blk-throttle.c 30.7 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12
/*
 * Interface for controlling IO bandwidth on a request queue
 *
 * Copyright (C) 2010 Vivek Goyal <vgoyal@redhat.com>
 */

#include <linux/module.h>
#include <linux/slab.h>
#include <linux/blkdev.h>
#include <linux/bio.h>
#include <linux/blktrace_api.h>
#include "blk-cgroup.h"
13
#include "blk.h"
14 15 16 17 18 19 20 21 22 23

/* Max dispatch from a group in 1 round */
static int throtl_grp_quantum = 8;

/* Total max dispatch from all groups in one round */
static int throtl_quantum = 32;

/* Throttling is performed over 100ms slice and after that slice is renewed */
static unsigned long throtl_slice = HZ/10;	/* 100 ms */

24 25 26 27 28
/* A workqueue to queue throttle related work */
static struct workqueue_struct *kthrotld_workqueue;
static void throtl_schedule_delayed_work(struct throtl_data *td,
				unsigned long delay);

29 30 31 32 33 34 35 36 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
struct throtl_rb_root {
	struct rb_root rb;
	struct rb_node *left;
	unsigned int count;
	unsigned long min_disptime;
};

#define THROTL_RB_ROOT	(struct throtl_rb_root) { .rb = RB_ROOT, .left = NULL, \
			.count = 0, .min_disptime = 0}

#define rb_entry_tg(node)	rb_entry((node), struct throtl_grp, rb_node)

struct throtl_grp {
	/* List of throtl groups on the request queue*/
	struct hlist_node tg_node;

	/* active throtl group service_tree member */
	struct rb_node rb_node;

	/*
	 * Dispatch time in jiffies. This is the estimated time when group
	 * will unthrottle and is ready to dispatch more bio. It is used as
	 * key to sort active groups in service tree.
	 */
	unsigned long disptime;

	struct blkio_group blkg;
	atomic_t ref;
	unsigned int flags;

	/* Two lists for READ and WRITE */
	struct bio_list bio_lists[2];

	/* Number of queued bios on READ and WRITE lists */
	unsigned int nr_queued[2];

	/* bytes per second rate limits */
	uint64_t bps[2];

68 69 70
	/* IOPS limits */
	unsigned int iops[2];

71 72
	/* Number of bytes disptached in current slice */
	uint64_t bytes_disp[2];
73 74
	/* Number of bio's dispatched in current slice */
	unsigned int io_disp[2];
75 76 77 78

	/* When did we start a new slice */
	unsigned long slice_start[2];
	unsigned long slice_end[2];
79 80

	/* Some throttle limits got updated for the group */
81
	int limits_changed;
82 83

	struct rcu_head rcu_head;
84 85 86 87 88 89 90 91 92 93
};

struct throtl_data
{
	/* List of throtl groups */
	struct hlist_head tg_list;

	/* service tree for active throtl groups */
	struct throtl_rb_root tg_service_tree;

94
	struct throtl_grp *root_tg;
95 96 97 98 99 100
	struct request_queue *queue;

	/* Total Number of queued bios on READ and WRITE lists */
	unsigned int nr_queued[2];

	/*
V
Vivek Goyal 已提交
101
	 * number of total undestroyed groups
102 103 104 105 106
	 */
	unsigned int nr_undestroyed_grps;

	/* Work for dispatching throttled bios */
	struct delayed_work throtl_work;
107

108
	int limits_changed;
109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145
};

enum tg_state_flags {
	THROTL_TG_FLAG_on_rr = 0,	/* on round-robin busy list */
};

#define THROTL_TG_FNS(name)						\
static inline void throtl_mark_tg_##name(struct throtl_grp *tg)		\
{									\
	(tg)->flags |= (1 << THROTL_TG_FLAG_##name);			\
}									\
static inline void throtl_clear_tg_##name(struct throtl_grp *tg)	\
{									\
	(tg)->flags &= ~(1 << THROTL_TG_FLAG_##name);			\
}									\
static inline int throtl_tg_##name(const struct throtl_grp *tg)		\
{									\
	return ((tg)->flags & (1 << THROTL_TG_FLAG_##name)) != 0;	\
}

THROTL_TG_FNS(on_rr);

#define throtl_log_tg(td, tg, fmt, args...)				\
	blk_add_trace_msg((td)->queue, "throtl %s " fmt,		\
				blkg_path(&(tg)->blkg), ##args);      	\

#define throtl_log(td, fmt, args...)	\
	blk_add_trace_msg((td)->queue, "throtl " fmt, ##args)

static inline struct throtl_grp *tg_of_blkg(struct blkio_group *blkg)
{
	if (blkg)
		return container_of(blkg, struct throtl_grp, blkg);

	return NULL;
}

146
static inline unsigned int total_nr_queued(struct throtl_data *td)
147
{
148
	return td->nr_queued[0] + td->nr_queued[1];
149 150 151 152 153 154 155 156
}

static inline struct throtl_grp *throtl_ref_get_tg(struct throtl_grp *tg)
{
	atomic_inc(&tg->ref);
	return tg;
}

157 158 159 160 161
static void throtl_free_tg(struct rcu_head *head)
{
	struct throtl_grp *tg;

	tg = container_of(head, struct throtl_grp, rcu_head);
162
	free_percpu(tg->blkg.stats_cpu);
163 164 165
	kfree(tg);
}

166 167 168 169 170
static void throtl_put_tg(struct throtl_grp *tg)
{
	BUG_ON(atomic_read(&tg->ref) <= 0);
	if (!atomic_dec_and_test(&tg->ref))
		return;
171

172 173 174
	/* release the extra blkcg reference this blkg has been holding */
	css_put(&tg->blkg.blkcg->css);

175 176 177 178 179 180 181 182 183 184
	/*
	 * A group is freed in rcu manner. But having an rcu lock does not
	 * mean that one can access all the fields of blkg and assume these
	 * are valid. For example, don't try to follow throtl_data and
	 * request queue links.
	 *
	 * Having a reference to blkg under an rcu allows acess to only
	 * values local to groups like group stats and group rate limits
	 */
	call_rcu(&tg->rcu_head, throtl_free_tg);
185 186
}

187 188
static struct blkio_group *throtl_alloc_blkio_group(struct request_queue *q,
						    struct blkio_cgroup *blkcg)
189
{
190 191 192 193 194 195
	struct throtl_grp *tg;

	tg = kzalloc_node(sizeof(*tg), GFP_ATOMIC, q->node);
	if (!tg)
		return NULL;

196 197 198 199 200 201
	INIT_HLIST_NODE(&tg->tg_node);
	RB_CLEAR_NODE(&tg->rb_node);
	bio_list_init(&tg->bio_lists[0]);
	bio_list_init(&tg->bio_lists[1]);
	tg->limits_changed = false;

202 203 204 205
	tg->bps[READ] = -1;
	tg->bps[WRITE] = -1;
	tg->iops[READ] = -1;
	tg->iops[WRITE] = -1;
206 207 208 209 210 211 212 213 214

	/*
	 * Take the initial reference that will be released on destroy
	 * This can be thought of a joint reference by cgroup and
	 * request queue which will be dropped by either request queue
	 * exit or cgroup deletion path depending on who is exiting first.
	 */
	atomic_set(&tg->ref, 1);

215
	return &tg->blkg;
216 217
}

218 219
static void throtl_link_blkio_group(struct request_queue *q,
				    struct blkio_group *blkg)
220
{
221 222
	struct throtl_data *td = q->td;
	struct throtl_grp *tg = tg_of_blkg(blkg);
223

224 225
	hlist_add_head(&tg->tg_node, &td->tg_list);
	td->nr_undestroyed_grps++;
226 227 228
}

static struct
229
throtl_grp *throtl_lookup_tg(struct throtl_data *td, struct blkio_cgroup *blkcg)
230
{
231 232
	/*
	 * This is the common case when there are no blkio cgroups.
233 234
	 * Avoid lookup in this case
	 */
235
	if (blkcg == &blkio_root_cgroup)
236
		return td->root_tg;
237

238
	return tg_of_blkg(blkg_lookup(blkcg, td->queue, BLKIO_POLICY_THROTL));
239 240
}

241 242
static struct throtl_grp *throtl_lookup_create_tg(struct throtl_data *td,
						  struct blkio_cgroup *blkcg)
243
{
244
	struct request_queue *q = td->queue;
245
	struct throtl_grp *tg = NULL;
246

247
	/*
248 249
	 * This is the common case when there are no blkio cgroups.
	 * Avoid lookup in this case
250
	 */
251 252 253 254
	if (blkcg == &blkio_root_cgroup) {
		tg = td->root_tg;
	} else {
		struct blkio_group *blkg;
255

256
		blkg = blkg_lookup_create(blkcg, q, BLKIO_POLICY_THROTL, false);
257

258 259 260 261 262
		/* if %NULL and @q is alive, fall back to root_tg */
		if (!IS_ERR(blkg))
			tg = tg_of_blkg(blkg);
		else if (!blk_queue_dead(q))
			tg = td->root_tg;
263 264
	}

265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377
	return tg;
}

static struct throtl_grp *throtl_rb_first(struct throtl_rb_root *root)
{
	/* Service tree is empty */
	if (!root->count)
		return NULL;

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

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

	return NULL;
}

static void rb_erase_init(struct rb_node *n, struct rb_root *root)
{
	rb_erase(n, root);
	RB_CLEAR_NODE(n);
}

static void throtl_rb_erase(struct rb_node *n, struct throtl_rb_root *root)
{
	if (root->left == n)
		root->left = NULL;
	rb_erase_init(n, &root->rb);
	--root->count;
}

static void update_min_dispatch_time(struct throtl_rb_root *st)
{
	struct throtl_grp *tg;

	tg = throtl_rb_first(st);
	if (!tg)
		return;

	st->min_disptime = tg->disptime;
}

static void
tg_service_tree_add(struct throtl_rb_root *st, struct throtl_grp *tg)
{
	struct rb_node **node = &st->rb.rb_node;
	struct rb_node *parent = NULL;
	struct throtl_grp *__tg;
	unsigned long key = tg->disptime;
	int left = 1;

	while (*node != NULL) {
		parent = *node;
		__tg = rb_entry_tg(parent);

		if (time_before(key, __tg->disptime))
			node = &parent->rb_left;
		else {
			node = &parent->rb_right;
			left = 0;
		}
	}

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

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

static void __throtl_enqueue_tg(struct throtl_data *td, struct throtl_grp *tg)
{
	struct throtl_rb_root *st = &td->tg_service_tree;

	tg_service_tree_add(st, tg);
	throtl_mark_tg_on_rr(tg);
	st->count++;
}

static void throtl_enqueue_tg(struct throtl_data *td, struct throtl_grp *tg)
{
	if (!throtl_tg_on_rr(tg))
		__throtl_enqueue_tg(td, tg);
}

static void __throtl_dequeue_tg(struct throtl_data *td, struct throtl_grp *tg)
{
	throtl_rb_erase(&tg->rb_node, &td->tg_service_tree);
	throtl_clear_tg_on_rr(tg);
}

static void throtl_dequeue_tg(struct throtl_data *td, struct throtl_grp *tg)
{
	if (throtl_tg_on_rr(tg))
		__throtl_dequeue_tg(td, tg);
}

static void throtl_schedule_next_dispatch(struct throtl_data *td)
{
	struct throtl_rb_root *st = &td->tg_service_tree;

	/*
	 * If there are more bios pending, schedule more work.
	 */
	if (!total_nr_queued(td))
		return;

	BUG_ON(!st->count);

	update_min_dispatch_time(st);

	if (time_before_eq(st->min_disptime, jiffies))
378
		throtl_schedule_delayed_work(td, 0);
379
	else
380
		throtl_schedule_delayed_work(td, (st->min_disptime - jiffies));
381 382 383 384 385 386
}

static inline void
throtl_start_new_slice(struct throtl_data *td, struct throtl_grp *tg, bool rw)
{
	tg->bytes_disp[rw] = 0;
387
	tg->io_disp[rw] = 0;
388 389 390 391 392 393 394
	tg->slice_start[rw] = jiffies;
	tg->slice_end[rw] = jiffies + throtl_slice;
	throtl_log_tg(td, tg, "[%c] new slice start=%lu end=%lu jiffies=%lu",
			rw == READ ? 'R' : 'W', tg->slice_start[rw],
			tg->slice_end[rw], jiffies);
}

395 396 397 398 399 400
static inline void throtl_set_slice_end(struct throtl_data *td,
		struct throtl_grp *tg, bool rw, unsigned long jiffy_end)
{
	tg->slice_end[rw] = roundup(jiffy_end, throtl_slice);
}

401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423
static inline void throtl_extend_slice(struct throtl_data *td,
		struct throtl_grp *tg, bool rw, unsigned long jiffy_end)
{
	tg->slice_end[rw] = roundup(jiffy_end, throtl_slice);
	throtl_log_tg(td, tg, "[%c] extend slice start=%lu end=%lu jiffies=%lu",
			rw == READ ? 'R' : 'W', tg->slice_start[rw],
			tg->slice_end[rw], jiffies);
}

/* Determine if previously allocated or extended slice is complete or not */
static bool
throtl_slice_used(struct throtl_data *td, struct throtl_grp *tg, bool rw)
{
	if (time_in_range(jiffies, tg->slice_start[rw], tg->slice_end[rw]))
		return 0;

	return 1;
}

/* Trim the used slices and adjust slice start accordingly */
static inline void
throtl_trim_slice(struct throtl_data *td, struct throtl_grp *tg, bool rw)
{
424 425
	unsigned long nr_slices, time_elapsed, io_trim;
	u64 bytes_trim, tmp;
426 427 428 429 430 431 432 433 434 435 436

	BUG_ON(time_before(tg->slice_end[rw], tg->slice_start[rw]));

	/*
	 * If bps are unlimited (-1), then time slice don't get
	 * renewed. Don't try to trim the slice if slice is used. A new
	 * slice will start when appropriate.
	 */
	if (throtl_slice_used(td, tg, rw))
		return;

437 438 439 440 441 442 443 444 445 446
	/*
	 * A bio has been dispatched. Also adjust slice_end. It might happen
	 * that initially cgroup limit was very low resulting in high
	 * slice_end, but later limit was bumped up and bio was dispached
	 * sooner, then we need to reduce slice_end. A high bogus slice_end
	 * is bad because it does not allow new slice to start.
	 */

	throtl_set_slice_end(td, tg, rw, jiffies + throtl_slice);

447 448 449 450 451 452
	time_elapsed = jiffies - tg->slice_start[rw];

	nr_slices = time_elapsed / throtl_slice;

	if (!nr_slices)
		return;
453 454 455
	tmp = tg->bps[rw] * throtl_slice * nr_slices;
	do_div(tmp, HZ);
	bytes_trim = tmp;
456

457
	io_trim = (tg->iops[rw] * throtl_slice * nr_slices)/HZ;
458

459
	if (!bytes_trim && !io_trim)
460 461 462 463 464 465 466
		return;

	if (tg->bytes_disp[rw] >= bytes_trim)
		tg->bytes_disp[rw] -= bytes_trim;
	else
		tg->bytes_disp[rw] = 0;

467 468 469 470 471
	if (tg->io_disp[rw] >= io_trim)
		tg->io_disp[rw] -= io_trim;
	else
		tg->io_disp[rw] = 0;

472 473
	tg->slice_start[rw] += nr_slices * throtl_slice;

474
	throtl_log_tg(td, tg, "[%c] trim slice nr=%lu bytes=%llu io=%lu"
475
			" start=%lu end=%lu jiffies=%lu",
476
			rw == READ ? 'R' : 'W', nr_slices, bytes_trim, io_trim,
477 478 479
			tg->slice_start[rw], tg->slice_end[rw], jiffies);
}

480 481
static bool tg_with_in_iops_limit(struct throtl_data *td, struct throtl_grp *tg,
		struct bio *bio, unsigned long *wait)
482 483
{
	bool rw = bio_data_dir(bio);
484
	unsigned int io_allowed;
485
	unsigned long jiffy_elapsed, jiffy_wait, jiffy_elapsed_rnd;
486
	u64 tmp;
487

488
	jiffy_elapsed = jiffy_elapsed_rnd = jiffies - tg->slice_start[rw];
489

490 491 492 493 494 495
	/* Slice has just started. Consider one slice interval */
	if (!jiffy_elapsed)
		jiffy_elapsed_rnd = throtl_slice;

	jiffy_elapsed_rnd = roundup(jiffy_elapsed_rnd, throtl_slice);

496 497 498 499 500 501 502 503 504 505 506 507 508 509
	/*
	 * jiffy_elapsed_rnd should not be a big value as minimum iops can be
	 * 1 then at max jiffy elapsed should be equivalent of 1 second as we
	 * will allow dispatch after 1 second and after that slice should
	 * have been trimmed.
	 */

	tmp = (u64)tg->iops[rw] * jiffy_elapsed_rnd;
	do_div(tmp, HZ);

	if (tmp > UINT_MAX)
		io_allowed = UINT_MAX;
	else
		io_allowed = tmp;
510 511

	if (tg->io_disp[rw] + 1 <= io_allowed) {
512 513 514 515 516
		if (wait)
			*wait = 0;
		return 1;
	}

517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533
	/* Calc approx time to dispatch */
	jiffy_wait = ((tg->io_disp[rw] + 1) * HZ)/tg->iops[rw] + 1;

	if (jiffy_wait > jiffy_elapsed)
		jiffy_wait = jiffy_wait - jiffy_elapsed;
	else
		jiffy_wait = 1;

	if (wait)
		*wait = jiffy_wait;
	return 0;
}

static bool tg_with_in_bps_limit(struct throtl_data *td, struct throtl_grp *tg,
		struct bio *bio, unsigned long *wait)
{
	bool rw = bio_data_dir(bio);
534
	u64 bytes_allowed, extra_bytes, tmp;
535
	unsigned long jiffy_elapsed, jiffy_wait, jiffy_elapsed_rnd;
536 537 538 539 540 541 542 543 544

	jiffy_elapsed = jiffy_elapsed_rnd = jiffies - tg->slice_start[rw];

	/* Slice has just started. Consider one slice interval */
	if (!jiffy_elapsed)
		jiffy_elapsed_rnd = throtl_slice;

	jiffy_elapsed_rnd = roundup(jiffy_elapsed_rnd, throtl_slice);

545 546
	tmp = tg->bps[rw] * jiffy_elapsed_rnd;
	do_div(tmp, HZ);
547
	bytes_allowed = tmp;
548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568

	if (tg->bytes_disp[rw] + bio->bi_size <= bytes_allowed) {
		if (wait)
			*wait = 0;
		return 1;
	}

	/* Calc approx time to dispatch */
	extra_bytes = tg->bytes_disp[rw] + bio->bi_size - bytes_allowed;
	jiffy_wait = div64_u64(extra_bytes * HZ, tg->bps[rw]);

	if (!jiffy_wait)
		jiffy_wait = 1;

	/*
	 * This wait time is without taking into consideration the rounding
	 * up we did. Add that time also.
	 */
	jiffy_wait = jiffy_wait + (jiffy_elapsed_rnd - jiffy_elapsed);
	if (wait)
		*wait = jiffy_wait;
569 570 571
	return 0;
}

572 573 574 575 576 577
static bool tg_no_rule_group(struct throtl_grp *tg, bool rw) {
	if (tg->bps[rw] == -1 && tg->iops[rw] == -1)
		return 1;
	return 0;
}

578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594
/*
 * Returns whether one can dispatch a bio or not. Also returns approx number
 * of jiffies to wait before this bio is with-in IO rate and can be dispatched
 */
static bool tg_may_dispatch(struct throtl_data *td, struct throtl_grp *tg,
				struct bio *bio, unsigned long *wait)
{
	bool rw = bio_data_dir(bio);
	unsigned long bps_wait = 0, iops_wait = 0, max_wait = 0;

	/*
 	 * Currently whole state machine of group depends on first bio
	 * queued in the group bio list. So one should not be calling
	 * this function with a different bio if there are other bios
	 * queued.
	 */
	BUG_ON(tg->nr_queued[rw] && bio != bio_list_peek(&tg->bio_lists[rw]));
595

596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628
	/* If tg->bps = -1, then BW is unlimited */
	if (tg->bps[rw] == -1 && tg->iops[rw] == -1) {
		if (wait)
			*wait = 0;
		return 1;
	}

	/*
	 * If previous slice expired, start a new one otherwise renew/extend
	 * existing slice to make sure it is at least throtl_slice interval
	 * long since now.
	 */
	if (throtl_slice_used(td, tg, rw))
		throtl_start_new_slice(td, tg, rw);
	else {
		if (time_before(tg->slice_end[rw], jiffies + throtl_slice))
			throtl_extend_slice(td, tg, rw, jiffies + throtl_slice);
	}

	if (tg_with_in_bps_limit(td, tg, bio, &bps_wait)
	    && tg_with_in_iops_limit(td, tg, bio, &iops_wait)) {
		if (wait)
			*wait = 0;
		return 1;
	}

	max_wait = max(bps_wait, iops_wait);

	if (wait)
		*wait = max_wait;

	if (time_before(tg->slice_end[rw], jiffies + max_wait))
		throtl_extend_slice(td, tg, rw, jiffies + max_wait);
629 630 631 632 633 634 635

	return 0;
}

static void throtl_charge_bio(struct throtl_grp *tg, struct bio *bio)
{
	bool rw = bio_data_dir(bio);
636
	bool sync = rw_is_sync(bio->bi_rw);
637 638 639

	/* Charge the bio to the group */
	tg->bytes_disp[rw] += bio->bi_size;
640
	tg->io_disp[rw]++;
641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702

	blkiocg_update_dispatch_stats(&tg->blkg, bio->bi_size, rw, sync);
}

static void throtl_add_bio_tg(struct throtl_data *td, struct throtl_grp *tg,
			struct bio *bio)
{
	bool rw = bio_data_dir(bio);

	bio_list_add(&tg->bio_lists[rw], bio);
	/* Take a bio reference on tg */
	throtl_ref_get_tg(tg);
	tg->nr_queued[rw]++;
	td->nr_queued[rw]++;
	throtl_enqueue_tg(td, tg);
}

static void tg_update_disptime(struct throtl_data *td, struct throtl_grp *tg)
{
	unsigned long read_wait = -1, write_wait = -1, min_wait = -1, disptime;
	struct bio *bio;

	if ((bio = bio_list_peek(&tg->bio_lists[READ])))
		tg_may_dispatch(td, tg, bio, &read_wait);

	if ((bio = bio_list_peek(&tg->bio_lists[WRITE])))
		tg_may_dispatch(td, tg, bio, &write_wait);

	min_wait = min(read_wait, write_wait);
	disptime = jiffies + min_wait;

	/* Update dispatch time */
	throtl_dequeue_tg(td, tg);
	tg->disptime = disptime;
	throtl_enqueue_tg(td, tg);
}

static void tg_dispatch_one_bio(struct throtl_data *td, struct throtl_grp *tg,
				bool rw, struct bio_list *bl)
{
	struct bio *bio;

	bio = bio_list_pop(&tg->bio_lists[rw]);
	tg->nr_queued[rw]--;
	/* Drop bio reference on tg */
	throtl_put_tg(tg);

	BUG_ON(td->nr_queued[rw] <= 0);
	td->nr_queued[rw]--;

	throtl_charge_bio(tg, bio);
	bio_list_add(bl, bio);
	bio->bi_rw |= REQ_THROTTLED;

	throtl_trim_slice(td, tg, rw);
}

static int throtl_dispatch_tg(struct throtl_data *td, struct throtl_grp *tg,
				struct bio_list *bl)
{
	unsigned int nr_reads = 0, nr_writes = 0;
	unsigned int max_nr_reads = throtl_grp_quantum*3/4;
703
	unsigned int max_nr_writes = throtl_grp_quantum - max_nr_reads;
704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761
	struct bio *bio;

	/* Try to dispatch 75% READS and 25% WRITES */

	while ((bio = bio_list_peek(&tg->bio_lists[READ]))
		&& tg_may_dispatch(td, tg, bio, NULL)) {

		tg_dispatch_one_bio(td, tg, bio_data_dir(bio), bl);
		nr_reads++;

		if (nr_reads >= max_nr_reads)
			break;
	}

	while ((bio = bio_list_peek(&tg->bio_lists[WRITE]))
		&& tg_may_dispatch(td, tg, bio, NULL)) {

		tg_dispatch_one_bio(td, tg, bio_data_dir(bio), bl);
		nr_writes++;

		if (nr_writes >= max_nr_writes)
			break;
	}

	return nr_reads + nr_writes;
}

static int throtl_select_dispatch(struct throtl_data *td, struct bio_list *bl)
{
	unsigned int nr_disp = 0;
	struct throtl_grp *tg;
	struct throtl_rb_root *st = &td->tg_service_tree;

	while (1) {
		tg = throtl_rb_first(st);

		if (!tg)
			break;

		if (time_before(jiffies, tg->disptime))
			break;

		throtl_dequeue_tg(td, tg);

		nr_disp += throtl_dispatch_tg(td, tg, bl);

		if (tg->nr_queued[0] || tg->nr_queued[1]) {
			tg_update_disptime(td, tg);
			throtl_enqueue_tg(td, tg);
		}

		if (nr_disp >= throtl_quantum)
			break;
	}

	return nr_disp;
}

762 763 764 765 766
static void throtl_process_limit_change(struct throtl_data *td)
{
	struct throtl_grp *tg;
	struct hlist_node *pos, *n;

767
	if (!td->limits_changed)
768 769
		return;

770
	xchg(&td->limits_changed, false);
771

772
	throtl_log(td, "limits changed");
773

774
	hlist_for_each_entry_safe(tg, pos, n, &td->tg_list, tg_node) {
775 776 777 778 779 780 781 782 783 784
		if (!tg->limits_changed)
			continue;

		if (!xchg(&tg->limits_changed, false))
			continue;

		throtl_log_tg(td, tg, "limit change rbps=%llu wbps=%llu"
			" riops=%u wiops=%u", tg->bps[READ], tg->bps[WRITE],
			tg->iops[READ], tg->iops[WRITE]);

785 786 787 788 789 790 791 792 793
		/*
		 * Restart the slices for both READ and WRITES. It
		 * might happen that a group's limit are dropped
		 * suddenly and we don't want to account recently
		 * dispatched IO with new low rate
		 */
		throtl_start_new_slice(td, tg, 0);
		throtl_start_new_slice(td, tg, 1);

794
		if (throtl_tg_on_rr(tg))
795 796 797 798
			tg_update_disptime(td, tg);
	}
}

799 800 801 802 803 804 805
/* Dispatch throttled bios. Should be called without queue lock held. */
static int throtl_dispatch(struct request_queue *q)
{
	struct throtl_data *td = q->td;
	unsigned int nr_disp = 0;
	struct bio_list bio_list_on_stack;
	struct bio *bio;
806
	struct blk_plug plug;
807 808 809

	spin_lock_irq(q->queue_lock);

810 811
	throtl_process_limit_change(td);

812 813 814 815 816
	if (!total_nr_queued(td))
		goto out;

	bio_list_init(&bio_list_on_stack);

817
	throtl_log(td, "dispatch nr_queued=%u read=%u write=%u",
818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834
			total_nr_queued(td), td->nr_queued[READ],
			td->nr_queued[WRITE]);

	nr_disp = throtl_select_dispatch(td, &bio_list_on_stack);

	if (nr_disp)
		throtl_log(td, "bios disp=%u", nr_disp);

	throtl_schedule_next_dispatch(td);
out:
	spin_unlock_irq(q->queue_lock);

	/*
	 * If we dispatched some requests, unplug the queue to make sure
	 * immediate dispatch
	 */
	if (nr_disp) {
835
		blk_start_plug(&plug);
836 837
		while((bio = bio_list_pop(&bio_list_on_stack)))
			generic_make_request(bio);
838
		blk_finish_plug(&plug);
839 840 841 842 843 844 845 846 847 848 849 850 851 852
	}
	return nr_disp;
}

void blk_throtl_work(struct work_struct *work)
{
	struct throtl_data *td = container_of(work, struct throtl_data,
					throtl_work.work);
	struct request_queue *q = td->queue;

	throtl_dispatch(q);
}

/* Call with queue lock held */
853 854
static void
throtl_schedule_delayed_work(struct throtl_data *td, unsigned long delay)
855 856 857 858
{

	struct delayed_work *dwork = &td->throtl_work;

859
	/* schedule work if limits changed even if no bio is queued */
860
	if (total_nr_queued(td) || td->limits_changed) {
861 862 863 864 865
		/*
		 * We might have a work scheduled to be executed in future.
		 * Cancel that and schedule a new one.
		 */
		__cancel_delayed_work(dwork);
866
		queue_delayed_work(kthrotld_workqueue, dwork, delay);
867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887
		throtl_log(td, "schedule work. delay=%lu jiffies=%lu",
				delay, jiffies);
	}
}

static void
throtl_destroy_tg(struct throtl_data *td, struct throtl_grp *tg)
{
	/* Something wrong if we are trying to remove same group twice */
	BUG_ON(hlist_unhashed(&tg->tg_node));

	hlist_del_init(&tg->tg_node);

	/*
	 * Put the reference taken at the time of creation so that when all
	 * queues are gone, group can be destroyed.
	 */
	throtl_put_tg(tg);
	td->nr_undestroyed_grps--;
}

888
static bool throtl_release_tgs(struct throtl_data *td, bool release_root)
889 890 891
{
	struct hlist_node *pos, *n;
	struct throtl_grp *tg;
892
	bool empty = true;
893 894

	hlist_for_each_entry_safe(tg, pos, n, &td->tg_list, tg_node) {
895 896 897 898
		/* skip root? */
		if (!release_root && tg == td->root_tg)
			continue;

899 900 901 902 903 904 905
		/*
		 * 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(&tg->blkg))
			throtl_destroy_tg(td, tg);
906 907
		else
			empty = false;
908
	}
909
	return empty;
910 911 912 913 914 915 916 917
}

/*
 * 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.
 *
918 919 920
 * This function is called under rcu_read_lock(). @q is the rcu protected
 * pointer. That means @q is a valid request_queue pointer as long as we
 * are rcu read lock.
921
 *
922
 * @q was fetched from blkio_group under blkio_cgroup->lock. That means
923 924 925
 * it should not be NULL as even if queue was going away, cgroup deltion
 * path got to it first.
 */
926 927
void throtl_unlink_blkio_group(struct request_queue *q,
			       struct blkio_group *blkg)
928 929 930
{
	unsigned long flags;

931 932 933
	spin_lock_irqsave(q->queue_lock, flags);
	throtl_destroy_tg(q->td, tg_of_blkg(blkg));
	spin_unlock_irqrestore(q->queue_lock, flags);
934 935
}

936 937 938 939 940 941 942 943 944 945 946 947 948 949
static bool throtl_clear_queue(struct request_queue *q)
{
	lockdep_assert_held(q->queue_lock);

	/*
	 * Clear tgs but leave the root one alone.  This is necessary
	 * because root_tg is expected to be persistent and safe because
	 * blk-throtl can never be disabled while @q is alive.  This is a
	 * kludge to prepare for unified blkg.  This whole function will be
	 * removed soon.
	 */
	return throtl_release_tgs(q->td, false);
}

950 951 952 953 954 955 956 957 958
static void throtl_update_blkio_group_common(struct throtl_data *td,
				struct throtl_grp *tg)
{
	xchg(&tg->limits_changed, true);
	xchg(&td->limits_changed, true);
	/* Schedule a work now to process the limit change */
	throtl_schedule_delayed_work(td, 0);
}

959
/*
960
 * For all update functions, @q should be a valid pointer because these
961
 * update functions are called under blkcg_lock, that means, blkg is
962
 * valid and in turn @q is valid. queue exit path can not race because
963 964 965 966 967
 * of blkcg_lock
 *
 * Can not take queue lock in update functions as queue lock under blkcg_lock
 * is not allowed. Under other paths we take blkcg_lock under queue_lock.
 */
968
static void throtl_update_blkio_group_read_bps(struct request_queue *q,
969
				struct blkio_group *blkg, u64 read_bps)
970
{
971
	struct throtl_grp *tg = tg_of_blkg(blkg);
972

973
	tg->bps[READ] = read_bps;
974
	throtl_update_blkio_group_common(q->td, tg);
975 976
}

977
static void throtl_update_blkio_group_write_bps(struct request_queue *q,
978
				struct blkio_group *blkg, u64 write_bps)
979
{
980
	struct throtl_grp *tg = tg_of_blkg(blkg);
981

982
	tg->bps[WRITE] = write_bps;
983
	throtl_update_blkio_group_common(q->td, tg);
984 985
}

986
static void throtl_update_blkio_group_read_iops(struct request_queue *q,
987
			struct blkio_group *blkg, unsigned int read_iops)
988
{
989
	struct throtl_grp *tg = tg_of_blkg(blkg);
990

991
	tg->iops[READ] = read_iops;
992
	throtl_update_blkio_group_common(q->td, tg);
993 994
}

995
static void throtl_update_blkio_group_write_iops(struct request_queue *q,
996
			struct blkio_group *blkg, unsigned int write_iops)
997
{
998
	struct throtl_grp *tg = tg_of_blkg(blkg);
999

1000
	tg->iops[WRITE] = write_iops;
1001
	throtl_update_blkio_group_common(q->td, tg);
1002 1003
}

1004
static void throtl_shutdown_wq(struct request_queue *q)
1005 1006 1007 1008 1009 1010 1011 1012
{
	struct throtl_data *td = q->td;

	cancel_delayed_work_sync(&td->throtl_work);
}

static struct blkio_policy_type blkio_policy_throtl = {
	.ops = {
1013 1014
		.blkio_alloc_group_fn = throtl_alloc_blkio_group,
		.blkio_link_group_fn = throtl_link_blkio_group,
1015
		.blkio_unlink_group_fn = throtl_unlink_blkio_group,
1016
		.blkio_clear_queue_fn = throtl_clear_queue,
1017 1018 1019 1020
		.blkio_update_group_read_bps_fn =
					throtl_update_blkio_group_read_bps,
		.blkio_update_group_write_bps_fn =
					throtl_update_blkio_group_write_bps,
1021 1022 1023 1024
		.blkio_update_group_read_iops_fn =
					throtl_update_blkio_group_read_iops,
		.blkio_update_group_write_iops_fn =
					throtl_update_blkio_group_write_iops,
1025
	},
1026
	.plid = BLKIO_POLICY_THROTL,
1027 1028
};

1029
bool blk_throtl_bio(struct request_queue *q, struct bio *bio)
1030 1031 1032 1033
{
	struct throtl_data *td = q->td;
	struct throtl_grp *tg;
	bool rw = bio_data_dir(bio), update_disptime = true;
1034
	struct blkio_cgroup *blkcg;
1035
	bool throttled = false;
1036 1037 1038

	if (bio->bi_rw & REQ_THROTTLED) {
		bio->bi_rw &= ~REQ_THROTTLED;
1039
		goto out;
1040 1041
	}

1042 1043 1044 1045 1046 1047 1048
	/*
	 * A throtl_grp pointer retrieved under rcu can be used to access
	 * basic fields like stats and io rates. If a group has no rules,
	 * just update the dispatch stats in lockless manner and return.
	 */
	rcu_read_lock();
	blkcg = task_blkio_cgroup(current);
1049
	tg = throtl_lookup_tg(td, blkcg);
1050 1051 1052
	if (tg) {
		if (tg_no_rule_group(tg, rw)) {
			blkiocg_update_dispatch_stats(&tg->blkg, bio->bi_size,
1053
					rw, rw_is_sync(bio->bi_rw));
1054
			goto out_unlock_rcu;
1055 1056 1057 1058 1059 1060 1061
		}
	}

	/*
	 * Either group has not been allocated yet or it is not an unlimited
	 * IO group
	 */
1062
	spin_lock_irq(q->queue_lock);
1063
	tg = throtl_lookup_create_tg(td, blkcg);
1064 1065
	if (unlikely(!tg))
		goto out_unlock;
1066

1067 1068 1069 1070 1071
	if (tg->nr_queued[rw]) {
		/*
		 * There is already another bio queued in same dir. No
		 * need to update dispatch time.
		 */
1072
		update_disptime = false;
1073
		goto queue_bio;
1074

1075 1076 1077 1078 1079
	}

	/* Bio is with-in rate limit of group */
	if (tg_may_dispatch(td, tg, bio, NULL)) {
		throtl_charge_bio(tg, bio);
1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092

		/*
		 * We need to trim slice even when bios are not being queued
		 * otherwise it might happen that a bio is not queued for
		 * a long time and slice keeps on extending and trim is not
		 * called for a long time. Now if limits are reduced suddenly
		 * we take into account all the IO dispatched so far at new
		 * low rate and * newly queued IO gets a really long dispatch
		 * time.
		 *
		 * So keep on trimming slice even if bio is not queued.
		 */
		throtl_trim_slice(td, tg, rw);
1093
		goto out_unlock;
1094 1095 1096
	}

queue_bio:
1097
	throtl_log_tg(td, tg, "[%c] bio. bdisp=%llu sz=%u bps=%llu"
1098 1099
			" iodisp=%u iops=%u queued=%d/%d",
			rw == READ ? 'R' : 'W',
1100
			tg->bytes_disp[rw], bio->bi_size, tg->bps[rw],
1101
			tg->io_disp[rw], tg->iops[rw],
1102 1103 1104
			tg->nr_queued[READ], tg->nr_queued[WRITE]);

	throtl_add_bio_tg(q->td, tg, bio);
1105
	throttled = true;
1106 1107 1108 1109 1110 1111

	if (update_disptime) {
		tg_update_disptime(td, tg);
		throtl_schedule_next_dispatch(td);
	}

1112
out_unlock:
1113
	spin_unlock_irq(q->queue_lock);
1114 1115
out_unlock_rcu:
	rcu_read_unlock();
1116 1117
out:
	return throttled;
1118 1119
}

1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134
/**
 * blk_throtl_drain - drain throttled bios
 * @q: request_queue to drain throttled bios for
 *
 * Dispatch all currently throttled bios on @q through ->make_request_fn().
 */
void blk_throtl_drain(struct request_queue *q)
	__releases(q->queue_lock) __acquires(q->queue_lock)
{
	struct throtl_data *td = q->td;
	struct throtl_rb_root *st = &td->tg_service_tree;
	struct throtl_grp *tg;
	struct bio_list bl;
	struct bio *bio;

1135
	WARN_ON_ONCE(!queue_is_locked(q));
1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154

	bio_list_init(&bl);

	while ((tg = throtl_rb_first(st))) {
		throtl_dequeue_tg(td, tg);

		while ((bio = bio_list_peek(&tg->bio_lists[READ])))
			tg_dispatch_one_bio(td, tg, bio_data_dir(bio), &bl);
		while ((bio = bio_list_peek(&tg->bio_lists[WRITE])))
			tg_dispatch_one_bio(td, tg, bio_data_dir(bio), &bl);
	}
	spin_unlock_irq(q->queue_lock);

	while ((bio = bio_list_pop(&bl)))
		generic_make_request(bio);

	spin_lock_irq(q->queue_lock);
}

1155 1156 1157
int blk_throtl_init(struct request_queue *q)
{
	struct throtl_data *td;
1158
	struct blkio_group *blkg;
1159 1160 1161 1162 1163 1164 1165

	td = kzalloc_node(sizeof(*td), GFP_KERNEL, q->node);
	if (!td)
		return -ENOMEM;

	INIT_HLIST_HEAD(&td->tg_list);
	td->tg_service_tree = THROTL_RB_ROOT;
1166
	td->limits_changed = false;
1167
	INIT_DELAYED_WORK(&td->throtl_work, blk_throtl_work);
1168

1169
	q->td = td;
1170
	td->queue = q;
V
Vivek Goyal 已提交
1171

1172
	/* alloc and init root group. */
1173 1174
	rcu_read_lock();
	spin_lock_irq(q->queue_lock);
1175

1176 1177 1178 1179
	blkg = blkg_lookup_create(&blkio_root_cgroup, q, BLKIO_POLICY_THROTL,
				  true);
	if (!IS_ERR(blkg))
		td->root_tg = tg_of_blkg(blkg);
1180

1181
	spin_unlock_irq(q->queue_lock);
1182 1183
	rcu_read_unlock();

1184 1185 1186 1187
	if (!td->root_tg) {
		kfree(td);
		return -ENOMEM;
	}
1188 1189 1190 1191 1192 1193 1194 1195 1196 1197
	return 0;
}

void blk_throtl_exit(struct request_queue *q)
{
	struct throtl_data *td = q->td;
	bool wait = false;

	BUG_ON(!td);

1198
	throtl_shutdown_wq(q);
1199 1200

	spin_lock_irq(q->queue_lock);
1201
	throtl_release_tgs(td, true);
1202 1203

	/* If there are other groups */
V
Vivek Goyal 已提交
1204
	if (td->nr_undestroyed_grps > 0)
1205 1206 1207 1208 1209
		wait = true;

	spin_unlock_irq(q->queue_lock);

	/*
1210
	 * Wait for tg->blkg->q accessors to exit their grace periods.
1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221
	 * Do this wait only if there are other undestroyed groups out
	 * there (other than root group). This can happen if cgroup deletion
	 * path claimed the responsibility of cleaning up a group before
	 * queue cleanup code get to the group.
	 *
	 * Do not call synchronize_rcu() unconditionally as there are drivers
	 * which create/delete request queue hundreds of times during scan/boot
	 * and synchronize_rcu() can take significant time and slow down boot.
	 */
	if (wait)
		synchronize_rcu();
1222 1223 1224 1225 1226 1227

	/*
	 * Just being safe to make sure after previous flush if some body did
	 * update limits through cgroup and another work got queued, cancel
	 * it.
	 */
1228
	throtl_shutdown_wq(q);
1229 1230 1231 1232 1233
}

void blk_throtl_release(struct request_queue *q)
{
	kfree(q->td);
1234 1235 1236 1237
}

static int __init throtl_init(void)
{
1238 1239 1240 1241
	kthrotld_workqueue = alloc_workqueue("kthrotld", WQ_MEM_RECLAIM, 0);
	if (!kthrotld_workqueue)
		panic("Failed to create kthrotld\n");

1242 1243 1244 1245 1246
	blkio_policy_register(&blkio_policy_throtl);
	return 0;
}

module_init(throtl_init);