[PATCH 1/2] null_blk: prevent timer handler running on a different CPU where started

[PATCH 1/2] null_blk: prevent timer handler running on a different CPU where started

[Akinobu Mita]

When irqmode=2 (IRQ completion handler is timer), timer handler should
be called on the same CPU where the timer has been started.

Since completion_queues are per-cpu and the completion handler only
touches completion_queue for local CPU, we need to prevent the handler
from running on a different CPU where the timer has been started.
Otherwise, the IO cannot be completed until another completion handler
is executed on that CPU.

Signed-off-by: Akinobu Mita <akinobu.mita@gmail.com>
Cc: Jens Axboe <axboe@fb.com>
---
 drivers/block/null_blk.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/block/null_blk.c b/drivers/block/null_blk.c
index 65cd61a..6f0a58e 100644
--- a/drivers/block/null_blk.c
+++ b/drivers/block/null_blk.c
@@ -257,7 +257,7 @@ static void null_cmd_end_timer(struct nullb_cmd *cmd)
 	if (llist_add(&cmd->ll_list, &cq->list)) {
 		ktime_t kt = ktime_set(0, completion_nsec);
 
-		hrtimer_start(&cq->timer, kt, HRTIMER_MODE_REL);
+		hrtimer_start(&cq->timer, kt, HRTIMER_MODE_REL_PINNED);
 	}
 
 	put_cpu();
-- 
1.9.1

[PATCH 2/2] null_blk: restart request processing on completion handler

[Akinobu Mita]

When irqmode=2 (IRQ completion handler is timer) and queue_mode=1
(Block interface to use is rq), the completion handler should restart
request handling for any pending requests on a queue because request
processing stops when the number of commands are queued more than
hw_queue_depth (null_rq_prep_fn returns BLKPREP_DEFER).

Without this change, the following command cannot finish.

	# modprobe null_blk irqmode=2 queue_mode=1 hw_queue_depth=1
	# fio --name=t --rw=read --size=1g --direct=1 \
	  --ioengine=libaio --iodepth=64 --filename=/dev/nullb0

Signed-off-by: Akinobu Mita <akinobu.mita@gmail.com>
Cc: Jens Axboe <axboe@fb.com>
---
 drivers/block/null_blk.c | 12 ++++++++++++
 1 file changed, 12 insertions(+)

diff --git a/drivers/block/null_blk.c b/drivers/block/null_blk.c
index 6f0a58e..6f9b753 100644
--- a/drivers/block/null_blk.c
+++ b/drivers/block/null_blk.c
@@ -243,6 +243,17 @@ static enum hrtimer_restart null_cmd_timer_expired(struct hrtimer *timer)
 			cmd = container_of(entry, struct nullb_cmd, ll_list);
 			entry = entry->next;
 			end_cmd(cmd);
+
+			if (cmd->rq) {
+				struct request_queue *q = cmd->rq->q;
+
+				if (!q->mq_ops && blk_queue_stopped(q)) {
+					spin_lock(q->queue_lock);
+					if (blk_queue_stopped(q))
+						blk_start_queue(q);
+					spin_unlock(q->queue_lock);
+				}
+			}
 		} while (entry);
 	}
 
@@ -334,6 +345,7 @@ static int null_rq_prep_fn(struct request_queue *q, struct request *req)
 		req->special = cmd;
 		return BLKPREP_OK;
 	}
+	blk_stop_queue(q);
 
 	return BLKPREP_DEFER;
 }
-- 
1.9.1

Re: [PATCH 1/2] null_blk: prevent timer handler running on a different CPU where started

[Jens Axboe]

On 06/01/2015 05:35 PM, Akinobu Mita wrote:
> When irqmode=2 (IRQ completion handler is timer), timer handler should
> be called on the same CPU where the timer has been started.
>
> Since completion_queues are per-cpu and the completion handler only
> touches completion_queue for local CPU, we need to prevent the handler
> from running on a different CPU where the timer has been started.
> Otherwise, the IO cannot be completed until another completion handler
> is executed on that CPU.

Thanks, applied this and 2/2, the latter seems to have been forgotten 
after our initial discussion about it.

-- 
Jens Axboe

Re: [PATCH 2/2] null_blk: restart request processing on completion handler

[Akinobu Mita]

2015-06-02 8:35 GMT+09:00 Akinobu Mita <akinobu.mita@gmail.com>:
> When irqmode=2 (IRQ completion handler is timer) and queue_mode=1
> (Block interface to use is rq), the completion handler should restart
> request handling for any pending requests on a queue because request
> processing stops when the number of commands are queued more than
> hw_queue_depth (null_rq_prep_fn returns BLKPREP_DEFER).
>
> Without this change, the following command cannot finish.
>
>         # modprobe null_blk irqmode=2 queue_mode=1 hw_queue_depth=1
>         # fio --name=t --rw=read --size=1g --direct=1 \
>           --ioengine=libaio --iodepth=64 --filename=/dev/nullb0
>
> Signed-off-by: Akinobu Mita <akinobu.mita@gmail.com>
> Cc: Jens Axboe <axboe@fb.com>
> ---
>  drivers/block/null_blk.c | 12 ++++++++++++
>  1 file changed, 12 insertions(+)
>
> diff --git a/drivers/block/null_blk.c b/drivers/block/null_blk.c
> index 6f0a58e..6f9b753 100644
> --- a/drivers/block/null_blk.c
> +++ b/drivers/block/null_blk.c
> @@ -243,6 +243,17 @@ static enum hrtimer_restart null_cmd_timer_expired(struct hrtimer *timer)
>                         cmd = container_of(entry, struct nullb_cmd, ll_list);
>                         entry = entry->next;
>                         end_cmd(cmd);
> +
> +                       if (cmd->rq) {
> +                               struct request_queue *q = cmd->rq->q;
> +
> +                               if (!q->mq_ops && blk_queue_stopped(q)) {
> +                                       spin_lock(q->queue_lock);
> +                                       if (blk_queue_stopped(q))
> +                                               blk_start_queue(q);

Should 'blk_start_queue(q)' be:

                queue_flag_clear(QUEUE_FLAG_STOPPED, q);
                blk_run_queue_async(q);
?

Otherwise, all pending requests on a queue are completed by this timer
handler execution without needing next timer expiry.

> +                                       spin_unlock(q->queue_lock);
> +                               }
> +                       }
>                 } while (entry);
>         }
>
> @@ -334,6 +345,7 @@ static int null_rq_prep_fn(struct request_queue *q, struct request *req)
>                 req->special = cmd;
>                 return BLKPREP_OK;
>         }
> +       blk_stop_queue(q);
>
>         return BLKPREP_DEFER;
>  }
> --
> 1.9.1
>