diff options
Diffstat (limited to 'block/bfq-iosched.c')
| -rw-r--r-- | block/bfq-iosched.c | 3740 |
1 files changed, 2792 insertions, 948 deletions
diff --git a/block/bfq-iosched.c b/block/bfq-iosched.c index cd307767a134..4a8d3d96bfe4 100644 --- a/block/bfq-iosched.c +++ b/block/bfq-iosched.c @@ -1,3 +1,4 @@ +// SPDX-License-Identifier: GPL-2.0-or-later /* * Budget Fair Queueing (BFQ) I/O scheduler. * @@ -12,21 +13,11 @@ * * Copyright (C) 2017 Paolo Valente <paolo.valente@linaro.org> * - * This program is free software; you can redistribute it and/or - * modify it under the terms of the GNU General Public License as - * published by the Free Software Foundation; either version 2 of the - * License, or (at your option) any later version. - * - * This program is distributed in the hope that it will be useful, - * but WITHOUT ANY WARRANTY; without even the implied warranty of - * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU - * General Public License for more details. - * * BFQ is a proportional-share I/O scheduler, with some extra * low-latency capabilities. BFQ also supports full hierarchical * scheduling through cgroups. Next paragraphs provide an introduction * on BFQ inner workings. Details on BFQ benefits, usage and - * limitations can be found in Documentation/block/bfq-iosched.txt. + * limitations can be found in Documentation/block/bfq-iosched.rst. * * BFQ is a proportional-share storage-I/O scheduling algorithm based * on the slice-by-slice service scheme of CFQ. But BFQ assigns @@ -126,16 +117,18 @@ #include <linux/slab.h> #include <linux/blkdev.h> #include <linux/cgroup.h> -#include <linux/elevator.h> #include <linux/ktime.h> #include <linux/rbtree.h> #include <linux/ioprio.h> #include <linux/sbitmap.h> #include <linux/delay.h> +#include <linux/backing-dev.h> +#include <trace/events/block.h> + +#include "elevator.h" #include "blk.h" #include "blk-mq.h" -#include "blk-mq-tag.h" #include "blk-mq-sched.h" #include "bfq-iosched.h" #include "blk-wbt.h" @@ -168,7 +161,7 @@ BFQ_BFQQ_FNS(split_coop); BFQ_BFQQ_FNS(softrt_update); #undef BFQ_BFQQ_FNS \ -/* Expiration time of sync (0) and async (1) requests, in ns. */ +/* Expiration time of async (0) and sync (1) requests, in ns. */ static const u64 bfq_fifo_expire[2] = { NSEC_PER_SEC / 4, NSEC_PER_SEC / 8 }; /* Maximum backwards seek (magic number lifted from CFQ), in KiB. */ @@ -189,7 +182,7 @@ static const int bfq_default_max_budget = 16 * 1024; /* * When a sync request is dispatched, the queue that contains that * request, and all the ancestor entities of that queue, are charged - * with the number of sectors of the request. In constrast, if the + * with the number of sectors of the request. In contrast, if the * request is async, then the queue and its ancestor entities are * charged with the number of sectors of the request, multiplied by * the factor below. This throttles the bandwidth for async I/O, @@ -217,7 +210,7 @@ const int bfq_timeout = HZ / 8; * queue merging. * * As can be deduced from the low time limit below, queue merging, if - * successful, happens at the very beggining of the I/O of the involved + * successful, happens at the very beginning of the I/O of the involved * cooperating processes, as a consequence of the arrival of the very * first requests from each cooperator. After that, there is very * little chance to find cooperators. @@ -230,13 +223,26 @@ static struct kmem_cache *bfq_pool; #define BFQ_MIN_TT (2 * NSEC_PER_MSEC) /* hw_tag detection: parallel requests threshold and min samples needed. */ -#define BFQ_HW_QUEUE_THRESHOLD 4 +#define BFQ_HW_QUEUE_THRESHOLD 3 #define BFQ_HW_QUEUE_SAMPLES 32 #define BFQQ_SEEK_THR (sector_t)(8 * 100) #define BFQQ_SECT_THR_NONROT (sector_t)(2 * 32) +#define BFQ_RQ_SEEKY(bfqd, last_pos, rq) \ + (get_sdist(last_pos, rq) > \ + BFQQ_SEEK_THR && \ + (!blk_queue_nonrot(bfqd->queue) || \ + blk_rq_sectors(rq) < BFQQ_SECT_THR_NONROT)) #define BFQQ_CLOSE_THR (sector_t)(8 * 1024) #define BFQQ_SEEKY(bfqq) (hweight32(bfqq->seek_history) > 19) +/* + * Sync random I/O is likely to be confused with soft real-time I/O, + * because it is characterized by limited throughput and apparently + * isochronous arrival pattern. To avoid false positives, queues + * containing only random (seeky) I/O are prevented from being tagged + * as soft real-time. + */ +#define BFQQ_TOTALLY_SEEKY(bfqq) (bfqq->seek_history == -1) /* Min number of samples required to perform peak-rate update */ #define BFQ_RATE_MIN_SAMPLES 32 @@ -357,17 +363,74 @@ static int ref_wr_duration[2]; */ static const unsigned long max_service_from_wr = 120000; -#define RQ_BIC(rq) icq_to_bic((rq)->elv.priv[0]) +/* + * Maximum time between the creation of two queues, for stable merge + * to be activated (in ms) + */ +static const unsigned long bfq_activation_stable_merging = 600; +/* + * Minimum time to be waited before evaluating delayed stable merge (in ms) + */ +static const unsigned long bfq_late_stable_merging = 600; + +#define RQ_BIC(rq) ((struct bfq_io_cq *)((rq)->elv.priv[0])) #define RQ_BFQQ(rq) ((rq)->elv.priv[1]) -struct bfq_queue *bic_to_bfqq(struct bfq_io_cq *bic, bool is_sync) +struct bfq_queue *bic_to_bfqq(struct bfq_io_cq *bic, bool is_sync, + unsigned int actuator_idx) { - return bic->bfqq[is_sync]; + if (is_sync) + return bic->bfqq[1][actuator_idx]; + + return bic->bfqq[0][actuator_idx]; } -void bic_set_bfqq(struct bfq_io_cq *bic, struct bfq_queue *bfqq, bool is_sync) +static void bfq_put_stable_ref(struct bfq_queue *bfqq); + +void bic_set_bfqq(struct bfq_io_cq *bic, + struct bfq_queue *bfqq, + bool is_sync, + unsigned int actuator_idx) { - bic->bfqq[is_sync] = bfqq; + struct bfq_queue *old_bfqq = bic->bfqq[is_sync][actuator_idx]; + + /* + * If bfqq != NULL, then a non-stable queue merge between + * bic->bfqq and bfqq is happening here. This causes troubles + * in the following case: bic->bfqq has also been scheduled + * for a possible stable merge with bic->stable_merge_bfqq, + * and bic->stable_merge_bfqq == bfqq happens to + * hold. Troubles occur because bfqq may then undergo a split, + * thereby becoming eligible for a stable merge. Yet, if + * bic->stable_merge_bfqq points exactly to bfqq, then bfqq + * would be stably merged with itself. To avoid this anomaly, + * we cancel the stable merge if + * bic->stable_merge_bfqq == bfqq. + */ + struct bfq_iocq_bfqq_data *bfqq_data = &bic->bfqq_data[actuator_idx]; + + /* Clear bic pointer if bfqq is detached from this bic */ + if (old_bfqq && old_bfqq->bic == bic) + old_bfqq->bic = NULL; + + if (is_sync) + bic->bfqq[1][actuator_idx] = bfqq; + else + bic->bfqq[0][actuator_idx] = bfqq; + + if (bfqq && bfqq_data->stable_merge_bfqq == bfqq) { + /* + * Actually, these same instructions are executed also + * in bfq_setup_cooperator, in case of abort or actual + * execution of a stable merge. We could avoid + * repeating these instructions there too, but if we + * did so, we would nest even more complexity in this + * function. + */ + bfq_put_stable_ref(bfqq_data->stable_merge_bfqq); + + bfqq_data->stable_merge_bfqq = NULL; + } } struct bfq_data *bic_to_bfqd(struct bfq_io_cq *bic) @@ -387,26 +450,14 @@ static struct bfq_io_cq *icq_to_bic(struct io_cq *icq) /** * bfq_bic_lookup - search into @ioc a bic associated to @bfqd. - * @bfqd: the lookup key. - * @ioc: the io_context of the process doing I/O. * @q: the request queue. */ -static struct bfq_io_cq *bfq_bic_lookup(struct bfq_data *bfqd, - struct io_context *ioc, - struct request_queue *q) +static struct bfq_io_cq *bfq_bic_lookup(struct request_queue *q) { - if (ioc) { - unsigned long flags; - struct bfq_io_cq *icq; - - spin_lock_irqsave(&q->queue_lock, flags); - icq = icq_to_bic(ioc_lookup_icq(ioc, q)); - spin_unlock_irqrestore(&q->queue_lock, flags); - - return icq; - } + if (!current->io_context) + return NULL; - return NULL; + return icq_to_bic(ioc_lookup_icq(q)); } /* @@ -415,6 +466,8 @@ static struct bfq_io_cq *bfq_bic_lookup(struct bfq_data *bfqd, */ void bfq_schedule_dispatch(struct bfq_data *bfqd) { + lockdep_assert_held(&bfqd->lock); + if (bfqd->queued != 0) { bfq_log(bfqd, "schedule dispatch"); blk_mq_run_hw_queues(bfqd->queue, true); @@ -422,13 +475,12 @@ void bfq_schedule_dispatch(struct bfq_data *bfqd) } #define bfq_class_idle(bfqq) ((bfqq)->ioprio_class == IOPRIO_CLASS_IDLE) -#define bfq_class_rt(bfqq) ((bfqq)->ioprio_class == IOPRIO_CLASS_RT) #define bfq_sample_valid(samples) ((samples) > 80) /* * Lifted from AS - choose which of rq1 and rq2 that is best served now. - * We choose the request that is closesr to the head right now. Distance + * We choose the request that is closer to the head right now. Distance * behind the head is penalized and only allowed to a certain extent. */ static struct request *bfq_choose_req(struct bfq_data *bfqd, @@ -520,26 +572,158 @@ static struct request *bfq_choose_req(struct bfq_data *bfqd, } } +#define BFQ_LIMIT_INLINE_DEPTH 16 + +#ifdef CONFIG_BFQ_GROUP_IOSCHED +static bool bfqq_request_over_limit(struct bfq_data *bfqd, + struct bfq_io_cq *bic, blk_opf_t opf, + unsigned int act_idx, int limit) +{ + struct bfq_entity *inline_entities[BFQ_LIMIT_INLINE_DEPTH]; + struct bfq_entity **entities = inline_entities; + int alloc_depth = BFQ_LIMIT_INLINE_DEPTH; + struct bfq_sched_data *sched_data; + struct bfq_entity *entity; + struct bfq_queue *bfqq; + unsigned long wsum; + bool ret = false; + int depth; + int level; + +retry: + spin_lock_irq(&bfqd->lock); + bfqq = bic_to_bfqq(bic, op_is_sync(opf), act_idx); + if (!bfqq) + goto out; + + entity = &bfqq->entity; + if (!entity->on_st_or_in_serv) + goto out; + + /* +1 for bfqq entity, root cgroup not included */ + depth = bfqg_to_blkg(bfqq_group(bfqq))->blkcg->css.cgroup->level + 1; + if (depth > alloc_depth) { + spin_unlock_irq(&bfqd->lock); + if (entities != inline_entities) + kfree(entities); + entities = kmalloc_array(depth, sizeof(*entities), GFP_NOIO); + if (!entities) + return false; + alloc_depth = depth; + goto retry; + } + + sched_data = entity->sched_data; + /* Gather our ancestors as we need to traverse them in reverse order */ + level = 0; + for_each_entity(entity) { + /* + * If at some level entity is not even active, allow request + * queueing so that BFQ knows there's work to do and activate + * entities. + */ + if (!entity->on_st_or_in_serv) + goto out; + /* Uh, more parents than cgroup subsystem thinks? */ + if (WARN_ON_ONCE(level >= depth)) + break; + entities[level++] = entity; + } + WARN_ON_ONCE(level != depth); + for (level--; level >= 0; level--) { + entity = entities[level]; + if (level > 0) { + wsum = bfq_entity_service_tree(entity)->wsum; + } else { + int i; + /* + * For bfqq itself we take into account service trees + * of all higher priority classes and multiply their + * weights so that low prio queue from higher class + * gets more requests than high prio queue from lower + * class. + */ + wsum = 0; + for (i = 0; i <= bfqq->ioprio_class - 1; i++) { + wsum = wsum * IOPRIO_BE_NR + + sched_data->service_tree[i].wsum; + } + } + if (!wsum) + continue; + limit = DIV_ROUND_CLOSEST(limit * entity->weight, wsum); + if (entity->allocated >= limit) { + bfq_log_bfqq(bfqq->bfqd, bfqq, + "too many requests: allocated %d limit %d level %d", + entity->allocated, limit, level); + ret = true; + break; + } + } +out: + spin_unlock_irq(&bfqd->lock); + if (entities != inline_entities) + kfree(entities); + return ret; +} +#else +static bool bfqq_request_over_limit(struct bfq_data *bfqd, + struct bfq_io_cq *bic, blk_opf_t opf, + unsigned int act_idx, int limit) +{ + return false; +} +#endif + /* * Async I/O can easily starve sync I/O (both sync reads and sync * writes), by consuming all tags. Similarly, storms of sync writes, * such as those that sync(2) may trigger, can starve sync reads. * Limit depths of async I/O and sync writes so as to counter both * problems. + * + * Also if a bfq queue or its parent cgroup consume more tags than would be + * appropriate for their weight, we trim the available tag depth to 1. This + * avoids a situation where one cgroup can starve another cgroup from tags and + * thus block service differentiation among cgroups. Note that because the + * queue / cgroup already has many requests allocated and queued, this does not + * significantly affect service guarantees coming from the BFQ scheduling + * algorithm. */ -static void bfq_limit_depth(unsigned int op, struct blk_mq_alloc_data *data) +static void bfq_limit_depth(blk_opf_t opf, struct blk_mq_alloc_data *data) { struct bfq_data *bfqd = data->q->elevator->elevator_data; + struct bfq_io_cq *bic = bfq_bic_lookup(data->q); + unsigned int limit, act_idx; - if (op_is_sync(op) && !op_is_write(op)) - return; + /* Sync reads have full depth available */ + if (op_is_sync(opf) && !op_is_write(opf)) + limit = data->q->nr_requests; + else + limit = bfqd->async_depths[!!bfqd->wr_busy_queues][op_is_sync(opf)]; - data->shallow_depth = - bfqd->word_depths[!!bfqd->wr_busy_queues][op_is_sync(op)]; + for (act_idx = 0; bic && act_idx < bfqd->num_actuators; act_idx++) { + /* Fast path to check if bfqq is already allocated. */ + if (!bic_to_bfqq(bic, op_is_sync(opf), act_idx)) + continue; + + /* + * Does queue (or any parent entity) exceed number of + * requests that should be available to it? Heavily + * limit depth so that it cannot consume more + * available requests and thus starve other entities. + */ + if (bfqq_request_over_limit(bfqd, bic, opf, act_idx, limit)) { + limit = 1; + break; + } + } bfq_log(bfqd, "[%s] wr_busy %d sync %d depth %u", - __func__, bfqd->wr_busy_queues, op_is_sync(op), - data->shallow_depth); + __func__, bfqd->wr_busy_queues, op_is_sync(opf), limit); + + if (limit < data->q->nr_requests) + data->shallow_depth = limit; } static struct bfq_queue * @@ -590,7 +774,16 @@ static bool bfq_too_late_for_merging(struct bfq_queue *bfqq) bfq_merge_time_limit); } -void bfq_pos_tree_add_move(struct bfq_data *bfqd, struct bfq_queue *bfqq) +/* + * The following function is not marked as __cold because it is + * actually cold, but for the same performance goal described in the + * comments on the likely() at the beginning of + * bfq_setup_cooperator(). Unexpectedly, to reach an even lower + * execution time for the case where this function is not invoked, we + * had to add an unlikely() in each involved if(). + */ +void __cold +bfq_pos_tree_add_move(struct bfq_data *bfqd, struct bfq_queue *bfqq) { struct rb_node **p, *parent; struct bfq_queue *__bfqq; @@ -600,6 +793,10 @@ void bfq_pos_tree_add_move(struct bfq_data *bfqd, struct bfq_queue *bfqq) bfqq->pos_root = NULL; } + /* oom_bfqq does not participate in queue merging */ + if (bfqq == &bfqd->oom_bfqq) + return; + /* * bfqq cannot be merged any longer (see comments in * bfq_setup_cooperator): no point in adding bfqq into the @@ -613,7 +810,7 @@ void bfq_pos_tree_add_move(struct bfq_data *bfqd, struct bfq_queue *bfqq) if (!bfqq->next_rq) return; - bfqq->pos_root = &bfq_bfqq_to_bfqg(bfqq)->rq_pos_tree; + bfqq->pos_root = &bfqq_group(bfqq)->rq_pos_tree; __bfqq = bfq_rq_pos_tree_lookup(bfqd, bfqq->pos_root, blk_rq_pos(bfqq->next_rq), &parent, &p); if (!__bfqq) { @@ -624,52 +821,68 @@ void bfq_pos_tree_add_move(struct bfq_data *bfqd, struct bfq_queue *bfqq) } /* - * Tell whether there are active queues with different weights or - * active groups. - */ -static bool bfq_varied_queue_weights_or_active_groups(struct bfq_data *bfqd) -{ - /* - * For queue weights to differ, queue_weights_tree must contain - * at least two nodes. - */ - return (!RB_EMPTY_ROOT(&bfqd->queue_weights_tree) && - (bfqd->queue_weights_tree.rb_node->rb_left || - bfqd->queue_weights_tree.rb_node->rb_right) -#ifdef CONFIG_BFQ_GROUP_IOSCHED - ) || - (bfqd->num_groups_with_pending_reqs > 0 -#endif - ); -} - -/* - * The following function returns true if every queue must receive the - * same share of the throughput (this condition is used when deciding - * whether idling may be disabled, see the comments in the function - * bfq_better_to_idle()). + * The following function returns false either if every active queue + * must receive the same share of the throughput (symmetric scenario), + * or, as a special case, if bfqq must receive a share of the + * throughput lower than or equal to the share that every other active + * queue must receive. If bfqq does sync I/O, then these are the only + * two cases where bfqq happens to be guaranteed its share of the + * throughput even if I/O dispatching is not plugged when bfqq remains + * temporarily empty (for more details, see the comments in the + * function bfq_better_to_idle()). For this reason, the return value + * of this function is used to check whether I/O-dispatch plugging can + * be avoided. * - * Such a scenario occurs when: + * The above first case (symmetric scenario) occurs when: * 1) all active queues have the same weight, - * 2) all active groups at the same level in the groups tree have the same - * weight, + * 2) all active queues belong to the same I/O-priority class, * 3) all active groups at the same level in the groups tree have the same + * weight, + * 4) all active groups at the same level in the groups tree have the same * number of children. * * Unfortunately, keeping the necessary state for evaluating exactly * the last two symmetry sub-conditions above would be quite complex - * and time consuming. Therefore this function evaluates, instead, - * only the following stronger two sub-conditions, for which it is + * and time consuming. Therefore this function evaluates, instead, + * only the following stronger three sub-conditions, for which it is * much easier to maintain the needed state: * 1) all active queues have the same weight, - * 2) there are no active groups. + * 2) all active queues belong to the same I/O-priority class, + * 3) there is at most one active group. * In particular, the last condition is always true if hierarchical * support or the cgroups interface are not enabled, thus no state * needs to be maintained in this case. */ -static bool bfq_symmetric_scenario(struct bfq_data *bfqd) +static bool bfq_asymmetric_scenario(struct bfq_data *bfqd, + struct bfq_queue *bfqq) { - return !bfq_varied_queue_weights_or_active_groups(bfqd); + bool smallest_weight = bfqq && + bfqq->weight_counter && + bfqq->weight_counter == + container_of( + rb_first_cached(&bfqd->queue_weights_tree), + struct bfq_weight_counter, + weights_node); + + /* + * For queue weights to differ, queue_weights_tree must contain + * at least two nodes. + */ + bool varied_queue_weights = !smallest_weight && + !RB_EMPTY_ROOT(&bfqd->queue_weights_tree.rb_root) && + (bfqd->queue_weights_tree.rb_root.rb_node->rb_left || + bfqd->queue_weights_tree.rb_root.rb_node->rb_right); + + bool multiple_classes_busy = + (bfqd->busy_queues[0] && bfqd->busy_queues[1]) || + (bfqd->busy_queues[0] && bfqd->busy_queues[2]) || + (bfqd->busy_queues[1] && bfqd->busy_queues[2]); + + return varied_queue_weights || multiple_classes_busy +#ifdef CONFIG_BFQ_GROUP_IOSCHED + || bfqd->num_groups_with_pending_reqs > 1 +#endif + ; } /* @@ -685,11 +898,12 @@ static bool bfq_symmetric_scenario(struct bfq_data *bfqd) * In most scenarios, the rate at which nodes are created/destroyed * should be low too. */ -void bfq_weights_tree_add(struct bfq_data *bfqd, struct bfq_queue *bfqq, - struct rb_root *root) +void bfq_weights_tree_add(struct bfq_queue *bfqq) { + struct rb_root_cached *root = &bfqq->bfqd->queue_weights_tree; struct bfq_entity *entity = &bfqq->entity; - struct rb_node **new = &(root->rb_node), *parent = NULL; + struct rb_node **new = &(root->rb_root.rb_node), *parent = NULL; + bool leftmost = true; /* * Do not insert if the queue is already associated with a @@ -718,8 +932,10 @@ void bfq_weights_tree_add(struct bfq_data *bfqd, struct bfq_queue *bfqq, } if (entity->weight < __counter->weight) new = &((*new)->rb_left); - else + else { new = &((*new)->rb_right); + leftmost = false; + } } bfqq->weight_counter = kzalloc(sizeof(struct bfq_weight_counter), @@ -728,25 +944,26 @@ void bfq_weights_tree_add(struct bfq_data *bfqd, struct bfq_queue *bfqq, /* * In the unlucky event of an allocation failure, we just * exit. This will cause the weight of queue to not be - * considered in bfq_varied_queue_weights_or_active_groups, - * which, in its turn, causes the scenario to be deemed - * wrongly symmetric in case bfqq's weight would have been - * the only weight making the scenario asymmetric. On the - * bright side, no unbalance will however occur when bfqq - * becomes inactive again (the invocation of this function - * is triggered by an activation of queue). In fact, - * bfq_weights_tree_remove does nothing if - * !bfqq->weight_counter. + * considered in bfq_asymmetric_scenario, which, in its turn, + * causes the scenario to be deemed wrongly symmetric in case + * bfqq's weight would have been the only weight making the + * scenario asymmetric. On the bright side, no unbalance will + * however occur when bfqq becomes inactive again (the + * invocation of this function is triggered by an activation + * of queue). In fact, bfq_weights_tree_remove does nothing + * if !bfqq->weight_counter. */ if (unlikely(!bfqq->weight_counter)) return; bfqq->weight_counter->weight = entity->weight; rb_link_node(&bfqq->weight_counter->weights_node, parent, new); - rb_insert_color(&bfqq->weight_counter->weights_node, root); + rb_insert_color_cached(&bfqq->weight_counter->weights_node, root, + leftmost); inc_counter: bfqq->weight_counter->num_active++; + bfqq->ref++; } /* @@ -755,69 +972,24 @@ inc_counter: * See the comments to the function bfq_weights_tree_add() for considerations * about overhead. */ -void __bfq_weights_tree_remove(struct bfq_data *bfqd, - struct bfq_queue *bfqq, - struct rb_root *root) +void bfq_weights_tree_remove(struct bfq_queue *bfqq) { + struct rb_root_cached *root; + if (!bfqq->weight_counter) return; + root = &bfqq->bfqd->queue_weights_tree; bfqq->weight_counter->num_active--; if (bfqq->weight_counter->num_active > 0) goto reset_entity_pointer; - rb_erase(&bfqq->weight_counter->weights_node, root); + rb_erase_cached(&bfqq->weight_counter->weights_node, root); kfree(bfqq->weight_counter); reset_entity_pointer: bfqq->weight_counter = NULL; -} - -/* - * Invoke __bfq_weights_tree_remove on bfqq and decrement the number - * of active groups for each queue's inactive parent entity. - */ -void bfq_weights_tree_remove(struct bfq_data *bfqd, - struct bfq_queue *bfqq) -{ - struct bfq_entity *entity = bfqq->entity.parent; - - __bfq_weights_tree_remove(bfqd, bfqq, - &bfqd->queue_weights_tree); - - for_each_entity(entity) { - struct bfq_sched_data *sd = entity->my_sched_data; - - if (sd->next_in_service || sd->in_service_entity) { - /* - * entity is still active, because either - * next_in_service or in_service_entity is not - * NULL (see the comments on the definition of - * next_in_service for details on why - * in_service_entity must be checked too). - * - * As a consequence, its parent entities are - * active as well, and thus this loop must - * stop here. - */ - break; - } - - /* - * 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--; - } - } + bfq_put_queue(bfqq); } /* @@ -835,7 +1007,7 @@ static struct request *bfq_check_fifo(struct bfq_queue *bfqq, rq = rq_entry_fifo(bfqq->fifo.next); - if (rq == last || ktime_get_ns() < rq->fifo_time) + if (rq == last || blk_time_get_ns() < rq->fifo_time) return NULL; bfq_log_bfqq(bfqq->bfqd, bfqq, "check_fifo: returned %p", rq); @@ -873,7 +1045,8 @@ static struct request *bfq_find_next_rq(struct bfq_data *bfqd, static unsigned long bfq_serv_to_charge(struct request *rq, struct bfq_queue *bfqq) { - if (bfq_bfqq_sync(bfqq) || bfqq->wr_coeff > 1) + if (bfq_bfqq_sync(bfqq) || bfqq->wr_coeff > 1 || + bfq_asymmetric_scenario(bfqq->bfqd, bfqq)) return blk_rq_sectors(rq); return blk_rq_sectors(rq) * bfq_async_charge_factor; @@ -907,8 +1080,10 @@ static void bfq_updated_next_req(struct bfq_data *bfqd, */ return; - new_budget = max_t(unsigned long, bfqq->max_budget, - bfq_serv_to_charge(next_rq, bfqq)); + new_budget = max_t(unsigned long, + max_t(unsigned long, bfqq->max_budget, + bfq_serv_to_charge(next_rq, bfqq)), + entity->service); if (entity->budget != new_budget) { entity->budget = new_budget; bfq_log_bfqq(bfqd, bfqq, "updated next rq: new budget %lu", @@ -921,9 +1096,6 @@ static unsigned int bfq_wr_duration(struct bfq_data *bfqd) { u64 dur; - if (bfqd->bfq_wr_max_time > 0) - return bfqd->bfq_wr_max_time; - dur = bfqd->rate_dur_prod; do_div(dur, bfqd->peak_rate); @@ -937,7 +1109,7 @@ static unsigned int bfq_wr_duration(struct bfq_data *bfqd) * of several files * mplayer took 23 seconds to start, if constantly weight-raised. * - * As for higher values than that accomodating the above bad + * As for higher values than that accommodating the above bad * scenario, tests show that higher values would often yield * the opposite of the desired result, i.e., would worsen * responsiveness by allowing non-interactive applications to @@ -963,24 +1135,41 @@ static void bfq_bfqq_resume_state(struct bfq_queue *bfqq, struct bfq_data *bfqd, struct bfq_io_cq *bic, bool bfq_already_existing) { - unsigned int old_wr_coeff = bfqq->wr_coeff; + unsigned int old_wr_coeff = 1; bool busy = bfq_already_existing && bfq_bfqq_busy(bfqq); + unsigned int a_idx = bfqq->actuator_idx; + struct bfq_iocq_bfqq_data *bfqq_data = &bic->bfqq_data[a_idx]; - if (bic->saved_has_short_ttime) + if (bfqq_data->saved_has_short_ttime) bfq_mark_bfqq_has_short_ttime(bfqq); else bfq_clear_bfqq_has_short_ttime(bfqq); - if (bic->saved_IO_bound) + if (bfqq_data->saved_IO_bound) bfq_mark_bfqq_IO_bound(bfqq); else bfq_clear_bfqq_IO_bound(bfqq); - bfqq->ttime = bic->saved_ttime; - bfqq->wr_coeff = bic->saved_wr_coeff; - bfqq->wr_start_at_switch_to_srt = bic->saved_wr_start_at_switch_to_srt; - bfqq->last_wr_start_finish = bic->saved_last_wr_start_finish; - bfqq->wr_cur_max_time = bic->saved_wr_cur_max_time; + bfqq->last_serv_time_ns = bfqq_data->saved_last_serv_time_ns; + bfqq->inject_limit = bfqq_data->saved_inject_limit; + bfqq->decrease_time_jif = bfqq_data->saved_decrease_time_jif; + + bfqq->entity.new_weight = bfqq_data->saved_weight; + bfqq->ttime = bfqq_data->saved_ttime; + bfqq->io_start_time = bfqq_data->saved_io_start_time; + bfqq->tot_idle_time = bfqq_data->saved_tot_idle_time; + /* + * Restore weight coefficient only if low_latency is on + */ + if (bfqd->low_latency) { + old_wr_coeff = bfqq->wr_coeff; + bfqq->wr_coeff = bfqq_data->saved_wr_coeff; + } + bfqq->service_from_wr = bfqq_data->saved_service_from_wr; + bfqq->wr_start_at_switch_to_srt = + bfqq_data->saved_wr_start_at_switch_to_srt; + bfqq->last_wr_start_finish = bfqq_data->saved_last_wr_start_finish; + bfqq->wr_cur_max_time = bfqq_data->saved_wr_cur_max_time; if (bfqq->wr_coeff > 1 && (bfq_bfqq_in_large_burst(bfqq) || time_is_before_jiffies(bfqq->last_wr_start_finish + @@ -1011,7 +1200,9 @@ bfq_bfqq_resume_state(struct bfq_queue *bfqq, struct bfq_data *bfqd, static int bfqq_process_refs(struct bfq_queue *bfqq) { - return bfqq->ref - bfqq->allocated - bfqq->entity.on_st; + return bfqq->ref - bfqq->entity.allocated - + bfqq->entity.on_st_or_in_serv - + (bfqq->weight_counter != NULL) - bfqq->stable_ref; } /* Empty burst list and add just bfqq (see comments on bfq_handle_burst) */ @@ -1022,8 +1213,18 @@ static void bfq_reset_burst_list(struct bfq_data *bfqd, struct bfq_queue *bfqq) hlist_for_each_entry_safe(item, n, &bfqd->burst_list, burst_list_node) hlist_del_init(&item->burst_list_node); - hlist_add_head(&bfqq->burst_list_node, &bfqd->burst_list); - bfqd->burst_size = 1; + + /* + * Start the creation of a new burst list only if there is no + * active queue. See comments on the conditional invocation of + * bfq_handle_burst(). + */ + if (bfq_tot_busy_queues(bfqd) == 0) { + hlist_add_head(&bfqq->burst_list_node, &bfqd->burst_list); + bfqd->burst_size = 1; + } else + bfqd->burst_size = 0; + bfqd->burst_parent_entity = bfqq->entity.parent; } @@ -1079,7 +1280,8 @@ static void bfq_add_to_burst(struct bfq_data *bfqd, struct bfq_queue *bfqq) * many parallel threads/processes. Examples are systemd during boot, * or git grep. To help these processes get their job done as soon as * possible, it is usually better to not grant either weight-raising - * or device idling to their queues. + * or device idling to their queues, unless these queues must be + * protected from the I/O flowing through other active queues. * * In this comment we describe, firstly, the reasons why this fact * holds, and, secondly, the next function, which implements the main @@ -1091,7 +1293,10 @@ static void bfq_add_to_burst(struct bfq_data *bfqd, struct bfq_queue *bfqq) * cumulatively served, the sooner the target job of these queues gets * completed. As a consequence, weight-raising any of these queues, * which also implies idling the device for it, is almost always - * counterproductive. In most cases it just lowers throughput. + * counterproductive, unless there are other active queues to isolate + * these new queues from. If there no other active queues, then + * weight-raising these new queues just lowers throughput in most + * cases. * * On the other hand, a burst of queue creations may be caused also by * the start of an application that does not consist of a lot of @@ -1125,14 +1330,16 @@ static void bfq_add_to_burst(struct bfq_data *bfqd, struct bfq_queue *bfqq) * are very rare. They typically occur if some service happens to * start doing I/O exactly when the interactive task starts. * - * Turning back to the next function, it implements all the steps - * needed to detect the occurrence of a large burst and to properly - * mark all the queues belonging to it (so that they can then be - * treated in a different way). This goal is achieved by maintaining a - * "burst list" that holds, temporarily, the queues that belong to the - * burst in progress. The list is then used to mark these queues as - * belonging to a large burst if the burst does become large. The main - * steps are the following. + * Turning back to the next function, it is invoked only if there are + * no active queues (apart from active queues that would belong to the + * same, possible burst bfqq would belong to), and it implements all + * the steps needed to detect the occurrence of a large burst and to + * properly mark all the queues belonging to it (so that they can then + * be treated in a different way). This goal is achieved by + * maintaining a "burst list" that holds, temporarily, the queues that + * belong to the burst in progress. The list is then used to mark + * these queues as belonging to a large burst if the burst does become + * large. The main steps are the following. * * . when the very first queue is created, the queue is inserted into the * list (as it could be the first queue in a possible burst) @@ -1366,21 +1573,31 @@ static int bfq_min_budget(struct bfq_data *bfqd) * mechanism may be re-designed in such a way to make it possible to * know whether preemption is needed without needing to update service * trees). In addition, queue preemptions almost always cause random - * I/O, and thus loss of throughput. Because of these facts, the next - * function adopts the following simple scheme to avoid both costly - * operations and too frequent preemptions: it requests the expiration - * of the in-service queue (unconditionally) only for queues that need - * to recover a hole, or that either are weight-raised or deserve to - * be weight-raised. + * I/O, which may in turn cause loss of throughput. Finally, there may + * even be no in-service queue when the next function is invoked (so, + * no queue to compare timestamps with). Because of these facts, the + * next function adopts the following simple scheme to avoid costly + * operations, too frequent preemptions and too many dependencies on + * the state of the scheduler: it requests the expiration of the + * in-service queue (unconditionally) only for queues that need to + * recover a hole. Then it delegates to other parts of the code the + * responsibility of handling the above case 2. */ static bool bfq_bfqq_update_budg_for_activation(struct bfq_data *bfqd, struct bfq_queue *bfqq, - bool arrived_in_time, - bool wr_or_deserves_wr) + bool arrived_in_time) { struct bfq_entity *entity = &bfqq->entity; - if (bfq_bfqq_non_blocking_wait_rq(bfqq) && arrived_in_time) { + /* + * In the next compound condition, we check also whether there + * is some budget left, because otherwise there is no point in + * trying to go on serving bfqq with this same budget: bfqq + * would be expired immediately after being selected for + * service. This would only cause useless overhead. + */ + if (bfq_bfqq_non_blocking_wait_rq(bfqq) && arrived_in_time && + bfq_bfqq_budget_left(bfqq) > 0) { /* * We do not clear the flag non_blocking_wait_rq here, as * the latter is used in bfq_activate_bfqq to signal @@ -1423,7 +1640,7 @@ static bool bfq_bfqq_update_budg_for_activation(struct bfq_data *bfqd, entity->budget = max_t(unsigned long, bfqq->max_budget, bfq_serv_to_charge(bfqq->next_rq, bfqq)); bfq_clear_bfqq_non_blocking_wait_rq(bfqq); - return wr_or_deserves_wr; + return false; } /* @@ -1541,6 +1758,65 @@ static bool bfq_bfqq_idle_for_long_time(struct bfq_data *bfqd, bfqd->bfq_wr_min_idle_time); } + +/* + * Return true if bfqq is in a higher priority class, or has a higher + * weight than the in-service queue. + */ +static bool bfq_bfqq_higher_class_or_weight(struct bfq_queue *bfqq, + struct bfq_queue *in_serv_bfqq) +{ + int bfqq_weight, in_serv_weight; + + if (bfqq->ioprio_class < in_serv_bfqq->ioprio_class) + return true; + + if (in_serv_bfqq->entity.parent == bfqq->entity.parent) { + bfqq_weight = bfqq->entity.weight; + in_serv_weight = in_serv_bfqq->entity.weight; + } else { + if (bfqq->entity.parent) + bfqq_weight = bfqq->entity.parent->weight; + else + bfqq_weight = bfqq->entity.weight; + if (in_serv_bfqq->entity.parent) + in_serv_weight = in_serv_bfqq->entity.parent->weight; + else + in_serv_weight = in_serv_bfqq->entity.weight; + } + + return bfqq_weight > in_serv_weight; +} + +/* + * Get the index of the actuator that will serve bio. + */ +static unsigned int bfq_actuator_index(struct bfq_data *bfqd, struct bio *bio) +{ + unsigned int i; + sector_t end; + + /* no search needed if one or zero ranges present */ + if (bfqd->num_actuators == 1) + return 0; + + /* bio_end_sector(bio) gives the sector after the last one */ + end = bio_end_sector(bio) - 1; + + for (i = 0; i < bfqd->num_actuators; i++) { + if (end >= bfqd->sector[i] && + end < bfqd->sector[i] + bfqd->nr_sectors[i]) + return i; + } + + WARN_ONCE(true, + "bfq_actuator_index: bio sector out of ranges: end=%llu\n", + end); + return 0; +} + +static bool bfq_better_to_idle(struct bfq_queue *bfqq); + static void bfq_bfqq_handle_idle_busy_switch(struct bfq_data *bfqd, struct bfq_queue *bfqq, int old_wr_coeff, @@ -1555,28 +1831,47 @@ static void bfq_bfqq_handle_idle_busy_switch(struct bfq_data *bfqd, * bfq_bfqq_update_budg_for_activation for * details on the usage of the next variable. */ - arrived_in_time = ktime_get_ns() <= + arrived_in_time = blk_time_get_ns() <= bfqq->ttime.last_end_request + bfqd->bfq_slice_idle * 3; - + unsigned int act_idx = bfq_actuator_index(bfqd, rq->bio); + bool bfqq_non_merged_or_stably_merged = + bfqq->bic || RQ_BIC(rq)->bfqq_data[act_idx].stably_merged; /* * bfqq deserves to be weight-raised if: * - it is sync, * - it does not belong to a large burst, * - it has been idle for enough time or is soft real-time, - * - is linked to a bfq_io_cq (it is not shared in any sense). + * - is linked to a bfq_io_cq (it is not shared in any sense), + * - has a default weight (otherwise we assume the user wanted + * to control its weight explicitly) */ in_burst = bfq_bfqq_in_large_burst(bfqq); soft_rt = bfqd->bfq_wr_max_softrt_rate > 0 && + !BFQQ_TOTALLY_SEEKY(bfqq) && !in_burst && time_is_before_jiffies(bfqq->soft_rt_next_start) && - bfqq->dispatched == 0; - *interactive = !in_burst && idle_for_long_time; + bfqq->dispatched == 0 && + bfqq->entity.new_weight == 40; + *interactive = !in_burst && idle_for_long_time && + bfqq->entity.new_weight == 40; + /* + * Merged bfq_queues are kept out of weight-raising + * (low-latency) mechanisms. The reason is that these queues + * are usually created for non-interactive and + * non-soft-real-time tasks. Yet this is not the case for + * stably-merged queues. These queues are merged just because + * they are created shortly after each other. So they may + * easily serve the I/O of an interactive or soft-real time + * application, if the application happens to spawn multiple + * processes. So let also stably-merged queued enjoy weight + * raising. + */ wr_or_deserves_wr = bfqd->low_latency && (bfqq->wr_coeff > 1 || - (bfq_bfqq_sync(bfqq) && - bfqq->bic && (*interactive || soft_rt))); + (bfq_bfqq_sync(bfqq) && bfqq_non_merged_or_stably_merged && + (*interactive || soft_rt))); /* * Using the last flag, update budget and check whether bfqq @@ -1584,8 +1879,7 @@ static void bfq_bfqq_handle_idle_busy_switch(struct bfq_data *bfqd, */ bfqq_wants_to_preempt = bfq_bfqq_update_budg_for_activation(bfqd, bfqq, - arrived_in_time, - wr_or_deserves_wr); + arrived_in_time); /* * If bfqq happened to be activated in a burst, but has been @@ -1611,17 +1905,6 @@ static void bfq_bfqq_handle_idle_busy_switch(struct bfq_data *bfqd, bfq_clear_bfqq_just_created(bfqq); - - if (!bfq_bfqq_IO_bound(bfqq)) { - if (arrived_in_time) { - bfqq->requests_within_timer++; - if (bfqq->requests_within_timer >= - bfqd->bfq_requests_within_timer) - bfq_mark_bfqq_IO_bound(bfqq); - } else - bfqq->requests_within_timer = 0; - } - if (bfqd->low_latency) { if (unlikely(time_is_after_jiffies(bfqq->split_time))) /* wraparound */ @@ -1646,25 +1929,280 @@ static void bfq_bfqq_handle_idle_busy_switch(struct bfq_data *bfqd, bfqq->service_from_backlogged = 0; bfq_clear_bfqq_softrt_update(bfqq); - bfq_add_bfqq_busy(bfqd, bfqq); + bfq_add_bfqq_busy(bfqq); /* - * Expire in-service queue only if preemption may be needed - * for guarantees. In this respect, the function - * next_queue_may_preempt just checks a simple, necessary - * condition, and not a sufficient condition based on - * timestamps. In fact, for the latter condition to be - * evaluated, timestamps would need first to be updated, and - * this operation is quite costly (see the comments on the - * function bfq_bfqq_update_budg_for_activation). + * Expire in-service queue if preemption may be needed for + * guarantees or throughput. As for guarantees, we care + * explicitly about two cases. The first is that bfqq has to + * recover a service hole, as explained in the comments on + * bfq_bfqq_update_budg_for_activation(), i.e., that + * bfqq_wants_to_preempt is true. However, if bfqq does not + * carry time-critical I/O, then bfqq's bandwidth is less + * important than that of queues that carry time-critical I/O. + * So, as a further constraint, we consider this case only if + * bfqq is at least as weight-raised, i.e., at least as time + * critical, as the in-service queue. + * + * The second case is that bfqq is in a higher priority class, + * or has a higher weight than the in-service queue. If this + * condition does not hold, we don't care because, even if + * bfqq does not start to be served immediately, the resulting + * delay for bfqq's I/O is however lower or much lower than + * the ideal completion time to be guaranteed to bfqq's I/O. + * + * In both cases, preemption is needed only if, according to + * the timestamps of both bfqq and of the in-service queue, + * bfqq actually is the next queue to serve. So, to reduce + * useless preemptions, the return value of + * next_queue_may_preempt() is considered in the next compound + * condition too. Yet next_queue_may_preempt() just checks a + * simple, necessary condition for bfqq to be the next queue + * to serve. In fact, to evaluate a sufficient condition, the + * timestamps of the in-service queue would need to be + * updated, and this operation is quite costly (see the + * comments on bfq_bfqq_update_budg_for_activation()). + * + * As for throughput, we ask bfq_better_to_idle() whether we + * still need to plug I/O dispatching. If bfq_better_to_idle() + * says no, then plugging is not needed any longer, either to + * boost throughput or to perserve service guarantees. Then + * the best option is to stop plugging I/O, as not doing so + * would certainly lower throughput. We may end up in this + * case if: (1) upon a dispatch attempt, we detected that it + * was better to plug I/O dispatch, and to wait for a new + * request to arrive for the currently in-service queue, but + * (2) this switch of bfqq to busy changes the scenario. */ - if (bfqd->in_service_queue && bfqq_wants_to_preempt && - bfqd->in_service_queue->wr_coeff < bfqq->wr_coeff && + if (bfqd->in_service_queue && + ((bfqq_wants_to_preempt && + bfqq->wr_coeff >= bfqd->in_service_queue->wr_coeff) || + bfq_bfqq_higher_class_or_weight(bfqq, bfqd->in_service_queue) || + !bfq_better_to_idle(bfqd->in_service_queue)) && next_queue_may_preempt(bfqd)) bfq_bfqq_expire(bfqd, bfqd->in_service_queue, false, BFQQE_PREEMPTED); } +static void bfq_reset_inject_limit(struct bfq_data *bfqd, + struct bfq_queue *bfqq) +{ + /* invalidate baseline total service time */ + bfqq->last_serv_time_ns = 0; + + /* + * Reset pointer in case we are waiting for + * some request completion. + */ + bfqd->waited_rq = NULL; + + /* + * If bfqq has a short think time, then start by setting the + * inject limit to 0 prudentially, because the service time of + * an injected I/O request may be higher than the think time + * of bfqq, and therefore, if one request was injected when + * bfqq remains empty, this injected request might delay the + * service of the next I/O request for bfqq significantly. In + * case bfqq can actually tolerate some injection, then the + * adaptive update will however raise the limit soon. This + * lucky circumstance holds exactly because bfqq has a short + * think time, and thus, after remaining empty, is likely to + * get new I/O enqueued---and then completed---before being + * expired. This is the very pattern that gives the + * limit-update algorithm the chance to measure the effect of + * injection on request service times, and then to update the + * limit accordingly. + * + * However, in the following special case, the inject limit is + * left to 1 even if the think time is short: bfqq's I/O is + * synchronized with that of some other queue, i.e., bfqq may + * receive new I/O only after the I/O of the other queue is + * completed. Keeping the inject limit to 1 allows the + * blocking I/O to be served while bfqq is in service. And + * this is very convenient both for bfqq and for overall + * throughput, as explained in detail in the comments in + * bfq_update_has_short_ttime(). + * + * On the opposite end, if bfqq has a long think time, then + * start directly by 1, because: + * a) on the bright side, keeping at most one request in + * service in the drive is unlikely to cause any harm to the + * latency of bfqq's requests, as the service time of a single + * request is likely to be lower than the think time of bfqq; + * b) on the downside, after becoming empty, bfqq is likely to + * expire before getting its next request. With this request + * arrival pattern, it is very hard to sample total service + * times and update the inject limit accordingly (see comments + * on bfq_update_inject_limit()). So the limit is likely to be + * never, or at least seldom, updated. As a consequence, by + * setting the limit to 1, we avoid that no injection ever + * occurs with bfqq. On the downside, this proactive step + * further reduces chances to actually compute the baseline + * total service time. Thus it reduces chances to execute the + * limit-update algorithm and possibly raise the limit to more + * than 1. + */ + if (bfq_bfqq_has_short_ttime(bfqq)) + bfqq->inject_limit = 0; + else + bfqq->inject_limit = 1; + + bfqq->decrease_time_jif = jiffies; +} + +static void bfq_update_io_intensity(struct bfq_queue *bfqq, u64 now_ns) +{ + u64 tot_io_time = now_ns - bfqq->io_start_time; + + if (RB_EMPTY_ROOT(&bfqq->sort_list) && bfqq->dispatched == 0) + bfqq->tot_idle_time += + now_ns - bfqq->ttime.last_end_request; + + if (unlikely(bfq_bfqq_just_created(bfqq))) + return; + + /* + * Must be busy for at least about 80% of the time to be + * considered I/O bound. + */ + if (bfqq->tot_idle_time * 5 > tot_io_time) + bfq_clear_bfqq_IO_bound(bfqq); + else + bfq_mark_bfqq_IO_bound(bfqq); + + /* + * Keep an observation window of at most 200 ms in the past + * from now. + */ + if (tot_io_time > 200 * NSEC_PER_MSEC) { + bfqq->io_start_time = now_ns - (tot_io_time>>1); + bfqq->tot_idle_time >>= 1; + } +} + +/* + * Detect whether bfqq's I/O seems synchronized with that of some + * other queue, i.e., whether bfqq, after remaining empty, happens to + * receive new I/O only right after some I/O request of the other + * queue has been completed. We call waker queue the other queue, and + * we assume, for simplicity, that bfqq may have at most one waker + * queue. + * + * A remarkable throughput boost can be reached by unconditionally + * injecting the I/O of the waker queue, every time a new + * bfq_dispatch_request happens to be invoked while I/O is being + * plugged for bfqq. In addition to boosting throughput, this + * unblocks bfqq's I/O, thereby improving bandwidth and latency for + * bfqq. Note that these same results may be achieved with the general + * injection mechanism, but less effectively. For details on this + * aspect, see the comments on the choice of the queue for injection + * in bfq_select_queue(). + * + * Turning back to the detection of a waker queue, a queue Q is deemed as a + * waker queue for bfqq if, for three consecutive times, bfqq happens to become + * non empty right after a request of Q has been completed within given + * timeout. In this respect, even if bfqq is empty, we do not check for a waker + * if it still has some in-flight I/O. In fact, in this case bfqq is actually + * still being served by the drive, and may receive new I/O on the completion + * of some of the in-flight requests. In particular, on the first time, Q is + * tentatively set as a candidate waker queue, while on the third consecutive + * time that Q is detected, the field waker_bfqq is set to Q, to confirm that Q + * is a waker queue for bfqq. These detection steps are performed only if bfqq + * has a long think time, so as to make it more likely that bfqq's I/O is + * actually being blocked by a synchronization. This last filter, plus the + * above three-times requirement and time limit for detection, make false + * positives less likely. + * + * NOTE + * + * The sooner a waker queue is detected, the sooner throughput can be + * boosted by injecting I/O from the waker queue. Fortunately, + * detection is likely to be actually fast, for the following + * reasons. While blocked by synchronization, bfqq has a long think + * time. This implies that bfqq's inject limit is at least equal to 1 + * (see the comments in bfq_update_inject_limit()). So, thanks to + * injection, the waker queue is likely to be served during the very + * first I/O-plugging time interval for bfqq. This triggers the first + * step of the detection mechanism. Thanks again to injection, the + * candidate waker queue is then likely to be confirmed no later than + * during the next I/O-plugging interval for bfqq. + * + * ISSUE + * + * On queue merging all waker information is lost. + */ +static void bfq_check_waker(struct bfq_data *bfqd, struct bfq_queue *bfqq, + u64 now_ns) +{ + char waker_name[MAX_BFQQ_NAME_LENGTH]; + + if (!bfqd->last_completed_rq_bfqq || + bfqd->last_completed_rq_bfqq == bfqq || + bfq_bfqq_has_short_ttime(bfqq) || + now_ns - bfqd->last_completion >= 4 * NSEC_PER_MSEC || + bfqd->last_completed_rq_bfqq == &bfqd->oom_bfqq || + bfqq == &bfqd->oom_bfqq) + return; + + /* + * We reset waker detection logic also if too much time has passed + * since the first detection. If wakeups are rare, pointless idling + * doesn't hurt throughput that much. The condition below makes sure + * we do not uselessly idle blocking waker in more than 1/64 cases. + */ + if (bfqd->last_completed_rq_bfqq != + bfqq->tentative_waker_bfqq || + now_ns > bfqq->waker_detection_started + + 128 * (u64)bfqd->bfq_slice_idle) { + /* + * First synchronization detected with a + * candidate waker queue, or with a different + * candidate waker queue from the current one. + */ + bfqq->tentative_waker_bfqq = + bfqd->last_completed_rq_bfqq; + bfqq->num_waker_detections = 1; + bfqq->waker_detection_started = now_ns; + bfq_bfqq_name(bfqq->tentative_waker_bfqq, waker_name, + MAX_BFQQ_NAME_LENGTH); + bfq_log_bfqq(bfqd, bfqq, "set tentative waker %s", waker_name); + } else /* Same tentative waker queue detected again */ + bfqq->num_waker_detections++; + + if (bfqq->num_waker_detections == 3) { + bfqq->waker_bfqq = bfqd->last_completed_rq_bfqq; + bfqq->tentative_waker_bfqq = NULL; + bfq_bfqq_name(bfqq->waker_bfqq, waker_name, + MAX_BFQQ_NAME_LENGTH); + bfq_log_bfqq(bfqd, bfqq, "set waker %s", waker_name); + + /* + * If the waker queue disappears, then + * bfqq->waker_bfqq must be reset. To + * this goal, we maintain in each + * waker queue a list, woken_list, of + * all the queues that reference the + * waker queue through their + * waker_bfqq pointer. When the waker + * queue exits, the waker_bfqq pointer + * of all the queues in the woken_list + * is reset. + * + * In addition, if bfqq is already in + * the woken_list of a waker queue, + * then, before being inserted into + * the woken_list of a new waker + * queue, bfqq must be removed from + * the woken_list of the old waker + * queue. + */ + if (!hlist_unhashed(&bfqq->woken_list_node)) + hlist_del_init(&bfqq->woken_list_node); + hlist_add_head(&bfqq->woken_list_node, + &bfqd->last_completed_rq_bfqq->woken_list); + } +} + static void bfq_add_request(struct request *rq) { struct bfq_queue *bfqq = RQ_BFQQ(rq); @@ -1672,10 +2210,88 @@ static void bfq_add_request(struct request *rq) struct request *next_rq, *prev; unsigned int old_wr_coeff = bfqq->wr_coeff; bool interactive = false; + u64 now_ns = blk_time_get_ns(); bfq_log_bfqq(bfqd, bfqq, "add_request %d", rq_is_sync(rq)); bfqq->queued[rq_is_sync(rq)]++; - bfqd->queued++; + /* + * Updating of 'bfqd->queued' is protected by 'bfqd->lock', however, it + * may be read without holding the lock in bfq_has_work(). + */ + WRITE_ONCE(bfqd->queued, bfqd->queued + 1); + + if (bfq_bfqq_sync(bfqq) && RQ_BIC(rq)->requests <= 1) { + bfq_check_waker(bfqd, bfqq, now_ns); + + /* + * Periodically reset inject limit, to make sure that + * the latter eventually drops in case workload + * changes, see step (3) in the comments on + * bfq_update_inject_limit(). + */ + if (time_is_before_eq_jiffies(bfqq->decrease_time_jif + + msecs_to_jiffies(1000))) + bfq_reset_inject_limit(bfqd, bfqq); + + /* + * The following conditions must hold to setup a new + * sampling of total service time, and then a new + * update of the inject limit: + * - bfqq is in service, because the total service + * time is evaluated only for the I/O requests of + * the queues in service; + * - this is the right occasion to compute or to + * lower the baseline total service time, because + * there are actually no requests in the drive, + * or + * the baseline total service time is available, and + * this is the right occasion to compute the other + * quantity needed to update the inject limit, i.e., + * the total service time caused by the amount of + * injection allowed by the current value of the + * limit. It is the right occasion because injection + * has actually been performed during the service + * hole, and there are still in-flight requests, + * which are very likely to be exactly the injected + * requests, or part of them; + * - the minimum interval for sampling the total + * service time and updating the inject limit has + * elapsed. + */ + if (bfqq == bfqd->in_service_queue && + (bfqd->tot_rq_in_driver == 0 || + (bfqq->last_serv_time_ns > 0 && + bfqd->rqs_injected && bfqd->tot_rq_in_driver > 0)) && + time_is_before_eq_jiffies(bfqq->decrease_time_jif + + msecs_to_jiffies(10))) { + bfqd->last_empty_occupied_ns = blk_time_get_ns(); + /* + * Start the state machine for measuring the + * total service time of rq: setting + * wait_dispatch will cause bfqd->waited_rq to + * be set when rq will be dispatched. + */ + bfqd->wait_dispatch = true; + /* + * If there is no I/O in service in the drive, + * then possible injection occurred before the + * arrival of rq will not affect the total + * service time of rq. So the injection limit + * must not be updated as a function of such + * total service time, unless new injection + * occurs before rq is completed. To have the + * injection limit updated only in the latter + * case, reset rqs_injected here (rqs_injected + * will be set in case injection is performed + * on bfqq before rq is completed). + */ + if (bfqd->tot_rq_in_driver == 0) + bfqd->rqs_injected = false; + } + } + + if (bfq_bfqq_sync(bfqq)) + bfq_update_io_intensity(bfqq, now_ns); elv_rb_add(&bfqq->sort_list, rq); @@ -1688,8 +2304,9 @@ static void bfq_add_request(struct request *rq) /* * Adjust priority tree position, if next_rq changes. + * See comments on bfq_pos_tree_add_move() for the unlikely(). */ - if (prev != bfqq->next_rq) + if (unlikely(!bfqd->nonrot_with_queueing && prev != bfqq->next_rq)) bfq_pos_tree_add_move(bfqd, bfqq); if (!bfq_bfqq_busy(bfqq)) /* switching to busy ... */ @@ -1762,22 +2379,6 @@ static sector_t get_sdist(sector_t last_pos, struct request *rq) return 0; } -#if 0 /* Still not clear if we can do without next two functions */ -static void bfq_activate_request(struct request_queue *q, struct request *rq) -{ - struct bfq_data *bfqd = q->elevator->elevator_data; - - bfqd->rq_in_driver++; -} - -static void bfq_deactivate_request(struct request_queue *q, struct request *rq) -{ - struct bfq_data *bfqd = q->elevator->elevator_data; - - bfqd->rq_in_driver--; -} -#endif - static void bfq_remove_request(struct request_queue *q, struct request *rq) { @@ -1793,7 +2394,11 @@ static void bfq_remove_request(struct request_queue *q, if (rq->queuelist.prev != &rq->queuelist) list_del_init(&rq->queuelist); bfqq->queued[sync]--; - bfqd->queued--; + /* + * Updating of 'bfqd->queued' is protected by 'bfqd->lock', however, it + * may be read without holding the lock in bfq_has_work(). + */ + WRITE_ONCE(bfqd->queued, bfqd->queued - 1); elv_rb_del(&bfqq->sort_list, rq); elv_rqhash_del(q, rq); @@ -1804,7 +2409,7 @@ static void bfq_remove_request(struct request_queue *q, bfqq->next_rq = NULL; if (bfq_bfqq_busy(bfqq) && bfqq != bfqd->in_service_queue) { - bfq_del_bfqq_busy(bfqd, bfqq, false); + bfq_del_bfqq_busy(bfqq, false); /* * bfqq emptied. In normal operation, when * bfqq is empty, bfqq->entity.service and @@ -1829,7 +2434,9 @@ static void bfq_remove_request(struct request_queue *q, bfqq->pos_root = NULL; } } else { - bfq_pos_tree_add_move(bfqd, bfqq); + /* see comments on bfq_pos_tree_add_move() for the unlikely() */ + if (unlikely(!bfqd->nonrot_with_queueing)) + bfq_pos_tree_add_move(bfqd, bfqq); } if (rq->cmd_flags & REQ_META) @@ -1837,34 +2444,35 @@ static void bfq_remove_request(struct request_queue *q, } -static bool bfq_bio_merge(struct blk_mq_hw_ctx *hctx, struct bio *bio) +static bool bfq_bio_merge(struct request_queue *q, struct bio *bio, + unsigned int nr_segs) { - struct request_queue *q = hctx->queue; struct bfq_data *bfqd = q->elevator->elevator_data; + struct bfq_io_cq *bic = bfq_bic_lookup(q); struct request *free = NULL; - /* - * bfq_bic_lookup grabs the queue_lock: invoke it now and - * store its return value for later use, to avoid nesting - * queue_lock inside the bfqd->lock. We assume that the bic - * returned by bfq_bic_lookup does not go away before - * bfqd->lock is taken. - */ - struct bfq_io_cq *bic = bfq_bic_lookup(bfqd, current->io_context, q); bool ret; spin_lock_irq(&bfqd->lock); - if (bic) - bfqd->bio_bfqq = bic_to_bfqq(bic, op_is_sync(bio->bi_opf)); - else + if (bic) { + /* + * Make sure cgroup info is uptodate for current process before + * considering the merge. + */ + bfq_bic_update_cgroup(bic, bio); + + bfqd->bio_bfqq = bic_to_bfqq(bic, op_is_sync(bio->bi_opf), + bfq_actuator_index(bfqd, bio)); + } else { bfqd->bio_bfqq = NULL; + } bfqd->bio_bic = bic; - ret = blk_mq_sched_try_merge(q, bio, &free); + ret = blk_mq_sched_try_merge(q, bio, nr_segs, &free); + spin_unlock_irq(&bfqd->lock); if (free) blk_mq_free_request(free); - spin_unlock_irq(&bfqd->lock); return ret; } @@ -1878,14 +2486,15 @@ static int bfq_request_merge(struct request_queue *q, struct request **req, __rq = bfq_find_rq_fmerge(bfqd, bio, q); if (__rq && elv_bio_merge_ok(__rq, bio)) { *req = __rq; + + if (blk_discard_mergable(__rq)) + return ELEVATOR_DISCARD_MERGE; return ELEVATOR_FRONT_MERGE; } return ELEVATOR_NO_MERGE; } -static struct bfq_queue *bfq_init_rq(struct request *rq); - static void bfq_request_merged(struct request_queue *q, struct request *req, enum elv_merge type) { @@ -1894,10 +2503,15 @@ static void bfq_request_merged(struct request_queue *q, struct request *req, blk_rq_pos(req) < blk_rq_pos(container_of(rb_prev(&req->rb_node), struct request, rb_node))) { - struct bfq_queue *bfqq = bfq_init_rq(req); - struct bfq_data *bfqd = bfqq->bfqd; + struct bfq_queue *bfqq = RQ_BFQQ(req); + struct bfq_data *bfqd; struct request *prev, *next_rq; + if (!bfqq) + return; + + bfqd = bfqq->bfqd; + /* Reposition request in its sort_list */ elv_rb_del(&bfqq->sort_list, req); elv_rb_add(&bfqq->sort_list, req); @@ -1914,7 +2528,12 @@ static void bfq_request_merged(struct request_queue *q, struct request *req, */ if (prev != bfqq->next_rq) { bfq_updated_next_req(bfqd, bfqq); - bfq_pos_tree_add_move(bfqd, bfqq); + /* + * See comments on bfq_pos_tree_add_move() for + * the unlikely(). + */ + if (unlikely(!bfqd->nonrot_with_queueing)) + bfq_pos_tree_add_move(bfqd, bfqq); } } } @@ -1936,8 +2555,11 @@ static void bfq_request_merged(struct request_queue *q, struct request *req, static void bfq_requests_merged(struct request_queue *q, struct request *rq, struct request *next) { - struct bfq_queue *bfqq = bfq_init_rq(rq), - *next_bfqq = bfq_init_rq(next); + struct bfq_queue *bfqq = RQ_BFQQ(rq), + *next_bfqq = RQ_BFQQ(next); + + if (!bfqq) + goto remove; /* * If next and rq belong to the same bfq_queue and next is older @@ -1960,11 +2582,37 @@ static void bfq_requests_merged(struct request_queue *q, struct request *rq, bfqq->next_rq = rq; bfqg_stats_update_io_merged(bfqq_group(bfqq), next->cmd_flags); +remove: + /* Merged request may be in the IO scheduler. Remove it. */ + if (!RB_EMPTY_NODE(&next->rb_node)) { + bfq_remove_request(next->q, next); + if (next_bfqq) + bfqg_stats_update_io_remove(bfqq_group(next_bfqq), + next->cmd_flags); + } } /* Must be called with bfqq != NULL */ static void bfq_bfqq_end_wr(struct bfq_queue *bfqq) { + /* + * If bfqq has been enjoying interactive weight-raising, then + * reset soft_rt_next_start. We do it for the following + * reason. bfqq may have been conveying the I/O needed to load + * a soft real-time application. Such an application actually + * exhibits a soft real-time I/O pattern after it finishes + * loading, and finally starts doing its job. But, if bfqq has + * been receiving a lot of bandwidth so far (likely to happen + * on a fast device), then soft_rt_next_start now contains a + * high value that. So, without this reset, bfqq would be + * prevented from being possibly considered as soft_rt for a + * very long time. + */ + + if (bfqq->wr_cur_max_time != + bfqq->bfqd->bfq_wr_rt_max_time) + bfqq->soft_rt_next_start = jiffies; + if (bfq_bfqq_busy(bfqq)) bfqq->bfqd->wr_busy_queues--; bfqq->wr_coeff = 1; @@ -1980,24 +2628,29 @@ static void bfq_bfqq_end_wr(struct bfq_queue *bfqq) void bfq_end_wr_async_queues(struct bfq_data *bfqd, struct bfq_group *bfqg) { - int i, j; + int i, j, k; - for (i = 0; i < 2; i++) - for (j = 0; j < IOPRIO_BE_NR; j++) - if (bfqg->async_bfqq[i][j]) - bfq_bfqq_end_wr(bfqg->async_bfqq[i][j]); - if (bfqg->async_idle_bfqq) - bfq_bfqq_end_wr(bfqg->async_idle_bfqq); + for (k = 0; k < bfqd->num_actuators; k++) { + for (i = 0; i < 2; i++) + for (j = 0; j < IOPRIO_NR_LEVELS; j++) + if (bfqg->async_bfqq[i][j][k]) + bfq_bfqq_end_wr(bfqg->async_bfqq[i][j][k]); + if (bfqg->async_idle_bfqq[k]) + bfq_bfqq_end_wr(bfqg->async_idle_bfqq[k]); + } } static void bfq_end_wr(struct bfq_data *bfqd) { struct bfq_queue *bfqq; + int i; spin_lock_irq(&bfqd->lock); - list_for_each_entry(bfqq, &bfqd->active_list, bfqq_list) - bfq_bfqq_end_wr(bfqq); + for (i = 0; i < bfqd->num_actuators; i++) { + list_for_each_entry(bfqq, &bfqd->active_list[i], bfqq_list) + bfq_bfqq_end_wr(bfqq); + } list_for_each_entry(bfqq, &bfqd->idle_list, bfqq_list) bfq_bfqq_end_wr(bfqq); bfq_end_wr_async(bfqd); @@ -2024,7 +2677,7 @@ static struct bfq_queue *bfqq_find_close(struct bfq_data *bfqd, struct bfq_queue *bfqq, sector_t sector) { - struct rb_root *root = &bfq_bfqq_to_bfqg(bfqq)->rq_pos_tree; + struct rb_root *root = &bfqq_group(bfqq)->rq_pos_tree; struct rb_node *parent, *node; struct bfq_queue *__bfqq; @@ -2113,6 +2766,14 @@ bfq_setup_merge(struct bfq_queue *bfqq, struct bfq_queue *new_bfqq) if (process_refs == 0 || new_process_refs == 0) return NULL; + /* + * Make sure merged queues belong to the same parent. Parents could + * have changed since the time we decided the two queues are suitable + * for merging. + */ + if (new_bfqq->entity.parent != bfqq->entity.parent) + return NULL; + bfq_log_bfqq(bfqq->bfqd, bfqq, "scheduling merge with queue %d", new_bfqq->pid); @@ -2137,6 +2798,15 @@ bfq_setup_merge(struct bfq_queue *bfqq, struct bfq_queue *new_bfqq) * are likely to increase the throughput. */ bfqq->new_bfqq = new_bfqq; + /* + * The above assignment schedules the following redirections: + * each time some I/O for bfqq arrives, the process that + * generated that I/O is disassociated from bfqq and + * associated with new_bfqq. Here we increases new_bfqq->ref + * in advance, adding the number of processes that are + * expected to be associated with new_bfqq as they happen to + * issue I/O. + */ new_bfqq->ref += process_refs; return new_bfqq; } @@ -2170,6 +2840,43 @@ static bool bfq_may_be_close_cooperator(struct bfq_queue *bfqq, return true; } +static bool idling_boosts_thr_without_issues(struct bfq_data *bfqd, + struct bfq_queue *bfqq); + +static struct bfq_queue * +bfq_setup_stable_merge(struct bfq_data *bfqd, struct bfq_queue *bfqq, + struct bfq_queue *stable_merge_bfqq, + struct bfq_iocq_bfqq_data *bfqq_data) +{ + int proc_ref = min(bfqq_process_refs(bfqq), + bfqq_process_refs(stable_merge_bfqq)); + struct bfq_queue *new_bfqq = NULL; + + bfqq_data->stable_merge_bfqq = NULL; + if (idling_boosts_thr_without_issues(bfqd, bfqq) || proc_ref == 0) + goto out; + + /* next function will take at least one ref */ + new_bfqq = bfq_setup_merge(bfqq, stable_merge_bfqq); + + if (new_bfqq) { + bfqq_data->stably_merged = true; + if (new_bfqq->bic) { + unsigned int new_a_idx = new_bfqq->actuator_idx; + struct bfq_iocq_bfqq_data *new_bfqq_data = + &new_bfqq->bic->bfqq_data[new_a_idx]; + + new_bfqq_data->stably_merged = true; + } + } + +out: + /* deschedule stable merge, because done or aborted here */ + bfq_put_stable_ref(stable_merge_bfqq); + + return new_bfqq; +} + /* * Attempt to schedule a merge of bfqq with the currently in-service * queue or with a close queue among the scheduled queues. Return @@ -2192,9 +2899,90 @@ static bool bfq_may_be_close_cooperator(struct bfq_queue *bfqq, */ static struct bfq_queue * bfq_setup_cooperator(struct bfq_data *bfqd, struct bfq_queue *bfqq, - void *io_struct, bool request) + void *io_struct, bool request, struct bfq_io_cq *bic) { struct bfq_queue *in_service_bfqq, *new_bfqq; + unsigned int a_idx = bfqq->actuator_idx; + struct bfq_iocq_bfqq_data *bfqq_data = &bic->bfqq_data[a_idx]; + + /* if a merge has already been setup, then proceed with that first */ + new_bfqq = bfqq->new_bfqq; + if (new_bfqq) { + while (new_bfqq->new_bfqq) + new_bfqq = new_bfqq->new_bfqq; + return new_bfqq; + } + + /* + * Check delayed stable merge for rotational or non-queueing + * devs. For this branch to be executed, bfqq must not be + * currently merged with some other queue (i.e., bfqq->bic + * must be non null). If we considered also merged queues, + * then we should also check whether bfqq has already been + * merged with bic->stable_merge_bfqq. But this would be + * costly and complicated. + */ + if (unlikely(!bfqd->nonrot_with_queueing)) { + /* + * Make sure also that bfqq is sync, because + * bic->stable_merge_bfqq may point to some queue (for + * stable merging) also if bic is associated with a + * sync queue, but this bfqq is async + */ + if (bfq_bfqq_sync(bfqq) && bfqq_data->stable_merge_bfqq && + !bfq_bfqq_just_created(bfqq) && + time_is_before_jiffies(bfqq->split_time + + msecs_to_jiffies(bfq_late_stable_merging)) && + time_is_before_jiffies(bfqq->creation_time + + msecs_to_jiffies(bfq_late_stable_merging))) { + struct bfq_queue *stable_merge_bfqq = + bfqq_data->stable_merge_bfqq; + + return bfq_setup_stable_merge(bfqd, bfqq, + stable_merge_bfqq, + bfqq_data); + } + } + + /* + * Do not perform queue merging if the device is non + * rotational and performs internal queueing. In fact, such a + * device reaches a high speed through internal parallelism + * and pipelining. This means that, to reach a high + * throughput, it must have many requests enqueued at the same + * time. But, in this configuration, the internal scheduling + * algorithm of the device does exactly the job of queue + * merging: it reorders requests so as to obtain as much as + * possible a sequential I/O pattern. As a consequence, with + * the workload generated by processes doing interleaved I/O, + * the throughput reached by the device is likely to be the + * same, with and without queue merging. + * + * Disabling merging also provides a remarkable benefit in + * terms of throughput. Merging tends to make many workloads + * artificially more uneven, because of shared queues + * remaining non empty for incomparably more time than + * non-merged queues. This may accentuate workload + * asymmetries. For example, if one of the queues in a set of + * merged queues has a higher weight than a normal queue, then + * the shared queue may inherit such a high weight and, by + * staying almost always active, may force BFQ to perform I/O + * plugging most of the time. This evidently makes it harder + * for BFQ to let the device reach a high throughput. + * + * Finally, the likely() macro below is not used because one + * of the two branches is more likely than the other, but to + * have the code path after the following if() executed as + * fast as possible for the case of a non rotational device + * with queueing. We want it because this is the fastest kind + * of device. On the opposite end, the likely() may lengthen + * the execution time of BFQ for the case of slower devices + * (rotational or at least without queueing). But in this case + * the execution time of BFQ matters very little, if not at + * all. + */ + if (likely(bfqd->nonrot_with_queueing)) + return NULL; /* * Prevent bfqq from being merged if it has been created too @@ -2210,21 +2998,19 @@ bfq_setup_cooperator(struct bfq_data *bfqd, struct bfq_queue *bfqq, if (bfq_too_late_for_merging(bfqq)) return NULL; - if (bfqq->new_bfqq) - return bfqq->new_bfqq; - if (!io_struct || unlikely(bfqq == &bfqd->oom_bfqq)) return NULL; /* If there is only one backlogged queue, don't search. */ - if (bfqd->busy_queues == 1) + if (bfq_tot_busy_queues(bfqd) == 1) return NULL; in_service_bfqq = bfqd->in_service_queue; if (in_service_bfqq && in_service_bfqq != bfqq && likely(in_service_bfqq != &bfqd->oom_bfqq) && - bfq_rq_close_to_sector(io_struct, request, bfqd->last_position) && + bfq_rq_close_to_sector(io_struct, request, + bfqd->in_serv_last_pos) && bfqq->entity.parent == in_service_bfqq->entity.parent && bfq_may_be_close_cooperator(bfqq, in_service_bfqq)) { new_bfqq = bfq_setup_merge(bfqq, in_service_bfqq); @@ -2249,6 +3035,8 @@ bfq_setup_cooperator(struct bfq_data *bfqd, struct bfq_queue *bfqq, static void bfq_bfqq_save_state(struct bfq_queue *bfqq) { struct bfq_io_cq *bic = bfqq->bic; + unsigned int a_idx = bfqq->actuator_idx; + struct bfq_iocq_bfqq_data *bfqq_data = &bic->bfqq_data[a_idx]; /* * If !bfqq->bic, the queue is already shared or its requests @@ -2258,11 +3046,21 @@ static void bfq_bfqq_save_state(struct bfq_queue *bfqq) if (!bic) return; - bic->saved_ttime = bfqq->ttime; - bic->saved_has_short_ttime = bfq_bfqq_has_short_ttime(bfqq); - bic->saved_IO_bound = bfq_bfqq_IO_bound(bfqq); - bic->saved_in_large_burst = bfq_bfqq_in_large_burst(bfqq); - bic->was_in_burst_list = !hlist_unhashed(&bfqq->burst_list_node); + bfqq_data->saved_last_serv_time_ns = bfqq->last_serv_time_ns; + bfqq_data->saved_inject_limit = bfqq->inject_limit; + bfqq_data->saved_decrease_time_jif = bfqq->decrease_time_jif; + + bfqq_data->saved_weight = bfqq->entity.orig_weight; + bfqq_data->saved_ttime = bfqq->ttime; + bfqq_data->saved_has_short_ttime = + bfq_bfqq_has_short_ttime(bfqq); + bfqq_data->saved_IO_bound = bfq_bfqq_IO_bound(bfqq); + bfqq_data->saved_io_start_time = bfqq->io_start_time; + bfqq_data->saved_tot_idle_time = bfqq->tot_idle_time; + bfqq_data->saved_in_large_burst = bfq_bfqq_in_large_burst(bfqq); + bfqq_data->was_in_burst_list = + !hlist_unhashed(&bfqq->burst_list_node); + if (unlikely(bfq_bfqq_just_created(bfqq) && !bfq_bfqq_in_large_burst(bfqq) && bfqq->bfqd->low_latency)) { @@ -2275,22 +3073,63 @@ static void bfq_bfqq_save_state(struct bfq_queue *bfqq) * to bfqq, so that to avoid that bfqq unjustly fails * to enjoy weight raising if split soon. */ - bic->saved_wr_coeff = bfqq->bfqd->bfq_wr_coeff; - bic->saved_wr_cur_max_time = bfq_wr_duration(bfqq->bfqd); - bic->saved_last_wr_start_finish = jiffies; + bfqq_data->saved_wr_coeff = bfqq->bfqd->bfq_wr_coeff; + bfqq_data->saved_wr_start_at_switch_to_srt = + bfq_smallest_from_now(); + bfqq_data->saved_wr_cur_max_time = + bfq_wr_duration(bfqq->bfqd); + bfqq_data->saved_last_wr_start_finish = jiffies; } else { - bic->saved_wr_coeff = bfqq->wr_coeff; - bic->saved_wr_start_at_switch_to_srt = + bfqq_data->saved_wr_coeff = bfqq->wr_coeff; + bfqq_data->saved_wr_start_at_switch_to_srt = bfqq->wr_start_at_switch_to_srt; - bic->saved_last_wr_start_finish = bfqq->last_wr_start_finish; - bic->saved_wr_cur_max_time = bfqq->wr_cur_max_time; + bfqq_data->saved_service_from_wr = + bfqq->service_from_wr; + bfqq_data->saved_last_wr_start_finish = + bfqq->last_wr_start_finish; + bfqq_data->saved_wr_cur_max_time = bfqq->wr_cur_max_time; } } -static void -bfq_merge_bfqqs(struct bfq_data *bfqd, struct bfq_io_cq *bic, - struct bfq_queue *bfqq, struct bfq_queue *new_bfqq) + +void bfq_reassign_last_bfqq(struct bfq_queue *cur_bfqq, + struct bfq_queue *new_bfqq) +{ + if (cur_bfqq->entity.parent && + cur_bfqq->entity.parent->last_bfqq_created == cur_bfqq) + cur_bfqq->entity.parent->last_bfqq_created = new_bfqq; + else if (cur_bfqq->bfqd && cur_bfqq->bfqd->last_bfqq_created == cur_bfqq) + cur_bfqq->bfqd->last_bfqq_created = new_bfqq; +} + +void bfq_release_process_ref(struct bfq_data *bfqd, struct bfq_queue *bfqq) +{ + /* + * To prevent bfqq's service guarantees from being violated, + * bfqq may be left busy, i.e., queued for service, even if + * empty (see comments in __bfq_bfqq_expire() for + * details). But, if no process will send requests to bfqq any + * longer, then there is no point in keeping bfqq queued for + * service. In addition, keeping bfqq queued for service, but + * with no process ref any longer, may have caused bfqq to be + * freed when dequeued from service. But this is assumed to + * never happen. + */ + if (bfq_bfqq_busy(bfqq) && RB_EMPTY_ROOT(&bfqq->sort_list) && + bfqq != bfqd->in_service_queue) + bfq_del_bfqq_busy(bfqq, false); + + bfq_reassign_last_bfqq(bfqq, NULL); + + bfq_put_queue(bfqq); +} + +static struct bfq_queue *bfq_merge_bfqqs(struct bfq_data *bfqd, + struct bfq_io_cq *bic, + struct bfq_queue *bfqq) { + struct bfq_queue *new_bfqq = bfqq->new_bfqq; + bfq_log_bfqq(bfqd, bfqq, "merging with queue %lu", (unsigned long)new_bfqq->pid); /* Save weight raising and idle window of the merged queues */ @@ -2301,6 +3140,29 @@ bfq_merge_bfqqs(struct bfq_data *bfqd, struct bfq_io_cq *bic, bfq_clear_bfqq_IO_bound(bfqq); /* + * The processes associated with bfqq are cooperators of the + * processes associated with new_bfqq. So, if bfqq has a + * waker, then assume that all these processes will be happy + * to let bfqq's waker freely inject I/O when they have no + * I/O. + */ + if (bfqq->waker_bfqq && !new_bfqq->waker_bfqq && + bfqq->waker_bfqq != new_bfqq) { + new_bfqq->waker_bfqq = bfqq->waker_bfqq; + new_bfqq->tentative_waker_bfqq = NULL; + + /* + * If the waker queue disappears, then + * new_bfqq->waker_bfqq must be reset. So insert + * new_bfqq into the woken_list of the waker. See + * bfq_check_waker for details. + */ + hlist_add_head(&new_bfqq->woken_list_node, + &new_bfqq->waker_bfqq->woken_list); + + } + + /* * If bfqq is weight-raised, then let new_bfqq inherit * weight-raising. To reduce false positives, neglect the case * where bfqq has just been created, but has not yet made it @@ -2333,7 +3195,7 @@ bfq_merge_bfqqs(struct bfq_data *bfqd, struct bfq_io_cq *bic, /* * Merge queues (that is, let bic redirect its requests to new_bfqq) */ - bic_set_bfqq(bic, new_bfqq, 1); + bic_set_bfqq(bic, new_bfqq, true, bfqq->actuator_idx); bfq_mark_bfqq_coop(new_bfqq); /* * new_bfqq now belongs to at least two bics (it is a shared queue): @@ -2346,9 +3208,23 @@ bfq_merge_bfqqs(struct bfq_data *bfqd, struct bfq_io_cq *bic, * assignment causes no harm). */ new_bfqq->bic = NULL; + /* + * If the queue is shared, the pid is the pid of one of the associated + * processes. Which pid depends on the exact sequence of merge events + * the queue underwent. So printing such a pid is useless and confusing + * because it reports a random pid between those of the associated + * processes. + * We mark such a queue with a pid -1, and then print SHARED instead of + * a pid in logging messages. + */ + new_bfqq->pid = -1; bfqq->bic = NULL; - /* release process reference to bfqq */ - bfq_put_queue(bfqq); + + bfq_reassign_last_bfqq(bfqq, new_bfqq); + + bfq_release_process_ref(bfqd, bfqq); + + return new_bfqq; } static bool bfq_allow_bio_merge(struct request_queue *q, struct request *rq, @@ -2375,23 +3251,17 @@ static bool bfq_allow_bio_merge(struct request_queue *q, struct request *rq, * We take advantage of this function to perform an early merge * of the queues of possible cooperating processes. */ - new_bfqq = bfq_setup_cooperator(bfqd, bfqq, bio, false); + new_bfqq = bfq_setup_cooperator(bfqd, bfqq, bio, false, bfqd->bio_bic); if (new_bfqq) { /* * bic still points to bfqq, then it has not yet been * redirected to some other bfq_queue, and a queue - * merge beween bfqq and new_bfqq can be safely - * fulfillled, i.e., bic can be redirected to new_bfqq + * merge between bfqq and new_bfqq can be safely + * fulfilled, i.e., bic can be redirected to new_bfqq * and bfqq can be put. */ - bfq_merge_bfqqs(bfqd, bfqd->bio_bic, bfqq, - new_bfqq); - /* - * If we get here, bio will be queued into new_queue, - * so use new_bfqq to decide whether bio and rq can be - * merged. - */ - bfqq = new_bfqq; + while (bfqq != new_bfqq) + bfqq = bfq_merge_bfqqs(bfqd, bfqd->bio_bic, bfqq); /* * Change also bqfd->bio_bfqq, as @@ -2421,7 +3291,7 @@ static void bfq_set_budget_timeout(struct bfq_data *bfqd, else timeout_coeff = bfqq->entity.weight / bfqq->entity.orig_weight; - bfqd->last_budget_start = ktime_get(); + bfqd->last_budget_start = blk_time_get(); bfqq->budget_timeout = jiffies + bfqd->bfq_timeout * timeout_coeff; @@ -2478,6 +3348,7 @@ static void __bfq_set_in_service_queue(struct bfq_data *bfqd, } bfqd->in_service_queue = bfqq; + bfqd->in_serv_last_pos = 0; } /* @@ -2515,10 +3386,14 @@ static void bfq_arm_slice_timer(struct bfq_data *bfqd) * queue). */ if (BFQQ_SEEKY(bfqq) && bfqq->wr_coeff == 1 && - bfq_symmetric_scenario(bfqd)) + !bfq_asymmetric_scenario(bfqd, bfqq)) sl = min_t(u64, sl, BFQ_MIN_TT); + else if (bfqq->wr_coeff > 1) + sl = max_t(u32, sl, 20ULL * NSEC_PER_MSEC); + + bfqd->last_idling_start = blk_time_get(); + bfqd->last_idling_start_jiffies = jiffies; - bfqd->last_idling_start = ktime_get(); hrtimer_start(&bfqd->idle_slice_timer, ns_to_ktime(sl), HRTIMER_MODE_REL); bfqg_stats_set_start_idle_time(bfqq_group(bfqq)); @@ -2555,7 +3430,7 @@ static void bfq_reset_rate_computation(struct bfq_data *bfqd, struct request *rq) { if (rq != NULL) { /* new rq dispatch now, reset accordingly */ - bfqd->last_dispatch = bfqd->first_dispatch = ktime_get_ns(); + bfqd->last_dispatch = bfqd->first_dispatch = blk_time_get_ns(); bfqd->peak_rate_samples = 1; bfqd->sequential_samples = 0; bfqd->tot_sectors_dispatched = bfqd->last_rq_max_size = @@ -2712,7 +3587,7 @@ reset_computation: */ static void bfq_update_peak_rate(struct bfq_data *bfqd, struct request *rq) { - u64 now_ns = ktime_get_ns(); + u64 now_ns = blk_time_get_ns(); if (bfqd->peak_rate_samples == 0) { /* first dispatch */ bfq_log(bfqd, "update_peak_rate: goto reset, samples %d", @@ -2734,15 +3609,15 @@ static void bfq_update_peak_rate(struct bfq_data *bfqd, struct request *rq) * - start a new observation interval with this dispatch */ if (now_ns - bfqd->last_dispatch > 100*NSEC_PER_MSEC && - bfqd->rq_in_driver == 0) + bfqd->tot_rq_in_driver == 0) goto update_rate_and_reset; /* Update sampling information */ bfqd->peak_rate_samples++; - if ((bfqd->rq_in_driver > 0 || + if ((bfqd->tot_rq_in_driver > 0 || now_ns - bfqd->last_completion < BFQ_MIN_TT) - && get_sdist(bfqd->last_position, rq) < BFQQ_SEEK_THR) + && !BFQ_RQ_SEEKY(bfqd, bfqd->last_position, rq)) bfqd->sequential_samples++; bfqd->tot_sectors_dispatched += blk_rq_sectors(rq); @@ -2764,6 +3639,8 @@ update_rate_and_reset: bfq_update_rate_reset(bfqd, rq); update_last_values: bfqd->last_position = blk_rq_pos(rq) + blk_rq_sectors(rq); + if (RQ_BFQQ(rq) == bfqd->in_service_queue) + bfqd->in_serv_last_pos = bfqd->last_position; bfqd->last_dispatch = now_ns; } @@ -2792,7 +3669,225 @@ static void bfq_dispatch_remove(struct request_queue *q, struct request *rq) bfq_remove_request(q, rq); } -static void __bfq_bfqq_expire(struct bfq_data *bfqd, struct bfq_queue *bfqq) +/* + * There is a case where idling does not have to be performed for + * throughput concerns, but to preserve the throughput share of + * the process associated with bfqq. + * + * To introduce this case, we can note that allowing the drive + * to enqueue more than one request at a time, and hence + * delegating de facto final scheduling decisions to the + * drive's internal scheduler, entails loss of control on the + * actual request service order. In particular, the critical + * situation is when requests from different processes happen + * to be present, at the same time, in the internal queue(s) + * of the drive. In such a situation, the drive, by deciding + * the service order of the internally-queued requests, does + * determine also the actual throughput distribution among + * these processes. But the drive typically has no notion or + * concern about per-process throughput distribution, and + * makes its decisions only on a per-request basis. Therefore, + * the service distribution enforced by the drive's internal + * scheduler is likely to coincide with the desired throughput + * distribution only in a completely symmetric, or favorably + * skewed scenario where: + * (i-a) each of these processes must get the same throughput as + * the others, + * (i-b) in case (i-a) does not hold, it holds that the process + * associated with bfqq must receive a lower or equal + * throughput than any of the other processes; + * (ii) the I/O of each process has the same properties, in + * terms of locality (sequential or random), direction + * (reads or writes), request sizes, greediness + * (from I/O-bound to sporadic), and so on; + + * In fact, in such a scenario, the drive tends to treat the requests + * of each process in about the same way as the requests of the + * others, and thus to provide each of these processes with about the + * same throughput. This is exactly the desired throughput + * distribution if (i-a) holds, or, if (i-b) holds instead, this is an + * even more convenient distribution for (the process associated with) + * bfqq. + * + * In contrast, in any asymmetric or unfavorable scenario, device + * idling (I/O-dispatch plugging) is certainly needed to guarantee + * that bfqq receives its assigned fraction of the device throughput + * (see [1] for details). + * + * The problem is that idling may significantly reduce throughput with + * certain combinations of types of I/O and devices. An important + * example is sync random I/O on flash storage with command + * queueing. So, unless bfqq falls in cases where idling also boosts + * throughput, it is important to check conditions (i-a), i(-b) and + * (ii) accurately, so as to avoid idling when not strictly needed for + * service guarantees. + * + * Unfortunately, it is extremely difficult to thoroughly check + * condition (ii). And, in case there are active groups, it becomes + * very difficult to check conditions (i-a) and (i-b) too. In fact, + * if there are active groups, then, for conditions (i-a) or (i-b) to + * become false 'indirectly', it is enough that an active group + * contains more active processes or sub-groups than some other active + * group. More precisely, for conditions (i-a) or (i-b) to become + * false 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_asymmetric_scenario(). + * + * 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-a), (i-b) or (ii). So the device is idled for bfqq. + * This behavior matches also the fact that groups are created + * exactly if controlling I/O is a primary concern (to + * preserve bandwidth and latency guarantees). + * + * On the opposite end, if there are no groups with requests waiting + * for completion, then only conditions (i-a) and (i-b) are actually + * controlled, i.e., provided that conditions (i-a) or (i-b) holds, + * idling is not performed, regardless of whether condition (ii) + * holds. In other words, only if conditions (i-a) and (i-b) do 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 conditions (i-a) and (i-b) 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. + * However, for queues with the same weight, a further + * mechanism, preemption, mitigates or even eliminates this + * problem. And it does so without consequences on overall + * throughput. This mechanism and its benefits are explained + * in the next three paragraphs. + * + * Even if a queue, say Q, is expired when it remains idle, Q + * can still preempt the new in-service queue if the next + * request of Q arrives soon (see the comments on + * bfq_bfqq_update_budg_for_activation). If all queues and + * groups have the same weight, this form of preemption, + * combined with the hole-recovery heuristic described in the + * comments on function bfq_bfqq_update_budg_for_activation, + * are enough to preserve a correct bandwidth distribution in + * the mid term, even without idling. In fact, even if not + * idling allows the internal queues of the device to contain + * many requests, and thus to reorder requests, we can rather + * safely assume that the internal scheduler still preserves a + * minimum of mid-term fairness. + * + * More precisely, this preemption-based, idleless approach + * provides fairness in terms of IOPS, and not sectors per + * second. This can be seen with a simple example. Suppose + * that there are two queues with the same weight, but that + * the first queue receives requests of 8 sectors, while the + * second queue receives requests of 1024 sectors. In + * addition, suppose that each of the two queues contains at + * most one request at a time, which implies that each queue + * always remains idle after it is served. Finally, after + * remaining idle, each queue receives very quickly a new + * request. It follows that the two queues are served + * alternatively, preempting each other if needed. This + * implies that, although both queues have the same weight, + * the queue with large requests receives a service that is + * 1024/8 times as high as the service received by the other + * queue. + * + * The motivation for using preemption instead of idling (for + * queues with the same weight) is that, by not idling, + * service guarantees are preserved (completely or at least in + * part) without minimally sacrificing throughput. And, if + * there is no active group, then the primary expectation for + * this device is probably a high throughput. + * + * We are now left only with explaining the two sub-conditions in the + * additional compound condition that is checked below for deciding + * whether the scenario is asymmetric. To explain the first + * sub-condition, we need to add that the function + * bfq_asymmetric_scenario checks the weights of only + * non-weight-raised queues, for efficiency reasons (see comments on + * bfq_weights_tree_add()). Then the fact that 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 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. + * + * The second sub-condition checked in the compound condition is + * whether there is a fair amount of already in-flight I/O not + * belonging to bfqq. If so, I/O dispatching is to be plugged, for the + * following reason. The drive may decide to serve in-flight + * non-bfqq's I/O requests before bfqq's ones, thereby delaying the + * arrival of new I/O requests for bfqq (recall that bfqq is sync). If + * I/O-dispatching is not plugged, then, while bfqq remains empty, a + * basically uncontrolled amount of I/O from other queues may be + * dispatched too, possibly causing the service of bfqq's I/O to be + * delayed even longer in the drive. This problem gets more and more + * serious as the speed and the queue depth of the drive grow, + * because, as these two quantities grow, the probability to find no + * queue busy but many requests in flight grows too. By contrast, + * plugging I/O dispatching minimizes the delay induced by already + * in-flight I/O, and enables bfqq to recover the bandwidth it may + * lose because of this delay. + * + * As a side note, it is worth considering that the above + * device-idling countermeasures may however fail in the following + * unlucky scenario: if I/O-dispatch plugging is (correctly) disabled + * in a time period during which all symmetry sub-conditions hold, and + * therefore the device is allowed to enqueue many requests, but at + * some later point in time some sub-condition stops to hold, then it + * may become impossible to make requests be served in the desired + * order until all the requests already queued in the device have been + * served. The last sub-condition commented above somewhat mitigates + * this problem for weight-raised queues. + * + * However, as an additional mitigation for this problem, we preserve + * plugging for a special symmetric case that may suddenly turn into + * asymmetric: the case where only bfqq is busy. In this case, not + * expiring bfqq does not cause any harm to any other queues in terms + * of service guarantees. In contrast, it avoids the following unlucky + * sequence of events: (1) bfqq is expired, (2) a new queue with a + * lower weight than bfqq becomes busy (or more queues), (3) the new + * queue is served until a new request arrives for bfqq, (4) when bfqq + * is finally served, there are so many requests of the new queue in + * the drive that the pending requests for bfqq take a lot of time to + * be served. In particular, event (2) may case even already + * dispatched requests of bfqq to be delayed, inside the drive. So, to + * avoid this series of events, the scenario is preventively declared + * as asymmetric also if bfqq is the only busy queues + */ +static bool idling_needed_for_service_guarantees(struct bfq_data *bfqd, + struct bfq_queue *bfqq) +{ + int tot_busy_queues = bfq_tot_busy_queues(bfqd); + + /* No point in idling for bfqq if it won't get requests any longer */ + if (unlikely(!bfqq_process_refs(bfqq))) + return false; + + return (bfqq->wr_coeff > 1 && + (bfqd->wr_busy_queues < tot_busy_queues || + bfqd->tot_rq_in_driver >= bfqq->dispatched + 4)) || + bfq_asymmetric_scenario(bfqd, bfqq) || + tot_busy_queues == 1; +} + +static bool __bfq_bfqq_expire(struct bfq_data *bfqd, struct bfq_queue *bfqq, + enum bfqq_expiration reason) { /* * If this bfqq is shared between multiple processes, check @@ -2803,7 +3898,22 @@ static void __bfq_bfqq_expire(struct bfq_data *bfqd, struct bfq_queue *bfqq) if (bfq_bfqq_coop(bfqq) && BFQQ_SEEKY(bfqq)) bfq_mark_bfqq_split_coop(bfqq); - if (RB_EMPTY_ROOT(&bfqq->sort_list)) { + /* + * Consider queues with a higher finish virtual time than + * bfqq. If idling_needed_for_service_guarantees(bfqq) returns + * true, then bfqq's bandwidth would be violated if an + * uncontrolled amount of I/O from these queues were + * dispatched while bfqq is waiting for its new I/O to + * arrive. This is exactly what may happen if this is a forced + * expiration caused by a preemption attempt, and if bfqq is + * not re-scheduled. To prevent this from happening, re-queue + * bfqq if it needs I/O-dispatch plugging, even if it is + * empty. By doing so, bfqq is granted to be served before the + * above queues (provided that bfqq is of course eligible). + */ + if (RB_EMPTY_ROOT(&bfqq->sort_list) && + !(reason == BFQQE_PREEMPTED && + idling_needed_for_service_guarantees(bfqd, bfqq))) { if (bfqq->dispatched == 0) /* * Overloading budget_timeout field to store @@ -2813,21 +3923,26 @@ static void __bfq_bfqq_expire(struct bfq_data *bfqd, struct bfq_queue *bfqq) */ bfqq->budget_timeout = jiffies; - bfq_del_bfqq_busy(bfqd, bfqq, true); + bfq_del_bfqq_busy(bfqq, true); } else { bfq_requeue_bfqq(bfqd, bfqq, true); /* * Resort priority tree of potential close cooperators. + * See comments on bfq_pos_tree_add_move() for the unlikely(). */ - bfq_pos_tree_add_move(bfqd, bfqq); + if (unlikely(!bfqd->nonrot_with_queueing && + !RB_EMPTY_ROOT(&bfqq->sort_list))) + bfq_pos_tree_add_move(bfqd, bfqq); } /* * All in-service entities must have been properly deactivated * or requeued before executing the next function, which - * resets all in-service entites as no more in service. + * resets all in-service entities as no more in service. This + * may cause bfqq to be freed. If this happens, the next + * function returns true. */ - __bfq_bfqd_reset_in_service(bfqd); + return __bfq_bfqd_reset_in_service(bfqd); } /** @@ -3032,8 +4147,7 @@ static void __bfq_bfqq_recalc_budget(struct bfq_data *bfqd, * function to evaluate the I/O speed of a process. */ static bool bfq_bfqq_is_slow(struct bfq_data *bfqd, struct bfq_queue *bfqq, - bool compensate, enum bfqq_expiration reason, - unsigned long *delta_ms) + bool compensate, unsigned long *delta_ms) { ktime_t delta_ktime; u32 delta_usecs; @@ -3045,7 +4159,7 @@ static bool bfq_bfqq_is_slow(struct bfq_data *bfqd, struct bfq_queue *bfqq, if (compensate) delta_ktime = bfqd->last_idling_start; else - delta_ktime = ktime_get(); + delta_ktime = blk_time_get(); delta_ktime = ktime_sub(delta_ktime, bfqd->last_budget_start); delta_usecs = ktime_to_us(delta_ktime); @@ -3191,13 +4305,6 @@ static unsigned long bfq_bfqq_softrt_next_start(struct bfq_data *bfqd, jiffies + nsecs_to_jiffies(bfqq->bfqd->bfq_slice_idle) + 4); } -static bool bfq_bfqq_injectable(struct bfq_queue *bfqq) -{ - return BFQQ_SEEKY(bfqq) && bfqq->wr_coeff == 1 && - blk_queue_nonrot(bfqq->bfqd->queue) && - bfqq->bfqd->hw_tag; -} - /** * bfq_bfqq_expire - expire a queue. * @bfqd: device owning the queue. @@ -3232,12 +4339,11 @@ void bfq_bfqq_expire(struct bfq_data *bfqd, bool slow; unsigned long delta = 0; struct bfq_entity *entity = &bfqq->entity; - int ref; /* * Check whether the process is slow (see bfq_bfqq_is_slow). */ - slow = bfq_bfqq_is_slow(bfqd, bfqq, compensate, reason, &delta); + slow = bfq_bfqq_is_slow(bfqd, bfqq, compensate, &delta); /* * As above explained, charge slow (typically seeky) and @@ -3260,10 +4366,6 @@ void bfq_bfqq_expire(struct bfq_data *bfqd, bfq_bfqq_budget_left(bfqq) >= entity->budget / 3))) bfq_bfqq_charge_time(bfqd, bfqq, delta); - if (reason == BFQQE_TOO_IDLE && - entity->service <= 2 * entity->budget / 10) - bfq_clear_bfqq_IO_bound(bfqq); - if (bfqd->low_latency && bfqq->wr_coeff == 1) bfqq->last_wr_start_finish = jiffies; @@ -3273,17 +4375,18 @@ void bfq_bfqq_expire(struct bfq_data *bfqd, * If we get here, and there are no outstanding * requests, then the request pattern is isochronous * (see the comments on the function - * bfq_bfqq_softrt_next_start()). Thus we can compute - * soft_rt_next_start. If, instead, the queue still - * has outstanding requests, then we have to wait for - * the completion of all the outstanding requests to - * discover whether the request pattern is actually - * isochronous. + * bfq_bfqq_softrt_next_start()). Therefore we can + * compute soft_rt_next_start. + * + * If, instead, the queue still has outstanding + * requests, then we have to wait for the completion + * of all the outstanding requests to discover whether + * the request pattern is actually isochronous. */ if (bfqq->dispatched == 0) bfqq->soft_rt_next_start = bfq_bfqq_softrt_next_start(bfqd, bfqq); - else { + else if (bfqq->dispatched > 0) { /* * Schedule an update of soft_rt_next_start to when * the task may be discovered to be isochronous. @@ -3297,18 +4400,22 @@ void bfq_bfqq_expire(struct bfq_data *bfqd, slow, bfqq->dispatched, bfq_bfqq_has_short_ttime(bfqq)); /* + * bfqq expired, so no total service time needs to be computed + * any longer: reset state machine for measuring total service + * times. + */ + bfqd->rqs_injected = bfqd->wait_dispatch = false; + bfqd->waited_rq = NULL; + + /* * Increase, decrease or leave budget unchanged according to * reason. */ __bfq_bfqq_recalc_budget(bfqd, bfqq, reason); - ref = bfqq->ref; - __bfq_bfqq_expire(bfqd, bfqq); - - if (ref == 1) /* bfqq is gone, no more actions on it */ + if (__bfq_bfqq_expire(bfqd, bfqq, reason)) + /* bfqq is gone, no more actions on it */ return; - bfqq->injected_service = 0; - /* mark bfqq as waiting a request only if a bic still points to it */ if (!bfq_bfqq_busy(bfqq) && reason != BFQQE_BUDGET_TIMEOUT && @@ -3376,52 +4483,16 @@ static bool bfq_may_expire_for_budg_timeout(struct bfq_queue *bfqq) bfq_bfqq_budget_timeout(bfqq); } -/* - * For a queue that becomes empty, device idling is allowed only if - * this function returns true for the queue. As a consequence, since - * device idling plays a critical role in both throughput boosting and - * service guarantees, the return value of this function plays a - * critical role in both these aspects as well. - * - * In a nutshell, this function returns true only if idling is - * beneficial for throughput or, even if detrimental for throughput, - * idling is however necessary to preserve service guarantees (low - * latency, desired throughput distribution, ...). In particular, on - * NCQ-capable devices, this function tries to return false, so as to - * help keep the drives' internal queues full, whenever this helps the - * device boost the throughput without causing any service-guarantee - * issue. - * - * In more detail, the return value of this function is obtained by, - * first, computing a number of boolean variables that take into - * account throughput and service-guarantee issues, and, then, - * combining these variables in a logical expression. Most of the - * issues taken into account are not trivial. We discuss these issues - * individually while introducing the variables. - */ -static bool bfq_better_to_idle(struct bfq_queue *bfqq) +static bool idling_boosts_thr_without_issues(struct bfq_data *bfqd, + struct bfq_queue *bfqq) { - struct bfq_data *bfqd = bfqq->bfqd; bool rot_without_queueing = !blk_queue_nonrot(bfqd->queue) && !bfqd->hw_tag, bfqq_sequential_and_IO_bound, - idling_boosts_thr, idling_boosts_thr_without_issues, - idling_needed_for_service_guarantees, - asymmetric_scenario; + idling_boosts_thr; - if (bfqd->strict_guarantees) - return true; - - /* - * Idling is performed only if slice_idle > 0. In addition, we - * do not idle if - * (a) bfqq is async - * (b) bfqq is in the idle io prio class: in this case we do - * not idle because we want to minimize the bandwidth that - * queues in this class can steal to higher-priority queues - */ - if (bfqd->bfq_slice_idle == 0 || !bfq_bfqq_sync(bfqq) || - bfq_class_idle(bfqq)) + /* No point in idling for bfqq if it won't get requests any longer */ + if (unlikely(!bfqq_process_refs(bfqq))) return false; bfqq_sequential_and_IO_bound = !BFQQ_SEEKY(bfqq) && @@ -3454,8 +4525,7 @@ static bool bfq_better_to_idle(struct bfq_queue *bfqq) bfqq_sequential_and_IO_bound); /* - * The value of the next variable, - * idling_boosts_thr_without_issues, is equal to that of + * The return value of this function is equal to that of * idling_boosts_thr, unless a special case holds. In this * special case, described below, idling may cause problems to * weight-raised queues. @@ -3472,217 +4542,85 @@ static bool bfq_better_to_idle(struct bfq_queue *bfqq) * which enqueue several requests in advance, and further * reorder internally-queued requests. * - * For this reason, we force to false the value of - * idling_boosts_thr_without_issues if there are weight-raised - * busy queues. In this case, and if bfqq is not weight-raised, - * this guarantees that the device is not idled for bfqq (if, - * instead, bfqq is weight-raised, then idling will be - * guaranteed by another variable, see below). Combined with - * the timestamping rules of BFQ (see [1] for details), this - * behavior causes bfqq, and hence any sync non-weight-raised - * queue, to get a lower number of requests served, and thus - * to ask for a lower number of requests from the request - * pool, before the busy weight-raised queues get served - * again. This often mitigates starvation problems in the - * presence of heavy write workloads and NCQ, thereby - * guaranteeing a higher application and system responsiveness - * in these hostile scenarios. - */ - idling_boosts_thr_without_issues = idling_boosts_thr && + * For this reason, we force to false the return value if + * there are weight-raised busy queues. In this case, and if + * bfqq is not weight-raised, this guarantees that the device + * is not idled for bfqq (if, instead, bfqq is weight-raised, + * then idling will be guaranteed by another variable, see + * below). Combined with the timestamping rules of BFQ (see + * [1] for details), this behavior causes bfqq, and hence any + * sync non-weight-raised queue, to get a lower number of + * requests served, and thus to ask for a lower number of + * requests from the request pool, before the busy + * weight-raised queues get served again. This often mitigates + * starvation problems in the presence of heavy write + * workloads and NCQ, thereby guaranteeing a higher + * application and system responsiveness in these hostile + * scenarios. + */ + return idling_boosts_thr && bfqd->wr_busy_queues == 0; +} + +/* + * For a queue that becomes empty, device idling is allowed only if + * this function returns true for that queue. As a consequence, since + * device idling plays a critical role for both throughput boosting + * and service guarantees, the return value of this function plays a + * critical role as well. + * + * In a nutshell, this function returns true only if idling is + * beneficial for throughput or, even if detrimental for throughput, + * idling is however necessary to preserve service guarantees (low + * latency, desired throughput distribution, ...). In particular, on + * NCQ-capable devices, this function tries to return false, so as to + * help keep the drives' internal queues full, whenever this helps the + * device boost the throughput without causing any service-guarantee + * issue. + * + * Most of the issues taken into account to get the return value of + * this function are not trivial. We discuss these issues in the two + * functions providing the main pieces of information needed by this + * function. + */ +static bool bfq_better_to_idle(struct bfq_queue *bfqq) +{ + struct bfq_data *bfqd = bfqq->bfqd; + bool idling_boosts_thr_with_no_issue, idling_needed_for_service_guar; + + /* No point in idling for bfqq if it won't get requests any longer */ + if (unlikely(!bfqq_process_refs(bfqq))) + return false; + + if (unlikely(bfqd->strict_guarantees)) + return true; /* - * There is then a case where idling must be performed not - * for throughput concerns, but to preserve service - * guarantees. - * - * To introduce this case, we can note that allowing the drive - * to enqueue more than one request at a time, and hence - * delegating de facto final scheduling decisions to the - * drive's internal scheduler, entails loss of control on the - * actual request service order. In particular, the critical - * situation is when requests from different processes happen - * to be present, at the same time, in the internal queue(s) - * of the drive. In such a situation, the drive, by deciding - * the service order of the internally-queued requests, does - * determine also the actual throughput distribution among - * these processes. But the drive typically has no notion or - * concern about per-process throughput distribution, and - * makes its decisions only on a per-request basis. Therefore, - * the service distribution enforced by the drive's internal - * scheduler is likely to coincide with the desired - * device-throughput distribution only in a completely - * symmetric scenario where: - * (i) each of these processes must get the same throughput as - * the others; - * (ii) the I/O of each process has the same properties, in - * terms of locality (sequential or random), direction - * (reads or writes), request sizes, greediness - * (from I/O-bound to sporadic), and so on. - * In fact, in such a scenario, the drive tends to treat - * the requests of each of these processes in about the same - * way as the requests of the others, and thus to provide - * each of these processes with about the same throughput - * (which is exactly the desired throughput distribution). In - * contrast, in any asymmetric scenario, device idling is - * certainly needed to guarantee that bfqq receives its - * assigned fraction of the device throughput (see [1] for - * details). - * The problem is that idling may significantly reduce - * throughput with certain combinations of types of I/O and - * devices. An important example is sync random I/O, on flash - * storage with command queueing. So, unless bfqq falls in the - * above cases where idling also boosts throughput, it would - * be important to check conditions (i) and (ii) accurately, - * so as to avoid idling when not strictly needed for service - * guarantees. - * - * Unfortunately, it is extremely difficult to thoroughly - * check condition (ii). And, in case there are active groups, - * it becomes very difficult to check condition (i) too. In - * 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. 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 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 is a primary concern (to - * preserve bandwidth and latency guarantees). - * - * 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. - * However, for queues with the same weight, a further - * mechanism, preemption, mitigates or even eliminates this - * problem. And it does so without consequences on overall - * throughput. This mechanism and its benefits are explained - * in the next three paragraphs. - * - * Even if a queue, say Q, is expired when it remains idle, Q - * can still preempt the new in-service queue if the next - * request of Q arrives soon (see the comments on - * bfq_bfqq_update_budg_for_activation). If all queues and - * groups have the same weight, this form of preemption, - * combined with the hole-recovery heuristic described in the - * comments on function bfq_bfqq_update_budg_for_activation, - * are enough to preserve a correct bandwidth distribution in - * the mid term, even without idling. In fact, even if not - * idling allows the internal queues of the device to contain - * many requests, and thus to reorder requests, we can rather - * safely assume that the internal scheduler still preserves a - * minimum of mid-term fairness. - * - * More precisely, this preemption-based, idleless approach - * provides fairness in terms of IOPS, and not sectors per - * second. This can be seen with a simple example. Suppose - * that there are two queues with the same weight, but that - * the first queue receives requests of 8 sectors, while the - * second queue receives requests of 1024 sectors. In - * addition, suppose that each of the two queues contains at - * most one request at a time, which implies that each queue - * always remains idle after it is served. Finally, after - * remaining idle, each queue receives very quickly a new - * request. It follows that the two queues are served - * alternatively, preempting each other if needed. This - * implies that, although both queues have the same weight, - * the queue with large requests receives a service that is - * 1024/8 times as high as the service received by the other - * queue. - * - * The motivation for using preemption instead of idling (for - * queues with the same weight) is that, by not idling, - * service guarantees are preserved (completely or at least in - * part) without minimally sacrificing throughput. And, if - * there is no active group, then the primary expectation for - * this device is probably a high throughput. - * - * We are now left only with explaining the additional - * compound condition that is checked below for deciding - * whether the scenario is asymmetric. To explain this - * compound condition, we need to add that the function - * bfq_symmetric_scenario checks the weights of only - * non-weight-raised queues, for efficiency reasons (see - * comments on bfq_weights_tree_add()). Then the fact that - * 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 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 - * following unlucky scenario: if idling is (correctly) - * disabled in a time period during which all symmetry - * sub-conditions hold, and hence the device is allowed to - * enqueue many requests, but at some later point in time some - * sub-condition stops to hold, then it may become impossible - * to let requests be served in the desired order until all - * the requests already queued in the device have been served. - */ - asymmetric_scenario = (bfqq->wr_coeff > 1 && - bfqd->wr_busy_queues < bfqd->busy_queues) || - !bfq_symmetric_scenario(bfqd); - - /* - * Finally, there is a case where maximizing throughput is the - * best choice even if it may cause unfairness toward - * bfqq. Such a case is when bfqq became active in a burst of - * queue activations. Queues that became active during a large - * burst benefit only from throughput, as discussed in the - * comments on bfq_handle_burst. Thus, if bfqq became active - * in a burst and not idling the device maximizes throughput, - * then the device must no be idled, because not idling the - * device provides bfqq and all other queues in the burst with - * maximum benefit. Combining this and the above case, we can - * now establish when idling is actually needed to preserve - * service guarantees. - */ - idling_needed_for_service_guarantees = - asymmetric_scenario && !bfq_bfqq_in_large_burst(bfqq); - - /* - * We have now all the components we need to compute the + * Idling is performed only if slice_idle > 0. In addition, we + * do not idle if + * (a) bfqq is async + * (b) bfqq is in the idle io prio class: in this case we do + * not idle because we want to minimize the bandwidth that + * queues in this class can steal to higher-priority queues + */ + if (bfqd->bfq_slice_idle == 0 || !bfq_bfqq_sync(bfqq) || + bfq_class_idle(bfqq)) + return false; + + idling_boosts_thr_with_no_issue = + idling_boosts_thr_without_issues(bfqd, bfqq); + + idling_needed_for_service_guar = + idling_needed_for_service_guarantees(bfqd, bfqq); + + /* + * We have now the two components we need to compute the * return value of the function, which is true only if idling * either boosts the throughput (without issues), or is * necessary to preserve service guarantees. */ - return idling_boosts_thr_without_issues || - idling_needed_for_service_guarantees; + return idling_boosts_thr_with_no_issue || + idling_needed_for_service_guar; } /* @@ -3701,37 +4639,165 @@ static bool bfq_bfqq_must_idle(struct bfq_queue *bfqq) return RB_EMPTY_ROOT(&bfqq->sort_list) && bfq_better_to_idle(bfqq); } -static struct bfq_queue *bfq_choose_bfqq_for_injection(struct bfq_data *bfqd) +/* + * This function chooses the queue from which to pick the next extra + * I/O request to inject, if it finds a compatible queue. See the + * comments on bfq_update_inject_limit() for details on the injection + * mechanism, and for the definitions of the quantities mentioned + * below. + */ +static struct bfq_queue * +bfq_choose_bfqq_for_injection(struct bfq_data *bfqd) { - struct bfq_queue *bfqq; + struct bfq_queue *bfqq, *in_serv_bfqq = bfqd->in_service_queue; + unsigned int limit = in_serv_bfqq->inject_limit; + int i; + + /* + * If + * - bfqq is not weight-raised and therefore does not carry + * time-critical I/O, + * or + * - regardless of whether bfqq is weight-raised, bfqq has + * however a long think time, during which it can absorb the + * effect of an appropriate number of extra I/O requests + * from other queues (see bfq_update_inject_limit for + * details on the computation of this number); + * then injection can be performed without restrictions. + */ + bool in_serv_always_inject = in_serv_bfqq->wr_coeff == 1 || + !bfq_bfqq_has_short_ttime(in_serv_bfqq); + + /* + * If + * - the baseline total service time could not be sampled yet, + * so the inject limit happens to be still 0, and + * - a lot of time has elapsed since the plugging of I/O + * dispatching started, so drive speed is being wasted + * significantly; + * then temporarily raise inject limit to one request. + */ + if (limit == 0 && in_serv_bfqq->last_serv_time_ns == 0 && + bfq_bfqq_wait_request(in_serv_bfqq) && + time_is_before_eq_jiffies(bfqd->last_idling_start_jiffies + + bfqd->bfq_slice_idle) + ) + limit = 1; + + if (bfqd->tot_rq_in_driver >= limit) + return NULL; /* - * A linear search; but, with a high probability, very few - * steps are needed to find a candidate queue, i.e., a queue - * with enough budget left for its next request. In fact: + * Linear search of the source queue for injection; but, with + * a high probability, very few steps are needed to find a + * candidate queue, i.e., a queue with enough budget left for + * its next request. In fact: * - BFQ dynamically updates the budget of every queue so as * to accommodate the expected backlog of the queue; * - if a queue gets all its requests dispatched as injected * service, then the queue is removed from the active list - * (and re-added only if it gets new requests, but with - * enough budget for its new backlog). + * (and re-added only if it gets new requests, but then it + * is assigned again enough budget for its new backlog). */ - list_for_each_entry(bfqq, &bfqd->active_list, bfqq_list) + for (i = 0; i < bfqd->num_actuators; i++) { + list_for_each_entry(bfqq, &bfqd->active_list[i], bfqq_list) + if (!RB_EMPTY_ROOT(&bfqq->sort_list) && + (in_serv_always_inject || bfqq->wr_coeff > 1) && + bfq_serv_to_charge(bfqq->next_rq, bfqq) <= + bfq_bfqq_budget_left(bfqq)) { + /* + * Allow for only one large in-flight request + * on non-rotational devices, for the + * following reason. On non-rotationl drives, + * large requests take much longer than + * smaller requests to be served. In addition, + * the drive prefers to serve large requests + * w.r.t. to small ones, if it can choose. So, + * having more than one large requests queued + * in the drive may easily make the next first + * request of the in-service queue wait for so + * long to break bfqq's service guarantees. On + * the bright side, large requests let the + * drive reach a very high throughput, even if + * there is only one in-flight large request + * at a time. + */ + if (blk_queue_nonrot(bfqd->queue) && + blk_rq_sectors(bfqq->next_rq) >= + BFQQ_SECT_THR_NONROT && + bfqd->tot_rq_in_driver >= 1) + continue; + else { + bfqd->rqs_injected = true; + return bfqq; + } + } + } + + return NULL; +} + +static struct bfq_queue * +bfq_find_active_bfqq_for_actuator(struct bfq_data *bfqd, int idx) +{ + struct bfq_queue *bfqq; + + if (bfqd->in_service_queue && + bfqd->in_service_queue->actuator_idx == idx) + return bfqd->in_service_queue; + + list_for_each_entry(bfqq, &bfqd->active_list[idx], bfqq_list) { if (!RB_EMPTY_ROOT(&bfqq->sort_list) && - bfq_serv_to_charge(bfqq->next_rq, bfqq) <= - bfq_bfqq_budget_left(bfqq)) + bfq_serv_to_charge(bfqq->next_rq, bfqq) <= + bfq_bfqq_budget_left(bfqq)) { return bfqq; + } + } return NULL; } /* + * Perform a linear scan of each actuator, until an actuator is found + * for which the following three conditions hold: the load of the + * actuator is below the threshold (see comments on + * actuator_load_threshold for details) and lower than that of the + * next actuator (comments on this extra condition below), and there + * is a queue that contains I/O for that actuator. On success, return + * that queue. + * + * Performing a plain linear scan entails a prioritization among + * actuators. The extra condition above breaks this prioritization and + * tends to distribute injection uniformly across actuators. + */ +static struct bfq_queue * +bfq_find_bfqq_for_underused_actuator(struct bfq_data *bfqd) +{ + int i; + + for (i = 0 ; i < bfqd->num_actuators; i++) { + if (bfqd->rq_in_driver[i] < bfqd->actuator_load_threshold && + (i == bfqd->num_actuators - 1 || + bfqd->rq_in_driver[i] < bfqd->rq_in_driver[i+1])) { + struct bfq_queue *bfqq = + bfq_find_active_bfqq_for_actuator(bfqd, i); + + if (bfqq) + return bfqq; + } + } + + return NULL; +} + + +/* * Select a queue for service. If we have a current queue in service, * check whether to continue servicing it, or retrieve and set a new one. */ static struct bfq_queue *bfq_select_queue(struct bfq_data *bfqd) { - struct bfq_queue *bfqq; + struct bfq_queue *bfqq, *inject_bfqq; struct request *next_rq; enum bfqq_expiration reason = BFQQE_BUDGET_TIMEOUT; @@ -3754,6 +4820,15 @@ static struct bfq_queue *bfq_select_queue(struct bfq_data *bfqd) check_queue: /* + * If some actuator is underutilized, but the in-service + * queue does not contain I/O for that actuator, then try to + * inject I/O for that actuator. + */ + inject_bfqq = bfq_find_bfqq_for_underused_actuator(bfqd); + if (inject_bfqq && inject_bfqq != bfqq) + return inject_bfqq; + + /* * This loop is rarely executed more than once. Even when it * happens, it is much more convenient to re-execute this loop * than to return NULL and trigger a new dispatch to get a @@ -3807,14 +4882,128 @@ check_queue: * for a new request, or has requests waiting for a completion and * may idle after their completion, then keep it anyway. * - * Yet, to boost throughput, inject service from other queues if - * possible. + * Yet, inject service from other queues if it boosts + * throughput and is possible. */ if (bfq_bfqq_wait_request(bfqq) || (bfqq->dispatched != 0 && bfq_better_to_idle(bfqq))) { - if (bfq_bfqq_injectable(bfqq) && - bfqq->injected_service * bfqq->inject_coeff < - bfqq->entity.service * 10) + unsigned int act_idx = bfqq->actuator_idx; + struct bfq_queue *async_bfqq = NULL; + struct bfq_queue *blocked_bfqq = + !hlist_empty(&bfqq->woken_list) ? + container_of(bfqq->woken_list.first, + struct bfq_queue, + woken_list_node) + : NULL; + + if (bfqq->bic && bfqq->bic->bfqq[0][act_idx] && + bfq_bfqq_busy(bfqq->bic->bfqq[0][act_idx]) && + bfqq->bic->bfqq[0][act_idx]->next_rq) + async_bfqq = bfqq->bic->bfqq[0][act_idx]; + /* + * The next four mutually-exclusive ifs decide + * whether to try injection, and choose the queue to + * pick an I/O request from. + * + * The first if checks whether the process associated + * with bfqq has also async I/O pending. If so, it + * injects such I/O unconditionally. Injecting async + * I/O from the same process can cause no harm to the + * process. On the contrary, it can only increase + * bandwidth and reduce latency for the process. + * + * The second if checks whether there happens to be a + * non-empty waker queue for bfqq, i.e., a queue whose + * I/O needs to be completed for bfqq to receive new + * I/O. This happens, e.g., if bfqq is associated with + * a process that does some sync. A sync generates + * extra blocking I/O, which must be completed before + * the process associated with bfqq can go on with its + * I/O. If the I/O of the waker queue is not served, + * then bfqq remains empty, and no I/O is dispatched, + * until the idle timeout fires for bfqq. This is + * likely to result in lower bandwidth and higher + * latencies for bfqq, and in a severe loss of total + * throughput. The best action to take is therefore to + * serve the waker queue as soon as possible. So do it + * (without relying on the third alternative below for + * eventually serving waker_bfqq's I/O; see the last + * paragraph for further details). This systematic + * injection of I/O from the waker queue does not + * cause any delay to bfqq's I/O. On the contrary, + * next bfqq's I/O is brought forward dramatically, + * for it is not blocked for milliseconds. + * + * The third if checks whether there is a queue woken + * by bfqq, and currently with pending I/O. Such a + * woken queue does not steal bandwidth from bfqq, + * because it remains soon without I/O if bfqq is not + * served. So there is virtually no risk of loss of + * bandwidth for bfqq if this woken queue has I/O + * dispatched while bfqq is waiting for new I/O. + * + * The fourth if checks whether bfqq is a queue for + * which it is better to avoid injection. It is so if + * bfqq delivers more throughput when served without + * any further I/O from other queues in the middle, or + * if the service times of bfqq's I/O requests both + * count more than overall throughput, and may be + * easily increased by injection (this happens if bfqq + * has a short think time). If none of these + * conditions holds, then a candidate queue for + * injection is looked for through + * bfq_choose_bfqq_for_injection(). Note that the + * latter may return NULL (for example if the inject + * limit for bfqq is currently 0). + * + * NOTE: motivation for the second alternative + * + * Thanks to the way the inject limit is updated in + * bfq_update_has_short_ttime(), it is rather likely + * that, if I/O is being plugged for bfqq and the + * waker queue has pending I/O requests that are + * blocking bfqq's I/O, then the fourth alternative + * above lets the waker queue get served before the + * I/O-plugging timeout fires. So one may deem the + * second alternative superfluous. It is not, because + * the fourth alternative may be way less effective in + * case of a synchronization. For two main + * reasons. First, throughput may be low because the + * inject limit may be too low to guarantee the same + * amount of injected I/O, from the waker queue or + * other queues, that the second alternative + * guarantees (the second alternative unconditionally + * injects a pending I/O request of the waker queue + * for each bfq_dispatch_request()). Second, with the + * fourth alternative, the duration of the plugging, + * i.e., the time before bfqq finally receives new I/O, + * may not be minimized, because the waker queue may + * happen to be served only after other queues. + */ + if (async_bfqq && + icq_to_bic(async_bfqq->next_rq->elv.icq) == bfqq->bic && + bfq_serv_to_charge(async_bfqq->next_rq, async_bfqq) <= + bfq_bfqq_budget_left(async_bfqq)) + bfqq = async_bfqq; + else if (bfqq->waker_bfqq && + bfq_bfqq_busy(bfqq->waker_bfqq) && + bfqq->waker_bfqq->next_rq && + bfq_serv_to_charge(bfqq->waker_bfqq->next_rq, + bfqq->waker_bfqq) <= + bfq_bfqq_budget_left(bfqq->waker_bfqq) + ) + bfqq = bfqq->waker_bfqq; + else if (blocked_bfqq && + bfq_bfqq_busy(blocked_bfqq) && + blocked_bfqq->next_rq && + bfq_serv_to_charge(blocked_bfqq->next_rq, + blocked_bfqq) <= + bfq_bfqq_budget_left(blocked_bfqq) + ) + bfqq = blocked_bfqq; + else if (!idling_boosts_thr_without_issues(bfqd, bfqq) && + (bfqq->wr_coeff == 1 || bfqd->wr_busy_queues > 1 || + !bfq_bfqq_has_short_ttime(bfqq))) bfqq = bfq_choose_bfqq_for_injection(bfqd); else bfqq = NULL; @@ -3866,9 +5055,21 @@ static void bfq_update_wr_data(struct bfq_data *bfqd, struct bfq_queue *bfqq) bfqq->wr_cur_max_time)) { if (bfqq->wr_cur_max_time != bfqd->bfq_wr_rt_max_time || time_is_before_jiffies(bfqq->wr_start_at_switch_to_srt + - bfq_wr_duration(bfqd))) + bfq_wr_duration(bfqd))) { + /* + * Either in interactive weight + * raising, or in soft_rt weight + * raising with the + * interactive-weight-raising period + * elapsed (so no switch back to + * interactive weight raising). + */ bfq_bfqq_end_wr(bfqq); - else { + } else { /* + * soft_rt finishing while still in + * interactive period, switch back to + * interactive weight raising + */ switch_back_to_interactive_wr(bfqq, bfqd); bfqq->entity.prio_changed = 1; } @@ -3906,15 +5107,15 @@ static struct request *bfq_dispatch_rq_from_bfqq(struct bfq_data *bfqd, bfq_bfqq_served(bfqq, service_to_charge); - bfq_dispatch_remove(bfqd->queue, rq); + if (bfqq == bfqd->in_service_queue && bfqd->wait_dispatch) { + bfqd->wait_dispatch = false; + bfqd->waited_rq = rq; + } - if (bfqq != bfqd->in_service_queue) { - if (likely(bfqd->in_service_queue)) - bfqd->in_service_queue->injected_service += - bfq_serv_to_charge(rq, bfqq); + bfq_dispatch_remove(bfqd->queue, rq); - goto return_rq; - } + if (bfqq != bfqd->in_service_queue) + return rq; /* * If weight raising has to terminate for bfqq, then next @@ -3934,12 +5135,9 @@ static struct request *bfq_dispatch_rq_from_bfqq(struct bfq_data *bfqd, * belongs to CLASS_IDLE and other queues are waiting for * service. */ - if (!(bfqd->busy_queues > 1 && bfq_class_idle(bfqq))) - goto return_rq; + if (bfq_tot_busy_queues(bfqd) > 1 && bfq_class_idle(bfqq)) + bfq_bfqq_expire(bfqd, bfqq, false, BFQQE_BUDGET_EXHAUSTED); - bfq_bfqq_expire(bfqd, bfqq, false, BFQQE_BUDGET_EXHAUSTED); - -return_rq: return rq; } @@ -3948,11 +5146,11 @@ static bool bfq_has_work(struct blk_mq_hw_ctx *hctx) struct bfq_data *bfqd = hctx->queue->elevator->elevator_data; /* - * Avoiding lock: a race on bfqd->busy_queues should cause at + * Avoiding lock: a race on bfqd->queued should cause at * most a call to dispatch for nothing */ return !list_empty_careful(&bfqd->dispatch) || - bfqd->busy_queues > 0; + READ_ONCE(bfqd->queued); } static struct request *__bfq_dispatch_request(struct blk_mq_hw_ctx *hctx) @@ -3982,11 +5180,11 @@ static struct request *__bfq_dispatch_request(struct blk_mq_hw_ctx *hctx) /* * We exploit the bfq_finish_requeue_request hook to - * decrement rq_in_driver, but + * decrement tot_rq_in_driver, but * bfq_finish_requeue_request will not be invoked on * this request. So, to avoid unbalance, just start - * this request, without incrementing rq_in_driver. As - * a negative consequence, rq_in_driver is deceptively + * this request, without incrementing tot_rq_in_driver. As + * a negative consequence, tot_rq_in_driver is deceptively * lower than it should be while this request is in * service. This may cause bfq_schedule_dispatch to be * invoked uselessly. @@ -3995,7 +5193,7 @@ static struct request *__bfq_dispatch_request(struct blk_mq_hw_ctx *hctx) * bfq_finish_requeue_request hook, if defined, is * probably invoked also on this request. So, by * exploiting this hook, we could 1) increment - * rq_in_driver here, and 2) decrement it in + * tot_rq_in_driver here, and 2) decrement it in * bfq_finish_requeue_request. Such a solution would * let the value of the counter be always accurate, * but it would entail using an extra interface @@ -4006,9 +5204,10 @@ static struct request *__bfq_dispatch_request(struct blk_mq_hw_ctx *hctx) goto start_rq; } - bfq_log(bfqd, "dispatch requests: %d busy queues", bfqd->busy_queues); + bfq_log(bfqd, "dispatch requests: %d busy queues", + bfq_tot_busy_queues(bfqd)); - if (bfqd->busy_queues == 0) + if (bfq_tot_busy_queues(bfqd) == 0) goto exit; /* @@ -4020,10 +5219,10 @@ static struct request *__bfq_dispatch_request(struct blk_mq_hw_ctx *hctx) * some unlucky request wait for as long as the device * wishes. * - * Of course, serving one request at at time may cause loss of + * Of course, serving one request at a time may cause loss of * throughput. */ - if (bfqd->strict_guarantees && bfqd->rq_in_driver > 0) + if (bfqd->strict_guarantees && bfqd->tot_rq_in_driver > 0) goto exit; bfqq = bfq_select_queue(bfqd); @@ -4034,7 +5233,8 @@ static struct request *__bfq_dispatch_request(struct blk_mq_hw_ctx *hctx) if (rq) { inc_in_driver_start_rq: - bfqd->rq_in_driver++; + bfqd->rq_in_driver[bfqq->actuator_idx]++; + bfqd->tot_rq_in_driver++; start_rq: rq->rq_flags |= RQF_STARTED; } @@ -4042,7 +5242,7 @@ exit: return rq; } -#if defined(CONFIG_BFQ_GROUP_IOSCHED) && defined(CONFIG_DEBUG_BLK_CGROUP) +#ifdef CONFIG_BFQ_CGROUP_DEBUG static void bfq_update_dispatch_stats(struct request_queue *q, struct request *rq, struct bfq_queue *in_serv_queue, @@ -4092,14 +5292,14 @@ static inline void bfq_update_dispatch_stats(struct request_queue *q, struct request *rq, struct bfq_queue *in_serv_queue, bool idle_timer_disabled) {} -#endif +#endif /* CONFIG_BFQ_CGROUP_DEBUG */ static struct request *bfq_dispatch_request(struct blk_mq_hw_ctx *hctx) { struct bfq_data *bfqd = hctx->queue->elevator->elevator_data; struct request *rq; struct bfq_queue *in_serv_queue; - bool waiting_rq, idle_timer_disabled; + bool waiting_rq, idle_timer_disabled = false; spin_lock_irq(&bfqd->lock); @@ -4107,14 +5307,15 @@ static struct request *bfq_dispatch_request(struct blk_mq_hw_ctx *hctx) waiting_rq = in_serv_queue && bfq_bfqq_wait_request(in_serv_queue); rq = __bfq_dispatch_request(hctx); - - idle_timer_disabled = - waiting_rq && !bfq_bfqq_wait_request(in_serv_queue); + if (in_serv_queue == bfqd->in_service_queue) { + idle_timer_disabled = + waiting_rq && !bfq_bfqq_wait_request(in_serv_queue); + } spin_unlock_irq(&bfqd->lock); - - bfq_update_dispatch_stats(hctx->queue, rq, in_serv_queue, - idle_timer_disabled); + bfq_update_dispatch_stats(hctx->queue, rq, + idle_timer_disabled ? in_serv_queue : NULL, + idle_timer_disabled); return rq; } @@ -4128,13 +5329,11 @@ static struct request *bfq_dispatch_request(struct blk_mq_hw_ctx *hctx) */ void bfq_put_queue(struct bfq_queue *bfqq) { -#ifdef CONFIG_BFQ_GROUP_IOSCHED + struct bfq_queue *item; + struct hlist_node *n; struct bfq_group *bfqg = bfqq_group(bfqq); -#endif - if (bfqq->bfqd) - bfq_log_bfqq(bfqq->bfqd, bfqq, "put_queue: %p %d", - bfqq, bfqq->ref); + bfq_log_bfqq(bfqq->bfqd, bfqq, "put_queue: %p %d", bfqq, bfqq->ref); bfqq->ref--; if (bfqq->ref) @@ -4172,13 +5371,50 @@ void bfq_put_queue(struct bfq_queue *bfqq) bfqq->bfqd->burst_size--; } + /* + * bfqq does not exist any longer, so it cannot be woken by + * any other queue, and cannot wake any other queue. Then bfqq + * must be removed from the woken list of its possible waker + * queue, and all queues in the woken list of bfqq must stop + * having a waker queue. Strictly speaking, these updates + * should be performed when bfqq remains with no I/O source + * attached to it, which happens before bfqq gets freed. In + * particular, this happens when the last process associated + * with bfqq exits or gets associated with a different + * queue. However, both events lead to bfqq being freed soon, + * and dangling references would come out only after bfqq gets + * freed. So these updates are done here, as a simple and safe + * way to handle all cases. + */ + /* remove bfqq from woken list */ + if (!hlist_unhashed(&bfqq->woken_list_node)) + hlist_del_init(&bfqq->woken_list_node); + + /* reset waker for all queues in woken list */ + hlist_for_each_entry_safe(item, n, &bfqq->woken_list, + woken_list_node) { + item->waker_bfqq = NULL; + hlist_del_init(&item->woken_list_node); + } + + if (bfqq->bfqd->last_completed_rq_bfqq == bfqq) + bfqq->bfqd->last_completed_rq_bfqq = NULL; + + WARN_ON_ONCE(!list_empty(&bfqq->fifo)); + WARN_ON_ONCE(!RB_EMPTY_ROOT(&bfqq->sort_list)); + WARN_ON_ONCE(bfqq->dispatched); + kmem_cache_free(bfq_pool, bfqq); -#ifdef CONFIG_BFQ_GROUP_IOSCHED bfqg_and_blkg_put(bfqg); -#endif } -static void bfq_put_cooperator(struct bfq_queue *bfqq) +static void bfq_put_stable_ref(struct bfq_queue *bfqq) +{ + bfqq->stable_ref--; + bfq_put_queue(bfqq); +} + +void bfq_put_cooperator(struct bfq_queue *bfqq) { struct bfq_queue *__bfqq, *next; @@ -4189,8 +5425,6 @@ static void bfq_put_cooperator(struct bfq_queue *bfqq) */ __bfqq = bfqq->new_bfqq; while (__bfqq) { - if (__bfqq == bfqq) - break; next = __bfqq->new_bfqq; bfq_put_queue(__bfqq); __bfqq = next; @@ -4200,7 +5434,7 @@ static void bfq_put_cooperator(struct bfq_queue *bfqq) static void bfq_exit_bfqq(struct bfq_data *bfqd, struct bfq_queue *bfqq) { if (bfqq == bfqd->in_service_queue) { - __bfq_bfqq_expire(bfqd, bfqq); + __bfq_bfqq_expire(bfqd, bfqq, BFQQE_BUDGET_TIMEOUT); bfq_schedule_dispatch(bfqd); } @@ -4208,33 +5442,60 @@ static void bfq_exit_bfqq(struct bfq_data *bfqd, struct bfq_queue *bfqq) bfq_put_cooperator(bfqq); - bfq_put_queue(bfqq); /* release process reference */ + bfq_release_process_ref(bfqd, bfqq); } -static void bfq_exit_icq_bfqq(struct bfq_io_cq *bic, bool is_sync) +static void bfq_exit_icq_bfqq(struct bfq_io_cq *bic, bool is_sync, + unsigned int actuator_idx) { - struct bfq_queue *bfqq = bic_to_bfqq(bic, is_sync); + struct bfq_queue *bfqq = bic_to_bfqq(bic, is_sync, actuator_idx); struct bfq_data *bfqd; if (bfqq) bfqd = bfqq->bfqd; /* NULL if scheduler already exited */ if (bfqq && bfqd) { - unsigned long flags; - - spin_lock_irqsave(&bfqd->lock, flags); + bic_set_bfqq(bic, NULL, is_sync, actuator_idx); bfq_exit_bfqq(bfqd, bfqq); - bic_set_bfqq(bic, NULL, is_sync); - spin_unlock_irqrestore(&bfqd->lock, flags); + } +} + +static void _bfq_exit_icq(struct bfq_io_cq *bic, unsigned int num_actuators) +{ + struct bfq_iocq_bfqq_data *bfqq_data = bic->bfqq_data; + unsigned int act_idx; + + for (act_idx = 0; act_idx < num_actuators; act_idx++) { + if (bfqq_data[act_idx].stable_merge_bfqq) + bfq_put_stable_ref(bfqq_data[act_idx].stable_merge_bfqq); + + bfq_exit_icq_bfqq(bic, true, act_idx); + bfq_exit_icq_bfqq(bic, false, act_idx); } } static void bfq_exit_icq(struct io_cq *icq) { struct bfq_io_cq *bic = icq_to_bic(icq); + struct bfq_data *bfqd = bic_to_bfqd(bic); + unsigned long flags; - bfq_exit_icq_bfqq(bic, true); - bfq_exit_icq_bfqq(bic, false); + /* + * If bfqd and thus bfqd->num_actuators is not available any + * longer, then cycle over all possible per-actuator bfqqs in + * next loop. We rely on bic being zeroed on creation, and + * therefore on its unused per-actuator fields being NULL. + * + * bfqd is NULL if scheduler already exited, and in that case + * this is the last time these queues are accessed. + */ + if (bfqd) { + spin_lock_irqsave(&bfqd->lock, flags); + _bfq_exit_icq(bic, bfqd->num_actuators); + spin_unlock_irqrestore(&bfqd->lock, flags); + } else { + _bfq_exit_icq(bic, BFQ_MAX_ACTUATORS); + } } /* @@ -4254,9 +5515,10 @@ bfq_set_next_ioprio_data(struct bfq_queue *bfqq, struct bfq_io_cq *bic) ioprio_class = IOPRIO_PRIO_CLASS(bic->ioprio); switch (ioprio_class) { default: - dev_err(bfqq->bfqd->queue->backing_dev_info->dev, - "bfq: bad prio class %d\n", ioprio_class); - /* fall through */ + pr_err("bdi %s: bfq: bad prio class %d\n", + bdi_dev_name(bfqq->bfqd->queue->disk->bdi), + ioprio_class); + fallthrough; case IOPRIO_CLASS_NONE: /* * No prio set, inherit CPU scheduling settings. @@ -4265,32 +5527,35 @@ bfq_set_next_ioprio_data(struct bfq_queue *bfqq, struct bfq_io_cq *bic) bfqq->new_ioprio_class = task_nice_ioclass(tsk); break; case IOPRIO_CLASS_RT: - bfqq->new_ioprio = IOPRIO_PRIO_DATA(bic->ioprio); + bfqq->new_ioprio = IOPRIO_PRIO_LEVEL(bic->ioprio); bfqq->new_ioprio_class = IOPRIO_CLASS_RT; break; case IOPRIO_CLASS_BE: - bfqq->new_ioprio = IOPRIO_PRIO_DATA(bic->ioprio); + bfqq->new_ioprio = IOPRIO_PRIO_LEVEL(bic->ioprio); bfqq->new_ioprio_class = IOPRIO_CLASS_BE; break; case IOPRIO_CLASS_IDLE: bfqq->new_ioprio_class = IOPRIO_CLASS_IDLE; - bfqq->new_ioprio = 7; + bfqq->new_ioprio = IOPRIO_NR_LEVELS - 1; break; } - if (bfqq->new_ioprio >= IOPRIO_BE_NR) { + if (bfqq->new_ioprio >= IOPRIO_NR_LEVELS) { pr_crit("bfq_set_next_ioprio_data: new_ioprio %d\n", bfqq->new_ioprio); - bfqq->new_ioprio = IOPRIO_BE_NR; + bfqq->new_ioprio = IOPRIO_NR_LEVELS - 1; } bfqq->entity.new_weight = bfq_ioprio_to_weight(bfqq->new_ioprio); + bfq_log_bfqq(bfqd, bfqq, "new_ioprio %d new_weight %d", + bfqq->new_ioprio, bfqq->entity.new_weight); bfqq->entity.prio_changed = 1; } static struct bfq_queue *bfq_get_queue(struct bfq_data *bfqd, struct bio *bio, bool is_sync, - struct bfq_io_cq *bic); + struct bfq_io_cq *bic, + bool respawn); static void bfq_check_ioprio_change(struct bfq_io_cq *bic, struct bio *bio) { @@ -4307,25 +5572,32 @@ static void bfq_check_ioprio_change(struct bfq_io_cq *bic, struct bio *bio) bic->ioprio = ioprio; - bfqq = bic_to_bfqq(bic, false); + bfqq = bic_to_bfqq(bic, false, bfq_actuator_index(bfqd, bio)); if (bfqq) { - /* release process reference on this queue */ - bfq_put_queue(bfqq); - bfqq = bfq_get_queue(bfqd, bio, BLK_RW_ASYNC, bic); - bic_set_bfqq(bic, bfqq, false); + struct bfq_queue *old_bfqq = bfqq; + + bfqq = bfq_get_queue(bfqd, bio, false, bic, true); + bic_set_bfqq(bic, bfqq, false, bfq_actuator_index(bfqd, bio)); + bfq_release_process_ref(bfqd, old_bfqq); } - bfqq = bic_to_bfqq(bic, true); + bfqq = bic_to_bfqq(bic, true, bfq_actuator_index(bfqd, bio)); if (bfqq) bfq_set_next_ioprio_data(bfqq, bic); } static void bfq_init_bfqq(struct bfq_data *bfqd, struct bfq_queue *bfqq, - struct bfq_io_cq *bic, pid_t pid, int is_sync) + struct bfq_io_cq *bic, pid_t pid, int is_sync, + unsigned int act_idx) { + u64 now_ns = blk_time_get_ns(); + + bfqq->actuator_idx = act_idx; RB_CLEAR_NODE(&bfqq->entity.rb_node); INIT_LIST_HEAD(&bfqq->fifo); INIT_HLIST_NODE(&bfqq->burst_list_node); + INIT_HLIST_NODE(&bfqq->woken_list_node); + INIT_HLIST_HEAD(&bfqq->woken_list); bfqq->ref = 0; bfqq->bfqd = bfqd; @@ -4344,18 +5616,15 @@ static void bfq_init_bfqq(struct bfq_data *bfqd, struct bfq_queue *bfqq, bfq_mark_bfqq_has_short_ttime(bfqq); bfq_mark_bfqq_sync(bfqq); bfq_mark_bfqq_just_created(bfqq); - /* - * Aggressively inject a lot of service: up to 90%. - * This coefficient remains constant during bfqq life, - * but this behavior might be changed, after enough - * testing and tuning. - */ - bfqq->inject_coeff = 1; } else bfq_clear_bfqq_sync(bfqq); /* set end request to minus infinity from now */ - bfqq->ttime.last_end_request = ktime_get_ns() + 1; + bfqq->ttime.last_end_request = now_ns + 1; + + bfqq->creation_time = jiffies; + + bfqq->io_start_time = now_ns; bfq_mark_bfqq_IO_bound(bfqq); @@ -4383,60 +5652,207 @@ static void bfq_init_bfqq(struct bfq_data *bfqd, struct bfq_queue *bfqq, /* first request is almost certainly seeky */ bfqq->seek_history = 1; + + bfqq->decrease_time_jif = jiffies; } static struct bfq_queue **bfq_async_queue_prio(struct bfq_data *bfqd, struct bfq_group *bfqg, - int ioprio_class, int ioprio) + int ioprio_class, int ioprio, int act_idx) { switch (ioprio_class) { case IOPRIO_CLASS_RT: - return &bfqg->async_bfqq[0][ioprio]; + return &bfqg->async_bfqq[0][ioprio][act_idx]; case IOPRIO_CLASS_NONE: - ioprio = IOPRIO_NORM; - /* fall through */ + ioprio = IOPRIO_BE_NORM; + fallthrough; case IOPRIO_CLASS_BE: - return &bfqg->async_bfqq[1][ioprio]; + return &bfqg->async_bfqq[1][ioprio][act_idx]; case IOPRIO_CLASS_IDLE: - return &bfqg->async_idle_bfqq; + return &bfqg->async_idle_bfqq[act_idx]; default: return NULL; } } +static struct bfq_queue * +bfq_do_early_stable_merge(struct bfq_data *bfqd, struct bfq_queue *bfqq, + struct bfq_io_cq *bic, + struct bfq_queue *last_bfqq_created) +{ + unsigned int a_idx = last_bfqq_created->actuator_idx; + struct bfq_queue *new_bfqq = + bfq_setup_merge(bfqq, last_bfqq_created); + + if (!new_bfqq) + return bfqq; + + if (new_bfqq->bic) + new_bfqq->bic->bfqq_data[a_idx].stably_merged = true; + bic->bfqq_data[a_idx].stably_merged = true; + + /* + * Reusing merge functions. This implies that + * bfqq->bic must be set too, for + * bfq_merge_bfqqs to correctly save bfqq's + * state before killing it. + */ + bfqq->bic = bic; + return bfq_merge_bfqqs(bfqd, bic, bfqq); +} + +/* + * Many throughput-sensitive workloads are made of several parallel + * I/O flows, with all flows generated by the same application, or + * more generically by the same task (e.g., system boot). The most + * counterproductive action with these workloads is plugging I/O + * dispatch when one of the bfq_queues associated with these flows + * remains temporarily empty. + * + * To avoid this plugging, BFQ has been using a burst-handling + * mechanism for years now. This mechanism has proven effective for + * throughput, and not detrimental for service guarantees. The + * following function pushes this mechanism a little bit further, + * basing on the following two facts. + * + * First, all the I/O flows of a the same application or task + * contribute to the execution/completion of that common application + * or task. So the performance figures that matter are total + * throughput of the flows and task-wide I/O latency. In particular, + * these flows do not need to be protected from each other, in terms + * of individual bandwidth or latency. + * + * Second, the above fact holds regardless of the number of flows. + * + * Putting these two facts together, this commits merges stably the + * bfq_queues associated with these I/O flows, i.e., with the + * processes that generate these IO/ flows, regardless of how many the + * involved processes are. + * + * To decide whether a set of bfq_queues is actually associated with + * the I/O flows of a common application or task, and to merge these + * queues stably, this function operates as follows: given a bfq_queue, + * say Q2, currently being created, and the last bfq_queue, say Q1, + * created before Q2, Q2 is merged stably with Q1 if + * - very little time has elapsed since when Q1 was created + * - Q2 has the same ioprio as Q1 + * - Q2 belongs to the same group as Q1 + * + * Merging bfq_queues also reduces scheduling overhead. A fio test + * with ten random readers on /dev/nullb shows a throughput boost of + * 40%, with a quadcore. Since BFQ's execution time amounts to ~50% of + * the total per-request processing time, the above throughput boost + * implies that BFQ's overhead is reduced by more than 50%. + * + * This new mechanism most certainly obsoletes the current + * burst-handling heuristics. We keep those heuristics for the moment. + */ +static struct bfq_queue *bfq_do_or_sched_stable_merge(struct bfq_data *bfqd, + struct bfq_queue *bfqq, + struct bfq_io_cq *bic) +{ + struct bfq_queue **source_bfqq = bfqq->entity.parent ? + &bfqq->entity.parent->last_bfqq_created : + &bfqd->last_bfqq_created; + + struct bfq_queue *last_bfqq_created = *source_bfqq; + + /* + * If last_bfqq_created has not been set yet, then init it. If + * it has been set already, but too long ago, then move it + * forward to bfqq. Finally, move also if bfqq belongs to a + * different group than last_bfqq_created, or if bfqq has a + * different ioprio, ioprio_class or actuator_idx. If none of + * these conditions holds true, then try an early stable merge + * or schedule a delayed stable merge. As for the condition on + * actuator_idx, the reason is that, if queues associated with + * different actuators are merged, then control is lost on + * each actuator. Therefore some actuator may be + * underutilized, and throughput may decrease. + * + * A delayed merge is scheduled (instead of performing an + * early merge), in case bfqq might soon prove to be more + * throughput-beneficial if not merged. Currently this is + * possible only if bfqd is rotational with no queueing. For + * such a drive, not merging bfqq is better for throughput if + * bfqq happens to contain sequential I/O. So, we wait a + * little bit for enough I/O to flow through bfqq. After that, + * if such an I/O is sequential, then the merge is + * canceled. Otherwise the merge is finally performed. + */ + if (!last_bfqq_created || + time_before(last_bfqq_created->creation_time + + msecs_to_jiffies(bfq_activation_stable_merging), + bfqq->creation_time) || + bfqq->entity.parent != last_bfqq_created->entity.parent || + bfqq->ioprio != last_bfqq_created->ioprio || + bfqq->ioprio_class != last_bfqq_created->ioprio_class || + bfqq->actuator_idx != last_bfqq_created->actuator_idx) + *source_bfqq = bfqq; + else if (time_after_eq(last_bfqq_created->creation_time + + bfqd->bfq_burst_interval, + bfqq->creation_time)) { + if (likely(bfqd->nonrot_with_queueing)) + /* + * With this type of drive, leaving + * bfqq alone may provide no + * throughput benefits compared with + * merging bfqq. So merge bfqq now. + */ + bfqq = bfq_do_early_stable_merge(bfqd, bfqq, + bic, + last_bfqq_created); + else { /* schedule tentative stable merge */ + /* + * get reference on last_bfqq_created, + * to prevent it from being freed, + * until we decide whether to merge + */ + last_bfqq_created->ref++; + /* + * need to keep track of stable refs, to + * compute process refs correctly + */ + last_bfqq_created->stable_ref++; + /* + * Record the bfqq to merge to. + */ + bic->bfqq_data[last_bfqq_created->actuator_idx].stable_merge_bfqq = + last_bfqq_created; + } + } + + return bfqq; +} + + static struct bfq_queue *bfq_get_queue(struct bfq_data *bfqd, struct bio *bio, bool is_sync, - struct bfq_io_cq *bic) + struct bfq_io_cq *bic, + bool respawn) { - const int ioprio = IOPRIO_PRIO_DATA(bic->ioprio); + const int ioprio = IOPRIO_PRIO_LEVEL(bic->ioprio); const int ioprio_class = IOPRIO_PRIO_CLASS(bic->ioprio); struct bfq_queue **async_bfqq = NULL; struct bfq_queue *bfqq; struct bfq_group *bfqg; - rcu_read_lock(); - - bfqg = bfq_find_set_group(bfqd, __bio_blkcg(bio)); - if (!bfqg) { - bfqq = &bfqd->oom_bfqq; - goto out; - } - + bfqg = bfq_bio_bfqg(bfqd, bio); if (!is_sync) { async_bfqq = bfq_async_queue_prio(bfqd, bfqg, ioprio_class, - ioprio); + ioprio, + bfq_actuator_index(bfqd, bio)); bfqq = *async_bfqq; if (bfqq) goto out; } - bfqq = kmem_cache_alloc_node(bfq_pool, - GFP_NOWAIT | __GFP_ZERO | __GFP_NOWARN, + bfqq = kmem_cache_alloc_node(bfq_pool, GFP_NOWAIT | __GFP_ZERO, bfqd->queue->node); if (bfqq) { bfq_init_bfqq(bfqd, bfqq, bic, current->pid, - is_sync); + is_sync, bfq_actuator_index(bfqd, bio)); bfq_init_entity(&bfqq->entity, bfqg); bfq_log_bfqq(bfqd, bfqq, "allocated"); } else { @@ -4464,8 +5880,9 @@ static struct bfq_queue *bfq_get_queue(struct bfq_data *bfqd, out: bfqq->ref++; /* get a process reference to this queue */ - bfq_log_bfqq(bfqd, bfqq, "get_queue, at end: %p, %d", bfqq, bfqq->ref); - rcu_read_unlock(); + + if (bfqq != &bfqd->oom_bfqq && is_sync && !respawn) + bfqq = bfq_do_or_sched_stable_merge(bfqd, bfqq, bic); return bfqq; } @@ -4473,11 +5890,19 @@ static void bfq_update_io_thinktime(struct bfq_data *bfqd, struct bfq_queue *bfqq) { struct bfq_ttime *ttime = &bfqq->ttime; - u64 elapsed = ktime_get_ns() - bfqq->ttime.last_end_request; + u64 elapsed; + /* + * We are really interested in how long it takes for the queue to + * become busy when there is no outstanding IO for this queue. So + * ignore cases when the bfq queue has already IO queued. + */ + if (bfqq->dispatched || bfq_bfqq_busy(bfqq)) + return; + elapsed = blk_time_get_ns() - bfqq->ttime.last_end_request; elapsed = min_t(u64, elapsed, 2ULL * bfqd->bfq_slice_idle); - ttime->ttime_samples = (7*bfqq->ttime.ttime_samples + 256) / 8; + ttime->ttime_samples = (7*ttime->ttime_samples + 256) / 8; ttime->ttime_total = div_u64(7*ttime->ttime_total + 256*elapsed, 8); ttime->ttime_mean = div64_ul(ttime->ttime_total + 128, ttime->ttime_samples); @@ -4488,17 +5913,37 @@ bfq_update_io_seektime(struct bfq_data *bfqd, struct bfq_queue *bfqq, struct request *rq) { bfqq->seek_history <<= 1; - bfqq->seek_history |= - get_sdist(bfqq->last_request_pos, rq) > BFQQ_SEEK_THR && - (!blk_queue_nonrot(bfqd->queue) || - blk_rq_sectors(rq) < BFQQ_SECT_THR_NONROT); + bfqq->seek_history |= BFQ_RQ_SEEKY(bfqd, bfqq->last_request_pos, rq); + + if (bfqq->wr_coeff > 1 && + bfqq->wr_cur_max_time == bfqd->bfq_wr_rt_max_time && + BFQQ_TOTALLY_SEEKY(bfqq)) { + if (time_is_before_jiffies(bfqq->wr_start_at_switch_to_srt + + bfq_wr_duration(bfqd))) { + /* + * In soft_rt weight raising with the + * interactive-weight-raising period + * elapsed (so no switch back to + * interactive weight raising). + */ + bfq_bfqq_end_wr(bfqq); + } else { /* + * stopping soft_rt weight raising + * while still in interactive period, + * switch back to interactive weight + * raising + */ + switch_back_to_interactive_wr(bfqq, bfqd); + bfqq->entity.prio_changed = 1; + } + } } static void bfq_update_has_short_ttime(struct bfq_data *bfqd, struct bfq_queue *bfqq, struct bfq_io_cq *bic) { - bool has_short_ttime = true; + bool has_short_ttime = true, state_changed; /* * No need to update has_short_ttime if bfqq is async or in @@ -4515,21 +5960,111 @@ static void bfq_update_has_short_ttime(struct bfq_data *bfqd, return; /* Think time is infinite if no process is linked to - * bfqq. Otherwise check average think time to - * decide whether to mark as has_short_ttime + * bfqq. Otherwise check average think time to decide whether + * to mark as has_short_ttime. To this goal, compare average + * think time with half the I/O-plugging timeout. */ if (atomic_read(&bic->icq.ioc->active_ref) == 0 || (bfq_sample_valid(bfqq->ttime.ttime_samples) && - bfqq->ttime.ttime_mean > bfqd->bfq_slice_idle)) + bfqq->ttime.ttime_mean > bfqd->bfq_slice_idle>>1)) has_short_ttime = false; - bfq_log_bfqq(bfqd, bfqq, "update_has_short_ttime: has_short_ttime %d", - has_short_ttime); + state_changed = has_short_ttime != bfq_bfqq_has_short_ttime(bfqq); if (has_short_ttime) bfq_mark_bfqq_has_short_ttime(bfqq); else bfq_clear_bfqq_has_short_ttime(bfqq); + + /* + * Until the base value for the total service time gets + * finally computed for bfqq, the inject limit does depend on + * the think-time state (short|long). In particular, the limit + * is 0 or 1 if the think time is deemed, respectively, as + * short or long (details in the comments in + * bfq_update_inject_limit()). Accordingly, the next + * instructions reset the inject limit if the think-time state + * has changed and the above base value is still to be + * computed. + * + * However, the reset is performed only if more than 100 ms + * have elapsed since the last update of the inject limit, or + * (inclusive) if the change is from short to long think + * time. The reason for this waiting is as follows. + * + * bfqq may have a long think time because of a + * synchronization with some other queue, i.e., because the + * I/O of some other queue may need to be completed for bfqq + * to receive new I/O. Details in the comments on the choice + * of the queue for injection in bfq_select_queue(). + * + * As stressed in those comments, if such a synchronization is + * actually in place, then, without injection on bfqq, the + * blocking I/O cannot happen to served while bfqq is in + * service. As a consequence, if bfqq is granted + * I/O-dispatch-plugging, then bfqq remains empty, and no I/O + * is dispatched, until the idle timeout fires. This is likely + * to result in lower bandwidth and higher latencies for bfqq, + * and in a severe loss of total throughput. + * + * On the opposite end, a non-zero inject limit may allow the + * I/O that blocks bfqq to be executed soon, and therefore + * bfqq to receive new I/O soon. + * + * But, if the blocking gets actually eliminated, then the + * next think-time sample for bfqq may be very low. This in + * turn may cause bfqq's think time to be deemed + * short. Without the 100 ms barrier, this new state change + * would cause the body of the next if to be executed + * immediately. But this would set to 0 the inject + * limit. Without injection, the blocking I/O would cause the + * think time of bfqq to become long again, and therefore the + * inject limit to be raised again, and so on. The only effect + * of such a steady oscillation between the two think-time + * states would be to prevent effective injection on bfqq. + * + * In contrast, if the inject limit is not reset during such a + * long time interval as 100 ms, then the number of short + * think time samples can grow significantly before the reset + * is performed. As a consequence, the think time state can + * become stable before the reset. Therefore there will be no + * state change when the 100 ms elapse, and no reset of the + * inject limit. The inject limit remains steadily equal to 1 + * both during and after the 100 ms. So injection can be + * performed at all times, and throughput gets boosted. + * + * An inject limit equal to 1 is however in conflict, in + * general, with the fact that the think time of bfqq is + * short, because injection may be likely to delay bfqq's I/O + * (as explained in the comments in + * bfq_update_inject_limit()). But this does not happen in + * this special case, because bfqq's low think time is due to + * an effective handling of a synchronization, through + * injection. In this special case, bfqq's I/O does not get + * delayed by injection; on the contrary, bfqq's I/O is + * brought forward, because it is not blocked for + * milliseconds. + * + * In addition, serving the blocking I/O much sooner, and much + * more frequently than once per I/O-plugging timeout, makes + * it much quicker to detect a waker queue (the concept of + * waker queue is defined in the comments in + * bfq_add_request()). This makes it possible to start sooner + * to boost throughput more effectively, by injecting the I/O + * of the waker queue unconditionally on every + * bfq_dispatch_request(). + * + * One last, important benefit of not resetting the inject + * limit before 100 ms is that, during this time interval, the + * base value for the total service time is likely to get + * finally computed for bfqq, freeing the inject limit from + * its relation with the think time. + */ + if (state_changed && bfqq->last_serv_time_ns == 0 && + (time_is_before_eq_jiffies(bfqq->decrease_time_jif + + msecs_to_jiffies(100)) || + !has_short_ttime)) + bfq_reset_inject_limit(bfqd, bfqq); } /* @@ -4539,19 +6074,9 @@ static void bfq_update_has_short_ttime(struct bfq_data *bfqd, static void bfq_rq_enqueued(struct bfq_data *bfqd, struct bfq_queue *bfqq, struct request *rq) { - struct bfq_io_cq *bic = RQ_BIC(rq); - if (rq->cmd_flags & REQ_META) bfqq->meta_pending++; - bfq_update_io_thinktime(bfqd, bfqq); - bfq_update_has_short_ttime(bfqd, bfqq, bic); - bfq_update_io_seektime(bfqd, bfqq, rq); - - bfq_log_bfqq(bfqd, bfqq, - "rq_enqueued: has_short_ttime=%d (seeky %d)", - bfq_bfqq_has_short_ttime(bfqq), BFQQ_SEEKY(bfqq)); - bfqq->last_request_pos = blk_rq_pos(rq) + blk_rq_sectors(rq); if (bfqq == bfqd->in_service_queue && bfq_bfqq_wait_request(bfqq)) { @@ -4560,28 +6085,31 @@ static void bfq_rq_enqueued(struct bfq_data *bfqd, struct bfq_queue *bfqq, bool budget_timeout = bfq_bfqq_budget_timeout(bfqq); /* - * There is just this request queued: if the request - * is small and the queue is not to be expired, then - * just exit. + * There is just this request queued: if + * - the request is small, and + * - we are idling to boost throughput, and + * - the queue is not to be expired, + * then just exit. * * In this way, if the device is being idled to wait * for a new request from the in-service queue, we * avoid unplugging the device and committing the - * device to serve just a small request. On the - * contrary, we wait for the block layer to decide - * when to unplug the device: hopefully, new requests - * will be merged to this one quickly, then the device - * will be unplugged and larger requests will be - * dispatched. + * device to serve just a small request. In contrast + * we wait for the block layer to decide when to + * unplug the device: hopefully, new requests will be + * merged to this one quickly, then the device will be + * unplugged and larger requests will be dispatched. */ - if (small_req && !budget_timeout) + if (small_req && idling_boosts_thr_without_issues(bfqd, bfqq) && + !budget_timeout) return; /* - * A large enough request arrived, or the queue is to - * be expired: in both cases disk idling is to be - * stopped, so clear wait_request flag and reset - * timer. + * A large enough request arrived, or idling is being + * performed to preserve service guarantees, or + * finally the queue is to be expired: in all these + * cases disk idling is to be stopped, so clear + * wait_request flag and reset timer. */ bfq_clear_bfqq_wait_request(bfqq); hrtimer_try_to_cancel(&bfqd->idle_slice_timer); @@ -4599,22 +6127,38 @@ static void bfq_rq_enqueued(struct bfq_data *bfqd, struct bfq_queue *bfqq, } } +static void bfqq_request_allocated(struct bfq_queue *bfqq) +{ + struct bfq_entity *entity = &bfqq->entity; + + for_each_entity(entity) + entity->allocated++; +} + +static void bfqq_request_freed(struct bfq_queue *bfqq) +{ + struct bfq_entity *entity = &bfqq->entity; + + for_each_entity(entity) + entity->allocated--; +} + /* returns true if it causes the idle timer to be disabled */ static bool __bfq_insert_request(struct bfq_data *bfqd, struct request *rq) { struct bfq_queue *bfqq = RQ_BFQQ(rq), - *new_bfqq = bfq_setup_cooperator(bfqd, bfqq, rq, true); + *new_bfqq = bfq_setup_cooperator(bfqd, bfqq, rq, true, + RQ_BIC(rq)); bool waiting, idle_timer_disabled = false; if (new_bfqq) { - if (bic_to_bfqq(RQ_BIC(rq), 1) != bfqq) - new_bfqq = bic_to_bfqq(RQ_BIC(rq), 1); + struct bfq_queue *old_bfqq = bfqq; /* * Release the request's reference to the old bfqq * and make sure one is taken to the shared queue. */ - new_bfqq->allocated++; - bfqq->allocated--; + bfqq_request_allocated(new_bfqq); + bfqq_request_freed(bfqq); new_bfqq->ref++; /* * If the bic associated with the process @@ -4624,25 +6168,30 @@ static bool __bfq_insert_request(struct bfq_data *bfqd, struct request *rq) * then complete the merge and redirect it to * new_bfqq. */ - if (bic_to_bfqq(RQ_BIC(rq), 1) == bfqq) - bfq_merge_bfqqs(bfqd, RQ_BIC(rq), - bfqq, new_bfqq); + if (bic_to_bfqq(RQ_BIC(rq), true, + bfq_actuator_index(bfqd, rq->bio)) == bfqq) { + while (bfqq != new_bfqq) + bfqq = bfq_merge_bfqqs(bfqd, RQ_BIC(rq), bfqq); + } - bfq_clear_bfqq_just_created(bfqq); + bfq_clear_bfqq_just_created(old_bfqq); /* * rq is about to be enqueued into new_bfqq, * release rq reference on bfqq */ - bfq_put_queue(bfqq); + bfq_put_queue(old_bfqq); rq->elv.priv[1] = new_bfqq; - bfqq = new_bfqq; } + bfq_update_io_thinktime(bfqd, bfqq); + bfq_update_has_short_ttime(bfqd, bfqq, RQ_BIC(rq)); + bfq_update_io_seektime(bfqd, bfqq, rq); + waiting = bfqq && bfq_bfqq_wait_request(bfqq); bfq_add_request(rq); idle_timer_disabled = waiting && !bfq_bfqq_wait_request(bfqq); - rq->fifo_time = ktime_get_ns() + bfqd->bfq_fifo_expire[rq_is_sync(rq)]; + rq->fifo_time = blk_time_get_ns() + bfqd->bfq_fifo_expire[rq_is_sync(rq)]; list_add_tail(&rq->queuelist, &bfqq->fifo); bfq_rq_enqueued(bfqd, bfqq, rq); @@ -4650,11 +6199,11 @@ static bool __bfq_insert_request(struct bfq_data *bfqd, struct request *rq) return idle_timer_disabled; } -#if defined(CONFIG_BFQ_GROUP_IOSCHED) && defined(CONFIG_DEBUG_BLK_CGROUP) +#ifdef CONFIG_BFQ_CGROUP_DEBUG static void bfq_update_insert_stats(struct request_queue *q, struct bfq_queue *bfqq, bool idle_timer_disabled, - unsigned int cmd_flags) + blk_opf_t cmd_flags) { if (!bfqq) return; @@ -4679,36 +6228,40 @@ static void bfq_update_insert_stats(struct request_queue *q, static inline void bfq_update_insert_stats(struct request_queue *q, struct bfq_queue *bfqq, bool idle_timer_disabled, - unsigned int cmd_flags) {} -#endif + blk_opf_t cmd_flags) {} +#endif /* CONFIG_BFQ_CGROUP_DEBUG */ + +static struct bfq_queue *bfq_init_rq(struct request *rq); static void bfq_insert_request(struct blk_mq_hw_ctx *hctx, struct request *rq, - bool at_head) + blk_insert_t flags) { struct request_queue *q = hctx->queue; struct bfq_data *bfqd = q->elevator->elevator_data; struct bfq_queue *bfqq; bool idle_timer_disabled = false; - unsigned int cmd_flags; + blk_opf_t cmd_flags; + LIST_HEAD(free); +#ifdef CONFIG_BFQ_GROUP_IOSCHED + if (!cgroup_subsys_on_dfl(io_cgrp_subsys) && rq->bio) + bfqg_stats_update_legacy_io(q, rq); +#endif spin_lock_irq(&bfqd->lock); - if (blk_mq_sched_try_insert_merge(q, rq)) { + bfqq = bfq_init_rq(rq); + if (blk_mq_sched_try_insert_merge(q, rq, &free)) { spin_unlock_irq(&bfqd->lock); + blk_mq_free_requests(&free); return; } - spin_unlock_irq(&bfqd->lock); - - blk_mq_sched_request_inserted(rq); + trace_block_rq_insert(rq); - spin_lock_irq(&bfqd->lock); - bfqq = bfq_init_rq(rq); - if (at_head || blk_rq_is_passthrough(rq)) { - if (at_head) - list_add(&rq->queuelist, &bfqd->dispatch); - else - list_add_tail(&rq->queuelist, &bfqd->dispatch); - } else { /* bfqq is assumed to be non null here */ + if (flags & BLK_MQ_INSERT_AT_HEAD) { + list_add(&rq->queuelist, &bfqd->dispatch); + } else if (!bfqq) { + list_add_tail(&rq->queuelist, &bfqd->dispatch); + } else { idle_timer_disabled = __bfq_insert_request(bfqd, rq); /* * Update bfqq, because, if a queue merge has occurred @@ -4730,7 +6283,6 @@ static void bfq_insert_request(struct blk_mq_hw_ctx *hctx, struct request *rq, * merge). */ cmd_flags = rq->cmd_flags; - spin_unlock_irq(&bfqd->lock); bfq_update_insert_stats(q, bfqq, idle_timer_disabled, @@ -4738,21 +6290,24 @@ static void bfq_insert_request(struct blk_mq_hw_ctx *hctx, struct request *rq, } static void bfq_insert_requests(struct blk_mq_hw_ctx *hctx, - struct list_head *list, bool at_head) + struct list_head *list, + blk_insert_t flags) { while (!list_empty(list)) { struct request *rq; rq = list_first_entry(list, struct request, queuelist); list_del_init(&rq->queuelist); - bfq_insert_request(hctx, rq, at_head); + bfq_insert_request(hctx, rq, flags); } } static void bfq_update_hw_tag(struct bfq_data *bfqd) { + struct bfq_queue *bfqq = bfqd->in_service_queue; + bfqd->max_rq_in_driver = max_t(int, bfqd->max_rq_in_driver, - bfqd->rq_in_driver); + bfqd->tot_rq_in_driver); if (bfqd->hw_tag == 1) return; @@ -4763,7 +6318,18 @@ static void bfq_update_hw_tag(struct bfq_data *bfqd) * sum is not exact, as it's not taking into account deactivated * requests. */ - if (bfqd->rq_in_driver + bfqd->queued < BFQ_HW_QUEUE_THRESHOLD) + if (bfqd->tot_rq_in_driver + bfqd->queued <= BFQ_HW_QUEUE_THRESHOLD) + return; + + /* + * If active queue hasn't enough requests and can idle, bfq might not + * dispatch sufficient requests to hardware. Don't zero hw_tag in this + * case + */ + if (bfqq && bfq_bfqq_has_short_ttime(bfqq) && + bfqq->dispatched + bfqq->queued[0] + bfqq->queued[1] < + BFQ_HW_QUEUE_THRESHOLD && + bfqd->tot_rq_in_driver < BFQ_HW_QUEUE_THRESHOLD) return; if (bfqd->hw_tag_samples++ < BFQ_HW_QUEUE_SAMPLES) @@ -4772,6 +6338,9 @@ static void bfq_update_hw_tag(struct bfq_data *bfqd) bfqd->hw_tag = bfqd->max_rq_in_driver > BFQ_HW_QUEUE_THRESHOLD; bfqd->max_rq_in_driver = 0; bfqd->hw_tag_samples = 0; + + bfqd->nonrot_with_queueing = + blk_queue_nonrot(bfqd->queue) && bfqd->hw_tag; } static void bfq_completed_request(struct bfq_queue *bfqq, struct bfq_data *bfqd) @@ -4781,7 +6350,8 @@ static void bfq_completed_request(struct bfq_queue *bfqq, struct bfq_data *bfqd) bfq_update_hw_tag(bfqd); - bfqd->rq_in_driver--; + bfqd->rq_in_driver[bfqq->actuator_idx]--; + bfqd->tot_rq_in_driver--; bfqq->dispatched--; if (!bfqq->dispatched && !bfq_bfqq_busy(bfqq)) { @@ -4793,10 +6363,11 @@ static void bfq_completed_request(struct bfq_queue *bfqq, struct bfq_data *bfqd) */ bfqq->budget_timeout = jiffies; - bfq_weights_tree_remove(bfqd, bfqq); + bfq_del_bfqq_in_groups_with_pending_reqs(bfqq); + bfq_weights_tree_remove(bfqq); } - now_ns = ktime_get_ns(); + now_ns = blk_time_get_ns(); bfqq->ttime.last_end_request = now_ns; @@ -4827,6 +6398,19 @@ static void bfq_completed_request(struct bfq_queue *bfqq, struct bfq_data *bfqd) 1UL<<(BFQ_RATE_SHIFT - 10)) bfq_update_rate_reset(bfqd, NULL); bfqd->last_completion = now_ns; + /* + * Shared queues are likely to receive I/O at a high + * rate. This may deceptively let them be considered as wakers + * of other queues. But a false waker will unjustly steal + * bandwidth to its supposedly woken queue. So considering + * also shared queues in the waking mechanism may cause more + * control troubles than throughput benefits. Then reset + * last_completed_rq_bfqq if bfqq is a shared queue. + */ + if (!bfq_bfqq_coop(bfqq)) + bfqd->last_completed_rq_bfqq = bfqq; + else + bfqd->last_completed_rq_bfqq = NULL; /* * If we are waiting to discover whether the request pattern @@ -4834,11 +6418,14 @@ static void bfq_completed_request(struct bfq_queue *bfqq, struct bfq_data *bfqd) * isochronous, and both requisites for this condition to hold * are now satisfied, then compute soft_rt_next_start (see the * comments on the function bfq_bfqq_softrt_next_start()). We - * schedule this delayed check when bfqq expires, if it still - * has in-flight requests. + * do not compute soft_rt_next_start if bfqq is in interactive + * weight raising (see the comments in bfq_bfqq_expire() for + * an explanation). We schedule this delayed update when bfqq + * expires, if it still has in-flight requests. */ if (bfq_bfqq_softrt_update(bfqq) && bfqq->dispatched == 0 && - RB_EMPTY_ROOT(&bfqq->sort_list)) + RB_EMPTY_ROOT(&bfqq->sort_list) && + bfqq->wr_coeff != bfqd->bfq_wr_coeff) bfqq->soft_rt_next_start = bfq_bfqq_softrt_next_start(bfqd, bfqq); @@ -4884,15 +6471,169 @@ static void bfq_completed_request(struct bfq_queue *bfqq, struct bfq_data *bfqd) BFQQE_NO_MORE_REQUESTS); } - if (!bfqd->rq_in_driver) + if (!bfqd->tot_rq_in_driver) bfq_schedule_dispatch(bfqd); } -static void bfq_finish_requeue_request_body(struct bfq_queue *bfqq) +/* + * The processes associated with bfqq may happen to generate their + * cumulative I/O at a lower rate than the rate at which the device + * could serve the same I/O. This is rather probable, e.g., if only + * one process is associated with bfqq and the device is an SSD. It + * results in bfqq becoming often empty while in service. In this + * respect, if BFQ is allowed to switch to another queue when bfqq + * remains empty, then the device goes on being fed with I/O requests, + * and the throughput is not affected. In contrast, if BFQ is not + * allowed to switch to another queue---because bfqq is sync and + * I/O-dispatch needs to be plugged while bfqq is temporarily + * empty---then, during the service of bfqq, there will be frequent + * "service holes", i.e., time intervals during which bfqq gets empty + * and the device can only consume the I/O already queued in its + * hardware queues. During service holes, the device may even get to + * remaining idle. In the end, during the service of bfqq, the device + * is driven at a lower speed than the one it can reach with the kind + * of I/O flowing through bfqq. + * + * To counter this loss of throughput, BFQ implements a "request + * injection mechanism", which tries to fill the above service holes + * with I/O requests taken from other queues. The hard part in this + * mechanism is finding the right amount of I/O to inject, so as to + * both boost throughput and not break bfqq's bandwidth and latency + * guarantees. In this respect, the mechanism maintains a per-queue + * inject limit, computed as below. While bfqq is empty, the injection + * mechanism dispatches extra I/O requests only until the total number + * of I/O requests in flight---i.e., already dispatched but not yet + * completed---remains lower than this limit. + * + * A first definition comes in handy to introduce the algorithm by + * which the inject limit is computed. We define as first request for + * bfqq, an I/O request for bfqq that arrives while bfqq is in + * service, and causes bfqq to switch from empty to non-empty. The + * algorithm updates the limit as a function of the effect of + * injection on the service times of only the first requests of + * bfqq. The reason for this restriction is that these are the + * requests whose service time is affected most, because they are the + * first to arrive after injection possibly occurred. + * + * To evaluate the effect of injection, the algorithm measures the + * "total service time" of first requests. We define as total service + * time of an I/O request, the time that elapses since when the + * request is enqueued into bfqq, to when it is completed. This + * quantity allows the whole effect of injection to be measured. It is + * easy to see why. Suppose that some requests of other queues are + * actually injected while bfqq is empty, and that a new request R + * then arrives for bfqq. If the device does start to serve all or + * part of the injected requests during the service hole, then, + * because of this extra service, it may delay the next invocation of + * the dispatch hook of BFQ. Then, even after R gets eventually + * dispatched, the device may delay the actual service of R if it is + * still busy serving the extra requests, or if it decides to serve, + * before R, some extra request still present in its queues. As a + * conclusion, the cumulative extra delay caused by injection can be + * easily evaluated by just comparing the total service time of first + * requests with and without injection. + * + * The limit-update algorithm works as follows. On the arrival of a + * first request of bfqq, the algorithm measures the total time of the + * request only if one of the three cases below holds, and, for each + * case, it updates the limit as described below: + * + * (1) If there is no in-flight request. This gives a baseline for the + * total service time of the requests of bfqq. If the baseline has + * not been computed yet, then, after computing it, the limit is + * set to 1, to start boosting throughput, and to prepare the + * ground for the next case. If the baseline has already been + * computed, then it is updated, in case it results to be lower + * than the previous value. + * + * (2) If the limit is higher than 0 and there are in-flight + * requests. By comparing the total service time in this case with + * the above baseline, it is possible to know at which extent the + * current value of the limit is inflating the total service + * time. If the inflation is below a certain threshold, then bfqq + * is assumed to be suffering from no perceivable loss of its + * service guarantees, and the limit is even tentatively + * increased. If the inflation is above the threshold, then the + * limit is decreased. Due to the lack of any hysteresis, this + * logic makes the limit oscillate even in steady workload + * conditions. Yet we opted for it, because it is fast in reaching + * the best value for the limit, as a function of the current I/O + * workload. To reduce oscillations, this step is disabled for a + * short time interval after the limit happens to be decreased. + * + * (3) Periodically, after resetting the limit, to make sure that the + * limit eventually drops in case the workload changes. This is + * needed because, after the limit has gone safely up for a + * certain workload, it is impossible to guess whether the + * baseline total service time may have changed, without measuring + * it again without injection. A more effective version of this + * step might be to just sample the baseline, by interrupting + * injection only once, and then to reset/lower the limit only if + * the total service time with the current limit does happen to be + * too large. + * + * More details on each step are provided in the comments on the + * pieces of code that implement these steps: the branch handling the + * transition from empty to non empty in bfq_add_request(), the branch + * handling injection in bfq_select_queue(), and the function + * bfq_choose_bfqq_for_injection(). These comments also explain some + * exceptions, made by the injection mechanism in some special cases. + */ +static void bfq_update_inject_limit(struct bfq_data *bfqd, + struct bfq_queue *bfqq) { - bfqq->allocated--; + u64 tot_time_ns = blk_time_get_ns() - bfqd->last_empty_occupied_ns; + unsigned int old_limit = bfqq->inject_limit; - bfq_put_queue(bfqq); + if (bfqq->last_serv_time_ns > 0 && bfqd->rqs_injected) { + u64 threshold = (bfqq->last_serv_time_ns * 3)>>1; + + if (tot_time_ns >= threshold && old_limit > 0) { + bfqq->inject_limit--; + bfqq->decrease_time_jif = jiffies; + } else if (tot_time_ns < threshold && + old_limit <= bfqd->max_rq_in_driver) + bfqq->inject_limit++; + } + + /* + * Either we still have to compute the base value for the + * total service time, and there seem to be the right + * conditions to do it, or we can lower the last base value + * computed. + * + * NOTE: (bfqd->tot_rq_in_driver == 1) means that there is no I/O + * request in flight, because this function is in the code + * path that handles the completion of a request of bfqq, and, + * in particular, this function is executed before + * bfqd->tot_rq_in_driver is decremented in such a code path. + */ + if ((bfqq->last_serv_time_ns == 0 && bfqd->tot_rq_in_driver == 1) || + tot_time_ns < bfqq->last_serv_time_ns) { + if (bfqq->last_serv_time_ns == 0) { + /* + * Now we certainly have a base value: make sure we + * start trying injection. + */ + bfqq->inject_limit = max_t(unsigned int, 1, old_limit); + } + bfqq->last_serv_time_ns = tot_time_ns; + } else if (!bfqd->rqs_injected && bfqd->tot_rq_in_driver == 1) + /* + * No I/O injected and no request still in service in + * the drive: these are the exact conditions for + * computing the base value of the total service time + * for bfqq. So let's update this value, because it is + * rather variable. For example, it varies if the size + * or the spatial locality of the I/O requests in bfqq + * change. + */ + bfqq->last_serv_time_ns = tot_time_ns; + + + /* update complete, not waiting for any request completion any longer */ + bfqd->waited_rq = NULL; + bfqd->rqs_injected = false; } /* @@ -4905,18 +6646,7 @@ static void bfq_finish_requeue_request(struct request *rq) { struct bfq_queue *bfqq = RQ_BFQQ(rq); struct bfq_data *bfqd; - - /* - * Requeue and finish hooks are invoked in blk-mq without - * checking whether the involved request is actually still - * referenced in the scheduler. To handle this fact, the - * following two checks make this function exit in case of - * spurious invocations, for which there is nothing to do. - * - * First, check whether rq has nothing to do with an elevator. - */ - if (unlikely(!(rq->rq_flags & RQF_ELVPRIV))) - return; + unsigned long flags; /* * rq either is not associated with any icq, or is an already @@ -4934,36 +6664,17 @@ static void bfq_finish_requeue_request(struct request *rq) rq->io_start_time_ns, rq->cmd_flags); + spin_lock_irqsave(&bfqd->lock, flags); if (likely(rq->rq_flags & RQF_STARTED)) { - unsigned long flags; - - spin_lock_irqsave(&bfqd->lock, flags); + if (rq == bfqd->waited_rq) + bfq_update_inject_limit(bfqd, bfqq); bfq_completed_request(bfqq, bfqd); - bfq_finish_requeue_request_body(bfqq); - - spin_unlock_irqrestore(&bfqd->lock, flags); - } else { - /* - * Request rq may be still/already in the scheduler, - * in which case we need to remove it (this should - * never happen in case of requeue). And we cannot - * defer such a check and removal, to avoid - * inconsistencies in the time interval from the end - * of this function to the start of the deferred work. - * This situation seems to occur only in process - * context, as a consequence of a merge. In the - * current version of the code, this implies that the - * lock is held. - */ - - if (!RB_EMPTY_NODE(&rq->rb_node)) { - bfq_remove_request(rq->q, rq); - bfqg_stats_update_io_remove(bfqq_group(bfqq), - rq->cmd_flags); - } - bfq_finish_requeue_request_body(bfqq); } + bfqq_request_freed(bfqq); + bfq_put_queue(bfqq); + RQ_BIC(rq)->requests--; + spin_unlock_irqrestore(&bfqd->lock, flags); /* * Reset private fields. In case of a requeue, this allows @@ -4986,7 +6697,19 @@ static void bfq_finish_requeue_request(struct request *rq) rq->elv.priv[1] = NULL; } +static void bfq_finish_request(struct request *rq) +{ + bfq_finish_requeue_request(rq); + + if (rq->elv.icq) { + put_io_context(rq->elv.icq->ioc); + rq->elv.icq = NULL; + } +} + /* + * Removes the association between the current task and bfqq, assuming + * that bic points to the bfq iocontext of the task. * Returns NULL if a new bfqq should be allocated, or the old bfqq if this * was the last process referring to that bfqq. */ @@ -4995,28 +6718,29 @@ bfq_split_bfqq(struct bfq_io_cq *bic, struct bfq_queue *bfqq) { bfq_log_bfqq(bfqq->bfqd, bfqq, "splitting queue"); - if (bfqq_process_refs(bfqq) == 1) { + if (bfqq_process_refs(bfqq) == 1 && !bfqq->new_bfqq) { bfqq->pid = current->pid; bfq_clear_bfqq_coop(bfqq); bfq_clear_bfqq_split_coop(bfqq); return bfqq; } - bic_set_bfqq(bic, NULL, 1); + bic_set_bfqq(bic, NULL, true, bfqq->actuator_idx); bfq_put_cooperator(bfqq); - bfq_put_queue(bfqq); + bfq_release_process_ref(bfqq->bfqd, bfqq); return NULL; } -static struct bfq_queue *bfq_get_bfqq_handle_split(struct bfq_data *bfqd, - struct bfq_io_cq *bic, - struct bio *bio, - bool split, bool is_sync, - bool *new_queue) +static struct bfq_queue * +__bfq_get_bfqq_handle_split(struct bfq_data *bfqd, struct bfq_io_cq *bic, + struct bio *bio, bool split, bool is_sync, + bool *new_queue) { - struct bfq_queue *bfqq = bic_to_bfqq(bic, is_sync); + unsigned int act_idx = bfq_actuator_index(bfqd, bio); + struct bfq_queue *bfqq = bic_to_bfqq(bic, is_sync, act_idx); + struct bfq_iocq_bfqq_data *bfqq_data = &bic->bfqq_data[act_idx]; if (likely(bfqq && bfqq != &bfqd->oom_bfqq)) return bfqq; @@ -5026,16 +6750,16 @@ static struct bfq_queue *bfq_get_bfqq_handle_split(struct bfq_data *bfqd, if (bfqq) bfq_put_queue(bfqq); - bfqq = bfq_get_queue(bfqd, bio, is_sync, bic); + bfqq = bfq_get_queue(bfqd, bio, is_sync, bic, split); - bic_set_bfqq(bic, bfqq, is_sync); + bic_set_bfqq(bic, bfqq, is_sync, act_idx); if (split && is_sync) { - if ((bic->was_in_burst_list && bfqd->large_burst) || - bic->saved_in_large_burst) + if ((bfqq_data->was_in_burst_list && bfqd->large_burst) || + bfqq_data->saved_in_large_burst) bfq_mark_bfqq_in_large_burst(bfqq); else { bfq_clear_bfqq_in_large_burst(bfqq); - if (bic->was_in_burst_list) + if (bfqq_data->was_in_burst_list) /* * If bfqq was in the current * burst list before being @@ -5079,8 +6803,10 @@ static struct bfq_queue *bfq_get_bfqq_handle_split(struct bfq_data *bfqd, * comments on bfq_init_rq for the reason behind this delayed * preparation. */ -static void bfq_prepare_request(struct request *rq, struct bio *bio) +static void bfq_prepare_request(struct request *rq) { + rq->elv.icq = ioc_find_get_icq(rq->q); + /* * Regardless of whether we have an icq attached, we have to * clear the scheduler pointers, as they might point to @@ -5089,6 +6815,92 @@ static void bfq_prepare_request(struct request *rq, struct bio *bio) rq->elv.priv[0] = rq->elv.priv[1] = NULL; } +static struct bfq_queue *bfq_waker_bfqq(struct bfq_queue *bfqq) +{ + struct bfq_queue *new_bfqq = bfqq->new_bfqq; + struct bfq_queue *waker_bfqq = bfqq->waker_bfqq; + + if (!waker_bfqq) + return NULL; + + while (new_bfqq) { + if (new_bfqq == waker_bfqq) { + /* + * If waker_bfqq is in the merge chain, and current + * is the only process, waker_bfqq can be freed. + */ + if (bfqq_process_refs(waker_bfqq) == 1) + return NULL; + + return waker_bfqq; + } + + new_bfqq = new_bfqq->new_bfqq; + } + + /* + * If waker_bfqq is not in the merge chain, and it's procress reference + * is 0, waker_bfqq can be freed. + */ + if (bfqq_process_refs(waker_bfqq) == 0) + return NULL; + + return waker_bfqq; +} + +static struct bfq_queue *bfq_get_bfqq_handle_split(struct bfq_data *bfqd, + struct bfq_io_cq *bic, + struct bio *bio, + unsigned int idx, + bool is_sync) +{ + struct bfq_queue *waker_bfqq; + struct bfq_queue *bfqq; + bool new_queue = false; + + bfqq = __bfq_get_bfqq_handle_split(bfqd, bic, bio, false, is_sync, + &new_queue); + if (unlikely(new_queue)) + return bfqq; + + /* If the queue was seeky for too long, break it apart. */ + if (!bfq_bfqq_coop(bfqq) || !bfq_bfqq_split_coop(bfqq) || + bic->bfqq_data[idx].stably_merged) + return bfqq; + + waker_bfqq = bfq_waker_bfqq(bfqq); + + /* Update bic before losing reference to bfqq */ + if (bfq_bfqq_in_large_burst(bfqq)) + bic->bfqq_data[idx].saved_in_large_burst = true; + + bfqq = bfq_split_bfqq(bic, bfqq); + if (bfqq) { + bfq_bfqq_resume_state(bfqq, bfqd, bic, true); + return bfqq; + } + + bfqq = __bfq_get_bfqq_handle_split(bfqd, bic, bio, true, is_sync, NULL); + if (unlikely(bfqq == &bfqd->oom_bfqq)) + return bfqq; + + bfq_bfqq_resume_state(bfqq, bfqd, bic, false); + bfqq->waker_bfqq = waker_bfqq; + bfqq->tentative_waker_bfqq = NULL; + + /* + * If the waker queue disappears, then new_bfqq->waker_bfqq must be + * reset. So insert new_bfqq into the + * woken_list of the waker. See + * bfq_check_waker for details. + */ + if (waker_bfqq) + hlist_add_head(&bfqq->woken_list_node, + &bfqq->waker_bfqq->woken_list); + + return bfqq; +} + /* * If needed, init rq, allocate bfq data structures associated with * rq, and increment reference counters in the destination bfq_queue @@ -5102,7 +6914,7 @@ static void bfq_prepare_request(struct request *rq, struct bio *bio) * preparation is that, after the prepare_request hook is invoked for * rq, rq may still be transformed into a request with no icq, i.e., a * request not associated with any queue. No bfq hook is invoked to - * signal this tranformation. As a consequence, should these + * signal this transformation. As a consequence, should these * preparation operations be performed when the prepare_request hook * is invoked, and should rq be transformed one moment later, bfq * would end up in an inconsistent state, because it would have @@ -5120,54 +6932,29 @@ static struct bfq_queue *bfq_init_rq(struct request *rq) struct bfq_io_cq *bic; const int is_sync = rq_is_sync(rq); struct bfq_queue *bfqq; - bool new_queue = false; - bool bfqq_already_existing = false, split = false; + unsigned int a_idx = bfq_actuator_index(bfqd, bio); if (unlikely(!rq->elv.icq)) return NULL; /* - * Assuming that elv.priv[1] is set only if everything is set + * Assuming that RQ_BFQQ(rq) is set only if everything is set * for this rq. This holds true, because this function is * invoked only for insertion or merging, and, after such * events, a request cannot be manipulated any longer before * being removed from bfq. */ - if (rq->elv.priv[1]) - return rq->elv.priv[1]; + if (RQ_BFQQ(rq)) + return RQ_BFQQ(rq); bic = icq_to_bic(rq->elv.icq); - bfq_check_ioprio_change(bic, bio); - bfq_bic_update_cgroup(bic, bio); + bfqq = bfq_get_bfqq_handle_split(bfqd, bic, bio, a_idx, is_sync); - bfqq = bfq_get_bfqq_handle_split(bfqd, bic, bio, false, is_sync, - &new_queue); - - if (likely(!new_queue)) { - /* If the queue was seeky for too long, break it apart. */ - if (bfq_bfqq_coop(bfqq) && bfq_bfqq_split_coop(bfqq)) { - bfq_log_bfqq(bfqd, bfqq, "breaking apart bfqq"); - - /* Update bic before losing reference to bfqq */ - if (bfq_bfqq_in_large_burst(bfqq)) - bic->saved_in_large_burst = true; - - bfqq = bfq_split_bfqq(bic, bfqq); - split = true; - - if (!bfqq) - bfqq = bfq_get_bfqq_handle_split(bfqd, bic, bio, - true, is_sync, - NULL); - else - bfqq_already_existing = true; - } - } - - bfqq->allocated++; + bfqq_request_allocated(bfqq); bfqq->ref++; + bic->requests++; bfq_log_bfqq(bfqd, bfqq, "get_request %p: bfqq %p, %d", rq, bfqq, bfqq->ref); @@ -5180,39 +6967,60 @@ static struct bfq_queue *bfq_init_rq(struct request *rq) * addition, if the queue has also just been split, we have to * resume its state. */ - if (likely(bfqq != &bfqd->oom_bfqq) && bfqq_process_refs(bfqq) == 1) { + if (likely(bfqq != &bfqd->oom_bfqq) && !bfqq->new_bfqq && + bfqq_process_refs(bfqq) == 1) bfqq->bic = bic; - if (split) { - /* - * The queue has just been split from a shared - * queue: restore the idle window and the - * possible weight raising period. - */ - bfq_bfqq_resume_state(bfqq, bfqd, bic, - bfqq_already_existing); - } - } - if (unlikely(bfq_bfqq_just_created(bfqq))) + /* + * Consider bfqq as possibly belonging to a burst of newly + * created queues only if: + * 1) A burst is actually happening (bfqd->burst_size > 0) + * or + * 2) There is no other active queue. In fact, if, in + * contrast, there are active queues not belonging to the + * possible burst bfqq may belong to, then there is no gain + * in considering bfqq as belonging to a burst, and + * therefore in not weight-raising bfqq. See comments on + * bfq_handle_burst(). + * + * This filtering also helps eliminating false positives, + * occurring when bfqq does not belong to an actual large + * burst, but some background task (e.g., a service) happens + * to trigger the creation of new queues very close to when + * bfqq and its possible companion queues are created. See + * comments on bfq_handle_burst() for further details also on + * this issue. + */ + if (unlikely(bfq_bfqq_just_created(bfqq) && + (bfqd->burst_size > 0 || + bfq_tot_busy_queues(bfqd) == 0))) bfq_handle_burst(bfqd, bfqq); return bfqq; } -static void bfq_idle_slice_timer_body(struct bfq_queue *bfqq) +static void +bfq_idle_slice_timer_body(struct bfq_data *bfqd, struct bfq_queue *bfqq) { - struct bfq_data *bfqd = bfqq->bfqd; enum bfqq_expiration reason; unsigned long flags; spin_lock_irqsave(&bfqd->lock, flags); - bfq_clear_bfqq_wait_request(bfqq); + /* + * Considering that bfqq may be in race, we should firstly check + * whether bfqq is in service before doing something on it. If + * the bfqq in race is not in service, it has already been expired + * through __bfq_bfqq_expire func and its wait_request flags has + * been cleared in __bfq_bfqd_reset_in_service func. + */ if (bfqq != bfqd->in_service_queue) { spin_unlock_irqrestore(&bfqd->lock, flags); return; } + bfq_clear_bfqq_wait_request(bfqq); + if (bfq_bfqq_budget_timeout(bfqq)) /* * Also here the queue can be safely expired @@ -5234,8 +7042,8 @@ static void bfq_idle_slice_timer_body(struct bfq_queue *bfqq) bfq_bfqq_expire(bfqd, bfqq, true, reason); schedule_dispatch: - spin_unlock_irqrestore(&bfqd->lock, flags); bfq_schedule_dispatch(bfqd); + spin_unlock_irqrestore(&bfqd->lock, flags); } /* @@ -5257,7 +7065,7 @@ static enum hrtimer_restart bfq_idle_slice_timer(struct hrtimer *timer) * early. */ if (bfqq) - bfq_idle_slice_timer_body(bfqq); + bfq_idle_slice_timer_body(bfqd, bfqq); return HRTIMER_NORESTART; } @@ -5286,23 +7094,25 @@ static void __bfq_put_async_bfqq(struct bfq_data *bfqd, */ void bfq_put_async_queues(struct bfq_data *bfqd, struct bfq_group *bfqg) { - int i, j; + int i, j, k; - for (i = 0; i < 2; i++) - for (j = 0; j < IOPRIO_BE_NR; j++) - __bfq_put_async_bfqq(bfqd, &bfqg->async_bfqq[i][j]); + for (k = 0; k < bfqd->num_actuators; k++) { + for (i = 0; i < 2; i++) + for (j = 0; j < IOPRIO_NR_LEVELS; j++) + __bfq_put_async_bfqq(bfqd, &bfqg->async_bfqq[i][j][k]); - __bfq_put_async_bfqq(bfqd, &bfqg->async_idle_bfqq); + __bfq_put_async_bfqq(bfqd, &bfqg->async_idle_bfqq[k]); + } } /* * See the comments on bfq_limit_depth for the purpose of * the depths set in the function. Return minimum shallow depth we'll use. */ -static unsigned int bfq_update_depths(struct bfq_data *bfqd, - struct sbitmap_queue *bt) +static void bfq_depth_updated(struct request_queue *q) { - unsigned int i, j, min_shallow = UINT_MAX; + struct bfq_data *bfqd = q->elevator->elevator_data; + unsigned int nr_requests = q->nr_requests; /* * In-word depths if no bfq_queue is being weight-raised: @@ -5315,13 +7125,13 @@ static unsigned int bfq_update_depths(struct bfq_data *bfqd, * limit 'something'. */ /* no more than 50% of tags for async I/O */ - bfqd->word_depths[0][0] = max((1U << bt->sb.shift) >> 1, 1U); + bfqd->async_depths[0][0] = max(nr_requests >> 1, 1U); /* * no more than 75% of tags for sync writes (25% extra tags * w.r.t. async I/O, to prevent async I/O from starving sync * writes) */ - bfqd->word_depths[0][1] = max(((1U << bt->sb.shift) * 3) >> 2, 1U); + bfqd->async_depths[0][1] = max((nr_requests * 3) >> 2, 1U); /* * In-word depths in case some bfq_queue is being weight- @@ -5331,32 +7141,18 @@ static unsigned int bfq_update_depths(struct bfq_data *bfqd, * shortage. */ /* no more than ~18% of tags for async I/O */ - bfqd->word_depths[1][0] = max(((1U << bt->sb.shift) * 3) >> 4, 1U); + bfqd->async_depths[1][0] = max((nr_requests * 3) >> 4, 1U); /* no more than ~37% of tags for sync writes (~20% extra tags) */ - bfqd->word_depths[1][1] = max(((1U << bt->sb.shift) * 6) >> 4, 1U); - - for (i = 0; i < 2; i++) - for (j = 0; j < 2; j++) - min_shallow = min(min_shallow, bfqd->word_depths[i][j]); + bfqd->async_depths[1][1] = max((nr_requests * 6) >> 4, 1U); - return min_shallow; -} - -static int bfq_init_hctx(struct blk_mq_hw_ctx *hctx, unsigned int index) -{ - struct bfq_data *bfqd = hctx->queue->elevator->elevator_data; - struct blk_mq_tags *tags = hctx->sched_tags; - unsigned int min_shallow; - - min_shallow = bfq_update_depths(bfqd, &tags->bitmap_tags); - sbitmap_queue_min_shallow_depth(&tags->bitmap_tags, min_shallow); - return 0; + blk_mq_set_min_shallow_depth(q, 1); } static void bfq_exit_queue(struct elevator_queue *e) { struct bfq_data *bfqd = e->elevator_data; struct bfq_queue *bfqq, *n; + unsigned int actuator; hrtimer_cancel(&bfqd->idle_slice_timer); @@ -5365,13 +7161,17 @@ static void bfq_exit_queue(struct elevator_queue *e) bfq_deactivate_bfqq(bfqd, bfqq, false, false); spin_unlock_irq(&bfqd->lock); + for (actuator = 0; actuator < bfqd->num_actuators; actuator++) + WARN_ON_ONCE(bfqd->rq_in_driver[actuator]); + WARN_ON_ONCE(bfqd->tot_rq_in_driver); + hrtimer_cancel(&bfqd->idle_slice_timer); -#ifdef CONFIG_BFQ_GROUP_IOSCHED /* release oom-queue reference to root group */ bfqg_and_blkg_put(bfqd->root_group); - blkcg_deactivate_policy(bfqd->queue, &blkcg_policy_bfq); +#ifdef CONFIG_BFQ_GROUP_IOSCHED + blkcg_deactivate_policy(bfqd->queue->disk, &blkcg_policy_bfq); #else spin_lock_irq(&bfqd->lock); bfq_put_async_queues(bfqd, bfqd->root_group); @@ -5379,6 +7179,10 @@ static void bfq_exit_queue(struct elevator_queue *e) spin_unlock_irq(&bfqd->lock); #endif + blk_stat_disable_accounting(bfqd->queue); + blk_queue_flag_clear(QUEUE_FLAG_DISABLE_WBT_DEF, bfqd->queue); + set_bit(ELEVATOR_FLAG_ENABLE_WBT_ON_EXIT, &e->flags); + kfree(bfqd); } @@ -5398,20 +7202,16 @@ static void bfq_init_root_group(struct bfq_group *root_group, root_group->sched_data.bfq_class_idle_last_service = jiffies; } -static int bfq_init_queue(struct request_queue *q, struct elevator_type *e) +static int bfq_init_queue(struct request_queue *q, struct elevator_queue *eq) { struct bfq_data *bfqd; - struct elevator_queue *eq; - - eq = elevator_alloc(q, e); - if (!eq) - return -ENOMEM; + unsigned int i; + struct blk_independent_access_ranges *ia_ranges = q->disk->ia_ranges; bfqd = kzalloc_node(sizeof(*bfqd), GFP_KERNEL, q->node); - if (!bfqd) { - kobject_put(&eq->kobj); + if (!bfqd) return -ENOMEM; - } + eq->elevator_data = bfqd; spin_lock_irq(&q->queue_lock); @@ -5422,8 +7222,10 @@ static int bfq_init_queue(struct request_queue *q, struct elevator_type *e) * Our fallback bfqq if bfq_find_alloc_queue() runs into OOM issues. * Grab a permanent reference to it, so that the normal code flow * will not attempt to free it. + * Set zero as actuator index: we will pretend that + * all I/O requests are for the same actuator. */ - bfq_init_bfqq(bfqd, &bfqd->oom_bfqq, NULL, 1, 0); + bfq_init_bfqq(bfqd, &bfqd->oom_bfqq, NULL, 1, 0, 0); bfqd->oom_bfqq.ref++; bfqd->oom_bfqq.new_ioprio = BFQ_DEFAULT_QUEUE_IOPRIO; bfqd->oom_bfqq.new_ioprio_class = IOPRIO_CLASS_BE; @@ -5442,20 +7244,56 @@ static int bfq_init_queue(struct request_queue *q, struct elevator_type *e) bfqd->queue = q; + bfqd->num_actuators = 1; + /* + * If the disk supports multiple actuators, copy independent + * access ranges from the request queue structure. + */ + spin_lock_irq(&q->queue_lock); + if (ia_ranges) { + /* + * Check if the disk ia_ranges size exceeds the current bfq + * actuator limit. + */ + if (ia_ranges->nr_ia_ranges > BFQ_MAX_ACTUATORS) { + pr_crit("nr_ia_ranges higher than act limit: iars=%d, max=%d.\n", + ia_ranges->nr_ia_ranges, BFQ_MAX_ACTUATORS); + pr_crit("Falling back to single actuator mode.\n"); + } else { + bfqd->num_actuators = ia_ranges->nr_ia_ranges; + + for (i = 0; i < bfqd->num_actuators; i++) { + bfqd->sector[i] = ia_ranges->ia_range[i].sector; + bfqd->nr_sectors[i] = + ia_ranges->ia_range[i].nr_sectors; + } + } + } + + /* Otherwise use single-actuator dev info */ + if (bfqd->num_actuators == 1) { + bfqd->sector[0] = 0; + bfqd->nr_sectors[0] = get_capacity(q->disk); + } + spin_unlock_irq(&q->queue_lock); + INIT_LIST_HEAD(&bfqd->dispatch); - hrtimer_init(&bfqd->idle_slice_timer, CLOCK_MONOTONIC, - HRTIMER_MODE_REL); - bfqd->idle_slice_timer.function = bfq_idle_slice_timer; + hrtimer_setup(&bfqd->idle_slice_timer, bfq_idle_slice_timer, CLOCK_MONOTONIC, + HRTIMER_MODE_REL); - bfqd->queue_weights_tree = RB_ROOT; + bfqd->queue_weights_tree = RB_ROOT_CACHED; +#ifdef CONFIG_BFQ_GROUP_IOSCHED bfqd->num_groups_with_pending_reqs = 0; +#endif - INIT_LIST_HEAD(&bfqd->active_list); + INIT_LIST_HEAD(&bfqd->active_list[0]); + INIT_LIST_HEAD(&bfqd->active_list[1]); INIT_LIST_HEAD(&bfqd->idle_list); INIT_HLIST_HEAD(&bfqd->burst_list); bfqd->hw_tag = -1; + bfqd->nonrot_with_queueing = blk_queue_nonrot(bfqd->queue); bfqd->bfq_max_budget = bfq_default_max_budget; @@ -5466,8 +7304,6 @@ static int bfq_init_queue(struct request_queue *q, struct elevator_type *e) bfqd->bfq_slice_idle = bfq_slice_idle; bfqd->bfq_timeout = bfq_timeout; - bfqd->bfq_requests_within_timer = 120; - bfqd->bfq_large_burst_thresh = 8; bfqd->bfq_burst_interval = msecs_to_jiffies(180); @@ -5478,7 +7314,6 @@ static int bfq_init_queue(struct request_queue *q, struct elevator_type *e) */ bfqd->bfq_wr_coeff = 30; bfqd->bfq_wr_rt_max_time = msecs_to_jiffies(300); - bfqd->bfq_wr_max_time = 0; bfqd->bfq_wr_min_idle_time = msecs_to_jiffies(2000); bfqd->bfq_wr_min_inter_arr_async = msecs_to_jiffies(500); bfqd->bfq_wr_max_softrt_rate = 7000; /* @@ -5497,6 +7332,9 @@ static int bfq_init_queue(struct request_queue *q, struct elevator_type *e) ref_wr_duration[blk_queue_nonrot(bfqd->queue)]; bfqd->peak_rate = ref_rate[blk_queue_nonrot(bfqd->queue)] * 2 / 3; + /* see comments on the definition of next field inside bfq_data */ + bfqd->actuator_load_threshold = 4; + spin_lock_init(&bfqd->lock); /* @@ -5519,13 +7357,19 @@ static int bfq_init_queue(struct request_queue *q, struct elevator_type *e) goto out_free; bfq_init_root_group(bfqd->root_group, bfqd); bfq_init_entity(&bfqd->oom_bfqq.entity, bfqd->root_group); + bfq_depth_updated(q); + + /* We dispatch from request queue wide instead of hw queue */ + blk_queue_flag_set(QUEUE_FLAG_SQ_SCHED, q); + + blk_queue_flag_set(QUEUE_FLAG_DISABLE_WBT_DEF, q); + wbt_disable_default(q->disk); + blk_stat_enable_accounting(q); - wbt_disable_default(q); return 0; out_free: kfree(bfqd); - kobject_put(&eq->kobj); return -ENOMEM; } @@ -5741,7 +7585,7 @@ static ssize_t bfq_low_latency_store(struct elevator_queue *e, #define BFQ_ATTR(name) \ __ATTR(name, 0644, bfq_##name##_show, bfq_##name##_store) -static struct elv_fs_entry bfq_attrs[] = { +static const struct elv_fs_entry bfq_attrs[] = { BFQ_ATTR(fifo_expire_sync), BFQ_ATTR(fifo_expire_async), BFQ_ATTR(back_seek_max), @@ -5760,7 +7604,7 @@ static struct elevator_type iosched_bfq_mq = { .limit_depth = bfq_limit_depth, .prepare_request = bfq_prepare_request, .requeue_request = bfq_finish_requeue_request, - .finish_request = bfq_finish_requeue_request, + .finish_request = bfq_finish_request, .exit_icq = bfq_exit_icq, .insert_requests = bfq_insert_requests, .dispatch_request = bfq_dispatch_request, @@ -5772,7 +7616,7 @@ static struct elevator_type iosched_bfq_mq = { .requests_merged = bfq_requests_merged, .request_merged = bfq_request_merged, .has_work = bfq_has_work, - .init_hctx = bfq_init_hctx, + .depth_updated = bfq_depth_updated, .init_sched = bfq_init_queue, .exit_sched = bfq_exit_queue, }, |