elevator.c 20.0 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34
/*
 *  Block device elevator/IO-scheduler.
 *
 *  Copyright (C) 2000 Andrea Arcangeli <andrea@suse.de> SuSE
 *
 * 30042000 Jens Axboe <axboe@suse.de> :
 *
 * Split the elevator a bit so that it is possible to choose a different
 * one or even write a new "plug in". There are three pieces:
 * - elevator_fn, inserts a new request in the queue list
 * - elevator_merge_fn, decides whether a new buffer can be merged with
 *   an existing request
 * - elevator_dequeue_fn, called when a request is taken off the active list
 *
 * 20082000 Dave Jones <davej@suse.de> :
 * Removed tests for max-bomb-segments, which was breaking elvtune
 *  when run without -bN
 *
 * Jens:
 * - Rework again to work with bio instead of buffer_heads
 * - loose bi_dev comparisons, partition handling is right now
 * - completely modularize elevator setup and teardown
 *
 */
#include <linux/kernel.h>
#include <linux/fs.h>
#include <linux/blkdev.h>
#include <linux/elevator.h>
#include <linux/bio.h>
#include <linux/config.h>
#include <linux/module.h>
#include <linux/slab.h>
#include <linux/init.h>
#include <linux/compiler.h>
T
Tejun Heo 已提交
35
#include <linux/delay.h>
36
#include <linux/blktrace_api.h>
L
Linus Torvalds 已提交
37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67

#include <asm/uaccess.h>

static DEFINE_SPINLOCK(elv_list_lock);
static LIST_HEAD(elv_list);

/*
 * can we safely merge with this request?
 */
inline int elv_rq_merge_ok(struct request *rq, struct bio *bio)
{
	if (!rq_mergeable(rq))
		return 0;

	/*
	 * different data direction or already started, don't merge
	 */
	if (bio_data_dir(bio) != rq_data_dir(rq))
		return 0;

	/*
	 * same device and no special stuff set, merge is ok
	 */
	if (rq->rq_disk == bio->bi_bdev->bd_disk &&
	    !rq->waiting && !rq->special)
		return 1;

	return 0;
}
EXPORT_SYMBOL(elv_rq_merge_ok);

68
static inline int elv_try_merge(struct request *__rq, struct bio *bio)
L
Linus Torvalds 已提交
69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110
{
	int ret = ELEVATOR_NO_MERGE;

	/*
	 * we can merge and sequence is ok, check if it's possible
	 */
	if (elv_rq_merge_ok(__rq, bio)) {
		if (__rq->sector + __rq->nr_sectors == bio->bi_sector)
			ret = ELEVATOR_BACK_MERGE;
		else if (__rq->sector - bio_sectors(bio) == bio->bi_sector)
			ret = ELEVATOR_FRONT_MERGE;
	}

	return ret;
}

static struct elevator_type *elevator_find(const char *name)
{
	struct elevator_type *e = NULL;
	struct list_head *entry;

	list_for_each(entry, &elv_list) {
		struct elevator_type *__e;

		__e = list_entry(entry, struct elevator_type, list);

		if (!strcmp(__e->elevator_name, name)) {
			e = __e;
			break;
		}
	}

	return e;
}

static void elevator_put(struct elevator_type *e)
{
	module_put(e->elevator_owner);
}

static struct elevator_type *elevator_get(const char *name)
{
111
	struct elevator_type *e;
L
Linus Torvalds 已提交
112

113 114 115 116 117 118 119
	spin_lock_irq(&elv_list_lock);

	e = elevator_find(name);
	if (e && !try_module_get(e->elevator_owner))
		e = NULL;

	spin_unlock_irq(&elv_list_lock);
L
Linus Torvalds 已提交
120 121 122 123

	return e;
}

124
static int elevator_attach(request_queue_t *q, struct elevator_queue *eq)
L
Linus Torvalds 已提交
125 126 127 128 129 130 131 132 133 134 135 136 137
{
	int ret = 0;

	q->elevator = eq;

	if (eq->ops->elevator_init_fn)
		ret = eq->ops->elevator_init_fn(q, eq);

	return ret;
}

static char chosen_elevator[16];

138
static int __init elevator_setup(char *str)
L
Linus Torvalds 已提交
139
{
140 141 142 143
	/*
	 * Be backwards-compatible with previous kernels, so users
	 * won't get the wrong elevator.
	 */
144
	if (!strcmp(str, "as"))
145
		strcpy(chosen_elevator, "anticipatory");
Z
Zachary Amsden 已提交
146
	else
147
		strncpy(chosen_elevator, str, sizeof(chosen_elevator) - 1);
L
Linus Torvalds 已提交
148 149 150 151 152
	return 0;
}

__setup("elevator=", elevator_setup);

153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178
static struct kobj_type elv_ktype;

static elevator_t *elevator_alloc(struct elevator_type *e)
{
	elevator_t *eq = kmalloc(sizeof(elevator_t), GFP_KERNEL);
	if (eq) {
		memset(eq, 0, sizeof(*eq));
		eq->ops = &e->ops;
		eq->elevator_type = e;
		kobject_init(&eq->kobj);
		snprintf(eq->kobj.name, KOBJ_NAME_LEN, "%s", "iosched");
		eq->kobj.ktype = &elv_ktype;
		mutex_init(&eq->sysfs_lock);
	} else {
		elevator_put(e);
	}
	return eq;
}

static void elevator_release(struct kobject *kobj)
{
	elevator_t *e = container_of(kobj, elevator_t, kobj);
	elevator_put(e->elevator_type);
	kfree(e);
}

L
Linus Torvalds 已提交
179 180 181 182 183 184
int elevator_init(request_queue_t *q, char *name)
{
	struct elevator_type *e = NULL;
	struct elevator_queue *eq;
	int ret = 0;

T
Tejun Heo 已提交
185 186 187 188 189
	INIT_LIST_HEAD(&q->queue_head);
	q->last_merge = NULL;
	q->end_sector = 0;
	q->boundary_rq = NULL;

190
	if (name && !(e = elevator_get(name)))
L
Linus Torvalds 已提交
191 192
		return -EINVAL;

193 194 195 196 197 198
	if (!e && *chosen_elevator && !(e = elevator_get(chosen_elevator)))
		printk("I/O scheduler %s not found\n", chosen_elevator);

	if (!e && !(e = elevator_get(CONFIG_DEFAULT_IOSCHED))) {
		printk("Default I/O scheduler not found, using no-op\n");
		e = elevator_get("noop");
199 200
	}

201 202
	eq = elevator_alloc(e);
	if (!eq)
L
Linus Torvalds 已提交
203 204
		return -ENOMEM;

205 206 207
	ret = elevator_attach(q, eq);
	if (ret)
		kobject_put(&eq->kobj);
L
Linus Torvalds 已提交
208 209 210 211 212 213

	return ret;
}

void elevator_exit(elevator_t *e)
{
214
	mutex_lock(&e->sysfs_lock);
L
Linus Torvalds 已提交
215 216
	if (e->ops->elevator_exit_fn)
		e->ops->elevator_exit_fn(e);
217 218
	e->ops = NULL;
	mutex_unlock(&e->sysfs_lock);
L
Linus Torvalds 已提交
219

220
	kobject_put(&e->kobj);
L
Linus Torvalds 已提交
221 222
}

223 224 225 226 227
/*
 * Insert rq into dispatch queue of q.  Queue lock must be held on
 * entry.  If sort != 0, rq is sort-inserted; otherwise, rq will be
 * appended to the dispatch queue.  To be used by specific elevators.
 */
J
Jens Axboe 已提交
228
void elv_dispatch_sort(request_queue_t *q, struct request *rq)
229 230 231 232
{
	sector_t boundary;
	struct list_head *entry;

233 234
	if (q->last_merge == rq)
		q->last_merge = NULL;
235
	q->nr_sorted--;
236

J
Jens Axboe 已提交
237
	boundary = q->end_sector;
238

239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257
	list_for_each_prev(entry, &q->queue_head) {
		struct request *pos = list_entry_rq(entry);

		if (pos->flags & (REQ_SOFTBARRIER|REQ_HARDBARRIER|REQ_STARTED))
			break;
		if (rq->sector >= boundary) {
			if (pos->sector < boundary)
				continue;
		} else {
			if (pos->sector >= boundary)
				break;
		}
		if (rq->sector >= pos->sector)
			break;
	}

	list_add(&rq->queuelist, entry);
}

L
Linus Torvalds 已提交
258 259 260
int elv_merge(request_queue_t *q, struct request **req, struct bio *bio)
{
	elevator_t *e = q->elevator;
261 262 263 264 265 266 267 268 269
	int ret;

	if (q->last_merge) {
		ret = elv_try_merge(q->last_merge, bio);
		if (ret != ELEVATOR_NO_MERGE) {
			*req = q->last_merge;
			return ret;
		}
	}
L
Linus Torvalds 已提交
270 271 272 273 274 275 276 277 278 279 280 281 282

	if (e->ops->elevator_merge_fn)
		return e->ops->elevator_merge_fn(q, req, bio);

	return ELEVATOR_NO_MERGE;
}

void elv_merged_request(request_queue_t *q, struct request *rq)
{
	elevator_t *e = q->elevator;

	if (e->ops->elevator_merged_fn)
		e->ops->elevator_merged_fn(q, rq);
283 284

	q->last_merge = rq;
L
Linus Torvalds 已提交
285 286 287 288 289 290 291 292 293
}

void elv_merge_requests(request_queue_t *q, struct request *rq,
			     struct request *next)
{
	elevator_t *e = q->elevator;

	if (e->ops->elevator_merge_req_fn)
		e->ops->elevator_merge_req_fn(q, rq, next);
294
	q->nr_sorted--;
295 296

	q->last_merge = rq;
L
Linus Torvalds 已提交
297 298
}

299
void elv_requeue_request(request_queue_t *q, struct request *rq)
L
Linus Torvalds 已提交
300 301 302 303 304 305 306
{
	elevator_t *e = q->elevator;

	/*
	 * it already went through dequeue, we need to decrement the
	 * in_flight count again
	 */
307
	if (blk_account_rq(rq)) {
L
Linus Torvalds 已提交
308
		q->in_flight--;
309 310 311
		if (blk_sorted_rq(rq) && e->ops->elevator_deactivate_req_fn)
			e->ops->elevator_deactivate_req_fn(q, rq);
	}
L
Linus Torvalds 已提交
312 313 314

	rq->flags &= ~REQ_STARTED;

315
	elv_insert(q, rq, ELEVATOR_INSERT_REQUEUE);
L
Linus Torvalds 已提交
316 317
}

318 319 320 321 322 323 324 325 326 327 328 329 330 331
static void elv_drain_elevator(request_queue_t *q)
{
	static int printed;
	while (q->elevator->ops->elevator_dispatch_fn(q, 1))
		;
	if (q->nr_sorted == 0)
		return;
	if (printed++ < 10) {
		printk(KERN_ERR "%s: forced dispatching is broken "
		       "(nr_sorted=%u), please report this\n",
		       q->elevator->elevator_type->elevator_name, q->nr_sorted);
	}
}

332
void elv_insert(request_queue_t *q, struct request *rq, int where)
L
Linus Torvalds 已提交
333
{
334 335 336
	struct list_head *pos;
	unsigned ordseq;

337 338
	blk_add_trace_rq(q, rq, BLK_TA_INSERT);

L
Linus Torvalds 已提交
339 340
	rq->q = q;

341 342 343 344 345 346 347 348 349
	switch (where) {
	case ELEVATOR_INSERT_FRONT:
		rq->flags |= REQ_SOFTBARRIER;

		list_add(&rq->queuelist, &q->queue_head);
		break;

	case ELEVATOR_INSERT_BACK:
		rq->flags |= REQ_SOFTBARRIER;
350
		elv_drain_elevator(q);
351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368
		list_add_tail(&rq->queuelist, &q->queue_head);
		/*
		 * We kick the queue here for the following reasons.
		 * - The elevator might have returned NULL previously
		 *   to delay requests and returned them now.  As the
		 *   queue wasn't empty before this request, ll_rw_blk
		 *   won't run the queue on return, resulting in hang.
		 * - Usually, back inserted requests won't be merged
		 *   with anything.  There's no point in delaying queue
		 *   processing.
		 */
		blk_remove_plug(q);
		q->request_fn(q);
		break;

	case ELEVATOR_INSERT_SORT:
		BUG_ON(!blk_fs_request(rq));
		rq->flags |= REQ_SORTED;
369
		q->nr_sorted++;
370 371
		if (q->last_merge == NULL && rq_mergeable(rq))
			q->last_merge = rq;
372 373 374 375 376 377
		/*
		 * Some ioscheds (cfq) run q->request_fn directly, so
		 * rq cannot be accessed after calling
		 * elevator_add_req_fn.
		 */
		q->elevator->ops->elevator_add_req_fn(q, rq);
378 379
		break;

380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403
	case ELEVATOR_INSERT_REQUEUE:
		/*
		 * If ordered flush isn't in progress, we do front
		 * insertion; otherwise, requests should be requeued
		 * in ordseq order.
		 */
		rq->flags |= REQ_SOFTBARRIER;

		if (q->ordseq == 0) {
			list_add(&rq->queuelist, &q->queue_head);
			break;
		}

		ordseq = blk_ordered_req_seq(rq);

		list_for_each(pos, &q->queue_head) {
			struct request *pos_rq = list_entry_rq(pos);
			if (ordseq <= blk_ordered_req_seq(pos_rq))
				break;
		}

		list_add_tail(&rq->queuelist, pos);
		break;

404 405 406 407 408 409 410 411 412 413 414 415 416
	default:
		printk(KERN_ERR "%s: bad insertion point %d\n",
		       __FUNCTION__, where);
		BUG();
	}

	if (blk_queue_plugged(q)) {
		int nrq = q->rq.count[READ] + q->rq.count[WRITE]
			- q->in_flight;

		if (nrq >= q->unplug_thresh)
			__generic_unplug_device(q);
	}
L
Linus Torvalds 已提交
417 418
}

419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454
void __elv_add_request(request_queue_t *q, struct request *rq, int where,
		       int plug)
{
	if (q->ordcolor)
		rq->flags |= REQ_ORDERED_COLOR;

	if (rq->flags & (REQ_SOFTBARRIER | REQ_HARDBARRIER)) {
		/*
		 * toggle ordered color
		 */
		if (blk_barrier_rq(rq))
			q->ordcolor ^= 1;

		/*
		 * barriers implicitly indicate back insertion
		 */
		if (where == ELEVATOR_INSERT_SORT)
			where = ELEVATOR_INSERT_BACK;

		/*
		 * this request is scheduling boundary, update
		 * end_sector
		 */
		if (blk_fs_request(rq)) {
			q->end_sector = rq_end_sector(rq);
			q->boundary_rq = rq;
		}
	} else if (!(rq->flags & REQ_ELVPRIV) && where == ELEVATOR_INSERT_SORT)
		where = ELEVATOR_INSERT_BACK;

	if (plug)
		blk_plug_device(q);

	elv_insert(q, rq, where);
}

L
Linus Torvalds 已提交
455 456 457 458 459 460 461 462 463 464 465 466
void elv_add_request(request_queue_t *q, struct request *rq, int where,
		     int plug)
{
	unsigned long flags;

	spin_lock_irqsave(q->queue_lock, flags);
	__elv_add_request(q, rq, where, plug);
	spin_unlock_irqrestore(q->queue_lock, flags);
}

static inline struct request *__elv_next_request(request_queue_t *q)
{
467 468
	struct request *rq;

469 470 471 472 473 474
	while (1) {
		while (!list_empty(&q->queue_head)) {
			rq = list_entry_rq(q->queue_head.next);
			if (blk_do_ordered(q, &rq))
				return rq;
		}
L
Linus Torvalds 已提交
475

476 477
		if (!q->elevator->ops->elevator_dispatch_fn(q, 0))
			return NULL;
L
Linus Torvalds 已提交
478 479 480 481 482 483 484 485 486
	}
}

struct request *elv_next_request(request_queue_t *q)
{
	struct request *rq;
	int ret;

	while ((rq = __elv_next_request(q)) != NULL) {
487 488 489 490 491 492 493 494 495 496 497
		if (!(rq->flags & REQ_STARTED)) {
			elevator_t *e = q->elevator;

			/*
			 * This is the first time the device driver
			 * sees this request (possibly after
			 * requeueing).  Notify IO scheduler.
			 */
			if (blk_sorted_rq(rq) &&
			    e->ops->elevator_activate_req_fn)
				e->ops->elevator_activate_req_fn(q, rq);
L
Linus Torvalds 已提交
498

499 500 501 502 503 504
			/*
			 * just mark as started even if we don't start
			 * it, a request that has been delayed should
			 * not be passed by new incoming requests
			 */
			rq->flags |= REQ_STARTED;
505
			blk_add_trace_rq(q, rq, BLK_TA_ISSUE);
506
		}
L
Linus Torvalds 已提交
507

508
		if (!q->boundary_rq || q->boundary_rq == rq) {
J
Jens Axboe 已提交
509
			q->end_sector = rq_end_sector(rq);
510 511
			q->boundary_rq = NULL;
		}
L
Linus Torvalds 已提交
512 513 514 515 516 517 518 519

		if ((rq->flags & REQ_DONTPREP) || !q->prep_rq_fn)
			break;

		ret = q->prep_rq_fn(q, rq);
		if (ret == BLKPREP_OK) {
			break;
		} else if (ret == BLKPREP_DEFER) {
520 521 522
			/*
			 * the request may have been (partially) prepped.
			 * we need to keep this request in the front to
523 524
			 * avoid resource deadlock.  REQ_STARTED will
			 * prevent other fs requests from passing this one.
525
			 */
L
Linus Torvalds 已提交
526 527 528 529 530 531 532 533 534 535 536
			rq = NULL;
			break;
		} else if (ret == BLKPREP_KILL) {
			int nr_bytes = rq->hard_nr_sectors << 9;

			if (!nr_bytes)
				nr_bytes = rq->data_len;

			blkdev_dequeue_request(rq);
			rq->flags |= REQ_QUIET;
			end_that_request_chunk(rq, 0, nr_bytes);
537
			end_that_request_last(rq, 0);
L
Linus Torvalds 已提交
538 539 540 541 542 543 544 545 546 547
		} else {
			printk(KERN_ERR "%s: bad return=%d\n", __FUNCTION__,
								ret);
			break;
		}
	}

	return rq;
}

548
void elv_dequeue_request(request_queue_t *q, struct request *rq)
L
Linus Torvalds 已提交
549
{
550 551 552
	BUG_ON(list_empty(&rq->queuelist));

	list_del_init(&rq->queuelist);
L
Linus Torvalds 已提交
553 554 555 556

	/*
	 * the time frame between a request being removed from the lists
	 * and to it is freed is accounted as io that is in progress at
557
	 * the driver side.
L
Linus Torvalds 已提交
558 559 560 561 562 563 564 565 566
	 */
	if (blk_account_rq(rq))
		q->in_flight++;
}

int elv_queue_empty(request_queue_t *q)
{
	elevator_t *e = q->elevator;

567 568 569
	if (!list_empty(&q->queue_head))
		return 0;

L
Linus Torvalds 已提交
570 571 572
	if (e->ops->elevator_queue_empty_fn)
		return e->ops->elevator_queue_empty_fn(q);

573
	return 1;
L
Linus Torvalds 已提交
574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593
}

struct request *elv_latter_request(request_queue_t *q, struct request *rq)
{
	elevator_t *e = q->elevator;

	if (e->ops->elevator_latter_req_fn)
		return e->ops->elevator_latter_req_fn(q, rq);
	return NULL;
}

struct request *elv_former_request(request_queue_t *q, struct request *rq)
{
	elevator_t *e = q->elevator;

	if (e->ops->elevator_former_req_fn)
		return e->ops->elevator_former_req_fn(q, rq);
	return NULL;
}

594
int elv_set_request(request_queue_t *q, struct request *rq, struct bio *bio,
A
Al Viro 已提交
595
		    gfp_t gfp_mask)
L
Linus Torvalds 已提交
596 597 598 599
{
	elevator_t *e = q->elevator;

	if (e->ops->elevator_set_req_fn)
600
		return e->ops->elevator_set_req_fn(q, rq, bio, gfp_mask);
L
Linus Torvalds 已提交
601 602 603 604 605 606 607 608 609 610 611 612 613

	rq->elevator_private = NULL;
	return 0;
}

void elv_put_request(request_queue_t *q, struct request *rq)
{
	elevator_t *e = q->elevator;

	if (e->ops->elevator_put_req_fn)
		e->ops->elevator_put_req_fn(q, rq);
}

614
int elv_may_queue(request_queue_t *q, int rw, struct bio *bio)
L
Linus Torvalds 已提交
615 616 617 618
{
	elevator_t *e = q->elevator;

	if (e->ops->elevator_may_queue_fn)
619
		return e->ops->elevator_may_queue_fn(q, rw, bio);
L
Linus Torvalds 已提交
620 621 622 623 624 625 626 627 628 629 630

	return ELV_MQUEUE_MAY;
}

void elv_completed_request(request_queue_t *q, struct request *rq)
{
	elevator_t *e = q->elevator;

	/*
	 * request is released from the driver, io must be done
	 */
631
	if (blk_account_rq(rq)) {
L
Linus Torvalds 已提交
632
		q->in_flight--;
633 634 635
		if (blk_sorted_rq(rq) && e->ops->elevator_completed_req_fn)
			e->ops->elevator_completed_req_fn(q, rq);
	}
636

637 638 639 640 641 642 643
	/*
	 * Check if the queue is waiting for fs requests to be
	 * drained for flush sequence.
	 */
	if (unlikely(q->ordseq)) {
		struct request *first_rq = list_entry_rq(q->queue_head.next);
		if (q->in_flight == 0 &&
644 645 646 647 648
		    blk_ordered_cur_seq(q) == QUEUE_ORDSEQ_DRAIN &&
		    blk_ordered_req_seq(first_rq) > QUEUE_ORDSEQ_DRAIN) {
			blk_ordered_complete_seq(q, QUEUE_ORDSEQ_DRAIN, 0);
			q->request_fn(q);
		}
649
	}
L
Linus Torvalds 已提交
650 651
}

652 653 654 655
#define to_elv(atr) container_of((atr), struct elv_fs_entry, attr)

static ssize_t
elv_attr_show(struct kobject *kobj, struct attribute *attr, char *page)
L
Linus Torvalds 已提交
656
{
657 658 659 660 661 662 663 664 665 666 667 668
	elevator_t *e = container_of(kobj, elevator_t, kobj);
	struct elv_fs_entry *entry = to_elv(attr);
	ssize_t error;

	if (!entry->show)
		return -EIO;

	mutex_lock(&e->sysfs_lock);
	error = e->ops ? entry->show(e, page) : -ENOENT;
	mutex_unlock(&e->sysfs_lock);
	return error;
}
L
Linus Torvalds 已提交
669

670 671 672 673 674 675 676
static ssize_t
elv_attr_store(struct kobject *kobj, struct attribute *attr,
	       const char *page, size_t length)
{
	elevator_t *e = container_of(kobj, elevator_t, kobj);
	struct elv_fs_entry *entry = to_elv(attr);
	ssize_t error;
L
Linus Torvalds 已提交
677

678 679
	if (!entry->store)
		return -EIO;
L
Linus Torvalds 已提交
680

681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705
	mutex_lock(&e->sysfs_lock);
	error = e->ops ? entry->store(e, page, length) : -ENOENT;
	mutex_unlock(&e->sysfs_lock);
	return error;
}

static struct sysfs_ops elv_sysfs_ops = {
	.show	= elv_attr_show,
	.store	= elv_attr_store,
};

static struct kobj_type elv_ktype = {
	.sysfs_ops	= &elv_sysfs_ops,
	.release	= elevator_release,
};

int elv_register_queue(struct request_queue *q)
{
	elevator_t *e = q->elevator;
	int error;

	e->kobj.parent = &q->kobj;

	error = kobject_add(&e->kobj);
	if (!error) {
706
		struct elv_fs_entry *attr = e->elevator_type->elevator_attrs;
707
		if (attr) {
708 709
			while (attr->attr.name) {
				if (sysfs_create_file(&e->kobj, &attr->attr))
710
					break;
711
				attr++;
712 713 714 715 716
			}
		}
		kobject_uevent(&e->kobj, KOBJ_ADD);
	}
	return error;
L
Linus Torvalds 已提交
717 718 719 720 721 722
}

void elv_unregister_queue(struct request_queue *q)
{
	if (q) {
		elevator_t *e = q->elevator;
723 724
		kobject_uevent(&e->kobj, KOBJ_REMOVE);
		kobject_del(&e->kobj);
L
Linus Torvalds 已提交
725 726 727 728 729
	}
}

int elv_register(struct elevator_type *e)
{
730
	spin_lock_irq(&elv_list_lock);
L
Linus Torvalds 已提交
731 732 733 734 735 736
	if (elevator_find(e->elevator_name))
		BUG();
	list_add_tail(&e->list, &elv_list);
	spin_unlock_irq(&elv_list_lock);

	printk(KERN_INFO "io scheduler %s registered", e->elevator_name);
737 738 739 740
	if (!strcmp(e->elevator_name, chosen_elevator) ||
			(!*chosen_elevator &&
			 !strcmp(e->elevator_name, CONFIG_DEFAULT_IOSCHED)))
				printk(" (default)");
L
Linus Torvalds 已提交
741 742 743 744 745 746 747
	printk("\n");
	return 0;
}
EXPORT_SYMBOL_GPL(elv_register);

void elv_unregister(struct elevator_type *e)
{
748 749 750 751 752
	struct task_struct *g, *p;

	/*
	 * Iterate every thread in the process to remove the io contexts.
	 */
753 754 755 756 757 758 759 760 761
	if (e->ops.trim) {
		read_lock(&tasklist_lock);
		do_each_thread(g, p) {
			task_lock(p);
			e->ops.trim(p->io_context);
			task_unlock(p);
		} while_each_thread(g, p);
		read_unlock(&tasklist_lock);
	}
762

L
Linus Torvalds 已提交
763 764 765 766 767 768 769 770 771 772
	spin_lock_irq(&elv_list_lock);
	list_del_init(&e->list);
	spin_unlock_irq(&elv_list_lock);
}
EXPORT_SYMBOL_GPL(elv_unregister);

/*
 * switch to new_e io scheduler. be careful not to introduce deadlocks -
 * we don't free the old io scheduler, before we have allocated what we
 * need for the new one. this way we have a chance of going back to the old
T
Tejun Heo 已提交
773
 * one, if the new one fails init for some reason.
L
Linus Torvalds 已提交
774
 */
775
static int elevator_switch(request_queue_t *q, struct elevator_type *new_e)
L
Linus Torvalds 已提交
776
{
T
Tejun Heo 已提交
777
	elevator_t *old_elevator, *e;
L
Linus Torvalds 已提交
778

T
Tejun Heo 已提交
779 780 781
	/*
	 * Allocate new elevator
	 */
782
	e = elevator_alloc(new_e);
L
Linus Torvalds 已提交
783
	if (!e)
784
		return 0;
L
Linus Torvalds 已提交
785 786

	/*
T
Tejun Heo 已提交
787
	 * Turn on BYPASS and drain all requests w/ elevator private data
L
Linus Torvalds 已提交
788
	 */
T
Tejun Heo 已提交
789 790
	spin_lock_irq(q->queue_lock);

J
Jens Axboe 已提交
791
	set_bit(QUEUE_FLAG_ELVSWITCH, &q->queue_flags);
T
Tejun Heo 已提交
792

793
	elv_drain_elevator(q);
T
Tejun Heo 已提交
794 795

	while (q->rq.elvpriv) {
796 797
		blk_remove_plug(q);
		q->request_fn(q);
T
Tejun Heo 已提交
798
		spin_unlock_irq(q->queue_lock);
J
Jens Axboe 已提交
799
		msleep(10);
T
Tejun Heo 已提交
800
		spin_lock_irq(q->queue_lock);
801
		elv_drain_elevator(q);
T
Tejun Heo 已提交
802 803 804
	}

	spin_unlock_irq(q->queue_lock);
L
Linus Torvalds 已提交
805 806 807 808 809 810 811 812 813 814

	/*
	 * unregister old elevator data
	 */
	elv_unregister_queue(q);
	old_elevator = q->elevator;

	/*
	 * attach and start new elevator
	 */
815
	if (elevator_attach(q, e))
L
Linus Torvalds 已提交
816 817 818 819 820 821
		goto fail;

	if (elv_register_queue(q))
		goto fail_register;

	/*
T
Tejun Heo 已提交
822
	 * finally exit old elevator and turn off BYPASS.
L
Linus Torvalds 已提交
823 824
	 */
	elevator_exit(old_elevator);
J
Jens Axboe 已提交
825
	clear_bit(QUEUE_FLAG_ELVSWITCH, &q->queue_flags);
826
	return 1;
L
Linus Torvalds 已提交
827 828 829 830 831 832 833

fail_register:
	/*
	 * switch failed, exit the new io scheduler and reattach the old
	 * one again (along with re-adding the sysfs dir)
	 */
	elevator_exit(e);
T
Tejun Heo 已提交
834
	e = NULL;
L
Linus Torvalds 已提交
835 836 837
fail:
	q->elevator = old_elevator;
	elv_register_queue(q);
J
Jens Axboe 已提交
838
	clear_bit(QUEUE_FLAG_ELVSWITCH, &q->queue_flags);
839 840 841
	if (e)
		kobject_put(&e->kobj);
	return 0;
L
Linus Torvalds 已提交
842 843 844 845 846
}

ssize_t elv_iosched_store(request_queue_t *q, const char *name, size_t count)
{
	char elevator_name[ELV_NAME_MAX];
847
	size_t len;
L
Linus Torvalds 已提交
848 849
	struct elevator_type *e;

850 851 852
	elevator_name[sizeof(elevator_name) - 1] = '\0';
	strncpy(elevator_name, name, sizeof(elevator_name) - 1);
	len = strlen(elevator_name);
L
Linus Torvalds 已提交
853

854 855
	if (len && elevator_name[len - 1] == '\n')
		elevator_name[len - 1] = '\0';
L
Linus Torvalds 已提交
856 857 858 859 860 861 862

	e = elevator_get(elevator_name);
	if (!e) {
		printk(KERN_ERR "elevator: type %s not found\n", elevator_name);
		return -EINVAL;
	}

863 864
	if (!strcmp(elevator_name, q->elevator->elevator_type->elevator_name)) {
		elevator_put(e);
L
Linus Torvalds 已提交
865
		return count;
866
	}
L
Linus Torvalds 已提交
867

868 869
	if (!elevator_switch(q, e))
		printk(KERN_ERR "elevator: switch to %s failed\n",elevator_name);
L
Linus Torvalds 已提交
870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895
	return count;
}

ssize_t elv_iosched_show(request_queue_t *q, char *name)
{
	elevator_t *e = q->elevator;
	struct elevator_type *elv = e->elevator_type;
	struct list_head *entry;
	int len = 0;

	spin_lock_irq(q->queue_lock);
	list_for_each(entry, &elv_list) {
		struct elevator_type *__e;

		__e = list_entry(entry, struct elevator_type, list);
		if (!strcmp(elv->elevator_name, __e->elevator_name))
			len += sprintf(name+len, "[%s] ", elv->elevator_name);
		else
			len += sprintf(name+len, "%s ", __e->elevator_name);
	}
	spin_unlock_irq(q->queue_lock);

	len += sprintf(len+name, "\n");
	return len;
}

J
Jens Axboe 已提交
896
EXPORT_SYMBOL(elv_dispatch_sort);
L
Linus Torvalds 已提交
897 898 899 900
EXPORT_SYMBOL(elv_add_request);
EXPORT_SYMBOL(__elv_add_request);
EXPORT_SYMBOL(elv_requeue_request);
EXPORT_SYMBOL(elv_next_request);
901
EXPORT_SYMBOL(elv_dequeue_request);
L
Linus Torvalds 已提交
902 903 904 905
EXPORT_SYMBOL(elv_queue_empty);
EXPORT_SYMBOL(elv_completed_request);
EXPORT_SYMBOL(elevator_exit);
EXPORT_SYMBOL(elevator_init);