cfq-iosched.c 56 KB
Newer Older
Linus Torvalds's avatar
Linus Torvalds committed
1 2 3 4 5 6 7 8 9 10
/*
 *  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.
 *
 *  Copyright (C) 2003 Jens Axboe <axboe@suse.de>
 */
#include <linux/config.h>
#include <linux/module.h>
Al Viro's avatar
Al Viro committed
11 12
#include <linux/blkdev.h>
#include <linux/elevator.h>
Linus Torvalds's avatar
Linus Torvalds committed
13 14
#include <linux/hash.h>
#include <linux/rbtree.h>
15
#include <linux/ioprio.h>
Linus Torvalds's avatar
Linus Torvalds committed
16 17 18 19

/*
 * tunables
 */
20 21 22 23 24
static const int cfq_quantum = 4;		/* max queue in one round of service */
static const int cfq_queued = 8;		/* minimum rq allocate limit per-queue*/
static const int cfq_fifo_expire[2] = { HZ / 4, HZ / 8 };
static const int cfq_back_max = 16 * 1024;	/* maximum backwards seek, in KiB */
static const int cfq_back_penalty = 2;		/* penalty of a backwards seek */
Linus Torvalds's avatar
Linus Torvalds committed
25

26
static const int cfq_slice_sync = HZ / 10;
Jens Axboe's avatar
Jens Axboe committed
27
static int cfq_slice_async = HZ / 25;
28
static const int cfq_slice_async_rq = 2;
Jens Axboe's avatar
Jens Axboe committed
29
static int cfq_slice_idle = HZ / 100;
30 31 32 33 34

#define CFQ_IDLE_GRACE		(HZ / 10)
#define CFQ_SLICE_SCALE		(5)

#define CFQ_KEY_ASYNC		(0)
Jens Axboe's avatar
Jens Axboe committed
35
#define CFQ_KEY_ANY		(0xffff)
36 37 38 39

/*
 * disable queueing at the driver/hardware level
 */
40
static const int cfq_max_depth = 2;
41

42 43
static DEFINE_RWLOCK(cfq_exit_lock);

Linus Torvalds's avatar
Linus Torvalds committed
44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61
/*
 * for the hash of cfqq inside the cfqd
 */
#define CFQ_QHASH_SHIFT		6
#define CFQ_QHASH_ENTRIES	(1 << CFQ_QHASH_SHIFT)
#define list_entry_qhash(entry)	hlist_entry((entry), struct cfq_queue, cfq_hash)

/*
 * for the hash of crq inside the cfqq
 */
#define CFQ_MHASH_SHIFT		6
#define CFQ_MHASH_BLOCK(sec)	((sec) >> 3)
#define CFQ_MHASH_ENTRIES	(1 << CFQ_MHASH_SHIFT)
#define CFQ_MHASH_FN(sec)	hash_long(CFQ_MHASH_BLOCK(sec), CFQ_MHASH_SHIFT)
#define rq_hash_key(rq)		((rq)->sector + (rq)->nr_sectors)
#define list_entry_hash(ptr)	hlist_entry((ptr), struct cfq_rq, hash)

#define list_entry_cfqq(ptr)	list_entry((ptr), struct cfq_queue, cfq_list)
62
#define list_entry_fifo(ptr)	list_entry((ptr), struct request, queuelist)
Linus Torvalds's avatar
Linus Torvalds committed
63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85

#define RQ_DATA(rq)		(rq)->elevator_private

/*
 * rb-tree defines
 */
#define RB_NONE			(2)
#define RB_EMPTY(node)		((node)->rb_node == NULL)
#define RB_CLEAR_COLOR(node)	(node)->rb_color = RB_NONE
#define RB_CLEAR(node)		do {	\
	(node)->rb_parent = NULL;	\
	RB_CLEAR_COLOR((node));		\
	(node)->rb_right = NULL;	\
	(node)->rb_left = NULL;		\
} while (0)
#define RB_CLEAR_ROOT(root)	((root)->rb_node = NULL)
#define rb_entry_crq(node)	rb_entry((node), struct cfq_rq, rb_node)
#define rq_rb_key(rq)		(rq)->sector

static kmem_cache_t *crq_pool;
static kmem_cache_t *cfq_pool;
static kmem_cache_t *cfq_ioc_pool;

86 87 88
static atomic_t ioc_count = ATOMIC_INIT(0);
static struct completion *ioc_gone;

89 90 91 92 93
#define CFQ_PRIO_LISTS		IOPRIO_BE_NR
#define cfq_class_idle(cfqq)	((cfqq)->ioprio_class == IOPRIO_CLASS_IDLE)
#define cfq_class_be(cfqq)	((cfqq)->ioprio_class == IOPRIO_CLASS_BE)
#define cfq_class_rt(cfqq)	((cfqq)->ioprio_class == IOPRIO_CLASS_RT)

Jens Axboe's avatar
Jens Axboe committed
94 95 96 97 98 99 100 101 102 103
#define ASYNC			(0)
#define SYNC			(1)

#define cfq_cfqq_dispatched(cfqq)	\
	((cfqq)->on_dispatch[ASYNC] + (cfqq)->on_dispatch[SYNC])

#define cfq_cfqq_class_sync(cfqq)	((cfqq)->key != CFQ_KEY_ASYNC)

#define cfq_cfqq_sync(cfqq)		\
	(cfq_cfqq_class_sync(cfqq) || (cfqq)->on_dispatch[SYNC])
104 105 106 107

/*
 * Per block device queue structure
 */
Linus Torvalds's avatar
Linus Torvalds committed
108
struct cfq_data {
109 110 111 112 113 114 115 116 117 118 119 120 121 122
	request_queue_t *queue;

	/*
	 * rr list of queues with requests and the count of them
	 */
	struct list_head rr_list[CFQ_PRIO_LISTS];
	struct list_head busy_rr;
	struct list_head cur_rr;
	struct list_head idle_rr;
	unsigned int busy_queues;

	/*
	 * non-ordered list of empty cfqq's
	 */
Linus Torvalds's avatar
Linus Torvalds committed
123 124
	struct list_head empty_list;

125 126 127
	/*
	 * cfqq lookup hash
	 */
Linus Torvalds's avatar
Linus Torvalds committed
128 129
	struct hlist_head *cfq_hash;

130 131 132 133
	/*
	 * global crq hash for all queues
	 */
	struct hlist_head *crq_hash;
Linus Torvalds's avatar
Linus Torvalds committed
134 135 136

	unsigned int max_queued;

137
	mempool_t *crq_pool;
Linus Torvalds's avatar
Linus Torvalds committed
138

139
	int rq_in_driver;
Linus Torvalds's avatar
Linus Torvalds committed
140

141 142 143 144 145 146 147 148
	/*
	 * schedule slice state info
	 */
	/*
	 * idle window management
	 */
	struct timer_list idle_slice_timer;
	struct work_struct unplug_work;
Linus Torvalds's avatar
Linus Torvalds committed
149

150 151 152 153 154 155
	struct cfq_queue *active_queue;
	struct cfq_io_context *active_cic;
	int cur_prio, cur_end_prio;
	unsigned int dispatch_slice;

	struct timer_list idle_class_timer;
Linus Torvalds's avatar
Linus Torvalds committed
156 157

	sector_t last_sector;
158
	unsigned long last_end_request;
Linus Torvalds's avatar
Linus Torvalds committed
159

160
	unsigned int rq_starved;
Linus Torvalds's avatar
Linus Torvalds committed
161 162 163 164 165 166

	/*
	 * tunables, see top of file
	 */
	unsigned int cfq_quantum;
	unsigned int cfq_queued;
167
	unsigned int cfq_fifo_expire[2];
Linus Torvalds's avatar
Linus Torvalds committed
168 169
	unsigned int cfq_back_penalty;
	unsigned int cfq_back_max;
170 171 172 173
	unsigned int cfq_slice[2];
	unsigned int cfq_slice_async_rq;
	unsigned int cfq_slice_idle;
	unsigned int cfq_max_depth;
174 175

	struct list_head cic_list;
Linus Torvalds's avatar
Linus Torvalds committed
176 177
};

178 179 180
/*
 * Per process-grouping structure
 */
Linus Torvalds's avatar
Linus Torvalds committed
181 182 183 184 185
struct cfq_queue {
	/* reference count */
	atomic_t ref;
	/* parent cfq_data */
	struct cfq_data *cfqd;
186
	/* cfqq lookup hash */
Linus Torvalds's avatar
Linus Torvalds committed
187 188
	struct hlist_node cfq_hash;
	/* hash key */
189
	unsigned int key;
Linus Torvalds's avatar
Linus Torvalds committed
190 191 192 193 194 195 196 197 198 199 200
	/* on either rr or empty list of cfqd */
	struct list_head cfq_list;
	/* sorted list of pending requests */
	struct rb_root sort_list;
	/* if fifo isn't expired, next request to serve */
	struct cfq_rq *next_crq;
	/* requests queued in sort_list */
	int queued[2];
	/* currently allocated requests */
	int allocated[2];
	/* fifo list of requests in sort_list */
201
	struct list_head fifo;
Linus Torvalds's avatar
Linus Torvalds committed
202

203 204 205 206
	unsigned long slice_start;
	unsigned long slice_end;
	unsigned long slice_left;
	unsigned long service_last;
Linus Torvalds's avatar
Linus Torvalds committed
207

Jens Axboe's avatar
Jens Axboe committed
208 209
	/* number of requests that are on the dispatch list */
	int on_dispatch[2];
210 211 212 213 214

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

Jens Axboe's avatar
Jens Axboe committed
215 216
	/* various state flags, see below */
	unsigned int flags;
Linus Torvalds's avatar
Linus Torvalds committed
217 218 219 220 221 222 223 224 225 226 227
};

struct cfq_rq {
	struct rb_node rb_node;
	sector_t rb_key;
	struct request *request;
	struct hlist_node hash;

	struct cfq_queue *cfq_queue;
	struct cfq_io_context *io_context;

Jens Axboe's avatar
Jens Axboe committed
228
	unsigned int crq_flags;
Linus Torvalds's avatar
Linus Torvalds committed
229 230
};

Jens Axboe's avatar
Jens Axboe committed
231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266
enum cfqq_state_flags {
	CFQ_CFQQ_FLAG_on_rr = 0,
	CFQ_CFQQ_FLAG_wait_request,
	CFQ_CFQQ_FLAG_must_alloc,
	CFQ_CFQQ_FLAG_must_alloc_slice,
	CFQ_CFQQ_FLAG_must_dispatch,
	CFQ_CFQQ_FLAG_fifo_expire,
	CFQ_CFQQ_FLAG_idle_window,
	CFQ_CFQQ_FLAG_prio_changed,
};

#define CFQ_CFQQ_FNS(name)						\
static inline void cfq_mark_cfqq_##name(struct cfq_queue *cfqq)		\
{									\
	cfqq->flags |= (1 << CFQ_CFQQ_FLAG_##name);			\
}									\
static inline void cfq_clear_cfqq_##name(struct cfq_queue *cfqq)	\
{									\
	cfqq->flags &= ~(1 << CFQ_CFQQ_FLAG_##name);			\
}									\
static inline int cfq_cfqq_##name(const struct cfq_queue *cfqq)		\
{									\
	return (cfqq->flags & (1 << CFQ_CFQQ_FLAG_##name)) != 0;	\
}

CFQ_CFQQ_FNS(on_rr);
CFQ_CFQQ_FNS(wait_request);
CFQ_CFQQ_FNS(must_alloc);
CFQ_CFQQ_FNS(must_alloc_slice);
CFQ_CFQQ_FNS(must_dispatch);
CFQ_CFQQ_FNS(fifo_expire);
CFQ_CFQQ_FNS(idle_window);
CFQ_CFQQ_FNS(prio_changed);
#undef CFQ_CFQQ_FNS

enum cfq_rq_state_flags {
267
	CFQ_CRQ_FLAG_is_sync = 0,
Jens Axboe's avatar
Jens Axboe committed
268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287
};

#define CFQ_CRQ_FNS(name)						\
static inline void cfq_mark_crq_##name(struct cfq_rq *crq)		\
{									\
	crq->crq_flags |= (1 << CFQ_CRQ_FLAG_##name);			\
}									\
static inline void cfq_clear_crq_##name(struct cfq_rq *crq)		\
{									\
	crq->crq_flags &= ~(1 << CFQ_CRQ_FLAG_##name);			\
}									\
static inline int cfq_crq_##name(const struct cfq_rq *crq)		\
{									\
	return (crq->crq_flags & (1 << CFQ_CRQ_FLAG_##name)) != 0;	\
}

CFQ_CRQ_FNS(is_sync);
#undef CFQ_CRQ_FNS

static struct cfq_queue *cfq_find_cfq_hash(struct cfq_data *, unsigned int, unsigned short);
288
static void cfq_dispatch_insert(request_queue_t *, struct cfq_rq *);
289
static struct cfq_queue *cfq_get_queue(struct cfq_data *cfqd, unsigned int key, struct task_struct *tsk, gfp_t gfp_mask);
Linus Torvalds's avatar
Linus Torvalds committed
290

291
#define process_sync(tsk)	((tsk)->flags & PF_SYNCWRITE)
Linus Torvalds's avatar
Linus Torvalds committed
292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328

/*
 * lots of deadline iosched dupes, can be abstracted later...
 */
static inline void cfq_del_crq_hash(struct cfq_rq *crq)
{
	hlist_del_init(&crq->hash);
}

static inline void cfq_add_crq_hash(struct cfq_data *cfqd, struct cfq_rq *crq)
{
	const int hash_idx = CFQ_MHASH_FN(rq_hash_key(crq->request));

	hlist_add_head(&crq->hash, &cfqd->crq_hash[hash_idx]);
}

static struct request *cfq_find_rq_hash(struct cfq_data *cfqd, sector_t offset)
{
	struct hlist_head *hash_list = &cfqd->crq_hash[CFQ_MHASH_FN(offset)];
	struct hlist_node *entry, *next;

	hlist_for_each_safe(entry, next, hash_list) {
		struct cfq_rq *crq = list_entry_hash(entry);
		struct request *__rq = crq->request;

		if (!rq_mergeable(__rq)) {
			cfq_del_crq_hash(crq);
			continue;
		}

		if (rq_hash_key(__rq) == offset)
			return __rq;
	}

	return NULL;
}

Andrew Morton's avatar
Andrew Morton committed
329 330 331 332 333 334
/*
 * scheduler run of queue, if there are requests pending and no one in the
 * driver that will restart queueing
 */
static inline void cfq_schedule_dispatch(struct cfq_data *cfqd)
{
335
	if (cfqd->busy_queues)
Andrew Morton's avatar
Andrew Morton committed
336 337 338 339 340 341 342
		kblockd_schedule_work(&cfqd->unplug_work);
}

static int cfq_queue_empty(request_queue_t *q)
{
	struct cfq_data *cfqd = q->elevator->elevator_data;

343
	return !cfqd->busy_queues;
Andrew Morton's avatar
Andrew Morton committed
344 345
}

Linus Torvalds's avatar
Linus Torvalds committed
346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361
/*
 * Lifted from AS - choose which of crq1 and crq2 that is best served now.
 * We choose the request that is closest to the head right now. Distance
 * behind the head are penalized and only allowed to a certain extent.
 */
static struct cfq_rq *
cfq_choose_req(struct cfq_data *cfqd, struct cfq_rq *crq1, struct cfq_rq *crq2)
{
	sector_t last, s1, s2, d1 = 0, d2 = 0;
	int r1_wrap = 0, r2_wrap = 0;	/* requests are behind the disk head */
	unsigned long back_max;

	if (crq1 == NULL || crq1 == crq2)
		return crq2;
	if (crq2 == NULL)
		return crq1;
362 363 364 365

	if (cfq_crq_is_sync(crq1) && !cfq_crq_is_sync(crq2))
		return crq1;
	else if (cfq_crq_is_sync(crq2) && !cfq_crq_is_sync(crq1))
366
		return crq2;
Linus Torvalds's avatar
Linus Torvalds committed
367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432

	s1 = crq1->request->sector;
	s2 = crq2->request->sector;

	last = cfqd->last_sector;

	/*
	 * 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
		r1_wrap = 1;

	if (s2 >= last)
		d2 = s2 - last;
	else if (s2 + back_max >= last)
		d2 = (last - s2) * cfqd->cfq_back_penalty;
	else
		r2_wrap = 1;

	/* Found required data */
	if (!r1_wrap && r2_wrap)
		return crq1;
	else if (!r2_wrap && r1_wrap)
		return crq2;
	else if (r1_wrap && r2_wrap) {
		/* both behind the head */
		if (s1 <= s2)
			return crq1;
		else
			return crq2;
	}

	/* Both requests in front of the head */
	if (d1 < d2)
		return crq1;
	else if (d2 < d1)
		return crq2;
	else {
		if (s1 >= s2)
			return crq1;
		else
			return crq2;
	}
}

/*
 * would be nice to take fifo expire time into account as well
 */
static struct cfq_rq *
cfq_find_next_crq(struct cfq_data *cfqd, struct cfq_queue *cfqq,
		  struct cfq_rq *last)
{
	struct cfq_rq *crq_next = NULL, *crq_prev = NULL;
	struct rb_node *rbnext, *rbprev;

433
	if (!(rbnext = rb_next(&last->rb_node))) {
Linus Torvalds's avatar
Linus Torvalds committed
434
		rbnext = rb_first(&cfqq->sort_list);
435 436 437
		if (rbnext == &last->rb_node)
			rbnext = NULL;
	}
Linus Torvalds's avatar
Linus Torvalds committed
438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456

	rbprev = rb_prev(&last->rb_node);

	if (rbprev)
		crq_prev = rb_entry_crq(rbprev);
	if (rbnext)
		crq_next = rb_entry_crq(rbnext);

	return cfq_choose_req(cfqd, crq_next, crq_prev);
}

static void cfq_update_next_crq(struct cfq_rq *crq)
{
	struct cfq_queue *cfqq = crq->cfq_queue;

	if (cfqq->next_crq == crq)
		cfqq->next_crq = cfq_find_next_crq(cfqq->cfqd, cfqq, crq);
}

457
static void cfq_resort_rr_list(struct cfq_queue *cfqq, int preempted)
Linus Torvalds's avatar
Linus Torvalds committed
458
{
459 460
	struct cfq_data *cfqd = cfqq->cfqd;
	struct list_head *list, *entry;
Linus Torvalds's avatar
Linus Torvalds committed
461

Jens Axboe's avatar
Jens Axboe committed
462
	BUG_ON(!cfq_cfqq_on_rr(cfqq));
Linus Torvalds's avatar
Linus Torvalds committed
463

464
	list_del(&cfqq->cfq_list);
Linus Torvalds's avatar
Linus Torvalds committed
465

466 467 468 469 470 471 472 473 474 475 476 477
	if (cfq_class_rt(cfqq))
		list = &cfqd->cur_rr;
	else if (cfq_class_idle(cfqq))
		list = &cfqd->idle_rr;
	else {
		/*
		 * if cfqq has requests in flight, don't allow it to be
		 * found in cfq_set_active_queue before it has finished them.
		 * this is done to increase fairness between a process that
		 * has lots of io pending vs one that only generates one
		 * sporadically or synchronously
		 */
Jens Axboe's avatar
Jens Axboe committed
478
		if (cfq_cfqq_dispatched(cfqq))
479 480 481
			list = &cfqd->busy_rr;
		else
			list = &cfqd->rr_list[cfqq->ioprio];
Linus Torvalds's avatar
Linus Torvalds committed
482 483
	}

484 485 486 487 488 489
	/*
	 * if queue was preempted, just add to front to be fair. busy_rr
	 * isn't sorted.
	 */
	if (preempted || list == &cfqd->busy_rr) {
		list_add(&cfqq->cfq_list, list);
Linus Torvalds's avatar
Linus Torvalds committed
490
		return;
491
	}
Linus Torvalds's avatar
Linus Torvalds committed
492 493

	/*
494
	 * sort by when queue was last serviced
Linus Torvalds's avatar
Linus Torvalds committed
495
	 */
496 497
	entry = list;
	while ((entry = entry->prev) != list) {
Linus Torvalds's avatar
Linus Torvalds committed
498 499
		struct cfq_queue *__cfqq = list_entry_cfqq(entry);

500 501 502
		if (!__cfqq->service_last)
			break;
		if (time_before(__cfqq->service_last, cfqq->service_last))
Linus Torvalds's avatar
Linus Torvalds committed
503 504 505 506 507 508 509 510
			break;
	}

	list_add(&cfqq->cfq_list, entry);
}

/*
 * add to busy list of queues for service, trying to be fair in ordering
511
 * the pending list according to last request service
Linus Torvalds's avatar
Linus Torvalds committed
512 513
 */
static inline void
514
cfq_add_cfqq_rr(struct cfq_data *cfqd, struct cfq_queue *cfqq)
Linus Torvalds's avatar
Linus Torvalds committed
515
{
Jens Axboe's avatar
Jens Axboe committed
516 517
	BUG_ON(cfq_cfqq_on_rr(cfqq));
	cfq_mark_cfqq_on_rr(cfqq);
Linus Torvalds's avatar
Linus Torvalds committed
518 519
	cfqd->busy_queues++;

520
	cfq_resort_rr_list(cfqq, 0);
Linus Torvalds's avatar
Linus Torvalds committed
521 522 523 524 525
}

static inline void
cfq_del_cfqq_rr(struct cfq_data *cfqd, struct cfq_queue *cfqq)
{
Jens Axboe's avatar
Jens Axboe committed
526 527
	BUG_ON(!cfq_cfqq_on_rr(cfqq));
	cfq_clear_cfqq_on_rr(cfqq);
528
	list_move(&cfqq->cfq_list, &cfqd->empty_list);
Linus Torvalds's avatar
Linus Torvalds committed
529 530 531 532 533 534 535 536 537 538 539

	BUG_ON(!cfqd->busy_queues);
	cfqd->busy_queues--;
}

/*
 * rb tree support functions
 */
static inline void cfq_del_crq_rb(struct cfq_rq *crq)
{
	struct cfq_queue *cfqq = crq->cfq_queue;
540 541
	struct cfq_data *cfqd = cfqq->cfqd;
	const int sync = cfq_crq_is_sync(crq);
Linus Torvalds's avatar
Linus Torvalds committed
542

543 544
	BUG_ON(!cfqq->queued[sync]);
	cfqq->queued[sync]--;
Linus Torvalds's avatar
Linus Torvalds committed
545

546
	cfq_update_next_crq(crq);
Linus Torvalds's avatar
Linus Torvalds committed
547

548 549
	rb_erase(&crq->rb_node, &cfqq->sort_list);
	RB_CLEAR_COLOR(&crq->rb_node);
Linus Torvalds's avatar
Linus Torvalds committed
550

551 552
	if (cfq_cfqq_on_rr(cfqq) && RB_EMPTY(&cfqq->sort_list))
		cfq_del_cfqq_rr(cfqd, cfqq);
Linus Torvalds's avatar
Linus Torvalds committed
553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585
}

static struct cfq_rq *
__cfq_add_crq_rb(struct cfq_rq *crq)
{
	struct rb_node **p = &crq->cfq_queue->sort_list.rb_node;
	struct rb_node *parent = NULL;
	struct cfq_rq *__crq;

	while (*p) {
		parent = *p;
		__crq = rb_entry_crq(parent);

		if (crq->rb_key < __crq->rb_key)
			p = &(*p)->rb_left;
		else if (crq->rb_key > __crq->rb_key)
			p = &(*p)->rb_right;
		else
			return __crq;
	}

	rb_link_node(&crq->rb_node, parent, p);
	return NULL;
}

static void cfq_add_crq_rb(struct cfq_rq *crq)
{
	struct cfq_queue *cfqq = crq->cfq_queue;
	struct cfq_data *cfqd = cfqq->cfqd;
	struct request *rq = crq->request;
	struct cfq_rq *__alias;

	crq->rb_key = rq_rb_key(rq);
Jens Axboe's avatar
Jens Axboe committed
586
	cfqq->queued[cfq_crq_is_sync(crq)]++;
Linus Torvalds's avatar
Linus Torvalds committed
587 588 589 590 591 592

	/*
	 * looks a little odd, but the first insert might return an alias.
	 * if that happens, put the alias on the dispatch list
	 */
	while ((__alias = __cfq_add_crq_rb(crq)) != NULL)
593
		cfq_dispatch_insert(cfqd->queue, __alias);
Linus Torvalds's avatar
Linus Torvalds committed
594 595 596

	rb_insert_color(&crq->rb_node, &cfqq->sort_list);

Jens Axboe's avatar
Jens Axboe committed
597
	if (!cfq_cfqq_on_rr(cfqq))
598
		cfq_add_cfqq_rr(cfqd, cfqq);
Linus Torvalds's avatar
Linus Torvalds committed
599 600 601 602 603 604 605 606 607 608

	/*
	 * check if this request is a better next-serve candidate
	 */
	cfqq->next_crq = cfq_choose_req(cfqd, cfqq->next_crq, crq);
}

static inline void
cfq_reposition_crq_rb(struct cfq_queue *cfqq, struct cfq_rq *crq)
{
609 610
	rb_erase(&crq->rb_node, &cfqq->sort_list);
	cfqq->queued[cfq_crq_is_sync(crq)]--;
Linus Torvalds's avatar
Linus Torvalds committed
611 612 613 614

	cfq_add_crq_rb(crq);
}

615 616
static struct request *cfq_find_rq_rb(struct cfq_data *cfqd, sector_t sector)

Linus Torvalds's avatar
Linus Torvalds committed
617
{
Jens Axboe's avatar
Jens Axboe committed
618
	struct cfq_queue *cfqq = cfq_find_cfq_hash(cfqd, current->pid, CFQ_KEY_ANY);
Linus Torvalds's avatar
Linus Torvalds committed
619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639
	struct rb_node *n;

	if (!cfqq)
		goto out;

	n = cfqq->sort_list.rb_node;
	while (n) {
		struct cfq_rq *crq = rb_entry_crq(n);

		if (sector < crq->rb_key)
			n = n->rb_left;
		else if (sector > crq->rb_key)
			n = n->rb_right;
		else
			return crq->request;
	}

out:
	return NULL;
}

640
static void cfq_activate_request(request_queue_t *q, struct request *rq)
Linus Torvalds's avatar
Linus Torvalds committed
641
{
642
	struct cfq_data *cfqd = q->elevator->elevator_data;
Jens Axboe's avatar
Jens Axboe committed
643

644
	cfqd->rq_in_driver++;
Linus Torvalds's avatar
Linus Torvalds committed
645 646
}

647
static void cfq_deactivate_request(request_queue_t *q, struct request *rq)
Linus Torvalds's avatar
Linus Torvalds committed
648
{
649 650 651 652
	struct cfq_data *cfqd = q->elevator->elevator_data;

	WARN_ON(!cfqd->rq_in_driver);
	cfqd->rq_in_driver--;
Linus Torvalds's avatar
Linus Torvalds committed
653 654
}

655
static void cfq_remove_request(struct request *rq)
Linus Torvalds's avatar
Linus Torvalds committed
656 657 658
{
	struct cfq_rq *crq = RQ_DATA(rq);

659 660
	list_del_init(&rq->queuelist);
	cfq_del_crq_rb(crq);
661
	cfq_del_crq_hash(crq);
Linus Torvalds's avatar
Linus Torvalds committed
662 663 664 665 666 667 668 669 670 671
}

static int
cfq_merge(request_queue_t *q, struct request **req, struct bio *bio)
{
	struct cfq_data *cfqd = q->elevator->elevator_data;
	struct request *__rq;
	int ret;

	__rq = cfq_find_rq_hash(cfqd, bio->bi_sector);
672 673 674
	if (__rq && elv_rq_merge_ok(__rq, bio)) {
		ret = ELEVATOR_BACK_MERGE;
		goto out;
Linus Torvalds's avatar
Linus Torvalds committed
675 676 677
	}

	__rq = cfq_find_rq_rb(cfqd, bio->bi_sector + bio_sectors(bio));
678 679 680
	if (__rq && elv_rq_merge_ok(__rq, bio)) {
		ret = ELEVATOR_FRONT_MERGE;
		goto out;
Linus Torvalds's avatar
Linus Torvalds committed
681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696
	}

	return ELEVATOR_NO_MERGE;
out:
	*req = __rq;
	return ret;
}

static void cfq_merged_request(request_queue_t *q, struct request *req)
{
	struct cfq_data *cfqd = q->elevator->elevator_data;
	struct cfq_rq *crq = RQ_DATA(req);

	cfq_del_crq_hash(crq);
	cfq_add_crq_hash(cfqd, crq);

697
	if (rq_rb_key(req) != crq->rb_key) {
Linus Torvalds's avatar
Linus Torvalds committed
698 699 700 701 702 703 704 705 706 707 708 709 710
		struct cfq_queue *cfqq = crq->cfq_queue;

		cfq_update_next_crq(crq);
		cfq_reposition_crq_rb(cfqq, crq);
	}
}

static void
cfq_merged_requests(request_queue_t *q, struct request *rq,
		    struct request *next)
{
	cfq_merged_request(q, rq);

711 712 713 714 715 716 717
	/*
	 * reposition in fifo if next is older than rq
	 */
	if (!list_empty(&rq->queuelist) && !list_empty(&next->queuelist) &&
	    time_before(next->start_time, rq->start_time))
		list_move(&rq->queuelist, &next->queuelist);

718
	cfq_remove_request(next);
719 720 721 722 723 724 725 726 727 728 729 730 731 732
}

static inline void
__cfq_set_active_queue(struct cfq_data *cfqd, struct cfq_queue *cfqq)
{
	if (cfqq) {
		/*
		 * stop potential idle class queues waiting service
		 */
		del_timer(&cfqd->idle_class_timer);

		cfqq->slice_start = jiffies;
		cfqq->slice_end = 0;
		cfqq->slice_left = 0;
Jens Axboe's avatar
Jens Axboe committed
733 734
		cfq_clear_cfqq_must_alloc_slice(cfqq);
		cfq_clear_cfqq_fifo_expire(cfqq);
735 736 737 738 739
	}

	cfqd->active_queue = cfqq;
}

740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790
/*
 * current cfqq expired its slice (or was too idle), select new one
 */
static void
__cfq_slice_expired(struct cfq_data *cfqd, struct cfq_queue *cfqq,
		    int preempted)
{
	unsigned long now = jiffies;

	if (cfq_cfqq_wait_request(cfqq))
		del_timer(&cfqd->idle_slice_timer);

	if (!preempted && !cfq_cfqq_dispatched(cfqq)) {
		cfqq->service_last = now;
		cfq_schedule_dispatch(cfqd);
	}

	cfq_clear_cfqq_must_dispatch(cfqq);
	cfq_clear_cfqq_wait_request(cfqq);

	/*
	 * store what was left of this slice, if the queue idled out
	 * or was preempted
	 */
	if (time_after(cfqq->slice_end, now))
		cfqq->slice_left = cfqq->slice_end - now;
	else
		cfqq->slice_left = 0;

	if (cfq_cfqq_on_rr(cfqq))
		cfq_resort_rr_list(cfqq, preempted);

	if (cfqq == cfqd->active_queue)
		cfqd->active_queue = NULL;

	if (cfqd->active_cic) {
		put_io_context(cfqd->active_cic->ioc);
		cfqd->active_cic = NULL;
	}

	cfqd->dispatch_slice = 0;
}

static inline void cfq_slice_expired(struct cfq_data *cfqd, int preempted)
{
	struct cfq_queue *cfqq = cfqd->active_queue;

	if (cfqq)
		__cfq_slice_expired(cfqd, cfqq, preempted);
}

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
/*
 * 0
 * 0,1
 * 0,1,2
 * 0,1,2,3
 * 0,1,2,3,4
 * 0,1,2,3,4,5
 * 0,1,2,3,4,5,6
 * 0,1,2,3,4,5,6,7
 */
static int cfq_get_next_prio_level(struct cfq_data *cfqd)
{
	int prio, wrap;

	prio = -1;
	wrap = 0;
	do {
		int p;

		for (p = cfqd->cur_prio; p <= cfqd->cur_end_prio; p++) {
			if (!list_empty(&cfqd->rr_list[p])) {
				prio = p;
				break;
			}
		}

		if (prio != -1)
			break;
		cfqd->cur_prio = 0;
		if (++cfqd->cur_end_prio == CFQ_PRIO_LISTS) {
			cfqd->cur_end_prio = 0;
			if (wrap)
				break;
			wrap = 1;
Linus Torvalds's avatar
Linus Torvalds committed
825
		}
826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842
	} while (1);

	if (unlikely(prio == -1))
		return -1;

	BUG_ON(prio >= CFQ_PRIO_LISTS);

	list_splice_init(&cfqd->rr_list[prio], &cfqd->cur_rr);

	cfqd->cur_prio = prio + 1;
	if (cfqd->cur_prio > cfqd->cur_end_prio) {
		cfqd->cur_end_prio = cfqd->cur_prio;
		cfqd->cur_prio = 0;
	}
	if (cfqd->cur_end_prio == CFQ_PRIO_LISTS) {
		cfqd->cur_prio = 0;
		cfqd->cur_end_prio = 0;
Linus Torvalds's avatar
Linus Torvalds committed
843 844
	}

845 846 847
	return prio;
}

Jens Axboe's avatar
Jens Axboe committed
848
static struct cfq_queue *cfq_set_active_queue(struct cfq_data *cfqd)
849
{
850
	struct cfq_queue *cfqq = NULL;
851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873

	/*
	 * if current list is non-empty, grab first entry. if it is empty,
	 * get next prio level and grab first entry then if any are spliced
	 */
	if (!list_empty(&cfqd->cur_rr) || cfq_get_next_prio_level(cfqd) != -1)
		cfqq = list_entry_cfqq(cfqd->cur_rr.next);

	/*
	 * if we have idle queues and no rt or be queues had pending
	 * requests, either allow immediate service if the grace period
	 * has passed or arm the idle grace timer
	 */
	if (!cfqq && !list_empty(&cfqd->idle_rr)) {
		unsigned long end = cfqd->last_end_request + CFQ_IDLE_GRACE;

		if (time_after_eq(jiffies, end))
			cfqq = list_entry_cfqq(cfqd->idle_rr.next);
		else
			mod_timer(&cfqd->idle_class_timer, end);
	}

	__cfq_set_active_queue(cfqd, cfqq);
Jens Axboe's avatar
Jens Axboe committed
874
	return cfqq;
875 876 877 878 879
}

static int cfq_arm_slice_timer(struct cfq_data *cfqd, struct cfq_queue *cfqq)

{
880 881
	unsigned long sl;

882 883 884 885 886 887 888 889
	WARN_ON(!RB_EMPTY(&cfqq->sort_list));
	WARN_ON(cfqq != cfqd->active_queue);

	/*
	 * idle is disabled, either manually or by past process history
	 */
	if (!cfqd->cfq_slice_idle)
		return 0;
Jens Axboe's avatar
Jens Axboe committed
890
	if (!cfq_cfqq_idle_window(cfqq))
891 892 893 894 895 896 897
		return 0;
	/*
	 * task has exited, don't wait
	 */
	if (cfqd->active_cic && !cfqd->active_cic->ioc->task)
		return 0;

Jens Axboe's avatar
Jens Axboe committed
898 899
	cfq_mark_cfqq_must_dispatch(cfqq);
	cfq_mark_cfqq_wait_request(cfqq);
900

901 902
	sl = min(cfqq->slice_end - 1, (unsigned long) cfqd->cfq_slice_idle);
	mod_timer(&cfqd->idle_slice_timer, jiffies + sl);
903
	return 1;
Linus Torvalds's avatar
Linus Torvalds committed
904 905
}

906
static void cfq_dispatch_insert(request_queue_t *q, struct cfq_rq *crq)
Linus Torvalds's avatar
Linus Torvalds committed
907 908 909
{
	struct cfq_data *cfqd = q->elevator->elevator_data;
	struct cfq_queue *cfqq = crq->cfq_queue;
910 911

	cfqq->next_crq = cfq_find_next_crq(cfqd, cfqq, crq);
912
	cfq_remove_request(crq->request);
Jens Axboe's avatar
Jens Axboe committed
913
	cfqq->on_dispatch[cfq_crq_is_sync(crq)]++;
914
	elv_dispatch_sort(q, crq->request);
Linus Torvalds's avatar
Linus Torvalds committed
915 916 917 918 919 920 921 922
}

/*
 * return expired entry, or NULL to just start from scratch in rbtree
 */
static inline struct cfq_rq *cfq_check_fifo(struct cfq_queue *cfqq)
{
	struct cfq_data *cfqd = cfqq->cfqd;
923
	struct request *rq;
Linus Torvalds's avatar
Linus Torvalds committed
924 925
	struct cfq_rq *crq;

Jens Axboe's avatar
Jens Axboe committed
926
	if (cfq_cfqq_fifo_expire(cfqq))
Linus Torvalds's avatar
Linus Torvalds committed
927 928
		return NULL;

929
	if (!list_empty(&cfqq->fifo)) {
Jens Axboe's avatar
Jens Axboe committed
930
		int fifo = cfq_cfqq_class_sync(cfqq);
Linus Torvalds's avatar
Linus Torvalds committed
931

932 933 934
		crq = RQ_DATA(list_entry_fifo(cfqq->fifo.next));
		rq = crq->request;
		if (time_after(jiffies, rq->start_time + cfqd->cfq_fifo_expire[fifo])) {
Jens Axboe's avatar
Jens Axboe committed
935
			cfq_mark_cfqq_fifo_expire(cfqq);
936 937
			return crq;
		}
Linus Torvalds's avatar
Linus Torvalds committed
938 939 940 941 942 943
	}

	return NULL;
}

/*
Jens Axboe's avatar
Jens Axboe committed
944 945 946
 * 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.
Linus Torvalds's avatar
Linus Torvalds committed
947
 */
948 949 950 951 952 953 954 955 956 957
static inline int
cfq_prio_to_slice(struct cfq_data *cfqd, struct cfq_queue *cfqq)
{
	const int base_slice = cfqd->cfq_slice[cfq_cfqq_sync(cfqq)];

	WARN_ON(cfqq->ioprio >= IOPRIO_BE_NR);

	return base_slice + (base_slice/CFQ_SLICE_SCALE * (4 - cfqq->ioprio));
}

Linus Torvalds's avatar
Linus Torvalds committed
958
static inline void
959
cfq_set_prio_slice(struct cfq_data *cfqd, struct cfq_queue *cfqq)
Linus Torvalds's avatar
Linus Torvalds committed
960
{
961 962
	cfqq->slice_end = cfq_prio_to_slice(cfqd, cfqq) + jiffies;
}
Linus Torvalds's avatar
Linus Torvalds committed
963

964 965 966 967
static inline int
cfq_prio_to_maxrq(struct cfq_data *cfqd, struct cfq_queue *cfqq)
{
	const int base_rq = cfqd->cfq_slice_async_rq;
Linus Torvalds's avatar
Linus Torvalds committed
968

969
	WARN_ON(cfqq->ioprio >= IOPRIO_BE_NR);
Linus Torvalds's avatar
Linus Torvalds committed
970

971
	return 2 * (base_rq + base_rq * (CFQ_PRIO_LISTS - 1 - cfqq->ioprio));
Linus Torvalds's avatar
Linus Torvalds committed
972 973
}

974 975 976
/*
 * get next queue for service
 */
977
static struct cfq_queue *cfq_select_queue(struct cfq_data *cfqd)
Linus Torvalds's avatar
Linus Torvalds committed
978
{
979
	unsigned long now = jiffies;
Linus Torvalds's avatar
Linus Torvalds committed
980 981
	struct cfq_queue *cfqq;

982 983 984
	cfqq = cfqd->active_queue;
	if (!cfqq)
		goto new_queue;
Linus Torvalds's avatar
Linus Torvalds committed
985

986 987 988
	/*
	 * slice has expired
	 */
Jens Axboe's avatar
Jens Axboe committed
989 990
	if (!cfq_cfqq_must_dispatch(cfqq) && time_after(now, cfqq->slice_end))
		goto expire;
Linus Torvalds's avatar
Linus Torvalds committed
991

992 993 994 995 996 997
	/*
	 * if queue has requests, dispatch one. if not, check if
	 * enough slice is left to wait for one
	 */
	if (!RB_EMPTY(&cfqq->sort_list))
		goto keep_queue;
998
	else if (cfq_cfqq_class_sync(cfqq) &&
999 1000 1001 1002 1003
		 time_before(now, cfqq->slice_end)) {
		if (cfq_arm_slice_timer(cfqd, cfqq))
			return NULL;
	}

Jens Axboe's avatar
Jens Axboe committed
1004
expire:
1005
	cfq_slice_expired(cfqd, 0);
Jens Axboe's avatar
Jens Axboe committed
1006 1007
new_queue:
	cfqq = cfq_set_active_queue(cfqd);
1008
keep_queue:
Jens Axboe's avatar
Jens Axboe committed
1009
	return cfqq;
1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021
}

static int
__cfq_dispatch_requests(struct cfq_data *cfqd, struct cfq_queue *cfqq,
			int max_dispatch)
{
	int dispatched = 0;

	BUG_ON(RB_EMPTY(&cfqq->sort_list));

	do {
		struct cfq_rq *crq;
Linus Torvalds's avatar
Linus Torvalds committed
1022 1023

		/*
1024
		 * follow expired path, else get first next available
Linus Torvalds's avatar
Linus Torvalds committed
1025
		 */
1026 1027 1028 1029 1030 1031
		if ((crq = cfq_check_fifo(cfqq)) == NULL)
			crq = cfqq->next_crq;

		/*
		 * finally, insert request into driver dispatch list
		 */
1032
		cfq_dispatch_insert(cfqd->queue, crq);
Linus Torvalds's avatar
Linus Torvalds committed
1033

1034 1035
		cfqd->dispatch_slice++;
		dispatched++;
Linus Torvalds's avatar
Linus Torvalds committed
1036

1037 1038 1039 1040
		if (!cfqd->active_cic) {
			atomic_inc(&crq->io_context->ioc->refcount);
			cfqd->active_cic = crq->io_context;
		}
Linus Torvalds's avatar
Linus Torvalds committed
1041

1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065
		if (RB_EMPTY(&cfqq->sort_list))
			break;

	} while (dispatched < max_dispatch);

	/*
	 * if slice end isn't set yet, set it. if at least one request was
	 * sync, use the sync time slice value
	 */
	if (!cfqq->slice_end)
		cfq_set_prio_slice(cfqd, cfqq);

	/*
	 * expire an async queue immediately if it has used up its slice. idle
	 * queue always expire after 1 dispatch round.
	 */
	if ((!cfq_cfqq_sync(cfqq) &&
	    cfqd->dispatch_slice >= cfq_prio_to_maxrq(cfqd, cfqq)) ||
	    cfq_class_idle(cfqq))
		cfq_slice_expired(cfqd, 0);

	return dispatched;
}

1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101
static int
cfq_forced_dispatch_cfqqs(struct list_head *list)
{
	int dispatched = 0;
	struct cfq_queue *cfqq, *next;
	struct cfq_rq *crq;

	list_for_each_entry_safe(cfqq, next, list, cfq_list) {
		while ((crq = cfqq->next_crq)) {
			cfq_dispatch_insert(cfqq->cfqd->queue, crq);
			dispatched++;
		}
		BUG_ON(!list_empty(&cfqq->fifo));
	}
	return dispatched;
}

static int
cfq_forced_dispatch(struct cfq_data *cfqd)
{
	int i, dispatched = 0;

	for (i = 0; i < CFQ_PRIO_LISTS; i++)
		dispatched += cfq_forced_dispatch_cfqqs(&cfqd->rr_list[i]);

	dispatched += cfq_forced_dispatch_cfqqs(&cfqd->busy_rr);
	dispatched += cfq_forced_dispatch_cfqqs(&cfqd->cur_rr);
	dispatched += cfq_forced_dispatch_cfqqs(&cfqd->idle_rr);

	cfq_slice_expired(cfqd, 0);

	BUG_ON(cfqd->busy_queues);

	return dispatched;
}

1102
static int
1103
cfq_dispatch_requests(request_queue_t *q, int force)
1104 1105 1106 1107 1108 1109 1110
{
	struct cfq_data *cfqd = q->elevator->elevator_data;
	struct cfq_queue *cfqq;

	if (!cfqd->busy_queues)
		return 0;

1111 1112 1113 1114
	if (unlikely(force))
		return cfq_forced_dispatch(cfqd);

	cfqq = cfq_select_queue(cfqd);
1115
	if (cfqq) {
1116 1117 1118 1119 1120 1121 1122 1123 1124
		int max_dispatch;

		/*
		 * if idle window is disabled, allow queue buildup
		 */
		if (!cfq_cfqq_idle_window(cfqq) &&
		    cfqd->rq_in_driver >= cfqd->cfq_max_depth)
			return 0;

Jens Axboe's avatar
Jens Axboe committed
1125 1126
		cfq_clear_cfqq_must_dispatch(cfqq);
		cfq_clear_cfqq_wait_request(cfqq);
1127 1128
		del_timer(&cfqd->idle_slice_timer);

1129 1130 1131
		max_dispatch = cfqd->cfq_quantum;
		if (cfq_class_idle(cfqq))
			max_dispatch = 1;
Linus Torvalds's avatar
Linus Torvalds committed
1132

1133
		return __cfq_dispatch_requests(cfqd, cfqq, max_dispatch);
Linus Torvalds's avatar
Linus Torvalds committed
1134 1135
	}

1136
	return 0;
Linus Torvalds's avatar
Linus Torvalds committed
1137 1138 1139 1140 1141 1142 1143 1144 1145 1146
}

/*
 * task holds one reference to the queue, dropped when task exits. each crq
 * in-flight on this queue also holds a reference, dropped when crq is freed.
 *
 * queue lock must be held here.
 */
static void cfq_put_queue(struct cfq_queue *cfqq)
{
1147 1148 1149
	struct cfq_data *cfqd = cfqq->cfqd;

	BUG_ON(atomic_read(&cfqq->ref) <= 0);
Linus Torvalds's avatar
Linus Torvalds committed
1150 1151 1152 1153 1154

	if (!atomic_dec_and_test(&cfqq->ref))
		return;

	BUG_ON(rb_first(&cfqq->sort_list));
1155
	BUG_ON(cfqq->allocated[READ] + cfqq->allocated[WRITE]);
Jens Axboe's avatar
Jens Axboe committed
1156
	BUG_ON(cfq_cfqq_on_rr(cfqq));
Linus Torvalds's avatar
Linus Torvalds committed
1157

1158
	if (unlikely(cfqd->active_queue == cfqq))
Jens Axboe's avatar
Jens Axboe committed
1159
		__cfq_slice_expired(cfqd, cfqq, 0);
1160

Linus Torvalds's avatar
Linus Torvalds committed
1161 1162 1163 1164 1165 1166 1167 1168 1169
	/*
	 * it's on the empty list and still hashed
	 */
	list_del(&cfqq->cfq_list);
	hlist_del(&cfqq->cfq_hash);
	kmem_cache_free(cfq_pool, cfqq);
}

static inline struct cfq_queue *
Jens Axboe's avatar
Jens Axboe committed
1170 1171
__cfq_find_cfq_hash(struct cfq_data *cfqd, unsigned int key, unsigned int prio,
		    const int hashval)
Linus Torvalds's avatar
Linus Torvalds committed
1172 1173 1174 1175 1176 1177
{
	struct hlist_head *hash_list = &cfqd->cfq_hash[hashval];
	struct hlist_node *entry, *next;

	hlist_for_each_safe(entry, next, hash_list) {
		struct cfq_queue *__cfqq = list_entry_qhash(entry);
Al Viro's avatar
Al Viro committed
1178
		const unsigned short __p = IOPRIO_PRIO_VALUE(__cfqq->org_ioprio_class, __cfqq->org_ioprio);
Linus Torvalds's avatar
Linus Torvalds committed
1179

Jens Axboe's avatar
Jens Axboe committed
1180
		if (__cfqq->key == key && (__p == prio || prio == CFQ_KEY_ANY))
Linus Torvalds's avatar
Linus Torvalds committed
1181 1182 1183 1184 1185 1186 1187
			return __cfqq;
	}

	return NULL;
}

static struct cfq_queue *
Jens Axboe's avatar
Jens Axboe committed
1188
cfq_find_cfq_hash(struct cfq_data *cfqd, unsigned int key, unsigned short prio)
Linus Torvalds's avatar
Linus Torvalds committed
1189
{
Jens Axboe's avatar
Jens Axboe committed
1190
	return __cfq_find_cfq_hash(cfqd, key, prio, hash_long(key, CFQ_QHASH_SHIFT));
Linus Torvalds's avatar
Linus Torvalds committed
1191 1192
}

1193
static void cfq_free_io_context(struct io_context *ioc)
Linus Torvalds's avatar
Linus Torvalds committed
1194
{
1195
	struct cfq_io_context *__cic;
1196 1197
	struct rb_node *n;
	int freed = 0;
Linus Torvalds's avatar
Linus Torvalds committed
1198

1199 1200 1201
	while ((n = rb_first(&ioc->cic_root)) != NULL) {
		__cic = rb_entry(n, struct cfq_io_context, rb_node);
		rb_erase(&__cic->rb_node, &ioc->cic_root);
1202
		kmem_cache_free(cfq_ioc_pool, __cic);
1203
		freed++;
Linus Torvalds's avatar
Linus Torvalds committed
1204 1205
	}

1206 1207
	if (atomic_sub_and_test(freed, &ioc_count) && ioc_gone)
		complete(ioc_gone);
Linus Torvalds's avatar
Linus Torvalds committed
1208 1209
}

1210 1211 1212
static void cfq_trim(struct io_context *ioc)
{
	ioc->set_ioprio = NULL;
1213
	cfq_free_io_context(ioc);
1214 1215
}

1216 1217 1218 1219
/*
 * Called with interrupts disabled
 */
static void cfq_exit_single_io_context(struct cfq_io_context *cic)
Linus Torvalds's avatar
Linus Torvalds committed
1220
{
1221
	struct cfq_data *cfqd = cic->key;
1222 1223 1224 1225 1226 1227
	request_queue_t *q;

	if (!cfqd)
		return;

	q = cfqd->queue;
1228 1229 1230 1231 1232

	WARN_ON(!irqs_disabled());

	spin_lock(q->queue_lock);

1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245
	if (cic->cfqq[ASYNC]) {
		if (unlikely(cic->cfqq[ASYNC] == cfqd->active_queue))
			__cfq_slice_expired(cfqd, cic->cfqq[ASYNC], 0);
		cfq_put_queue(cic->cfqq[ASYNC]);
		cic->cfqq[ASYNC] = NULL;
	}

	if (cic->cfqq[SYNC]) {
		if (unlikely(cic->cfqq[SYNC] == cfqd->active_queue))
			__cfq_slice_expired(cfqd, cic->cfqq[SYNC], 0);
		cfq_put_queue(cic->cfqq[SYNC]);
		cic->cfqq[SYNC] = NULL;
	}
1246

1247
	cic->key = NULL;
1248
	list_del_init(&cic->queue_list);
1249
	spin_unlock(q->queue_lock);
Linus Torvalds's avatar
Linus Torvalds committed
1250 1251
}

1252
static void cfq_exit_io_context(struct io_context *ioc)
Linus Torvalds's avatar
Linus Torvalds committed
1253
{
1254
	struct cfq_io_context *__cic;
Linus Torvalds's avatar
Linus Torvalds committed
1255
	unsigned long flags;
1256
	struct rb_node *n;
1257

Linus Torvalds's avatar
Linus Torvalds committed
1258 1259 1260
	/*
	 * put the reference this task is holding to the various queues
	 */
1261 1262 1263 1264 1265 1266
	read_lock_irqsave(&cfq_exit_lock, flags);

	n = rb_first(&ioc->cic_root);
	while (n != NULL) {
		__cic = rb_entry(n, struct cfq_io_context, rb_node);

1267
		cfq_exit_single_io_context(__cic);
1268
		n = rb_next(n);
Linus Torvalds's avatar
Linus Torvalds committed
1269 1270
	}

1271
	read_unlock_irqrestore(&cfq_exit_lock, flags);
Linus Torvalds's avatar
Linus Torvalds committed
1272 1273
}

1274
static struct cfq_io_context *
Al Viro's avatar
Al Viro committed
1275
cfq_alloc_io_context(struct cfq_data *cfqd, gfp_t gfp_mask)
Linus Torvalds's avatar
Linus Torvalds committed
1276
{
1277
	struct cfq_io_context *cic = kmem_cache_alloc(cfq_ioc_pool, gfp_mask);
Linus Torvalds's avatar
Linus Torvalds committed
1278 1279

	if (cic) {
1280 1281
		RB_CLEAR(&cic->rb_node);
		cic->key = NULL;
1282 1283
		cic->cfqq[ASYNC] = NULL;
		cic->cfqq[SYNC] = NULL;
1284 1285 1286 1287 1288 1289
		cic->last_end_request = jiffies;
		cic->ttime_total = 0;
		cic->ttime_samples = 0;
		cic->ttime_mean = 0;
		cic->dtor = cfq_free_io_context;
		cic->exit = cfq_exit_io_context;
1290
		INIT_LIST_HEAD(&cic->queue_list);
1291
		atomic_inc(&ioc_count);
Linus Torvalds's avatar
Linus Torvalds committed
1292 1293 1294 1295 1296
	}

	return cic;
}

1297 1298 1299 1300 1301
static void cfq_init_prio_data(struct cfq_queue *cfqq)
{
	struct task_struct *tsk = current;
	int ioprio_class;

Jens Axboe's avatar
Jens Axboe committed
1302
	if (!cfq_cfqq_prio_changed(cfqq))
1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326
		return;

	ioprio_class = IOPRIO_PRIO_CLASS(tsk->ioprio);
	switch (ioprio_class) {
		default:
			printk(KERN_ERR "cfq: bad prio %x\n", ioprio_class);
		case IOPRIO_CLASS_NONE:
			/*
			 * no prio set, place us in the middle of the BE classes
			 */
			cfqq->ioprio = task_nice_ioprio(tsk);
			cfqq->ioprio_class = IOPRIO_CLASS_BE;
			break;
		case IOPRIO_CLASS_RT:
			cfqq->ioprio = task_ioprio(tsk);
			cfqq->ioprio_class = IOPRIO_CLASS_RT;
			break;
		case IOPRIO_CLASS_BE:
			cfqq->ioprio = task_ioprio(tsk);
			cfqq->ioprio_class = IOPRIO_CLASS_BE;
			break;
		case IOPRIO_CLASS_IDLE:
			cfqq->ioprio_class = IOPRIO_CLASS_IDLE;
			cfqq->ioprio = 7;
Jens Axboe's avatar
Jens Axboe committed
1327
			cfq_clear_cfqq_idle_window(cfqq);
1328 1329 1330 1331 1332 1333 1334 1335 1336 1337
			break;
	}

	/*
	 * keep track of original prio settings in case we have to temporarily
	 * elevate the priority of this queue
	 */
	cfqq->org_ioprio = cfqq->ioprio;
	cfqq->org_ioprio_class = cfqq->ioprio_class;

Jens Axboe's avatar
Jens Axboe committed
1338
	if (cfq_cfqq_on_rr(cfqq))
1339 1340
		cfq_resort_rr_list(cfqq, 0);

Jens Axboe's avatar
Jens Axboe committed
1341
	cfq_clear_cfqq_prio_changed(cfqq);
1342 1343
}

1344
static inline void changed_ioprio(struct cfq_io_context *cic)
1345
{
1346 1347 1348
	struct cfq_data *cfqd = cic->key;
	struct cfq_queue *cfqq;
	if (cfqd) {
1349
		spin_lock(cfqd->queue->queue_lock);
1350 1351
		cfqq = cic->cfqq[ASYNC];
		if (cfqq) {
1352 1353 1354 1355 1356 1357 1358
			struct cfq_queue *new_cfqq;
			new_cfqq = cfq_get_queue(cfqd, CFQ_KEY_ASYNC,
						cic->ioc->task, GFP_ATOMIC);
			if (new_cfqq) {
				cic->cfqq[ASYNC] = new_cfqq;
				cfq_put_queue(cfqq);
			}
1359 1360
		}
		cfqq = cic->cfqq[SYNC];
1361 1362 1363 1364
		if (cfqq) {
			cfq_mark_cfqq_prio_changed(cfqq);
			cfq_init_prio_data(cfqq);
		}
1365 1366 1367 1368 1369 1370 1371 1372 1373
		spin_unlock(cfqd->queue->queue_lock);
	}
}

/*
 * callback from sys_ioprio_set, irqs are disabled
 */
static int cfq_ioc_set_ioprio(struct io_context *ioc, unsigned int ioprio)
{
1374
	struct cfq_io_context *cic;
1375
	struct rb_node *n;
1376 1377 1378

	write_lock(&cfq_exit_lock);

Jens Axboe's avatar