Loading...
Note: File does not exist in v6.13.7.
1/*
2 * CFQ, or complete fairness queueing, disk scheduler.
3 *
4 * Based on ideas from a previously unfinished io
5 * scheduler (round robin per-process disk scheduling) and Andrea Arcangeli.
6 *
7 * Copyright (C) 2003 Jens Axboe <axboe@kernel.dk>
8 */
9#include <linux/module.h>
10#include <linux/slab.h>
11#include <linux/blkdev.h>
12#include <linux/elevator.h>
13#include <linux/jiffies.h>
14#include <linux/rbtree.h>
15#include <linux/ioprio.h>
16#include <linux/blktrace_api.h>
17#include "blk.h"
18#include "blk-cgroup.h"
19
20/*
21 * tunables
22 */
23/* max queue in one round of service */
24static const int cfq_quantum = 8;
25static const int cfq_fifo_expire[2] = { HZ / 4, HZ / 8 };
26/* maximum backwards seek, in KiB */
27static const int cfq_back_max = 16 * 1024;
28/* penalty of a backwards seek */
29static const int cfq_back_penalty = 2;
30static const int cfq_slice_sync = HZ / 10;
31static int cfq_slice_async = HZ / 25;
32static const int cfq_slice_async_rq = 2;
33static int cfq_slice_idle = HZ / 125;
34static int cfq_group_idle = HZ / 125;
35static const int cfq_target_latency = HZ * 3/10; /* 300 ms */
36static const int cfq_hist_divisor = 4;
37
38/*
39 * offset from end of service tree
40 */
41#define CFQ_IDLE_DELAY (HZ / 5)
42
43/*
44 * below this threshold, we consider thinktime immediate
45 */
46#define CFQ_MIN_TT (2)
47
48#define CFQ_SLICE_SCALE (5)
49#define CFQ_HW_QUEUE_MIN (5)
50#define CFQ_SERVICE_SHIFT 12
51
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)
56
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])
60
61static struct kmem_cache *cfq_pool;
62
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)
66
67#define sample_valid(samples) ((samples) > 80)
68#define rb_entry_cfqg(node) rb_entry((node), struct cfq_group, rb_node)
69
70struct cfq_ttime {
71 unsigned long last_end_request;
72
73 unsigned long ttime_total;
74 unsigned long ttime_samples;
75 unsigned long ttime_mean;
76};
77
78/*
79 * Most of our rbtree usage is for sorting with min extraction, so
80 * if we cache the leftmost node we don't have to walk down the tree
81 * to find it. Idea borrowed from Ingo Molnars CFS scheduler. We should
82 * move this into the elevator for the rq sorting as well.
83 */
84struct cfq_rb_root {
85 struct rb_root rb;
86 struct rb_node *left;
87 unsigned count;
88 u64 min_vdisktime;
89 struct cfq_ttime ttime;
90};
91#define CFQ_RB_ROOT (struct cfq_rb_root) { .rb = RB_ROOT, \
92 .ttime = {.last_end_request = jiffies,},}
93
94/*
95 * Per process-grouping structure
96 */
97struct cfq_queue {
98 /* reference count */
99 int ref;
100 /* various state flags, see below */
101 unsigned int flags;
102 /* parent cfq_data */
103 struct cfq_data *cfqd;
104 /* service_tree member */
105 struct rb_node rb_node;
106 /* service_tree key */
107 unsigned long rb_key;
108 /* prio tree member */
109 struct rb_node p_node;
110 /* prio tree root we belong to, if any */
111 struct rb_root *p_root;
112 /* sorted list of pending requests */
113 struct rb_root sort_list;
114 /* if fifo isn't expired, next request to serve */
115 struct request *next_rq;
116 /* requests queued in sort_list */
117 int queued[2];
118 /* currently allocated requests */
119 int allocated[2];
120 /* fifo list of requests in sort_list */
121 struct list_head fifo;
122
123 /* time when queue got scheduled in to dispatch first request. */
124 unsigned long dispatch_start;
125 unsigned int allocated_slice;
126 unsigned int slice_dispatch;
127 /* time when first request from queue completed and slice started. */
128 unsigned long slice_start;
129 unsigned long slice_end;
130 long slice_resid;
131
132 /* pending priority requests */
133 int prio_pending;
134 /* number of requests that are on the dispatch list or inside driver */
135 int dispatched;
136
137 /* io prio of this group */
138 unsigned short ioprio, org_ioprio;
139 unsigned short ioprio_class;
140
141 pid_t pid;
142
143 u32 seek_history;
144 sector_t last_request_pos;
145
146 struct cfq_rb_root *service_tree;
147 struct cfq_queue *new_cfqq;
148 struct cfq_group *cfqg;
149 /* Number of sectors dispatched from queue in single dispatch round */
150 unsigned long nr_sectors;
151};
152
153/*
154 * First index in the service_trees.
155 * IDLE is handled separately, so it has negative index
156 */
157enum wl_class_t {
158 BE_WORKLOAD = 0,
159 RT_WORKLOAD = 1,
160 IDLE_WORKLOAD = 2,
161 CFQ_PRIO_NR,
162};
163
164/*
165 * Second index in the service_trees.
166 */
167enum wl_type_t {
168 ASYNC_WORKLOAD = 0,
169 SYNC_NOIDLE_WORKLOAD = 1,
170 SYNC_WORKLOAD = 2
171};
172
173struct cfqg_stats {
174#ifdef CONFIG_CFQ_GROUP_IOSCHED
175 /* total bytes transferred */
176 struct blkg_rwstat service_bytes;
177 /* total IOs serviced, post merge */
178 struct blkg_rwstat serviced;
179 /* number of ios merged */
180 struct blkg_rwstat merged;
181 /* total time spent on device in ns, may not be accurate w/ queueing */
182 struct blkg_rwstat service_time;
183 /* total time spent waiting in scheduler queue in ns */
184 struct blkg_rwstat wait_time;
185 /* number of IOs queued up */
186 struct blkg_rwstat queued;
187 /* total sectors transferred */
188 struct blkg_stat sectors;
189 /* total disk time and nr sectors dispatched by this group */
190 struct blkg_stat time;
191#ifdef CONFIG_DEBUG_BLK_CGROUP
192 /* time not charged to this cgroup */
193 struct blkg_stat unaccounted_time;
194 /* sum of number of ios queued across all samples */
195 struct blkg_stat avg_queue_size_sum;
196 /* count of samples taken for average */
197 struct blkg_stat avg_queue_size_samples;
198 /* how many times this group has been removed from service tree */
199 struct blkg_stat dequeue;
200 /* total time spent waiting for it to be assigned a timeslice. */
201 struct blkg_stat group_wait_time;
202 /* time spent idling for this blkcg_gq */
203 struct blkg_stat idle_time;
204 /* total time with empty current active q with other requests queued */
205 struct blkg_stat empty_time;
206 /* fields after this shouldn't be cleared on stat reset */
207 uint64_t start_group_wait_time;
208 uint64_t start_idle_time;
209 uint64_t start_empty_time;
210 uint16_t flags;
211#endif /* CONFIG_DEBUG_BLK_CGROUP */
212#endif /* CONFIG_CFQ_GROUP_IOSCHED */
213};
214
215/* This is per cgroup per device grouping structure */
216struct cfq_group {
217 /* must be the first member */
218 struct blkg_policy_data pd;
219
220 /* group service_tree member */
221 struct rb_node rb_node;
222
223 /* group service_tree key */
224 u64 vdisktime;
225
226 /*
227 * The number of active cfqgs and sum of their weights under this
228 * cfqg. This covers this cfqg's leaf_weight and all children's
229 * weights, but does not cover weights of further descendants.
230 *
231 * If a cfqg is on the service tree, it's active. An active cfqg
232 * also activates its parent and contributes to the children_weight
233 * of the parent.
234 */
235 int nr_active;
236 unsigned int children_weight;
237
238 /*
239 * vfraction is the fraction of vdisktime that the tasks in this
240 * cfqg are entitled to. This is determined by compounding the
241 * ratios walking up from this cfqg to the root.
242 *
243 * It is in fixed point w/ CFQ_SERVICE_SHIFT and the sum of all
244 * vfractions on a service tree is approximately 1. The sum may
245 * deviate a bit due to rounding errors and fluctuations caused by
246 * cfqgs entering and leaving the service tree.
247 */
248 unsigned int vfraction;
249
250 /*
251 * There are two weights - (internal) weight is the weight of this
252 * cfqg against the sibling cfqgs. leaf_weight is the wight of
253 * this cfqg against the child cfqgs. For the root cfqg, both
254 * weights are kept in sync for backward compatibility.
255 */
256 unsigned int weight;
257 unsigned int new_weight;
258 unsigned int dev_weight;
259
260 unsigned int leaf_weight;
261 unsigned int new_leaf_weight;
262 unsigned int dev_leaf_weight;
263
264 /* number of cfqq currently on this group */
265 int nr_cfqq;
266
267 /*
268 * Per group busy queues average. Useful for workload slice calc. We
269 * create the array for each prio class but at run time it is used
270 * only for RT and BE class and slot for IDLE class remains unused.
271 * This is primarily done to avoid confusion and a gcc warning.
272 */
273 unsigned int busy_queues_avg[CFQ_PRIO_NR];
274 /*
275 * rr lists of queues with requests. We maintain service trees for
276 * RT and BE classes. These trees are subdivided in subclasses
277 * of SYNC, SYNC_NOIDLE and ASYNC based on workload type. For IDLE
278 * class there is no subclassification and all the cfq queues go on
279 * a single tree service_tree_idle.
280 * Counts are embedded in the cfq_rb_root
281 */
282 struct cfq_rb_root service_trees[2][3];
283 struct cfq_rb_root service_tree_idle;
284
285 unsigned long saved_wl_slice;
286 enum wl_type_t saved_wl_type;
287 enum wl_class_t saved_wl_class;
288
289 /* number of requests that are on the dispatch list or inside driver */
290 int dispatched;
291 struct cfq_ttime ttime;
292 struct cfqg_stats stats; /* stats for this cfqg */
293 struct cfqg_stats dead_stats; /* stats pushed from dead children */
294};
295
296struct cfq_io_cq {
297 struct io_cq icq; /* must be the first member */
298 struct cfq_queue *cfqq[2];
299 struct cfq_ttime ttime;
300 int ioprio; /* the current ioprio */
301#ifdef CONFIG_CFQ_GROUP_IOSCHED
302 uint64_t blkcg_id; /* the current blkcg ID */
303#endif
304};
305
306/*
307 * Per block device queue structure
308 */
309struct cfq_data {
310 struct request_queue *queue;
311 /* Root service tree for cfq_groups */
312 struct cfq_rb_root grp_service_tree;
313 struct cfq_group *root_group;
314
315 /*
316 * The priority currently being served
317 */
318 enum wl_class_t serving_wl_class;
319 enum wl_type_t serving_wl_type;
320 unsigned long workload_expires;
321 struct cfq_group *serving_group;
322
323 /*
324 * Each priority tree is sorted by next_request position. These
325 * trees are used when determining if two or more queues are
326 * interleaving requests (see cfq_close_cooperator).
327 */
328 struct rb_root prio_trees[CFQ_PRIO_LISTS];
329
330 unsigned int busy_queues;
331 unsigned int busy_sync_queues;
332
333 int rq_in_driver;
334 int rq_in_flight[2];
335
336 /*
337 * queue-depth detection
338 */
339 int rq_queued;
340 int hw_tag;
341 /*
342 * hw_tag can be
343 * -1 => indeterminate, (cfq will behave as if NCQ is present, to allow better detection)
344 * 1 => NCQ is present (hw_tag_est_depth is the estimated max depth)
345 * 0 => no NCQ
346 */
347 int hw_tag_est_depth;
348 unsigned int hw_tag_samples;
349
350 /*
351 * idle window management
352 */
353 struct timer_list idle_slice_timer;
354 struct work_struct unplug_work;
355
356 struct cfq_queue *active_queue;
357 struct cfq_io_cq *active_cic;
358
359 /*
360 * async queue for each priority case
361 */
362 struct cfq_queue *async_cfqq[2][IOPRIO_BE_NR];
363 struct cfq_queue *async_idle_cfqq;
364
365 sector_t last_position;
366
367 /*
368 * tunables, see top of file
369 */
370 unsigned int cfq_quantum;
371 unsigned int cfq_fifo_expire[2];
372 unsigned int cfq_back_penalty;
373 unsigned int cfq_back_max;
374 unsigned int cfq_slice[2];
375 unsigned int cfq_slice_async_rq;
376 unsigned int cfq_slice_idle;
377 unsigned int cfq_group_idle;
378 unsigned int cfq_latency;
379 unsigned int cfq_target_latency;
380
381 /*
382 * Fallback dummy cfqq for extreme OOM conditions
383 */
384 struct cfq_queue oom_cfqq;
385
386 unsigned long last_delayed_sync;
387};
388
389static struct cfq_group *cfq_get_next_cfqg(struct cfq_data *cfqd);
390
391static struct cfq_rb_root *st_for(struct cfq_group *cfqg,
392 enum wl_class_t class,
393 enum wl_type_t type)
394{
395 if (!cfqg)
396 return NULL;
397
398 if (class == IDLE_WORKLOAD)
399 return &cfqg->service_tree_idle;
400
401 return &cfqg->service_trees[class][type];
402}
403
404enum cfqq_state_flags {
405 CFQ_CFQQ_FLAG_on_rr = 0, /* on round-robin busy list */
406 CFQ_CFQQ_FLAG_wait_request, /* waiting for a request */
407 CFQ_CFQQ_FLAG_must_dispatch, /* must be allowed a dispatch */
408 CFQ_CFQQ_FLAG_must_alloc_slice, /* per-slice must_alloc flag */
409 CFQ_CFQQ_FLAG_fifo_expire, /* FIFO checked in this slice */
410 CFQ_CFQQ_FLAG_idle_window, /* slice idling enabled */
411 CFQ_CFQQ_FLAG_prio_changed, /* task priority has changed */
412 CFQ_CFQQ_FLAG_slice_new, /* no requests dispatched in slice */
413 CFQ_CFQQ_FLAG_sync, /* synchronous queue */
414 CFQ_CFQQ_FLAG_coop, /* cfqq is shared */
415 CFQ_CFQQ_FLAG_split_coop, /* shared cfqq will be splitted */
416 CFQ_CFQQ_FLAG_deep, /* sync cfqq experienced large depth */
417 CFQ_CFQQ_FLAG_wait_busy, /* Waiting for next request */
418};
419
420#define CFQ_CFQQ_FNS(name) \
421static inline void cfq_mark_cfqq_##name(struct cfq_queue *cfqq) \
422{ \
423 (cfqq)->flags |= (1 << CFQ_CFQQ_FLAG_##name); \
424} \
425static inline void cfq_clear_cfqq_##name(struct cfq_queue *cfqq) \
426{ \
427 (cfqq)->flags &= ~(1 << CFQ_CFQQ_FLAG_##name); \
428} \
429static inline int cfq_cfqq_##name(const struct cfq_queue *cfqq) \
430{ \
431 return ((cfqq)->flags & (1 << CFQ_CFQQ_FLAG_##name)) != 0; \
432}
433
434CFQ_CFQQ_FNS(on_rr);
435CFQ_CFQQ_FNS(wait_request);
436CFQ_CFQQ_FNS(must_dispatch);
437CFQ_CFQQ_FNS(must_alloc_slice);
438CFQ_CFQQ_FNS(fifo_expire);
439CFQ_CFQQ_FNS(idle_window);
440CFQ_CFQQ_FNS(prio_changed);
441CFQ_CFQQ_FNS(slice_new);
442CFQ_CFQQ_FNS(sync);
443CFQ_CFQQ_FNS(coop);
444CFQ_CFQQ_FNS(split_coop);
445CFQ_CFQQ_FNS(deep);
446CFQ_CFQQ_FNS(wait_busy);
447#undef CFQ_CFQQ_FNS
448
449static inline struct cfq_group *pd_to_cfqg(struct blkg_policy_data *pd)
450{
451 return pd ? container_of(pd, struct cfq_group, pd) : NULL;
452}
453
454static inline struct blkcg_gq *cfqg_to_blkg(struct cfq_group *cfqg)
455{
456 return pd_to_blkg(&cfqg->pd);
457}
458
459#if defined(CONFIG_CFQ_GROUP_IOSCHED) && defined(CONFIG_DEBUG_BLK_CGROUP)
460
461/* cfqg stats flags */
462enum cfqg_stats_flags {
463 CFQG_stats_waiting = 0,
464 CFQG_stats_idling,
465 CFQG_stats_empty,
466};
467
468#define CFQG_FLAG_FNS(name) \
469static inline void cfqg_stats_mark_##name(struct cfqg_stats *stats) \
470{ \
471 stats->flags |= (1 << CFQG_stats_##name); \
472} \
473static inline void cfqg_stats_clear_##name(struct cfqg_stats *stats) \
474{ \
475 stats->flags &= ~(1 << CFQG_stats_##name); \
476} \
477static inline int cfqg_stats_##name(struct cfqg_stats *stats) \
478{ \
479 return (stats->flags & (1 << CFQG_stats_##name)) != 0; \
480} \
481
482CFQG_FLAG_FNS(waiting)
483CFQG_FLAG_FNS(idling)
484CFQG_FLAG_FNS(empty)
485#undef CFQG_FLAG_FNS
486
487/* This should be called with the queue_lock held. */
488static void cfqg_stats_update_group_wait_time(struct cfqg_stats *stats)
489{
490 unsigned long long now;
491
492 if (!cfqg_stats_waiting(stats))
493 return;
494
495 now = sched_clock();
496 if (time_after64(now, stats->start_group_wait_time))
497 blkg_stat_add(&stats->group_wait_time,
498 now - stats->start_group_wait_time);
499 cfqg_stats_clear_waiting(stats);
500}
501
502/* This should be called with the queue_lock held. */
503static void cfqg_stats_set_start_group_wait_time(struct cfq_group *cfqg,
504 struct cfq_group *curr_cfqg)
505{
506 struct cfqg_stats *stats = &cfqg->stats;
507
508 if (cfqg_stats_waiting(stats))
509 return;
510 if (cfqg == curr_cfqg)
511 return;
512 stats->start_group_wait_time = sched_clock();
513 cfqg_stats_mark_waiting(stats);
514}
515
516/* This should be called with the queue_lock held. */
517static void cfqg_stats_end_empty_time(struct cfqg_stats *stats)
518{
519 unsigned long long now;
520
521 if (!cfqg_stats_empty(stats))
522 return;
523
524 now = sched_clock();
525 if (time_after64(now, stats->start_empty_time))
526 blkg_stat_add(&stats->empty_time,
527 now - stats->start_empty_time);
528 cfqg_stats_clear_empty(stats);
529}
530
531static void cfqg_stats_update_dequeue(struct cfq_group *cfqg)
532{
533 blkg_stat_add(&cfqg->stats.dequeue, 1);
534}
535
536static void cfqg_stats_set_start_empty_time(struct cfq_group *cfqg)
537{
538 struct cfqg_stats *stats = &cfqg->stats;
539
540 if (blkg_rwstat_total(&stats->queued))
541 return;
542
543 /*
544 * group is already marked empty. This can happen if cfqq got new
545 * request in parent group and moved to this group while being added
546 * to service tree. Just ignore the event and move on.
547 */
548 if (cfqg_stats_empty(stats))
549 return;
550
551 stats->start_empty_time = sched_clock();
552 cfqg_stats_mark_empty(stats);
553}
554
555static void cfqg_stats_update_idle_time(struct cfq_group *cfqg)
556{
557 struct cfqg_stats *stats = &cfqg->stats;
558
559 if (cfqg_stats_idling(stats)) {
560 unsigned long long now = sched_clock();
561
562 if (time_after64(now, stats->start_idle_time))
563 blkg_stat_add(&stats->idle_time,
564 now - stats->start_idle_time);
565 cfqg_stats_clear_idling(stats);
566 }
567}
568
569static void cfqg_stats_set_start_idle_time(struct cfq_group *cfqg)
570{
571 struct cfqg_stats *stats = &cfqg->stats;
572
573 BUG_ON(cfqg_stats_idling(stats));
574
575 stats->start_idle_time = sched_clock();
576 cfqg_stats_mark_idling(stats);
577}
578
579static void cfqg_stats_update_avg_queue_size(struct cfq_group *cfqg)
580{
581 struct cfqg_stats *stats = &cfqg->stats;
582
583 blkg_stat_add(&stats->avg_queue_size_sum,
584 blkg_rwstat_total(&stats->queued));
585 blkg_stat_add(&stats->avg_queue_size_samples, 1);
586 cfqg_stats_update_group_wait_time(stats);
587}
588
589#else /* CONFIG_CFQ_GROUP_IOSCHED && CONFIG_DEBUG_BLK_CGROUP */
590
591static inline void cfqg_stats_set_start_group_wait_time(struct cfq_group *cfqg, struct cfq_group *curr_cfqg) { }
592static inline void cfqg_stats_end_empty_time(struct cfqg_stats *stats) { }
593static inline void cfqg_stats_update_dequeue(struct cfq_group *cfqg) { }
594static inline void cfqg_stats_set_start_empty_time(struct cfq_group *cfqg) { }
595static inline void cfqg_stats_update_idle_time(struct cfq_group *cfqg) { }
596static inline void cfqg_stats_set_start_idle_time(struct cfq_group *cfqg) { }
597static inline void cfqg_stats_update_avg_queue_size(struct cfq_group *cfqg) { }
598
599#endif /* CONFIG_CFQ_GROUP_IOSCHED && CONFIG_DEBUG_BLK_CGROUP */
600
601#ifdef CONFIG_CFQ_GROUP_IOSCHED
602
603static struct blkcg_policy blkcg_policy_cfq;
604
605static inline struct cfq_group *blkg_to_cfqg(struct blkcg_gq *blkg)
606{
607 return pd_to_cfqg(blkg_to_pd(blkg, &blkcg_policy_cfq));
608}
609
610static inline struct cfq_group *cfqg_parent(struct cfq_group *cfqg)
611{
612 struct blkcg_gq *pblkg = cfqg_to_blkg(cfqg)->parent;
613
614 return pblkg ? blkg_to_cfqg(pblkg) : NULL;
615}
616
617static inline void cfqg_get(struct cfq_group *cfqg)
618{
619 return blkg_get(cfqg_to_blkg(cfqg));
620}
621
622static inline void cfqg_put(struct cfq_group *cfqg)
623{
624 return blkg_put(cfqg_to_blkg(cfqg));
625}
626
627#define cfq_log_cfqq(cfqd, cfqq, fmt, args...) do { \
628 char __pbuf[128]; \
629 \
630 blkg_path(cfqg_to_blkg((cfqq)->cfqg), __pbuf, sizeof(__pbuf)); \
631 blk_add_trace_msg((cfqd)->queue, "cfq%d%c%c %s " fmt, (cfqq)->pid, \
632 cfq_cfqq_sync((cfqq)) ? 'S' : 'A', \
633 cfqq_type((cfqq)) == SYNC_NOIDLE_WORKLOAD ? 'N' : ' ',\
634 __pbuf, ##args); \
635} while (0)
636
637#define cfq_log_cfqg(cfqd, cfqg, fmt, args...) do { \
638 char __pbuf[128]; \
639 \
640 blkg_path(cfqg_to_blkg(cfqg), __pbuf, sizeof(__pbuf)); \
641 blk_add_trace_msg((cfqd)->queue, "%s " fmt, __pbuf, ##args); \
642} while (0)
643
644static inline void cfqg_stats_update_io_add(struct cfq_group *cfqg,
645 struct cfq_group *curr_cfqg, int rw)
646{
647 blkg_rwstat_add(&cfqg->stats.queued, rw, 1);
648 cfqg_stats_end_empty_time(&cfqg->stats);
649 cfqg_stats_set_start_group_wait_time(cfqg, curr_cfqg);
650}
651
652static inline void cfqg_stats_update_timeslice_used(struct cfq_group *cfqg,
653 unsigned long time, unsigned long unaccounted_time)
654{
655 blkg_stat_add(&cfqg->stats.time, time);
656#ifdef CONFIG_DEBUG_BLK_CGROUP
657 blkg_stat_add(&cfqg->stats.unaccounted_time, unaccounted_time);
658#endif
659}
660
661static inline void cfqg_stats_update_io_remove(struct cfq_group *cfqg, int rw)
662{
663 blkg_rwstat_add(&cfqg->stats.queued, rw, -1);
664}
665
666static inline void cfqg_stats_update_io_merged(struct cfq_group *cfqg, int rw)
667{
668 blkg_rwstat_add(&cfqg->stats.merged, rw, 1);
669}
670
671static inline void cfqg_stats_update_dispatch(struct cfq_group *cfqg,
672 uint64_t bytes, int rw)
673{
674 blkg_stat_add(&cfqg->stats.sectors, bytes >> 9);
675 blkg_rwstat_add(&cfqg->stats.serviced, rw, 1);
676 blkg_rwstat_add(&cfqg->stats.service_bytes, rw, bytes);
677}
678
679static inline void cfqg_stats_update_completion(struct cfq_group *cfqg,
680 uint64_t start_time, uint64_t io_start_time, int rw)
681{
682 struct cfqg_stats *stats = &cfqg->stats;
683 unsigned long long now = sched_clock();
684
685 if (time_after64(now, io_start_time))
686 blkg_rwstat_add(&stats->service_time, rw, now - io_start_time);
687 if (time_after64(io_start_time, start_time))
688 blkg_rwstat_add(&stats->wait_time, rw,
689 io_start_time - start_time);
690}
691
692/* @stats = 0 */
693static void cfqg_stats_reset(struct cfqg_stats *stats)
694{
695 /* queued stats shouldn't be cleared */
696 blkg_rwstat_reset(&stats->service_bytes);
697 blkg_rwstat_reset(&stats->serviced);
698 blkg_rwstat_reset(&stats->merged);
699 blkg_rwstat_reset(&stats->service_time);
700 blkg_rwstat_reset(&stats->wait_time);
701 blkg_stat_reset(&stats->time);
702#ifdef CONFIG_DEBUG_BLK_CGROUP
703 blkg_stat_reset(&stats->unaccounted_time);
704 blkg_stat_reset(&stats->avg_queue_size_sum);
705 blkg_stat_reset(&stats->avg_queue_size_samples);
706 blkg_stat_reset(&stats->dequeue);
707 blkg_stat_reset(&stats->group_wait_time);
708 blkg_stat_reset(&stats->idle_time);
709 blkg_stat_reset(&stats->empty_time);
710#endif
711}
712
713/* @to += @from */
714static void cfqg_stats_merge(struct cfqg_stats *to, struct cfqg_stats *from)
715{
716 /* queued stats shouldn't be cleared */
717 blkg_rwstat_merge(&to->service_bytes, &from->service_bytes);
718 blkg_rwstat_merge(&to->serviced, &from->serviced);
719 blkg_rwstat_merge(&to->merged, &from->merged);
720 blkg_rwstat_merge(&to->service_time, &from->service_time);
721 blkg_rwstat_merge(&to->wait_time, &from->wait_time);
722 blkg_stat_merge(&from->time, &from->time);
723#ifdef CONFIG_DEBUG_BLK_CGROUP
724 blkg_stat_merge(&to->unaccounted_time, &from->unaccounted_time);
725 blkg_stat_merge(&to->avg_queue_size_sum, &from->avg_queue_size_sum);
726 blkg_stat_merge(&to->avg_queue_size_samples, &from->avg_queue_size_samples);
727 blkg_stat_merge(&to->dequeue, &from->dequeue);
728 blkg_stat_merge(&to->group_wait_time, &from->group_wait_time);
729 blkg_stat_merge(&to->idle_time, &from->idle_time);
730 blkg_stat_merge(&to->empty_time, &from->empty_time);
731#endif
732}
733
734/*
735 * Transfer @cfqg's stats to its parent's dead_stats so that the ancestors'
736 * recursive stats can still account for the amount used by this cfqg after
737 * it's gone.
738 */
739static void cfqg_stats_xfer_dead(struct cfq_group *cfqg)
740{
741 struct cfq_group *parent = cfqg_parent(cfqg);
742
743 lockdep_assert_held(cfqg_to_blkg(cfqg)->q->queue_lock);
744
745 if (unlikely(!parent))
746 return;
747
748 cfqg_stats_merge(&parent->dead_stats, &cfqg->stats);
749 cfqg_stats_merge(&parent->dead_stats, &cfqg->dead_stats);
750 cfqg_stats_reset(&cfqg->stats);
751 cfqg_stats_reset(&cfqg->dead_stats);
752}
753
754#else /* CONFIG_CFQ_GROUP_IOSCHED */
755
756static inline struct cfq_group *cfqg_parent(struct cfq_group *cfqg) { return NULL; }
757static inline void cfqg_get(struct cfq_group *cfqg) { }
758static inline void cfqg_put(struct cfq_group *cfqg) { }
759
760#define cfq_log_cfqq(cfqd, cfqq, fmt, args...) \
761 blk_add_trace_msg((cfqd)->queue, "cfq%d%c%c " fmt, (cfqq)->pid, \
762 cfq_cfqq_sync((cfqq)) ? 'S' : 'A', \
763 cfqq_type((cfqq)) == SYNC_NOIDLE_WORKLOAD ? 'N' : ' ',\
764 ##args)
765#define cfq_log_cfqg(cfqd, cfqg, fmt, args...) do {} while (0)
766
767static inline void cfqg_stats_update_io_add(struct cfq_group *cfqg,
768 struct cfq_group *curr_cfqg, int rw) { }
769static inline void cfqg_stats_update_timeslice_used(struct cfq_group *cfqg,
770 unsigned long time, unsigned long unaccounted_time) { }
771static inline void cfqg_stats_update_io_remove(struct cfq_group *cfqg, int rw) { }
772static inline void cfqg_stats_update_io_merged(struct cfq_group *cfqg, int rw) { }
773static inline void cfqg_stats_update_dispatch(struct cfq_group *cfqg,
774 uint64_t bytes, int rw) { }
775static inline void cfqg_stats_update_completion(struct cfq_group *cfqg,
776 uint64_t start_time, uint64_t io_start_time, int rw) { }
777
778#endif /* CONFIG_CFQ_GROUP_IOSCHED */
779
780#define cfq_log(cfqd, fmt, args...) \
781 blk_add_trace_msg((cfqd)->queue, "cfq " fmt, ##args)
782
783/* Traverses through cfq group service trees */
784#define for_each_cfqg_st(cfqg, i, j, st) \
785 for (i = 0; i <= IDLE_WORKLOAD; i++) \
786 for (j = 0, st = i < IDLE_WORKLOAD ? &cfqg->service_trees[i][j]\
787 : &cfqg->service_tree_idle; \
788 (i < IDLE_WORKLOAD && j <= SYNC_WORKLOAD) || \
789 (i == IDLE_WORKLOAD && j == 0); \
790 j++, st = i < IDLE_WORKLOAD ? \
791 &cfqg->service_trees[i][j]: NULL) \
792
793static inline bool cfq_io_thinktime_big(struct cfq_data *cfqd,
794 struct cfq_ttime *ttime, bool group_idle)
795{
796 unsigned long slice;
797 if (!sample_valid(ttime->ttime_samples))
798 return false;
799 if (group_idle)
800 slice = cfqd->cfq_group_idle;
801 else
802 slice = cfqd->cfq_slice_idle;
803 return ttime->ttime_mean > slice;
804}
805
806static inline bool iops_mode(struct cfq_data *cfqd)
807{
808 /*
809 * If we are not idling on queues and it is a NCQ drive, parallel
810 * execution of requests is on and measuring time is not possible
811 * in most of the cases until and unless we drive shallower queue
812 * depths and that becomes a performance bottleneck. In such cases
813 * switch to start providing fairness in terms of number of IOs.
814 */
815 if (!cfqd->cfq_slice_idle && cfqd->hw_tag)
816 return true;
817 else
818 return false;
819}
820
821static inline enum wl_class_t cfqq_class(struct cfq_queue *cfqq)
822{
823 if (cfq_class_idle(cfqq))
824 return IDLE_WORKLOAD;
825 if (cfq_class_rt(cfqq))
826 return RT_WORKLOAD;
827 return BE_WORKLOAD;
828}
829
830
831static enum wl_type_t cfqq_type(struct cfq_queue *cfqq)
832{
833 if (!cfq_cfqq_sync(cfqq))
834 return ASYNC_WORKLOAD;
835 if (!cfq_cfqq_idle_window(cfqq))
836 return SYNC_NOIDLE_WORKLOAD;
837 return SYNC_WORKLOAD;
838}
839
840static inline int cfq_group_busy_queues_wl(enum wl_class_t wl_class,
841 struct cfq_data *cfqd,
842 struct cfq_group *cfqg)
843{
844 if (wl_class == IDLE_WORKLOAD)
845 return cfqg->service_tree_idle.count;
846
847 return cfqg->service_trees[wl_class][ASYNC_WORKLOAD].count +
848 cfqg->service_trees[wl_class][SYNC_NOIDLE_WORKLOAD].count +
849 cfqg->service_trees[wl_class][SYNC_WORKLOAD].count;
850}
851
852static inline int cfqg_busy_async_queues(struct cfq_data *cfqd,
853 struct cfq_group *cfqg)
854{
855 return cfqg->service_trees[RT_WORKLOAD][ASYNC_WORKLOAD].count +
856 cfqg->service_trees[BE_WORKLOAD][ASYNC_WORKLOAD].count;
857}
858
859static void cfq_dispatch_insert(struct request_queue *, struct request *);
860static struct cfq_queue *cfq_get_queue(struct cfq_data *cfqd, bool is_sync,
861 struct cfq_io_cq *cic, struct bio *bio,
862 gfp_t gfp_mask);
863
864static inline struct cfq_io_cq *icq_to_cic(struct io_cq *icq)
865{
866 /* cic->icq is the first member, %NULL will convert to %NULL */
867 return container_of(icq, struct cfq_io_cq, icq);
868}
869
870static inline struct cfq_io_cq *cfq_cic_lookup(struct cfq_data *cfqd,
871 struct io_context *ioc)
872{
873 if (ioc)
874 return icq_to_cic(ioc_lookup_icq(ioc, cfqd->queue));
875 return NULL;
876}
877
878static inline struct cfq_queue *cic_to_cfqq(struct cfq_io_cq *cic, bool is_sync)
879{
880 return cic->cfqq[is_sync];
881}
882
883static inline void cic_set_cfqq(struct cfq_io_cq *cic, struct cfq_queue *cfqq,
884 bool is_sync)
885{
886 cic->cfqq[is_sync] = cfqq;
887}
888
889static inline struct cfq_data *cic_to_cfqd(struct cfq_io_cq *cic)
890{
891 return cic->icq.q->elevator->elevator_data;
892}
893
894/*
895 * We regard a request as SYNC, if it's either a read or has the SYNC bit
896 * set (in which case it could also be direct WRITE).
897 */
898static inline bool cfq_bio_sync(struct bio *bio)
899{
900 return bio_data_dir(bio) == READ || (bio->bi_rw & REQ_SYNC);
901}
902
903/*
904 * scheduler run of queue, if there are requests pending and no one in the
905 * driver that will restart queueing
906 */
907static inline void cfq_schedule_dispatch(struct cfq_data *cfqd)
908{
909 if (cfqd->busy_queues) {
910 cfq_log(cfqd, "schedule dispatch");
911 kblockd_schedule_work(cfqd->queue, &cfqd->unplug_work);
912 }
913}
914
915/*
916 * Scale schedule slice based on io priority. Use the sync time slice only
917 * if a queue is marked sync and has sync io queued. A sync queue with async
918 * io only, should not get full sync slice length.
919 */
920static inline int cfq_prio_slice(struct cfq_data *cfqd, bool sync,
921 unsigned short prio)
922{
923 const int base_slice = cfqd->cfq_slice[sync];
924
925 WARN_ON(prio >= IOPRIO_BE_NR);
926
927 return base_slice + (base_slice/CFQ_SLICE_SCALE * (4 - prio));
928}
929
930static inline int
931cfq_prio_to_slice(struct cfq_data *cfqd, struct cfq_queue *cfqq)
932{
933 return cfq_prio_slice(cfqd, cfq_cfqq_sync(cfqq), cfqq->ioprio);
934}
935
936/**
937 * cfqg_scale_charge - scale disk time charge according to cfqg weight
938 * @charge: disk time being charged
939 * @vfraction: vfraction of the cfqg, fixed point w/ CFQ_SERVICE_SHIFT
940 *
941 * Scale @charge according to @vfraction, which is in range (0, 1]. The
942 * scaling is inversely proportional.
943 *
944 * scaled = charge / vfraction
945 *
946 * The result is also in fixed point w/ CFQ_SERVICE_SHIFT.
947 */
948static inline u64 cfqg_scale_charge(unsigned long charge,
949 unsigned int vfraction)
950{
951 u64 c = charge << CFQ_SERVICE_SHIFT; /* make it fixed point */
952
953 /* charge / vfraction */
954 c <<= CFQ_SERVICE_SHIFT;
955 do_div(c, vfraction);
956 return c;
957}
958
959static inline u64 max_vdisktime(u64 min_vdisktime, u64 vdisktime)
960{
961 s64 delta = (s64)(vdisktime - min_vdisktime);
962 if (delta > 0)
963 min_vdisktime = vdisktime;
964
965 return min_vdisktime;
966}
967
968static inline u64 min_vdisktime(u64 min_vdisktime, u64 vdisktime)
969{
970 s64 delta = (s64)(vdisktime - min_vdisktime);
971 if (delta < 0)
972 min_vdisktime = vdisktime;
973
974 return min_vdisktime;
975}
976
977static void update_min_vdisktime(struct cfq_rb_root *st)
978{
979 struct cfq_group *cfqg;
980
981 if (st->left) {
982 cfqg = rb_entry_cfqg(st->left);
983 st->min_vdisktime = max_vdisktime(st->min_vdisktime,
984 cfqg->vdisktime);
985 }
986}
987
988/*
989 * get averaged number of queues of RT/BE priority.
990 * average is updated, with a formula that gives more weight to higher numbers,
991 * to quickly follows sudden increases and decrease slowly
992 */
993
994static inline unsigned cfq_group_get_avg_queues(struct cfq_data *cfqd,
995 struct cfq_group *cfqg, bool rt)
996{
997 unsigned min_q, max_q;
998 unsigned mult = cfq_hist_divisor - 1;
999 unsigned round = cfq_hist_divisor / 2;
1000 unsigned busy = cfq_group_busy_queues_wl(rt, cfqd, cfqg);
1001
1002 min_q = min(cfqg->busy_queues_avg[rt], busy);
1003 max_q = max(cfqg->busy_queues_avg[rt], busy);
1004 cfqg->busy_queues_avg[rt] = (mult * max_q + min_q + round) /
1005 cfq_hist_divisor;
1006 return cfqg->busy_queues_avg[rt];
1007}
1008
1009static inline unsigned
1010cfq_group_slice(struct cfq_data *cfqd, struct cfq_group *cfqg)
1011{
1012 return cfqd->cfq_target_latency * cfqg->vfraction >> CFQ_SERVICE_SHIFT;
1013}
1014
1015static inline unsigned
1016cfq_scaled_cfqq_slice(struct cfq_data *cfqd, struct cfq_queue *cfqq)
1017{
1018 unsigned slice = cfq_prio_to_slice(cfqd, cfqq);
1019 if (cfqd->cfq_latency) {
1020 /*
1021 * interested queues (we consider only the ones with the same
1022 * priority class in the cfq group)
1023 */
1024 unsigned iq = cfq_group_get_avg_queues(cfqd, cfqq->cfqg,
1025 cfq_class_rt(cfqq));
1026 unsigned sync_slice = cfqd->cfq_slice[1];
1027 unsigned expect_latency = sync_slice * iq;
1028 unsigned group_slice = cfq_group_slice(cfqd, cfqq->cfqg);
1029
1030 if (expect_latency > group_slice) {
1031 unsigned base_low_slice = 2 * cfqd->cfq_slice_idle;
1032 /* scale low_slice according to IO priority
1033 * and sync vs async */
1034 unsigned low_slice =
1035 min(slice, base_low_slice * slice / sync_slice);
1036 /* the adapted slice value is scaled to fit all iqs
1037 * into the target latency */
1038 slice = max(slice * group_slice / expect_latency,
1039 low_slice);
1040 }
1041 }
1042 return slice;
1043}
1044
1045static inline void
1046cfq_set_prio_slice(struct cfq_data *cfqd, struct cfq_queue *cfqq)
1047{
1048 unsigned slice = cfq_scaled_cfqq_slice(cfqd, cfqq);
1049
1050 cfqq->slice_start = jiffies;
1051 cfqq->slice_end = jiffies + slice;
1052 cfqq->allocated_slice = slice;
1053 cfq_log_cfqq(cfqd, cfqq, "set_slice=%lu", cfqq->slice_end - jiffies);
1054}
1055
1056/*
1057 * We need to wrap this check in cfq_cfqq_slice_new(), since ->slice_end
1058 * isn't valid until the first request from the dispatch is activated
1059 * and the slice time set.
1060 */
1061static inline bool cfq_slice_used(struct cfq_queue *cfqq)
1062{
1063 if (cfq_cfqq_slice_new(cfqq))
1064 return false;
1065 if (time_before(jiffies, cfqq->slice_end))
1066 return false;
1067
1068 return true;
1069}
1070
1071/*
1072 * Lifted from AS - choose which of rq1 and rq2 that is best served now.
1073 * We choose the request that is closest to the head right now. Distance
1074 * behind the head is penalized and only allowed to a certain extent.
1075 */
1076static struct request *
1077cfq_choose_req(struct cfq_data *cfqd, struct request *rq1, struct request *rq2, sector_t last)
1078{
1079 sector_t s1, s2, d1 = 0, d2 = 0;
1080 unsigned long back_max;
1081#define CFQ_RQ1_WRAP 0x01 /* request 1 wraps */
1082#define CFQ_RQ2_WRAP 0x02 /* request 2 wraps */
1083 unsigned wrap = 0; /* bit mask: requests behind the disk head? */
1084
1085 if (rq1 == NULL || rq1 == rq2)
1086 return rq2;
1087 if (rq2 == NULL)
1088 return rq1;
1089
1090 if (rq_is_sync(rq1) != rq_is_sync(rq2))
1091 return rq_is_sync(rq1) ? rq1 : rq2;
1092
1093 if ((rq1->cmd_flags ^ rq2->cmd_flags) & REQ_PRIO)
1094 return rq1->cmd_flags & REQ_PRIO ? rq1 : rq2;
1095
1096 s1 = blk_rq_pos(rq1);
1097 s2 = blk_rq_pos(rq2);
1098
1099 /*
1100 * by definition, 1KiB is 2 sectors
1101 */
1102 back_max = cfqd->cfq_back_max * 2;
1103
1104 /*
1105 * Strict one way elevator _except_ in the case where we allow
1106 * short backward seeks which are biased as twice the cost of a
1107 * similar forward seek.
1108 */
1109 if (s1 >= last)
1110 d1 = s1 - last;
1111 else if (s1 + back_max >= last)
1112 d1 = (last - s1) * cfqd->cfq_back_penalty;
1113 else
1114 wrap |= CFQ_RQ1_WRAP;
1115
1116 if (s2 >= last)
1117 d2 = s2 - last;
1118 else if (s2 + back_max >= last)
1119 d2 = (last - s2) * cfqd->cfq_back_penalty;
1120 else
1121 wrap |= CFQ_RQ2_WRAP;
1122
1123 /* Found required data */
1124
1125 /*
1126 * By doing switch() on the bit mask "wrap" we avoid having to
1127 * check two variables for all permutations: --> faster!
1128 */
1129 switch (wrap) {
1130 case 0: /* common case for CFQ: rq1 and rq2 not wrapped */
1131 if (d1 < d2)
1132 return rq1;
1133 else if (d2 < d1)
1134 return rq2;
1135 else {
1136 if (s1 >= s2)
1137 return rq1;
1138 else
1139 return rq2;
1140 }
1141
1142 case CFQ_RQ2_WRAP:
1143 return rq1;
1144 case CFQ_RQ1_WRAP:
1145 return rq2;
1146 case (CFQ_RQ1_WRAP|CFQ_RQ2_WRAP): /* both rqs wrapped */
1147 default:
1148 /*
1149 * Since both rqs are wrapped,
1150 * start with the one that's further behind head
1151 * (--> only *one* back seek required),
1152 * since back seek takes more time than forward.
1153 */
1154 if (s1 <= s2)
1155 return rq1;
1156 else
1157 return rq2;
1158 }
1159}
1160
1161/*
1162 * The below is leftmost cache rbtree addon
1163 */
1164static struct cfq_queue *cfq_rb_first(struct cfq_rb_root *root)
1165{
1166 /* Service tree is empty */
1167 if (!root->count)
1168 return NULL;
1169
1170 if (!root->left)
1171 root->left = rb_first(&root->rb);
1172
1173 if (root->left)
1174 return rb_entry(root->left, struct cfq_queue, rb_node);
1175
1176 return NULL;
1177}
1178
1179static struct cfq_group *cfq_rb_first_group(struct cfq_rb_root *root)
1180{
1181 if (!root->left)
1182 root->left = rb_first(&root->rb);
1183
1184 if (root->left)
1185 return rb_entry_cfqg(root->left);
1186
1187 return NULL;
1188}
1189
1190static void rb_erase_init(struct rb_node *n, struct rb_root *root)
1191{
1192 rb_erase(n, root);
1193 RB_CLEAR_NODE(n);
1194}
1195
1196static void cfq_rb_erase(struct rb_node *n, struct cfq_rb_root *root)
1197{
1198 if (root->left == n)
1199 root->left = NULL;
1200 rb_erase_init(n, &root->rb);
1201 --root->count;
1202}
1203
1204/*
1205 * would be nice to take fifo expire time into account as well
1206 */
1207static struct request *
1208cfq_find_next_rq(struct cfq_data *cfqd, struct cfq_queue *cfqq,
1209 struct request *last)
1210{
1211 struct rb_node *rbnext = rb_next(&last->rb_node);
1212 struct rb_node *rbprev = rb_prev(&last->rb_node);
1213 struct request *next = NULL, *prev = NULL;
1214
1215 BUG_ON(RB_EMPTY_NODE(&last->rb_node));
1216
1217 if (rbprev)
1218 prev = rb_entry_rq(rbprev);
1219
1220 if (rbnext)
1221 next = rb_entry_rq(rbnext);
1222 else {
1223 rbnext = rb_first(&cfqq->sort_list);
1224 if (rbnext && rbnext != &last->rb_node)
1225 next = rb_entry_rq(rbnext);
1226 }
1227
1228 return cfq_choose_req(cfqd, next, prev, blk_rq_pos(last));
1229}
1230
1231static unsigned long cfq_slice_offset(struct cfq_data *cfqd,
1232 struct cfq_queue *cfqq)
1233{
1234 /*
1235 * just an approximation, should be ok.
1236 */
1237 return (cfqq->cfqg->nr_cfqq - 1) * (cfq_prio_slice(cfqd, 1, 0) -
1238 cfq_prio_slice(cfqd, cfq_cfqq_sync(cfqq), cfqq->ioprio));
1239}
1240
1241static inline s64
1242cfqg_key(struct cfq_rb_root *st, struct cfq_group *cfqg)
1243{
1244 return cfqg->vdisktime - st->min_vdisktime;
1245}
1246
1247static void
1248__cfq_group_service_tree_add(struct cfq_rb_root *st, struct cfq_group *cfqg)
1249{
1250 struct rb_node **node = &st->rb.rb_node;
1251 struct rb_node *parent = NULL;
1252 struct cfq_group *__cfqg;
1253 s64 key = cfqg_key(st, cfqg);
1254 int left = 1;
1255
1256 while (*node != NULL) {
1257 parent = *node;
1258 __cfqg = rb_entry_cfqg(parent);
1259
1260 if (key < cfqg_key(st, __cfqg))
1261 node = &parent->rb_left;
1262 else {
1263 node = &parent->rb_right;
1264 left = 0;
1265 }
1266 }
1267
1268 if (left)
1269 st->left = &cfqg->rb_node;
1270
1271 rb_link_node(&cfqg->rb_node, parent, node);
1272 rb_insert_color(&cfqg->rb_node, &st->rb);
1273}
1274
1275static void
1276cfq_update_group_weight(struct cfq_group *cfqg)
1277{
1278 BUG_ON(!RB_EMPTY_NODE(&cfqg->rb_node));
1279
1280 if (cfqg->new_weight) {
1281 cfqg->weight = cfqg->new_weight;
1282 cfqg->new_weight = 0;
1283 }
1284
1285 if (cfqg->new_leaf_weight) {
1286 cfqg->leaf_weight = cfqg->new_leaf_weight;
1287 cfqg->new_leaf_weight = 0;
1288 }
1289}
1290
1291static void
1292cfq_group_service_tree_add(struct cfq_rb_root *st, struct cfq_group *cfqg)
1293{
1294 unsigned int vfr = 1 << CFQ_SERVICE_SHIFT; /* start with 1 */
1295 struct cfq_group *pos = cfqg;
1296 struct cfq_group *parent;
1297 bool propagate;
1298
1299 /* add to the service tree */
1300 BUG_ON(!RB_EMPTY_NODE(&cfqg->rb_node));
1301
1302 cfq_update_group_weight(cfqg);
1303 __cfq_group_service_tree_add(st, cfqg);
1304
1305 /*
1306 * Activate @cfqg and calculate the portion of vfraction @cfqg is
1307 * entitled to. vfraction is calculated by walking the tree
1308 * towards the root calculating the fraction it has at each level.
1309 * The compounded ratio is how much vfraction @cfqg owns.
1310 *
1311 * Start with the proportion tasks in this cfqg has against active
1312 * children cfqgs - its leaf_weight against children_weight.
1313 */
1314 propagate = !pos->nr_active++;
1315 pos->children_weight += pos->leaf_weight;
1316 vfr = vfr * pos->leaf_weight / pos->children_weight;
1317
1318 /*
1319 * Compound ->weight walking up the tree. Both activation and
1320 * vfraction calculation are done in the same loop. Propagation
1321 * stops once an already activated node is met. vfraction
1322 * calculation should always continue to the root.
1323 */
1324 while ((parent = cfqg_parent(pos))) {
1325 if (propagate) {
1326 propagate = !parent->nr_active++;
1327 parent->children_weight += pos->weight;
1328 }
1329 vfr = vfr * pos->weight / parent->children_weight;
1330 pos = parent;
1331 }
1332
1333 cfqg->vfraction = max_t(unsigned, vfr, 1);
1334}
1335
1336static void
1337cfq_group_notify_queue_add(struct cfq_data *cfqd, struct cfq_group *cfqg)
1338{
1339 struct cfq_rb_root *st = &cfqd->grp_service_tree;
1340 struct cfq_group *__cfqg;
1341 struct rb_node *n;
1342
1343 cfqg->nr_cfqq++;
1344 if (!RB_EMPTY_NODE(&cfqg->rb_node))
1345 return;
1346
1347 /*
1348 * Currently put the group at the end. Later implement something
1349 * so that groups get lesser vtime based on their weights, so that
1350 * if group does not loose all if it was not continuously backlogged.
1351 */
1352 n = rb_last(&st->rb);
1353 if (n) {
1354 __cfqg = rb_entry_cfqg(n);
1355 cfqg->vdisktime = __cfqg->vdisktime + CFQ_IDLE_DELAY;
1356 } else
1357 cfqg->vdisktime = st->min_vdisktime;
1358 cfq_group_service_tree_add(st, cfqg);
1359}
1360
1361static void
1362cfq_group_service_tree_del(struct cfq_rb_root *st, struct cfq_group *cfqg)
1363{
1364 struct cfq_group *pos = cfqg;
1365 bool propagate;
1366
1367 /*
1368 * Undo activation from cfq_group_service_tree_add(). Deactivate
1369 * @cfqg and propagate deactivation upwards.
1370 */
1371 propagate = !--pos->nr_active;
1372 pos->children_weight -= pos->leaf_weight;
1373
1374 while (propagate) {
1375 struct cfq_group *parent = cfqg_parent(pos);
1376
1377 /* @pos has 0 nr_active at this point */
1378 WARN_ON_ONCE(pos->children_weight);
1379 pos->vfraction = 0;
1380
1381 if (!parent)
1382 break;
1383
1384 propagate = !--parent->nr_active;
1385 parent->children_weight -= pos->weight;
1386 pos = parent;
1387 }
1388
1389 /* remove from the service tree */
1390 if (!RB_EMPTY_NODE(&cfqg->rb_node))
1391 cfq_rb_erase(&cfqg->rb_node, st);
1392}
1393
1394static void
1395cfq_group_notify_queue_del(struct cfq_data *cfqd, struct cfq_group *cfqg)
1396{
1397 struct cfq_rb_root *st = &cfqd->grp_service_tree;
1398
1399 BUG_ON(cfqg->nr_cfqq < 1);
1400 cfqg->nr_cfqq--;
1401
1402 /* If there are other cfq queues under this group, don't delete it */
1403 if (cfqg->nr_cfqq)
1404 return;
1405
1406 cfq_log_cfqg(cfqd, cfqg, "del_from_rr group");
1407 cfq_group_service_tree_del(st, cfqg);
1408 cfqg->saved_wl_slice = 0;
1409 cfqg_stats_update_dequeue(cfqg);
1410}
1411
1412static inline unsigned int cfq_cfqq_slice_usage(struct cfq_queue *cfqq,
1413 unsigned int *unaccounted_time)
1414{
1415 unsigned int slice_used;
1416
1417 /*
1418 * Queue got expired before even a single request completed or
1419 * got expired immediately after first request completion.
1420 */
1421 if (!cfqq->slice_start || cfqq->slice_start == jiffies) {
1422 /*
1423 * Also charge the seek time incurred to the group, otherwise
1424 * if there are mutiple queues in the group, each can dispatch
1425 * a single request on seeky media and cause lots of seek time
1426 * and group will never know it.
1427 */
1428 slice_used = max_t(unsigned, (jiffies - cfqq->dispatch_start),
1429 1);
1430 } else {
1431 slice_used = jiffies - cfqq->slice_start;
1432 if (slice_used > cfqq->allocated_slice) {
1433 *unaccounted_time = slice_used - cfqq->allocated_slice;
1434 slice_used = cfqq->allocated_slice;
1435 }
1436 if (time_after(cfqq->slice_start, cfqq->dispatch_start))
1437 *unaccounted_time += cfqq->slice_start -
1438 cfqq->dispatch_start;
1439 }
1440
1441 return slice_used;
1442}
1443
1444static void cfq_group_served(struct cfq_data *cfqd, struct cfq_group *cfqg,
1445 struct cfq_queue *cfqq)
1446{
1447 struct cfq_rb_root *st = &cfqd->grp_service_tree;
1448 unsigned int used_sl, charge, unaccounted_sl = 0;
1449 int nr_sync = cfqg->nr_cfqq - cfqg_busy_async_queues(cfqd, cfqg)
1450 - cfqg->service_tree_idle.count;
1451 unsigned int vfr;
1452
1453 BUG_ON(nr_sync < 0);
1454 used_sl = charge = cfq_cfqq_slice_usage(cfqq, &unaccounted_sl);
1455
1456 if (iops_mode(cfqd))
1457 charge = cfqq->slice_dispatch;
1458 else if (!cfq_cfqq_sync(cfqq) && !nr_sync)
1459 charge = cfqq->allocated_slice;
1460
1461 /*
1462 * Can't update vdisktime while on service tree and cfqg->vfraction
1463 * is valid only while on it. Cache vfr, leave the service tree,
1464 * update vdisktime and go back on. The re-addition to the tree
1465 * will also update the weights as necessary.
1466 */
1467 vfr = cfqg->vfraction;
1468 cfq_group_service_tree_del(st, cfqg);
1469 cfqg->vdisktime += cfqg_scale_charge(charge, vfr);
1470 cfq_group_service_tree_add(st, cfqg);
1471
1472 /* This group is being expired. Save the context */
1473 if (time_after(cfqd->workload_expires, jiffies)) {
1474 cfqg->saved_wl_slice = cfqd->workload_expires
1475 - jiffies;
1476 cfqg->saved_wl_type = cfqd->serving_wl_type;
1477 cfqg->saved_wl_class = cfqd->serving_wl_class;
1478 } else
1479 cfqg->saved_wl_slice = 0;
1480
1481 cfq_log_cfqg(cfqd, cfqg, "served: vt=%llu min_vt=%llu", cfqg->vdisktime,
1482 st->min_vdisktime);
1483 cfq_log_cfqq(cfqq->cfqd, cfqq,
1484 "sl_used=%u disp=%u charge=%u iops=%u sect=%lu",
1485 used_sl, cfqq->slice_dispatch, charge,
1486 iops_mode(cfqd), cfqq->nr_sectors);
1487 cfqg_stats_update_timeslice_used(cfqg, used_sl, unaccounted_sl);
1488 cfqg_stats_set_start_empty_time(cfqg);
1489}
1490
1491/**
1492 * cfq_init_cfqg_base - initialize base part of a cfq_group
1493 * @cfqg: cfq_group to initialize
1494 *
1495 * Initialize the base part which is used whether %CONFIG_CFQ_GROUP_IOSCHED
1496 * is enabled or not.
1497 */
1498static void cfq_init_cfqg_base(struct cfq_group *cfqg)
1499{
1500 struct cfq_rb_root *st;
1501 int i, j;
1502
1503 for_each_cfqg_st(cfqg, i, j, st)
1504 *st = CFQ_RB_ROOT;
1505 RB_CLEAR_NODE(&cfqg->rb_node);
1506
1507 cfqg->ttime.last_end_request = jiffies;
1508}
1509
1510#ifdef CONFIG_CFQ_GROUP_IOSCHED
1511static void cfqg_stats_init(struct cfqg_stats *stats)
1512{
1513 blkg_rwstat_init(&stats->service_bytes);
1514 blkg_rwstat_init(&stats->serviced);
1515 blkg_rwstat_init(&stats->merged);
1516 blkg_rwstat_init(&stats->service_time);
1517 blkg_rwstat_init(&stats->wait_time);
1518 blkg_rwstat_init(&stats->queued);
1519
1520 blkg_stat_init(&stats->sectors);
1521 blkg_stat_init(&stats->time);
1522
1523#ifdef CONFIG_DEBUG_BLK_CGROUP
1524 blkg_stat_init(&stats->unaccounted_time);
1525 blkg_stat_init(&stats->avg_queue_size_sum);
1526 blkg_stat_init(&stats->avg_queue_size_samples);
1527 blkg_stat_init(&stats->dequeue);
1528 blkg_stat_init(&stats->group_wait_time);
1529 blkg_stat_init(&stats->idle_time);
1530 blkg_stat_init(&stats->empty_time);
1531#endif
1532}
1533
1534static void cfq_pd_init(struct blkcg_gq *blkg)
1535{
1536 struct cfq_group *cfqg = blkg_to_cfqg(blkg);
1537
1538 cfq_init_cfqg_base(cfqg);
1539 cfqg->weight = blkg->blkcg->cfq_weight;
1540 cfqg->leaf_weight = blkg->blkcg->cfq_leaf_weight;
1541 cfqg_stats_init(&cfqg->stats);
1542 cfqg_stats_init(&cfqg->dead_stats);
1543}
1544
1545static void cfq_pd_offline(struct blkcg_gq *blkg)
1546{
1547 /*
1548 * @blkg is going offline and will be ignored by
1549 * blkg_[rw]stat_recursive_sum(). Transfer stats to the parent so
1550 * that they don't get lost. If IOs complete after this point, the
1551 * stats for them will be lost. Oh well...
1552 */
1553 cfqg_stats_xfer_dead(blkg_to_cfqg(blkg));
1554}
1555
1556/* offset delta from cfqg->stats to cfqg->dead_stats */
1557static const int dead_stats_off_delta = offsetof(struct cfq_group, dead_stats) -
1558 offsetof(struct cfq_group, stats);
1559
1560/* to be used by recursive prfill, sums live and dead stats recursively */
1561static u64 cfqg_stat_pd_recursive_sum(struct blkg_policy_data *pd, int off)
1562{
1563 u64 sum = 0;
1564
1565 sum += blkg_stat_recursive_sum(pd, off);
1566 sum += blkg_stat_recursive_sum(pd, off + dead_stats_off_delta);
1567 return sum;
1568}
1569
1570/* to be used by recursive prfill, sums live and dead rwstats recursively */
1571static struct blkg_rwstat cfqg_rwstat_pd_recursive_sum(struct blkg_policy_data *pd,
1572 int off)
1573{
1574 struct blkg_rwstat a, b;
1575
1576 a = blkg_rwstat_recursive_sum(pd, off);
1577 b = blkg_rwstat_recursive_sum(pd, off + dead_stats_off_delta);
1578 blkg_rwstat_merge(&a, &b);
1579 return a;
1580}
1581
1582static void cfq_pd_reset_stats(struct blkcg_gq *blkg)
1583{
1584 struct cfq_group *cfqg = blkg_to_cfqg(blkg);
1585
1586 cfqg_stats_reset(&cfqg->stats);
1587 cfqg_stats_reset(&cfqg->dead_stats);
1588}
1589
1590/*
1591 * Search for the cfq group current task belongs to. request_queue lock must
1592 * be held.
1593 */
1594static struct cfq_group *cfq_lookup_create_cfqg(struct cfq_data *cfqd,
1595 struct blkcg *blkcg)
1596{
1597 struct request_queue *q = cfqd->queue;
1598 struct cfq_group *cfqg = NULL;
1599
1600 /* avoid lookup for the common case where there's no blkcg */
1601 if (blkcg == &blkcg_root) {
1602 cfqg = cfqd->root_group;
1603 } else {
1604 struct blkcg_gq *blkg;
1605
1606 blkg = blkg_lookup_create(blkcg, q);
1607 if (!IS_ERR(blkg))
1608 cfqg = blkg_to_cfqg(blkg);
1609 }
1610
1611 return cfqg;
1612}
1613
1614static void cfq_link_cfqq_cfqg(struct cfq_queue *cfqq, struct cfq_group *cfqg)
1615{
1616 /* Currently, all async queues are mapped to root group */
1617 if (!cfq_cfqq_sync(cfqq))
1618 cfqg = cfqq->cfqd->root_group;
1619
1620 cfqq->cfqg = cfqg;
1621 /* cfqq reference on cfqg */
1622 cfqg_get(cfqg);
1623}
1624
1625static u64 cfqg_prfill_weight_device(struct seq_file *sf,
1626 struct blkg_policy_data *pd, int off)
1627{
1628 struct cfq_group *cfqg = pd_to_cfqg(pd);
1629
1630 if (!cfqg->dev_weight)
1631 return 0;
1632 return __blkg_prfill_u64(sf, pd, cfqg->dev_weight);
1633}
1634
1635static int cfqg_print_weight_device(struct seq_file *sf, void *v)
1636{
1637 blkcg_print_blkgs(sf, css_to_blkcg(seq_css(sf)),
1638 cfqg_prfill_weight_device, &blkcg_policy_cfq,
1639 0, false);
1640 return 0;
1641}
1642
1643static u64 cfqg_prfill_leaf_weight_device(struct seq_file *sf,
1644 struct blkg_policy_data *pd, int off)
1645{
1646 struct cfq_group *cfqg = pd_to_cfqg(pd);
1647
1648 if (!cfqg->dev_leaf_weight)
1649 return 0;
1650 return __blkg_prfill_u64(sf, pd, cfqg->dev_leaf_weight);
1651}
1652
1653static int cfqg_print_leaf_weight_device(struct seq_file *sf, void *v)
1654{
1655 blkcg_print_blkgs(sf, css_to_blkcg(seq_css(sf)),
1656 cfqg_prfill_leaf_weight_device, &blkcg_policy_cfq,
1657 0, false);
1658 return 0;
1659}
1660
1661static int cfq_print_weight(struct seq_file *sf, void *v)
1662{
1663 seq_printf(sf, "%u\n", css_to_blkcg(seq_css(sf))->cfq_weight);
1664 return 0;
1665}
1666
1667static int cfq_print_leaf_weight(struct seq_file *sf, void *v)
1668{
1669 seq_printf(sf, "%u\n", css_to_blkcg(seq_css(sf))->cfq_leaf_weight);
1670 return 0;
1671}
1672
1673static int __cfqg_set_weight_device(struct cgroup_subsys_state *css,
1674 struct cftype *cft, const char *buf,
1675 bool is_leaf_weight)
1676{
1677 struct blkcg *blkcg = css_to_blkcg(css);
1678 struct blkg_conf_ctx ctx;
1679 struct cfq_group *cfqg;
1680 int ret;
1681
1682 ret = blkg_conf_prep(blkcg, &blkcg_policy_cfq, buf, &ctx);
1683 if (ret)
1684 return ret;
1685
1686 ret = -EINVAL;
1687 cfqg = blkg_to_cfqg(ctx.blkg);
1688 if (!ctx.v || (ctx.v >= CFQ_WEIGHT_MIN && ctx.v <= CFQ_WEIGHT_MAX)) {
1689 if (!is_leaf_weight) {
1690 cfqg->dev_weight = ctx.v;
1691 cfqg->new_weight = ctx.v ?: blkcg->cfq_weight;
1692 } else {
1693 cfqg->dev_leaf_weight = ctx.v;
1694 cfqg->new_leaf_weight = ctx.v ?: blkcg->cfq_leaf_weight;
1695 }
1696 ret = 0;
1697 }
1698
1699 blkg_conf_finish(&ctx);
1700 return ret;
1701}
1702
1703static int cfqg_set_weight_device(struct cgroup_subsys_state *css,
1704 struct cftype *cft, char *buf)
1705{
1706 return __cfqg_set_weight_device(css, cft, buf, false);
1707}
1708
1709static int cfqg_set_leaf_weight_device(struct cgroup_subsys_state *css,
1710 struct cftype *cft, char *buf)
1711{
1712 return __cfqg_set_weight_device(css, cft, buf, true);
1713}
1714
1715static int __cfq_set_weight(struct cgroup_subsys_state *css, struct cftype *cft,
1716 u64 val, bool is_leaf_weight)
1717{
1718 struct blkcg *blkcg = css_to_blkcg(css);
1719 struct blkcg_gq *blkg;
1720
1721 if (val < CFQ_WEIGHT_MIN || val > CFQ_WEIGHT_MAX)
1722 return -EINVAL;
1723
1724 spin_lock_irq(&blkcg->lock);
1725
1726 if (!is_leaf_weight)
1727 blkcg->cfq_weight = val;
1728 else
1729 blkcg->cfq_leaf_weight = val;
1730
1731 hlist_for_each_entry(blkg, &blkcg->blkg_list, blkcg_node) {
1732 struct cfq_group *cfqg = blkg_to_cfqg(blkg);
1733
1734 if (!cfqg)
1735 continue;
1736
1737 if (!is_leaf_weight) {
1738 if (!cfqg->dev_weight)
1739 cfqg->new_weight = blkcg->cfq_weight;
1740 } else {
1741 if (!cfqg->dev_leaf_weight)
1742 cfqg->new_leaf_weight = blkcg->cfq_leaf_weight;
1743 }
1744 }
1745
1746 spin_unlock_irq(&blkcg->lock);
1747 return 0;
1748}
1749
1750static int cfq_set_weight(struct cgroup_subsys_state *css, struct cftype *cft,
1751 u64 val)
1752{
1753 return __cfq_set_weight(css, cft, val, false);
1754}
1755
1756static int cfq_set_leaf_weight(struct cgroup_subsys_state *css,
1757 struct cftype *cft, u64 val)
1758{
1759 return __cfq_set_weight(css, cft, val, true);
1760}
1761
1762static int cfqg_print_stat(struct seq_file *sf, void *v)
1763{
1764 blkcg_print_blkgs(sf, css_to_blkcg(seq_css(sf)), blkg_prfill_stat,
1765 &blkcg_policy_cfq, seq_cft(sf)->private, false);
1766 return 0;
1767}
1768
1769static int cfqg_print_rwstat(struct seq_file *sf, void *v)
1770{
1771 blkcg_print_blkgs(sf, css_to_blkcg(seq_css(sf)), blkg_prfill_rwstat,
1772 &blkcg_policy_cfq, seq_cft(sf)->private, true);
1773 return 0;
1774}
1775
1776static u64 cfqg_prfill_stat_recursive(struct seq_file *sf,
1777 struct blkg_policy_data *pd, int off)
1778{
1779 u64 sum = cfqg_stat_pd_recursive_sum(pd, off);
1780
1781 return __blkg_prfill_u64(sf, pd, sum);
1782}
1783
1784static u64 cfqg_prfill_rwstat_recursive(struct seq_file *sf,
1785 struct blkg_policy_data *pd, int off)
1786{
1787 struct blkg_rwstat sum = cfqg_rwstat_pd_recursive_sum(pd, off);
1788
1789 return __blkg_prfill_rwstat(sf, pd, &sum);
1790}
1791
1792static int cfqg_print_stat_recursive(struct seq_file *sf, void *v)
1793{
1794 blkcg_print_blkgs(sf, css_to_blkcg(seq_css(sf)),
1795 cfqg_prfill_stat_recursive, &blkcg_policy_cfq,
1796 seq_cft(sf)->private, false);
1797 return 0;
1798}
1799
1800static int cfqg_print_rwstat_recursive(struct seq_file *sf, void *v)
1801{
1802 blkcg_print_blkgs(sf, css_to_blkcg(seq_css(sf)),
1803 cfqg_prfill_rwstat_recursive, &blkcg_policy_cfq,
1804 seq_cft(sf)->private, true);
1805 return 0;
1806}
1807
1808#ifdef CONFIG_DEBUG_BLK_CGROUP
1809static u64 cfqg_prfill_avg_queue_size(struct seq_file *sf,
1810 struct blkg_policy_data *pd, int off)
1811{
1812 struct cfq_group *cfqg = pd_to_cfqg(pd);
1813 u64 samples = blkg_stat_read(&cfqg->stats.avg_queue_size_samples);
1814 u64 v = 0;
1815
1816 if (samples) {
1817 v = blkg_stat_read(&cfqg->stats.avg_queue_size_sum);
1818 v = div64_u64(v, samples);
1819 }
1820 __blkg_prfill_u64(sf, pd, v);
1821 return 0;
1822}
1823
1824/* print avg_queue_size */
1825static int cfqg_print_avg_queue_size(struct seq_file *sf, void *v)
1826{
1827 blkcg_print_blkgs(sf, css_to_blkcg(seq_css(sf)),
1828 cfqg_prfill_avg_queue_size, &blkcg_policy_cfq,
1829 0, false);
1830 return 0;
1831}
1832#endif /* CONFIG_DEBUG_BLK_CGROUP */
1833
1834static struct cftype cfq_blkcg_files[] = {
1835 /* on root, weight is mapped to leaf_weight */
1836 {
1837 .name = "weight_device",
1838 .flags = CFTYPE_ONLY_ON_ROOT,
1839 .seq_show = cfqg_print_leaf_weight_device,
1840 .write_string = cfqg_set_leaf_weight_device,
1841 },
1842 {
1843 .name = "weight",
1844 .flags = CFTYPE_ONLY_ON_ROOT,
1845 .seq_show = cfq_print_leaf_weight,
1846 .write_u64 = cfq_set_leaf_weight,
1847 },
1848
1849 /* no such mapping necessary for !roots */
1850 {
1851 .name = "weight_device",
1852 .flags = CFTYPE_NOT_ON_ROOT,
1853 .seq_show = cfqg_print_weight_device,
1854 .write_string = cfqg_set_weight_device,
1855 },
1856 {
1857 .name = "weight",
1858 .flags = CFTYPE_NOT_ON_ROOT,
1859 .seq_show = cfq_print_weight,
1860 .write_u64 = cfq_set_weight,
1861 },
1862
1863 {
1864 .name = "leaf_weight_device",
1865 .seq_show = cfqg_print_leaf_weight_device,
1866 .write_string = cfqg_set_leaf_weight_device,
1867 },
1868 {
1869 .name = "leaf_weight",
1870 .seq_show = cfq_print_leaf_weight,
1871 .write_u64 = cfq_set_leaf_weight,
1872 },
1873
1874 /* statistics, covers only the tasks in the cfqg */
1875 {
1876 .name = "time",
1877 .private = offsetof(struct cfq_group, stats.time),
1878 .seq_show = cfqg_print_stat,
1879 },
1880 {
1881 .name = "sectors",
1882 .private = offsetof(struct cfq_group, stats.sectors),
1883 .seq_show = cfqg_print_stat,
1884 },
1885 {
1886 .name = "io_service_bytes",
1887 .private = offsetof(struct cfq_group, stats.service_bytes),
1888 .seq_show = cfqg_print_rwstat,
1889 },
1890 {
1891 .name = "io_serviced",
1892 .private = offsetof(struct cfq_group, stats.serviced),
1893 .seq_show = cfqg_print_rwstat,
1894 },
1895 {
1896 .name = "io_service_time",
1897 .private = offsetof(struct cfq_group, stats.service_time),
1898 .seq_show = cfqg_print_rwstat,
1899 },
1900 {
1901 .name = "io_wait_time",
1902 .private = offsetof(struct cfq_group, stats.wait_time),
1903 .seq_show = cfqg_print_rwstat,
1904 },
1905 {
1906 .name = "io_merged",
1907 .private = offsetof(struct cfq_group, stats.merged),
1908 .seq_show = cfqg_print_rwstat,
1909 },
1910 {
1911 .name = "io_queued",
1912 .private = offsetof(struct cfq_group, stats.queued),
1913 .seq_show = cfqg_print_rwstat,
1914 },
1915
1916 /* the same statictics which cover the cfqg and its descendants */
1917 {
1918 .name = "time_recursive",
1919 .private = offsetof(struct cfq_group, stats.time),
1920 .seq_show = cfqg_print_stat_recursive,
1921 },
1922 {
1923 .name = "sectors_recursive",
1924 .private = offsetof(struct cfq_group, stats.sectors),
1925 .seq_show = cfqg_print_stat_recursive,
1926 },
1927 {
1928 .name = "io_service_bytes_recursive",
1929 .private = offsetof(struct cfq_group, stats.service_bytes),
1930 .seq_show = cfqg_print_rwstat_recursive,
1931 },
1932 {
1933 .name = "io_serviced_recursive",
1934 .private = offsetof(struct cfq_group, stats.serviced),
1935 .seq_show = cfqg_print_rwstat_recursive,
1936 },
1937 {
1938 .name = "io_service_time_recursive",
1939 .private = offsetof(struct cfq_group, stats.service_time),
1940 .seq_show = cfqg_print_rwstat_recursive,
1941 },
1942 {
1943 .name = "io_wait_time_recursive",
1944 .private = offsetof(struct cfq_group, stats.wait_time),
1945 .seq_show = cfqg_print_rwstat_recursive,
1946 },
1947 {
1948 .name = "io_merged_recursive",
1949 .private = offsetof(struct cfq_group, stats.merged),
1950 .seq_show = cfqg_print_rwstat_recursive,
1951 },
1952 {
1953 .name = "io_queued_recursive",
1954 .private = offsetof(struct cfq_group, stats.queued),
1955 .seq_show = cfqg_print_rwstat_recursive,
1956 },
1957#ifdef CONFIG_DEBUG_BLK_CGROUP
1958 {
1959 .name = "avg_queue_size",
1960 .seq_show = cfqg_print_avg_queue_size,
1961 },
1962 {
1963 .name = "group_wait_time",
1964 .private = offsetof(struct cfq_group, stats.group_wait_time),
1965 .seq_show = cfqg_print_stat,
1966 },
1967 {
1968 .name = "idle_time",
1969 .private = offsetof(struct cfq_group, stats.idle_time),
1970 .seq_show = cfqg_print_stat,
1971 },
1972 {
1973 .name = "empty_time",
1974 .private = offsetof(struct cfq_group, stats.empty_time),
1975 .seq_show = cfqg_print_stat,
1976 },
1977 {
1978 .name = "dequeue",
1979 .private = offsetof(struct cfq_group, stats.dequeue),
1980 .seq_show = cfqg_print_stat,
1981 },
1982 {
1983 .name = "unaccounted_time",
1984 .private = offsetof(struct cfq_group, stats.unaccounted_time),
1985 .seq_show = cfqg_print_stat,
1986 },
1987#endif /* CONFIG_DEBUG_BLK_CGROUP */
1988 { } /* terminate */
1989};
1990#else /* GROUP_IOSCHED */
1991static struct cfq_group *cfq_lookup_create_cfqg(struct cfq_data *cfqd,
1992 struct blkcg *blkcg)
1993{
1994 return cfqd->root_group;
1995}
1996
1997static inline void
1998cfq_link_cfqq_cfqg(struct cfq_queue *cfqq, struct cfq_group *cfqg) {
1999 cfqq->cfqg = cfqg;
2000}
2001
2002#endif /* GROUP_IOSCHED */
2003
2004/*
2005 * The cfqd->service_trees holds all pending cfq_queue's that have
2006 * requests waiting to be processed. It is sorted in the order that
2007 * we will service the queues.
2008 */
2009static void cfq_service_tree_add(struct cfq_data *cfqd, struct cfq_queue *cfqq,
2010 bool add_front)
2011{
2012 struct rb_node **p, *parent;
2013 struct cfq_queue *__cfqq;
2014 unsigned long rb_key;
2015 struct cfq_rb_root *st;
2016 int left;
2017 int new_cfqq = 1;
2018
2019 st = st_for(cfqq->cfqg, cfqq_class(cfqq), cfqq_type(cfqq));
2020 if (cfq_class_idle(cfqq)) {
2021 rb_key = CFQ_IDLE_DELAY;
2022 parent = rb_last(&st->rb);
2023 if (parent && parent != &cfqq->rb_node) {
2024 __cfqq = rb_entry(parent, struct cfq_queue, rb_node);
2025 rb_key += __cfqq->rb_key;
2026 } else
2027 rb_key += jiffies;
2028 } else if (!add_front) {
2029 /*
2030 * Get our rb key offset. Subtract any residual slice
2031 * value carried from last service. A negative resid
2032 * count indicates slice overrun, and this should position
2033 * the next service time further away in the tree.
2034 */
2035 rb_key = cfq_slice_offset(cfqd, cfqq) + jiffies;
2036 rb_key -= cfqq->slice_resid;
2037 cfqq->slice_resid = 0;
2038 } else {
2039 rb_key = -HZ;
2040 __cfqq = cfq_rb_first(st);
2041 rb_key += __cfqq ? __cfqq->rb_key : jiffies;
2042 }
2043
2044 if (!RB_EMPTY_NODE(&cfqq->rb_node)) {
2045 new_cfqq = 0;
2046 /*
2047 * same position, nothing more to do
2048 */
2049 if (rb_key == cfqq->rb_key && cfqq->service_tree == st)
2050 return;
2051
2052 cfq_rb_erase(&cfqq->rb_node, cfqq->service_tree);
2053 cfqq->service_tree = NULL;
2054 }
2055
2056 left = 1;
2057 parent = NULL;
2058 cfqq->service_tree = st;
2059 p = &st->rb.rb_node;
2060 while (*p) {
2061 parent = *p;
2062 __cfqq = rb_entry(parent, struct cfq_queue, rb_node);
2063
2064 /*
2065 * sort by key, that represents service time.
2066 */
2067 if (time_before(rb_key, __cfqq->rb_key))
2068 p = &parent->rb_left;
2069 else {
2070 p = &parent->rb_right;
2071 left = 0;
2072 }
2073 }
2074
2075 if (left)
2076 st->left = &cfqq->rb_node;
2077
2078 cfqq->rb_key = rb_key;
2079 rb_link_node(&cfqq->rb_node, parent, p);
2080 rb_insert_color(&cfqq->rb_node, &st->rb);
2081 st->count++;
2082 if (add_front || !new_cfqq)
2083 return;
2084 cfq_group_notify_queue_add(cfqd, cfqq->cfqg);
2085}
2086
2087static struct cfq_queue *
2088cfq_prio_tree_lookup(struct cfq_data *cfqd, struct rb_root *root,
2089 sector_t sector, struct rb_node **ret_parent,
2090 struct rb_node ***rb_link)
2091{
2092 struct rb_node **p, *parent;
2093 struct cfq_queue *cfqq = NULL;
2094
2095 parent = NULL;
2096 p = &root->rb_node;
2097 while (*p) {
2098 struct rb_node **n;
2099
2100 parent = *p;
2101 cfqq = rb_entry(parent, struct cfq_queue, p_node);
2102
2103 /*
2104 * Sort strictly based on sector. Smallest to the left,
2105 * largest to the right.
2106 */
2107 if (sector > blk_rq_pos(cfqq->next_rq))
2108 n = &(*p)->rb_right;
2109 else if (sector < blk_rq_pos(cfqq->next_rq))
2110 n = &(*p)->rb_left;
2111 else
2112 break;
2113 p = n;
2114 cfqq = NULL;
2115 }
2116
2117 *ret_parent = parent;
2118 if (rb_link)
2119 *rb_link = p;
2120 return cfqq;
2121}
2122
2123static void cfq_prio_tree_add(struct cfq_data *cfqd, struct cfq_queue *cfqq)
2124{
2125 struct rb_node **p, *parent;
2126 struct cfq_queue *__cfqq;
2127
2128 if (cfqq->p_root) {
2129 rb_erase(&cfqq->p_node, cfqq->p_root);
2130 cfqq->p_root = NULL;
2131 }
2132
2133 if (cfq_class_idle(cfqq))
2134 return;
2135 if (!cfqq->next_rq)
2136 return;
2137
2138 cfqq->p_root = &cfqd->prio_trees[cfqq->org_ioprio];
2139 __cfqq = cfq_prio_tree_lookup(cfqd, cfqq->p_root,
2140 blk_rq_pos(cfqq->next_rq), &parent, &p);
2141 if (!__cfqq) {
2142 rb_link_node(&cfqq->p_node, parent, p);
2143 rb_insert_color(&cfqq->p_node, cfqq->p_root);
2144 } else
2145 cfqq->p_root = NULL;
2146}
2147
2148/*
2149 * Update cfqq's position in the service tree.
2150 */
2151static void cfq_resort_rr_list(struct cfq_data *cfqd, struct cfq_queue *cfqq)
2152{
2153 /*
2154 * Resorting requires the cfqq to be on the RR list already.
2155 */
2156 if (cfq_cfqq_on_rr(cfqq)) {
2157 cfq_service_tree_add(cfqd, cfqq, 0);
2158 cfq_prio_tree_add(cfqd, cfqq);
2159 }
2160}
2161
2162/*
2163 * add to busy list of queues for service, trying to be fair in ordering
2164 * the pending list according to last request service
2165 */
2166static void cfq_add_cfqq_rr(struct cfq_data *cfqd, struct cfq_queue *cfqq)
2167{
2168 cfq_log_cfqq(cfqd, cfqq, "add_to_rr");
2169 BUG_ON(cfq_cfqq_on_rr(cfqq));
2170 cfq_mark_cfqq_on_rr(cfqq);
2171 cfqd->busy_queues++;
2172 if (cfq_cfqq_sync(cfqq))
2173 cfqd->busy_sync_queues++;
2174
2175 cfq_resort_rr_list(cfqd, cfqq);
2176}
2177
2178/*
2179 * Called when the cfqq no longer has requests pending, remove it from
2180 * the service tree.
2181 */
2182static void cfq_del_cfqq_rr(struct cfq_data *cfqd, struct cfq_queue *cfqq)
2183{
2184 cfq_log_cfqq(cfqd, cfqq, "del_from_rr");
2185 BUG_ON(!cfq_cfqq_on_rr(cfqq));
2186 cfq_clear_cfqq_on_rr(cfqq);
2187
2188 if (!RB_EMPTY_NODE(&cfqq->rb_node)) {
2189 cfq_rb_erase(&cfqq->rb_node, cfqq->service_tree);
2190 cfqq->service_tree = NULL;
2191 }
2192 if (cfqq->p_root) {
2193 rb_erase(&cfqq->p_node, cfqq->p_root);
2194 cfqq->p_root = NULL;
2195 }
2196
2197 cfq_group_notify_queue_del(cfqd, cfqq->cfqg);
2198 BUG_ON(!cfqd->busy_queues);
2199 cfqd->busy_queues--;
2200 if (cfq_cfqq_sync(cfqq))
2201 cfqd->busy_sync_queues--;
2202}
2203
2204/*
2205 * rb tree support functions
2206 */
2207static void cfq_del_rq_rb(struct request *rq)
2208{
2209 struct cfq_queue *cfqq = RQ_CFQQ(rq);
2210 const int sync = rq_is_sync(rq);
2211
2212 BUG_ON(!cfqq->queued[sync]);
2213 cfqq->queued[sync]--;
2214
2215 elv_rb_del(&cfqq->sort_list, rq);
2216
2217 if (cfq_cfqq_on_rr(cfqq) && RB_EMPTY_ROOT(&cfqq->sort_list)) {
2218 /*
2219 * Queue will be deleted from service tree when we actually
2220 * expire it later. Right now just remove it from prio tree
2221 * as it is empty.
2222 */
2223 if (cfqq->p_root) {
2224 rb_erase(&cfqq->p_node, cfqq->p_root);
2225 cfqq->p_root = NULL;
2226 }
2227 }
2228}
2229
2230static void cfq_add_rq_rb(struct request *rq)
2231{
2232 struct cfq_queue *cfqq = RQ_CFQQ(rq);
2233 struct cfq_data *cfqd = cfqq->cfqd;
2234 struct request *prev;
2235
2236 cfqq->queued[rq_is_sync(rq)]++;
2237
2238 elv_rb_add(&cfqq->sort_list, rq);
2239
2240 if (!cfq_cfqq_on_rr(cfqq))
2241 cfq_add_cfqq_rr(cfqd, cfqq);
2242
2243 /*
2244 * check if this request is a better next-serve candidate
2245 */
2246 prev = cfqq->next_rq;
2247 cfqq->next_rq = cfq_choose_req(cfqd, cfqq->next_rq, rq, cfqd->last_position);
2248
2249 /*
2250 * adjust priority tree position, if ->next_rq changes
2251 */
2252 if (prev != cfqq->next_rq)
2253 cfq_prio_tree_add(cfqd, cfqq);
2254
2255 BUG_ON(!cfqq->next_rq);
2256}
2257
2258static void cfq_reposition_rq_rb(struct cfq_queue *cfqq, struct request *rq)
2259{
2260 elv_rb_del(&cfqq->sort_list, rq);
2261 cfqq->queued[rq_is_sync(rq)]--;
2262 cfqg_stats_update_io_remove(RQ_CFQG(rq), rq->cmd_flags);
2263 cfq_add_rq_rb(rq);
2264 cfqg_stats_update_io_add(RQ_CFQG(rq), cfqq->cfqd->serving_group,
2265 rq->cmd_flags);
2266}
2267
2268static struct request *
2269cfq_find_rq_fmerge(struct cfq_data *cfqd, struct bio *bio)
2270{
2271 struct task_struct *tsk = current;
2272 struct cfq_io_cq *cic;
2273 struct cfq_queue *cfqq;
2274
2275 cic = cfq_cic_lookup(cfqd, tsk->io_context);
2276 if (!cic)
2277 return NULL;
2278
2279 cfqq = cic_to_cfqq(cic, cfq_bio_sync(bio));
2280 if (cfqq)
2281 return elv_rb_find(&cfqq->sort_list, bio_end_sector(bio));
2282
2283 return NULL;
2284}
2285
2286static void cfq_activate_request(struct request_queue *q, struct request *rq)
2287{
2288 struct cfq_data *cfqd = q->elevator->elevator_data;
2289
2290 cfqd->rq_in_driver++;
2291 cfq_log_cfqq(cfqd, RQ_CFQQ(rq), "activate rq, drv=%d",
2292 cfqd->rq_in_driver);
2293
2294 cfqd->last_position = blk_rq_pos(rq) + blk_rq_sectors(rq);
2295}
2296
2297static void cfq_deactivate_request(struct request_queue *q, struct request *rq)
2298{
2299 struct cfq_data *cfqd = q->elevator->elevator_data;
2300
2301 WARN_ON(!cfqd->rq_in_driver);
2302 cfqd->rq_in_driver--;
2303 cfq_log_cfqq(cfqd, RQ_CFQQ(rq), "deactivate rq, drv=%d",
2304 cfqd->rq_in_driver);
2305}
2306
2307static void cfq_remove_request(struct request *rq)
2308{
2309 struct cfq_queue *cfqq = RQ_CFQQ(rq);
2310
2311 if (cfqq->next_rq == rq)
2312 cfqq->next_rq = cfq_find_next_rq(cfqq->cfqd, cfqq, rq);
2313
2314 list_del_init(&rq->queuelist);
2315 cfq_del_rq_rb(rq);
2316
2317 cfqq->cfqd->rq_queued--;
2318 cfqg_stats_update_io_remove(RQ_CFQG(rq), rq->cmd_flags);
2319 if (rq->cmd_flags & REQ_PRIO) {
2320 WARN_ON(!cfqq->prio_pending);
2321 cfqq->prio_pending--;
2322 }
2323}
2324
2325static int cfq_merge(struct request_queue *q, struct request **req,
2326 struct bio *bio)
2327{
2328 struct cfq_data *cfqd = q->elevator->elevator_data;
2329 struct request *__rq;
2330
2331 __rq = cfq_find_rq_fmerge(cfqd, bio);
2332 if (__rq && elv_rq_merge_ok(__rq, bio)) {
2333 *req = __rq;
2334 return ELEVATOR_FRONT_MERGE;
2335 }
2336
2337 return ELEVATOR_NO_MERGE;
2338}
2339
2340static void cfq_merged_request(struct request_queue *q, struct request *req,
2341 int type)
2342{
2343 if (type == ELEVATOR_FRONT_MERGE) {
2344 struct cfq_queue *cfqq = RQ_CFQQ(req);
2345
2346 cfq_reposition_rq_rb(cfqq, req);
2347 }
2348}
2349
2350static void cfq_bio_merged(struct request_queue *q, struct request *req,
2351 struct bio *bio)
2352{
2353 cfqg_stats_update_io_merged(RQ_CFQG(req), bio->bi_rw);
2354}
2355
2356static void
2357cfq_merged_requests(struct request_queue *q, struct request *rq,
2358 struct request *next)
2359{
2360 struct cfq_queue *cfqq = RQ_CFQQ(rq);
2361 struct cfq_data *cfqd = q->elevator->elevator_data;
2362
2363 /*
2364 * reposition in fifo if next is older than rq
2365 */
2366 if (!list_empty(&rq->queuelist) && !list_empty(&next->queuelist) &&
2367 time_before(next->fifo_time, rq->fifo_time) &&
2368 cfqq == RQ_CFQQ(next)) {
2369 list_move(&rq->queuelist, &next->queuelist);
2370 rq->fifo_time = next->fifo_time;
2371 }
2372
2373 if (cfqq->next_rq == next)
2374 cfqq->next_rq = rq;
2375 cfq_remove_request(next);
2376 cfqg_stats_update_io_merged(RQ_CFQG(rq), next->cmd_flags);
2377
2378 cfqq = RQ_CFQQ(next);
2379 /*
2380 * all requests of this queue are merged to other queues, delete it
2381 * from the service tree. If it's the active_queue,
2382 * cfq_dispatch_requests() will choose to expire it or do idle
2383 */
2384 if (cfq_cfqq_on_rr(cfqq) && RB_EMPTY_ROOT(&cfqq->sort_list) &&
2385 cfqq != cfqd->active_queue)
2386 cfq_del_cfqq_rr(cfqd, cfqq);
2387}
2388
2389static int cfq_allow_merge(struct request_queue *q, struct request *rq,
2390 struct bio *bio)
2391{
2392 struct cfq_data *cfqd = q->elevator->elevator_data;
2393 struct cfq_io_cq *cic;
2394 struct cfq_queue *cfqq;
2395
2396 /*
2397 * Disallow merge of a sync bio into an async request.
2398 */
2399 if (cfq_bio_sync(bio) && !rq_is_sync(rq))
2400 return false;
2401
2402 /*
2403 * Lookup the cfqq that this bio will be queued with and allow
2404 * merge only if rq is queued there.
2405 */
2406 cic = cfq_cic_lookup(cfqd, current->io_context);
2407 if (!cic)
2408 return false;
2409
2410 cfqq = cic_to_cfqq(cic, cfq_bio_sync(bio));
2411 return cfqq == RQ_CFQQ(rq);
2412}
2413
2414static inline void cfq_del_timer(struct cfq_data *cfqd, struct cfq_queue *cfqq)
2415{
2416 del_timer(&cfqd->idle_slice_timer);
2417 cfqg_stats_update_idle_time(cfqq->cfqg);
2418}
2419
2420static void __cfq_set_active_queue(struct cfq_data *cfqd,
2421 struct cfq_queue *cfqq)
2422{
2423 if (cfqq) {
2424 cfq_log_cfqq(cfqd, cfqq, "set_active wl_class:%d wl_type:%d",
2425 cfqd->serving_wl_class, cfqd->serving_wl_type);
2426 cfqg_stats_update_avg_queue_size(cfqq->cfqg);
2427 cfqq->slice_start = 0;
2428 cfqq->dispatch_start = jiffies;
2429 cfqq->allocated_slice = 0;
2430 cfqq->slice_end = 0;
2431 cfqq->slice_dispatch = 0;
2432 cfqq->nr_sectors = 0;
2433
2434 cfq_clear_cfqq_wait_request(cfqq);
2435 cfq_clear_cfqq_must_dispatch(cfqq);
2436 cfq_clear_cfqq_must_alloc_slice(cfqq);
2437 cfq_clear_cfqq_fifo_expire(cfqq);
2438 cfq_mark_cfqq_slice_new(cfqq);
2439
2440 cfq_del_timer(cfqd, cfqq);
2441 }
2442
2443 cfqd->active_queue = cfqq;
2444}
2445
2446/*
2447 * current cfqq expired its slice (or was too idle), select new one
2448 */
2449static void
2450__cfq_slice_expired(struct cfq_data *cfqd, struct cfq_queue *cfqq,
2451 bool timed_out)
2452{
2453 cfq_log_cfqq(cfqd, cfqq, "slice expired t=%d", timed_out);
2454
2455 if (cfq_cfqq_wait_request(cfqq))
2456 cfq_del_timer(cfqd, cfqq);
2457
2458 cfq_clear_cfqq_wait_request(cfqq);
2459 cfq_clear_cfqq_wait_busy(cfqq);
2460
2461 /*
2462 * If this cfqq is shared between multiple processes, check to
2463 * make sure that those processes are still issuing I/Os within
2464 * the mean seek distance. If not, it may be time to break the
2465 * queues apart again.
2466 */
2467 if (cfq_cfqq_coop(cfqq) && CFQQ_SEEKY(cfqq))
2468 cfq_mark_cfqq_split_coop(cfqq);
2469
2470 /*
2471 * store what was left of this slice, if the queue idled/timed out
2472 */
2473 if (timed_out) {
2474 if (cfq_cfqq_slice_new(cfqq))
2475 cfqq->slice_resid = cfq_scaled_cfqq_slice(cfqd, cfqq);
2476 else
2477 cfqq->slice_resid = cfqq->slice_end - jiffies;
2478 cfq_log_cfqq(cfqd, cfqq, "resid=%ld", cfqq->slice_resid);
2479 }
2480
2481 cfq_group_served(cfqd, cfqq->cfqg, cfqq);
2482
2483 if (cfq_cfqq_on_rr(cfqq) && RB_EMPTY_ROOT(&cfqq->sort_list))
2484 cfq_del_cfqq_rr(cfqd, cfqq);
2485
2486 cfq_resort_rr_list(cfqd, cfqq);
2487
2488 if (cfqq == cfqd->active_queue)
2489 cfqd->active_queue = NULL;
2490
2491 if (cfqd->active_cic) {
2492 put_io_context(cfqd->active_cic->icq.ioc);
2493 cfqd->active_cic = NULL;
2494 }
2495}
2496
2497static inline void cfq_slice_expired(struct cfq_data *cfqd, bool timed_out)
2498{
2499 struct cfq_queue *cfqq = cfqd->active_queue;
2500
2501 if (cfqq)
2502 __cfq_slice_expired(cfqd, cfqq, timed_out);
2503}
2504
2505/*
2506 * Get next queue for service. Unless we have a queue preemption,
2507 * we'll simply select the first cfqq in the service tree.
2508 */
2509static struct cfq_queue *cfq_get_next_queue(struct cfq_data *cfqd)
2510{
2511 struct cfq_rb_root *st = st_for(cfqd->serving_group,
2512 cfqd->serving_wl_class, cfqd->serving_wl_type);
2513
2514 if (!cfqd->rq_queued)
2515 return NULL;
2516
2517 /* There is nothing to dispatch */
2518 if (!st)
2519 return NULL;
2520 if (RB_EMPTY_ROOT(&st->rb))
2521 return NULL;
2522 return cfq_rb_first(st);
2523}
2524
2525static struct cfq_queue *cfq_get_next_queue_forced(struct cfq_data *cfqd)
2526{
2527 struct cfq_group *cfqg;
2528 struct cfq_queue *cfqq;
2529 int i, j;
2530 struct cfq_rb_root *st;
2531
2532 if (!cfqd->rq_queued)
2533 return NULL;
2534
2535 cfqg = cfq_get_next_cfqg(cfqd);
2536 if (!cfqg)
2537 return NULL;
2538
2539 for_each_cfqg_st(cfqg, i, j, st)
2540 if ((cfqq = cfq_rb_first(st)) != NULL)
2541 return cfqq;
2542 return NULL;
2543}
2544
2545/*
2546 * Get and set a new active queue for service.
2547 */
2548static struct cfq_queue *cfq_set_active_queue(struct cfq_data *cfqd,
2549 struct cfq_queue *cfqq)
2550{
2551 if (!cfqq)
2552 cfqq = cfq_get_next_queue(cfqd);
2553
2554 __cfq_set_active_queue(cfqd, cfqq);
2555 return cfqq;
2556}
2557
2558static inline sector_t cfq_dist_from_last(struct cfq_data *cfqd,
2559 struct request *rq)
2560{
2561 if (blk_rq_pos(rq) >= cfqd->last_position)
2562 return blk_rq_pos(rq) - cfqd->last_position;
2563 else
2564 return cfqd->last_position - blk_rq_pos(rq);
2565}
2566
2567static inline int cfq_rq_close(struct cfq_data *cfqd, struct cfq_queue *cfqq,
2568 struct request *rq)
2569{
2570 return cfq_dist_from_last(cfqd, rq) <= CFQQ_CLOSE_THR;
2571}
2572
2573static struct cfq_queue *cfqq_close(struct cfq_data *cfqd,
2574 struct cfq_queue *cur_cfqq)
2575{
2576 struct rb_root *root = &cfqd->prio_trees[cur_cfqq->org_ioprio];
2577 struct rb_node *parent, *node;
2578 struct cfq_queue *__cfqq;
2579 sector_t sector = cfqd->last_position;
2580
2581 if (RB_EMPTY_ROOT(root))
2582 return NULL;
2583
2584 /*
2585 * First, if we find a request starting at the end of the last
2586 * request, choose it.
2587 */
2588 __cfqq = cfq_prio_tree_lookup(cfqd, root, sector, &parent, NULL);
2589 if (__cfqq)
2590 return __cfqq;
2591
2592 /*
2593 * If the exact sector wasn't found, the parent of the NULL leaf
2594 * will contain the closest sector.
2595 */
2596 __cfqq = rb_entry(parent, struct cfq_queue, p_node);
2597 if (cfq_rq_close(cfqd, cur_cfqq, __cfqq->next_rq))
2598 return __cfqq;
2599
2600 if (blk_rq_pos(__cfqq->next_rq) < sector)
2601 node = rb_next(&__cfqq->p_node);
2602 else
2603 node = rb_prev(&__cfqq->p_node);
2604 if (!node)
2605 return NULL;
2606
2607 __cfqq = rb_entry(node, struct cfq_queue, p_node);
2608 if (cfq_rq_close(cfqd, cur_cfqq, __cfqq->next_rq))
2609 return __cfqq;
2610
2611 return NULL;
2612}
2613
2614/*
2615 * cfqd - obvious
2616 * cur_cfqq - passed in so that we don't decide that the current queue is
2617 * closely cooperating with itself.
2618 *
2619 * So, basically we're assuming that that cur_cfqq has dispatched at least
2620 * one request, and that cfqd->last_position reflects a position on the disk
2621 * associated with the I/O issued by cur_cfqq. I'm not sure this is a valid
2622 * assumption.
2623 */
2624static struct cfq_queue *cfq_close_cooperator(struct cfq_data *cfqd,
2625 struct cfq_queue *cur_cfqq)
2626{
2627 struct cfq_queue *cfqq;
2628
2629 if (cfq_class_idle(cur_cfqq))
2630 return NULL;
2631 if (!cfq_cfqq_sync(cur_cfqq))
2632 return NULL;
2633 if (CFQQ_SEEKY(cur_cfqq))
2634 return NULL;
2635
2636 /*
2637 * Don't search priority tree if it's the only queue in the group.
2638 */
2639 if (cur_cfqq->cfqg->nr_cfqq == 1)
2640 return NULL;
2641
2642 /*
2643 * We should notice if some of the queues are cooperating, eg
2644 * working closely on the same area of the disk. In that case,
2645 * we can group them together and don't waste time idling.
2646 */
2647 cfqq = cfqq_close(cfqd, cur_cfqq);
2648 if (!cfqq)
2649 return NULL;
2650
2651 /* If new queue belongs to different cfq_group, don't choose it */
2652 if (cur_cfqq->cfqg != cfqq->cfqg)
2653 return NULL;
2654
2655 /*
2656 * It only makes sense to merge sync queues.
2657 */
2658 if (!cfq_cfqq_sync(cfqq))
2659 return NULL;
2660 if (CFQQ_SEEKY(cfqq))
2661 return NULL;
2662
2663 /*
2664 * Do not merge queues of different priority classes
2665 */
2666 if (cfq_class_rt(cfqq) != cfq_class_rt(cur_cfqq))
2667 return NULL;
2668
2669 return cfqq;
2670}
2671
2672/*
2673 * Determine whether we should enforce idle window for this queue.
2674 */
2675
2676static bool cfq_should_idle(struct cfq_data *cfqd, struct cfq_queue *cfqq)
2677{
2678 enum wl_class_t wl_class = cfqq_class(cfqq);
2679 struct cfq_rb_root *st = cfqq->service_tree;
2680
2681 BUG_ON(!st);
2682 BUG_ON(!st->count);
2683
2684 if (!cfqd->cfq_slice_idle)
2685 return false;
2686
2687 /* We never do for idle class queues. */
2688 if (wl_class == IDLE_WORKLOAD)
2689 return false;
2690
2691 /* We do for queues that were marked with idle window flag. */
2692 if (cfq_cfqq_idle_window(cfqq) &&
2693 !(blk_queue_nonrot(cfqd->queue) && cfqd->hw_tag))
2694 return true;
2695
2696 /*
2697 * Otherwise, we do only if they are the last ones
2698 * in their service tree.
2699 */
2700 if (st->count == 1 && cfq_cfqq_sync(cfqq) &&
2701 !cfq_io_thinktime_big(cfqd, &st->ttime, false))
2702 return true;
2703 cfq_log_cfqq(cfqd, cfqq, "Not idling. st->count:%d", st->count);
2704 return false;
2705}
2706
2707static void cfq_arm_slice_timer(struct cfq_data *cfqd)
2708{
2709 struct cfq_queue *cfqq = cfqd->active_queue;
2710 struct cfq_io_cq *cic;
2711 unsigned long sl, group_idle = 0;
2712
2713 /*
2714 * SSD device without seek penalty, disable idling. But only do so
2715 * for devices that support queuing, otherwise we still have a problem
2716 * with sync vs async workloads.
2717 */
2718 if (blk_queue_nonrot(cfqd->queue) && cfqd->hw_tag)
2719 return;
2720
2721 WARN_ON(!RB_EMPTY_ROOT(&cfqq->sort_list));
2722 WARN_ON(cfq_cfqq_slice_new(cfqq));
2723
2724 /*
2725 * idle is disabled, either manually or by past process history
2726 */
2727 if (!cfq_should_idle(cfqd, cfqq)) {
2728 /* no queue idling. Check for group idling */
2729 if (cfqd->cfq_group_idle)
2730 group_idle = cfqd->cfq_group_idle;
2731 else
2732 return;
2733 }
2734
2735 /*
2736 * still active requests from this queue, don't idle
2737 */
2738 if (cfqq->dispatched)
2739 return;
2740
2741 /*
2742 * task has exited, don't wait
2743 */
2744 cic = cfqd->active_cic;
2745 if (!cic || !atomic_read(&cic->icq.ioc->active_ref))
2746 return;
2747
2748 /*
2749 * If our average think time is larger than the remaining time
2750 * slice, then don't idle. This avoids overrunning the allotted
2751 * time slice.
2752 */
2753 if (sample_valid(cic->ttime.ttime_samples) &&
2754 (cfqq->slice_end - jiffies < cic->ttime.ttime_mean)) {
2755 cfq_log_cfqq(cfqd, cfqq, "Not idling. think_time:%lu",
2756 cic->ttime.ttime_mean);
2757 return;
2758 }
2759
2760 /* There are other queues in the group, don't do group idle */
2761 if (group_idle && cfqq->cfqg->nr_cfqq > 1)
2762 return;
2763
2764 cfq_mark_cfqq_wait_request(cfqq);
2765
2766 if (group_idle)
2767 sl = cfqd->cfq_group_idle;
2768 else
2769 sl = cfqd->cfq_slice_idle;
2770
2771 mod_timer(&cfqd->idle_slice_timer, jiffies + sl);
2772 cfqg_stats_set_start_idle_time(cfqq->cfqg);
2773 cfq_log_cfqq(cfqd, cfqq, "arm_idle: %lu group_idle: %d", sl,
2774 group_idle ? 1 : 0);
2775}
2776
2777/*
2778 * Move request from internal lists to the request queue dispatch list.
2779 */
2780static void cfq_dispatch_insert(struct request_queue *q, struct request *rq)
2781{
2782 struct cfq_data *cfqd = q->elevator->elevator_data;
2783 struct cfq_queue *cfqq = RQ_CFQQ(rq);
2784
2785 cfq_log_cfqq(cfqd, cfqq, "dispatch_insert");
2786
2787 cfqq->next_rq = cfq_find_next_rq(cfqd, cfqq, rq);
2788 cfq_remove_request(rq);
2789 cfqq->dispatched++;
2790 (RQ_CFQG(rq))->dispatched++;
2791 elv_dispatch_sort(q, rq);
2792
2793 cfqd->rq_in_flight[cfq_cfqq_sync(cfqq)]++;
2794 cfqq->nr_sectors += blk_rq_sectors(rq);
2795 cfqg_stats_update_dispatch(cfqq->cfqg, blk_rq_bytes(rq), rq->cmd_flags);
2796}
2797
2798/*
2799 * return expired entry, or NULL to just start from scratch in rbtree
2800 */
2801static struct request *cfq_check_fifo(struct cfq_queue *cfqq)
2802{
2803 struct request *rq = NULL;
2804
2805 if (cfq_cfqq_fifo_expire(cfqq))
2806 return NULL;
2807
2808 cfq_mark_cfqq_fifo_expire(cfqq);
2809
2810 if (list_empty(&cfqq->fifo))
2811 return NULL;
2812
2813 rq = rq_entry_fifo(cfqq->fifo.next);
2814 if (time_before(jiffies, rq->fifo_time))
2815 rq = NULL;
2816
2817 cfq_log_cfqq(cfqq->cfqd, cfqq, "fifo=%p", rq);
2818 return rq;
2819}
2820
2821static inline int
2822cfq_prio_to_maxrq(struct cfq_data *cfqd, struct cfq_queue *cfqq)
2823{
2824 const int base_rq = cfqd->cfq_slice_async_rq;
2825
2826 WARN_ON(cfqq->ioprio >= IOPRIO_BE_NR);
2827
2828 return 2 * base_rq * (IOPRIO_BE_NR - cfqq->ioprio);
2829}
2830
2831/*
2832 * Must be called with the queue_lock held.
2833 */
2834static int cfqq_process_refs(struct cfq_queue *cfqq)
2835{
2836 int process_refs, io_refs;
2837
2838 io_refs = cfqq->allocated[READ] + cfqq->allocated[WRITE];
2839 process_refs = cfqq->ref - io_refs;
2840 BUG_ON(process_refs < 0);
2841 return process_refs;
2842}
2843
2844static void cfq_setup_merge(struct cfq_queue *cfqq, struct cfq_queue *new_cfqq)
2845{
2846 int process_refs, new_process_refs;
2847 struct cfq_queue *__cfqq;
2848
2849 /*
2850 * If there are no process references on the new_cfqq, then it is
2851 * unsafe to follow the ->new_cfqq chain as other cfqq's in the
2852 * chain may have dropped their last reference (not just their
2853 * last process reference).
2854 */
2855 if (!cfqq_process_refs(new_cfqq))
2856 return;
2857
2858 /* Avoid a circular list and skip interim queue merges */
2859 while ((__cfqq = new_cfqq->new_cfqq)) {
2860 if (__cfqq == cfqq)
2861 return;
2862 new_cfqq = __cfqq;
2863 }
2864
2865 process_refs = cfqq_process_refs(cfqq);
2866 new_process_refs = cfqq_process_refs(new_cfqq);
2867 /*
2868 * If the process for the cfqq has gone away, there is no
2869 * sense in merging the queues.
2870 */
2871 if (process_refs == 0 || new_process_refs == 0)
2872 return;
2873
2874 /*
2875 * Merge in the direction of the lesser amount of work.
2876 */
2877 if (new_process_refs >= process_refs) {
2878 cfqq->new_cfqq = new_cfqq;
2879 new_cfqq->ref += process_refs;
2880 } else {
2881 new_cfqq->new_cfqq = cfqq;
2882 cfqq->ref += new_process_refs;
2883 }
2884}
2885
2886static enum wl_type_t cfq_choose_wl_type(struct cfq_data *cfqd,
2887 struct cfq_group *cfqg, enum wl_class_t wl_class)
2888{
2889 struct cfq_queue *queue;
2890 int i;
2891 bool key_valid = false;
2892 unsigned long lowest_key = 0;
2893 enum wl_type_t cur_best = SYNC_NOIDLE_WORKLOAD;
2894
2895 for (i = 0; i <= SYNC_WORKLOAD; ++i) {
2896 /* select the one with lowest rb_key */
2897 queue = cfq_rb_first(st_for(cfqg, wl_class, i));
2898 if (queue &&
2899 (!key_valid || time_before(queue->rb_key, lowest_key))) {
2900 lowest_key = queue->rb_key;
2901 cur_best = i;
2902 key_valid = true;
2903 }
2904 }
2905
2906 return cur_best;
2907}
2908
2909static void
2910choose_wl_class_and_type(struct cfq_data *cfqd, struct cfq_group *cfqg)
2911{
2912 unsigned slice;
2913 unsigned count;
2914 struct cfq_rb_root *st;
2915 unsigned group_slice;
2916 enum wl_class_t original_class = cfqd->serving_wl_class;
2917
2918 /* Choose next priority. RT > BE > IDLE */
2919 if (cfq_group_busy_queues_wl(RT_WORKLOAD, cfqd, cfqg))
2920 cfqd->serving_wl_class = RT_WORKLOAD;
2921 else if (cfq_group_busy_queues_wl(BE_WORKLOAD, cfqd, cfqg))
2922 cfqd->serving_wl_class = BE_WORKLOAD;
2923 else {
2924 cfqd->serving_wl_class = IDLE_WORKLOAD;
2925 cfqd->workload_expires = jiffies + 1;
2926 return;
2927 }
2928
2929 if (original_class != cfqd->serving_wl_class)
2930 goto new_workload;
2931
2932 /*
2933 * For RT and BE, we have to choose also the type
2934 * (SYNC, SYNC_NOIDLE, ASYNC), and to compute a workload
2935 * expiration time
2936 */
2937 st = st_for(cfqg, cfqd->serving_wl_class, cfqd->serving_wl_type);
2938 count = st->count;
2939
2940 /*
2941 * check workload expiration, and that we still have other queues ready
2942 */
2943 if (count && !time_after(jiffies, cfqd->workload_expires))
2944 return;
2945
2946new_workload:
2947 /* otherwise select new workload type */
2948 cfqd->serving_wl_type = cfq_choose_wl_type(cfqd, cfqg,
2949 cfqd->serving_wl_class);
2950 st = st_for(cfqg, cfqd->serving_wl_class, cfqd->serving_wl_type);
2951 count = st->count;
2952
2953 /*
2954 * the workload slice is computed as a fraction of target latency
2955 * proportional to the number of queues in that workload, over
2956 * all the queues in the same priority class
2957 */
2958 group_slice = cfq_group_slice(cfqd, cfqg);
2959
2960 slice = group_slice * count /
2961 max_t(unsigned, cfqg->busy_queues_avg[cfqd->serving_wl_class],
2962 cfq_group_busy_queues_wl(cfqd->serving_wl_class, cfqd,
2963 cfqg));
2964
2965 if (cfqd->serving_wl_type == ASYNC_WORKLOAD) {
2966 unsigned int tmp;
2967
2968 /*
2969 * Async queues are currently system wide. Just taking
2970 * proportion of queues with-in same group will lead to higher
2971 * async ratio system wide as generally root group is going
2972 * to have higher weight. A more accurate thing would be to
2973 * calculate system wide asnc/sync ratio.
2974 */
2975 tmp = cfqd->cfq_target_latency *
2976 cfqg_busy_async_queues(cfqd, cfqg);
2977 tmp = tmp/cfqd->busy_queues;
2978 slice = min_t(unsigned, slice, tmp);
2979
2980 /* async workload slice is scaled down according to
2981 * the sync/async slice ratio. */
2982 slice = slice * cfqd->cfq_slice[0] / cfqd->cfq_slice[1];
2983 } else
2984 /* sync workload slice is at least 2 * cfq_slice_idle */
2985 slice = max(slice, 2 * cfqd->cfq_slice_idle);
2986
2987 slice = max_t(unsigned, slice, CFQ_MIN_TT);
2988 cfq_log(cfqd, "workload slice:%d", slice);
2989 cfqd->workload_expires = jiffies + slice;
2990}
2991
2992static struct cfq_group *cfq_get_next_cfqg(struct cfq_data *cfqd)
2993{
2994 struct cfq_rb_root *st = &cfqd->grp_service_tree;
2995 struct cfq_group *cfqg;
2996
2997 if (RB_EMPTY_ROOT(&st->rb))
2998 return NULL;
2999 cfqg = cfq_rb_first_group(st);
3000 update_min_vdisktime(st);
3001 return cfqg;
3002}
3003
3004static void cfq_choose_cfqg(struct cfq_data *cfqd)
3005{
3006 struct cfq_group *cfqg = cfq_get_next_cfqg(cfqd);
3007
3008 cfqd->serving_group = cfqg;
3009
3010 /* Restore the workload type data */
3011 if (cfqg->saved_wl_slice) {
3012 cfqd->workload_expires = jiffies + cfqg->saved_wl_slice;
3013 cfqd->serving_wl_type = cfqg->saved_wl_type;
3014 cfqd->serving_wl_class = cfqg->saved_wl_class;
3015 } else
3016 cfqd->workload_expires = jiffies - 1;
3017
3018 choose_wl_class_and_type(cfqd, cfqg);
3019}
3020
3021/*
3022 * Select a queue for service. If we have a current active queue,
3023 * check whether to continue servicing it, or retrieve and set a new one.
3024 */
3025static struct cfq_queue *cfq_select_queue(struct cfq_data *cfqd)
3026{
3027 struct cfq_queue *cfqq, *new_cfqq = NULL;
3028
3029 cfqq = cfqd->active_queue;
3030 if (!cfqq)
3031 goto new_queue;
3032
3033 if (!cfqd->rq_queued)
3034 return NULL;
3035
3036 /*
3037 * We were waiting for group to get backlogged. Expire the queue
3038 */
3039 if (cfq_cfqq_wait_busy(cfqq) && !RB_EMPTY_ROOT(&cfqq->sort_list))
3040 goto expire;
3041
3042 /*
3043 * The active queue has run out of time, expire it and select new.
3044 */
3045 if (cfq_slice_used(cfqq) && !cfq_cfqq_must_dispatch(cfqq)) {
3046 /*
3047 * If slice had not expired at the completion of last request
3048 * we might not have turned on wait_busy flag. Don't expire
3049 * the queue yet. Allow the group to get backlogged.
3050 *
3051 * The very fact that we have used the slice, that means we
3052 * have been idling all along on this queue and it should be
3053 * ok to wait for this request to complete.
3054 */
3055 if (cfqq->cfqg->nr_cfqq == 1 && RB_EMPTY_ROOT(&cfqq->sort_list)
3056 && cfqq->dispatched && cfq_should_idle(cfqd, cfqq)) {
3057 cfqq = NULL;
3058 goto keep_queue;
3059 } else
3060 goto check_group_idle;
3061 }
3062
3063 /*
3064 * The active queue has requests and isn't expired, allow it to
3065 * dispatch.
3066 */
3067 if (!RB_EMPTY_ROOT(&cfqq->sort_list))
3068 goto keep_queue;
3069
3070 /*
3071 * If another queue has a request waiting within our mean seek
3072 * distance, let it run. The expire code will check for close
3073 * cooperators and put the close queue at the front of the service
3074 * tree. If possible, merge the expiring queue with the new cfqq.
3075 */
3076 new_cfqq = cfq_close_cooperator(cfqd, cfqq);
3077 if (new_cfqq) {
3078 if (!cfqq->new_cfqq)
3079 cfq_setup_merge(cfqq, new_cfqq);
3080 goto expire;
3081 }
3082
3083 /*
3084 * No requests pending. If the active queue still has requests in
3085 * flight or is idling for a new request, allow either of these
3086 * conditions to happen (or time out) before selecting a new queue.
3087 */
3088 if (timer_pending(&cfqd->idle_slice_timer)) {
3089 cfqq = NULL;
3090 goto keep_queue;
3091 }
3092
3093 /*
3094 * This is a deep seek queue, but the device is much faster than
3095 * the queue can deliver, don't idle
3096 **/
3097 if (CFQQ_SEEKY(cfqq) && cfq_cfqq_idle_window(cfqq) &&
3098 (cfq_cfqq_slice_new(cfqq) ||
3099 (cfqq->slice_end - jiffies > jiffies - cfqq->slice_start))) {
3100 cfq_clear_cfqq_deep(cfqq);
3101 cfq_clear_cfqq_idle_window(cfqq);
3102 }
3103
3104 if (cfqq->dispatched && cfq_should_idle(cfqd, cfqq)) {
3105 cfqq = NULL;
3106 goto keep_queue;
3107 }
3108
3109 /*
3110 * If group idle is enabled and there are requests dispatched from
3111 * this group, wait for requests to complete.
3112 */
3113check_group_idle:
3114 if (cfqd->cfq_group_idle && cfqq->cfqg->nr_cfqq == 1 &&
3115 cfqq->cfqg->dispatched &&
3116 !cfq_io_thinktime_big(cfqd, &cfqq->cfqg->ttime, true)) {
3117 cfqq = NULL;
3118 goto keep_queue;
3119 }
3120
3121expire:
3122 cfq_slice_expired(cfqd, 0);
3123new_queue:
3124 /*
3125 * Current queue expired. Check if we have to switch to a new
3126 * service tree
3127 */
3128 if (!new_cfqq)
3129 cfq_choose_cfqg(cfqd);
3130
3131 cfqq = cfq_set_active_queue(cfqd, new_cfqq);
3132keep_queue:
3133 return cfqq;
3134}
3135
3136static int __cfq_forced_dispatch_cfqq(struct cfq_queue *cfqq)
3137{
3138 int dispatched = 0;
3139
3140 while (cfqq->next_rq) {
3141 cfq_dispatch_insert(cfqq->cfqd->queue, cfqq->next_rq);
3142 dispatched++;
3143 }
3144
3145 BUG_ON(!list_empty(&cfqq->fifo));
3146
3147 /* By default cfqq is not expired if it is empty. Do it explicitly */
3148 __cfq_slice_expired(cfqq->cfqd, cfqq, 0);
3149 return dispatched;
3150}
3151
3152/*
3153 * Drain our current requests. Used for barriers and when switching
3154 * io schedulers on-the-fly.
3155 */
3156static int cfq_forced_dispatch(struct cfq_data *cfqd)
3157{
3158 struct cfq_queue *cfqq;
3159 int dispatched = 0;
3160
3161 /* Expire the timeslice of the current active queue first */
3162 cfq_slice_expired(cfqd, 0);
3163 while ((cfqq = cfq_get_next_queue_forced(cfqd)) != NULL) {
3164 __cfq_set_active_queue(cfqd, cfqq);
3165 dispatched += __cfq_forced_dispatch_cfqq(cfqq);
3166 }
3167
3168 BUG_ON(cfqd->busy_queues);
3169
3170 cfq_log(cfqd, "forced_dispatch=%d", dispatched);
3171 return dispatched;
3172}
3173
3174static inline bool cfq_slice_used_soon(struct cfq_data *cfqd,
3175 struct cfq_queue *cfqq)
3176{
3177 /* the queue hasn't finished any request, can't estimate */
3178 if (cfq_cfqq_slice_new(cfqq))
3179 return true;
3180 if (time_after(jiffies + cfqd->cfq_slice_idle * cfqq->dispatched,
3181 cfqq->slice_end))
3182 return true;
3183
3184 return false;
3185}
3186
3187static bool cfq_may_dispatch(struct cfq_data *cfqd, struct cfq_queue *cfqq)
3188{
3189 unsigned int max_dispatch;
3190
3191 /*
3192 * Drain async requests before we start sync IO
3193 */
3194 if (cfq_should_idle(cfqd, cfqq) && cfqd->rq_in_flight[BLK_RW_ASYNC])
3195 return false;
3196
3197 /*
3198 * If this is an async queue and we have sync IO in flight, let it wait
3199 */
3200 if (cfqd->rq_in_flight[BLK_RW_SYNC] && !cfq_cfqq_sync(cfqq))
3201 return false;
3202
3203 max_dispatch = max_t(unsigned int, cfqd->cfq_quantum / 2, 1);
3204 if (cfq_class_idle(cfqq))
3205 max_dispatch = 1;
3206
3207 /*
3208 * Does this cfqq already have too much IO in flight?
3209 */
3210 if (cfqq->dispatched >= max_dispatch) {
3211 bool promote_sync = false;
3212 /*
3213 * idle queue must always only have a single IO in flight
3214 */
3215 if (cfq_class_idle(cfqq))
3216 return false;
3217
3218 /*
3219 * If there is only one sync queue
3220 * we can ignore async queue here and give the sync
3221 * queue no dispatch limit. The reason is a sync queue can
3222 * preempt async queue, limiting the sync queue doesn't make
3223 * sense. This is useful for aiostress test.
3224 */
3225 if (cfq_cfqq_sync(cfqq) && cfqd->busy_sync_queues == 1)
3226 promote_sync = true;
3227
3228 /*
3229 * We have other queues, don't allow more IO from this one
3230 */
3231 if (cfqd->busy_queues > 1 && cfq_slice_used_soon(cfqd, cfqq) &&
3232 !promote_sync)
3233 return false;
3234
3235 /*
3236 * Sole queue user, no limit
3237 */
3238 if (cfqd->busy_queues == 1 || promote_sync)
3239 max_dispatch = -1;
3240 else
3241 /*
3242 * Normally we start throttling cfqq when cfq_quantum/2
3243 * requests have been dispatched. But we can drive
3244 * deeper queue depths at the beginning of slice
3245 * subjected to upper limit of cfq_quantum.
3246 * */
3247 max_dispatch = cfqd->cfq_quantum;
3248 }
3249
3250 /*
3251 * Async queues must wait a bit before being allowed dispatch.
3252 * We also ramp up the dispatch depth gradually for async IO,
3253 * based on the last sync IO we serviced
3254 */
3255 if (!cfq_cfqq_sync(cfqq) && cfqd->cfq_latency) {
3256 unsigned long last_sync = jiffies - cfqd->last_delayed_sync;
3257 unsigned int depth;
3258
3259 depth = last_sync / cfqd->cfq_slice[1];
3260 if (!depth && !cfqq->dispatched)
3261 depth = 1;
3262 if (depth < max_dispatch)
3263 max_dispatch = depth;
3264 }
3265
3266 /*
3267 * If we're below the current max, allow a dispatch
3268 */
3269 return cfqq->dispatched < max_dispatch;
3270}
3271
3272/*
3273 * Dispatch a request from cfqq, moving them to the request queue
3274 * dispatch list.
3275 */
3276static bool cfq_dispatch_request(struct cfq_data *cfqd, struct cfq_queue *cfqq)
3277{
3278 struct request *rq;
3279
3280 BUG_ON(RB_EMPTY_ROOT(&cfqq->sort_list));
3281
3282 if (!cfq_may_dispatch(cfqd, cfqq))
3283 return false;
3284
3285 /*
3286 * follow expired path, else get first next available
3287 */
3288 rq = cfq_check_fifo(cfqq);
3289 if (!rq)
3290 rq = cfqq->next_rq;
3291
3292 /*
3293 * insert request into driver dispatch list
3294 */
3295 cfq_dispatch_insert(cfqd->queue, rq);
3296
3297 if (!cfqd->active_cic) {
3298 struct cfq_io_cq *cic = RQ_CIC(rq);
3299
3300 atomic_long_inc(&cic->icq.ioc->refcount);
3301 cfqd->active_cic = cic;
3302 }
3303
3304 return true;
3305}
3306
3307/*
3308 * Find the cfqq that we need to service and move a request from that to the
3309 * dispatch list
3310 */
3311static int cfq_dispatch_requests(struct request_queue *q, int force)
3312{
3313 struct cfq_data *cfqd = q->elevator->elevator_data;
3314 struct cfq_queue *cfqq;
3315
3316 if (!cfqd->busy_queues)
3317 return 0;
3318
3319 if (unlikely(force))
3320 return cfq_forced_dispatch(cfqd);
3321
3322 cfqq = cfq_select_queue(cfqd);
3323 if (!cfqq)
3324 return 0;
3325
3326 /*
3327 * Dispatch a request from this cfqq, if it is allowed
3328 */
3329 if (!cfq_dispatch_request(cfqd, cfqq))
3330 return 0;
3331
3332 cfqq->slice_dispatch++;
3333 cfq_clear_cfqq_must_dispatch(cfqq);
3334
3335 /*
3336 * expire an async queue immediately if it has used up its slice. idle
3337 * queue always expire after 1 dispatch round.
3338 */
3339 if (cfqd->busy_queues > 1 && ((!cfq_cfqq_sync(cfqq) &&
3340 cfqq->slice_dispatch >= cfq_prio_to_maxrq(cfqd, cfqq)) ||
3341 cfq_class_idle(cfqq))) {
3342 cfqq->slice_end = jiffies + 1;
3343 cfq_slice_expired(cfqd, 0);
3344 }
3345
3346 cfq_log_cfqq(cfqd, cfqq, "dispatched a request");
3347 return 1;
3348}
3349
3350/*
3351 * task holds one reference to the queue, dropped when task exits. each rq
3352 * in-flight on this queue also holds a reference, dropped when rq is freed.
3353 *
3354 * Each cfq queue took a reference on the parent group. Drop it now.
3355 * queue lock must be held here.
3356 */
3357static void cfq_put_queue(struct cfq_queue *cfqq)
3358{
3359 struct cfq_data *cfqd = cfqq->cfqd;
3360 struct cfq_group *cfqg;
3361
3362 BUG_ON(cfqq->ref <= 0);
3363
3364 cfqq->ref--;
3365 if (cfqq->ref)
3366 return;
3367
3368 cfq_log_cfqq(cfqd, cfqq, "put_queue");
3369 BUG_ON(rb_first(&cfqq->sort_list));
3370 BUG_ON(cfqq->allocated[READ] + cfqq->allocated[WRITE]);
3371 cfqg = cfqq->cfqg;
3372
3373 if (unlikely(cfqd->active_queue == cfqq)) {
3374 __cfq_slice_expired(cfqd, cfqq, 0);
3375 cfq_schedule_dispatch(cfqd);
3376 }
3377
3378 BUG_ON(cfq_cfqq_on_rr(cfqq));
3379 kmem_cache_free(cfq_pool, cfqq);
3380 cfqg_put(cfqg);
3381}
3382
3383static void cfq_put_cooperator(struct cfq_queue *cfqq)
3384{
3385 struct cfq_queue *__cfqq, *next;
3386
3387 /*
3388 * If this queue was scheduled to merge with another queue, be
3389 * sure to drop the reference taken on that queue (and others in
3390 * the merge chain). See cfq_setup_merge and cfq_merge_cfqqs.
3391 */
3392 __cfqq = cfqq->new_cfqq;
3393 while (__cfqq) {
3394 if (__cfqq == cfqq) {
3395 WARN(1, "cfqq->new_cfqq loop detected\n");
3396 break;
3397 }
3398 next = __cfqq->new_cfqq;
3399 cfq_put_queue(__cfqq);
3400 __cfqq = next;
3401 }
3402}
3403
3404static void cfq_exit_cfqq(struct cfq_data *cfqd, struct cfq_queue *cfqq)
3405{
3406 if (unlikely(cfqq == cfqd->active_queue)) {
3407 __cfq_slice_expired(cfqd, cfqq, 0);
3408 cfq_schedule_dispatch(cfqd);
3409 }
3410
3411 cfq_put_cooperator(cfqq);
3412
3413 cfq_put_queue(cfqq);
3414}
3415
3416static void cfq_init_icq(struct io_cq *icq)
3417{
3418 struct cfq_io_cq *cic = icq_to_cic(icq);
3419
3420 cic->ttime.last_end_request = jiffies;
3421}
3422
3423static void cfq_exit_icq(struct io_cq *icq)
3424{
3425 struct cfq_io_cq *cic = icq_to_cic(icq);
3426 struct cfq_data *cfqd = cic_to_cfqd(cic);
3427
3428 if (cic->cfqq[BLK_RW_ASYNC]) {
3429 cfq_exit_cfqq(cfqd, cic->cfqq[BLK_RW_ASYNC]);
3430 cic->cfqq[BLK_RW_ASYNC] = NULL;
3431 }
3432
3433 if (cic->cfqq[BLK_RW_SYNC]) {
3434 cfq_exit_cfqq(cfqd, cic->cfqq[BLK_RW_SYNC]);
3435 cic->cfqq[BLK_RW_SYNC] = NULL;
3436 }
3437}
3438
3439static void cfq_init_prio_data(struct cfq_queue *cfqq, struct cfq_io_cq *cic)
3440{
3441 struct task_struct *tsk = current;
3442 int ioprio_class;
3443
3444 if (!cfq_cfqq_prio_changed(cfqq))
3445 return;
3446
3447 ioprio_class = IOPRIO_PRIO_CLASS(cic->ioprio);
3448 switch (ioprio_class) {
3449 default:
3450 printk(KERN_ERR "cfq: bad prio %x\n", ioprio_class);
3451 case IOPRIO_CLASS_NONE:
3452 /*
3453 * no prio set, inherit CPU scheduling settings
3454 */
3455 cfqq->ioprio = task_nice_ioprio(tsk);
3456 cfqq->ioprio_class = task_nice_ioclass(tsk);
3457 break;
3458 case IOPRIO_CLASS_RT:
3459 cfqq->ioprio = IOPRIO_PRIO_DATA(cic->ioprio);
3460 cfqq->ioprio_class = IOPRIO_CLASS_RT;
3461 break;
3462 case IOPRIO_CLASS_BE:
3463 cfqq->ioprio = IOPRIO_PRIO_DATA(cic->ioprio);
3464 cfqq->ioprio_class = IOPRIO_CLASS_BE;
3465 break;
3466 case IOPRIO_CLASS_IDLE:
3467 cfqq->ioprio_class = IOPRIO_CLASS_IDLE;
3468 cfqq->ioprio = 7;
3469 cfq_clear_cfqq_idle_window(cfqq);
3470 break;
3471 }
3472
3473 /*
3474 * keep track of original prio settings in case we have to temporarily
3475 * elevate the priority of this queue
3476 */
3477 cfqq->org_ioprio = cfqq->ioprio;
3478 cfq_clear_cfqq_prio_changed(cfqq);
3479}
3480
3481static void check_ioprio_changed(struct cfq_io_cq *cic, struct bio *bio)
3482{
3483 int ioprio = cic->icq.ioc->ioprio;
3484 struct cfq_data *cfqd = cic_to_cfqd(cic);
3485 struct cfq_queue *cfqq;
3486
3487 /*
3488 * Check whether ioprio has changed. The condition may trigger
3489 * spuriously on a newly created cic but there's no harm.
3490 */
3491 if (unlikely(!cfqd) || likely(cic->ioprio == ioprio))
3492 return;
3493
3494 cfqq = cic->cfqq[BLK_RW_ASYNC];
3495 if (cfqq) {
3496 struct cfq_queue *new_cfqq;
3497 new_cfqq = cfq_get_queue(cfqd, BLK_RW_ASYNC, cic, bio,
3498 GFP_ATOMIC);
3499 if (new_cfqq) {
3500 cic->cfqq[BLK_RW_ASYNC] = new_cfqq;
3501 cfq_put_queue(cfqq);
3502 }
3503 }
3504
3505 cfqq = cic->cfqq[BLK_RW_SYNC];
3506 if (cfqq)
3507 cfq_mark_cfqq_prio_changed(cfqq);
3508
3509 cic->ioprio = ioprio;
3510}
3511
3512static void cfq_init_cfqq(struct cfq_data *cfqd, struct cfq_queue *cfqq,
3513 pid_t pid, bool is_sync)
3514{
3515 RB_CLEAR_NODE(&cfqq->rb_node);
3516 RB_CLEAR_NODE(&cfqq->p_node);
3517 INIT_LIST_HEAD(&cfqq->fifo);
3518
3519 cfqq->ref = 0;
3520 cfqq->cfqd = cfqd;
3521
3522 cfq_mark_cfqq_prio_changed(cfqq);
3523
3524 if (is_sync) {
3525 if (!cfq_class_idle(cfqq))
3526 cfq_mark_cfqq_idle_window(cfqq);
3527 cfq_mark_cfqq_sync(cfqq);
3528 }
3529 cfqq->pid = pid;
3530}
3531
3532#ifdef CONFIG_CFQ_GROUP_IOSCHED
3533static void check_blkcg_changed(struct cfq_io_cq *cic, struct bio *bio)
3534{
3535 struct cfq_data *cfqd = cic_to_cfqd(cic);
3536 struct cfq_queue *sync_cfqq;
3537 uint64_t id;
3538
3539 rcu_read_lock();
3540 id = bio_blkcg(bio)->id;
3541 rcu_read_unlock();
3542
3543 /*
3544 * Check whether blkcg has changed. The condition may trigger
3545 * spuriously on a newly created cic but there's no harm.
3546 */
3547 if (unlikely(!cfqd) || likely(cic->blkcg_id == id))
3548 return;
3549
3550 sync_cfqq = cic_to_cfqq(cic, 1);
3551 if (sync_cfqq) {
3552 /*
3553 * Drop reference to sync queue. A new sync queue will be
3554 * assigned in new group upon arrival of a fresh request.
3555 */
3556 cfq_log_cfqq(cfqd, sync_cfqq, "changed cgroup");
3557 cic_set_cfqq(cic, NULL, 1);
3558 cfq_put_queue(sync_cfqq);
3559 }
3560
3561 cic->blkcg_id = id;
3562}
3563#else
3564static inline void check_blkcg_changed(struct cfq_io_cq *cic, struct bio *bio) { }
3565#endif /* CONFIG_CFQ_GROUP_IOSCHED */
3566
3567static struct cfq_queue *
3568cfq_find_alloc_queue(struct cfq_data *cfqd, bool is_sync, struct cfq_io_cq *cic,
3569 struct bio *bio, gfp_t gfp_mask)
3570{
3571 struct blkcg *blkcg;
3572 struct cfq_queue *cfqq, *new_cfqq = NULL;
3573 struct cfq_group *cfqg;
3574
3575retry:
3576 rcu_read_lock();
3577
3578 blkcg = bio_blkcg(bio);
3579 cfqg = cfq_lookup_create_cfqg(cfqd, blkcg);
3580 cfqq = cic_to_cfqq(cic, is_sync);
3581
3582 /*
3583 * Always try a new alloc if we fell back to the OOM cfqq
3584 * originally, since it should just be a temporary situation.
3585 */
3586 if (!cfqq || cfqq == &cfqd->oom_cfqq) {
3587 cfqq = NULL;
3588 if (new_cfqq) {
3589 cfqq = new_cfqq;
3590 new_cfqq = NULL;
3591 } else if (gfp_mask & __GFP_WAIT) {
3592 rcu_read_unlock();
3593 spin_unlock_irq(cfqd->queue->queue_lock);
3594 new_cfqq = kmem_cache_alloc_node(cfq_pool,
3595 gfp_mask | __GFP_ZERO,
3596 cfqd->queue->node);
3597 spin_lock_irq(cfqd->queue->queue_lock);
3598 if (new_cfqq)
3599 goto retry;
3600 else
3601 return &cfqd->oom_cfqq;
3602 } else {
3603 cfqq = kmem_cache_alloc_node(cfq_pool,
3604 gfp_mask | __GFP_ZERO,
3605 cfqd->queue->node);
3606 }
3607
3608 if (cfqq) {
3609 cfq_init_cfqq(cfqd, cfqq, current->pid, is_sync);
3610 cfq_init_prio_data(cfqq, cic);
3611 cfq_link_cfqq_cfqg(cfqq, cfqg);
3612 cfq_log_cfqq(cfqd, cfqq, "alloced");
3613 } else
3614 cfqq = &cfqd->oom_cfqq;
3615 }
3616
3617 if (new_cfqq)
3618 kmem_cache_free(cfq_pool, new_cfqq);
3619
3620 rcu_read_unlock();
3621 return cfqq;
3622}
3623
3624static struct cfq_queue **
3625cfq_async_queue_prio(struct cfq_data *cfqd, int ioprio_class, int ioprio)
3626{
3627 switch (ioprio_class) {
3628 case IOPRIO_CLASS_RT:
3629 return &cfqd->async_cfqq[0][ioprio];
3630 case IOPRIO_CLASS_NONE:
3631 ioprio = IOPRIO_NORM;
3632 /* fall through */
3633 case IOPRIO_CLASS_BE:
3634 return &cfqd->async_cfqq[1][ioprio];
3635 case IOPRIO_CLASS_IDLE:
3636 return &cfqd->async_idle_cfqq;
3637 default:
3638 BUG();
3639 }
3640}
3641
3642static struct cfq_queue *
3643cfq_get_queue(struct cfq_data *cfqd, bool is_sync, struct cfq_io_cq *cic,
3644 struct bio *bio, gfp_t gfp_mask)
3645{
3646 const int ioprio_class = IOPRIO_PRIO_CLASS(cic->ioprio);
3647 const int ioprio = IOPRIO_PRIO_DATA(cic->ioprio);
3648 struct cfq_queue **async_cfqq = NULL;
3649 struct cfq_queue *cfqq = NULL;
3650
3651 if (!is_sync) {
3652 async_cfqq = cfq_async_queue_prio(cfqd, ioprio_class, ioprio);
3653 cfqq = *async_cfqq;
3654 }
3655
3656 if (!cfqq)
3657 cfqq = cfq_find_alloc_queue(cfqd, is_sync, cic, bio, gfp_mask);
3658
3659 /*
3660 * pin the queue now that it's allocated, scheduler exit will prune it
3661 */
3662 if (!is_sync && !(*async_cfqq)) {
3663 cfqq->ref++;
3664 *async_cfqq = cfqq;
3665 }
3666
3667 cfqq->ref++;
3668 return cfqq;
3669}
3670
3671static void
3672__cfq_update_io_thinktime(struct cfq_ttime *ttime, unsigned long slice_idle)
3673{
3674 unsigned long elapsed = jiffies - ttime->last_end_request;
3675 elapsed = min(elapsed, 2UL * slice_idle);
3676
3677 ttime->ttime_samples = (7*ttime->ttime_samples + 256) / 8;
3678 ttime->ttime_total = (7*ttime->ttime_total + 256*elapsed) / 8;
3679 ttime->ttime_mean = (ttime->ttime_total + 128) / ttime->ttime_samples;
3680}
3681
3682static void
3683cfq_update_io_thinktime(struct cfq_data *cfqd, struct cfq_queue *cfqq,
3684 struct cfq_io_cq *cic)
3685{
3686 if (cfq_cfqq_sync(cfqq)) {
3687 __cfq_update_io_thinktime(&cic->ttime, cfqd->cfq_slice_idle);
3688 __cfq_update_io_thinktime(&cfqq->service_tree->ttime,
3689 cfqd->cfq_slice_idle);
3690 }
3691#ifdef CONFIG_CFQ_GROUP_IOSCHED
3692 __cfq_update_io_thinktime(&cfqq->cfqg->ttime, cfqd->cfq_group_idle);
3693#endif
3694}
3695
3696static void
3697cfq_update_io_seektime(struct cfq_data *cfqd, struct cfq_queue *cfqq,
3698 struct request *rq)
3699{
3700 sector_t sdist = 0;
3701 sector_t n_sec = blk_rq_sectors(rq);
3702 if (cfqq->last_request_pos) {
3703 if (cfqq->last_request_pos < blk_rq_pos(rq))
3704 sdist = blk_rq_pos(rq) - cfqq->last_request_pos;
3705 else
3706 sdist = cfqq->last_request_pos - blk_rq_pos(rq);
3707 }
3708
3709 cfqq->seek_history <<= 1;
3710 if (blk_queue_nonrot(cfqd->queue))
3711 cfqq->seek_history |= (n_sec < CFQQ_SECT_THR_NONROT);
3712 else
3713 cfqq->seek_history |= (sdist > CFQQ_SEEK_THR);
3714}
3715
3716/*
3717 * Disable idle window if the process thinks too long or seeks so much that
3718 * it doesn't matter
3719 */
3720static void
3721cfq_update_idle_window(struct cfq_data *cfqd, struct cfq_queue *cfqq,
3722 struct cfq_io_cq *cic)
3723{
3724 int old_idle, enable_idle;
3725
3726 /*
3727 * Don't idle for async or idle io prio class
3728 */
3729 if (!cfq_cfqq_sync(cfqq) || cfq_class_idle(cfqq))
3730 return;
3731
3732 enable_idle = old_idle = cfq_cfqq_idle_window(cfqq);
3733
3734 if (cfqq->queued[0] + cfqq->queued[1] >= 4)
3735 cfq_mark_cfqq_deep(cfqq);
3736
3737 if (cfqq->next_rq && (cfqq->next_rq->cmd_flags & REQ_NOIDLE))
3738 enable_idle = 0;
3739 else if (!atomic_read(&cic->icq.ioc->active_ref) ||
3740 !cfqd->cfq_slice_idle ||
3741 (!cfq_cfqq_deep(cfqq) && CFQQ_SEEKY(cfqq)))
3742 enable_idle = 0;
3743 else if (sample_valid(cic->ttime.ttime_samples)) {
3744 if (cic->ttime.ttime_mean > cfqd->cfq_slice_idle)
3745 enable_idle = 0;
3746 else
3747 enable_idle = 1;
3748 }
3749
3750 if (old_idle != enable_idle) {
3751 cfq_log_cfqq(cfqd, cfqq, "idle=%d", enable_idle);
3752 if (enable_idle)
3753 cfq_mark_cfqq_idle_window(cfqq);
3754 else
3755 cfq_clear_cfqq_idle_window(cfqq);
3756 }
3757}
3758
3759/*
3760 * Check if new_cfqq should preempt the currently active queue. Return 0 for
3761 * no or if we aren't sure, a 1 will cause a preempt.
3762 */
3763static bool
3764cfq_should_preempt(struct cfq_data *cfqd, struct cfq_queue *new_cfqq,
3765 struct request *rq)
3766{
3767 struct cfq_queue *cfqq;
3768
3769 cfqq = cfqd->active_queue;
3770 if (!cfqq)
3771 return false;
3772
3773 if (cfq_class_idle(new_cfqq))
3774 return false;
3775
3776 if (cfq_class_idle(cfqq))
3777 return true;
3778
3779 /*
3780 * Don't allow a non-RT request to preempt an ongoing RT cfqq timeslice.
3781 */
3782 if (cfq_class_rt(cfqq) && !cfq_class_rt(new_cfqq))
3783 return false;
3784
3785 /*
3786 * if the new request is sync, but the currently running queue is
3787 * not, let the sync request have priority.
3788 */
3789 if (rq_is_sync(rq) && !cfq_cfqq_sync(cfqq))
3790 return true;
3791
3792 if (new_cfqq->cfqg != cfqq->cfqg)
3793 return false;
3794
3795 if (cfq_slice_used(cfqq))
3796 return true;
3797
3798 /* Allow preemption only if we are idling on sync-noidle tree */
3799 if (cfqd->serving_wl_type == SYNC_NOIDLE_WORKLOAD &&
3800 cfqq_type(new_cfqq) == SYNC_NOIDLE_WORKLOAD &&
3801 new_cfqq->service_tree->count == 2 &&
3802 RB_EMPTY_ROOT(&cfqq->sort_list))
3803 return true;
3804
3805 /*
3806 * So both queues are sync. Let the new request get disk time if
3807 * it's a metadata request and the current queue is doing regular IO.
3808 */
3809 if ((rq->cmd_flags & REQ_PRIO) && !cfqq->prio_pending)
3810 return true;
3811
3812 /*
3813 * Allow an RT request to pre-empt an ongoing non-RT cfqq timeslice.
3814 */
3815 if (cfq_class_rt(new_cfqq) && !cfq_class_rt(cfqq))
3816 return true;
3817
3818 /* An idle queue should not be idle now for some reason */
3819 if (RB_EMPTY_ROOT(&cfqq->sort_list) && !cfq_should_idle(cfqd, cfqq))
3820 return true;
3821
3822 if (!cfqd->active_cic || !cfq_cfqq_wait_request(cfqq))
3823 return false;
3824
3825 /*
3826 * if this request is as-good as one we would expect from the
3827 * current cfqq, let it preempt
3828 */
3829 if (cfq_rq_close(cfqd, cfqq, rq))
3830 return true;
3831
3832 return false;
3833}
3834
3835/*
3836 * cfqq preempts the active queue. if we allowed preempt with no slice left,
3837 * let it have half of its nominal slice.
3838 */
3839static void cfq_preempt_queue(struct cfq_data *cfqd, struct cfq_queue *cfqq)
3840{
3841 enum wl_type_t old_type = cfqq_type(cfqd->active_queue);
3842
3843 cfq_log_cfqq(cfqd, cfqq, "preempt");
3844 cfq_slice_expired(cfqd, 1);
3845
3846 /*
3847 * workload type is changed, don't save slice, otherwise preempt
3848 * doesn't happen
3849 */
3850 if (old_type != cfqq_type(cfqq))
3851 cfqq->cfqg->saved_wl_slice = 0;
3852
3853 /*
3854 * Put the new queue at the front of the of the current list,
3855 * so we know that it will be selected next.
3856 */
3857 BUG_ON(!cfq_cfqq_on_rr(cfqq));
3858
3859 cfq_service_tree_add(cfqd, cfqq, 1);
3860
3861 cfqq->slice_end = 0;
3862 cfq_mark_cfqq_slice_new(cfqq);
3863}
3864
3865/*
3866 * Called when a new fs request (rq) is added (to cfqq). Check if there's
3867 * something we should do about it
3868 */
3869static void
3870cfq_rq_enqueued(struct cfq_data *cfqd, struct cfq_queue *cfqq,
3871 struct request *rq)
3872{
3873 struct cfq_io_cq *cic = RQ_CIC(rq);
3874
3875 cfqd->rq_queued++;
3876 if (rq->cmd_flags & REQ_PRIO)
3877 cfqq->prio_pending++;
3878
3879 cfq_update_io_thinktime(cfqd, cfqq, cic);
3880 cfq_update_io_seektime(cfqd, cfqq, rq);
3881 cfq_update_idle_window(cfqd, cfqq, cic);
3882
3883 cfqq->last_request_pos = blk_rq_pos(rq) + blk_rq_sectors(rq);
3884
3885 if (cfqq == cfqd->active_queue) {
3886 /*
3887 * Remember that we saw a request from this process, but
3888 * don't start queuing just yet. Otherwise we risk seeing lots
3889 * of tiny requests, because we disrupt the normal plugging
3890 * and merging. If the request is already larger than a single
3891 * page, let it rip immediately. For that case we assume that
3892 * merging is already done. Ditto for a busy system that
3893 * has other work pending, don't risk delaying until the
3894 * idle timer unplug to continue working.
3895 */
3896 if (cfq_cfqq_wait_request(cfqq)) {
3897 if (blk_rq_bytes(rq) > PAGE_CACHE_SIZE ||
3898 cfqd->busy_queues > 1) {
3899 cfq_del_timer(cfqd, cfqq);
3900 cfq_clear_cfqq_wait_request(cfqq);
3901 __blk_run_queue(cfqd->queue);
3902 } else {
3903 cfqg_stats_update_idle_time(cfqq->cfqg);
3904 cfq_mark_cfqq_must_dispatch(cfqq);
3905 }
3906 }
3907 } else if (cfq_should_preempt(cfqd, cfqq, rq)) {
3908 /*
3909 * not the active queue - expire current slice if it is
3910 * idle and has expired it's mean thinktime or this new queue
3911 * has some old slice time left and is of higher priority or
3912 * this new queue is RT and the current one is BE
3913 */
3914 cfq_preempt_queue(cfqd, cfqq);
3915 __blk_run_queue(cfqd->queue);
3916 }
3917}
3918
3919static void cfq_insert_request(struct request_queue *q, struct request *rq)
3920{
3921 struct cfq_data *cfqd = q->elevator->elevator_data;
3922 struct cfq_queue *cfqq = RQ_CFQQ(rq);
3923
3924 cfq_log_cfqq(cfqd, cfqq, "insert_request");
3925 cfq_init_prio_data(cfqq, RQ_CIC(rq));
3926
3927 rq->fifo_time = jiffies + cfqd->cfq_fifo_expire[rq_is_sync(rq)];
3928 list_add_tail(&rq->queuelist, &cfqq->fifo);
3929 cfq_add_rq_rb(rq);
3930 cfqg_stats_update_io_add(RQ_CFQG(rq), cfqd->serving_group,
3931 rq->cmd_flags);
3932 cfq_rq_enqueued(cfqd, cfqq, rq);
3933}
3934
3935/*
3936 * Update hw_tag based on peak queue depth over 50 samples under
3937 * sufficient load.
3938 */
3939static void cfq_update_hw_tag(struct cfq_data *cfqd)
3940{
3941 struct cfq_queue *cfqq = cfqd->active_queue;
3942
3943 if (cfqd->rq_in_driver > cfqd->hw_tag_est_depth)
3944 cfqd->hw_tag_est_depth = cfqd->rq_in_driver;
3945
3946 if (cfqd->hw_tag == 1)
3947 return;
3948
3949 if (cfqd->rq_queued <= CFQ_HW_QUEUE_MIN &&
3950 cfqd->rq_in_driver <= CFQ_HW_QUEUE_MIN)
3951 return;
3952
3953 /*
3954 * If active queue hasn't enough requests and can idle, cfq might not
3955 * dispatch sufficient requests to hardware. Don't zero hw_tag in this
3956 * case
3957 */
3958 if (cfqq && cfq_cfqq_idle_window(cfqq) &&
3959 cfqq->dispatched + cfqq->queued[0] + cfqq->queued[1] <
3960 CFQ_HW_QUEUE_MIN && cfqd->rq_in_driver < CFQ_HW_QUEUE_MIN)
3961 return;
3962
3963 if (cfqd->hw_tag_samples++ < 50)
3964 return;
3965
3966 if (cfqd->hw_tag_est_depth >= CFQ_HW_QUEUE_MIN)
3967 cfqd->hw_tag = 1;
3968 else
3969 cfqd->hw_tag = 0;
3970}
3971
3972static bool cfq_should_wait_busy(struct cfq_data *cfqd, struct cfq_queue *cfqq)
3973{
3974 struct cfq_io_cq *cic = cfqd->active_cic;
3975
3976 /* If the queue already has requests, don't wait */
3977 if (!RB_EMPTY_ROOT(&cfqq->sort_list))
3978 return false;
3979
3980 /* If there are other queues in the group, don't wait */
3981 if (cfqq->cfqg->nr_cfqq > 1)
3982 return false;
3983
3984 /* the only queue in the group, but think time is big */
3985 if (cfq_io_thinktime_big(cfqd, &cfqq->cfqg->ttime, true))
3986 return false;
3987
3988 if (cfq_slice_used(cfqq))
3989 return true;
3990
3991 /* if slice left is less than think time, wait busy */
3992 if (cic && sample_valid(cic->ttime.ttime_samples)
3993 && (cfqq->slice_end - jiffies < cic->ttime.ttime_mean))
3994 return true;
3995
3996 /*
3997 * If think times is less than a jiffy than ttime_mean=0 and above
3998 * will not be true. It might happen that slice has not expired yet
3999 * but will expire soon (4-5 ns) during select_queue(). To cover the
4000 * case where think time is less than a jiffy, mark the queue wait
4001 * busy if only 1 jiffy is left in the slice.
4002 */
4003 if (cfqq->slice_end - jiffies == 1)
4004 return true;
4005
4006 return false;
4007}
4008
4009static void cfq_completed_request(struct request_queue *q, struct request *rq)
4010{
4011 struct cfq_queue *cfqq = RQ_CFQQ(rq);
4012 struct cfq_data *cfqd = cfqq->cfqd;
4013 const int sync = rq_is_sync(rq);
4014 unsigned long now;
4015
4016 now = jiffies;
4017 cfq_log_cfqq(cfqd, cfqq, "complete rqnoidle %d",
4018 !!(rq->cmd_flags & REQ_NOIDLE));
4019
4020 cfq_update_hw_tag(cfqd);
4021
4022 WARN_ON(!cfqd->rq_in_driver);
4023 WARN_ON(!cfqq->dispatched);
4024 cfqd->rq_in_driver--;
4025 cfqq->dispatched--;
4026 (RQ_CFQG(rq))->dispatched--;
4027 cfqg_stats_update_completion(cfqq->cfqg, rq_start_time_ns(rq),
4028 rq_io_start_time_ns(rq), rq->cmd_flags);
4029
4030 cfqd->rq_in_flight[cfq_cfqq_sync(cfqq)]--;
4031
4032 if (sync) {
4033 struct cfq_rb_root *st;
4034
4035 RQ_CIC(rq)->ttime.last_end_request = now;
4036
4037 if (cfq_cfqq_on_rr(cfqq))
4038 st = cfqq->service_tree;
4039 else
4040 st = st_for(cfqq->cfqg, cfqq_class(cfqq),
4041 cfqq_type(cfqq));
4042
4043 st->ttime.last_end_request = now;
4044 if (!time_after(rq->start_time + cfqd->cfq_fifo_expire[1], now))
4045 cfqd->last_delayed_sync = now;
4046 }
4047
4048#ifdef CONFIG_CFQ_GROUP_IOSCHED
4049 cfqq->cfqg->ttime.last_end_request = now;
4050#endif
4051
4052 /*
4053 * If this is the active queue, check if it needs to be expired,
4054 * or if we want to idle in case it has no pending requests.
4055 */
4056 if (cfqd->active_queue == cfqq) {
4057 const bool cfqq_empty = RB_EMPTY_ROOT(&cfqq->sort_list);
4058
4059 if (cfq_cfqq_slice_new(cfqq)) {
4060 cfq_set_prio_slice(cfqd, cfqq);
4061 cfq_clear_cfqq_slice_new(cfqq);
4062 }
4063
4064 /*
4065 * Should we wait for next request to come in before we expire
4066 * the queue.
4067 */
4068 if (cfq_should_wait_busy(cfqd, cfqq)) {
4069 unsigned long extend_sl = cfqd->cfq_slice_idle;
4070 if (!cfqd->cfq_slice_idle)
4071 extend_sl = cfqd->cfq_group_idle;
4072 cfqq->slice_end = jiffies + extend_sl;
4073 cfq_mark_cfqq_wait_busy(cfqq);
4074 cfq_log_cfqq(cfqd, cfqq, "will busy wait");
4075 }
4076
4077 /*
4078 * Idling is not enabled on:
4079 * - expired queues
4080 * - idle-priority queues
4081 * - async queues
4082 * - queues with still some requests queued
4083 * - when there is a close cooperator
4084 */
4085 if (cfq_slice_used(cfqq) || cfq_class_idle(cfqq))
4086 cfq_slice_expired(cfqd, 1);
4087 else if (sync && cfqq_empty &&
4088 !cfq_close_cooperator(cfqd, cfqq)) {
4089 cfq_arm_slice_timer(cfqd);
4090 }
4091 }
4092
4093 if (!cfqd->rq_in_driver)
4094 cfq_schedule_dispatch(cfqd);
4095}
4096
4097static inline int __cfq_may_queue(struct cfq_queue *cfqq)
4098{
4099 if (cfq_cfqq_wait_request(cfqq) && !cfq_cfqq_must_alloc_slice(cfqq)) {
4100 cfq_mark_cfqq_must_alloc_slice(cfqq);
4101 return ELV_MQUEUE_MUST;
4102 }
4103
4104 return ELV_MQUEUE_MAY;
4105}
4106
4107static int cfq_may_queue(struct request_queue *q, int rw)
4108{
4109 struct cfq_data *cfqd = q->elevator->elevator_data;
4110 struct task_struct *tsk = current;
4111 struct cfq_io_cq *cic;
4112 struct cfq_queue *cfqq;
4113
4114 /*
4115 * don't force setup of a queue from here, as a call to may_queue
4116 * does not necessarily imply that a request actually will be queued.
4117 * so just lookup a possibly existing queue, or return 'may queue'
4118 * if that fails
4119 */
4120 cic = cfq_cic_lookup(cfqd, tsk->io_context);
4121 if (!cic)
4122 return ELV_MQUEUE_MAY;
4123
4124 cfqq = cic_to_cfqq(cic, rw_is_sync(rw));
4125 if (cfqq) {
4126 cfq_init_prio_data(cfqq, cic);
4127
4128 return __cfq_may_queue(cfqq);
4129 }
4130
4131 return ELV_MQUEUE_MAY;
4132}
4133
4134/*
4135 * queue lock held here
4136 */
4137static void cfq_put_request(struct request *rq)
4138{
4139 struct cfq_queue *cfqq = RQ_CFQQ(rq);
4140
4141 if (cfqq) {
4142 const int rw = rq_data_dir(rq);
4143
4144 BUG_ON(!cfqq->allocated[rw]);
4145 cfqq->allocated[rw]--;
4146
4147 /* Put down rq reference on cfqg */
4148 cfqg_put(RQ_CFQG(rq));
4149 rq->elv.priv[0] = NULL;
4150 rq->elv.priv[1] = NULL;
4151
4152 cfq_put_queue(cfqq);
4153 }
4154}
4155
4156static struct cfq_queue *
4157cfq_merge_cfqqs(struct cfq_data *cfqd, struct cfq_io_cq *cic,
4158 struct cfq_queue *cfqq)
4159{
4160 cfq_log_cfqq(cfqd, cfqq, "merging with queue %p", cfqq->new_cfqq);
4161 cic_set_cfqq(cic, cfqq->new_cfqq, 1);
4162 cfq_mark_cfqq_coop(cfqq->new_cfqq);
4163 cfq_put_queue(cfqq);
4164 return cic_to_cfqq(cic, 1);
4165}
4166
4167/*
4168 * Returns NULL if a new cfqq should be allocated, or the old cfqq if this
4169 * was the last process referring to said cfqq.
4170 */
4171static struct cfq_queue *
4172split_cfqq(struct cfq_io_cq *cic, struct cfq_queue *cfqq)
4173{
4174 if (cfqq_process_refs(cfqq) == 1) {
4175 cfqq->pid = current->pid;
4176 cfq_clear_cfqq_coop(cfqq);
4177 cfq_clear_cfqq_split_coop(cfqq);
4178 return cfqq;
4179 }
4180
4181 cic_set_cfqq(cic, NULL, 1);
4182
4183 cfq_put_cooperator(cfqq);
4184
4185 cfq_put_queue(cfqq);
4186 return NULL;
4187}
4188/*
4189 * Allocate cfq data structures associated with this request.
4190 */
4191static int
4192cfq_set_request(struct request_queue *q, struct request *rq, struct bio *bio,
4193 gfp_t gfp_mask)
4194{
4195 struct cfq_data *cfqd = q->elevator->elevator_data;
4196 struct cfq_io_cq *cic = icq_to_cic(rq->elv.icq);
4197 const int rw = rq_data_dir(rq);
4198 const bool is_sync = rq_is_sync(rq);
4199 struct cfq_queue *cfqq;
4200
4201 might_sleep_if(gfp_mask & __GFP_WAIT);
4202
4203 spin_lock_irq(q->queue_lock);
4204
4205 check_ioprio_changed(cic, bio);
4206 check_blkcg_changed(cic, bio);
4207new_queue:
4208 cfqq = cic_to_cfqq(cic, is_sync);
4209 if (!cfqq || cfqq == &cfqd->oom_cfqq) {
4210 cfqq = cfq_get_queue(cfqd, is_sync, cic, bio, gfp_mask);
4211 cic_set_cfqq(cic, cfqq, is_sync);
4212 } else {
4213 /*
4214 * If the queue was seeky for too long, break it apart.
4215 */
4216 if (cfq_cfqq_coop(cfqq) && cfq_cfqq_split_coop(cfqq)) {
4217 cfq_log_cfqq(cfqd, cfqq, "breaking apart cfqq");
4218 cfqq = split_cfqq(cic, cfqq);
4219 if (!cfqq)
4220 goto new_queue;
4221 }
4222
4223 /*
4224 * Check to see if this queue is scheduled to merge with
4225 * another, closely cooperating queue. The merging of
4226 * queues happens here as it must be done in process context.
4227 * The reference on new_cfqq was taken in merge_cfqqs.
4228 */
4229 if (cfqq->new_cfqq)
4230 cfqq = cfq_merge_cfqqs(cfqd, cic, cfqq);
4231 }
4232
4233 cfqq->allocated[rw]++;
4234
4235 cfqq->ref++;
4236 cfqg_get(cfqq->cfqg);
4237 rq->elv.priv[0] = cfqq;
4238 rq->elv.priv[1] = cfqq->cfqg;
4239 spin_unlock_irq(q->queue_lock);
4240 return 0;
4241}
4242
4243static void cfq_kick_queue(struct work_struct *work)
4244{
4245 struct cfq_data *cfqd =
4246 container_of(work, struct cfq_data, unplug_work);
4247 struct request_queue *q = cfqd->queue;
4248
4249 spin_lock_irq(q->queue_lock);
4250 __blk_run_queue(cfqd->queue);
4251 spin_unlock_irq(q->queue_lock);
4252}
4253
4254/*
4255 * Timer running if the active_queue is currently idling inside its time slice
4256 */
4257static void cfq_idle_slice_timer(unsigned long data)
4258{
4259 struct cfq_data *cfqd = (struct cfq_data *) data;
4260 struct cfq_queue *cfqq;
4261 unsigned long flags;
4262 int timed_out = 1;
4263
4264 cfq_log(cfqd, "idle timer fired");
4265
4266 spin_lock_irqsave(cfqd->queue->queue_lock, flags);
4267
4268 cfqq = cfqd->active_queue;
4269 if (cfqq) {
4270 timed_out = 0;
4271
4272 /*
4273 * We saw a request before the queue expired, let it through
4274 */
4275 if (cfq_cfqq_must_dispatch(cfqq))
4276 goto out_kick;
4277
4278 /*
4279 * expired
4280 */
4281 if (cfq_slice_used(cfqq))
4282 goto expire;
4283
4284 /*
4285 * only expire and reinvoke request handler, if there are
4286 * other queues with pending requests
4287 */
4288 if (!cfqd->busy_queues)
4289 goto out_cont;
4290
4291 /*
4292 * not expired and it has a request pending, let it dispatch
4293 */
4294 if (!RB_EMPTY_ROOT(&cfqq->sort_list))
4295 goto out_kick;
4296
4297 /*
4298 * Queue depth flag is reset only when the idle didn't succeed
4299 */
4300 cfq_clear_cfqq_deep(cfqq);
4301 }
4302expire:
4303 cfq_slice_expired(cfqd, timed_out);
4304out_kick:
4305 cfq_schedule_dispatch(cfqd);
4306out_cont:
4307 spin_unlock_irqrestore(cfqd->queue->queue_lock, flags);
4308}
4309
4310static void cfq_shutdown_timer_wq(struct cfq_data *cfqd)
4311{
4312 del_timer_sync(&cfqd->idle_slice_timer);
4313 cancel_work_sync(&cfqd->unplug_work);
4314}
4315
4316static void cfq_put_async_queues(struct cfq_data *cfqd)
4317{
4318 int i;
4319
4320 for (i = 0; i < IOPRIO_BE_NR; i++) {
4321 if (cfqd->async_cfqq[0][i])
4322 cfq_put_queue(cfqd->async_cfqq[0][i]);
4323 if (cfqd->async_cfqq[1][i])
4324 cfq_put_queue(cfqd->async_cfqq[1][i]);
4325 }
4326
4327 if (cfqd->async_idle_cfqq)
4328 cfq_put_queue(cfqd->async_idle_cfqq);
4329}
4330
4331static void cfq_exit_queue(struct elevator_queue *e)
4332{
4333 struct cfq_data *cfqd = e->elevator_data;
4334 struct request_queue *q = cfqd->queue;
4335
4336 cfq_shutdown_timer_wq(cfqd);
4337
4338 spin_lock_irq(q->queue_lock);
4339
4340 if (cfqd->active_queue)
4341 __cfq_slice_expired(cfqd, cfqd->active_queue, 0);
4342
4343 cfq_put_async_queues(cfqd);
4344
4345 spin_unlock_irq(q->queue_lock);
4346
4347 cfq_shutdown_timer_wq(cfqd);
4348
4349#ifdef CONFIG_CFQ_GROUP_IOSCHED
4350 blkcg_deactivate_policy(q, &blkcg_policy_cfq);
4351#else
4352 kfree(cfqd->root_group);
4353#endif
4354 kfree(cfqd);
4355}
4356
4357static int cfq_init_queue(struct request_queue *q, struct elevator_type *e)
4358{
4359 struct cfq_data *cfqd;
4360 struct blkcg_gq *blkg __maybe_unused;
4361 int i, ret;
4362 struct elevator_queue *eq;
4363
4364 eq = elevator_alloc(q, e);
4365 if (!eq)
4366 return -ENOMEM;
4367
4368 cfqd = kzalloc_node(sizeof(*cfqd), GFP_KERNEL, q->node);
4369 if (!cfqd) {
4370 kobject_put(&eq->kobj);
4371 return -ENOMEM;
4372 }
4373 eq->elevator_data = cfqd;
4374
4375 cfqd->queue = q;
4376 spin_lock_irq(q->queue_lock);
4377 q->elevator = eq;
4378 spin_unlock_irq(q->queue_lock);
4379
4380 /* Init root service tree */
4381 cfqd->grp_service_tree = CFQ_RB_ROOT;
4382
4383 /* Init root group and prefer root group over other groups by default */
4384#ifdef CONFIG_CFQ_GROUP_IOSCHED
4385 ret = blkcg_activate_policy(q, &blkcg_policy_cfq);
4386 if (ret)
4387 goto out_free;
4388
4389 cfqd->root_group = blkg_to_cfqg(q->root_blkg);
4390#else
4391 ret = -ENOMEM;
4392 cfqd->root_group = kzalloc_node(sizeof(*cfqd->root_group),
4393 GFP_KERNEL, cfqd->queue->node);
4394 if (!cfqd->root_group)
4395 goto out_free;
4396
4397 cfq_init_cfqg_base(cfqd->root_group);
4398#endif
4399 cfqd->root_group->weight = 2 * CFQ_WEIGHT_DEFAULT;
4400 cfqd->root_group->leaf_weight = 2 * CFQ_WEIGHT_DEFAULT;
4401
4402 /*
4403 * Not strictly needed (since RB_ROOT just clears the node and we
4404 * zeroed cfqd on alloc), but better be safe in case someone decides
4405 * to add magic to the rb code
4406 */
4407 for (i = 0; i < CFQ_PRIO_LISTS; i++)
4408 cfqd->prio_trees[i] = RB_ROOT;
4409
4410 /*
4411 * Our fallback cfqq if cfq_find_alloc_queue() runs into OOM issues.
4412 * Grab a permanent reference to it, so that the normal code flow
4413 * will not attempt to free it. oom_cfqq is linked to root_group
4414 * but shouldn't hold a reference as it'll never be unlinked. Lose
4415 * the reference from linking right away.
4416 */
4417 cfq_init_cfqq(cfqd, &cfqd->oom_cfqq, 1, 0);
4418 cfqd->oom_cfqq.ref++;
4419
4420 spin_lock_irq(q->queue_lock);
4421 cfq_link_cfqq_cfqg(&cfqd->oom_cfqq, cfqd->root_group);
4422 cfqg_put(cfqd->root_group);
4423 spin_unlock_irq(q->queue_lock);
4424
4425 init_timer(&cfqd->idle_slice_timer);
4426 cfqd->idle_slice_timer.function = cfq_idle_slice_timer;
4427 cfqd->idle_slice_timer.data = (unsigned long) cfqd;
4428
4429 INIT_WORK(&cfqd->unplug_work, cfq_kick_queue);
4430
4431 cfqd->cfq_quantum = cfq_quantum;
4432 cfqd->cfq_fifo_expire[0] = cfq_fifo_expire[0];
4433 cfqd->cfq_fifo_expire[1] = cfq_fifo_expire[1];
4434 cfqd->cfq_back_max = cfq_back_max;
4435 cfqd->cfq_back_penalty = cfq_back_penalty;
4436 cfqd->cfq_slice[0] = cfq_slice_async;
4437 cfqd->cfq_slice[1] = cfq_slice_sync;
4438 cfqd->cfq_target_latency = cfq_target_latency;
4439 cfqd->cfq_slice_async_rq = cfq_slice_async_rq;
4440 cfqd->cfq_slice_idle = cfq_slice_idle;
4441 cfqd->cfq_group_idle = cfq_group_idle;
4442 cfqd->cfq_latency = 1;
4443 cfqd->hw_tag = -1;
4444 /*
4445 * we optimistically start assuming sync ops weren't delayed in last
4446 * second, in order to have larger depth for async operations.
4447 */
4448 cfqd->last_delayed_sync = jiffies - HZ;
4449 return 0;
4450
4451out_free:
4452 kfree(cfqd);
4453 kobject_put(&eq->kobj);
4454 return ret;
4455}
4456
4457/*
4458 * sysfs parts below -->
4459 */
4460static ssize_t
4461cfq_var_show(unsigned int var, char *page)
4462{
4463 return sprintf(page, "%d\n", var);
4464}
4465
4466static ssize_t
4467cfq_var_store(unsigned int *var, const char *page, size_t count)
4468{
4469 char *p = (char *) page;
4470
4471 *var = simple_strtoul(p, &p, 10);
4472 return count;
4473}
4474
4475#define SHOW_FUNCTION(__FUNC, __VAR, __CONV) \
4476static ssize_t __FUNC(struct elevator_queue *e, char *page) \
4477{ \
4478 struct cfq_data *cfqd = e->elevator_data; \
4479 unsigned int __data = __VAR; \
4480 if (__CONV) \
4481 __data = jiffies_to_msecs(__data); \
4482 return cfq_var_show(__data, (page)); \
4483}
4484SHOW_FUNCTION(cfq_quantum_show, cfqd->cfq_quantum, 0);
4485SHOW_FUNCTION(cfq_fifo_expire_sync_show, cfqd->cfq_fifo_expire[1], 1);
4486SHOW_FUNCTION(cfq_fifo_expire_async_show, cfqd->cfq_fifo_expire[0], 1);
4487SHOW_FUNCTION(cfq_back_seek_max_show, cfqd->cfq_back_max, 0);
4488SHOW_FUNCTION(cfq_back_seek_penalty_show, cfqd->cfq_back_penalty, 0);
4489SHOW_FUNCTION(cfq_slice_idle_show, cfqd->cfq_slice_idle, 1);
4490SHOW_FUNCTION(cfq_group_idle_show, cfqd->cfq_group_idle, 1);
4491SHOW_FUNCTION(cfq_slice_sync_show, cfqd->cfq_slice[1], 1);
4492SHOW_FUNCTION(cfq_slice_async_show, cfqd->cfq_slice[0], 1);
4493SHOW_FUNCTION(cfq_slice_async_rq_show, cfqd->cfq_slice_async_rq, 0);
4494SHOW_FUNCTION(cfq_low_latency_show, cfqd->cfq_latency, 0);
4495SHOW_FUNCTION(cfq_target_latency_show, cfqd->cfq_target_latency, 1);
4496#undef SHOW_FUNCTION
4497
4498#define STORE_FUNCTION(__FUNC, __PTR, MIN, MAX, __CONV) \
4499static ssize_t __FUNC(struct elevator_queue *e, const char *page, size_t count) \
4500{ \
4501 struct cfq_data *cfqd = e->elevator_data; \
4502 unsigned int __data; \
4503 int ret = cfq_var_store(&__data, (page), count); \
4504 if (__data < (MIN)) \
4505 __data = (MIN); \
4506 else if (__data > (MAX)) \
4507 __data = (MAX); \
4508 if (__CONV) \
4509 *(__PTR) = msecs_to_jiffies(__data); \
4510 else \
4511 *(__PTR) = __data; \
4512 return ret; \
4513}
4514STORE_FUNCTION(cfq_quantum_store, &cfqd->cfq_quantum, 1, UINT_MAX, 0);
4515STORE_FUNCTION(cfq_fifo_expire_sync_store, &cfqd->cfq_fifo_expire[1], 1,
4516 UINT_MAX, 1);
4517STORE_FUNCTION(cfq_fifo_expire_async_store, &cfqd->cfq_fifo_expire[0], 1,
4518 UINT_MAX, 1);
4519STORE_FUNCTION(cfq_back_seek_max_store, &cfqd->cfq_back_max, 0, UINT_MAX, 0);
4520STORE_FUNCTION(cfq_back_seek_penalty_store, &cfqd->cfq_back_penalty, 1,
4521 UINT_MAX, 0);
4522STORE_FUNCTION(cfq_slice_idle_store, &cfqd->cfq_slice_idle, 0, UINT_MAX, 1);
4523STORE_FUNCTION(cfq_group_idle_store, &cfqd->cfq_group_idle, 0, UINT_MAX, 1);
4524STORE_FUNCTION(cfq_slice_sync_store, &cfqd->cfq_slice[1], 1, UINT_MAX, 1);
4525STORE_FUNCTION(cfq_slice_async_store, &cfqd->cfq_slice[0], 1, UINT_MAX, 1);
4526STORE_FUNCTION(cfq_slice_async_rq_store, &cfqd->cfq_slice_async_rq, 1,
4527 UINT_MAX, 0);
4528STORE_FUNCTION(cfq_low_latency_store, &cfqd->cfq_latency, 0, 1, 0);
4529STORE_FUNCTION(cfq_target_latency_store, &cfqd->cfq_target_latency, 1, UINT_MAX, 1);
4530#undef STORE_FUNCTION
4531
4532#define CFQ_ATTR(name) \
4533 __ATTR(name, S_IRUGO|S_IWUSR, cfq_##name##_show, cfq_##name##_store)
4534
4535static struct elv_fs_entry cfq_attrs[] = {
4536 CFQ_ATTR(quantum),
4537 CFQ_ATTR(fifo_expire_sync),
4538 CFQ_ATTR(fifo_expire_async),
4539 CFQ_ATTR(back_seek_max),
4540 CFQ_ATTR(back_seek_penalty),
4541 CFQ_ATTR(slice_sync),
4542 CFQ_ATTR(slice_async),
4543 CFQ_ATTR(slice_async_rq),
4544 CFQ_ATTR(slice_idle),
4545 CFQ_ATTR(group_idle),
4546 CFQ_ATTR(low_latency),
4547 CFQ_ATTR(target_latency),
4548 __ATTR_NULL
4549};
4550
4551static struct elevator_type iosched_cfq = {
4552 .ops = {
4553 .elevator_merge_fn = cfq_merge,
4554 .elevator_merged_fn = cfq_merged_request,
4555 .elevator_merge_req_fn = cfq_merged_requests,
4556 .elevator_allow_merge_fn = cfq_allow_merge,
4557 .elevator_bio_merged_fn = cfq_bio_merged,
4558 .elevator_dispatch_fn = cfq_dispatch_requests,
4559 .elevator_add_req_fn = cfq_insert_request,
4560 .elevator_activate_req_fn = cfq_activate_request,
4561 .elevator_deactivate_req_fn = cfq_deactivate_request,
4562 .elevator_completed_req_fn = cfq_completed_request,
4563 .elevator_former_req_fn = elv_rb_former_request,
4564 .elevator_latter_req_fn = elv_rb_latter_request,
4565 .elevator_init_icq_fn = cfq_init_icq,
4566 .elevator_exit_icq_fn = cfq_exit_icq,
4567 .elevator_set_req_fn = cfq_set_request,
4568 .elevator_put_req_fn = cfq_put_request,
4569 .elevator_may_queue_fn = cfq_may_queue,
4570 .elevator_init_fn = cfq_init_queue,
4571 .elevator_exit_fn = cfq_exit_queue,
4572 },
4573 .icq_size = sizeof(struct cfq_io_cq),
4574 .icq_align = __alignof__(struct cfq_io_cq),
4575 .elevator_attrs = cfq_attrs,
4576 .elevator_name = "cfq",
4577 .elevator_owner = THIS_MODULE,
4578};
4579
4580#ifdef CONFIG_CFQ_GROUP_IOSCHED
4581static struct blkcg_policy blkcg_policy_cfq = {
4582 .pd_size = sizeof(struct cfq_group),
4583 .cftypes = cfq_blkcg_files,
4584
4585 .pd_init_fn = cfq_pd_init,
4586 .pd_offline_fn = cfq_pd_offline,
4587 .pd_reset_stats_fn = cfq_pd_reset_stats,
4588};
4589#endif
4590
4591static int __init cfq_init(void)
4592{
4593 int ret;
4594
4595 /*
4596 * could be 0 on HZ < 1000 setups
4597 */
4598 if (!cfq_slice_async)
4599 cfq_slice_async = 1;
4600 if (!cfq_slice_idle)
4601 cfq_slice_idle = 1;
4602
4603#ifdef CONFIG_CFQ_GROUP_IOSCHED
4604 if (!cfq_group_idle)
4605 cfq_group_idle = 1;
4606
4607 ret = blkcg_policy_register(&blkcg_policy_cfq);
4608 if (ret)
4609 return ret;
4610#else
4611 cfq_group_idle = 0;
4612#endif
4613
4614 ret = -ENOMEM;
4615 cfq_pool = KMEM_CACHE(cfq_queue, 0);
4616 if (!cfq_pool)
4617 goto err_pol_unreg;
4618
4619 ret = elv_register(&iosched_cfq);
4620 if (ret)
4621 goto err_free_pool;
4622
4623 return 0;
4624
4625err_free_pool:
4626 kmem_cache_destroy(cfq_pool);
4627err_pol_unreg:
4628#ifdef CONFIG_CFQ_GROUP_IOSCHED
4629 blkcg_policy_unregister(&blkcg_policy_cfq);
4630#endif
4631 return ret;
4632}
4633
4634static void __exit cfq_exit(void)
4635{
4636#ifdef CONFIG_CFQ_GROUP_IOSCHED
4637 blkcg_policy_unregister(&blkcg_policy_cfq);
4638#endif
4639 elv_unregister(&iosched_cfq);
4640 kmem_cache_destroy(cfq_pool);
4641}
4642
4643module_init(cfq_init);
4644module_exit(cfq_exit);
4645
4646MODULE_AUTHOR("Jens Axboe");
4647MODULE_LICENSE("GPL");
4648MODULE_DESCRIPTION("Completely Fair Queueing IO scheduler");