cfq-iosched.c 114 KB
Newer Older
Linus Torvalds's avatar
Linus Torvalds committed
1 2 3 4 5 6
/*
 *  CFQ, or complete fairness queueing, disk scheduler.
 *
 *  Based on ideas from a previously unfinished io
 *  scheduler (round robin per-process disk scheduling) and Andrea Arcangeli.
 *
7
 *  Copyright (C) 2003 Jens Axboe <axboe@kernel.dk>
Linus Torvalds's avatar
Linus Torvalds committed
8 9
 */
#include <linux/module.h>
10
#include <linux/slab.h>
Al Viro's avatar
Al Viro committed
11 12
#include <linux/blkdev.h>
#include <linux/elevator.h>
Randy Dunlap's avatar
Randy Dunlap committed
13
#include <linux/jiffies.h>
Linus Torvalds's avatar
Linus Torvalds committed
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"
Linus Torvalds's avatar
Linus Torvalds committed
19 20 21 22

/*
 * tunables
 */
23
/* max queue in one round of service */
Shaohua Li's avatar
Shaohua Li committed
24
static const int cfq_quantum = 8;
25
static const int cfq_fifo_expire[2] = { HZ / 4, HZ / 8 };
26 27 28 29
/* maximum backwards seek, in KiB */
static const int cfq_back_max = 16 * 1024;
/* penalty of a backwards seek */
static const int cfq_back_penalty = 2;
30
static const int cfq_slice_sync = HZ / 10;
Jens Axboe's avatar
Jens Axboe committed
31
static int cfq_slice_async = HZ / 25;
32
static const int cfq_slice_async_rq = 2;
33
static int cfq_slice_idle = HZ / 125;
34
static int cfq_group_idle = HZ / 125;
35 36
static const int cfq_target_latency = HZ * 3/10; /* 300 ms */
static 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 45 46 47

/*
 * below this threshold, we consider thinktime immediate
 */
#define CFQ_MIN_TT		(2)

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 58 59
#define RQ_CIC(rq)		icq_to_cic((rq)->elv.icq)
#define RQ_CFQQ(rq)		(struct cfq_queue *) ((rq)->elv.priv[0])
#define RQ_CFQG(rq)		(struct cfq_group *) ((rq)->elv.priv[1])
Linus Torvalds's avatar
Linus Torvalds committed
60

61
static struct kmem_cache *cfq_pool;
Linus Torvalds's avatar
Linus Torvalds committed
62

63 64 65 66
#define CFQ_PRIO_LISTS		IOPRIO_BE_NR
#define cfq_class_idle(cfqq)	((cfqq)->ioprio_class == IOPRIO_CLASS_IDLE)
#define cfq_class_rt(cfqq)	((cfqq)->ioprio_class == IOPRIO_CLASS_RT)

67
#define sample_valid(samples)	((samples) > 80)
68
#define rb_entry_cfqg(node)	rb_entry((node), struct cfq_group, rb_node)
69

70 71 72 73 74 75 76 77
struct cfq_ttime {
	unsigned long last_end_request;

	unsigned long ttime_total;
	unsigned long ttime_samples;
	unsigned long ttime_mean;
};

78 79 80 81 82 83 84 85 86
/*
 * Most of our rbtree usage is for sorting with min extraction, so
 * if we cache the leftmost node we don't have to walk down the tree
 * to find it. Idea borrowed from Ingo Molnars CFS scheduler. We should
 * move this into the elevator for the rq sorting as well.
 */
struct cfq_rb_root {
	struct rb_root rb;
	struct rb_node *left;
87
	unsigned count;
88
	u64 min_vdisktime;
89
	struct cfq_ttime ttime;
90
};
91 92
#define CFQ_RB_ROOT	(struct cfq_rb_root) { .rb = RB_ROOT, \
			.ttime = {.last_end_request = jiffies,},}
93

94 95 96 97 98
/*
 * Per process-grouping structure
 */
struct cfq_queue {
	/* reference count */
99
	int ref;
100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122
	/* various state flags, see below */
	unsigned int flags;
	/* parent cfq_data */
	struct cfq_data *cfqd;
	/* service_tree member */
	struct rb_node rb_node;
	/* service_tree key */
	unsigned long rb_key;
	/* prio tree member */
	struct rb_node p_node;
	/* prio tree root we belong to, if any */
	struct rb_root *p_root;
	/* sorted list of pending requests */
	struct rb_root sort_list;
	/* if fifo isn't expired, next request to serve */
	struct request *next_rq;
	/* requests queued in sort_list */
	int queued[2];
	/* currently allocated requests */
	int allocated[2];
	/* fifo list of requests in sort_list */
	struct list_head fifo;

123 124
	/* time when queue got scheduled in to dispatch first request. */
	unsigned long dispatch_start;
125
	unsigned int allocated_slice;
126
	unsigned int slice_dispatch;
127 128
	/* time when first request from queue completed and slice started. */
	unsigned long slice_start;
129 130 131
	unsigned long slice_end;
	long slice_resid;

132 133
	/* pending priority requests */
	int prio_pending;
134 135 136 137 138
	/* number of requests that are on the dispatch list or inside driver */
	int dispatched;

	/* io prio of this group */
	unsigned short ioprio, org_ioprio;
139
	unsigned short ioprio_class;
140

141 142
	pid_t pid;

143
	u32 seek_history;
144 145
	sector_t last_request_pos;

146
	struct cfq_rb_root *service_tree;
Jeff Moyer's avatar
Jeff Moyer committed
147
	struct cfq_queue *new_cfqq;
148
	struct cfq_group *cfqg;
149 150
	/* Number of sectors dispatched from queue in single dispatch round */
	unsigned long nr_sectors;
151 152
};

153
/*
154
 * First index in the service_trees.
155 156
 * IDLE is handled separately, so it has negative index
 */
157
enum wl_class_t {
158
	BE_WORKLOAD = 0,
159 160
	RT_WORKLOAD = 1,
	IDLE_WORKLOAD = 2,
161
	CFQ_PRIO_NR,
162 163
};

164 165 166 167 168 169 170 171 172
/*
 * Second index in the service_trees.
 */
enum wl_type_t {
	ASYNC_WORKLOAD = 0,
	SYNC_NOIDLE_WORKLOAD = 1,
	SYNC_WORKLOAD = 2
};

173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201
struct cfqg_stats {
#ifdef CONFIG_CFQ_GROUP_IOSCHED
	/* total bytes transferred */
	struct blkg_rwstat		service_bytes;
	/* total IOs serviced, post merge */
	struct blkg_rwstat		serviced;
	/* number of ios merged */
	struct blkg_rwstat		merged;
	/* total time spent on device in ns, may not be accurate w/ queueing */
	struct blkg_rwstat		service_time;
	/* total time spent waiting in scheduler queue in ns */
	struct blkg_rwstat		wait_time;
	/* number of IOs queued up */
	struct blkg_rwstat		queued;
	/* total sectors transferred */
	struct blkg_stat		sectors;
	/* total disk time and nr sectors dispatched by this group */
	struct blkg_stat		time;
#ifdef CONFIG_DEBUG_BLK_CGROUP
	/* time not charged to this cgroup */
	struct blkg_stat		unaccounted_time;
	/* sum of number of ios queued across all samples */
	struct blkg_stat		avg_queue_size_sum;
	/* count of samples taken for average */
	struct blkg_stat		avg_queue_size_samples;
	/* how many times this group has been removed from service tree */
	struct blkg_stat		dequeue;
	/* total time spent waiting for it to be assigned a timeslice. */
	struct blkg_stat		group_wait_time;
Tejun Heo's avatar
Tejun Heo committed
202
	/* time spent idling for this blkcg_gq */
203 204 205 206 207 208 209 210 211 212 213 214
	struct blkg_stat		idle_time;
	/* total time with empty current active q with other requests queued */
	struct blkg_stat		empty_time;
	/* fields after this shouldn't be cleared on stat reset */
	uint64_t			start_group_wait_time;
	uint64_t			start_idle_time;
	uint64_t			start_empty_time;
	uint16_t			flags;
#endif	/* CONFIG_DEBUG_BLK_CGROUP */
#endif	/* CONFIG_CFQ_GROUP_IOSCHED */
};

215 216
/* This is per cgroup per device grouping structure */
struct cfq_group {
217 218 219
	/* must be the first member */
	struct blkg_policy_data pd;

220 221 222 223 224
	/* group service_tree member */
	struct rb_node rb_node;

	/* group service_tree key */
	u64 vdisktime;
Tejun Heo's avatar
Tejun Heo committed
225

226 227 228 229 230 231 232 233 234 235 236 237
	/*
	 * The number of active cfqgs and sum of their weights under this
	 * cfqg.  This covers this cfqg's leaf_weight and all children's
	 * weights, but does not cover weights of further descendants.
	 *
	 * If a cfqg is on the service tree, it's active.  An active cfqg
	 * also activates its parent and contributes to the children_weight
	 * of the parent.
	 */
	int nr_active;
	unsigned int children_weight;

238 239 240 241 242 243 244 245 246 247 248 249
	/*
	 * vfraction is the fraction of vdisktime that the tasks in this
	 * cfqg are entitled to.  This is determined by compounding the
	 * ratios walking up from this cfqg to the root.
	 *
	 * It is in fixed point w/ CFQ_SERVICE_SHIFT and the sum of all
	 * vfractions on a service tree is approximately 1.  The sum may
	 * deviate a bit due to rounding errors and fluctuations caused by
	 * cfqgs entering and leaving the service tree.
	 */
	unsigned int vfraction;

Tejun Heo's avatar
Tejun Heo committed
250 251 252 253 254 255
	/*
	 * There are two weights - (internal) weight is the weight of this
	 * cfqg against the sibling cfqgs.  leaf_weight is the wight of
	 * this cfqg against the child cfqgs.  For the root cfqg, both
	 * weights are kept in sync for backward compatibility.
	 */
256
	unsigned int weight;
257
	unsigned int new_weight;
258
	unsigned int dev_weight;
259

Tejun Heo's avatar
Tejun Heo committed
260 261 262 263
	unsigned int leaf_weight;
	unsigned int new_leaf_weight;
	unsigned int dev_leaf_weight;

264 265 266
	/* number of cfqq currently on this group */
	int nr_cfqq;

267
	/*
268
	 * Per group busy queues average. Useful for workload slice calc. We
269 270 271 272 273 274 275 276 277 278 279
	 * create the array for each prio class but at run time it is used
	 * only for RT and BE class and slot for IDLE class remains unused.
	 * This is primarily done to avoid confusion and a gcc warning.
	 */
	unsigned int busy_queues_avg[CFQ_PRIO_NR];
	/*
	 * rr lists of queues with requests. We maintain service trees for
	 * RT and BE classes. These trees are subdivided in subclasses
	 * of SYNC, SYNC_NOIDLE and ASYNC based on workload type. For IDLE
	 * class there is no subclassification and all the cfq queues go on
	 * a single tree service_tree_idle.
280 281 282 283
	 * Counts are embedded in the cfq_rb_root
	 */
	struct cfq_rb_root service_trees[2][3];
	struct cfq_rb_root service_tree_idle;
284

285 286 287
	unsigned long saved_wl_slice;
	enum wl_type_t saved_wl_type;
	enum wl_class_t saved_wl_class;
288

289 290
	/* number of requests that are on the dispatch list or inside driver */
	int dispatched;
291
	struct cfq_ttime ttime;
292
	struct cfqg_stats stats;
293
};
294

295 296 297 298
struct cfq_io_cq {
	struct io_cq		icq;		/* must be the first member */
	struct cfq_queue	*cfqq[2];
	struct cfq_ttime	ttime;
Tejun Heo's avatar
Tejun Heo committed
299 300 301 302
	int			ioprio;		/* the current ioprio */
#ifdef CONFIG_CFQ_GROUP_IOSCHED
	uint64_t		blkcg_id;	/* the current blkcg ID */
#endif
303 304
};

305 306 307
/*
 * Per block device queue structure
 */
Linus Torvalds's avatar
Linus Torvalds committed
308
struct cfq_data {
309
	struct request_queue *queue;
310 311
	/* Root service tree for cfq_groups */
	struct cfq_rb_root grp_service_tree;
312
	struct cfq_group *root_group;
313

314 315
	/*
	 * The priority currently being served
316
	 */
317 318
	enum wl_class_t serving_wl_class;
	enum wl_type_t serving_wl_type;
319
	unsigned long workload_expires;
320
	struct cfq_group *serving_group;
321 322 323 324 325 326 327 328

	/*
	 * Each priority tree is sorted by next_request position.  These
	 * trees are used when determining if two or more queues are
	 * interleaving requests (see cfq_close_cooperator).
	 */
	struct rb_root prio_trees[CFQ_PRIO_LISTS];

329
	unsigned int busy_queues;
330
	unsigned int busy_sync_queues;
331

332 333
	int rq_in_driver;
	int rq_in_flight[2];
334 335 336 337 338

	/*
	 * queue-depth detection
	 */
	int rq_queued;
339
	int hw_tag;
340 341 342 343 344 345 346 347
	/*
	 * hw_tag can be
	 * -1 => indeterminate, (cfq will behave as if NCQ is present, to allow better detection)
	 *  1 => NCQ is present (hw_tag_est_depth is the estimated max depth)
	 *  0 => no NCQ
	 */
	int hw_tag_est_depth;
	unsigned int hw_tag_samples;
Linus Torvalds's avatar
Linus Torvalds committed
348

349 350 351 352
	/*
	 * idle window management
	 */
	struct timer_list idle_slice_timer;
353
	struct work_struct unplug_work;
Linus Torvalds's avatar
Linus Torvalds committed
354

355
	struct cfq_queue *active_queue;
356
	struct cfq_io_cq *active_cic;
357

358 359 360 361 362
	/*
	 * async queue for each priority case
	 */
	struct cfq_queue *async_cfqq[2][IOPRIO_BE_NR];
	struct cfq_queue *async_idle_cfqq;
363

Jens Axboe's avatar
Jens Axboe committed
364
	sector_t last_position;
Linus Torvalds's avatar
Linus Torvalds committed
365 366 367 368 369

	/*
	 * tunables, see top of file
	 */
	unsigned int cfq_quantum;
370
	unsigned int cfq_fifo_expire[2];
Linus Torvalds's avatar
Linus Torvalds committed
371 372
	unsigned int cfq_back_penalty;
	unsigned int cfq_back_max;
373 374 375
	unsigned int cfq_slice[2];
	unsigned int cfq_slice_async_rq;
	unsigned int cfq_slice_idle;
376
	unsigned int cfq_group_idle;
377
	unsigned int cfq_latency;
378
	unsigned int cfq_target_latency;
379

380 381 382 383
	/*
	 * Fallback dummy cfqq for extreme OOM conditions
	 */
	struct cfq_queue oom_cfqq;
384

385
	unsigned long last_delayed_sync;
Linus Torvalds's avatar
Linus Torvalds committed
386 387
};

388 389
static struct cfq_group *cfq_get_next_cfqg(struct cfq_data *cfqd);

390
static struct cfq_rb_root *st_for(struct cfq_group *cfqg,
391
					    enum wl_class_t class,
392
					    enum wl_type_t type)
393
{
394 395 396
	if (!cfqg)
		return NULL;

397
	if (class == IDLE_WORKLOAD)
398
		return &cfqg->service_tree_idle;
399

400
	return &cfqg->service_trees[class][type];
401 402
}

Jens Axboe's avatar
Jens Axboe committed
403
enum cfqq_state_flags {
404 405
	CFQ_CFQQ_FLAG_on_rr = 0,	/* on round-robin busy list */
	CFQ_CFQQ_FLAG_wait_request,	/* waiting for a request */
406
	CFQ_CFQQ_FLAG_must_dispatch,	/* must be allowed a dispatch */
407 408 409 410
	CFQ_CFQQ_FLAG_must_alloc_slice,	/* per-slice must_alloc flag */
	CFQ_CFQQ_FLAG_fifo_expire,	/* FIFO checked in this slice */
	CFQ_CFQQ_FLAG_idle_window,	/* slice idling enabled */
	CFQ_CFQQ_FLAG_prio_changed,	/* task priority has changed */
411
	CFQ_CFQQ_FLAG_slice_new,	/* no requests dispatched in slice */
412
	CFQ_CFQQ_FLAG_sync,		/* synchronous queue */
413
	CFQ_CFQQ_FLAG_coop,		/* cfqq is shared */
414
	CFQ_CFQQ_FLAG_split_coop,	/* shared cfqq will be splitted */
415
	CFQ_CFQQ_FLAG_deep,		/* sync cfqq experienced large depth */
416
	CFQ_CFQQ_FLAG_wait_busy,	/* Waiting for next request */
Jens Axboe's avatar
Jens Axboe committed
417 418 419 420 421
};

#define CFQ_CFQQ_FNS(name)						\
static inline void cfq_mark_cfqq_##name(struct cfq_queue *cfqq)		\
{									\
422
	(cfqq)->flags |= (1 << CFQ_CFQQ_FLAG_##name);			\
Jens Axboe's avatar
Jens Axboe committed
423 424 425
}									\
static inline void cfq_clear_cfqq_##name(struct cfq_queue *cfqq)	\
{									\
426
	(cfqq)->flags &= ~(1 << CFQ_CFQQ_FLAG_##name);			\
Jens Axboe's avatar
Jens Axboe committed
427 428 429
}									\
static inline int cfq_cfqq_##name(const struct cfq_queue *cfqq)		\
{									\
430
	return ((cfqq)->flags & (1 << CFQ_CFQQ_FLAG_##name)) != 0;	\
Jens Axboe's avatar
Jens Axboe committed
431 432 433 434
}

CFQ_CFQQ_FNS(on_rr);
CFQ_CFQQ_FNS(wait_request);
435
CFQ_CFQQ_FNS(must_dispatch);
Jens Axboe's avatar
Jens Axboe committed
436 437 438 439
CFQ_CFQQ_FNS(must_alloc_slice);
CFQ_CFQQ_FNS(fifo_expire);
CFQ_CFQQ_FNS(idle_window);
CFQ_CFQQ_FNS(prio_changed);
440
CFQ_CFQQ_FNS(slice_new);
441
CFQ_CFQQ_FNS(sync);
442
CFQ_CFQQ_FNS(coop);
443
CFQ_CFQQ_FNS(split_coop);
444
CFQ_CFQQ_FNS(deep);
445
CFQ_CFQQ_FNS(wait_busy);
Jens Axboe's avatar
Jens Axboe committed
446 447
#undef CFQ_CFQQ_FNS

448 449 450 451 452 453 454 455 456 457
static inline struct cfq_group *pd_to_cfqg(struct blkg_policy_data *pd)
{
	return pd ? container_of(pd, struct cfq_group, pd) : NULL;
}

static inline struct blkcg_gq *cfqg_to_blkg(struct cfq_group *cfqg)
{
	return pd_to_blkg(&cfqg->pd);
}

458
#if defined(CONFIG_CFQ_GROUP_IOSCHED) && defined(CONFIG_DEBUG_BLK_CGROUP)
459

460 461 462 463 464
/* cfqg stats flags */
enum cfqg_stats_flags {
	CFQG_stats_waiting = 0,
	CFQG_stats_idling,
	CFQG_stats_empty,
465 466
};

467 468
#define CFQG_FLAG_FNS(name)						\
static inline void cfqg_stats_mark_##name(struct cfqg_stats *stats)	\
469
{									\
470
	stats->flags |= (1 << CFQG_stats_##name);			\
471
}									\
472
static inline void cfqg_stats_clear_##name(struct cfqg_stats *stats)	\
473
{									\
474
	stats->flags &= ~(1 << CFQG_stats_##name);			\
475
}									\
476
static inline int cfqg_stats_##name(struct cfqg_stats *stats)		\
477
{									\
478
	return (stats->flags & (1 << CFQG_stats_##name)) != 0;		\
479 480
}									\

481 482 483 484
CFQG_FLAG_FNS(waiting)
CFQG_FLAG_FNS(idling)
CFQG_FLAG_FNS(empty)
#undef CFQG_FLAG_FNS
485 486

/* This should be called with the queue_lock held. */
487
static void cfqg_stats_update_group_wait_time(struct cfqg_stats *stats)
488 489 490
{
	unsigned long long now;

491
	if (!cfqg_stats_waiting(stats))
492 493 494 495 496 497
		return;

	now = sched_clock();
	if (time_after64(now, stats->start_group_wait_time))
		blkg_stat_add(&stats->group_wait_time,
			      now - stats->start_group_wait_time);
498
	cfqg_stats_clear_waiting(stats);
499 500 501
}

/* This should be called with the queue_lock held. */
502 503
static void cfqg_stats_set_start_group_wait_time(struct cfq_group *cfqg,
						 struct cfq_group *curr_cfqg)
504
{
505
	struct cfqg_stats *stats = &cfqg->stats;
506

507
	if (cfqg_stats_waiting(stats))
508
		return;
509
	if (cfqg == curr_cfqg)
510
		return;
511 512
	stats->start_group_wait_time = sched_clock();
	cfqg_stats_mark_waiting(stats);
513 514 515
}

/* This should be called with the queue_lock held. */
516
static void cfqg_stats_end_empty_time(struct cfqg_stats *stats)
517 518 519
{
	unsigned long long now;

520
	if (!cfqg_stats_empty(stats))
521 522 523 524 525 526
		return;

	now = sched_clock();
	if (time_after64(now, stats->start_empty_time))
		blkg_stat_add(&stats->empty_time,
			      now - stats->start_empty_time);
527
	cfqg_stats_clear_empty(stats);
528 529
}

530
static void cfqg_stats_update_dequeue(struct cfq_group *cfqg)
531
{
532
	blkg_stat_add(&cfqg->stats.dequeue, 1);
533 534
}

535
static void cfqg_stats_set_start_empty_time(struct cfq_group *cfqg)
536
{
537
	struct cfqg_stats *stats = &cfqg->stats;
538

539
	if (blkg_rwstat_total(&stats->queued))
540 541 542 543 544 545 546
		return;

	/*
	 * group is already marked empty. This can happen if cfqq got new
	 * request in parent group and moved to this group while being added
	 * to service tree. Just ignore the event and move on.
	 */
547
	if (cfqg_stats_empty(stats))
548 549 550
		return;

	stats->start_empty_time = sched_clock();
551
	cfqg_stats_mark_empty(stats);
552 553
}

554
static void cfqg_stats_update_idle_time(struct cfq_group *cfqg)
555
{
556
	struct cfqg_stats *stats = &cfqg->stats;
557

558
	if (cfqg_stats_idling(stats)) {
559 560 561 562 563
		unsigned long long now = sched_clock();

		if (time_after64(now, stats->start_idle_time))
			blkg_stat_add(&stats->idle_time,
				      now - stats->start_idle_time);
564
		cfqg_stats_clear_idling(stats);
565 566 567
	}
}

568
static void cfqg_stats_set_start_idle_time(struct cfq_group *cfqg)
569
{
570
	struct cfqg_stats *stats = &cfqg->stats;
571

572
	BUG_ON(cfqg_stats_idling(stats));
573 574

	stats->start_idle_time = sched_clock();
575
	cfqg_stats_mark_idling(stats);
576 577
}

578
static void cfqg_stats_update_avg_queue_size(struct cfq_group *cfqg)
579
{
580
	struct cfqg_stats *stats = &cfqg->stats;
581 582

	blkg_stat_add(&stats->avg_queue_size_sum,
583
		      blkg_rwstat_total(&stats->queued));
584
	blkg_stat_add(&stats->avg_queue_size_samples, 1);
585
	cfqg_stats_update_group_wait_time(stats);
586 587 588 589
}

#else	/* CONFIG_CFQ_GROUP_IOSCHED && CONFIG_DEBUG_BLK_CGROUP */

Tejun Heo's avatar
Tejun Heo committed
590 591 592 593 594 595 596
static inline void cfqg_stats_set_start_group_wait_time(struct cfq_group *cfqg, struct cfq_group *curr_cfqg) { }
static inline void cfqg_stats_end_empty_time(struct cfqg_stats *stats) { }
static inline void cfqg_stats_update_dequeue(struct cfq_group *cfqg) { }
static inline void cfqg_stats_set_start_empty_time(struct cfq_group *cfqg) { }
static inline void cfqg_stats_update_idle_time(struct cfq_group *cfqg) { }
static inline void cfqg_stats_set_start_idle_time(struct cfq_group *cfqg) { }
static inline void cfqg_stats_update_avg_queue_size(struct cfq_group *cfqg) { }
597 598 599 600

#endif	/* CONFIG_CFQ_GROUP_IOSCHED && CONFIG_DEBUG_BLK_CGROUP */

#ifdef CONFIG_CFQ_GROUP_IOSCHED
601

602 603 604 605 606 607 608
static struct blkcg_policy blkcg_policy_cfq;

static inline struct cfq_group *blkg_to_cfqg(struct blkcg_gq *blkg)
{
	return pd_to_cfqg(blkg_to_pd(blkg, &blkcg_policy_cfq));
}

609
static inline struct cfq_group *cfqg_parent(struct cfq_group *cfqg)
610
{
611
	struct blkcg_gq *pblkg = cfqg_to_blkg(cfqg)->parent;
612

613
	return pblkg ? blkg_to_cfqg(pblkg) : NULL;
614 615
}

616 617 618 619 620 621 622 623 624 625
static inline void cfqg_get(struct cfq_group *cfqg)
{
	return blkg_get(cfqg_to_blkg(cfqg));
}

static inline void cfqg_put(struct cfq_group *cfqg)
{
	return blkg_put(cfqg_to_blkg(cfqg));
}

Tejun Heo's avatar
Tejun Heo committed
626 627 628 629
#define cfq_log_cfqq(cfqd, cfqq, fmt, args...)	do {			\
	char __pbuf[128];						\
									\
	blkg_path(cfqg_to_blkg((cfqq)->cfqg), __pbuf, sizeof(__pbuf));	\
630 631 632
	blk_add_trace_msg((cfqd)->queue, "cfq%d%c%c %s " fmt, (cfqq)->pid, \
			cfq_cfqq_sync((cfqq)) ? 'S' : 'A',		\
			cfqq_type((cfqq)) == SYNC_NOIDLE_WORKLOAD ? 'N' : ' ',\
Tejun Heo's avatar
Tejun Heo committed
633 634 635 636 637 638 639 640 641
			  __pbuf, ##args);				\
} while (0)

#define cfq_log_cfqg(cfqd, cfqg, fmt, args...)	do {			\
	char __pbuf[128];						\
									\
	blkg_path(cfqg_to_blkg(cfqg), __pbuf, sizeof(__pbuf));		\
	blk_add_trace_msg((cfqd)->queue, "%s " fmt, __pbuf, ##args);	\
} while (0)
642

643 644
static inline void cfqg_stats_update_io_add(struct cfq_group *cfqg,
					    struct cfq_group *curr_cfqg, int rw)
645
{
646 647 648
	blkg_rwstat_add(&cfqg->stats.queued, rw, 1);
	cfqg_stats_end_empty_time(&cfqg->stats);
	cfqg_stats_set_start_group_wait_time(cfqg, curr_cfqg);
649 650
}

651 652
static inline void cfqg_stats_update_timeslice_used(struct cfq_group *cfqg,
			unsigned long time, unsigned long unaccounted_time)
653
{
654
	blkg_stat_add(&cfqg->stats.time, time);
655
#ifdef CONFIG_DEBUG_BLK_CGROUP
656
	blkg_stat_add(&cfqg->stats.unaccounted_time, unaccounted_time);
657
#endif
658 659
}

660
static inline void cfqg_stats_update_io_remove(struct cfq_group *cfqg, int rw)
661
{
662
	blkg_rwstat_add(&cfqg->stats.queued, rw, -1);
663 664
}

665
static inline void cfqg_stats_update_io_merged(struct cfq_group *cfqg, int rw)
666
{
667
	blkg_rwstat_add(&cfqg->stats.merged, rw, 1);
668 669
}

670 671
static inline void cfqg_stats_update_dispatch(struct cfq_group *cfqg,
					      uint64_t bytes, int rw)
672
{
673 674 675
	blkg_stat_add(&cfqg->stats.sectors, bytes >> 9);
	blkg_rwstat_add(&cfqg->stats.serviced, rw, 1);
	blkg_rwstat_add(&cfqg->stats.service_bytes, rw, bytes);
676 677
}

678 679
static inline void cfqg_stats_update_completion(struct cfq_group *cfqg,
			uint64_t start_time, uint64_t io_start_time, int rw)
680
{
681
	struct cfqg_stats *stats = &cfqg->stats;
682 683 684 685 686 687 688
	unsigned long long now = sched_clock();

	if (time_after64(now, io_start_time))
		blkg_rwstat_add(&stats->service_time, rw, now - io_start_time);
	if (time_after64(io_start_time, start_time))
		blkg_rwstat_add(&stats->wait_time, rw,
				io_start_time - start_time);
689 690
}

Tejun Heo's avatar
Tejun Heo committed
691
static void cfq_pd_reset_stats(struct blkcg_gq *blkg)
692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713
{
	struct cfq_group *cfqg = blkg_to_cfqg(blkg);
	struct cfqg_stats *stats = &cfqg->stats;

	/* queued stats shouldn't be cleared */
	blkg_rwstat_reset(&stats->service_bytes);
	blkg_rwstat_reset(&stats->serviced);
	blkg_rwstat_reset(&stats->merged);
	blkg_rwstat_reset(&stats->service_time);
	blkg_rwstat_reset(&stats->wait_time);
	blkg_stat_reset(&stats->time);
#ifdef CONFIG_DEBUG_BLK_CGROUP
	blkg_stat_reset(&stats->unaccounted_time);
	blkg_stat_reset(&stats->avg_queue_size_sum);
	blkg_stat_reset(&stats->avg_queue_size_samples);
	blkg_stat_reset(&stats->dequeue);
	blkg_stat_reset(&stats->group_wait_time);
	blkg_stat_reset(&stats->idle_time);
	blkg_stat_reset(&stats->empty_time);
#endif
}

714 715
#else	/* CONFIG_CFQ_GROUP_IOSCHED */

716
static inline struct cfq_group *cfqg_parent(struct cfq_group *cfqg) { return NULL; }
717 718 719
static inline void cfqg_get(struct cfq_group *cfqg) { }
static inline void cfqg_put(struct cfq_group *cfqg) { }

720
#define cfq_log_cfqq(cfqd, cfqq, fmt, args...)	\
721 722 723 724
	blk_add_trace_msg((cfqd)->queue, "cfq%d%c%c " fmt, (cfqq)->pid,	\
			cfq_cfqq_sync((cfqq)) ? 'S' : 'A',		\
			cfqq_type((cfqq)) == SYNC_NOIDLE_WORKLOAD ? 'N' : ' ',\
				##args)
725
#define cfq_log_cfqg(cfqd, cfqg, fmt, args...)		do {} while (0)
726

727 728 729 730 731 732 733 734 735 736
static inline void cfqg_stats_update_io_add(struct cfq_group *cfqg,
			struct cfq_group *curr_cfqg, int rw) { }
static inline void cfqg_stats_update_timeslice_used(struct cfq_group *cfqg,
			unsigned long time, unsigned long unaccounted_time) { }
static inline void cfqg_stats_update_io_remove(struct cfq_group *cfqg, int rw) { }
static inline void cfqg_stats_update_io_merged(struct cfq_group *cfqg, int rw) { }
static inline void cfqg_stats_update_dispatch(struct cfq_group *cfqg,
					      uint64_t bytes, int rw) { }
static inline void cfqg_stats_update_completion(struct cfq_group *cfqg,
			uint64_t start_time, uint64_t io_start_time, int rw) { }
737

738 739
#endif	/* CONFIG_CFQ_GROUP_IOSCHED */

740 741 742
#define cfq_log(cfqd, fmt, args...)	\
	blk_add_trace_msg((cfqd)->queue, "cfq " fmt, ##args)

743 744 745 746 747 748 749 750 751 752
/* Traverses through cfq group service trees */
#define for_each_cfqg_st(cfqg, i, j, st) \
	for (i = 0; i <= IDLE_WORKLOAD; i++) \
		for (j = 0, st = i < IDLE_WORKLOAD ? &cfqg->service_trees[i][j]\
			: &cfqg->service_tree_idle; \
			(i < IDLE_WORKLOAD && j <= SYNC_WORKLOAD) || \
			(i == IDLE_WORKLOAD && j == 0); \
			j++, st = i < IDLE_WORKLOAD ? \
			&cfqg->service_trees[i][j]: NULL) \

753 754 755 756 757 758 759 760 761 762 763 764
static inline bool cfq_io_thinktime_big(struct cfq_data *cfqd,
	struct cfq_ttime *ttime, bool group_idle)
{
	unsigned long slice;
	if (!sample_valid(ttime->ttime_samples))
		return false;
	if (group_idle)
		slice = cfqd->cfq_group_idle;
	else
		slice = cfqd->cfq_slice_idle;
	return ttime->ttime_mean > slice;
}
765

766 767 768 769 770 771 772 773 774 775 776 777 778 779 780
static inline bool iops_mode(struct cfq_data *cfqd)
{
	/*
	 * If we are not idling on queues and it is a NCQ drive, parallel
	 * execution of requests is on and measuring time is not possible
	 * in most of the cases until and unless we drive shallower queue
	 * depths and that becomes a performance bottleneck. In such cases
	 * switch to start providing fairness in terms of number of IOs.
	 */
	if (!cfqd->cfq_slice_idle && cfqd->hw_tag)
		return true;
	else
		return false;
}

781
static inline enum wl_class_t cfqq_class(struct cfq_queue *cfqq)
782 783 784 785 786 787 788 789
{
	if (cfq_class_idle(cfqq))
		return IDLE_WORKLOAD;
	if (cfq_class_rt(cfqq))
		return RT_WORKLOAD;
	return BE_WORKLOAD;
}

790 791 792 793 794 795 796 797 798 799

static enum wl_type_t cfqq_type(struct cfq_queue *cfqq)
{
	if (!cfq_cfqq_sync(cfqq))
		return ASYNC_WORKLOAD;
	if (!cfq_cfqq_idle_window(cfqq))
		return SYNC_NOIDLE_WORKLOAD;
	return SYNC_WORKLOAD;
}

800
static inline int cfq_group_busy_queues_wl(enum wl_class_t wl_class,
801 802
					struct cfq_data *cfqd,
					struct cfq_group *cfqg)
803
{
804
	if (wl_class == IDLE_WORKLOAD)
805
		return cfqg->service_tree_idle.count;
806

807 808 809
	return cfqg->service_trees[wl_class][ASYNC_WORKLOAD].count +
		cfqg->service_trees[wl_class][SYNC_NOIDLE_WORKLOAD].count +
		cfqg->service_trees[wl_class][SYNC_WORKLOAD].count;
810 811
}

812 813 814
static inline int cfqg_busy_async_queues(struct cfq_data *cfqd,
					struct cfq_group *cfqg)
{
815 816
	return cfqg->service_trees[RT_WORKLOAD][ASYNC_WORKLOAD].count +
		cfqg->service_trees[BE_WORKLOAD][ASYNC_WORKLOAD].count;
817 818
}

819
static void cfq_dispatch_insert(struct request_queue *, struct request *);
820
static struct cfq_queue *cfq_get_queue(struct cfq_data *cfqd, bool is_sync,
821
				       struct cfq_io_cq *cic, struct bio *bio,
822
				       gfp_t gfp_mask);
823

824 825 826 827 828 829
static inline struct cfq_io_cq *icq_to_cic(struct io_cq *icq)
{
	/* cic->icq is the first member, %NULL will convert to %NULL */
	return container_of(icq, struct cfq_io_cq, icq);
}

830 831 832 833 834 835 836 837
static inline struct cfq_io_cq *cfq_cic_lookup(struct cfq_data *cfqd,
					       struct io_context *ioc)
{
	if (ioc)
		return icq_to_cic(ioc_lookup_icq(ioc, cfqd->queue));
	return NULL;
}

838
static inline struct cfq_queue *cic_to_cfqq(struct cfq_io_cq *cic, bool is_sync)
839
{
840
	return cic->cfqq[is_sync];
841 842
}

843 844
static inline void cic_set_cfqq(struct cfq_io_cq *cic, struct cfq_queue *cfqq,
				bool is_sync)
845
{
846
	cic->cfqq[is_sync] = cfqq;
847 848
}

849
static inline struct cfq_data *cic_to_cfqd(struct cfq_io_cq *cic)
850
{
851
	return cic->icq.q->elevator->elevator_data;
852 853
}

854 855 856 857
/*
 * We regard a request as SYNC, if it's either a read or has the SYNC bit
 * set (in which case it could also be direct WRITE).
 */
858
static inline bool cfq_bio_sync(struct bio *bio)
859
{
860
	return bio_data_dir(bio) == READ || (bio->bi_rw & REQ_SYNC);
861
}
Linus Torvalds's avatar
Linus Torvalds committed
862

Andrew Morton's avatar
Andrew Morton committed
863 864 865 866
/*
 * scheduler run of queue, if there are requests pending and no one in the
 * driver that will restart queueing
 */
867
static inline void cfq_schedule_dispatch(struct cfq_data *cfqd)
Andrew Morton's avatar
Andrew Morton committed
868
{
869 870
	if (cfqd->busy_queues) {
		cfq_log(cfqd, "schedule dispatch");
871
		kblockd_schedule_work(cfqd->queue, &cfqd->unplug_work);
872
	}
Andrew Morton's avatar
Andrew Morton committed
873 874
}

875 876 877 878 879
/*
 * Scale schedule slice based on io priority. Use the sync time slice only
 * if a queue is marked sync and has sync io queued. A sync queue with async
 * io only, should not get full sync slice length.
 */
880
static inline int cfq_prio_slice(struct cfq_data *cfqd, bool sync,
881
				 unsigned short prio)
882
{
883
	const int base_slice = cfqd->cfq_slice[sync];
884

885 886 887 888
	WARN_ON(prio >= IOPRIO_BE_NR);

	return base_slice + (base_slice/CFQ_SLICE_SCALE * (4 - prio));
}
889

890 891 892 893
static inline int
cfq_prio_to_slice(struct cfq_data *cfqd, struct cfq_queue *cfqq)
{
	return cfq_prio_slice(cfqd, cfq_cfqq_sync(cfqq), cfqq->ioprio);
894 895
}

896 897 898 899 900 901 902 903 904 905 906 907 908 909
/**
 * cfqg_scale_charge - scale disk time charge according to cfqg weight
 * @charge: disk time being charged
 * @vfraction: vfraction of the cfqg, fixed point w/ CFQ_SERVICE_SHIFT
 *
 * Scale @charge according to @vfraction, which is in range (0, 1].  The
 * scaling is inversely proportional.
 *
 * scaled = charge / vfraction
 *
 * The result is also in fixed point w/ CFQ_SERVICE_SHIFT.
 */
static inline u64 cfqg_scale_charge(unsigned long charge,
				    unsigned int vfraction)
910
{
911
	u64 c = charge << CFQ_SERVICE_SHIFT;	/* make it fixed point */
912

913 914 915 916
	/* charge / vfraction */
	c <<= CFQ_SERVICE_SHIFT;
	do_div(c, vfraction);
	return c;
917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942
}

static inline u64 max_vdisktime(u64 min_vdisktime, u64 vdisktime)
{
	s64 delta = (s64)(vdisktime - min_vdisktime);
	if (delta > 0)
		min_vdisktime = vdisktime;

	return min_vdisktime;
}

static inline u64 min_vdisktime(u64 min_vdisktime, u64 vdisktime)
{
	s64 delta = (s64)(vdisktime - min_vdisktime);
	if (delta < 0)
		min_vdisktime = vdisktime;

	return min_vdisktime;
}

static void update_min_vdisktime(struct cfq_rb_root *st)
{
	struct cfq_group *cfqg;

	if (st->left) {
		cfqg = rb_entry_cfqg(st->left);
943 944
		st->min_vdisktime = max_vdisktime(st->min_vdisktime,
						  cfqg->vdisktime);
945 946 947
	}
}

948 949 950 951 952 953
/*
 * get averaged number of queues of RT/BE priority.
 * average is updated, with a formula that gives more weight to higher numbers,
 * to quickly follows sudden increases and decrease slowly
 */

954 955
static inline unsigned cfq_group_get_avg_queues(struct cfq_data *cfqd,
					struct cfq_group *cfqg, bool rt)
956
{
957 958 959
	unsigned min_q, max_q;
	unsigned mult  = cfq_hist_divisor - 1;
	unsigned round = cfq_hist_divisor / 2;
960
	unsigned busy = cfq_group_busy_queues_wl(rt, cfqd, cfqg);
961

962 963 964
	min_q = min(cfqg->busy_queues_avg[rt], busy);
	max_q = max(cfqg->busy_queues_avg[rt], busy);
	cfqg->busy_queues_avg[rt] = (mult * max_q + min_q + round) /
965
		cfq_hist_divisor;
966 967 968 969 970 971
	return cfqg->busy_queues_avg[rt];
}

static inline unsigned
cfq_group_slice(struct cfq_data *cfqd, struct cfq_group *cfqg)
{
972
	return cfqd->cfq_target_latency * cfqg->vfraction >> CFQ_SERVICE_SHIFT;
973 974
}

975
static inline unsigned
976
cfq_scaled_cfqq_slice(struct cfq_data *cfqd, struct cfq_queue *cfqq)
977
{
978 979
	unsigned slice = cfq_prio_to_slice(cfqd, cfqq);
	if (cfqd->cfq_latency) {
980 981 982 983 984 985
		/*
		 * interested queues (we consider only the ones with the same
		 * priority class in the cfq group)
		 */
		unsigned iq = cfq_group_get_avg_queues(cfqd, cfqq->cfqg,
						cfq_class_rt(cfqq));
986 987
		unsigned sync_slice = cfqd->cfq_slice[1];
		unsigned expect_latency = sync_slice * iq;
988 989 990
		unsigned group_slice = cfq_group_slice(cfqd, cfqq->cfqg);

		if (expect_latency > group_slice) {
991 992 993 994 995 996 997
			unsigned base_low_slice = 2 * cfqd->cfq_slice_idle;
			/* scale low_slice according to IO priority
			 * and sync vs async */
			unsigned low_slice =
				min(slice, base_low_slice * slice / sync_slice);
			/* the adapted slice value is scaled to fit all iqs
			 * into the target latency */
998
			slice = max(slice * group_slice / expect_latency,
999 1000 1001
				    low_slice);
		}
	}
1002 1003 1004 1005 1006 1007
	return slice;
}

static inline void
cfq_set_prio_slice(struct cfq_data *cfqd, struct cfq_queue *cfqq)
{
1008
	unsigned slice = cfq_scaled_cfqq_slice(cfqd, cfqq);
1009

1010
	cfqq->slice_start = jiffies;
1011
	cfqq->slice_end = jiffies + slice;
1012
	cfqq->allocated_slice = slice;