blk-throttle.c 30.9 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
static struct blkio_policy_type blkio_policy_throtl;

26 27 28 29 30
/* 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);

31 32 33 34 35 36 37 38 39 40 41 42
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)

43 44 45 46 47 48 49 50
/* Per-cpu group stats */
struct tg_stats_cpu {
	/* total bytes transferred */
	struct blkg_rwstat		service_bytes;
	/* total IOs serviced, post merge */
	struct blkg_rwstat		serviced;
};

51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72
struct throtl_grp {
	/* 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;

	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];

73 74 75
	/* IOPS limits */
	unsigned int iops[2];

76 77
	/* Number of bytes disptached in current slice */
	uint64_t bytes_disp[2];
78 79
	/* Number of bio's dispatched in current slice */
	unsigned int io_disp[2];
80 81 82 83

	/* When did we start a new slice */
	unsigned long slice_start[2];
	unsigned long slice_end[2];
84 85

	/* Some throttle limits got updated for the group */
86
	int limits_changed;
87 88 89 90 91 92

	/* Per cpu stats pointer */
	struct tg_stats_cpu __percpu *stats_cpu;

	/* List of tgs waiting for per cpu stats memory to be allocated */
	struct list_head stats_alloc_node;
93 94 95 96 97 98 99 100 101 102 103 104 105
};

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

	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
106
	 * number of total undestroyed groups
107 108 109 110 111
	 */
	unsigned int nr_undestroyed_grps;

	/* Work for dispatching throttled bios */
	struct delayed_work throtl_work;
112

113
	int limits_changed;
114 115
};

116 117 118 119 120 121 122
/* list and work item to allocate percpu group stats */
static DEFINE_SPINLOCK(tg_stats_alloc_lock);
static LIST_HEAD(tg_stats_alloc_list);

static void tg_stats_alloc_fn(struct work_struct *);
static DECLARE_DELAYED_WORK(tg_stats_alloc_work, tg_stats_alloc_fn);

123 124 125 126 127 128 129
static inline struct throtl_grp *blkg_to_tg(struct blkio_group *blkg)
{
	return blkg_to_pdata(blkg, &blkio_policy_throtl);
}

static inline struct blkio_group *tg_to_blkg(struct throtl_grp *tg)
{
130
	return pdata_to_blkg(tg);
131 132
}

T
Tejun Heo 已提交
133 134 135 136 137
static inline struct throtl_grp *td_root_tg(struct throtl_data *td)
{
	return blkg_to_tg(td->queue->root_blkg);
}

138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159
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,		\
160
			  blkg_path(tg_to_blkg(tg)), ##args);		\
161 162 163 164

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

165
static inline unsigned int total_nr_queued(struct throtl_data *td)
166
{
167
	return td->nr_queued[0] + td->nr_queued[1];
168 169
}

170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207
/*
 * Worker for allocating per cpu stat for tgs. This is scheduled on the
 * system_nrt_wq once there are some groups on the alloc_list waiting for
 * allocation.
 */
static void tg_stats_alloc_fn(struct work_struct *work)
{
	static struct tg_stats_cpu *stats_cpu;	/* this fn is non-reentrant */
	struct delayed_work *dwork = to_delayed_work(work);
	bool empty = false;

alloc_stats:
	if (!stats_cpu) {
		stats_cpu = alloc_percpu(struct tg_stats_cpu);
		if (!stats_cpu) {
			/* allocation failed, try again after some time */
			queue_delayed_work(system_nrt_wq, dwork,
					   msecs_to_jiffies(10));
			return;
		}
	}

	spin_lock_irq(&tg_stats_alloc_lock);

	if (!list_empty(&tg_stats_alloc_list)) {
		struct throtl_grp *tg = list_first_entry(&tg_stats_alloc_list,
							 struct throtl_grp,
							 stats_alloc_node);
		swap(tg->stats_cpu, stats_cpu);
		list_del_init(&tg->stats_alloc_node);
	}

	empty = list_empty(&tg_stats_alloc_list);
	spin_unlock_irq(&tg_stats_alloc_lock);
	if (!empty)
		goto alloc_stats;
}

208
static void throtl_init_blkio_group(struct blkio_group *blkg)
209
{
210
	struct throtl_grp *tg = blkg_to_tg(blkg);
211

212 213 214 215 216
	RB_CLEAR_NODE(&tg->rb_node);
	bio_list_init(&tg->bio_lists[0]);
	bio_list_init(&tg->bio_lists[1]);
	tg->limits_changed = false;

217 218 219 220
	tg->bps[READ] = -1;
	tg->bps[WRITE] = -1;
	tg->iops[READ] = -1;
	tg->iops[WRITE] = -1;
221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257

	/*
	 * Ugh... We need to perform per-cpu allocation for tg->stats_cpu
	 * but percpu allocator can't be called from IO path.  Queue tg on
	 * tg_stats_alloc_list and allocate from work item.
	 */
	spin_lock(&tg_stats_alloc_lock);
	list_add(&tg->stats_alloc_node, &tg_stats_alloc_list);
	queue_delayed_work(system_nrt_wq, &tg_stats_alloc_work, 0);
	spin_unlock(&tg_stats_alloc_lock);
}

static void throtl_exit_blkio_group(struct blkio_group *blkg)
{
	struct throtl_grp *tg = blkg_to_tg(blkg);

	spin_lock(&tg_stats_alloc_lock);
	list_del_init(&tg->stats_alloc_node);
	spin_unlock(&tg_stats_alloc_lock);

	free_percpu(tg->stats_cpu);
}

static void throtl_reset_group_stats(struct blkio_group *blkg)
{
	struct throtl_grp *tg = blkg_to_tg(blkg);
	int cpu;

	if (tg->stats_cpu == NULL)
		return;

	for_each_possible_cpu(cpu) {
		struct tg_stats_cpu *sc = per_cpu_ptr(tg->stats_cpu, cpu);

		blkg_rwstat_reset(&sc->service_bytes);
		blkg_rwstat_reset(&sc->serviced);
	}
258 259
}

260
static struct
261
throtl_grp *throtl_lookup_tg(struct throtl_data *td, struct blkio_cgroup *blkcg)
262
{
263 264
	/*
	 * This is the common case when there are no blkio cgroups.
265 266
	 * Avoid lookup in this case
	 */
267
	if (blkcg == &blkio_root_cgroup)
T
Tejun Heo 已提交
268
		return td_root_tg(td);
269

270
	return blkg_to_tg(blkg_lookup(blkcg, td->queue));
271 272
}

273 274
static struct throtl_grp *throtl_lookup_create_tg(struct throtl_data *td,
						  struct blkio_cgroup *blkcg)
275
{
276
	struct request_queue *q = td->queue;
277
	struct throtl_grp *tg = NULL;
278

279
	/*
280 281
	 * This is the common case when there are no blkio cgroups.
	 * Avoid lookup in this case
282
	 */
283
	if (blkcg == &blkio_root_cgroup) {
T
Tejun Heo 已提交
284
		tg = td_root_tg(td);
285 286
	} else {
		struct blkio_group *blkg;
287

288
		blkg = blkg_lookup_create(blkcg, q, false);
289

290 291
		/* if %NULL and @q is alive, fall back to root_tg */
		if (!IS_ERR(blkg))
292
			tg = blkg_to_tg(blkg);
293
		else if (!blk_queue_dead(q))
T
Tejun Heo 已提交
294
			tg = td_root_tg(td);
295 296
	}

297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409
	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))
410
		throtl_schedule_delayed_work(td, 0);
411
	else
412
		throtl_schedule_delayed_work(td, (st->min_disptime - jiffies));
413 414 415 416 417 418
}

static inline void
throtl_start_new_slice(struct throtl_data *td, struct throtl_grp *tg, bool rw)
{
	tg->bytes_disp[rw] = 0;
419
	tg->io_disp[rw] = 0;
420 421 422 423 424 425 426
	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);
}

427 428 429 430 431 432
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);
}

433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455
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)
{
456 457
	unsigned long nr_slices, time_elapsed, io_trim;
	u64 bytes_trim, tmp;
458 459 460 461 462 463 464 465 466 467 468

	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;

469 470 471 472 473 474 475 476 477 478
	/*
	 * 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);

479 480 481 482 483 484
	time_elapsed = jiffies - tg->slice_start[rw];

	nr_slices = time_elapsed / throtl_slice;

	if (!nr_slices)
		return;
485 486 487
	tmp = tg->bps[rw] * throtl_slice * nr_slices;
	do_div(tmp, HZ);
	bytes_trim = tmp;
488

489
	io_trim = (tg->iops[rw] * throtl_slice * nr_slices)/HZ;
490

491
	if (!bytes_trim && !io_trim)
492 493 494 495 496 497 498
		return;

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

499 500 501 502 503
	if (tg->io_disp[rw] >= io_trim)
		tg->io_disp[rw] -= io_trim;
	else
		tg->io_disp[rw] = 0;

504 505
	tg->slice_start[rw] += nr_slices * throtl_slice;

506
	throtl_log_tg(td, tg, "[%c] trim slice nr=%lu bytes=%llu io=%lu"
507
			" start=%lu end=%lu jiffies=%lu",
508
			rw == READ ? 'R' : 'W', nr_slices, bytes_trim, io_trim,
509 510 511
			tg->slice_start[rw], tg->slice_end[rw], jiffies);
}

512 513
static bool tg_with_in_iops_limit(struct throtl_data *td, struct throtl_grp *tg,
		struct bio *bio, unsigned long *wait)
514 515
{
	bool rw = bio_data_dir(bio);
516
	unsigned int io_allowed;
517
	unsigned long jiffy_elapsed, jiffy_wait, jiffy_elapsed_rnd;
518
	u64 tmp;
519

520
	jiffy_elapsed = jiffy_elapsed_rnd = jiffies - tg->slice_start[rw];
521

522 523 524 525 526 527
	/* 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);

528 529 530 531 532 533 534 535 536 537 538 539 540 541
	/*
	 * 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;
542 543

	if (tg->io_disp[rw] + 1 <= io_allowed) {
544 545 546 547 548
		if (wait)
			*wait = 0;
		return 1;
	}

549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565
	/* 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);
566
	u64 bytes_allowed, extra_bytes, tmp;
567
	unsigned long jiffy_elapsed, jiffy_wait, jiffy_elapsed_rnd;
568 569 570 571 572 573 574 575 576

	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);

577 578
	tmp = tg->bps[rw] * jiffy_elapsed_rnd;
	do_div(tmp, HZ);
579
	bytes_allowed = tmp;
580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600

	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;
601 602 603
	return 0;
}

604 605 606 607 608 609
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;
}

610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626
/*
 * 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]));
627

628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660
	/* 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);
661 662 663 664

	return 0;
}

665 666 667
static void throtl_update_dispatch_stats(struct blkio_group *blkg, u64 bytes,
					 int rw)
{
668 669
	struct throtl_grp *tg = blkg_to_tg(blkg);
	struct tg_stats_cpu *stats_cpu;
670 671 672
	unsigned long flags;

	/* If per cpu stats are not allocated yet, don't do any accounting. */
673
	if (tg->stats_cpu == NULL)
674 675 676 677 678 679 680 681 682
		return;

	/*
	 * Disabling interrupts to provide mutual exclusion between two
	 * writes on same cpu. It probably is not needed for 64bit. Not
	 * optimizing that case yet.
	 */
	local_irq_save(flags);

683
	stats_cpu = this_cpu_ptr(tg->stats_cpu);
684 685 686 687 688 689 690

	blkg_rwstat_add(&stats_cpu->serviced, rw, 1);
	blkg_rwstat_add(&stats_cpu->service_bytes, rw, bytes);

	local_irq_restore(flags);
}

691 692 693 694 695 696
static void throtl_charge_bio(struct throtl_grp *tg, struct bio *bio)
{
	bool rw = bio_data_dir(bio);

	/* Charge the bio to the group */
	tg->bytes_disp[rw] += bio->bi_size;
697
	tg->io_disp[rw]++;
698

699
	throtl_update_dispatch_stats(tg_to_blkg(tg), bio->bi_size, bio->bi_rw);
700 701 702 703 704 705 706 707 708
}

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 */
T
Tejun Heo 已提交
709
	blkg_get(tg_to_blkg(tg));
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
	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]--;
T
Tejun Heo 已提交
742 743
	/* Drop bio reference on blkg */
	blkg_put(tg_to_blkg(tg));
744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759

	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;
760
	unsigned int max_nr_writes = throtl_grp_quantum - max_nr_reads;
761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818
	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;
}

819 820
static void throtl_process_limit_change(struct throtl_data *td)
{
821 822
	struct request_queue *q = td->queue;
	struct blkio_group *blkg, *n;
823

824
	if (!td->limits_changed)
825 826
		return;

827
	xchg(&td->limits_changed, false);
828

829
	throtl_log(td, "limits changed");
830

831
	list_for_each_entry_safe(blkg, n, &q->blkg_list, q_node) {
832 833
		struct throtl_grp *tg = blkg_to_tg(blkg);

834 835 836 837 838 839 840 841 842 843
		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]);

844 845 846 847 848 849 850 851 852
		/*
		 * 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);

853
		if (throtl_tg_on_rr(tg))
854 855 856 857
			tg_update_disptime(td, tg);
	}
}

858 859 860 861 862 863 864
/* 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;
865
	struct blk_plug plug;
866 867 868

	spin_lock_irq(q->queue_lock);

869 870
	throtl_process_limit_change(td);

871 872 873 874 875
	if (!total_nr_queued(td))
		goto out;

	bio_list_init(&bio_list_on_stack);

876
	throtl_log(td, "dispatch nr_queued=%u read=%u write=%u",
877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893
			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) {
894
		blk_start_plug(&plug);
895 896
		while((bio = bio_list_pop(&bio_list_on_stack)))
			generic_make_request(bio);
897
		blk_finish_plug(&plug);
898 899 900 901 902 903 904 905 906 907 908 909 910 911
	}
	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 */
912 913
static void
throtl_schedule_delayed_work(struct throtl_data *td, unsigned long delay)
914 915 916 917
{

	struct delayed_work *dwork = &td->throtl_work;

918
	/* schedule work if limits changed even if no bio is queued */
919
	if (total_nr_queued(td) || td->limits_changed) {
920 921 922 923 924
		/*
		 * We might have a work scheduled to be executed in future.
		 * Cancel that and schedule a new one.
		 */
		__cancel_delayed_work(dwork);
925
		queue_delayed_work(kthrotld_workqueue, dwork, delay);
926 927 928 929 930
		throtl_log(td, "schedule work. delay=%lu jiffies=%lu",
				delay, jiffies);
	}
}

931
static u64 tg_prfill_cpu_rwstat(struct seq_file *sf, void *pdata, int off)
932
{
933
	struct throtl_grp *tg = pdata;
934 935 936 937
	struct blkg_rwstat rwstat = { }, tmp;
	int i, cpu;

	for_each_possible_cpu(cpu) {
938
		struct tg_stats_cpu *sc = per_cpu_ptr(tg->stats_cpu, cpu);
939 940 941 942 943 944

		tmp = blkg_rwstat_read((void *)sc + off);
		for (i = 0; i < BLKG_RWSTAT_NR; i++)
			rwstat.cnt[i] += tmp.cnt[i];
	}

945
	return __blkg_prfill_rwstat(sf, pdata, &rwstat);
946 947
}

948 949
static int tg_print_cpu_rwstat(struct cgroup *cgrp, struct cftype *cft,
			       struct seq_file *sf)
950 951 952
{
	struct blkio_cgroup *blkcg = cgroup_to_blkio_cgroup(cgrp);

953
	blkcg_print_blkgs(sf, blkcg, tg_prfill_cpu_rwstat, &blkio_policy_throtl,
954
			  cft->private, true);
955 956 957
	return 0;
}

958
static u64 tg_prfill_conf_u64(struct seq_file *sf, void *pdata, int off)
959
{
960
	u64 v = *(u64 *)(pdata + off);
961

962
	if (v == -1)
963
		return 0;
964
	return __blkg_prfill_u64(sf, pdata, v);
965 966
}

967
static u64 tg_prfill_conf_uint(struct seq_file *sf, void *pdata, int off)
968
{
969
	unsigned int v = *(unsigned int *)(pdata + off);
970

971 972
	if (v == -1)
		return 0;
973
	return __blkg_prfill_u64(sf, pdata, v);
974 975
}

976 977
static int tg_print_conf_u64(struct cgroup *cgrp, struct cftype *cft,
			     struct seq_file *sf)
978
{
979
	blkcg_print_blkgs(sf, cgroup_to_blkio_cgroup(cgrp), tg_prfill_conf_u64,
980
			  &blkio_policy_throtl, cft->private, false);
981
	return 0;
982 983
}

984 985
static int tg_print_conf_uint(struct cgroup *cgrp, struct cftype *cft,
			      struct seq_file *sf)
986
{
987
	blkcg_print_blkgs(sf, cgroup_to_blkio_cgroup(cgrp), tg_prfill_conf_uint,
988
			  &blkio_policy_throtl, cft->private, false);
989
	return 0;
990 991
}

992 993
static int tg_set_conf(struct cgroup *cgrp, struct cftype *cft, const char *buf,
		       bool is_u64)
994 995 996
{
	struct blkio_cgroup *blkcg = cgroup_to_blkio_cgroup(cgrp);
	struct blkg_conf_ctx ctx;
997
	struct throtl_grp *tg;
998 999
	int ret;

1000
	ret = blkg_conf_prep(blkcg, &blkio_policy_throtl, buf, &ctx);
1001 1002 1003 1004
	if (ret)
		return ret;

	ret = -EINVAL;
1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021
	tg = blkg_to_tg(ctx.blkg);
	if (tg) {
		struct throtl_data *td = ctx.blkg->q->td;

		if (!ctx.v)
			ctx.v = -1;

		if (is_u64)
			*(u64 *)((void *)tg + cft->private) = ctx.v;
		else
			*(unsigned int *)((void *)tg + cft->private) = ctx.v;

		/* XXX: we don't need the following deferred processing */
		xchg(&tg->limits_changed, true);
		xchg(&td->limits_changed, true);
		throtl_schedule_delayed_work(td, 0);

1022 1023 1024 1025 1026
		ret = 0;
	}

	blkg_conf_finish(&ctx);
	return ret;
1027 1028
}

1029 1030
static int tg_set_conf_u64(struct cgroup *cgrp, struct cftype *cft,
			   const char *buf)
1031
{
1032
	return tg_set_conf(cgrp, cft, buf, true);
1033 1034
}

1035 1036
static int tg_set_conf_uint(struct cgroup *cgrp, struct cftype *cft,
			    const char *buf)
1037
{
1038
	return tg_set_conf(cgrp, cft, buf, false);
1039 1040 1041 1042 1043
}

static struct cftype throtl_files[] = {
	{
		.name = "throttle.read_bps_device",
1044 1045 1046
		.private = offsetof(struct throtl_grp, bps[READ]),
		.read_seq_string = tg_print_conf_u64,
		.write_string = tg_set_conf_u64,
1047 1048 1049 1050
		.max_write_len = 256,
	},
	{
		.name = "throttle.write_bps_device",
1051 1052 1053
		.private = offsetof(struct throtl_grp, bps[WRITE]),
		.read_seq_string = tg_print_conf_u64,
		.write_string = tg_set_conf_u64,
1054 1055 1056 1057
		.max_write_len = 256,
	},
	{
		.name = "throttle.read_iops_device",
1058 1059 1060
		.private = offsetof(struct throtl_grp, iops[READ]),
		.read_seq_string = tg_print_conf_uint,
		.write_string = tg_set_conf_uint,
1061 1062 1063 1064
		.max_write_len = 256,
	},
	{
		.name = "throttle.write_iops_device",
1065 1066 1067
		.private = offsetof(struct throtl_grp, iops[WRITE]),
		.read_seq_string = tg_print_conf_uint,
		.write_string = tg_set_conf_uint,
1068 1069 1070 1071
		.max_write_len = 256,
	},
	{
		.name = "throttle.io_service_bytes",
1072
		.private = offsetof(struct tg_stats_cpu, service_bytes),
1073
		.read_seq_string = tg_print_cpu_rwstat,
1074 1075 1076
	},
	{
		.name = "throttle.io_serviced",
1077
		.private = offsetof(struct tg_stats_cpu, serviced),
1078
		.read_seq_string = tg_print_cpu_rwstat,
1079 1080 1081 1082
	},
	{ }	/* terminate */
};

1083
static void throtl_shutdown_wq(struct request_queue *q)
1084 1085 1086 1087 1088 1089 1090 1091
{
	struct throtl_data *td = q->td;

	cancel_delayed_work_sync(&td->throtl_work);
}

static struct blkio_policy_type blkio_policy_throtl = {
	.ops = {
1092
		.blkio_init_group_fn = throtl_init_blkio_group,
1093 1094
		.blkio_exit_group_fn = throtl_exit_blkio_group,
		.blkio_reset_group_stats_fn = throtl_reset_group_stats,
1095
	},
1096
	.pdata_size = sizeof(struct throtl_grp),
1097
	.cftypes = throtl_files,
1098 1099
};

1100
bool blk_throtl_bio(struct request_queue *q, struct bio *bio)
1101 1102 1103 1104
{
	struct throtl_data *td = q->td;
	struct throtl_grp *tg;
	bool rw = bio_data_dir(bio), update_disptime = true;
1105
	struct blkio_cgroup *blkcg;
1106
	bool throttled = false;
1107 1108 1109

	if (bio->bi_rw & REQ_THROTTLED) {
		bio->bi_rw &= ~REQ_THROTTLED;
1110
		goto out;
1111 1112
	}

1113 1114 1115
	/* bio_associate_current() needs ioc, try creating */
	create_io_context(GFP_ATOMIC, q->node);

1116 1117 1118 1119 1120 1121
	/*
	 * 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();
1122
	blkcg = bio_blkio_cgroup(bio);
1123
	tg = throtl_lookup_tg(td, blkcg);
1124 1125
	if (tg) {
		if (tg_no_rule_group(tg, rw)) {
1126 1127
			throtl_update_dispatch_stats(tg_to_blkg(tg),
						     bio->bi_size, bio->bi_rw);
1128
			goto out_unlock_rcu;
1129 1130 1131 1132 1133 1134 1135
		}
	}

	/*
	 * Either group has not been allocated yet or it is not an unlimited
	 * IO group
	 */
1136
	spin_lock_irq(q->queue_lock);
1137
	tg = throtl_lookup_create_tg(td, blkcg);
1138 1139
	if (unlikely(!tg))
		goto out_unlock;
1140

1141 1142 1143 1144 1145
	if (tg->nr_queued[rw]) {
		/*
		 * There is already another bio queued in same dir. No
		 * need to update dispatch time.
		 */
1146
		update_disptime = false;
1147
		goto queue_bio;
1148

1149 1150 1151 1152 1153
	}

	/* Bio is with-in rate limit of group */
	if (tg_may_dispatch(td, tg, bio, NULL)) {
		throtl_charge_bio(tg, bio);
1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166

		/*
		 * 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);
1167
		goto out_unlock;
1168 1169 1170
	}

queue_bio:
1171
	throtl_log_tg(td, tg, "[%c] bio. bdisp=%llu sz=%u bps=%llu"
1172 1173
			" iodisp=%u iops=%u queued=%d/%d",
			rw == READ ? 'R' : 'W',
1174
			tg->bytes_disp[rw], bio->bi_size, tg->bps[rw],
1175
			tg->io_disp[rw], tg->iops[rw],
1176 1177
			tg->nr_queued[READ], tg->nr_queued[WRITE]);

1178
	bio_associate_current(bio);
1179
	throtl_add_bio_tg(q->td, tg, bio);
1180
	throttled = true;
1181 1182 1183 1184 1185 1186

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

1187
out_unlock:
1188
	spin_unlock_irq(q->queue_lock);
1189 1190
out_unlock_rcu:
	rcu_read_unlock();
1191 1192
out:
	return throttled;
1193 1194
}

1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209
/**
 * 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;

1210
	WARN_ON_ONCE(!queue_is_locked(q));
1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229

	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);
}

1230 1231 1232
int blk_throtl_init(struct request_queue *q)
{
	struct throtl_data *td;
1233
	struct blkio_group *blkg;
1234 1235 1236 1237 1238 1239

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

	td->tg_service_tree = THROTL_RB_ROOT;
1240
	td->limits_changed = false;
1241
	INIT_DELAYED_WORK(&td->throtl_work, blk_throtl_work);
1242

1243
	q->td = td;
1244
	td->queue = q;
V
Vivek Goyal 已提交
1245

1246
	/* alloc and init root group. */
1247 1248
	rcu_read_lock();
	spin_lock_irq(q->queue_lock);
1249

1250
	blkg = blkg_lookup_create(&blkio_root_cgroup, q, true);
1251
	if (!IS_ERR(blkg))
T
Tejun Heo 已提交
1252
		q->root_blkg = blkg;
1253

1254
	spin_unlock_irq(q->queue_lock);
1255 1256
	rcu_read_unlock();

T
Tejun Heo 已提交
1257
	if (!q->root_blkg) {
1258 1259 1260
		kfree(td);
		return -ENOMEM;
	}
1261 1262 1263 1264 1265
	return 0;
}

void blk_throtl_exit(struct request_queue *q)
{
T
Tejun Heo 已提交
1266
	BUG_ON(!q->td);
1267
	throtl_shutdown_wq(q);
1268
	kfree(q->td);
1269 1270 1271 1272
}

static int __init throtl_init(void)
{
1273 1274 1275 1276
	kthrotld_workqueue = alloc_workqueue("kthrotld", WQ_MEM_RECLAIM, 0);
	if (!kthrotld_workqueue)
		panic("Failed to create kthrotld\n");

T
Tejun Heo 已提交
1277
	return blkio_policy_register(&blkio_policy_throtl);
1278 1279 1280
}

module_init(throtl_init);