cfq-iosched.c 93.6 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>
Al Viro's avatar
Al Viro committed
10 11
#include <linux/blkdev.h>
#include <linux/elevator.h>
Randy Dunlap's avatar
Randy Dunlap committed
12
#include <linux/jiffies.h>
Linus Torvalds's avatar
Linus Torvalds committed
13
#include <linux/rbtree.h>
14
#include <linux/ioprio.h>
15
#include <linux/blktrace_api.h>
16
#include "blk-cgroup.h"
Linus Torvalds's avatar
Linus Torvalds committed
17 18 19 20

/*
 * tunables
 */
21 22
/* max queue in one round of service */
static const int cfq_quantum = 4;
23
static const int cfq_fifo_expire[2] = { HZ / 4, HZ / 8 };
24 25 26 27
/* 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;
28
static const int cfq_slice_sync = HZ / 10;
Jens Axboe's avatar
Jens Axboe committed
29
static int cfq_slice_async = HZ / 25;
30
static const int cfq_slice_async_rq = 2;
31
static int cfq_slice_idle = HZ / 125;
32 33
static const int cfq_target_latency = HZ * 3/10; /* 300 ms */
static const int cfq_hist_divisor = 4;
34

35
/*
36
 * offset from end of service tree
37
 */
38
#define CFQ_IDLE_DELAY		(HZ / 5)
39 40 41 42 43 44

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

45 46 47 48 49 50
/*
 * Allow merged cfqqs to perform this amount of seeky I/O before
 * deciding to break the queues up again.
 */
#define CFQQ_COOP_TOUT		(HZ)

51
#define CFQ_SLICE_SCALE		(5)
52
#define CFQ_HW_QUEUE_MIN	(5)
53
#define CFQ_SERVICE_SHIFT       12
54

55 56
#define RQ_CIC(rq)		\
	((struct cfq_io_context *) (rq)->elevator_private)
57
#define RQ_CFQQ(rq)		(struct cfq_queue *) ((rq)->elevator_private2)
Linus Torvalds's avatar
Linus Torvalds committed
58

59 60
static struct kmem_cache *cfq_pool;
static struct kmem_cache *cfq_ioc_pool;
Linus Torvalds's avatar
Linus Torvalds committed
61

62
static DEFINE_PER_CPU(unsigned long, cfq_ioc_count);
63
static struct completion *ioc_gone;
64
static DEFINE_SPINLOCK(ioc_gone_lock);
65

66 67 68 69
#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)

70
#define sample_valid(samples)	((samples) > 80)
71
#define rb_entry_cfqg(node)	rb_entry((node), struct cfq_group, rb_node)
72

73 74 75 76 77 78 79 80 81
/*
 * 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;
82
	unsigned count;
83
	u64 min_vdisktime;
84
	struct rb_node *active;
85
	unsigned total_weight;
86
};
87
#define CFQ_RB_ROOT	(struct cfq_rb_root) { RB_ROOT, NULL, 0, 0, }
88

89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117
/*
 * Per process-grouping structure
 */
struct cfq_queue {
	/* reference count */
	atomic_t ref;
	/* 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;

118 119 120 121
	/* time when queue got scheduled in to dispatch first request. */
	unsigned long dispatch_start;
	/* time when first request from queue completed and slice started. */
	unsigned long slice_start;
122 123 124 125 126 127 128 129 130 131 132 133 134
	unsigned long slice_end;
	long slice_resid;
	unsigned int slice_dispatch;

	/* pending metadata requests */
	int meta_pending;
	/* number of requests that are on the dispatch list or inside driver */
	int dispatched;

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

135 136 137 138
	unsigned int seek_samples;
	u64 seek_total;
	sector_t seek_mean;
	sector_t last_request_pos;
139
	unsigned long seeky_start;
140

141
	pid_t pid;
Jeff Moyer's avatar
Jeff Moyer committed
142

143
	struct cfq_rb_root *service_tree;
Jeff Moyer's avatar
Jeff Moyer committed
144
	struct cfq_queue *new_cfqq;
145
	struct cfq_group *cfqg;
146 147
	/* Sectors dispatched in current dispatch round */
	unsigned long nr_sectors;
148 149
};

150
/*
151
 * First index in the service_trees.
152 153 154 155
 * IDLE is handled separately, so it has negative index
 */
enum wl_prio_t {
	BE_WORKLOAD = 0,
156 157
	RT_WORKLOAD = 1,
	IDLE_WORKLOAD = 2,
158 159
};

160 161 162 163 164 165 166 167 168
/*
 * Second index in the service_trees.
 */
enum wl_type_t {
	ASYNC_WORKLOAD = 0,
	SYNC_NOIDLE_WORKLOAD = 1,
	SYNC_WORKLOAD = 2
};

169 170
/* This is per cgroup per device grouping structure */
struct cfq_group {
171 172 173 174 175
	/* group service_tree member */
	struct rb_node rb_node;

	/* group service_tree key */
	u64 vdisktime;
176
	unsigned int weight;
177 178 179 180 181
	bool on_st;

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

182 183
	/* Per group busy queus average. Useful for workload slice calc. */
	unsigned int busy_queues_avg[2];
184 185 186 187 188 189
	/*
	 * rr lists of queues with requests, onle rr for each priority class.
	 * Counts are embedded in the cfq_rb_root
	 */
	struct cfq_rb_root service_trees[2][3];
	struct cfq_rb_root service_tree_idle;
190 191 192 193

	unsigned long saved_workload_slice;
	enum wl_type_t saved_workload;
	enum wl_prio_t saved_serving_prio;
194 195 196
	struct blkio_group blkg;
#ifdef CONFIG_CFQ_GROUP_IOSCHED
	struct hlist_node cfqd_node;
197
	atomic_t ref;
198
#endif
199
};
200

201 202 203
/*
 * Per block device queue structure
 */
Linus Torvalds's avatar
Linus Torvalds committed
204
struct cfq_data {
205
	struct request_queue *queue;
206 207
	/* Root service tree for cfq_groups */
	struct cfq_rb_root grp_service_tree;
208
	struct cfq_group root_group;
209 210
	/* Number of active cfq groups on group service tree */
	int nr_groups;
211

212 213
	/*
	 * The priority currently being served
214
	 */
215
	enum wl_prio_t serving_prio;
216 217
	enum wl_type_t serving_type;
	unsigned long workload_expires;
218
	struct cfq_group *serving_group;
219
	bool noidle_tree_requires_idle;
220 221 222 223 224 225 226 227

	/*
	 * 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];

228 229
	unsigned int busy_queues;

230
	int rq_in_driver[2];
231
	int sync_flight;
232 233 234 235 236

	/*
	 * queue-depth detection
	 */
	int rq_queued;
237
	int hw_tag;
238 239 240 241 242 243 244 245
	/*
	 * 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
246

247 248 249 250
	/*
	 * idle window management
	 */
	struct timer_list idle_slice_timer;
251
	struct work_struct unplug_work;
Linus Torvalds's avatar
Linus Torvalds committed
252

253 254 255
	struct cfq_queue *active_queue;
	struct cfq_io_context *active_cic;

256 257 258 259 260
	/*
	 * async queue for each priority case
	 */
	struct cfq_queue *async_cfqq[2][IOPRIO_BE_NR];
	struct cfq_queue *async_idle_cfqq;
261

Jens Axboe's avatar
Jens Axboe committed
262
	sector_t last_position;
Linus Torvalds's avatar
Linus Torvalds committed
263 264 265 266 267

	/*
	 * tunables, see top of file
	 */
	unsigned int cfq_quantum;
268
	unsigned int cfq_fifo_expire[2];
Linus Torvalds's avatar
Linus Torvalds committed
269 270
	unsigned int cfq_back_penalty;
	unsigned int cfq_back_max;
271 272 273
	unsigned int cfq_slice[2];
	unsigned int cfq_slice_async_rq;
	unsigned int cfq_slice_idle;
274
	unsigned int cfq_latency;
275 276

	struct list_head cic_list;
Linus Torvalds's avatar
Linus Torvalds committed
277

278 279 280 281
	/*
	 * Fallback dummy cfqq for extreme OOM conditions
	 */
	struct cfq_queue oom_cfqq;
282 283

	unsigned long last_end_sync_rq;
284 285 286

	/* List of cfq groups being managed on this device*/
	struct hlist_head cfqg_list;
Linus Torvalds's avatar
Linus Torvalds committed
287 288
};

289 290
static struct cfq_group *cfq_get_next_cfqg(struct cfq_data *cfqd);

291 292
static struct cfq_rb_root *service_tree_for(struct cfq_group *cfqg,
					    enum wl_prio_t prio,
293
					    enum wl_type_t type,
294 295
					    struct cfq_data *cfqd)
{
296 297 298
	if (!cfqg)
		return NULL;

299
	if (prio == IDLE_WORKLOAD)
300
		return &cfqg->service_tree_idle;
301

302
	return &cfqg->service_trees[prio][type];
303 304
}

Jens Axboe's avatar
Jens Axboe committed
305
enum cfqq_state_flags {
306 307
	CFQ_CFQQ_FLAG_on_rr = 0,	/* on round-robin busy list */
	CFQ_CFQQ_FLAG_wait_request,	/* waiting for a request */
308
	CFQ_CFQQ_FLAG_must_dispatch,	/* must be allowed a dispatch */
309 310 311 312
	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 */
313
	CFQ_CFQQ_FLAG_slice_new,	/* no requests dispatched in slice */
314
	CFQ_CFQQ_FLAG_sync,		/* synchronous queue */
315
	CFQ_CFQQ_FLAG_coop,		/* cfqq is shared */
316
	CFQ_CFQQ_FLAG_deep,		/* sync cfqq experienced large depth */
Jens Axboe's avatar
Jens Axboe committed
317 318 319 320 321
};

#define CFQ_CFQQ_FNS(name)						\
static inline void cfq_mark_cfqq_##name(struct cfq_queue *cfqq)		\
{									\
322
	(cfqq)->flags |= (1 << CFQ_CFQQ_FLAG_##name);			\
Jens Axboe's avatar
Jens Axboe committed
323 324 325
}									\
static inline void cfq_clear_cfqq_##name(struct cfq_queue *cfqq)	\
{									\
326
	(cfqq)->flags &= ~(1 << CFQ_CFQQ_FLAG_##name);			\
Jens Axboe's avatar
Jens Axboe committed
327 328 329
}									\
static inline int cfq_cfqq_##name(const struct cfq_queue *cfqq)		\
{									\
330
	return ((cfqq)->flags & (1 << CFQ_CFQQ_FLAG_##name)) != 0;	\
Jens Axboe's avatar
Jens Axboe committed
331 332 333 334
}

CFQ_CFQQ_FNS(on_rr);
CFQ_CFQQ_FNS(wait_request);
335
CFQ_CFQQ_FNS(must_dispatch);
Jens Axboe's avatar
Jens Axboe committed
336 337 338 339
CFQ_CFQQ_FNS(must_alloc_slice);
CFQ_CFQQ_FNS(fifo_expire);
CFQ_CFQQ_FNS(idle_window);
CFQ_CFQQ_FNS(prio_changed);
340
CFQ_CFQQ_FNS(slice_new);
341
CFQ_CFQQ_FNS(sync);
342
CFQ_CFQQ_FNS(coop);
343
CFQ_CFQQ_FNS(deep);
Jens Axboe's avatar
Jens Axboe committed
344 345
#undef CFQ_CFQQ_FNS

346 347 348 349 350 351 352 353 354 355 356
#ifdef CONFIG_DEBUG_CFQ_IOSCHED
#define cfq_log_cfqq(cfqd, cfqq, fmt, args...)	\
	blk_add_trace_msg((cfqd)->queue, "cfq%d%c %s " fmt, (cfqq)->pid, \
			cfq_cfqq_sync((cfqq)) ? 'S' : 'A', \
			blkg_path(&(cfqq)->cfqg->blkg), ##args);

#define cfq_log_cfqg(cfqd, cfqg, fmt, args...)				\
	blk_add_trace_msg((cfqd)->queue, "%s " fmt,			\
				blkg_path(&(cfqg)->blkg), ##args);      \

#else
357 358
#define cfq_log_cfqq(cfqd, cfqq, fmt, args...)	\
	blk_add_trace_msg((cfqd)->queue, "cfq%d " fmt, (cfqq)->pid, ##args)
359 360
#define cfq_log_cfqg(cfqd, cfqg, fmt, args...)		do {} while (0);
#endif
361 362 363
#define cfq_log(cfqd, fmt, args...)	\
	blk_add_trace_msg((cfqd)->queue, "cfq " fmt, ##args)

364 365 366 367 368 369 370 371 372 373 374
/* 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) \


375 376 377 378 379 380 381 382 383
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;
}

384 385 386 387 388 389 390 391 392 393

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;
}

394 395 396
static inline int cfq_group_busy_queues_wl(enum wl_prio_t wl,
					struct cfq_data *cfqd,
					struct cfq_group *cfqg)
397 398
{
	if (wl == IDLE_WORKLOAD)
399
		return cfqg->service_tree_idle.count;
400

401 402 403
	return cfqg->service_trees[wl][ASYNC_WORKLOAD].count
		+ cfqg->service_trees[wl][SYNC_NOIDLE_WORKLOAD].count
		+ cfqg->service_trees[wl][SYNC_WORKLOAD].count;
404 405
}

406
static void cfq_dispatch_insert(struct request_queue *, struct request *);
407
static struct cfq_queue *cfq_get_queue(struct cfq_data *, bool,
408
				       struct io_context *, gfp_t);
409
static struct cfq_io_context *cfq_cic_lookup(struct cfq_data *,
410 411
						struct io_context *);

412 413 414 415 416
static inline int rq_in_driver(struct cfq_data *cfqd)
{
	return cfqd->rq_in_driver[0] + cfqd->rq_in_driver[1];
}

417
static inline struct cfq_queue *cic_to_cfqq(struct cfq_io_context *cic,
418
					    bool is_sync)
419
{
420
	return cic->cfqq[is_sync];
421 422 423
}

static inline void cic_set_cfqq(struct cfq_io_context *cic,
424
				struct cfq_queue *cfqq, bool is_sync)
425
{
426
	cic->cfqq[is_sync] = cfqq;
427 428 429 430 431 432
}

/*
 * 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).
 */
433
static inline bool cfq_bio_sync(struct bio *bio)
434
{
435
	return bio_data_dir(bio) == READ || bio_rw_flagged(bio, BIO_RW_SYNCIO);
436
}
Linus Torvalds's avatar
Linus Torvalds committed
437

Andrew Morton's avatar
Andrew Morton committed
438 439 440 441
/*
 * scheduler run of queue, if there are requests pending and no one in the
 * driver that will restart queueing
 */
442
static inline void cfq_schedule_dispatch(struct cfq_data *cfqd)
Andrew Morton's avatar
Andrew Morton committed
443
{
444 445
	if (cfqd->busy_queues) {
		cfq_log(cfqd, "schedule dispatch");
446
		kblockd_schedule_work(cfqd->queue, &cfqd->unplug_work);
447
	}
Andrew Morton's avatar
Andrew Morton committed
448 449
}

450
static int cfq_queue_empty(struct request_queue *q)
Andrew Morton's avatar
Andrew Morton committed
451 452 453
{
	struct cfq_data *cfqd = q->elevator->elevator_data;

454
	return !cfqd->rq_queued;
Andrew Morton's avatar
Andrew Morton committed
455 456
}

457 458 459 460 461
/*
 * 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.
 */
462
static inline int cfq_prio_slice(struct cfq_data *cfqd, bool sync,
463
				 unsigned short prio)
464
{
465
	const int base_slice = cfqd->cfq_slice[sync];
466

467 468 469 470
	WARN_ON(prio >= IOPRIO_BE_NR);

	return base_slice + (base_slice/CFQ_SLICE_SCALE * (4 - prio));
}
471

472 473 474 475
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);
476 477
}

478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522
static inline u64 cfq_scale_slice(unsigned long delta, struct cfq_group *cfqg)
{
	u64 d = delta << CFQ_SERVICE_SHIFT;

	d = d * BLKIO_WEIGHT_DEFAULT;
	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)
{
	u64 vdisktime = st->min_vdisktime;
	struct cfq_group *cfqg;

	if (st->active) {
		cfqg = rb_entry_cfqg(st->active);
		vdisktime = cfqg->vdisktime;
	}

	if (st->left) {
		cfqg = rb_entry_cfqg(st->left);
		vdisktime = min_vdisktime(vdisktime, cfqg->vdisktime);
	}

	st->min_vdisktime = max_vdisktime(st->min_vdisktime, vdisktime);
}

523 524 525 526 527 528
/*
 * 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
 */

529 530
static inline unsigned cfq_group_get_avg_queues(struct cfq_data *cfqd,
					struct cfq_group *cfqg, bool rt)
531
{
532 533 534
	unsigned min_q, max_q;
	unsigned mult  = cfq_hist_divisor - 1;
	unsigned round = cfq_hist_divisor / 2;
535
	unsigned busy = cfq_group_busy_queues_wl(rt, cfqd, cfqg);
536

537 538 539
	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) /
540
		cfq_hist_divisor;
541 542 543 544 545 546 547 548 549
	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;
550 551
}

552 553 554
static inline void
cfq_set_prio_slice(struct cfq_data *cfqd, struct cfq_queue *cfqq)
{
555 556
	unsigned slice = cfq_prio_to_slice(cfqd, cfqq);
	if (cfqd->cfq_latency) {
557 558 559 560 561 562
		/*
		 * 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));
563 564
		unsigned sync_slice = cfqd->cfq_slice[1];
		unsigned expect_latency = sync_slice * iq;
565 566 567
		unsigned group_slice = cfq_group_slice(cfqd, cfqq->cfqg);

		if (expect_latency > group_slice) {
568 569 570 571 572 573 574
			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 */
575
			slice = max(slice * group_slice / expect_latency,
576 577 578
				    low_slice);
		}
	}
579
	cfqq->slice_start = jiffies;
580
	cfqq->slice_end = jiffies + slice;
581
	cfq_log_cfqq(cfqd, cfqq, "set_slice=%lu", cfqq->slice_end - jiffies);
582 583 584 585 586 587 588
}

/*
 * 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.
 */
589
static inline bool cfq_slice_used(struct cfq_queue *cfqq)
590 591 592 593 594 595 596 597 598
{
	if (cfq_cfqq_slice_new(cfqq))
		return 0;
	if (time_before(jiffies, cfqq->slice_end))
		return 0;

	return 1;
}

Linus Torvalds's avatar
Linus Torvalds committed
599
/*
Jens Axboe's avatar
Jens Axboe committed
600
 * Lifted from AS - choose which of rq1 and rq2 that is best served now.
Linus Torvalds's avatar
Linus Torvalds committed
601
 * We choose the request that is closest to the head right now. Distance
602
 * behind the head is penalized and only allowed to a certain extent.
Linus Torvalds's avatar
Linus Torvalds committed
603
 */
Jens Axboe's avatar
Jens Axboe committed
604
static struct request *
605
cfq_choose_req(struct cfq_data *cfqd, struct request *rq1, struct request *rq2, sector_t last)
Linus Torvalds's avatar
Linus Torvalds committed
606
{
607
	sector_t s1, s2, d1 = 0, d2 = 0;
Linus Torvalds's avatar
Linus Torvalds committed
608
	unsigned long back_max;
609 610 611
#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
612

Jens Axboe's avatar
Jens Axboe committed
613 614 615 616
	if (rq1 == NULL || rq1 == rq2)
		return rq2;
	if (rq2 == NULL)
		return rq1;
617

Jens Axboe's avatar
Jens Axboe committed
618 619 620 621
	if (rq_is_sync(rq1) && !rq_is_sync(rq2))
		return rq1;
	else if (rq_is_sync(rq2) && !rq_is_sync(rq1))
		return rq2;
622 623 624 625
	if (rq_is_meta(rq1) && !rq_is_meta(rq2))
		return rq1;
	else if (rq_is_meta(rq2) && !rq_is_meta(rq1))
		return rq2;
Linus Torvalds's avatar
Linus Torvalds committed
626

627 628
	s1 = blk_rq_pos(rq1);
	s2 = blk_rq_pos(rq2);
Linus Torvalds's avatar
Linus Torvalds committed
629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644

	/*
	 * 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
645
		wrap |= CFQ_RQ1_WRAP;
Linus Torvalds's avatar
Linus Torvalds committed
646 647 648 649 650 651

	if (s2 >= last)
		d2 = s2 - last;
	else if (s2 + back_max >= last)
		d2 = (last - s2) * cfqd->cfq_back_penalty;
	else
652
		wrap |= CFQ_RQ2_WRAP;
Linus Torvalds's avatar
Linus Torvalds committed
653 654

	/* Found required data */
655 656 657 658 659 660

	/*
	 * By doing switch() on the bit mask "wrap" we avoid having to
	 * check two variables for all permutations: --> faster!
	 */
	switch (wrap) {
Jens Axboe's avatar
Jens Axboe committed
661
	case 0: /* common case for CFQ: rq1 and rq2 not wrapped */
662
		if (d1 < d2)
Jens Axboe's avatar
Jens Axboe committed
663
			return rq1;
664
		else if (d2 < d1)
Jens Axboe's avatar
Jens Axboe committed
665
			return rq2;
666 667
		else {
			if (s1 >= s2)
Jens Axboe's avatar
Jens Axboe committed
668
				return rq1;
669
			else
Jens Axboe's avatar
Jens Axboe committed
670
				return rq2;
671
		}
Linus Torvalds's avatar
Linus Torvalds committed
672

673
	case CFQ_RQ2_WRAP:
Jens Axboe's avatar
Jens Axboe committed
674
		return rq1;
675
	case CFQ_RQ1_WRAP:
Jens Axboe's avatar
Jens Axboe committed
676 677
		return rq2;
	case (CFQ_RQ1_WRAP|CFQ_RQ2_WRAP): /* both rqs wrapped */
678 679 680 681 682 683 684 685
	default:
		/*
		 * Since both rqs are wrapped,
		 * start with the one that's further behind head
		 * (--> only *one* back seek required),
		 * since back seek takes more time than forward.
		 */
		if (s1 <= s2)
Jens Axboe's avatar
Jens Axboe committed
686
			return rq1;
Linus Torvalds's avatar
Linus Torvalds committed
687
		else
Jens Axboe's avatar
Jens Axboe committed
688
			return rq2;
Linus Torvalds's avatar
Linus Torvalds committed
689 690 691
	}
}

692 693 694
/*
 * The below is leftmost cache rbtree addon
 */
695
static struct cfq_queue *cfq_rb_first(struct cfq_rb_root *root)
696
{
697 698 699 700
	/* Service tree is empty */
	if (!root->count)
		return NULL;

701 702 703
	if (!root->left)
		root->left = rb_first(&root->rb);

704 705 706 707
	if (root->left)
		return rb_entry(root->left, struct cfq_queue, rb_node);

	return NULL;
708 709
}

710 711 712 713 714 715 716 717 718 719 720
static struct cfq_group *cfq_rb_first_group(struct cfq_rb_root *root)
{
	if (!root->left)
		root->left = rb_first(&root->rb);

	if (root->left)
		return rb_entry_cfqg(root->left);

	return NULL;
}

721 722 723 724 725 726
static void rb_erase_init(struct rb_node *n, struct rb_root *root)
{
	rb_erase(n, root);
	RB_CLEAR_NODE(n);
}

727 728 729 730
static void cfq_rb_erase(struct rb_node *n, struct cfq_rb_root *root)
{
	if (root->left == n)
		root->left = NULL;
731
	rb_erase_init(n, &root->rb);
732
	--root->count;
733 734
}

Linus Torvalds's avatar
Linus Torvalds committed
735 736 737
/*
 * would be nice to take fifo expire time into account as well
 */
Jens Axboe's avatar
Jens Axboe committed
738 739 740
static struct request *
cfq_find_next_rq(struct cfq_data *cfqd, struct cfq_queue *cfqq,
		  struct request *last)
Linus Torvalds's avatar
Linus Torvalds committed
741
{
742 743
	struct rb_node *rbnext = rb_next(&last->rb_node);
	struct rb_node *rbprev = rb_prev(&last->rb_node);
Jens Axboe's avatar
Jens Axboe committed
744
	struct request *next = NULL, *prev = NULL;
Linus Torvalds's avatar
Linus Torvalds committed
745

746
	BUG_ON(RB_EMPTY_NODE(&last->rb_node));
Linus Torvalds's avatar
Linus Torvalds committed
747 748

	if (rbprev)
Jens Axboe's avatar
Jens Axboe committed
749
		prev = rb_entry_rq(rbprev);
Linus Torvalds's avatar
Linus Torvalds committed
750

751
	if (rbnext)
Jens Axboe's avatar
Jens Axboe committed
752
		next = rb_entry_rq(rbnext);
753 754 755
	else {
		rbnext = rb_first(&cfqq->sort_list);
		if (rbnext && rbnext != &last->rb_node)
Jens Axboe's avatar
Jens Axboe committed
756
			next = rb_entry_rq(rbnext);
757
	}
Linus Torvalds's avatar
Linus Torvalds committed
758

759
	return cfq_choose_req(cfqd, next, prev, blk_rq_pos(last));
Linus Torvalds's avatar
Linus Torvalds committed
760 761
}

762 763
static unsigned long cfq_slice_offset(struct cfq_data *cfqd,
				      struct cfq_queue *cfqq)
Linus Torvalds's avatar
Linus Torvalds committed
764
{
765 766 767
	/*
	 * just an approximation, should be ok.
	 */
768
	return (cfqq->cfqg->nr_cfqq - 1) * (cfq_prio_slice(cfqd, 1, 0) -
769
		       cfq_prio_slice(cfqd, cfq_cfqq_sync(cfqq), cfqq->ioprio));
770 771
}

772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830
static inline s64
cfqg_key(struct cfq_rb_root *st, struct cfq_group *cfqg)
{
	return cfqg->vdisktime - st->min_vdisktime;
}

static void
__cfq_group_service_tree_add(struct cfq_rb_root *st, struct cfq_group *cfqg)
{
	struct rb_node **node = &st->rb.rb_node;
	struct rb_node *parent = NULL;
	struct cfq_group *__cfqg;
	s64 key = cfqg_key(st, cfqg);
	int left = 1;

	while (*node != NULL) {
		parent = *node;
		__cfqg = rb_entry_cfqg(parent);

		if (key < cfqg_key(st, __cfqg))
			node = &parent->rb_left;
		else {
			node = &parent->rb_right;
			left = 0;
		}
	}

	if (left)
		st->left = &cfqg->rb_node;

	rb_link_node(&cfqg->rb_node, parent, node);
	rb_insert_color(&cfqg->rb_node, &st->rb);
}

static void
cfq_group_service_tree_add(struct cfq_data *cfqd, struct cfq_group *cfqg)
{
	struct cfq_rb_root *st = &cfqd->grp_service_tree;
	struct cfq_group *__cfqg;
	struct rb_node *n;

	cfqg->nr_cfqq++;
	if (cfqg->on_st)
		return;

	/*
	 * Currently put the group at the end. Later implement something
	 * so that groups get lesser vtime based on their weights, so that
	 * if group does not loose all if it was not continously backlogged.
	 */
	n = rb_last(&st->rb);
	if (n) {
		__cfqg = rb_entry_cfqg(n);
		cfqg->vdisktime = __cfqg->vdisktime + CFQ_IDLE_DELAY;
	} else
		cfqg->vdisktime = st->min_vdisktime;

	__cfq_group_service_tree_add(st, cfqg);
	cfqg->on_st = true;
831 832
	cfqd->nr_groups++;
	st->total_weight += cfqg->weight;
833 834 835 836 837 838 839
}

static void
cfq_group_service_tree_del(struct cfq_data *cfqd, struct cfq_group *cfqg)
{
	struct cfq_rb_root *st = &cfqd->grp_service_tree;

840 841 842
	if (st->active == &cfqg->rb_node)
		st->active = NULL;

843 844
	BUG_ON(cfqg->nr_cfqq < 1);
	cfqg->nr_cfqq--;
845

846 847 848 849
	/* If there are other cfq queues under this group, don't delete it */
	if (cfqg->nr_cfqq)
		return;

850
	cfq_log_cfqg(cfqd, cfqg, "del_from_rr group");
851
	cfqg->on_st = false;
852 853
	cfqd->nr_groups--;
	st->total_weight -= cfqg->weight;
854 855
	if (!RB_EMPTY_NODE(&cfqg->rb_node))
		cfq_rb_erase(&cfqg->rb_node, st);
856
	cfqg->saved_workload_slice = 0;
857
	blkiocg_update_blkio_group_dequeue_stats(&cfqg->blkg, 1);
858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883
}

static inline unsigned int cfq_cfqq_slice_usage(struct cfq_queue *cfqq)
{
	unsigned int slice_used, allocated_slice;

	/*
	 * Queue got expired before even a single request completed or
	 * got expired immediately after first request completion.
	 */
	if (!cfqq->slice_start || cfqq->slice_start == jiffies) {
		/*
		 * Also charge the seek time incurred to the group, otherwise
		 * if there are mutiple queues in the group, each can dispatch
		 * a single request on seeky media and cause lots of seek time
		 * and group will never know it.
		 */
		slice_used = max_t(unsigned, (jiffies - cfqq->dispatch_start),
					1);
	} else {
		slice_used = jiffies - cfqq->slice_start;
		allocated_slice = cfqq->slice_end - cfqq->slice_start;
		if (slice_used > allocated_slice)
			slice_used = allocated_slice;
	}

884 885
	cfq_log_cfqq(cfqq->cfqd, cfqq, "sl_used=%u sect=%lu", slice_used,
				cfqq->nr_sectors);
886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909
	return slice_used;
}

static void cfq_group_served(struct cfq_data *cfqd, struct cfq_group *cfqg,
				struct cfq_queue *cfqq)
{
	struct cfq_rb_root *st = &cfqd->grp_service_tree;
	unsigned int used_sl;

	used_sl = cfq_cfqq_slice_usage(cfqq);

	/* Can't update vdisktime while group is on service tree */
	cfq_rb_erase(&cfqg->rb_node, st);
	cfqg->vdisktime += cfq_scale_slice(used_sl, cfqg);
	__cfq_group_service_tree_add(st, cfqg);

	/* This group is being expired. Save the context */
	if (time_after(cfqd->workload_expires, jiffies)) {
		cfqg->saved_workload_slice = cfqd->workload_expires
						- jiffies;
		cfqg->saved_workload = cfqd->serving_type;
		cfqg->saved_serving_prio = cfqd->serving_prio;
	} else
		cfqg->saved_workload_slice = 0;
910 911 912

	cfq_log_cfqg(cfqd, cfqg, "served: vt=%llu min_vt=%llu", cfqg->vdisktime,
					st->min_vdisktime);
913 914
	blkiocg_update_blkio_group_stats(&cfqg->blkg, used_sl,
						cfqq->nr_sectors);
915 916
}

917 918 919 920 921 922 923 924
#ifdef CONFIG_CFQ_GROUP_IOSCHED
static inline struct cfq_group *cfqg_of_blkg(struct blkio_group *blkg)
{
	if (blkg)
		return container_of(blkg, struct cfq_group, blkg);
	return NULL;
}

925 926 927 928 929 930
void
cfq_update_blkio_group_weight(struct blkio_group *blkg, unsigned int weight)
{
	cfqg_of_blkg(blkg)->weight = weight;
}

931 932 933 934 935 936 937 938
static struct cfq_group *
cfq_find_alloc_cfqg(struct cfq_data *cfqd, struct cgroup *cgroup, int create)
{
	struct blkio_cgroup *blkcg = cgroup_to_blkio_cgroup(cgroup);
	struct cfq_group *cfqg = NULL;
	void *key = cfqd;
	int i, j;
	struct cfq_rb_root *st;
939 940
	struct backing_dev_info *bdi = &cfqd->queue->backing_dev_info;
	unsigned int major, minor;
941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958

	/* Do we need to take this reference */
	if (!css_tryget(&blkcg->css))
		return NULL;;

	cfqg = cfqg_of_blkg(blkiocg_lookup_group(blkcg, key));
	if (cfqg || !create)
		goto done;

	cfqg = kzalloc_node(sizeof(*cfqg), GFP_ATOMIC, cfqd->queue->node);
	if (!cfqg)
		goto done;

	cfqg->weight = blkcg->weight;
	for_each_cfqg_st(cfqg, i, j, st)
		*st = CFQ_RB_ROOT;
	RB_CLEAR_NODE(&cfqg->rb_node);

959 960 961 962 963 964 965 966
	/*
	 * Take the initial reference that will be released on destroy
	 * This can be thought of a joint reference by cgroup and
	 * elevator which will be dropped by either elevator exit
	 * or cgroup deletion path depending on who is exiting first.
	 */
	atomic_set(&cfqg->ref, 1);

967
	/* Add group onto cgroup list */
968 969 970
	sscanf(dev_name(bdi->dev), "%u:%u", &major, &minor);
	blkiocg_add_blkio_group(blkcg, &cfqg->blkg, (void *)cfqd,
					MKDEV(major, minor));
971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004

	/* Add group on cfqd list */
	hlist_add_head(&cfqg->cfqd_node, &cfqd->cfqg_list);

done:
	css_put(&blkcg->css);
	return cfqg;
}

/*
 * Search for the cfq group current task belongs to. If create = 1, then also
 * create the cfq group if it does not exist. request_queue lock must be held.
 */
static struct cfq_group *cfq_get_cfqg(struct cfq_data *cfqd, int create)
{
	struct cgroup *cgroup;
	struct cfq_group *cfqg = NULL;

	rcu_read_lock();
	cgroup = task_cgroup(current, blkio_subsys_id);
	cfqg = cfq_find_alloc_cfqg(cfqd, cgroup, create);
	if (!cfqg && create)
		cfqg = &cfqd->root_group;
	rcu_read_unlock();
	return cfqg;
}

static void cfq_link_cfqq_cfqg(struct cfq_queue *cfqq, struct cfq_group *cfqg)
{
	/* Currently, all async queues are mapped to root group */
	if (!cfq_cfqq_sync(cfqq))
		cfqg = &cfqq->cfqd->root_group;

	cfqq->cfqg = cfqg;
1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049
	/* cfqq reference on cfqg */
	atomic_inc(&cfqq->cfqg->ref);
}

static void cfq_put_cfqg(struct cfq_group *cfqg)
{
	struct cfq_rb_root *st;
	int i, j;

	BUG_ON(atomic_read(&cfqg->ref) <= 0);
	if (!atomic_dec_and_test(&cfqg->ref))
		return;
	for_each_cfqg_st(cfqg, i, j, st)
		BUG_ON(!RB_EMPTY_ROOT(&st->rb) || st->active != NULL);
	kfree(cfqg);
}

static void cfq_destroy_cfqg(struct cfq_data *cfqd, struct cfq_group *cfqg)
{
	/* Something wrong if we are trying to remove same group twice */
	BUG_ON(hlist_unhashed(&cfqg->cfqd_node));

	hlist_del_init(&cfqg->cfqd_node);

	/*
	 * Put the reference taken at the time of creation so that when all
	 * queues are gone, group can be destroyed.
	 */
	cfq_put_cfqg(cfqg);
}

static void cfq_release_cfq_groups(struct cfq_data *cfqd)
{
	struct hlist_node *pos, *n;
	struct cfq_group *cfqg;

	hlist_for_each_entry_safe(cfqg, pos, n, &cfqd->cfqg_list, cfqd_node) {
		/*
		 * If cgroup removal path got to blk_group first and removed
		 * it from cgroup list, then it will take care of destroying
		 * cfqg also.
		 */
		if (!blkiocg_del_blkio_group(&cfqg->blkg))
			cfq_destroy_cfqg(cfqd, cfqg);
	}
1050
}
1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075

/*
 * Blk cgroup controller notification saying that blkio_group object is being
 * delinked as associated cgroup object is going away. That also means that
 * no new IO will come in this group. So get rid of this group as soon as
 * any pending IO in the group is finished.
 *
 * This function is called under rcu_read_lock(). key is the rcu protected
 * pointer. That means "key" is a valid cfq_data pointer as long as we are rcu
 * read lock.
 *
 * "key" was fetched from blkio_group under blkio_cgroup->lock. That means
 * it should not be NULL as even if elevator was exiting, cgroup deltion
 * path got to it first.
 */
void cfq_unlink_blkio_group(void *key, struct blkio_group *blkg)
{
	unsigned long  flags;
	struct cfq_data *cfqd = key;

	spin_lock_irqsave(cfqd->queue->queue_lock, flags);
	cfq_destroy_cfqg(cfqd, cfqg_of_blkg(blkg));
	spin_unlock_irqrestore(cfqd->queue->queue_lock, flags);
}

1076 1077 1078 1079 1080 1081 1082 1083 1084 1085
#else /* GROUP_IOSCHED */
static struct cfq_group *cfq_get_cfqg(struct cfq_data *cfqd, int create)
{
	return &cfqd->root_group;
}
static inline void
cfq_link_cfqq_cfqg(struct cfq_queue *cfqq, struct cfq_group *cfqg) {
	cfqq->cfqg = cfqg;
}

1086 1087 1088
static void cfq_release_cfq_groups(struct cfq_data *cfqd) {}
static inline void cfq_put_cfqg(struct cfq_group *cfqg) {}

1089 1090
#endif /* GROUP_IOSCHED */

1091
/*
1092
 * The cfqd->service_trees holds all pending cfq_queue's that have
1093 1094 1095
 * requests waiting to be processed. It is sorted in the order that
 * we will service the queues.
 */
1096
static void cfq_service_tree_add(struct cfq_data *cfqd, struct cfq_queue *cfqq,
1097
				 bool add_front)
1098
{
1099 1100
	struct rb_node **p, *parent;
	struct cfq_queue *__cfqq;
1101
	unsigned long rb_key;
1102
	struct cfq_rb_root *service_tree;
1103
	int left;
1104
	int new_cfqq = 1;
1105

1106 1107
	service_tree = service_tree_for(cfqq->cfqg, cfqq_prio(cfqq),
						cfqq_type(cfqq), cfqd);
1108 1109
	if (cfq_class_idle(cfqq)) {
		rb_key = CFQ_IDLE_DELAY;
1110
		parent = rb_last(&service_tree->rb);
1111 1112 1113 1114 1115 1116
		if (parent && parent != &cfqq->rb_node) {
			__cfqq = rb_entry(parent, struct cfq_queue, rb_node);
			rb_key += __cfqq->rb_key;
		} else
			rb_key += jiffies;
	} else if (!add_front) {
1117 1118 1119 1120 1121 1122
		/*
		 * Get our rb key offset. Subtract any residual slice
		 * value carried from last service. A negative resid
		 * count indicates slice overrun, and this should position
		 * the next service time further away in the tree.
		 */
1123
		rb_key = cfq_slice_offset(cfqd, cfqq) + jiffies;
1124
		rb_key -= cfqq->slice_resid;
1125
		cfqq->slice_resid = 0;
1126 1127
	} else {
		rb_key = -HZ;
1128
		__cfqq = cfq_rb_first(service_tree);
1129 1130
		rb_key += __cfqq ? __cfqq->rb_key : jiffies;
	}
Linus Torvalds's avatar
Linus Torvalds committed
1131

1132
	if (!RB_EMPTY_NODE(&cfqq->rb_node)) {
1133
		new_cfqq = 0;
1134
		/*
1135
		 * same position, nothing more to do
1136
		 */
1137 1138
		if (rb_key == cfqq->rb_key &&
		    cfqq->service_tree == service_tree)
1139
			return;
Linus Torvalds's avatar
Linus Torvalds committed
1140

1141 1142
		cfq_rb_erase(&cfqq->rb_node, cfqq->service_tree);
		cfqq->service_tree = NULL;
Linus Torvalds's avatar
Linus Torvalds committed
1143
	}
1144

1145
	left = 1;
1146
	parent = NULL;
1147 1148
	cfqq->service_tree = service_tree;
	p = &service_tree->rb.rb_node;
1149
	while (*p) {
1150
		struct rb_node **n;
1151