blk-cgroup.c 46.6 KB
Newer Older
1
// SPDX-License-Identifier: GPL-2.0
2 3 4 5 6 7 8 9 10 11 12
/*
 * 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>
13 14 15 16
 *
 * For policy-specific per-blkcg data:
 * Copyright (C) 2015 Paolo Valente <paolo.valente@unimore.it>
 *                    Arianna Avanzini <avanzini.arianna@gmail.com>
17 18
 */
#include <linux/ioprio.h>
19
#include <linux/kdev_t.h>
20
#include <linux/module.h>
21
#include <linux/sched/signal.h>
22
#include <linux/err.h>
23
#include <linux/blkdev.h>
24
#include <linux/backing-dev.h>
25
#include <linux/slab.h>
26
#include <linux/genhd.h>
27
#include <linux/delay.h>
T
Tejun Heo 已提交
28
#include <linux/atomic.h>
29
#include <linux/ctype.h>
30
#include <linux/blk-cgroup.h>
31
#include <linux/tracehook.h>
32
#include "blk.h"
33

34 35
#define MAX_KEY_LEN 100

36 37 38 39 40 41 42 43
/*
 * blkcg_pol_mutex protects blkcg_policy[] and policy [de]activation.
 * blkcg_pol_register_mutex nests outside of it and synchronizes entire
 * policy [un]register operations including cgroup file additions /
 * removals.  Putting cgroup file registration outside blkcg_pol_mutex
 * allows grabbing it from cgroup callbacks.
 */
static DEFINE_MUTEX(blkcg_pol_register_mutex);
44
static DEFINE_MUTEX(blkcg_pol_mutex);
45

46
struct blkcg blkcg_root;
T
Tejun Heo 已提交
47
EXPORT_SYMBOL_GPL(blkcg_root);
48

T
Tejun Heo 已提交
49 50
struct cgroup_subsys_state * const blkcg_root_css = &blkcg_root.css;

T
Tejun Heo 已提交
51
static struct blkcg_policy *blkcg_policy[BLKCG_MAX_POLS];
52

T
Tejun Heo 已提交
53 54
static LIST_HEAD(all_blkcgs);		/* protected by blkcg_pol_mutex */

55 56
static bool blkcg_debug_stats = false;

57
static bool blkcg_policy_enabled(struct request_queue *q,
T
Tejun Heo 已提交
58
				 const struct blkcg_policy *pol)
59 60 61 62
{
	return pol && test_bit(pol->plid, q->blkcg_pols);
}

63 64 65 66 67 68
/**
 * blkg_free - free a blkg
 * @blkg: blkg to free
 *
 * Free @blkg which may be partially allocated.
 */
T
Tejun Heo 已提交
69
static void blkg_free(struct blkcg_gq *blkg)
70
{
71
	int i;
72 73 74 75

	if (!blkg)
		return;

76
	for (i = 0; i < BLKCG_MAX_POLS; i++)
77 78
		if (blkg->pd[i])
			blkcg_policy[i]->pd_free_fn(blkg->pd[i]);
79

80 81
	blkg_rwstat_exit(&blkg->stat_ios);
	blkg_rwstat_exit(&blkg->stat_bytes);
82
	kfree(blkg);
83 84
}

85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115
static void __blkg_release(struct rcu_head *rcu)
{
	struct blkcg_gq *blkg = container_of(rcu, struct blkcg_gq, rcu_head);

	percpu_ref_exit(&blkg->refcnt);

	/* release the blkcg and parent blkg refs this blkg has been holding */
	css_put(&blkg->blkcg->css);
	if (blkg->parent)
		blkg_put(blkg->parent);

	wb_congested_put(blkg->wb_congested);

	blkg_free(blkg);
}

/*
 * A group is RCU protected, 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 accesses to only values
 * local to groups like group stats and group rate limits.
 */
static void blkg_release(struct percpu_ref *ref)
{
	struct blkcg_gq *blkg = container_of(ref, struct blkcg_gq, refcnt);

	call_rcu(&blkg->rcu_head, __blkg_release);
}

116 117 118 119
/**
 * blkg_alloc - allocate a blkg
 * @blkcg: block cgroup the new blkg is associated with
 * @q: request_queue the new blkg is associated with
120
 * @gfp_mask: allocation mask to use
121
 *
122
 * Allocate a new blkg assocating @blkcg and @q.
123
 */
124 125
static struct blkcg_gq *blkg_alloc(struct blkcg *blkcg, struct request_queue *q,
				   gfp_t gfp_mask)
126
{
T
Tejun Heo 已提交
127
	struct blkcg_gq *blkg;
128
	int i;
129 130

	/* alloc and init base part */
131
	blkg = kzalloc_node(sizeof(*blkg), gfp_mask, q->node);
132 133 134
	if (!blkg)
		return NULL;

135 136 137 138
	if (blkg_rwstat_init(&blkg->stat_bytes, gfp_mask) ||
	    blkg_rwstat_init(&blkg->stat_ios, gfp_mask))
		goto err_free;

T
Tejun Heo 已提交
139
	blkg->q = q;
140
	INIT_LIST_HEAD(&blkg->q_node);
141 142
	blkg->blkcg = blkcg;

T
Tejun Heo 已提交
143
	for (i = 0; i < BLKCG_MAX_POLS; i++) {
T
Tejun Heo 已提交
144
		struct blkcg_policy *pol = blkcg_policy[i];
145
		struct blkg_policy_data *pd;
146

147
		if (!blkcg_policy_enabled(q, pol))
148 149 150
			continue;

		/* alloc per-policy data and attach it to blkg */
151
		pd = pol->pd_alloc_fn(gfp_mask, q->node);
152 153
		if (!pd)
			goto err_free;
154

155 156
		blkg->pd[i] = pd;
		pd->blkg = blkg;
T
Tejun Heo 已提交
157
		pd->plid = i;
158 159
	}

160
	return blkg;
161 162 163 164

err_free:
	blkg_free(blkg);
	return NULL;
165 166
}

T
Tejun Heo 已提交
167 168
struct blkcg_gq *blkg_lookup_slowpath(struct blkcg *blkcg,
				      struct request_queue *q, bool update_hint)
169
{
T
Tejun Heo 已提交
170
	struct blkcg_gq *blkg;
171

172
	/*
173 174 175 176
	 * 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.
177 178
	 */
	blkg = radix_tree_lookup(&blkcg->blkg_tree, q->id);
179 180
	if (blkg && blkg->q == q) {
		if (update_hint) {
181
			lockdep_assert_held(&q->queue_lock);
182 183
			rcu_assign_pointer(blkcg->blkg_hint, blkg);
		}
184
		return blkg;
185
	}
186

187 188
	return NULL;
}
189
EXPORT_SYMBOL_GPL(blkg_lookup_slowpath);
190

191
/*
192 193
 * If @new_blkg is %NULL, this function tries to allocate a new one as
 * necessary using %GFP_NOWAIT.  @new_blkg is always consumed on return.
194
 */
195
static struct blkcg_gq *blkg_create(struct blkcg *blkcg,
196 197
				    struct request_queue *q,
				    struct blkcg_gq *new_blkg)
198
{
199
	struct blkcg_gq *blkg;
200
	struct bdi_writeback_congested *wb_congested;
201
	int i, ret;
202

203
	WARN_ON_ONCE(!rcu_read_lock_held());
204
	lockdep_assert_held(&q->queue_lock);
205

206 207 208 209 210 211
	/* request_queue is dying, do not create/recreate a blkg */
	if (blk_queue_dying(q)) {
		ret = -ENODEV;
		goto err_free_blkg;
	}

212
	/* blkg holds a reference to blkcg */
213
	if (!css_tryget_online(&blkcg->css)) {
214
		ret = -ENODEV;
215
		goto err_free_blkg;
216
	}
217

218
	wb_congested = wb_congested_get_create(q->backing_dev_info,
219 220 221
					       blkcg->css.id,
					       GFP_NOWAIT | __GFP_NOWARN);
	if (!wb_congested) {
222
		ret = -ENOMEM;
223
		goto err_put_css;
224 225
	}

226 227 228 229 230 231
	/* allocate */
	if (!new_blkg) {
		new_blkg = blkg_alloc(blkcg, q, GFP_NOWAIT | __GFP_NOWARN);
		if (unlikely(!new_blkg)) {
			ret = -ENOMEM;
			goto err_put_congested;
232 233
		}
	}
234 235
	blkg = new_blkg;
	blkg->wb_congested = wb_congested;
236

237
	/* link parent */
T
Tejun Heo 已提交
238 239 240
	if (blkcg_parent(blkcg)) {
		blkg->parent = __blkg_lookup(blkcg_parent(blkcg), q, false);
		if (WARN_ON_ONCE(!blkg->parent)) {
241
			ret = -ENODEV;
242
			goto err_put_congested;
T
Tejun Heo 已提交
243 244 245 246
		}
		blkg_get(blkg->parent);
	}

247 248 249 250 251
	ret = percpu_ref_init(&blkg->refcnt, blkg_release, 0,
			      GFP_NOWAIT | __GFP_NOWARN);
	if (ret)
		goto err_cancel_ref;

252 253 254 255 256
	/* invoke per-policy init */
	for (i = 0; i < BLKCG_MAX_POLS; i++) {
		struct blkcg_policy *pol = blkcg_policy[i];

		if (blkg->pd[i] && pol->pd_init_fn)
257
			pol->pd_init_fn(blkg->pd[i]);
258 259 260
	}

	/* insert */
261
	spin_lock(&blkcg->lock);
262 263 264 265
	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);
266 267 268 269 270

		for (i = 0; i < BLKCG_MAX_POLS; i++) {
			struct blkcg_policy *pol = blkcg_policy[i];

			if (blkg->pd[i] && pol->pd_online_fn)
271
				pol->pd_online_fn(blkg->pd[i]);
272
		}
273
	}
274
	blkg->online = true;
275
	spin_unlock(&blkcg->lock);
276

277
	if (!ret)
278
		return blkg;
279

T
Tejun Heo 已提交
280 281 282 283
	/* @blkg failed fully initialized, use the usual release path */
	blkg_put(blkg);
	return ERR_PTR(ret);

284 285
err_cancel_ref:
	percpu_ref_exit(&blkg->refcnt);
286 287 288
err_put_congested:
	wb_congested_put(wb_congested);
err_put_css:
289
	css_put(&blkcg->css);
290
err_free_blkg:
291
	blkg_free(new_blkg);
292
	return ERR_PTR(ret);
293
}
294

295
/**
296
 * __blkg_lookup_create - lookup blkg, try to create one if not there
297 298 299 300
 * @blkcg: blkcg of interest
 * @q: request_queue of interest
 *
 * Lookup blkg for the @blkcg - @q pair.  If it doesn't exist, try to
T
Tejun Heo 已提交
301 302 303
 * create one.  blkg creation is performed recursively from blkcg_root such
 * that all non-root blkg's have access to the parent blkg.  This function
 * should be called under RCU read lock and @q->queue_lock.
304
 *
305 306
 * Returns the blkg or the closest blkg if blkg_create() fails as it walks
 * down from root.
307
 */
308 309
struct blkcg_gq *__blkg_lookup_create(struct blkcg *blkcg,
				      struct request_queue *q)
310
{
311 312 313
	struct blkcg_gq *blkg;

	WARN_ON_ONCE(!rcu_read_lock_held());
314
	lockdep_assert_held(&q->queue_lock);
315 316 317 318 319

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

T
Tejun Heo 已提交
320 321
	/*
	 * Create blkgs walking down from blkcg_root to @blkcg, so that all
322 323
	 * non-root blkgs have access to their parents.  Returns the closest
	 * blkg to the intended blkg should blkg_create() fail.
T
Tejun Heo 已提交
324 325 326 327
	 */
	while (true) {
		struct blkcg *pos = blkcg;
		struct blkcg *parent = blkcg_parent(blkcg);
328 329 330 331 332 333 334 335 336
		struct blkcg_gq *ret_blkg = q->root_blkg;

		while (parent) {
			blkg = __blkg_lookup(parent, q, false);
			if (blkg) {
				/* remember closest blkg */
				ret_blkg = blkg;
				break;
			}
T
Tejun Heo 已提交
337 338 339 340
			pos = parent;
			parent = blkcg_parent(parent);
		}

341
		blkg = blkg_create(pos, q, NULL);
342 343 344
		if (IS_ERR(blkg))
			return ret_blkg;
		if (pos == blkcg)
T
Tejun Heo 已提交
345 346
			return blkg;
	}
347
}
348

349 350 351 352 353 354 355 356 357 358 359 360 361 362
/**
 * blkg_lookup_create - find or create a blkg
 * @blkcg: target block cgroup
 * @q: target request_queue
 *
 * This looks up or creates the blkg representing the unique pair
 * of the blkcg and the request_queue.
 */
struct blkcg_gq *blkg_lookup_create(struct blkcg *blkcg,
				    struct request_queue *q)
{
	struct blkcg_gq *blkg = blkg_lookup(blkcg, q);

	if (unlikely(!blkg)) {
363 364 365
		unsigned long flags;

		spin_lock_irqsave(&q->queue_lock, flags);
366
		blkg = __blkg_lookup_create(blkcg, q);
367
		spin_unlock_irqrestore(&q->queue_lock, flags);
368 369 370 371 372
	}

	return blkg;
}

T
Tejun Heo 已提交
373
static void blkg_destroy(struct blkcg_gq *blkg)
374
{
T
Tejun Heo 已提交
375
	struct blkcg *blkcg = blkg->blkcg;
376
	struct blkcg_gq *parent = blkg->parent;
377
	int i;
378

379
	lockdep_assert_held(&blkg->q->queue_lock);
380
	lockdep_assert_held(&blkcg->lock);
381 382

	/* Something wrong if we are trying to remove same group twice */
383
	WARN_ON_ONCE(list_empty(&blkg->q_node));
384
	WARN_ON_ONCE(hlist_unhashed(&blkg->blkcg_node));
385

386 387 388 389 390 391 392
	for (i = 0; i < BLKCG_MAX_POLS; i++) {
		struct blkcg_policy *pol = blkcg_policy[i];

		if (blkg->pd[i] && pol->pd_offline_fn)
			pol->pd_offline_fn(blkg->pd[i]);
	}

393 394 395 396 397
	if (parent) {
		blkg_rwstat_add_aux(&parent->stat_bytes, &blkg->stat_bytes);
		blkg_rwstat_add_aux(&parent->stat_ios, &blkg->stat_ios);
	}

398 399
	blkg->online = false;

400
	radix_tree_delete(&blkcg->blkg_tree, blkg->q->id);
401
	list_del_init(&blkg->q_node);
402
	hlist_del_init_rcu(&blkg->blkcg_node);
403

404 405 406 407 408
	/*
	 * 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.
	 */
409
	if (rcu_access_pointer(blkcg->blkg_hint) == blkg)
410 411
		rcu_assign_pointer(blkcg->blkg_hint, NULL);

412 413 414 415
	/*
	 * Put the reference taken at the time of creation so that when all
	 * queues are gone, group can be destroyed.
	 */
416
	percpu_ref_kill(&blkg->refcnt);
417 418
}

419 420 421 422
/**
 * blkg_destroy_all - destroy all blkgs associated with a request_queue
 * @q: request_queue of interest
 *
423
 * Destroy all blkgs associated with @q.
424
 */
425
static void blkg_destroy_all(struct request_queue *q)
426
{
T
Tejun Heo 已提交
427
	struct blkcg_gq *blkg, *n;
428

429
	spin_lock_irq(&q->queue_lock);
430
	list_for_each_entry_safe(blkg, n, &q->blkg_list, q_node) {
T
Tejun Heo 已提交
431
		struct blkcg *blkcg = blkg->blkcg;
432

433 434 435
		spin_lock(&blkcg->lock);
		blkg_destroy(blkg);
		spin_unlock(&blkcg->lock);
436
	}
437 438

	q->root_blkg = NULL;
439
	spin_unlock_irq(&q->queue_lock);
440 441
}

442 443
static int blkcg_reset_stats(struct cgroup_subsys_state *css,
			     struct cftype *cftype, u64 val)
444
{
445
	struct blkcg *blkcg = css_to_blkcg(css);
T
Tejun Heo 已提交
446
	struct blkcg_gq *blkg;
447
	int i;
448

449
	mutex_lock(&blkcg_pol_mutex);
450
	spin_lock_irq(&blkcg->lock);
T
Tejun Heo 已提交
451 452 453 454 455 456

	/*
	 * 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.
	 */
457
	hlist_for_each_entry(blkg, &blkcg->blkg_list, blkcg_node) {
458 459 460
		blkg_rwstat_reset(&blkg->stat_bytes);
		blkg_rwstat_reset(&blkg->stat_ios);

T
Tejun Heo 已提交
461
		for (i = 0; i < BLKCG_MAX_POLS; i++) {
T
Tejun Heo 已提交
462
			struct blkcg_policy *pol = blkcg_policy[i];
463

464 465
			if (blkg->pd[i] && pol->pd_reset_stats_fn)
				pol->pd_reset_stats_fn(blkg->pd[i]);
466
		}
467
	}
468

469
	spin_unlock_irq(&blkcg->lock);
470
	mutex_unlock(&blkcg_pol_mutex);
471 472 473
	return 0;
}

474
const char *blkg_dev_name(struct blkcg_gq *blkg)
475
{
476
	/* some drivers (floppy) instantiate a queue w/o disk registered */
477 478
	if (blkg->q->backing_dev_info->dev)
		return dev_name(blkg->q->backing_dev_info->dev);
479
	return NULL;
480 481
}

482 483 484 485 486 487 488 489 490 491 492
/**
 * 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
493 494 495
 * policy data and @data and the matching queue lock held.  If @show_total
 * is %true, the sum of the return values from @prfill is printed with
 * "Total" label at the end.
496 497 498 499
 *
 * This is to be used to construct print functions for
 * cftype->read_seq_string method.
 */
T
Tejun Heo 已提交
500
void blkcg_print_blkgs(struct seq_file *sf, struct blkcg *blkcg,
501 502
		       u64 (*prfill)(struct seq_file *,
				     struct blkg_policy_data *, int),
T
Tejun Heo 已提交
503
		       const struct blkcg_policy *pol, int data,
504
		       bool show_total)
505
{
T
Tejun Heo 已提交
506
	struct blkcg_gq *blkg;
507
	u64 total = 0;
508

509
	rcu_read_lock();
510
	hlist_for_each_entry_rcu(blkg, &blkcg->blkg_list, blkcg_node) {
511
		spin_lock_irq(&blkg->q->queue_lock);
512
		if (blkcg_policy_enabled(blkg->q, pol))
513
			total += prfill(sf, blkg->pd[pol->plid], data);
514
		spin_unlock_irq(&blkg->q->queue_lock);
515 516
	}
	rcu_read_unlock();
517 518 519 520

	if (show_total)
		seq_printf(sf, "Total %llu\n", (unsigned long long)total);
}
521
EXPORT_SYMBOL_GPL(blkcg_print_blkgs);
522 523 524 525

/**
 * __blkg_prfill_u64 - prfill helper for a single u64 value
 * @sf: seq_file to print to
526
 * @pd: policy private data of interest
527 528
 * @v: value to print
 *
529
 * Print @v to @sf for the device assocaited with @pd.
530
 */
531
u64 __blkg_prfill_u64(struct seq_file *sf, struct blkg_policy_data *pd, u64 v)
532
{
533
	const char *dname = blkg_dev_name(pd->blkg);
534 535 536 537 538 539 540

	if (!dname)
		return 0;

	seq_printf(sf, "%s %llu\n", dname, (unsigned long long)v);
	return v;
}
541
EXPORT_SYMBOL_GPL(__blkg_prfill_u64);
542 543 544 545

/**
 * __blkg_prfill_rwstat - prfill helper for a blkg_rwstat
 * @sf: seq_file to print to
546
 * @pd: policy private data of interest
547 548
 * @rwstat: rwstat to print
 *
549
 * Print @rwstat to @sf for the device assocaited with @pd.
550
 */
551
u64 __blkg_prfill_rwstat(struct seq_file *sf, struct blkg_policy_data *pd,
552
			 const struct blkg_rwstat *rwstat)
553 554 555 556 557 558
{
	static const char *rwstr[] = {
		[BLKG_RWSTAT_READ]	= "Read",
		[BLKG_RWSTAT_WRITE]	= "Write",
		[BLKG_RWSTAT_SYNC]	= "Sync",
		[BLKG_RWSTAT_ASYNC]	= "Async",
559
		[BLKG_RWSTAT_DISCARD]	= "Discard",
560
	};
561
	const char *dname = blkg_dev_name(pd->blkg);
562 563 564 565 566 567 568 569
	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],
T
Tejun Heo 已提交
570
			   (unsigned long long)atomic64_read(&rwstat->aux_cnt[i]));
571

T
Tejun Heo 已提交
572
	v = atomic64_read(&rwstat->aux_cnt[BLKG_RWSTAT_READ]) +
573 574
		atomic64_read(&rwstat->aux_cnt[BLKG_RWSTAT_WRITE]) +
		atomic64_read(&rwstat->aux_cnt[BLKG_RWSTAT_DISCARD]);
575 576 577
	seq_printf(sf, "%s Total %llu\n", dname, (unsigned long long)v);
	return v;
}
T
Tejun Heo 已提交
578
EXPORT_SYMBOL_GPL(__blkg_prfill_rwstat);
579

580 581 582
/**
 * blkg_prfill_stat - prfill callback for blkg_stat
 * @sf: seq_file to print to
583 584
 * @pd: policy private data of interest
 * @off: offset to the blkg_stat in @pd
585 586 587
 *
 * prfill callback for printing a blkg_stat.
 */
588
u64 blkg_prfill_stat(struct seq_file *sf, struct blkg_policy_data *pd, int off)
589
{
590
	return __blkg_prfill_u64(sf, pd, blkg_stat_read((void *)pd + off));
591
}
592
EXPORT_SYMBOL_GPL(blkg_prfill_stat);
593

594 595 596
/**
 * blkg_prfill_rwstat - prfill callback for blkg_rwstat
 * @sf: seq_file to print to
597 598
 * @pd: policy private data of interest
 * @off: offset to the blkg_rwstat in @pd
599 600 601
 *
 * prfill callback for printing a blkg_rwstat.
 */
602 603
u64 blkg_prfill_rwstat(struct seq_file *sf, struct blkg_policy_data *pd,
		       int off)
604
{
605
	struct blkg_rwstat rwstat = blkg_rwstat_read((void *)pd + off);
606

607
	return __blkg_prfill_rwstat(sf, pd, &rwstat);
608
}
609
EXPORT_SYMBOL_GPL(blkg_prfill_rwstat);
610

611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691
static u64 blkg_prfill_rwstat_field(struct seq_file *sf,
				    struct blkg_policy_data *pd, int off)
{
	struct blkg_rwstat rwstat = blkg_rwstat_read((void *)pd->blkg + off);

	return __blkg_prfill_rwstat(sf, pd, &rwstat);
}

/**
 * blkg_print_stat_bytes - seq_show callback for blkg->stat_bytes
 * @sf: seq_file to print to
 * @v: unused
 *
 * To be used as cftype->seq_show to print blkg->stat_bytes.
 * cftype->private must be set to the blkcg_policy.
 */
int blkg_print_stat_bytes(struct seq_file *sf, void *v)
{
	blkcg_print_blkgs(sf, css_to_blkcg(seq_css(sf)),
			  blkg_prfill_rwstat_field, (void *)seq_cft(sf)->private,
			  offsetof(struct blkcg_gq, stat_bytes), true);
	return 0;
}
EXPORT_SYMBOL_GPL(blkg_print_stat_bytes);

/**
 * blkg_print_stat_bytes - seq_show callback for blkg->stat_ios
 * @sf: seq_file to print to
 * @v: unused
 *
 * To be used as cftype->seq_show to print blkg->stat_ios.  cftype->private
 * must be set to the blkcg_policy.
 */
int blkg_print_stat_ios(struct seq_file *sf, void *v)
{
	blkcg_print_blkgs(sf, css_to_blkcg(seq_css(sf)),
			  blkg_prfill_rwstat_field, (void *)seq_cft(sf)->private,
			  offsetof(struct blkcg_gq, stat_ios), true);
	return 0;
}
EXPORT_SYMBOL_GPL(blkg_print_stat_ios);

static u64 blkg_prfill_rwstat_field_recursive(struct seq_file *sf,
					      struct blkg_policy_data *pd,
					      int off)
{
	struct blkg_rwstat rwstat = blkg_rwstat_recursive_sum(pd->blkg,
							      NULL, off);
	return __blkg_prfill_rwstat(sf, pd, &rwstat);
}

/**
 * blkg_print_stat_bytes_recursive - recursive version of blkg_print_stat_bytes
 * @sf: seq_file to print to
 * @v: unused
 */
int blkg_print_stat_bytes_recursive(struct seq_file *sf, void *v)
{
	blkcg_print_blkgs(sf, css_to_blkcg(seq_css(sf)),
			  blkg_prfill_rwstat_field_recursive,
			  (void *)seq_cft(sf)->private,
			  offsetof(struct blkcg_gq, stat_bytes), true);
	return 0;
}
EXPORT_SYMBOL_GPL(blkg_print_stat_bytes_recursive);

/**
 * blkg_print_stat_ios_recursive - recursive version of blkg_print_stat_ios
 * @sf: seq_file to print to
 * @v: unused
 */
int blkg_print_stat_ios_recursive(struct seq_file *sf, void *v)
{
	blkcg_print_blkgs(sf, css_to_blkcg(seq_css(sf)),
			  blkg_prfill_rwstat_field_recursive,
			  (void *)seq_cft(sf)->private,
			  offsetof(struct blkcg_gq, stat_ios), true);
	return 0;
}
EXPORT_SYMBOL_GPL(blkg_print_stat_ios_recursive);

692 693
/**
 * blkg_stat_recursive_sum - collect hierarchical blkg_stat
694 695 696
 * @blkg: blkg of interest
 * @pol: blkcg_policy which contains the blkg_stat
 * @off: offset to the blkg_stat in blkg_policy_data or @blkg
697
 *
698 699 700 701 702 703
 * Collect the blkg_stat specified by @blkg, @pol and @off and all its
 * online descendants and their aux counts.  The caller must be holding the
 * queue lock for online tests.
 *
 * If @pol is NULL, blkg_stat is at @off bytes into @blkg; otherwise, it is
 * at @off bytes into @blkg's blkg_policy_data of the policy.
704
 */
705 706
u64 blkg_stat_recursive_sum(struct blkcg_gq *blkg,
			    struct blkcg_policy *pol, int off)
707 708
{
	struct blkcg_gq *pos_blkg;
709
	struct cgroup_subsys_state *pos_css;
710
	u64 sum = 0;
711

712
	lockdep_assert_held(&blkg->q->queue_lock);
713 714

	rcu_read_lock();
715 716 717 718 719
	blkg_for_each_descendant_pre(pos_blkg, pos_css, blkg) {
		struct blkg_stat *stat;

		if (!pos_blkg->online)
			continue;
720

721 722 723 724 725 726
		if (pol)
			stat = (void *)blkg_to_pd(pos_blkg, pol) + off;
		else
			stat = (void *)blkg + off;

		sum += blkg_stat_read(stat) + atomic64_read(&stat->aux_cnt);
727 728 729 730 731 732 733 734 735
	}
	rcu_read_unlock();

	return sum;
}
EXPORT_SYMBOL_GPL(blkg_stat_recursive_sum);

/**
 * blkg_rwstat_recursive_sum - collect hierarchical blkg_rwstat
736 737 738
 * @blkg: blkg of interest
 * @pol: blkcg_policy which contains the blkg_rwstat
 * @off: offset to the blkg_rwstat in blkg_policy_data or @blkg
739
 *
740 741 742 743 744 745
 * Collect the blkg_rwstat specified by @blkg, @pol and @off and all its
 * online descendants and their aux counts.  The caller must be holding the
 * queue lock for online tests.
 *
 * If @pol is NULL, blkg_rwstat is at @off bytes into @blkg; otherwise, it
 * is at @off bytes into @blkg's blkg_policy_data of the policy.
746
 */
747 748
struct blkg_rwstat blkg_rwstat_recursive_sum(struct blkcg_gq *blkg,
					     struct blkcg_policy *pol, int off)
749 750
{
	struct blkcg_gq *pos_blkg;
751
	struct cgroup_subsys_state *pos_css;
752
	struct blkg_rwstat sum = { };
753 754
	int i;

755
	lockdep_assert_held(&blkg->q->queue_lock);
756 757

	rcu_read_lock();
758
	blkg_for_each_descendant_pre(pos_blkg, pos_css, blkg) {
759
		struct blkg_rwstat *rwstat;
760 761 762 763

		if (!pos_blkg->online)
			continue;

764 765 766 767 768
		if (pol)
			rwstat = (void *)blkg_to_pd(pos_blkg, pol) + off;
		else
			rwstat = (void *)pos_blkg + off;

769
		for (i = 0; i < BLKG_RWSTAT_NR; i++)
770 771 772
			atomic64_add(atomic64_read(&rwstat->aux_cnt[i]) +
				percpu_counter_sum_positive(&rwstat->cpu_cnt[i]),
				&sum.aux_cnt[i]);
773 774 775 776 777 778 779
	}
	rcu_read_unlock();

	return sum;
}
EXPORT_SYMBOL_GPL(blkg_rwstat_recursive_sum);

780 781 782 783 784 785
/* Performs queue bypass and policy enabled checks then looks up blkg. */
static struct blkcg_gq *blkg_lookup_check(struct blkcg *blkcg,
					  const struct blkcg_policy *pol,
					  struct request_queue *q)
{
	WARN_ON_ONCE(!rcu_read_lock_held());
786
	lockdep_assert_held(&q->queue_lock);
787 788 789 790 791 792

	if (!blkcg_policy_enabled(q, pol))
		return ERR_PTR(-EOPNOTSUPP);
	return __blkg_lookup(blkcg, q, true /* update_hint */);
}

793 794 795
/**
 * blkg_conf_prep - parse and prepare for per-blkg config update
 * @blkcg: target block cgroup
796
 * @pol: target policy
797 798 799 800
 * @input: input string
 * @ctx: blkg_conf_ctx to be filled
 *
 * Parse per-blkg config update from @input and initialize @ctx with the
801 802 803
 * result.  @ctx->blkg points to the blkg to be updated and @ctx->body the
 * part of @input following MAJ:MIN.  This function returns with RCU read
 * lock and queue lock held and must be paired with blkg_conf_finish().
804
 */
T
Tejun Heo 已提交
805
int blkg_conf_prep(struct blkcg *blkcg, const struct blkcg_policy *pol,
806
		   char *input, struct blkg_conf_ctx *ctx)
807
	__acquires(rcu) __acquires(&disk->queue->queue_lock)
808
{
809
	struct gendisk *disk;
810
	struct request_queue *q;
T
Tejun Heo 已提交
811
	struct blkcg_gq *blkg;
T
Tejun Heo 已提交
812
	unsigned int major, minor;
813 814
	int key_len, part, ret;
	char *body;
815

816
	if (sscanf(input, "%u:%u%n", &major, &minor, &key_len) != 2)
T
Tejun Heo 已提交
817
		return -EINVAL;
818

819 820 821 822 823
	body = input + key_len;
	if (!isspace(*body))
		return -EINVAL;
	body = skip_spaces(body);

T
Tejun Heo 已提交
824
	disk = get_gendisk(MKDEV(major, minor), &part);
825
	if (!disk)
826
		return -ENODEV;
827
	if (part) {
828 829
		ret = -ENODEV;
		goto fail;
830
	}
831

832
	q = disk->queue;
833

834
	rcu_read_lock();
835
	spin_lock_irq(&q->queue_lock);
836

837
	blkg = blkg_lookup_check(blkcg, pol, q);
T
Tejun Heo 已提交
838 839
	if (IS_ERR(blkg)) {
		ret = PTR_ERR(blkg);
840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861
		goto fail_unlock;
	}

	if (blkg)
		goto success;

	/*
	 * Create blkgs walking down from blkcg_root to @blkcg, so that all
	 * non-root blkgs have access to their parents.
	 */
	while (true) {
		struct blkcg *pos = blkcg;
		struct blkcg *parent;
		struct blkcg_gq *new_blkg;

		parent = blkcg_parent(blkcg);
		while (parent && !__blkg_lookup(parent, q, false)) {
			pos = parent;
			parent = blkcg_parent(parent);
		}

		/* Drop locks to do new blkg allocation with GFP_KERNEL. */
862
		spin_unlock_irq(&q->queue_lock);
863
		rcu_read_unlock();
864 865 866 867 868

		new_blkg = blkg_alloc(pos, q, GFP_KERNEL);
		if (unlikely(!new_blkg)) {
			ret = -ENOMEM;
			goto fail;
869
		}
870

871
		rcu_read_lock();
872
		spin_lock_irq(&q->queue_lock);
873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893

		blkg = blkg_lookup_check(pos, pol, q);
		if (IS_ERR(blkg)) {
			ret = PTR_ERR(blkg);
			goto fail_unlock;
		}

		if (blkg) {
			blkg_free(new_blkg);
		} else {
			blkg = blkg_create(pos, q, new_blkg);
			if (unlikely(IS_ERR(blkg))) {
				ret = PTR_ERR(blkg);
				goto fail_unlock;
			}
		}

		if (pos == blkcg)
			goto success;
	}
success:
894 895
	ctx->disk = disk;
	ctx->blkg = blkg;
896
	ctx->body = body;
T
Tejun Heo 已提交
897
	return 0;
898 899

fail_unlock:
900
	spin_unlock_irq(&q->queue_lock);
901 902
	rcu_read_unlock();
fail:
903
	put_disk_and_module(disk);
904 905 906 907 908 909 910 911 912 913 914
	/*
	 * 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();
	}
	return ret;
915 916
}

917 918 919 920 921 922 923
/**
 * 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().
 */
924
void blkg_conf_finish(struct blkg_conf_ctx *ctx)
925
	__releases(&ctx->disk->queue->queue_lock) __releases(rcu)
926
{
927
	spin_unlock_irq(&ctx->disk->queue->queue_lock);
928
	rcu_read_unlock();
929
	put_disk_and_module(ctx->disk);
930 931
}

932 933 934 935 936 937 938 939 940
static int blkcg_print_stat(struct seq_file *sf, void *v)
{
	struct blkcg *blkcg = css_to_blkcg(seq_css(sf));
	struct blkcg_gq *blkg;

	rcu_read_lock();

	hlist_for_each_entry_rcu(blkg, &blkcg->blkg_list, blkcg_node) {
		const char *dname;
941
		char *buf;
942
		struct blkg_rwstat rwstat;
943
		u64 rbytes, wbytes, rios, wios, dbytes, dios;
944 945 946
		size_t size = seq_get_buf(sf, &buf), off = 0;
		int i;
		bool has_stats = false;
947 948 949 950 951

		dname = blkg_dev_name(blkg);
		if (!dname)
			continue;

952 953 954 955 956 957 958 959
		/*
		 * Hooray string manipulation, count is the size written NOT
		 * INCLUDING THE \0, so size is now count+1 less than what we
		 * had before, but we want to start writing the next bit from
		 * the \0 so we only add count to buf.
		 */
		off += scnprintf(buf+off, size-off, "%s ", dname);

960
		spin_lock_irq(&blkg->q->queue_lock);
961 962 963 964 965

		rwstat = blkg_rwstat_recursive_sum(blkg, NULL,
					offsetof(struct blkcg_gq, stat_bytes));
		rbytes = atomic64_read(&rwstat.aux_cnt[BLKG_RWSTAT_READ]);
		wbytes = atomic64_read(&rwstat.aux_cnt[BLKG_RWSTAT_WRITE]);
966
		dbytes = atomic64_read(&rwstat.aux_cnt[BLKG_RWSTAT_DISCARD]);
967 968 969 970 971

		rwstat = blkg_rwstat_recursive_sum(blkg, NULL,
					offsetof(struct blkcg_gq, stat_ios));
		rios = atomic64_read(&rwstat.aux_cnt[BLKG_RWSTAT_READ]);
		wios = atomic64_read(&rwstat.aux_cnt[BLKG_RWSTAT_WRITE]);
972
		dios = atomic64_read(&rwstat.aux_cnt[BLKG_RWSTAT_DISCARD]);
973

974
		spin_unlock_irq(&blkg->q->queue_lock);
975

976 977 978
		if (rbytes || wbytes || rios || wios) {
			has_stats = true;
			off += scnprintf(buf+off, size-off,
979 980 981
					 "rbytes=%llu wbytes=%llu rios=%llu wios=%llu dbytes=%llu dios=%llu",
					 rbytes, wbytes, rios, wios,
					 dbytes, dios);
982 983 984 985 986
		}

		if (!blkcg_debug_stats)
			goto next;

987 988 989 990 991 992 993 994
		if (atomic_read(&blkg->use_delay)) {
			has_stats = true;
			off += scnprintf(buf+off, size-off,
					 " use_delay=%d delay_nsec=%llu",
					 atomic_read(&blkg->use_delay),
					(unsigned long long)atomic64_read(&blkg->delay_nsec));
		}

995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011
		for (i = 0; i < BLKCG_MAX_POLS; i++) {
			struct blkcg_policy *pol = blkcg_policy[i];
			size_t written;

			if (!blkg->pd[i] || !pol->pd_stat_fn)
				continue;

			written = pol->pd_stat_fn(blkg->pd[i], buf+off, size-off);
			if (written)
				has_stats = true;
			off += written;
		}
next:
		if (has_stats) {
			off += scnprintf(buf+off, size-off, "\n");
			seq_commit(sf, off);
		}
1012 1013 1014 1015 1016 1017
	}

	rcu_read_unlock();
	return 0;
}

1018
static struct cftype blkcg_files[] = {
1019 1020
	{
		.name = "stat",
1021
		.flags = CFTYPE_NOT_ON_ROOT,
1022 1023 1024 1025 1026
		.seq_show = blkcg_print_stat,
	},
	{ }	/* terminate */
};

1027
static struct cftype blkcg_legacy_files[] = {
1028 1029
	{
		.name = "reset_stats",
T
Tejun Heo 已提交
1030
		.write_u64 = blkcg_reset_stats,
1031
	},
1032
	{ }	/* terminate */
1033 1034
};

1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055
/*
 * blkcg destruction is a three-stage process.
 *
 * 1. Destruction starts.  The blkcg_css_offline() callback is invoked
 *    which offlines writeback.  Here we tie the next stage of blkg destruction
 *    to the completion of writeback associated with the blkcg.  This lets us
 *    avoid punting potentially large amounts of outstanding writeback to root
 *    while maintaining any ongoing policies.  The next stage is triggered when
 *    the nr_cgwbs count goes to zero.
 *
 * 2. When the nr_cgwbs count goes to zero, blkcg_destroy_blkgs() is called
 *    and handles the destruction of blkgs.  Here the css reference held by
 *    the blkg is put back eventually allowing blkcg_css_free() to be called.
 *    This work may occur in cgwb_release_workfn() on the cgwb_release
 *    workqueue.  Any submitted ios that fail to get the blkg ref will be
 *    punted to the root_blkg.
 *
 * 3. Once the blkcg ref count goes to zero, blkcg_css_free() is called.
 *    This finally frees the blkcg.
 */

1056
/**
1057
 * blkcg_css_offline - cgroup css_offline callback
1058
 * @css: css of interest
1059
 *
1060 1061 1062
 * This function is called when @css is about to go away.  Here the cgwbs are
 * offlined first and only once writeback associated with the blkcg has
 * finished do we start step 2 (see above).
1063
 */
1064
static void blkcg_css_offline(struct cgroup_subsys_state *css)
1065
{
1066
	struct blkcg *blkcg = css_to_blkcg(css);
1067

1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087
	/* this prevents anyone from attaching or migrating to this blkcg */
	wb_blkcg_offline(blkcg);

	/* put the base cgwb reference allowing step 2 to be triggered */
	blkcg_cgwb_put(blkcg);
}

/**
 * blkcg_destroy_blkgs - responsible for shooting down blkgs
 * @blkcg: blkcg of interest
 *
 * 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.
 * Destroying the blkgs releases the reference held on the blkcg's css allowing
 * blkcg_css_free to eventually be called.
 *
 * This is the blkcg counterpart of ioc_release_fn().
 */
void blkcg_destroy_blkgs(struct blkcg *blkcg)
{
1088
	spin_lock_irq(&blkcg->lock);
1089

1090 1091
	while (!hlist_empty(&blkcg->blkg_list)) {
		struct blkcg_gq *blkg = hlist_entry(blkcg->blkg_list.first,
1092
						struct blkcg_gq, blkcg_node);
1093 1094
		struct request_queue *q = blkg->q;

1095
		if (spin_trylock(&q->queue_lock)) {
1096
			blkg_destroy(blkg);
1097
			spin_unlock(&q->queue_lock);
1098 1099 1100 1101 1102 1103
		} else {
			spin_unlock_irq(&blkcg->lock);
			cpu_relax();
			spin_lock_irq(&blkcg->lock);
		}
	}
1104

1105 1106 1107
	spin_unlock_irq(&blkcg->lock);
}

1108
static void blkcg_css_free(struct cgroup_subsys_state *css)
1109
{
1110
	struct blkcg *blkcg = css_to_blkcg(css);
1111
	int i;
1112

T
Tejun Heo 已提交
1113
	mutex_lock(&blkcg_pol_mutex);
1114

T
Tejun Heo 已提交
1115 1116
	list_del(&blkcg->all_blkcgs_node);

1117
	for (i = 0; i < BLKCG_MAX_POLS; i++)
1118 1119 1120 1121 1122
		if (blkcg->cpd[i])
			blkcg_policy[i]->cpd_free_fn(blkcg->cpd[i]);

	mutex_unlock(&blkcg_pol_mutex);

1123
	kfree(blkcg);
1124 1125
}

1126 1127
static struct cgroup_subsys_state *
blkcg_css_alloc(struct cgroup_subsys_state *parent_css)
1128
{
T
Tejun Heo 已提交
1129
	struct blkcg *blkcg;
1130 1131
	struct cgroup_subsys_state *ret;
	int i;
1132

T
Tejun Heo 已提交
1133 1134
	mutex_lock(&blkcg_pol_mutex);

1135
	if (!parent_css) {
T
Tejun Heo 已提交
1136
		blkcg = &blkcg_root;
1137 1138 1139 1140
	} else {
		blkcg = kzalloc(sizeof(*blkcg), GFP_KERNEL);
		if (!blkcg) {
			ret = ERR_PTR(-ENOMEM);
1141
			goto unlock;
1142
		}
1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154
	}

	for (i = 0; i < BLKCG_MAX_POLS ; i++) {
		struct blkcg_policy *pol = blkcg_policy[i];
		struct blkcg_policy_data *cpd;

		/*
		 * If the policy hasn't been attached yet, wait for it
		 * to be attached before doing anything else. Otherwise,
		 * check if the policy requires any specific per-cgroup
		 * data: if it does, allocate and initialize it.
		 */
1155
		if (!pol || !pol->cpd_alloc_fn)
1156 1157
			continue;

1158
		cpd = pol->cpd_alloc_fn(GFP_KERNEL);
1159 1160 1161 1162
		if (!cpd) {
			ret = ERR_PTR(-ENOMEM);
			goto free_pd_blkcg;
		}
1163 1164
		blkcg->cpd[i] = cpd;
		cpd->blkcg = blkcg;
1165
		cpd->plid = i;
1166 1167
		if (pol->cpd_init_fn)
			pol->cpd_init_fn(cpd);
1168
	}
1169 1170

	spin_lock_init(&blkcg->lock);
1171
	INIT_RADIX_TREE(&blkcg->blkg_tree, GFP_NOWAIT | __GFP_NOWARN);
1172
	INIT_HLIST_HEAD(&blkcg->blkg_list);
1173 1174
#ifdef CONFIG_CGROUP_WRITEBACK
	INIT_LIST_HEAD(&blkcg->cgwb_list);
1175
	refcount_set(&blkcg->cgwb_refcnt, 1);
1176
#endif
T
Tejun Heo 已提交
1177 1178 1179
	list_add_tail(&blkcg->all_blkcgs_node, &all_blkcgs);

	mutex_unlock(&blkcg_pol_mutex);
1180
	return &blkcg->css;
1181 1182 1183

free_pd_blkcg:
	for (i--; i >= 0; i--)
1184 1185
		if (blkcg->cpd[i])
			blkcg_policy[i]->cpd_free_fn(blkcg->cpd[i]);
1186 1187 1188 1189

	if (blkcg != &blkcg_root)
		kfree(blkcg);
unlock:
T
Tejun Heo 已提交
1190
	mutex_unlock(&blkcg_pol_mutex);
1191
	return ret;
1192 1193
}

1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205
/**
 * 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)
{
1206 1207
	struct blkcg_gq *new_blkg, *blkg;
	bool preloaded;
1208 1209
	int ret;

1210 1211 1212 1213 1214 1215
	new_blkg = blkg_alloc(&blkcg_root, q, GFP_KERNEL);
	if (!new_blkg)
		return -ENOMEM;

	preloaded = !radix_tree_preload(GFP_KERNEL);

1216
	/* Make sure the root blkg exists. */
1217
	rcu_read_lock();
1218
	spin_lock_irq(&q->queue_lock);
1219
	blkg = blkg_create(&blkcg_root, q, new_blkg);
1220 1221 1222
	if (IS_ERR(blkg))
		goto err_unlock;
	q->root_blkg = blkg;
1223
	spin_unlock_irq(&q->queue_lock);
1224 1225
	rcu_read_unlock();

1226 1227 1228
	if (preloaded)
		radix_tree_preload_end();

1229
	ret = blk_iolatency_init(q);
1230 1231
	if (ret)
		goto err_destroy_all;
1232

1233
	ret = blk_throtl_init(q);
1234 1235 1236
	if (ret)
		goto err_destroy_all;
	return 0;
1237

1238 1239 1240
err_destroy_all:
	blkg_destroy_all(q);
	return ret;
1241
err_unlock:
1242
	spin_unlock_irq(&q->queue_lock);
1243 1244 1245 1246
	rcu_read_unlock();
	if (preloaded)
		radix_tree_preload_end();
	return PTR_ERR(blkg);
1247 1248 1249 1250 1251 1252 1253 1254 1255 1256
}

/**
 * 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)
{
1257
	lockdep_assert_held(&q->queue_lock);
1258

1259 1260 1261 1262 1263 1264 1265
	/*
	 * @q could be exiting and already have destroyed all blkgs as
	 * indicated by NULL root_blkg.  If so, don't confuse policies.
	 */
	if (!q->root_blkg)
		return;

1266 1267 1268 1269 1270 1271 1272
	blk_throtl_drain(q);
}

/**
 * blkcg_exit_queue - exit and release blkcg part of request_queue
 * @q: request_queue being released
 *
1273
 * Called from blk_exit_queue().  Responsible for exiting blkcg part.
1274 1275 1276
 */
void blkcg_exit_queue(struct request_queue *q)
{
1277
	blkg_destroy_all(q);
1278 1279 1280
	blk_throtl_exit(q);
}

1281 1282 1283 1284 1285 1286
/*
 * 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.
 */
1287
static int blkcg_can_attach(struct cgroup_taskset *tset)
1288
{
1289
	struct task_struct *task;
1290
	struct cgroup_subsys_state *dst_css;
1291 1292 1293 1294
	struct io_context *ioc;
	int ret = 0;

	/* task_lock() is needed to avoid races with exit_io_context() */
1295
	cgroup_taskset_for_each(task, dst_css, tset) {
1296 1297 1298 1299 1300 1301 1302 1303
		task_lock(task);
		ioc = task->io_context;
		if (ioc && atomic_read(&ioc->nr_tasks) > 1)
			ret = -EINVAL;
		task_unlock(task);
		if (ret)
			break;
	}
1304 1305 1306
	return ret;
}

1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326
static void blkcg_bind(struct cgroup_subsys_state *root_css)
{
	int i;

	mutex_lock(&blkcg_pol_mutex);

	for (i = 0; i < BLKCG_MAX_POLS; i++) {
		struct blkcg_policy *pol = blkcg_policy[i];
		struct blkcg *blkcg;

		if (!pol || !pol->cpd_bind_fn)
			continue;

		list_for_each_entry(blkcg, &all_blkcgs, all_blkcgs_node)
			if (blkcg->cpd[pol->plid])
				pol->cpd_bind_fn(blkcg->cpd[pol->plid]);
	}
	mutex_unlock(&blkcg_pol_mutex);
}

1327 1328 1329 1330 1331 1332 1333
static void blkcg_exit(struct task_struct *tsk)
{
	if (tsk->throttle_queue)
		blk_put_queue(tsk->throttle_queue);
	tsk->throttle_queue = NULL;
}

1334
struct cgroup_subsys io_cgrp_subsys = {
1335 1336 1337
	.css_alloc = blkcg_css_alloc,
	.css_offline = blkcg_css_offline,
	.css_free = blkcg_css_free,
T
Tejun Heo 已提交
1338
	.can_attach = blkcg_can_attach,
1339
	.bind = blkcg_bind,
1340
	.dfl_cftypes = blkcg_files,
1341
	.legacy_cftypes = blkcg_legacy_files,
1342
	.legacy_name = "blkio",
1343
	.exit = blkcg_exit,
1344 1345 1346 1347 1348 1349 1350 1351
#ifdef CONFIG_MEMCG
	/*
	 * This ensures that, if available, memcg is automatically enabled
	 * together on the default hierarchy so that the owner cgroup can
	 * be retrieved from writeback pages.
	 */
	.depends_on = 1 << memory_cgrp_id,
#endif
1352
};
1353
EXPORT_SYMBOL_GPL(io_cgrp_subsys);
1354

1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371
/**
 * 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
1372
			  const struct blkcg_policy *pol)
1373
{
1374
	struct blkg_policy_data *pd_prealloc = NULL;
1375
	struct blkcg_gq *blkg;
1376
	int ret;
1377 1378 1379 1380

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

J
Jens Axboe 已提交
1381
	if (queue_is_mq(q))
1382
		blk_mq_freeze_queue(q);
1383 1384
pd_prealloc:
	if (!pd_prealloc) {
1385
		pd_prealloc = pol->pd_alloc_fn(GFP_KERNEL, q->node);
1386
		if (!pd_prealloc) {
1387
			ret = -ENOMEM;
1388
			goto out_bypass_end;
1389 1390 1391
		}
	}

1392
	spin_lock_irq(&q->queue_lock);
1393 1394

	list_for_each_entry(blkg, &q->blkg_list, q_node) {
1395 1396 1397 1398
		struct blkg_policy_data *pd;

		if (blkg->pd[pol->plid])
			continue;
1399

1400
		pd = pol->pd_alloc_fn(GFP_NOWAIT | __GFP_NOWARN, q->node);
1401 1402 1403
		if (!pd)
			swap(pd, pd_prealloc);
		if (!pd) {
1404
			spin_unlock_irq(&q->queue_lock);
1405 1406
			goto pd_prealloc;
		}
1407 1408 1409

		blkg->pd[pol->plid] = pd;
		pd->blkg = blkg;
T
Tejun Heo 已提交
1410
		pd->plid = pol->plid;
1411
		if (pol->pd_init_fn)
1412
			pol->pd_init_fn(pd);
1413 1414 1415 1416
	}

	__set_bit(pol->plid, q->blkcg_pols);
	ret = 0;
1417

1418
	spin_unlock_irq(&q->queue_lock);
1419
out_bypass_end:
J
Jens Axboe 已提交
1420
	if (queue_is_mq(q))
1421
		blk_mq_unfreeze_queue(q);
1422 1423
	if (pd_prealloc)
		pol->pd_free_fn(pd_prealloc);
1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436
	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
1437
			     const struct blkcg_policy *pol)
1438
{
T
Tejun Heo 已提交
1439
	struct blkcg_gq *blkg;
1440 1441 1442 1443

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

J
Jens Axboe 已提交
1444
	if (queue_is_mq(q))
1445 1446
		blk_mq_freeze_queue(q);

1447
	spin_lock_irq(&q->queue_lock);
1448 1449 1450 1451

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

	list_for_each_entry(blkg, &q->blkg_list, q_node) {
1452
		if (blkg->pd[pol->plid]) {
1453
			if (pol->pd_offline_fn)
1454
				pol->pd_offline_fn(blkg->pd[pol->plid]);
1455 1456 1457
			pol->pd_free_fn(blkg->pd[pol->plid]);
			blkg->pd[pol->plid] = NULL;
		}
1458 1459
	}

1460
	spin_unlock_irq(&q->queue_lock);
1461

J
Jens Axboe 已提交
1462
	if (queue_is_mq(q))
1463
		blk_mq_unfreeze_queue(q);
1464 1465 1466
}
EXPORT_SYMBOL_GPL(blkcg_deactivate_policy);

T
Tejun Heo 已提交
1467
/**
T
Tejun Heo 已提交
1468 1469
 * blkcg_policy_register - register a blkcg policy
 * @pol: blkcg policy to register
T
Tejun Heo 已提交
1470
 *
T
Tejun Heo 已提交
1471 1472
 * 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
1473
 */
1474
int blkcg_policy_register(struct blkcg_policy *pol)
1475
{
1476
	struct blkcg *blkcg;
T
Tejun Heo 已提交
1477
	int i, ret;
1478

1479
	mutex_lock(&blkcg_pol_register_mutex);
1480 1481
	mutex_lock(&blkcg_pol_mutex);

T
Tejun Heo 已提交
1482 1483 1484
	/* find an empty slot */
	ret = -ENOSPC;
	for (i = 0; i < BLKCG_MAX_POLS; i++)
T
Tejun Heo 已提交
1485
		if (!blkcg_policy[i])
T
Tejun Heo 已提交
1486
			break;
1487 1488
	if (i >= BLKCG_MAX_POLS) {
		pr_warn("blkcg_policy_register: BLKCG_MAX_POLS too small\n");
1489
		goto err_unlock;
1490
	}
1491

1492 1493 1494 1495 1496
	/* Make sure cpd/pd_alloc_fn and cpd/pd_free_fn in pairs */
	if ((!pol->cpd_alloc_fn ^ !pol->cpd_free_fn) ||
		(!pol->pd_alloc_fn ^ !pol->pd_free_fn))
		goto err_unlock;

1497
	/* register @pol */
T
Tejun Heo 已提交
1498
	pol->plid = i;
1499 1500 1501
	blkcg_policy[pol->plid] = pol;

	/* allocate and install cpd's */
1502
	if (pol->cpd_alloc_fn) {
1503 1504 1505
		list_for_each_entry(blkcg, &all_blkcgs, all_blkcgs_node) {
			struct blkcg_policy_data *cpd;

1506
			cpd = pol->cpd_alloc_fn(GFP_KERNEL);
1507
			if (!cpd)
1508 1509
				goto err_free_cpds;

1510 1511
			blkcg->cpd[pol->plid] = cpd;
			cpd->blkcg = blkcg;
1512
			cpd->plid = pol->plid;
1513
			pol->cpd_init_fn(cpd);
1514 1515 1516
		}
	}

1517
	mutex_unlock(&blkcg_pol_mutex);
T
Tejun Heo 已提交
1518 1519

	/* everything is in place, add intf files for the new policy */
1520 1521 1522
	if (pol->dfl_cftypes)
		WARN_ON(cgroup_add_dfl_cftypes(&io_cgrp_subsys,
					       pol->dfl_cftypes));
1523
	if (pol->legacy_cftypes)
1524
		WARN_ON(cgroup_add_legacy_cftypes(&io_cgrp_subsys,
1525
						  pol->legacy_cftypes));
1526 1527 1528
	mutex_unlock(&blkcg_pol_register_mutex);
	return 0;

1529
err_free_cpds:
1530
	if (pol->cpd_free_fn) {
1531
		list_for_each_entry(blkcg, &all_blkcgs, all_blkcgs_node) {
1532 1533 1534 1535
			if (blkcg->cpd[pol->plid]) {
				pol->cpd_free_fn(blkcg->cpd[pol->plid]);
				blkcg->cpd[pol->plid] = NULL;
			}
1536 1537 1538
		}
	}
	blkcg_policy[pol->plid] = NULL;
1539
err_unlock:
1540
	mutex_unlock(&blkcg_pol_mutex);
1541
	mutex_unlock(&blkcg_pol_register_mutex);
T
Tejun Heo 已提交
1542
	return ret;
1543
}
T
Tejun Heo 已提交
1544
EXPORT_SYMBOL_GPL(blkcg_policy_register);
1545

T
Tejun Heo 已提交
1546
/**
T
Tejun Heo 已提交
1547 1548
 * blkcg_policy_unregister - unregister a blkcg policy
 * @pol: blkcg policy to unregister
T
Tejun Heo 已提交
1549
 *
T
Tejun Heo 已提交
1550
 * Undo blkcg_policy_register(@pol).  Might sleep.
T
Tejun Heo 已提交
1551
 */
T
Tejun Heo 已提交
1552
void blkcg_policy_unregister(struct blkcg_policy *pol)
1553
{
1554 1555
	struct blkcg *blkcg;

1556
	mutex_lock(&blkcg_pol_register_mutex);
1557

T
Tejun Heo 已提交
1558
	if (WARN_ON(blkcg_policy[pol->plid] != pol))
T
Tejun Heo 已提交
1559 1560 1561
		goto out_unlock;

	/* kill the intf files first */
1562 1563
	if (pol->dfl_cftypes)
		cgroup_rm_cftypes(pol->dfl_cftypes);
1564 1565
	if (pol->legacy_cftypes)
		cgroup_rm_cftypes(pol->legacy_cftypes);
1566

1567
	/* remove cpds and unregister */
1568
	mutex_lock(&blkcg_pol_mutex);
1569

1570
	if (pol->cpd_free_fn) {
1571
		list_for_each_entry(blkcg, &all_blkcgs, all_blkcgs_node) {
1572 1573 1574 1575
			if (blkcg->cpd[pol->plid]) {
				pol->cpd_free_fn(blkcg->cpd[pol->plid]);
				blkcg->cpd[pol->plid] = NULL;
			}
1576 1577
		}
	}
T
Tejun Heo 已提交
1578
	blkcg_policy[pol->plid] = NULL;
1579

1580
	mutex_unlock(&blkcg_pol_mutex);
1581 1582
out_unlock:
	mutex_unlock(&blkcg_pol_register_mutex);
1583
}
T
Tejun Heo 已提交
1584
EXPORT_SYMBOL_GPL(blkcg_policy_unregister);
1585

1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622 1623 1624 1625 1626 1627 1628 1629 1630 1631 1632 1633 1634 1635 1636 1637 1638 1639 1640 1641 1642 1643 1644 1645 1646 1647 1648 1649 1650 1651 1652 1653 1654 1655 1656 1657 1658 1659 1660 1661 1662 1663 1664 1665 1666 1667 1668 1669 1670 1671 1672 1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692 1693 1694 1695 1696 1697 1698 1699 1700 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724
/*
 * Scale the accumulated delay based on how long it has been since we updated
 * the delay.  We only call this when we are adding delay, in case it's been a
 * while since we added delay, and when we are checking to see if we need to
 * delay a task, to account for any delays that may have occurred.
 */
static void blkcg_scale_delay(struct blkcg_gq *blkg, u64 now)
{
	u64 old = atomic64_read(&blkg->delay_start);

	/*
	 * We only want to scale down every second.  The idea here is that we
	 * want to delay people for min(delay_nsec, NSEC_PER_SEC) in a certain
	 * time window.  We only want to throttle tasks for recent delay that
	 * has occurred, in 1 second time windows since that's the maximum
	 * things can be throttled.  We save the current delay window in
	 * blkg->last_delay so we know what amount is still left to be charged
	 * to the blkg from this point onward.  blkg->last_use keeps track of
	 * the use_delay counter.  The idea is if we're unthrottling the blkg we
	 * are ok with whatever is happening now, and we can take away more of
	 * the accumulated delay as we've already throttled enough that
	 * everybody is happy with their IO latencies.
	 */
	if (time_before64(old + NSEC_PER_SEC, now) &&
	    atomic64_cmpxchg(&blkg->delay_start, old, now) == old) {
		u64 cur = atomic64_read(&blkg->delay_nsec);
		u64 sub = min_t(u64, blkg->last_delay, now - old);
		int cur_use = atomic_read(&blkg->use_delay);

		/*
		 * We've been unthrottled, subtract a larger chunk of our
		 * accumulated delay.
		 */
		if (cur_use < blkg->last_use)
			sub = max_t(u64, sub, blkg->last_delay >> 1);

		/*
		 * This shouldn't happen, but handle it anyway.  Our delay_nsec
		 * should only ever be growing except here where we subtract out
		 * min(last_delay, 1 second), but lord knows bugs happen and I'd
		 * rather not end up with negative numbers.
		 */
		if (unlikely(cur < sub)) {
			atomic64_set(&blkg->delay_nsec, 0);
			blkg->last_delay = 0;
		} else {
			atomic64_sub(sub, &blkg->delay_nsec);
			blkg->last_delay = cur - sub;
		}
		blkg->last_use = cur_use;
	}
}

/*
 * This is called when we want to actually walk up the hierarchy and check to
 * see if we need to throttle, and then actually throttle if there is some
 * accumulated delay.  This should only be called upon return to user space so
 * we're not holding some lock that would induce a priority inversion.
 */
static void blkcg_maybe_throttle_blkg(struct blkcg_gq *blkg, bool use_memdelay)
{
	u64 now = ktime_to_ns(ktime_get());
	u64 exp;
	u64 delay_nsec = 0;
	int tok;

	while (blkg->parent) {
		if (atomic_read(&blkg->use_delay)) {
			blkcg_scale_delay(blkg, now);
			delay_nsec = max_t(u64, delay_nsec,
					   atomic64_read(&blkg->delay_nsec));
		}
		blkg = blkg->parent;
	}

	if (!delay_nsec)
		return;

	/*
	 * Let's not sleep for all eternity if we've amassed a huge delay.
	 * Swapping or metadata IO can accumulate 10's of seconds worth of
	 * delay, and we want userspace to be able to do _something_ so cap the
	 * delays at 1 second.  If there's 10's of seconds worth of delay then
	 * the tasks will be delayed for 1 second for every syscall.
	 */
	delay_nsec = min_t(u64, delay_nsec, 250 * NSEC_PER_MSEC);

	/*
	 * TODO: the use_memdelay flag is going to be for the upcoming psi stuff
	 * that hasn't landed upstream yet.  Once that stuff is in place we need
	 * to do a psi_memstall_enter/leave if memdelay is set.
	 */

	exp = ktime_add_ns(now, delay_nsec);
	tok = io_schedule_prepare();
	do {
		__set_current_state(TASK_KILLABLE);
		if (!schedule_hrtimeout(&exp, HRTIMER_MODE_ABS))
			break;
	} while (!fatal_signal_pending(current));
	io_schedule_finish(tok);
}

/**
 * blkcg_maybe_throttle_current - throttle the current task if it has been marked
 *
 * This is only called if we've been marked with set_notify_resume().  Obviously
 * we can be set_notify_resume() for reasons other than blkcg throttling, so we
 * check to see if current->throttle_queue is set and if not this doesn't do
 * anything.  This should only ever be called by the resume code, it's not meant
 * to be called by people willy-nilly as it will actually do the work to
 * throttle the task if it is setup for throttling.
 */
void blkcg_maybe_throttle_current(void)
{
	struct request_queue *q = current->throttle_queue;
	struct cgroup_subsys_state *css;
	struct blkcg *blkcg;
	struct blkcg_gq *blkg;
	bool use_memdelay = current->use_memdelay;

	if (!q)
		return;

	current->throttle_queue = NULL;
	current->use_memdelay = false;

	rcu_read_lock();
	css = kthread_blkcg();
	if (css)
		blkcg = css_to_blkcg(css);
	else
		blkcg = css_to_blkcg(task_css(current, io_cgrp_id));

	if (!blkcg)
		goto out;
	blkg = blkg_lookup(blkcg, q);
	if (!blkg)
		goto out;
1725
	if (!blkg_tryget(blkg))
1726 1727 1728 1729 1730
		goto out;
	rcu_read_unlock();

	blkcg_maybe_throttle_blkg(blkg, use_memdelay);
	blkg_put(blkg);
1731
	blk_put_queue(q);
1732 1733 1734 1735 1736 1737 1738 1739
	return;
out:
	rcu_read_unlock();
	blk_put_queue(q);
}

/**
 * blkcg_schedule_throttle - this task needs to check for throttling
B
Bart Van Assche 已提交
1740 1741
 * @q: the request queue IO was submitted on
 * @use_memdelay: do we charge this to memory delay for PSI
1742 1743 1744 1745 1746 1747 1748 1749 1750 1751 1752 1753 1754 1755 1756 1757 1758 1759 1760 1761 1762 1763 1764 1765 1766 1767 1768 1769 1770 1771 1772
 *
 * This is called by the IO controller when we know there's delay accumulated
 * for the blkg for this task.  We do not pass the blkg because there are places
 * we call this that may not have that information, the swapping code for
 * instance will only have a request_queue at that point.  This set's the
 * notify_resume for the task to check and see if it requires throttling before
 * returning to user space.
 *
 * We will only schedule once per syscall.  You can call this over and over
 * again and it will only do the check once upon return to user space, and only
 * throttle once.  If the task needs to be throttled again it'll need to be
 * re-set at the next time we see the task.
 */
void blkcg_schedule_throttle(struct request_queue *q, bool use_memdelay)
{
	if (unlikely(current->flags & PF_KTHREAD))
		return;

	if (!blk_get_queue(q))
		return;

	if (current->throttle_queue)
		blk_put_queue(current->throttle_queue);
	current->throttle_queue = q;
	if (use_memdelay)
		current->use_memdelay = use_memdelay;
	set_notify_resume(current);
}

/**
 * blkcg_add_delay - add delay to this blkg
B
Bart Van Assche 已提交
1773 1774 1775
 * @blkg: blkg of interest
 * @now: the current time in nanoseconds
 * @delta: how many nanoseconds of delay to add
1776 1777 1778 1779 1780 1781 1782 1783 1784 1785
 *
 * Charge @delta to the blkg's current delay accumulation.  This is used to
 * throttle tasks if an IO controller thinks we need more throttling.
 */
void blkcg_add_delay(struct blkcg_gq *blkg, u64 now, u64 delta)
{
	blkcg_scale_delay(blkg, now);
	atomic64_add(delta, &blkg->delay_nsec);
}

1786 1787
module_param(blkcg_debug_stats, bool, 0644);
MODULE_PARM_DESC(blkcg_debug_stats, "True if you want debug stats, false if not");