blk-cgroup.c 25 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>
Tejun Heo's avatar
Tejun Heo committed
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

Tejun Heo's avatar
Tejun Heo committed
29
30
struct blkcg blkcg_root = { .cfq_weight = 2 * CFQ_WEIGHT_DEFAULT };
EXPORT_SYMBOL_GPL(blkcg_root);
31

Tejun Heo's avatar
Tejun Heo committed
32
static struct blkcg_policy *blkcg_policy[BLKCG_MAX_POLS];
33

34
static bool blkcg_policy_enabled(struct request_queue *q,
Tejun Heo's avatar
Tejun Heo committed
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.
 */
Tejun Heo's avatar
Tejun Heo committed
46
static void blkg_free(struct blkcg_gq *blkg)
47
{
48
	int i;
49
50
51
52

	if (!blkg)
		return;

53
	for (i = 0; i < BLKCG_MAX_POLS; i++) {
Tejun Heo's avatar
Tejun Heo committed
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
{
Tejun Heo's avatar
Tejun Heo committed
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;

89
	blkg->q = q;
90
	INIT_LIST_HEAD(&blkg->q_node);
91
	blkg->blkcg = blkcg;
Tejun Heo's avatar
Tejun Heo committed
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;
	}

101
	for (i = 0; i < BLKCG_MAX_POLS; i++) {
Tejun Heo's avatar
Tejun Heo committed
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;

116
		/* invoke per-policy init */
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
}

Tejun Heo's avatar
Tejun Heo committed
128
static struct blkcg_gq *__blkg_lookup(struct blkcg *blkcg,
129
				      struct request_queue *q, bool update_hint)
130
{
Tejun Heo's avatar
Tejun Heo committed
131
	struct blkcg_gq *blkg;
132

133
134
135
136
137
	blkg = rcu_dereference(blkcg->blkg_hint);
	if (blkg && blkg->q == q)
		return blkg;

	/*
138
139
140
141
	 * Hint didn't match.  Look up from the radix tree.  Note that the
	 * hint can only be updated under queue_lock as otherwise @blkg
	 * could have already been removed from blkg_tree.  The caller is
	 * responsible for grabbing queue_lock if @update_hint.
142
143
	 */
	blkg = radix_tree_lookup(&blkcg->blkg_tree, q->id);
144
145
146
147
148
	if (blkg && blkg->q == q) {
		if (update_hint) {
			lockdep_assert_held(q->queue_lock);
			rcu_assign_pointer(blkcg->blkg_hint, blkg);
		}
149
		return blkg;
150
	}
151

152
153
154
155
156
157
158
159
160
161
162
163
	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.
 */
Tejun Heo's avatar
Tejun Heo committed
164
struct blkcg_gq *blkg_lookup(struct blkcg *blkcg, struct request_queue *q)
165
166
167
168
169
{
	WARN_ON_ONCE(!rcu_read_lock_held());

	if (unlikely(blk_queue_bypass(q)))
		return NULL;
170
	return __blkg_lookup(blkcg, q, false);
171
172
173
}
EXPORT_SYMBOL_GPL(blkg_lookup);

174
175
176
177
/*
 * 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.
 */
178
179
180
static struct blkcg_gq *blkg_create(struct blkcg *blkcg,
				    struct request_queue *q,
				    struct blkcg_gq *new_blkg)
181
{
Tejun Heo's avatar
Tejun Heo committed
182
	struct blkcg_gq *blkg;
183
	int ret;
184

185
186
187
	WARN_ON_ONCE(!rcu_read_lock_held());
	lockdep_assert_held(q->queue_lock);

188
	/* blkg holds a reference to blkcg */
189
190
191
192
	if (!css_tryget(&blkcg->css)) {
		blkg = ERR_PTR(-EINVAL);
		goto out_free;
	}
193

194
	/* allocate */
195
196
197
198
199
200
201
202
	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;
203
204
205

	/* insert */
	spin_lock(&blkcg->lock);
206
207
208
209
210
	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);
	}
211
	spin_unlock(&blkcg->lock);
212

213
214
	if (!ret)
		return blkg;
215
216
217

	blkg = ERR_PTR(ret);
out_put:
218
	css_put(&blkcg->css);
219
220
221
out_free:
	blkg_free(new_blkg);
	return blkg;
222
}
223

224
225
226
227
228
229
230
231
232
233
234
235
236
/**
 * blkg_lookup_create - lookup blkg, try to create one if not there
 * @blkcg: blkcg of interest
 * @q: request_queue of interest
 *
 * Lookup blkg for the @blkcg - @q pair.  If it doesn't exist, try to
 * create one.  This function should be called under RCU read lock and
 * @q->queue_lock.
 *
 * Returns pointer to the looked up or created blkg on success, ERR_PTR()
 * value on error.  If @q is dead, returns ERR_PTR(-EINVAL).  If @q is not
 * dead and bypassing, returns ERR_PTR(-EBUSY).
 */
Tejun Heo's avatar
Tejun Heo committed
237
238
struct blkcg_gq *blkg_lookup_create(struct blkcg *blkcg,
				    struct request_queue *q)
239
{
240
241
242
243
244
	struct blkcg_gq *blkg;

	WARN_ON_ONCE(!rcu_read_lock_held());
	lockdep_assert_held(q->queue_lock);

245
246
247
248
249
	/*
	 * 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)))
250
		return ERR_PTR(blk_queue_dying(q) ? -EINVAL : -EBUSY);
251
252
253
254
255
256

	blkg = __blkg_lookup(blkcg, q, true);
	if (blkg)
		return blkg;

	return blkg_create(blkcg, q, NULL);
257
}
258
EXPORT_SYMBOL_GPL(blkg_lookup_create);
259

Tejun Heo's avatar
Tejun Heo committed
260
static void blkg_destroy(struct blkcg_gq *blkg)
261
{
Tejun Heo's avatar
Tejun Heo committed
262
	struct blkcg *blkcg = blkg->blkcg;
263

264
	lockdep_assert_held(blkg->q->queue_lock);
265
	lockdep_assert_held(&blkcg->lock);
266
267

	/* Something wrong if we are trying to remove same group twice */
268
	WARN_ON_ONCE(list_empty(&blkg->q_node));
269
	WARN_ON_ONCE(hlist_unhashed(&blkg->blkcg_node));
270
271

	radix_tree_delete(&blkcg->blkg_tree, blkg->q->id);
272
	list_del_init(&blkg->q_node);
273
	hlist_del_init_rcu(&blkg->blkcg_node);
274

275
276
277
278
279
280
281
282
	/*
	 * 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);

283
284
285
286
287
288
289
	/*
	 * Put the reference taken at the time of creation so that when all
	 * queues are gone, group can be destroyed.
	 */
	blkg_put(blkg);
}

290
291
292
293
/**
 * blkg_destroy_all - destroy all blkgs associated with a request_queue
 * @q: request_queue of interest
 *
294
 * Destroy all blkgs associated with @q.
295
 */
296
static void blkg_destroy_all(struct request_queue *q)
297
{
Tejun Heo's avatar
Tejun Heo committed
298
	struct blkcg_gq *blkg, *n;
299

300
	lockdep_assert_held(q->queue_lock);
301

302
	list_for_each_entry_safe(blkg, n, &q->blkg_list, q_node) {
Tejun Heo's avatar
Tejun Heo committed
303
		struct blkcg *blkcg = blkg->blkcg;
304

305
306
307
		spin_lock(&blkcg->lock);
		blkg_destroy(blkg);
		spin_unlock(&blkcg->lock);
308
	}
309
310
311
312
313
314
315

	/*
	 * 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;
316
317
}

Tejun Heo's avatar
Tejun Heo committed
318
319
static void blkg_rcu_free(struct rcu_head *rcu_head)
{
Tejun Heo's avatar
Tejun Heo committed
320
	blkg_free(container_of(rcu_head, struct blkcg_gq, rcu_head));
Tejun Heo's avatar
Tejun Heo committed
321
322
}

Tejun Heo's avatar
Tejun Heo committed
323
void __blkg_release(struct blkcg_gq *blkg)
Tejun Heo's avatar
Tejun Heo committed
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
{
	/* 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);

341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
/*
 * 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;
357
358
359
		/* There are no more block groups, hence no request lists */
		if (list_empty(ent))
			return NULL;
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
	} 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;
}

Tejun Heo's avatar
Tejun Heo committed
376
377
static int blkcg_reset_stats(struct cgroup *cgroup, struct cftype *cftype,
			     u64 val)
378
{
Tejun Heo's avatar
Tejun Heo committed
379
380
	struct blkcg *blkcg = cgroup_to_blkcg(cgroup);
	struct blkcg_gq *blkg;
381
	struct hlist_node *n;
382
	int i;
383

384
	mutex_lock(&blkcg_pol_mutex);
385
	spin_lock_irq(&blkcg->lock);
Tejun Heo's avatar
Tejun Heo committed
386
387
388
389
390
391

	/*
	 * 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.
	 */
392
	hlist_for_each_entry(blkg, n, &blkcg->blkg_list, blkcg_node) {
393
		for (i = 0; i < BLKCG_MAX_POLS; i++) {
Tejun Heo's avatar
Tejun Heo committed
394
			struct blkcg_policy *pol = blkcg_policy[i];
395

396
			if (blkcg_policy_enabled(blkg->q, pol) &&
397
398
			    pol->pd_reset_stats_fn)
				pol->pd_reset_stats_fn(blkg);
399
		}
400
	}
401

402
	spin_unlock_irq(&blkcg->lock);
403
	mutex_unlock(&blkcg_pol_mutex);
404
405
406
	return 0;
}

Tejun Heo's avatar
Tejun Heo committed
407
static const char *blkg_dev_name(struct blkcg_gq *blkg)
408
{
409
410
411
412
	/* 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;
413
414
}

415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
/**
 * 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.
 */
Tejun Heo's avatar
Tejun Heo committed
432
void blkcg_print_blkgs(struct seq_file *sf, struct blkcg *blkcg,
433
434
		       u64 (*prfill)(struct seq_file *,
				     struct blkg_policy_data *, int),
Tejun Heo's avatar
Tejun Heo committed
435
		       const struct blkcg_policy *pol, int data,
436
		       bool show_total)
437
{
Tejun Heo's avatar
Tejun Heo committed
438
	struct blkcg_gq *blkg;
439
440
	struct hlist_node *n;
	u64 total = 0;
441

442
443
	spin_lock_irq(&blkcg->lock);
	hlist_for_each_entry(blkg, n, &blkcg->blkg_list, blkcg_node)
444
		if (blkcg_policy_enabled(blkg->q, pol))
445
			total += prfill(sf, blkg->pd[pol->plid], data);
446
447
448
449
450
	spin_unlock_irq(&blkcg->lock);

	if (show_total)
		seq_printf(sf, "Total %llu\n", (unsigned long long)total);
}
451
EXPORT_SYMBOL_GPL(blkcg_print_blkgs);
452
453
454
455

/**
 * __blkg_prfill_u64 - prfill helper for a single u64 value
 * @sf: seq_file to print to
456
 * @pd: policy private data of interest
457
458
 * @v: value to print
 *
459
 * Print @v to @sf for the device assocaited with @pd.
460
 */
461
u64 __blkg_prfill_u64(struct seq_file *sf, struct blkg_policy_data *pd, u64 v)
462
{
463
	const char *dname = blkg_dev_name(pd->blkg);
464
465
466
467
468
469
470

	if (!dname)
		return 0;

	seq_printf(sf, "%s %llu\n", dname, (unsigned long long)v);
	return v;
}
471
EXPORT_SYMBOL_GPL(__blkg_prfill_u64);
472
473
474
475

/**
 * __blkg_prfill_rwstat - prfill helper for a blkg_rwstat
 * @sf: seq_file to print to
476
 * @pd: policy private data of interest
477
478
 * @rwstat: rwstat to print
 *
479
 * Print @rwstat to @sf for the device assocaited with @pd.
480
 */
481
u64 __blkg_prfill_rwstat(struct seq_file *sf, struct blkg_policy_data *pd,
482
			 const struct blkg_rwstat *rwstat)
483
484
485
486
487
488
489
{
	static const char *rwstr[] = {
		[BLKG_RWSTAT_READ]	= "Read",
		[BLKG_RWSTAT_WRITE]	= "Write",
		[BLKG_RWSTAT_SYNC]	= "Sync",
		[BLKG_RWSTAT_ASYNC]	= "Async",
	};
490
	const char *dname = blkg_dev_name(pd->blkg);
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
	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;
}

506
507
508
/**
 * blkg_prfill_stat - prfill callback for blkg_stat
 * @sf: seq_file to print to
509
510
 * @pd: policy private data of interest
 * @off: offset to the blkg_stat in @pd
511
512
513
 *
 * prfill callback for printing a blkg_stat.
 */
514
u64 blkg_prfill_stat(struct seq_file *sf, struct blkg_policy_data *pd, int off)
515
{
516
	return __blkg_prfill_u64(sf, pd, blkg_stat_read((void *)pd + off));
517
}
518
EXPORT_SYMBOL_GPL(blkg_prfill_stat);
519

520
521
522
/**
 * blkg_prfill_rwstat - prfill callback for blkg_rwstat
 * @sf: seq_file to print to
523
524
 * @pd: policy private data of interest
 * @off: offset to the blkg_rwstat in @pd
525
526
527
 *
 * prfill callback for printing a blkg_rwstat.
 */
528
529
u64 blkg_prfill_rwstat(struct seq_file *sf, struct blkg_policy_data *pd,
		       int off)
530
{
531
	struct blkg_rwstat rwstat = blkg_rwstat_read((void *)pd + off);
532

533
	return __blkg_prfill_rwstat(sf, pd, &rwstat);
534
}
535
EXPORT_SYMBOL_GPL(blkg_prfill_rwstat);
536

537
538
539
/**
 * blkg_conf_prep - parse and prepare for per-blkg config update
 * @blkcg: target block cgroup
540
 * @pol: target policy
541
542
543
544
545
 * @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
546
547
 * value.  This function returns with RCU read lock and queue lock held and
 * must be paired with blkg_conf_finish().
548
 */
Tejun Heo's avatar
Tejun Heo committed
549
550
int blkg_conf_prep(struct blkcg *blkcg, const struct blkcg_policy *pol,
		   const char *input, struct blkg_conf_ctx *ctx)
551
	__acquires(rcu) __acquires(disk->queue->queue_lock)
552
{
553
	struct gendisk *disk;
Tejun Heo's avatar
Tejun Heo committed
554
	struct blkcg_gq *blkg;
Tejun Heo's avatar
Tejun Heo committed
555
556
557
	unsigned int major, minor;
	unsigned long long v;
	int part, ret;
558

Tejun Heo's avatar
Tejun Heo committed
559
560
	if (sscanf(input, "%u:%u %llu", &major, &minor, &v) != 3)
		return -EINVAL;
561

Tejun Heo's avatar
Tejun Heo committed
562
	disk = get_gendisk(MKDEV(major, minor), &part);
Tejun Heo's avatar
Tejun Heo committed
563
	if (!disk || part)
Tejun Heo's avatar
Tejun Heo committed
564
		return -EINVAL;
565
566

	rcu_read_lock();
Tejun Heo's avatar
Tejun Heo committed
567
	spin_lock_irq(disk->queue->queue_lock);
568

569
	if (blkcg_policy_enabled(disk->queue, pol))
570
		blkg = blkg_lookup_create(blkcg, disk->queue);
571
572
	else
		blkg = ERR_PTR(-EINVAL);
573

Tejun Heo's avatar
Tejun Heo committed
574
575
	if (IS_ERR(blkg)) {
		ret = PTR_ERR(blkg);
576
		rcu_read_unlock();
577
		spin_unlock_irq(disk->queue->queue_lock);
578
579
580
581
582
583
584
585
586
587
		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();
588
		}
Tejun Heo's avatar
Tejun Heo committed
589
		return ret;
590
	}
591
592
593

	ctx->disk = disk;
	ctx->blkg = blkg;
Tejun Heo's avatar
Tejun Heo committed
594
595
	ctx->v = v;
	return 0;
596
}
597
EXPORT_SYMBOL_GPL(blkg_conf_prep);
598

599
600
601
602
603
604
605
/**
 * 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().
 */
606
void blkg_conf_finish(struct blkg_conf_ctx *ctx)
607
	__releases(ctx->disk->queue->queue_lock) __releases(rcu)
608
{
609
	spin_unlock_irq(ctx->disk->queue->queue_lock);
610
611
	rcu_read_unlock();
	put_disk(ctx->disk);
612
}
613
EXPORT_SYMBOL_GPL(blkg_conf_finish);
614

Tejun Heo's avatar
Tejun Heo committed
615
struct cftype blkcg_files[] = {
616
617
	{
		.name = "reset_stats",
Tejun Heo's avatar
Tejun Heo committed
618
		.write_u64 = blkcg_reset_stats,
619
	},
620
	{ }	/* terminate */
621
622
};

623
/**
624
 * blkcg_css_offline - cgroup css_offline callback
625
626
627
628
629
630
631
632
633
 * @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().
 */
634
static void blkcg_css_offline(struct cgroup *cgroup)
635
{
Tejun Heo's avatar
Tejun Heo committed
636
	struct blkcg *blkcg = cgroup_to_blkcg(cgroup);
637

638
	spin_lock_irq(&blkcg->lock);
639

640
	while (!hlist_empty(&blkcg->blkg_list)) {
Tejun Heo's avatar
Tejun Heo committed
641
642
		struct blkcg_gq *blkg = hlist_entry(blkcg->blkg_list.first,
						struct blkcg_gq, blkcg_node);
643
		struct request_queue *q = blkg->q;
644

645
646
647
648
649
650
		if (spin_trylock(q->queue_lock)) {
			blkg_destroy(blkg);
			spin_unlock(q->queue_lock);
		} else {
			spin_unlock_irq(&blkcg->lock);
			cpu_relax();
651
			spin_lock_irq(&blkcg->lock);
652
		}
653
	}
654

655
	spin_unlock_irq(&blkcg->lock);
656
657
}

658
static void blkcg_css_free(struct cgroup *cgroup)
659
{
Tejun Heo's avatar
Tejun Heo committed
660
	struct blkcg *blkcg = cgroup_to_blkcg(cgroup);
661

Tejun Heo's avatar
Tejun Heo committed
662
	if (blkcg != &blkcg_root)
Ben Blum's avatar
Ben Blum committed
663
		kfree(blkcg);
664
665
}

666
static struct cgroup_subsys_state *blkcg_css_alloc(struct cgroup *cgroup)
667
{
Tejun Heo's avatar
Tejun Heo committed
668
	static atomic64_t id_seq = ATOMIC64_INIT(0);
Tejun Heo's avatar
Tejun Heo committed
669
	struct blkcg *blkcg;
670
	struct cgroup *parent = cgroup->parent;
671

672
	if (!parent) {
Tejun Heo's avatar
Tejun Heo committed
673
		blkcg = &blkcg_root;
674
675
676
677
678
679
680
		goto done;
	}

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

681
	blkcg->cfq_weight = CFQ_WEIGHT_DEFAULT;
Tejun Heo's avatar
Tejun Heo committed
682
	blkcg->id = atomic64_inc_return(&id_seq); /* root is 0, start from 1 */
683
684
done:
	spin_lock_init(&blkcg->lock);
685
	INIT_RADIX_TREE(&blkcg->blkg_tree, GFP_ATOMIC);
686
687
688
689
690
	INIT_HLIST_HEAD(&blkcg->blkg_list);

	return &blkcg->css;
}

691
692
693
694
695
696
697
698
699
700
701
702
703
704
/**
 * 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();

705
	return blk_throtl_init(q);
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
}

/**
 * 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)
{
729
	spin_lock_irq(q->queue_lock);
730
	blkg_destroy_all(q);
731
732
	spin_unlock_irq(q->queue_lock);

733
734
735
	blk_throtl_exit(q);
}

736
737
738
739
740
741
/*
 * 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.
 */
Tejun Heo's avatar
Tejun Heo committed
742
static int blkcg_can_attach(struct cgroup *cgrp, struct cgroup_taskset *tset)
743
{
744
	struct task_struct *task;
745
746
747
748
	struct io_context *ioc;
	int ret = 0;

	/* task_lock() is needed to avoid races with exit_io_context() */
749
750
751
752
753
754
755
756
757
	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;
	}
758
759
760
	return ret;
}

761
762
struct cgroup_subsys blkio_subsys = {
	.name = "blkio",
763
764
765
	.css_alloc = blkcg_css_alloc,
	.css_offline = blkcg_css_offline,
	.css_free = blkcg_css_free,
Tejun Heo's avatar
Tejun Heo committed
766
	.can_attach = blkcg_can_attach,
767
	.subsys_id = blkio_subsys_id,
Tejun Heo's avatar
Tejun Heo committed
768
	.base_cftypes = blkcg_files,
769
	.module = THIS_MODULE,
770
771
772
773
774
775
776
777

	/*
	 * 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,
778
779
780
};
EXPORT_SYMBOL_GPL(blkio_subsys);

781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
/**
 * 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,
Tejun Heo's avatar
Tejun Heo committed
798
			  const struct blkcg_policy *pol)
799
800
{
	LIST_HEAD(pds);
801
	struct blkcg_gq *blkg, *new_blkg;
802
803
	struct blkg_policy_data *pd, *n;
	int cnt = 0, ret;
804
	bool preloaded;
805
806
807
808

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

809
	/* preallocations for root blkg */
810
811
	new_blkg = blkg_alloc(&blkcg_root, q, GFP_KERNEL);
	if (!new_blkg)
812
813
814
815
		return -ENOMEM;

	preloaded = !radix_tree_preload(GFP_KERNEL);

816
817
	blk_queue_bypass_start(q);

818
819
820
821
822
	/*
	 * Make sure the root blkg exists and count the existing blkgs.  As
	 * @q is bypassing at this point, blkg_lookup_create() can't be
	 * used.  Open code it.
	 */
823
824
825
	spin_lock_irq(q->queue_lock);

	rcu_read_lock();
826
827
828
829
830
	blkg = __blkg_lookup(&blkcg_root, q, false);
	if (blkg)
		blkg_free(new_blkg);
	else
		blkg = blkg_create(&blkcg_root, q, new_blkg);
831
832
	rcu_read_unlock();

833
834
835
	if (preloaded)
		radix_tree_preload_end();

836
837
838
839
840
	if (IS_ERR(blkg)) {
		ret = PTR_ERR(blkg);
		goto out_unlock;
	}
	q->root_blkg = blkg;
841
	q->root_rl.blkg = blkg;
842
843
844
845
846
847
848
849

	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--) {
850
		pd = kzalloc_node(pol->pd_size, GFP_KERNEL, q->node);
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
		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;
878
		pol->pd_init_fn(blkg);
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903

		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,
Tejun Heo's avatar
Tejun Heo committed
904
			     const struct blkcg_policy *pol)
905
{
Tejun Heo's avatar
Tejun Heo committed
906
	struct blkcg_gq *blkg;
907
908
909
910
911
912
913
914
915

	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);

916
917
918
919
	/* 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);

920
921
922
923
	list_for_each_entry(blkg, &q->blkg_list, q_node) {
		/* grab blkcg lock too while removing @pd from @blkg */
		spin_lock(&blkg->blkcg->lock);

924
925
		if (pol->pd_exit_fn)
			pol->pd_exit_fn(blkg);
926
927
928
929
930
931
932
933
934
935
936
937

		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);

938
/**
Tejun Heo's avatar
Tejun Heo committed
939
940
 * blkcg_policy_register - register a blkcg policy
 * @pol: blkcg policy to register
941
 *
Tejun Heo's avatar
Tejun Heo committed
942
943
 * Register @pol with blkcg core.  Might sleep and @pol may be modified on
 * successful registration.  Returns 0 on success and -errno on failure.
944
 */
Tejun Heo's avatar
Tejun Heo committed
945
int blkcg_policy_register(struct blkcg_policy *pol)
946
{
947
	int i, ret;
948

949
950
951
	if (WARN_ON(pol->pd_size < sizeof(struct blkg_policy_data)))
		return -EINVAL;

952
953
	mutex_lock(&blkcg_pol_mutex);

954
955
956
	/* find an empty slot */
	ret = -ENOSPC;
	for (i = 0; i < BLKCG_MAX_POLS; i++)
Tejun Heo's avatar
Tejun Heo committed
957
		if (!blkcg_policy[i])
958
959
960
			break;
	if (i >= BLKCG_MAX_POLS)
		goto out_unlock;
961

962
	/* register and update blkgs */
Tejun Heo's avatar
Tejun Heo committed
963
964
	pol->plid = i;
	blkcg_policy[i] = pol;
965
966

	/* everything is in place, add intf files for the new policy */
Tejun Heo's avatar
Tejun Heo committed
967
968
	if (pol->cftypes)
		WARN_ON(cgroup_add_cftypes(&blkio_subsys, pol->cftypes));
969
970
	ret = 0;
out_unlock:
971
	mutex_unlock(&blkcg_pol_mutex);
972
	return ret;
973
}
Tejun Heo's avatar
Tejun Heo committed
974
EXPORT_SYMBOL_GPL(blkcg_policy_register);
975

976
/**
Tejun Heo's avatar
Tejun Heo committed
977
978
 * blkcg_policy_unregister - unregister a blkcg policy
 * @pol: blkcg policy to unregister
979
 *
Tejun Heo's avatar
Tejun Heo committed
980
 * Undo blkcg_policy_register(@pol).  Might sleep.
981
 */
Tejun Heo's avatar
Tejun Heo committed
982
void blkcg_policy_unregister(struct blkcg_policy *pol)
983
{
984
985
	mutex_lock(&blkcg_pol_mutex);

Tejun Heo's avatar
Tejun Heo committed
986
	if (WARN_ON(blkcg_policy[pol->plid] != pol))
987
988
989
		goto out_unlock;

	/* kill the intf files first */
Tejun Heo's avatar
Tejun Heo committed
990
991
	if (pol->cftypes)
		cgroup_rm_cftypes(&blkio_subsys, pol->cftypes);
992

993
	/* unregister and update blkgs */
Tejun Heo's avatar
Tejun Heo committed
994
	blkcg_policy[pol->plid] = NULL;
995
out_unlock:
996
	mutex_unlock(&blkcg_pol_mutex);
997
}
Tejun Heo's avatar
Tejun Heo committed
998
EXPORT_SYMBOL_GPL(blkcg_policy_unregister);