_wq); + /* dispatch requests which are inserted during quiescing */ blk_mq_run_hw_queues(q, true); } EXPORT_SYMBOL_GPL(blk_mq_unquiesce_queue); +/** + * blk_wait_if_quiesced() - wait if a queue is quiesced (blk-mq) or stopped (legacy block layer) + * @q: Request queue pointer. + * + * Some block drivers, e.g. the SCSI core, can bypass the block layer core + * request submission mechanism. Surround such code with blk_wait_if_quiesced() + * / blk_finish_wait_if_quiesced() to avoid that request submission can happen + * while a queue is quiesced or stopped. + * + * Returns with the RCU read lock held (blk-mq) or with q->queue_lock held + * (legacy block layer). + * + * Note: this function does not support block drivers whose .queue_rq() + * implementation can sleep (BLK_MQ_F_BLOCKING). + */ +int blk_wait_if_quiesced(struct request_queue *q) +{ + struct blk_mq_hw_ctx *hctx; + unsigned int i; + + might_sleep(); + + if (q->mq_ops) { + queue_for_each_hw_ctx(q, hctx, i) + WARN_ON(hctx->flags & BLK_MQ_F_BLOCKING); + + rcu_read_lock(); + while (!blk_queue_dying(q) && blk_queue_quiesced(q)) { + rcu_read_unlock(); + wait_event(q->mq_wq, blk_queue_dying(q) || + !blk_queue_quiesced(q)); + rcu_read_lock(); + } + } else { + spin_lock_irq(q->queue_lock); + wait_event_lock_irq(q->mq_wq, + blk_queue_dying(q) || !blk_queue_stopped(q), + *q->queue_lock); + q->request_fn_active++; + } + return blk_queue_dying(q) ? -ENODEV : 0; +} +EXPORT_SYMBOL(blk_wait_if_quiesced); + +/** + * blk_finish_wait_if_quiesced() - counterpart of blk_wait_if_quiesced() + * @q: Request queue pointer. + */ +void blk_finish_wait_if_quiesced(struct request_queue *q) +{ + if (q->mq_ops) { + rcu_read_unlock(); + } else { + q->request_fn_active--; + spin_unlock_irq(q->queue_lock); + } +} +EXPORT_SYMBOL(blk_finish_wait_if_quiesced); + void blk_mq_wake_waiters(struct request_queue *q) { struct blk_mq_hw_ctx *hctx; diff --git a/include/linux/blk-mq.h b/include/linux/blk-mq.h index 95c9a5c862e2..f6b787bd244e 100644 --- a/include/linux/blk-mq.h +++ b/include/linux/blk-mq.h @@ -266,6 +266,8 @@ void blk_mq_start_stopped_hw_queue(struct blk_mq_hw_ctx *hctx, bool async); void blk_mq_start_stopped_hw_queues(struct request_queue *q, bool async); void blk_mq_quiesce_queue(struct request_queue *q); void blk_mq_unquiesce_queue(struct request_queue *q); +int blk_wait_if_quiesced(struct request_queue *q); +void blk_finish_wait_if_quiesced(struct request_queue *q); void blk_mq_delay_run_hw_queue(struct blk_mq_hw_ctx *hctx, unsigned long msecs); bool blk_mq_run_hw_queue(struct blk_mq_hw_ctx *hctx, bool async); void blk_mq_run_hw_queues(struct request_queue *q, bool async);
Introduce functions that allow block drivers to wait while a request queue is in the quiesced state (blk-mq) or in the stopped state (legacy block layer). The next patch will add calls to these functions in the SCSI core. Signed-off-by: Bart Van Assche <bart.vanassche@wdc.com> Cc: Martin K. Petersen <martin.petersen@oracle.com> Cc: Christoph Hellwig <hch@lst.de> Cc: Hannes Reinecke <hare@suse.de> Cc: Johannes Thumshirn <jthumshirn@suse.de> Cc: Ming Lei <ming.lei@redhat.com> --- block/blk-core.c | 1 + block/blk-mq.c | 61 ++++++++++++++++++++++++++++++++++++++++++++++++++ include/linux/blk-mq.h | 2 ++ 3 files changed, 64 insertions(+)