blk-cgroup.c 24.2 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13
/*
 * Common Block IO controller cgroup interface
 *
 * Based on ideas and code from CFQ, CFS and BFQ:
 * Copyright (C) 2003 Jens Axboe <axboe@kernel.dk>
 *
 * Copyright (C) 2008 Fabio Checconi <fabio@gandalf.sssup.it>
 *		      Paolo Valente <paolo.valente@unimore.it>
 *
 * Copyright (C) 2009 Vivek Goyal <vgoyal@redhat.com>
 * 	              Nauman Rafique <nauman@google.com>
 */
#include <linux/ioprio.h>
14
#include <linux/kdev_t.h>
15
#include <linux/module.h>
16
#include <linux/err.h>
17
#include <linux/blkdev.h>
18
#include <linux/slab.h>
19
#include <linux/genhd.h>
20
#include <linux/delay.h>
T
Tejun Heo 已提交
21
#include <linux/atomic.h>
22
#include "blk-cgroup.h"
23
#include "blk.h"
24

25 26
#define MAX_KEY_LEN 100

27
static DEFINE_MUTEX(blkcg_pol_mutex);
28

T
Tejun Heo 已提交
29 30
struct blkcg blkcg_root = { .cfq_weight = 2 * CFQ_WEIGHT_DEFAULT };
EXPORT_SYMBOL_GPL(blkcg_root);
31

T
Tejun Heo 已提交
32
static struct blkcg_policy *blkcg_policy[BLKCG_MAX_POLS];
33

34
static bool blkcg_policy_enabled(struct request_queue *q,
T
Tejun Heo 已提交
35
				 const struct blkcg_policy *pol)
36 37 38 39
{
	return pol && test_bit(pol->plid, q->blkcg_pols);
}

40 41 42 43 44 45
/**
 * blkg_free - free a blkg
 * @blkg: blkg to free
 *
 * Free @blkg which may be partially allocated.
 */
T
Tejun Heo 已提交
46
static void blkg_free(struct blkcg_gq *blkg)
47
{
48
	int i;
49 50 51 52

	if (!blkg)
		return;

T
Tejun Heo 已提交
53
	for (i = 0; i < BLKCG_MAX_POLS; i++) {
T
Tejun Heo 已提交
54
		struct blkcg_policy *pol = blkcg_policy[i];
55 56
		struct blkg_policy_data *pd = blkg->pd[i];

57 58 59
		if (!pd)
			continue;

60 61
		if (pol && pol->pd_exit_fn)
			pol->pd_exit_fn(blkg);
62 63

		kfree(pd);
64
	}
65

66
	blk_exit_rl(&blkg->rl);
67
	kfree(blkg);
68 69 70 71 72 73
}

/**
 * blkg_alloc - allocate a blkg
 * @blkcg: block cgroup the new blkg is associated with
 * @q: request_queue the new blkg is associated with
74
 * @gfp_mask: allocation mask to use
75
 *
76
 * Allocate a new blkg assocating @blkcg and @q.
77
 */
78 79
static struct blkcg_gq *blkg_alloc(struct blkcg *blkcg, struct request_queue *q,
				   gfp_t gfp_mask)
80
{
T
Tejun Heo 已提交
81
	struct blkcg_gq *blkg;
82
	int i;
83 84

	/* alloc and init base part */
85
	blkg = kzalloc_node(sizeof(*blkg), gfp_mask, q->node);
86 87 88
	if (!blkg)
		return NULL;

T
Tejun Heo 已提交
89
	blkg->q = q;
90
	INIT_LIST_HEAD(&blkg->q_node);
91
	blkg->blkcg = blkcg;
T
Tejun Heo 已提交
92
	blkg->refcnt = 1;
93

94 95 96 97 98 99 100
	/* root blkg uses @q->root_rl, init rl only for !root blkgs */
	if (blkcg != &blkcg_root) {
		if (blk_init_rl(&blkg->rl, q, gfp_mask))
			goto err_free;
		blkg->rl.blkg = blkg;
	}

T
Tejun Heo 已提交
101
	for (i = 0; i < BLKCG_MAX_POLS; i++) {
T
Tejun Heo 已提交
102
		struct blkcg_policy *pol = blkcg_policy[i];
103
		struct blkg_policy_data *pd;
104

105
		if (!blkcg_policy_enabled(q, pol))
106 107 108
			continue;

		/* alloc per-policy data and attach it to blkg */
109
		pd = kzalloc_node(pol->pd_size, gfp_mask, q->node);
110 111
		if (!pd)
			goto err_free;
112

113 114 115
		blkg->pd[i] = pd;
		pd->blkg = blkg;

T
Tejun Heo 已提交
116
		/* invoke per-policy init */
T
Tejun Heo 已提交
117
		if (pol->pd_init_fn)
118
			pol->pd_init_fn(blkg);
119 120
	}

121
	return blkg;
122 123 124 125

err_free:
	blkg_free(blkg);
	return NULL;
126 127
}

T
Tejun Heo 已提交
128 129
static struct blkcg_gq *__blkg_lookup(struct blkcg *blkcg,
				      struct request_queue *q)
130
{
T
Tejun Heo 已提交
131
	struct blkcg_gq *blkg;
132

133 134 135 136 137 138 139 140 141 142 143 144 145 146
	blkg = rcu_dereference(blkcg->blkg_hint);
	if (blkg && blkg->q == q)
		return blkg;

	/*
	 * Hint didn't match.  Look up from the radix tree.  Note that we
	 * may not be holding queue_lock and thus are not sure whether
	 * @blkg from blkg_tree has already been removed or not, so we
	 * can't update hint to the lookup result.  Leave it to the caller.
	 */
	blkg = radix_tree_lookup(&blkcg->blkg_tree, q->id);
	if (blkg && blkg->q == q)
		return blkg;

147 148 149 150 151 152 153 154 155 156 157 158
	return NULL;
}

/**
 * blkg_lookup - lookup blkg for the specified blkcg - q pair
 * @blkcg: blkcg of interest
 * @q: request_queue of interest
 *
 * Lookup blkg for the @blkcg - @q pair.  This function should be called
 * under RCU read lock and is guaranteed to return %NULL if @q is bypassing
 * - see blk_queue_bypass_start() for details.
 */
T
Tejun Heo 已提交
159
struct blkcg_gq *blkg_lookup(struct blkcg *blkcg, struct request_queue *q)
160 161 162 163 164 165 166 167 168
{
	WARN_ON_ONCE(!rcu_read_lock_held());

	if (unlikely(blk_queue_bypass(q)))
		return NULL;
	return __blkg_lookup(blkcg, q);
}
EXPORT_SYMBOL_GPL(blkg_lookup);

169 170 171 172
/*
 * If @new_blkg is %NULL, this function tries to allocate a new one as
 * necessary using %GFP_ATOMIC.  @new_blkg is always consumed on return.
 */
T
Tejun Heo 已提交
173
static struct blkcg_gq *__blkg_lookup_create(struct blkcg *blkcg,
174 175
					     struct request_queue *q,
					     struct blkcg_gq *new_blkg)
176
{
T
Tejun Heo 已提交
177
	struct blkcg_gq *blkg;
178
	int ret;
179

180 181 182
	WARN_ON_ONCE(!rcu_read_lock_held());
	lockdep_assert_held(q->queue_lock);

183
	/* lookup and update hint on success, see __blkg_lookup() for details */
184
	blkg = __blkg_lookup(blkcg, q);
185 186
	if (blkg) {
		rcu_assign_pointer(blkcg->blkg_hint, blkg);
187
		goto out_free;
188
	}
189

190
	/* blkg holds a reference to blkcg */
191 192 193 194
	if (!css_tryget(&blkcg->css)) {
		blkg = ERR_PTR(-EINVAL);
		goto out_free;
	}
195

196
	/* allocate */
197 198 199 200 201 202 203 204
	if (!new_blkg) {
		new_blkg = blkg_alloc(blkcg, q, GFP_ATOMIC);
		if (unlikely(!new_blkg)) {
			blkg = ERR_PTR(-ENOMEM);
			goto out_put;
		}
	}
	blkg = new_blkg;
205 206 207

	/* insert */
	spin_lock(&blkcg->lock);
208 209 210 211 212
	ret = radix_tree_insert(&blkcg->blkg_tree, q->id, blkg);
	if (likely(!ret)) {
		hlist_add_head_rcu(&blkg->blkcg_node, &blkcg->blkg_list);
		list_add(&blkg->q_node, &q->blkg_list);
	}
213
	spin_unlock(&blkcg->lock);
214

215 216
	if (!ret)
		return blkg;
217 218 219

	blkg = ERR_PTR(ret);
out_put:
220
	css_put(&blkcg->css);
221 222 223
out_free:
	blkg_free(new_blkg);
	return blkg;
224
}
225

T
Tejun Heo 已提交
226 227
struct blkcg_gq *blkg_lookup_create(struct blkcg *blkcg,
				    struct request_queue *q)
228 229 230 231 232 233
{
	/*
	 * This could be the first entry point of blkcg implementation and
	 * we shouldn't allow anything to go through for a bypassing queue.
	 */
	if (unlikely(blk_queue_bypass(q)))
B
Bart Van Assche 已提交
234
		return ERR_PTR(blk_queue_dying(q) ? -EINVAL : -EBUSY);
235
	return __blkg_lookup_create(blkcg, q, NULL);
236
}
237
EXPORT_SYMBOL_GPL(blkg_lookup_create);
238

T
Tejun Heo 已提交
239
static void blkg_destroy(struct blkcg_gq *blkg)
240
{
T
Tejun Heo 已提交
241
	struct blkcg *blkcg = blkg->blkcg;
242

243
	lockdep_assert_held(blkg->q->queue_lock);
244
	lockdep_assert_held(&blkcg->lock);
245 246

	/* Something wrong if we are trying to remove same group twice */
247
	WARN_ON_ONCE(list_empty(&blkg->q_node));
248
	WARN_ON_ONCE(hlist_unhashed(&blkg->blkcg_node));
249 250

	radix_tree_delete(&blkcg->blkg_tree, blkg->q->id);
251
	list_del_init(&blkg->q_node);
252
	hlist_del_init_rcu(&blkg->blkcg_node);
253

254 255 256 257 258 259 260 261
	/*
	 * Both setting lookup hint to and clearing it from @blkg are done
	 * under queue_lock.  If it's not pointing to @blkg now, it never
	 * will.  Hint assignment itself can race safely.
	 */
	if (rcu_dereference_raw(blkcg->blkg_hint) == blkg)
		rcu_assign_pointer(blkcg->blkg_hint, NULL);

262 263 264 265 266 267 268
	/*
	 * Put the reference taken at the time of creation so that when all
	 * queues are gone, group can be destroyed.
	 */
	blkg_put(blkg);
}

269 270 271 272
/**
 * blkg_destroy_all - destroy all blkgs associated with a request_queue
 * @q: request_queue of interest
 *
273
 * Destroy all blkgs associated with @q.
274
 */
275
static void blkg_destroy_all(struct request_queue *q)
276
{
T
Tejun Heo 已提交
277
	struct blkcg_gq *blkg, *n;
278

279
	lockdep_assert_held(q->queue_lock);
280

281
	list_for_each_entry_safe(blkg, n, &q->blkg_list, q_node) {
T
Tejun Heo 已提交
282
		struct blkcg *blkcg = blkg->blkcg;
283

284 285 286
		spin_lock(&blkcg->lock);
		blkg_destroy(blkg);
		spin_unlock(&blkcg->lock);
287
	}
288 289 290 291 292 293 294

	/*
	 * root blkg is destroyed.  Just clear the pointer since
	 * root_rl does not take reference on root blkg.
	 */
	q->root_blkg = NULL;
	q->root_rl.blkg = NULL;
295 296
}

T
Tejun Heo 已提交
297 298
static void blkg_rcu_free(struct rcu_head *rcu_head)
{
T
Tejun Heo 已提交
299
	blkg_free(container_of(rcu_head, struct blkcg_gq, rcu_head));
T
Tejun Heo 已提交
300 301
}

T
Tejun Heo 已提交
302
void __blkg_release(struct blkcg_gq *blkg)
T
Tejun Heo 已提交
303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319
{
	/* release the extra blkcg reference this blkg has been holding */
	css_put(&blkg->blkcg->css);

	/*
	 * A group is freed in rcu manner. But having an rcu lock does not
	 * mean that one can access all the fields of blkg and assume these
	 * are valid. For example, don't try to follow throtl_data and
	 * request queue links.
	 *
	 * Having a reference to blkg under an rcu allows acess to only
	 * values local to groups like group stats and group rate limits
	 */
	call_rcu(&blkg->rcu_head, blkg_rcu_free);
}
EXPORT_SYMBOL_GPL(__blkg_release);

320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335
/*
 * The next function used by blk_queue_for_each_rl().  It's a bit tricky
 * because the root blkg uses @q->root_rl instead of its own rl.
 */
struct request_list *__blk_queue_next_rl(struct request_list *rl,
					 struct request_queue *q)
{
	struct list_head *ent;
	struct blkcg_gq *blkg;

	/*
	 * Determine the current blkg list_head.  The first entry is
	 * root_rl which is off @q->blkg_list and mapped to the head.
	 */
	if (rl == &q->root_rl) {
		ent = &q->blkg_list;
336 337 338
		/* There are no more block groups, hence no request lists */
		if (list_empty(ent))
			return NULL;
339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354
	} else {
		blkg = container_of(rl, struct blkcg_gq, rl);
		ent = &blkg->q_node;
	}

	/* walk to the next list_head, skip root blkcg */
	ent = ent->next;
	if (ent == &q->root_blkg->q_node)
		ent = ent->next;
	if (ent == &q->blkg_list)
		return NULL;

	blkg = container_of(ent, struct blkcg_gq, q_node);
	return &blkg->rl;
}

T
Tejun Heo 已提交
355 356
static int blkcg_reset_stats(struct cgroup *cgroup, struct cftype *cftype,
			     u64 val)
357
{
T
Tejun Heo 已提交
358 359
	struct blkcg *blkcg = cgroup_to_blkcg(cgroup);
	struct blkcg_gq *blkg;
360
	struct hlist_node *n;
361
	int i;
362

363
	mutex_lock(&blkcg_pol_mutex);
364
	spin_lock_irq(&blkcg->lock);
T
Tejun Heo 已提交
365 366 367 368 369 370

	/*
	 * Note that stat reset is racy - it doesn't synchronize against
	 * stat updates.  This is a debug feature which shouldn't exist
	 * anyway.  If you get hit by a race, retry.
	 */
371
	hlist_for_each_entry(blkg, n, &blkcg->blkg_list, blkcg_node) {
T
Tejun Heo 已提交
372
		for (i = 0; i < BLKCG_MAX_POLS; i++) {
T
Tejun Heo 已提交
373
			struct blkcg_policy *pol = blkcg_policy[i];
374

375
			if (blkcg_policy_enabled(blkg->q, pol) &&
376 377
			    pol->pd_reset_stats_fn)
				pol->pd_reset_stats_fn(blkg);
378
		}
379
	}
380

381
	spin_unlock_irq(&blkcg->lock);
382
	mutex_unlock(&blkcg_pol_mutex);
383 384 385
	return 0;
}

T
Tejun Heo 已提交
386
static const char *blkg_dev_name(struct blkcg_gq *blkg)
387
{
388 389 390 391
	/* some drivers (floppy) instantiate a queue w/o disk registered */
	if (blkg->q->backing_dev_info.dev)
		return dev_name(blkg->q->backing_dev_info.dev);
	return NULL;
392 393
}

394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410
/**
 * blkcg_print_blkgs - helper for printing per-blkg data
 * @sf: seq_file to print to
 * @blkcg: blkcg of interest
 * @prfill: fill function to print out a blkg
 * @pol: policy in question
 * @data: data to be passed to @prfill
 * @show_total: to print out sum of prfill return values or not
 *
 * This function invokes @prfill on each blkg of @blkcg if pd for the
 * policy specified by @pol exists.  @prfill is invoked with @sf, the
 * policy data and @data.  If @show_total is %true, the sum of the return
 * values from @prfill is printed with "Total" label at the end.
 *
 * This is to be used to construct print functions for
 * cftype->read_seq_string method.
 */
T
Tejun Heo 已提交
411
void blkcg_print_blkgs(struct seq_file *sf, struct blkcg *blkcg,
412 413
		       u64 (*prfill)(struct seq_file *,
				     struct blkg_policy_data *, int),
T
Tejun Heo 已提交
414
		       const struct blkcg_policy *pol, int data,
415
		       bool show_total)
416
{
T
Tejun Heo 已提交
417
	struct blkcg_gq *blkg;
418 419
	struct hlist_node *n;
	u64 total = 0;
420

421 422
	spin_lock_irq(&blkcg->lock);
	hlist_for_each_entry(blkg, n, &blkcg->blkg_list, blkcg_node)
423
		if (blkcg_policy_enabled(blkg->q, pol))
424
			total += prfill(sf, blkg->pd[pol->plid], data);
425 426 427 428 429
	spin_unlock_irq(&blkcg->lock);

	if (show_total)
		seq_printf(sf, "Total %llu\n", (unsigned long long)total);
}
430
EXPORT_SYMBOL_GPL(blkcg_print_blkgs);
431 432 433 434

/**
 * __blkg_prfill_u64 - prfill helper for a single u64 value
 * @sf: seq_file to print to
435
 * @pd: policy private data of interest
436 437
 * @v: value to print
 *
438
 * Print @v to @sf for the device assocaited with @pd.
439
 */
440
u64 __blkg_prfill_u64(struct seq_file *sf, struct blkg_policy_data *pd, u64 v)
441
{
442
	const char *dname = blkg_dev_name(pd->blkg);
443 444 445 446 447 448 449

	if (!dname)
		return 0;

	seq_printf(sf, "%s %llu\n", dname, (unsigned long long)v);
	return v;
}
450
EXPORT_SYMBOL_GPL(__blkg_prfill_u64);
451 452 453 454

/**
 * __blkg_prfill_rwstat - prfill helper for a blkg_rwstat
 * @sf: seq_file to print to
455
 * @pd: policy private data of interest
456 457
 * @rwstat: rwstat to print
 *
458
 * Print @rwstat to @sf for the device assocaited with @pd.
459
 */
460
u64 __blkg_prfill_rwstat(struct seq_file *sf, struct blkg_policy_data *pd,
461
			 const struct blkg_rwstat *rwstat)
462 463 464 465 466 467 468
{
	static const char *rwstr[] = {
		[BLKG_RWSTAT_READ]	= "Read",
		[BLKG_RWSTAT_WRITE]	= "Write",
		[BLKG_RWSTAT_SYNC]	= "Sync",
		[BLKG_RWSTAT_ASYNC]	= "Async",
	};
469
	const char *dname = blkg_dev_name(pd->blkg);
470 471 472 473 474 475 476 477 478 479 480 481 482 483 484
	u64 v;
	int i;

	if (!dname)
		return 0;

	for (i = 0; i < BLKG_RWSTAT_NR; i++)
		seq_printf(sf, "%s %s %llu\n", dname, rwstr[i],
			   (unsigned long long)rwstat->cnt[i]);

	v = rwstat->cnt[BLKG_RWSTAT_READ] + rwstat->cnt[BLKG_RWSTAT_WRITE];
	seq_printf(sf, "%s Total %llu\n", dname, (unsigned long long)v);
	return v;
}

485 486 487
/**
 * blkg_prfill_stat - prfill callback for blkg_stat
 * @sf: seq_file to print to
488 489
 * @pd: policy private data of interest
 * @off: offset to the blkg_stat in @pd
490 491 492
 *
 * prfill callback for printing a blkg_stat.
 */
493
u64 blkg_prfill_stat(struct seq_file *sf, struct blkg_policy_data *pd, int off)
494
{
495
	return __blkg_prfill_u64(sf, pd, blkg_stat_read((void *)pd + off));
496
}
497
EXPORT_SYMBOL_GPL(blkg_prfill_stat);
498

499 500 501
/**
 * blkg_prfill_rwstat - prfill callback for blkg_rwstat
 * @sf: seq_file to print to
502 503
 * @pd: policy private data of interest
 * @off: offset to the blkg_rwstat in @pd
504 505 506
 *
 * prfill callback for printing a blkg_rwstat.
 */
507 508
u64 blkg_prfill_rwstat(struct seq_file *sf, struct blkg_policy_data *pd,
		       int off)
509
{
510
	struct blkg_rwstat rwstat = blkg_rwstat_read((void *)pd + off);
511

512
	return __blkg_prfill_rwstat(sf, pd, &rwstat);
513
}
514
EXPORT_SYMBOL_GPL(blkg_prfill_rwstat);
515

516 517 518
/**
 * blkg_conf_prep - parse and prepare for per-blkg config update
 * @blkcg: target block cgroup
519
 * @pol: target policy
520 521 522 523 524
 * @input: input string
 * @ctx: blkg_conf_ctx to be filled
 *
 * Parse per-blkg config update from @input and initialize @ctx with the
 * result.  @ctx->blkg points to the blkg to be updated and @ctx->v the new
525 526
 * value.  This function returns with RCU read lock and queue lock held and
 * must be paired with blkg_conf_finish().
527
 */
T
Tejun Heo 已提交
528 529
int blkg_conf_prep(struct blkcg *blkcg, const struct blkcg_policy *pol,
		   const char *input, struct blkg_conf_ctx *ctx)
530
	__acquires(rcu) __acquires(disk->queue->queue_lock)
531
{
532
	struct gendisk *disk;
T
Tejun Heo 已提交
533
	struct blkcg_gq *blkg;
T
Tejun Heo 已提交
534 535 536
	unsigned int major, minor;
	unsigned long long v;
	int part, ret;
537

T
Tejun Heo 已提交
538 539
	if (sscanf(input, "%u:%u %llu", &major, &minor, &v) != 3)
		return -EINVAL;
540

T
Tejun Heo 已提交
541
	disk = get_gendisk(MKDEV(major, minor), &part);
T
Tejun Heo 已提交
542
	if (!disk || part)
T
Tejun Heo 已提交
543
		return -EINVAL;
544 545

	rcu_read_lock();
T
Tejun Heo 已提交
546
	spin_lock_irq(disk->queue->queue_lock);
547

548
	if (blkcg_policy_enabled(disk->queue, pol))
549
		blkg = blkg_lookup_create(blkcg, disk->queue);
550 551
	else
		blkg = ERR_PTR(-EINVAL);
552

T
Tejun Heo 已提交
553 554
	if (IS_ERR(blkg)) {
		ret = PTR_ERR(blkg);
555
		rcu_read_unlock();
556
		spin_unlock_irq(disk->queue->queue_lock);
557 558 559 560 561 562 563 564 565 566
		put_disk(disk);
		/*
		 * If queue was bypassing, we should retry.  Do so after a
		 * short msleep().  It isn't strictly necessary but queue
		 * can be bypassing for some time and it's always nice to
		 * avoid busy looping.
		 */
		if (ret == -EBUSY) {
			msleep(10);
			ret = restart_syscall();
567
		}
T
Tejun Heo 已提交
568
		return ret;
569
	}
570 571 572

	ctx->disk = disk;
	ctx->blkg = blkg;
T
Tejun Heo 已提交
573 574
	ctx->v = v;
	return 0;
575
}
576
EXPORT_SYMBOL_GPL(blkg_conf_prep);
577

578 579 580 581 582 583 584
/**
 * blkg_conf_finish - finish up per-blkg config update
 * @ctx: blkg_conf_ctx intiailized by blkg_conf_prep()
 *
 * Finish up after per-blkg config update.  This function must be paired
 * with blkg_conf_prep().
 */
585
void blkg_conf_finish(struct blkg_conf_ctx *ctx)
586
	__releases(ctx->disk->queue->queue_lock) __releases(rcu)
587
{
588
	spin_unlock_irq(ctx->disk->queue->queue_lock);
589 590
	rcu_read_unlock();
	put_disk(ctx->disk);
591
}
592
EXPORT_SYMBOL_GPL(blkg_conf_finish);
593

T
Tejun Heo 已提交
594
struct cftype blkcg_files[] = {
595 596
	{
		.name = "reset_stats",
T
Tejun Heo 已提交
597
		.write_u64 = blkcg_reset_stats,
598
	},
599
	{ }	/* terminate */
600 601
};

602
/**
603
 * blkcg_css_offline - cgroup css_offline callback
604 605 606 607 608 609 610 611 612
 * @cgroup: cgroup of interest
 *
 * This function is called when @cgroup is about to go away and responsible
 * for shooting down all blkgs associated with @cgroup.  blkgs should be
 * removed while holding both q and blkcg locks.  As blkcg lock is nested
 * inside q lock, this function performs reverse double lock dancing.
 *
 * This is the blkcg counterpart of ioc_release_fn().
 */
613
static void blkcg_css_offline(struct cgroup *cgroup)
614
{
T
Tejun Heo 已提交
615
	struct blkcg *blkcg = cgroup_to_blkcg(cgroup);
616

617
	spin_lock_irq(&blkcg->lock);
618

619
	while (!hlist_empty(&blkcg->blkg_list)) {
T
Tejun Heo 已提交
620 621
		struct blkcg_gq *blkg = hlist_entry(blkcg->blkg_list.first,
						struct blkcg_gq, blkcg_node);
T
Tejun Heo 已提交
622
		struct request_queue *q = blkg->q;
623

624 625 626 627 628 629
		if (spin_trylock(q->queue_lock)) {
			blkg_destroy(blkg);
			spin_unlock(q->queue_lock);
		} else {
			spin_unlock_irq(&blkcg->lock);
			cpu_relax();
630
			spin_lock_irq(&blkcg->lock);
631
		}
632
	}
633

634
	spin_unlock_irq(&blkcg->lock);
635 636
}

637
static void blkcg_css_free(struct cgroup *cgroup)
638
{
T
Tejun Heo 已提交
639
	struct blkcg *blkcg = cgroup_to_blkcg(cgroup);
640

T
Tejun Heo 已提交
641
	if (blkcg != &blkcg_root)
B
Ben Blum 已提交
642
		kfree(blkcg);
643 644
}

645
static struct cgroup_subsys_state *blkcg_css_alloc(struct cgroup *cgroup)
646
{
T
Tejun Heo 已提交
647
	static atomic64_t id_seq = ATOMIC64_INIT(0);
T
Tejun Heo 已提交
648
	struct blkcg *blkcg;
649
	struct cgroup *parent = cgroup->parent;
650

651
	if (!parent) {
T
Tejun Heo 已提交
652
		blkcg = &blkcg_root;
653 654 655 656 657 658 659
		goto done;
	}

	blkcg = kzalloc(sizeof(*blkcg), GFP_KERNEL);
	if (!blkcg)
		return ERR_PTR(-ENOMEM);

660
	blkcg->cfq_weight = CFQ_WEIGHT_DEFAULT;
T
Tejun Heo 已提交
661
	blkcg->id = atomic64_inc_return(&id_seq); /* root is 0, start from 1 */
662 663
done:
	spin_lock_init(&blkcg->lock);
664
	INIT_RADIX_TREE(&blkcg->blkg_tree, GFP_ATOMIC);
665 666 667 668 669
	INIT_HLIST_HEAD(&blkcg->blkg_list);

	return &blkcg->css;
}

670 671 672 673 674 675 676 677 678 679 680 681 682 683
/**
 * blkcg_init_queue - initialize blkcg part of request queue
 * @q: request_queue to initialize
 *
 * Called from blk_alloc_queue_node(). Responsible for initializing blkcg
 * part of new request_queue @q.
 *
 * RETURNS:
 * 0 on success, -errno on failure.
 */
int blkcg_init_queue(struct request_queue *q)
{
	might_sleep();

684
	return blk_throtl_init(q);
685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707
}

/**
 * blkcg_drain_queue - drain blkcg part of request_queue
 * @q: request_queue to drain
 *
 * Called from blk_drain_queue().  Responsible for draining blkcg part.
 */
void blkcg_drain_queue(struct request_queue *q)
{
	lockdep_assert_held(q->queue_lock);

	blk_throtl_drain(q);
}

/**
 * blkcg_exit_queue - exit and release blkcg part of request_queue
 * @q: request_queue being released
 *
 * Called from blk_release_queue().  Responsible for exiting blkcg part.
 */
void blkcg_exit_queue(struct request_queue *q)
{
708
	spin_lock_irq(q->queue_lock);
709
	blkg_destroy_all(q);
710 711
	spin_unlock_irq(q->queue_lock);

712 713 714
	blk_throtl_exit(q);
}

715 716 717 718 719 720
/*
 * We cannot support shared io contexts, as we have no mean to support
 * two tasks with the same ioc in two different groups without major rework
 * of the main cic data structures.  For now we allow a task to change
 * its cgroup only if it's the only owner of its ioc.
 */
T
Tejun Heo 已提交
721
static int blkcg_can_attach(struct cgroup *cgrp, struct cgroup_taskset *tset)
722
{
723
	struct task_struct *task;
724 725 726 727
	struct io_context *ioc;
	int ret = 0;

	/* task_lock() is needed to avoid races with exit_io_context() */
728 729 730 731 732 733 734 735 736
	cgroup_taskset_for_each(task, cgrp, tset) {
		task_lock(task);
		ioc = task->io_context;
		if (ioc && atomic_read(&ioc->nr_tasks) > 1)
			ret = -EINVAL;
		task_unlock(task);
		if (ret)
			break;
	}
737 738 739
	return ret;
}

740 741
struct cgroup_subsys blkio_subsys = {
	.name = "blkio",
742 743 744
	.css_alloc = blkcg_css_alloc,
	.css_offline = blkcg_css_offline,
	.css_free = blkcg_css_free,
T
Tejun Heo 已提交
745
	.can_attach = blkcg_can_attach,
746
	.subsys_id = blkio_subsys_id,
T
Tejun Heo 已提交
747
	.base_cftypes = blkcg_files,
748
	.module = THIS_MODULE,
749 750 751 752 753 754 755 756

	/*
	 * blkio subsystem is utterly broken in terms of hierarchy support.
	 * It treats all cgroups equally regardless of where they're
	 * located in the hierarchy - all cgroups are treated as if they're
	 * right below the root.  Fix it and remove the following.
	 */
	.broken_hierarchy = true,
757 758 759
};
EXPORT_SYMBOL_GPL(blkio_subsys);

760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776
/**
 * blkcg_activate_policy - activate a blkcg policy on a request_queue
 * @q: request_queue of interest
 * @pol: blkcg policy to activate
 *
 * Activate @pol on @q.  Requires %GFP_KERNEL context.  @q goes through
 * bypass mode to populate its blkgs with policy_data for @pol.
 *
 * Activation happens with @q bypassed, so nobody would be accessing blkgs
 * from IO path.  Update of each blkg is protected by both queue and blkcg
 * locks so that holding either lock and testing blkcg_policy_enabled() is
 * always enough for dereferencing policy data.
 *
 * The caller is responsible for synchronizing [de]activations and policy
 * [un]registerations.  Returns 0 on success, -errno on failure.
 */
int blkcg_activate_policy(struct request_queue *q,
T
Tejun Heo 已提交
777
			  const struct blkcg_policy *pol)
778 779
{
	LIST_HEAD(pds);
T
Tejun Heo 已提交
780
	struct blkcg_gq *blkg;
781 782
	struct blkg_policy_data *pd, *n;
	int cnt = 0, ret;
783
	bool preloaded;
784 785 786 787

	if (blkcg_policy_enabled(q, pol))
		return 0;

788 789 790 791 792 793 794
	/* preallocations for root blkg */
	blkg = blkg_alloc(&blkcg_root, q, GFP_KERNEL);
	if (!blkg)
		return -ENOMEM;

	preloaded = !radix_tree_preload(GFP_KERNEL);

795 796 797 798 799 800
	blk_queue_bypass_start(q);

	/* make sure the root blkg exists and count the existing blkgs */
	spin_lock_irq(q->queue_lock);

	rcu_read_lock();
801
	blkg = __blkg_lookup_create(&blkcg_root, q, blkg);
802 803
	rcu_read_unlock();

804 805 806
	if (preloaded)
		radix_tree_preload_end();

807 808 809 810 811
	if (IS_ERR(blkg)) {
		ret = PTR_ERR(blkg);
		goto out_unlock;
	}
	q->root_blkg = blkg;
812
	q->root_rl.blkg = blkg;
813 814 815 816 817 818 819 820

	list_for_each_entry(blkg, &q->blkg_list, q_node)
		cnt++;

	spin_unlock_irq(q->queue_lock);

	/* allocate policy_data for all existing blkgs */
	while (cnt--) {
821
		pd = kzalloc_node(pol->pd_size, GFP_KERNEL, q->node);
822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848
		if (!pd) {
			ret = -ENOMEM;
			goto out_free;
		}
		list_add_tail(&pd->alloc_node, &pds);
	}

	/*
	 * Install the allocated pds.  With @q bypassing, no new blkg
	 * should have been created while the queue lock was dropped.
	 */
	spin_lock_irq(q->queue_lock);

	list_for_each_entry(blkg, &q->blkg_list, q_node) {
		if (WARN_ON(list_empty(&pds))) {
			/* umm... this shouldn't happen, just abort */
			ret = -ENOMEM;
			goto out_unlock;
		}
		pd = list_first_entry(&pds, struct blkg_policy_data, alloc_node);
		list_del_init(&pd->alloc_node);

		/* grab blkcg lock too while installing @pd on @blkg */
		spin_lock(&blkg->blkcg->lock);

		blkg->pd[pol->plid] = pd;
		pd->blkg = blkg;
849
		pol->pd_init_fn(blkg);
850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874

		spin_unlock(&blkg->blkcg->lock);
	}

	__set_bit(pol->plid, q->blkcg_pols);
	ret = 0;
out_unlock:
	spin_unlock_irq(q->queue_lock);
out_free:
	blk_queue_bypass_end(q);
	list_for_each_entry_safe(pd, n, &pds, alloc_node)
		kfree(pd);
	return ret;
}
EXPORT_SYMBOL_GPL(blkcg_activate_policy);

/**
 * blkcg_deactivate_policy - deactivate a blkcg policy on a request_queue
 * @q: request_queue of interest
 * @pol: blkcg policy to deactivate
 *
 * Deactivate @pol on @q.  Follows the same synchronization rules as
 * blkcg_activate_policy().
 */
void blkcg_deactivate_policy(struct request_queue *q,
T
Tejun Heo 已提交
875
			     const struct blkcg_policy *pol)
876
{
T
Tejun Heo 已提交
877
	struct blkcg_gq *blkg;
878 879 880 881 882 883 884 885 886

	if (!blkcg_policy_enabled(q, pol))
		return;

	blk_queue_bypass_start(q);
	spin_lock_irq(q->queue_lock);

	__clear_bit(pol->plid, q->blkcg_pols);

887 888 889 890
	/* if no policy is left, no need for blkgs - shoot them down */
	if (bitmap_empty(q->blkcg_pols, BLKCG_MAX_POLS))
		blkg_destroy_all(q);

891 892 893 894
	list_for_each_entry(blkg, &q->blkg_list, q_node) {
		/* grab blkcg lock too while removing @pd from @blkg */
		spin_lock(&blkg->blkcg->lock);

895 896
		if (pol->pd_exit_fn)
			pol->pd_exit_fn(blkg);
897 898 899 900 901 902 903 904 905 906 907 908

		kfree(blkg->pd[pol->plid]);
		blkg->pd[pol->plid] = NULL;

		spin_unlock(&blkg->blkcg->lock);
	}

	spin_unlock_irq(q->queue_lock);
	blk_queue_bypass_end(q);
}
EXPORT_SYMBOL_GPL(blkcg_deactivate_policy);

T
Tejun Heo 已提交
909
/**
T
Tejun Heo 已提交
910 911
 * blkcg_policy_register - register a blkcg policy
 * @pol: blkcg policy to register
T
Tejun Heo 已提交
912
 *
T
Tejun Heo 已提交
913 914
 * Register @pol with blkcg core.  Might sleep and @pol may be modified on
 * successful registration.  Returns 0 on success and -errno on failure.
T
Tejun Heo 已提交
915
 */
T
Tejun Heo 已提交
916
int blkcg_policy_register(struct blkcg_policy *pol)
917
{
T
Tejun Heo 已提交
918
	int i, ret;
919

920 921 922
	if (WARN_ON(pol->pd_size < sizeof(struct blkg_policy_data)))
		return -EINVAL;

923 924
	mutex_lock(&blkcg_pol_mutex);

T
Tejun Heo 已提交
925 926 927
	/* find an empty slot */
	ret = -ENOSPC;
	for (i = 0; i < BLKCG_MAX_POLS; i++)
T
Tejun Heo 已提交
928
		if (!blkcg_policy[i])
T
Tejun Heo 已提交
929 930 931
			break;
	if (i >= BLKCG_MAX_POLS)
		goto out_unlock;
932

T
Tejun Heo 已提交
933
	/* register and update blkgs */
T
Tejun Heo 已提交
934 935
	pol->plid = i;
	blkcg_policy[i] = pol;
T
Tejun Heo 已提交
936 937

	/* everything is in place, add intf files for the new policy */
T
Tejun Heo 已提交
938 939
	if (pol->cftypes)
		WARN_ON(cgroup_add_cftypes(&blkio_subsys, pol->cftypes));
T
Tejun Heo 已提交
940 941
	ret = 0;
out_unlock:
942
	mutex_unlock(&blkcg_pol_mutex);
T
Tejun Heo 已提交
943
	return ret;
944
}
T
Tejun Heo 已提交
945
EXPORT_SYMBOL_GPL(blkcg_policy_register);
946

T
Tejun Heo 已提交
947
/**
T
Tejun Heo 已提交
948 949
 * blkcg_policy_unregister - unregister a blkcg policy
 * @pol: blkcg policy to unregister
T
Tejun Heo 已提交
950
 *
T
Tejun Heo 已提交
951
 * Undo blkcg_policy_register(@pol).  Might sleep.
T
Tejun Heo 已提交
952
 */
T
Tejun Heo 已提交
953
void blkcg_policy_unregister(struct blkcg_policy *pol)
954
{
955 956
	mutex_lock(&blkcg_pol_mutex);

T
Tejun Heo 已提交
957
	if (WARN_ON(blkcg_policy[pol->plid] != pol))
T
Tejun Heo 已提交
958 959 960
		goto out_unlock;

	/* kill the intf files first */
T
Tejun Heo 已提交
961 962
	if (pol->cftypes)
		cgroup_rm_cftypes(&blkio_subsys, pol->cftypes);
963

T
Tejun Heo 已提交
964
	/* unregister and update blkgs */
T
Tejun Heo 已提交
965
	blkcg_policy[pol->plid] = NULL;
T
Tejun Heo 已提交
966
out_unlock:
967
	mutex_unlock(&blkcg_pol_mutex);
968
}
T
Tejun Heo 已提交
969
EXPORT_SYMBOL_GPL(blkcg_policy_unregister);