blk-throttle.c 53.5 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11
/*
 * 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>
12
#include <linux/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 */

T
Tejun Heo 已提交
24
static struct blkcg_policy blkcg_policy_throtl;
25

26 27 28
/* A workqueue to queue throttle related work */
static struct workqueue_struct *kthrotld_workqueue;

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
/*
 * To implement hierarchical throttling, throtl_grps form a tree and bios
 * are dispatched upwards level by level until they reach the top and get
 * issued.  When dispatching bios from the children and local group at each
 * level, if the bios are dispatched into a single bio_list, there's a risk
 * of a local or child group which can queue many bios at once filling up
 * the list starving others.
 *
 * To avoid such starvation, dispatched bios are queued separately
 * according to where they came from.  When they are again dispatched to
 * the parent, they're popped in round-robin order so that no single source
 * hogs the dispatch window.
 *
 * throtl_qnode is used to keep the queued bios separated by their sources.
 * Bios are queued to throtl_qnode which in turn is queued to
 * throtl_service_queue and then dispatched in round-robin order.
 *
 * It's also used to track the reference counts on blkg's.  A qnode always
 * belongs to a throtl_grp and gets queued on itself or the parent, so
 * incrementing the reference of the associated throtl_grp when a qnode is
 * queued and decrementing when dequeued is enough to keep the whole blkg
 * tree pinned while bios are in flight.
 */
struct throtl_qnode {
	struct list_head	node;		/* service_queue->queued[] */
	struct bio_list		bios;		/* queued bios */
	struct throtl_grp	*tg;		/* tg this qnode belongs to */
};

58
struct throtl_service_queue {
59 60
	struct throtl_service_queue *parent_sq;	/* the parent service_queue */

61 62 63 64
	/*
	 * Bios queued directly to this service_queue or dispatched from
	 * children throtl_grp's.
	 */
65
	struct list_head	queued[2];	/* throtl_qnode [READ/WRITE] */
66 67 68 69 70 71
	unsigned int		nr_queued[2];	/* number of queued bios */

	/*
	 * RB tree of active children throtl_grp's, which are sorted by
	 * their ->disptime.
	 */
72 73 74 75
	struct rb_root		pending_tree;	/* RB tree of active tgs */
	struct rb_node		*first_pending;	/* first node in the tree */
	unsigned int		nr_pending;	/* # queued in the tree */
	unsigned long		first_pending_disptime;	/* disptime of the first tg */
76
	struct timer_list	pending_timer;	/* fires on first_pending_disptime */
77 78
};

79 80
enum tg_state_flags {
	THROTL_TG_PENDING	= 1 << 0,	/* on parent's pending tree */
81
	THROTL_TG_WAS_EMPTY	= 1 << 1,	/* bio_lists[] became non-empty */
82 83
};

84 85
#define rb_entry_tg(node)	rb_entry((node), struct throtl_grp, rb_node)

86
enum {
S
Shaohua Li 已提交
87
	LIMIT_LOW,
88 89 90 91
	LIMIT_MAX,
	LIMIT_CNT,
};

92
struct throtl_grp {
93 94 95
	/* must be the first member */
	struct blkg_policy_data pd;

96
	/* active throtl group service_queue member */
97 98
	struct rb_node rb_node;

99 100 101
	/* throtl_data this group belongs to */
	struct throtl_data *td;

102 103 104
	/* this group's service queue */
	struct throtl_service_queue service_queue;

105 106 107 108 109 110 111 112 113 114 115
	/*
	 * qnode_on_self is used when bios are directly queued to this
	 * throtl_grp so that local bios compete fairly with bios
	 * dispatched from children.  qnode_on_parent is used when bios are
	 * dispatched from this throtl_grp into its parent and will compete
	 * with the sibling qnode_on_parents and the parent's
	 * qnode_on_self.
	 */
	struct throtl_qnode qnode_on_self[2];
	struct throtl_qnode qnode_on_parent[2];

116 117 118 119 120 121 122 123 124
	/*
	 * 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;

125 126 127
	/* are there any throtl rules between this group and td? */
	bool has_rules[2];

S
Shaohua Li 已提交
128
	/* internally used bytes per second rate limits */
129
	uint64_t bps[2][LIMIT_CNT];
S
Shaohua Li 已提交
130 131
	/* user configured bps limits */
	uint64_t bps_conf[2][LIMIT_CNT];
132

S
Shaohua Li 已提交
133
	/* internally used IOPS limits */
134
	unsigned int iops[2][LIMIT_CNT];
S
Shaohua Li 已提交
135 136
	/* user configured IOPS limits */
	unsigned int iops_conf[2][LIMIT_CNT];
137

138 139
	/* Number of bytes disptached in current slice */
	uint64_t bytes_disp[2];
140 141
	/* Number of bio's dispatched in current slice */
	unsigned int io_disp[2];
142

S
Shaohua Li 已提交
143 144 145 146 147 148 149
	unsigned long last_low_overflow_time[2];

	uint64_t last_bytes_disp[2];
	unsigned int last_io_disp[2];

	unsigned long last_check_time;

150 151 152 153 154 155 156 157
	/* When did we start a new slice */
	unsigned long slice_start[2];
	unsigned long slice_end[2];
};

struct throtl_data
{
	/* service tree for active throtl groups */
158
	struct throtl_service_queue service_queue;
159 160 161 162 163 164 165

	struct request_queue *queue;

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

	/* Work for dispatching throttled bios */
166
	struct work_struct dispatch_work;
167 168
	unsigned int limit_index;
	bool limit_valid[LIMIT_CNT];
S
Shaohua Li 已提交
169 170 171

	unsigned long low_upgrade_time;
	unsigned long low_downgrade_time;
172 173
};

174 175
static void throtl_pending_timer_fn(unsigned long arg);

176 177 178 179 180
static inline struct throtl_grp *pd_to_tg(struct blkg_policy_data *pd)
{
	return pd ? container_of(pd, struct throtl_grp, pd) : NULL;
}

T
Tejun Heo 已提交
181
static inline struct throtl_grp *blkg_to_tg(struct blkcg_gq *blkg)
182
{
183
	return pd_to_tg(blkg_to_pd(blkg, &blkcg_policy_throtl));
184 185
}

T
Tejun Heo 已提交
186
static inline struct blkcg_gq *tg_to_blkg(struct throtl_grp *tg)
187
{
188
	return pd_to_blkg(&tg->pd);
189 190
}

191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209
/**
 * sq_to_tg - return the throl_grp the specified service queue belongs to
 * @sq: the throtl_service_queue of interest
 *
 * Return the throtl_grp @sq belongs to.  If @sq is the top-level one
 * embedded in throtl_data, %NULL is returned.
 */
static struct throtl_grp *sq_to_tg(struct throtl_service_queue *sq)
{
	if (sq && sq->parent_sq)
		return container_of(sq, struct throtl_grp, service_queue);
	else
		return NULL;
}

/**
 * sq_to_td - return throtl_data the specified service queue belongs to
 * @sq: the throtl_service_queue of interest
 *
210
 * A service_queue can be embedded in either a throtl_grp or throtl_data.
211 212 213 214 215 216 217 218 219 220 221 222
 * Determine the associated throtl_data accordingly and return it.
 */
static struct throtl_data *sq_to_td(struct throtl_service_queue *sq)
{
	struct throtl_grp *tg = sq_to_tg(sq);

	if (tg)
		return tg->td;
	else
		return container_of(sq, struct throtl_data, service_queue);
}

223 224
static uint64_t tg_bps_limit(struct throtl_grp *tg, int rw)
{
225 226 227 228 229 230 231 232 233
	struct blkcg_gq *blkg = tg_to_blkg(tg);
	uint64_t ret;

	if (cgroup_subsys_on_dfl(io_cgrp_subsys) && !blkg->parent)
		return U64_MAX;
	ret = tg->bps[rw][tg->td->limit_index];
	if (ret == 0 && tg->td->limit_index == LIMIT_LOW)
		return tg->bps[rw][LIMIT_MAX];
	return ret;
234 235 236 237
}

static unsigned int tg_iops_limit(struct throtl_grp *tg, int rw)
{
238 239 240 241 242 243 244 245 246
	struct blkcg_gq *blkg = tg_to_blkg(tg);
	unsigned int ret;

	if (cgroup_subsys_on_dfl(io_cgrp_subsys) && !blkg->parent)
		return UINT_MAX;
	ret = tg->iops[rw][tg->td->limit_index];
	if (ret == 0 && tg->td->limit_index == LIMIT_LOW)
		return tg->iops[rw][LIMIT_MAX];
	return ret;
247 248
}

249 250 251 252 253 254 255 256 257 258 259 260 261 262
/**
 * throtl_log - log debug message via blktrace
 * @sq: the service_queue being reported
 * @fmt: printf format string
 * @args: printf args
 *
 * The messages are prefixed with "throtl BLKG_NAME" if @sq belongs to a
 * throtl_grp; otherwise, just "throtl".
 */
#define throtl_log(sq, fmt, args...)	do {				\
	struct throtl_grp *__tg = sq_to_tg((sq));			\
	struct throtl_data *__td = sq_to_td((sq));			\
									\
	(void)__td;							\
263 264
	if (likely(!blk_trace_note_message_enabled(__td->queue)))	\
		break;							\
265 266
	if ((__tg)) {							\
		char __pbuf[128];					\
T
Tejun Heo 已提交
267
									\
268 269 270 271 272
		blkg_path(tg_to_blkg(__tg), __pbuf, sizeof(__pbuf));	\
		blk_add_trace_msg(__td->queue, "throtl %s " fmt, __pbuf, ##args); \
	} else {							\
		blk_add_trace_msg(__td->queue, "throtl " fmt, ##args);	\
	}								\
T
Tejun Heo 已提交
273
} while (0)
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
static void throtl_qnode_init(struct throtl_qnode *qn, struct throtl_grp *tg)
{
	INIT_LIST_HEAD(&qn->node);
	bio_list_init(&qn->bios);
	qn->tg = tg;
}

/**
 * throtl_qnode_add_bio - add a bio to a throtl_qnode and activate it
 * @bio: bio being added
 * @qn: qnode to add bio to
 * @queued: the service_queue->queued[] list @qn belongs to
 *
 * Add @bio to @qn and put @qn on @queued if it's not already on.
 * @qn->tg's reference count is bumped when @qn is activated.  See the
 * comment on top of throtl_qnode definition for details.
 */
static void throtl_qnode_add_bio(struct bio *bio, struct throtl_qnode *qn,
				 struct list_head *queued)
{
	bio_list_add(&qn->bios, bio);
	if (list_empty(&qn->node)) {
		list_add_tail(&qn->node, queued);
		blkg_get(tg_to_blkg(qn->tg));
	}
}

/**
 * throtl_peek_queued - peek the first bio on a qnode list
 * @queued: the qnode list to peek
 */
static struct bio *throtl_peek_queued(struct list_head *queued)
{
	struct throtl_qnode *qn = list_first_entry(queued, struct throtl_qnode, node);
	struct bio *bio;

	if (list_empty(queued))
		return NULL;

	bio = bio_list_peek(&qn->bios);
	WARN_ON_ONCE(!bio);
	return bio;
}

/**
 * throtl_pop_queued - pop the first bio form a qnode list
 * @queued: the qnode list to pop a bio from
 * @tg_to_put: optional out argument for throtl_grp to put
 *
 * Pop the first bio from the qnode list @queued.  After popping, the first
 * qnode is removed from @queued if empty or moved to the end of @queued so
 * that the popping order is round-robin.
 *
 * When the first qnode is removed, its associated throtl_grp should be put
 * too.  If @tg_to_put is NULL, this function automatically puts it;
 * otherwise, *@tg_to_put is set to the throtl_grp to put and the caller is
 * responsible for putting it.
 */
static struct bio *throtl_pop_queued(struct list_head *queued,
				     struct throtl_grp **tg_to_put)
{
	struct throtl_qnode *qn = list_first_entry(queued, struct throtl_qnode, node);
	struct bio *bio;

	if (list_empty(queued))
		return NULL;

	bio = bio_list_pop(&qn->bios);
	WARN_ON_ONCE(!bio);

	if (bio_list_empty(&qn->bios)) {
		list_del_init(&qn->node);
		if (tg_to_put)
			*tg_to_put = qn->tg;
		else
			blkg_put(tg_to_blkg(qn->tg));
	} else {
		list_move_tail(&qn->node, queued);
	}

	return bio;
}

358
/* init a service_queue, assumes the caller zeroed it */
359
static void throtl_service_queue_init(struct throtl_service_queue *sq)
360
{
361 362
	INIT_LIST_HEAD(&sq->queued[0]);
	INIT_LIST_HEAD(&sq->queued[1]);
363
	sq->pending_tree = RB_ROOT;
364 365 366 367
	setup_timer(&sq->pending_timer, throtl_pending_timer_fn,
		    (unsigned long)sq);
}

368 369
static struct blkg_policy_data *throtl_pd_alloc(gfp_t gfp, int node)
{
370
	struct throtl_grp *tg;
T
Tejun Heo 已提交
371
	int rw;
372 373 374

	tg = kzalloc_node(sizeof(*tg), gfp, node);
	if (!tg)
375
		return NULL;
376

377 378 379 380 381 382 383 384
	throtl_service_queue_init(&tg->service_queue);

	for (rw = READ; rw <= WRITE; rw++) {
		throtl_qnode_init(&tg->qnode_on_self[rw], tg);
		throtl_qnode_init(&tg->qnode_on_parent[rw], tg);
	}

	RB_CLEAR_NODE(&tg->rb_node);
385 386 387 388
	tg->bps[READ][LIMIT_MAX] = U64_MAX;
	tg->bps[WRITE][LIMIT_MAX] = U64_MAX;
	tg->iops[READ][LIMIT_MAX] = UINT_MAX;
	tg->iops[WRITE][LIMIT_MAX] = UINT_MAX;
S
Shaohua Li 已提交
389 390 391 392 393
	tg->bps_conf[READ][LIMIT_MAX] = U64_MAX;
	tg->bps_conf[WRITE][LIMIT_MAX] = U64_MAX;
	tg->iops_conf[READ][LIMIT_MAX] = UINT_MAX;
	tg->iops_conf[WRITE][LIMIT_MAX] = UINT_MAX;
	/* LIMIT_LOW will have default value 0 */
394

395
	return &tg->pd;
396 397
}

398
static void throtl_pd_init(struct blkg_policy_data *pd)
399
{
400 401
	struct throtl_grp *tg = pd_to_tg(pd);
	struct blkcg_gq *blkg = tg_to_blkg(tg);
402
	struct throtl_data *td = blkg->q->td;
403
	struct throtl_service_queue *sq = &tg->service_queue;
404

405
	/*
406
	 * If on the default hierarchy, we switch to properly hierarchical
407 408 409 410 411
	 * behavior where limits on a given throtl_grp are applied to the
	 * whole subtree rather than just the group itself.  e.g. If 16M
	 * read_bps limit is set on the root group, the whole system can't
	 * exceed 16M for the device.
	 *
412
	 * If not on the default hierarchy, the broken flat hierarchy
413 414 415 416 417
	 * behavior is retained where all throtl_grps are treated as if
	 * they're all separate root groups right below throtl_data.
	 * Limits of a group don't interact with limits of other groups
	 * regardless of the position of the group in the hierarchy.
	 */
418
	sq->parent_sq = &td->service_queue;
419
	if (cgroup_subsys_on_dfl(io_cgrp_subsys) && blkg->parent)
420
		sq->parent_sq = &blkg_to_tg(blkg->parent)->service_queue;
421
	tg->td = td;
422 423
}

424 425 426 427 428 429 430 431
/*
 * Set has_rules[] if @tg or any of its parents have limits configured.
 * This doesn't require walking up to the top of the hierarchy as the
 * parent's has_rules[] is guaranteed to be correct.
 */
static void tg_update_has_rules(struct throtl_grp *tg)
{
	struct throtl_grp *parent_tg = sq_to_tg(tg->service_queue.parent_sq);
432
	struct throtl_data *td = tg->td;
433 434 435 436
	int rw;

	for (rw = READ; rw <= WRITE; rw++)
		tg->has_rules[rw] = (parent_tg && parent_tg->has_rules[rw]) ||
437 438 439
			(td->limit_valid[td->limit_index] &&
			 (tg_bps_limit(tg, rw) != U64_MAX ||
			  tg_iops_limit(tg, rw) != UINT_MAX));
440 441
}

442
static void throtl_pd_online(struct blkg_policy_data *pd)
443 444 445 446 447
{
	/*
	 * We don't want new groups to escape the limits of its ancestors.
	 * Update has_rules[] after a new group is brought online.
	 */
448
	tg_update_has_rules(pd_to_tg(pd));
449 450
}

S
Shaohua Li 已提交
451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469
static void blk_throtl_update_limit_valid(struct throtl_data *td)
{
	struct cgroup_subsys_state *pos_css;
	struct blkcg_gq *blkg;
	bool low_valid = false;

	rcu_read_lock();
	blkg_for_each_descendant_post(blkg, pos_css, td->queue->root_blkg) {
		struct throtl_grp *tg = blkg_to_tg(blkg);

		if (tg->bps[READ][LIMIT_LOW] || tg->bps[WRITE][LIMIT_LOW] ||
		    tg->iops[READ][LIMIT_LOW] || tg->iops[WRITE][LIMIT_LOW])
			low_valid = true;
	}
	rcu_read_unlock();

	td->limit_valid[LIMIT_LOW] = low_valid;
}

470
static void throtl_upgrade_state(struct throtl_data *td);
S
Shaohua Li 已提交
471 472 473 474 475 476 477 478 479 480 481
static void throtl_pd_offline(struct blkg_policy_data *pd)
{
	struct throtl_grp *tg = pd_to_tg(pd);

	tg->bps[READ][LIMIT_LOW] = 0;
	tg->bps[WRITE][LIMIT_LOW] = 0;
	tg->iops[READ][LIMIT_LOW] = 0;
	tg->iops[WRITE][LIMIT_LOW] = 0;

	blk_throtl_update_limit_valid(tg->td);

482 483
	if (!tg->td->limit_valid[tg->td->limit_index])
		throtl_upgrade_state(tg->td);
S
Shaohua Li 已提交
484 485
}

486 487
static void throtl_pd_free(struct blkg_policy_data *pd)
{
488 489
	struct throtl_grp *tg = pd_to_tg(pd);

490
	del_timer_sync(&tg->service_queue.pending_timer);
491
	kfree(tg);
492 493
}

494 495
static struct throtl_grp *
throtl_rb_first(struct throtl_service_queue *parent_sq)
496 497
{
	/* Service tree is empty */
498
	if (!parent_sq->nr_pending)
499 500
		return NULL;

501 502
	if (!parent_sq->first_pending)
		parent_sq->first_pending = rb_first(&parent_sq->pending_tree);
503

504 505
	if (parent_sq->first_pending)
		return rb_entry_tg(parent_sq->first_pending);
506 507 508 509 510 511 512 513 514 515

	return NULL;
}

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

516 517
static void throtl_rb_erase(struct rb_node *n,
			    struct throtl_service_queue *parent_sq)
518
{
519 520 521 522
	if (parent_sq->first_pending == n)
		parent_sq->first_pending = NULL;
	rb_erase_init(n, &parent_sq->pending_tree);
	--parent_sq->nr_pending;
523 524
}

525
static void update_min_dispatch_time(struct throtl_service_queue *parent_sq)
526 527 528
{
	struct throtl_grp *tg;

529
	tg = throtl_rb_first(parent_sq);
530 531 532
	if (!tg)
		return;

533
	parent_sq->first_pending_disptime = tg->disptime;
534 535
}

536
static void tg_service_queue_add(struct throtl_grp *tg)
537
{
538
	struct throtl_service_queue *parent_sq = tg->service_queue.parent_sq;
539
	struct rb_node **node = &parent_sq->pending_tree.rb_node;
540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557
	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)
558
		parent_sq->first_pending = &tg->rb_node;
559 560

	rb_link_node(&tg->rb_node, parent, node);
561
	rb_insert_color(&tg->rb_node, &parent_sq->pending_tree);
562 563
}

564
static void __throtl_enqueue_tg(struct throtl_grp *tg)
565
{
566
	tg_service_queue_add(tg);
567
	tg->flags |= THROTL_TG_PENDING;
568
	tg->service_queue.parent_sq->nr_pending++;
569 570
}

571
static void throtl_enqueue_tg(struct throtl_grp *tg)
572
{
573
	if (!(tg->flags & THROTL_TG_PENDING))
574
		__throtl_enqueue_tg(tg);
575 576
}

577
static void __throtl_dequeue_tg(struct throtl_grp *tg)
578
{
579
	throtl_rb_erase(&tg->rb_node, tg->service_queue.parent_sq);
580
	tg->flags &= ~THROTL_TG_PENDING;
581 582
}

583
static void throtl_dequeue_tg(struct throtl_grp *tg)
584
{
585
	if (tg->flags & THROTL_TG_PENDING)
586
		__throtl_dequeue_tg(tg);
587 588
}

589
/* Call with queue lock held */
590 591
static void throtl_schedule_pending_timer(struct throtl_service_queue *sq,
					  unsigned long expires)
592
{
593 594 595 596 597 598 599 600 601 602 603
	unsigned long max_expire = jiffies + 8 * throtl_slice;

	/*
	 * Since we are adjusting the throttle limit dynamically, the sleep
	 * time calculated according to previous limit might be invalid. It's
	 * possible the cgroup sleep time is very long and no other cgroups
	 * have IO running so notify the limit changes. Make sure the cgroup
	 * doesn't sleep too long to avoid the missed notification.
	 */
	if (time_after(expires, max_expire))
		expires = max_expire;
604 605 606
	mod_timer(&sq->pending_timer, expires);
	throtl_log(sq, "schedule timer. delay=%lu jiffies=%lu",
		   expires - jiffies, jiffies);
607 608
}

609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628
/**
 * throtl_schedule_next_dispatch - schedule the next dispatch cycle
 * @sq: the service_queue to schedule dispatch for
 * @force: force scheduling
 *
 * Arm @sq->pending_timer so that the next dispatch cycle starts on the
 * dispatch time of the first pending child.  Returns %true if either timer
 * is armed or there's no pending child left.  %false if the current
 * dispatch window is still open and the caller should continue
 * dispatching.
 *
 * If @force is %true, the dispatch timer is always scheduled and this
 * function is guaranteed to return %true.  This is to be used when the
 * caller can't dispatch itself and needs to invoke pending_timer
 * unconditionally.  Note that forced scheduling is likely to induce short
 * delay before dispatch starts even if @sq->first_pending_disptime is not
 * in the future and thus shouldn't be used in hot paths.
 */
static bool throtl_schedule_next_dispatch(struct throtl_service_queue *sq,
					  bool force)
629
{
630
	/* any pending children left? */
631
	if (!sq->nr_pending)
632
		return true;
633

634
	update_min_dispatch_time(sq);
635

636
	/* is the next dispatch time in the future? */
637
	if (force || time_after(sq->first_pending_disptime, jiffies)) {
638
		throtl_schedule_pending_timer(sq, sq->first_pending_disptime);
639
		return true;
640 641
	}

642 643
	/* tell the caller to continue dispatching */
	return false;
644 645
}

646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667
static inline void throtl_start_new_slice_with_credit(struct throtl_grp *tg,
		bool rw, unsigned long start)
{
	tg->bytes_disp[rw] = 0;
	tg->io_disp[rw] = 0;

	/*
	 * Previous slice has expired. We must have trimmed it after last
	 * bio dispatch. That means since start of last slice, we never used
	 * that bandwidth. Do try to make use of that bandwidth while giving
	 * credit.
	 */
	if (time_after_eq(start, tg->slice_start[rw]))
		tg->slice_start[rw] = start;

	tg->slice_end[rw] = jiffies + throtl_slice;
	throtl_log(&tg->service_queue,
		   "[%c] new slice with credit start=%lu end=%lu jiffies=%lu",
		   rw == READ ? 'R' : 'W', tg->slice_start[rw],
		   tg->slice_end[rw], jiffies);
}

668
static inline void throtl_start_new_slice(struct throtl_grp *tg, bool rw)
669 670
{
	tg->bytes_disp[rw] = 0;
671
	tg->io_disp[rw] = 0;
672 673
	tg->slice_start[rw] = jiffies;
	tg->slice_end[rw] = jiffies + throtl_slice;
674 675 676 677
	throtl_log(&tg->service_queue,
		   "[%c] new slice start=%lu end=%lu jiffies=%lu",
		   rw == READ ? 'R' : 'W', tg->slice_start[rw],
		   tg->slice_end[rw], jiffies);
678 679
}

680 681
static inline void throtl_set_slice_end(struct throtl_grp *tg, bool rw,
					unsigned long jiffy_end)
682 683 684 685
{
	tg->slice_end[rw] = roundup(jiffy_end, throtl_slice);
}

686 687
static inline void throtl_extend_slice(struct throtl_grp *tg, bool rw,
				       unsigned long jiffy_end)
688 689
{
	tg->slice_end[rw] = roundup(jiffy_end, throtl_slice);
690 691 692 693
	throtl_log(&tg->service_queue,
		   "[%c] extend slice start=%lu end=%lu jiffies=%lu",
		   rw == READ ? 'R' : 'W', tg->slice_start[rw],
		   tg->slice_end[rw], jiffies);
694 695 696
}

/* Determine if previously allocated or extended slice is complete or not */
697
static bool throtl_slice_used(struct throtl_grp *tg, bool rw)
698 699
{
	if (time_in_range(jiffies, tg->slice_start[rw], tg->slice_end[rw]))
700
		return false;
701 702 703 704 705

	return 1;
}

/* Trim the used slices and adjust slice start accordingly */
706
static inline void throtl_trim_slice(struct throtl_grp *tg, bool rw)
707
{
708 709
	unsigned long nr_slices, time_elapsed, io_trim;
	u64 bytes_trim, tmp;
710 711 712 713 714 715 716 717

	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.
	 */
718
	if (throtl_slice_used(tg, rw))
719 720
		return;

721 722 723 724 725 726 727 728
	/*
	 * 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.
	 */

729
	throtl_set_slice_end(tg, rw, jiffies + throtl_slice);
730

731 732 733 734 735 736
	time_elapsed = jiffies - tg->slice_start[rw];

	nr_slices = time_elapsed / throtl_slice;

	if (!nr_slices)
		return;
737
	tmp = tg_bps_limit(tg, rw) * throtl_slice * nr_slices;
738 739
	do_div(tmp, HZ);
	bytes_trim = tmp;
740

741
	io_trim = (tg_iops_limit(tg, rw) * throtl_slice * nr_slices) / HZ;
742

743
	if (!bytes_trim && !io_trim)
744 745 746 747 748 749 750
		return;

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

751 752 753 754 755
	if (tg->io_disp[rw] >= io_trim)
		tg->io_disp[rw] -= io_trim;
	else
		tg->io_disp[rw] = 0;

756 757
	tg->slice_start[rw] += nr_slices * throtl_slice;

758 759 760 761
	throtl_log(&tg->service_queue,
		   "[%c] trim slice nr=%lu bytes=%llu io=%lu start=%lu end=%lu jiffies=%lu",
		   rw == READ ? 'R' : 'W', nr_slices, bytes_trim, io_trim,
		   tg->slice_start[rw], tg->slice_end[rw], jiffies);
762 763
}

764 765
static bool tg_with_in_iops_limit(struct throtl_grp *tg, struct bio *bio,
				  unsigned long *wait)
766 767
{
	bool rw = bio_data_dir(bio);
768
	unsigned int io_allowed;
769
	unsigned long jiffy_elapsed, jiffy_wait, jiffy_elapsed_rnd;
770
	u64 tmp;
771

772
	jiffy_elapsed = jiffy_elapsed_rnd = jiffies - tg->slice_start[rw];
773

774 775 776 777 778 779
	/* 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);

780 781 782 783 784 785 786
	/*
	 * 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.
	 */

787
	tmp = (u64)tg_iops_limit(tg, rw) * jiffy_elapsed_rnd;
788 789 790 791 792 793
	do_div(tmp, HZ);

	if (tmp > UINT_MAX)
		io_allowed = UINT_MAX;
	else
		io_allowed = tmp;
794 795

	if (tg->io_disp[rw] + 1 <= io_allowed) {
796 797
		if (wait)
			*wait = 0;
798
		return true;
799 800
	}

801
	/* Calc approx time to dispatch */
802
	jiffy_wait = ((tg->io_disp[rw] + 1) * HZ) / tg_iops_limit(tg, rw) + 1;
803 804 805 806 807 808 809 810 811 812 813

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

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

814 815
static bool tg_with_in_bps_limit(struct throtl_grp *tg, struct bio *bio,
				 unsigned long *wait)
816 817
{
	bool rw = bio_data_dir(bio);
818
	u64 bytes_allowed, extra_bytes, tmp;
819
	unsigned long jiffy_elapsed, jiffy_wait, jiffy_elapsed_rnd;
820 821 822 823 824 825 826 827 828

	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);

829
	tmp = tg_bps_limit(tg, rw) * jiffy_elapsed_rnd;
830
	do_div(tmp, HZ);
831
	bytes_allowed = tmp;
832

833
	if (tg->bytes_disp[rw] + bio->bi_iter.bi_size <= bytes_allowed) {
834 835
		if (wait)
			*wait = 0;
836
		return true;
837 838 839
	}

	/* Calc approx time to dispatch */
840
	extra_bytes = tg->bytes_disp[rw] + bio->bi_iter.bi_size - bytes_allowed;
841
	jiffy_wait = div64_u64(extra_bytes * HZ, tg_bps_limit(tg, rw));
842 843 844 845 846 847 848 849 850 851 852

	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;
853 854 855 856 857 858 859
	return 0;
}

/*
 * 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
 */
860 861
static bool tg_may_dispatch(struct throtl_grp *tg, struct bio *bio,
			    unsigned long *wait)
862 863 864 865 866 867 868 869 870 871
{
	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.
	 */
872
	BUG_ON(tg->service_queue.nr_queued[rw] &&
873
	       bio != throtl_peek_queued(&tg->service_queue.queued[rw]));
874

875
	/* If tg->bps = -1, then BW is unlimited */
876 877
	if (tg_bps_limit(tg, rw) == U64_MAX &&
	    tg_iops_limit(tg, rw) == UINT_MAX) {
878 879
		if (wait)
			*wait = 0;
880
		return true;
881 882 883 884 885
	}

	/*
	 * If previous slice expired, start a new one otherwise renew/extend
	 * existing slice to make sure it is at least throtl_slice interval
886 887 888
	 * long since now. New slice is started only for empty throttle group.
	 * If there is queued bio, that means there should be an active
	 * slice and it should be extended instead.
889
	 */
890
	if (throtl_slice_used(tg, rw) && !(tg->service_queue.nr_queued[rw]))
891
		throtl_start_new_slice(tg, rw);
892 893
	else {
		if (time_before(tg->slice_end[rw], jiffies + throtl_slice))
894
			throtl_extend_slice(tg, rw, jiffies + throtl_slice);
895 896
	}

897 898
	if (tg_with_in_bps_limit(tg, bio, &bps_wait) &&
	    tg_with_in_iops_limit(tg, bio, &iops_wait)) {
899 900 901 902 903 904 905 906 907 908 909
		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))
910
		throtl_extend_slice(tg, rw, jiffies + max_wait);
911 912 913 914 915 916 917 918 919

	return 0;
}

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

	/* Charge the bio to the group */
920
	tg->bytes_disp[rw] += bio->bi_iter.bi_size;
921
	tg->io_disp[rw]++;
S
Shaohua Li 已提交
922 923
	tg->last_bytes_disp[rw] += bio->bi_iter.bi_size;
	tg->last_io_disp[rw]++;
924

925
	/*
926
	 * BIO_THROTTLED is used to prevent the same bio to be throttled
927 928 929 930
	 * more than once as a throttled bio will go through blk-throtl the
	 * second time when it eventually gets issued.  Set it when a bio
	 * is being charged to a tg.
	 */
931 932
	if (!bio_flagged(bio, BIO_THROTTLED))
		bio_set_flag(bio, BIO_THROTTLED);
933 934
}

935 936 937 938 939 940 941 942 943 944 945
/**
 * throtl_add_bio_tg - add a bio to the specified throtl_grp
 * @bio: bio to add
 * @qn: qnode to use
 * @tg: the target throtl_grp
 *
 * Add @bio to @tg's service_queue using @qn.  If @qn is not specified,
 * tg->qnode_on_self[] is used.
 */
static void throtl_add_bio_tg(struct bio *bio, struct throtl_qnode *qn,
			      struct throtl_grp *tg)
946
{
947
	struct throtl_service_queue *sq = &tg->service_queue;
948 949
	bool rw = bio_data_dir(bio);

950 951 952
	if (!qn)
		qn = &tg->qnode_on_self[rw];

953 954 955 956 957 958 959 960 961
	/*
	 * If @tg doesn't currently have any bios queued in the same
	 * direction, queueing @bio can change when @tg should be
	 * dispatched.  Mark that @tg was empty.  This is automatically
	 * cleaered on the next tg_update_disptime().
	 */
	if (!sq->nr_queued[rw])
		tg->flags |= THROTL_TG_WAS_EMPTY;

962 963
	throtl_qnode_add_bio(bio, qn, &sq->queued[rw]);

964
	sq->nr_queued[rw]++;
965
	throtl_enqueue_tg(tg);
966 967
}

968
static void tg_update_disptime(struct throtl_grp *tg)
969
{
970
	struct throtl_service_queue *sq = &tg->service_queue;
971 972 973
	unsigned long read_wait = -1, write_wait = -1, min_wait = -1, disptime;
	struct bio *bio;

974 975
	bio = throtl_peek_queued(&sq->queued[READ]);
	if (bio)
976
		tg_may_dispatch(tg, bio, &read_wait);
977

978 979
	bio = throtl_peek_queued(&sq->queued[WRITE]);
	if (bio)
980
		tg_may_dispatch(tg, bio, &write_wait);
981 982 983 984 985

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

	/* Update dispatch time */
986
	throtl_dequeue_tg(tg);
987
	tg->disptime = disptime;
988
	throtl_enqueue_tg(tg);
989 990 991

	/* see throtl_add_bio_tg() */
	tg->flags &= ~THROTL_TG_WAS_EMPTY;
992 993
}

994 995 996 997 998 999 1000 1001 1002 1003
static void start_parent_slice_with_credit(struct throtl_grp *child_tg,
					struct throtl_grp *parent_tg, bool rw)
{
	if (throtl_slice_used(parent_tg, rw)) {
		throtl_start_new_slice_with_credit(parent_tg, rw,
				child_tg->slice_start[rw]);
	}

}

1004
static void tg_dispatch_one_bio(struct throtl_grp *tg, bool rw)
1005
{
1006
	struct throtl_service_queue *sq = &tg->service_queue;
1007 1008
	struct throtl_service_queue *parent_sq = sq->parent_sq;
	struct throtl_grp *parent_tg = sq_to_tg(parent_sq);
1009
	struct throtl_grp *tg_to_put = NULL;
1010 1011
	struct bio *bio;

1012 1013 1014 1015 1016 1017 1018
	/*
	 * @bio is being transferred from @tg to @parent_sq.  Popping a bio
	 * from @tg may put its reference and @parent_sq might end up
	 * getting released prematurely.  Remember the tg to put and put it
	 * after @bio is transferred to @parent_sq.
	 */
	bio = throtl_pop_queued(&sq->queued[rw], &tg_to_put);
1019
	sq->nr_queued[rw]--;
1020 1021

	throtl_charge_bio(tg, bio);
1022 1023 1024 1025 1026 1027 1028 1029 1030

	/*
	 * If our parent is another tg, we just need to transfer @bio to
	 * the parent using throtl_add_bio_tg().  If our parent is
	 * @td->service_queue, @bio is ready to be issued.  Put it on its
	 * bio_lists[] and decrease total number queued.  The caller is
	 * responsible for issuing these bios.
	 */
	if (parent_tg) {
1031
		throtl_add_bio_tg(bio, &tg->qnode_on_parent[rw], parent_tg);
1032
		start_parent_slice_with_credit(tg, parent_tg, rw);
1033
	} else {
1034 1035
		throtl_qnode_add_bio(bio, &tg->qnode_on_parent[rw],
				     &parent_sq->queued[rw]);
1036 1037 1038
		BUG_ON(tg->td->nr_queued[rw] <= 0);
		tg->td->nr_queued[rw]--;
	}
1039

1040
	throtl_trim_slice(tg, rw);
1041

1042 1043
	if (tg_to_put)
		blkg_put(tg_to_blkg(tg_to_put));
1044 1045
}

1046
static int throtl_dispatch_tg(struct throtl_grp *tg)
1047
{
1048
	struct throtl_service_queue *sq = &tg->service_queue;
1049 1050
	unsigned int nr_reads = 0, nr_writes = 0;
	unsigned int max_nr_reads = throtl_grp_quantum*3/4;
1051
	unsigned int max_nr_writes = throtl_grp_quantum - max_nr_reads;
1052 1053 1054 1055
	struct bio *bio;

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

1056
	while ((bio = throtl_peek_queued(&sq->queued[READ])) &&
1057
	       tg_may_dispatch(tg, bio, NULL)) {
1058

1059
		tg_dispatch_one_bio(tg, bio_data_dir(bio));
1060 1061 1062 1063 1064 1065
		nr_reads++;

		if (nr_reads >= max_nr_reads)
			break;
	}

1066
	while ((bio = throtl_peek_queued(&sq->queued[WRITE])) &&
1067
	       tg_may_dispatch(tg, bio, NULL)) {
1068

1069
		tg_dispatch_one_bio(tg, bio_data_dir(bio));
1070 1071 1072 1073 1074 1075 1076 1077 1078
		nr_writes++;

		if (nr_writes >= max_nr_writes)
			break;
	}

	return nr_reads + nr_writes;
}

1079
static int throtl_select_dispatch(struct throtl_service_queue *parent_sq)
1080 1081 1082 1083
{
	unsigned int nr_disp = 0;

	while (1) {
1084 1085
		struct throtl_grp *tg = throtl_rb_first(parent_sq);
		struct throtl_service_queue *sq = &tg->service_queue;
1086 1087 1088 1089 1090 1091 1092

		if (!tg)
			break;

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

1093
		throtl_dequeue_tg(tg);
1094

1095
		nr_disp += throtl_dispatch_tg(tg);
1096

1097
		if (sq->nr_queued[0] || sq->nr_queued[1])
1098
			tg_update_disptime(tg);
1099 1100 1101 1102 1103 1104 1105 1106

		if (nr_disp >= throtl_quantum)
			break;
	}

	return nr_disp;
}

1107 1108
static bool throtl_can_upgrade(struct throtl_data *td,
	struct throtl_grp *this_tg);
1109 1110 1111 1112 1113 1114 1115
/**
 * throtl_pending_timer_fn - timer function for service_queue->pending_timer
 * @arg: the throtl_service_queue being serviced
 *
 * This timer is armed when a child throtl_grp with active bio's become
 * pending and queued on the service_queue's pending_tree and expires when
 * the first child throtl_grp should be dispatched.  This function
1116 1117 1118 1119 1120 1121 1122
 * dispatches bio's from the children throtl_grps to the parent
 * service_queue.
 *
 * If the parent's parent is another throtl_grp, dispatching is propagated
 * by either arming its pending_timer or repeating dispatch directly.  If
 * the top-level service_tree is reached, throtl_data->dispatch_work is
 * kicked so that the ready bio's are issued.
1123
 */
1124 1125 1126
static void throtl_pending_timer_fn(unsigned long arg)
{
	struct throtl_service_queue *sq = (void *)arg;
1127
	struct throtl_grp *tg = sq_to_tg(sq);
1128
	struct throtl_data *td = sq_to_td(sq);
1129
	struct request_queue *q = td->queue;
1130 1131
	struct throtl_service_queue *parent_sq;
	bool dispatched;
1132
	int ret;
1133 1134

	spin_lock_irq(q->queue_lock);
1135 1136 1137
	if (throtl_can_upgrade(td, NULL))
		throtl_upgrade_state(td);

1138 1139 1140
again:
	parent_sq = sq->parent_sq;
	dispatched = false;
1141

1142 1143
	while (true) {
		throtl_log(sq, "dispatch nr_queued=%u read=%u write=%u",
1144 1145
			   sq->nr_queued[READ] + sq->nr_queued[WRITE],
			   sq->nr_queued[READ], sq->nr_queued[WRITE]);
1146 1147 1148 1149 1150 1151

		ret = throtl_select_dispatch(sq);
		if (ret) {
			throtl_log(sq, "bios disp=%u", ret);
			dispatched = true;
		}
1152

1153 1154
		if (throtl_schedule_next_dispatch(sq, false))
			break;
1155

1156 1157 1158 1159
		/* this dispatch windows is still open, relax and repeat */
		spin_unlock_irq(q->queue_lock);
		cpu_relax();
		spin_lock_irq(q->queue_lock);
1160
	}
1161

1162 1163
	if (!dispatched)
		goto out_unlock;
1164

1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180
	if (parent_sq) {
		/* @parent_sq is another throl_grp, propagate dispatch */
		if (tg->flags & THROTL_TG_WAS_EMPTY) {
			tg_update_disptime(tg);
			if (!throtl_schedule_next_dispatch(parent_sq, false)) {
				/* window is already open, repeat dispatching */
				sq = parent_sq;
				tg = sq_to_tg(sq);
				goto again;
			}
		}
	} else {
		/* reached the top-level, queue issueing */
		queue_work(kthrotld_workqueue, &td->dispatch_work);
	}
out_unlock:
1181
	spin_unlock_irq(q->queue_lock);
1182
}
1183

1184 1185 1186 1187 1188 1189 1190 1191
/**
 * blk_throtl_dispatch_work_fn - work function for throtl_data->dispatch_work
 * @work: work item being executed
 *
 * This function is queued for execution when bio's reach the bio_lists[]
 * of throtl_data->service_queue.  Those bio's are ready and issued by this
 * function.
 */
1192
static void blk_throtl_dispatch_work_fn(struct work_struct *work)
1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205
{
	struct throtl_data *td = container_of(work, struct throtl_data,
					      dispatch_work);
	struct throtl_service_queue *td_sq = &td->service_queue;
	struct request_queue *q = td->queue;
	struct bio_list bio_list_on_stack;
	struct bio *bio;
	struct blk_plug plug;
	int rw;

	bio_list_init(&bio_list_on_stack);

	spin_lock_irq(q->queue_lock);
1206 1207 1208
	for (rw = READ; rw <= WRITE; rw++)
		while ((bio = throtl_pop_queued(&td_sq->queued[rw], NULL)))
			bio_list_add(&bio_list_on_stack, bio);
1209 1210 1211
	spin_unlock_irq(q->queue_lock);

	if (!bio_list_empty(&bio_list_on_stack)) {
1212
		blk_start_plug(&plug);
1213 1214
		while((bio = bio_list_pop(&bio_list_on_stack)))
			generic_make_request(bio);
1215
		blk_finish_plug(&plug);
1216 1217 1218
	}
}

1219 1220
static u64 tg_prfill_conf_u64(struct seq_file *sf, struct blkg_policy_data *pd,
			      int off)
1221
{
1222 1223
	struct throtl_grp *tg = pd_to_tg(pd);
	u64 v = *(u64 *)((void *)tg + off);
1224

1225
	if (v == U64_MAX)
1226
		return 0;
1227
	return __blkg_prfill_u64(sf, pd, v);
1228 1229
}

1230 1231
static u64 tg_prfill_conf_uint(struct seq_file *sf, struct blkg_policy_data *pd,
			       int off)
1232
{
1233 1234
	struct throtl_grp *tg = pd_to_tg(pd);
	unsigned int v = *(unsigned int *)((void *)tg + off);
1235

1236
	if (v == UINT_MAX)
1237
		return 0;
1238
	return __blkg_prfill_u64(sf, pd, v);
1239 1240
}

1241
static int tg_print_conf_u64(struct seq_file *sf, void *v)
1242
{
1243 1244
	blkcg_print_blkgs(sf, css_to_blkcg(seq_css(sf)), tg_prfill_conf_u64,
			  &blkcg_policy_throtl, seq_cft(sf)->private, false);
1245
	return 0;
1246 1247
}

1248
static int tg_print_conf_uint(struct seq_file *sf, void *v)
1249
{
1250 1251
	blkcg_print_blkgs(sf, css_to_blkcg(seq_css(sf)), tg_prfill_conf_uint,
			  &blkcg_policy_throtl, seq_cft(sf)->private, false);
1252
	return 0;
1253 1254
}

1255
static void tg_conf_updated(struct throtl_grp *tg)
1256
{
1257
	struct throtl_service_queue *sq = &tg->service_queue;
1258
	struct cgroup_subsys_state *pos_css;
1259
	struct blkcg_gq *blkg;
1260

1261 1262
	throtl_log(&tg->service_queue,
		   "limit change rbps=%llu wbps=%llu riops=%u wiops=%u",
1263 1264
		   tg_bps_limit(tg, READ), tg_bps_limit(tg, WRITE),
		   tg_iops_limit(tg, READ), tg_iops_limit(tg, WRITE));
1265

1266 1267 1268 1269 1270 1271 1272
	/*
	 * Update has_rules[] flags for the updated tg's subtree.  A tg is
	 * considered to have rules if either the tg itself or any of its
	 * ancestors has rules.  This identifies groups without any
	 * restrictions in the whole hierarchy and allows them to bypass
	 * blk-throttle.
	 */
1273
	blkg_for_each_descendant_pre(blkg, pos_css, tg_to_blkg(tg))
1274 1275
		tg_update_has_rules(blkg_to_tg(blkg));

1276 1277 1278 1279 1280 1281 1282 1283
	/*
	 * We're already holding queue_lock and know @tg is valid.  Let's
	 * apply the new config directly.
	 *
	 * 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.
	 */
1284 1285
	throtl_start_new_slice(tg, 0);
	throtl_start_new_slice(tg, 1);
1286

1287
	if (tg->flags & THROTL_TG_PENDING) {
1288
		tg_update_disptime(tg);
1289
		throtl_schedule_next_dispatch(sq->parent_sq, true);
1290
	}
1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309
}

static ssize_t tg_set_conf(struct kernfs_open_file *of,
			   char *buf, size_t nbytes, loff_t off, bool is_u64)
{
	struct blkcg *blkcg = css_to_blkcg(of_css(of));
	struct blkg_conf_ctx ctx;
	struct throtl_grp *tg;
	int ret;
	u64 v;

	ret = blkg_conf_prep(blkcg, &blkcg_policy_throtl, buf, &ctx);
	if (ret)
		return ret;

	ret = -EINVAL;
	if (sscanf(ctx.body, "%llu", &v) != 1)
		goto out_finish;
	if (!v)
1310
		v = U64_MAX;
1311 1312 1313 1314 1315 1316 1317

	tg = blkg_to_tg(ctx.blkg);

	if (is_u64)
		*(u64 *)((void *)tg + of_cft(of)->private) = v;
	else
		*(unsigned int *)((void *)tg + of_cft(of)->private) = v;
1318

1319
	tg_conf_updated(tg);
1320 1321
	ret = 0;
out_finish:
1322
	blkg_conf_finish(&ctx);
1323
	return ret ?: nbytes;
1324 1325
}

1326 1327
static ssize_t tg_set_conf_u64(struct kernfs_open_file *of,
			       char *buf, size_t nbytes, loff_t off)
1328
{
1329
	return tg_set_conf(of, buf, nbytes, off, true);
1330 1331
}

1332 1333
static ssize_t tg_set_conf_uint(struct kernfs_open_file *of,
				char *buf, size_t nbytes, loff_t off)
1334
{
1335
	return tg_set_conf(of, buf, nbytes, off, false);
1336 1337
}

1338
static struct cftype throtl_legacy_files[] = {
1339 1340
	{
		.name = "throttle.read_bps_device",
1341
		.private = offsetof(struct throtl_grp, bps[READ][LIMIT_MAX]),
1342
		.seq_show = tg_print_conf_u64,
1343
		.write = tg_set_conf_u64,
1344 1345 1346
	},
	{
		.name = "throttle.write_bps_device",
1347
		.private = offsetof(struct throtl_grp, bps[WRITE][LIMIT_MAX]),
1348
		.seq_show = tg_print_conf_u64,
1349
		.write = tg_set_conf_u64,
1350 1351 1352
	},
	{
		.name = "throttle.read_iops_device",
1353
		.private = offsetof(struct throtl_grp, iops[READ][LIMIT_MAX]),
1354
		.seq_show = tg_print_conf_uint,
1355
		.write = tg_set_conf_uint,
1356 1357 1358
	},
	{
		.name = "throttle.write_iops_device",
1359
		.private = offsetof(struct throtl_grp, iops[WRITE][LIMIT_MAX]),
1360
		.seq_show = tg_print_conf_uint,
1361
		.write = tg_set_conf_uint,
1362 1363 1364
	},
	{
		.name = "throttle.io_service_bytes",
1365 1366
		.private = (unsigned long)&blkcg_policy_throtl,
		.seq_show = blkg_print_stat_bytes,
1367 1368 1369
	},
	{
		.name = "throttle.io_serviced",
1370 1371
		.private = (unsigned long)&blkcg_policy_throtl,
		.seq_show = blkg_print_stat_ios,
1372 1373 1374 1375
	},
	{ }	/* terminate */
};

S
Shaohua Li 已提交
1376
static u64 tg_prfill_limit(struct seq_file *sf, struct blkg_policy_data *pd,
1377 1378 1379 1380 1381
			 int off)
{
	struct throtl_grp *tg = pd_to_tg(pd);
	const char *dname = blkg_dev_name(pd->blkg);
	char bufs[4][21] = { "max", "max", "max", "max" };
S
Shaohua Li 已提交
1382 1383
	u64 bps_dft;
	unsigned int iops_dft;
1384 1385 1386

	if (!dname)
		return 0;
1387

S
Shaohua Li 已提交
1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399
	if (off == LIMIT_LOW) {
		bps_dft = 0;
		iops_dft = 0;
	} else {
		bps_dft = U64_MAX;
		iops_dft = UINT_MAX;
	}

	if (tg->bps_conf[READ][off] == bps_dft &&
	    tg->bps_conf[WRITE][off] == bps_dft &&
	    tg->iops_conf[READ][off] == iops_dft &&
	    tg->iops_conf[WRITE][off] == iops_dft)
1400 1401
		return 0;

S
Shaohua Li 已提交
1402
	if (tg->bps_conf[READ][off] != bps_dft)
1403
		snprintf(bufs[0], sizeof(bufs[0]), "%llu",
S
Shaohua Li 已提交
1404 1405
			tg->bps_conf[READ][off]);
	if (tg->bps_conf[WRITE][off] != bps_dft)
1406
		snprintf(bufs[1], sizeof(bufs[1]), "%llu",
S
Shaohua Li 已提交
1407 1408
			tg->bps_conf[WRITE][off]);
	if (tg->iops_conf[READ][off] != iops_dft)
1409
		snprintf(bufs[2], sizeof(bufs[2]), "%u",
S
Shaohua Li 已提交
1410 1411
			tg->iops_conf[READ][off]);
	if (tg->iops_conf[WRITE][off] != iops_dft)
1412
		snprintf(bufs[3], sizeof(bufs[3]), "%u",
S
Shaohua Li 已提交
1413
			tg->iops_conf[WRITE][off]);
1414 1415 1416 1417 1418 1419

	seq_printf(sf, "%s rbps=%s wbps=%s riops=%s wiops=%s\n",
		   dname, bufs[0], bufs[1], bufs[2], bufs[3]);
	return 0;
}

S
Shaohua Li 已提交
1420
static int tg_print_limit(struct seq_file *sf, void *v)
1421
{
S
Shaohua Li 已提交
1422
	blkcg_print_blkgs(sf, css_to_blkcg(seq_css(sf)), tg_prfill_limit,
1423 1424 1425 1426
			  &blkcg_policy_throtl, seq_cft(sf)->private, false);
	return 0;
}

S
Shaohua Li 已提交
1427
static ssize_t tg_set_limit(struct kernfs_open_file *of,
1428 1429 1430 1431 1432 1433 1434
			  char *buf, size_t nbytes, loff_t off)
{
	struct blkcg *blkcg = css_to_blkcg(of_css(of));
	struct blkg_conf_ctx ctx;
	struct throtl_grp *tg;
	u64 v[4];
	int ret;
S
Shaohua Li 已提交
1435
	int index = of_cft(of)->private;
1436 1437 1438 1439 1440 1441 1442

	ret = blkg_conf_prep(blkcg, &blkcg_policy_throtl, buf, &ctx);
	if (ret)
		return ret;

	tg = blkg_to_tg(ctx.blkg);

S
Shaohua Li 已提交
1443 1444 1445 1446
	v[0] = tg->bps_conf[READ][index];
	v[1] = tg->bps_conf[WRITE][index];
	v[2] = tg->iops_conf[READ][index];
	v[3] = tg->iops_conf[WRITE][index];
1447 1448 1449 1450

	while (true) {
		char tok[27];	/* wiops=18446744073709551616 */
		char *p;
1451
		u64 val = U64_MAX;
1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482
		int len;

		if (sscanf(ctx.body, "%26s%n", tok, &len) != 1)
			break;
		if (tok[0] == '\0')
			break;
		ctx.body += len;

		ret = -EINVAL;
		p = tok;
		strsep(&p, "=");
		if (!p || (sscanf(p, "%llu", &val) != 1 && strcmp(p, "max")))
			goto out_finish;

		ret = -ERANGE;
		if (!val)
			goto out_finish;

		ret = -EINVAL;
		if (!strcmp(tok, "rbps"))
			v[0] = val;
		else if (!strcmp(tok, "wbps"))
			v[1] = val;
		else if (!strcmp(tok, "riops"))
			v[2] = min_t(u64, val, UINT_MAX);
		else if (!strcmp(tok, "wiops"))
			v[3] = min_t(u64, val, UINT_MAX);
		else
			goto out_finish;
	}

S
Shaohua Li 已提交
1483 1484 1485 1486
	tg->bps_conf[READ][index] = v[0];
	tg->bps_conf[WRITE][index] = v[1];
	tg->iops_conf[READ][index] = v[2];
	tg->iops_conf[WRITE][index] = v[3];
1487

S
Shaohua Li 已提交
1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507
	if (index == LIMIT_MAX) {
		tg->bps[READ][index] = v[0];
		tg->bps[WRITE][index] = v[1];
		tg->iops[READ][index] = v[2];
		tg->iops[WRITE][index] = v[3];
	}
	tg->bps[READ][LIMIT_LOW] = min(tg->bps_conf[READ][LIMIT_LOW],
		tg->bps_conf[READ][LIMIT_MAX]);
	tg->bps[WRITE][LIMIT_LOW] = min(tg->bps_conf[WRITE][LIMIT_LOW],
		tg->bps_conf[WRITE][LIMIT_MAX]);
	tg->iops[READ][LIMIT_LOW] = min(tg->iops_conf[READ][LIMIT_LOW],
		tg->iops_conf[READ][LIMIT_MAX]);
	tg->iops[WRITE][LIMIT_LOW] = min(tg->iops_conf[WRITE][LIMIT_LOW],
		tg->iops_conf[WRITE][LIMIT_MAX]);

	if (index == LIMIT_LOW) {
		blk_throtl_update_limit_valid(tg->td);
		if (tg->td->limit_valid[LIMIT_LOW])
			tg->td->limit_index = LIMIT_LOW;
	}
1508 1509 1510 1511 1512 1513 1514 1515
	tg_conf_updated(tg);
	ret = 0;
out_finish:
	blkg_conf_finish(&ctx);
	return ret ?: nbytes;
}

static struct cftype throtl_files[] = {
S
Shaohua Li 已提交
1516 1517 1518 1519 1520 1521 1522 1523 1524
#ifdef CONFIG_BLK_DEV_THROTTLING_LOW
	{
		.name = "low",
		.flags = CFTYPE_NOT_ON_ROOT,
		.seq_show = tg_print_limit,
		.write = tg_set_limit,
		.private = LIMIT_LOW,
	},
#endif
1525 1526 1527
	{
		.name = "max",
		.flags = CFTYPE_NOT_ON_ROOT,
S
Shaohua Li 已提交
1528 1529 1530
		.seq_show = tg_print_limit,
		.write = tg_set_limit,
		.private = LIMIT_MAX,
1531 1532 1533 1534
	},
	{ }	/* terminate */
};

1535
static void throtl_shutdown_wq(struct request_queue *q)
1536 1537 1538
{
	struct throtl_data *td = q->td;

1539
	cancel_work_sync(&td->dispatch_work);
1540 1541
}

T
Tejun Heo 已提交
1542
static struct blkcg_policy blkcg_policy_throtl = {
1543
	.dfl_cftypes		= throtl_files,
1544
	.legacy_cftypes		= throtl_legacy_files,
1545

1546
	.pd_alloc_fn		= throtl_pd_alloc,
1547
	.pd_init_fn		= throtl_pd_init,
1548
	.pd_online_fn		= throtl_pd_online,
S
Shaohua Li 已提交
1549
	.pd_offline_fn		= throtl_pd_offline,
1550
	.pd_free_fn		= throtl_pd_free,
1551 1552
};

S
Shaohua Li 已提交
1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591
static unsigned long __tg_last_low_overflow_time(struct throtl_grp *tg)
{
	unsigned long rtime = jiffies, wtime = jiffies;

	if (tg->bps[READ][LIMIT_LOW] || tg->iops[READ][LIMIT_LOW])
		rtime = tg->last_low_overflow_time[READ];
	if (tg->bps[WRITE][LIMIT_LOW] || tg->iops[WRITE][LIMIT_LOW])
		wtime = tg->last_low_overflow_time[WRITE];
	return min(rtime, wtime);
}

/* tg should not be an intermediate node */
static unsigned long tg_last_low_overflow_time(struct throtl_grp *tg)
{
	struct throtl_service_queue *parent_sq;
	struct throtl_grp *parent = tg;
	unsigned long ret = __tg_last_low_overflow_time(tg);

	while (true) {
		parent_sq = parent->service_queue.parent_sq;
		parent = sq_to_tg(parent_sq);
		if (!parent)
			break;

		/*
		 * The parent doesn't have low limit, it always reaches low
		 * limit. Its overflow time is useless for children
		 */
		if (!parent->bps[READ][LIMIT_LOW] &&
		    !parent->iops[READ][LIMIT_LOW] &&
		    !parent->bps[WRITE][LIMIT_LOW] &&
		    !parent->iops[WRITE][LIMIT_LOW])
			continue;
		if (time_after(__tg_last_low_overflow_time(parent), ret))
			ret = __tg_last_low_overflow_time(parent);
	}
	return ret;
}

1592 1593 1594 1595 1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622 1623 1624 1625 1626 1627 1628 1629 1630 1631 1632 1633 1634
static bool throtl_tg_can_upgrade(struct throtl_grp *tg)
{
	struct throtl_service_queue *sq = &tg->service_queue;
	bool read_limit, write_limit;

	/*
	 * if cgroup reaches low limit (if low limit is 0, the cgroup always
	 * reaches), it's ok to upgrade to next limit
	 */
	read_limit = tg->bps[READ][LIMIT_LOW] || tg->iops[READ][LIMIT_LOW];
	write_limit = tg->bps[WRITE][LIMIT_LOW] || tg->iops[WRITE][LIMIT_LOW];
	if (!read_limit && !write_limit)
		return true;
	if (read_limit && sq->nr_queued[READ] &&
	    (!write_limit || sq->nr_queued[WRITE]))
		return true;
	if (write_limit && sq->nr_queued[WRITE] &&
	    (!read_limit || sq->nr_queued[READ]))
		return true;
	return false;
}

static bool throtl_hierarchy_can_upgrade(struct throtl_grp *tg)
{
	while (true) {
		if (throtl_tg_can_upgrade(tg))
			return true;
		tg = sq_to_tg(tg->service_queue.parent_sq);
		if (!tg || !tg_to_blkg(tg)->parent)
			return false;
	}
	return false;
}

static bool throtl_can_upgrade(struct throtl_data *td,
	struct throtl_grp *this_tg)
{
	struct cgroup_subsys_state *pos_css;
	struct blkcg_gq *blkg;

	if (td->limit_index != LIMIT_LOW)
		return false;

S
Shaohua Li 已提交
1635 1636 1637
	if (time_before(jiffies, td->low_downgrade_time + throtl_slice))
		return false;

1638 1639 1640 1641 1642 1643 1644 1645 1646 1647 1648 1649 1650 1651 1652 1653 1654 1655 1656 1657 1658 1659 1660
	rcu_read_lock();
	blkg_for_each_descendant_post(blkg, pos_css, td->queue->root_blkg) {
		struct throtl_grp *tg = blkg_to_tg(blkg);

		if (tg == this_tg)
			continue;
		if (!list_empty(&tg_to_blkg(tg)->blkcg->css.children))
			continue;
		if (!throtl_hierarchy_can_upgrade(tg)) {
			rcu_read_unlock();
			return false;
		}
	}
	rcu_read_unlock();
	return true;
}

static void throtl_upgrade_state(struct throtl_data *td)
{
	struct cgroup_subsys_state *pos_css;
	struct blkcg_gq *blkg;

	td->limit_index = LIMIT_MAX;
S
Shaohua Li 已提交
1661
	td->low_upgrade_time = jiffies;
1662 1663 1664 1665 1666 1667 1668 1669 1670 1671 1672 1673 1674 1675 1676
	rcu_read_lock();
	blkg_for_each_descendant_post(blkg, pos_css, td->queue->root_blkg) {
		struct throtl_grp *tg = blkg_to_tg(blkg);
		struct throtl_service_queue *sq = &tg->service_queue;

		tg->disptime = jiffies - 1;
		throtl_select_dispatch(sq);
		throtl_schedule_next_dispatch(sq, false);
	}
	rcu_read_unlock();
	throtl_select_dispatch(&td->service_queue);
	throtl_schedule_next_dispatch(&td->service_queue, false);
	queue_work(kthrotld_workqueue, &td->dispatch_work);
}

S
Shaohua Li 已提交
1677 1678 1679 1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692 1693 1694 1695 1696 1697 1698 1699 1700 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724 1725 1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746 1747 1748 1749 1750 1751 1752 1753 1754 1755 1756 1757 1758 1759 1760 1761 1762 1763 1764 1765 1766 1767 1768 1769
static void throtl_downgrade_state(struct throtl_data *td, int new)
{
	td->limit_index = new;
	td->low_downgrade_time = jiffies;
}

static bool throtl_tg_can_downgrade(struct throtl_grp *tg)
{
	struct throtl_data *td = tg->td;
	unsigned long now = jiffies;

	/*
	 * If cgroup is below low limit, consider downgrade and throttle other
	 * cgroups
	 */
	if (time_after_eq(now, td->low_upgrade_time + throtl_slice) &&
	    time_after_eq(now, tg_last_low_overflow_time(tg) + throtl_slice))
		return true;
	return false;
}

static bool throtl_hierarchy_can_downgrade(struct throtl_grp *tg)
{
	while (true) {
		if (!throtl_tg_can_downgrade(tg))
			return false;
		tg = sq_to_tg(tg->service_queue.parent_sq);
		if (!tg || !tg_to_blkg(tg)->parent)
			break;
	}
	return true;
}

static void throtl_downgrade_check(struct throtl_grp *tg)
{
	uint64_t bps;
	unsigned int iops;
	unsigned long elapsed_time;
	unsigned long now = jiffies;

	if (tg->td->limit_index != LIMIT_MAX ||
	    !tg->td->limit_valid[LIMIT_LOW])
		return;
	if (!list_empty(&tg_to_blkg(tg)->blkcg->css.children))
		return;
	if (time_after(tg->last_check_time + throtl_slice, now))
		return;

	elapsed_time = now - tg->last_check_time;
	tg->last_check_time = now;

	if (time_before(now, tg_last_low_overflow_time(tg) + throtl_slice))
		return;

	if (tg->bps[READ][LIMIT_LOW]) {
		bps = tg->last_bytes_disp[READ] * HZ;
		do_div(bps, elapsed_time);
		if (bps >= tg->bps[READ][LIMIT_LOW])
			tg->last_low_overflow_time[READ] = now;
	}

	if (tg->bps[WRITE][LIMIT_LOW]) {
		bps = tg->last_bytes_disp[WRITE] * HZ;
		do_div(bps, elapsed_time);
		if (bps >= tg->bps[WRITE][LIMIT_LOW])
			tg->last_low_overflow_time[WRITE] = now;
	}

	if (tg->iops[READ][LIMIT_LOW]) {
		iops = tg->last_io_disp[READ] * HZ / elapsed_time;
		if (iops >= tg->iops[READ][LIMIT_LOW])
			tg->last_low_overflow_time[READ] = now;
	}

	if (tg->iops[WRITE][LIMIT_LOW]) {
		iops = tg->last_io_disp[WRITE] * HZ / elapsed_time;
		if (iops >= tg->iops[WRITE][LIMIT_LOW])
			tg->last_low_overflow_time[WRITE] = now;
	}

	/*
	 * If cgroup is below low limit, consider downgrade and throttle other
	 * cgroups
	 */
	if (throtl_hierarchy_can_downgrade(tg))
		throtl_downgrade_state(tg->td, LIMIT_LOW);

	tg->last_bytes_disp[READ] = 0;
	tg->last_bytes_disp[WRITE] = 0;
	tg->last_io_disp[READ] = 0;
	tg->last_io_disp[WRITE] = 0;
}

1770 1771
bool blk_throtl_bio(struct request_queue *q, struct blkcg_gq *blkg,
		    struct bio *bio)
1772
{
1773
	struct throtl_qnode *qn = NULL;
1774
	struct throtl_grp *tg = blkg_to_tg(blkg ?: q->root_blkg);
1775
	struct throtl_service_queue *sq;
1776
	bool rw = bio_data_dir(bio);
1777
	bool throttled = false;
1778

1779 1780
	WARN_ON_ONCE(!rcu_read_lock_held());

1781
	/* see throtl_charge_bio() */
1782
	if (bio_flagged(bio, BIO_THROTTLED) || !tg->has_rules[rw])
1783
		goto out;
1784 1785

	spin_lock_irq(q->queue_lock);
1786 1787

	if (unlikely(blk_queue_bypass(q)))
1788
		goto out_unlock;
1789

1790 1791
	sq = &tg->service_queue;

1792
again:
1793
	while (true) {
S
Shaohua Li 已提交
1794 1795 1796
		if (tg->last_low_overflow_time[rw] == 0)
			tg->last_low_overflow_time[rw] = jiffies;
		throtl_downgrade_check(tg);
1797 1798 1799
		/* throtl is FIFO - if bios are already queued, should queue */
		if (sq->nr_queued[rw])
			break;
1800

1801
		/* if above limits, break to queue */
1802
		if (!tg_may_dispatch(tg, bio, NULL)) {
S
Shaohua Li 已提交
1803
			tg->last_low_overflow_time[rw] = jiffies;
1804 1805 1806 1807
			if (throtl_can_upgrade(tg->td, tg)) {
				throtl_upgrade_state(tg->td);
				goto again;
			}
1808
			break;
1809
		}
1810 1811

		/* within limits, let's charge and dispatch directly */
1812
		throtl_charge_bio(tg, bio);
1813 1814 1815 1816 1817 1818 1819 1820 1821 1822 1823 1824

		/*
		 * 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.
		 */
1825
		throtl_trim_slice(tg, rw);
1826 1827 1828 1829 1830 1831

		/*
		 * @bio passed through this layer without being throttled.
		 * Climb up the ladder.  If we''re already at the top, it
		 * can be executed directly.
		 */
1832
		qn = &tg->qnode_on_parent[rw];
1833 1834 1835 1836
		sq = sq->parent_sq;
		tg = sq_to_tg(sq);
		if (!tg)
			goto out_unlock;
1837 1838
	}

1839
	/* out-of-limit, queue to @tg */
1840 1841
	throtl_log(sq, "[%c] bio. bdisp=%llu sz=%u bps=%llu iodisp=%u iops=%u queued=%d/%d",
		   rw == READ ? 'R' : 'W',
1842 1843 1844
		   tg->bytes_disp[rw], bio->bi_iter.bi_size,
		   tg_bps_limit(tg, rw),
		   tg->io_disp[rw], tg_iops_limit(tg, rw),
1845
		   sq->nr_queued[READ], sq->nr_queued[WRITE]);
1846

S
Shaohua Li 已提交
1847 1848
	tg->last_low_overflow_time[rw] = jiffies;

1849
	bio_associate_current(bio);
1850
	tg->td->nr_queued[rw]++;
1851
	throtl_add_bio_tg(bio, qn, tg);
1852
	throttled = true;
1853

1854 1855 1856 1857 1858 1859
	/*
	 * Update @tg's dispatch time and force schedule dispatch if @tg
	 * was empty before @bio.  The forced scheduling isn't likely to
	 * cause undue delay as @bio is likely to be dispatched directly if
	 * its @tg's disptime is not in the future.
	 */
1860
	if (tg->flags & THROTL_TG_WAS_EMPTY) {
1861
		tg_update_disptime(tg);
1862
		throtl_schedule_next_dispatch(tg->service_queue.parent_sq, true);
1863 1864
	}

1865
out_unlock:
1866
	spin_unlock_irq(q->queue_lock);
1867
out:
1868 1869 1870 1871 1872 1873
	/*
	 * As multiple blk-throtls may stack in the same issue path, we
	 * don't want bios to leave with the flag set.  Clear the flag if
	 * being issued.
	 */
	if (!throttled)
1874
		bio_clear_flag(bio, BIO_THROTTLED);
1875
	return throttled;
1876 1877
}

1878 1879 1880 1881 1882 1883 1884 1885 1886 1887 1888 1889 1890 1891 1892
/*
 * Dispatch all bios from all children tg's queued on @parent_sq.  On
 * return, @parent_sq is guaranteed to not have any active children tg's
 * and all bios from previously active tg's are on @parent_sq->bio_lists[].
 */
static void tg_drain_bios(struct throtl_service_queue *parent_sq)
{
	struct throtl_grp *tg;

	while ((tg = throtl_rb_first(parent_sq))) {
		struct throtl_service_queue *sq = &tg->service_queue;
		struct bio *bio;

		throtl_dequeue_tg(tg);

1893
		while ((bio = throtl_peek_queued(&sq->queued[READ])))
1894
			tg_dispatch_one_bio(tg, bio_data_dir(bio));
1895
		while ((bio = throtl_peek_queued(&sq->queued[WRITE])))
1896 1897 1898 1899
			tg_dispatch_one_bio(tg, bio_data_dir(bio));
	}
}

1900 1901 1902 1903 1904 1905 1906 1907 1908 1909
/**
 * 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;
1910
	struct blkcg_gq *blkg;
1911
	struct cgroup_subsys_state *pos_css;
1912
	struct bio *bio;
1913
	int rw;
1914

1915
	queue_lockdep_assert_held(q);
1916
	rcu_read_lock();
1917

1918 1919 1920 1921 1922 1923
	/*
	 * Drain each tg while doing post-order walk on the blkg tree, so
	 * that all bios are propagated to td->service_queue.  It'd be
	 * better to walk service_queue tree directly but blkg walk is
	 * easier.
	 */
1924
	blkg_for_each_descendant_post(blkg, pos_css, td->queue->root_blkg)
1925
		tg_drain_bios(&blkg_to_tg(blkg)->service_queue);
1926

1927 1928 1929 1930
	/* finally, transfer bios from top-level tg's into the td */
	tg_drain_bios(&td->service_queue);

	rcu_read_unlock();
1931 1932
	spin_unlock_irq(q->queue_lock);

1933
	/* all bios now should be in td->service_queue, issue them */
1934
	for (rw = READ; rw <= WRITE; rw++)
1935 1936
		while ((bio = throtl_pop_queued(&td->service_queue.queued[rw],
						NULL)))
1937
			generic_make_request(bio);
1938 1939 1940 1941

	spin_lock_irq(q->queue_lock);
}

1942 1943 1944
int blk_throtl_init(struct request_queue *q)
{
	struct throtl_data *td;
1945
	int ret;
1946 1947 1948 1949 1950

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

1951
	INIT_WORK(&td->dispatch_work, blk_throtl_dispatch_work_fn);
1952
	throtl_service_queue_init(&td->service_queue);
1953

1954
	q->td = td;
1955
	td->queue = q;
V
Vivek Goyal 已提交
1956

1957
	td->limit_valid[LIMIT_MAX] = true;
S
Shaohua Li 已提交
1958
	td->limit_index = LIMIT_MAX;
S
Shaohua Li 已提交
1959 1960
	td->low_upgrade_time = jiffies;
	td->low_downgrade_time = jiffies;
1961
	/* activate policy */
T
Tejun Heo 已提交
1962
	ret = blkcg_activate_policy(q, &blkcg_policy_throtl);
1963
	if (ret)
1964
		kfree(td);
1965
	return ret;
1966 1967 1968 1969
}

void blk_throtl_exit(struct request_queue *q)
{
T
Tejun Heo 已提交
1970
	BUG_ON(!q->td);
1971
	throtl_shutdown_wq(q);
T
Tejun Heo 已提交
1972
	blkcg_deactivate_policy(q, &blkcg_policy_throtl);
1973
	kfree(q->td);
1974 1975 1976 1977
}

static int __init throtl_init(void)
{
1978 1979 1980 1981
	kthrotld_workqueue = alloc_workqueue("kthrotld", WQ_MEM_RECLAIM, 0);
	if (!kthrotld_workqueue)
		panic("Failed to create kthrotld\n");

T
Tejun Heo 已提交
1982
	return blkcg_policy_register(&blkcg_policy_throtl);
1983 1984 1985
}

module_init(throtl_init);