2 * Header file for the BFQ I/O scheduler: data structures and
3 * prototypes of interface functions among BFQ components.
5 * This program is free software; you can redistribute it and/or
6 * modify it under the terms of the GNU General Public License as
7 * published by the Free Software Foundation; either version 2 of the
8 * License, or (at your option) any later version.
10 * This program is distributed in the hope that it will be useful,
11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
13 * General Public License for more details.
18 #include <linux/blktrace_api.h>
19 #include <linux/hrtimer.h>
20 #include <linux/blk-cgroup.h>
22 #define BFQ_IOPRIO_CLASSES 3
23 #define BFQ_CL_IDLE_TIMEOUT (HZ/5)
25 #define BFQ_MIN_WEIGHT 1
26 #define BFQ_MAX_WEIGHT 1000
27 #define BFQ_WEIGHT_CONVERSION_COEFF 10
29 #define BFQ_DEFAULT_QUEUE_IOPRIO 4
31 #define BFQ_WEIGHT_LEGACY_DFL 100
32 #define BFQ_DEFAULT_GRP_IOPRIO 0
33 #define BFQ_DEFAULT_GRP_CLASS IOPRIO_CLASS_BE
36 * Soft real-time applications are extremely more latency sensitive
37 * than interactive ones. Over-raise the weight of the former to
38 * privilege them against the latter.
40 #define BFQ_SOFTRT_WEIGHT_FACTOR 100
45 * struct bfq_service_tree - per ioprio_class service tree.
47 * Each service tree represents a B-WF2Q+ scheduler on its own. Each
48 * ioprio_class has its own independent scheduler, and so its own
49 * bfq_service_tree. All the fields are protected by the queue lock
50 * of the containing bfqd.
52 struct bfq_service_tree {
53 /* tree for active entities (i.e., those backlogged) */
54 struct rb_root active;
55 /* tree for idle entities (i.e., not backlogged, with V < F_i)*/
58 /* idle entity with minimum F_i */
59 struct bfq_entity *first_idle;
60 /* idle entity with maximum F_i */
61 struct bfq_entity *last_idle;
63 /* scheduler virtual time */
65 /* scheduler weight sum; active and idle entities contribute to it */
70 * struct bfq_sched_data - multi-class scheduler.
72 * bfq_sched_data is the basic scheduler queue. It supports three
73 * ioprio_classes, and can be used either as a toplevel queue or as an
74 * intermediate queue in a hierarchical setup.
76 * The supported ioprio_classes are the same as in CFQ, in descending
77 * priority order, IOPRIO_CLASS_RT, IOPRIO_CLASS_BE, IOPRIO_CLASS_IDLE.
78 * Requests from higher priority queues are served before all the
79 * requests from lower priority queues; among requests of the same
80 * queue requests are served according to B-WF2Q+.
82 * The schedule is implemented by the service trees, plus the field
83 * @next_in_service, which points to the entity on the active trees
84 * that will be served next, if 1) no changes in the schedule occurs
85 * before the current in-service entity is expired, 2) the in-service
86 * queue becomes idle when it expires, and 3) if the entity pointed by
87 * in_service_entity is not a queue, then the in-service child entity
88 * of the entity pointed by in_service_entity becomes idle on
89 * expiration. This peculiar definition allows for the following
90 * optimization, not yet exploited: while a given entity is still in
91 * service, we already know which is the best candidate for next
92 * service among the other active entitities in the same parent
93 * entity. We can then quickly compare the timestamps of the
94 * in-service entity with those of such best candidate.
96 * All fields are protected by the lock of the containing bfqd.
98 struct bfq_sched_data {
99 /* entity in service */
100 struct bfq_entity *in_service_entity;
101 /* head-of-line entity (see comments above) */
102 struct bfq_entity *next_in_service;
103 /* array of service trees, one per ioprio_class */
104 struct bfq_service_tree service_tree[BFQ_IOPRIO_CLASSES];
105 /* last time CLASS_IDLE was served */
106 unsigned long bfq_class_idle_last_service;
111 * struct bfq_weight_counter - counter of the number of all active entities
112 * with a given weight.
114 struct bfq_weight_counter {
115 unsigned int weight; /* weight of the entities this counter refers to */
116 unsigned int num_active; /* nr of active entities with this weight */
118 * Weights tree member (see bfq_data's @queue_weights_tree and
119 * @group_weights_tree)
121 struct rb_node weights_node;
125 * struct bfq_entity - schedulable entity.
127 * A bfq_entity is used to represent either a bfq_queue (leaf node in the
128 * cgroup hierarchy) or a bfq_group into the upper level scheduler. Each
129 * entity belongs to the sched_data of the parent group in the cgroup
130 * hierarchy. Non-leaf entities have also their own sched_data, stored
133 * Each entity stores independently its priority values; this would
134 * allow different weights on different devices, but this
135 * functionality is not exported to userspace by now. Priorities and
136 * weights are updated lazily, first storing the new values into the
137 * new_* fields, then setting the @prio_changed flag. As soon as
138 * there is a transition in the entity state that allows the priority
139 * update to take place the effective and the requested priority
140 * values are synchronized.
142 * Unless cgroups are used, the weight value is calculated from the
143 * ioprio to export the same interface as CFQ. When dealing with
144 * ``well-behaved'' queues (i.e., queues that do not spend too much
145 * time to consume their budget and have true sequential behavior, and
146 * when there are no external factors breaking anticipation) the
147 * relative weights at each level of the cgroups hierarchy should be
148 * guaranteed. All the fields are protected by the queue lock of the
152 /* service_tree member */
153 struct rb_node rb_node;
154 /* pointer to the weight counter associated with this entity */
155 struct bfq_weight_counter *weight_counter;
158 * Flag, true if the entity is on a tree (either the active or
159 * the idle one of its service_tree) or is in service.
163 /* B-WF2Q+ start and finish timestamps [sectors/weight] */
166 /* tree the entity is enqueued into; %NULL if not on a tree */
167 struct rb_root *tree;
170 * minimum start time of the (active) subtree rooted at this
171 * entity; used for O(log N) lookups into active trees
175 /* amount of service received during the last service slot */
178 /* budget, used also to calculate F_i: F_i = S_i + @budget / @weight */
181 /* weight of the queue */
183 /* next weight if a change is in progress */
186 /* original weight, used to implement weight boosting */
189 /* parent entity, for hierarchical scheduling */
190 struct bfq_entity *parent;
193 * For non-leaf nodes in the hierarchy, the associated
194 * scheduler queue, %NULL on leaf nodes.
196 struct bfq_sched_data *my_sched_data;
197 /* the scheduler queue this entity belongs to */
198 struct bfq_sched_data *sched_data;
200 /* flag, set to request a weight, ioprio or ioprio_class change */
207 * struct bfq_ttime - per process thinktime stats.
210 /* completion time of the last request */
211 u64 last_end_request;
213 /* total process thinktime */
215 /* number of thinktime samples */
216 unsigned long ttime_samples;
217 /* average process thinktime */
222 * struct bfq_queue - leaf schedulable entity.
224 * A bfq_queue is a leaf request queue; it can be associated with an
225 * io_context or more, if it is async or shared between cooperating
226 * processes. @cgroup holds a reference to the cgroup, to be sure that it
227 * does not disappear while a bfqq still references it (mostly to avoid
228 * races between request issuing and task migration followed by cgroup
230 * All the fields are protected by the queue lock of the containing bfqd.
233 /* reference counter */
235 /* parent bfq_data */
236 struct bfq_data *bfqd;
238 /* current ioprio and ioprio class */
239 unsigned short ioprio, ioprio_class;
240 /* next ioprio and ioprio class if a change is in progress */
241 unsigned short new_ioprio, new_ioprio_class;
244 * Shared bfq_queue if queue is cooperating with one or more
247 struct bfq_queue *new_bfqq;
248 /* request-position tree member (see bfq_group's @rq_pos_tree) */
249 struct rb_node pos_node;
250 /* request-position tree root (see bfq_group's @rq_pos_tree) */
251 struct rb_root *pos_root;
253 /* sorted list of pending requests */
254 struct rb_root sort_list;
255 /* if fifo isn't expired, next request to serve */
256 struct request *next_rq;
257 /* number of sync and async requests queued */
259 /* number of requests currently allocated */
261 /* number of pending metadata requests */
263 /* fifo list of requests in sort_list */
264 struct list_head fifo;
266 /* entity representing this queue in the scheduler */
267 struct bfq_entity entity;
269 /* maximum budget allowed from the feedback mechanism */
271 /* budget expiration (in jiffies) */
272 unsigned long budget_timeout;
274 /* number of requests on the dispatch list or inside driver */
280 /* node for active/idle bfqq list inside parent bfqd */
281 struct list_head bfqq_list;
283 /* associated @bfq_ttime struct */
284 struct bfq_ttime ttime;
286 /* bit vector: a 1 for each seeky requests in history */
289 /* node for the device's burst list */
290 struct hlist_node burst_list_node;
292 /* position of the last request enqueued */
293 sector_t last_request_pos;
295 /* Number of consecutive pairs of request completion and
296 * arrival, such that the queue becomes idle after the
297 * completion, but the next request arrives within an idle
298 * time slice; used only if the queue's IO_bound flag has been
301 unsigned int requests_within_timer;
303 /* pid of the process owning the queue, used for logging purposes */
307 * Pointer to the bfq_io_cq owning the bfq_queue, set to %NULL
308 * if the queue is shared.
310 struct bfq_io_cq *bic;
312 /* current maximum weight-raising time for this queue */
313 unsigned long wr_cur_max_time;
315 * Minimum time instant such that, only if a new request is
316 * enqueued after this time instant in an idle @bfq_queue with
317 * no outstanding requests, then the task associated with the
318 * queue it is deemed as soft real-time (see the comments on
319 * the function bfq_bfqq_softrt_next_start())
321 unsigned long soft_rt_next_start;
323 * Start time of the current weight-raising period if
324 * the @bfq-queue is being weight-raised, otherwise
325 * finish time of the last weight-raising period.
327 unsigned long last_wr_start_finish;
328 /* factor by which the weight of this queue is multiplied */
329 unsigned int wr_coeff;
331 * Time of the last transition of the @bfq_queue from idle to
334 unsigned long last_idle_bklogged;
336 * Cumulative service received from the @bfq_queue since the
337 * last transition from idle to backlogged.
339 unsigned long service_from_backlogged;
341 * Cumulative service received from the @bfq_queue since its
342 * last transition to weight-raised state.
344 unsigned long service_from_wr;
347 * Value of wr start time when switching to soft rt
349 unsigned long wr_start_at_switch_to_srt;
351 unsigned long split_time; /* time of last split */
353 unsigned long first_IO_time; /* time of first I/O for this queue */
357 * struct bfq_io_cq - per (request_queue, io_context) structure.
360 /* associated io_cq structure */
361 struct io_cq icq; /* must be the first member */
362 /* array of two process queues, the sync and the async */
363 struct bfq_queue *bfqq[2];
364 /* per (request_queue, blkcg) ioprio */
366 #ifdef CONFIG_BFQ_GROUP_IOSCHED
367 uint64_t blkcg_serial_nr; /* the current blkcg serial */
370 * Snapshot of the has_short_time flag before merging; taken
371 * to remember its value while the queue is merged, so as to
372 * be able to restore it in case of split.
374 bool saved_has_short_ttime;
376 * Same purpose as the previous two fields for the I/O bound
377 * classification of a queue.
382 * Same purpose as the previous fields for the value of the
383 * field keeping the queue's belonging to a large burst
385 bool saved_in_large_burst;
387 * True if the queue belonged to a burst list before its merge
388 * with another cooperating queue.
390 bool was_in_burst_list;
393 * Similar to previous fields: save wr information.
395 unsigned long saved_wr_coeff;
396 unsigned long saved_last_wr_start_finish;
397 unsigned long saved_wr_start_at_switch_to_srt;
398 unsigned int saved_wr_cur_max_time;
399 struct bfq_ttime saved_ttime;
403 * struct bfq_data - per-device data structure.
405 * All the fields are protected by @lock.
408 /* device request queue */
409 struct request_queue *queue;
411 struct list_head dispatch;
413 /* root bfq_group for the device */
414 struct bfq_group *root_group;
417 * rbtree of weight counters of @bfq_queues, sorted by
418 * weight. Used to keep track of whether all @bfq_queues have
419 * the same weight. The tree contains one counter for each
420 * distinct weight associated to some active and not
421 * weight-raised @bfq_queue (see the comments to the functions
422 * bfq_weights_tree_[add|remove] for further details).
424 struct rb_root queue_weights_tree;
426 * rbtree of non-queue @bfq_entity weight counters, sorted by
427 * weight. Used to keep track of whether all @bfq_groups have
428 * the same weight. The tree contains one counter for each
429 * distinct weight associated to some active @bfq_group (see
430 * the comments to the functions bfq_weights_tree_[add|remove]
431 * for further details).
433 struct rb_root group_weights_tree;
436 * Number of bfq_queues containing requests (including the
437 * queue in service, even if it is idling).
440 /* number of weight-raised busy @bfq_queues */
442 /* number of queued requests */
444 /* number of requests dispatched and waiting for completion */
448 * Maximum number of requests in driver in the last
449 * @hw_tag_samples completed requests.
451 int max_rq_in_driver;
452 /* number of samples used to calculate hw_tag */
454 /* flag set to one if the driver is showing a queueing behavior */
457 /* number of budgets assigned */
458 int budgets_assigned;
461 * Timer set when idling (waiting) for the next request from
462 * the queue in service.
464 struct hrtimer idle_slice_timer;
466 /* bfq_queue in service */
467 struct bfq_queue *in_service_queue;
469 /* on-disk position of the last served request */
470 sector_t last_position;
472 /* position of the last served request for the in-service queue */
473 sector_t in_serv_last_pos;
475 /* time of last request completion (ns) */
478 /* time of first rq dispatch in current observation interval (ns) */
480 /* time of last rq dispatch in current observation interval (ns) */
483 /* beginning of the last budget */
484 ktime_t last_budget_start;
485 /* beginning of the last idle slice */
486 ktime_t last_idling_start;
488 /* number of samples in current observation interval */
489 int peak_rate_samples;
490 /* num of samples of seq dispatches in current observation interval */
491 u32 sequential_samples;
492 /* total num of sectors transferred in current observation interval */
493 u64 tot_sectors_dispatched;
494 /* max rq size seen during current observation interval (sectors) */
495 u32 last_rq_max_size;
496 /* time elapsed from first dispatch in current observ. interval (us) */
497 u64 delta_from_first;
499 * Current estimate of the device peak rate, measured in
500 * [(sectors/usec) / 2^BFQ_RATE_SHIFT]. The left-shift by
501 * BFQ_RATE_SHIFT is performed to increase precision in
502 * fixed-point calculations.
506 /* maximum budget allotted to a bfq_queue before rescheduling */
509 /* list of all the bfq_queues active on the device */
510 struct list_head active_list;
511 /* list of all the bfq_queues idle on the device */
512 struct list_head idle_list;
515 * Timeout for async/sync requests; when it fires, requests
516 * are served in fifo order.
518 u64 bfq_fifo_expire[2];
519 /* weight of backward seeks wrt forward ones */
520 unsigned int bfq_back_penalty;
521 /* maximum allowed backward seek */
522 unsigned int bfq_back_max;
523 /* maximum idling time */
526 /* user-configured max budget value (0 for auto-tuning) */
527 int bfq_user_max_budget;
529 * Timeout for bfq_queues to consume their budget; used to
530 * prevent seeky queues from imposing long latencies to
531 * sequential or quasi-sequential ones (this also implies that
532 * seeky queues cannot receive guarantees in the service
533 * domain; after a timeout they are charged for the time they
534 * have been in service, to preserve fairness among them, but
535 * without service-domain guarantees).
537 unsigned int bfq_timeout;
540 * Number of consecutive requests that must be issued within
541 * the idle time slice to set again idling to a queue which
542 * was marked as non-I/O-bound (see the definition of the
543 * IO_bound flag for further details).
545 unsigned int bfq_requests_within_timer;
548 * Force device idling whenever needed to provide accurate
549 * service guarantees, without caring about throughput
550 * issues. CAVEAT: this may even increase latencies, in case
551 * of useless idling for processes that did stop doing I/O.
553 bool strict_guarantees;
556 * Last time at which a queue entered the current burst of
557 * queues being activated shortly after each other; for more
558 * details about this and the following parameters related to
559 * a burst of activations, see the comments on the function
562 unsigned long last_ins_in_burst;
564 * Reference time interval used to decide whether a queue has
565 * been activated shortly after @last_ins_in_burst.
567 unsigned long bfq_burst_interval;
568 /* number of queues in the current burst of queue activations */
571 /* common parent entity for the queues in the burst */
572 struct bfq_entity *burst_parent_entity;
573 /* Maximum burst size above which the current queue-activation
574 * burst is deemed as 'large'.
576 unsigned long bfq_large_burst_thresh;
577 /* true if a large queue-activation burst is in progress */
580 * Head of the burst list (as for the above fields, more
581 * details in the comments on the function bfq_handle_burst).
583 struct hlist_head burst_list;
585 /* if set to true, low-latency heuristics are enabled */
588 * Maximum factor by which the weight of a weight-raised queue
591 unsigned int bfq_wr_coeff;
592 /* maximum duration of a weight-raising period (jiffies) */
593 unsigned int bfq_wr_max_time;
595 /* Maximum weight-raising duration for soft real-time processes */
596 unsigned int bfq_wr_rt_max_time;
598 * Minimum idle period after which weight-raising may be
599 * reactivated for a queue (in jiffies).
601 unsigned int bfq_wr_min_idle_time;
603 * Minimum period between request arrivals after which
604 * weight-raising may be reactivated for an already busy async
605 * queue (in jiffies).
607 unsigned long bfq_wr_min_inter_arr_async;
609 /* Max service-rate for a soft real-time queue, in sectors/sec */
610 unsigned int bfq_wr_max_softrt_rate;
612 * Cached value of the product ref_rate*ref_wr_duration, used
613 * for computing the maximum duration of weight raising
618 /* fallback dummy bfqq for extreme OOM conditions */
619 struct bfq_queue oom_bfqq;
624 * bic associated with the task issuing current bio for
625 * merging. This and the next field are used as a support to
626 * be able to perform the bic lookup, needed by bio-merge
627 * functions, before the scheduler lock is taken, and thus
628 * avoid taking the request-queue lock while the scheduler
629 * lock is being held.
631 struct bfq_io_cq *bio_bic;
632 /* bfqq associated with the task issuing current bio for merging */
633 struct bfq_queue *bio_bfqq;
636 * Depth limits used in bfq_limit_depth (see comments on the
639 unsigned int word_depths[2][2];
642 enum bfqq_state_flags {
643 BFQQF_just_created = 0, /* queue just allocated */
644 BFQQF_busy, /* has requests or is in service */
645 BFQQF_wait_request, /* waiting for a request */
646 BFQQF_non_blocking_wait_rq, /*
647 * waiting for a request
648 * without idling the device
650 BFQQF_fifo_expire, /* FIFO checked in this slice */
651 BFQQF_has_short_ttime, /* queue has a short think time */
652 BFQQF_sync, /* synchronous queue */
654 * bfqq has timed-out at least once
655 * having consumed at most 2/10 of
658 BFQQF_in_large_burst, /*
659 * bfqq activated in a large burst,
660 * see comments to bfq_handle_burst.
662 BFQQF_softrt_update, /*
663 * may need softrt-next-start
666 BFQQF_coop, /* bfqq is shared */
667 BFQQF_split_coop /* shared bfqq will be split */
670 #define BFQ_BFQQ_FNS(name) \
671 void bfq_mark_bfqq_##name(struct bfq_queue *bfqq); \
672 void bfq_clear_bfqq_##name(struct bfq_queue *bfqq); \
673 int bfq_bfqq_##name(const struct bfq_queue *bfqq);
675 BFQ_BFQQ_FNS(just_created);
677 BFQ_BFQQ_FNS(wait_request);
678 BFQ_BFQQ_FNS(non_blocking_wait_rq);
679 BFQ_BFQQ_FNS(fifo_expire);
680 BFQ_BFQQ_FNS(has_short_ttime);
682 BFQ_BFQQ_FNS(IO_bound);
683 BFQ_BFQQ_FNS(in_large_burst);
685 BFQ_BFQQ_FNS(split_coop);
686 BFQ_BFQQ_FNS(softrt_update);
689 /* Expiration reasons. */
690 enum bfqq_expiration {
691 BFQQE_TOO_IDLE = 0, /*
692 * queue has been idling for
695 BFQQE_BUDGET_TIMEOUT, /* budget took too long to be used */
696 BFQQE_BUDGET_EXHAUSTED, /* budget consumed */
697 BFQQE_NO_MORE_REQUESTS, /* the queue has no more requests */
698 BFQQE_PREEMPTED /* preemption in progress */
702 #if defined(CONFIG_BFQ_GROUP_IOSCHED) && defined(CONFIG_DEBUG_BLK_CGROUP)
703 /* number of ios merged */
704 struct blkg_rwstat merged;
705 /* total time spent on device in ns, may not be accurate w/ queueing */
706 struct blkg_rwstat service_time;
707 /* total time spent waiting in scheduler queue in ns */
708 struct blkg_rwstat wait_time;
709 /* number of IOs queued up */
710 struct blkg_rwstat queued;
711 /* total disk time and nr sectors dispatched by this group */
712 struct blkg_stat time;
713 /* sum of number of ios queued across all samples */
714 struct blkg_stat avg_queue_size_sum;
715 /* count of samples taken for average */
716 struct blkg_stat avg_queue_size_samples;
717 /* how many times this group has been removed from service tree */
718 struct blkg_stat dequeue;
719 /* total time spent waiting for it to be assigned a timeslice. */
720 struct blkg_stat group_wait_time;
721 /* time spent idling for this blkcg_gq */
722 struct blkg_stat idle_time;
723 /* total time with empty current active q with other requests queued */
724 struct blkg_stat empty_time;
725 /* fields after this shouldn't be cleared on stat reset */
726 u64 start_group_wait_time;
728 u64 start_empty_time;
730 #endif /* CONFIG_BFQ_GROUP_IOSCHED && CONFIG_DEBUG_BLK_CGROUP */
733 #ifdef CONFIG_BFQ_GROUP_IOSCHED
736 * struct bfq_group_data - per-blkcg storage for the blkio subsystem.
738 * @ps: @blkcg_policy_storage that this structure inherits
739 * @weight: weight of the bfq_group
741 struct bfq_group_data {
742 /* must be the first member */
743 struct blkcg_policy_data pd;
749 * struct bfq_group - per (device, cgroup) data structure.
750 * @entity: schedulable entity to insert into the parent group sched_data.
751 * @sched_data: own sched_data, to contain child entities (they may be
752 * both bfq_queues and bfq_groups).
753 * @bfqd: the bfq_data for the device this group acts upon.
754 * @async_bfqq: array of async queues for all the tasks belonging to
755 * the group, one queue per ioprio value per ioprio_class,
756 * except for the idle class that has only one queue.
757 * @async_idle_bfqq: async queue for the idle class (ioprio is ignored).
758 * @my_entity: pointer to @entity, %NULL for the toplevel group; used
759 * to avoid too many special cases during group creation/
761 * @stats: stats for this bfqg.
762 * @active_entities: number of active entities belonging to the group;
763 * unused for the root group. Used to know whether there
764 * are groups with more than one active @bfq_entity
765 * (see the comments to the function
766 * bfq_bfqq_may_idle()).
767 * @rq_pos_tree: rbtree sorted by next_request position, used when
768 * determining if two or more queues have interleaving
769 * requests (see bfq_find_close_cooperator()).
771 * Each (device, cgroup) pair has its own bfq_group, i.e., for each cgroup
772 * there is a set of bfq_groups, each one collecting the lower-level
773 * entities belonging to the group that are acting on the same device.
775 * Locking works as follows:
776 * o @bfqd is protected by the queue lock, RCU is used to access it
778 * o All the other fields are protected by the @bfqd queue lock.
781 /* must be the first member */
782 struct blkg_policy_data pd;
784 /* cached path for this blkg (see comments in bfq_bic_update_cgroup) */
787 /* reference counter (see comments in bfq_bic_update_cgroup) */
790 struct bfq_entity entity;
791 struct bfq_sched_data sched_data;
795 struct bfq_queue *async_bfqq[2][IOPRIO_BE_NR];
796 struct bfq_queue *async_idle_bfqq;
798 struct bfq_entity *my_entity;
802 struct rb_root rq_pos_tree;
804 struct bfqg_stats stats;
809 struct bfq_sched_data sched_data;
811 struct bfq_queue *async_bfqq[2][IOPRIO_BE_NR];
812 struct bfq_queue *async_idle_bfqq;
814 struct rb_root rq_pos_tree;
818 struct bfq_queue *bfq_entity_to_bfqq(struct bfq_entity *entity);
820 /* --------------- main algorithm interface ----------------- */
822 #define BFQ_SERVICE_TREE_INIT ((struct bfq_service_tree) \
823 { RB_ROOT, RB_ROOT, NULL, NULL, 0, 0 })
825 extern const int bfq_timeout;
827 struct bfq_queue *bic_to_bfqq(struct bfq_io_cq *bic, bool is_sync);
828 void bic_set_bfqq(struct bfq_io_cq *bic, struct bfq_queue *bfqq, bool is_sync);
829 struct bfq_data *bic_to_bfqd(struct bfq_io_cq *bic);
830 void bfq_pos_tree_add_move(struct bfq_data *bfqd, struct bfq_queue *bfqq);
831 void bfq_weights_tree_add(struct bfq_data *bfqd, struct bfq_entity *entity,
832 struct rb_root *root);
833 void __bfq_weights_tree_remove(struct bfq_data *bfqd,
834 struct bfq_entity *entity,
835 struct rb_root *root);
836 void bfq_weights_tree_remove(struct bfq_data *bfqd,
837 struct bfq_queue *bfqq);
838 void bfq_bfqq_expire(struct bfq_data *bfqd, struct bfq_queue *bfqq,
839 bool compensate, enum bfqq_expiration reason);
840 void bfq_put_queue(struct bfq_queue *bfqq);
841 void bfq_end_wr_async_queues(struct bfq_data *bfqd, struct bfq_group *bfqg);
842 void bfq_schedule_dispatch(struct bfq_data *bfqd);
843 void bfq_put_async_queues(struct bfq_data *bfqd, struct bfq_group *bfqg);
845 /* ------------ end of main algorithm interface -------------- */
847 /* ---------------- cgroups-support interface ---------------- */
849 void bfqg_stats_update_io_add(struct bfq_group *bfqg, struct bfq_queue *bfqq,
851 void bfqg_stats_update_io_remove(struct bfq_group *bfqg, unsigned int op);
852 void bfqg_stats_update_io_merged(struct bfq_group *bfqg, unsigned int op);
853 void bfqg_stats_update_completion(struct bfq_group *bfqg, u64 start_time_ns,
854 u64 io_start_time_ns, unsigned int op);
855 void bfqg_stats_update_dequeue(struct bfq_group *bfqg);
856 void bfqg_stats_set_start_empty_time(struct bfq_group *bfqg);
857 void bfqg_stats_update_idle_time(struct bfq_group *bfqg);
858 void bfqg_stats_set_start_idle_time(struct bfq_group *bfqg);
859 void bfqg_stats_update_avg_queue_size(struct bfq_group *bfqg);
860 void bfq_bfqq_move(struct bfq_data *bfqd, struct bfq_queue *bfqq,
861 struct bfq_group *bfqg);
863 void bfq_init_entity(struct bfq_entity *entity, struct bfq_group *bfqg);
864 void bfq_bic_update_cgroup(struct bfq_io_cq *bic, struct bio *bio);
865 void bfq_end_wr_async(struct bfq_data *bfqd);
866 struct bfq_group *bfq_find_set_group(struct bfq_data *bfqd,
867 struct blkcg *blkcg);
868 struct blkcg_gq *bfqg_to_blkg(struct bfq_group *bfqg);
869 struct bfq_group *bfqq_group(struct bfq_queue *bfqq);
870 struct bfq_group *bfq_create_group_hierarchy(struct bfq_data *bfqd, int node);
871 void bfqg_and_blkg_put(struct bfq_group *bfqg);
873 #ifdef CONFIG_BFQ_GROUP_IOSCHED
874 extern struct cftype bfq_blkcg_legacy_files[];
875 extern struct cftype bfq_blkg_files[];
876 extern struct blkcg_policy blkcg_policy_bfq;
879 /* ------------- end of cgroups-support interface ------------- */
881 /* - interface of the internal hierarchical B-WF2Q+ scheduler - */
883 #ifdef CONFIG_BFQ_GROUP_IOSCHED
884 /* both next loops stop at one of the child entities of the root group */
885 #define for_each_entity(entity) \
886 for (; entity ; entity = entity->parent)
889 * For each iteration, compute parent in advance, so as to be safe if
890 * entity is deallocated during the iteration. Such a deallocation may
891 * happen as a consequence of a bfq_put_queue that frees the bfq_queue
894 #define for_each_entity_safe(entity, parent) \
895 for (; entity && ({ parent = entity->parent; 1; }); entity = parent)
897 #else /* CONFIG_BFQ_GROUP_IOSCHED */
899 * Next two macros are fake loops when cgroups support is not
900 * enabled. I fact, in such a case, there is only one level to go up
901 * (to reach the root group).
903 #define for_each_entity(entity) \
904 for (; entity ; entity = NULL)
906 #define for_each_entity_safe(entity, parent) \
907 for (parent = NULL; entity ; entity = parent)
908 #endif /* CONFIG_BFQ_GROUP_IOSCHED */
910 struct bfq_group *bfq_bfqq_to_bfqg(struct bfq_queue *bfqq);
911 struct bfq_queue *bfq_entity_to_bfqq(struct bfq_entity *entity);
912 struct bfq_service_tree *bfq_entity_service_tree(struct bfq_entity *entity);
913 struct bfq_entity *bfq_entity_of(struct rb_node *node);
914 unsigned short bfq_ioprio_to_weight(int ioprio);
915 void bfq_put_idle_entity(struct bfq_service_tree *st,
916 struct bfq_entity *entity);
917 struct bfq_service_tree *
918 __bfq_entity_update_weight_prio(struct bfq_service_tree *old_st,
919 struct bfq_entity *entity,
920 bool update_class_too);
921 void bfq_bfqq_served(struct bfq_queue *bfqq, int served);
922 void bfq_bfqq_charge_time(struct bfq_data *bfqd, struct bfq_queue *bfqq,
923 unsigned long time_ms);
924 bool __bfq_deactivate_entity(struct bfq_entity *entity,
925 bool ins_into_idle_tree);
926 bool next_queue_may_preempt(struct bfq_data *bfqd);
927 struct bfq_queue *bfq_get_next_queue(struct bfq_data *bfqd);
928 void __bfq_bfqd_reset_in_service(struct bfq_data *bfqd);
929 void bfq_deactivate_bfqq(struct bfq_data *bfqd, struct bfq_queue *bfqq,
930 bool ins_into_idle_tree, bool expiration);
931 void bfq_activate_bfqq(struct bfq_data *bfqd, struct bfq_queue *bfqq);
932 void bfq_requeue_bfqq(struct bfq_data *bfqd, struct bfq_queue *bfqq,
934 void bfq_del_bfqq_busy(struct bfq_data *bfqd, struct bfq_queue *bfqq,
936 void bfq_add_bfqq_busy(struct bfq_data *bfqd, struct bfq_queue *bfqq);
938 /* --------------- end of interface of B-WF2Q+ ---------------- */
940 /* Logging facilities. */
941 #ifdef CONFIG_BFQ_GROUP_IOSCHED
942 struct bfq_group *bfqq_group(struct bfq_queue *bfqq);
944 #define bfq_log_bfqq(bfqd, bfqq, fmt, args...) do { \
945 blk_add_cgroup_trace_msg((bfqd)->queue, \
946 bfqg_to_blkg(bfqq_group(bfqq))->blkcg, \
947 "bfq%d%c " fmt, (bfqq)->pid, \
948 bfq_bfqq_sync((bfqq)) ? 'S' : 'A', ##args); \
951 #define bfq_log_bfqg(bfqd, bfqg, fmt, args...) do { \
952 blk_add_cgroup_trace_msg((bfqd)->queue, \
953 bfqg_to_blkg(bfqg)->blkcg, fmt, ##args); \
956 #else /* CONFIG_BFQ_GROUP_IOSCHED */
958 #define bfq_log_bfqq(bfqd, bfqq, fmt, args...) \
959 blk_add_trace_msg((bfqd)->queue, "bfq%d%c " fmt, (bfqq)->pid, \
960 bfq_bfqq_sync((bfqq)) ? 'S' : 'A', \
962 #define bfq_log_bfqg(bfqd, bfqg, fmt, args...) do {} while (0)
964 #endif /* CONFIG_BFQ_GROUP_IOSCHED */
966 #define bfq_log(bfqd, fmt, args...) \
967 blk_add_trace_msg((bfqd)->queue, "bfq " fmt, ##args)