blk-throttle.c 43.7 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 87 88 89 90
enum {
	LIMIT_MAX,
	LIMIT_CNT,
};

91
struct throtl_grp {
92 93 94
	/* must be the first member */
	struct blkg_policy_data pd;

95
	/* active throtl group service_queue member */
96 97
	struct rb_node rb_node;

98 99 100
	/* throtl_data this group belongs to */
	struct throtl_data *td;

101 102 103
	/* this group's service queue */
	struct throtl_service_queue service_queue;

104 105 106 107 108 109 110 111 112 113 114
	/*
	 * 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];

115 116 117 118 119 120 121 122 123
	/*
	 * 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;

124 125 126
	/* are there any throtl rules between this group and td? */
	bool has_rules[2];

127
	/* bytes per second rate limits */
128
	uint64_t bps[2][LIMIT_CNT];
129

130
	/* IOPS limits */
131
	unsigned int iops[2][LIMIT_CNT];
132

133 134
	/* Number of bytes disptached in current slice */
	uint64_t bytes_disp[2];
135 136
	/* Number of bio's dispatched in current slice */
	unsigned int io_disp[2];
137 138 139 140 141 142 143 144 145

	/* 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 */
146
	struct throtl_service_queue service_queue;
147 148 149 150 151 152 153

	struct request_queue *queue;

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

	/* Work for dispatching throttled bios */
154
	struct work_struct dispatch_work;
155 156
	unsigned int limit_index;
	bool limit_valid[LIMIT_CNT];
157 158
};

159 160
static void throtl_pending_timer_fn(unsigned long arg);

161 162 163 164 165
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
166
static inline struct throtl_grp *blkg_to_tg(struct blkcg_gq *blkg)
167
{
168
	return pd_to_tg(blkg_to_pd(blkg, &blkcg_policy_throtl));
169 170
}

T
Tejun Heo 已提交
171
static inline struct blkcg_gq *tg_to_blkg(struct throtl_grp *tg)
172
{
173
	return pd_to_blkg(&tg->pd);
174 175
}

176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194
/**
 * 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
 *
195
 * A service_queue can be embedded in either a throtl_grp or throtl_data.
196 197 198 199 200 201 202 203 204 205 206 207
 * 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);
}

208 209 210 211 212 213 214 215 216 217
static uint64_t tg_bps_limit(struct throtl_grp *tg, int rw)
{
	return tg->bps[rw][tg->td->limit_index];
}

static unsigned int tg_iops_limit(struct throtl_grp *tg, int rw)
{
	return tg->iops[rw][tg->td->limit_index];
}

218 219 220 221 222 223 224 225 226 227 228 229 230 231
/**
 * 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;							\
232 233
	if (likely(!blk_trace_note_message_enabled(__td->queue)))	\
		break;							\
234 235
	if ((__tg)) {							\
		char __pbuf[128];					\
T
Tejun Heo 已提交
236
									\
237 238 239 240 241
		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
242
} while (0)
243

244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326
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;
}

327
/* init a service_queue, assumes the caller zeroed it */
328
static void throtl_service_queue_init(struct throtl_service_queue *sq)
329
{
330 331
	INIT_LIST_HEAD(&sq->queued[0]);
	INIT_LIST_HEAD(&sq->queued[1]);
332
	sq->pending_tree = RB_ROOT;
333 334 335 336
	setup_timer(&sq->pending_timer, throtl_pending_timer_fn,
		    (unsigned long)sq);
}

337 338
static struct blkg_policy_data *throtl_pd_alloc(gfp_t gfp, int node)
{
339
	struct throtl_grp *tg;
T
Tejun Heo 已提交
340
	int rw;
341 342 343

	tg = kzalloc_node(sizeof(*tg), gfp, node);
	if (!tg)
344
		return NULL;
345

346 347 348 349 350 351 352 353
	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);
354 355 356 357
	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;
358

359
	return &tg->pd;
360 361
}

362
static void throtl_pd_init(struct blkg_policy_data *pd)
363
{
364 365
	struct throtl_grp *tg = pd_to_tg(pd);
	struct blkcg_gq *blkg = tg_to_blkg(tg);
366
	struct throtl_data *td = blkg->q->td;
367
	struct throtl_service_queue *sq = &tg->service_queue;
368

369
	/*
370
	 * If on the default hierarchy, we switch to properly hierarchical
371 372 373 374 375
	 * 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.
	 *
376
	 * If not on the default hierarchy, the broken flat hierarchy
377 378 379 380 381
	 * 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.
	 */
382
	sq->parent_sq = &td->service_queue;
383
	if (cgroup_subsys_on_dfl(io_cgrp_subsys) && blkg->parent)
384
		sq->parent_sq = &blkg_to_tg(blkg->parent)->service_queue;
385
	tg->td = td;
386 387
}

388 389 390 391 392 393 394 395
/*
 * 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);
396
	struct throtl_data *td = tg->td;
397 398 399 400
	int rw;

	for (rw = READ; rw <= WRITE; rw++)
		tg->has_rules[rw] = (parent_tg && parent_tg->has_rules[rw]) ||
401 402 403
			(td->limit_valid[td->limit_index] &&
			 (tg_bps_limit(tg, rw) != U64_MAX ||
			  tg_iops_limit(tg, rw) != UINT_MAX));
404 405
}

406
static void throtl_pd_online(struct blkg_policy_data *pd)
407 408 409 410 411
{
	/*
	 * We don't want new groups to escape the limits of its ancestors.
	 * Update has_rules[] after a new group is brought online.
	 */
412
	tg_update_has_rules(pd_to_tg(pd));
413 414
}

415 416
static void throtl_pd_free(struct blkg_policy_data *pd)
{
417 418
	struct throtl_grp *tg = pd_to_tg(pd);

419
	del_timer_sync(&tg->service_queue.pending_timer);
420
	kfree(tg);
421 422
}

423 424
static struct throtl_grp *
throtl_rb_first(struct throtl_service_queue *parent_sq)
425 426
{
	/* Service tree is empty */
427
	if (!parent_sq->nr_pending)
428 429
		return NULL;

430 431
	if (!parent_sq->first_pending)
		parent_sq->first_pending = rb_first(&parent_sq->pending_tree);
432

433 434
	if (parent_sq->first_pending)
		return rb_entry_tg(parent_sq->first_pending);
435 436 437 438 439 440 441 442 443 444

	return NULL;
}

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

445 446
static void throtl_rb_erase(struct rb_node *n,
			    struct throtl_service_queue *parent_sq)
447
{
448 449 450 451
	if (parent_sq->first_pending == n)
		parent_sq->first_pending = NULL;
	rb_erase_init(n, &parent_sq->pending_tree);
	--parent_sq->nr_pending;
452 453
}

454
static void update_min_dispatch_time(struct throtl_service_queue *parent_sq)
455 456 457
{
	struct throtl_grp *tg;

458
	tg = throtl_rb_first(parent_sq);
459 460 461
	if (!tg)
		return;

462
	parent_sq->first_pending_disptime = tg->disptime;
463 464
}

465
static void tg_service_queue_add(struct throtl_grp *tg)
466
{
467
	struct throtl_service_queue *parent_sq = tg->service_queue.parent_sq;
468
	struct rb_node **node = &parent_sq->pending_tree.rb_node;
469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486
	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)
487
		parent_sq->first_pending = &tg->rb_node;
488 489

	rb_link_node(&tg->rb_node, parent, node);
490
	rb_insert_color(&tg->rb_node, &parent_sq->pending_tree);
491 492
}

493
static void __throtl_enqueue_tg(struct throtl_grp *tg)
494
{
495
	tg_service_queue_add(tg);
496
	tg->flags |= THROTL_TG_PENDING;
497
	tg->service_queue.parent_sq->nr_pending++;
498 499
}

500
static void throtl_enqueue_tg(struct throtl_grp *tg)
501
{
502
	if (!(tg->flags & THROTL_TG_PENDING))
503
		__throtl_enqueue_tg(tg);
504 505
}

506
static void __throtl_dequeue_tg(struct throtl_grp *tg)
507
{
508
	throtl_rb_erase(&tg->rb_node, tg->service_queue.parent_sq);
509
	tg->flags &= ~THROTL_TG_PENDING;
510 511
}

512
static void throtl_dequeue_tg(struct throtl_grp *tg)
513
{
514
	if (tg->flags & THROTL_TG_PENDING)
515
		__throtl_dequeue_tg(tg);
516 517
}

518
/* Call with queue lock held */
519 520
static void throtl_schedule_pending_timer(struct throtl_service_queue *sq,
					  unsigned long expires)
521
{
522 523 524
	mod_timer(&sq->pending_timer, expires);
	throtl_log(sq, "schedule timer. delay=%lu jiffies=%lu",
		   expires - jiffies, jiffies);
525 526
}

527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546
/**
 * 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)
547
{
548
	/* any pending children left? */
549
	if (!sq->nr_pending)
550
		return true;
551

552
	update_min_dispatch_time(sq);
553

554
	/* is the next dispatch time in the future? */
555
	if (force || time_after(sq->first_pending_disptime, jiffies)) {
556
		throtl_schedule_pending_timer(sq, sq->first_pending_disptime);
557
		return true;
558 559
	}

560 561
	/* tell the caller to continue dispatching */
	return false;
562 563
}

564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585
static 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);
}

586
static inline void throtl_start_new_slice(struct throtl_grp *tg, bool rw)
587 588
{
	tg->bytes_disp[rw] = 0;
589
	tg->io_disp[rw] = 0;
590 591
	tg->slice_start[rw] = jiffies;
	tg->slice_end[rw] = jiffies + throtl_slice;
592 593 594 595
	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);
596 597
}

598 599
static inline void throtl_set_slice_end(struct throtl_grp *tg, bool rw,
					unsigned long jiffy_end)
600 601 602 603
{
	tg->slice_end[rw] = roundup(jiffy_end, throtl_slice);
}

604 605
static inline void throtl_extend_slice(struct throtl_grp *tg, bool rw,
				       unsigned long jiffy_end)
606 607
{
	tg->slice_end[rw] = roundup(jiffy_end, throtl_slice);
608 609 610 611
	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);
612 613 614
}

/* Determine if previously allocated or extended slice is complete or not */
615
static bool throtl_slice_used(struct throtl_grp *tg, bool rw)
616 617
{
	if (time_in_range(jiffies, tg->slice_start[rw], tg->slice_end[rw]))
618
		return false;
619 620 621 622 623

	return 1;
}

/* Trim the used slices and adjust slice start accordingly */
624
static inline void throtl_trim_slice(struct throtl_grp *tg, bool rw)
625
{
626 627
	unsigned long nr_slices, time_elapsed, io_trim;
	u64 bytes_trim, tmp;
628 629 630 631 632 633 634 635

	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.
	 */
636
	if (throtl_slice_used(tg, rw))
637 638
		return;

639 640 641 642 643 644 645 646
	/*
	 * 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.
	 */

647
	throtl_set_slice_end(tg, rw, jiffies + throtl_slice);
648

649 650 651 652 653 654
	time_elapsed = jiffies - tg->slice_start[rw];

	nr_slices = time_elapsed / throtl_slice;

	if (!nr_slices)
		return;
655
	tmp = tg_bps_limit(tg, rw) * throtl_slice * nr_slices;
656 657
	do_div(tmp, HZ);
	bytes_trim = tmp;
658

659
	io_trim = (tg_iops_limit(tg, rw) * throtl_slice * nr_slices) / HZ;
660

661
	if (!bytes_trim && !io_trim)
662 663 664 665 666 667 668
		return;

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

669 670 671 672 673
	if (tg->io_disp[rw] >= io_trim)
		tg->io_disp[rw] -= io_trim;
	else
		tg->io_disp[rw] = 0;

674 675
	tg->slice_start[rw] += nr_slices * throtl_slice;

676 677 678 679
	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);
680 681
}

682 683
static bool tg_with_in_iops_limit(struct throtl_grp *tg, struct bio *bio,
				  unsigned long *wait)
684 685
{
	bool rw = bio_data_dir(bio);
686
	unsigned int io_allowed;
687
	unsigned long jiffy_elapsed, jiffy_wait, jiffy_elapsed_rnd;
688
	u64 tmp;
689

690
	jiffy_elapsed = jiffy_elapsed_rnd = jiffies - tg->slice_start[rw];
691

692 693 694 695 696 697
	/* 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);

698 699 700 701 702 703 704
	/*
	 * 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.
	 */

705
	tmp = (u64)tg_iops_limit(tg, rw) * jiffy_elapsed_rnd;
706 707 708 709 710 711
	do_div(tmp, HZ);

	if (tmp > UINT_MAX)
		io_allowed = UINT_MAX;
	else
		io_allowed = tmp;
712 713

	if (tg->io_disp[rw] + 1 <= io_allowed) {
714 715
		if (wait)
			*wait = 0;
716
		return true;
717 718
	}

719
	/* Calc approx time to dispatch */
720
	jiffy_wait = ((tg->io_disp[rw] + 1) * HZ) / tg_iops_limit(tg, rw) + 1;
721 722 723 724 725 726 727 728 729 730 731

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

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

732 733
static bool tg_with_in_bps_limit(struct throtl_grp *tg, struct bio *bio,
				 unsigned long *wait)
734 735
{
	bool rw = bio_data_dir(bio);
736
	u64 bytes_allowed, extra_bytes, tmp;
737
	unsigned long jiffy_elapsed, jiffy_wait, jiffy_elapsed_rnd;
738 739 740 741 742 743 744 745 746

	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);

747
	tmp = tg_bps_limit(tg, rw) * jiffy_elapsed_rnd;
748
	do_div(tmp, HZ);
749
	bytes_allowed = tmp;
750

751
	if (tg->bytes_disp[rw] + bio->bi_iter.bi_size <= bytes_allowed) {
752 753
		if (wait)
			*wait = 0;
754
		return true;
755 756 757
	}

	/* Calc approx time to dispatch */
758
	extra_bytes = tg->bytes_disp[rw] + bio->bi_iter.bi_size - bytes_allowed;
759
	jiffy_wait = div64_u64(extra_bytes * HZ, tg_bps_limit(tg, rw));
760 761 762 763 764 765 766 767 768 769 770

	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;
771 772 773 774 775 776 777
	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
 */
778 779
static bool tg_may_dispatch(struct throtl_grp *tg, struct bio *bio,
			    unsigned long *wait)
780 781 782 783 784 785 786 787 788 789
{
	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.
	 */
790
	BUG_ON(tg->service_queue.nr_queued[rw] &&
791
	       bio != throtl_peek_queued(&tg->service_queue.queued[rw]));
792

793
	/* If tg->bps = -1, then BW is unlimited */
794 795
	if (tg_bps_limit(tg, rw) == U64_MAX &&
	    tg_iops_limit(tg, rw) == UINT_MAX) {
796 797
		if (wait)
			*wait = 0;
798
		return true;
799 800 801 802 803
	}

	/*
	 * If previous slice expired, start a new one otherwise renew/extend
	 * existing slice to make sure it is at least throtl_slice interval
804 805 806
	 * 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.
807
	 */
808
	if (throtl_slice_used(tg, rw) && !(tg->service_queue.nr_queued[rw]))
809
		throtl_start_new_slice(tg, rw);
810 811
	else {
		if (time_before(tg->slice_end[rw], jiffies + throtl_slice))
812
			throtl_extend_slice(tg, rw, jiffies + throtl_slice);
813 814
	}

815 816
	if (tg_with_in_bps_limit(tg, bio, &bps_wait) &&
	    tg_with_in_iops_limit(tg, bio, &iops_wait)) {
817 818 819 820 821 822 823 824 825 826 827
		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))
828
		throtl_extend_slice(tg, rw, jiffies + max_wait);
829 830 831 832 833 834 835 836 837

	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 */
838
	tg->bytes_disp[rw] += bio->bi_iter.bi_size;
839
	tg->io_disp[rw]++;
840

841
	/*
842
	 * BIO_THROTTLED is used to prevent the same bio to be throttled
843 844 845 846
	 * 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.
	 */
847 848
	if (!bio_flagged(bio, BIO_THROTTLED))
		bio_set_flag(bio, BIO_THROTTLED);
849 850
}

851 852 853 854 855 856 857 858 859 860 861
/**
 * 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)
862
{
863
	struct throtl_service_queue *sq = &tg->service_queue;
864 865
	bool rw = bio_data_dir(bio);

866 867 868
	if (!qn)
		qn = &tg->qnode_on_self[rw];

869 870 871 872 873 874 875 876 877
	/*
	 * 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;

878 879
	throtl_qnode_add_bio(bio, qn, &sq->queued[rw]);

880
	sq->nr_queued[rw]++;
881
	throtl_enqueue_tg(tg);
882 883
}

884
static void tg_update_disptime(struct throtl_grp *tg)
885
{
886
	struct throtl_service_queue *sq = &tg->service_queue;
887 888 889
	unsigned long read_wait = -1, write_wait = -1, min_wait = -1, disptime;
	struct bio *bio;

890 891
	bio = throtl_peek_queued(&sq->queued[READ]);
	if (bio)
892
		tg_may_dispatch(tg, bio, &read_wait);
893

894 895
	bio = throtl_peek_queued(&sq->queued[WRITE]);
	if (bio)
896
		tg_may_dispatch(tg, bio, &write_wait);
897 898 899 900 901

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

	/* Update dispatch time */
902
	throtl_dequeue_tg(tg);
903
	tg->disptime = disptime;
904
	throtl_enqueue_tg(tg);
905 906 907

	/* see throtl_add_bio_tg() */
	tg->flags &= ~THROTL_TG_WAS_EMPTY;
908 909
}

910 911 912 913 914 915 916 917 918 919
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]);
	}

}

920
static void tg_dispatch_one_bio(struct throtl_grp *tg, bool rw)
921
{
922
	struct throtl_service_queue *sq = &tg->service_queue;
923 924
	struct throtl_service_queue *parent_sq = sq->parent_sq;
	struct throtl_grp *parent_tg = sq_to_tg(parent_sq);
925
	struct throtl_grp *tg_to_put = NULL;
926 927
	struct bio *bio;

928 929 930 931 932 933 934
	/*
	 * @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);
935
	sq->nr_queued[rw]--;
936 937

	throtl_charge_bio(tg, bio);
938 939 940 941 942 943 944 945 946

	/*
	 * 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) {
947
		throtl_add_bio_tg(bio, &tg->qnode_on_parent[rw], parent_tg);
948
		start_parent_slice_with_credit(tg, parent_tg, rw);
949
	} else {
950 951
		throtl_qnode_add_bio(bio, &tg->qnode_on_parent[rw],
				     &parent_sq->queued[rw]);
952 953 954
		BUG_ON(tg->td->nr_queued[rw] <= 0);
		tg->td->nr_queued[rw]--;
	}
955

956
	throtl_trim_slice(tg, rw);
957

958 959
	if (tg_to_put)
		blkg_put(tg_to_blkg(tg_to_put));
960 961
}

962
static int throtl_dispatch_tg(struct throtl_grp *tg)
963
{
964
	struct throtl_service_queue *sq = &tg->service_queue;
965 966
	unsigned int nr_reads = 0, nr_writes = 0;
	unsigned int max_nr_reads = throtl_grp_quantum*3/4;
967
	unsigned int max_nr_writes = throtl_grp_quantum - max_nr_reads;
968 969 970 971
	struct bio *bio;

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

972
	while ((bio = throtl_peek_queued(&sq->queued[READ])) &&
973
	       tg_may_dispatch(tg, bio, NULL)) {
974

975
		tg_dispatch_one_bio(tg, bio_data_dir(bio));
976 977 978 979 980 981
		nr_reads++;

		if (nr_reads >= max_nr_reads)
			break;
	}

982
	while ((bio = throtl_peek_queued(&sq->queued[WRITE])) &&
983
	       tg_may_dispatch(tg, bio, NULL)) {
984

985
		tg_dispatch_one_bio(tg, bio_data_dir(bio));
986 987 988 989 990 991 992 993 994
		nr_writes++;

		if (nr_writes >= max_nr_writes)
			break;
	}

	return nr_reads + nr_writes;
}

995
static int throtl_select_dispatch(struct throtl_service_queue *parent_sq)
996 997 998 999
{
	unsigned int nr_disp = 0;

	while (1) {
1000 1001
		struct throtl_grp *tg = throtl_rb_first(parent_sq);
		struct throtl_service_queue *sq = &tg->service_queue;
1002 1003 1004 1005 1006 1007 1008

		if (!tg)
			break;

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

1009
		throtl_dequeue_tg(tg);
1010

1011
		nr_disp += throtl_dispatch_tg(tg);
1012

1013
		if (sq->nr_queued[0] || sq->nr_queued[1])
1014
			tg_update_disptime(tg);
1015 1016 1017 1018 1019 1020 1021 1022

		if (nr_disp >= throtl_quantum)
			break;
	}

	return nr_disp;
}

1023 1024 1025 1026 1027 1028 1029
/**
 * 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
1030 1031 1032 1033 1034 1035 1036
 * 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.
1037
 */
1038 1039 1040
static void throtl_pending_timer_fn(unsigned long arg)
{
	struct throtl_service_queue *sq = (void *)arg;
1041
	struct throtl_grp *tg = sq_to_tg(sq);
1042
	struct throtl_data *td = sq_to_td(sq);
1043
	struct request_queue *q = td->queue;
1044 1045
	struct throtl_service_queue *parent_sq;
	bool dispatched;
1046
	int ret;
1047 1048

	spin_lock_irq(q->queue_lock);
1049 1050 1051
again:
	parent_sq = sq->parent_sq;
	dispatched = false;
1052

1053 1054
	while (true) {
		throtl_log(sq, "dispatch nr_queued=%u read=%u write=%u",
1055 1056
			   sq->nr_queued[READ] + sq->nr_queued[WRITE],
			   sq->nr_queued[READ], sq->nr_queued[WRITE]);
1057 1058 1059 1060 1061 1062

		ret = throtl_select_dispatch(sq);
		if (ret) {
			throtl_log(sq, "bios disp=%u", ret);
			dispatched = true;
		}
1063

1064 1065
		if (throtl_schedule_next_dispatch(sq, false))
			break;
1066

1067 1068 1069 1070
		/* this dispatch windows is still open, relax and repeat */
		spin_unlock_irq(q->queue_lock);
		cpu_relax();
		spin_lock_irq(q->queue_lock);
1071
	}
1072

1073 1074
	if (!dispatched)
		goto out_unlock;
1075

1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091
	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:
1092
	spin_unlock_irq(q->queue_lock);
1093
}
1094

1095 1096 1097 1098 1099 1100 1101 1102
/**
 * 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.
 */
1103
static void blk_throtl_dispatch_work_fn(struct work_struct *work)
1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116
{
	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);
1117 1118 1119
	for (rw = READ; rw <= WRITE; rw++)
		while ((bio = throtl_pop_queued(&td_sq->queued[rw], NULL)))
			bio_list_add(&bio_list_on_stack, bio);
1120 1121 1122
	spin_unlock_irq(q->queue_lock);

	if (!bio_list_empty(&bio_list_on_stack)) {
1123
		blk_start_plug(&plug);
1124 1125
		while((bio = bio_list_pop(&bio_list_on_stack)))
			generic_make_request(bio);
1126
		blk_finish_plug(&plug);
1127 1128 1129
	}
}

1130 1131
static u64 tg_prfill_conf_u64(struct seq_file *sf, struct blkg_policy_data *pd,
			      int off)
1132
{
1133 1134
	struct throtl_grp *tg = pd_to_tg(pd);
	u64 v = *(u64 *)((void *)tg + off);
1135

1136
	if (v == U64_MAX)
1137
		return 0;
1138
	return __blkg_prfill_u64(sf, pd, v);
1139 1140
}

1141 1142
static u64 tg_prfill_conf_uint(struct seq_file *sf, struct blkg_policy_data *pd,
			       int off)
1143
{
1144 1145
	struct throtl_grp *tg = pd_to_tg(pd);
	unsigned int v = *(unsigned int *)((void *)tg + off);
1146

1147
	if (v == UINT_MAX)
1148
		return 0;
1149
	return __blkg_prfill_u64(sf, pd, v);
1150 1151
}

1152
static int tg_print_conf_u64(struct seq_file *sf, void *v)
1153
{
1154 1155
	blkcg_print_blkgs(sf, css_to_blkcg(seq_css(sf)), tg_prfill_conf_u64,
			  &blkcg_policy_throtl, seq_cft(sf)->private, false);
1156
	return 0;
1157 1158
}

1159
static int tg_print_conf_uint(struct seq_file *sf, void *v)
1160
{
1161 1162
	blkcg_print_blkgs(sf, css_to_blkcg(seq_css(sf)), tg_prfill_conf_uint,
			  &blkcg_policy_throtl, seq_cft(sf)->private, false);
1163
	return 0;
1164 1165
}

1166
static void tg_conf_updated(struct throtl_grp *tg)
1167
{
1168
	struct throtl_service_queue *sq = &tg->service_queue;
1169
	struct cgroup_subsys_state *pos_css;
1170
	struct blkcg_gq *blkg;
1171

1172 1173
	throtl_log(&tg->service_queue,
		   "limit change rbps=%llu wbps=%llu riops=%u wiops=%u",
1174 1175
		   tg_bps_limit(tg, READ), tg_bps_limit(tg, WRITE),
		   tg_iops_limit(tg, READ), tg_iops_limit(tg, WRITE));
1176

1177 1178 1179 1180 1181 1182 1183
	/*
	 * 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.
	 */
1184
	blkg_for_each_descendant_pre(blkg, pos_css, tg_to_blkg(tg))
1185 1186
		tg_update_has_rules(blkg_to_tg(blkg));

1187 1188 1189 1190 1191 1192 1193 1194
	/*
	 * 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.
	 */
1195 1196
	throtl_start_new_slice(tg, 0);
	throtl_start_new_slice(tg, 1);
1197

1198
	if (tg->flags & THROTL_TG_PENDING) {
1199
		tg_update_disptime(tg);
1200
		throtl_schedule_next_dispatch(sq->parent_sq, true);
1201
	}
1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220
}

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)
1221
		v = U64_MAX;
1222 1223 1224 1225 1226 1227 1228

	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;
1229

1230
	tg_conf_updated(tg);
1231 1232
	ret = 0;
out_finish:
1233
	blkg_conf_finish(&ctx);
1234
	return ret ?: nbytes;
1235 1236
}

1237 1238
static ssize_t tg_set_conf_u64(struct kernfs_open_file *of,
			       char *buf, size_t nbytes, loff_t off)
1239
{
1240
	return tg_set_conf(of, buf, nbytes, off, true);
1241 1242
}

1243 1244
static ssize_t tg_set_conf_uint(struct kernfs_open_file *of,
				char *buf, size_t nbytes, loff_t off)
1245
{
1246
	return tg_set_conf(of, buf, nbytes, off, false);
1247 1248
}

1249
static struct cftype throtl_legacy_files[] = {
1250 1251
	{
		.name = "throttle.read_bps_device",
1252
		.private = offsetof(struct throtl_grp, bps[READ][LIMIT_MAX]),
1253
		.seq_show = tg_print_conf_u64,
1254
		.write = tg_set_conf_u64,
1255 1256 1257
	},
	{
		.name = "throttle.write_bps_device",
1258
		.private = offsetof(struct throtl_grp, bps[WRITE][LIMIT_MAX]),
1259
		.seq_show = tg_print_conf_u64,
1260
		.write = tg_set_conf_u64,
1261 1262 1263
	},
	{
		.name = "throttle.read_iops_device",
1264
		.private = offsetof(struct throtl_grp, iops[READ][LIMIT_MAX]),
1265
		.seq_show = tg_print_conf_uint,
1266
		.write = tg_set_conf_uint,
1267 1268 1269
	},
	{
		.name = "throttle.write_iops_device",
1270
		.private = offsetof(struct throtl_grp, iops[WRITE][LIMIT_MAX]),
1271
		.seq_show = tg_print_conf_uint,
1272
		.write = tg_set_conf_uint,
1273 1274 1275
	},
	{
		.name = "throttle.io_service_bytes",
1276 1277
		.private = (unsigned long)&blkcg_policy_throtl,
		.seq_show = blkg_print_stat_bytes,
1278 1279 1280
	},
	{
		.name = "throttle.io_serviced",
1281 1282
		.private = (unsigned long)&blkcg_policy_throtl,
		.seq_show = blkg_print_stat_ios,
1283 1284 1285 1286
	},
	{ }	/* terminate */
};

1287 1288 1289 1290 1291 1292 1293 1294 1295
static u64 tg_prfill_max(struct seq_file *sf, struct blkg_policy_data *pd,
			 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" };

	if (!dname)
		return 0;
1296 1297 1298 1299 1300

	if (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)
1301 1302
		return 0;

1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314
	if (tg->bps[READ][LIMIT_MAX] != U64_MAX)
		snprintf(bufs[0], sizeof(bufs[0]), "%llu",
			tg->bps[READ][LIMIT_MAX]);
	if (tg->bps[WRITE][LIMIT_MAX] != U64_MAX)
		snprintf(bufs[1], sizeof(bufs[1]), "%llu",
			tg->bps[WRITE][LIMIT_MAX]);
	if (tg->iops[READ][LIMIT_MAX] != UINT_MAX)
		snprintf(bufs[2], sizeof(bufs[2]), "%u",
			tg->iops[READ][LIMIT_MAX]);
	if (tg->iops[WRITE][LIMIT_MAX] != UINT_MAX)
		snprintf(bufs[3], sizeof(bufs[3]), "%u",
			tg->iops[WRITE][LIMIT_MAX]);
1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342

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

static int tg_print_max(struct seq_file *sf, void *v)
{
	blkcg_print_blkgs(sf, css_to_blkcg(seq_css(sf)), tg_prfill_max,
			  &blkcg_policy_throtl, seq_cft(sf)->private, false);
	return 0;
}

static ssize_t tg_set_max(struct kernfs_open_file *of,
			  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;

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

	tg = blkg_to_tg(ctx.blkg);

1343 1344 1345 1346
	v[0] = tg->bps[READ][LIMIT_MAX];
	v[1] = tg->bps[WRITE][LIMIT_MAX];
	v[2] = tg->iops[READ][LIMIT_MAX];
	v[3] = tg->iops[WRITE][LIMIT_MAX];
1347 1348 1349 1350

	while (true) {
		char tok[27];	/* wiops=18446744073709551616 */
		char *p;
1351
		u64 val = U64_MAX;
1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382
		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;
	}

1383 1384 1385 1386
	tg->bps[READ][LIMIT_MAX] = v[0];
	tg->bps[WRITE][LIMIT_MAX] = v[1];
	tg->iops[READ][LIMIT_MAX] = v[2];
	tg->iops[WRITE][LIMIT_MAX] = v[3];
1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404

	tg_conf_updated(tg);
	ret = 0;
out_finish:
	blkg_conf_finish(&ctx);
	return ret ?: nbytes;
}

static struct cftype throtl_files[] = {
	{
		.name = "max",
		.flags = CFTYPE_NOT_ON_ROOT,
		.seq_show = tg_print_max,
		.write = tg_set_max,
	},
	{ }	/* terminate */
};

1405
static void throtl_shutdown_wq(struct request_queue *q)
1406 1407 1408
{
	struct throtl_data *td = q->td;

1409
	cancel_work_sync(&td->dispatch_work);
1410 1411
}

T
Tejun Heo 已提交
1412
static struct blkcg_policy blkcg_policy_throtl = {
1413
	.dfl_cftypes		= throtl_files,
1414
	.legacy_cftypes		= throtl_legacy_files,
1415

1416
	.pd_alloc_fn		= throtl_pd_alloc,
1417
	.pd_init_fn		= throtl_pd_init,
1418
	.pd_online_fn		= throtl_pd_online,
1419
	.pd_free_fn		= throtl_pd_free,
1420 1421
};

1422 1423
bool blk_throtl_bio(struct request_queue *q, struct blkcg_gq *blkg,
		    struct bio *bio)
1424
{
1425
	struct throtl_qnode *qn = NULL;
1426
	struct throtl_grp *tg = blkg_to_tg(blkg ?: q->root_blkg);
1427
	struct throtl_service_queue *sq;
1428
	bool rw = bio_data_dir(bio);
1429
	bool throttled = false;
1430

1431 1432
	WARN_ON_ONCE(!rcu_read_lock_held());

1433
	/* see throtl_charge_bio() */
1434
	if (bio_flagged(bio, BIO_THROTTLED) || !tg->has_rules[rw])
1435
		goto out;
1436 1437

	spin_lock_irq(q->queue_lock);
1438 1439

	if (unlikely(blk_queue_bypass(q)))
1440
		goto out_unlock;
1441

1442 1443
	sq = &tg->service_queue;

1444 1445 1446 1447
	while (true) {
		/* throtl is FIFO - if bios are already queued, should queue */
		if (sq->nr_queued[rw])
			break;
1448

1449 1450 1451 1452 1453
		/* if above limits, break to queue */
		if (!tg_may_dispatch(tg, bio, NULL))
			break;

		/* within limits, let's charge and dispatch directly */
1454
		throtl_charge_bio(tg, bio);
1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466

		/*
		 * 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.
		 */
1467
		throtl_trim_slice(tg, rw);
1468 1469 1470 1471 1472 1473

		/*
		 * @bio passed through this layer without being throttled.
		 * Climb up the ladder.  If we''re already at the top, it
		 * can be executed directly.
		 */
1474
		qn = &tg->qnode_on_parent[rw];
1475 1476 1477 1478
		sq = sq->parent_sq;
		tg = sq_to_tg(sq);
		if (!tg)
			goto out_unlock;
1479 1480
	}

1481
	/* out-of-limit, queue to @tg */
1482 1483
	throtl_log(sq, "[%c] bio. bdisp=%llu sz=%u bps=%llu iodisp=%u iops=%u queued=%d/%d",
		   rw == READ ? 'R' : 'W',
1484 1485 1486
		   tg->bytes_disp[rw], bio->bi_iter.bi_size,
		   tg_bps_limit(tg, rw),
		   tg->io_disp[rw], tg_iops_limit(tg, rw),
1487
		   sq->nr_queued[READ], sq->nr_queued[WRITE]);
1488

1489
	bio_associate_current(bio);
1490
	tg->td->nr_queued[rw]++;
1491
	throtl_add_bio_tg(bio, qn, tg);
1492
	throttled = true;
1493

1494 1495 1496 1497 1498 1499
	/*
	 * 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.
	 */
1500
	if (tg->flags & THROTL_TG_WAS_EMPTY) {
1501
		tg_update_disptime(tg);
1502
		throtl_schedule_next_dispatch(tg->service_queue.parent_sq, true);
1503 1504
	}

1505
out_unlock:
1506
	spin_unlock_irq(q->queue_lock);
1507
out:
1508 1509 1510 1511 1512 1513
	/*
	 * 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)
1514
		bio_clear_flag(bio, BIO_THROTTLED);
1515
	return throttled;
1516 1517
}

1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532
/*
 * 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);

1533
		while ((bio = throtl_peek_queued(&sq->queued[READ])))
1534
			tg_dispatch_one_bio(tg, bio_data_dir(bio));
1535
		while ((bio = throtl_peek_queued(&sq->queued[WRITE])))
1536 1537 1538 1539
			tg_dispatch_one_bio(tg, bio_data_dir(bio));
	}
}

1540 1541 1542 1543 1544 1545 1546 1547 1548 1549
/**
 * 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;
1550
	struct blkcg_gq *blkg;
1551
	struct cgroup_subsys_state *pos_css;
1552
	struct bio *bio;
1553
	int rw;
1554

1555
	queue_lockdep_assert_held(q);
1556
	rcu_read_lock();
1557

1558 1559 1560 1561 1562 1563
	/*
	 * 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.
	 */
1564
	blkg_for_each_descendant_post(blkg, pos_css, td->queue->root_blkg)
1565
		tg_drain_bios(&blkg_to_tg(blkg)->service_queue);
1566

1567 1568 1569 1570
	/* finally, transfer bios from top-level tg's into the td */
	tg_drain_bios(&td->service_queue);

	rcu_read_unlock();
1571 1572
	spin_unlock_irq(q->queue_lock);

1573
	/* all bios now should be in td->service_queue, issue them */
1574
	for (rw = READ; rw <= WRITE; rw++)
1575 1576
		while ((bio = throtl_pop_queued(&td->service_queue.queued[rw],
						NULL)))
1577
			generic_make_request(bio);
1578 1579 1580 1581

	spin_lock_irq(q->queue_lock);
}

1582 1583 1584
int blk_throtl_init(struct request_queue *q)
{
	struct throtl_data *td;
1585
	int ret;
1586 1587 1588 1589 1590

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

1591
	INIT_WORK(&td->dispatch_work, blk_throtl_dispatch_work_fn);
1592
	throtl_service_queue_init(&td->service_queue);
1593

1594
	q->td = td;
1595
	td->queue = q;
V
Vivek Goyal 已提交
1596

1597
	td->limit_valid[LIMIT_MAX] = true;
1598
	/* activate policy */
T
Tejun Heo 已提交
1599
	ret = blkcg_activate_policy(q, &blkcg_policy_throtl);
1600
	if (ret)
1601
		kfree(td);
1602
	return ret;
1603 1604 1605 1606
}

void blk_throtl_exit(struct request_queue *q)
{
T
Tejun Heo 已提交
1607
	BUG_ON(!q->td);
1608
	throtl_shutdown_wq(q);
T
Tejun Heo 已提交
1609
	blkcg_deactivate_policy(q, &blkcg_policy_throtl);
1610
	kfree(q->td);
1611 1612 1613 1614
}

static int __init throtl_init(void)
{
1615 1616 1617 1618
	kthrotld_workqueue = alloc_workqueue("kthrotld", WQ_MEM_RECLAIM, 0);
	if (!kthrotld_workqueue)
		panic("Failed to create kthrotld\n");

T
Tejun Heo 已提交
1619
	return blkcg_policy_register(&blkcg_policy_throtl);
1620 1621 1622
}

module_init(throtl_init);