2 * CFQ, or complete fairness queueing, disk scheduler.
4 * Based on ideas from a previously unfinished io
5 * scheduler (round robin per-process disk scheduling) and Andrea Arcangeli.
7 * Copyright (C) 2003 Jens Axboe <axboe@kernel.dk>
9 #include <linux/module.h>
10 #include <linux/slab.h>
11 #include <linux/blkdev.h>
12 #include <linux/elevator.h>
13 #include <linux/ktime.h>
14 #include <linux/rbtree.h>
15 #include <linux/ioprio.h>
16 #include <linux/blktrace_api.h>
17 #include <linux/blk-cgroup.h>
23 /* max queue in one round of service */
24 static const int cfq_quantum = 8;
25 static const u64 cfq_fifo_expire[2] = { NSEC_PER_SEC / 4, NSEC_PER_SEC / 8 };
26 /* maximum backwards seek, in KiB */
27 static const int cfq_back_max = 16 * 1024;
28 /* penalty of a backwards seek */
29 static const int cfq_back_penalty = 2;
30 static const u64 cfq_slice_sync = NSEC_PER_SEC / 10;
31 static u64 cfq_slice_async = NSEC_PER_SEC / 25;
32 static const int cfq_slice_async_rq = 2;
33 static u64 cfq_slice_idle = NSEC_PER_SEC / 125;
34 static u64 cfq_group_idle = NSEC_PER_SEC / 125;
35 static const u64 cfq_target_latency = (u64)NSEC_PER_SEC * 3/10; /* 300 ms */
36 static const int cfq_hist_divisor = 4;
39 * offset from end of service tree
41 #define CFQ_IDLE_DELAY (NSEC_PER_SEC / 5)
44 * below this threshold, we consider thinktime immediate
46 #define CFQ_MIN_TT (2 * NSEC_PER_SEC / HZ)
48 #define CFQ_SLICE_SCALE (5)
49 #define CFQ_HW_QUEUE_MIN (5)
50 #define CFQ_SERVICE_SHIFT 12
52 #define CFQQ_SEEK_THR (sector_t)(8 * 100)
53 #define CFQQ_CLOSE_THR (sector_t)(8 * 1024)
54 #define CFQQ_SECT_THR_NONROT (sector_t)(2 * 32)
55 #define CFQQ_SEEKY(cfqq) (hweight32(cfqq->seek_history) > 32/8)
57 #define RQ_CIC(rq) icq_to_cic((rq)->elv.icq)
58 #define RQ_CFQQ(rq) (struct cfq_queue *) ((rq)->elv.priv[0])
59 #define RQ_CFQG(rq) (struct cfq_group *) ((rq)->elv.priv[1])
61 static struct kmem_cache *cfq_pool;
63 #define CFQ_PRIO_LISTS IOPRIO_BE_NR
64 #define cfq_class_idle(cfqq) ((cfqq)->ioprio_class == IOPRIO_CLASS_IDLE)
65 #define cfq_class_rt(cfqq) ((cfqq)->ioprio_class == IOPRIO_CLASS_RT)
67 #define sample_valid(samples) ((samples) > 80)
68 #define rb_entry_cfqg(node) rb_entry((node), struct cfq_group, rb_node)
70 /* blkio-related constants */
71 #define CFQ_WEIGHT_LEGACY_MIN 10
72 #define CFQ_WEIGHT_LEGACY_DFL 500
73 #define CFQ_WEIGHT_LEGACY_MAX 1000
80 unsigned long ttime_samples;
84 * Most of our rbtree usage is for sorting with min extraction, so
85 * if we cache the leftmost node we don't have to walk down the tree
86 * to find it. Idea borrowed from Ingo Molnars CFS scheduler. We should
87 * move this into the elevator for the rq sorting as well.
94 struct cfq_ttime ttime;
96 #define CFQ_RB_ROOT (struct cfq_rb_root) { .rb = RB_ROOT, \
97 .ttime = {.last_end_request = ktime_get_ns(),},}
100 * Per process-grouping structure
103 /* reference count */
105 /* various state flags, see below */
107 /* parent cfq_data */
108 struct cfq_data *cfqd;
109 /* service_tree member */
110 struct rb_node rb_node;
111 /* service_tree key */
113 /* prio tree member */
114 struct rb_node p_node;
115 /* prio tree root we belong to, if any */
116 struct rb_root *p_root;
117 /* sorted list of pending requests */
118 struct rb_root sort_list;
119 /* if fifo isn't expired, next request to serve */
120 struct request *next_rq;
121 /* requests queued in sort_list */
123 /* currently allocated requests */
125 /* fifo list of requests in sort_list */
126 struct list_head fifo;
128 /* time when queue got scheduled in to dispatch first request. */
132 /* time when first request from queue completed and slice started. */
137 /* pending priority requests */
139 /* number of requests that are on the dispatch list or inside driver */
142 /* io prio of this group */
143 unsigned short ioprio, org_ioprio;
144 unsigned short ioprio_class, org_ioprio_class;
149 sector_t last_request_pos;
151 struct cfq_rb_root *service_tree;
152 struct cfq_queue *new_cfqq;
153 struct cfq_group *cfqg;
154 /* Number of sectors dispatched from queue in single dispatch round */
155 unsigned long nr_sectors;
159 * First index in the service_trees.
160 * IDLE is handled separately, so it has negative index
170 * Second index in the service_trees.
174 SYNC_NOIDLE_WORKLOAD = 1,
179 #ifdef CONFIG_CFQ_GROUP_IOSCHED
180 /* number of ios merged */
181 struct blkg_rwstat merged;
182 /* total time spent on device in ns, may not be accurate w/ queueing */
183 struct blkg_rwstat service_time;
184 /* total time spent waiting in scheduler queue in ns */
185 struct blkg_rwstat wait_time;
186 /* number of IOs queued up */
187 struct blkg_rwstat queued;
188 /* total disk time and nr sectors dispatched by this group */
189 struct blkg_stat time;
190 #ifdef CONFIG_DEBUG_BLK_CGROUP
191 /* time not charged to this cgroup */
192 struct blkg_stat unaccounted_time;
193 /* sum of number of ios queued across all samples */
194 struct blkg_stat avg_queue_size_sum;
195 /* count of samples taken for average */
196 struct blkg_stat avg_queue_size_samples;
197 /* how many times this group has been removed from service tree */
198 struct blkg_stat dequeue;
199 /* total time spent waiting for it to be assigned a timeslice. */
200 struct blkg_stat group_wait_time;
201 /* time spent idling for this blkcg_gq */
202 struct blkg_stat idle_time;
203 /* total time with empty current active q with other requests queued */
204 struct blkg_stat empty_time;
205 /* fields after this shouldn't be cleared on stat reset */
206 uint64_t start_group_wait_time;
207 uint64_t start_idle_time;
208 uint64_t start_empty_time;
210 #endif /* CONFIG_DEBUG_BLK_CGROUP */
211 #endif /* CONFIG_CFQ_GROUP_IOSCHED */
214 /* Per-cgroup data */
215 struct cfq_group_data {
216 /* must be the first member */
217 struct blkcg_policy_data cpd;
220 unsigned int leaf_weight;
223 /* This is per cgroup per device grouping structure */
225 /* must be the first member */
226 struct blkg_policy_data pd;
228 /* group service_tree member */
229 struct rb_node rb_node;
231 /* group service_tree key */
235 * The number of active cfqgs and sum of their weights under this
236 * cfqg. This covers this cfqg's leaf_weight and all children's
237 * weights, but does not cover weights of further descendants.
239 * If a cfqg is on the service tree, it's active. An active cfqg
240 * also activates its parent and contributes to the children_weight
244 unsigned int children_weight;
247 * vfraction is the fraction of vdisktime that the tasks in this
248 * cfqg are entitled to. This is determined by compounding the
249 * ratios walking up from this cfqg to the root.
251 * It is in fixed point w/ CFQ_SERVICE_SHIFT and the sum of all
252 * vfractions on a service tree is approximately 1. The sum may
253 * deviate a bit due to rounding errors and fluctuations caused by
254 * cfqgs entering and leaving the service tree.
256 unsigned int vfraction;
259 * There are two weights - (internal) weight is the weight of this
260 * cfqg against the sibling cfqgs. leaf_weight is the wight of
261 * this cfqg against the child cfqgs. For the root cfqg, both
262 * weights are kept in sync for backward compatibility.
265 unsigned int new_weight;
266 unsigned int dev_weight;
268 unsigned int leaf_weight;
269 unsigned int new_leaf_weight;
270 unsigned int dev_leaf_weight;
272 /* number of cfqq currently on this group */
276 * Per group busy queues average. Useful for workload slice calc. We
277 * create the array for each prio class but at run time it is used
278 * only for RT and BE class and slot for IDLE class remains unused.
279 * This is primarily done to avoid confusion and a gcc warning.
281 unsigned int busy_queues_avg[CFQ_PRIO_NR];
283 * rr lists of queues with requests. We maintain service trees for
284 * RT and BE classes. These trees are subdivided in subclasses
285 * of SYNC, SYNC_NOIDLE and ASYNC based on workload type. For IDLE
286 * class there is no subclassification and all the cfq queues go on
287 * a single tree service_tree_idle.
288 * Counts are embedded in the cfq_rb_root
290 struct cfq_rb_root service_trees[2][3];
291 struct cfq_rb_root service_tree_idle;
294 enum wl_type_t saved_wl_type;
295 enum wl_class_t saved_wl_class;
297 /* number of requests that are on the dispatch list or inside driver */
299 struct cfq_ttime ttime;
300 struct cfqg_stats stats; /* stats for this cfqg */
302 /* async queue for each priority case */
303 struct cfq_queue *async_cfqq[2][IOPRIO_BE_NR];
304 struct cfq_queue *async_idle_cfqq;
309 struct io_cq icq; /* must be the first member */
310 struct cfq_queue *cfqq[2];
311 struct cfq_ttime ttime;
312 int ioprio; /* the current ioprio */
313 #ifdef CONFIG_CFQ_GROUP_IOSCHED
314 uint64_t blkcg_serial_nr; /* the current blkcg serial */
319 * Per block device queue structure
322 struct request_queue *queue;
323 /* Root service tree for cfq_groups */
324 struct cfq_rb_root grp_service_tree;
325 struct cfq_group *root_group;
328 * The priority currently being served
330 enum wl_class_t serving_wl_class;
331 enum wl_type_t serving_wl_type;
332 u64 workload_expires;
333 struct cfq_group *serving_group;
336 * Each priority tree is sorted by next_request position. These
337 * trees are used when determining if two or more queues are
338 * interleaving requests (see cfq_close_cooperator).
340 struct rb_root prio_trees[CFQ_PRIO_LISTS];
342 unsigned int busy_queues;
343 unsigned int busy_sync_queues;
349 * queue-depth detection
355 * -1 => indeterminate, (cfq will behave as if NCQ is present, to allow better detection)
356 * 1 => NCQ is present (hw_tag_est_depth is the estimated max depth)
359 int hw_tag_est_depth;
360 unsigned int hw_tag_samples;
363 * idle window management
365 struct hrtimer idle_slice_timer;
366 struct work_struct unplug_work;
368 struct cfq_queue *active_queue;
369 struct cfq_io_cq *active_cic;
371 sector_t last_position;
374 * tunables, see top of file
376 unsigned int cfq_quantum;
377 unsigned int cfq_back_penalty;
378 unsigned int cfq_back_max;
379 unsigned int cfq_slice_async_rq;
380 unsigned int cfq_latency;
381 u64 cfq_fifo_expire[2];
385 u64 cfq_target_latency;
388 * Fallback dummy cfqq for extreme OOM conditions
390 struct cfq_queue oom_cfqq;
392 u64 last_delayed_sync;
395 static struct cfq_group *cfq_get_next_cfqg(struct cfq_data *cfqd);
396 static void cfq_put_queue(struct cfq_queue *cfqq);
398 static struct cfq_rb_root *st_for(struct cfq_group *cfqg,
399 enum wl_class_t class,
405 if (class == IDLE_WORKLOAD)
406 return &cfqg->service_tree_idle;
408 return &cfqg->service_trees[class][type];
411 enum cfqq_state_flags {
412 CFQ_CFQQ_FLAG_on_rr = 0, /* on round-robin busy list */
413 CFQ_CFQQ_FLAG_wait_request, /* waiting for a request */
414 CFQ_CFQQ_FLAG_must_dispatch, /* must be allowed a dispatch */
415 CFQ_CFQQ_FLAG_must_alloc_slice, /* per-slice must_alloc flag */
416 CFQ_CFQQ_FLAG_fifo_expire, /* FIFO checked in this slice */
417 CFQ_CFQQ_FLAG_idle_window, /* slice idling enabled */
418 CFQ_CFQQ_FLAG_prio_changed, /* task priority has changed */
419 CFQ_CFQQ_FLAG_slice_new, /* no requests dispatched in slice */
420 CFQ_CFQQ_FLAG_sync, /* synchronous queue */
421 CFQ_CFQQ_FLAG_coop, /* cfqq is shared */
422 CFQ_CFQQ_FLAG_split_coop, /* shared cfqq will be splitted */
423 CFQ_CFQQ_FLAG_deep, /* sync cfqq experienced large depth */
424 CFQ_CFQQ_FLAG_wait_busy, /* Waiting for next request */
427 #define CFQ_CFQQ_FNS(name) \
428 static inline void cfq_mark_cfqq_##name(struct cfq_queue *cfqq) \
430 (cfqq)->flags |= (1 << CFQ_CFQQ_FLAG_##name); \
432 static inline void cfq_clear_cfqq_##name(struct cfq_queue *cfqq) \
434 (cfqq)->flags &= ~(1 << CFQ_CFQQ_FLAG_##name); \
436 static inline int cfq_cfqq_##name(const struct cfq_queue *cfqq) \
438 return ((cfqq)->flags & (1 << CFQ_CFQQ_FLAG_##name)) != 0; \
442 CFQ_CFQQ_FNS(wait_request);
443 CFQ_CFQQ_FNS(must_dispatch);
444 CFQ_CFQQ_FNS(must_alloc_slice);
445 CFQ_CFQQ_FNS(fifo_expire);
446 CFQ_CFQQ_FNS(idle_window);
447 CFQ_CFQQ_FNS(prio_changed);
448 CFQ_CFQQ_FNS(slice_new);
451 CFQ_CFQQ_FNS(split_coop);
453 CFQ_CFQQ_FNS(wait_busy);
456 #if defined(CONFIG_CFQ_GROUP_IOSCHED) && defined(CONFIG_DEBUG_BLK_CGROUP)
458 /* cfqg stats flags */
459 enum cfqg_stats_flags {
460 CFQG_stats_waiting = 0,
465 #define CFQG_FLAG_FNS(name) \
466 static inline void cfqg_stats_mark_##name(struct cfqg_stats *stats) \
468 stats->flags |= (1 << CFQG_stats_##name); \
470 static inline void cfqg_stats_clear_##name(struct cfqg_stats *stats) \
472 stats->flags &= ~(1 << CFQG_stats_##name); \
474 static inline int cfqg_stats_##name(struct cfqg_stats *stats) \
476 return (stats->flags & (1 << CFQG_stats_##name)) != 0; \
479 CFQG_FLAG_FNS(waiting)
480 CFQG_FLAG_FNS(idling)
484 /* This should be called with the queue_lock held. */
485 static void cfqg_stats_update_group_wait_time(struct cfqg_stats *stats)
487 unsigned long long now;
489 if (!cfqg_stats_waiting(stats))
493 if (time_after64(now, stats->start_group_wait_time))
494 blkg_stat_add(&stats->group_wait_time,
495 now - stats->start_group_wait_time);
496 cfqg_stats_clear_waiting(stats);
499 /* This should be called with the queue_lock held. */
500 static void cfqg_stats_set_start_group_wait_time(struct cfq_group *cfqg,
501 struct cfq_group *curr_cfqg)
503 struct cfqg_stats *stats = &cfqg->stats;
505 if (cfqg_stats_waiting(stats))
507 if (cfqg == curr_cfqg)
509 stats->start_group_wait_time = sched_clock();
510 cfqg_stats_mark_waiting(stats);
513 /* This should be called with the queue_lock held. */
514 static void cfqg_stats_end_empty_time(struct cfqg_stats *stats)
516 unsigned long long now;
518 if (!cfqg_stats_empty(stats))
522 if (time_after64(now, stats->start_empty_time))
523 blkg_stat_add(&stats->empty_time,
524 now - stats->start_empty_time);
525 cfqg_stats_clear_empty(stats);
528 static void cfqg_stats_update_dequeue(struct cfq_group *cfqg)
530 blkg_stat_add(&cfqg->stats.dequeue, 1);
533 static void cfqg_stats_set_start_empty_time(struct cfq_group *cfqg)
535 struct cfqg_stats *stats = &cfqg->stats;
537 if (blkg_rwstat_total(&stats->queued))
541 * group is already marked empty. This can happen if cfqq got new
542 * request in parent group and moved to this group while being added
543 * to service tree. Just ignore the event and move on.
545 if (cfqg_stats_empty(stats))
548 stats->start_empty_time = sched_clock();
549 cfqg_stats_mark_empty(stats);
552 static void cfqg_stats_update_idle_time(struct cfq_group *cfqg)
554 struct cfqg_stats *stats = &cfqg->stats;
556 if (cfqg_stats_idling(stats)) {
557 unsigned long long now = sched_clock();
559 if (time_after64(now, stats->start_idle_time))
560 blkg_stat_add(&stats->idle_time,
561 now - stats->start_idle_time);
562 cfqg_stats_clear_idling(stats);
566 static void cfqg_stats_set_start_idle_time(struct cfq_group *cfqg)
568 struct cfqg_stats *stats = &cfqg->stats;
570 BUG_ON(cfqg_stats_idling(stats));
572 stats->start_idle_time = sched_clock();
573 cfqg_stats_mark_idling(stats);
576 static void cfqg_stats_update_avg_queue_size(struct cfq_group *cfqg)
578 struct cfqg_stats *stats = &cfqg->stats;
580 blkg_stat_add(&stats->avg_queue_size_sum,
581 blkg_rwstat_total(&stats->queued));
582 blkg_stat_add(&stats->avg_queue_size_samples, 1);
583 cfqg_stats_update_group_wait_time(stats);
586 #else /* CONFIG_CFQ_GROUP_IOSCHED && CONFIG_DEBUG_BLK_CGROUP */
588 static inline void cfqg_stats_set_start_group_wait_time(struct cfq_group *cfqg, struct cfq_group *curr_cfqg) { }
589 static inline void cfqg_stats_end_empty_time(struct cfqg_stats *stats) { }
590 static inline void cfqg_stats_update_dequeue(struct cfq_group *cfqg) { }
591 static inline void cfqg_stats_set_start_empty_time(struct cfq_group *cfqg) { }
592 static inline void cfqg_stats_update_idle_time(struct cfq_group *cfqg) { }
593 static inline void cfqg_stats_set_start_idle_time(struct cfq_group *cfqg) { }
594 static inline void cfqg_stats_update_avg_queue_size(struct cfq_group *cfqg) { }
596 #endif /* CONFIG_CFQ_GROUP_IOSCHED && CONFIG_DEBUG_BLK_CGROUP */
598 #ifdef CONFIG_CFQ_GROUP_IOSCHED
600 static inline struct cfq_group *pd_to_cfqg(struct blkg_policy_data *pd)
602 return pd ? container_of(pd, struct cfq_group, pd) : NULL;
605 static struct cfq_group_data
606 *cpd_to_cfqgd(struct blkcg_policy_data *cpd)
608 return cpd ? container_of(cpd, struct cfq_group_data, cpd) : NULL;
611 static inline struct blkcg_gq *cfqg_to_blkg(struct cfq_group *cfqg)
613 return pd_to_blkg(&cfqg->pd);
616 static struct blkcg_policy blkcg_policy_cfq;
618 static inline struct cfq_group *blkg_to_cfqg(struct blkcg_gq *blkg)
620 return pd_to_cfqg(blkg_to_pd(blkg, &blkcg_policy_cfq));
623 static struct cfq_group_data *blkcg_to_cfqgd(struct blkcg *blkcg)
625 return cpd_to_cfqgd(blkcg_to_cpd(blkcg, &blkcg_policy_cfq));
628 static inline struct cfq_group *cfqg_parent(struct cfq_group *cfqg)
630 struct blkcg_gq *pblkg = cfqg_to_blkg(cfqg)->parent;
632 return pblkg ? blkg_to_cfqg(pblkg) : NULL;
635 static inline bool cfqg_is_descendant(struct cfq_group *cfqg,
636 struct cfq_group *ancestor)
638 return cgroup_is_descendant(cfqg_to_blkg(cfqg)->blkcg->css.cgroup,
639 cfqg_to_blkg(ancestor)->blkcg->css.cgroup);
642 static inline void cfqg_get(struct cfq_group *cfqg)
644 return blkg_get(cfqg_to_blkg(cfqg));
647 static inline void cfqg_put(struct cfq_group *cfqg)
649 return blkg_put(cfqg_to_blkg(cfqg));
652 #define cfq_log_cfqq(cfqd, cfqq, fmt, args...) do { \
655 blkg_path(cfqg_to_blkg((cfqq)->cfqg), __pbuf, sizeof(__pbuf)); \
656 blk_add_trace_msg((cfqd)->queue, "cfq%d%c%c %s " fmt, (cfqq)->pid, \
657 cfq_cfqq_sync((cfqq)) ? 'S' : 'A', \
658 cfqq_type((cfqq)) == SYNC_NOIDLE_WORKLOAD ? 'N' : ' ',\
662 #define cfq_log_cfqg(cfqd, cfqg, fmt, args...) do { \
665 blkg_path(cfqg_to_blkg(cfqg), __pbuf, sizeof(__pbuf)); \
666 blk_add_trace_msg((cfqd)->queue, "%s " fmt, __pbuf, ##args); \
669 static inline void cfqg_stats_update_io_add(struct cfq_group *cfqg,
670 struct cfq_group *curr_cfqg, int op,
673 blkg_rwstat_add(&cfqg->stats.queued, op, op_flags, 1);
674 cfqg_stats_end_empty_time(&cfqg->stats);
675 cfqg_stats_set_start_group_wait_time(cfqg, curr_cfqg);
678 static inline void cfqg_stats_update_timeslice_used(struct cfq_group *cfqg,
679 uint64_t time, unsigned long unaccounted_time)
681 blkg_stat_add(&cfqg->stats.time, time);
682 #ifdef CONFIG_DEBUG_BLK_CGROUP
683 blkg_stat_add(&cfqg->stats.unaccounted_time, unaccounted_time);
687 static inline void cfqg_stats_update_io_remove(struct cfq_group *cfqg, int op,
690 blkg_rwstat_add(&cfqg->stats.queued, op, op_flags, -1);
693 static inline void cfqg_stats_update_io_merged(struct cfq_group *cfqg, int op,
696 blkg_rwstat_add(&cfqg->stats.merged, op, op_flags, 1);
699 static inline void cfqg_stats_update_completion(struct cfq_group *cfqg,
700 uint64_t start_time, uint64_t io_start_time, int op,
703 struct cfqg_stats *stats = &cfqg->stats;
704 unsigned long long now = sched_clock();
706 if (time_after64(now, io_start_time))
707 blkg_rwstat_add(&stats->service_time, op, op_flags,
708 now - io_start_time);
709 if (time_after64(io_start_time, start_time))
710 blkg_rwstat_add(&stats->wait_time, op, op_flags,
711 io_start_time - start_time);
715 static void cfqg_stats_reset(struct cfqg_stats *stats)
717 /* queued stats shouldn't be cleared */
718 blkg_rwstat_reset(&stats->merged);
719 blkg_rwstat_reset(&stats->service_time);
720 blkg_rwstat_reset(&stats->wait_time);
721 blkg_stat_reset(&stats->time);
722 #ifdef CONFIG_DEBUG_BLK_CGROUP
723 blkg_stat_reset(&stats->unaccounted_time);
724 blkg_stat_reset(&stats->avg_queue_size_sum);
725 blkg_stat_reset(&stats->avg_queue_size_samples);
726 blkg_stat_reset(&stats->dequeue);
727 blkg_stat_reset(&stats->group_wait_time);
728 blkg_stat_reset(&stats->idle_time);
729 blkg_stat_reset(&stats->empty_time);
734 static void cfqg_stats_add_aux(struct cfqg_stats *to, struct cfqg_stats *from)
736 /* queued stats shouldn't be cleared */
737 blkg_rwstat_add_aux(&to->merged, &from->merged);
738 blkg_rwstat_add_aux(&to->service_time, &from->service_time);
739 blkg_rwstat_add_aux(&to->wait_time, &from->wait_time);
740 blkg_stat_add_aux(&from->time, &from->time);
741 #ifdef CONFIG_DEBUG_BLK_CGROUP
742 blkg_stat_add_aux(&to->unaccounted_time, &from->unaccounted_time);
743 blkg_stat_add_aux(&to->avg_queue_size_sum, &from->avg_queue_size_sum);
744 blkg_stat_add_aux(&to->avg_queue_size_samples, &from->avg_queue_size_samples);
745 blkg_stat_add_aux(&to->dequeue, &from->dequeue);
746 blkg_stat_add_aux(&to->group_wait_time, &from->group_wait_time);
747 blkg_stat_add_aux(&to->idle_time, &from->idle_time);
748 blkg_stat_add_aux(&to->empty_time, &from->empty_time);
753 * Transfer @cfqg's stats to its parent's aux counts so that the ancestors'
754 * recursive stats can still account for the amount used by this cfqg after
757 static void cfqg_stats_xfer_dead(struct cfq_group *cfqg)
759 struct cfq_group *parent = cfqg_parent(cfqg);
761 lockdep_assert_held(cfqg_to_blkg(cfqg)->q->queue_lock);
763 if (unlikely(!parent))
766 cfqg_stats_add_aux(&parent->stats, &cfqg->stats);
767 cfqg_stats_reset(&cfqg->stats);
770 #else /* CONFIG_CFQ_GROUP_IOSCHED */
772 static inline struct cfq_group *cfqg_parent(struct cfq_group *cfqg) { return NULL; }
773 static inline bool cfqg_is_descendant(struct cfq_group *cfqg,
774 struct cfq_group *ancestor)
778 static inline void cfqg_get(struct cfq_group *cfqg) { }
779 static inline void cfqg_put(struct cfq_group *cfqg) { }
781 #define cfq_log_cfqq(cfqd, cfqq, fmt, args...) \
782 blk_add_trace_msg((cfqd)->queue, "cfq%d%c%c " fmt, (cfqq)->pid, \
783 cfq_cfqq_sync((cfqq)) ? 'S' : 'A', \
784 cfqq_type((cfqq)) == SYNC_NOIDLE_WORKLOAD ? 'N' : ' ',\
786 #define cfq_log_cfqg(cfqd, cfqg, fmt, args...) do {} while (0)
788 static inline void cfqg_stats_update_io_add(struct cfq_group *cfqg,
789 struct cfq_group *curr_cfqg, int op, int op_flags) { }
790 static inline void cfqg_stats_update_timeslice_used(struct cfq_group *cfqg,
791 uint64_t time, unsigned long unaccounted_time) { }
792 static inline void cfqg_stats_update_io_remove(struct cfq_group *cfqg, int op,
794 static inline void cfqg_stats_update_io_merged(struct cfq_group *cfqg, int op,
796 static inline void cfqg_stats_update_completion(struct cfq_group *cfqg,
797 uint64_t start_time, uint64_t io_start_time, int op,
800 #endif /* CONFIG_CFQ_GROUP_IOSCHED */
802 #define cfq_log(cfqd, fmt, args...) \
803 blk_add_trace_msg((cfqd)->queue, "cfq " fmt, ##args)
805 /* Traverses through cfq group service trees */
806 #define for_each_cfqg_st(cfqg, i, j, st) \
807 for (i = 0; i <= IDLE_WORKLOAD; i++) \
808 for (j = 0, st = i < IDLE_WORKLOAD ? &cfqg->service_trees[i][j]\
809 : &cfqg->service_tree_idle; \
810 (i < IDLE_WORKLOAD && j <= SYNC_WORKLOAD) || \
811 (i == IDLE_WORKLOAD && j == 0); \
812 j++, st = i < IDLE_WORKLOAD ? \
813 &cfqg->service_trees[i][j]: NULL) \
815 static inline bool cfq_io_thinktime_big(struct cfq_data *cfqd,
816 struct cfq_ttime *ttime, bool group_idle)
819 if (!sample_valid(ttime->ttime_samples))
822 slice = cfqd->cfq_group_idle;
824 slice = cfqd->cfq_slice_idle;
825 return ttime->ttime_mean > slice;
828 static inline bool iops_mode(struct cfq_data *cfqd)
831 * If we are not idling on queues and it is a NCQ drive, parallel
832 * execution of requests is on and measuring time is not possible
833 * in most of the cases until and unless we drive shallower queue
834 * depths and that becomes a performance bottleneck. In such cases
835 * switch to start providing fairness in terms of number of IOs.
837 if (!cfqd->cfq_slice_idle && cfqd->hw_tag)
843 static inline enum wl_class_t cfqq_class(struct cfq_queue *cfqq)
845 if (cfq_class_idle(cfqq))
846 return IDLE_WORKLOAD;
847 if (cfq_class_rt(cfqq))
853 static enum wl_type_t cfqq_type(struct cfq_queue *cfqq)
855 if (!cfq_cfqq_sync(cfqq))
856 return ASYNC_WORKLOAD;
857 if (!cfq_cfqq_idle_window(cfqq))
858 return SYNC_NOIDLE_WORKLOAD;
859 return SYNC_WORKLOAD;
862 static inline int cfq_group_busy_queues_wl(enum wl_class_t wl_class,
863 struct cfq_data *cfqd,
864 struct cfq_group *cfqg)
866 if (wl_class == IDLE_WORKLOAD)
867 return cfqg->service_tree_idle.count;
869 return cfqg->service_trees[wl_class][ASYNC_WORKLOAD].count +
870 cfqg->service_trees[wl_class][SYNC_NOIDLE_WORKLOAD].count +
871 cfqg->service_trees[wl_class][SYNC_WORKLOAD].count;
874 static inline int cfqg_busy_async_queues(struct cfq_data *cfqd,
875 struct cfq_group *cfqg)
877 return cfqg->service_trees[RT_WORKLOAD][ASYNC_WORKLOAD].count +
878 cfqg->service_trees[BE_WORKLOAD][ASYNC_WORKLOAD].count;
881 static void cfq_dispatch_insert(struct request_queue *, struct request *);
882 static struct cfq_queue *cfq_get_queue(struct cfq_data *cfqd, bool is_sync,
883 struct cfq_io_cq *cic, struct bio *bio);
885 static inline struct cfq_io_cq *icq_to_cic(struct io_cq *icq)
887 /* cic->icq is the first member, %NULL will convert to %NULL */
888 return container_of(icq, struct cfq_io_cq, icq);
891 static inline struct cfq_io_cq *cfq_cic_lookup(struct cfq_data *cfqd,
892 struct io_context *ioc)
895 return icq_to_cic(ioc_lookup_icq(ioc, cfqd->queue));
899 static inline struct cfq_queue *cic_to_cfqq(struct cfq_io_cq *cic, bool is_sync)
901 return cic->cfqq[is_sync];
904 static inline void cic_set_cfqq(struct cfq_io_cq *cic, struct cfq_queue *cfqq,
907 cic->cfqq[is_sync] = cfqq;
910 static inline struct cfq_data *cic_to_cfqd(struct cfq_io_cq *cic)
912 return cic->icq.q->elevator->elevator_data;
916 * We regard a request as SYNC, if it's either a read or has the SYNC bit
917 * set (in which case it could also be direct WRITE).
919 static inline bool cfq_bio_sync(struct bio *bio)
921 return bio_data_dir(bio) == READ || (bio->bi_rw & REQ_SYNC);
925 * scheduler run of queue, if there are requests pending and no one in the
926 * driver that will restart queueing
928 static inline void cfq_schedule_dispatch(struct cfq_data *cfqd)
930 if (cfqd->busy_queues) {
931 cfq_log(cfqd, "schedule dispatch");
932 kblockd_schedule_work(&cfqd->unplug_work);
937 * Scale schedule slice based on io priority. Use the sync time slice only
938 * if a queue is marked sync and has sync io queued. A sync queue with async
939 * io only, should not get full sync slice length.
941 static inline u64 cfq_prio_slice(struct cfq_data *cfqd, bool sync,
944 u64 base_slice = cfqd->cfq_slice[sync];
945 u64 slice = div_u64(base_slice, CFQ_SLICE_SCALE);
947 WARN_ON(prio >= IOPRIO_BE_NR);
949 return base_slice + (slice * (4 - prio));
953 cfq_prio_to_slice(struct cfq_data *cfqd, struct cfq_queue *cfqq)
955 return cfq_prio_slice(cfqd, cfq_cfqq_sync(cfqq), cfqq->ioprio);
959 * cfqg_scale_charge - scale disk time charge according to cfqg weight
960 * @charge: disk time being charged
961 * @vfraction: vfraction of the cfqg, fixed point w/ CFQ_SERVICE_SHIFT
963 * Scale @charge according to @vfraction, which is in range (0, 1]. The
964 * scaling is inversely proportional.
966 * scaled = charge / vfraction
968 * The result is also in fixed point w/ CFQ_SERVICE_SHIFT.
970 static inline u64 cfqg_scale_charge(u64 charge,
971 unsigned int vfraction)
973 u64 c = charge << CFQ_SERVICE_SHIFT; /* make it fixed point */
975 /* charge / vfraction */
976 c <<= CFQ_SERVICE_SHIFT;
977 return div_u64(c, vfraction);
980 static inline u64 max_vdisktime(u64 min_vdisktime, u64 vdisktime)
982 s64 delta = (s64)(vdisktime - min_vdisktime);
984 min_vdisktime = vdisktime;
986 return min_vdisktime;
989 static inline u64 min_vdisktime(u64 min_vdisktime, u64 vdisktime)
991 s64 delta = (s64)(vdisktime - min_vdisktime);
993 min_vdisktime = vdisktime;
995 return min_vdisktime;
998 static void update_min_vdisktime(struct cfq_rb_root *st)
1000 struct cfq_group *cfqg;
1003 cfqg = rb_entry_cfqg(st->left);
1004 st->min_vdisktime = max_vdisktime(st->min_vdisktime,
1010 * get averaged number of queues of RT/BE priority.
1011 * average is updated, with a formula that gives more weight to higher numbers,
1012 * to quickly follows sudden increases and decrease slowly
1015 static inline unsigned cfq_group_get_avg_queues(struct cfq_data *cfqd,
1016 struct cfq_group *cfqg, bool rt)
1018 unsigned min_q, max_q;
1019 unsigned mult = cfq_hist_divisor - 1;
1020 unsigned round = cfq_hist_divisor / 2;
1021 unsigned busy = cfq_group_busy_queues_wl(rt, cfqd, cfqg);
1023 min_q = min(cfqg->busy_queues_avg[rt], busy);
1024 max_q = max(cfqg->busy_queues_avg[rt], busy);
1025 cfqg->busy_queues_avg[rt] = (mult * max_q + min_q + round) /
1027 return cfqg->busy_queues_avg[rt];
1031 cfq_group_slice(struct cfq_data *cfqd, struct cfq_group *cfqg)
1033 return cfqd->cfq_target_latency * cfqg->vfraction >> CFQ_SERVICE_SHIFT;
1037 cfq_scaled_cfqq_slice(struct cfq_data *cfqd, struct cfq_queue *cfqq)
1039 u64 slice = cfq_prio_to_slice(cfqd, cfqq);
1040 if (cfqd->cfq_latency) {
1042 * interested queues (we consider only the ones with the same
1043 * priority class in the cfq group)
1045 unsigned iq = cfq_group_get_avg_queues(cfqd, cfqq->cfqg,
1046 cfq_class_rt(cfqq));
1047 u64 sync_slice = cfqd->cfq_slice[1];
1048 u64 expect_latency = sync_slice * iq;
1049 u64 group_slice = cfq_group_slice(cfqd, cfqq->cfqg);
1051 if (expect_latency > group_slice) {
1052 u64 base_low_slice = 2 * cfqd->cfq_slice_idle;
1055 /* scale low_slice according to IO priority
1056 * and sync vs async */
1057 low_slice = div64_u64(base_low_slice*slice, sync_slice);
1058 low_slice = min(slice, low_slice);
1059 /* the adapted slice value is scaled to fit all iqs
1060 * into the target latency */
1061 slice = div64_u64(slice*group_slice, expect_latency);
1062 slice = max(slice, low_slice);
1069 cfq_set_prio_slice(struct cfq_data *cfqd, struct cfq_queue *cfqq)
1071 u64 slice = cfq_scaled_cfqq_slice(cfqd, cfqq);
1072 u64 now = ktime_get_ns();
1074 cfqq->slice_start = now;
1075 cfqq->slice_end = now + slice;
1076 cfqq->allocated_slice = slice;
1077 cfq_log_cfqq(cfqd, cfqq, "set_slice=%llu", cfqq->slice_end - now);
1081 * We need to wrap this check in cfq_cfqq_slice_new(), since ->slice_end
1082 * isn't valid until the first request from the dispatch is activated
1083 * and the slice time set.
1085 static inline bool cfq_slice_used(struct cfq_queue *cfqq)
1087 if (cfq_cfqq_slice_new(cfqq))
1089 if (ktime_get_ns() < cfqq->slice_end)
1096 * Lifted from AS - choose which of rq1 and rq2 that is best served now.
1097 * We choose the request that is closest to the head right now. Distance
1098 * behind the head is penalized and only allowed to a certain extent.
1100 static struct request *
1101 cfq_choose_req(struct cfq_data *cfqd, struct request *rq1, struct request *rq2, sector_t last)
1103 sector_t s1, s2, d1 = 0, d2 = 0;
1104 unsigned long back_max;
1105 #define CFQ_RQ1_WRAP 0x01 /* request 1 wraps */
1106 #define CFQ_RQ2_WRAP 0x02 /* request 2 wraps */
1107 unsigned wrap = 0; /* bit mask: requests behind the disk head? */
1109 if (rq1 == NULL || rq1 == rq2)
1114 if (rq_is_sync(rq1) != rq_is_sync(rq2))
1115 return rq_is_sync(rq1) ? rq1 : rq2;
1117 if ((rq1->cmd_flags ^ rq2->cmd_flags) & REQ_PRIO)
1118 return rq1->cmd_flags & REQ_PRIO ? rq1 : rq2;
1120 s1 = blk_rq_pos(rq1);
1121 s2 = blk_rq_pos(rq2);
1124 * by definition, 1KiB is 2 sectors
1126 back_max = cfqd->cfq_back_max * 2;
1129 * Strict one way elevator _except_ in the case where we allow
1130 * short backward seeks which are biased as twice the cost of a
1131 * similar forward seek.
1135 else if (s1 + back_max >= last)
1136 d1 = (last - s1) * cfqd->cfq_back_penalty;
1138 wrap |= CFQ_RQ1_WRAP;
1142 else if (s2 + back_max >= last)
1143 d2 = (last - s2) * cfqd->cfq_back_penalty;
1145 wrap |= CFQ_RQ2_WRAP;
1147 /* Found required data */
1150 * By doing switch() on the bit mask "wrap" we avoid having to
1151 * check two variables for all permutations: --> faster!
1154 case 0: /* common case for CFQ: rq1 and rq2 not wrapped */
1170 case (CFQ_RQ1_WRAP|CFQ_RQ2_WRAP): /* both rqs wrapped */
1173 * Since both rqs are wrapped,
1174 * start with the one that's further behind head
1175 * (--> only *one* back seek required),
1176 * since back seek takes more time than forward.
1186 * The below is leftmost cache rbtree addon
1188 static struct cfq_queue *cfq_rb_first(struct cfq_rb_root *root)
1190 /* Service tree is empty */
1195 root->left = rb_first(&root->rb);
1198 return rb_entry(root->left, struct cfq_queue, rb_node);
1203 static struct cfq_group *cfq_rb_first_group(struct cfq_rb_root *root)
1206 root->left = rb_first(&root->rb);
1209 return rb_entry_cfqg(root->left);
1214 static void rb_erase_init(struct rb_node *n, struct rb_root *root)
1220 static void cfq_rb_erase(struct rb_node *n, struct cfq_rb_root *root)
1222 if (root->left == n)
1224 rb_erase_init(n, &root->rb);
1229 * would be nice to take fifo expire time into account as well
1231 static struct request *
1232 cfq_find_next_rq(struct cfq_data *cfqd, struct cfq_queue *cfqq,
1233 struct request *last)
1235 struct rb_node *rbnext = rb_next(&last->rb_node);
1236 struct rb_node *rbprev = rb_prev(&last->rb_node);
1237 struct request *next = NULL, *prev = NULL;
1239 BUG_ON(RB_EMPTY_NODE(&last->rb_node));
1242 prev = rb_entry_rq(rbprev);
1245 next = rb_entry_rq(rbnext);
1247 rbnext = rb_first(&cfqq->sort_list);
1248 if (rbnext && rbnext != &last->rb_node)
1249 next = rb_entry_rq(rbnext);
1252 return cfq_choose_req(cfqd, next, prev, blk_rq_pos(last));
1255 static u64 cfq_slice_offset(struct cfq_data *cfqd,
1256 struct cfq_queue *cfqq)
1259 * just an approximation, should be ok.
1261 return (cfqq->cfqg->nr_cfqq - 1) * (cfq_prio_slice(cfqd, 1, 0) -
1262 cfq_prio_slice(cfqd, cfq_cfqq_sync(cfqq), cfqq->ioprio));
1266 cfqg_key(struct cfq_rb_root *st, struct cfq_group *cfqg)
1268 return cfqg->vdisktime - st->min_vdisktime;
1272 __cfq_group_service_tree_add(struct cfq_rb_root *st, struct cfq_group *cfqg)
1274 struct rb_node **node = &st->rb.rb_node;
1275 struct rb_node *parent = NULL;
1276 struct cfq_group *__cfqg;
1277 s64 key = cfqg_key(st, cfqg);
1280 while (*node != NULL) {
1282 __cfqg = rb_entry_cfqg(parent);
1284 if (key < cfqg_key(st, __cfqg))
1285 node = &parent->rb_left;
1287 node = &parent->rb_right;
1293 st->left = &cfqg->rb_node;
1295 rb_link_node(&cfqg->rb_node, parent, node);
1296 rb_insert_color(&cfqg->rb_node, &st->rb);
1300 * This has to be called only on activation of cfqg
1303 cfq_update_group_weight(struct cfq_group *cfqg)
1305 if (cfqg->new_weight) {
1306 cfqg->weight = cfqg->new_weight;
1307 cfqg->new_weight = 0;
1312 cfq_update_group_leaf_weight(struct cfq_group *cfqg)
1314 BUG_ON(!RB_EMPTY_NODE(&cfqg->rb_node));
1316 if (cfqg->new_leaf_weight) {
1317 cfqg->leaf_weight = cfqg->new_leaf_weight;
1318 cfqg->new_leaf_weight = 0;
1323 cfq_group_service_tree_add(struct cfq_rb_root *st, struct cfq_group *cfqg)
1325 unsigned int vfr = 1 << CFQ_SERVICE_SHIFT; /* start with 1 */
1326 struct cfq_group *pos = cfqg;
1327 struct cfq_group *parent;
1330 /* add to the service tree */
1331 BUG_ON(!RB_EMPTY_NODE(&cfqg->rb_node));
1334 * Update leaf_weight. We cannot update weight at this point
1335 * because cfqg might already have been activated and is
1336 * contributing its current weight to the parent's child_weight.
1338 cfq_update_group_leaf_weight(cfqg);
1339 __cfq_group_service_tree_add(st, cfqg);
1342 * Activate @cfqg and calculate the portion of vfraction @cfqg is
1343 * entitled to. vfraction is calculated by walking the tree
1344 * towards the root calculating the fraction it has at each level.
1345 * The compounded ratio is how much vfraction @cfqg owns.
1347 * Start with the proportion tasks in this cfqg has against active
1348 * children cfqgs - its leaf_weight against children_weight.
1350 propagate = !pos->nr_active++;
1351 pos->children_weight += pos->leaf_weight;
1352 vfr = vfr * pos->leaf_weight / pos->children_weight;
1355 * Compound ->weight walking up the tree. Both activation and
1356 * vfraction calculation are done in the same loop. Propagation
1357 * stops once an already activated node is met. vfraction
1358 * calculation should always continue to the root.
1360 while ((parent = cfqg_parent(pos))) {
1362 cfq_update_group_weight(pos);
1363 propagate = !parent->nr_active++;
1364 parent->children_weight += pos->weight;
1366 vfr = vfr * pos->weight / parent->children_weight;
1370 cfqg->vfraction = max_t(unsigned, vfr, 1);
1374 cfq_group_notify_queue_add(struct cfq_data *cfqd, struct cfq_group *cfqg)
1376 struct cfq_rb_root *st = &cfqd->grp_service_tree;
1377 struct cfq_group *__cfqg;
1381 if (!RB_EMPTY_NODE(&cfqg->rb_node))
1385 * Currently put the group at the end. Later implement something
1386 * so that groups get lesser vtime based on their weights, so that
1387 * if group does not loose all if it was not continuously backlogged.
1389 n = rb_last(&st->rb);
1391 __cfqg = rb_entry_cfqg(n);
1392 cfqg->vdisktime = __cfqg->vdisktime + CFQ_IDLE_DELAY;
1394 cfqg->vdisktime = st->min_vdisktime;
1395 cfq_group_service_tree_add(st, cfqg);
1399 cfq_group_service_tree_del(struct cfq_rb_root *st, struct cfq_group *cfqg)
1401 struct cfq_group *pos = cfqg;
1405 * Undo activation from cfq_group_service_tree_add(). Deactivate
1406 * @cfqg and propagate deactivation upwards.
1408 propagate = !--pos->nr_active;
1409 pos->children_weight -= pos->leaf_weight;
1412 struct cfq_group *parent = cfqg_parent(pos);
1414 /* @pos has 0 nr_active at this point */
1415 WARN_ON_ONCE(pos->children_weight);
1421 propagate = !--parent->nr_active;
1422 parent->children_weight -= pos->weight;
1426 /* remove from the service tree */
1427 if (!RB_EMPTY_NODE(&cfqg->rb_node))
1428 cfq_rb_erase(&cfqg->rb_node, st);
1432 cfq_group_notify_queue_del(struct cfq_data *cfqd, struct cfq_group *cfqg)
1434 struct cfq_rb_root *st = &cfqd->grp_service_tree;
1436 BUG_ON(cfqg->nr_cfqq < 1);
1439 /* If there are other cfq queues under this group, don't delete it */
1443 cfq_log_cfqg(cfqd, cfqg, "del_from_rr group");
1444 cfq_group_service_tree_del(st, cfqg);
1445 cfqg->saved_wl_slice = 0;
1446 cfqg_stats_update_dequeue(cfqg);
1449 static inline u64 cfq_cfqq_slice_usage(struct cfq_queue *cfqq,
1450 u64 *unaccounted_time)
1453 u64 now = ktime_get_ns();
1456 * Queue got expired before even a single request completed or
1457 * got expired immediately after first request completion.
1459 if (!cfqq->slice_start || cfqq->slice_start == now) {
1461 * Also charge the seek time incurred to the group, otherwise
1462 * if there are mutiple queues in the group, each can dispatch
1463 * a single request on seeky media and cause lots of seek time
1464 * and group will never know it.
1466 slice_used = max_t(u64, (now - cfqq->dispatch_start),
1467 jiffies_to_nsecs(1));
1469 slice_used = now - cfqq->slice_start;
1470 if (slice_used > cfqq->allocated_slice) {
1471 *unaccounted_time = slice_used - cfqq->allocated_slice;
1472 slice_used = cfqq->allocated_slice;
1474 if (cfqq->slice_start > cfqq->dispatch_start)
1475 *unaccounted_time += cfqq->slice_start -
1476 cfqq->dispatch_start;
1482 static void cfq_group_served(struct cfq_data *cfqd, struct cfq_group *cfqg,
1483 struct cfq_queue *cfqq)
1485 struct cfq_rb_root *st = &cfqd->grp_service_tree;
1486 u64 used_sl, charge, unaccounted_sl = 0;
1487 int nr_sync = cfqg->nr_cfqq - cfqg_busy_async_queues(cfqd, cfqg)
1488 - cfqg->service_tree_idle.count;
1490 u64 now = ktime_get_ns();
1492 BUG_ON(nr_sync < 0);
1493 used_sl = charge = cfq_cfqq_slice_usage(cfqq, &unaccounted_sl);
1495 if (iops_mode(cfqd))
1496 charge = cfqq->slice_dispatch;
1497 else if (!cfq_cfqq_sync(cfqq) && !nr_sync)
1498 charge = cfqq->allocated_slice;
1501 * Can't update vdisktime while on service tree and cfqg->vfraction
1502 * is valid only while on it. Cache vfr, leave the service tree,
1503 * update vdisktime and go back on. The re-addition to the tree
1504 * will also update the weights as necessary.
1506 vfr = cfqg->vfraction;
1507 cfq_group_service_tree_del(st, cfqg);
1508 cfqg->vdisktime += cfqg_scale_charge(charge, vfr);
1509 cfq_group_service_tree_add(st, cfqg);
1511 /* This group is being expired. Save the context */
1512 if (cfqd->workload_expires > now) {
1513 cfqg->saved_wl_slice = cfqd->workload_expires - now;
1514 cfqg->saved_wl_type = cfqd->serving_wl_type;
1515 cfqg->saved_wl_class = cfqd->serving_wl_class;
1517 cfqg->saved_wl_slice = 0;
1519 cfq_log_cfqg(cfqd, cfqg, "served: vt=%llu min_vt=%llu", cfqg->vdisktime,
1521 cfq_log_cfqq(cfqq->cfqd, cfqq,
1522 "sl_used=%llu disp=%llu charge=%llu iops=%u sect=%lu",
1523 used_sl, cfqq->slice_dispatch, charge,
1524 iops_mode(cfqd), cfqq->nr_sectors);
1525 cfqg_stats_update_timeslice_used(cfqg, used_sl, unaccounted_sl);
1526 cfqg_stats_set_start_empty_time(cfqg);
1530 * cfq_init_cfqg_base - initialize base part of a cfq_group
1531 * @cfqg: cfq_group to initialize
1533 * Initialize the base part which is used whether %CONFIG_CFQ_GROUP_IOSCHED
1534 * is enabled or not.
1536 static void cfq_init_cfqg_base(struct cfq_group *cfqg)
1538 struct cfq_rb_root *st;
1541 for_each_cfqg_st(cfqg, i, j, st)
1543 RB_CLEAR_NODE(&cfqg->rb_node);
1545 cfqg->ttime.last_end_request = ktime_get_ns();
1548 #ifdef CONFIG_CFQ_GROUP_IOSCHED
1549 static int __cfq_set_weight(struct cgroup_subsys_state *css, u64 val,
1550 bool on_dfl, bool reset_dev, bool is_leaf_weight);
1552 static void cfqg_stats_exit(struct cfqg_stats *stats)
1554 blkg_rwstat_exit(&stats->merged);
1555 blkg_rwstat_exit(&stats->service_time);
1556 blkg_rwstat_exit(&stats->wait_time);
1557 blkg_rwstat_exit(&stats->queued);
1558 blkg_stat_exit(&stats->time);
1559 #ifdef CONFIG_DEBUG_BLK_CGROUP
1560 blkg_stat_exit(&stats->unaccounted_time);
1561 blkg_stat_exit(&stats->avg_queue_size_sum);
1562 blkg_stat_exit(&stats->avg_queue_size_samples);
1563 blkg_stat_exit(&stats->dequeue);
1564 blkg_stat_exit(&stats->group_wait_time);
1565 blkg_stat_exit(&stats->idle_time);
1566 blkg_stat_exit(&stats->empty_time);
1570 static int cfqg_stats_init(struct cfqg_stats *stats, gfp_t gfp)
1572 if (blkg_rwstat_init(&stats->merged, gfp) ||
1573 blkg_rwstat_init(&stats->service_time, gfp) ||
1574 blkg_rwstat_init(&stats->wait_time, gfp) ||
1575 blkg_rwstat_init(&stats->queued, gfp) ||
1576 blkg_stat_init(&stats->time, gfp))
1579 #ifdef CONFIG_DEBUG_BLK_CGROUP
1580 if (blkg_stat_init(&stats->unaccounted_time, gfp) ||
1581 blkg_stat_init(&stats->avg_queue_size_sum, gfp) ||
1582 blkg_stat_init(&stats->avg_queue_size_samples, gfp) ||
1583 blkg_stat_init(&stats->dequeue, gfp) ||
1584 blkg_stat_init(&stats->group_wait_time, gfp) ||
1585 blkg_stat_init(&stats->idle_time, gfp) ||
1586 blkg_stat_init(&stats->empty_time, gfp))
1591 cfqg_stats_exit(stats);
1595 static struct blkcg_policy_data *cfq_cpd_alloc(gfp_t gfp)
1597 struct cfq_group_data *cgd;
1599 cgd = kzalloc(sizeof(*cgd), GFP_KERNEL);
1605 static void cfq_cpd_init(struct blkcg_policy_data *cpd)
1607 struct cfq_group_data *cgd = cpd_to_cfqgd(cpd);
1608 unsigned int weight = cgroup_subsys_on_dfl(io_cgrp_subsys) ?
1609 CGROUP_WEIGHT_DFL : CFQ_WEIGHT_LEGACY_DFL;
1611 if (cpd_to_blkcg(cpd) == &blkcg_root)
1614 cgd->weight = weight;
1615 cgd->leaf_weight = weight;
1618 static void cfq_cpd_free(struct blkcg_policy_data *cpd)
1620 kfree(cpd_to_cfqgd(cpd));
1623 static void cfq_cpd_bind(struct blkcg_policy_data *cpd)
1625 struct blkcg *blkcg = cpd_to_blkcg(cpd);
1626 bool on_dfl = cgroup_subsys_on_dfl(io_cgrp_subsys);
1627 unsigned int weight = on_dfl ? CGROUP_WEIGHT_DFL : CFQ_WEIGHT_LEGACY_DFL;
1629 if (blkcg == &blkcg_root)
1632 WARN_ON_ONCE(__cfq_set_weight(&blkcg->css, weight, on_dfl, true, false));
1633 WARN_ON_ONCE(__cfq_set_weight(&blkcg->css, weight, on_dfl, true, true));
1636 static struct blkg_policy_data *cfq_pd_alloc(gfp_t gfp, int node)
1638 struct cfq_group *cfqg;
1640 cfqg = kzalloc_node(sizeof(*cfqg), gfp, node);
1644 cfq_init_cfqg_base(cfqg);
1645 if (cfqg_stats_init(&cfqg->stats, gfp)) {
1653 static void cfq_pd_init(struct blkg_policy_data *pd)
1655 struct cfq_group *cfqg = pd_to_cfqg(pd);
1656 struct cfq_group_data *cgd = blkcg_to_cfqgd(pd->blkg->blkcg);
1658 cfqg->weight = cgd->weight;
1659 cfqg->leaf_weight = cgd->leaf_weight;
1662 static void cfq_pd_offline(struct blkg_policy_data *pd)
1664 struct cfq_group *cfqg = pd_to_cfqg(pd);
1667 for (i = 0; i < IOPRIO_BE_NR; i++) {
1668 if (cfqg->async_cfqq[0][i])
1669 cfq_put_queue(cfqg->async_cfqq[0][i]);
1670 if (cfqg->async_cfqq[1][i])
1671 cfq_put_queue(cfqg->async_cfqq[1][i]);
1674 if (cfqg->async_idle_cfqq)
1675 cfq_put_queue(cfqg->async_idle_cfqq);
1678 * @blkg is going offline and will be ignored by
1679 * blkg_[rw]stat_recursive_sum(). Transfer stats to the parent so
1680 * that they don't get lost. If IOs complete after this point, the
1681 * stats for them will be lost. Oh well...
1683 cfqg_stats_xfer_dead(cfqg);
1686 static void cfq_pd_free(struct blkg_policy_data *pd)
1688 struct cfq_group *cfqg = pd_to_cfqg(pd);
1690 cfqg_stats_exit(&cfqg->stats);
1694 static void cfq_pd_reset_stats(struct blkg_policy_data *pd)
1696 struct cfq_group *cfqg = pd_to_cfqg(pd);
1698 cfqg_stats_reset(&cfqg->stats);
1701 static struct cfq_group *cfq_lookup_cfqg(struct cfq_data *cfqd,
1702 struct blkcg *blkcg)
1704 struct blkcg_gq *blkg;
1706 blkg = blkg_lookup(blkcg, cfqd->queue);
1708 return blkg_to_cfqg(blkg);
1712 static void cfq_link_cfqq_cfqg(struct cfq_queue *cfqq, struct cfq_group *cfqg)
1715 /* cfqq reference on cfqg */
1719 static u64 cfqg_prfill_weight_device(struct seq_file *sf,
1720 struct blkg_policy_data *pd, int off)
1722 struct cfq_group *cfqg = pd_to_cfqg(pd);
1724 if (!cfqg->dev_weight)
1726 return __blkg_prfill_u64(sf, pd, cfqg->dev_weight);
1729 static int cfqg_print_weight_device(struct seq_file *sf, void *v)
1731 blkcg_print_blkgs(sf, css_to_blkcg(seq_css(sf)),
1732 cfqg_prfill_weight_device, &blkcg_policy_cfq,
1737 static u64 cfqg_prfill_leaf_weight_device(struct seq_file *sf,
1738 struct blkg_policy_data *pd, int off)
1740 struct cfq_group *cfqg = pd_to_cfqg(pd);
1742 if (!cfqg->dev_leaf_weight)
1744 return __blkg_prfill_u64(sf, pd, cfqg->dev_leaf_weight);
1747 static int cfqg_print_leaf_weight_device(struct seq_file *sf, void *v)
1749 blkcg_print_blkgs(sf, css_to_blkcg(seq_css(sf)),
1750 cfqg_prfill_leaf_weight_device, &blkcg_policy_cfq,
1755 static int cfq_print_weight(struct seq_file *sf, void *v)
1757 struct blkcg *blkcg = css_to_blkcg(seq_css(sf));
1758 struct cfq_group_data *cgd = blkcg_to_cfqgd(blkcg);
1759 unsigned int val = 0;
1764 seq_printf(sf, "%u\n", val);
1768 static int cfq_print_leaf_weight(struct seq_file *sf, void *v)
1770 struct blkcg *blkcg = css_to_blkcg(seq_css(sf));
1771 struct cfq_group_data *cgd = blkcg_to_cfqgd(blkcg);
1772 unsigned int val = 0;
1775 val = cgd->leaf_weight;
1777 seq_printf(sf, "%u\n", val);
1781 static ssize_t __cfqg_set_weight_device(struct kernfs_open_file *of,
1782 char *buf, size_t nbytes, loff_t off,
1783 bool on_dfl, bool is_leaf_weight)
1785 unsigned int min = on_dfl ? CGROUP_WEIGHT_MIN : CFQ_WEIGHT_LEGACY_MIN;
1786 unsigned int max = on_dfl ? CGROUP_WEIGHT_MAX : CFQ_WEIGHT_LEGACY_MAX;
1787 struct blkcg *blkcg = css_to_blkcg(of_css(of));
1788 struct blkg_conf_ctx ctx;
1789 struct cfq_group *cfqg;
1790 struct cfq_group_data *cfqgd;
1794 ret = blkg_conf_prep(blkcg, &blkcg_policy_cfq, buf, &ctx);
1798 if (sscanf(ctx.body, "%llu", &v) == 1) {
1799 /* require "default" on dfl */
1803 } else if (!strcmp(strim(ctx.body), "default")) {
1810 cfqg = blkg_to_cfqg(ctx.blkg);
1811 cfqgd = blkcg_to_cfqgd(blkcg);
1814 if (!v || (v >= min && v <= max)) {
1815 if (!is_leaf_weight) {
1816 cfqg->dev_weight = v;
1817 cfqg->new_weight = v ?: cfqgd->weight;
1819 cfqg->dev_leaf_weight = v;
1820 cfqg->new_leaf_weight = v ?: cfqgd->leaf_weight;
1825 blkg_conf_finish(&ctx);
1826 return ret ?: nbytes;
1829 static ssize_t cfqg_set_weight_device(struct kernfs_open_file *of,
1830 char *buf, size_t nbytes, loff_t off)
1832 return __cfqg_set_weight_device(of, buf, nbytes, off, false, false);
1835 static ssize_t cfqg_set_leaf_weight_device(struct kernfs_open_file *of,
1836 char *buf, size_t nbytes, loff_t off)
1838 return __cfqg_set_weight_device(of, buf, nbytes, off, false, true);
1841 static int __cfq_set_weight(struct cgroup_subsys_state *css, u64 val,
1842 bool on_dfl, bool reset_dev, bool is_leaf_weight)
1844 unsigned int min = on_dfl ? CGROUP_WEIGHT_MIN : CFQ_WEIGHT_LEGACY_MIN;
1845 unsigned int max = on_dfl ? CGROUP_WEIGHT_MAX : CFQ_WEIGHT_LEGACY_MAX;
1846 struct blkcg *blkcg = css_to_blkcg(css);
1847 struct blkcg_gq *blkg;
1848 struct cfq_group_data *cfqgd;
1851 if (val < min || val > max)
1854 spin_lock_irq(&blkcg->lock);
1855 cfqgd = blkcg_to_cfqgd(blkcg);
1861 if (!is_leaf_weight)
1862 cfqgd->weight = val;
1864 cfqgd->leaf_weight = val;
1866 hlist_for_each_entry(blkg, &blkcg->blkg_list, blkcg_node) {
1867 struct cfq_group *cfqg = blkg_to_cfqg(blkg);
1872 if (!is_leaf_weight) {
1874 cfqg->dev_weight = 0;
1875 if (!cfqg->dev_weight)
1876 cfqg->new_weight = cfqgd->weight;
1879 cfqg->dev_leaf_weight = 0;
1880 if (!cfqg->dev_leaf_weight)
1881 cfqg->new_leaf_weight = cfqgd->leaf_weight;
1886 spin_unlock_irq(&blkcg->lock);
1890 static int cfq_set_weight(struct cgroup_subsys_state *css, struct cftype *cft,
1893 return __cfq_set_weight(css, val, false, false, false);
1896 static int cfq_set_leaf_weight(struct cgroup_subsys_state *css,
1897 struct cftype *cft, u64 val)
1899 return __cfq_set_weight(css, val, false, false, true);
1902 static int cfqg_print_stat(struct seq_file *sf, void *v)
1904 blkcg_print_blkgs(sf, css_to_blkcg(seq_css(sf)), blkg_prfill_stat,
1905 &blkcg_policy_cfq, seq_cft(sf)->private, false);
1909 static int cfqg_print_rwstat(struct seq_file *sf, void *v)
1911 blkcg_print_blkgs(sf, css_to_blkcg(seq_css(sf)), blkg_prfill_rwstat,
1912 &blkcg_policy_cfq, seq_cft(sf)->private, true);
1916 static u64 cfqg_prfill_stat_recursive(struct seq_file *sf,
1917 struct blkg_policy_data *pd, int off)
1919 u64 sum = blkg_stat_recursive_sum(pd_to_blkg(pd),
1920 &blkcg_policy_cfq, off);
1921 return __blkg_prfill_u64(sf, pd, sum);
1924 static u64 cfqg_prfill_rwstat_recursive(struct seq_file *sf,
1925 struct blkg_policy_data *pd, int off)
1927 struct blkg_rwstat sum = blkg_rwstat_recursive_sum(pd_to_blkg(pd),
1928 &blkcg_policy_cfq, off);
1929 return __blkg_prfill_rwstat(sf, pd, &sum);
1932 static int cfqg_print_stat_recursive(struct seq_file *sf, void *v)
1934 blkcg_print_blkgs(sf, css_to_blkcg(seq_css(sf)),
1935 cfqg_prfill_stat_recursive, &blkcg_policy_cfq,
1936 seq_cft(sf)->private, false);
1940 static int cfqg_print_rwstat_recursive(struct seq_file *sf, void *v)
1942 blkcg_print_blkgs(sf, css_to_blkcg(seq_css(sf)),
1943 cfqg_prfill_rwstat_recursive, &blkcg_policy_cfq,
1944 seq_cft(sf)->private, true);
1948 static u64 cfqg_prfill_sectors(struct seq_file *sf, struct blkg_policy_data *pd,
1951 u64 sum = blkg_rwstat_total(&pd->blkg->stat_bytes);
1953 return __blkg_prfill_u64(sf, pd, sum >> 9);
1956 static int cfqg_print_stat_sectors(struct seq_file *sf, void *v)
1958 blkcg_print_blkgs(sf, css_to_blkcg(seq_css(sf)),
1959 cfqg_prfill_sectors, &blkcg_policy_cfq, 0, false);
1963 static u64 cfqg_prfill_sectors_recursive(struct seq_file *sf,
1964 struct blkg_policy_data *pd, int off)
1966 struct blkg_rwstat tmp = blkg_rwstat_recursive_sum(pd->blkg, NULL,
1967 offsetof(struct blkcg_gq, stat_bytes));
1968 u64 sum = atomic64_read(&tmp.aux_cnt[BLKG_RWSTAT_READ]) +
1969 atomic64_read(&tmp.aux_cnt[BLKG_RWSTAT_WRITE]);
1971 return __blkg_prfill_u64(sf, pd, sum >> 9);
1974 static int cfqg_print_stat_sectors_recursive(struct seq_file *sf, void *v)
1976 blkcg_print_blkgs(sf, css_to_blkcg(seq_css(sf)),
1977 cfqg_prfill_sectors_recursive, &blkcg_policy_cfq, 0,
1982 #ifdef CONFIG_DEBUG_BLK_CGROUP
1983 static u64 cfqg_prfill_avg_queue_size(struct seq_file *sf,
1984 struct blkg_policy_data *pd, int off)
1986 struct cfq_group *cfqg = pd_to_cfqg(pd);
1987 u64 samples = blkg_stat_read(&cfqg->stats.avg_queue_size_samples);
1991 v = blkg_stat_read(&cfqg->stats.avg_queue_size_sum);
1992 v = div64_u64(v, samples);
1994 __blkg_prfill_u64(sf, pd, v);
1998 /* print avg_queue_size */
1999 static int cfqg_print_avg_queue_size(struct seq_file *sf, void *v)
2001 blkcg_print_blkgs(sf, css_to_blkcg(seq_css(sf)),
2002 cfqg_prfill_avg_queue_size, &blkcg_policy_cfq,
2006 #endif /* CONFIG_DEBUG_BLK_CGROUP */
2008 static struct cftype cfq_blkcg_legacy_files[] = {
2009 /* on root, weight is mapped to leaf_weight */
2011 .name = "weight_device",
2012 .flags = CFTYPE_ONLY_ON_ROOT,
2013 .seq_show = cfqg_print_leaf_weight_device,
2014 .write = cfqg_set_leaf_weight_device,
2018 .flags = CFTYPE_ONLY_ON_ROOT,
2019 .seq_show = cfq_print_leaf_weight,
2020 .write_u64 = cfq_set_leaf_weight,
2023 /* no such mapping necessary for !roots */
2025 .name = "weight_device",
2026 .flags = CFTYPE_NOT_ON_ROOT,
2027 .seq_show = cfqg_print_weight_device,
2028 .write = cfqg_set_weight_device,
2032 .flags = CFTYPE_NOT_ON_ROOT,
2033 .seq_show = cfq_print_weight,
2034 .write_u64 = cfq_set_weight,
2038 .name = "leaf_weight_device",
2039 .seq_show = cfqg_print_leaf_weight_device,
2040 .write = cfqg_set_leaf_weight_device,
2043 .name = "leaf_weight",
2044 .seq_show = cfq_print_leaf_weight,
2045 .write_u64 = cfq_set_leaf_weight,
2048 /* statistics, covers only the tasks in the cfqg */
2051 .private = offsetof(struct cfq_group, stats.time),
2052 .seq_show = cfqg_print_stat,
2056 .seq_show = cfqg_print_stat_sectors,
2059 .name = "io_service_bytes",
2060 .private = (unsigned long)&blkcg_policy_cfq,
2061 .seq_show = blkg_print_stat_bytes,
2064 .name = "io_serviced",
2065 .private = (unsigned long)&blkcg_policy_cfq,
2066 .seq_show = blkg_print_stat_ios,
2069 .name = "io_service_time",
2070 .private = offsetof(struct cfq_group, stats.service_time),
2071 .seq_show = cfqg_print_rwstat,
2074 .name = "io_wait_time",
2075 .private = offsetof(struct cfq_group, stats.wait_time),
2076 .seq_show = cfqg_print_rwstat,
2079 .name = "io_merged",
2080 .private = offsetof(struct cfq_group, stats.merged),
2081 .seq_show = cfqg_print_rwstat,
2084 .name = "io_queued",
2085 .private = offsetof(struct cfq_group, stats.queued),
2086 .seq_show = cfqg_print_rwstat,
2089 /* the same statictics which cover the cfqg and its descendants */
2091 .name = "time_recursive",
2092 .private = offsetof(struct cfq_group, stats.time),
2093 .seq_show = cfqg_print_stat_recursive,
2096 .name = "sectors_recursive",
2097 .seq_show = cfqg_print_stat_sectors_recursive,
2100 .name = "io_service_bytes_recursive",
2101 .private = (unsigned long)&blkcg_policy_cfq,
2102 .seq_show = blkg_print_stat_bytes_recursive,
2105 .name = "io_serviced_recursive",
2106 .private = (unsigned long)&blkcg_policy_cfq,
2107 .seq_show = blkg_print_stat_ios_recursive,
2110 .name = "io_service_time_recursive",
2111 .private = offsetof(struct cfq_group, stats.service_time),
2112 .seq_show = cfqg_print_rwstat_recursive,
2115 .name = "io_wait_time_recursive",
2116 .private = offsetof(struct cfq_group, stats.wait_time),
2117 .seq_show = cfqg_print_rwstat_recursive,
2120 .name = "io_merged_recursive",
2121 .private = offsetof(struct cfq_group, stats.merged),
2122 .seq_show = cfqg_print_rwstat_recursive,
2125 .name = "io_queued_recursive",
2126 .private = offsetof(struct cfq_group, stats.queued),
2127 .seq_show = cfqg_print_rwstat_recursive,
2129 #ifdef CONFIG_DEBUG_BLK_CGROUP
2131 .name = "avg_queue_size",
2132 .seq_show = cfqg_print_avg_queue_size,
2135 .name = "group_wait_time",
2136 .private = offsetof(struct cfq_group, stats.group_wait_time),
2137 .seq_show = cfqg_print_stat,
2140 .name = "idle_time",
2141 .private = offsetof(struct cfq_group, stats.idle_time),
2142 .seq_show = cfqg_print_stat,
2145 .name = "empty_time",
2146 .private = offsetof(struct cfq_group, stats.empty_time),
2147 .seq_show = cfqg_print_stat,
2151 .private = offsetof(struct cfq_group, stats.dequeue),
2152 .seq_show = cfqg_print_stat,
2155 .name = "unaccounted_time",
2156 .private = offsetof(struct cfq_group, stats.unaccounted_time),
2157 .seq_show = cfqg_print_stat,
2159 #endif /* CONFIG_DEBUG_BLK_CGROUP */
2163 static int cfq_print_weight_on_dfl(struct seq_file *sf, void *v)
2165 struct blkcg *blkcg = css_to_blkcg(seq_css(sf));
2166 struct cfq_group_data *cgd = blkcg_to_cfqgd(blkcg);
2168 seq_printf(sf, "default %u\n", cgd->weight);
2169 blkcg_print_blkgs(sf, blkcg, cfqg_prfill_weight_device,
2170 &blkcg_policy_cfq, 0, false);
2174 static ssize_t cfq_set_weight_on_dfl(struct kernfs_open_file *of,
2175 char *buf, size_t nbytes, loff_t off)
2183 /* "WEIGHT" or "default WEIGHT" sets the default weight */
2184 v = simple_strtoull(buf, &endp, 0);
2185 if (*endp == '\0' || sscanf(buf, "default %llu", &v) == 1) {
2186 ret = __cfq_set_weight(of_css(of), v, true, false, false);
2187 return ret ?: nbytes;
2190 /* "MAJ:MIN WEIGHT" */
2191 return __cfqg_set_weight_device(of, buf, nbytes, off, true, false);
2194 static struct cftype cfq_blkcg_files[] = {
2197 .flags = CFTYPE_NOT_ON_ROOT,
2198 .seq_show = cfq_print_weight_on_dfl,
2199 .write = cfq_set_weight_on_dfl,
2204 #else /* GROUP_IOSCHED */
2205 static struct cfq_group *cfq_lookup_cfqg(struct cfq_data *cfqd,
2206 struct blkcg *blkcg)
2208 return cfqd->root_group;
2212 cfq_link_cfqq_cfqg(struct cfq_queue *cfqq, struct cfq_group *cfqg) {
2216 #endif /* GROUP_IOSCHED */
2219 * The cfqd->service_trees holds all pending cfq_queue's that have
2220 * requests waiting to be processed. It is sorted in the order that
2221 * we will service the queues.
2223 static void cfq_service_tree_add(struct cfq_data *cfqd, struct cfq_queue *cfqq,
2226 struct rb_node **p, *parent;
2227 struct cfq_queue *__cfqq;
2229 struct cfq_rb_root *st;
2232 u64 now = ktime_get_ns();
2234 st = st_for(cfqq->cfqg, cfqq_class(cfqq), cfqq_type(cfqq));
2235 if (cfq_class_idle(cfqq)) {
2236 rb_key = CFQ_IDLE_DELAY;
2237 parent = rb_last(&st->rb);
2238 if (parent && parent != &cfqq->rb_node) {
2239 __cfqq = rb_entry(parent, struct cfq_queue, rb_node);
2240 rb_key += __cfqq->rb_key;
2243 } else if (!add_front) {
2245 * Get our rb key offset. Subtract any residual slice
2246 * value carried from last service. A negative resid
2247 * count indicates slice overrun, and this should position
2248 * the next service time further away in the tree.
2250 rb_key = cfq_slice_offset(cfqd, cfqq) + now;
2251 rb_key -= cfqq->slice_resid;
2252 cfqq->slice_resid = 0;
2254 rb_key = -NSEC_PER_SEC;
2255 __cfqq = cfq_rb_first(st);
2256 rb_key += __cfqq ? __cfqq->rb_key : now;
2259 if (!RB_EMPTY_NODE(&cfqq->rb_node)) {
2262 * same position, nothing more to do
2264 if (rb_key == cfqq->rb_key && cfqq->service_tree == st)
2267 cfq_rb_erase(&cfqq->rb_node, cfqq->service_tree);
2268 cfqq->service_tree = NULL;
2273 cfqq->service_tree = st;
2274 p = &st->rb.rb_node;
2277 __cfqq = rb_entry(parent, struct cfq_queue, rb_node);
2280 * sort by key, that represents service time.
2282 if (rb_key < __cfqq->rb_key)
2283 p = &parent->rb_left;
2285 p = &parent->rb_right;
2291 st->left = &cfqq->rb_node;
2293 cfqq->rb_key = rb_key;
2294 rb_link_node(&cfqq->rb_node, parent, p);
2295 rb_insert_color(&cfqq->rb_node, &st->rb);
2297 if (add_front || !new_cfqq)
2299 cfq_group_notify_queue_add(cfqd, cfqq->cfqg);
2302 static struct cfq_queue *
2303 cfq_prio_tree_lookup(struct cfq_data *cfqd, struct rb_root *root,
2304 sector_t sector, struct rb_node **ret_parent,
2305 struct rb_node ***rb_link)
2307 struct rb_node **p, *parent;
2308 struct cfq_queue *cfqq = NULL;
2316 cfqq = rb_entry(parent, struct cfq_queue, p_node);
2319 * Sort strictly based on sector. Smallest to the left,
2320 * largest to the right.
2322 if (sector > blk_rq_pos(cfqq->next_rq))
2323 n = &(*p)->rb_right;
2324 else if (sector < blk_rq_pos(cfqq->next_rq))
2332 *ret_parent = parent;
2338 static void cfq_prio_tree_add(struct cfq_data *cfqd, struct cfq_queue *cfqq)
2340 struct rb_node **p, *parent;
2341 struct cfq_queue *__cfqq;
2344 rb_erase(&cfqq->p_node, cfqq->p_root);
2345 cfqq->p_root = NULL;
2348 if (cfq_class_idle(cfqq))
2353 cfqq->p_root = &cfqd->prio_trees[cfqq->org_ioprio];
2354 __cfqq = cfq_prio_tree_lookup(cfqd, cfqq->p_root,
2355 blk_rq_pos(cfqq->next_rq), &parent, &p);
2357 rb_link_node(&cfqq->p_node, parent, p);
2358 rb_insert_color(&cfqq->p_node, cfqq->p_root);
2360 cfqq->p_root = NULL;
2364 * Update cfqq's position in the service tree.
2366 static void cfq_resort_rr_list(struct cfq_data *cfqd, struct cfq_queue *cfqq)
2369 * Resorting requires the cfqq to be on the RR list already.
2371 if (cfq_cfqq_on_rr(cfqq)) {
2372 cfq_service_tree_add(cfqd, cfqq, 0);
2373 cfq_prio_tree_add(cfqd, cfqq);
2378 * add to busy list of queues for service, trying to be fair in ordering
2379 * the pending list according to last request service
2381 static void cfq_add_cfqq_rr(struct cfq_data *cfqd, struct cfq_queue *cfqq)
2383 cfq_log_cfqq(cfqd, cfqq, "add_to_rr");
2384 BUG_ON(cfq_cfqq_on_rr(cfqq));
2385 cfq_mark_cfqq_on_rr(cfqq);
2386 cfqd->busy_queues++;
2387 if (cfq_cfqq_sync(cfqq))
2388 cfqd->busy_sync_queues++;
2390 cfq_resort_rr_list(cfqd, cfqq);
2394 * Called when the cfqq no longer has requests pending, remove it from
2397 static void cfq_del_cfqq_rr(struct cfq_data *cfqd, struct cfq_queue *cfqq)
2399 cfq_log_cfqq(cfqd, cfqq, "del_from_rr");
2400 BUG_ON(!cfq_cfqq_on_rr(cfqq));
2401 cfq_clear_cfqq_on_rr(cfqq);
2403 if (!RB_EMPTY_NODE(&cfqq->rb_node)) {
2404 cfq_rb_erase(&cfqq->rb_node, cfqq->service_tree);
2405 cfqq->service_tree = NULL;
2408 rb_erase(&cfqq->p_node, cfqq->p_root);
2409 cfqq->p_root = NULL;
2412 cfq_group_notify_queue_del(cfqd, cfqq->cfqg);
2413 BUG_ON(!cfqd->busy_queues);
2414 cfqd->busy_queues--;
2415 if (cfq_cfqq_sync(cfqq))
2416 cfqd->busy_sync_queues--;
2420 * rb tree support functions
2422 static void cfq_del_rq_rb(struct request *rq)
2424 struct cfq_queue *cfqq = RQ_CFQQ(rq);
2425 const int sync = rq_is_sync(rq);
2427 BUG_ON(!cfqq->queued[sync]);
2428 cfqq->queued[sync]--;
2430 elv_rb_del(&cfqq->sort_list, rq);
2432 if (cfq_cfqq_on_rr(cfqq) && RB_EMPTY_ROOT(&cfqq->sort_list)) {
2434 * Queue will be deleted from service tree when we actually
2435 * expire it later. Right now just remove it from prio tree
2439 rb_erase(&cfqq->p_node, cfqq->p_root);
2440 cfqq->p_root = NULL;
2445 static void cfq_add_rq_rb(struct request *rq)
2447 struct cfq_queue *cfqq = RQ_CFQQ(rq);
2448 struct cfq_data *cfqd = cfqq->cfqd;
2449 struct request *prev;
2451 cfqq->queued[rq_is_sync(rq)]++;
2453 elv_rb_add(&cfqq->sort_list, rq);
2455 if (!cfq_cfqq_on_rr(cfqq))
2456 cfq_add_cfqq_rr(cfqd, cfqq);
2459 * check if this request is a better next-serve candidate
2461 prev = cfqq->next_rq;
2462 cfqq->next_rq = cfq_choose_req(cfqd, cfqq->next_rq, rq, cfqd->last_position);
2465 * adjust priority tree position, if ->next_rq changes
2467 if (prev != cfqq->next_rq)
2468 cfq_prio_tree_add(cfqd, cfqq);
2470 BUG_ON(!cfqq->next_rq);
2473 static void cfq_reposition_rq_rb(struct cfq_queue *cfqq, struct request *rq)
2475 elv_rb_del(&cfqq->sort_list, rq);
2476 cfqq->queued[rq_is_sync(rq)]--;
2477 cfqg_stats_update_io_remove(RQ_CFQG(rq), req_op(rq), rq->cmd_flags);
2479 cfqg_stats_update_io_add(RQ_CFQG(rq), cfqq->cfqd->serving_group,
2480 req_op(rq), rq->cmd_flags);
2483 static struct request *
2484 cfq_find_rq_fmerge(struct cfq_data *cfqd, struct bio *bio)
2486 struct task_struct *tsk = current;
2487 struct cfq_io_cq *cic;
2488 struct cfq_queue *cfqq;
2490 cic = cfq_cic_lookup(cfqd, tsk->io_context);
2494 cfqq = cic_to_cfqq(cic, cfq_bio_sync(bio));
2496 return elv_rb_find(&cfqq->sort_list, bio_end_sector(bio));
2501 static void cfq_activate_request(struct request_queue *q, struct request *rq)
2503 struct cfq_data *cfqd = q->elevator->elevator_data;
2505 cfqd->rq_in_driver++;
2506 cfq_log_cfqq(cfqd, RQ_CFQQ(rq), "activate rq, drv=%d",
2507 cfqd->rq_in_driver);
2509 cfqd->last_position = blk_rq_pos(rq) + blk_rq_sectors(rq);
2512 static void cfq_deactivate_request(struct request_queue *q, struct request *rq)
2514 struct cfq_data *cfqd = q->elevator->elevator_data;
2516 WARN_ON(!cfqd->rq_in_driver);
2517 cfqd->rq_in_driver--;
2518 cfq_log_cfqq(cfqd, RQ_CFQQ(rq), "deactivate rq, drv=%d",
2519 cfqd->rq_in_driver);
2522 static void cfq_remove_request(struct request *rq)
2524 struct cfq_queue *cfqq = RQ_CFQQ(rq);
2526 if (cfqq->next_rq == rq)
2527 cfqq->next_rq = cfq_find_next_rq(cfqq->cfqd, cfqq, rq);
2529 list_del_init(&rq->queuelist);
2532 cfqq->cfqd->rq_queued--;
2533 cfqg_stats_update_io_remove(RQ_CFQG(rq), req_op(rq), rq->cmd_flags);
2534 if (rq->cmd_flags & REQ_PRIO) {
2535 WARN_ON(!cfqq->prio_pending);
2536 cfqq->prio_pending--;
2540 static int cfq_merge(struct request_queue *q, struct request **req,
2543 struct cfq_data *cfqd = q->elevator->elevator_data;
2544 struct request *__rq;
2546 __rq = cfq_find_rq_fmerge(cfqd, bio);
2547 if (__rq && elv_bio_merge_ok(__rq, bio)) {
2549 return ELEVATOR_FRONT_MERGE;
2552 return ELEVATOR_NO_MERGE;
2555 static void cfq_merged_request(struct request_queue *q, struct request *req,
2558 if (type == ELEVATOR_FRONT_MERGE) {
2559 struct cfq_queue *cfqq = RQ_CFQQ(req);
2561 cfq_reposition_rq_rb(cfqq, req);
2565 static void cfq_bio_merged(struct request_queue *q, struct request *req,
2568 cfqg_stats_update_io_merged(RQ_CFQG(req), bio_op(bio), bio->bi_rw);
2572 cfq_merged_requests(struct request_queue *q, struct request *rq,
2573 struct request *next)
2575 struct cfq_queue *cfqq = RQ_CFQQ(rq);
2576 struct cfq_data *cfqd = q->elevator->elevator_data;
2579 * reposition in fifo if next is older than rq
2581 if (!list_empty(&rq->queuelist) && !list_empty(&next->queuelist) &&
2582 next->fifo_time < rq->fifo_time &&
2583 cfqq == RQ_CFQQ(next)) {
2584 list_move(&rq->queuelist, &next->queuelist);
2585 rq->fifo_time = next->fifo_time;
2588 if (cfqq->next_rq == next)
2590 cfq_remove_request(next);
2591 cfqg_stats_update_io_merged(RQ_CFQG(rq), req_op(next), next->cmd_flags);
2593 cfqq = RQ_CFQQ(next);
2595 * all requests of this queue are merged to other queues, delete it
2596 * from the service tree. If it's the active_queue,
2597 * cfq_dispatch_requests() will choose to expire it or do idle
2599 if (cfq_cfqq_on_rr(cfqq) && RB_EMPTY_ROOT(&cfqq->sort_list) &&
2600 cfqq != cfqd->active_queue)
2601 cfq_del_cfqq_rr(cfqd, cfqq);
2604 static int cfq_allow_bio_merge(struct request_queue *q, struct request *rq,
2607 struct cfq_data *cfqd = q->elevator->elevator_data;
2608 struct cfq_io_cq *cic;
2609 struct cfq_queue *cfqq;
2612 * Disallow merge of a sync bio into an async request.
2614 if (cfq_bio_sync(bio) && !rq_is_sync(rq))
2618 * Lookup the cfqq that this bio will be queued with and allow
2619 * merge only if rq is queued there.
2621 cic = cfq_cic_lookup(cfqd, current->io_context);
2625 cfqq = cic_to_cfqq(cic, cfq_bio_sync(bio));
2626 return cfqq == RQ_CFQQ(rq);
2629 static int cfq_allow_rq_merge(struct request_queue *q, struct request *rq,
2630 struct request *next)
2632 return RQ_CFQQ(rq) == RQ_CFQQ(next);
2635 static inline void cfq_del_timer(struct cfq_data *cfqd, struct cfq_queue *cfqq)
2637 hrtimer_try_to_cancel(&cfqd->idle_slice_timer);
2638 cfqg_stats_update_idle_time(cfqq->cfqg);
2641 static void __cfq_set_active_queue(struct cfq_data *cfqd,
2642 struct cfq_queue *cfqq)
2645 cfq_log_cfqq(cfqd, cfqq, "set_active wl_class:%d wl_type:%d",
2646 cfqd->serving_wl_class, cfqd->serving_wl_type);
2647 cfqg_stats_update_avg_queue_size(cfqq->cfqg);
2648 cfqq->slice_start = 0;
2649 cfqq->dispatch_start = ktime_get_ns();
2650 cfqq->allocated_slice = 0;
2651 cfqq->slice_end = 0;
2652 cfqq->slice_dispatch = 0;
2653 cfqq->nr_sectors = 0;
2655 cfq_clear_cfqq_wait_request(cfqq);
2656 cfq_clear_cfqq_must_dispatch(cfqq);
2657 cfq_clear_cfqq_must_alloc_slice(cfqq);
2658 cfq_clear_cfqq_fifo_expire(cfqq);
2659 cfq_mark_cfqq_slice_new(cfqq);
2661 cfq_del_timer(cfqd, cfqq);
2664 cfqd->active_queue = cfqq;
2668 * current cfqq expired its slice (or was too idle), select new one
2671 __cfq_slice_expired(struct cfq_data *cfqd, struct cfq_queue *cfqq,
2674 cfq_log_cfqq(cfqd, cfqq, "slice expired t=%d", timed_out);
2676 if (cfq_cfqq_wait_request(cfqq))
2677 cfq_del_timer(cfqd, cfqq);
2679 cfq_clear_cfqq_wait_request(cfqq);
2680 cfq_clear_cfqq_wait_busy(cfqq);
2683 * If this cfqq is shared between multiple processes, check to
2684 * make sure that those processes are still issuing I/Os within
2685 * the mean seek distance. If not, it may be time to break the
2686 * queues apart again.
2688 if (cfq_cfqq_coop(cfqq) && CFQQ_SEEKY(cfqq))
2689 cfq_mark_cfqq_split_coop(cfqq);
2692 * store what was left of this slice, if the queue idled/timed out
2695 if (cfq_cfqq_slice_new(cfqq))
2696 cfqq->slice_resid = cfq_scaled_cfqq_slice(cfqd, cfqq);
2698 cfqq->slice_resid = cfqq->slice_end - ktime_get_ns();
2699 cfq_log_cfqq(cfqd, cfqq, "resid=%lld", cfqq->slice_resid);
2702 cfq_group_served(cfqd, cfqq->cfqg, cfqq);
2704 if (cfq_cfqq_on_rr(cfqq) && RB_EMPTY_ROOT(&cfqq->sort_list))
2705 cfq_del_cfqq_rr(cfqd, cfqq);
2707 cfq_resort_rr_list(cfqd, cfqq);
2709 if (cfqq == cfqd->active_queue)
2710 cfqd->active_queue = NULL;
2712 if (cfqd->active_cic) {
2713 put_io_context(cfqd->active_cic->icq.ioc);
2714 cfqd->active_cic = NULL;
2718 static inline void cfq_slice_expired(struct cfq_data *cfqd, bool timed_out)
2720 struct cfq_queue *cfqq = cfqd->active_queue;
2723 __cfq_slice_expired(cfqd, cfqq, timed_out);
2727 * Get next queue for service. Unless we have a queue preemption,
2728 * we'll simply select the first cfqq in the service tree.
2730 static struct cfq_queue *cfq_get_next_queue(struct cfq_data *cfqd)
2732 struct cfq_rb_root *st = st_for(cfqd->serving_group,
2733 cfqd->serving_wl_class, cfqd->serving_wl_type);
2735 if (!cfqd->rq_queued)
2738 /* There is nothing to dispatch */
2741 if (RB_EMPTY_ROOT(&st->rb))
2743 return cfq_rb_first(st);
2746 static struct cfq_queue *cfq_get_next_queue_forced(struct cfq_data *cfqd)
2748 struct cfq_group *cfqg;
2749 struct cfq_queue *cfqq;
2751 struct cfq_rb_root *st;
2753 if (!cfqd->rq_queued)
2756 cfqg = cfq_get_next_cfqg(cfqd);
2760 for_each_cfqg_st(cfqg, i, j, st)
2761 if ((cfqq = cfq_rb_first(st)) != NULL)
2767 * Get and set a new active queue for service.
2769 static struct cfq_queue *cfq_set_active_queue(struct cfq_data *cfqd,
2770 struct cfq_queue *cfqq)
2773 cfqq = cfq_get_next_queue(cfqd);
2775 __cfq_set_active_queue(cfqd, cfqq);
2779 static inline sector_t cfq_dist_from_last(struct cfq_data *cfqd,
2782 if (blk_rq_pos(rq) >= cfqd->last_position)
2783 return blk_rq_pos(rq) - cfqd->last_position;
2785 return cfqd->last_position - blk_rq_pos(rq);
2788 static inline int cfq_rq_close(struct cfq_data *cfqd, struct cfq_queue *cfqq,
2791 return cfq_dist_from_last(cfqd, rq) <= CFQQ_CLOSE_THR;
2794 static struct cfq_queue *cfqq_close(struct cfq_data *cfqd,
2795 struct cfq_queue *cur_cfqq)
2797 struct rb_root *root = &cfqd->prio_trees[cur_cfqq->org_ioprio];
2798 struct rb_node *parent, *node;
2799 struct cfq_queue *__cfqq;
2800 sector_t sector = cfqd->last_position;
2802 if (RB_EMPTY_ROOT(root))
2806 * First, if we find a request starting at the end of the last
2807 * request, choose it.
2809 __cfqq = cfq_prio_tree_lookup(cfqd, root, sector, &parent, NULL);
2814 * If the exact sector wasn't found, the parent of the NULL leaf
2815 * will contain the closest sector.
2817 __cfqq = rb_entry(parent, struct cfq_queue, p_node);
2818 if (cfq_rq_close(cfqd, cur_cfqq, __cfqq->next_rq))
2821 if (blk_rq_pos(__cfqq->next_rq) < sector)
2822 node = rb_next(&__cfqq->p_node);
2824 node = rb_prev(&__cfqq->p_node);
2828 __cfqq = rb_entry(node, struct cfq_queue, p_node);
2829 if (cfq_rq_close(cfqd, cur_cfqq, __cfqq->next_rq))
2837 * cur_cfqq - passed in so that we don't decide that the current queue is
2838 * closely cooperating with itself.
2840 * So, basically we're assuming that that cur_cfqq has dispatched at least
2841 * one request, and that cfqd->last_position reflects a position on the disk
2842 * associated with the I/O issued by cur_cfqq. I'm not sure this is a valid
2845 static struct cfq_queue *cfq_close_cooperator(struct cfq_data *cfqd,
2846 struct cfq_queue *cur_cfqq)
2848 struct cfq_queue *cfqq;
2850 if (cfq_class_idle(cur_cfqq))
2852 if (!cfq_cfqq_sync(cur_cfqq))
2854 if (CFQQ_SEEKY(cur_cfqq))
2858 * Don't search priority tree if it's the only queue in the group.
2860 if (cur_cfqq->cfqg->nr_cfqq == 1)
2864 * We should notice if some of the queues are cooperating, eg
2865 * working closely on the same area of the disk. In that case,
2866 * we can group them together and don't waste time idling.
2868 cfqq = cfqq_close(cfqd, cur_cfqq);
2872 /* If new queue belongs to different cfq_group, don't choose it */
2873 if (cur_cfqq->cfqg != cfqq->cfqg)
2877 * It only makes sense to merge sync queues.
2879 if (!cfq_cfqq_sync(cfqq))
2881 if (CFQQ_SEEKY(cfqq))
2885 * Do not merge queues of different priority classes
2887 if (cfq_class_rt(cfqq) != cfq_class_rt(cur_cfqq))
2894 * Determine whether we should enforce idle window for this queue.
2897 static bool cfq_should_idle(struct cfq_data *cfqd, struct cfq_queue *cfqq)
2899 enum wl_class_t wl_class = cfqq_class(cfqq);
2900 struct cfq_rb_root *st = cfqq->service_tree;
2905 if (!cfqd->cfq_slice_idle)
2908 /* We never do for idle class queues. */
2909 if (wl_class == IDLE_WORKLOAD)
2912 /* We do for queues that were marked with idle window flag. */
2913 if (cfq_cfqq_idle_window(cfqq) &&
2914 !(blk_queue_nonrot(cfqd->queue) && cfqd->hw_tag))
2918 * Otherwise, we do only if they are the last ones
2919 * in their service tree.
2921 if (st->count == 1 && cfq_cfqq_sync(cfqq) &&
2922 !cfq_io_thinktime_big(cfqd, &st->ttime, false))
2924 cfq_log_cfqq(cfqd, cfqq, "Not idling. st->count:%d", st->count);
2928 static void cfq_arm_slice_timer(struct cfq_data *cfqd)
2930 struct cfq_queue *cfqq = cfqd->active_queue;
2931 struct cfq_rb_root *st = cfqq->service_tree;
2932 struct cfq_io_cq *cic;
2933 u64 sl, group_idle = 0;
2934 u64 now = ktime_get_ns();
2937 * SSD device without seek penalty, disable idling. But only do so
2938 * for devices that support queuing, otherwise we still have a problem
2939 * with sync vs async workloads.
2941 if (blk_queue_nonrot(cfqd->queue) && cfqd->hw_tag)
2944 WARN_ON(!RB_EMPTY_ROOT(&cfqq->sort_list));
2945 WARN_ON(cfq_cfqq_slice_new(cfqq));
2948 * idle is disabled, either manually or by past process history
2950 if (!cfq_should_idle(cfqd, cfqq)) {
2951 /* no queue idling. Check for group idling */
2952 if (cfqd->cfq_group_idle)
2953 group_idle = cfqd->cfq_group_idle;
2959 * still active requests from this queue, don't idle
2961 if (cfqq->dispatched)
2965 * task has exited, don't wait
2967 cic = cfqd->active_cic;
2968 if (!cic || !atomic_read(&cic->icq.ioc->active_ref))
2972 * If our average think time is larger than the remaining time
2973 * slice, then don't idle. This avoids overrunning the allotted
2976 if (sample_valid(cic->ttime.ttime_samples) &&
2977 (cfqq->slice_end - now < cic->ttime.ttime_mean)) {
2978 cfq_log_cfqq(cfqd, cfqq, "Not idling. think_time:%llu",
2979 cic->ttime.ttime_mean);
2984 * There are other queues in the group or this is the only group and
2985 * it has too big thinktime, don't do group idle.
2988 (cfqq->cfqg->nr_cfqq > 1 ||
2989 cfq_io_thinktime_big(cfqd, &st->ttime, true)))
2992 cfq_mark_cfqq_wait_request(cfqq);
2995 sl = cfqd->cfq_group_idle;
2997 sl = cfqd->cfq_slice_idle;
2999 hrtimer_start(&cfqd->idle_slice_timer, ns_to_ktime(sl),
3001 cfqg_stats_set_start_idle_time(cfqq->cfqg);
3002 cfq_log_cfqq(cfqd, cfqq, "arm_idle: %llu group_idle: %d", sl,
3003 group_idle ? 1 : 0);
3007 * Move request from internal lists to the request queue dispatch list.
3009 static void cfq_dispatch_insert(struct request_queue *q, struct request *rq)
3011 struct cfq_data *cfqd = q->elevator->elevator_data;
3012 struct cfq_queue *cfqq = RQ_CFQQ(rq);
3014 cfq_log_cfqq(cfqd, cfqq, "dispatch_insert");
3016 cfqq->next_rq = cfq_find_next_rq(cfqd, cfqq, rq);
3017 cfq_remove_request(rq);
3019 (RQ_CFQG(rq))->dispatched++;
3020 elv_dispatch_sort(q, rq);
3022 cfqd->rq_in_flight[cfq_cfqq_sync(cfqq)]++;
3023 cfqq->nr_sectors += blk_rq_sectors(rq);
3027 * return expired entry, or NULL to just start from scratch in rbtree
3029 static struct request *cfq_check_fifo(struct cfq_queue *cfqq)
3031 struct request *rq = NULL;
3033 if (cfq_cfqq_fifo_expire(cfqq))
3036 cfq_mark_cfqq_fifo_expire(cfqq);
3038 if (list_empty(&cfqq->fifo))
3041 rq = rq_entry_fifo(cfqq->fifo.next);
3042 if (ktime_get_ns() < rq->fifo_time)
3045 cfq_log_cfqq(cfqq->cfqd, cfqq, "fifo=%p", rq);
3050 cfq_prio_to_maxrq(struct cfq_data *cfqd, struct cfq_queue *cfqq)
3052 const int base_rq = cfqd->cfq_slice_async_rq;
3054 WARN_ON(cfqq->ioprio >= IOPRIO_BE_NR);
3056 return 2 * base_rq * (IOPRIO_BE_NR - cfqq->ioprio);
3060 * Must be called with the queue_lock held.
3062 static int cfqq_process_refs(struct cfq_queue *cfqq)
3064 int process_refs, io_refs;
3066 io_refs = cfqq->allocated[READ] + cfqq->allocated[WRITE];
3067 process_refs = cfqq->ref - io_refs;
3068 BUG_ON(process_refs < 0);
3069 return process_refs;
3072 static void cfq_setup_merge(struct cfq_queue *cfqq, struct cfq_queue *new_cfqq)
3074 int process_refs, new_process_refs;
3075 struct cfq_queue *__cfqq;
3078 * If there are no process references on the new_cfqq, then it is
3079 * unsafe to follow the ->new_cfqq chain as other cfqq's in the
3080 * chain may have dropped their last reference (not just their
3081 * last process reference).
3083 if (!cfqq_process_refs(new_cfqq))
3086 /* Avoid a circular list and skip interim queue merges */
3087 while ((__cfqq = new_cfqq->new_cfqq)) {
3093 process_refs = cfqq_process_refs(cfqq);
3094 new_process_refs = cfqq_process_refs(new_cfqq);
3096 * If the process for the cfqq has gone away, there is no
3097 * sense in merging the queues.
3099 if (process_refs == 0 || new_process_refs == 0)
3103 * Merge in the direction of the lesser amount of work.
3105 if (new_process_refs >= process_refs) {
3106 cfqq->new_cfqq = new_cfqq;
3107 new_cfqq->ref += process_refs;
3109 new_cfqq->new_cfqq = cfqq;
3110 cfqq->ref += new_process_refs;
3114 static enum wl_type_t cfq_choose_wl_type(struct cfq_data *cfqd,
3115 struct cfq_group *cfqg, enum wl_class_t wl_class)
3117 struct cfq_queue *queue;
3119 bool key_valid = false;
3121 enum wl_type_t cur_best = SYNC_NOIDLE_WORKLOAD;
3123 for (i = 0; i <= SYNC_WORKLOAD; ++i) {
3124 /* select the one with lowest rb_key */
3125 queue = cfq_rb_first(st_for(cfqg, wl_class, i));
3127 (!key_valid || queue->rb_key < lowest_key)) {
3128 lowest_key = queue->rb_key;
3138 choose_wl_class_and_type(struct cfq_data *cfqd, struct cfq_group *cfqg)
3142 struct cfq_rb_root *st;
3144 enum wl_class_t original_class = cfqd->serving_wl_class;
3145 u64 now = ktime_get_ns();
3147 /* Choose next priority. RT > BE > IDLE */
3148 if (cfq_group_busy_queues_wl(RT_WORKLOAD, cfqd, cfqg))
3149 cfqd->serving_wl_class = RT_WORKLOAD;
3150 else if (cfq_group_busy_queues_wl(BE_WORKLOAD, cfqd, cfqg))
3151 cfqd->serving_wl_class = BE_WORKLOAD;
3153 cfqd->serving_wl_class = IDLE_WORKLOAD;
3154 cfqd->workload_expires = now + jiffies_to_nsecs(1);
3158 if (original_class != cfqd->serving_wl_class)
3162 * For RT and BE, we have to choose also the type
3163 * (SYNC, SYNC_NOIDLE, ASYNC), and to compute a workload
3166 st = st_for(cfqg, cfqd->serving_wl_class, cfqd->serving_wl_type);
3170 * check workload expiration, and that we still have other queues ready
3172 if (count && !(now > cfqd->workload_expires))
3176 /* otherwise select new workload type */
3177 cfqd->serving_wl_type = cfq_choose_wl_type(cfqd, cfqg,
3178 cfqd->serving_wl_class);
3179 st = st_for(cfqg, cfqd->serving_wl_class, cfqd->serving_wl_type);
3183 * the workload slice is computed as a fraction of target latency
3184 * proportional to the number of queues in that workload, over
3185 * all the queues in the same priority class
3187 group_slice = cfq_group_slice(cfqd, cfqg);
3189 slice = div_u64(group_slice * count,
3190 max_t(unsigned, cfqg->busy_queues_avg[cfqd->serving_wl_class],
3191 cfq_group_busy_queues_wl(cfqd->serving_wl_class, cfqd,
3194 if (cfqd->serving_wl_type == ASYNC_WORKLOAD) {
3198 * Async queues are currently system wide. Just taking
3199 * proportion of queues with-in same group will lead to higher
3200 * async ratio system wide as generally root group is going
3201 * to have higher weight. A more accurate thing would be to
3202 * calculate system wide asnc/sync ratio.
3204 tmp = cfqd->cfq_target_latency *
3205 cfqg_busy_async_queues(cfqd, cfqg);
3206 tmp = div_u64(tmp, cfqd->busy_queues);
3207 slice = min_t(u64, slice, tmp);
3209 /* async workload slice is scaled down according to
3210 * the sync/async slice ratio. */
3211 slice = div64_u64(slice*cfqd->cfq_slice[0], cfqd->cfq_slice[1]);
3213 /* sync workload slice is at least 2 * cfq_slice_idle */
3214 slice = max(slice, 2 * cfqd->cfq_slice_idle);
3216 slice = max_t(u64, slice, CFQ_MIN_TT);
3217 cfq_log(cfqd, "workload slice:%llu", slice);
3218 cfqd->workload_expires = now + slice;
3221 static struct cfq_group *cfq_get_next_cfqg(struct cfq_data *cfqd)
3223 struct cfq_rb_root *st = &cfqd->grp_service_tree;
3224 struct cfq_group *cfqg;
3226 if (RB_EMPTY_ROOT(&st->rb))
3228 cfqg = cfq_rb_first_group(st);
3229 update_min_vdisktime(st);
3233 static void cfq_choose_cfqg(struct cfq_data *cfqd)
3235 struct cfq_group *cfqg = cfq_get_next_cfqg(cfqd);
3236 u64 now = ktime_get_ns();
3238 cfqd->serving_group = cfqg;
3240 /* Restore the workload type data */
3241 if (cfqg->saved_wl_slice) {
3242 cfqd->workload_expires = now + cfqg->saved_wl_slice;
3243 cfqd->serving_wl_type = cfqg->saved_wl_type;
3244 cfqd->serving_wl_class = cfqg->saved_wl_class;
3246 cfqd->workload_expires = now - 1;
3248 choose_wl_class_and_type(cfqd, cfqg);
3252 * Select a queue for service. If we have a current active queue,
3253 * check whether to continue servicing it, or retrieve and set a new one.
3255 static struct cfq_queue *cfq_select_queue(struct cfq_data *cfqd)
3257 struct cfq_queue *cfqq, *new_cfqq = NULL;
3258 u64 now = ktime_get_ns();
3260 cfqq = cfqd->active_queue;
3264 if (!cfqd->rq_queued)
3268 * We were waiting for group to get backlogged. Expire the queue
3270 if (cfq_cfqq_wait_busy(cfqq) && !RB_EMPTY_ROOT(&cfqq->sort_list))
3274 * The active queue has run out of time, expire it and select new.
3276 if (cfq_slice_used(cfqq) && !cfq_cfqq_must_dispatch(cfqq)) {
3278 * If slice had not expired at the completion of last request
3279 * we might not have turned on wait_busy flag. Don't expire
3280 * the queue yet. Allow the group to get backlogged.
3282 * The very fact that we have used the slice, that means we
3283 * have been idling all along on this queue and it should be
3284 * ok to wait for this request to complete.
3286 if (cfqq->cfqg->nr_cfqq == 1 && RB_EMPTY_ROOT(&cfqq->sort_list)
3287 && cfqq->dispatched && cfq_should_idle(cfqd, cfqq)) {
3291 goto check_group_idle;
3295 * The active queue has requests and isn't expired, allow it to
3298 if (!RB_EMPTY_ROOT(&cfqq->sort_list))
3302 * If another queue has a request waiting within our mean seek
3303 * distance, let it run. The expire code will check for close
3304 * cooperators and put the close queue at the front of the service
3305 * tree. If possible, merge the expiring queue with the new cfqq.
3307 new_cfqq = cfq_close_cooperator(cfqd, cfqq);
3309 if (!cfqq->new_cfqq)
3310 cfq_setup_merge(cfqq, new_cfqq);
3315 * No requests pending. If the active queue still has requests in
3316 * flight or is idling for a new request, allow either of these
3317 * conditions to happen (or time out) before selecting a new queue.
3319 if (hrtimer_active(&cfqd->idle_slice_timer)) {
3325 * This is a deep seek queue, but the device is much faster than
3326 * the queue can deliver, don't idle
3328 if (CFQQ_SEEKY(cfqq) && cfq_cfqq_idle_window(cfqq) &&
3329 (cfq_cfqq_slice_new(cfqq) ||
3330 (cfqq->slice_end - now > now - cfqq->slice_start))) {
3331 cfq_clear_cfqq_deep(cfqq);
3332 cfq_clear_cfqq_idle_window(cfqq);
3335 if (cfqq->dispatched && cfq_should_idle(cfqd, cfqq)) {
3341 * If group idle is enabled and there are requests dispatched from
3342 * this group, wait for requests to complete.
3345 if (cfqd->cfq_group_idle && cfqq->cfqg->nr_cfqq == 1 &&
3346 cfqq->cfqg->dispatched &&
3347 !cfq_io_thinktime_big(cfqd, &cfqq->cfqg->ttime, true)) {
3353 cfq_slice_expired(cfqd, 0);
3356 * Current queue expired. Check if we have to switch to a new
3360 cfq_choose_cfqg(cfqd);
3362 cfqq = cfq_set_active_queue(cfqd, new_cfqq);
3367 static int __cfq_forced_dispatch_cfqq(struct cfq_queue *cfqq)
3371 while (cfqq->next_rq) {
3372 cfq_dispatch_insert(cfqq->cfqd->queue, cfqq->next_rq);
3376 BUG_ON(!list_empty(&cfqq->fifo));
3378 /* By default cfqq is not expired if it is empty. Do it explicitly */
3379 __cfq_slice_expired(cfqq->cfqd, cfqq, 0);
3384 * Drain our current requests. Used for barriers and when switching
3385 * io schedulers on-the-fly.
3387 static int cfq_forced_dispatch(struct cfq_data *cfqd)
3389 struct cfq_queue *cfqq;
3392 /* Expire the timeslice of the current active queue first */
3393 cfq_slice_expired(cfqd, 0);
3394 while ((cfqq = cfq_get_next_queue_forced(cfqd)) != NULL) {
3395 __cfq_set_active_queue(cfqd, cfqq);
3396 dispatched += __cfq_forced_dispatch_cfqq(cfqq);
3399 BUG_ON(cfqd->busy_queues);
3401 cfq_log(cfqd, "forced_dispatch=%d", dispatched);
3405 static inline bool cfq_slice_used_soon(struct cfq_data *cfqd,
3406 struct cfq_queue *cfqq)
3408 u64 now = ktime_get_ns();
3410 /* the queue hasn't finished any request, can't estimate */
3411 if (cfq_cfqq_slice_new(cfqq))
3413 if (now + cfqd->cfq_slice_idle * cfqq->dispatched > cfqq->slice_end)
3419 static bool cfq_may_dispatch(struct cfq_data *cfqd, struct cfq_queue *cfqq)
3421 unsigned int max_dispatch;
3424 * Drain async requests before we start sync IO
3426 if (cfq_should_idle(cfqd, cfqq) && cfqd->rq_in_flight[BLK_RW_ASYNC])
3430 * If this is an async queue and we have sync IO in flight, let it wait
3432 if (cfqd->rq_in_flight[BLK_RW_SYNC] && !cfq_cfqq_sync(cfqq))
3435 max_dispatch = max_t(unsigned int, cfqd->cfq_quantum / 2, 1);
3436 if (cfq_class_idle(cfqq))
3440 * Does this cfqq already have too much IO in flight?
3442 if (cfqq->dispatched >= max_dispatch) {
3443 bool promote_sync = false;
3445 * idle queue must always only have a single IO in flight
3447 if (cfq_class_idle(cfqq))
3451 * If there is only one sync queue
3452 * we can ignore async queue here and give the sync
3453 * queue no dispatch limit. The reason is a sync queue can
3454 * preempt async queue, limiting the sync queue doesn't make
3455 * sense. This is useful for aiostress test.
3457 if (cfq_cfqq_sync(cfqq) && cfqd->busy_sync_queues == 1)
3458 promote_sync = true;
3461 * We have other queues, don't allow more IO from this one
3463 if (cfqd->busy_queues > 1 && cfq_slice_used_soon(cfqd, cfqq) &&
3468 * Sole queue user, no limit
3470 if (cfqd->busy_queues == 1 || promote_sync)
3474 * Normally we start throttling cfqq when cfq_quantum/2
3475 * requests have been dispatched. But we can drive
3476 * deeper queue depths at the beginning of slice
3477 * subjected to upper limit of cfq_quantum.
3479 max_dispatch = cfqd->cfq_quantum;
3483 * Async queues must wait a bit before being allowed dispatch.
3484 * We also ramp up the dispatch depth gradually for async IO,
3485 * based on the last sync IO we serviced
3487 if (!cfq_cfqq_sync(cfqq) && cfqd->cfq_latency) {
3488 u64 last_sync = ktime_get_ns() - cfqd->last_delayed_sync;
3491 depth = div64_u64(last_sync, cfqd->cfq_slice[1]);
3492 if (!depth && !cfqq->dispatched)
3494 if (depth < max_dispatch)
3495 max_dispatch = depth;
3499 * If we're below the current max, allow a dispatch
3501 return cfqq->dispatched < max_dispatch;
3505 * Dispatch a request from cfqq, moving them to the request queue
3508 static bool cfq_dispatch_request(struct cfq_data *cfqd, struct cfq_queue *cfqq)
3512 BUG_ON(RB_EMPTY_ROOT(&cfqq->sort_list));
3514 if (!cfq_may_dispatch(cfqd, cfqq))
3518 * follow expired path, else get first next available
3520 rq = cfq_check_fifo(cfqq);
3525 * insert request into driver dispatch list
3527 cfq_dispatch_insert(cfqd->queue, rq);
3529 if (!cfqd->active_cic) {
3530 struct cfq_io_cq *cic = RQ_CIC(rq);
3532 atomic_long_inc(&cic->icq.ioc->refcount);
3533 cfqd->active_cic = cic;
3540 * Find the cfqq that we need to service and move a request from that to the
3543 static int cfq_dispatch_requests(struct request_queue *q, int force)
3545 struct cfq_data *cfqd = q->elevator->elevator_data;
3546 struct cfq_queue *cfqq;
3548 if (!cfqd->busy_queues)
3551 if (unlikely(force))
3552 return cfq_forced_dispatch(cfqd);
3554 cfqq = cfq_select_queue(cfqd);
3559 * Dispatch a request from this cfqq, if it is allowed
3561 if (!cfq_dispatch_request(cfqd, cfqq))
3564 cfqq->slice_dispatch++;
3565 cfq_clear_cfqq_must_dispatch(cfqq);
3568 * expire an async queue immediately if it has used up its slice. idle
3569 * queue always expire after 1 dispatch round.
3571 if (cfqd->busy_queues > 1 && ((!cfq_cfqq_sync(cfqq) &&
3572 cfqq->slice_dispatch >= cfq_prio_to_maxrq(cfqd, cfqq)) ||
3573 cfq_class_idle(cfqq))) {
3574 cfqq->slice_end = ktime_get_ns() + 1;
3575 cfq_slice_expired(cfqd, 0);
3578 cfq_log_cfqq(cfqd, cfqq, "dispatched a request");
3583 * task holds one reference to the queue, dropped when task exits. each rq
3584 * in-flight on this queue also holds a reference, dropped when rq is freed.
3586 * Each cfq queue took a reference on the parent group. Drop it now.
3587 * queue lock must be held here.
3589 static void cfq_put_queue(struct cfq_queue *cfqq)
3591 struct cfq_data *cfqd = cfqq->cfqd;
3592 struct cfq_group *cfqg;
3594 BUG_ON(cfqq->ref <= 0);
3600 cfq_log_cfqq(cfqd, cfqq, "put_queue");
3601 BUG_ON(rb_first(&cfqq->sort_list));
3602 BUG_ON(cfqq->allocated[READ] + cfqq->allocated[WRITE]);
3605 if (unlikely(cfqd->active_queue == cfqq)) {
3606 __cfq_slice_expired(cfqd, cfqq, 0);
3607 cfq_schedule_dispatch(cfqd);
3610 BUG_ON(cfq_cfqq_on_rr(cfqq));
3611 kmem_cache_free(cfq_pool, cfqq);
3615 static void cfq_put_cooperator(struct cfq_queue *cfqq)
3617 struct cfq_queue *__cfqq, *next;
3620 * If this queue was scheduled to merge with another queue, be
3621 * sure to drop the reference taken on that queue (and others in
3622 * the merge chain). See cfq_setup_merge and cfq_merge_cfqqs.
3624 __cfqq = cfqq->new_cfqq;
3626 if (__cfqq == cfqq) {
3627 WARN(1, "cfqq->new_cfqq loop detected\n");
3630 next = __cfqq->new_cfqq;
3631 cfq_put_queue(__cfqq);
3636 static void cfq_exit_cfqq(struct cfq_data *cfqd, struct cfq_queue *cfqq)
3638 if (unlikely(cfqq == cfqd->active_queue)) {
3639 __cfq_slice_expired(cfqd, cfqq, 0);
3640 cfq_schedule_dispatch(cfqd);
3643 cfq_put_cooperator(cfqq);
3645 cfq_put_queue(cfqq);
3648 static void cfq_init_icq(struct io_cq *icq)
3650 struct cfq_io_cq *cic = icq_to_cic(icq);
3652 cic->ttime.last_end_request = ktime_get_ns();
3655 static void cfq_exit_icq(struct io_cq *icq)
3657 struct cfq_io_cq *cic = icq_to_cic(icq);
3658 struct cfq_data *cfqd = cic_to_cfqd(cic);
3660 if (cic_to_cfqq(cic, false)) {
3661 cfq_exit_cfqq(cfqd, cic_to_cfqq(cic, false));
3662 cic_set_cfqq(cic, NULL, false);
3665 if (cic_to_cfqq(cic, true)) {
3666 cfq_exit_cfqq(cfqd, cic_to_cfqq(cic, true));
3667 cic_set_cfqq(cic, NULL, true);
3671 static void cfq_init_prio_data(struct cfq_queue *cfqq, struct cfq_io_cq *cic)
3673 struct task_struct *tsk = current;
3676 if (!cfq_cfqq_prio_changed(cfqq))
3679 ioprio_class = IOPRIO_PRIO_CLASS(cic->ioprio);
3680 switch (ioprio_class) {
3682 printk(KERN_ERR "cfq: bad prio %x\n", ioprio_class);
3683 case IOPRIO_CLASS_NONE:
3685 * no prio set, inherit CPU scheduling settings
3687 cfqq->ioprio = task_nice_ioprio(tsk);
3688 cfqq->ioprio_class = task_nice_ioclass(tsk);
3690 case IOPRIO_CLASS_RT:
3691 cfqq->ioprio = IOPRIO_PRIO_DATA(cic->ioprio);
3692 cfqq->ioprio_class = IOPRIO_CLASS_RT;
3694 case IOPRIO_CLASS_BE:
3695 cfqq->ioprio = IOPRIO_PRIO_DATA(cic->ioprio);
3696 cfqq->ioprio_class = IOPRIO_CLASS_BE;
3698 case IOPRIO_CLASS_IDLE:
3699 cfqq->ioprio_class = IOPRIO_CLASS_IDLE;
3701 cfq_clear_cfqq_idle_window(cfqq);
3706 * keep track of original prio settings in case we have to temporarily
3707 * elevate the priority of this queue
3709 cfqq->org_ioprio = cfqq->ioprio;
3710 cfqq->org_ioprio_class = cfqq->ioprio_class;
3711 cfq_clear_cfqq_prio_changed(cfqq);
3714 static void check_ioprio_changed(struct cfq_io_cq *cic, struct bio *bio)
3716 int ioprio = cic->icq.ioc->ioprio;
3717 struct cfq_data *cfqd = cic_to_cfqd(cic);
3718 struct cfq_queue *cfqq;
3721 * Check whether ioprio has changed. The condition may trigger
3722 * spuriously on a newly created cic but there's no harm.
3724 if (unlikely(!cfqd) || likely(cic->ioprio == ioprio))
3727 cfqq = cic_to_cfqq(cic, false);
3729 cfq_put_queue(cfqq);
3730 cfqq = cfq_get_queue(cfqd, BLK_RW_ASYNC, cic, bio);
3731 cic_set_cfqq(cic, cfqq, false);
3734 cfqq = cic_to_cfqq(cic, true);
3736 cfq_mark_cfqq_prio_changed(cfqq);
3738 cic->ioprio = ioprio;
3741 static void cfq_init_cfqq(struct cfq_data *cfqd, struct cfq_queue *cfqq,
3742 pid_t pid, bool is_sync)
3744 RB_CLEAR_NODE(&cfqq->rb_node);
3745 RB_CLEAR_NODE(&cfqq->p_node);
3746 INIT_LIST_HEAD(&cfqq->fifo);
3751 cfq_mark_cfqq_prio_changed(cfqq);
3754 if (!cfq_class_idle(cfqq))
3755 cfq_mark_cfqq_idle_window(cfqq);
3756 cfq_mark_cfqq_sync(cfqq);
3761 #ifdef CONFIG_CFQ_GROUP_IOSCHED
3762 static void check_blkcg_changed(struct cfq_io_cq *cic, struct bio *bio)
3764 struct cfq_data *cfqd = cic_to_cfqd(cic);
3765 struct cfq_queue *cfqq;
3769 serial_nr = bio_blkcg(bio)->css.serial_nr;
3773 * Check whether blkcg has changed. The condition may trigger
3774 * spuriously on a newly created cic but there's no harm.
3776 if (unlikely(!cfqd) || likely(cic->blkcg_serial_nr == serial_nr))
3780 * Drop reference to queues. New queues will be assigned in new
3781 * group upon arrival of fresh requests.
3783 cfqq = cic_to_cfqq(cic, false);
3785 cfq_log_cfqq(cfqd, cfqq, "changed cgroup");
3786 cic_set_cfqq(cic, NULL, false);
3787 cfq_put_queue(cfqq);
3790 cfqq = cic_to_cfqq(cic, true);
3792 cfq_log_cfqq(cfqd, cfqq, "changed cgroup");
3793 cic_set_cfqq(cic, NULL, true);
3794 cfq_put_queue(cfqq);
3797 cic->blkcg_serial_nr = serial_nr;
3800 static inline void check_blkcg_changed(struct cfq_io_cq *cic, struct bio *bio) { }
3801 #endif /* CONFIG_CFQ_GROUP_IOSCHED */
3803 static struct cfq_queue **
3804 cfq_async_queue_prio(struct cfq_group *cfqg, int ioprio_class, int ioprio)
3806 switch (ioprio_class) {
3807 case IOPRIO_CLASS_RT:
3808 return &cfqg->async_cfqq[0][ioprio];
3809 case IOPRIO_CLASS_NONE:
3810 ioprio = IOPRIO_NORM;
3812 case IOPRIO_CLASS_BE:
3813 return &cfqg->async_cfqq[1][ioprio];
3814 case IOPRIO_CLASS_IDLE:
3815 return &cfqg->async_idle_cfqq;
3821 static struct cfq_queue *
3822 cfq_get_queue(struct cfq_data *cfqd, bool is_sync, struct cfq_io_cq *cic,
3825 int ioprio_class = IOPRIO_PRIO_CLASS(cic->ioprio);
3826 int ioprio = IOPRIO_PRIO_DATA(cic->ioprio);
3827 struct cfq_queue **async_cfqq = NULL;
3828 struct cfq_queue *cfqq;
3829 struct cfq_group *cfqg;
3832 cfqg = cfq_lookup_cfqg(cfqd, bio_blkcg(bio));
3834 cfqq = &cfqd->oom_cfqq;
3839 if (!ioprio_valid(cic->ioprio)) {
3840 struct task_struct *tsk = current;
3841 ioprio = task_nice_ioprio(tsk);
3842 ioprio_class = task_nice_ioclass(tsk);
3844 async_cfqq = cfq_async_queue_prio(cfqg, ioprio_class, ioprio);
3850 cfqq = kmem_cache_alloc_node(cfq_pool, GFP_NOWAIT | __GFP_ZERO,
3853 cfqq = &cfqd->oom_cfqq;
3857 cfq_init_cfqq(cfqd, cfqq, current->pid, is_sync);
3858 cfq_init_prio_data(cfqq, cic);
3859 cfq_link_cfqq_cfqg(cfqq, cfqg);
3860 cfq_log_cfqq(cfqd, cfqq, "alloced");
3863 /* a new async queue is created, pin and remember */
3874 __cfq_update_io_thinktime(struct cfq_ttime *ttime, u64 slice_idle)
3876 u64 elapsed = ktime_get_ns() - ttime->last_end_request;
3877 elapsed = min(elapsed, 2UL * slice_idle);
3879 ttime->ttime_samples = (7*ttime->ttime_samples + 256) / 8;
3880 ttime->ttime_total = div_u64(7*ttime->ttime_total + 256*elapsed, 8);
3881 ttime->ttime_mean = div64_ul(ttime->ttime_total + 128,
3882 ttime->ttime_samples);
3886 cfq_update_io_thinktime(struct cfq_data *cfqd, struct cfq_queue *cfqq,
3887 struct cfq_io_cq *cic)
3889 if (cfq_cfqq_sync(cfqq)) {
3890 __cfq_update_io_thinktime(&cic->ttime, cfqd->cfq_slice_idle);
3891 __cfq_update_io_thinktime(&cfqq->service_tree->ttime,
3892 cfqd->cfq_slice_idle);
3894 #ifdef CONFIG_CFQ_GROUP_IOSCHED
3895 __cfq_update_io_thinktime(&cfqq->cfqg->ttime, cfqd->cfq_group_idle);
3900 cfq_update_io_seektime(struct cfq_data *cfqd, struct cfq_queue *cfqq,
3904 sector_t n_sec = blk_rq_sectors(rq);
3905 if (cfqq->last_request_pos) {
3906 if (cfqq->last_request_pos < blk_rq_pos(rq))
3907 sdist = blk_rq_pos(rq) - cfqq->last_request_pos;
3909 sdist = cfqq->last_request_pos - blk_rq_pos(rq);
3912 cfqq->seek_history <<= 1;
3913 if (blk_queue_nonrot(cfqd->queue))
3914 cfqq->seek_history |= (n_sec < CFQQ_SECT_THR_NONROT);
3916 cfqq->seek_history |= (sdist > CFQQ_SEEK_THR);
3920 * Disable idle window if the process thinks too long or seeks so much that
3924 cfq_update_idle_window(struct cfq_data *cfqd, struct cfq_queue *cfqq,
3925 struct cfq_io_cq *cic)
3927 int old_idle, enable_idle;
3930 * Don't idle for async or idle io prio class
3932 if (!cfq_cfqq_sync(cfqq) || cfq_class_idle(cfqq))
3935 enable_idle = old_idle = cfq_cfqq_idle_window(cfqq);
3937 if (cfqq->queued[0] + cfqq->queued[1] >= 4)
3938 cfq_mark_cfqq_deep(cfqq);
3940 if (cfqq->next_rq && (cfqq->next_rq->cmd_flags & REQ_NOIDLE))
3942 else if (!atomic_read(&cic->icq.ioc->active_ref) ||
3943 !cfqd->cfq_slice_idle ||
3944 (!cfq_cfqq_deep(cfqq) && CFQQ_SEEKY(cfqq)))
3946 else if (sample_valid(cic->ttime.ttime_samples)) {
3947 if (cic->ttime.ttime_mean > cfqd->cfq_slice_idle)
3953 if (old_idle != enable_idle) {
3954 cfq_log_cfqq(cfqd, cfqq, "idle=%d", enable_idle);
3956 cfq_mark_cfqq_idle_window(cfqq);
3958 cfq_clear_cfqq_idle_window(cfqq);
3963 * Check if new_cfqq should preempt the currently active queue. Return 0 for
3964 * no or if we aren't sure, a 1 will cause a preempt.
3967 cfq_should_preempt(struct cfq_data *cfqd, struct cfq_queue *new_cfqq,
3970 struct cfq_queue *cfqq;
3972 cfqq = cfqd->active_queue;
3976 if (cfq_class_idle(new_cfqq))
3979 if (cfq_class_idle(cfqq))
3983 * Don't allow a non-RT request to preempt an ongoing RT cfqq timeslice.
3985 if (cfq_class_rt(cfqq) && !cfq_class_rt(new_cfqq))
3989 * if the new request is sync, but the currently running queue is
3990 * not, let the sync request have priority.
3992 if (rq_is_sync(rq) && !cfq_cfqq_sync(cfqq))
3996 * Treat ancestors of current cgroup the same way as current cgroup.
3997 * For anybody else we disallow preemption to guarantee service
3998 * fairness among cgroups.
4000 if (!cfqg_is_descendant(cfqq->cfqg, new_cfqq->cfqg))
4003 if (cfq_slice_used(cfqq))
4007 * Allow an RT request to pre-empt an ongoing non-RT cfqq timeslice.
4009 if (cfq_class_rt(new_cfqq) && !cfq_class_rt(cfqq))
4012 WARN_ON_ONCE(cfqq->ioprio_class != new_cfqq->ioprio_class);
4013 /* Allow preemption only if we are idling on sync-noidle tree */
4014 if (cfqd->serving_wl_type == SYNC_NOIDLE_WORKLOAD &&
4015 cfqq_type(new_cfqq) == SYNC_NOIDLE_WORKLOAD &&
4016 RB_EMPTY_ROOT(&cfqq->sort_list))
4020 * So both queues are sync. Let the new request get disk time if
4021 * it's a metadata request and the current queue is doing regular IO.
4023 if ((rq->cmd_flags & REQ_PRIO) && !cfqq->prio_pending)
4026 /* An idle queue should not be idle now for some reason */
4027 if (RB_EMPTY_ROOT(&cfqq->sort_list) && !cfq_should_idle(cfqd, cfqq))
4030 if (!cfqd->active_cic || !cfq_cfqq_wait_request(cfqq))
4034 * if this request is as-good as one we would expect from the
4035 * current cfqq, let it preempt
4037 if (cfq_rq_close(cfqd, cfqq, rq))
4044 * cfqq preempts the active queue. if we allowed preempt with no slice left,
4045 * let it have half of its nominal slice.
4047 static void cfq_preempt_queue(struct cfq_data *cfqd, struct cfq_queue *cfqq)
4049 enum wl_type_t old_type = cfqq_type(cfqd->active_queue);
4051 cfq_log_cfqq(cfqd, cfqq, "preempt");
4052 cfq_slice_expired(cfqd, 1);
4055 * workload type is changed, don't save slice, otherwise preempt
4058 if (old_type != cfqq_type(cfqq))
4059 cfqq->cfqg->saved_wl_slice = 0;
4062 * Put the new queue at the front of the of the current list,
4063 * so we know that it will be selected next.
4065 BUG_ON(!cfq_cfqq_on_rr(cfqq));
4067 cfq_service_tree_add(cfqd, cfqq, 1);
4069 cfqq->slice_end = 0;
4070 cfq_mark_cfqq_slice_new(cfqq);
4074 * Called when a new fs request (rq) is added (to cfqq). Check if there's
4075 * something we should do about it
4078 cfq_rq_enqueued(struct cfq_data *cfqd, struct cfq_queue *cfqq,
4081 struct cfq_io_cq *cic = RQ_CIC(rq);
4084 if (rq->cmd_flags & REQ_PRIO)
4085 cfqq->prio_pending++;
4087 cfq_update_io_thinktime(cfqd, cfqq, cic);
4088 cfq_update_io_seektime(cfqd, cfqq, rq);
4089 cfq_update_idle_window(cfqd, cfqq, cic);
4091 cfqq->last_request_pos = blk_rq_pos(rq) + blk_rq_sectors(rq);
4093 if (cfqq == cfqd->active_queue) {
4095 * Remember that we saw a request from this process, but
4096 * don't start queuing just yet. Otherwise we risk seeing lots
4097 * of tiny requests, because we disrupt the normal plugging
4098 * and merging. If the request is already larger than a single
4099 * page, let it rip immediately. For that case we assume that
4100 * merging is already done. Ditto for a busy system that
4101 * has other work pending, don't risk delaying until the
4102 * idle timer unplug to continue working.
4104 if (cfq_cfqq_wait_request(cfqq)) {
4105 if (blk_rq_bytes(rq) > PAGE_SIZE ||
4106 cfqd->busy_queues > 1) {
4107 cfq_del_timer(cfqd, cfqq);
4108 cfq_clear_cfqq_wait_request(cfqq);
4109 __blk_run_queue(cfqd->queue);
4111 cfqg_stats_update_idle_time(cfqq->cfqg);
4112 cfq_mark_cfqq_must_dispatch(cfqq);
4115 } else if (cfq_should_preempt(cfqd, cfqq, rq)) {
4117 * not the active queue - expire current slice if it is
4118 * idle and has expired it's mean thinktime or this new queue
4119 * has some old slice time left and is of higher priority or
4120 * this new queue is RT and the current one is BE
4122 cfq_preempt_queue(cfqd, cfqq);
4123 __blk_run_queue(cfqd->queue);
4127 static void cfq_insert_request(struct request_queue *q, struct request *rq)
4129 struct cfq_data *cfqd = q->elevator->elevator_data;
4130 struct cfq_queue *cfqq = RQ_CFQQ(rq);
4132 cfq_log_cfqq(cfqd, cfqq, "insert_request");
4133 cfq_init_prio_data(cfqq, RQ_CIC(rq));
4135 rq->fifo_time = ktime_get_ns() + cfqd->cfq_fifo_expire[rq_is_sync(rq)];
4136 list_add_tail(&rq->queuelist, &cfqq->fifo);
4138 cfqg_stats_update_io_add(RQ_CFQG(rq), cfqd->serving_group, req_op(rq),
4140 cfq_rq_enqueued(cfqd, cfqq, rq);
4144 * Update hw_tag based on peak queue depth over 50 samples under
4147 static void cfq_update_hw_tag(struct cfq_data *cfqd)
4149 struct cfq_queue *cfqq = cfqd->active_queue;
4151 if (cfqd->rq_in_driver > cfqd->hw_tag_est_depth)
4152 cfqd->hw_tag_est_depth = cfqd->rq_in_driver;
4154 if (cfqd->hw_tag == 1)
4157 if (cfqd->rq_queued <= CFQ_HW_QUEUE_MIN &&
4158 cfqd->rq_in_driver <= CFQ_HW_QUEUE_MIN)
4162 * If active queue hasn't enough requests and can idle, cfq might not
4163 * dispatch sufficient requests to hardware. Don't zero hw_tag in this
4166 if (cfqq && cfq_cfqq_idle_window(cfqq) &&
4167 cfqq->dispatched + cfqq->queued[0] + cfqq->queued[1] <
4168 CFQ_HW_QUEUE_MIN && cfqd->rq_in_driver < CFQ_HW_QUEUE_MIN)
4171 if (cfqd->hw_tag_samples++ < 50)
4174 if (cfqd->hw_tag_est_depth >= CFQ_HW_QUEUE_MIN)
4180 static bool cfq_should_wait_busy(struct cfq_data *cfqd, struct cfq_queue *cfqq)
4182 struct cfq_io_cq *cic = cfqd->active_cic;
4183 u64 now = ktime_get_ns();
4185 /* If the queue already has requests, don't wait */
4186 if (!RB_EMPTY_ROOT(&cfqq->sort_list))
4189 /* If there are other queues in the group, don't wait */
4190 if (cfqq->cfqg->nr_cfqq > 1)
4193 /* the only queue in the group, but think time is big */
4194 if (cfq_io_thinktime_big(cfqd, &cfqq->cfqg->ttime, true))
4197 if (cfq_slice_used(cfqq))
4200 /* if slice left is less than think time, wait busy */
4201 if (cic && sample_valid(cic->ttime.ttime_samples)
4202 && (cfqq->slice_end - now < cic->ttime.ttime_mean))
4206 * If think times is less than a jiffy than ttime_mean=0 and above
4207 * will not be true. It might happen that slice has not expired yet
4208 * but will expire soon (4-5 ns) during select_queue(). To cover the
4209 * case where think time is less than a jiffy, mark the queue wait
4210 * busy if only 1 jiffy is left in the slice.
4212 if (cfqq->slice_end - now <= jiffies_to_nsecs(1))
4218 static void cfq_completed_request(struct request_queue *q, struct request *rq)
4220 struct cfq_queue *cfqq = RQ_CFQQ(rq);
4221 struct cfq_data *cfqd = cfqq->cfqd;
4222 const int sync = rq_is_sync(rq);
4223 u64 now = ktime_get_ns();
4225 cfq_log_cfqq(cfqd, cfqq, "complete rqnoidle %d",
4226 !!(rq->cmd_flags & REQ_NOIDLE));
4228 cfq_update_hw_tag(cfqd);
4230 WARN_ON(!cfqd->rq_in_driver);
4231 WARN_ON(!cfqq->dispatched);
4232 cfqd->rq_in_driver--;
4234 (RQ_CFQG(rq))->dispatched--;
4235 cfqg_stats_update_completion(cfqq->cfqg, rq_start_time_ns(rq),
4236 rq_io_start_time_ns(rq), req_op(rq),
4239 cfqd->rq_in_flight[cfq_cfqq_sync(cfqq)]--;
4242 struct cfq_rb_root *st;
4244 RQ_CIC(rq)->ttime.last_end_request = now;
4246 if (cfq_cfqq_on_rr(cfqq))
4247 st = cfqq->service_tree;
4249 st = st_for(cfqq->cfqg, cfqq_class(cfqq),
4252 st->ttime.last_end_request = now;
4254 * We have to do this check in jiffies since start_time is in
4255 * jiffies and it is not trivial to convert to ns. If
4256 * cfq_fifo_expire[1] ever comes close to 1 jiffie, this test
4257 * will become problematic but so far we are fine (the default
4260 if (!time_after(rq->start_time +
4261 nsecs_to_jiffies(cfqd->cfq_fifo_expire[1]),
4263 cfqd->last_delayed_sync = now;
4266 #ifdef CONFIG_CFQ_GROUP_IOSCHED
4267 cfqq->cfqg->ttime.last_end_request = now;
4271 * If this is the active queue, check if it needs to be expired,
4272 * or if we want to idle in case it has no pending requests.
4274 if (cfqd->active_queue == cfqq) {
4275 const bool cfqq_empty = RB_EMPTY_ROOT(&cfqq->sort_list);
4277 if (cfq_cfqq_slice_new(cfqq)) {
4278 cfq_set_prio_slice(cfqd, cfqq);
4279 cfq_clear_cfqq_slice_new(cfqq);
4283 * Should we wait for next request to come in before we expire
4286 if (cfq_should_wait_busy(cfqd, cfqq)) {
4287 u64 extend_sl = cfqd->cfq_slice_idle;
4288 if (!cfqd->cfq_slice_idle)
4289 extend_sl = cfqd->cfq_group_idle;
4290 cfqq->slice_end = now + extend_sl;
4291 cfq_mark_cfqq_wait_busy(cfqq);
4292 cfq_log_cfqq(cfqd, cfqq, "will busy wait");
4296 * Idling is not enabled on:
4298 * - idle-priority queues
4300 * - queues with still some requests queued
4301 * - when there is a close cooperator
4303 if (cfq_slice_used(cfqq) || cfq_class_idle(cfqq))
4304 cfq_slice_expired(cfqd, 1);
4305 else if (sync && cfqq_empty &&
4306 !cfq_close_cooperator(cfqd, cfqq)) {
4307 cfq_arm_slice_timer(cfqd);
4311 if (!cfqd->rq_in_driver)
4312 cfq_schedule_dispatch(cfqd);
4315 static void cfqq_boost_on_prio(struct cfq_queue *cfqq, int op_flags)
4318 * If REQ_PRIO is set, boost class and prio level, if it's below
4319 * BE/NORM. If prio is not set, restore the potentially boosted
4322 if (!(op_flags & REQ_PRIO)) {
4323 cfqq->ioprio_class = cfqq->org_ioprio_class;
4324 cfqq->ioprio = cfqq->org_ioprio;
4326 if (cfq_class_idle(cfqq))
4327 cfqq->ioprio_class = IOPRIO_CLASS_BE;
4328 if (cfqq->ioprio > IOPRIO_NORM)
4329 cfqq->ioprio = IOPRIO_NORM;
4333 static inline int __cfq_may_queue(struct cfq_queue *cfqq)
4335 if (cfq_cfqq_wait_request(cfqq) && !cfq_cfqq_must_alloc_slice(cfqq)) {
4336 cfq_mark_cfqq_must_alloc_slice(cfqq);
4337 return ELV_MQUEUE_MUST;
4340 return ELV_MQUEUE_MAY;
4343 static int cfq_may_queue(struct request_queue *q, int op, int op_flags)
4345 struct cfq_data *cfqd = q->elevator->elevator_data;
4346 struct task_struct *tsk = current;
4347 struct cfq_io_cq *cic;
4348 struct cfq_queue *cfqq;
4351 * don't force setup of a queue from here, as a call to may_queue
4352 * does not necessarily imply that a request actually will be queued.
4353 * so just lookup a possibly existing queue, or return 'may queue'
4356 cic = cfq_cic_lookup(cfqd, tsk->io_context);
4358 return ELV_MQUEUE_MAY;
4360 cfqq = cic_to_cfqq(cic, rw_is_sync(op, op_flags));
4362 cfq_init_prio_data(cfqq, cic);
4363 cfqq_boost_on_prio(cfqq, op_flags);
4365 return __cfq_may_queue(cfqq);
4368 return ELV_MQUEUE_MAY;
4372 * queue lock held here
4374 static void cfq_put_request(struct request *rq)
4376 struct cfq_queue *cfqq = RQ_CFQQ(rq);
4379 const int rw = rq_data_dir(rq);
4381 BUG_ON(!cfqq->allocated[rw]);
4382 cfqq->allocated[rw]--;
4384 /* Put down rq reference on cfqg */
4385 cfqg_put(RQ_CFQG(rq));
4386 rq->elv.priv[0] = NULL;
4387 rq->elv.priv[1] = NULL;
4389 cfq_put_queue(cfqq);
4393 static struct cfq_queue *
4394 cfq_merge_cfqqs(struct cfq_data *cfqd, struct cfq_io_cq *cic,
4395 struct cfq_queue *cfqq)
4397 cfq_log_cfqq(cfqd, cfqq, "merging with queue %p", cfqq->new_cfqq);
4398 cic_set_cfqq(cic, cfqq->new_cfqq, 1);
4399 cfq_mark_cfqq_coop(cfqq->new_cfqq);
4400 cfq_put_queue(cfqq);
4401 return cic_to_cfqq(cic, 1);
4405 * Returns NULL if a new cfqq should be allocated, or the old cfqq if this
4406 * was the last process referring to said cfqq.
4408 static struct cfq_queue *
4409 split_cfqq(struct cfq_io_cq *cic, struct cfq_queue *cfqq)
4411 if (cfqq_process_refs(cfqq) == 1) {
4412 cfqq->pid = current->pid;
4413 cfq_clear_cfqq_coop(cfqq);
4414 cfq_clear_cfqq_split_coop(cfqq);
4418 cic_set_cfqq(cic, NULL, 1);
4420 cfq_put_cooperator(cfqq);
4422 cfq_put_queue(cfqq);
4426 * Allocate cfq data structures associated with this request.
4429 cfq_set_request(struct request_queue *q, struct request *rq, struct bio *bio,
4432 struct cfq_data *cfqd = q->elevator->elevator_data;
4433 struct cfq_io_cq *cic = icq_to_cic(rq->elv.icq);
4434 const int rw = rq_data_dir(rq);
4435 const bool is_sync = rq_is_sync(rq);
4436 struct cfq_queue *cfqq;
4438 spin_lock_irq(q->queue_lock);
4440 check_ioprio_changed(cic, bio);
4441 check_blkcg_changed(cic, bio);
4443 cfqq = cic_to_cfqq(cic, is_sync);
4444 if (!cfqq || cfqq == &cfqd->oom_cfqq) {
4446 cfq_put_queue(cfqq);
4447 cfqq = cfq_get_queue(cfqd, is_sync, cic, bio);
4448 cic_set_cfqq(cic, cfqq, is_sync);
4451 * If the queue was seeky for too long, break it apart.
4453 if (cfq_cfqq_coop(cfqq) && cfq_cfqq_split_coop(cfqq)) {
4454 cfq_log_cfqq(cfqd, cfqq, "breaking apart cfqq");
4455 cfqq = split_cfqq(cic, cfqq);
4461 * Check to see if this queue is scheduled to merge with
4462 * another, closely cooperating queue. The merging of
4463 * queues happens here as it must be done in process context.
4464 * The reference on new_cfqq was taken in merge_cfqqs.
4467 cfqq = cfq_merge_cfqqs(cfqd, cic, cfqq);
4470 cfqq->allocated[rw]++;
4473 cfqg_get(cfqq->cfqg);
4474 rq->elv.priv[0] = cfqq;
4475 rq->elv.priv[1] = cfqq->cfqg;
4476 spin_unlock_irq(q->queue_lock);
4480 static void cfq_kick_queue(struct work_struct *work)
4482 struct cfq_data *cfqd =
4483 container_of(work, struct cfq_data, unplug_work);
4484 struct request_queue *q = cfqd->queue;
4486 spin_lock_irq(q->queue_lock);
4487 __blk_run_queue(cfqd->queue);
4488 spin_unlock_irq(q->queue_lock);
4492 * Timer running if the active_queue is currently idling inside its time slice
4494 static enum hrtimer_restart cfq_idle_slice_timer(struct hrtimer *timer)
4496 struct cfq_data *cfqd = container_of(timer, struct cfq_data,
4498 struct cfq_queue *cfqq;
4499 unsigned long flags;
4502 cfq_log(cfqd, "idle timer fired");
4504 spin_lock_irqsave(cfqd->queue->queue_lock, flags);
4506 cfqq = cfqd->active_queue;
4511 * We saw a request before the queue expired, let it through
4513 if (cfq_cfqq_must_dispatch(cfqq))
4519 if (cfq_slice_used(cfqq))
4523 * only expire and reinvoke request handler, if there are
4524 * other queues with pending requests
4526 if (!cfqd->busy_queues)
4530 * not expired and it has a request pending, let it dispatch
4532 if (!RB_EMPTY_ROOT(&cfqq->sort_list))
4536 * Queue depth flag is reset only when the idle didn't succeed
4538 cfq_clear_cfqq_deep(cfqq);
4541 cfq_slice_expired(cfqd, timed_out);
4543 cfq_schedule_dispatch(cfqd);
4545 spin_unlock_irqrestore(cfqd->queue->queue_lock, flags);
4546 return HRTIMER_NORESTART;
4549 static void cfq_shutdown_timer_wq(struct cfq_data *cfqd)
4551 hrtimer_cancel(&cfqd->idle_slice_timer);
4552 cancel_work_sync(&cfqd->unplug_work);
4555 static void cfq_exit_queue(struct elevator_queue *e)
4557 struct cfq_data *cfqd = e->elevator_data;
4558 struct request_queue *q = cfqd->queue;
4560 cfq_shutdown_timer_wq(cfqd);
4562 spin_lock_irq(q->queue_lock);
4564 if (cfqd->active_queue)
4565 __cfq_slice_expired(cfqd, cfqd->active_queue, 0);
4567 spin_unlock_irq(q->queue_lock);
4569 cfq_shutdown_timer_wq(cfqd);
4571 #ifdef CONFIG_CFQ_GROUP_IOSCHED
4572 blkcg_deactivate_policy(q, &blkcg_policy_cfq);
4574 kfree(cfqd->root_group);
4579 static int cfq_init_queue(struct request_queue *q, struct elevator_type *e)
4581 struct cfq_data *cfqd;
4582 struct blkcg_gq *blkg __maybe_unused;
4584 struct elevator_queue *eq;
4586 eq = elevator_alloc(q, e);
4590 cfqd = kzalloc_node(sizeof(*cfqd), GFP_KERNEL, q->node);
4592 kobject_put(&eq->kobj);
4595 eq->elevator_data = cfqd;
4598 spin_lock_irq(q->queue_lock);
4600 spin_unlock_irq(q->queue_lock);
4602 /* Init root service tree */
4603 cfqd->grp_service_tree = CFQ_RB_ROOT;
4605 /* Init root group and prefer root group over other groups by default */
4606 #ifdef CONFIG_CFQ_GROUP_IOSCHED
4607 ret = blkcg_activate_policy(q, &blkcg_policy_cfq);
4611 cfqd->root_group = blkg_to_cfqg(q->root_blkg);
4614 cfqd->root_group = kzalloc_node(sizeof(*cfqd->root_group),
4615 GFP_KERNEL, cfqd->queue->node);
4616 if (!cfqd->root_group)
4619 cfq_init_cfqg_base(cfqd->root_group);
4620 cfqd->root_group->weight = 2 * CFQ_WEIGHT_LEGACY_DFL;
4621 cfqd->root_group->leaf_weight = 2 * CFQ_WEIGHT_LEGACY_DFL;
4625 * Not strictly needed (since RB_ROOT just clears the node and we
4626 * zeroed cfqd on alloc), but better be safe in case someone decides
4627 * to add magic to the rb code
4629 for (i = 0; i < CFQ_PRIO_LISTS; i++)
4630 cfqd->prio_trees[i] = RB_ROOT;
4633 * Our fallback cfqq if cfq_get_queue() runs into OOM issues.
4634 * Grab a permanent reference to it, so that the normal code flow
4635 * will not attempt to free it. oom_cfqq is linked to root_group
4636 * but shouldn't hold a reference as it'll never be unlinked. Lose
4637 * the reference from linking right away.
4639 cfq_init_cfqq(cfqd, &cfqd->oom_cfqq, 1, 0);
4640 cfqd->oom_cfqq.ref++;
4642 spin_lock_irq(q->queue_lock);
4643 cfq_link_cfqq_cfqg(&cfqd->oom_cfqq, cfqd->root_group);
4644 cfqg_put(cfqd->root_group);
4645 spin_unlock_irq(q->queue_lock);
4647 hrtimer_init(&cfqd->idle_slice_timer, CLOCK_MONOTONIC,
4649 cfqd->idle_slice_timer.function = cfq_idle_slice_timer;
4651 INIT_WORK(&cfqd->unplug_work, cfq_kick_queue);
4653 cfqd->cfq_quantum = cfq_quantum;
4654 cfqd->cfq_fifo_expire[0] = cfq_fifo_expire[0];
4655 cfqd->cfq_fifo_expire[1] = cfq_fifo_expire[1];
4656 cfqd->cfq_back_max = cfq_back_max;
4657 cfqd->cfq_back_penalty = cfq_back_penalty;
4658 cfqd->cfq_slice[0] = cfq_slice_async;
4659 cfqd->cfq_slice[1] = cfq_slice_sync;
4660 cfqd->cfq_target_latency = cfq_target_latency;
4661 cfqd->cfq_slice_async_rq = cfq_slice_async_rq;
4662 cfqd->cfq_slice_idle = cfq_slice_idle;
4663 cfqd->cfq_group_idle = cfq_group_idle;
4664 cfqd->cfq_latency = 1;
4667 * we optimistically start assuming sync ops weren't delayed in last
4668 * second, in order to have larger depth for async operations.
4670 cfqd->last_delayed_sync = ktime_get_ns() - NSEC_PER_SEC;
4675 kobject_put(&eq->kobj);
4679 static void cfq_registered_queue(struct request_queue *q)
4681 struct elevator_queue *e = q->elevator;
4682 struct cfq_data *cfqd = e->elevator_data;
4685 * Default to IOPS mode with no idling for SSDs
4687 if (blk_queue_nonrot(q))
4688 cfqd->cfq_slice_idle = 0;
4692 * sysfs parts below -->
4695 cfq_var_show(unsigned int var, char *page)
4697 return sprintf(page, "%u\n", var);
4701 cfq_var_store(unsigned int *var, const char *page, size_t count)
4703 char *p = (char *) page;
4705 *var = simple_strtoul(p, &p, 10);
4709 #define SHOW_FUNCTION(__FUNC, __VAR, __CONV) \
4710 static ssize_t __FUNC(struct elevator_queue *e, char *page) \
4712 struct cfq_data *cfqd = e->elevator_data; \
4713 u64 __data = __VAR; \
4715 __data = div_u64(__data, NSEC_PER_MSEC); \
4716 return cfq_var_show(__data, (page)); \
4718 SHOW_FUNCTION(cfq_quantum_show, cfqd->cfq_quantum, 0);
4719 SHOW_FUNCTION(cfq_fifo_expire_sync_show, cfqd->cfq_fifo_expire[1], 1);
4720 SHOW_FUNCTION(cfq_fifo_expire_async_show, cfqd->cfq_fifo_expire[0], 1);
4721 SHOW_FUNCTION(cfq_back_seek_max_show, cfqd->cfq_back_max, 0);
4722 SHOW_FUNCTION(cfq_back_seek_penalty_show, cfqd->cfq_back_penalty, 0);
4723 SHOW_FUNCTION(cfq_slice_idle_show, cfqd->cfq_slice_idle, 1);
4724 SHOW_FUNCTION(cfq_group_idle_show, cfqd->cfq_group_idle, 1);
4725 SHOW_FUNCTION(cfq_slice_sync_show, cfqd->cfq_slice[1], 1);
4726 SHOW_FUNCTION(cfq_slice_async_show, cfqd->cfq_slice[0], 1);
4727 SHOW_FUNCTION(cfq_slice_async_rq_show, cfqd->cfq_slice_async_rq, 0);
4728 SHOW_FUNCTION(cfq_low_latency_show, cfqd->cfq_latency, 0);
4729 SHOW_FUNCTION(cfq_target_latency_show, cfqd->cfq_target_latency, 1);
4730 #undef SHOW_FUNCTION
4732 #define USEC_SHOW_FUNCTION(__FUNC, __VAR) \
4733 static ssize_t __FUNC(struct elevator_queue *e, char *page) \
4735 struct cfq_data *cfqd = e->elevator_data; \
4736 u64 __data = __VAR; \
4737 __data = div_u64(__data, NSEC_PER_USEC); \
4738 return cfq_var_show(__data, (page)); \
4740 USEC_SHOW_FUNCTION(cfq_slice_idle_us_show, cfqd->cfq_slice_idle);
4741 USEC_SHOW_FUNCTION(cfq_group_idle_us_show, cfqd->cfq_group_idle);
4742 USEC_SHOW_FUNCTION(cfq_slice_sync_us_show, cfqd->cfq_slice[1]);
4743 USEC_SHOW_FUNCTION(cfq_slice_async_us_show, cfqd->cfq_slice[0]);
4744 USEC_SHOW_FUNCTION(cfq_target_latency_us_show, cfqd->cfq_target_latency);
4745 #undef USEC_SHOW_FUNCTION
4747 #define STORE_FUNCTION(__FUNC, __PTR, MIN, MAX, __CONV) \
4748 static ssize_t __FUNC(struct elevator_queue *e, const char *page, size_t count) \
4750 struct cfq_data *cfqd = e->elevator_data; \
4751 unsigned int __data; \
4752 int ret = cfq_var_store(&__data, (page), count); \
4753 if (__data < (MIN)) \
4755 else if (__data > (MAX)) \
4758 *(__PTR) = (u64)__data * NSEC_PER_MSEC; \
4760 *(__PTR) = __data; \
4763 STORE_FUNCTION(cfq_quantum_store, &cfqd->cfq_quantum, 1, UINT_MAX, 0);
4764 STORE_FUNCTION(cfq_fifo_expire_sync_store, &cfqd->cfq_fifo_expire[1], 1,
4766 STORE_FUNCTION(cfq_fifo_expire_async_store, &cfqd->cfq_fifo_expire[0], 1,
4768 STORE_FUNCTION(cfq_back_seek_max_store, &cfqd->cfq_back_max, 0, UINT_MAX, 0);
4769 STORE_FUNCTION(cfq_back_seek_penalty_store, &cfqd->cfq_back_penalty, 1,
4771 STORE_FUNCTION(cfq_slice_idle_store, &cfqd->cfq_slice_idle, 0, UINT_MAX, 1);
4772 STORE_FUNCTION(cfq_group_idle_store, &cfqd->cfq_group_idle, 0, UINT_MAX, 1);
4773 STORE_FUNCTION(cfq_slice_sync_store, &cfqd->cfq_slice[1], 1, UINT_MAX, 1);
4774 STORE_FUNCTION(cfq_slice_async_store, &cfqd->cfq_slice[0], 1, UINT_MAX, 1);
4775 STORE_FUNCTION(cfq_slice_async_rq_store, &cfqd->cfq_slice_async_rq, 1,
4777 STORE_FUNCTION(cfq_low_latency_store, &cfqd->cfq_latency, 0, 1, 0);
4778 STORE_FUNCTION(cfq_target_latency_store, &cfqd->cfq_target_latency, 1, UINT_MAX, 1);
4779 #undef STORE_FUNCTION
4781 #define USEC_STORE_FUNCTION(__FUNC, __PTR, MIN, MAX) \
4782 static ssize_t __FUNC(struct elevator_queue *e, const char *page, size_t count) \
4784 struct cfq_data *cfqd = e->elevator_data; \
4785 unsigned int __data; \
4786 int ret = cfq_var_store(&__data, (page), count); \
4787 if (__data < (MIN)) \
4789 else if (__data > (MAX)) \
4791 *(__PTR) = (u64)__data * NSEC_PER_USEC; \
4794 USEC_STORE_FUNCTION(cfq_slice_idle_us_store, &cfqd->cfq_slice_idle, 0, UINT_MAX);
4795 USEC_STORE_FUNCTION(cfq_group_idle_us_store, &cfqd->cfq_group_idle, 0, UINT_MAX);
4796 USEC_STORE_FUNCTION(cfq_slice_sync_us_store, &cfqd->cfq_slice[1], 1, UINT_MAX);
4797 USEC_STORE_FUNCTION(cfq_slice_async_us_store, &cfqd->cfq_slice[0], 1, UINT_MAX);
4798 USEC_STORE_FUNCTION(cfq_target_latency_us_store, &cfqd->cfq_target_latency, 1, UINT_MAX);
4799 #undef USEC_STORE_FUNCTION
4801 #define CFQ_ATTR(name) \
4802 __ATTR(name, S_IRUGO|S_IWUSR, cfq_##name##_show, cfq_##name##_store)
4804 static struct elv_fs_entry cfq_attrs[] = {
4806 CFQ_ATTR(fifo_expire_sync),
4807 CFQ_ATTR(fifo_expire_async),
4808 CFQ_ATTR(back_seek_max),
4809 CFQ_ATTR(back_seek_penalty),
4810 CFQ_ATTR(slice_sync),
4811 CFQ_ATTR(slice_sync_us),
4812 CFQ_ATTR(slice_async),
4813 CFQ_ATTR(slice_async_us),
4814 CFQ_ATTR(slice_async_rq),
4815 CFQ_ATTR(slice_idle),
4816 CFQ_ATTR(slice_idle_us),
4817 CFQ_ATTR(group_idle),
4818 CFQ_ATTR(group_idle_us),
4819 CFQ_ATTR(low_latency),
4820 CFQ_ATTR(target_latency),
4821 CFQ_ATTR(target_latency_us),
4825 static struct elevator_type iosched_cfq = {
4827 .elevator_merge_fn = cfq_merge,
4828 .elevator_merged_fn = cfq_merged_request,
4829 .elevator_merge_req_fn = cfq_merged_requests,
4830 .elevator_allow_bio_merge_fn = cfq_allow_bio_merge,
4831 .elevator_allow_rq_merge_fn = cfq_allow_rq_merge,
4832 .elevator_bio_merged_fn = cfq_bio_merged,
4833 .elevator_dispatch_fn = cfq_dispatch_requests,
4834 .elevator_add_req_fn = cfq_insert_request,
4835 .elevator_activate_req_fn = cfq_activate_request,
4836 .elevator_deactivate_req_fn = cfq_deactivate_request,
4837 .elevator_completed_req_fn = cfq_completed_request,
4838 .elevator_former_req_fn = elv_rb_former_request,
4839 .elevator_latter_req_fn = elv_rb_latter_request,
4840 .elevator_init_icq_fn = cfq_init_icq,
4841 .elevator_exit_icq_fn = cfq_exit_icq,
4842 .elevator_set_req_fn = cfq_set_request,
4843 .elevator_put_req_fn = cfq_put_request,
4844 .elevator_may_queue_fn = cfq_may_queue,
4845 .elevator_init_fn = cfq_init_queue,
4846 .elevator_exit_fn = cfq_exit_queue,
4847 .elevator_registered_fn = cfq_registered_queue,
4849 .icq_size = sizeof(struct cfq_io_cq),
4850 .icq_align = __alignof__(struct cfq_io_cq),
4851 .elevator_attrs = cfq_attrs,
4852 .elevator_name = "cfq",
4853 .elevator_owner = THIS_MODULE,
4856 #ifdef CONFIG_CFQ_GROUP_IOSCHED
4857 static struct blkcg_policy blkcg_policy_cfq = {
4858 .dfl_cftypes = cfq_blkcg_files,
4859 .legacy_cftypes = cfq_blkcg_legacy_files,
4861 .cpd_alloc_fn = cfq_cpd_alloc,
4862 .cpd_init_fn = cfq_cpd_init,
4863 .cpd_free_fn = cfq_cpd_free,
4864 .cpd_bind_fn = cfq_cpd_bind,
4866 .pd_alloc_fn = cfq_pd_alloc,
4867 .pd_init_fn = cfq_pd_init,
4868 .pd_offline_fn = cfq_pd_offline,
4869 .pd_free_fn = cfq_pd_free,
4870 .pd_reset_stats_fn = cfq_pd_reset_stats,
4874 static int __init cfq_init(void)
4878 #ifdef CONFIG_CFQ_GROUP_IOSCHED
4879 ret = blkcg_policy_register(&blkcg_policy_cfq);
4887 cfq_pool = KMEM_CACHE(cfq_queue, 0);
4891 ret = elv_register(&iosched_cfq);
4898 kmem_cache_destroy(cfq_pool);
4900 #ifdef CONFIG_CFQ_GROUP_IOSCHED
4901 blkcg_policy_unregister(&blkcg_policy_cfq);
4906 static void __exit cfq_exit(void)
4908 #ifdef CONFIG_CFQ_GROUP_IOSCHED
4909 blkcg_policy_unregister(&blkcg_policy_cfq);
4911 elv_unregister(&iosched_cfq);
4912 kmem_cache_destroy(cfq_pool);
4915 module_init(cfq_init);
4916 module_exit(cfq_exit);
4918 MODULE_AUTHOR("Jens Axboe");
4919 MODULE_LICENSE("GPL");
4920 MODULE_DESCRIPTION("Completely Fair Queueing IO scheduler");