page-writeback.c 36.2 KB
Newer Older
L
Linus Torvalds 已提交
1
/*
2
 * mm/page-writeback.c
L
Linus Torvalds 已提交
3 4
 *
 * Copyright (C) 2002, Linus Torvalds.
P
Peter Zijlstra 已提交
5
 * Copyright (C) 2007 Red Hat, Inc., Peter Zijlstra <pzijlstr@redhat.com>
L
Linus Torvalds 已提交
6 7 8 9
 *
 * Contains functions related to writing back dirty pages at the
 * address_space level.
 *
10
 * 10Apr2002	Andrew Morton
L
Linus Torvalds 已提交
11 12 13 14 15 16 17 18 19 20 21 22 23 24
 *		Initial version
 */

#include <linux/kernel.h>
#include <linux/module.h>
#include <linux/spinlock.h>
#include <linux/fs.h>
#include <linux/mm.h>
#include <linux/swap.h>
#include <linux/slab.h>
#include <linux/pagemap.h>
#include <linux/writeback.h>
#include <linux/init.h>
#include <linux/backing-dev.h>
25
#include <linux/task_io_accounting_ops.h>
L
Linus Torvalds 已提交
26 27
#include <linux/blkdev.h>
#include <linux/mpage.h>
28
#include <linux/rmap.h>
L
Linus Torvalds 已提交
29 30 31 32 33 34
#include <linux/percpu.h>
#include <linux/notifier.h>
#include <linux/smp.h>
#include <linux/sysctl.h>
#include <linux/cpu.h>
#include <linux/syscalls.h>
35
#include <linux/buffer_head.h>
36
#include <linux/pagevec.h>
L
Linus Torvalds 已提交
37 38 39 40 41 42 43 44 45 46

/*
 * After a CPU has dirtied this many pages, balance_dirty_pages_ratelimited
 * will look to see if it needs to force writeback or throttling.
 */
static long ratelimit_pages = 32;

/*
 * When balance_dirty_pages decides that the caller needs to perform some
 * non-background writeback, this is how many pages it will attempt to write.
47
 * It should be somewhat larger than dirtied pages to ensure that reasonably
L
Linus Torvalds 已提交
48 49
 * large amounts of I/O are submitted.
 */
50
static inline long sync_writeback_pages(unsigned long dirtied)
L
Linus Torvalds 已提交
51
{
52 53 54 55
	if (dirtied < ratelimit_pages)
		dirtied = ratelimit_pages;

	return dirtied + dirtied / 2;
L
Linus Torvalds 已提交
56 57 58 59 60 61 62
}

/* The following parameters are exported via /proc/sys/vm */

/*
 * Start background writeback (via pdflush) at this percentage
 */
63
int dirty_background_ratio = 10;
L
Linus Torvalds 已提交
64

65 66 67 68 69 70
/*
 * dirty_background_bytes starts at 0 (disabled) so that it is a function of
 * dirty_background_ratio * the amount of dirtyable memory
 */
unsigned long dirty_background_bytes;

71 72 73 74 75 76
/*
 * free highmem will not be subtracted from the total free memory
 * for calculating free ratios if vm_highmem_is_dirtyable is true
 */
int vm_highmem_is_dirtyable;

L
Linus Torvalds 已提交
77 78 79
/*
 * The generator of dirty data starts writeback at this percentage
 */
80
int vm_dirty_ratio = 20;
L
Linus Torvalds 已提交
81

82 83 84 85 86 87
/*
 * vm_dirty_bytes starts at 0 (disabled) so that it is a function of
 * vm_dirty_ratio * the amount of dirtyable memory
 */
unsigned long vm_dirty_bytes;

L
Linus Torvalds 已提交
88
/*
89
 * The interval between `kupdate'-style writebacks
L
Linus Torvalds 已提交
90
 */
91
unsigned int dirty_writeback_interval = 5 * 100; /* centiseconds */
L
Linus Torvalds 已提交
92 93

/*
94
 * The longest time for which data is allowed to remain dirty
L
Linus Torvalds 已提交
95
 */
96
unsigned int dirty_expire_interval = 30 * 100; /* centiseconds */
L
Linus Torvalds 已提交
97 98 99 100 101 102 103

/*
 * Flag that makes the machine dump writes/reads and block dirtyings.
 */
int block_dump;

/*
104 105
 * Flag that puts the machine in "laptop mode". Doubles as a timeout in jiffies:
 * a full sync is triggered after this time elapses without any disk activity.
L
Linus Torvalds 已提交
106 107 108 109 110 111 112 113
 */
int laptop_mode;

EXPORT_SYMBOL(laptop_mode);

/* End of sysctl-exported parameters */


P
Peter Zijlstra 已提交
114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130
/*
 * Scale the writeback cache size proportional to the relative writeout speeds.
 *
 * We do this by keeping a floating proportion between BDIs, based on page
 * writeback completions [end_page_writeback()]. Those devices that write out
 * pages fastest will get the larger share, while the slower will get a smaller
 * share.
 *
 * We use page writeout completions because we are interested in getting rid of
 * dirty pages. Having them written out is the primary goal.
 *
 * We introduce a concept of time, a period over which we measure these events,
 * because demand can/will vary over time. The length of this period itself is
 * measured in page writeback completions.
 *
 */
static struct prop_descriptor vm_completions;
P
Peter Zijlstra 已提交
131
static struct prop_descriptor vm_dirties;
P
Peter Zijlstra 已提交
132 133 134 135 136 137 138 139 140 141

/*
 * couple the period to the dirty_ratio:
 *
 *   period/2 ~ roundup_pow_of_two(dirty limit)
 */
static int calc_period_shift(void)
{
	unsigned long dirty_total;

142 143 144 145 146
	if (vm_dirty_bytes)
		dirty_total = vm_dirty_bytes / PAGE_SIZE;
	else
		dirty_total = (vm_dirty_ratio * determine_dirtyable_memory()) /
				100;
P
Peter Zijlstra 已提交
147 148 149 150
	return 2 + ilog2(dirty_total - 1);
}

/*
151
 * update the period when the dirty threshold changes.
P
Peter Zijlstra 已提交
152
 */
153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183
static void update_completion_period(void)
{
	int shift = calc_period_shift();
	prop_change_shift(&vm_completions, shift);
	prop_change_shift(&vm_dirties, shift);
}

int dirty_background_ratio_handler(struct ctl_table *table, int write,
		struct file *filp, void __user *buffer, size_t *lenp,
		loff_t *ppos)
{
	int ret;

	ret = proc_dointvec_minmax(table, write, filp, buffer, lenp, ppos);
	if (ret == 0 && write)
		dirty_background_bytes = 0;
	return ret;
}

int dirty_background_bytes_handler(struct ctl_table *table, int write,
		struct file *filp, void __user *buffer, size_t *lenp,
		loff_t *ppos)
{
	int ret;

	ret = proc_doulongvec_minmax(table, write, filp, buffer, lenp, ppos);
	if (ret == 0 && write)
		dirty_background_ratio = 0;
	return ret;
}

P
Peter Zijlstra 已提交
184 185 186 187 188
int dirty_ratio_handler(struct ctl_table *table, int write,
		struct file *filp, void __user *buffer, size_t *lenp,
		loff_t *ppos)
{
	int old_ratio = vm_dirty_ratio;
189 190 191
	int ret;

	ret = proc_dointvec_minmax(table, write, filp, buffer, lenp, ppos);
P
Peter Zijlstra 已提交
192
	if (ret == 0 && write && vm_dirty_ratio != old_ratio) {
193 194 195 196 197 198 199 200 201 202 203
		update_completion_period();
		vm_dirty_bytes = 0;
	}
	return ret;
}


int dirty_bytes_handler(struct ctl_table *table, int write,
		struct file *filp, void __user *buffer, size_t *lenp,
		loff_t *ppos)
{
204
	unsigned long old_bytes = vm_dirty_bytes;
205 206 207 208 209 210
	int ret;

	ret = proc_doulongvec_minmax(table, write, filp, buffer, lenp, ppos);
	if (ret == 0 && write && vm_dirty_bytes != old_bytes) {
		update_completion_period();
		vm_dirty_ratio = 0;
P
Peter Zijlstra 已提交
211 212 213 214 215 216 217 218 219 220
	}
	return ret;
}

/*
 * Increment the BDI's writeout completion count and the global writeout
 * completion count. Called from test_clear_page_writeback().
 */
static inline void __bdi_writeout_inc(struct backing_dev_info *bdi)
{
221 222
	__prop_inc_percpu_max(&vm_completions, &bdi->completions,
			      bdi->max_prop_frac);
P
Peter Zijlstra 已提交
223 224
}

225 226 227 228 229 230 231 232 233 234
void bdi_writeout_inc(struct backing_dev_info *bdi)
{
	unsigned long flags;

	local_irq_save(flags);
	__bdi_writeout_inc(bdi);
	local_irq_restore(flags);
}
EXPORT_SYMBOL_GPL(bdi_writeout_inc);

N
Nick Piggin 已提交
235
void task_dirty_inc(struct task_struct *tsk)
P
Peter Zijlstra 已提交
236 237 238 239
{
	prop_inc_single(&vm_dirties, &tsk->dirties);
}

P
Peter Zijlstra 已提交
240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259
/*
 * Obtain an accurate fraction of the BDI's portion.
 */
static void bdi_writeout_fraction(struct backing_dev_info *bdi,
		long *numerator, long *denominator)
{
	if (bdi_cap_writeback_dirty(bdi)) {
		prop_fraction_percpu(&vm_completions, &bdi->completions,
				numerator, denominator);
	} else {
		*numerator = 0;
		*denominator = 1;
	}
}

/*
 * Clip the earned share of dirty pages to that which is actually available.
 * This avoids exceeding the total dirty_limit when the floating averages
 * fluctuate too quickly.
 */
260 261
static void clip_bdi_dirty_limit(struct backing_dev_info *bdi,
		unsigned long dirty, unsigned long *pbdi_dirty)
P
Peter Zijlstra 已提交
262
{
263
	unsigned long avail_dirty;
P
Peter Zijlstra 已提交
264

265
	avail_dirty = global_page_state(NR_FILE_DIRTY) +
P
Peter Zijlstra 已提交
266
		 global_page_state(NR_WRITEBACK) +
267
		 global_page_state(NR_UNSTABLE_NFS) +
268
		 global_page_state(NR_WRITEBACK_TEMP);
P
Peter Zijlstra 已提交
269

270 271 272
	if (avail_dirty < dirty)
		avail_dirty = dirty - avail_dirty;
	else
P
Peter Zijlstra 已提交
273 274 275 276 277 278 279 280
		avail_dirty = 0;

	avail_dirty += bdi_stat(bdi, BDI_RECLAIMABLE) +
		bdi_stat(bdi, BDI_WRITEBACK);

	*pbdi_dirty = min(*pbdi_dirty, avail_dirty);
}

P
Peter Zijlstra 已提交
281 282 283 284 285 286 287 288 289 290 291 292 293 294
static inline void task_dirties_fraction(struct task_struct *tsk,
		long *numerator, long *denominator)
{
	prop_fraction_single(&vm_dirties, &tsk->dirties,
				numerator, denominator);
}

/*
 * scale the dirty limit
 *
 * task specific dirty limit:
 *
 *   dirty -= (dirty/8) * p_{t}
 */
295
static void task_dirty_limit(struct task_struct *tsk, unsigned long *pdirty)
P
Peter Zijlstra 已提交
296 297
{
	long numerator, denominator;
298
	unsigned long dirty = *pdirty;
P
Peter Zijlstra 已提交
299 300 301 302 303 304 305 306 307 308 309 310 311
	u64 inv = dirty >> 3;

	task_dirties_fraction(tsk, &numerator, &denominator);
	inv *= numerator;
	do_div(inv, denominator);

	dirty -= inv;
	if (dirty < *pdirty/2)
		dirty = *pdirty/2;

	*pdirty = dirty;
}

312 313 314 315 316 317 318 319 320
/*
 *
 */
static unsigned int bdi_min_ratio;

int bdi_set_min_ratio(struct backing_dev_info *bdi, unsigned int min_ratio)
{
	int ret = 0;

321
	spin_lock_bh(&bdi_lock);
322
	if (min_ratio > bdi->max_ratio) {
323
		ret = -EINVAL;
324 325 326 327 328 329 330 331 332
	} else {
		min_ratio -= bdi->min_ratio;
		if (bdi_min_ratio + min_ratio < 100) {
			bdi_min_ratio += min_ratio;
			bdi->min_ratio += min_ratio;
		} else {
			ret = -EINVAL;
		}
	}
333
	spin_unlock_bh(&bdi_lock);
334 335 336 337 338 339 340 341 342 343 344

	return ret;
}

int bdi_set_max_ratio(struct backing_dev_info *bdi, unsigned max_ratio)
{
	int ret = 0;

	if (max_ratio > 100)
		return -EINVAL;

345
	spin_lock_bh(&bdi_lock);
346 347 348 349 350 351
	if (bdi->min_ratio > max_ratio) {
		ret = -EINVAL;
	} else {
		bdi->max_ratio = max_ratio;
		bdi->max_prop_frac = (PROP_FRAC_BASE * max_ratio) / 100;
	}
352
	spin_unlock_bh(&bdi_lock);
353 354 355

	return ret;
}
356
EXPORT_SYMBOL(bdi_set_max_ratio);
357

L
Linus Torvalds 已提交
358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374
/*
 * Work out the current dirty-memory clamping and background writeout
 * thresholds.
 *
 * The main aim here is to lower them aggressively if there is a lot of mapped
 * memory around.  To avoid stressing page reclaim with lots of unreclaimable
 * pages.  It is better to clamp down on writers than to start swapping, and
 * performing lots of scanning.
 *
 * We only allow 1/2 of the currently-unmapped memory to be dirtied.
 *
 * We don't permit the clamping level to fall below 5% - that is getting rather
 * excessive.
 *
 * We make sure that the background writeout level is below the adjusted
 * clamping level.
 */
375 376 377 378 379 380 381

static unsigned long highmem_dirtyable_memory(unsigned long total)
{
#ifdef CONFIG_HIGHMEM
	int node;
	unsigned long x = 0;

382
	for_each_node_state(node, N_HIGH_MEMORY) {
383 384 385
		struct zone *z =
			&NODE_DATA(node)->node_zones[ZONE_HIGHMEM];

386 387
		x += zone_page_state(z, NR_FREE_PAGES) +
		     zone_reclaimable_pages(z);
388 389 390 391 392 393 394 395 396 397 398 399 400
	}
	/*
	 * Make sure that the number of highmem pages is never larger
	 * than the number of the total dirtyable memory. This can only
	 * occur in very strange VM situations but we want to make sure
	 * that this does not occur.
	 */
	return min(x, total);
#else
	return 0;
#endif
}

S
Steven Rostedt 已提交
401 402 403 404 405 406 407
/**
 * determine_dirtyable_memory - amount of memory that may be used
 *
 * Returns the numebr of pages that can currently be freed and used
 * by the kernel for direct mappings.
 */
unsigned long determine_dirtyable_memory(void)
408 409 410
{
	unsigned long x;

411
	x = global_page_state(NR_FREE_PAGES) + global_reclaimable_pages();
412 413 414 415

	if (!vm_highmem_is_dirtyable)
		x -= highmem_dirtyable_memory(x);

416 417 418
	return x + 1;	/* Ensure that we never return 0 */
}

419
void
420 421
get_dirty_limits(unsigned long *pbackground, unsigned long *pdirty,
		 unsigned long *pbdi_dirty, struct backing_dev_info *bdi)
L
Linus Torvalds 已提交
422
{
423 424
	unsigned long background;
	unsigned long dirty;
425
	unsigned long available_memory = determine_dirtyable_memory();
L
Linus Torvalds 已提交
426 427
	struct task_struct *tsk;

428 429 430 431 432 433 434 435 436 437
	if (vm_dirty_bytes)
		dirty = DIV_ROUND_UP(vm_dirty_bytes, PAGE_SIZE);
	else {
		int dirty_ratio;

		dirty_ratio = vm_dirty_ratio;
		if (dirty_ratio < 5)
			dirty_ratio = 5;
		dirty = (dirty_ratio * available_memory) / 100;
	}
L
Linus Torvalds 已提交
438

439 440 441 442
	if (dirty_background_bytes)
		background = DIV_ROUND_UP(dirty_background_bytes, PAGE_SIZE);
	else
		background = (dirty_background_ratio * available_memory) / 100;
L
Linus Torvalds 已提交
443

444 445
	if (background >= dirty)
		background = dirty / 2;
L
Linus Torvalds 已提交
446 447 448 449 450 451 452
	tsk = current;
	if (tsk->flags & PF_LESS_THROTTLE || rt_task(tsk)) {
		background += background / 4;
		dirty += dirty / 4;
	}
	*pbackground = background;
	*pdirty = dirty;
P
Peter Zijlstra 已提交
453 454

	if (bdi) {
455
		u64 bdi_dirty;
P
Peter Zijlstra 已提交
456 457 458 459 460 461 462
		long numerator, denominator;

		/*
		 * Calculate this BDI's share of the dirty ratio.
		 */
		bdi_writeout_fraction(bdi, &numerator, &denominator);

463
		bdi_dirty = (dirty * (100 - bdi_min_ratio)) / 100;
P
Peter Zijlstra 已提交
464 465
		bdi_dirty *= numerator;
		do_div(bdi_dirty, denominator);
466
		bdi_dirty += (dirty * bdi->min_ratio) / 100;
467 468
		if (bdi_dirty > (dirty * bdi->max_ratio) / 100)
			bdi_dirty = dirty * bdi->max_ratio / 100;
P
Peter Zijlstra 已提交
469 470 471

		*pbdi_dirty = bdi_dirty;
		clip_bdi_dirty_limit(bdi, dirty, pbdi_dirty);
P
Peter Zijlstra 已提交
472
		task_dirty_limit(current, pbdi_dirty);
P
Peter Zijlstra 已提交
473
	}
L
Linus Torvalds 已提交
474 475 476 477 478 479 480 481 482
}

/*
 * balance_dirty_pages() must be called by processes which are generating dirty
 * data.  It looks at the number of dirty pages in the machine and will force
 * the caller to perform writeback if the system is over `vm_dirty_ratio'.
 * If we're over `background_thresh' then pdflush is woken to perform some
 * writeout.
 */
483 484
static void balance_dirty_pages(struct address_space *mapping,
				unsigned long write_chunk)
L
Linus Torvalds 已提交
485
{
486 487
	long nr_reclaimable, bdi_nr_reclaimable;
	long nr_writeback, bdi_nr_writeback;
488 489 490
	unsigned long background_thresh;
	unsigned long dirty_thresh;
	unsigned long bdi_thresh;
L
Linus Torvalds 已提交
491
	unsigned long pages_written = 0;
492
	unsigned long pause = 1;
L
Linus Torvalds 已提交
493 494 495 496 497 498 499 500 501

	struct backing_dev_info *bdi = mapping->backing_dev_info;

	for (;;) {
		struct writeback_control wbc = {
			.bdi		= bdi,
			.sync_mode	= WB_SYNC_NONE,
			.older_than_this = NULL,
			.nr_to_write	= write_chunk,
502
			.range_cyclic	= 1,
L
Linus Torvalds 已提交
503 504
		};

P
Peter Zijlstra 已提交
505 506
		get_dirty_limits(&background_thresh, &dirty_thresh,
				&bdi_thresh, bdi);
507 508 509 510 511

		nr_reclaimable = global_page_state(NR_FILE_DIRTY) +
					global_page_state(NR_UNSTABLE_NFS);
		nr_writeback = global_page_state(NR_WRITEBACK);

P
Peter Zijlstra 已提交
512 513
		bdi_nr_reclaimable = bdi_stat(bdi, BDI_RECLAIMABLE);
		bdi_nr_writeback = bdi_stat(bdi, BDI_WRITEBACK);
514

P
Peter Zijlstra 已提交
515 516
		if (bdi_nr_reclaimable + bdi_nr_writeback <= bdi_thresh)
			break;
L
Linus Torvalds 已提交
517

518 519 520 521 522 523 524 525 526
		/*
		 * Throttle it only when the background writeback cannot
		 * catch-up. This avoids (excessively) small writeouts
		 * when the bdi limits are ramping up.
		 */
		if (nr_reclaimable + nr_writeback <
				(background_thresh + dirty_thresh) / 2)
			break;

P
Peter Zijlstra 已提交
527 528
		if (!bdi->dirty_exceeded)
			bdi->dirty_exceeded = 1;
L
Linus Torvalds 已提交
529 530 531 532 533 534

		/* Note: nr_reclaimable denotes nr_dirty + nr_unstable.
		 * Unstable writes are a feature of certain networked
		 * filesystems (i.e. NFS) in which data may have been
		 * written to the server's write cache, but has not yet
		 * been flushed to permanent storage.
535 536 537
		 * Only move pages to writeback if this bdi is over its
		 * threshold otherwise wait until the disk writes catch
		 * up.
L
Linus Torvalds 已提交
538
		 */
539
		if (bdi_nr_reclaimable > bdi_thresh) {
540
			writeback_inodes_wbc(&wbc);
L
Linus Torvalds 已提交
541
			pages_written += write_chunk - wbc.nr_to_write;
P
Peter Zijlstra 已提交
542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561
			get_dirty_limits(&background_thresh, &dirty_thresh,
				       &bdi_thresh, bdi);
		}

		/*
		 * In order to avoid the stacked BDI deadlock we need
		 * to ensure we accurately count the 'dirty' pages when
		 * the threshold is low.
		 *
		 * Otherwise it would be possible to get thresh+n pages
		 * reported dirty, even though there are thresh-m pages
		 * actually dirty; with m+n sitting in the percpu
		 * deltas.
		 */
		if (bdi_thresh < 2*bdi_stat_error(bdi)) {
			bdi_nr_reclaimable = bdi_stat_sum(bdi, BDI_RECLAIMABLE);
			bdi_nr_writeback = bdi_stat_sum(bdi, BDI_WRITEBACK);
		} else if (bdi_nr_reclaimable) {
			bdi_nr_reclaimable = bdi_stat(bdi, BDI_RECLAIMABLE);
			bdi_nr_writeback = bdi_stat(bdi, BDI_WRITEBACK);
L
Linus Torvalds 已提交
562
		}
P
Peter Zijlstra 已提交
563 564 565 566 567 568

		if (bdi_nr_reclaimable + bdi_nr_writeback <= bdi_thresh)
			break;
		if (pages_written >= write_chunk)
			break;		/* We've done our duty */

569 570 571 572 573 574 575 576 577
		schedule_timeout_interruptible(pause);

		/*
		 * Increase the delay for each loop, up to our previous
		 * default of taking a 100ms nap.
		 */
		pause <<= 1;
		if (pause > HZ / 10)
			pause = HZ / 10;
L
Linus Torvalds 已提交
578 579
	}

P
Peter Zijlstra 已提交
580 581 582
	if (bdi_nr_reclaimable + bdi_nr_writeback < bdi_thresh &&
			bdi->dirty_exceeded)
		bdi->dirty_exceeded = 0;
L
Linus Torvalds 已提交
583 584 585 586 587 588 589 590 591 592 593 594 595

	if (writeback_in_progress(bdi))
		return;		/* pdflush is already working this queue */

	/*
	 * In laptop mode, we wait until hitting the higher threshold before
	 * starting background writeout, and then write out all the way down
	 * to the lower threshold.  So slow writers cause minimal disk activity.
	 *
	 * In normal mode, we start background writeout at the lower
	 * background_thresh, to keep the amount of dirty memory low.
	 */
	if ((laptop_mode && pages_written) ||
596 597
	    (!laptop_mode && ((nr_writeback = global_page_state(NR_FILE_DIRTY)
					  + global_page_state(NR_UNSTABLE_NFS))
598 599
					  > background_thresh)))
		bdi_start_writeback(bdi, nr_writeback);
L
Linus Torvalds 已提交
600 601
}

602
void set_page_dirty_balance(struct page *page, int page_mkwrite)
P
Peter Zijlstra 已提交
603
{
604
	if (set_page_dirty(page) || page_mkwrite) {
P
Peter Zijlstra 已提交
605 606 607 608 609 610 611
		struct address_space *mapping = page_mapping(page);

		if (mapping)
			balance_dirty_pages_ratelimited(mapping);
	}
}

612 613
static DEFINE_PER_CPU(unsigned long, bdp_ratelimits) = 0;

L
Linus Torvalds 已提交
614
/**
615
 * balance_dirty_pages_ratelimited_nr - balance dirty memory state
616
 * @mapping: address_space which was dirtied
617
 * @nr_pages_dirtied: number of pages which the caller has just dirtied
L
Linus Torvalds 已提交
618 619 620 621 622 623 624 625 626 627
 *
 * Processes which are dirtying memory should call in here once for each page
 * which was newly dirtied.  The function will periodically check the system's
 * dirty state and will initiate writeback if needed.
 *
 * On really big machines, get_writeback_state is expensive, so try to avoid
 * calling it too often (ratelimiting).  But once we're over the dirty memory
 * limit we decrease the ratelimiting by a lot, to prevent individual processes
 * from overshooting the limit by (ratelimit_pages) each.
 */
628 629
void balance_dirty_pages_ratelimited_nr(struct address_space *mapping,
					unsigned long nr_pages_dirtied)
L
Linus Torvalds 已提交
630
{
631 632
	unsigned long ratelimit;
	unsigned long *p;
L
Linus Torvalds 已提交
633 634

	ratelimit = ratelimit_pages;
P
Peter Zijlstra 已提交
635
	if (mapping->backing_dev_info->dirty_exceeded)
L
Linus Torvalds 已提交
636 637 638 639 640 641
		ratelimit = 8;

	/*
	 * Check the rate limiting. Also, we do not want to throttle real-time
	 * tasks in balance_dirty_pages(). Period.
	 */
642
	preempt_disable();
643
	p =  &__get_cpu_var(bdp_ratelimits);
644 645
	*p += nr_pages_dirtied;
	if (unlikely(*p >= ratelimit)) {
646
		ratelimit = sync_writeback_pages(*p);
647 648
		*p = 0;
		preempt_enable();
649
		balance_dirty_pages(mapping, ratelimit);
L
Linus Torvalds 已提交
650 651
		return;
	}
652
	preempt_enable();
L
Linus Torvalds 已提交
653
}
654
EXPORT_SYMBOL(balance_dirty_pages_ratelimited_nr);
L
Linus Torvalds 已提交
655

656
void throttle_vm_writeout(gfp_t gfp_mask)
L
Linus Torvalds 已提交
657
{
658 659
	unsigned long background_thresh;
	unsigned long dirty_thresh;
L
Linus Torvalds 已提交
660 661

        for ( ; ; ) {
P
Peter Zijlstra 已提交
662
		get_dirty_limits(&background_thresh, &dirty_thresh, NULL, NULL);
L
Linus Torvalds 已提交
663 664 665 666 667 668 669

                /*
                 * Boost the allowable dirty threshold a bit for page
                 * allocators so they don't get DoS'ed by heavy writers
                 */
                dirty_thresh += dirty_thresh / 10;      /* wheeee... */

670 671 672
                if (global_page_state(NR_UNSTABLE_NFS) +
			global_page_state(NR_WRITEBACK) <= dirty_thresh)
                        	break;
673
                congestion_wait(BLK_RW_ASYNC, HZ/10);
674 675 676 677 678 679 680 681

		/*
		 * The caller might hold locks which can prevent IO completion
		 * or progress in the filesystem.  So we cannot just sit here
		 * waiting for IO to complete.
		 */
		if ((gfp_mask & (__GFP_FS|__GFP_IO)) != (__GFP_FS|__GFP_IO))
			break;
L
Linus Torvalds 已提交
682 683 684 685 686
        }
}

static void laptop_timer_fn(unsigned long unused);

687
static DEFINE_TIMER(laptop_mode_wb_timer, laptop_timer_fn, 0, 0);
L
Linus Torvalds 已提交
688 689 690 691 692

/*
 * sysctl handler for /proc/sys/vm/dirty_writeback_centisecs
 */
int dirty_writeback_centisecs_handler(ctl_table *table, int write,
693
	struct file *file, void __user *buffer, size_t *length, loff_t *ppos)
L
Linus Torvalds 已提交
694
{
695
	proc_dointvec(table, write, file, buffer, length, ppos);
L
Linus Torvalds 已提交
696 697 698
	return 0;
}

699
static void do_laptop_sync(struct work_struct *work)
L
Linus Torvalds 已提交
700
{
701 702
	wakeup_flusher_threads(0);
	kfree(work);
L
Linus Torvalds 已提交
703 704 705 706
}

static void laptop_timer_fn(unsigned long unused)
{
707 708 709 710 711 712 713
	struct work_struct *work;

	work = kmalloc(sizeof(*work), GFP_ATOMIC);
	if (work) {
		INIT_WORK(work, do_laptop_sync);
		schedule_work(work);
	}
L
Linus Torvalds 已提交
714 715 716 717 718 719 720 721 722
}

/*
 * We've spun up the disk and we're in laptop mode: schedule writeback
 * of all dirty data a few seconds from now.  If the flush is already scheduled
 * then push it back - the user is still using the disk.
 */
void laptop_io_completion(void)
{
723
	mod_timer(&laptop_mode_wb_timer, jiffies + laptop_mode);
L
Linus Torvalds 已提交
724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752
}

/*
 * We're in laptop mode and we've just synced. The sync's writes will have
 * caused another writeback to be scheduled by laptop_io_completion.
 * Nothing needs to be written back anymore, so we unschedule the writeback.
 */
void laptop_sync_completion(void)
{
	del_timer(&laptop_mode_wb_timer);
}

/*
 * If ratelimit_pages is too high then we can get into dirty-data overload
 * if a large number of processes all perform writes at the same time.
 * If it is too low then SMP machines will call the (expensive)
 * get_writeback_state too often.
 *
 * Here we set ratelimit_pages to a level which ensures that when all CPUs are
 * dirtying in parallel, we cannot go more than 3% (1/32) over the dirty memory
 * thresholds before writeback cuts in.
 *
 * But the limit should not be set too high.  Because it also controls the
 * amount of memory which the balance_dirty_pages() caller has to write back.
 * If this is too large then the caller will block on the IO queue all the
 * time.  So limit it to four megabytes - the balance_dirty_pages() caller
 * will write six megabyte chunks, max.
 */

753
void writeback_set_ratelimit(void)
L
Linus Torvalds 已提交
754
{
755
	ratelimit_pages = vm_total_pages / (num_online_cpus() * 32);
L
Linus Torvalds 已提交
756 757 758 759 760 761
	if (ratelimit_pages < 16)
		ratelimit_pages = 16;
	if (ratelimit_pages * PAGE_CACHE_SIZE > 4096 * 1024)
		ratelimit_pages = (4096 * 1024) / PAGE_CACHE_SIZE;
}

762
static int __cpuinit
L
Linus Torvalds 已提交
763 764
ratelimit_handler(struct notifier_block *self, unsigned long u, void *v)
{
765
	writeback_set_ratelimit();
766
	return NOTIFY_DONE;
L
Linus Torvalds 已提交
767 768
}

769
static struct notifier_block __cpuinitdata ratelimit_nb = {
L
Linus Torvalds 已提交
770 771 772 773 774
	.notifier_call	= ratelimit_handler,
	.next		= NULL,
};

/*
775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790
 * Called early on to tune the page writeback dirty limits.
 *
 * We used to scale dirty pages according to how total memory
 * related to pages that could be allocated for buffers (by
 * comparing nr_free_buffer_pages() to vm_total_pages.
 *
 * However, that was when we used "dirty_ratio" to scale with
 * all memory, and we don't do that any more. "dirty_ratio"
 * is now applied to total non-HIGHPAGE memory (by subtracting
 * totalhigh_pages from vm_total_pages), and as such we can't
 * get into the old insane situation any more where we had
 * large amounts of dirty pages compared to a small amount of
 * non-HIGHMEM memory.
 *
 * But we might still want to scale the dirty_ratio by how
 * much memory the box has..
L
Linus Torvalds 已提交
791 792 793
 */
void __init page_writeback_init(void)
{
P
Peter Zijlstra 已提交
794 795
	int shift;

796
	writeback_set_ratelimit();
L
Linus Torvalds 已提交
797
	register_cpu_notifier(&ratelimit_nb);
P
Peter Zijlstra 已提交
798 799 800

	shift = calc_period_shift();
	prop_descriptor_init(&vm_completions, shift);
P
Peter Zijlstra 已提交
801
	prop_descriptor_init(&vm_dirties, shift);
L
Linus Torvalds 已提交
802 803
}

804
/**
805
 * write_cache_pages - walk the list of dirty pages of the given address space and write all of them.
806 807
 * @mapping: address space structure to write
 * @wbc: subtract the number of written pages from *@wbc->nr_to_write
808 809
 * @writepage: function called for each page
 * @data: data passed to writepage function
810
 *
811
 * If a page is already under I/O, write_cache_pages() skips it, even
812 813 814 815 816 817 818
 * if it's dirty.  This is desirable behaviour for memory-cleaning writeback,
 * but it is INCORRECT for data-integrity system calls such as fsync().  fsync()
 * and msync() need to guarantee that all the data which was dirty at the time
 * the call was made get new I/O started against them.  If wbc->sync_mode is
 * WB_SYNC_ALL then we were called for data integrity and we must wait for
 * existing IO to complete.
 */
819 820 821
int write_cache_pages(struct address_space *mapping,
		      struct writeback_control *wbc, writepage_t writepage,
		      void *data)
822 823 824 825 826 827
{
	struct backing_dev_info *bdi = mapping->backing_dev_info;
	int ret = 0;
	int done = 0;
	struct pagevec pvec;
	int nr_pages;
N
Nick Piggin 已提交
828
	pgoff_t uninitialized_var(writeback_index);
829 830
	pgoff_t index;
	pgoff_t end;		/* Inclusive */
831
	pgoff_t done_index;
N
Nick Piggin 已提交
832
	int cycled;
833
	int range_whole = 0;
834
	long nr_to_write = wbc->nr_to_write;
835 836 837 838 839 840 841 842

	if (wbc->nonblocking && bdi_write_congested(bdi)) {
		wbc->encountered_congestion = 1;
		return 0;
	}

	pagevec_init(&pvec, 0);
	if (wbc->range_cyclic) {
N
Nick Piggin 已提交
843 844 845 846 847 848
		writeback_index = mapping->writeback_index; /* prev offset */
		index = writeback_index;
		if (index == 0)
			cycled = 1;
		else
			cycled = 0;
849 850 851 852 853 854
		end = -1;
	} else {
		index = wbc->range_start >> PAGE_CACHE_SHIFT;
		end = wbc->range_end >> PAGE_CACHE_SHIFT;
		if (wbc->range_start == 0 && wbc->range_end == LLONG_MAX)
			range_whole = 1;
N
Nick Piggin 已提交
855
		cycled = 1; /* ignore range_cyclic tests */
856 857
	}
retry:
858
	done_index = index;
N
Nick Piggin 已提交
859 860 861 862 863 864 865 866
	while (!done && (index <= end)) {
		int i;

		nr_pages = pagevec_lookup_tag(&pvec, mapping, &index,
			      PAGECACHE_TAG_DIRTY,
			      min(end - index, (pgoff_t)PAGEVEC_SIZE-1) + 1);
		if (nr_pages == 0)
			break;
867 868 869 870 871

		for (i = 0; i < nr_pages; i++) {
			struct page *page = pvec.pages[i];

			/*
872 873 874 875 876
			 * At this point, the page may be truncated or
			 * invalidated (changing page->mapping to NULL), or
			 * even swizzled back from swapper_space to tmpfs file
			 * mapping. However, page->index will not change
			 * because we have a reference on the page.
877
			 */
878 879 880 881 882 883 884 885 886 887 888
			if (page->index > end) {
				/*
				 * can't be range_cyclic (1st pass) because
				 * end == -1 in that case.
				 */
				done = 1;
				break;
			}

			done_index = page->index + 1;

889 890
			lock_page(page);

N
Nick Piggin 已提交
891 892 893 894 895 896 897 898
			/*
			 * Page truncated or invalidated. We can freely skip it
			 * then, even for data integrity operations: the page
			 * has disappeared concurrently, so there could be no
			 * real expectation of this data interity operation
			 * even if there is now a new, dirty page at the same
			 * pagecache address.
			 */
899
			if (unlikely(page->mapping != mapping)) {
N
Nick Piggin 已提交
900
continue_unlock:
901 902 903 904
				unlock_page(page);
				continue;
			}

905 906 907 908 909 910 911 912 913 914 915
			if (!PageDirty(page)) {
				/* someone wrote it for us */
				goto continue_unlock;
			}

			if (PageWriteback(page)) {
				if (wbc->sync_mode != WB_SYNC_NONE)
					wait_on_page_writeback(page);
				else
					goto continue_unlock;
			}
916

917 918
			BUG_ON(PageWriteback(page));
			if (!clear_page_dirty_for_io(page))
N
Nick Piggin 已提交
919
				goto continue_unlock;
920

921
			ret = (*writepage)(page, wbc, data);
922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940
			if (unlikely(ret)) {
				if (ret == AOP_WRITEPAGE_ACTIVATE) {
					unlock_page(page);
					ret = 0;
				} else {
					/*
					 * done_index is set past this page,
					 * so media errors will not choke
					 * background writeout for the entire
					 * file. This has consequences for
					 * range_cyclic semantics (ie. it may
					 * not be suitable for data integrity
					 * writeout).
					 */
					done = 1;
					break;
				}
 			}

F
Federico Cuello 已提交
941
			if (nr_to_write > 0) {
942
				nr_to_write--;
F
Federico Cuello 已提交
943 944 945 946 947 948 949 950 951 952 953 954 955 956 957
				if (nr_to_write == 0 &&
				    wbc->sync_mode == WB_SYNC_NONE) {
					/*
					 * We stop writing back only if we are
					 * not doing integrity sync. In case of
					 * integrity sync we have to keep going
					 * because someone may be concurrently
					 * dirtying pages, and we might have
					 * synced a lot of newly appeared dirty
					 * pages, but have not synced all of the
					 * old dirty pages.
					 */
					done = 1;
					break;
				}
958
			}
959

960 961 962
			if (wbc->nonblocking && bdi_write_congested(bdi)) {
				wbc->encountered_congestion = 1;
				done = 1;
963
				break;
964 965 966 967 968
			}
		}
		pagevec_release(&pvec);
		cond_resched();
	}
969
	if (!cycled && !done) {
970
		/*
N
Nick Piggin 已提交
971
		 * range_cyclic:
972 973 974
		 * We hit the last page and there is more work to be done: wrap
		 * back to the start of the file
		 */
N
Nick Piggin 已提交
975
		cycled = 1;
976
		index = 0;
N
Nick Piggin 已提交
977
		end = writeback_index - 1;
978 979
		goto retry;
	}
980 981
	if (!wbc->no_nrwrite_index_update) {
		if (wbc->range_cyclic || (range_whole && nr_to_write > 0))
982
			mapping->writeback_index = done_index;
983 984
		wbc->nr_to_write = nr_to_write;
	}
985

986 987
	return ret;
}
988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019
EXPORT_SYMBOL(write_cache_pages);

/*
 * Function used by generic_writepages to call the real writepage
 * function and set the mapping flags on error
 */
static int __writepage(struct page *page, struct writeback_control *wbc,
		       void *data)
{
	struct address_space *mapping = data;
	int ret = mapping->a_ops->writepage(page, wbc);
	mapping_set_error(mapping, ret);
	return ret;
}

/**
 * generic_writepages - walk the list of dirty pages of the given address space and writepage() all of them.
 * @mapping: address space structure to write
 * @wbc: subtract the number of written pages from *@wbc->nr_to_write
 *
 * This is a library function, which implements the writepages()
 * address_space_operation.
 */
int generic_writepages(struct address_space *mapping,
		       struct writeback_control *wbc)
{
	/* deal with chardevs and other special file */
	if (!mapping->a_ops->writepage)
		return 0;

	return write_cache_pages(mapping, wbc, __writepage, mapping);
}
1020 1021 1022

EXPORT_SYMBOL(generic_writepages);

L
Linus Torvalds 已提交
1023 1024
int do_writepages(struct address_space *mapping, struct writeback_control *wbc)
{
1025 1026
	int ret;

L
Linus Torvalds 已提交
1027 1028 1029
	if (wbc->nr_to_write <= 0)
		return 0;
	if (mapping->a_ops->writepages)
1030
		ret = mapping->a_ops->writepages(mapping, wbc);
1031 1032 1033
	else
		ret = generic_writepages(mapping, wbc);
	return ret;
L
Linus Torvalds 已提交
1034 1035 1036 1037
}

/**
 * write_one_page - write out a single page and optionally wait on I/O
1038 1039
 * @page: the page to write
 * @wait: if true, wait on writeout
L
Linus Torvalds 已提交
1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074
 *
 * The page must be locked by the caller and will be unlocked upon return.
 *
 * write_one_page() returns a negative error code if I/O failed.
 */
int write_one_page(struct page *page, int wait)
{
	struct address_space *mapping = page->mapping;
	int ret = 0;
	struct writeback_control wbc = {
		.sync_mode = WB_SYNC_ALL,
		.nr_to_write = 1,
	};

	BUG_ON(!PageLocked(page));

	if (wait)
		wait_on_page_writeback(page);

	if (clear_page_dirty_for_io(page)) {
		page_cache_get(page);
		ret = mapping->a_ops->writepage(page, &wbc);
		if (ret == 0 && wait) {
			wait_on_page_writeback(page);
			if (PageError(page))
				ret = -EIO;
		}
		page_cache_release(page);
	} else {
		unlock_page(page);
	}
	return ret;
}
EXPORT_SYMBOL(write_one_page);

1075 1076 1077 1078 1079 1080 1081 1082 1083 1084
/*
 * For address_spaces which do not use buffers nor write back.
 */
int __set_page_dirty_no_writeback(struct page *page)
{
	if (!PageDirty(page))
		SetPageDirty(page);
	return 0;
}

1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098
/*
 * Helper function for set_page_dirty family.
 * NOTE: This relies on being atomic wrt interrupts.
 */
void account_page_dirtied(struct page *page, struct address_space *mapping)
{
	if (mapping_cap_account_dirty(mapping)) {
		__inc_zone_page_state(page, NR_FILE_DIRTY);
		__inc_bdi_stat(mapping->backing_dev_info, BDI_RECLAIMABLE);
		task_dirty_inc(current);
		task_io_account_write(PAGE_CACHE_SIZE);
	}
}

L
Linus Torvalds 已提交
1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111
/*
 * For address_spaces which do not use buffers.  Just tag the page as dirty in
 * its radix tree.
 *
 * This is also used when a single buffer is being dirtied: we want to set the
 * page dirty in that case, but not all the buffers.  This is a "bottom-up"
 * dirtying, whereas __set_page_dirty_buffers() is a "top-down" dirtying.
 *
 * Most callers have locked the page, which pins the address_space in memory.
 * But zap_pte_range() does not lock the page, however in that case the
 * mapping is pinned by the vma's ->vm_file reference.
 *
 * We take care to handle the case where the page was truncated from the
S
Simon Arlott 已提交
1112
 * mapping by re-checking page_mapping() inside tree_lock.
L
Linus Torvalds 已提交
1113 1114 1115 1116 1117 1118 1119
 */
int __set_page_dirty_nobuffers(struct page *page)
{
	if (!TestSetPageDirty(page)) {
		struct address_space *mapping = page_mapping(page);
		struct address_space *mapping2;

1120 1121 1122
		if (!mapping)
			return 1;

N
Nick Piggin 已提交
1123
		spin_lock_irq(&mapping->tree_lock);
1124 1125 1126
		mapping2 = page_mapping(page);
		if (mapping2) { /* Race with truncate? */
			BUG_ON(mapping2 != mapping);
1127
			WARN_ON_ONCE(!PagePrivate(page) && !PageUptodate(page));
1128
			account_page_dirtied(page, mapping);
1129 1130 1131
			radix_tree_tag_set(&mapping->page_tree,
				page_index(page), PAGECACHE_TAG_DIRTY);
		}
N
Nick Piggin 已提交
1132
		spin_unlock_irq(&mapping->tree_lock);
1133 1134 1135
		if (mapping->host) {
			/* !PageAnon && !swapper_space */
			__mark_inode_dirty(mapping->host, I_DIRTY_PAGES);
L
Linus Torvalds 已提交
1136
		}
1137
		return 1;
L
Linus Torvalds 已提交
1138
	}
1139
	return 0;
L
Linus Torvalds 已提交
1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158
}
EXPORT_SYMBOL(__set_page_dirty_nobuffers);

/*
 * When a writepage implementation decides that it doesn't want to write this
 * page for some reason, it should redirty the locked page via
 * redirty_page_for_writepage() and it should then unlock the page and return 0
 */
int redirty_page_for_writepage(struct writeback_control *wbc, struct page *page)
{
	wbc->pages_skipped++;
	return __set_page_dirty_nobuffers(page);
}
EXPORT_SYMBOL(redirty_page_for_writepage);

/*
 * If the mapping doesn't provide a set_page_dirty a_op, then
 * just fall through and assume that it wants buffer_heads.
 */
N
Nick Piggin 已提交
1159
int set_page_dirty(struct page *page)
L
Linus Torvalds 已提交
1160 1161 1162 1163 1164
{
	struct address_space *mapping = page_mapping(page);

	if (likely(mapping)) {
		int (*spd)(struct page *) = mapping->a_ops->set_page_dirty;
1165 1166 1167 1168 1169
#ifdef CONFIG_BLOCK
		if (!spd)
			spd = __set_page_dirty_buffers;
#endif
		return (*spd)(page);
L
Linus Torvalds 已提交
1170
	}
1171 1172 1173 1174
	if (!PageDirty(page)) {
		if (!TestSetPageDirty(page))
			return 1;
	}
L
Linus Torvalds 已提交
1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192
	return 0;
}
EXPORT_SYMBOL(set_page_dirty);

/*
 * set_page_dirty() is racy if the caller has no reference against
 * page->mapping->host, and if the page is unlocked.  This is because another
 * CPU could truncate the page off the mapping and then free the mapping.
 *
 * Usually, the page _is_ locked, or the caller is a user-space process which
 * holds a reference on the inode by having an open file.
 *
 * In other cases, the page should be locked before running set_page_dirty().
 */
int set_page_dirty_lock(struct page *page)
{
	int ret;

1193
	lock_page_nosync(page);
L
Linus Torvalds 已提交
1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217
	ret = set_page_dirty(page);
	unlock_page(page);
	return ret;
}
EXPORT_SYMBOL(set_page_dirty_lock);

/*
 * Clear a page's dirty flag, while caring for dirty memory accounting.
 * Returns true if the page was previously dirty.
 *
 * This is for preparing to put the page under writeout.  We leave the page
 * tagged as dirty in the radix tree so that a concurrent write-for-sync
 * can discover it via a PAGECACHE_TAG_DIRTY walk.  The ->writepage
 * implementation will run either set_page_writeback() or set_page_dirty(),
 * at which stage we bring the page's dirty flag and radix-tree dirty tag
 * back into sync.
 *
 * This incoherency between the page's dirty flag and radix-tree tag is
 * unfortunate, but it only exists while the page is locked.
 */
int clear_page_dirty_for_io(struct page *page)
{
	struct address_space *mapping = page_mapping(page);

1218 1219
	BUG_ON(!PageLocked(page));

1220
	ClearPageReclaim(page);
1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248
	if (mapping && mapping_cap_account_dirty(mapping)) {
		/*
		 * Yes, Virginia, this is indeed insane.
		 *
		 * We use this sequence to make sure that
		 *  (a) we account for dirty stats properly
		 *  (b) we tell the low-level filesystem to
		 *      mark the whole page dirty if it was
		 *      dirty in a pagetable. Only to then
		 *  (c) clean the page again and return 1 to
		 *      cause the writeback.
		 *
		 * This way we avoid all nasty races with the
		 * dirty bit in multiple places and clearing
		 * them concurrently from different threads.
		 *
		 * Note! Normally the "set_page_dirty(page)"
		 * has no effect on the actual dirty bit - since
		 * that will already usually be set. But we
		 * need the side effects, and it can help us
		 * avoid races.
		 *
		 * We basically use the page "master dirty bit"
		 * as a serialization point for all the different
		 * threads doing their things.
		 */
		if (page_mkclean(page))
			set_page_dirty(page);
1249 1250 1251 1252 1253 1254 1255 1256 1257 1258
		/*
		 * We carefully synchronise fault handlers against
		 * installing a dirty pte and marking the page dirty
		 * at this point. We do this by having them hold the
		 * page lock at some point after installing their
		 * pte, but before marking the page dirty.
		 * Pages are always locked coming in here, so we get
		 * the desired exclusion. See mm/memory.c:do_wp_page()
		 * for more comments.
		 */
1259
		if (TestClearPageDirty(page)) {
1260
			dec_zone_page_state(page, NR_FILE_DIRTY);
1261 1262
			dec_bdi_stat(mapping->backing_dev_info,
					BDI_RECLAIMABLE);
1263
			return 1;
L
Linus Torvalds 已提交
1264
		}
1265
		return 0;
L
Linus Torvalds 已提交
1266
	}
1267
	return TestClearPageDirty(page);
L
Linus Torvalds 已提交
1268
}
1269
EXPORT_SYMBOL(clear_page_dirty_for_io);
L
Linus Torvalds 已提交
1270 1271 1272 1273 1274 1275 1276

int test_clear_page_writeback(struct page *page)
{
	struct address_space *mapping = page_mapping(page);
	int ret;

	if (mapping) {
P
Peter Zijlstra 已提交
1277
		struct backing_dev_info *bdi = mapping->backing_dev_info;
L
Linus Torvalds 已提交
1278 1279
		unsigned long flags;

N
Nick Piggin 已提交
1280
		spin_lock_irqsave(&mapping->tree_lock, flags);
L
Linus Torvalds 已提交
1281
		ret = TestClearPageWriteback(page);
P
Peter Zijlstra 已提交
1282
		if (ret) {
L
Linus Torvalds 已提交
1283 1284 1285
			radix_tree_tag_clear(&mapping->page_tree,
						page_index(page),
						PAGECACHE_TAG_WRITEBACK);
1286
			if (bdi_cap_account_writeback(bdi)) {
P
Peter Zijlstra 已提交
1287
				__dec_bdi_stat(bdi, BDI_WRITEBACK);
P
Peter Zijlstra 已提交
1288 1289
				__bdi_writeout_inc(bdi);
			}
P
Peter Zijlstra 已提交
1290
		}
N
Nick Piggin 已提交
1291
		spin_unlock_irqrestore(&mapping->tree_lock, flags);
L
Linus Torvalds 已提交
1292 1293 1294
	} else {
		ret = TestClearPageWriteback(page);
	}
1295 1296
	if (ret)
		dec_zone_page_state(page, NR_WRITEBACK);
L
Linus Torvalds 已提交
1297 1298 1299 1300 1301 1302 1303 1304 1305
	return ret;
}

int test_set_page_writeback(struct page *page)
{
	struct address_space *mapping = page_mapping(page);
	int ret;

	if (mapping) {
P
Peter Zijlstra 已提交
1306
		struct backing_dev_info *bdi = mapping->backing_dev_info;
L
Linus Torvalds 已提交
1307 1308
		unsigned long flags;

N
Nick Piggin 已提交
1309
		spin_lock_irqsave(&mapping->tree_lock, flags);
L
Linus Torvalds 已提交
1310
		ret = TestSetPageWriteback(page);
P
Peter Zijlstra 已提交
1311
		if (!ret) {
L
Linus Torvalds 已提交
1312 1313 1314
			radix_tree_tag_set(&mapping->page_tree,
						page_index(page),
						PAGECACHE_TAG_WRITEBACK);
1315
			if (bdi_cap_account_writeback(bdi))
P
Peter Zijlstra 已提交
1316 1317
				__inc_bdi_stat(bdi, BDI_WRITEBACK);
		}
L
Linus Torvalds 已提交
1318 1319 1320 1321
		if (!PageDirty(page))
			radix_tree_tag_clear(&mapping->page_tree,
						page_index(page),
						PAGECACHE_TAG_DIRTY);
N
Nick Piggin 已提交
1322
		spin_unlock_irqrestore(&mapping->tree_lock, flags);
L
Linus Torvalds 已提交
1323 1324 1325
	} else {
		ret = TestSetPageWriteback(page);
	}
1326 1327
	if (!ret)
		inc_zone_page_state(page, NR_WRITEBACK);
L
Linus Torvalds 已提交
1328 1329 1330 1331 1332 1333
	return ret;

}
EXPORT_SYMBOL(test_set_page_writeback);

/*
N
Nick Piggin 已提交
1334
 * Return true if any of the pages in the mapping are marked with the
L
Linus Torvalds 已提交
1335 1336 1337 1338 1339
 * passed tag.
 */
int mapping_tagged(struct address_space *mapping, int tag)
{
	int ret;
N
Nick Piggin 已提交
1340
	rcu_read_lock();
L
Linus Torvalds 已提交
1341
	ret = radix_tree_tagged(&mapping->page_tree, tag);
N
Nick Piggin 已提交
1342
	rcu_read_unlock();
L
Linus Torvalds 已提交
1343 1344 1345
	return ret;
}
EXPORT_SYMBOL(mapping_tagged);