linux-block.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Damien Le Moal <Damien.LeMoal@wdc.com>
To: Bart Van Assche <bvanassche@acm.org>, Jens Axboe <axboe@kernel.dk>
Cc: "linux-block@vger.kernel.org" <linux-block@vger.kernel.org>,
	Christoph Hellwig <hch@lst.de>, Jaegeuk Kim <jaegeuk@kernel.org>,
	Hannes Reinecke <hare@suse.de>, Ming Lei <ming.lei@redhat.com>,
	Johannes Thumshirn <Johannes.Thumshirn@wdc.com>,
	Himanshu Madhani <himanshu.madhani@oracle.com>
Subject: Re: [PATCH 14/14] block/mq-deadline: Prioritize high-priority requests
Date: Wed, 9 Jun 2021 05:10:53 +0000	[thread overview]
Message-ID: <DM6PR04MB7081853B525156C4F4FF0C0AE7369@DM6PR04MB7081.namprd04.prod.outlook.com> (raw)
In-Reply-To: <20210608230703.19510-15-bvanassche@acm.org>

On 2021/06/09 8:07, Bart Van Assche wrote:
> While one or more requests with a certain I/O priority are pending, do not
> dispatch lower priority requests. Dispatch lower priority requests anyway
> after the "aging" time has expired.
> 
> This patch has been tested as follows:
> 
> modprobe scsi_debug ndelay=1000000 max_queue=16 &&
> sd='' &&
> while [ -z "$sd" ]; do
>   sd=/dev/$(basename /sys/bus/pseudo/drivers/scsi_debug/adapter*/host*/target*/*/block/*)
> done &&
> echo $((100*1000)) > /sys/block/$sd/queue/iosched/aging_expire &&
> cd /sys/fs/cgroup/blkio/ &&
> echo $$ >cgroup.procs &&
> echo 2 >blkio.prio.class &&
> mkdir -p hipri &&
> cd hipri &&
> echo 1 >blkio.prio.class &&
> { max-iops -a1 -d32 -j1 -e mq-deadline $sd >& ~/low-pri.txt & } &&
> echo $$ >cgroup.procs &&
> max-iops -a1 -d32 -j1 -e mq-deadline $sd >& ~/hi-pri.txt
> 
> Result:
> * 11000 IOPS for the high-priority job
> *   400 IOPS for the low-priority job
> 
> If the aging expiry time is changed from 100s into 0, the IOPS results change
> into 6712 and 6796 IOPS.
> 
> The max-iops script is a script that runs fio with the following arguments:
> --bs=4K --gtod_reduce=1 --ioengine=libaio --ioscheduler=${arg_e} --runtime=60
> --norandommap --rw=read --thread --buffered=0 --numjobs=${arg_j}
> --iodepth=${arg_d} --iodepth_batch_submit=${arg_a}
> --iodepth_batch_complete=$((arg_d / 2)) --name=${positional_argument_1}
> --filename=${positional_argument_1}
> 
> Cc: Damien Le Moal <damien.lemoal@wdc.com>
> Cc: Hannes Reinecke <hare@suse.de>
> Cc: Christoph Hellwig <hch@lst.de>
> Cc: Ming Lei <ming.lei@redhat.com>
> Cc: Johannes Thumshirn <johannes.thumshirn@wdc.com>
> Cc: Himanshu Madhani <himanshu.madhani@oracle.com>
> Signed-off-by: Bart Van Assche <bvanassche@acm.org>
> ---
>  block/mq-deadline-main.c | 42 +++++++++++++++++++++++++++++++++++-----
>  1 file changed, 37 insertions(+), 5 deletions(-)
> 
> diff --git a/block/mq-deadline-main.c b/block/mq-deadline-main.c
> index 1b2b6c1de5b8..cecdb475a610 100644
> --- a/block/mq-deadline-main.c
> +++ b/block/mq-deadline-main.c
> @@ -32,6 +32,11 @@
>   */
>  static const int read_expire = HZ / 2;  /* max time before a read is submitted. */
>  static const int write_expire = 5 * HZ; /* ditto for writes, these limits are SOFT! */
> +/*
> + * Time after which to dispatch lower priority requests even if higher
> + * priority requests are pending.
> + */
> +static const int aging_expire = 10 * HZ;
>  static const int writes_starved = 2;    /* max times reads can starve a write */
>  static const int fifo_batch = 16;       /* # of sequential requests treated as one
>  				     by the above parameters. For throughput. */
> @@ -90,6 +95,7 @@ struct deadline_data {
>  	int writes_starved;
>  	int front_merges;
>  	u32 async_depth;
> +	int aging_expire;
>  
>  	spinlock_t lock;
>  	spinlock_t zone_lock;
> @@ -363,10 +369,10 @@ deadline_next_request(struct deadline_data *dd, enum dd_prio prio,
>  
>  /*
>   * deadline_dispatch_requests selects the best request according to
> - * read/write expire, fifo_batch, etc
> + * read/write expire, fifo_batch, etc and with a start time <= @latest.
>   */
>  static struct request *__dd_dispatch_request(struct deadline_data *dd,
> -					     enum dd_prio prio)
> +					enum dd_prio prio, u64 latest_start_ns)
>  {
>  	struct request *rq, *next_rq;
>  	enum dd_data_dir data_dir;
> @@ -378,6 +384,8 @@ static struct request *__dd_dispatch_request(struct deadline_data *dd,
>  	if (!list_empty(&dd->dispatch[prio])) {
>  		rq = list_first_entry(&dd->dispatch[prio], struct request,
>  				      queuelist);
> +		if (rq->start_time_ns > latest_start_ns)
> +			return NULL;
>  		list_del_init(&rq->queuelist);
>  		goto done;
>  	}
> @@ -457,6 +465,8 @@ static struct request *__dd_dispatch_request(struct deadline_data *dd,
>  	dd->batching = 0;
>  
>  dispatch_request:
> +	if (rq->start_time_ns > latest_start_ns)
> +		return NULL;
>  	/*
>  	 * rq is the selected appropriate request.
>  	 */
> @@ -493,15 +503,32 @@ static struct request *__dd_dispatch_request(struct deadline_data *dd,
>  static struct request *dd_dispatch_request(struct blk_mq_hw_ctx *hctx)
>  {
>  	struct deadline_data *dd = hctx->queue->elevator->elevator_data;
> -	struct request *rq;
> +	const u64 now_ns = ktime_get_ns();
> +	struct request *rq = NULL;
>  	enum dd_prio prio;
>  
>  	spin_lock(&dd->lock);
> -	for (prio = 0; prio <= DD_PRIO_MAX; prio++) {
> -		rq = __dd_dispatch_request(dd, prio);
> +	/*
> +	 * Start with dispatching requests whose deadline expired more than
> +	 * aging_expire jiffies ago.
> +	 */
> +	for (prio = DD_BE_PRIO; prio <= DD_PRIO_MAX; prio++) {
> +		rq = __dd_dispatch_request(dd, prio, now_ns -
> +					   jiffies_to_nsecs(dd->aging_expire));
>  		if (rq)
> +			goto unlock;
> +	}
> +	/*
> +	 * Next, dispatch requests in priority order. Ignore lower priority
> +	 * requests if any higher priority requests are pending.
> +	 */
> +	for (prio = 0; prio <= DD_PRIO_MAX; prio++) {
> +		rq = __dd_dispatch_request(dd, prio, now_ns);
> +		if (rq || dd_queued(dd, prio))
>  			break;
>  	}
> +
> +unlock:
>  	spin_unlock(&dd->lock);
>  
>  	return rq;
> @@ -607,6 +634,7 @@ static int dd_init_sched(struct request_queue *q, struct elevator_type *e)
>  	dd->writes_starved = writes_starved;
>  	dd->front_merges = 1;
>  	dd->fifo_batch = fifo_batch;
> +	dd->aging_expire = aging_expire;
>  	spin_lock_init(&dd->lock);
>  	spin_lock_init(&dd->zone_lock);
>  
> @@ -725,6 +753,7 @@ static void dd_insert_request(struct blk_mq_hw_ctx *hctx, struct request *rq,
>  	trace_block_rq_insert(rq);
>  
>  	if (at_head) {
> +		rq->fifo_time = jiffies;
>  		list_add(&rq->queuelist, &dd->dispatch[prio]);
>  	} else {
>  		deadline_add_rq_rb(dd, rq);
> @@ -841,6 +870,7 @@ static ssize_t __FUNC(struct elevator_queue *e, char *page)		\
>  #define SHOW_JIFFIES(__FUNC, __VAR) SHOW_INT(__FUNC, jiffies_to_msecs(__VAR))
>  SHOW_JIFFIES(deadline_read_expire_show, dd->fifo_expire[DD_READ]);
>  SHOW_JIFFIES(deadline_write_expire_show, dd->fifo_expire[DD_WRITE]);
> +SHOW_JIFFIES(deadline_aging_expire_show, dd->aging_expire);
>  SHOW_INT(deadline_writes_starved_show, dd->writes_starved);
>  SHOW_INT(deadline_front_merges_show, dd->front_merges);
>  SHOW_INT(deadline_fifo_batch_show, dd->fifo_batch);
> @@ -869,6 +899,7 @@ static ssize_t __FUNC(struct elevator_queue *e, const char *page, size_t count)
>  	STORE_FUNCTION(__FUNC, __PTR, MIN, MAX, msecs_to_jiffies)
>  STORE_JIFFIES(deadline_read_expire_store, &dd->fifo_expire[DD_READ], 0, INT_MAX);
>  STORE_JIFFIES(deadline_write_expire_store, &dd->fifo_expire[DD_WRITE], 0, INT_MAX);
> +STORE_JIFFIES(deadline_aging_expire_store, &dd->aging_expire, 0, INT_MAX);
>  STORE_INT(deadline_writes_starved_store, &dd->writes_starved, INT_MIN, INT_MAX);
>  STORE_INT(deadline_front_merges_store, &dd->front_merges, 0, 1);
>  STORE_INT(deadline_fifo_batch_store, &dd->fifo_batch, 0, INT_MAX);
> @@ -885,6 +916,7 @@ static struct elv_fs_entry deadline_attrs[] = {
>  	DD_ATTR(writes_starved),
>  	DD_ATTR(front_merges),
>  	DD_ATTR(fifo_batch),
> +	DD_ATTR(aging_expire),
>  	__ATTR_NULL
>  };
>  
> 

Looks good.

Reviewed-by: Damien Le Moal <damien.lemoal@wdc.com>


-- 
Damien Le Moal
Western Digital Research

  reply	other threads:[~2021-06-09  5:10 UTC|newest]

Thread overview: 45+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-06-08 23:06 [PATCH 00/14] Improve I/O priority support Bart Van Assche
2021-06-08 23:06 ` [PATCH 01/14] block/Kconfig: Make the BLK_WBT and BLK_WBT_MQ entries consecutive Bart Van Assche
2021-06-09  4:14   ` Damien Le Moal
2021-06-09  7:03   ` Johannes Thumshirn
2021-06-10  6:02   ` Hannes Reinecke
2021-06-08 23:06 ` [PATCH 02/14] block/blk-cgroup: Swap the blk_throtl_init() and blk_iolatency_init() calls Bart Van Assche
2021-06-09  4:19   ` Damien Le Moal
2021-06-09  7:05   ` Johannes Thumshirn
2021-06-10  6:02   ` Hannes Reinecke
2021-06-08 23:06 ` [PATCH 03/14] block/blk-rq-qos: Move a function from a header file into a C file Bart Van Assche
2021-06-09  4:22   ` Damien Le Moal
2021-06-10  6:03   ` Hannes Reinecke
2021-06-08 23:06 ` [PATCH 04/14] block: Introduce the ioprio rq-qos policy Bart Van Assche
2021-06-09  4:40   ` Damien Le Moal
2021-06-09 16:48     ` Bart Van Assche
2021-06-10  0:29       ` Damien Le Moal
2021-06-10  6:20   ` Hannes Reinecke
2021-06-10 17:14     ` Bart Van Assche
2021-06-08 23:06 ` [PATCH 05/14] block/mq-deadline: Add several comments Bart Van Assche
2021-06-10  6:21   ` Hannes Reinecke
2021-06-08 23:06 ` [PATCH 06/14] block/mq-deadline: Add two lockdep_assert_held() statements Bart Van Assche
2021-06-10  6:21   ` Hannes Reinecke
2021-06-08 23:06 ` [PATCH 07/14] block/mq-deadline: Remove two local variables Bart Van Assche
2021-06-10  6:22   ` Hannes Reinecke
2021-06-08 23:06 ` [PATCH 08/14] block/mq-deadline: Rename dd_init_queue() and dd_exit_queue() Bart Van Assche
2021-06-08 23:06 ` [PATCH 09/14] block/mq-deadline: Improve compile-time argument checking Bart Van Assche
2021-06-08 23:06 ` [PATCH 10/14] block/mq-deadline: Improve the sysfs show and store macros Bart Van Assche
2021-06-09  4:46   ` Damien Le Moal
2021-06-09 16:54     ` Bart Van Assche
2021-06-10  0:31       ` Damien Le Moal
2021-06-09  7:11   ` Johannes Thumshirn
2021-06-09 16:59     ` Bart Van Assche
2021-06-10  6:23   ` Hannes Reinecke
2021-06-08 23:07 ` [PATCH 11/14] block/mq-deadline: Reserve 25% of scheduler tags for synchronous requests Bart Van Assche
2021-06-10  6:26   ` Hannes Reinecke
2021-06-08 23:07 ` [PATCH 12/14] block/mq-deadline: Add I/O priority support Bart Van Assche
2021-06-09  5:03   ` Damien Le Moal
2021-06-09 17:25     ` Bart Van Assche
2021-06-10  0:39       ` Damien Le Moal
2021-06-10  6:35   ` Hannes Reinecke
2021-06-08 23:07 ` [PATCH 13/14] block/mq-deadline: Add cgroup support Bart Van Assche
2021-06-10  6:37   ` Hannes Reinecke
2021-06-08 23:07 ` [PATCH 14/14] block/mq-deadline: Prioritize high-priority requests Bart Van Assche
2021-06-09  5:10   ` Damien Le Moal [this message]
2021-06-10  6:43   ` Hannes Reinecke

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=DM6PR04MB7081853B525156C4F4FF0C0AE7369@DM6PR04MB7081.namprd04.prod.outlook.com \
    --to=damien.lemoal@wdc.com \
    --cc=Johannes.Thumshirn@wdc.com \
    --cc=axboe@kernel.dk \
    --cc=bvanassche@acm.org \
    --cc=hare@suse.de \
    --cc=hch@lst.de \
    --cc=himanshu.madhani@oracle.com \
    --cc=jaegeuk@kernel.org \
    --cc=linux-block@vger.kernel.org \
    --cc=ming.lei@redhat.com \
    --subject='Re: [PATCH 14/14] block/mq-deadline: Prioritize high-priority requests' \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).