cfq-iosched: fix sequential write regression (a9938006) · Commits · 戴 / test

We have a 10-15% performance regression for sequential writes on TCQ/NCQ enabled drives in 2.6.21-rcX after the CFQ update went in. It has been reported by Valerie Clement <valerie.clement@bull.net> and the Intel testing folks. The regression is because of CFQ's now more aggressive queue control, limiting the depth available to the device. This patches fixes that regression by allowing a greater depth when only one queue is busy. It has been tested to not impact sync-vs-async workloads too much - we still do a lot better than 2.6.20. Signed-off-by:

Jens Axboe <jens.axboe@oracle.com> Signed-off-by:

Linus Torvalds <torvalds@linux-foundation.org>

block/cfq-iosched.c

+19 −15

Original line number	Diff line number	Diff line
		@@ -986,9 +986,9 @@ __cfq_dispatch_requests(struct cfq_data cfqd, struct cfq_queue cfqq,
		* expire an async queue immediately if it has used up its slice. idle
		* queue always expire after 1 dispatch round.
		*/
		if ((!cfq_cfqq_sync(cfqq) &&
		if (cfqd->busy_queues > 1 && ((!cfq_cfqq_sync(cfqq) &&
		cfqd->dispatch_slice >= cfq_prio_to_maxrq(cfqd, cfqq)) \|\|
		cfq_class_idle(cfqq)) {
		cfq_class_idle(cfqq))) {
		cfqq->slice_end = jiffies + 1;
		cfq_slice_expired(cfqd, 0, 0);
		}
		@@ -1051,6 +1051,7 @@ cfq_dispatch_requests(request_queue_t *q, int force)
		while ((cfqq = cfq_select_queue(cfqd)) != NULL) {
		int max_dispatch;

		if (cfqd->busy_queues > 1) {
		/*
		* Don't repeat dispatch from the previous queue.
		*/
		@@ -1064,6 +1065,7 @@ cfq_dispatch_requests(request_queue_t *q, int force)
		*/
		if (cfqd->rq_in_driver && cfq_cfqq_idle_window(cfqq))
		break;
		}

		cfq_clear_cfqq_must_dispatch(cfqq);
		cfq_clear_cfqq_wait_request(cfqq);
		@@ -1370,7 +1372,9 @@ retry:
		atomic_set(&cfqq->ref, 0);
		cfqq->cfqd = cfqd;

		if (key != CFQ_KEY_ASYNC)
		cfq_mark_cfqq_idle_window(cfqq);

		cfq_mark_cfqq_prio_changed(cfqq);
		cfq_mark_cfqq_queue_new(cfqq);
		cfq_init_prio_data(cfqq);

Admin message