cfq-iosched.c 113 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;
Tejun Heo's avatar
Tejun Heo committed
226

227 228 229 230 231 232 233 234 235 236 237 238
	/*
	 * 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;

Tejun Heo's avatar
Tejun Heo committed
239 240 241 242 243 244
	/*
	 * 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.
	 */
245
	unsigned int weight;
246
	unsigned int new_weight;
247
	unsigned int dev_weight;
248

Tejun Heo's avatar
Tejun Heo committed
249 250 251 252
	unsigned int leaf_weight;
	unsigned int new_leaf_weight;
	unsigned int dev_leaf_weight;

253 254 255
	/* number of cfqq currently on this group */
	int nr_cfqq;

256
	/*
257
	 * Per group busy queues average. Useful for workload slice calc. We
258 259 260 261 262 263 264 265 266 267 268
	 * 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.
269 270 271 272
	 * Counts are embedded in the cfq_rb_root
	 */
	struct cfq_rb_root service_trees[2][3];
	struct cfq_rb_root service_tree_idle;
273

274 275 276
	unsigned long saved_wl_slice;
	enum wl_type_t saved_wl_type;
	enum wl_class_t saved_wl_class;
277

278 279
	/* number of requests that are on the dispatch list or inside driver */
	int dispatched;
280
	struct cfq_ttime ttime;
281
	struct cfqg_stats stats;
282
};
283

284 285 286 287
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
288 289 290 291
	int			ioprio;		/* the current ioprio */
#ifdef CONFIG_CFQ_GROUP_IOSCHED
	uint64_t		blkcg_id;	/* the current blkcg ID */
#endif
292 293
};

294 295 296
/*
 * Per block device queue structure
 */
Linus Torvalds's avatar
Linus Torvalds committed
297
struct cfq_data {
298
	struct request_queue *queue;
299 300
	/* Root service tree for cfq_groups */
	struct cfq_rb_root grp_service_tree;
301
	struct cfq_group *root_group;
302

303 304
	/*
	 * The priority currently being served
305
	 */
306 307
	enum wl_class_t serving_wl_class;
	enum wl_type_t serving_wl_type;
308
	unsigned long workload_expires;
309
	struct cfq_group *serving_group;
310 311 312 313 314 315 316 317

	/*
	 * 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];

318
	unsigned int busy_queues;
319
	unsigned int busy_sync_queues;
320

321 322
	int rq_in_driver;
	int rq_in_flight[2];
323 324 325 326 327

	/*
	 * queue-depth detection
	 */
	int rq_queued;
328
	int hw_tag;
329 330 331 332 333 334 335 336
	/*
	 * 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
337

338 339 340 341
	/*
	 * idle window management
	 */
	struct timer_list idle_slice_timer;
342
	struct work_struct unplug_work;
Linus Torvalds's avatar
Linus Torvalds committed
343

344
	struct cfq_queue *active_queue;
345
	struct cfq_io_cq *active_cic;
346

347 348 349 350 351
	/*
	 * async queue for each priority case
	 */
	struct cfq_queue *async_cfqq[2][IOPRIO_BE_NR];
	struct cfq_queue *async_idle_cfqq;
352

Jens Axboe's avatar
Jens Axboe committed
353
	sector_t last_position;
Linus Torvalds's avatar
Linus Torvalds committed
354 355 356 357 358

	/*
	 * tunables, see top of file
	 */
	unsigned int cfq_quantum;
359
	unsigned int cfq_fifo_expire[2];
Linus Torvalds's avatar
Linus Torvalds committed
360 361
	unsigned int cfq_back_penalty;
	unsigned int cfq_back_max;
362 363 364
	unsigned int cfq_slice[2];
	unsigned int cfq_slice_async_rq;
	unsigned int cfq_slice_idle;
365
	unsigned int cfq_group_idle;
366
	unsigned int cfq_latency;
367
	unsigned int cfq_target_latency;
368

369 370 371 372
	/*
	 * Fallback dummy cfqq for extreme OOM conditions
	 */
	struct cfq_queue oom_cfqq;
373

374
	unsigned long last_delayed_sync;
Linus Torvalds's avatar
Linus Torvalds committed
375 376
};

377 378
static struct cfq_group *cfq_get_next_cfqg(struct cfq_data *cfqd);

379
static struct cfq_rb_root *st_for(struct cfq_group *cfqg,
380
					    enum wl_class_t class,
381
					    enum wl_type_t type)
382
{
383 384 385
	if (!cfqg)
		return NULL;

386
	if (class == IDLE_WORKLOAD)
387
		return &cfqg->service_tree_idle;
388

389
	return &cfqg->service_trees[class][type];
390 391
}

Jens Axboe's avatar
Jens Axboe committed
392
enum cfqq_state_flags {
393 394
	CFQ_CFQQ_FLAG_on_rr = 0,	/* on round-robin busy list */
	CFQ_CFQQ_FLAG_wait_request,	/* waiting for a request */
395
	CFQ_CFQQ_FLAG_must_dispatch,	/* must be allowed a dispatch */
396 397 398 399
	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 */
400
	CFQ_CFQQ_FLAG_slice_new,	/* no requests dispatched in slice */
401
	CFQ_CFQQ_FLAG_sync,		/* synchronous queue */
402
	CFQ_CFQQ_FLAG_coop,		/* cfqq is shared */
403
	CFQ_CFQQ_FLAG_split_coop,	/* shared cfqq will be splitted */
404
	CFQ_CFQQ_FLAG_deep,		/* sync cfqq experienced large depth */
405
	CFQ_CFQQ_FLAG_wait_busy,	/* Waiting for next request */
Jens Axboe's avatar
Jens Axboe committed
406 407 408 409 410
};

#define CFQ_CFQQ_FNS(name)						\
static inline void cfq_mark_cfqq_##name(struct cfq_queue *cfqq)		\
{									\
411
	(cfqq)->flags |= (1 << CFQ_CFQQ_FLAG_##name);			\
Jens Axboe's avatar
Jens Axboe committed
412 413 414
}									\
static inline void cfq_clear_cfqq_##name(struct cfq_queue *cfqq)	\
{									\
415
	(cfqq)->flags &= ~(1 << CFQ_CFQQ_FLAG_##name);			\
Jens Axboe's avatar
Jens Axboe committed
416 417 418
}									\
static inline int cfq_cfqq_##name(const struct cfq_queue *cfqq)		\
{									\
419
	return ((cfqq)->flags & (1 << CFQ_CFQQ_FLAG_##name)) != 0;	\
Jens Axboe's avatar
Jens Axboe committed
420 421 422 423
}

CFQ_CFQQ_FNS(on_rr);
CFQ_CFQQ_FNS(wait_request);
424
CFQ_CFQQ_FNS(must_dispatch);
Jens Axboe's avatar
Jens Axboe committed
425 426 427 428
CFQ_CFQQ_FNS(must_alloc_slice);
CFQ_CFQQ_FNS(fifo_expire);
CFQ_CFQQ_FNS(idle_window);
CFQ_CFQQ_FNS(prio_changed);
429
CFQ_CFQQ_FNS(slice_new);
430
CFQ_CFQQ_FNS(sync);
431
CFQ_CFQQ_FNS(coop);
432
CFQ_CFQQ_FNS(split_coop);
433
CFQ_CFQQ_FNS(deep);
434
CFQ_CFQQ_FNS(wait_busy);
Jens Axboe's avatar
Jens Axboe committed
435 436
#undef CFQ_CFQQ_FNS

437 438 439 440 441 442 443 444 445 446
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);
}

447
#if defined(CONFIG_CFQ_GROUP_IOSCHED) && defined(CONFIG_DEBUG_BLK_CGROUP)
448

449 450 451 452 453
/* cfqg stats flags */
enum cfqg_stats_flags {
	CFQG_stats_waiting = 0,
	CFQG_stats_idling,
	CFQG_stats_empty,
454 455
};

456 457
#define CFQG_FLAG_FNS(name)						\
static inline void cfqg_stats_mark_##name(struct cfqg_stats *stats)	\
458
{									\
459
	stats->flags |= (1 << CFQG_stats_##name);			\
460
}									\
461
static inline void cfqg_stats_clear_##name(struct cfqg_stats *stats)	\
462
{									\
463
	stats->flags &= ~(1 << CFQG_stats_##name);			\
464
}									\
465
static inline int cfqg_stats_##name(struct cfqg_stats *stats)		\
466
{									\
467
	return (stats->flags & (1 << CFQG_stats_##name)) != 0;		\
468 469
}									\

470 471 472 473
CFQG_FLAG_FNS(waiting)
CFQG_FLAG_FNS(idling)
CFQG_FLAG_FNS(empty)
#undef CFQG_FLAG_FNS
474 475

/* This should be called with the queue_lock held. */
476
static void cfqg_stats_update_group_wait_time(struct cfqg_stats *stats)
477 478 479
{
	unsigned long long now;

480
	if (!cfqg_stats_waiting(stats))
481 482 483 484 485 486
		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);
487
	cfqg_stats_clear_waiting(stats);
488 489 490
}

/* This should be called with the queue_lock held. */
491 492
static void cfqg_stats_set_start_group_wait_time(struct cfq_group *cfqg,
						 struct cfq_group *curr_cfqg)
493
{
494
	struct cfqg_stats *stats = &cfqg->stats;
495

496
	if (cfqg_stats_waiting(stats))
497
		return;
498
	if (cfqg == curr_cfqg)
499
		return;
500 501
	stats->start_group_wait_time = sched_clock();
	cfqg_stats_mark_waiting(stats);
502 503 504
}

/* This should be called with the queue_lock held. */
505
static void cfqg_stats_end_empty_time(struct cfqg_stats *stats)
506 507 508
{
	unsigned long long now;

509
	if (!cfqg_stats_empty(stats))
510 511 512 513 514 515
		return;

	now = sched_clock();
	if (time_after64(now, stats->start_empty_time))
		blkg_stat_add(&stats->empty_time,
			      now - stats->start_empty_time);
516
	cfqg_stats_clear_empty(stats);
517 518
}

519
static void cfqg_stats_update_dequeue(struct cfq_group *cfqg)
520
{
521
	blkg_stat_add(&cfqg->stats.dequeue, 1);
522 523
}

524
static void cfqg_stats_set_start_empty_time(struct cfq_group *cfqg)
525
{
526
	struct cfqg_stats *stats = &cfqg->stats;
527 528 529 530 531 532 533 534 535

	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.
	 */
536
	if (cfqg_stats_empty(stats))
537 538 539
		return;

	stats->start_empty_time = sched_clock();
540
	cfqg_stats_mark_empty(stats);
541 542
}

543
static void cfqg_stats_update_idle_time(struct cfq_group *cfqg)
544
{
545
	struct cfqg_stats *stats = &cfqg->stats;
546

547
	if (cfqg_stats_idling(stats)) {
548 549 550 551 552
		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);
553
		cfqg_stats_clear_idling(stats);
554 555 556
	}
}

557
static void cfqg_stats_set_start_idle_time(struct cfq_group *cfqg)
558
{
559
	struct cfqg_stats *stats = &cfqg->stats;
560

561
	BUG_ON(cfqg_stats_idling(stats));
562 563

	stats->start_idle_time = sched_clock();
564
	cfqg_stats_mark_idling(stats);
565 566
}

567
static void cfqg_stats_update_avg_queue_size(struct cfq_group *cfqg)
568
{
569
	struct cfqg_stats *stats = &cfqg->stats;
570 571 572 573

	blkg_stat_add(&stats->avg_queue_size_sum,
		      blkg_rwstat_sum(&stats->queued));
	blkg_stat_add(&stats->avg_queue_size_samples, 1);
574
	cfqg_stats_update_group_wait_time(stats);
575 576 577 578
}

#else	/* CONFIG_CFQ_GROUP_IOSCHED && CONFIG_DEBUG_BLK_CGROUP */

Tejun Heo's avatar
Tejun Heo committed
579 580 581 582 583 584 585
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) { }
586 587 588 589

#endif	/* CONFIG_CFQ_GROUP_IOSCHED && CONFIG_DEBUG_BLK_CGROUP */

#ifdef CONFIG_CFQ_GROUP_IOSCHED
590

591 592 593 594 595 596 597
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));
}

598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613
/*
 * Determine the parent cfqg for weight calculation.  Currently, cfqg
 * scheduling is flat and the root is the parent of everyone else.
 */
static inline struct cfq_group *cfqg_flat_parent(struct cfq_group *cfqg)
{
	struct blkcg_gq *blkg = cfqg_to_blkg(cfqg);
	struct cfq_group *root;

	while (blkg->parent)
		blkg = blkg->parent;
	root = blkg_to_cfqg(blkg);

	return root != cfqg ? root : NULL;
}

614 615 616 617 618 619 620 621 622 623
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
624 625 626 627
#define cfq_log_cfqq(cfqd, cfqq, fmt, args...)	do {			\
	char __pbuf[128];						\
									\
	blkg_path(cfqg_to_blkg((cfqq)->cfqg), __pbuf, sizeof(__pbuf));	\
628 629 630
	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
631 632 633 634 635 636 637 638 639
			  __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)
640

641 642
static inline void cfqg_stats_update_io_add(struct cfq_group *cfqg,
					    struct cfq_group *curr_cfqg, int rw)
643
{
644 645 646
	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);
647 648
}

649 650
static inline void cfqg_stats_update_timeslice_used(struct cfq_group *cfqg,
			unsigned long time, unsigned long unaccounted_time)
651
{
652
	blkg_stat_add(&cfqg->stats.time, time);
653
#ifdef CONFIG_DEBUG_BLK_CGROUP
654
	blkg_stat_add(&cfqg->stats.unaccounted_time, unaccounted_time);
655
#endif
656 657
}

658
static inline void cfqg_stats_update_io_remove(struct cfq_group *cfqg, int rw)
659
{
660
	blkg_rwstat_add(&cfqg->stats.queued, rw, -1);
661 662
}

663
static inline void cfqg_stats_update_io_merged(struct cfq_group *cfqg, int rw)
664
{
665
	blkg_rwstat_add(&cfqg->stats.merged, rw, 1);
666 667
}

668 669
static inline void cfqg_stats_update_dispatch(struct cfq_group *cfqg,
					      uint64_t bytes, int rw)
670
{
671 672 673
	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);
674 675
}

676 677
static inline void cfqg_stats_update_completion(struct cfq_group *cfqg,
			uint64_t start_time, uint64_t io_start_time, int rw)
678
{
679
	struct cfqg_stats *stats = &cfqg->stats;
680 681 682 683 684 685 686
	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);
687 688
}

Tejun Heo's avatar
Tejun Heo committed
689
static void cfq_pd_reset_stats(struct blkcg_gq *blkg)
690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711
{
	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
}

712 713
#else	/* CONFIG_CFQ_GROUP_IOSCHED */

714
static inline struct cfq_group *cfqg_flat_parent(struct cfq_group *cfqg) { return NULL; }
715 716 717
static inline void cfqg_get(struct cfq_group *cfqg) { }
static inline void cfqg_put(struct cfq_group *cfqg) { }

718
#define cfq_log_cfqq(cfqd, cfqq, fmt, args...)	\
719 720 721 722
	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)
723
#define cfq_log_cfqg(cfqd, cfqg, fmt, args...)		do {} while (0)
724

725 726 727 728 729 730 731 732 733 734
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) { }
735

736 737
#endif	/* CONFIG_CFQ_GROUP_IOSCHED */

738 739 740
#define cfq_log(cfqd, fmt, args...)	\
	blk_add_trace_msg((cfqd)->queue, "cfq " fmt, ##args)

741 742 743 744 745 746 747 748 749 750
/* 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) \

751 752 753 754 755 756 757 758 759 760 761 762
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;
}
763

764 765 766 767 768 769 770 771 772 773 774 775 776 777 778
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;
}

779
static inline enum wl_class_t cfqq_class(struct cfq_queue *cfqq)
780 781 782 783 784 785 786 787
{
	if (cfq_class_idle(cfqq))
		return IDLE_WORKLOAD;
	if (cfq_class_rt(cfqq))
		return RT_WORKLOAD;
	return BE_WORKLOAD;
}

788 789 790 791 792 793 794 795 796 797

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;
}

798
static inline int cfq_group_busy_queues_wl(enum wl_class_t wl_class,
799 800
					struct cfq_data *cfqd,
					struct cfq_group *cfqg)
801
{
802
	if (wl_class == IDLE_WORKLOAD)
803
		return cfqg->service_tree_idle.count;
804

805 806 807
	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;
808 809
}

810 811 812
static inline int cfqg_busy_async_queues(struct cfq_data *cfqd,
					struct cfq_group *cfqg)
{
813 814
	return cfqg->service_trees[RT_WORKLOAD][ASYNC_WORKLOAD].count +
		cfqg->service_trees[BE_WORKLOAD][ASYNC_WORKLOAD].count;
815 816
}

817
static void cfq_dispatch_insert(struct request_queue *, struct request *);
818
static struct cfq_queue *cfq_get_queue(struct cfq_data *cfqd, bool is_sync,
819
				       struct cfq_io_cq *cic, struct bio *bio,
820
				       gfp_t gfp_mask);
821

822 823 824 825 826 827
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);
}

828 829 830 831 832 833 834 835
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;
}

836
static inline struct cfq_queue *cic_to_cfqq(struct cfq_io_cq *cic, bool is_sync)
837
{
838
	return cic->cfqq[is_sync];
839 840
}

841 842
static inline void cic_set_cfqq(struct cfq_io_cq *cic, struct cfq_queue *cfqq,
				bool is_sync)
843
{
844
	cic->cfqq[is_sync] = cfqq;
845 846
}

847
static inline struct cfq_data *cic_to_cfqd(struct cfq_io_cq *cic)
848
{
849
	return cic->icq.q->elevator->elevator_data;
850 851
}

852 853 854 855
/*
 * 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).
 */
856
static inline bool cfq_bio_sync(struct bio *bio)
857
{
858
	return bio_data_dir(bio) == READ || (bio->bi_rw & REQ_SYNC);
859
}
Linus Torvalds's avatar
Linus Torvalds committed
860

Andrew Morton's avatar
Andrew Morton committed
861 862 863 864
/*
 * scheduler run of queue, if there are requests pending and no one in the
 * driver that will restart queueing
 */
865
static inline void cfq_schedule_dispatch(struct cfq_data *cfqd)
Andrew Morton's avatar
Andrew Morton committed
866
{
867 868
	if (cfqd->busy_queues) {
		cfq_log(cfqd, "schedule dispatch");
869
		kblockd_schedule_work(cfqd->queue, &cfqd->unplug_work);
870
	}
Andrew Morton's avatar
Andrew Morton committed
871 872
}

873 874 875 876 877
/*
 * 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.
 */
878
static inline int cfq_prio_slice(struct cfq_data *cfqd, bool sync,
879
				 unsigned short prio)
880
{
881
	const int base_slice = cfqd->cfq_slice[sync];
882

883 884 885 886
	WARN_ON(prio >= IOPRIO_BE_NR);

	return base_slice + (base_slice/CFQ_SLICE_SCALE * (4 - prio));
}
887

888 889 890 891
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);
892 893
}

894 895 896 897
static inline u64 cfq_scale_slice(unsigned long delta, struct cfq_group *cfqg)
{
	u64 d = delta << CFQ_SERVICE_SHIFT;

898
	d = d * CFQ_WEIGHT_DEFAULT;
899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926
	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);
927 928
		st->min_vdisktime = max_vdisktime(st->min_vdisktime,
						  cfqg->vdisktime);
929 930 931
	}
}

932 933 934 935 936 937
/*
 * 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
 */

938 939
static inline unsigned cfq_group_get_avg_queues(struct cfq_data *cfqd,
					struct cfq_group *cfqg, bool rt)
940
{
941 942 943
	unsigned min_q, max_q;
	unsigned mult  = cfq_hist_divisor - 1;
	unsigned round = cfq_hist_divisor / 2;
944
	unsigned busy = cfq_group_busy_queues_wl(rt, cfqd, cfqg);
945

946 947 948
	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) /
949
		cfq_hist_divisor;
950 951 952 953 954 955 956 957
	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;

958
	return cfqd->cfq_target_latency * cfqg->weight / st->total_weight;
959 960
}

961
static inline unsigned
962
cfq_scaled_cfqq_slice(struct cfq_data *cfqd, struct cfq_queue *cfqq)
963
{
964 965
	unsigned slice = cfq_prio_to_slice(cfqd, cfqq);
	if (cfqd->cfq_latency) {
966 967 968 969 970 971
		/*
		 * 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));
972 973
		unsigned sync_slice = cfqd->cfq_slice[1];
		unsigned expect_latency = sync_slice * iq;
974 975 976
		unsigned group_slice = cfq_group_slice(cfqd, cfqq->cfqg);

		if (expect_latency > group_slice) {
977 978 979 980 981 982 983
			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 */
984
			slice = max(slice * group_slice / expect_latency,
985 986 987
				    low_slice);
		}
	}
988 989 990 991 992 993
	return slice;
}

static inline void
cfq_set_prio_slice(struct cfq_data *cfqd, struct cfq_queue *cfqq)
{
994
	unsigned slice = cfq_scaled_cfqq_slice(cfqd, cfqq);
995

996
	cfqq->slice_start = jiffies;
997
	cfqq->slice_end = jiffies + slice;
998
	cfqq->allocated_slice = slice;
999
	cfq_log_cfqq(cfqd, cfqq, "set_slice=%lu", cfqq->slice_end - jiffies);
1000 1001 1002 1003 1004 1005 1006
}

/*
 * 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.
 */
1007
static inline bool cfq_slice_used(struct cfq_queue *cfqq)
1008 1009
{
	if (cfq_cfqq_slice_new(cfqq))
Shaohua Li's avatar
Shaohua Li committed
1010
		return false;
1011
	if (time_before(jiffies, cfqq->slice_end))
Shaohua Li's avatar
Shaohua Li committed
1012
		return false;
1013

Shaohua Li's avatar
Shaohua Li committed
1014
	return true;
1015 1016
}