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

Tejun Heo's avatar
Tejun Heo committed
20
static struct blkio_policy_type blkio_policy_cfq __maybe_unused;
21

Linus Torvalds's avatar
Linus Torvalds committed
22 23 24
/*
 * tunables
 */
25
/* max queue in one round of service */
Shaohua Li's avatar
Shaohua Li committed
26
static const int cfq_quantum = 8;
27
static const int cfq_fifo_expire[2] = { HZ / 4, HZ / 8 };
28 29 30 31
/* 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;
32
static const int cfq_slice_sync = HZ / 10;
Jens Axboe's avatar
Jens Axboe committed
33
static int cfq_slice_async = HZ / 25;
34
static const int cfq_slice_async_rq = 2;
35
static int cfq_slice_idle = HZ / 125;
36
static int cfq_group_idle = HZ / 125;
37 38
static const int cfq_target_latency = HZ * 3/10; /* 300 ms */
static const int cfq_hist_divisor = 4;
39

40
/*
41
 * offset from end of service tree
42
 */
43
#define CFQ_IDLE_DELAY		(HZ / 5)
44 45 46 47 48 49

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

50
#define CFQ_SLICE_SCALE		(5)
51
#define CFQ_HW_QUEUE_MIN	(5)
52
#define CFQ_SERVICE_SHIFT       12
53

54
#define CFQQ_SEEK_THR		(sector_t)(8 * 100)
55
#define CFQQ_CLOSE_THR		(sector_t)(8 * 1024)
56
#define CFQQ_SECT_THR_NONROT	(sector_t)(2 * 32)
57
#define CFQQ_SEEKY(cfqq)	(hweight32(cfqq->seek_history) > 32/8)
58

59 60 61
#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
62

63
static struct kmem_cache *cfq_pool;
Linus Torvalds's avatar
Linus Torvalds committed
64

65 66 67 68
#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)

69
#define sample_valid(samples)	((samples) > 80)
70
#define rb_entry_cfqg(node)	rb_entry((node), struct cfq_group, rb_node)
71

72 73 74 75 76 77 78 79
struct cfq_ttime {
	unsigned long last_end_request;

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

80 81 82 83 84 85 86 87 88
/*
 * 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;
89
	unsigned count;
90
	unsigned total_weight;
91
	u64 min_vdisktime;
92
	struct cfq_ttime ttime;
93
};
94 95
#define CFQ_RB_ROOT	(struct cfq_rb_root) { .rb = RB_ROOT, \
			.ttime = {.last_end_request = jiffies,},}
96

97 98 99 100 101
/*
 * Per process-grouping structure
 */
struct cfq_queue {
	/* reference count */
102
	int ref;
103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125
	/* 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;

126 127
	/* time when queue got scheduled in to dispatch first request. */
	unsigned long dispatch_start;
128
	unsigned int allocated_slice;
129
	unsigned int slice_dispatch;
130 131
	/* time when first request from queue completed and slice started. */
	unsigned long slice_start;
132 133 134
	unsigned long slice_end;
	long slice_resid;

135 136
	/* pending priority requests */
	int prio_pending;
137 138 139 140 141
	/* number of requests that are on the dispatch list or inside driver */
	int dispatched;

	/* io prio of this group */
	unsigned short ioprio, org_ioprio;
142
	unsigned short ioprio_class;
143

144 145
	pid_t pid;

146
	u32 seek_history;
147 148
	sector_t last_request_pos;

149
	struct cfq_rb_root *service_tree;
Jeff Moyer's avatar
Jeff Moyer committed
150
	struct cfq_queue *new_cfqq;
151
	struct cfq_group *cfqg;
152 153
	/* Number of sectors dispatched from queue in single dispatch round */
	unsigned long nr_sectors;
154 155
};

156
/*
157
 * First index in the service_trees.
158 159 160 161
 * IDLE is handled separately, so it has negative index
 */
enum wl_prio_t {
	BE_WORKLOAD = 0,
162 163
	RT_WORKLOAD = 1,
	IDLE_WORKLOAD = 2,
164
	CFQ_PRIO_NR,
165 166
};

167 168 169 170 171 172 173 174 175
/*
 * Second index in the service_trees.
 */
enum wl_type_t {
	ASYNC_WORKLOAD = 0,
	SYNC_NOIDLE_WORKLOAD = 1,
	SYNC_WORKLOAD = 2
};

176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217
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;
	/* time spent idling for this blkio_group */
	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 */
};

218 219
/* This is per cgroup per device grouping structure */
struct cfq_group {
220 221 222 223 224
	/* group service_tree member */
	struct rb_node rb_node;

	/* group service_tree key */
	u64 vdisktime;
225
	unsigned int weight;
226
	unsigned int new_weight;
227
	unsigned int dev_weight;
228 229 230 231

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

232
	/*
233
	 * Per group busy queues average. Useful for workload slice calc. We
234 235 236 237 238 239 240 241 242 243 244
	 * 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.
245 246 247 248
	 * Counts are embedded in the cfq_rb_root
	 */
	struct cfq_rb_root service_trees[2][3];
	struct cfq_rb_root service_tree_idle;
249 250 251 252

	unsigned long saved_workload_slice;
	enum wl_type_t saved_workload;
	enum wl_prio_t saved_serving_prio;
253

254 255
	/* number of requests that are on the dispatch list or inside driver */
	int dispatched;
256
	struct cfq_ttime ttime;
257
	struct cfqg_stats stats;
258
};
259

260 261 262 263
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
264 265 266 267
	int			ioprio;		/* the current ioprio */
#ifdef CONFIG_CFQ_GROUP_IOSCHED
	uint64_t		blkcg_id;	/* the current blkcg ID */
#endif
268 269
};

270 271 272
/*
 * Per block device queue structure
 */
Linus Torvalds's avatar
Linus Torvalds committed
273
struct cfq_data {
274
	struct request_queue *queue;
275 276
	/* Root service tree for cfq_groups */
	struct cfq_rb_root grp_service_tree;
277
	struct cfq_group *root_group;
278

279 280
	/*
	 * The priority currently being served
281
	 */
282
	enum wl_prio_t serving_prio;
283 284
	enum wl_type_t serving_type;
	unsigned long workload_expires;
285
	struct cfq_group *serving_group;
286 287 288 289 290 291 292 293

	/*
	 * 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];

294
	unsigned int busy_queues;
295
	unsigned int busy_sync_queues;
296

297 298
	int rq_in_driver;
	int rq_in_flight[2];
299 300 301 302 303

	/*
	 * queue-depth detection
	 */
	int rq_queued;
304
	int hw_tag;
305 306 307 308 309 310 311 312
	/*
	 * 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
313

314 315 316 317
	/*
	 * idle window management
	 */
	struct timer_list idle_slice_timer;
318
	struct work_struct unplug_work;
Linus Torvalds's avatar
Linus Torvalds committed
319

320
	struct cfq_queue *active_queue;
321
	struct cfq_io_cq *active_cic;
322

323 324 325 326 327
	/*
	 * async queue for each priority case
	 */
	struct cfq_queue *async_cfqq[2][IOPRIO_BE_NR];
	struct cfq_queue *async_idle_cfqq;
328

Jens Axboe's avatar
Jens Axboe committed
329
	sector_t last_position;
Linus Torvalds's avatar
Linus Torvalds committed
330 331 332 333 334

	/*
	 * tunables, see top of file
	 */
	unsigned int cfq_quantum;
335
	unsigned int cfq_fifo_expire[2];
Linus Torvalds's avatar
Linus Torvalds committed
336 337
	unsigned int cfq_back_penalty;
	unsigned int cfq_back_max;
338 339 340
	unsigned int cfq_slice[2];
	unsigned int cfq_slice_async_rq;
	unsigned int cfq_slice_idle;
341
	unsigned int cfq_group_idle;
342
	unsigned int cfq_latency;
343

344 345 346 347
	/*
	 * Fallback dummy cfqq for extreme OOM conditions
	 */
	struct cfq_queue oom_cfqq;
348

349
	unsigned long last_delayed_sync;
Linus Torvalds's avatar
Linus Torvalds committed
350 351
};

352 353
static struct cfq_group *cfq_get_next_cfqg(struct cfq_data *cfqd);

354 355
static struct cfq_rb_root *service_tree_for(struct cfq_group *cfqg,
					    enum wl_prio_t prio,
356
					    enum wl_type_t type)
357
{
358 359 360
	if (!cfqg)
		return NULL;

361
	if (prio == IDLE_WORKLOAD)
362
		return &cfqg->service_tree_idle;
363

364
	return &cfqg->service_trees[prio][type];
365 366
}

Jens Axboe's avatar
Jens Axboe committed
367
enum cfqq_state_flags {
368 369
	CFQ_CFQQ_FLAG_on_rr = 0,	/* on round-robin busy list */
	CFQ_CFQQ_FLAG_wait_request,	/* waiting for a request */
370
	CFQ_CFQQ_FLAG_must_dispatch,	/* must be allowed a dispatch */
371 372 373 374
	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 */
375
	CFQ_CFQQ_FLAG_slice_new,	/* no requests dispatched in slice */
376
	CFQ_CFQQ_FLAG_sync,		/* synchronous queue */
377
	CFQ_CFQQ_FLAG_coop,		/* cfqq is shared */
378
	CFQ_CFQQ_FLAG_split_coop,	/* shared cfqq will be splitted */
379
	CFQ_CFQQ_FLAG_deep,		/* sync cfqq experienced large depth */
380
	CFQ_CFQQ_FLAG_wait_busy,	/* Waiting for next request */
Jens Axboe's avatar
Jens Axboe committed
381 382 383 384 385
};

#define CFQ_CFQQ_FNS(name)						\
static inline void cfq_mark_cfqq_##name(struct cfq_queue *cfqq)		\
{									\
386
	(cfqq)->flags |= (1 << CFQ_CFQQ_FLAG_##name);			\
Jens Axboe's avatar
Jens Axboe committed
387 388 389
}									\
static inline void cfq_clear_cfqq_##name(struct cfq_queue *cfqq)	\
{									\
390
	(cfqq)->flags &= ~(1 << CFQ_CFQQ_FLAG_##name);			\
Jens Axboe's avatar
Jens Axboe committed
391 392 393
}									\
static inline int cfq_cfqq_##name(const struct cfq_queue *cfqq)		\
{									\
394
	return ((cfqq)->flags & (1 << CFQ_CFQQ_FLAG_##name)) != 0;	\
Jens Axboe's avatar
Jens Axboe committed
395 396 397 398
}

CFQ_CFQQ_FNS(on_rr);
CFQ_CFQQ_FNS(wait_request);
399
CFQ_CFQQ_FNS(must_dispatch);
Jens Axboe's avatar
Jens Axboe committed
400 401 402 403
CFQ_CFQQ_FNS(must_alloc_slice);
CFQ_CFQQ_FNS(fifo_expire);
CFQ_CFQQ_FNS(idle_window);
CFQ_CFQQ_FNS(prio_changed);
404
CFQ_CFQQ_FNS(slice_new);
405
CFQ_CFQQ_FNS(sync);
406
CFQ_CFQQ_FNS(coop);
407
CFQ_CFQQ_FNS(split_coop);
408
CFQ_CFQQ_FNS(deep);
409
CFQ_CFQQ_FNS(wait_busy);
Jens Axboe's avatar
Jens Axboe committed
410 411
#undef CFQ_CFQQ_FNS

412
#if defined(CONFIG_CFQ_GROUP_IOSCHED) && defined(CONFIG_DEBUG_BLK_CGROUP)
413

414 415 416 417 418
/* cfqg stats flags */
enum cfqg_stats_flags {
	CFQG_stats_waiting = 0,
	CFQG_stats_idling,
	CFQG_stats_empty,
419 420
};

421 422
#define CFQG_FLAG_FNS(name)						\
static inline void cfqg_stats_mark_##name(struct cfqg_stats *stats)	\
423
{									\
424
	stats->flags |= (1 << CFQG_stats_##name);			\
425
}									\
426
static inline void cfqg_stats_clear_##name(struct cfqg_stats *stats)	\
427
{									\
428
	stats->flags &= ~(1 << CFQG_stats_##name);			\
429
}									\
430
static inline int cfqg_stats_##name(struct cfqg_stats *stats)		\
431
{									\
432
	return (stats->flags & (1 << CFQG_stats_##name)) != 0;		\
433 434
}									\

435 436 437 438
CFQG_FLAG_FNS(waiting)
CFQG_FLAG_FNS(idling)
CFQG_FLAG_FNS(empty)
#undef CFQG_FLAG_FNS
439 440

/* This should be called with the queue_lock held. */
441
static void cfqg_stats_update_group_wait_time(struct cfqg_stats *stats)
442 443 444
{
	unsigned long long now;

445
	if (!cfqg_stats_waiting(stats))
446 447 448 449 450 451
		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);
452
	cfqg_stats_clear_waiting(stats);
453 454 455
}

/* This should be called with the queue_lock held. */
456 457
static void cfqg_stats_set_start_group_wait_time(struct cfq_group *cfqg,
						 struct cfq_group *curr_cfqg)
458
{
459
	struct cfqg_stats *stats = &cfqg->stats;
460

461
	if (cfqg_stats_waiting(stats))
462
		return;
463
	if (cfqg == curr_cfqg)
464
		return;
465 466
	stats->start_group_wait_time = sched_clock();
	cfqg_stats_mark_waiting(stats);
467 468 469
}

/* This should be called with the queue_lock held. */
470
static void cfqg_stats_end_empty_time(struct cfqg_stats *stats)
471 472 473
{
	unsigned long long now;

474
	if (!cfqg_stats_empty(stats))
475 476 477 478 479 480
		return;

	now = sched_clock();
	if (time_after64(now, stats->start_empty_time))
		blkg_stat_add(&stats->empty_time,
			      now - stats->start_empty_time);
481
	cfqg_stats_clear_empty(stats);
482 483
}

484
static void cfqg_stats_update_dequeue(struct cfq_group *cfqg)
485
{
486
	blkg_stat_add(&cfqg->stats.dequeue, 1);
487 488
}

489
static void cfqg_stats_set_start_empty_time(struct cfq_group *cfqg)
490
{
491
	struct cfqg_stats *stats = &cfqg->stats;
492 493 494 495 496 497 498 499 500

	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.
	 */
501
	if (cfqg_stats_empty(stats))
502 503 504
		return;

	stats->start_empty_time = sched_clock();
505
	cfqg_stats_mark_empty(stats);
506 507
}

508
static void cfqg_stats_update_idle_time(struct cfq_group *cfqg)
509
{
510
	struct cfqg_stats *stats = &cfqg->stats;
511

512
	if (cfqg_stats_idling(stats)) {
513 514 515 516 517
		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);
518
		cfqg_stats_clear_idling(stats);
519 520 521
	}
}

522
static void cfqg_stats_set_start_idle_time(struct cfq_group *cfqg)
523
{
524
	struct cfqg_stats *stats = &cfqg->stats;
525

526
	BUG_ON(cfqg_stats_idling(stats));
527 528

	stats->start_idle_time = sched_clock();
529
	cfqg_stats_mark_idling(stats);
530 531
}

532
static void cfqg_stats_update_avg_queue_size(struct cfq_group *cfqg)
533
{
534
	struct cfqg_stats *stats = &cfqg->stats;
535 536 537 538

	blkg_stat_add(&stats->avg_queue_size_sum,
		      blkg_rwstat_sum(&stats->queued));
	blkg_stat_add(&stats->avg_queue_size_samples, 1);
539
	cfqg_stats_update_group_wait_time(stats);
540 541 542 543
}

#else	/* CONFIG_CFQ_GROUP_IOSCHED && CONFIG_DEBUG_BLK_CGROUP */

Tejun Heo's avatar
Tejun Heo committed
544 545 546 547 548 549 550
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) { }
551 552 553 554

#endif	/* CONFIG_CFQ_GROUP_IOSCHED && CONFIG_DEBUG_BLK_CGROUP */

#ifdef CONFIG_CFQ_GROUP_IOSCHED
555

556 557 558 559 560 561 562
static inline struct cfq_group *blkg_to_cfqg(struct blkio_group *blkg)
{
	return blkg_to_pdata(blkg, &blkio_policy_cfq);
}

static inline struct blkio_group *cfqg_to_blkg(struct cfq_group *cfqg)
{
563
	return pdata_to_blkg(cfqg);
564 565 566 567 568 569 570 571 572 573 574 575
}

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
576 577 578 579
#define cfq_log_cfqq(cfqd, cfqq, fmt, args...)	do {			\
	char __pbuf[128];						\
									\
	blkg_path(cfqg_to_blkg((cfqq)->cfqg), __pbuf, sizeof(__pbuf));	\
580
	blk_add_trace_msg((cfqd)->queue, "cfq%d%c %s " fmt, (cfqq)->pid, \
Tejun Heo's avatar
Tejun Heo committed
581 582 583 584 585 586 587 588 589 590
			  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)
591

592 593
static inline void cfqg_stats_update_io_add(struct cfq_group *cfqg,
					    struct cfq_group *curr_cfqg, int rw)
594
{
595 596 597
	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);
598 599
}

600 601
static inline void cfqg_stats_update_timeslice_used(struct cfq_group *cfqg,
			unsigned long time, unsigned long unaccounted_time)
602
{
603
	blkg_stat_add(&cfqg->stats.time, time);
604
#ifdef CONFIG_DEBUG_BLK_CGROUP
605
	blkg_stat_add(&cfqg->stats.unaccounted_time, unaccounted_time);
606
#endif
607 608
}

609
static inline void cfqg_stats_update_io_remove(struct cfq_group *cfqg, int rw)
610
{
611
	blkg_rwstat_add(&cfqg->stats.queued, rw, -1);
612 613
}

614
static inline void cfqg_stats_update_io_merged(struct cfq_group *cfqg, int rw)
615
{
616
	blkg_rwstat_add(&cfqg->stats.merged, rw, 1);
617 618
}

619 620
static inline void cfqg_stats_update_dispatch(struct cfq_group *cfqg,
					      uint64_t bytes, int rw)
621
{
622 623 624
	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);
625 626
}

627 628
static inline void cfqg_stats_update_completion(struct cfq_group *cfqg,
			uint64_t start_time, uint64_t io_start_time, int rw)
629
{
630
	struct cfqg_stats *stats = &cfqg->stats;
631 632 633 634 635 636 637
	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);
638 639
}

640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662
static void cfqg_stats_reset(struct blkio_group *blkg)
{
	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
}

663 664 665 666 667 668 669
#else	/* CONFIG_CFQ_GROUP_IOSCHED */

static inline struct cfq_group *blkg_to_cfqg(struct blkio_group *blkg) { return NULL; }
static inline struct blkio_group *cfqg_to_blkg(struct cfq_group *cfqg) { return NULL; }
static inline void cfqg_get(struct cfq_group *cfqg) { }
static inline void cfqg_put(struct cfq_group *cfqg) { }

670 671
#define cfq_log_cfqq(cfqd, cfqq, fmt, args...)	\
	blk_add_trace_msg((cfqd)->queue, "cfq%d " fmt, (cfqq)->pid, ##args)
672
#define cfq_log_cfqg(cfqd, cfqg, fmt, args...)		do {} while (0)
673

674 675 676 677 678 679 680 681 682 683
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) { }
684

685 686
#endif	/* CONFIG_CFQ_GROUP_IOSCHED */

687 688 689
#define cfq_log(cfqd, fmt, args...)	\
	blk_add_trace_msg((cfqd)->queue, "cfq " fmt, ##args)

690 691 692 693 694 695 696 697 698 699
/* 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) \

700 701 702 703 704 705 706 707 708 709 710 711
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;
}
712

713 714 715 716 717 718 719 720 721 722 723 724 725 726 727
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;
}

728 729 730 731 732 733 734 735 736
static inline enum wl_prio_t cfqq_prio(struct cfq_queue *cfqq)
{
	if (cfq_class_idle(cfqq))
		return IDLE_WORKLOAD;
	if (cfq_class_rt(cfqq))
		return RT_WORKLOAD;
	return BE_WORKLOAD;
}

737 738 739 740 741 742 743 744 745 746

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;
}

747 748 749
static inline int cfq_group_busy_queues_wl(enum wl_prio_t wl,
					struct cfq_data *cfqd,
					struct cfq_group *cfqg)
750 751
{
	if (wl == IDLE_WORKLOAD)
752
		return cfqg->service_tree_idle.count;
753

754 755 756
	return cfqg->service_trees[wl][ASYNC_WORKLOAD].count
		+ cfqg->service_trees[wl][SYNC_NOIDLE_WORKLOAD].count
		+ cfqg->service_trees[wl][SYNC_WORKLOAD].count;
757 758
}

759 760 761 762 763 764 765
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;
}

766
static void cfq_dispatch_insert(struct request_queue *, struct request *);
767
static struct cfq_queue *cfq_get_queue(struct cfq_data *cfqd, bool is_sync,
768
				       struct cfq_io_cq *cic, struct bio *bio,
769
				       gfp_t gfp_mask);
770

771 772 773 774 775 776
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);
}

777 778 779 780 781 782 783 784
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;
}

785
static inline struct cfq_queue *cic_to_cfqq(struct cfq_io_cq *cic, bool is_sync)
786
{
787
	return cic->cfqq[is_sync];
788 789
}

790 791
static inline void cic_set_cfqq(struct cfq_io_cq *cic, struct cfq_queue *cfqq,
				bool is_sync)
792
{
793
	cic->cfqq[is_sync] = cfqq;
794 795
}

796
static inline struct cfq_data *cic_to_cfqd(struct cfq_io_cq *cic)
797
{
798
	return cic->icq.q->elevator->elevator_data;
799 800
}

801 802 803 804
/*
 * 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).
 */
805
static inline bool cfq_bio_sync(struct bio *bio)
806
{
807
	return bio_data_dir(bio) == READ || (bio->bi_rw & REQ_SYNC);
808
}
Linus Torvalds's avatar
Linus Torvalds committed
809

Andrew Morton's avatar
Andrew Morton committed
810 811 812 813
/*
 * scheduler run of queue, if there are requests pending and no one in the
 * driver that will restart queueing
 */
814
static inline void cfq_schedule_dispatch(struct cfq_data *cfqd)
Andrew Morton's avatar
Andrew Morton committed
815
{
816 817
	if (cfqd->busy_queues) {
		cfq_log(cfqd, "schedule dispatch");
818
		kblockd_schedule_work(cfqd->queue, &cfqd->unplug_work);
819
	}
Andrew Morton's avatar
Andrew Morton committed
820 821
}

822 823 824 825 826
/*
 * 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.
 */
827
static inline int cfq_prio_slice(struct cfq_data *cfqd, bool sync,
828
				 unsigned short prio)
829
{
830
	const int base_slice = cfqd->cfq_slice[sync];
831

832 833 834 835
	WARN_ON(prio >= IOPRIO_BE_NR);

	return base_slice + (base_slice/CFQ_SLICE_SCALE * (4 - prio));
}
836

837 838 839 840
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);
841 842
}

843 844 845 846
static inline u64 cfq_scale_slice(unsigned long delta, struct cfq_group *cfqg)
{
	u64 d = delta << CFQ_SERVICE_SHIFT;

847
	d = d * CFQ_WEIGHT_DEFAULT;
848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875
	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);
876 877
		st->min_vdisktime = max_vdisktime(st->min_vdisktime,
						  cfqg->vdisktime);
878 879 880
	}
}

881 882 883 884 885 886
/*
 * 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
 */

887 888
static inline unsigned cfq_group_get_avg_queues(struct cfq_data *cfqd,
					struct cfq_group *cfqg, bool rt)
889
{
890 891 892
	unsigned min_q, max_q;
	unsigned mult  = cfq_hist_divisor - 1;
	unsigned round = cfq_hist_divisor / 2;
893
	unsigned busy = cfq_group_busy_queues_wl(rt, cfqd, cfqg);
894

895 896 897
	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) /
898
		cfq_hist_divisor;
899 900 901 902 903 904 905 906 907
	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;

	return cfq_target_latency * cfqg->weight / st->total_weight;
908 909
}

910
static inline unsigned
911
cfq_scaled_cfqq_slice(struct cfq_data *cfqd, struct cfq_queue *cfqq)
912
{
913 914
	unsigned slice = cfq_prio_to_slice(cfqd, cfqq);
	if (cfqd->cfq_latency) {
915 916 917 918 919 920
		/*
		 * 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));
921 922
		unsigned sync_slice = cfqd->cfq_slice[1];
		unsigned expect_latency = sync_slice * iq;
923 924 925
		unsigned group_slice = cfq_group_slice(cfqd, cfqq->cfqg);

		if (expect_latency > group_slice) {
926 927 928 929 930 931 932
			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 */
933
			slice = max(slice * group_slice / expect_latency,
934 935 936
				    low_slice);
		}
	}
937 938 939 940 941 942
	return slice;
}

static inline void
cfq_set_prio_slice(struct cfq_data *cfqd, struct cfq_queue *cfqq)
{
943
	unsigned slice = cfq_scaled_cfqq_slice(cfqd, cfqq);
944

945
	cfqq->slice_start = jiffies;
946
	cfqq->slice_end = jiffies + slice;
947
	cfqq->allocated_slice = slice;
948
	cfq_log_cfqq(cfqd, cfqq, "set_slice=%lu", cfqq->slice_end - jiffies);
949 950 951 952 953 954 955
}

/*
 * 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.
 */
956
static inline bool cfq_slice_used(struct cfq_queue *cfqq)
957 958
{
	if (cfq_cfqq_slice_new(cfqq))
Shaohua Li's avatar
Shaohua Li committed
959
		return false;
960
	if (time_before(jiffies, cfqq->slice_end))
Shaohua Li's avatar
Shaohua Li committed
961
		return false;
962

Shaohua Li's avatar
Shaohua Li committed
963
	return true;
964 965
}

Linus Torvalds's avatar
Linus Torvalds committed
966
/*
Jens Axboe's avatar
Jens Axboe committed
967
 * Lifted from AS - choose which of rq1 and rq2 that is best served now.
Linus Torvalds's avatar
Linus Torvalds committed
968
 * We choose the request that is closest to the head right now. Distance
969
 * behind the head is penalized and only allowed to a certain extent.
Linus Torvalds's avatar
Linus Torvalds committed
970
 */
Jens Axboe's avatar
Jens Axboe committed
971
static struct request *
972
cfq_choose_req(struct cfq_data *cfqd, struct request *rq1, struct request *rq2, sector_t last)
Linus Torvalds's avatar
Linus Torvalds committed
973
{
974
	sector_t s1, s2, d1 = 0, d2 = 0;
Linus Torvalds's avatar
Linus Torvalds committed
975
	unsigned long back_max;
976 977 978
#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
979

Jens Axboe's avatar
Jens Axboe committed
980 981 982 983
	if (rq1 == NULL || rq1 == rq2)
		return rq2;
	if (rq2 == NULL)
		return rq1;
984

985 986 987
	if (rq_is_sync(rq1) != rq_is_sync(rq2))
		return rq_is_sync(rq1) ? rq1 : rq2;

988 989
	if ((rq1->cmd_flags ^ rq2->cmd_flags) & REQ_PRIO)
		return rq1->cmd_flags & REQ_PRIO ? rq1 : rq2;
990

991 992
	s1 = blk_rq_pos(rq1);
	s2 = blk_rq_pos(rq2);
Linus Torvalds's avatar
Linus Torvalds committed
993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008

	/*
	 * by definition, 1KiB is 2 sectors
	 */
	back_max = cfqd->cfq_back_max * 2;

	/*
	 * Strict one way elevator _except_ in the case where we allow
	 * short backward seeks which are biased as twice the cost of a
	 * similar forward seek.
	 */
	if (s1 >= last)
		d1 = s1 - last;
	else if (s1 + back_max >= last)
		d1 = (last - s1) * cfqd->cfq_back_penalty;
	else
1009
		wrap |= CFQ_RQ1_WRAP;
Linus Torvalds's avatar
Linus Torvalds committed
1010 1011 1012 1013 1014 1015

	if (s2 >= last)
		d2 = s2 - last;
	else if (s2 + back_max >= last)
		d2 = (last - s2) * cfqd->cfq_back_penalty;
	else
1016
		wrap |= CFQ_RQ2_WRAP;
Linus Torvalds's avatar
Linus Torvalds committed
1017