]> err.no Git - linux-2.6/commitdiff
[PATCH] cfq-iosched: Don't set the queue batching limits
authorJens Axboe <axboe@suse.de>
Tue, 13 Jun 2006 06:08:38 +0000 (08:08 +0200)
committerJens Axboe <axboe@nelson.home.kernel.dk>
Fri, 23 Jun 2006 15:10:38 +0000 (17:10 +0200)
We cannot update them if the user changes nr_requests, so don't
set it in the first place. The gains are pretty questionable as
well. The batching loss has been shown to decrease throughput.

Signed-off-by: Jens Axboe <axboe@suse.de>
block/cfq-iosched.c

index e2e6ad0a158e2201d81af14f90008a7b538d5ed7..c88f161d3fb3192d18fdc109e23e6b774a8fcb39 100644 (file)
@@ -123,8 +123,6 @@ struct cfq_data {
         */
        struct hlist_head *crq_hash;
 
-       unsigned int max_queued;
-
        mempool_t *crq_pool;
 
        int rq_in_driver;
@@ -1910,7 +1908,6 @@ static inline int
 __cfq_may_queue(struct cfq_data *cfqd, struct cfq_queue *cfqq,
                struct task_struct *task, int rw)
 {
-#if 1
        if ((cfq_cfqq_wait_request(cfqq) || cfq_cfqq_must_alloc(cfqq)) &&
            !cfq_cfqq_must_alloc_slice(cfqq)) {
                cfq_mark_cfqq_must_alloc_slice(cfqq);
@@ -1918,39 +1915,6 @@ __cfq_may_queue(struct cfq_data *cfqd, struct cfq_queue *cfqq,
        }
 
        return ELV_MQUEUE_MAY;
-#else
-       if (!cfqq || task->flags & PF_MEMALLOC)
-               return ELV_MQUEUE_MAY;
-       if (!cfqq->allocated[rw] || cfq_cfqq_must_alloc(cfqq)) {
-               if (cfq_cfqq_wait_request(cfqq))
-                       return ELV_MQUEUE_MUST;
-
-               /*
-                * only allow 1 ELV_MQUEUE_MUST per slice, otherwise we
-                * can quickly flood the queue with writes from a single task
-                */
-               if (rw == READ || !cfq_cfqq_must_alloc_slice(cfqq)) {
-                       cfq_mark_cfqq_must_alloc_slice(cfqq);
-                       return ELV_MQUEUE_MUST;
-               }
-
-               return ELV_MQUEUE_MAY;
-       }
-       if (cfq_class_idle(cfqq))
-               return ELV_MQUEUE_NO;
-       if (cfqq->allocated[rw] >= cfqd->max_queued) {
-               struct io_context *ioc = get_io_context(GFP_ATOMIC);
-               int ret = ELV_MQUEUE_NO;
-
-               if (ioc && ioc->nr_batch_requests)
-                       ret = ELV_MQUEUE_MAY;
-
-               put_io_context(ioc);
-               return ret;
-       }
-
-       return ELV_MQUEUE_MAY;
-#endif
 }
 
 static int cfq_may_queue(request_queue_t *q, int rw, struct bio *bio)
@@ -1979,16 +1943,13 @@ static int cfq_may_queue(request_queue_t *q, int rw, struct bio *bio)
 static void cfq_check_waiters(request_queue_t *q, struct cfq_queue *cfqq)
 {
        struct cfq_data *cfqd = q->elevator->elevator_data;
-       struct request_list *rl = &q->rq;
 
-       if (cfqq->allocated[READ] <= cfqd->max_queued || cfqd->rq_starved) {
+       if (unlikely(cfqd->rq_starved)) {
+               struct request_list *rl = &q->rq;
+
                smp_mb();
                if (waitqueue_active(&rl->wait[READ]))
                        wake_up(&rl->wait[READ]);
-       }
-
-       if (cfqq->allocated[WRITE] <= cfqd->max_queued || cfqd->rq_starved) {
-               smp_mb();
                if (waitqueue_active(&rl->wait[WRITE]))
                        wake_up(&rl->wait[WRITE]);
        }
@@ -2278,9 +2239,6 @@ static void *cfq_init_queue(request_queue_t *q, elevator_t *e)
 
        cfqd->queue = q;
 
-       cfqd->max_queued = q->nr_requests / 4;
-       q->nr_batching = cfq_queued;
-
        init_timer(&cfqd->idle_slice_timer);
        cfqd->idle_slice_timer.function = cfq_idle_slice_timer;
        cfqd->idle_slice_timer.data = (unsigned long) cfqd;