cfq-iosched.c 108 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
	unsigned total_weight;
89
	u64 min_vdisktime;
90
	struct cfq_ttime ttime;
91
};
92 93
#define CFQ_RB_ROOT	(struct cfq_rb_root) { .rb = RB_ROOT, \
			.ttime = {.last_end_request = jiffies,},}
94

95 96 97 98 99
/*
 * Per process-grouping structure
 */
struct cfq_queue {
	/* reference count */
100
	int ref;
101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123
	/* 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;

124 125
	/* time when queue got scheduled in to dispatch first request. */
	unsigned long dispatch_start;
126
	unsigned int allocated_slice;
127
	unsigned int slice_dispatch;
128 129
	/* time when first request from queue completed and slice started. */
	unsigned long slice_start;
130 131 132
	unsigned long slice_end;
	long slice_resid;

133 134
	/* pending priority requests */
	int prio_pending;
135 136 137 138 139
	/* number of requests that are on the dispatch list or inside driver */
	int dispatched;

	/* io prio of this group */
	unsigned short ioprio, org_ioprio;
140
	unsigned short ioprio_class;
141

142 143
	pid_t pid;

144
	u32 seek_history;
145 146
	sector_t last_request_pos;

147
	struct cfq_rb_root *service_tree;
Jeff Moyer's avatar
Jeff Moyer committed
148
	struct cfq_queue *new_cfqq;
149
	struct cfq_group *cfqg;
150 151
	/* Number of sectors dispatched from queue in single dispatch round */
	unsigned long nr_sectors;
152 153
};

154
/*
155
 * First index in the service_trees.
156 157
 * IDLE is handled separately, so it has negative index
 */
158
enum wl_class_t {
159
	BE_WORKLOAD = 0,
160 161
	RT_WORKLOAD = 1,
	IDLE_WORKLOAD = 2,
162
	CFQ_PRIO_NR,
163 164
};

165 166 167 168 169 170 171 172 173
/*
 * Second index in the service_trees.
 */
enum wl_type_t {
	ASYNC_WORKLOAD = 0,
	SYNC_NOIDLE_WORKLOAD = 1,
	SYNC_WORKLOAD = 2
};

174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202
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
203
	/* time spent idling for this blkcg_gq */
204 205 206 207 208 209 210 211 212 213 214 215
	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 */
};

216 217
/* This is per cgroup per device grouping structure */
struct cfq_group {
218 219 220
	/* must be the first member */
	struct blkg_policy_data pd;

221 222 223 224 225
	/* group service_tree member */
	struct rb_node rb_node;

	/* group service_tree key */
	u64 vdisktime;
226
	unsigned int weight;
227
	unsigned int new_weight;
228
	unsigned int dev_weight;
229 230 231 232

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

233
	/*
234
	 * Per group busy queues average. Useful for workload slice calc. We
235 236 237 238 239 240 241 242 243 244 245
	 * 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.
246 247 248 249
	 * Counts are embedded in the cfq_rb_root
	 */
	struct cfq_rb_root service_trees[2][3];
	struct cfq_rb_root service_tree_idle;
250

251 252 253
	unsigned long saved_wl_slice;
	enum wl_type_t saved_wl_type;
	enum wl_class_t saved_wl_class;
254

255 256
	/* number of requests that are on the dispatch list or inside driver */
	int dispatched;
257
	struct cfq_ttime ttime;
258
	struct cfqg_stats stats;
259
};
260

261 262 263 264
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
265 266 267 268
	int			ioprio;		/* the current ioprio */
#ifdef CONFIG_CFQ_GROUP_IOSCHED
	uint64_t		blkcg_id;	/* the current blkcg ID */
#endif
269 270
};

271 272 273
/*
 * Per block device queue structure
 */
Linus Torvalds's avatar
Linus Torvalds committed
274
struct cfq_data {
275
	struct request_queue *queue;
276 277
	/* Root service tree for cfq_groups */
	struct cfq_rb_root grp_service_tree;
278
	struct cfq_group *root_group;
279

280 281
	/*
	 * The priority currently being served
282
	 */
283 284
	enum wl_class_t serving_wl_class;
	enum wl_type_t serving_wl_type;
285
	unsigned long workload_expires;
286
	struct cfq_group *serving_group;
287 288 289 290 291 292 293 294

	/*
	 * 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];

295
	unsigned int busy_queues;
296
	unsigned int busy_sync_queues;
297

298 299
	int rq_in_driver;
	int rq_in_flight[2];
300 301 302 303 304

	/*
	 * queue-depth detection
	 */
	int rq_queued;
305
	int hw_tag;
306 307 308 309 310 311 312 313
	/*
	 * 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
314

315 316 317 318
	/*
	 * idle window management
	 */
	struct timer_list idle_slice_timer;
319
	struct work_struct unplug_work;
Linus Torvalds's avatar
Linus Torvalds committed
320

321
	struct cfq_queue *active_queue;
322
	struct cfq_io_cq *active_cic;
323

324 325 326 327 328
	/*
	 * async queue for each priority case
	 */
	struct cfq_queue *async_cfqq[2][IOPRIO_BE_NR];
	struct cfq_queue *async_idle_cfqq;
329

Jens Axboe's avatar
Jens Axboe committed
330
	sector_t last_position;
Linus Torvalds's avatar
Linus Torvalds committed
331 332 333 334 335

	/*
	 * tunables, see top of file
	 */
	unsigned int cfq_quantum;
336
	unsigned int cfq_fifo_expire[2];
Linus Torvalds's avatar
Linus Torvalds committed
337 338
	unsigned int cfq_back_penalty;
	unsigned int cfq_back_max;
339 340 341
	unsigned int cfq_slice[2];
	unsigned int cfq_slice_async_rq;
	unsigned int cfq_slice_idle;
342
	unsigned int cfq_group_idle;
343
	unsigned int cfq_latency;
344
	unsigned int cfq_target_latency;
345

346 347 348 349
	/*
	 * Fallback dummy cfqq for extreme OOM conditions
	 */
	struct cfq_queue oom_cfqq;
350

351
	unsigned long last_delayed_sync;
Linus Torvalds's avatar
Linus Torvalds committed
352 353
};

354 355
static struct cfq_group *cfq_get_next_cfqg(struct cfq_data *cfqd);

356
static struct cfq_rb_root *service_tree_for(struct cfq_group *cfqg,
357
					    enum wl_class_t class,
358
					    enum wl_type_t type)
359
{
360 361 362
	if (!cfqg)
		return NULL;

363
	if (class == IDLE_WORKLOAD)
364
		return &cfqg->service_tree_idle;
365

366
	return &cfqg->service_trees[class][type];
367 368
}

Jens Axboe's avatar
Jens Axboe committed
369
enum cfqq_state_flags {
370 371
	CFQ_CFQQ_FLAG_on_rr = 0,	/* on round-robin busy list */
	CFQ_CFQQ_FLAG_wait_request,	/* waiting for a request */
372
	CFQ_CFQQ_FLAG_must_dispatch,	/* must be allowed a dispatch */
373 374 375 376
	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 */
377
	CFQ_CFQQ_FLAG_slice_new,	/* no requests dispatched in slice */
378
	CFQ_CFQQ_FLAG_sync,		/* synchronous queue */
379
	CFQ_CFQQ_FLAG_coop,		/* cfqq is shared */
380
	CFQ_CFQQ_FLAG_split_coop,	/* shared cfqq will be splitted */
381
	CFQ_CFQQ_FLAG_deep,		/* sync cfqq experienced large depth */
382
	CFQ_CFQQ_FLAG_wait_busy,	/* Waiting for next request */
Jens Axboe's avatar
Jens Axboe committed
383 384 385 386 387
};

#define CFQ_CFQQ_FNS(name)						\
static inline void cfq_mark_cfqq_##name(struct cfq_queue *cfqq)		\
{									\
388
	(cfqq)->flags |= (1 << CFQ_CFQQ_FLAG_##name);			\
Jens Axboe's avatar
Jens Axboe committed
389 390 391
}									\
static inline void cfq_clear_cfqq_##name(struct cfq_queue *cfqq)	\
{									\
392
	(cfqq)->flags &= ~(1 << CFQ_CFQQ_FLAG_##name);			\
Jens Axboe's avatar
Jens Axboe committed
393 394 395
}									\
static inline int cfq_cfqq_##name(const struct cfq_queue *cfqq)		\
{									\
396
	return ((cfqq)->flags & (1 << CFQ_CFQQ_FLAG_##name)) != 0;	\
Jens Axboe's avatar
Jens Axboe committed
397 398 399 400
}

CFQ_CFQQ_FNS(on_rr);
CFQ_CFQQ_FNS(wait_request);
401
CFQ_CFQQ_FNS(must_dispatch);
Jens Axboe's avatar
Jens Axboe committed
402 403 404 405
CFQ_CFQQ_FNS(must_alloc_slice);
CFQ_CFQQ_FNS(fifo_expire);
CFQ_CFQQ_FNS(idle_window);
CFQ_CFQQ_FNS(prio_changed);
406
CFQ_CFQQ_FNS(slice_new);
407
CFQ_CFQQ_FNS(sync);
408
CFQ_CFQQ_FNS(coop);
409
CFQ_CFQQ_FNS(split_coop);
410
CFQ_CFQQ_FNS(deep);
411
CFQ_CFQQ_FNS(wait_busy);
Jens Axboe's avatar
Jens Axboe committed
412 413
#undef CFQ_CFQQ_FNS

414 415 416 417 418 419 420 421 422 423
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);
}

424
#if defined(CONFIG_CFQ_GROUP_IOSCHED) && defined(CONFIG_DEBUG_BLK_CGROUP)
425

426 427 428 429 430
/* cfqg stats flags */
enum cfqg_stats_flags {
	CFQG_stats_waiting = 0,
	CFQG_stats_idling,
	CFQG_stats_empty,
431 432
};

433 434
#define CFQG_FLAG_FNS(name)						\
static inline void cfqg_stats_mark_##name(struct cfqg_stats *stats)	\
435
{									\
436
	stats->flags |= (1 << CFQG_stats_##name);			\
437
}									\
438
static inline void cfqg_stats_clear_##name(struct cfqg_stats *stats)	\
439
{									\
440
	stats->flags &= ~(1 << CFQG_stats_##name);			\
441
}									\
442
static inline int cfqg_stats_##name(struct cfqg_stats *stats)		\
443
{									\
444
	return (stats->flags & (1 << CFQG_stats_##name)) != 0;		\
445 446
}									\

447 448 449 450
CFQG_FLAG_FNS(waiting)
CFQG_FLAG_FNS(idling)
CFQG_FLAG_FNS(empty)
#undef CFQG_FLAG_FNS
451 452

/* This should be called with the queue_lock held. */
453
static void cfqg_stats_update_group_wait_time(struct cfqg_stats *stats)
454 455 456
{
	unsigned long long now;

457
	if (!cfqg_stats_waiting(stats))
458 459 460 461 462 463
		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);
464
	cfqg_stats_clear_waiting(stats);
465 466 467
}

/* This should be called with the queue_lock held. */
468 469
static void cfqg_stats_set_start_group_wait_time(struct cfq_group *cfqg,
						 struct cfq_group *curr_cfqg)
470
{
471
	struct cfqg_stats *stats = &cfqg->stats;
472

473
	if (cfqg_stats_waiting(stats))
474
		return;
475
	if (cfqg == curr_cfqg)
476
		return;
477 478
	stats->start_group_wait_time = sched_clock();
	cfqg_stats_mark_waiting(stats);
479 480 481
}

/* This should be called with the queue_lock held. */
482
static void cfqg_stats_end_empty_time(struct cfqg_stats *stats)
483 484 485
{
	unsigned long long now;

486
	if (!cfqg_stats_empty(stats))
487 488 489 490 491 492
		return;

	now = sched_clock();
	if (time_after64(now, stats->start_empty_time))
		blkg_stat_add(&stats->empty_time,
			      now - stats->start_empty_time);
493
	cfqg_stats_clear_empty(stats);
494 495
}

496
static void cfqg_stats_update_dequeue(struct cfq_group *cfqg)
497
{
498
	blkg_stat_add(&cfqg->stats.dequeue, 1);
499 500
}

501
static void cfqg_stats_set_start_empty_time(struct cfq_group *cfqg)
502
{
503
	struct cfqg_stats *stats = &cfqg->stats;
504 505 506 507 508 509 510 511 512

	if (blkg_rwstat_sum(&stats->queued))
		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.
	 */
513
	if (cfqg_stats_empty(stats))
514 515 516
		return;

	stats->start_empty_time = sched_clock();
517
	cfqg_stats_mark_empty(stats);
518 519
}

520
static void cfqg_stats_update_idle_time(struct cfq_group *cfqg)
521
{
522
	struct cfqg_stats *stats = &cfqg->stats;
523

524
	if (cfqg_stats_idling(stats)) {
525 526 527 528 529
		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);
530
		cfqg_stats_clear_idling(stats);
531 532 533
	}
}

534
static void cfqg_stats_set_start_idle_time(struct cfq_group *cfqg)
535
{
536
	struct cfqg_stats *stats = &cfqg->stats;
537

538
	BUG_ON(cfqg_stats_idling(stats));
539 540

	stats->start_idle_time = sched_clock();
541
	cfqg_stats_mark_idling(stats);
542 543
}

544
static void cfqg_stats_update_avg_queue_size(struct cfq_group *cfqg)
545
{
546
	struct cfqg_stats *stats = &cfqg->stats;
547 548 549 550

	blkg_stat_add(&stats->avg_queue_size_sum,
		      blkg_rwstat_sum(&stats->queued));
	blkg_stat_add(&stats->avg_queue_size_samples, 1);
551
	cfqg_stats_update_group_wait_time(stats);
552 553 554 555
}

#else	/* CONFIG_CFQ_GROUP_IOSCHED && CONFIG_DEBUG_BLK_CGROUP */

Tejun Heo's avatar
Tejun Heo committed
556 557 558 559 560 561 562
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) { }
563 564 565 566

#endif	/* CONFIG_CFQ_GROUP_IOSCHED && CONFIG_DEBUG_BLK_CGROUP */

#ifdef CONFIG_CFQ_GROUP_IOSCHED
567

568 569 570 571 572 573 574
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));
}

575 576 577 578 579 580 581 582 583 584
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
585 586 587 588
#define cfq_log_cfqq(cfqd, cfqq, fmt, args...)	do {			\
	char __pbuf[128];						\
									\
	blkg_path(cfqg_to_blkg((cfqq)->cfqg), __pbuf, sizeof(__pbuf));	\
589
	blk_add_trace_msg((cfqd)->queue, "cfq%d%c %s " fmt, (cfqq)->pid, \
Tejun Heo's avatar
Tejun Heo committed
590 591 592 593 594 595 596 597 598 599
			  cfq_cfqq_sync((cfqq)) ? 'S' : 'A',		\
			  __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)
600

601 602
static inline void cfqg_stats_update_io_add(struct cfq_group *cfqg,
					    struct cfq_group *curr_cfqg, int rw)
603
{
604 605 606
	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);
607 608
}

609 610
static inline void cfqg_stats_update_timeslice_used(struct cfq_group *cfqg,
			unsigned long time, unsigned long unaccounted_time)
611
{
612
	blkg_stat_add(&cfqg->stats.time, time);
613
#ifdef CONFIG_DEBUG_BLK_CGROUP
614
	blkg_stat_add(&cfqg->stats.unaccounted_time, unaccounted_time);
615
#endif
616 617
}

618
static inline void cfqg_stats_update_io_remove(struct cfq_group *cfqg, int rw)
619
{
620
	blkg_rwstat_add(&cfqg->stats.queued, rw, -1);
621 622
}

623
static inline void cfqg_stats_update_io_merged(struct cfq_group *cfqg, int rw)
624
{
625
	blkg_rwstat_add(&cfqg->stats.merged, rw, 1);
626 627
}

628 629
static inline void cfqg_stats_update_dispatch(struct cfq_group *cfqg,
					      uint64_t bytes, int rw)
630
{
631 632 633
	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);
634 635
}

636 637
static inline void cfqg_stats_update_completion(struct cfq_group *cfqg,
			uint64_t start_time, uint64_t io_start_time, int rw)
638
{
639
	struct cfqg_stats *stats = &cfqg->stats;
640 641 642 643 644 645 646
	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);
647 648
}

Tejun Heo's avatar
Tejun Heo committed
649
static void cfq_pd_reset_stats(struct blkcg_gq *blkg)
650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671
{
	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
}

672 673 674 675 676
#else	/* CONFIG_CFQ_GROUP_IOSCHED */

static inline void cfqg_get(struct cfq_group *cfqg) { }
static inline void cfqg_put(struct cfq_group *cfqg) { }

677 678
#define cfq_log_cfqq(cfqd, cfqq, fmt, args...)	\
	blk_add_trace_msg((cfqd)->queue, "cfq%d " fmt, (cfqq)->pid, ##args)
679
#define cfq_log_cfqg(cfqd, cfqg, fmt, args...)		do {} while (0)
680

681 682 683 684 685 686 687 688 689 690
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) { }
691

692 693
#endif	/* CONFIG_CFQ_GROUP_IOSCHED */

694 695 696
#define cfq_log(cfqd, fmt, args...)	\
	blk_add_trace_msg((cfqd)->queue, "cfq " fmt, ##args)

697 698 699 700 701 702 703 704 705 706
/* 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) \

707 708 709 710 711 712 713 714 715 716 717 718
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;
}
719

720 721 722 723 724 725 726 727 728 729 730 731 732 733 734
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;
}

735
static inline enum wl_class_t cfqq_class(struct cfq_queue *cfqq)
736 737 738 739 740 741 742 743
{
	if (cfq_class_idle(cfqq))
		return IDLE_WORKLOAD;
	if (cfq_class_rt(cfqq))
		return RT_WORKLOAD;
	return BE_WORKLOAD;
}

744 745 746 747 748 749 750 751 752 753

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;
}

754
static inline int cfq_group_busy_queues_wl(enum wl_class_t wl_class,
755 756
					struct cfq_data *cfqd,
					struct cfq_group *cfqg)
757
{
758
	if (wl_class == IDLE_WORKLOAD)
759
		return cfqg->service_tree_idle.count;
760

761 762 763
	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;
764 765
}

766 767 768 769 770 771 772
static inline int cfqg_busy_async_queues(struct cfq_data *cfqd,
					struct cfq_group *cfqg)
{
	return cfqg->service_trees[RT_WORKLOAD][ASYNC_WORKLOAD].count
		+ cfqg->service_trees[BE_WORKLOAD][ASYNC_WORKLOAD].count;
}

773
static void cfq_dispatch_insert(struct request_queue *, struct request *);
774
static struct cfq_queue *cfq_get_queue(struct cfq_data *cfqd, bool is_sync,
775
				       struct cfq_io_cq *cic, struct bio *bio,
776
				       gfp_t gfp_mask);
777

778 779 780 781 782 783
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);
}

784 785 786 787 788 789 790 791
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;
}

792
static inline struct cfq_queue *cic_to_cfqq(struct cfq_io_cq *cic, bool is_sync)
793
{
794
	return cic->cfqq[is_sync];
795 796
}

797 798
static inline void cic_set_cfqq(struct cfq_io_cq *cic, struct cfq_queue *cfqq,
				bool is_sync)
799
{
800
	cic->cfqq[is_sync] = cfqq;
801 802
}

803
static inline struct cfq_data *cic_to_cfqd(struct cfq_io_cq *cic)
804
{
805
	return cic->icq.q->elevator->elevator_data;
806 807
}

808 809 810 811
/*
 * 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).
 */
812
static inline bool cfq_bio_sync(struct bio *bio)
813
{
814
	return bio_data_dir(bio) == READ || (bio->bi_rw & REQ_SYNC);
815
}
Linus Torvalds's avatar
Linus Torvalds committed
816

Andrew Morton's avatar
Andrew Morton committed
817 818 819 820
/*
 * scheduler run of queue, if there are requests pending and no one in the
 * driver that will restart queueing
 */
821
static inline void cfq_schedule_dispatch(struct cfq_data *cfqd)
Andrew Morton's avatar
Andrew Morton committed
822
{
823 824
	if (cfqd->busy_queues) {
		cfq_log(cfqd, "schedule dispatch");
825
		kblockd_schedule_work(cfqd->queue, &cfqd->unplug_work);
826
	}
Andrew Morton's avatar
Andrew Morton committed
827 828
}

829 830 831 832 833
/*
 * 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.
 */
834
static inline int cfq_prio_slice(struct cfq_data *cfqd, bool sync,
835
				 unsigned short prio)
836
{
837
	const int base_slice = cfqd->cfq_slice[sync];
838

839 840 841 842
	WARN_ON(prio >= IOPRIO_BE_NR);

	return base_slice + (base_slice/CFQ_SLICE_SCALE * (4 - prio));
}
843

844 845 846 847
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);
848 849
}

850 851 852 853
static inline u64 cfq_scale_slice(unsigned long delta, struct cfq_group *cfqg)
{
	u64 d = delta << CFQ_SERVICE_SHIFT;

854
	d = d * CFQ_WEIGHT_DEFAULT;
855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882
	do_div(d, cfqg->weight);
	return d;
}

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);
883 884
		st->min_vdisktime = max_vdisktime(st->min_vdisktime,
						  cfqg->vdisktime);
885 886 887
	}
}

888 889 890 891 892 893
/*
 * 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
 */

894 895
static inline unsigned cfq_group_get_avg_queues(struct cfq_data *cfqd,
					struct cfq_group *cfqg, bool rt)
896
{
897 898 899
	unsigned min_q, max_q;
	unsigned mult  = cfq_hist_divisor - 1;
	unsigned round = cfq_hist_divisor / 2;
900
	unsigned busy = cfq_group_busy_queues_wl(rt, cfqd, cfqg);
901

902 903 904
	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) /
905
		cfq_hist_divisor;
906 907 908 909 910 911 912 913
	return cfqg->busy_queues_avg[rt];
}

static inline unsigned
cfq_group_slice(struct cfq_data *cfqd, struct cfq_group *cfqg)
{
	struct cfq_rb_root *st = &cfqd->grp_service_tree;

914
	return cfqd->cfq_target_latency * cfqg->weight / st->total_weight;
915 916
}

917
static inline unsigned
918
cfq_scaled_cfqq_slice(struct cfq_data *cfqd, struct cfq_queue *cfqq)
919
{
920 921
	unsigned slice = cfq_prio_to_slice(cfqd, cfqq);
	if (cfqd->cfq_latency) {
922 923 924 925 926 927
		/*
		 * 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));
928 929
		unsigned sync_slice = cfqd->cfq_slice[1];
		unsigned expect_latency = sync_slice * iq;
930 931 932
		unsigned group_slice = cfq_group_slice(cfqd, cfqq->cfqg);

		if (expect_latency > group_slice) {
933 934 935 936 937 938 939
			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 */
940
			slice = max(slice * group_slice / expect_latency,
941 942 943
				    low_slice);
		}
	}
944 945 946 947 948 949
	return slice;
}

static inline void
cfq_set_prio_slice(struct cfq_data *cfqd, struct cfq_queue *cfqq)
{
950
	unsigned slice = cfq_scaled_cfqq_slice(cfqd, cfqq);
951

952
	cfqq->slice_start = jiffies;
953
	cfqq->slice_end = jiffies + slice;
954
	cfqq->allocated_slice = slice;
955
	cfq_log_cfqq(cfqd, cfqq, "set_slice=%lu", cfqq->slice_end - jiffies);
956 957 958 959 960 961 962
}

/*
 * We need to wrap this check in cfq_cfqq_slice_new(), since ->slice_end
 * isn't valid until the first request from the dispatch is activated
 * and the slice time set.
 */
963
static inline bool cfq_slice_used(struct cfq_queue *cfqq)
964 965
{
	if (cfq_cfqq_slice_new(cfqq))
Shaohua Li's avatar
Shaohua Li committed
966
		return false;
967
	if (time_before(jiffies, cfqq->slice_end))
Shaohua Li's avatar
Shaohua Li committed
968
		return false;
969

Shaohua Li's avatar
Shaohua Li committed
970
	return true;
971 972
}

Linus Torvalds's avatar
Linus Torvalds committed
973
/*
Jens Axboe's avatar
Jens Axboe committed
974
 * Lifted from AS - choose which of rq1 and rq2 that is best served now.
Linus Torvalds's avatar
Linus Torvalds committed
975
 * We choose the request that is closest to the head right now. Distance
976
 * behind the head is penalized and only allowed to a certain extent.
Linus Torvalds's avatar
Linus Torvalds committed
977
 */
Jens Axboe's avatar
Jens Axboe committed
978
static struct request *
979
cfq_choose_req(struct cfq_data *cfqd, struct request *rq1, struct request *rq2, sector_t last)
Linus Torvalds's avatar
Linus Torvalds committed
980
{
981
	sector_t s1, s2, d1 = 0, d2 = 0;
Linus Torvalds's avatar
Linus Torvalds committed
982
	unsigned long back_max;
983 984 985
#define CFQ_RQ1_WRAP	0x01 /* request 1 wraps */
#define CFQ_RQ2_WRAP	0x02 /* request 2 wraps */
	unsigned wrap = 0; /* bit mask: requests behind the disk head? */
Linus Torvalds's avatar
Linus Torvalds committed
986

Jens Axboe's avatar
Jens Axboe committed
987 988 989 990
	if (rq1 == NULL || rq1 == rq2)
		return rq2;
	if (rq2 == NULL)
		return rq1;
991

992 993 994
	if (rq_is_sync(rq1) != rq_is_sync(rq2))
		return rq_is_sync(rq1) ? rq1 : rq2;

995 996
	if ((rq1->cmd_flags ^ rq2->cmd_flags) & REQ_PRIO)
		return rq1->cmd_flags & REQ_PRIO ? rq1 : rq2;
997