diff options
Diffstat (limited to 'block/bfq-iosched.c')
| -rw-r--r-- | block/bfq-iosched.c | 97 |
1 files changed, 64 insertions, 33 deletions
diff --git a/block/bfq-iosched.c b/block/bfq-iosched.c index 3a27d31fcda6..cd307767a134 100644 --- a/block/bfq-iosched.c +++ b/block/bfq-iosched.c @@ -399,9 +399,9 @@ static struct bfq_io_cq *bfq_bic_lookup(struct bfq_data *bfqd, unsigned long flags; struct bfq_io_cq *icq; - spin_lock_irqsave(q->queue_lock, flags); + spin_lock_irqsave(&q->queue_lock, flags); icq = icq_to_bic(ioc_lookup_icq(ioc, q)); - spin_unlock_irqrestore(q->queue_lock, flags); + spin_unlock_irqrestore(&q->queue_lock, flags); return icq; } @@ -638,7 +638,7 @@ static bool bfq_varied_queue_weights_or_active_groups(struct bfq_data *bfqd) bfqd->queue_weights_tree.rb_node->rb_right) #ifdef CONFIG_BFQ_GROUP_IOSCHED ) || - (bfqd->num_active_groups > 0 + (bfqd->num_groups_with_pending_reqs > 0 #endif ); } @@ -802,7 +802,21 @@ void bfq_weights_tree_remove(struct bfq_data *bfqd, */ break; } - bfqd->num_active_groups--; + + /* + * The decrement of num_groups_with_pending_reqs is + * not performed immediately upon the deactivation of + * entity, but it is delayed to when it also happens + * that the first leaf descendant bfqq of entity gets + * all its pending requests completed. The following + * instructions perform this delayed decrement, if + * needed. See the comments on + * num_groups_with_pending_reqs for details. + */ + if (entity->in_groups_with_pending_reqs) { + entity->in_groups_with_pending_reqs = false; + bfqd->num_groups_with_pending_reqs--; + } } } @@ -3529,27 +3543,44 @@ static bool bfq_better_to_idle(struct bfq_queue *bfqq) * fact, if there are active groups, then, for condition (i) * to become false, it is enough that an active group contains * more active processes or sub-groups than some other active - * group. We address this issue with the following bi-modal - * behavior, implemented in the function + * group. More precisely, for condition (i) to hold because of + * such a group, it is not even necessary that the group is + * (still) active: it is sufficient that, even if the group + * has become inactive, some of its descendant processes still + * have some request already dispatched but still waiting for + * completion. In fact, requests have still to be guaranteed + * their share of the throughput even after being + * dispatched. In this respect, it is easy to show that, if a + * group frequently becomes inactive while still having + * in-flight requests, and if, when this happens, the group is + * not considered in the calculation of whether the scenario + * is asymmetric, then the group may fail to be guaranteed its + * fair share of the throughput (basically because idling may + * not be performed for the descendant processes of the group, + * but it had to be). We address this issue with the + * following bi-modal behavior, implemented in the function * bfq_symmetric_scenario(). * - * If there are active groups, then the scenario is tagged as + * If there are groups with requests waiting for completion + * (as commented above, some of these groups may even be + * already inactive), then the scenario is tagged as * asymmetric, conservatively, without checking any of the * conditions (i) and (ii). So the device is idled for bfqq. * This behavior matches also the fact that groups are created - * exactly if controlling I/O (to preserve bandwidth and - * latency guarantees) is a primary concern. + * exactly if controlling I/O is a primary concern (to + * preserve bandwidth and latency guarantees). * - * On the opposite end, if there are no active groups, then - * only condition (i) is actually controlled, i.e., provided - * that condition (i) holds, idling is not performed, - * regardless of whether condition (ii) holds. In other words, - * only if condition (i) does not hold, then idling is - * allowed, and the device tends to be prevented from queueing - * many requests, possibly of several processes. Since there - * are no active groups, then, to control condition (i) it is - * enough to check whether all active queues have the same - * weight. + * On the opposite end, if there are no groups with requests + * waiting for completion, then only condition (i) is actually + * controlled, i.e., provided that condition (i) holds, idling + * is not performed, regardless of whether condition (ii) + * holds. In other words, only if condition (i) does not hold, + * then idling is allowed, and the device tends to be + * prevented from queueing many requests, possibly of several + * processes. Since there are no groups with requests waiting + * for completion, then, to control condition (i) it is enough + * to check just whether all the queues with requests waiting + * for completion also have the same weight. * * Not checking condition (ii) evidently exposes bfqq to the * risk of getting less throughput than its fair share. @@ -3607,10 +3638,11 @@ static bool bfq_better_to_idle(struct bfq_queue *bfqq) * bfqq is weight-raised is checked explicitly here. More * precisely, the compound condition below takes into account * also the fact that, even if bfqq is being weight-raised, - * the scenario is still symmetric if all active queues happen - * to be weight-raised. Actually, we should be even more - * precise here, and differentiate between interactive weight - * raising and soft real-time weight raising. + * the scenario is still symmetric if all queues with requests + * waiting for completion happen to be + * weight-raised. Actually, we should be even more precise + * here, and differentiate between interactive weight raising + * and soft real-time weight raising. * * As a side note, it is worth considering that the above * device-idling countermeasures may however fail in the @@ -4034,7 +4066,7 @@ static void bfq_update_dispatch_stats(struct request_queue *q, * In addition, the following queue lock guarantees that * bfqq_group(bfqq) exists as well. */ - spin_lock_irq(q->queue_lock); + spin_lock_irq(&q->queue_lock); if (idle_timer_disabled) /* * Since the idle timer has been disabled, @@ -4053,7 +4085,7 @@ static void bfq_update_dispatch_stats(struct request_queue *q, bfqg_stats_set_start_empty_time(bfqg); bfqg_stats_update_io_remove(bfqg, rq->cmd_flags); } - spin_unlock_irq(q->queue_lock); + spin_unlock_irq(&q->queue_lock); } #else static inline void bfq_update_dispatch_stats(struct request_queue *q, @@ -4384,7 +4416,7 @@ static struct bfq_queue *bfq_get_queue(struct bfq_data *bfqd, rcu_read_lock(); - bfqg = bfq_find_set_group(bfqd, bio_blkcg(bio)); + bfqg = bfq_find_set_group(bfqd, __bio_blkcg(bio)); if (!bfqg) { bfqq = &bfqd->oom_bfqq; goto out; @@ -4637,11 +4669,11 @@ static void bfq_update_insert_stats(struct request_queue *q, * In addition, the following queue lock guarantees that * bfqq_group(bfqq) exists as well. */ - spin_lock_irq(q->queue_lock); + spin_lock_irq(&q->queue_lock); bfqg_stats_update_io_add(bfqq_group(bfqq), bfqq, cmd_flags); if (idle_timer_disabled) bfqg_stats_update_idle_time(bfqq_group(bfqq)); - spin_unlock_irq(q->queue_lock); + spin_unlock_irq(&q->queue_lock); } #else static inline void bfq_update_insert_stats(struct request_queue *q, @@ -5382,9 +5414,9 @@ static int bfq_init_queue(struct request_queue *q, struct elevator_type *e) } eq->elevator_data = bfqd; - spin_lock_irq(q->queue_lock); + spin_lock_irq(&q->queue_lock); q->elevator = eq; - spin_unlock_irq(q->queue_lock); + spin_unlock_irq(&q->queue_lock); /* * Our fallback bfqq if bfq_find_alloc_queue() runs into OOM issues. @@ -5417,7 +5449,7 @@ static int bfq_init_queue(struct request_queue *q, struct elevator_type *e) bfqd->idle_slice_timer.function = bfq_idle_slice_timer; bfqd->queue_weights_tree = RB_ROOT; - bfqd->num_active_groups = 0; + bfqd->num_groups_with_pending_reqs = 0; INIT_LIST_HEAD(&bfqd->active_list); INIT_LIST_HEAD(&bfqd->idle_list); @@ -5724,7 +5756,7 @@ static struct elv_fs_entry bfq_attrs[] = { }; static struct elevator_type iosched_bfq_mq = { - .ops.mq = { + .ops = { .limit_depth = bfq_limit_depth, .prepare_request = bfq_prepare_request, .requeue_request = bfq_finish_requeue_request, @@ -5745,7 +5777,6 @@ static struct elevator_type iosched_bfq_mq = { .exit_sched = bfq_exit_queue, }, - .uses_mq = true, .icq_size = sizeof(struct bfq_io_cq), .icq_align = __alignof__(struct bfq_io_cq), .elevator_attrs = bfq_attrs, |