blk-cgroup.c 22.6 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
	kfree(blkg);
67
68
69
70
71
72
}

/**
 * blkg_alloc - allocate a blkg
 * @blkcg: block cgroup the new blkg is associated with
 * @q: request_queue the new blkg is associated with
73
 * @gfp_mask: allocation mask to use
74
 *
75
 * Allocate a new blkg assocating @blkcg and @q.
76
 */
77
78
static struct blkcg_gq *blkg_alloc(struct blkcg *blkcg, struct request_queue *q,
				   gfp_t gfp_mask)
79
{
Tejun Heo's avatar
Tejun Heo committed
80
	struct blkcg_gq *blkg;
81
	int i;
82
83

	/* alloc and init base part */
84
	blkg = kzalloc_node(sizeof(*blkg), gfp_mask, q->node);
85
86
87
	if (!blkg)
		return NULL;

88
	blkg->q = q;
89
	INIT_LIST_HEAD(&blkg->q_node);
90
	blkg->blkcg = blkcg;
Tejun Heo's avatar
Tejun Heo committed
91
	blkg->refcnt = 1;
92

93
	for (i = 0; i < BLKCG_MAX_POLS; i++) {
Tejun Heo's avatar
Tejun Heo committed
94
		struct blkcg_policy *pol = blkcg_policy[i];
95
		struct blkg_policy_data *pd;
96

97
		if (!blkcg_policy_enabled(q, pol))
98
99
100
			continue;

		/* alloc per-policy data and attach it to blkg */
101
		pd = kzalloc_node(pol->pd_size, gfp_mask, q->node);
102
103
104
105
		if (!pd) {
			blkg_free(blkg);
			return NULL;
		}
106

107
108
109
		blkg->pd[i] = pd;
		pd->blkg = blkg;

110
		/* invoke per-policy init */
111
		if (blkcg_policy_enabled(blkg->q, pol))
112
			pol->pd_init_fn(blkg);
113
114
	}

115
116
117
	return blkg;
}

Tejun Heo's avatar
Tejun Heo committed
118
119
static struct blkcg_gq *__blkg_lookup(struct blkcg *blkcg,
				      struct request_queue *q)
120
{
Tejun Heo's avatar
Tejun Heo committed
121
	struct blkcg_gq *blkg;
122

123
124
125
126
127
128
129
130
131
132
133
134
135
136
	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;

137
138
139
140
141
142
143
144
145
146
147
148
	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
149
struct blkcg_gq *blkg_lookup(struct blkcg *blkcg, struct request_queue *q)
150
151
152
153
154
155
156
157
158
{
	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);

159
160
161
162
/*
 * 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.
 */
Tejun Heo's avatar
Tejun Heo committed
163
static struct blkcg_gq *__blkg_lookup_create(struct blkcg *blkcg,
164
165
					     struct request_queue *q,
					     struct blkcg_gq *new_blkg)
166
{
Tejun Heo's avatar
Tejun Heo committed
167
	struct blkcg_gq *blkg;
168
	int ret;
169

170
171
172
	WARN_ON_ONCE(!rcu_read_lock_held());
	lockdep_assert_held(q->queue_lock);

173
	/* lookup and update hint on success, see __blkg_lookup() for details */
174
	blkg = __blkg_lookup(blkcg, q);
175
176
	if (blkg) {
		rcu_assign_pointer(blkcg->blkg_hint, blkg);
177
		goto out_free;
178
	}
179

180
	/* blkg holds a reference to blkcg */
181
182
183
184
	if (!css_tryget(&blkcg->css)) {
		blkg = ERR_PTR(-EINVAL);
		goto out_free;
	}
185

186
	/* allocate */
187
188
189
190
191
192
193
194
	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;
195
196
197

	/* insert */
	spin_lock(&blkcg->lock);
198
199
200
201
202
	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);
	}
203
	spin_unlock(&blkcg->lock);
204

205
206
	if (!ret)
		return blkg;
207
208
209

	blkg = ERR_PTR(ret);
out_put:
210
	css_put(&blkcg->css);
211
212
213
out_free:
	blkg_free(new_blkg);
	return blkg;
214
}
215

Tejun Heo's avatar
Tejun Heo committed
216
217
struct blkcg_gq *blkg_lookup_create(struct blkcg *blkcg,
				    struct request_queue *q)
218
219
220
221
222
223
224
{
	/*
	 * 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)))
		return ERR_PTR(blk_queue_dead(q) ? -EINVAL : -EBUSY);
225
	return __blkg_lookup_create(blkcg, q, NULL);
226
}
227
EXPORT_SYMBOL_GPL(blkg_lookup_create);
228

Tejun Heo's avatar
Tejun Heo committed
229
static void blkg_destroy(struct blkcg_gq *blkg)
230
{
Tejun Heo's avatar
Tejun Heo committed
231
	struct blkcg *blkcg = blkg->blkcg;
232

233
	lockdep_assert_held(blkg->q->queue_lock);
234
	lockdep_assert_held(&blkcg->lock);
235
236

	/* Something wrong if we are trying to remove same group twice */
237
	WARN_ON_ONCE(list_empty(&blkg->q_node));
238
	WARN_ON_ONCE(hlist_unhashed(&blkg->blkcg_node));
239
240

	radix_tree_delete(&blkcg->blkg_tree, blkg->q->id);
241
	list_del_init(&blkg->q_node);
242
	hlist_del_init_rcu(&blkg->blkcg_node);
243

244
245
246
247
248
249
250
251
	/*
	 * 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);

252
253
254
255
256
257
258
	/*
	 * Put the reference taken at the time of creation so that when all
	 * queues are gone, group can be destroyed.
	 */
	blkg_put(blkg);
}

259
260
261
262
/**
 * blkg_destroy_all - destroy all blkgs associated with a request_queue
 * @q: request_queue of interest
 *
263
 * Destroy all blkgs associated with @q.
264
 */
265
static void blkg_destroy_all(struct request_queue *q)
266
{
Tejun Heo's avatar
Tejun Heo committed
267
	struct blkcg_gq *blkg, *n;
268

269
	lockdep_assert_held(q->queue_lock);
270

271
	list_for_each_entry_safe(blkg, n, &q->blkg_list, q_node) {
Tejun Heo's avatar
Tejun Heo committed
272
		struct blkcg *blkcg = blkg->blkcg;
273

274
275
276
		spin_lock(&blkcg->lock);
		blkg_destroy(blkg);
		spin_unlock(&blkcg->lock);
277
278
279
	}
}

Tejun Heo's avatar
Tejun Heo committed
280
281
static void blkg_rcu_free(struct rcu_head *rcu_head)
{
Tejun Heo's avatar
Tejun Heo committed
282
	blkg_free(container_of(rcu_head, struct blkcg_gq, rcu_head));
Tejun Heo's avatar
Tejun Heo committed
283
284
}

Tejun Heo's avatar
Tejun Heo committed
285
void __blkg_release(struct blkcg_gq *blkg)
Tejun Heo's avatar
Tejun Heo committed
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
{
	/* 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);

Tejun Heo's avatar
Tejun Heo committed
303
304
static int blkcg_reset_stats(struct cgroup *cgroup, struct cftype *cftype,
			     u64 val)
305
{
Tejun Heo's avatar
Tejun Heo committed
306
307
	struct blkcg *blkcg = cgroup_to_blkcg(cgroup);
	struct blkcg_gq *blkg;
308
	struct hlist_node *n;
309
	int i;
310

311
	mutex_lock(&blkcg_pol_mutex);
312
	spin_lock_irq(&blkcg->lock);
Tejun Heo's avatar
Tejun Heo committed
313
314
315
316
317
318

	/*
	 * 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.
	 */
319
	hlist_for_each_entry(blkg, n, &blkcg->blkg_list, blkcg_node) {
320
		for (i = 0; i < BLKCG_MAX_POLS; i++) {
Tejun Heo's avatar
Tejun Heo committed
321
			struct blkcg_policy *pol = blkcg_policy[i];
322

323
			if (blkcg_policy_enabled(blkg->q, pol) &&
324
325
			    pol->pd_reset_stats_fn)
				pol->pd_reset_stats_fn(blkg);
326
		}
327
	}
328

329
	spin_unlock_irq(&blkcg->lock);
330
	mutex_unlock(&blkcg_pol_mutex);
331
332
333
	return 0;
}

Tejun Heo's avatar
Tejun Heo committed
334
static const char *blkg_dev_name(struct blkcg_gq *blkg)
335
{
336
337
338
339
	/* 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;
340
341
}

342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
/**
 * 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
359
void blkcg_print_blkgs(struct seq_file *sf, struct blkcg *blkcg,
360
361
		       u64 (*prfill)(struct seq_file *,
				     struct blkg_policy_data *, int),
Tejun Heo's avatar
Tejun Heo committed
362
		       const struct blkcg_policy *pol, int data,
363
		       bool show_total)
364
{
Tejun Heo's avatar
Tejun Heo committed
365
	struct blkcg_gq *blkg;
366
367
	struct hlist_node *n;
	u64 total = 0;
368

369
370
	spin_lock_irq(&blkcg->lock);
	hlist_for_each_entry(blkg, n, &blkcg->blkg_list, blkcg_node)
371
		if (blkcg_policy_enabled(blkg->q, pol))
372
			total += prfill(sf, blkg->pd[pol->plid], data);
373
374
375
376
377
	spin_unlock_irq(&blkcg->lock);

	if (show_total)
		seq_printf(sf, "Total %llu\n", (unsigned long long)total);
}
378
EXPORT_SYMBOL_GPL(blkcg_print_blkgs);
379
380
381
382

/**
 * __blkg_prfill_u64 - prfill helper for a single u64 value
 * @sf: seq_file to print to
383
 * @pd: policy private data of interest
384
385
 * @v: value to print
 *
386
 * Print @v to @sf for the device assocaited with @pd.
387
 */
388
u64 __blkg_prfill_u64(struct seq_file *sf, struct blkg_policy_data *pd, u64 v)
389
{
390
	const char *dname = blkg_dev_name(pd->blkg);
391
392
393
394
395
396
397

	if (!dname)
		return 0;

	seq_printf(sf, "%s %llu\n", dname, (unsigned long long)v);
	return v;
}
398
EXPORT_SYMBOL_GPL(__blkg_prfill_u64);
399
400
401
402

/**
 * __blkg_prfill_rwstat - prfill helper for a blkg_rwstat
 * @sf: seq_file to print to
403
 * @pd: policy private data of interest
404
405
 * @rwstat: rwstat to print
 *
406
 * Print @rwstat to @sf for the device assocaited with @pd.
407
 */
408
u64 __blkg_prfill_rwstat(struct seq_file *sf, struct blkg_policy_data *pd,
409
			 const struct blkg_rwstat *rwstat)
410
411
412
413
414
415
416
{
	static const char *rwstr[] = {
		[BLKG_RWSTAT_READ]	= "Read",
		[BLKG_RWSTAT_WRITE]	= "Write",
		[BLKG_RWSTAT_SYNC]	= "Sync",
		[BLKG_RWSTAT_ASYNC]	= "Async",
	};
417
	const char *dname = blkg_dev_name(pd->blkg);
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
	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;
}

433
434
435
/**
 * blkg_prfill_stat - prfill callback for blkg_stat
 * @sf: seq_file to print to
436
437
 * @pd: policy private data of interest
 * @off: offset to the blkg_stat in @pd
438
439
440
 *
 * prfill callback for printing a blkg_stat.
 */
441
u64 blkg_prfill_stat(struct seq_file *sf, struct blkg_policy_data *pd, int off)
442
{
443
	return __blkg_prfill_u64(sf, pd, blkg_stat_read((void *)pd + off));
444
}
445
EXPORT_SYMBOL_GPL(blkg_prfill_stat);
446

447
448
449
/**
 * blkg_prfill_rwstat - prfill callback for blkg_rwstat
 * @sf: seq_file to print to
450
451
 * @pd: policy private data of interest
 * @off: offset to the blkg_rwstat in @pd
452
453
454
 *
 * prfill callback for printing a blkg_rwstat.
 */
455
456
u64 blkg_prfill_rwstat(struct seq_file *sf, struct blkg_policy_data *pd,
		       int off)
457
{
458
	struct blkg_rwstat rwstat = blkg_rwstat_read((void *)pd + off);
459

460
	return __blkg_prfill_rwstat(sf, pd, &rwstat);
461
}
462
EXPORT_SYMBOL_GPL(blkg_prfill_rwstat);
463

464
465
466
/**
 * blkg_conf_prep - parse and prepare for per-blkg config update
 * @blkcg: target block cgroup
467
 * @pol: target policy
468
469
470
471
472
 * @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
473
474
 * value.  This function returns with RCU read lock and queue lock held and
 * must be paired with blkg_conf_finish().
475
 */
Tejun Heo's avatar
Tejun Heo committed
476
477
int blkg_conf_prep(struct blkcg *blkcg, const struct blkcg_policy *pol,
		   const char *input, struct blkg_conf_ctx *ctx)
478
	__acquires(rcu) __acquires(disk->queue->queue_lock)
479
{
480
	struct gendisk *disk;
Tejun Heo's avatar
Tejun Heo committed
481
	struct blkcg_gq *blkg;
Tejun Heo's avatar
Tejun Heo committed
482
483
484
	unsigned int major, minor;
	unsigned long long v;
	int part, ret;
485

Tejun Heo's avatar
Tejun Heo committed
486
487
	if (sscanf(input, "%u:%u %llu", &major, &minor, &v) != 3)
		return -EINVAL;
488

Tejun Heo's avatar
Tejun Heo committed
489
	disk = get_gendisk(MKDEV(major, minor), &part);
Tejun Heo's avatar
Tejun Heo committed
490
	if (!disk || part)
Tejun Heo's avatar
Tejun Heo committed
491
		return -EINVAL;
492
493

	rcu_read_lock();
Tejun Heo's avatar
Tejun Heo committed
494
	spin_lock_irq(disk->queue->queue_lock);
495

496
	if (blkcg_policy_enabled(disk->queue, pol))
497
		blkg = blkg_lookup_create(blkcg, disk->queue);
498
499
	else
		blkg = ERR_PTR(-EINVAL);
500

Tejun Heo's avatar
Tejun Heo committed
501
502
	if (IS_ERR(blkg)) {
		ret = PTR_ERR(blkg);
503
		rcu_read_unlock();
504
		spin_unlock_irq(disk->queue->queue_lock);
505
506
507
508
509
510
511
512
513
514
		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();
515
		}
Tejun Heo's avatar
Tejun Heo committed
516
		return ret;
517
	}
518
519
520

	ctx->disk = disk;
	ctx->blkg = blkg;
Tejun Heo's avatar
Tejun Heo committed
521
522
	ctx->v = v;
	return 0;
523
}
524
EXPORT_SYMBOL_GPL(blkg_conf_prep);
525

526
527
528
529
530
531
532
/**
 * 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().
 */
533
void blkg_conf_finish(struct blkg_conf_ctx *ctx)
534
	__releases(ctx->disk->queue->queue_lock) __releases(rcu)
535
{
536
	spin_unlock_irq(ctx->disk->queue->queue_lock);
537
538
	rcu_read_unlock();
	put_disk(ctx->disk);
539
}
540
EXPORT_SYMBOL_GPL(blkg_conf_finish);
541

Tejun Heo's avatar
Tejun Heo committed
542
struct cftype blkcg_files[] = {
543
544
	{
		.name = "reset_stats",
Tejun Heo's avatar
Tejun Heo committed
545
		.write_u64 = blkcg_reset_stats,
546
	},
547
	{ }	/* terminate */
548
549
};

550
/**
Tejun Heo's avatar
Tejun Heo committed
551
 * blkcg_pre_destroy - cgroup pre_destroy callback
552
553
554
555
556
557
558
559
560
 * @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().
 */
Tejun Heo's avatar
Tejun Heo committed
561
static int blkcg_pre_destroy(struct cgroup *cgroup)
562
{
Tejun Heo's avatar
Tejun Heo committed
563
	struct blkcg *blkcg = cgroup_to_blkcg(cgroup);
564

565
	spin_lock_irq(&blkcg->lock);
566

567
	while (!hlist_empty(&blkcg->blkg_list)) {
Tejun Heo's avatar
Tejun Heo committed
568
569
		struct blkcg_gq *blkg = hlist_entry(blkcg->blkg_list.first,
						struct blkcg_gq, blkcg_node);
570
		struct request_queue *q = blkg->q;
571

572
573
574
575
576
577
		if (spin_trylock(q->queue_lock)) {
			blkg_destroy(blkg);
			spin_unlock(q->queue_lock);
		} else {
			spin_unlock_irq(&blkcg->lock);
			cpu_relax();
578
			spin_lock_irq(&blkcg->lock);
579
		}
580
	}
581

582
	spin_unlock_irq(&blkcg->lock);
583
584
585
	return 0;
}

Tejun Heo's avatar
Tejun Heo committed
586
static void blkcg_destroy(struct cgroup *cgroup)
587
{
Tejun Heo's avatar
Tejun Heo committed
588
	struct blkcg *blkcg = cgroup_to_blkcg(cgroup);
589

Tejun Heo's avatar
Tejun Heo committed
590
	if (blkcg != &blkcg_root)
Ben Blum's avatar
Ben Blum committed
591
		kfree(blkcg);
592
593
}

Tejun Heo's avatar
Tejun Heo committed
594
static struct cgroup_subsys_state *blkcg_create(struct cgroup *cgroup)
595
{
Tejun Heo's avatar
Tejun Heo committed
596
	static atomic64_t id_seq = ATOMIC64_INIT(0);
Tejun Heo's avatar
Tejun Heo committed
597
	struct blkcg *blkcg;
598
	struct cgroup *parent = cgroup->parent;
599

600
	if (!parent) {
Tejun Heo's avatar
Tejun Heo committed
601
		blkcg = &blkcg_root;
602
603
604
605
606
607
608
		goto done;
	}

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

609
	blkcg->cfq_weight = CFQ_WEIGHT_DEFAULT;
Tejun Heo's avatar
Tejun Heo committed
610
	blkcg->id = atomic64_inc_return(&id_seq); /* root is 0, start from 1 */
611
612
done:
	spin_lock_init(&blkcg->lock);
613
	INIT_RADIX_TREE(&blkcg->blkg_tree, GFP_ATOMIC);
614
615
616
617
618
	INIT_HLIST_HEAD(&blkcg->blkg_list);

	return &blkcg->css;
}

619
620
621
622
623
624
625
626
627
628
629
630
631
632
/**
 * 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();

633
	return blk_throtl_init(q);
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
}

/**
 * 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)
{
657
	spin_lock_irq(q->queue_lock);
658
	blkg_destroy_all(q);
659
660
	spin_unlock_irq(q->queue_lock);

661
662
663
	blk_throtl_exit(q);
}

664
665
666
667
668
669
/*
 * 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
670
static int blkcg_can_attach(struct cgroup *cgrp, struct cgroup_taskset *tset)
671
{
672
	struct task_struct *task;
673
674
675
676
	struct io_context *ioc;
	int ret = 0;

	/* task_lock() is needed to avoid races with exit_io_context() */
677
678
679
680
681
682
683
684
685
	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;
	}
686
687
688
	return ret;
}

689
690
struct cgroup_subsys blkio_subsys = {
	.name = "blkio",
Tejun Heo's avatar
Tejun Heo committed
691
692
693
694
	.create = blkcg_create,
	.can_attach = blkcg_can_attach,
	.pre_destroy = blkcg_pre_destroy,
	.destroy = blkcg_destroy,
695
	.subsys_id = blkio_subsys_id,
Tejun Heo's avatar
Tejun Heo committed
696
	.base_cftypes = blkcg_files,
697
698
699
700
	.module = THIS_MODULE,
};
EXPORT_SYMBOL_GPL(blkio_subsys);

701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
/**
 * 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
718
			  const struct blkcg_policy *pol)
719
720
{
	LIST_HEAD(pds);
Tejun Heo's avatar
Tejun Heo committed
721
	struct blkcg_gq *blkg;
722
723
	struct blkg_policy_data *pd, *n;
	int cnt = 0, ret;
724
	bool preloaded;
725
726
727
728

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

729
730
731
732
733
734
735
	/* preallocations for root blkg */
	blkg = blkg_alloc(&blkcg_root, q, GFP_KERNEL);
	if (!blkg)
		return -ENOMEM;

	preloaded = !radix_tree_preload(GFP_KERNEL);

736
737
738
739
740
741
	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();
742
	blkg = __blkg_lookup_create(&blkcg_root, q, blkg);
743
744
	rcu_read_unlock();

745
746
747
	if (preloaded)
		radix_tree_preload_end();

748
749
750
751
752
753
754
755
756
757
758
759
760
	if (IS_ERR(blkg)) {
		ret = PTR_ERR(blkg);
		goto out_unlock;
	}
	q->root_blkg = blkg;

	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--) {
761
		pd = kzalloc_node(pol->pd_size, GFP_KERNEL, q->node);
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
		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;
789
		pol->pd_init_fn(blkg);
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814

		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
815
			     const struct blkcg_policy *pol)
816
{
Tejun Heo's avatar
Tejun Heo committed
817
	struct blkcg_gq *blkg;
818
819
820
821
822
823
824
825
826

	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);

827
828
829
830
	/* 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);

831
832
833
834
	list_for_each_entry(blkg, &q->blkg_list, q_node) {
		/* grab blkcg lock too while removing @pd from @blkg */
		spin_lock(&blkg->blkcg->lock);

835
836
		if (pol->pd_exit_fn)
			pol->pd_exit_fn(blkg);
837
838
839
840
841
842
843
844
845
846
847
848

		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);

849
/**
Tejun Heo's avatar
Tejun Heo committed
850
851
 * blkcg_policy_register - register a blkcg policy
 * @pol: blkcg policy to register
852
 *
Tejun Heo's avatar
Tejun Heo committed
853
854
 * Register @pol with blkcg core.  Might sleep and @pol may be modified on
 * successful registration.  Returns 0 on success and -errno on failure.
855
 */
Tejun Heo's avatar
Tejun Heo committed
856
int blkcg_policy_register(struct blkcg_policy *pol)
857
{
858
	int i, ret;
859

860
861
862
	if (WARN_ON(pol->pd_size < sizeof(struct blkg_policy_data)))
		return -EINVAL;

863
864
	mutex_lock(&blkcg_pol_mutex);

865
866
867
	/* find an empty slot */
	ret = -ENOSPC;
	for (i = 0; i < BLKCG_MAX_POLS; i++)
Tejun Heo's avatar
Tejun Heo committed
868
		if (!blkcg_policy[i])
869
870
871
			break;
	if (i >= BLKCG_MAX_POLS)
		goto out_unlock;
872

873
	/* register and update blkgs */
Tejun Heo's avatar
Tejun Heo committed
874
875
	pol->plid = i;
	blkcg_policy[i] = pol;
876
877

	/* everything is in place, add intf files for the new policy */
Tejun Heo's avatar
Tejun Heo committed
878
879
	if (pol->cftypes)
		WARN_ON(cgroup_add_cftypes(&blkio_subsys, pol->cftypes));
880
881
	ret = 0;
out_unlock:
882
	mutex_unlock(&blkcg_pol_mutex);
883
	return ret;
884
}
Tejun Heo's avatar
Tejun Heo committed
885
EXPORT_SYMBOL_GPL(blkcg_policy_register);
886

887
/**
Tejun Heo's avatar
Tejun Heo committed
888
889
 * blkcg_policy_unregister - unregister a blkcg policy
 * @pol: blkcg policy to unregister
890
 *
Tejun Heo's avatar
Tejun Heo committed
891
 * Undo blkcg_policy_register(@pol).  Might sleep.
892
 */
Tejun Heo's avatar
Tejun Heo committed
893
void blkcg_policy_unregister(struct blkcg_policy *pol)
894
{
895
896
	mutex_lock(&blkcg_pol_mutex);

Tejun Heo's avatar
Tejun Heo committed
897
	if (WARN_ON(blkcg_policy[pol->plid] != pol))
898
899
900
		goto out_unlock;

	/* kill the intf files first */
Tejun Heo's avatar
Tejun Heo committed
901
902
	if (pol->cftypes)
		cgroup_rm_cftypes(&blkio_subsys, pol->cftypes);
903

904
	/* unregister and update blkgs */
Tejun Heo's avatar
Tejun Heo committed
905
	blkcg_policy[pol->plid] = NULL;
906
out_unlock:
907
	mutex_unlock(&blkcg_pol_mutex);
908
}
Tejun Heo's avatar
Tejun Heo committed
909
EXPORT_SYMBOL_GPL(blkcg_policy_unregister);