backing-dev.c 26.8 KB
Newer Older
1 2 3

#include <linux/wait.h>
#include <linux/backing-dev.h>
4 5
#include <linux/kthread.h>
#include <linux/freezer.h>
6
#include <linux/fs.h>
7
#include <linux/pagemap.h>
8
#include <linux/mm.h>
9 10
#include <linux/sched.h>
#include <linux/module.h>
11 12
#include <linux/writeback.h>
#include <linux/device.h>
13
#include <trace/events/writeback.h>
14

15 16
struct backing_dev_info noop_backing_dev_info = {
	.name		= "noop",
17
	.capabilities	= BDI_CAP_NO_ACCT_AND_WRITEBACK,
18
};
19
EXPORT_SYMBOL_GPL(noop_backing_dev_info);
20

21
static struct class *bdi_class;
22 23

/*
24
 * bdi_lock protects updates to bdi_list. bdi_list has RCU reader side
25 26
 * locking.
 */
27
DEFINE_SPINLOCK(bdi_lock);
28
LIST_HEAD(bdi_list);
29

30 31 32
/* bdi_wq serves all asynchronous writeback tasks */
struct workqueue_struct *bdi_wq;

33 34 35 36 37 38 39 40 41 42 43 44 45 46
#ifdef CONFIG_DEBUG_FS
#include <linux/debugfs.h>
#include <linux/seq_file.h>

static struct dentry *bdi_debug_root;

static void bdi_debug_init(void)
{
	bdi_debug_root = debugfs_create_dir("bdi", NULL);
}

static int bdi_debug_stats_show(struct seq_file *m, void *v)
{
	struct backing_dev_info *bdi = m->private;
47
	struct bdi_writeback *wb = &bdi->wb;
48 49
	unsigned long background_thresh;
	unsigned long dirty_thresh;
50
	unsigned long wb_thresh;
51
	unsigned long nr_dirty, nr_io, nr_more_io, nr_dirty_time;
52 53
	struct inode *inode;

54
	nr_dirty = nr_io = nr_more_io = nr_dirty_time = 0;
55
	spin_lock(&wb->list_lock);
56
	list_for_each_entry(inode, &wb->b_dirty, i_io_list)
57
		nr_dirty++;
58
	list_for_each_entry(inode, &wb->b_io, i_io_list)
59
		nr_io++;
60
	list_for_each_entry(inode, &wb->b_more_io, i_io_list)
61
		nr_more_io++;
62
	list_for_each_entry(inode, &wb->b_dirty_time, i_io_list)
63 64
		if (inode->i_state & I_DIRTY_TIME)
			nr_dirty_time++;
65
	spin_unlock(&wb->list_lock);
66

67
	global_dirty_limits(&background_thresh, &dirty_thresh);
68
	wb_thresh = wb_calc_thresh(wb, dirty_thresh);
69 70 71

#define K(x) ((x) << (PAGE_SHIFT - 10))
	seq_printf(m,
72 73 74 75 76
		   "BdiWriteback:       %10lu kB\n"
		   "BdiReclaimable:     %10lu kB\n"
		   "BdiDirtyThresh:     %10lu kB\n"
		   "DirtyThresh:        %10lu kB\n"
		   "BackgroundThresh:   %10lu kB\n"
77
		   "BdiDirtied:         %10lu kB\n"
78 79 80 81 82
		   "BdiWritten:         %10lu kB\n"
		   "BdiWriteBandwidth:  %10lu kBps\n"
		   "b_dirty:            %10lu\n"
		   "b_io:               %10lu\n"
		   "b_more_io:          %10lu\n"
83
		   "b_dirty_time:       %10lu\n"
84 85
		   "bdi_list:           %10u\n"
		   "state:              %10lx\n",
86 87
		   (unsigned long) K(wb_stat(wb, WB_WRITEBACK)),
		   (unsigned long) K(wb_stat(wb, WB_RECLAIMABLE)),
88
		   K(wb_thresh),
89 90
		   K(dirty_thresh),
		   K(background_thresh),
91 92
		   (unsigned long) K(wb_stat(wb, WB_DIRTIED)),
		   (unsigned long) K(wb_stat(wb, WB_WRITTEN)),
93
		   (unsigned long) K(wb->write_bandwidth),
94 95 96
		   nr_dirty,
		   nr_io,
		   nr_more_io,
97
		   nr_dirty_time,
98
		   !list_empty(&bdi->bdi_list), bdi->wb.state);
99 100 101 102
#undef K

	return 0;
}
103
DEFINE_SHOW_ATTRIBUTE(bdi_debug_stats);
104

105
static int bdi_debug_register(struct backing_dev_info *bdi, const char *name)
106
{
107 108 109
	if (!bdi_debug_root)
		return -ENOMEM;

110
	bdi->debug_dir = debugfs_create_dir(name, bdi_debug_root);
111 112 113
	if (!bdi->debug_dir)
		return -ENOMEM;

114 115
	bdi->debug_stats = debugfs_create_file("stats", 0444, bdi->debug_dir,
					       bdi, &bdi_debug_stats_fops);
116 117
	if (!bdi->debug_stats) {
		debugfs_remove(bdi->debug_dir);
118
		bdi->debug_dir = NULL;
119 120 121 122
		return -ENOMEM;
	}

	return 0;
123 124 125 126 127 128 129 130 131 132 133
}

static void bdi_debug_unregister(struct backing_dev_info *bdi)
{
	debugfs_remove(bdi->debug_stats);
	debugfs_remove(bdi->debug_dir);
}
#else
static inline void bdi_debug_init(void)
{
}
134
static inline int bdi_debug_register(struct backing_dev_info *bdi,
135 136
				      const char *name)
{
137
	return 0;
138 139 140 141 142 143
}
static inline void bdi_debug_unregister(struct backing_dev_info *bdi)
{
}
#endif

144 145 146 147 148 149
static ssize_t read_ahead_kb_store(struct device *dev,
				  struct device_attribute *attr,
				  const char *buf, size_t count)
{
	struct backing_dev_info *bdi = dev_get_drvdata(dev);
	unsigned long read_ahead_kb;
150
	ssize_t ret;
151

152 153 154 155 156 157 158
	ret = kstrtoul(buf, 10, &read_ahead_kb);
	if (ret < 0)
		return ret;

	bdi->ra_pages = read_ahead_kb >> (PAGE_SHIFT - 10);

	return count;
159 160 161 162 163 164 165 166 167 168 169
}

#define K(pages) ((pages) << (PAGE_SHIFT - 10))

#define BDI_SHOW(name, expr)						\
static ssize_t name##_show(struct device *dev,				\
			   struct device_attribute *attr, char *page)	\
{									\
	struct backing_dev_info *bdi = dev_get_drvdata(dev);		\
									\
	return snprintf(page, PAGE_SIZE-1, "%lld\n", (long long)expr);	\
170 171
}									\
static DEVICE_ATTR_RW(name);
172 173 174

BDI_SHOW(read_ahead_kb, K(bdi->ra_pages))

175 176 177 178 179
static ssize_t min_ratio_store(struct device *dev,
		struct device_attribute *attr, const char *buf, size_t count)
{
	struct backing_dev_info *bdi = dev_get_drvdata(dev);
	unsigned int ratio;
180 181 182 183 184 185 186 187 188
	ssize_t ret;

	ret = kstrtouint(buf, 10, &ratio);
	if (ret < 0)
		return ret;

	ret = bdi_set_min_ratio(bdi, ratio);
	if (!ret)
		ret = count;
189 190 191 192 193

	return ret;
}
BDI_SHOW(min_ratio, bdi->min_ratio)

194 195 196 197 198
static ssize_t max_ratio_store(struct device *dev,
		struct device_attribute *attr, const char *buf, size_t count)
{
	struct backing_dev_info *bdi = dev_get_drvdata(dev);
	unsigned int ratio;
199 200 201 202 203 204 205 206 207
	ssize_t ret;

	ret = kstrtouint(buf, 10, &ratio);
	if (ret < 0)
		return ret;

	ret = bdi_set_max_ratio(bdi, ratio);
	if (!ret)
		ret = count;
208 209 210 211 212

	return ret;
}
BDI_SHOW(max_ratio, bdi->max_ratio)

213 214 215 216 217 218 219 220 221
static ssize_t stable_pages_required_show(struct device *dev,
					  struct device_attribute *attr,
					  char *page)
{
	struct backing_dev_info *bdi = dev_get_drvdata(dev);

	return snprintf(page, PAGE_SIZE-1, "%d\n",
			bdi_cap_stable_pages_required(bdi) ? 1 : 0);
}
222 223 224 225 226 227 228 229
static DEVICE_ATTR_RO(stable_pages_required);

static struct attribute *bdi_dev_attrs[] = {
	&dev_attr_read_ahead_kb.attr,
	&dev_attr_min_ratio.attr,
	&dev_attr_max_ratio.attr,
	&dev_attr_stable_pages_required.attr,
	NULL,
230
};
231
ATTRIBUTE_GROUPS(bdi_dev);
232 233 234 235

static __init int bdi_class_init(void)
{
	bdi_class = class_create(THIS_MODULE, "bdi");
236 237 238
	if (IS_ERR(bdi_class))
		return PTR_ERR(bdi_class);

239
	bdi_class->dev_groups = bdi_dev_groups;
240
	bdi_debug_init();
241

242 243
	return 0;
}
244
postcore_initcall(bdi_class_init);
245

246 247
static int bdi_init(struct backing_dev_info *bdi);

248 249 250 251
static int __init default_bdi_init(void)
{
	int err;

252
	bdi_wq = alloc_workqueue("writeback", WQ_MEM_RECLAIM | WQ_FREEZABLE |
253
					      WQ_UNBOUND | WQ_SYSFS, 0);
254 255 256
	if (!bdi_wq)
		return -ENOMEM;

257
	err = bdi_init(&noop_backing_dev_info);
258 259 260 261 262

	return err;
}
subsys_initcall(default_bdi_init);

263
/*
264
 * This function is used when the first inode for this wb is marked dirty. It
265 266 267 268 269 270 271 272
 * wakes-up the corresponding bdi thread which should then take care of the
 * periodic background write-out of dirty inodes. Since the write-out would
 * starts only 'dirty_writeback_interval' centisecs from now anyway, we just
 * set up a timer which wakes the bdi thread up later.
 *
 * Note, we wouldn't bother setting up the timer, but this function is on the
 * fast-path (used by '__mark_inode_dirty()'), so we save few context switches
 * by delaying the wake-up.
273 274 275
 *
 * We have to be careful not to postpone flush work if it is scheduled for
 * earlier. Thus we use queue_delayed_work().
276
 */
277
void wb_wakeup_delayed(struct bdi_writeback *wb)
278 279 280 281
{
	unsigned long timeout;

	timeout = msecs_to_jiffies(dirty_writeback_interval * 10);
282 283 284 285
	spin_lock_bh(&wb->work_lock);
	if (test_bit(WB_registered, &wb->state))
		queue_delayed_work(bdi_wq, &wb->dwork, timeout);
	spin_unlock_bh(&wb->work_lock);
286 287
}

288
/*
289
 * Initial write bandwidth: 100 MB/s
290
 */
291
#define INIT_BW		(100 << (20 - PAGE_SHIFT))
292

293
static int wb_init(struct bdi_writeback *wb, struct backing_dev_info *bdi,
294
		   int blkcg_id, gfp_t gfp)
295
{
296
	int i, err;
297

298
	memset(wb, 0, sizeof(*wb));
299

300 301
	if (wb != &bdi->wb)
		bdi_get(bdi);
302 303 304 305 306
	wb->bdi = bdi;
	wb->last_old_flush = jiffies;
	INIT_LIST_HEAD(&wb->b_dirty);
	INIT_LIST_HEAD(&wb->b_io);
	INIT_LIST_HEAD(&wb->b_more_io);
307
	INIT_LIST_HEAD(&wb->b_dirty_time);
308
	spin_lock_init(&wb->list_lock);
309

310 311 312 313 314
	wb->bw_time_stamp = jiffies;
	wb->balanced_dirty_ratelimit = INIT_BW;
	wb->dirty_ratelimit = INIT_BW;
	wb->write_bandwidth = INIT_BW;
	wb->avg_write_bandwidth = INIT_BW;
315

316 317 318
	spin_lock_init(&wb->work_lock);
	INIT_LIST_HEAD(&wb->work_list);
	INIT_DELAYED_WORK(&wb->dwork, wb_workfn);
319
	wb->dirty_sleep = jiffies;
320

321
	wb->congested = wb_congested_get_create(bdi, blkcg_id, gfp);
322 323 324 325
	if (!wb->congested) {
		err = -ENOMEM;
		goto out_put_bdi;
	}
326

327
	err = fprop_local_init_percpu(&wb->completions, gfp);
328
	if (err)
329
		goto out_put_cong;
330

331
	for (i = 0; i < NR_WB_STAT_ITEMS; i++) {
332
		err = percpu_counter_init(&wb->stat[i], 0, gfp);
333 334
		if (err)
			goto out_destroy_stat;
335
	}
336

337
	return 0;
338 339

out_destroy_stat:
340
	while (i--)
341 342 343 344
		percpu_counter_destroy(&wb->stat[i]);
	fprop_local_destroy_percpu(&wb->completions);
out_put_cong:
	wb_congested_put(wb->congested);
345 346 347
out_put_bdi:
	if (wb != &bdi->wb)
		bdi_put(bdi);
348
	return err;
349 350
}

351 352
static void cgwb_remove_from_bdi_list(struct bdi_writeback *wb);

353 354 355
/*
 * Remove bdi from the global list and shutdown any threads we have running
 */
356
static void wb_shutdown(struct bdi_writeback *wb)
357
{
358
	/* Make sure nobody queues further work */
359 360 361
	spin_lock_bh(&wb->work_lock);
	if (!test_and_clear_bit(WB_registered, &wb->state)) {
		spin_unlock_bh(&wb->work_lock);
362
		return;
363
	}
364
	spin_unlock_bh(&wb->work_lock);
365

366
	cgwb_remove_from_bdi_list(wb);
367
	/*
368 369 370
	 * Drain work list and shutdown the delayed_work.  !WB_registered
	 * tells wb_workfn() that @wb is dying and its work_list needs to
	 * be drained no matter what.
371
	 */
372 373 374 375 376
	mod_delayed_work(bdi_wq, &wb->dwork, 0);
	flush_delayed_work(&wb->dwork);
	WARN_ON(!list_empty(&wb->work_list));
}

377
static void wb_exit(struct bdi_writeback *wb)
378 379 380 381 382 383 384
{
	int i;

	WARN_ON(delayed_work_pending(&wb->dwork));

	for (i = 0; i < NR_WB_STAT_ITEMS; i++)
		percpu_counter_destroy(&wb->stat[i]);
385

386
	fprop_local_destroy_percpu(&wb->completions);
387
	wb_congested_put(wb->congested);
388 389
	if (wb != &wb->bdi->wb)
		bdi_put(wb->bdi);
390
}
391

392 393 394 395 396 397 398
#ifdef CONFIG_CGROUP_WRITEBACK

#include <linux/memcontrol.h>

/*
 * cgwb_lock protects bdi->cgwb_tree, bdi->cgwb_congested_tree,
 * blkcg->cgwb_list, and memcg->cgwb_list.  bdi->cgwb_tree is also RCU
399
 * protected.
400 401
 */
static DEFINE_SPINLOCK(cgwb_lock);
402
static struct workqueue_struct *cgwb_release_wq;
403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427

/**
 * wb_congested_get_create - get or create a wb_congested
 * @bdi: associated bdi
 * @blkcg_id: ID of the associated blkcg
 * @gfp: allocation mask
 *
 * Look up the wb_congested for @blkcg_id on @bdi.  If missing, create one.
 * The returned wb_congested has its reference count incremented.  Returns
 * NULL on failure.
 */
struct bdi_writeback_congested *
wb_congested_get_create(struct backing_dev_info *bdi, int blkcg_id, gfp_t gfp)
{
	struct bdi_writeback_congested *new_congested = NULL, *congested;
	struct rb_node **node, *parent;
	unsigned long flags;
retry:
	spin_lock_irqsave(&cgwb_lock, flags);

	node = &bdi->cgwb_congested_tree.rb_node;
	parent = NULL;

	while (*node != NULL) {
		parent = *node;
428 429
		congested = rb_entry(parent, struct bdi_writeback_congested,
				     rb_node);
430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454
		if (congested->blkcg_id < blkcg_id)
			node = &parent->rb_left;
		else if (congested->blkcg_id > blkcg_id)
			node = &parent->rb_right;
		else
			goto found;
	}

	if (new_congested) {
		/* !found and storage for new one already allocated, insert */
		congested = new_congested;
		new_congested = NULL;
		rb_link_node(&congested->rb_node, parent, node);
		rb_insert_color(&congested->rb_node, &bdi->cgwb_congested_tree);
		goto found;
	}

	spin_unlock_irqrestore(&cgwb_lock, flags);

	/* allocate storage for new one and retry */
	new_congested = kzalloc(sizeof(*new_congested), gfp);
	if (!new_congested)
		return NULL;

	atomic_set(&new_congested->refcnt, 0);
455
	new_congested->__bdi = bdi;
456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481
	new_congested->blkcg_id = blkcg_id;
	goto retry;

found:
	atomic_inc(&congested->refcnt);
	spin_unlock_irqrestore(&cgwb_lock, flags);
	kfree(new_congested);
	return congested;
}

/**
 * wb_congested_put - put a wb_congested
 * @congested: wb_congested to put
 *
 * Put @congested and destroy it if the refcnt reaches zero.
 */
void wb_congested_put(struct bdi_writeback_congested *congested)
{
	unsigned long flags;

	local_irq_save(flags);
	if (!atomic_dec_and_lock(&congested->refcnt, &cgwb_lock)) {
		local_irq_restore(flags);
		return;
	}

482
	/* bdi might already have been destroyed leaving @congested unlinked */
483
	if (congested->__bdi) {
484
		rb_erase(&congested->rb_node,
485 486
			 &congested->__bdi->cgwb_congested_tree);
		congested->__bdi = NULL;
487 488
	}

489 490 491 492 493 494 495 496 497
	spin_unlock_irqrestore(&cgwb_lock, flags);
	kfree(congested);
}

static void cgwb_release_workfn(struct work_struct *work)
{
	struct bdi_writeback *wb = container_of(work, struct bdi_writeback,
						release_work);

498
	mutex_lock(&wb->bdi->cgwb_release_mutex);
499 500 501 502
	wb_shutdown(wb);

	css_put(wb->memcg_css);
	css_put(wb->blkcg_css);
503
	mutex_unlock(&wb->bdi->cgwb_release_mutex);
504

505
	fprop_local_destroy_percpu(&wb->memcg_completions);
506 507 508 509 510 511 512 513 514
	percpu_ref_exit(&wb->refcnt);
	wb_exit(wb);
	kfree_rcu(wb, rcu);
}

static void cgwb_release(struct percpu_ref *refcnt)
{
	struct bdi_writeback *wb = container_of(refcnt, struct bdi_writeback,
						refcnt);
515
	queue_work(cgwb_release_wq, &wb->release_work);
516 517 518 519 520 521 522 523 524 525 526 527
}

static void cgwb_kill(struct bdi_writeback *wb)
{
	lockdep_assert_held(&cgwb_lock);

	WARN_ON(!radix_tree_delete(&wb->bdi->cgwb_tree, wb->memcg_css->id));
	list_del(&wb->memcg_node);
	list_del(&wb->blkcg_node);
	percpu_ref_kill(&wb->refcnt);
}

528 529 530 531 532 533 534
static void cgwb_remove_from_bdi_list(struct bdi_writeback *wb)
{
	spin_lock_irq(&cgwb_lock);
	list_del_rcu(&wb->bdi_node);
	spin_unlock_irq(&cgwb_lock);
}

535 536 537 538 539 540 541 542 543 544 545 546
static int cgwb_create(struct backing_dev_info *bdi,
		       struct cgroup_subsys_state *memcg_css, gfp_t gfp)
{
	struct mem_cgroup *memcg;
	struct cgroup_subsys_state *blkcg_css;
	struct blkcg *blkcg;
	struct list_head *memcg_cgwb_list, *blkcg_cgwb_list;
	struct bdi_writeback *wb;
	unsigned long flags;
	int ret = 0;

	memcg = mem_cgroup_from_css(memcg_css);
547
	blkcg_css = cgroup_get_e_css(memcg_css->cgroup, &io_cgrp_subsys);
548
	blkcg = css_to_blkcg(blkcg_css);
549
	memcg_cgwb_list = &memcg->cgwb_list;
550 551 552 553 554 555 556 557 558 559 560 561 562 563 564
	blkcg_cgwb_list = &blkcg->cgwb_list;

	/* look up again under lock and discard on blkcg mismatch */
	spin_lock_irqsave(&cgwb_lock, flags);
	wb = radix_tree_lookup(&bdi->cgwb_tree, memcg_css->id);
	if (wb && wb->blkcg_css != blkcg_css) {
		cgwb_kill(wb);
		wb = NULL;
	}
	spin_unlock_irqrestore(&cgwb_lock, flags);
	if (wb)
		goto out_put;

	/* need to create a new one */
	wb = kmalloc(sizeof(*wb), gfp);
565 566 567 568
	if (!wb) {
		ret = -ENOMEM;
		goto out_put;
	}
569

570
	ret = wb_init(wb, bdi, blkcg_css->id, gfp);
571 572 573 574 575 576 577
	if (ret)
		goto err_free;

	ret = percpu_ref_init(&wb->refcnt, cgwb_release, 0, gfp);
	if (ret)
		goto err_wb_exit;

578 579 580 581
	ret = fprop_local_init_percpu(&wb->memcg_completions, gfp);
	if (ret)
		goto err_ref_exit;

582 583 584 585
	wb->memcg_css = memcg_css;
	wb->blkcg_css = blkcg_css;
	INIT_WORK(&wb->release_work, cgwb_release_workfn);
	set_bit(WB_registered, &wb->state);
586 587

	/*
588 589 590 591
	 * The root wb determines the registered state of the whole bdi and
	 * memcg_cgwb_list and blkcg_cgwb_list's next pointers indicate
	 * whether they're still online.  Don't link @wb if any is dead.
	 * See wb_memcg_offline() and wb_blkcg_offline().
592
	 */
593 594 595 596 597 598 599
	ret = -ENODEV;
	spin_lock_irqsave(&cgwb_lock, flags);
	if (test_bit(WB_registered, &bdi->wb.state) &&
	    blkcg_cgwb_list->next && memcg_cgwb_list->next) {
		/* we might have raced another instance of this function */
		ret = radix_tree_insert(&bdi->cgwb_tree, memcg_css->id, wb);
		if (!ret) {
600
			list_add_tail_rcu(&wb->bdi_node, &bdi->wb_list);
601 602 603 604 605 606 607 608 609 610
			list_add(&wb->memcg_node, memcg_cgwb_list);
			list_add(&wb->blkcg_node, blkcg_cgwb_list);
			css_get(memcg_css);
			css_get(blkcg_css);
		}
	}
	spin_unlock_irqrestore(&cgwb_lock, flags);
	if (ret) {
		if (ret == -EEXIST)
			ret = 0;
611
		goto err_fprop_exit;
612 613 614
	}
	goto out_put;

615 616
err_fprop_exit:
	fprop_local_destroy_percpu(&wb->memcg_completions);
617 618 619 620 621 622 623 624 625
err_ref_exit:
	percpu_ref_exit(&wb->refcnt);
err_wb_exit:
	wb_exit(wb);
err_free:
	kfree(wb);
out_put:
	css_put(blkcg_css);
	return ret;
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
/**
 * wb_get_create - get wb for a given memcg, create if necessary
 * @bdi: target bdi
 * @memcg_css: cgroup_subsys_state of the target memcg (must have positive ref)
 * @gfp: allocation mask to use
 *
 * Try to get the wb for @memcg_css on @bdi.  If it doesn't exist, try to
 * create one.  The returned wb has its refcount incremented.
 *
 * This function uses css_get() on @memcg_css and thus expects its refcnt
 * to be positive on invocation.  IOW, rcu_read_lock() protection on
 * @memcg_css isn't enough.  try_get it before calling this function.
 *
 * A wb is keyed by its associated memcg.  As blkcg implicitly enables
 * memcg on the default hierarchy, memcg association is guaranteed to be
 * more specific (equal or descendant to the associated blkcg) and thus can
 * identify both the memcg and blkcg associations.
 *
 * Because the blkcg associated with a memcg may change as blkcg is enabled
 * and disabled closer to root in the hierarchy, each wb keeps track of
 * both the memcg and blkcg associated with it and verifies the blkcg on
 * each lookup.  On mismatch, the existing wb is discarded and a new one is
 * created.
 */
struct bdi_writeback *wb_get_create(struct backing_dev_info *bdi,
				    struct cgroup_subsys_state *memcg_css,
				    gfp_t gfp)
655
{
656 657
	struct bdi_writeback *wb;

658
	might_sleep_if(gfpflags_allow_blocking(gfp));
659 660 661 662 663 664 665 666 667 668 669 670

	if (!memcg_css->parent)
		return &bdi->wb;

	do {
		rcu_read_lock();
		wb = radix_tree_lookup(&bdi->cgwb_tree, memcg_css->id);
		if (wb) {
			struct cgroup_subsys_state *blkcg_css;

			/* see whether the blkcg association has changed */
			blkcg_css = cgroup_get_e_css(memcg_css->cgroup,
671
						     &io_cgrp_subsys);
672 673 674 675 676 677 678 679 680 681
			if (unlikely(wb->blkcg_css != blkcg_css ||
				     !wb_tryget(wb)))
				wb = NULL;
			css_put(blkcg_css);
		}
		rcu_read_unlock();
	} while (!wb && !cgwb_create(bdi, memcg_css, gfp));

	return wb;
}
682

683
static int cgwb_bdi_init(struct backing_dev_info *bdi)
684
{
685 686
	int ret;

687 688
	INIT_RADIX_TREE(&bdi->cgwb_tree, GFP_ATOMIC);
	bdi->cgwb_congested_tree = RB_ROOT;
689
	mutex_init(&bdi->cgwb_release_mutex);
690 691 692

	ret = wb_init(&bdi->wb, bdi, 1, GFP_KERNEL);
	if (!ret) {
693
		bdi->wb.memcg_css = &root_mem_cgroup->css;
694 695 696
		bdi->wb.blkcg_css = blkcg_root_css;
	}
	return ret;
697 698
}

699
static void cgwb_bdi_unregister(struct backing_dev_info *bdi)
700 701 702
{
	struct radix_tree_iter iter;
	void **slot;
703
	struct bdi_writeback *wb;
704 705 706 707 708 709

	WARN_ON(test_bit(WB_registered, &bdi->wb.state));

	spin_lock_irq(&cgwb_lock);
	radix_tree_for_each_slot(slot, &bdi->cgwb_tree, &iter, 0)
		cgwb_kill(*slot);
710
	spin_unlock_irq(&cgwb_lock);
711

712 713
	mutex_lock(&bdi->cgwb_release_mutex);
	spin_lock_irq(&cgwb_lock);
714 715 716 717 718 719 720
	while (!list_empty(&bdi->wb_list)) {
		wb = list_first_entry(&bdi->wb_list, struct bdi_writeback,
				      bdi_node);
		spin_unlock_irq(&cgwb_lock);
		wb_shutdown(wb);
		spin_lock_irq(&cgwb_lock);
	}
721
	spin_unlock_irq(&cgwb_lock);
722
	mutex_unlock(&bdi->cgwb_release_mutex);
723 724 725 726 727 728 729
}

/**
 * wb_memcg_offline - kill all wb's associated with a memcg being offlined
 * @memcg: memcg being offlined
 *
 * Also prevents creation of any new wb's associated with @memcg.
730
 */
731 732
void wb_memcg_offline(struct mem_cgroup *memcg)
{
733
	struct list_head *memcg_cgwb_list = &memcg->cgwb_list;
734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759
	struct bdi_writeback *wb, *next;

	spin_lock_irq(&cgwb_lock);
	list_for_each_entry_safe(wb, next, memcg_cgwb_list, memcg_node)
		cgwb_kill(wb);
	memcg_cgwb_list->next = NULL;	/* prevent new wb's */
	spin_unlock_irq(&cgwb_lock);
}

/**
 * wb_blkcg_offline - kill all wb's associated with a blkcg being offlined
 * @blkcg: blkcg being offlined
 *
 * Also prevents creation of any new wb's associated with @blkcg.
 */
void wb_blkcg_offline(struct blkcg *blkcg)
{
	struct bdi_writeback *wb, *next;

	spin_lock_irq(&cgwb_lock);
	list_for_each_entry_safe(wb, next, &blkcg->cgwb_list, blkcg_node)
		cgwb_kill(wb);
	blkcg->cgwb_list.next = NULL;	/* prevent new wb's */
	spin_unlock_irq(&cgwb_lock);
}

760 761 762 763 764 765 766 767 768 769
static void cgwb_bdi_exit(struct backing_dev_info *bdi)
{
	struct rb_node *rbn;

	spin_lock_irq(&cgwb_lock);
	while ((rbn = rb_first(&bdi->cgwb_congested_tree))) {
		struct bdi_writeback_congested *congested =
			rb_entry(rbn, struct bdi_writeback_congested, rb_node);

		rb_erase(rbn, &bdi->cgwb_congested_tree);
770
		congested->__bdi = NULL;	/* mark @congested unlinked */
771 772 773 774
	}
	spin_unlock_irq(&cgwb_lock);
}

775 776 777 778 779 780 781
static void cgwb_bdi_register(struct backing_dev_info *bdi)
{
	spin_lock_irq(&cgwb_lock);
	list_add_tail_rcu(&bdi->wb.bdi_node, &bdi->wb_list);
	spin_unlock_irq(&cgwb_lock);
}

782 783 784 785 786 787 788 789 790 791 792 793 794 795 796
static int __init cgwb_init(void)
{
	/*
	 * There can be many concurrent release work items overwhelming
	 * system_wq.  Put them in a separate wq and limit concurrency.
	 * There's no point in executing many of these in parallel.
	 */
	cgwb_release_wq = alloc_workqueue("cgwb_release", 0, 1);
	if (!cgwb_release_wq)
		return -ENOMEM;

	return 0;
}
subsys_initcall(cgwb_init);

797 798
#else	/* CONFIG_CGROUP_WRITEBACK */

799 800 801 802 803 804 805 806
static int cgwb_bdi_init(struct backing_dev_info *bdi)
{
	int err;

	bdi->wb_congested = kzalloc(sizeof(*bdi->wb_congested), GFP_KERNEL);
	if (!bdi->wb_congested)
		return -ENOMEM;

807 808
	atomic_set(&bdi->wb_congested->refcnt, 1);

809 810
	err = wb_init(&bdi->wb, bdi, 1, GFP_KERNEL);
	if (err) {
811
		wb_congested_put(bdi->wb_congested);
812 813 814 815 816
		return err;
	}
	return 0;
}

817
static void cgwb_bdi_unregister(struct backing_dev_info *bdi) { }
818 819

static void cgwb_bdi_exit(struct backing_dev_info *bdi)
820 821 822
{
	wb_congested_put(bdi->wb_congested);
}
823

824 825 826 827 828 829 830 831 832 833
static void cgwb_bdi_register(struct backing_dev_info *bdi)
{
	list_add_tail_rcu(&bdi->wb.bdi_node, &bdi->wb_list);
}

static void cgwb_remove_from_bdi_list(struct bdi_writeback *wb)
{
	list_del_rcu(&wb->bdi_node);
}

834
#endif	/* CONFIG_CGROUP_WRITEBACK */
835

836
static int bdi_init(struct backing_dev_info *bdi)
837
{
838 839
	int ret;

840 841
	bdi->dev = NULL;

842
	kref_init(&bdi->refcnt);
843
	bdi->min_ratio = 0;
844
	bdi->max_ratio = 100;
845
	bdi->max_prop_frac = FPROP_FRAC_BASE;
846
	INIT_LIST_HEAD(&bdi->bdi_list);
847
	INIT_LIST_HEAD(&bdi->wb_list);
848
	init_waitqueue_head(&bdi->wb_waitq);
849

850 851 852
	ret = cgwb_bdi_init(bdi);

	return ret;
853
}
854

855 856 857 858 859 860 861 862 863 864 865 866 867 868 869
struct backing_dev_info *bdi_alloc_node(gfp_t gfp_mask, int node_id)
{
	struct backing_dev_info *bdi;

	bdi = kmalloc_node(sizeof(struct backing_dev_info),
			   gfp_mask | __GFP_ZERO, node_id);
	if (!bdi)
		return NULL;

	if (bdi_init(bdi)) {
		kfree(bdi);
		return NULL;
	}
	return bdi;
}
870
EXPORT_SYMBOL(bdi_alloc_node);
871

872
int bdi_register_va(struct backing_dev_info *bdi, const char *fmt, va_list args)
873 874
{
	struct device *dev;
875

876 877
	if (bdi->dev)	/* The driver needs to use separate queues per device */
		return 0;
878

879
	dev = device_create_vargs(bdi_class, NULL, MKDEV(0, 0), bdi, fmt, args);
880 881
	if (IS_ERR(dev))
		return PTR_ERR(dev);
882

883
	cgwb_bdi_register(bdi);
884
	bdi->dev = dev;
885

886
	bdi_debug_register(bdi, dev_name(dev));
887 888 889 890 891 892 893 894
	set_bit(WB_registered, &bdi->wb.state);

	spin_lock_bh(&bdi_lock);
	list_add_tail_rcu(&bdi->bdi_list, &bdi_list);
	spin_unlock_bh(&bdi_lock);

	trace_writeback_bdi_register(bdi);
	return 0;
895
}
896 897
EXPORT_SYMBOL(bdi_register_va);

898
int bdi_register(struct backing_dev_info *bdi, const char *fmt, ...)
899 900 901 902 903
{
	va_list args;
	int ret;

	va_start(args, fmt);
904
	ret = bdi_register_va(bdi, fmt, args);
905 906 907
	va_end(args);
	return ret;
}
908
EXPORT_SYMBOL(bdi_register);
909

910 911 912 913
int bdi_register_owner(struct backing_dev_info *bdi, struct device *owner)
{
	int rc;

914
	rc = bdi_register(bdi, "%u:%u", MAJOR(owner->devt), MINOR(owner->devt));
915 916
	if (rc)
		return rc;
917 918
	/* Leaking owner reference... */
	WARN_ON(bdi->owner);
919 920 921 922 923 924
	bdi->owner = owner;
	get_device(owner);
	return 0;
}
EXPORT_SYMBOL(bdi_register_owner);

925 926 927 928 929 930 931 932
/*
 * Remove bdi from bdi_list, and ensure that it is no longer visible
 */
static void bdi_remove_from_list(struct backing_dev_info *bdi)
{
	spin_lock_bh(&bdi_lock);
	list_del_rcu(&bdi->bdi_list);
	spin_unlock_bh(&bdi_lock);
933

934 935
	synchronize_rcu_expedited();
}
936

937
void bdi_unregister(struct backing_dev_info *bdi)
938
{
939 940 941
	/* make sure nobody finds us on the bdi_list anymore */
	bdi_remove_from_list(bdi);
	wb_shutdown(&bdi->wb);
942
	cgwb_bdi_unregister(bdi);
943

944 945 946 947 948
	if (bdi->dev) {
		bdi_debug_unregister(bdi);
		device_unregister(bdi->dev);
		bdi->dev = NULL;
	}
949 950 951 952 953

	if (bdi->owner) {
		put_device(bdi->owner);
		bdi->owner = NULL;
	}
954
}
955

956 957 958 959 960
static void release_bdi(struct kref *ref)
{
	struct backing_dev_info *bdi =
			container_of(ref, struct backing_dev_info, refcnt);

961 962
	if (test_bit(WB_registered, &bdi->wb.state))
		bdi_unregister(bdi);
963 964 965
	WARN_ON_ONCE(bdi->dev);
	wb_exit(&bdi->wb);
	cgwb_bdi_exit(bdi);
966 967 968 969 970 971 972
	kfree(bdi);
}

void bdi_put(struct backing_dev_info *bdi)
{
	kref_put(&bdi->refcnt, release_bdi);
}
973
EXPORT_SYMBOL(bdi_put);
974

975 976 977 978
static wait_queue_head_t congestion_wqh[2] = {
		__WAIT_QUEUE_HEAD_INITIALIZER(congestion_wqh[0]),
		__WAIT_QUEUE_HEAD_INITIALIZER(congestion_wqh[1])
	};
979
static atomic_t nr_wb_congested[2];
980

981
void clear_wb_congested(struct bdi_writeback_congested *congested, int sync)
982
{
983
	wait_queue_head_t *wqh = &congestion_wqh[sync];
984
	enum wb_congested_state bit;
985

986
	bit = sync ? WB_sync_congested : WB_async_congested;
987 988
	if (test_and_clear_bit(bit, &congested->state))
		atomic_dec(&nr_wb_congested[sync]);
989
	smp_mb__after_atomic();
990 991 992
	if (waitqueue_active(wqh))
		wake_up(wqh);
}
993
EXPORT_SYMBOL(clear_wb_congested);
994

995
void set_wb_congested(struct bdi_writeback_congested *congested, int sync)
996
{
997
	enum wb_congested_state bit;
998

999
	bit = sync ? WB_sync_congested : WB_async_congested;
1000 1001
	if (!test_and_set_bit(bit, &congested->state))
		atomic_inc(&nr_wb_congested[sync]);
1002
}
1003
EXPORT_SYMBOL(set_wb_congested);
1004 1005 1006

/**
 * congestion_wait - wait for a backing_dev to become uncongested
1007
 * @sync: SYNC or ASYNC IO
1008 1009 1010 1011 1012 1013
 * @timeout: timeout in jiffies
 *
 * Waits for up to @timeout jiffies for a backing_dev (any backing_dev) to exit
 * write congestion.  If no backing_devs are congested then just wait for the
 * next write to be completed.
 */
1014
long congestion_wait(int sync, long timeout)
1015 1016
{
	long ret;
1017
	unsigned long start = jiffies;
1018
	DEFINE_WAIT(wait);
1019
	wait_queue_head_t *wqh = &congestion_wqh[sync];
1020 1021 1022 1023

	prepare_to_wait(wqh, &wait, TASK_UNINTERRUPTIBLE);
	ret = io_schedule_timeout(timeout);
	finish_wait(wqh, &wait);
1024 1025 1026 1027

	trace_writeback_congestion_wait(jiffies_to_usecs(timeout),
					jiffies_to_usecs(jiffies - start));

1028 1029 1030
	return ret;
}
EXPORT_SYMBOL(congestion_wait);
1031

1032
/**
1033
 * wait_iff_congested - Conditionally wait for a backing_dev to become uncongested or a pgdat to complete writes
1034 1035 1036
 * @sync: SYNC or ASYNC IO
 * @timeout: timeout in jiffies
 *
1037 1038 1039
 * In the event of a congested backing_dev (any backing_dev) this waits
 * for up to @timeout jiffies for either a BDI to exit congestion of the
 * given @sync queue or a write to complete.
1040 1041 1042 1043 1044
 *
 * The return value is 0 if the sleep is for the full timeout. Otherwise,
 * it is the number of jiffies that were still remaining when the function
 * returned. return_value == timeout implies the function did not sleep.
 */
1045
long wait_iff_congested(int sync, long timeout)
1046 1047 1048 1049 1050 1051 1052
{
	long ret;
	unsigned long start = jiffies;
	DEFINE_WAIT(wait);
	wait_queue_head_t *wqh = &congestion_wqh[sync];

	/*
1053
	 * If there is no congestion, yield if necessary instead
1054 1055
	 * of sleeping on the congestion queue
	 */
1056
	if (atomic_read(&nr_wb_congested[sync]) == 0) {
1057
		cond_resched();
1058

1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078
		/* In case we scheduled, work out time remaining */
		ret = timeout - (jiffies - start);
		if (ret < 0)
			ret = 0;

		goto out;
	}

	/* Sleep until uncongested or a write happens */
	prepare_to_wait(wqh, &wait, TASK_UNINTERRUPTIBLE);
	ret = io_schedule_timeout(timeout);
	finish_wait(wqh, &wait);

out:
	trace_writeback_wait_iff_congested(jiffies_to_usecs(timeout),
					jiffies_to_usecs(jiffies - start));

	return ret;
}
EXPORT_SYMBOL(wait_iff_congested);