1 /* SPDX-License-Identifier: GPL-2.0-or-later */
2 /*
3  * Header file for the BFQ I/O scheduler: data structures and
4  * prototypes of interface functions among BFQ components.
5  */
6 #ifndef _BFQ_H
7 #define _BFQ_H
8 
9 #include <linux/blktrace_api.h>
10 #include <linux/hrtimer.h>
11 
12 #include "blk-cgroup-rwstat.h"
13 
14 #define BFQ_IOPRIO_CLASSES	3
15 #define BFQ_CL_IDLE_TIMEOUT	(HZ/5)
16 
17 #define BFQ_MIN_WEIGHT			1
18 #define BFQ_MAX_WEIGHT			1000
19 #define BFQ_WEIGHT_CONVERSION_COEFF	10
20 
21 #define BFQ_DEFAULT_QUEUE_IOPRIO	4
22 
23 #define BFQ_WEIGHT_LEGACY_DFL	100
24 #define BFQ_DEFAULT_GRP_IOPRIO	0
25 #define BFQ_DEFAULT_GRP_CLASS	IOPRIO_CLASS_BE
26 
27 #define MAX_BFQQ_NAME_LENGTH 16
28 
29 /*
30  * Soft real-time applications are extremely more latency sensitive
31  * than interactive ones. Over-raise the weight of the former to
32  * privilege them against the latter.
33  */
34 #define BFQ_SOFTRT_WEIGHT_FACTOR	100
35 
36 /*
37  * Maximum number of actuators supported. This constant is used simply
38  * to define the size of the static array that will contain
39  * per-actuator data. The current value is hopefully a good upper
40  * bound to the possible number of actuators of any actual drive.
41  */
42 #define BFQ_MAX_ACTUATORS 8
43 
44 struct bfq_entity;
45 
46 /**
47  * struct bfq_service_tree - per ioprio_class service tree.
48  *
49  * Each service tree represents a B-WF2Q+ scheduler on its own.  Each
50  * ioprio_class has its own independent scheduler, and so its own
51  * bfq_service_tree.  All the fields are protected by the queue lock
52  * of the containing bfqd.
53  */
54 struct bfq_service_tree {
55 	/* tree for active entities (i.e., those backlogged) */
56 	struct rb_root active;
57 	/* tree for idle entities (i.e., not backlogged, with V < F_i)*/
58 	struct rb_root idle;
59 
60 	/* idle entity with minimum F_i */
61 	struct bfq_entity *first_idle;
62 	/* idle entity with maximum F_i */
63 	struct bfq_entity *last_idle;
64 
65 	/* scheduler virtual time */
66 	u64 vtime;
67 	/* scheduler weight sum; active and idle entities contribute to it */
68 	unsigned long wsum;
69 };
70 
71 /**
72  * struct bfq_sched_data - multi-class scheduler.
73  *
74  * bfq_sched_data is the basic scheduler queue.  It supports three
75  * ioprio_classes, and can be used either as a toplevel queue or as an
76  * intermediate queue in a hierarchical setup.
77  *
78  * The supported ioprio_classes are the same as in CFQ, in descending
79  * priority order, IOPRIO_CLASS_RT, IOPRIO_CLASS_BE, IOPRIO_CLASS_IDLE.
80  * Requests from higher priority queues are served before all the
81  * requests from lower priority queues; among requests of the same
82  * queue requests are served according to B-WF2Q+.
83  *
84  * The schedule is implemented by the service trees, plus the field
85  * @next_in_service, which points to the entity on the active trees
86  * that will be served next, if 1) no changes in the schedule occurs
87  * before the current in-service entity is expired, 2) the in-service
88  * queue becomes idle when it expires, and 3) if the entity pointed by
89  * in_service_entity is not a queue, then the in-service child entity
90  * of the entity pointed by in_service_entity becomes idle on
91  * expiration. This peculiar definition allows for the following
92  * optimization, not yet exploited: while a given entity is still in
93  * service, we already know which is the best candidate for next
94  * service among the other active entities in the same parent
95  * entity. We can then quickly compare the timestamps of the
96  * in-service entity with those of such best candidate.
97  *
98  * All fields are protected by the lock of the containing bfqd.
99  */
100 struct bfq_sched_data {
101 	/* entity in service */
102 	struct bfq_entity *in_service_entity;
103 	/* head-of-line entity (see comments above) */
104 	struct bfq_entity *next_in_service;
105 	/* array of service trees, one per ioprio_class */
106 	struct bfq_service_tree service_tree[BFQ_IOPRIO_CLASSES];
107 	/* last time CLASS_IDLE was served */
108 	unsigned long bfq_class_idle_last_service;
109 
110 };
111 
112 /**
113  * struct bfq_weight_counter - counter of the number of all active queues
114  *                             with a given weight.
115  */
116 struct bfq_weight_counter {
117 	unsigned int weight; /* weight of the queues this counter refers to */
118 	unsigned int num_active; /* nr of active queues with this weight */
119 	/*
120 	 * Weights tree member (see bfq_data's @queue_weights_tree)
121 	 */
122 	struct rb_node weights_node;
123 };
124 
125 /**
126  * struct bfq_entity - schedulable entity.
127  *
128  * A bfq_entity is used to represent either a bfq_queue (leaf node in the
129  * cgroup hierarchy) or a bfq_group into the upper level scheduler.  Each
130  * entity belongs to the sched_data of the parent group in the cgroup
131  * hierarchy.  Non-leaf entities have also their own sched_data, stored
132  * in @my_sched_data.
133  *
134  * Each entity stores independently its priority values; this would
135  * allow different weights on different devices, but this
136  * functionality is not exported to userspace by now.  Priorities and
137  * weights are updated lazily, first storing the new values into the
138  * new_* fields, then setting the @prio_changed flag.  As soon as
139  * there is a transition in the entity state that allows the priority
140  * update to take place the effective and the requested priority
141  * values are synchronized.
142  *
143  * Unless cgroups are used, the weight value is calculated from the
144  * ioprio to export the same interface as CFQ.  When dealing with
145  * "well-behaved" queues (i.e., queues that do not spend too much
146  * time to consume their budget and have true sequential behavior, and
147  * when there are no external factors breaking anticipation) the
148  * relative weights at each level of the cgroups hierarchy should be
149  * guaranteed.  All the fields are protected by the queue lock of the
150  * containing bfqd.
151  */
152 struct bfq_entity {
153 	/* service_tree member */
154 	struct rb_node rb_node;
155 
156 	/*
157 	 * Flag, true if the entity is on a tree (either the active or
158 	 * the idle one of its service_tree) or is in service.
159 	 */
160 	bool on_st_or_in_serv;
161 
162 	/* B-WF2Q+ start and finish timestamps [sectors/weight] */
163 	u64 start, finish;
164 
165 	/* tree the entity is enqueued into; %NULL if not on a tree */
166 	struct rb_root *tree;
167 
168 	/*
169 	 * minimum start time of the (active) subtree rooted at this
170 	 * entity; used for O(log N) lookups into active trees
171 	 */
172 	u64 min_start;
173 
174 	/* amount of service received during the last service slot */
175 	int service;
176 
177 	/* budget, used also to calculate F_i: F_i = S_i + @budget / @weight */
178 	int budget;
179 
180 	/* Number of requests allocated in the subtree of this entity */
181 	int allocated;
182 
183 	/* device weight, if non-zero, it overrides the default weight of
184 	 * bfq_group_data */
185 	int dev_weight;
186 	/* weight of the queue */
187 	int weight;
188 	/* next weight if a change is in progress */
189 	int new_weight;
190 
191 	/* original weight, used to implement weight boosting */
192 	int orig_weight;
193 
194 	/* parent entity, for hierarchical scheduling */
195 	struct bfq_entity *parent;
196 
197 	/*
198 	 * For non-leaf nodes in the hierarchy, the associated
199 	 * scheduler queue, %NULL on leaf nodes.
200 	 */
201 	struct bfq_sched_data *my_sched_data;
202 	/* the scheduler queue this entity belongs to */
203 	struct bfq_sched_data *sched_data;
204 
205 	/* flag, set to request a weight, ioprio or ioprio_class change  */
206 	int prio_changed;
207 
208 #ifdef CONFIG_BFQ_GROUP_IOSCHED
209 	/* flag, set if the entity is counted in groups_with_pending_reqs */
210 	bool in_groups_with_pending_reqs;
211 #endif
212 
213 	/* last child queue of entity created (for non-leaf entities) */
214 	struct bfq_queue *last_bfqq_created;
215 };
216 
217 struct bfq_group;
218 
219 /**
220  * struct bfq_ttime - per process thinktime stats.
221  */
222 struct bfq_ttime {
223 	/* completion time of the last request */
224 	u64 last_end_request;
225 
226 	/* total process thinktime */
227 	u64 ttime_total;
228 	/* number of thinktime samples */
229 	unsigned long ttime_samples;
230 	/* average process thinktime */
231 	u64 ttime_mean;
232 };
233 
234 /**
235  * struct bfq_queue - leaf schedulable entity.
236  *
237  * A bfq_queue is a leaf request queue; it can be associated with an
238  * io_context or more, if it is async or shared between cooperating
239  * processes. Besides, it contains I/O requests for only one actuator
240  * (an io_context is associated with a different bfq_queue for each
241  * actuator it generates I/O for). @cgroup holds a reference to the
242  * cgroup, to be sure that it does not disappear while a bfqq still
243  * references it (mostly to avoid races between request issuing and
244  * task migration followed by cgroup destruction).  All the fields are
245  * protected by the queue lock of the containing bfqd.
246  */
247 struct bfq_queue {
248 	/* reference counter */
249 	int ref;
250 	/* counter of references from other queues for delayed stable merge */
251 	int stable_ref;
252 	/* parent bfq_data */
253 	struct bfq_data *bfqd;
254 
255 	/* current ioprio and ioprio class */
256 	unsigned short ioprio, ioprio_class;
257 	/* next ioprio and ioprio class if a change is in progress */
258 	unsigned short new_ioprio, new_ioprio_class;
259 
260 	/* last total-service-time sample, see bfq_update_inject_limit() */
261 	u64 last_serv_time_ns;
262 	/* limit for request injection */
263 	unsigned int inject_limit;
264 	/* last time the inject limit has been decreased, in jiffies */
265 	unsigned long decrease_time_jif;
266 
267 	/*
268 	 * Shared bfq_queue if queue is cooperating with one or more
269 	 * other queues.
270 	 */
271 	struct bfq_queue *new_bfqq;
272 	/* request-position tree member (see bfq_group's @rq_pos_tree) */
273 	struct rb_node pos_node;
274 	/* request-position tree root (see bfq_group's @rq_pos_tree) */
275 	struct rb_root *pos_root;
276 
277 	/* sorted list of pending requests */
278 	struct rb_root sort_list;
279 	/* if fifo isn't expired, next request to serve */
280 	struct request *next_rq;
281 	/* number of sync and async requests queued */
282 	int queued[2];
283 	/* number of pending metadata requests */
284 	int meta_pending;
285 	/* fifo list of requests in sort_list */
286 	struct list_head fifo;
287 
288 	/* entity representing this queue in the scheduler */
289 	struct bfq_entity entity;
290 
291 	/* pointer to the weight counter associated with this entity */
292 	struct bfq_weight_counter *weight_counter;
293 
294 	/* maximum budget allowed from the feedback mechanism */
295 	int max_budget;
296 	/* budget expiration (in jiffies) */
297 	unsigned long budget_timeout;
298 
299 	/* number of requests on the dispatch list or inside driver */
300 	int dispatched;
301 
302 	/* status flags */
303 	unsigned long flags;
304 
305 	/* node for active/idle bfqq list inside parent bfqd */
306 	struct list_head bfqq_list;
307 
308 	/* associated @bfq_ttime struct */
309 	struct bfq_ttime ttime;
310 
311 	/* when bfqq started to do I/O within the last observation window */
312 	u64 io_start_time;
313 	/* how long bfqq has remained empty during the last observ. window */
314 	u64 tot_idle_time;
315 
316 	/* bit vector: a 1 for each seeky requests in history */
317 	u32 seek_history;
318 
319 	/* node for the device's burst list */
320 	struct hlist_node burst_list_node;
321 
322 	/* position of the last request enqueued */
323 	sector_t last_request_pos;
324 
325 	/* Number of consecutive pairs of request completion and
326 	 * arrival, such that the queue becomes idle after the
327 	 * completion, but the next request arrives within an idle
328 	 * time slice; used only if the queue's IO_bound flag has been
329 	 * cleared.
330 	 */
331 	unsigned int requests_within_timer;
332 
333 	/* pid of the process owning the queue, used for logging purposes */
334 	pid_t pid;
335 
336 	/*
337 	 * Pointer to the bfq_io_cq owning the bfq_queue, set to %NULL
338 	 * if the queue is shared.
339 	 */
340 	struct bfq_io_cq *bic;
341 
342 	/* current maximum weight-raising time for this queue */
343 	unsigned long wr_cur_max_time;
344 	/*
345 	 * Minimum time instant such that, only if a new request is
346 	 * enqueued after this time instant in an idle @bfq_queue with
347 	 * no outstanding requests, then the task associated with the
348 	 * queue it is deemed as soft real-time (see the comments on
349 	 * the function bfq_bfqq_softrt_next_start())
350 	 */
351 	unsigned long soft_rt_next_start;
352 	/*
353 	 * Start time of the current weight-raising period if
354 	 * the @bfq-queue is being weight-raised, otherwise
355 	 * finish time of the last weight-raising period.
356 	 */
357 	unsigned long last_wr_start_finish;
358 	/* factor by which the weight of this queue is multiplied */
359 	unsigned int wr_coeff;
360 	/*
361 	 * Time of the last transition of the @bfq_queue from idle to
362 	 * backlogged.
363 	 */
364 	unsigned long last_idle_bklogged;
365 	/*
366 	 * Cumulative service received from the @bfq_queue since the
367 	 * last transition from idle to backlogged.
368 	 */
369 	unsigned long service_from_backlogged;
370 	/*
371 	 * Cumulative service received from the @bfq_queue since its
372 	 * last transition to weight-raised state.
373 	 */
374 	unsigned long service_from_wr;
375 
376 	/*
377 	 * Value of wr start time when switching to soft rt
378 	 */
379 	unsigned long wr_start_at_switch_to_srt;
380 
381 	unsigned long split_time; /* time of last split */
382 
383 	unsigned long first_IO_time; /* time of first I/O for this queue */
384 	unsigned long creation_time; /* when this queue is created */
385 
386 	/*
387 	 * Pointer to the waker queue for this queue, i.e., to the
388 	 * queue Q such that this queue happens to get new I/O right
389 	 * after some I/O request of Q is completed. For details, see
390 	 * the comments on the choice of the queue for injection in
391 	 * bfq_select_queue().
392 	 */
393 	struct bfq_queue *waker_bfqq;
394 	/* pointer to the curr. tentative waker queue, see bfq_check_waker() */
395 	struct bfq_queue *tentative_waker_bfqq;
396 	/* number of times the same tentative waker has been detected */
397 	unsigned int num_waker_detections;
398 	/* time when we started considering this waker */
399 	u64 waker_detection_started;
400 
401 	/* node for woken_list, see below */
402 	struct hlist_node woken_list_node;
403 	/*
404 	 * Head of the list of the woken queues for this queue, i.e.,
405 	 * of the list of the queues for which this queue is a waker
406 	 * queue. This list is used to reset the waker_bfqq pointer in
407 	 * the woken queues when this queue exits.
408 	 */
409 	struct hlist_head woken_list;
410 
411 	/* index of the actuator this queue is associated with */
412 	unsigned int actuator_idx;
413 };
414 
415 /**
416 * struct bfq_data - bfqq data unique and persistent for associated bfq_io_cq
417 */
418 struct bfq_iocq_bfqq_data {
419 	/*
420 	 * Snapshot of the has_short_time flag before merging; taken
421 	 * to remember its values while the queue is merged, so as to
422 	 * be able to restore it in case of split.
423 	 */
424 	bool saved_has_short_ttime;
425 	/*
426 	 * Same purpose as the previous two fields for the I/O bound
427 	 * classification of a queue.
428 	 */
429 	bool saved_IO_bound;
430 
431 	u64 saved_io_start_time;
432 	u64 saved_tot_idle_time;
433 
434 	/*
435 	 * Same purpose as the previous fields for the values of the
436 	 * field keeping the queue's belonging to a large burst
437 	 */
438 	bool saved_in_large_burst;
439 	/*
440 	 * True if the queue belonged to a burst list before its merge
441 	 * with another cooperating queue.
442 	 */
443 	bool was_in_burst_list;
444 
445 	/*
446 	 * Save the weight when a merge occurs, to be able
447 	 * to restore it in case of split. If the weight is not
448 	 * correctly resumed when the queue is recycled,
449 	 * then the weight of the recycled queue could differ
450 	 * from the weight of the original queue.
451 	 */
452 	unsigned int saved_weight;
453 
454 	/*
455 	 * Similar to previous fields: save wr information.
456 	 */
457 	unsigned long saved_wr_coeff;
458 	unsigned long saved_last_wr_start_finish;
459 	unsigned long saved_service_from_wr;
460 	unsigned long saved_wr_start_at_switch_to_srt;
461 	unsigned int saved_wr_cur_max_time;
462 	struct bfq_ttime saved_ttime;
463 
464 	/* Save also injection state */
465 	u64 saved_last_serv_time_ns;
466 	unsigned int saved_inject_limit;
467 	unsigned long saved_decrease_time_jif;
468 
469 	/* candidate queue for a stable merge (due to close creation time) */
470 	struct bfq_queue *stable_merge_bfqq;
471 
472 	bool stably_merged;	/* non splittable if true */
473 };
474 
475 /**
476  * struct bfq_io_cq - per (request_queue, io_context) structure.
477  */
478 struct bfq_io_cq {
479 	/* associated io_cq structure */
480 	struct io_cq icq; /* must be the first member */
481 	/*
482 	 * Matrix of associated process queues: first row for async
483 	 * queues, second row sync queues. Each row contains one
484 	 * column for each actuator. An I/O request generated by the
485 	 * process is inserted into the queue pointed by bfqq[i][j] if
486 	 * the request is to be served by the j-th actuator of the
487 	 * drive, where i==0 or i==1, depending on whether the request
488 	 * is async or sync. So there is a distinct queue for each
489 	 * actuator.
490 	 */
491 	struct bfq_queue *bfqq[2][BFQ_MAX_ACTUATORS];
492 	/* per (request_queue, blkcg) ioprio */
493 	int ioprio;
494 #ifdef CONFIG_BFQ_GROUP_IOSCHED
495 	uint64_t blkcg_serial_nr; /* the current blkcg serial */
496 #endif
497 
498 	/*
499 	 * Persistent data for associated synchronous process queues
500 	 * (one queue per actuator, see field bfqq above). In
501 	 * particular, each of these queues may undergo a merge.
502 	 */
503 	struct bfq_iocq_bfqq_data bfqq_data[BFQ_MAX_ACTUATORS];
504 
505 	unsigned int requests;	/* Number of requests this process has in flight */
506 };
507 
508 /**
509  * struct bfq_data - per-device data structure.
510  *
511  * All the fields are protected by @lock.
512  */
513 struct bfq_data {
514 	/* device request queue */
515 	struct request_queue *queue;
516 	/* dispatch queue */
517 	struct list_head dispatch;
518 
519 	/* root bfq_group for the device */
520 	struct bfq_group *root_group;
521 
522 	/*
523 	 * rbtree of weight counters of @bfq_queues, sorted by
524 	 * weight. Used to keep track of whether all @bfq_queues have
525 	 * the same weight. The tree contains one counter for each
526 	 * distinct weight associated to some active and not
527 	 * weight-raised @bfq_queue (see the comments to the functions
528 	 * bfq_weights_tree_[add|remove] for further details).
529 	 */
530 	struct rb_root_cached queue_weights_tree;
531 
532 #ifdef CONFIG_BFQ_GROUP_IOSCHED
533 	/*
534 	 * Number of groups with at least one process that
535 	 * has at least one request waiting for completion. Note that
536 	 * this accounts for also requests already dispatched, but not
537 	 * yet completed. Therefore this number of groups may differ
538 	 * (be larger) than the number of active groups, as a group is
539 	 * considered active only if its corresponding entity has
540 	 * queues with at least one request queued. This
541 	 * number is used to decide whether a scenario is symmetric.
542 	 * For a detailed explanation see comments on the computation
543 	 * of the variable asymmetric_scenario in the function
544 	 * bfq_better_to_idle().
545 	 *
546 	 * However, it is hard to compute this number exactly, for
547 	 * groups with multiple processes. Consider a group
548 	 * that is inactive, i.e., that has no process with
549 	 * pending I/O inside BFQ queues. Then suppose that
550 	 * num_groups_with_pending_reqs is still accounting for this
551 	 * group, because the group has processes with some
552 	 * I/O request still in flight. num_groups_with_pending_reqs
553 	 * should be decremented when the in-flight request of the
554 	 * last process is finally completed (assuming that
555 	 * nothing else has changed for the group in the meantime, in
556 	 * terms of composition of the group and active/inactive state of child
557 	 * groups and processes). To accomplish this, an additional
558 	 * pending-request counter must be added to entities, and must
559 	 * be updated correctly. To avoid this additional field and operations,
560 	 * we resort to the following tradeoff between simplicity and
561 	 * accuracy: for an inactive group that is still counted in
562 	 * num_groups_with_pending_reqs, we decrement
563 	 * num_groups_with_pending_reqs when the first
564 	 * process of the group remains with no request waiting for
565 	 * completion.
566 	 *
567 	 * Even this simpler decrement strategy requires a little
568 	 * carefulness: to avoid multiple decrements, we flag a group,
569 	 * more precisely an entity representing a group, as still
570 	 * counted in num_groups_with_pending_reqs when it becomes
571 	 * inactive. Then, when the first queue of the
572 	 * entity remains with no request waiting for completion,
573 	 * num_groups_with_pending_reqs is decremented, and this flag
574 	 * is reset. After this flag is reset for the entity,
575 	 * num_groups_with_pending_reqs won't be decremented any
576 	 * longer in case a new queue of the entity remains
577 	 * with no request waiting for completion.
578 	 */
579 	unsigned int num_groups_with_pending_reqs;
580 #endif
581 
582 	/*
583 	 * Per-class (RT, BE, IDLE) number of bfq_queues containing
584 	 * requests (including the queue in service, even if it is
585 	 * idling).
586 	 */
587 	unsigned int busy_queues[3];
588 	/* number of weight-raised busy @bfq_queues */
589 	int wr_busy_queues;
590 	/* number of queued requests */
591 	int queued;
592 	/* number of requests dispatched and waiting for completion */
593 	int tot_rq_in_driver;
594 	/*
595 	 * number of requests dispatched and waiting for completion
596 	 * for each actuator
597 	 */
598 	int rq_in_driver[BFQ_MAX_ACTUATORS];
599 
600 	/* true if the device is non rotational and performs queueing */
601 	bool nonrot_with_queueing;
602 
603 	/*
604 	 * Maximum number of requests in driver in the last
605 	 * @hw_tag_samples completed requests.
606 	 */
607 	int max_rq_in_driver;
608 	/* number of samples used to calculate hw_tag */
609 	int hw_tag_samples;
610 	/* flag set to one if the driver is showing a queueing behavior */
611 	int hw_tag;
612 
613 	/* number of budgets assigned */
614 	int budgets_assigned;
615 
616 	/*
617 	 * Timer set when idling (waiting) for the next request from
618 	 * the queue in service.
619 	 */
620 	struct hrtimer idle_slice_timer;
621 
622 	/* bfq_queue in service */
623 	struct bfq_queue *in_service_queue;
624 
625 	/* on-disk position of the last served request */
626 	sector_t last_position;
627 
628 	/* position of the last served request for the in-service queue */
629 	sector_t in_serv_last_pos;
630 
631 	/* time of last request completion (ns) */
632 	u64 last_completion;
633 
634 	/* bfqq owning the last completed rq */
635 	struct bfq_queue *last_completed_rq_bfqq;
636 
637 	/* last bfqq created, among those in the root group */
638 	struct bfq_queue *last_bfqq_created;
639 
640 	/* time of last transition from empty to non-empty (ns) */
641 	u64 last_empty_occupied_ns;
642 
643 	/*
644 	 * Flag set to activate the sampling of the total service time
645 	 * of a just-arrived first I/O request (see
646 	 * bfq_update_inject_limit()). This will cause the setting of
647 	 * waited_rq when the request is finally dispatched.
648 	 */
649 	bool wait_dispatch;
650 	/*
651 	 *  If set, then bfq_update_inject_limit() is invoked when
652 	 *  waited_rq is eventually completed.
653 	 */
654 	struct request *waited_rq;
655 	/*
656 	 * True if some request has been injected during the last service hole.
657 	 */
658 	bool rqs_injected;
659 
660 	/* time of first rq dispatch in current observation interval (ns) */
661 	u64 first_dispatch;
662 	/* time of last rq dispatch in current observation interval (ns) */
663 	u64 last_dispatch;
664 
665 	/* beginning of the last budget */
666 	ktime_t last_budget_start;
667 	/* beginning of the last idle slice */
668 	ktime_t last_idling_start;
669 	unsigned long last_idling_start_jiffies;
670 
671 	/* number of samples in current observation interval */
672 	int peak_rate_samples;
673 	/* num of samples of seq dispatches in current observation interval */
674 	u32 sequential_samples;
675 	/* total num of sectors transferred in current observation interval */
676 	u64 tot_sectors_dispatched;
677 	/* max rq size seen during current observation interval (sectors) */
678 	u32 last_rq_max_size;
679 	/* time elapsed from first dispatch in current observ. interval (us) */
680 	u64 delta_from_first;
681 	/*
682 	 * Current estimate of the device peak rate, measured in
683 	 * [(sectors/usec) / 2^BFQ_RATE_SHIFT]. The left-shift by
684 	 * BFQ_RATE_SHIFT is performed to increase precision in
685 	 * fixed-point calculations.
686 	 */
687 	u32 peak_rate;
688 
689 	/* maximum budget allotted to a bfq_queue before rescheduling */
690 	int bfq_max_budget;
691 
692 	/*
693 	 * List of all the bfq_queues active for a specific actuator
694 	 * on the device. Keeping active queues separate on a
695 	 * per-actuator basis helps implementing per-actuator
696 	 * injection more efficiently.
697 	 */
698 	struct list_head active_list[BFQ_MAX_ACTUATORS];
699 	/* list of all the bfq_queues idle on the device */
700 	struct list_head idle_list;
701 
702 	/*
703 	 * Timeout for async/sync requests; when it fires, requests
704 	 * are served in fifo order.
705 	 */
706 	u64 bfq_fifo_expire[2];
707 	/* weight of backward seeks wrt forward ones */
708 	unsigned int bfq_back_penalty;
709 	/* maximum allowed backward seek */
710 	unsigned int bfq_back_max;
711 	/* maximum idling time */
712 	u32 bfq_slice_idle;
713 
714 	/* user-configured max budget value (0 for auto-tuning) */
715 	int bfq_user_max_budget;
716 	/*
717 	 * Timeout for bfq_queues to consume their budget; used to
718 	 * prevent seeky queues from imposing long latencies to
719 	 * sequential or quasi-sequential ones (this also implies that
720 	 * seeky queues cannot receive guarantees in the service
721 	 * domain; after a timeout they are charged for the time they
722 	 * have been in service, to preserve fairness among them, but
723 	 * without service-domain guarantees).
724 	 */
725 	unsigned int bfq_timeout;
726 
727 	/*
728 	 * Force device idling whenever needed to provide accurate
729 	 * service guarantees, without caring about throughput
730 	 * issues. CAVEAT: this may even increase latencies, in case
731 	 * of useless idling for processes that did stop doing I/O.
732 	 */
733 	bool strict_guarantees;
734 
735 	/*
736 	 * Last time at which a queue entered the current burst of
737 	 * queues being activated shortly after each other; for more
738 	 * details about this and the following parameters related to
739 	 * a burst of activations, see the comments on the function
740 	 * bfq_handle_burst.
741 	 */
742 	unsigned long last_ins_in_burst;
743 	/*
744 	 * Reference time interval used to decide whether a queue has
745 	 * been activated shortly after @last_ins_in_burst.
746 	 */
747 	unsigned long bfq_burst_interval;
748 	/* number of queues in the current burst of queue activations */
749 	int burst_size;
750 
751 	/* common parent entity for the queues in the burst */
752 	struct bfq_entity *burst_parent_entity;
753 	/* Maximum burst size above which the current queue-activation
754 	 * burst is deemed as 'large'.
755 	 */
756 	unsigned long bfq_large_burst_thresh;
757 	/* true if a large queue-activation burst is in progress */
758 	bool large_burst;
759 	/*
760 	 * Head of the burst list (as for the above fields, more
761 	 * details in the comments on the function bfq_handle_burst).
762 	 */
763 	struct hlist_head burst_list;
764 
765 	/* if set to true, low-latency heuristics are enabled */
766 	bool low_latency;
767 	/*
768 	 * Maximum factor by which the weight of a weight-raised queue
769 	 * is multiplied.
770 	 */
771 	unsigned int bfq_wr_coeff;
772 
773 	/* Maximum weight-raising duration for soft real-time processes */
774 	unsigned int bfq_wr_rt_max_time;
775 	/*
776 	 * Minimum idle period after which weight-raising may be
777 	 * reactivated for a queue (in jiffies).
778 	 */
779 	unsigned int bfq_wr_min_idle_time;
780 	/*
781 	 * Minimum period between request arrivals after which
782 	 * weight-raising may be reactivated for an already busy async
783 	 * queue (in jiffies).
784 	 */
785 	unsigned long bfq_wr_min_inter_arr_async;
786 
787 	/* Max service-rate for a soft real-time queue, in sectors/sec */
788 	unsigned int bfq_wr_max_softrt_rate;
789 	/*
790 	 * Cached value of the product ref_rate*ref_wr_duration, used
791 	 * for computing the maximum duration of weight raising
792 	 * automatically.
793 	 */
794 	u64 rate_dur_prod;
795 
796 	/* fallback dummy bfqq for extreme OOM conditions */
797 	struct bfq_queue oom_bfqq;
798 
799 	spinlock_t lock;
800 
801 	/*
802 	 * bic associated with the task issuing current bio for
803 	 * merging. This and the next field are used as a support to
804 	 * be able to perform the bic lookup, needed by bio-merge
805 	 * functions, before the scheduler lock is taken, and thus
806 	 * avoid taking the request-queue lock while the scheduler
807 	 * lock is being held.
808 	 */
809 	struct bfq_io_cq *bio_bic;
810 	/* bfqq associated with the task issuing current bio for merging */
811 	struct bfq_queue *bio_bfqq;
812 
813 	/*
814 	 * Depth limits used in bfq_limit_depth (see comments on the
815 	 * function)
816 	 */
817 	unsigned int word_depths[2][2];
818 	unsigned int full_depth_shift;
819 
820 	/*
821 	 * Number of independent actuators. This is equal to 1 in
822 	 * case of single-actuator drives.
823 	 */
824 	unsigned int num_actuators;
825 	/*
826 	 * Disk independent access ranges for each actuator
827 	 * in this device.
828 	 */
829 	sector_t sector[BFQ_MAX_ACTUATORS];
830 	sector_t nr_sectors[BFQ_MAX_ACTUATORS];
831 	struct blk_independent_access_range ia_ranges[BFQ_MAX_ACTUATORS];
832 
833 	/*
834 	 * If the number of I/O requests queued in the device for a
835 	 * given actuator is below next threshold, then the actuator
836 	 * is deemed as underutilized. If this condition is found to
837 	 * hold for some actuator upon a dispatch, but (i) the
838 	 * in-service queue does not contain I/O for that actuator,
839 	 * while (ii) some other queue does contain I/O for that
840 	 * actuator, then the head I/O request of the latter queue is
841 	 * returned (injected), instead of the head request of the
842 	 * currently in-service queue.
843 	 *
844 	 * We set the threshold, empirically, to the minimum possible
845 	 * value for which an actuator is fully utilized, or close to
846 	 * be fully utilized. By doing so, injected I/O 'steals' as
847 	 * few drive-queue slots as possibile to the in-service
848 	 * queue. This reduces as much as possible the probability
849 	 * that the service of I/O from the in-service bfq_queue gets
850 	 * delayed because of slot exhaustion, i.e., because all the
851 	 * slots of the drive queue are filled with I/O injected from
852 	 * other queues (NCQ provides for 32 slots).
853 	 */
854 	unsigned int actuator_load_threshold;
855 };
856 
857 enum bfqq_state_flags {
858 	BFQQF_just_created = 0,	/* queue just allocated */
859 	BFQQF_busy,		/* has requests or is in service */
860 	BFQQF_wait_request,	/* waiting for a request */
861 	BFQQF_non_blocking_wait_rq, /*
862 				     * waiting for a request
863 				     * without idling the device
864 				     */
865 	BFQQF_fifo_expire,	/* FIFO checked in this slice */
866 	BFQQF_has_short_ttime,	/* queue has a short think time */
867 	BFQQF_sync,		/* synchronous queue */
868 	BFQQF_IO_bound,		/*
869 				 * bfqq has timed-out at least once
870 				 * having consumed at most 2/10 of
871 				 * its budget
872 				 */
873 	BFQQF_in_large_burst,	/*
874 				 * bfqq activated in a large burst,
875 				 * see comments to bfq_handle_burst.
876 				 */
877 	BFQQF_softrt_update,	/*
878 				 * may need softrt-next-start
879 				 * update
880 				 */
881 	BFQQF_coop,		/* bfqq is shared */
882 	BFQQF_split_coop,	/* shared bfqq will be split */
883 };
884 
885 #define BFQ_BFQQ_FNS(name)						\
886 void bfq_mark_bfqq_##name(struct bfq_queue *bfqq);			\
887 void bfq_clear_bfqq_##name(struct bfq_queue *bfqq);			\
888 int bfq_bfqq_##name(const struct bfq_queue *bfqq);
889 
890 BFQ_BFQQ_FNS(just_created);
891 BFQ_BFQQ_FNS(busy);
892 BFQ_BFQQ_FNS(wait_request);
893 BFQ_BFQQ_FNS(non_blocking_wait_rq);
894 BFQ_BFQQ_FNS(fifo_expire);
895 BFQ_BFQQ_FNS(has_short_ttime);
896 BFQ_BFQQ_FNS(sync);
897 BFQ_BFQQ_FNS(IO_bound);
898 BFQ_BFQQ_FNS(in_large_burst);
899 BFQ_BFQQ_FNS(coop);
900 BFQ_BFQQ_FNS(split_coop);
901 BFQ_BFQQ_FNS(softrt_update);
902 #undef BFQ_BFQQ_FNS
903 
904 /* Expiration reasons. */
905 enum bfqq_expiration {
906 	BFQQE_TOO_IDLE = 0,		/*
907 					 * queue has been idling for
908 					 * too long
909 					 */
910 	BFQQE_BUDGET_TIMEOUT,	/* budget took too long to be used */
911 	BFQQE_BUDGET_EXHAUSTED,	/* budget consumed */
912 	BFQQE_NO_MORE_REQUESTS,	/* the queue has no more requests */
913 	BFQQE_PREEMPTED		/* preemption in progress */
914 };
915 
916 struct bfq_stat {
917 	struct percpu_counter		cpu_cnt;
918 	atomic64_t			aux_cnt;
919 };
920 
921 struct bfqg_stats {
922 	/* basic stats */
923 	struct blkg_rwstat		bytes;
924 	struct blkg_rwstat		ios;
925 #ifdef CONFIG_BFQ_CGROUP_DEBUG
926 	/* number of ios merged */
927 	struct blkg_rwstat		merged;
928 	/* total time spent on device in ns, may not be accurate w/ queueing */
929 	struct blkg_rwstat		service_time;
930 	/* total time spent waiting in scheduler queue in ns */
931 	struct blkg_rwstat		wait_time;
932 	/* number of IOs queued up */
933 	struct blkg_rwstat		queued;
934 	/* total disk time and nr sectors dispatched by this group */
935 	struct bfq_stat		time;
936 	/* sum of number of ios queued across all samples */
937 	struct bfq_stat		avg_queue_size_sum;
938 	/* count of samples taken for average */
939 	struct bfq_stat		avg_queue_size_samples;
940 	/* how many times this group has been removed from service tree */
941 	struct bfq_stat		dequeue;
942 	/* total time spent waiting for it to be assigned a timeslice. */
943 	struct bfq_stat		group_wait_time;
944 	/* time spent idling for this blkcg_gq */
945 	struct bfq_stat		idle_time;
946 	/* total time with empty current active q with other requests queued */
947 	struct bfq_stat		empty_time;
948 	/* fields after this shouldn't be cleared on stat reset */
949 	u64				start_group_wait_time;
950 	u64				start_idle_time;
951 	u64				start_empty_time;
952 	uint16_t			flags;
953 #endif /* CONFIG_BFQ_CGROUP_DEBUG */
954 };
955 
956 #ifdef CONFIG_BFQ_GROUP_IOSCHED
957 
958 /*
959  * struct bfq_group_data - per-blkcg storage for the blkio subsystem.
960  *
961  * @ps: @blkcg_policy_storage that this structure inherits
962  * @weight: weight of the bfq_group
963  */
964 struct bfq_group_data {
965 	/* must be the first member */
966 	struct blkcg_policy_data pd;
967 
968 	unsigned int weight;
969 };
970 
971 /**
972  * struct bfq_group - per (device, cgroup) data structure.
973  * @entity: schedulable entity to insert into the parent group sched_data.
974  * @sched_data: own sched_data, to contain child entities (they may be
975  *              both bfq_queues and bfq_groups).
976  * @bfqd: the bfq_data for the device this group acts upon.
977  * @async_bfqq: array of async queues for all the tasks belonging to
978  *              the group, one queue per ioprio value per ioprio_class,
979  *              except for the idle class that has only one queue.
980  * @async_idle_bfqq: async queue for the idle class (ioprio is ignored).
981  * @my_entity: pointer to @entity, %NULL for the toplevel group; used
982  *             to avoid too many special cases during group creation/
983  *             migration.
984  * @stats: stats for this bfqg.
985  * @active_entities: number of active entities belonging to the group;
986  *                   unused for the root group. Used to know whether there
987  *                   are groups with more than one active @bfq_entity
988  *                   (see the comments to the function
989  *                   bfq_bfqq_may_idle()).
990  * @rq_pos_tree: rbtree sorted by next_request position, used when
991  *               determining if two or more queues have interleaving
992  *               requests (see bfq_find_close_cooperator()).
993  *
994  * Each (device, cgroup) pair has its own bfq_group, i.e., for each cgroup
995  * there is a set of bfq_groups, each one collecting the lower-level
996  * entities belonging to the group that are acting on the same device.
997  *
998  * Locking works as follows:
999  *    o @bfqd is protected by the queue lock, RCU is used to access it
1000  *      from the readers.
1001  *    o All the other fields are protected by the @bfqd queue lock.
1002  */
1003 struct bfq_group {
1004 	/* must be the first member */
1005 	struct blkg_policy_data pd;
1006 
1007 	/* cached path for this blkg (see comments in bfq_bic_update_cgroup) */
1008 	char blkg_path[128];
1009 
1010 	/* reference counter (see comments in bfq_bic_update_cgroup) */
1011 	refcount_t ref;
1012 
1013 	struct bfq_entity entity;
1014 	struct bfq_sched_data sched_data;
1015 
1016 	struct bfq_data *bfqd;
1017 
1018 	struct bfq_queue *async_bfqq[2][IOPRIO_NR_LEVELS][BFQ_MAX_ACTUATORS];
1019 	struct bfq_queue *async_idle_bfqq[BFQ_MAX_ACTUATORS];
1020 
1021 	struct bfq_entity *my_entity;
1022 
1023 	int active_entities;
1024 	int num_queues_with_pending_reqs;
1025 
1026 	struct rb_root rq_pos_tree;
1027 
1028 	struct bfqg_stats stats;
1029 };
1030 
1031 #else
1032 struct bfq_group {
1033 	struct bfq_entity entity;
1034 	struct bfq_sched_data sched_data;
1035 
1036 	struct bfq_queue *async_bfqq[2][IOPRIO_NR_LEVELS][BFQ_MAX_ACTUATORS];
1037 	struct bfq_queue *async_idle_bfqq[BFQ_MAX_ACTUATORS];
1038 
1039 	struct rb_root rq_pos_tree;
1040 };
1041 #endif
1042 
1043 /* --------------- main algorithm interface ----------------- */
1044 
1045 #define BFQ_SERVICE_TREE_INIT	((struct bfq_service_tree)		\
1046 				{ RB_ROOT, RB_ROOT, NULL, NULL, 0, 0 })
1047 
1048 extern const int bfq_timeout;
1049 
1050 struct bfq_queue *bic_to_bfqq(struct bfq_io_cq *bic, bool is_sync,
1051 				unsigned int actuator_idx);
1052 void bic_set_bfqq(struct bfq_io_cq *bic, struct bfq_queue *bfqq, bool is_sync,
1053 				unsigned int actuator_idx);
1054 struct bfq_data *bic_to_bfqd(struct bfq_io_cq *bic);
1055 void bfq_pos_tree_add_move(struct bfq_data *bfqd, struct bfq_queue *bfqq);
1056 void bfq_weights_tree_add(struct bfq_queue *bfqq);
1057 void bfq_weights_tree_remove(struct bfq_queue *bfqq);
1058 void bfq_bfqq_expire(struct bfq_data *bfqd, struct bfq_queue *bfqq,
1059 		     bool compensate, enum bfqq_expiration reason);
1060 void bfq_put_queue(struct bfq_queue *bfqq);
1061 void bfq_put_cooperator(struct bfq_queue *bfqq);
1062 void bfq_end_wr_async_queues(struct bfq_data *bfqd, struct bfq_group *bfqg);
1063 void bfq_release_process_ref(struct bfq_data *bfqd, struct bfq_queue *bfqq);
1064 void bfq_schedule_dispatch(struct bfq_data *bfqd);
1065 void bfq_put_async_queues(struct bfq_data *bfqd, struct bfq_group *bfqg);
1066 
1067 /* ------------ end of main algorithm interface -------------- */
1068 
1069 /* ---------------- cgroups-support interface ---------------- */
1070 
1071 void bfqg_stats_update_legacy_io(struct request_queue *q, struct request *rq);
1072 void bfqg_stats_update_io_remove(struct bfq_group *bfqg, blk_opf_t opf);
1073 void bfqg_stats_update_io_merged(struct bfq_group *bfqg, blk_opf_t opf);
1074 void bfqg_stats_update_completion(struct bfq_group *bfqg, u64 start_time_ns,
1075 				  u64 io_start_time_ns, blk_opf_t opf);
1076 void bfqg_stats_update_dequeue(struct bfq_group *bfqg);
1077 void bfqg_stats_set_start_idle_time(struct bfq_group *bfqg);
1078 void bfq_bfqq_move(struct bfq_data *bfqd, struct bfq_queue *bfqq,
1079 		   struct bfq_group *bfqg);
1080 
1081 #ifdef CONFIG_BFQ_CGROUP_DEBUG
1082 void bfqg_stats_update_io_add(struct bfq_group *bfqg, struct bfq_queue *bfqq,
1083 			      blk_opf_t opf);
1084 void bfqg_stats_set_start_empty_time(struct bfq_group *bfqg);
1085 void bfqg_stats_update_idle_time(struct bfq_group *bfqg);
1086 void bfqg_stats_update_avg_queue_size(struct bfq_group *bfqg);
1087 #endif
1088 
1089 void bfq_init_entity(struct bfq_entity *entity, struct bfq_group *bfqg);
1090 void bfq_bic_update_cgroup(struct bfq_io_cq *bic, struct bio *bio);
1091 void bfq_end_wr_async(struct bfq_data *bfqd);
1092 struct bfq_group *bfq_bio_bfqg(struct bfq_data *bfqd, struct bio *bio);
1093 struct blkcg_gq *bfqg_to_blkg(struct bfq_group *bfqg);
1094 struct bfq_group *bfqq_group(struct bfq_queue *bfqq);
1095 struct bfq_group *bfq_create_group_hierarchy(struct bfq_data *bfqd, int node);
1096 void bfqg_and_blkg_put(struct bfq_group *bfqg);
1097 
1098 #ifdef CONFIG_BFQ_GROUP_IOSCHED
1099 extern struct cftype bfq_blkcg_legacy_files[];
1100 extern struct cftype bfq_blkg_files[];
1101 extern struct blkcg_policy blkcg_policy_bfq;
1102 #endif
1103 
1104 /* ------------- end of cgroups-support interface ------------- */
1105 
1106 /* - interface of the internal hierarchical B-WF2Q+ scheduler - */
1107 
1108 #ifdef CONFIG_BFQ_GROUP_IOSCHED
1109 /* both next loops stop at one of the child entities of the root group */
1110 #define for_each_entity(entity)	\
1111 	for (; entity ; entity = entity->parent)
1112 
1113 /*
1114  * For each iteration, compute parent in advance, so as to be safe if
1115  * entity is deallocated during the iteration. Such a deallocation may
1116  * happen as a consequence of a bfq_put_queue that frees the bfq_queue
1117  * containing entity.
1118  */
1119 #define for_each_entity_safe(entity, parent) \
1120 	for (; entity && ({ parent = entity->parent; 1; }); entity = parent)
1121 
1122 #else /* CONFIG_BFQ_GROUP_IOSCHED */
1123 /*
1124  * Next two macros are fake loops when cgroups support is not
1125  * enabled. I fact, in such a case, there is only one level to go up
1126  * (to reach the root group).
1127  */
1128 #define for_each_entity(entity)	\
1129 	for (; entity ; entity = NULL)
1130 
1131 #define for_each_entity_safe(entity, parent) \
1132 	for (parent = NULL; entity ; entity = parent)
1133 #endif /* CONFIG_BFQ_GROUP_IOSCHED */
1134 
1135 struct bfq_queue *bfq_entity_to_bfqq(struct bfq_entity *entity);
1136 unsigned int bfq_tot_busy_queues(struct bfq_data *bfqd);
1137 struct bfq_service_tree *bfq_entity_service_tree(struct bfq_entity *entity);
1138 struct bfq_entity *bfq_entity_of(struct rb_node *node);
1139 unsigned short bfq_ioprio_to_weight(int ioprio);
1140 void bfq_put_idle_entity(struct bfq_service_tree *st,
1141 			 struct bfq_entity *entity);
1142 struct bfq_service_tree *
1143 __bfq_entity_update_weight_prio(struct bfq_service_tree *old_st,
1144 				struct bfq_entity *entity,
1145 				bool update_class_too);
1146 void bfq_bfqq_served(struct bfq_queue *bfqq, int served);
1147 void bfq_bfqq_charge_time(struct bfq_data *bfqd, struct bfq_queue *bfqq,
1148 			  unsigned long time_ms);
1149 bool __bfq_deactivate_entity(struct bfq_entity *entity,
1150 			     bool ins_into_idle_tree);
1151 bool next_queue_may_preempt(struct bfq_data *bfqd);
1152 struct bfq_queue *bfq_get_next_queue(struct bfq_data *bfqd);
1153 bool __bfq_bfqd_reset_in_service(struct bfq_data *bfqd);
1154 void bfq_deactivate_bfqq(struct bfq_data *bfqd, struct bfq_queue *bfqq,
1155 			 bool ins_into_idle_tree, bool expiration);
1156 void bfq_activate_bfqq(struct bfq_data *bfqd, struct bfq_queue *bfqq);
1157 void bfq_requeue_bfqq(struct bfq_data *bfqd, struct bfq_queue *bfqq,
1158 		      bool expiration);
1159 void bfq_del_bfqq_busy(struct bfq_queue *bfqq, bool expiration);
1160 void bfq_add_bfqq_busy(struct bfq_queue *bfqq);
1161 void bfq_add_bfqq_in_groups_with_pending_reqs(struct bfq_queue *bfqq);
1162 void bfq_del_bfqq_in_groups_with_pending_reqs(struct bfq_queue *bfqq);
1163 
1164 /* --------------- end of interface of B-WF2Q+ ---------------- */
1165 
1166 /* Logging facilities. */
bfq_bfqq_name(struct bfq_queue * bfqq,char * str,int len)1167 static inline void bfq_bfqq_name(struct bfq_queue *bfqq, char *str, int len)
1168 {
1169 	char type = bfq_bfqq_sync(bfqq) ? 'S' : 'A';
1170 
1171 	if (bfqq->pid != -1)
1172 		snprintf(str, len, "bfq%d%c", bfqq->pid, type);
1173 	else
1174 		snprintf(str, len, "bfqSHARED-%c", type);
1175 }
1176 
1177 #ifdef CONFIG_BFQ_GROUP_IOSCHED
1178 struct bfq_group *bfqq_group(struct bfq_queue *bfqq);
1179 
1180 #define bfq_log_bfqq(bfqd, bfqq, fmt, args...)	do {			\
1181 	char pid_str[MAX_BFQQ_NAME_LENGTH];				\
1182 	if (likely(!blk_trace_note_message_enabled((bfqd)->queue)))	\
1183 		break;							\
1184 	bfq_bfqq_name((bfqq), pid_str, MAX_BFQQ_NAME_LENGTH);		\
1185 	blk_add_cgroup_trace_msg((bfqd)->queue,				\
1186 			&bfqg_to_blkg(bfqq_group(bfqq))->blkcg->css,	\
1187 			"%s " fmt, pid_str, ##args);			\
1188 } while (0)
1189 
1190 #define bfq_log_bfqg(bfqd, bfqg, fmt, args...)	do {			\
1191 	blk_add_cgroup_trace_msg((bfqd)->queue,				\
1192 		&bfqg_to_blkg(bfqg)->blkcg->css, fmt, ##args);		\
1193 } while (0)
1194 
1195 #else /* CONFIG_BFQ_GROUP_IOSCHED */
1196 
1197 #define bfq_log_bfqq(bfqd, bfqq, fmt, args...) do {	\
1198 	char pid_str[MAX_BFQQ_NAME_LENGTH];				\
1199 	if (likely(!blk_trace_note_message_enabled((bfqd)->queue)))	\
1200 		break;							\
1201 	bfq_bfqq_name((bfqq), pid_str, MAX_BFQQ_NAME_LENGTH);		\
1202 	blk_add_trace_msg((bfqd)->queue, "%s " fmt, pid_str, ##args);	\
1203 } while (0)
1204 #define bfq_log_bfqg(bfqd, bfqg, fmt, args...)		do {} while (0)
1205 
1206 #endif /* CONFIG_BFQ_GROUP_IOSCHED */
1207 
1208 #define bfq_log(bfqd, fmt, args...) \
1209 	blk_add_trace_msg((bfqd)->queue, "bfq " fmt, ##args)
1210 
1211 #endif /* _BFQ_H */
1212