public inbox for linux-block@vger.kernel.org
 help / color / mirror / Atom feed
From: Ming Lei <ming.lei@redhat.com>
To: Jens Axboe <axboe@fb.com>,
	linux-block@vger.kernel.org,
	Christoph Hellwig <hch@infradead.org>
Cc: Bart Van Assche <bart.vanassche@sandisk.com>,
	Sagi Grimberg <sagi@grimberg.me>, Ming Lei <ming.lei@redhat.com>
Subject: [PATCH 4/6] blk-mq: use EWMA to estimate congestion threshold
Date: Wed, 12 Jul 2017 02:21:01 +0800	[thread overview]
Message-ID: <20170711182103.11461-5-ming.lei@redhat.com> (raw)
In-Reply-To: <20170711182103.11461-1-ming.lei@redhat.com>

When .queue_rq() returns BLK_STS_RESOURCE(BUSY), we can
consider that there is congestion in either low level
driver or hardware.

This patch uses EWMA to estimate this congestion threshold,
then this threshold can be used to detect/avoid congestion.

Signed-off-by: Ming Lei <ming.lei@redhat.com>
---
 block/blk-mq.c         | 14 ++++++++++++++
 block/blk-mq.h         |  9 +++++++++
 include/linux/blk-mq.h |  2 ++
 3 files changed, 25 insertions(+)

diff --git a/block/blk-mq.c b/block/blk-mq.c
index 6e0fc80aa151..da50c187c508 100644
--- a/block/blk-mq.c
+++ b/block/blk-mq.c
@@ -976,6 +976,18 @@ static bool blk_mq_dispatch_wait_add(struct blk_mq_hw_ctx *hctx)
 	return true;
 }
 
+static void blk_mq_update_req_dispatch_busy(struct blk_mq_hw_ctx *hctx)
+{
+	struct sbitmap_queue *sbq;
+	unsigned depth;
+
+	sbq = &hctx->tags->bitmap_tags;
+	depth = sbitmap_weight(&sbq->sb);
+
+	/* use EWMA to estimate a threshold for detecting congestion */
+	ewma_add(hctx->avg_busy_threshold, depth, 8, 0);
+}
+
 bool blk_mq_dispatch_rq_list(struct request_queue *q, struct list_head *list)
 {
 	struct blk_mq_hw_ctx *hctx;
@@ -1064,6 +1076,7 @@ bool blk_mq_dispatch_rq_list(struct request_queue *q, struct list_head *list)
 
 		spin_lock(&hctx->lock);
 		list_splice_init(list, &hctx->dispatch);
+		blk_mq_update_req_dispatch_busy(hctx);
 		spin_unlock(&hctx->lock);
 
 		/*
@@ -1468,6 +1481,7 @@ static void blk_mq_direct_dispatch(struct blk_mq_hw_ctx *hctx,
 {
 	spin_lock(&hctx->lock);
 	list_add(&rq->queuelist, &hctx->dispatch);
+	blk_mq_update_req_dispatch_busy(hctx);
 	spin_unlock(&hctx->lock);
 
 	blk_mq_run_hw_queue(hctx, false);
diff --git a/block/blk-mq.h b/block/blk-mq.h
index 60b01c0309bc..c4516d2a2d2c 100644
--- a/block/blk-mq.h
+++ b/block/blk-mq.h
@@ -133,4 +133,13 @@ static inline bool blk_mq_hw_queue_mapped(struct blk_mq_hw_ctx *hctx)
 	return hctx->nr_ctx && hctx->tags;
 }
 
+/* borrowed from bcache */
+#define ewma_add(ewma, val, weight, factor)                             \
+({                                                                      \
+        (ewma) *= (weight) - 1;                                         \
+        (ewma) += (val) << factor;                                      \
+        (ewma) /= (weight);                                             \
+        (ewma) >> factor;                                               \
+})
+
 #endif
diff --git a/include/linux/blk-mq.h b/include/linux/blk-mq.h
index 14542308d25b..8694fb39cd80 100644
--- a/include/linux/blk-mq.h
+++ b/include/linux/blk-mq.h
@@ -22,6 +22,8 @@ struct blk_mq_hw_ctx {
 
 	unsigned long		flags;		/* BLK_MQ_F_* flags */
 
+	unsigned long		avg_busy_threshold;
+
 	void			*sched_data;
 	struct request_queue	*queue;
 	struct blk_flush_queue	*fq;
-- 
2.9.4

  parent reply	other threads:[~2017-07-11 18:22 UTC|newest]

Thread overview: 32+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2017-07-11 18:20 [PATCH 0/6] blk-mq: introduce congestion control Ming Lei
2017-07-11 18:20 ` [PATCH 1/6] xen-blkfront: avoid to use start/stop queue Ming Lei
2017-07-11 18:41   ` Konrad Rzeszutek Wilk
2017-07-12  2:52     ` Ming Lei
2017-07-11 18:41   ` Bart Van Assche
2017-07-12  2:59     ` Ming Lei
2017-07-12  3:05     ` Ming Lei
2017-07-11 21:24   ` Roger Pau Monné
2017-07-12  3:12     ` Ming Lei
2017-07-11 18:20 ` [PATCH 2/6] SCSI: use blk_mq_run_hw_queues() in scsi_kick_queue() Ming Lei
2017-07-11 19:57   ` Bart Van Assche
2017-07-12  3:15     ` Ming Lei
2017-07-12 15:12       ` Bart Van Assche
2017-07-13 10:23         ` Ming Lei
2017-07-13 17:44           ` Bart Van Assche
2017-07-11 18:21 ` [PATCH 3/6] blk-mq: send the request to dispatch list if direct issue returns busy Ming Lei
2017-07-11 20:18   ` Bart Van Assche
2017-07-12  3:45     ` Ming Lei
2017-07-11 18:21 ` Ming Lei [this message]
2017-07-11 18:25   ` [PATCH 4/6] blk-mq: use EWMA to estimate congestion threshold Jens Axboe
2017-07-12  2:30     ` Ming Lei
2017-07-12 15:39       ` Bart Van Assche
2017-07-13 10:43         ` Ming Lei
2017-07-13 14:56           ` Bart Van Assche
2017-07-13 15:32             ` Ming Lei
2017-07-13 17:35               ` Bart Van Assche
2017-07-11 18:39   ` Jens Axboe
2017-07-12  3:20     ` Ming Lei
2017-07-11 21:02   ` Bart Van Assche
2017-07-12  3:43     ` Ming Lei
2017-07-11 18:21 ` [PATCH 5/6] blk-mq: introduce basic congestion control Ming Lei
2017-07-11 18:21 ` [PATCH 6/6] blk-mq: unexport APIs for start/stop queues Ming Lei

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=20170711182103.11461-5-ming.lei@redhat.com \
    --to=ming.lei@redhat.com \
    --cc=axboe@fb.com \
    --cc=bart.vanassche@sandisk.com \
    --cc=hch@infradead.org \
    --cc=linux-block@vger.kernel.org \
    --cc=sagi@grimberg.me \
    /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
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox