[RFC PATCH] blk-throtl: Introduce sync queue for write ios

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



From: Jinke Han <hanjinke.666@xxxxxxxxxxxxx>

Now we don't distinguish sync write ios from normal buffer write ios
in blk-throtl. A bio with REQ_SYNC tagged always mean it will be wait
until write completion soon after it submit. So it's reasonable for sync
io to complete as soon as possible.

In our test, fio writes a 100g file in sequential 4k blocksize in
a container with low bps limit configured (wbps=10M). More than 1200
ios were throttled in blk-throtl queue and the avarage throtle time
of each io is 140s. At the same time, the operation of saving a small
file by vim will be blocked amolst 140s. As a fsync will be send by vim,
the sync ios of fsync will be blocked by a huge amount of buffer write
ios ahead. This is also a priority inversion problem within one cgroup.
In the database scene, things got really bad with blk-throtle enabled
as fsync is called very often.

This patch introduces a independent sync queue for write ios and gives
a huge priority to sync write ios. I think it's a nice respond to the
semantics of REQ_SYNC. Bios with REQ_META and REQ_PRIO gains the same
priority as they are important to fs. This may avoid some potential
priority inversion problems.

Signed-off-by: Jinke Han <hanjinke.666@xxxxxxxxxxxxx>
---
 block/blk-throttle.c | 83 ++++++++++++++++++++++++++++++++++++++++----
 block/blk-throttle.h |  9 ++++-
 2 files changed, 85 insertions(+), 7 deletions(-)

diff --git a/block/blk-throttle.c b/block/blk-throttle.c
index 847721dc2b2b..5d290fb95f5a 100644
--- a/block/blk-throttle.c
+++ b/block/blk-throttle.c
@@ -21,6 +21,9 @@
 /* Total max dispatch from all groups in one round */
 #define THROTL_QUANTUM 32
 
+/* For write ios, dispatch 4 sync ios and 1 normal io in one loop */
+#define THROTL_SYNC_FACTOR 4
+
 /* Throttling is performed over a slice and after that slice is renewed */
 #define DFL_THROTL_SLICE_HD (HZ / 10)
 #define DFL_THROTL_SLICE_SSD (HZ / 50)
@@ -241,11 +244,27 @@ static inline unsigned int throtl_bio_data_size(struct bio *bio)
 	return bio->bi_iter.bi_size;
 }
 
-static void throtl_qnode_init(struct throtl_qnode *qn, struct throtl_grp *tg)
+static void throtl_qnode_init(struct throtl_qnode *qn, struct throtl_grp *tg,
+			      bool rw)
 {
 	INIT_LIST_HEAD(&qn->node);
 	bio_list_init(&qn->bios);
+	bio_list_init(&qn->sync_bios);
 	qn->tg = tg;
+	qn->dispatch_sync_cnt = (rw == READ) ? UINT_MAX : 0;
+}
+
+#define BLK_THROTL_SYNC(bio) (bio->bi_opf & (REQ_SYNC | REQ_META | REQ_PRIO))
+
+static inline void throtl_qnode_add_bio_list(struct throtl_qnode *qn,
+					     struct bio *bio)
+{
+	bool rw = bio_data_dir(bio);
+
+	if ((rw == WRITE) && (BLK_THROTL_SYNC(bio)))
+		bio_list_add(&qn->sync_bios, bio);
+	else
+		bio_list_add(&qn->bios, bio);
 }
 
 /**
@@ -261,13 +280,34 @@ static void throtl_qnode_init(struct throtl_qnode *qn, struct throtl_grp *tg)
 static void throtl_qnode_add_bio(struct bio *bio, struct throtl_qnode *qn,
 				 struct list_head *queued)
 {
-	bio_list_add(&qn->bios, bio);
+	throtl_qnode_add_bio_list(qn, bio);
 	if (list_empty(&qn->node)) {
 		list_add_tail(&qn->node, queued);
 		blkg_get(tg_to_blkg(qn->tg));
 	}
 }
 
+static inline struct bio *throtl_qnode_bio_peek(struct throtl_qnode *qn)
+{
+	struct bio *bio1, *bio2;
+
+	/* qn for read ios */
+	if (qn->dispatch_sync_cnt == UINT_MAX)
+		return bio_list_peek(&qn->bios);
+
+	/* qn for write ios */
+	bio1 = bio_list_peek(&qn->bios);
+	bio2 = bio_list_peek(&qn->sync_bios);
+
+	if (bio1 && bio2) {
+		if (qn->dispatch_sync_cnt == THROTL_SYNC_FACTOR)
+			return bio1;
+		return bio2;
+	}
+
+	return bio1 ?: bio2;
+}
+
 /**
  * throtl_peek_queued - peek the first bio on a qnode list
  * @queued: the qnode list to peek
@@ -281,11 +321,42 @@ static struct bio *throtl_peek_queued(struct list_head *queued)
 		return NULL;
 
 	qn = list_first_entry(queued, struct throtl_qnode, node);
-	bio = bio_list_peek(&qn->bios);
+	bio = throtl_qnode_bio_peek(qn);
 	WARN_ON_ONCE(!bio);
 	return bio;
 }
 
+static inline struct bio *throtl_qnode_bio_pop(struct throtl_qnode *qn)
+{
+	struct bio *bio;
+
+	/* qn for read ios */
+	if (qn->dispatch_sync_cnt == UINT_MAX)
+		return bio_list_pop(&qn->bios);
+
+	/* try to dispatch sync io */
+	if (qn->dispatch_sync_cnt < THROTL_SYNC_FACTOR) {
+		bio = bio_list_pop(&qn->sync_bios);
+		if (bio) {
+			qn->dispatch_sync_cnt++;
+			return bio;
+		}
+		bio = bio_list_pop(&qn->bios);
+		qn->dispatch_sync_cnt = 0;
+		return bio;
+	}
+
+	/* try to dispatch normal io */
+	bio = bio_list_pop(&qn->bios);
+	if (bio) {
+		qn->dispatch_sync_cnt = 0;
+		return bio;
+	}
+	bio = bio_list_pop(&qn->sync_bios);
+
+	return bio;
+}
+
 /**
  * throtl_pop_queued - pop the first bio form a qnode list
  * @queued: the qnode list to pop a bio from
@@ -310,7 +381,7 @@ static struct bio *throtl_pop_queued(struct list_head *queued,
 		return NULL;
 
 	qn = list_first_entry(queued, struct throtl_qnode, node);
-	bio = bio_list_pop(&qn->bios);
+	bio = throtl_qnode_bio_pop(qn);
 	WARN_ON_ONCE(!bio);
 
 	if (bio_list_empty(&qn->bios)) {
@@ -355,8 +426,8 @@ static struct blkg_policy_data *throtl_pd_alloc(gfp_t gfp,
 	throtl_service_queue_init(&tg->service_queue);
 
 	for (rw = READ; rw <= WRITE; rw++) {
-		throtl_qnode_init(&tg->qnode_on_self[rw], tg);
-		throtl_qnode_init(&tg->qnode_on_parent[rw], tg);
+		throtl_qnode_init(&tg->qnode_on_self[rw], tg, rw);
+		throtl_qnode_init(&tg->qnode_on_parent[rw], tg, rw);
 	}
 
 	RB_CLEAR_NODE(&tg->rb_node);
diff --git a/block/blk-throttle.h b/block/blk-throttle.h
index ef4b7a4de987..6a857c9420f9 100644
--- a/block/blk-throttle.h
+++ b/block/blk-throttle.h
@@ -28,8 +28,15 @@
  */
 struct throtl_qnode {
 	struct list_head	node;		/* service_queue->queued[] */
-	struct bio_list		bios;		/* queued bios */
+	struct bio_list		bios;		/* queued normal bios */
+	struct bio_list		sync_bios;	/* queued sync write bios */
 	struct throtl_grp	*tg;		/* tg this qnode belongs to */
+
+	/*
+	 * [0, THROTL_SYNC_FACTOR-1] dispatch sync io, THROTL_SYNC_FACTOR
+	 * dispatch normal io
+	 */
+	unsigned int dispatch_sync_cnt;         /* sync io dispatch counter */
 };
 
 struct throtl_service_queue {
-- 
2.20.1




[Index of Archives]     [Linux RAID]     [Linux SCSI]     [Linux ATA RAID]     [IDE]     [Linux Wireless]     [Linux Kernel]     [ATH6KL]     [Linux Bluetooth]     [Linux Netdev]     [Kernel Newbies]     [Security]     [Git]     [Netfilter]     [Bugtraq]     [Yosemite News]     [MIPS Linux]     [ARM Linux]     [Linux Security]     [Device Mapper]

  Powered by Linux