cfq-iosched.c 94.3 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
	/* time when queue got scheduled in to dispatch first request. */
	unsigned long dispatch_start;
120
	unsigned int allocated_slice;
121 122
	/* time when first request from queue completed and slice started. */
	unsigned long slice_start;
123 124 125 126 127 128 129 130 131 132 133 134 135
	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;

136 137 138 139
	unsigned int seek_samples;
	u64 seek_total;
	sector_t seek_mean;
	sector_t last_request_pos;
140
	unsigned long seeky_start;
141

142
	pid_t pid;
Jeff Moyer's avatar
Jeff Moyer committed
143

144
	struct cfq_rb_root *service_tree;
Jeff Moyer's avatar
Jeff Moyer committed
145
	struct cfq_queue *new_cfqq;
146
	struct cfq_group *cfqg;
147 148
	/* Sectors dispatched in current dispatch round */
	unsigned long nr_sectors;
149 150
};

151
/*
152
 * First index in the service_trees.
153 154 155 156
 * IDLE is handled separately, so it has negative index
 */
enum wl_prio_t {
	BE_WORKLOAD = 0,
157 158
	RT_WORKLOAD = 1,
	IDLE_WORKLOAD = 2,
159 160
};

161 162 163 164 165 166 167 168 169
/*
 * Second index in the service_trees.
 */
enum wl_type_t {
	ASYNC_WORKLOAD = 0,
	SYNC_NOIDLE_WORKLOAD = 1,
	SYNC_WORKLOAD = 2
};

170 171
/* This is per cgroup per device grouping structure */
struct cfq_group {
172 173 174 175 176
	/* group service_tree member */
	struct rb_node rb_node;

	/* group service_tree key */
	u64 vdisktime;
177
	unsigned int weight;
178 179 180 181 182
	bool on_st;

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

183 184
	/* Per group busy queus average. Useful for workload slice calc. */
	unsigned int busy_queues_avg[2];
185 186 187 188 189 190
	/*
	 * 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;
191 192 193 194

	unsigned long saved_workload_slice;
	enum wl_type_t saved_workload;
	enum wl_prio_t saved_serving_prio;
195 196 197
	struct blkio_group blkg;
#ifdef CONFIG_CFQ_GROUP_IOSCHED
	struct hlist_node cfqd_node;
198
	atomic_t ref;
199
#endif
200
};
201

202 203 204
/*
 * Per block device queue structure
 */
Linus Torvalds's avatar
Linus Torvalds committed
205
struct cfq_data {
206
	struct request_queue *queue;
207 208
	/* Root service tree for cfq_groups */
	struct cfq_rb_root grp_service_tree;
209
	struct cfq_group root_group;
210 211
	/* Number of active cfq groups on group service tree */
	int nr_groups;
212

213 214
	/*
	 * The priority currently being served
215
	 */
216
	enum wl_prio_t serving_prio;
217 218
	enum wl_type_t serving_type;
	unsigned long workload_expires;
219
	struct cfq_group *serving_group;
220
	bool noidle_tree_requires_idle;
221 222 223 224 225 226 227 228

	/*
	 * 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];

229 230
	unsigned int busy_queues;

231
	int rq_in_driver[2];
232
	int sync_flight;
233 234 235 236 237

	/*
	 * queue-depth detection
	 */
	int rq_queued;
238
	int hw_tag;
239 240 241 242 243 244 245 246
	/*
	 * 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
247

248 249 250 251
	/*
	 * idle window management
	 */
	struct timer_list idle_slice_timer;
252
	struct work_struct unplug_work;
Linus Torvalds's avatar
Linus Torvalds committed
253

254 255 256
	struct cfq_queue *active_queue;
	struct cfq_io_context *active_cic;

257 258 259 260 261
	/*
	 * async queue for each priority case
	 */
	struct cfq_queue *async_cfqq[2][IOPRIO_BE_NR];
	struct cfq_queue *async_idle_cfqq;
262

Jens Axboe's avatar
Jens Axboe committed
263
	sector_t last_position;
Linus Torvalds's avatar
Linus Torvalds committed
264 265 266 267 268

	/*
	 * tunables, see top of file
	 */
	unsigned int cfq_quantum;
269
	unsigned int cfq_fifo_expire[2];
Linus Torvalds's avatar
Linus Torvalds committed
270 271
	unsigned int cfq_back_penalty;
	unsigned int cfq_back_max;
272 273 274
	unsigned int cfq_slice[2];
	unsigned int cfq_slice_async_rq;
	unsigned int cfq_slice_idle;
275
	unsigned int cfq_latency;
276 277

	struct list_head cic_list;
Linus Torvalds's avatar
Linus Torvalds committed
278

279 280 281 282
	/*
	 * Fallback dummy cfqq for extreme OOM conditions
	 */
	struct cfq_queue oom_cfqq;
283 284

	unsigned long last_end_sync_rq;
285 286 287

	/* List of cfq groups being managed on this device*/
	struct hlist_head cfqg_list;
Linus Torvalds's avatar
Linus Torvalds committed
288 289
};

290 291
static struct cfq_group *cfq_get_next_cfqg(struct cfq_data *cfqd);

292 293
static struct cfq_rb_root *service_tree_for(struct cfq_group *cfqg,
					    enum wl_prio_t prio,
294
					    enum wl_type_t type,
295 296
					    struct cfq_data *cfqd)
{
297 298 299
	if (!cfqg)
		return NULL;

300
	if (prio == IDLE_WORKLOAD)
301
		return &cfqg->service_tree_idle;
302

303
	return &cfqg->service_trees[prio][type];
304 305
}

Jens Axboe's avatar
Jens Axboe committed
306
enum cfqq_state_flags {
307 308
	CFQ_CFQQ_FLAG_on_rr = 0,	/* on round-robin busy list */
	CFQ_CFQQ_FLAG_wait_request,	/* waiting for a request */
309
	CFQ_CFQQ_FLAG_must_dispatch,	/* must be allowed a dispatch */
310 311 312 313
	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 */
314
	CFQ_CFQQ_FLAG_slice_new,	/* no requests dispatched in slice */
315
	CFQ_CFQQ_FLAG_sync,		/* synchronous queue */
316
	CFQ_CFQQ_FLAG_coop,		/* cfqq is shared */
317
	CFQ_CFQQ_FLAG_deep,		/* sync cfqq experienced large depth */
318 319
	CFQ_CFQQ_FLAG_wait_busy,	/* Waiting for next request */
	CFQ_CFQQ_FLAG_wait_busy_done,	/* Got new request. Expire the queue */
Jens Axboe's avatar
Jens Axboe committed
320 321 322 323 324
};

#define CFQ_CFQQ_FNS(name)						\
static inline void cfq_mark_cfqq_##name(struct cfq_queue *cfqq)		\
{									\
325
	(cfqq)->flags |= (1 << CFQ_CFQQ_FLAG_##name);			\
Jens Axboe's avatar
Jens Axboe committed
326 327 328
}									\
static inline void cfq_clear_cfqq_##name(struct cfq_queue *cfqq)	\
{									\
329
	(cfqq)->flags &= ~(1 << CFQ_CFQQ_FLAG_##name);			\
Jens Axboe's avatar
Jens Axboe committed
330 331 332
}									\
static inline int cfq_cfqq_##name(const struct cfq_queue *cfqq)		\
{									\
333
	return ((cfqq)->flags & (1 << CFQ_CFQQ_FLAG_##name)) != 0;	\
Jens Axboe's avatar
Jens Axboe committed
334 335 336 337
}

CFQ_CFQQ_FNS(on_rr);
CFQ_CFQQ_FNS(wait_request);
338
CFQ_CFQQ_FNS(must_dispatch);
Jens Axboe's avatar
Jens Axboe committed
339 340 341 342
CFQ_CFQQ_FNS(must_alloc_slice);
CFQ_CFQQ_FNS(fifo_expire);
CFQ_CFQQ_FNS(idle_window);
CFQ_CFQQ_FNS(prio_changed);
343
CFQ_CFQQ_FNS(slice_new);
344
CFQ_CFQQ_FNS(sync);
345
CFQ_CFQQ_FNS(coop);
346
CFQ_CFQQ_FNS(deep);
347 348
CFQ_CFQQ_FNS(wait_busy);
CFQ_CFQQ_FNS(wait_busy_done);
Jens Axboe's avatar
Jens Axboe committed
349 350
#undef CFQ_CFQQ_FNS

351 352 353 354 355 356 357 358 359 360 361
#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
362 363
#define cfq_log_cfqq(cfqd, cfqq, fmt, args...)	\
	blk_add_trace_msg((cfqd)->queue, "cfq%d " fmt, (cfqq)->pid, ##args)
364 365
#define cfq_log_cfqg(cfqd, cfqg, fmt, args...)		do {} while (0);
#endif
366 367 368
#define cfq_log(cfqd, fmt, args...)	\
	blk_add_trace_msg((cfqd)->queue, "cfq " fmt, ##args)

369 370 371 372 373 374 375 376 377 378 379
/* 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) \


380 381 382 383 384 385 386 387 388
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;
}

389 390 391 392 393 394 395 396 397 398

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;
}

399 400 401
static inline int cfq_group_busy_queues_wl(enum wl_prio_t wl,
					struct cfq_data *cfqd,
					struct cfq_group *cfqg)
402 403
{
	if (wl == IDLE_WORKLOAD)
404
		return cfqg->service_tree_idle.count;
405

406 407 408
	return cfqg->service_trees[wl][ASYNC_WORKLOAD].count
		+ cfqg->service_trees[wl][SYNC_NOIDLE_WORKLOAD].count
		+ cfqg->service_trees[wl][SYNC_WORKLOAD].count;
409 410
}

411
static void cfq_dispatch_insert(struct request_queue *, struct request *);
412
static struct cfq_queue *cfq_get_queue(struct cfq_data *, bool,
413
				       struct io_context *, gfp_t);
414
static struct cfq_io_context *cfq_cic_lookup(struct cfq_data *,
415 416
						struct io_context *);

417 418 419 420 421
static inline int rq_in_driver(struct cfq_data *cfqd)
{
	return cfqd->rq_in_driver[0] + cfqd->rq_in_driver[1];
}

422
static inline struct cfq_queue *cic_to_cfqq(struct cfq_io_context *cic,
423
					    bool is_sync)
424
{
425
	return cic->cfqq[is_sync];
426 427 428
}

static inline void cic_set_cfqq(struct cfq_io_context *cic,
429
				struct cfq_queue *cfqq, bool is_sync)
430
{
431
	cic->cfqq[is_sync] = cfqq;
432 433 434 435 436 437
}

/*
 * 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).
 */
438
static inline bool cfq_bio_sync(struct bio *bio)
439
{
440
	return bio_data_dir(bio) == READ || bio_rw_flagged(bio, BIO_RW_SYNCIO);
441
}
Linus Torvalds's avatar
Linus Torvalds committed
442

Andrew Morton's avatar
Andrew Morton committed
443 444 445 446
/*
 * scheduler run of queue, if there are requests pending and no one in the
 * driver that will restart queueing
 */
447
static inline void cfq_schedule_dispatch(struct cfq_data *cfqd)
Andrew Morton's avatar
Andrew Morton committed
448
{
449 450
	if (cfqd->busy_queues) {
		cfq_log(cfqd, "schedule dispatch");
451
		kblockd_schedule_work(cfqd->queue, &cfqd->unplug_work);
452
	}
Andrew Morton's avatar
Andrew Morton committed
453 454
}

455
static int cfq_queue_empty(struct request_queue *q)
Andrew Morton's avatar
Andrew Morton committed
456 457 458
{
	struct cfq_data *cfqd = q->elevator->elevator_data;

459
	return !cfqd->rq_queued;
Andrew Morton's avatar
Andrew Morton committed
460 461
}

462 463 464 465 466
/*
 * 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.
 */
467
static inline int cfq_prio_slice(struct cfq_data *cfqd, bool sync,
468
				 unsigned short prio)
469
{
470
	const int base_slice = cfqd->cfq_slice[sync];
471

472 473 474 475
	WARN_ON(prio >= IOPRIO_BE_NR);

	return base_slice + (base_slice/CFQ_SLICE_SCALE * (4 - prio));
}
476

477 478 479 480
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);
481 482
}

483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527
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);
}

528 529 530 531 532 533
/*
 * 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
 */

534 535
static inline unsigned cfq_group_get_avg_queues(struct cfq_data *cfqd,
					struct cfq_group *cfqg, bool rt)
536
{
537 538 539
	unsigned min_q, max_q;
	unsigned mult  = cfq_hist_divisor - 1;
	unsigned round = cfq_hist_divisor / 2;
540
	unsigned busy = cfq_group_busy_queues_wl(rt, cfqd, cfqg);
541

542 543 544
	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) /
545
		cfq_hist_divisor;
546 547 548 549 550 551 552 553 554
	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;
555 556
}

557 558 559
static inline void
cfq_set_prio_slice(struct cfq_data *cfqd, struct cfq_queue *cfqq)
{
560 561
	unsigned slice = cfq_prio_to_slice(cfqd, cfqq);
	if (cfqd->cfq_latency) {
562 563 564 565 566 567
		/*
		 * 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));
568 569
		unsigned sync_slice = cfqd->cfq_slice[1];
		unsigned expect_latency = sync_slice * iq;
570 571 572
		unsigned group_slice = cfq_group_slice(cfqd, cfqq->cfqg);

		if (expect_latency > group_slice) {
573 574 575 576 577 578 579
			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 */
580
			slice = max(slice * group_slice / expect_latency,
581 582 583
				    low_slice);
		}
	}
584
	cfqq->slice_start = jiffies;
585
	cfqq->slice_end = jiffies + slice;
586
	cfqq->allocated_slice = slice;
587
	cfq_log_cfqq(cfqd, cfqq, "set_slice=%lu", cfqq->slice_end - jiffies);
588 589 590 591 592 593 594
}

/*
 * 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.
 */
595
static inline bool cfq_slice_used(struct cfq_queue *cfqq)
596 597 598 599 600 601 602 603 604
{
	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
605
/*
Jens Axboe's avatar
Jens Axboe committed
606
 * Lifted from AS - choose which of rq1 and rq2 that is best served now.
Linus Torvalds's avatar
Linus Torvalds committed
607
 * We choose the request that is closest to the head right now. Distance
608
 * behind the head is penalized and only allowed to a certain extent.
Linus Torvalds's avatar
Linus Torvalds committed
609
 */
Jens Axboe's avatar
Jens Axboe committed
610
static struct request *
611
cfq_choose_req(struct cfq_data *cfqd, struct request *rq1, struct request *rq2, sector_t last)
Linus Torvalds's avatar
Linus Torvalds committed
612
{
613
	sector_t s1, s2, d1 = 0, d2 = 0;
Linus Torvalds's avatar
Linus Torvalds committed
614
	unsigned long back_max;
615 616 617
#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
618

Jens Axboe's avatar
Jens Axboe committed
619 620 621 622
	if (rq1 == NULL || rq1 == rq2)
		return rq2;
	if (rq2 == NULL)
		return rq1;
623

Jens Axboe's avatar
Jens Axboe committed
624 625 626 627
	if (rq_is_sync(rq1) && !rq_is_sync(rq2))
		return rq1;
	else if (rq_is_sync(rq2) && !rq_is_sync(rq1))
		return rq2;
628 629 630 631
	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
632

633 634
	s1 = blk_rq_pos(rq1);
	s2 = blk_rq_pos(rq2);
Linus Torvalds's avatar
Linus Torvalds committed
635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650

	/*
	 * 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
651
		wrap |= CFQ_RQ1_WRAP;
Linus Torvalds's avatar
Linus Torvalds committed
652 653 654 655 656 657

	if (s2 >= last)
		d2 = s2 - last;
	else if (s2 + back_max >= last)
		d2 = (last - s2) * cfqd->cfq_back_penalty;
	else
658
		wrap |= CFQ_RQ2_WRAP;
Linus Torvalds's avatar
Linus Torvalds committed
659 660

	/* Found required data */
661 662 663 664 665 666

	/*
	 * 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
667
	case 0: /* common case for CFQ: rq1 and rq2 not wrapped */
668
		if (d1 < d2)
Jens Axboe's avatar
Jens Axboe committed
669
			return rq1;
670
		else if (d2 < d1)
Jens Axboe's avatar
Jens Axboe committed
671
			return rq2;
672 673
		else {
			if (s1 >= s2)
Jens Axboe's avatar
Jens Axboe committed
674
				return rq1;
675
			else
Jens Axboe's avatar
Jens Axboe committed
676
				return rq2;
677
		}
Linus Torvalds's avatar
Linus Torvalds committed
678

679
	case CFQ_RQ2_WRAP:
Jens Axboe's avatar
Jens Axboe committed
680
		return rq1;
681
	case CFQ_RQ1_WRAP:
Jens Axboe's avatar
Jens Axboe committed
682 683
		return rq2;
	case (CFQ_RQ1_WRAP|CFQ_RQ2_WRAP): /* both rqs wrapped */
684 685 686 687 688 689 690 691
	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
692
			return rq1;
Linus Torvalds's avatar
Linus Torvalds committed
693
		else
Jens Axboe's avatar
Jens Axboe committed
694
			return rq2;
Linus Torvalds's avatar
Linus Torvalds committed
695 696 697
	}
}

698 699 700
/*
 * The below is leftmost cache rbtree addon
 */
701
static struct cfq_queue *cfq_rb_first(struct cfq_rb_root *root)
702
{
703 704 705 706
	/* Service tree is empty */
	if (!root->count)
		return NULL;

707 708 709
	if (!root->left)
		root->left = rb_first(&root->rb);

710 711 712 713
	if (root->left)
		return rb_entry(root->left, struct cfq_queue, rb_node);

	return NULL;
714 715
}

716 717 718 719 720 721 722 723 724 725 726
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;
}

727 728 729 730 731 732
static void rb_erase_init(struct rb_node *n, struct rb_root *root)
{
	rb_erase(n, root);
	RB_CLEAR_NODE(n);
}

733 734 735 736
static void cfq_rb_erase(struct rb_node *n, struct cfq_rb_root *root)
{
	if (root->left == n)
		root->left = NULL;
737
	rb_erase_init(n, &root->rb);
738
	--root->count;
739 740
}

Linus Torvalds's avatar
Linus Torvalds committed
741 742 743
/*
 * would be nice to take fifo expire time into account as well
 */
Jens Axboe's avatar
Jens Axboe committed
744 745 746
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
747
{
748 749
	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
750
	struct request *next = NULL, *prev = NULL;
Linus Torvalds's avatar
Linus Torvalds committed
751

752
	BUG_ON(RB_EMPTY_NODE(&last->rb_node));
Linus Torvalds's avatar
Linus Torvalds committed
753 754

	if (rbprev)
Jens Axboe's avatar
Jens Axboe committed
755
		prev = rb_entry_rq(rbprev);
Linus Torvalds's avatar
Linus Torvalds committed
756

757
	if (rbnext)
Jens Axboe's avatar
Jens Axboe committed
758
		next = rb_entry_rq(rbnext);
759 760 761
	else {
		rbnext = rb_first(&cfqq->sort_list);
		if (rbnext && rbnext != &last->rb_node)
Jens Axboe's avatar
Jens Axboe committed
762
			next = rb_entry_rq(rbnext);
763
	}
Linus Torvalds's avatar
Linus Torvalds committed
764

765
	return cfq_choose_req(cfqd, next, prev, blk_rq_pos(last));
Linus Torvalds's avatar
Linus Torvalds committed
766 767
}

768 769
static unsigned long cfq_slice_offset(struct cfq_data *cfqd,
				      struct cfq_queue *cfqq)
Linus Torvalds's avatar
Linus Torvalds committed
770
{
771 772 773
	/*
	 * just an approximation, should be ok.
	 */
774
	return (cfqq->cfqg->nr_cfqq - 1) * (cfq_prio_slice(cfqd, 1, 0) -
775
		       cfq_prio_slice(cfqd, cfq_cfqq_sync(cfqq), cfqq->ioprio));
776 777
}

778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836
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;
837 838
	cfqd->nr_groups++;
	st->total_weight += cfqg->weight;
839 840 841 842 843 844 845
}

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

846 847 848
	if (st->active == &cfqg->rb_node)
		st->active = NULL;

849 850
	BUG_ON(cfqg->nr_cfqq < 1);
	cfqg->nr_cfqq--;
851

852 853 854 855
	/* If there are other cfq queues under this group, don't delete it */
	if (cfqg->nr_cfqq)
		return;

856
	cfq_log_cfqg(cfqd, cfqg, "del_from_rr group");
857
	cfqg->on_st = false;
858 859
	cfqd->nr_groups--;
	st->total_weight -= cfqg->weight;
860 861
	if (!RB_EMPTY_NODE(&cfqg->rb_node))
		cfq_rb_erase(&cfqg->rb_node, st);
862
	cfqg->saved_workload_slice = 0;
863
	blkiocg_update_blkio_group_dequeue_stats(&cfqg->blkg, 1);
864 865 866 867
}

static inline unsigned int cfq_cfqq_slice_usage(struct cfq_queue *cfqq)
{
868
	unsigned int slice_used;
869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884

	/*
	 * 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;
885 886
		if (slice_used > cfqq->allocated_slice)
			slice_used = cfqq->allocated_slice;
887 888
	}

889 890
	cfq_log_cfqq(cfqq->cfqd, cfqq, "sl_used=%u sect=%lu", slice_used,
				cfqq->nr_sectors);
891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914
	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;
915 916 917

	cfq_log_cfqg(cfqd, cfqg, "served: vt=%llu min_vt=%llu", cfqg->vdisktime,
					st->min_vdisktime);
918 919
	blkiocg_update_blkio_group_stats(&cfqg->blkg, used_sl,
						cfqq->nr_sectors);
920 921
}

922 923 924 925 926 927 928 929
#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;
}

930 931 932 933 934 935
void
cfq_update_blkio_group_weight(struct blkio_group *blkg, unsigned int weight)
{
	cfqg_of_blkg(blkg)->weight = weight;
}

936 937 938 939 940 941 942 943
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;
944 945
	struct backing_dev_info *bdi = &cfqd->queue->backing_dev_info;
	unsigned int major, minor;
946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963

	/* 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);

964 965 966 967 968 969 970 971
	/*
	 * 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);

972
	/* Add group onto cgroup list */
973 974 975
	sscanf(dev_name(bdi->dev), "%u:%u", &major, &minor);
	blkiocg_add_blkio_group(blkcg, &cfqg->blkg, (void *)cfqd,
					MKDEV(major, minor));
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 1005 1006 1007 1008 1009

	/* 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;
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 1050 1051 1052 1053 1054
	/* 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);
	}
1055
}
1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080

/*
 * 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);
}

1081 1082 1083 1084 1085 1086 1087 1088 1089 1090
#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;
}

1091 1092 1093
static void cfq_release_cfq_groups(struct cfq_data *cfqd) {}
static inline void cfq_put_cfqg(struct cfq_group *cfqg) {}

1094 1095
#endif /* GROUP_IOSCHED */

1096
/*
1097
 * The cfqd->service_trees holds all pending cfq_queue's that have
1098 1099 1100
 * requests waiting to be processed. It is sorted in the order that
 * we will service the queues.
 */
1101
static void cfq_service_tree_add(struct cfq_data *cfqd, struct cfq_queue *cfqq,
1102
				 bool add_front)
1103
{
1104 1105
	struct rb_node **p, *parent;
	struct cfq_queue *__cfqq;
1106
	unsigned long rb_key;
1107
	struct cfq_rb_root *service_tree;
1108
	int left;
1109
	int new_cfqq = 1;
1110

1111 1112
	service_tree = service_tree_for(cfqq->cfqg, cfqq_prio(cfqq),
						cfqq_type(cfqq), cfqd);
1113 1114
	if (cfq_class_idle(cfqq)) {
		rb_key = CFQ_IDLE_DELAY;
1115
		parent = rb_last(&service_tree->rb);
1116 1117 1118 1119 1120 1121
		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) {
1122 1123 1124 1125 1126 1127
		/*
		 * 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.
		 */
1128
		rb_key = cfq_slice_offset(cfqd, cfqq) + jiffies;
1129
		rb_key -= cfqq->slice_resid;
1130
		cfqq->slice_resid = 0;
1131 1132
	} else {
		rb_key = -HZ;
1133
		__cfqq = cfq_rb_first(service_tree);
1134 1135
		rb_key += __cfqq ? __cfqq->rb_key : jiffies;
	}
Linus Torvalds's avatar
Linus Torvalds committed
1136

1137
	if (!RB_EMPTY_NODE(&cfqq->rb_node)) {
1138
		new_cfqq = 0;
1139
		/*
1140
		 * same position, nothing more to do
1141
		 */
1142 1143
		if (rb_key == cfqq->rb_key &&
		    cfqq->service_tree == service_tree)
1144
			return;
Linus Torvalds's avatar
Linus Torvalds committed
1145

1146 1147
		cfq_rb_erase