vmscan.c 102.5 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
/*
 *  linux/mm/vmscan.c
 *
 *  Copyright (C) 1991, 1992, 1993, 1994  Linus Torvalds
 *
 *  Swap reorganised 29.12.95, Stephen Tweedie.
 *  kswapd added: 7.1.96  sct
 *  Removed kswapd_ctl limits, and swap out as many pages as needed
 *  to bring the system back to freepages.high: 2.4.97, Rik van Riel.
 *  Zone aware kswapd started 02/00, Kanoj Sarcar (kanoj@sgi.com).
 *  Multiqueue VM started 5.8.00, Rik van Riel.
 */

#include <linux/mm.h>
#include <linux/module.h>
16
#include <linux/gfp.h>
L
Linus Torvalds 已提交
17 18 19 20 21
#include <linux/kernel_stat.h>
#include <linux/swap.h>
#include <linux/pagemap.h>
#include <linux/init.h>
#include <linux/highmem.h>
22
#include <linux/vmstat.h>
L
Linus Torvalds 已提交
23 24 25 26 27 28 29 30 31 32 33 34
#include <linux/file.h>
#include <linux/writeback.h>
#include <linux/blkdev.h>
#include <linux/buffer_head.h>	/* for try_to_release_page(),
					buffer_heads_over_limit */
#include <linux/mm_inline.h>
#include <linux/pagevec.h>
#include <linux/backing-dev.h>
#include <linux/rmap.h>
#include <linux/topology.h>
#include <linux/cpu.h>
#include <linux/cpuset.h>
35
#include <linux/compaction.h>
L
Linus Torvalds 已提交
36 37
#include <linux/notifier.h>
#include <linux/rwsem.h>
38
#include <linux/delay.h>
39
#include <linux/kthread.h>
40
#include <linux/freezer.h>
41
#include <linux/memcontrol.h>
42
#include <linux/delayacct.h>
43
#include <linux/sysctl.h>
44
#include <linux/oom.h>
45
#include <linux/prefetch.h>
L
Linus Torvalds 已提交
46 47 48 49 50 51

#include <asm/tlbflush.h>
#include <asm/div64.h>

#include <linux/swapops.h>

52 53
#include "internal.h"

54 55 56
#define CREATE_TRACE_POINTS
#include <trace/events/vmscan.h>

57
/*
58 59 60 61 62
 * reclaim_mode determines how the inactive list is shrunk
 * RECLAIM_MODE_SINGLE: Reclaim only order-0 pages
 * RECLAIM_MODE_ASYNC:  Do not block
 * RECLAIM_MODE_SYNC:   Allow blocking e.g. call wait_on_page_writeback
 * RECLAIM_MODE_LUMPYRECLAIM: For high-order allocations, take a reference
63 64
 *			page from the LRU and reclaim all pages within a
 *			naturally aligned range
65
 * RECLAIM_MODE_COMPACTION: For high-order allocations, reclaim a number of
66
 *			order-0 pages and then compact the zone
67
 */
68 69 70 71 72 73
typedef unsigned __bitwise__ reclaim_mode_t;
#define RECLAIM_MODE_SINGLE		((__force reclaim_mode_t)0x01u)
#define RECLAIM_MODE_ASYNC		((__force reclaim_mode_t)0x02u)
#define RECLAIM_MODE_SYNC		((__force reclaim_mode_t)0x04u)
#define RECLAIM_MODE_LUMPYRECLAIM	((__force reclaim_mode_t)0x08u)
#define RECLAIM_MODE_COMPACTION		((__force reclaim_mode_t)0x10u)
74

L
Linus Torvalds 已提交
75 76 77 78
struct scan_control {
	/* Incremented by the number of inactive pages that were scanned */
	unsigned long nr_scanned;

79 80 81
	/* Number of pages freed so far during a call to shrink_zones() */
	unsigned long nr_reclaimed;

82 83 84
	/* How many pages shrink_list() should reclaim */
	unsigned long nr_to_reclaim;

85 86
	unsigned long hibernation_mode;

L
Linus Torvalds 已提交
87
	/* This context's GFP mask */
A
Al Viro 已提交
88
	gfp_t gfp_mask;
L
Linus Torvalds 已提交
89 90 91

	int may_writepage;

92 93
	/* Can mapped pages be reclaimed? */
	int may_unmap;
94

95 96 97
	/* Can pages be swapped as part of reclaim? */
	int may_swap;

A
Andy Whitcroft 已提交
98
	int order;
99

100
	/*
101 102
	 * Intend to reclaim enough continuous memory rather than reclaim
	 * enough amount of memory. i.e, mode for high order allocation.
103
	 */
104
	reclaim_mode_t reclaim_mode;
105

106 107 108 109 110
	/*
	 * The memory cgroup that hit its limit and as a result is the
	 * primary target of this reclaim invocation.
	 */
	struct mem_cgroup *target_mem_cgroup;
111

112 113 114 115 116
	/*
	 * Nodemask of nodes allowed by the caller. If NULL, all nodes
	 * are scanned.
	 */
	nodemask_t	*nodemask;
L
Linus Torvalds 已提交
117 118
};

119 120 121 122 123
struct mem_cgroup_zone {
	struct mem_cgroup *mem_cgroup;
	struct zone *zone;
};

L
Linus Torvalds 已提交
124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157
#define lru_to_page(_head) (list_entry((_head)->prev, struct page, lru))

#ifdef ARCH_HAS_PREFETCH
#define prefetch_prev_lru_page(_page, _base, _field)			\
	do {								\
		if ((_page)->lru.prev != _base) {			\
			struct page *prev;				\
									\
			prev = lru_to_page(&(_page->lru));		\
			prefetch(&prev->_field);			\
		}							\
	} while (0)
#else
#define prefetch_prev_lru_page(_page, _base, _field) do { } while (0)
#endif

#ifdef ARCH_HAS_PREFETCHW
#define prefetchw_prev_lru_page(_page, _base, _field)			\
	do {								\
		if ((_page)->lru.prev != _base) {			\
			struct page *prev;				\
									\
			prev = lru_to_page(&(_page->lru));		\
			prefetchw(&prev->_field);			\
		}							\
	} while (0)
#else
#define prefetchw_prev_lru_page(_page, _base, _field) do { } while (0)
#endif

/*
 * From 0 .. 100.  Higher means more swappy.
 */
int vm_swappiness = 60;
158
long vm_total_pages;	/* The total number of pages which the VM controls */
L
Linus Torvalds 已提交
159 160 161 162

static LIST_HEAD(shrinker_list);
static DECLARE_RWSEM(shrinker_rwsem);

163
#ifdef CONFIG_CGROUP_MEM_RES_CTLR
164 165
static bool global_reclaim(struct scan_control *sc)
{
166
	return !sc->target_mem_cgroup;
167 168
}

169
static bool scanning_global_lru(struct mem_cgroup_zone *mz)
170
{
171
	return !mz->mem_cgroup;
172
}
173
#else
174 175 176 177 178
static bool global_reclaim(struct scan_control *sc)
{
	return true;
}

179
static bool scanning_global_lru(struct mem_cgroup_zone *mz)
180 181 182
{
	return true;
}
183 184
#endif

185
static struct zone_reclaim_stat *get_reclaim_stat(struct mem_cgroup_zone *mz)
186
{
187 188
	if (!scanning_global_lru(mz))
		return mem_cgroup_get_reclaim_stat(mz->mem_cgroup, mz->zone);
K
KOSAKI Motohiro 已提交
189

190
	return &mz->zone->reclaim_stat;
191 192
}

193 194
static unsigned long zone_nr_lru_pages(struct mem_cgroup_zone *mz,
				       enum lru_list lru)
195
{
196 197 198 199 200
	if (!scanning_global_lru(mz))
		return mem_cgroup_zone_nr_lru_pages(mz->mem_cgroup,
						    zone_to_nid(mz->zone),
						    zone_idx(mz->zone),
						    BIT(lru));
201

202
	return zone_page_state(mz->zone, NR_LRU_BASE + lru);
203 204 205
}


L
Linus Torvalds 已提交
206 207 208
/*
 * Add a shrinker callback to be called from the vm
 */
209
void register_shrinker(struct shrinker *shrinker)
L
Linus Torvalds 已提交
210
{
211
	atomic_long_set(&shrinker->nr_in_batch, 0);
212 213 214
	down_write(&shrinker_rwsem);
	list_add_tail(&shrinker->list, &shrinker_list);
	up_write(&shrinker_rwsem);
L
Linus Torvalds 已提交
215
}
216
EXPORT_SYMBOL(register_shrinker);
L
Linus Torvalds 已提交
217 218 219 220

/*
 * Remove one
 */
221
void unregister_shrinker(struct shrinker *shrinker)
L
Linus Torvalds 已提交
222 223 224 225 226
{
	down_write(&shrinker_rwsem);
	list_del(&shrinker->list);
	up_write(&shrinker_rwsem);
}
227
EXPORT_SYMBOL(unregister_shrinker);
L
Linus Torvalds 已提交
228

229 230 231 232 233 234 235 236
static inline int do_shrinker_shrink(struct shrinker *shrinker,
				     struct shrink_control *sc,
				     unsigned long nr_to_scan)
{
	sc->nr_to_scan = nr_to_scan;
	return (*shrinker->shrink)(shrinker, sc);
}

L
Linus Torvalds 已提交
237 238 239 240 241 242 243 244 245
#define SHRINK_BATCH 128
/*
 * Call the shrink functions to age shrinkable caches
 *
 * Here we assume it costs one seek to replace a lru page and that it also
 * takes a seek to recreate a cache object.  With this in mind we age equal
 * percentages of the lru and ageable caches.  This should balance the seeks
 * generated by these structures.
 *
S
Simon Arlott 已提交
246
 * If the vm encountered mapped pages on the LRU it increase the pressure on
L
Linus Torvalds 已提交
247 248 249 250 251 252 253
 * slab to avoid swapping.
 *
 * We do weird things to avoid (scanned*seeks*entries) overflowing 32 bits.
 *
 * `lru_pages' represents the number of on-LRU pages in all the zones which
 * are eligible for the caller's allocation attempt.  It is used for balancing
 * slab reclaim versus page reclaim.
254 255
 *
 * Returns the number of slab objects which we shrunk.
L
Linus Torvalds 已提交
256
 */
257
unsigned long shrink_slab(struct shrink_control *shrink,
258
			  unsigned long nr_pages_scanned,
259
			  unsigned long lru_pages)
L
Linus Torvalds 已提交
260 261
{
	struct shrinker *shrinker;
262
	unsigned long ret = 0;
L
Linus Torvalds 已提交
263

264 265
	if (nr_pages_scanned == 0)
		nr_pages_scanned = SWAP_CLUSTER_MAX;
L
Linus Torvalds 已提交
266

267 268 269 270 271
	if (!down_read_trylock(&shrinker_rwsem)) {
		/* Assume we'll be able to shrink next time */
		ret = 1;
		goto out;
	}
L
Linus Torvalds 已提交
272 273 274

	list_for_each_entry(shrinker, &shrinker_list, list) {
		unsigned long long delta;
275 276
		long total_scan;
		long max_pass;
277
		int shrink_ret = 0;
278 279
		long nr;
		long new_nr;
280 281
		long batch_size = shrinker->batch ? shrinker->batch
						  : SHRINK_BATCH;
L
Linus Torvalds 已提交
282

283 284 285 286
		max_pass = do_shrinker_shrink(shrinker, shrink, 0);
		if (max_pass <= 0)
			continue;

287 288 289 290 291
		/*
		 * copy the current shrinker scan count into a local variable
		 * and zero it so that other concurrent shrinker invocations
		 * don't also do this scanning work.
		 */
292
		nr = atomic_long_xchg(&shrinker->nr_in_batch, 0);
293 294

		total_scan = nr;
295
		delta = (4 * nr_pages_scanned) / shrinker->seeks;
296
		delta *= max_pass;
L
Linus Torvalds 已提交
297
		do_div(delta, lru_pages + 1);
298 299
		total_scan += delta;
		if (total_scan < 0) {
300 301
			printk(KERN_ERR "shrink_slab: %pF negative objects to "
			       "delete nr=%ld\n",
302 303
			       shrinker->shrink, total_scan);
			total_scan = max_pass;
304 305
		}

306 307 308 309 310 311 312 313 314 315 316 317 318 319 320
		/*
		 * We need to avoid excessive windup on filesystem shrinkers
		 * due to large numbers of GFP_NOFS allocations causing the
		 * shrinkers to return -1 all the time. This results in a large
		 * nr being built up so when a shrink that can do some work
		 * comes along it empties the entire cache due to nr >>>
		 * max_pass.  This is bad for sustaining a working set in
		 * memory.
		 *
		 * Hence only allow the shrinker to scan the entire cache when
		 * a large delta change is calculated directly.
		 */
		if (delta < max_pass / 4)
			total_scan = min(total_scan, max_pass / 2);

321 322 323 324 325
		/*
		 * Avoid risking looping forever due to too large nr value:
		 * never try to free more than twice the estimate number of
		 * freeable entries.
		 */
326 327
		if (total_scan > max_pass * 2)
			total_scan = max_pass * 2;
L
Linus Torvalds 已提交
328

329
		trace_mm_shrink_slab_start(shrinker, shrink, nr,
330 331 332
					nr_pages_scanned, lru_pages,
					max_pass, delta, total_scan);

333
		while (total_scan >= batch_size) {
334
			int nr_before;
L
Linus Torvalds 已提交
335

336 337
			nr_before = do_shrinker_shrink(shrinker, shrink, 0);
			shrink_ret = do_shrinker_shrink(shrinker, shrink,
338
							batch_size);
L
Linus Torvalds 已提交
339 340
			if (shrink_ret == -1)
				break;
341 342
			if (shrink_ret < nr_before)
				ret += nr_before - shrink_ret;
343 344
			count_vm_events(SLABS_SCANNED, batch_size);
			total_scan -= batch_size;
L
Linus Torvalds 已提交
345 346 347 348

			cond_resched();
		}

349 350 351 352 353
		/*
		 * move the unused scan count back into the shrinker in a
		 * manner that handles concurrent updates. If we exhausted the
		 * scan, there is no need to do an update.
		 */
354 355 356 357 358
		if (total_scan > 0)
			new_nr = atomic_long_add_return(total_scan,
					&shrinker->nr_in_batch);
		else
			new_nr = atomic_long_read(&shrinker->nr_in_batch);
359 360

		trace_mm_shrink_slab_end(shrinker, shrink_ret, nr, new_nr);
L
Linus Torvalds 已提交
361 362
	}
	up_read(&shrinker_rwsem);
363 364
out:
	cond_resched();
365
	return ret;
L
Linus Torvalds 已提交
366 367
}

368
static void set_reclaim_mode(int priority, struct scan_control *sc,
369 370
				   bool sync)
{
371
	reclaim_mode_t syncmode = sync ? RECLAIM_MODE_SYNC : RECLAIM_MODE_ASYNC;
372 373

	/*
374 375 376
	 * Initially assume we are entering either lumpy reclaim or
	 * reclaim/compaction.Depending on the order, we will either set the
	 * sync mode or just reclaim order-0 pages later.
377
	 */
378
	if (COMPACTION_BUILD)
379
		sc->reclaim_mode = RECLAIM_MODE_COMPACTION;
380
	else
381
		sc->reclaim_mode = RECLAIM_MODE_LUMPYRECLAIM;
382 383

	/*
384 385 386
	 * Avoid using lumpy reclaim or reclaim/compaction if possible by
	 * restricting when its set to either costly allocations or when
	 * under memory pressure
387 388
	 */
	if (sc->order > PAGE_ALLOC_COSTLY_ORDER)
389
		sc->reclaim_mode |= syncmode;
390
	else if (sc->order && priority < DEF_PRIORITY - 2)
391
		sc->reclaim_mode |= syncmode;
392
	else
393
		sc->reclaim_mode = RECLAIM_MODE_SINGLE | RECLAIM_MODE_ASYNC;
394 395
}

396
static void reset_reclaim_mode(struct scan_control *sc)
397
{
398
	sc->reclaim_mode = RECLAIM_MODE_SINGLE | RECLAIM_MODE_ASYNC;
399 400
}

L
Linus Torvalds 已提交
401 402
static inline int is_page_cache_freeable(struct page *page)
{
403 404 405 406 407
	/*
	 * A freeable page cache page is referenced only by the caller
	 * that isolated the page, the page cache radix tree and
	 * optional buffer heads at page->private.
	 */
408
	return page_count(page) - page_has_private(page) == 2;
L
Linus Torvalds 已提交
409 410
}

411 412
static int may_write_to_queue(struct backing_dev_info *bdi,
			      struct scan_control *sc)
L
Linus Torvalds 已提交
413
{
414
	if (current->flags & PF_SWAPWRITE)
L
Linus Torvalds 已提交
415 416 417 418 419
		return 1;
	if (!bdi_write_congested(bdi))
		return 1;
	if (bdi == current->backing_dev_info)
		return 1;
420 421 422 423

	/* lumpy reclaim for hugepage often need a lot of write */
	if (sc->order > PAGE_ALLOC_COSTLY_ORDER)
		return 1;
L
Linus Torvalds 已提交
424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441
	return 0;
}

/*
 * We detected a synchronous write error writing a page out.  Probably
 * -ENOSPC.  We need to propagate that into the address_space for a subsequent
 * fsync(), msync() or close().
 *
 * The tricky part is that after writepage we cannot touch the mapping: nothing
 * prevents it from being freed up.  But we have a ref on the page and once
 * that page is locked, the mapping is pinned.
 *
 * We're allowed to run sleeping lock_page() here because we know the caller has
 * __GFP_FS.
 */
static void handle_write_error(struct address_space *mapping,
				struct page *page, int error)
{
J
Jens Axboe 已提交
442
	lock_page(page);
443 444
	if (page_mapping(page) == mapping)
		mapping_set_error(mapping, error);
L
Linus Torvalds 已提交
445 446 447
	unlock_page(page);
}

448 449 450 451 452 453 454 455 456 457 458 459
/* possible outcome of pageout() */
typedef enum {
	/* failed to write page out, page is locked */
	PAGE_KEEP,
	/* move page to the active list, page is locked */
	PAGE_ACTIVATE,
	/* page has been sent to the disk successfully, page is unlocked */
	PAGE_SUCCESS,
	/* page is clean and locked */
	PAGE_CLEAN,
} pageout_t;

L
Linus Torvalds 已提交
460
/*
A
Andrew Morton 已提交
461 462
 * pageout is called by shrink_page_list() for each dirty page.
 * Calls ->writepage().
L
Linus Torvalds 已提交
463
 */
464
static pageout_t pageout(struct page *page, struct address_space *mapping,
465
			 struct scan_control *sc)
L
Linus Torvalds 已提交
466 467 468 469 470 471 472 473
{
	/*
	 * If the page is dirty, only perform writeback if that write
	 * will be non-blocking.  To prevent this allocation from being
	 * stalled by pagecache activity.  But note that there may be
	 * stalls if we need to run get_block().  We could test
	 * PagePrivate for that.
	 *
474
	 * If this process is currently in __generic_file_aio_write() against
L
Linus Torvalds 已提交
475 476 477 478 479 480 481 482 483 484 485 486 487 488 489
	 * this page's queue, we can perform writeback even if that
	 * will block.
	 *
	 * If the page is swapcache, write it back even if that would
	 * block, for some throttling. This happens by accident, because
	 * swap_backing_dev_info is bust: it doesn't reflect the
	 * congestion state of the swapdevs.  Easy to fix, if needed.
	 */
	if (!is_page_cache_freeable(page))
		return PAGE_KEEP;
	if (!mapping) {
		/*
		 * Some data journaling orphaned pages can have
		 * page->mapping == NULL while being dirty with clean buffers.
		 */
490
		if (page_has_private(page)) {
L
Linus Torvalds 已提交
491 492
			if (try_to_free_buffers(page)) {
				ClearPageDirty(page);
493
				printk("%s: orphaned page\n", __func__);
L
Linus Torvalds 已提交
494 495 496 497 498 499 500
				return PAGE_CLEAN;
			}
		}
		return PAGE_KEEP;
	}
	if (mapping->a_ops->writepage == NULL)
		return PAGE_ACTIVATE;
501
	if (!may_write_to_queue(mapping->backing_dev_info, sc))
L
Linus Torvalds 已提交
502 503 504 505 506 507 508
		return PAGE_KEEP;

	if (clear_page_dirty_for_io(page)) {
		int res;
		struct writeback_control wbc = {
			.sync_mode = WB_SYNC_NONE,
			.nr_to_write = SWAP_CLUSTER_MAX,
509 510
			.range_start = 0,
			.range_end = LLONG_MAX,
L
Linus Torvalds 已提交
511 512 513 514 515 516 517
			.for_reclaim = 1,
		};

		SetPageReclaim(page);
		res = mapping->a_ops->writepage(page, &wbc);
		if (res < 0)
			handle_write_error(mapping, page, res);
518
		if (res == AOP_WRITEPAGE_ACTIVATE) {
L
Linus Torvalds 已提交
519 520 521
			ClearPageReclaim(page);
			return PAGE_ACTIVATE;
		}
522

L
Linus Torvalds 已提交
523 524 525 526
		if (!PageWriteback(page)) {
			/* synchronous write or broken a_ops? */
			ClearPageReclaim(page);
		}
527
		trace_mm_vmscan_writepage(page,
528
			trace_reclaim_flags(page, sc->reclaim_mode));
529
		inc_zone_page_state(page, NR_VMSCAN_WRITE);
L
Linus Torvalds 已提交
530 531 532 533 534 535
		return PAGE_SUCCESS;
	}

	return PAGE_CLEAN;
}

536
/*
N
Nick Piggin 已提交
537 538
 * Same as remove_mapping, but if the page is removed from the mapping, it
 * gets returned with a refcount of 0.
539
 */
N
Nick Piggin 已提交
540
static int __remove_mapping(struct address_space *mapping, struct page *page)
541
{
542 543
	BUG_ON(!PageLocked(page));
	BUG_ON(mapping != page_mapping(page));
544

N
Nick Piggin 已提交
545
	spin_lock_irq(&mapping->tree_lock);
546
	/*
N
Nick Piggin 已提交
547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569
	 * The non racy check for a busy page.
	 *
	 * Must be careful with the order of the tests. When someone has
	 * a ref to the page, it may be possible that they dirty it then
	 * drop the reference. So if PageDirty is tested before page_count
	 * here, then the following race may occur:
	 *
	 * get_user_pages(&page);
	 * [user mapping goes away]
	 * write_to(page);
	 *				!PageDirty(page)    [good]
	 * SetPageDirty(page);
	 * put_page(page);
	 *				!page_count(page)   [good, discard it]
	 *
	 * [oops, our write_to data is lost]
	 *
	 * Reversing the order of the tests ensures such a situation cannot
	 * escape unnoticed. The smp_rmb is needed to ensure the page->flags
	 * load is not satisfied before that of page->_count.
	 *
	 * Note that if SetPageDirty is always performed via set_page_dirty,
	 * and thus under tree_lock, then this ordering is not required.
570
	 */
N
Nick Piggin 已提交
571
	if (!page_freeze_refs(page, 2))
572
		goto cannot_free;
N
Nick Piggin 已提交
573 574 575
	/* note: atomic_cmpxchg in page_freeze_refs provides the smp_rmb */
	if (unlikely(PageDirty(page))) {
		page_unfreeze_refs(page, 2);
576
		goto cannot_free;
N
Nick Piggin 已提交
577
	}
578 579 580 581

	if (PageSwapCache(page)) {
		swp_entry_t swap = { .val = page_private(page) };
		__delete_from_swap_cache(page);
N
Nick Piggin 已提交
582
		spin_unlock_irq(&mapping->tree_lock);
583
		swapcache_free(swap, page);
N
Nick Piggin 已提交
584
	} else {
585 586 587 588
		void (*freepage)(struct page *);

		freepage = mapping->a_ops->freepage;

589
		__delete_from_page_cache(page);
N
Nick Piggin 已提交
590
		spin_unlock_irq(&mapping->tree_lock);
591
		mem_cgroup_uncharge_cache_page(page);
592 593 594

		if (freepage != NULL)
			freepage(page);
595 596 597 598 599
	}

	return 1;

cannot_free:
N
Nick Piggin 已提交
600
	spin_unlock_irq(&mapping->tree_lock);
601 602 603
	return 0;
}

N
Nick Piggin 已提交
604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623
/*
 * Attempt to detach a locked page from its ->mapping.  If it is dirty or if
 * someone else has a ref on the page, abort and return 0.  If it was
 * successfully detached, return 1.  Assumes the caller has a single ref on
 * this page.
 */
int remove_mapping(struct address_space *mapping, struct page *page)
{
	if (__remove_mapping(mapping, page)) {
		/*
		 * Unfreezing the refcount with 1 rather than 2 effectively
		 * drops the pagecache ref for us without requiring another
		 * atomic operation.
		 */
		page_unfreeze_refs(page, 1);
		return 1;
	}
	return 0;
}

L
Lee Schermerhorn 已提交
624 625 626 627 628 629 630 631 632 633 634 635 636
/**
 * putback_lru_page - put previously isolated page onto appropriate LRU list
 * @page: page to be put back to appropriate lru list
 *
 * Add previously isolated @page to appropriate LRU list.
 * Page may still be unevictable for other reasons.
 *
 * lru_lock must not be held, interrupts must be enabled.
 */
void putback_lru_page(struct page *page)
{
	int lru;
	int active = !!TestClearPageActive(page);
637
	int was_unevictable = PageUnevictable(page);
L
Lee Schermerhorn 已提交
638 639 640 641 642 643 644 645 646 647 648 649 650

	VM_BUG_ON(PageLRU(page));

redo:
	ClearPageUnevictable(page);

	if (page_evictable(page, NULL)) {
		/*
		 * For evictable pages, we can use the cache.
		 * In event of a race, worst case is we end up with an
		 * unevictable page on [in]active list.
		 * We know how to handle that.
		 */
651
		lru = active + page_lru_base_type(page);
L
Lee Schermerhorn 已提交
652 653 654 655 656 657 658 659
		lru_cache_add_lru(page, lru);
	} else {
		/*
		 * Put unevictable pages directly on zone's unevictable
		 * list.
		 */
		lru = LRU_UNEVICTABLE;
		add_page_to_unevictable_list(page);
660
		/*
661 662 663 664 665
		 * When racing with an mlock or AS_UNEVICTABLE clearing
		 * (page is unlocked) make sure that if the other thread
		 * does not observe our setting of PG_lru and fails
		 * isolation/check_move_unevictable_page,
		 * we see PG_mlocked/AS_UNEVICTABLE cleared below and move
666 667
		 * the page back to the evictable list.
		 *
668
		 * The other side is TestClearPageMlocked() or shmem_lock().
669 670
		 */
		smp_mb();
L
Lee Schermerhorn 已提交
671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688
	}

	/*
	 * page's status can change while we move it among lru. If an evictable
	 * page is on unevictable list, it never be freed. To avoid that,
	 * check after we added it to the list, again.
	 */
	if (lru == LRU_UNEVICTABLE && page_evictable(page, NULL)) {
		if (!isolate_lru_page(page)) {
			put_page(page);
			goto redo;
		}
		/* This means someone else dropped this page from LRU
		 * So, it will be freed or putback to LRU again. There is
		 * nothing to do here.
		 */
	}

689 690 691 692 693
	if (was_unevictable && lru != LRU_UNEVICTABLE)
		count_vm_event(UNEVICTABLE_PGRESCUED);
	else if (!was_unevictable && lru == LRU_UNEVICTABLE)
		count_vm_event(UNEVICTABLE_PGCULLED);

L
Lee Schermerhorn 已提交
694 695 696
	put_page(page);		/* drop ref from isolate */
}

697 698 699
enum page_references {
	PAGEREF_RECLAIM,
	PAGEREF_RECLAIM_CLEAN,
700
	PAGEREF_KEEP,
701 702 703 704
	PAGEREF_ACTIVATE,
};

static enum page_references page_check_references(struct page *page,
705
						  struct mem_cgroup_zone *mz,
706 707
						  struct scan_control *sc)
{
708
	int referenced_ptes, referenced_page;
709 710
	unsigned long vm_flags;

711
	referenced_ptes = page_referenced(page, 1, mz->mem_cgroup, &vm_flags);
712
	referenced_page = TestClearPageReferenced(page);
713 714

	/* Lumpy reclaim - ignore references */
715
	if (sc->reclaim_mode & RECLAIM_MODE_LUMPYRECLAIM)
716 717 718 719 720 721 722 723 724
		return PAGEREF_RECLAIM;

	/*
	 * Mlock lost the isolation race with us.  Let try_to_unmap()
	 * move the page to the unevictable list.
	 */
	if (vm_flags & VM_LOCKED)
		return PAGEREF_RECLAIM;

725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743
	if (referenced_ptes) {
		if (PageAnon(page))
			return PAGEREF_ACTIVATE;
		/*
		 * All mapped pages start out with page table
		 * references from the instantiating fault, so we need
		 * to look twice if a mapped file page is used more
		 * than once.
		 *
		 * Mark it and spare it for another trip around the
		 * inactive list.  Another page table reference will
		 * lead to its activation.
		 *
		 * Note: the mark is set for activated pages as well
		 * so that recently deactivated but used pages are
		 * quickly recovered.
		 */
		SetPageReferenced(page);

744
		if (referenced_page || referenced_ptes > 1)
745 746
			return PAGEREF_ACTIVATE;

747 748 749 750 751 752
		/*
		 * Activate file-backed executable pages after first usage.
		 */
		if (vm_flags & VM_EXEC)
			return PAGEREF_ACTIVATE;

753 754
		return PAGEREF_KEEP;
	}
755 756

	/* Reclaim if clean, defer dirty pages to writeback */
757
	if (referenced_page && !PageSwapBacked(page))
758 759 760
		return PAGEREF_RECLAIM_CLEAN;

	return PAGEREF_RECLAIM;
761 762
}

L
Linus Torvalds 已提交
763
/*
A
Andrew Morton 已提交
764
 * shrink_page_list() returns the number of reclaimed pages
L
Linus Torvalds 已提交
765
 */
A
Andrew Morton 已提交
766
static unsigned long shrink_page_list(struct list_head *page_list,
767
				      struct mem_cgroup_zone *mz,
768
				      struct scan_control *sc,
769 770 771
				      int priority,
				      unsigned long *ret_nr_dirty,
				      unsigned long *ret_nr_writeback)
L
Linus Torvalds 已提交
772 773
{
	LIST_HEAD(ret_pages);
774
	LIST_HEAD(free_pages);
L
Linus Torvalds 已提交
775
	int pgactivate = 0;
776 777
	unsigned long nr_dirty = 0;
	unsigned long nr_congested = 0;
778
	unsigned long nr_reclaimed = 0;
779
	unsigned long nr_writeback = 0;
L
Linus Torvalds 已提交
780 781 782 783

	cond_resched();

	while (!list_empty(page_list)) {
784
		enum page_references references;
L
Linus Torvalds 已提交
785 786 787 788 789 790 791 792 793
		struct address_space *mapping;
		struct page *page;
		int may_enter_fs;

		cond_resched();

		page = lru_to_page(page_list);
		list_del(&page->lru);

N
Nick Piggin 已提交
794
		if (!trylock_page(page))
L
Linus Torvalds 已提交
795 796
			goto keep;

N
Nick Piggin 已提交
797
		VM_BUG_ON(PageActive(page));
798
		VM_BUG_ON(page_zone(page) != mz->zone);
L
Linus Torvalds 已提交
799 800

		sc->nr_scanned++;
801

N
Nick Piggin 已提交
802 803
		if (unlikely(!page_evictable(page, NULL)))
			goto cull_mlocked;
L
Lee Schermerhorn 已提交
804

805
		if (!sc->may_unmap && page_mapped(page))
806 807
			goto keep_locked;

L
Linus Torvalds 已提交
808 809 810 811
		/* Double the slab pressure for mapped and swapcache pages */
		if (page_mapped(page) || PageSwapCache(page))
			sc->nr_scanned++;

812 813 814 815
		may_enter_fs = (sc->gfp_mask & __GFP_FS) ||
			(PageSwapCache(page) && (sc->gfp_mask & __GFP_IO));

		if (PageWriteback(page)) {
816
			nr_writeback++;
817
			/*
818 819 820 821
			 * Synchronous reclaim cannot queue pages for
			 * writeback due to the possibility of stack overflow
			 * but if it encounters a page under writeback, wait
			 * for the IO to complete.
822
			 */
823
			if ((sc->reclaim_mode & RECLAIM_MODE_SYNC) &&
824
			    may_enter_fs)
825
				wait_on_page_writeback(page);
826 827 828 829
			else {
				unlock_page(page);
				goto keep_lumpy;
			}
830
		}
L
Linus Torvalds 已提交
831

832
		references = page_check_references(page, mz, sc);
833 834
		switch (references) {
		case PAGEREF_ACTIVATE:
L
Linus Torvalds 已提交
835
			goto activate_locked;
836 837
		case PAGEREF_KEEP:
			goto keep_locked;
838 839 840 841
		case PAGEREF_RECLAIM:
		case PAGEREF_RECLAIM_CLEAN:
			; /* try to reclaim the page below */
		}
L
Linus Torvalds 已提交
842 843 844 845 846

		/*
		 * Anonymous process memory has backing store?
		 * Try to allocate it some swap space here.
		 */
N
Nick Piggin 已提交
847
		if (PageAnon(page) && !PageSwapCache(page)) {
848 849
			if (!(sc->gfp_mask & __GFP_IO))
				goto keep_locked;
850
			if (!add_to_swap(page))
L
Linus Torvalds 已提交
851
				goto activate_locked;
852
			may_enter_fs = 1;
N
Nick Piggin 已提交
853
		}
L
Linus Torvalds 已提交
854 855 856 857 858 859 860 861

		mapping = page_mapping(page);

		/*
		 * The page is mapped into the page tables of one or more
		 * processes. Try to unmap it here.
		 */
		if (page_mapped(page) && mapping) {
862
			switch (try_to_unmap(page, TTU_UNMAP)) {
L
Linus Torvalds 已提交
863 864 865 866
			case SWAP_FAIL:
				goto activate_locked;
			case SWAP_AGAIN:
				goto keep_locked;
N
Nick Piggin 已提交
867 868
			case SWAP_MLOCK:
				goto cull_mlocked;
L
Linus Torvalds 已提交
869 870 871 872 873 874
			case SWAP_SUCCESS:
				; /* try to free the page below */
			}
		}

		if (PageDirty(page)) {
875 876
			nr_dirty++;

877 878
			/*
			 * Only kswapd can writeback filesystem pages to
879 880
			 * avoid risk of stack overflow but do not writeback
			 * unless under significant pressure.
881
			 */
882 883
			if (page_is_file_cache(page) &&
					(!current_is_kswapd() || priority >= DEF_PRIORITY - 2)) {
884 885 886 887 888 889 890 891 892
				/*
				 * Immediately reclaim when written back.
				 * Similar in principal to deactivate_page()
				 * except we already have the page isolated
				 * and know it's dirty
				 */
				inc_zone_page_state(page, NR_VMSCAN_IMMEDIATE);
				SetPageReclaim(page);

893 894 895
				goto keep_locked;
			}

896
			if (references == PAGEREF_RECLAIM_CLEAN)
L
Linus Torvalds 已提交
897
				goto keep_locked;
898
			if (!may_enter_fs)
L
Linus Torvalds 已提交
899
				goto keep_locked;
900
			if (!sc->may_writepage)
L
Linus Torvalds 已提交
901 902 903
				goto keep_locked;

			/* Page is dirty, try to write it out here */
904
			switch (pageout(page, mapping, sc)) {
L
Linus Torvalds 已提交
905
			case PAGE_KEEP:
906
				nr_congested++;
L
Linus Torvalds 已提交
907 908 909 910
				goto keep_locked;
			case PAGE_ACTIVATE:
				goto activate_locked;
			case PAGE_SUCCESS:
911 912 913
				if (PageWriteback(page))
					goto keep_lumpy;
				if (PageDirty(page))
L
Linus Torvalds 已提交
914
					goto keep;
915

L
Linus Torvalds 已提交
916 917 918 919
				/*
				 * A synchronous write - probably a ramdisk.  Go
				 * ahead and try to reclaim the page.
				 */
N
Nick Piggin 已提交
920
				if (!trylock_page(page))
L
Linus Torvalds 已提交
921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939
					goto keep;
				if (PageDirty(page) || PageWriteback(page))
					goto keep_locked;
				mapping = page_mapping(page);
			case PAGE_CLEAN:
				; /* try to free the page below */
			}
		}

		/*
		 * If the page has buffers, try to free the buffer mappings
		 * associated with this page. If we succeed we try to free
		 * the page as well.
		 *
		 * We do this even if the page is PageDirty().
		 * try_to_release_page() does not perform I/O, but it is
		 * possible for a page to have PageDirty set, but it is actually
		 * clean (all its buffers are clean).  This happens if the
		 * buffers were written out directly, with submit_bh(). ext3
L
Lee Schermerhorn 已提交
940
		 * will do this, as well as the blockdev mapping.
L
Linus Torvalds 已提交
941 942 943 944 945 946 947 948 949 950
		 * try_to_release_page() will discover that cleanness and will
		 * drop the buffers and mark the page clean - it can be freed.
		 *
		 * Rarely, pages can have buffers and no ->mapping.  These are
		 * the pages which were not successfully invalidated in
		 * truncate_complete_page().  We try to drop those buffers here
		 * and if that worked, and the page is no longer mapped into
		 * process address space (page_count == 1) it can be freed.
		 * Otherwise, leave the page on the LRU so it is swappable.
		 */
951
		if (page_has_private(page)) {
L
Linus Torvalds 已提交
952 953
			if (!try_to_release_page(page, sc->gfp_mask))
				goto activate_locked;
N
Nick Piggin 已提交
954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969
			if (!mapping && page_count(page) == 1) {
				unlock_page(page);
				if (put_page_testzero(page))
					goto free_it;
				else {
					/*
					 * rare race with speculative reference.
					 * the speculative reference will free
					 * this page shortly, so we may
					 * increment nr_reclaimed here (and
					 * leave it off the LRU).
					 */
					nr_reclaimed++;
					continue;
				}
			}
L
Linus Torvalds 已提交
970 971
		}

N
Nick Piggin 已提交
972
		if (!mapping || !__remove_mapping(mapping, page))
973
			goto keep_locked;
L
Linus Torvalds 已提交
974

N
Nick Piggin 已提交
975 976 977 978 979 980 981 982
		/*
		 * At this point, we have no other references and there is
		 * no way to pick any more up (removed from LRU, removed
		 * from pagecache). Can use non-atomic bitops now (and
		 * we obviously don't have to worry about waking up a process
		 * waiting on the page lock, because there are no references.
		 */
		__clear_page_locked(page);
N
Nick Piggin 已提交
983
free_it:
984
		nr_reclaimed++;
985 986 987 988 989 990

		/*
		 * Is there need to periodically free_page_list? It would
		 * appear not as the counts should be low
		 */
		list_add(&page->lru, &free_pages);
L
Linus Torvalds 已提交
991 992
		continue;

N
Nick Piggin 已提交
993
cull_mlocked:
994 995
		if (PageSwapCache(page))
			try_to_free_swap(page);
N
Nick Piggin 已提交
996 997
		unlock_page(page);
		putback_lru_page(page);
998
		reset_reclaim_mode(sc);
N
Nick Piggin 已提交
999 1000
		continue;

L
Linus Torvalds 已提交
1001
activate_locked:
1002 1003
		/* Not a candidate for swapping, so reclaim swap space. */
		if (PageSwapCache(page) && vm_swap_full())
1004
			try_to_free_swap(page);
L
Lee Schermerhorn 已提交
1005
		VM_BUG_ON(PageActive(page));
L
Linus Torvalds 已提交
1006 1007 1008 1009 1010
		SetPageActive(page);
		pgactivate++;
keep_locked:
		unlock_page(page);
keep:
1011
		reset_reclaim_mode(sc);
1012
keep_lumpy:
L
Linus Torvalds 已提交
1013
		list_add(&page->lru, &ret_pages);
N
Nick Piggin 已提交
1014
		VM_BUG_ON(PageLRU(page) || PageUnevictable(page));
L
Linus Torvalds 已提交
1015
	}
1016

1017 1018 1019 1020 1021 1022
	/*
	 * Tag a zone as congested if all the dirty pages encountered were
	 * backed by a congested BDI. In this case, reclaimers should just
	 * back off and wait for congestion to clear because further reclaim
	 * will encounter the same problem
	 */
1023
	if (nr_dirty && nr_dirty == nr_congested && global_reclaim(sc))
1024
		zone_set_flag(mz->zone, ZONE_CONGESTED);
1025

1026
	free_hot_cold_page_list(&free_pages, 1);
1027

L
Linus Torvalds 已提交
1028
	list_splice(&ret_pages, page_list);
1029
	count_vm_events(PGACTIVATE, pgactivate);
1030 1031
	*ret_nr_dirty += nr_dirty;
	*ret_nr_writeback += nr_writeback;
1032
	return nr_reclaimed;
L
Linus Torvalds 已提交
1033 1034
}

A
Andy Whitcroft 已提交
1035 1036 1037 1038 1039 1040 1041 1042 1043 1044
/*
 * Attempt to remove the specified page from its LRU.  Only take this page
 * if it is of the appropriate PageActive status.  Pages which are being
 * freed elsewhere are also ignored.
 *
 * page:	page to consider
 * mode:	one of the LRU isolation modes defined above
 *
 * returns 0 on success, -ve errno on failure.
 */
1045
int __isolate_lru_page(struct page *page, isolate_mode_t mode, int file)
A
Andy Whitcroft 已提交
1046
{
1047
	bool all_lru_mode;
A
Andy Whitcroft 已提交
1048 1049 1050 1051 1052 1053
	int ret = -EINVAL;

	/* Only take pages on the LRU. */
	if (!PageLRU(page))
		return ret;

1054 1055 1056
	all_lru_mode = (mode & (ISOLATE_ACTIVE|ISOLATE_INACTIVE)) ==
		(ISOLATE_ACTIVE|ISOLATE_INACTIVE);

A
Andy Whitcroft 已提交
1057 1058 1059 1060 1061
	/*
	 * When checking the active state, we need to be sure we are
	 * dealing with comparible boolean values.  Take the logical not
	 * of each.
	 */
1062
	if (!all_lru_mode && !PageActive(page) != !(mode & ISOLATE_ACTIVE))
A
Andy Whitcroft 已提交
1063 1064
		return ret;

1065
	if (!all_lru_mode && !!page_is_file_cache(page) != file)
1066 1067
		return ret;

L
Lee Schermerhorn 已提交
1068 1069 1070 1071 1072 1073 1074 1075
	/*
	 * When this function is being called for lumpy reclaim, we
	 * initially look into all LRU pages, active, inactive and
	 * unevictable; only give shrink_page_list evictable pages.
	 */
	if (PageUnevictable(page))
		return ret;

A
Andy Whitcroft 已提交
1076
	ret = -EBUSY;
K
KAMEZAWA Hiroyuki 已提交
1077

1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110
	/*
	 * To minimise LRU disruption, the caller can indicate that it only
	 * wants to isolate pages it will be able to operate on without
	 * blocking - clean pages for the most part.
	 *
	 * ISOLATE_CLEAN means that only clean pages should be isolated. This
	 * is used by reclaim when it is cannot write to backing storage
	 *
	 * ISOLATE_ASYNC_MIGRATE is used to indicate that it only wants to pages
	 * that it is possible to migrate without blocking
	 */
	if (mode & (ISOLATE_CLEAN|ISOLATE_ASYNC_MIGRATE)) {
		/* All the caller can do on PageWriteback is block */
		if (PageWriteback(page))
			return ret;

		if (PageDirty(page)) {
			struct address_space *mapping;

			/* ISOLATE_CLEAN means only clean pages */
			if (mode & ISOLATE_CLEAN)
				return ret;

			/*
			 * Only pages without mappings or that have a
			 * ->migratepage callback are possible to migrate
			 * without blocking
			 */
			mapping = page_mapping(page);
			if (mapping && !mapping->a_ops->migratepage)
				return ret;
		}
	}
1111

1112 1113 1114
	if ((mode & ISOLATE_UNMAPPED) && page_mapped(page))
		return ret;

A
Andy Whitcroft 已提交
1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127
	if (likely(get_page_unless_zero(page))) {
		/*
		 * Be careful not to clear PageLRU until after we're
		 * sure the page is not being freed elsewhere -- the
		 * page release code relies on it.
		 */
		ClearPageLRU(page);
		ret = 0;
	}

	return ret;
}

L
Linus Torvalds 已提交
1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138
/*
 * zone->lru_lock is heavily contended.  Some of the functions that
 * shrink the lists perform better by taking out a batch of pages
 * and working on them outside the LRU lock.
 *
 * For pagecache intensive workloads, this function is the hottest
 * spot in the kernel (apart from copy_*_user functions).
 *
 * Appropriate locks must be held before calling this function.
 *
 * @nr_to_scan:	The number of pages to look through on the list.
H
Hugh Dickins 已提交
1139
 * @mz:		The mem_cgroup_zone to pull pages from.
L
Linus Torvalds 已提交
1140
 * @dst:	The temp list to put pages on to.
H
Hugh Dickins 已提交
1141
 * @nr_scanned:	The number of pages that were scanned.
A
Andy Whitcroft 已提交
1142 1143
 * @order:	The caller's attempted allocation order
 * @mode:	One of the LRU isolation modes
H
Hugh Dickins 已提交
1144
 * @active:	True [1] if isolating active pages
1145
 * @file:	True [1] if isolating file [!anon] pages
L
Linus Torvalds 已提交
1146 1147 1148
 *
 * returns how many pages were moved onto *@dst.
 */
1149
static unsigned long isolate_lru_pages(unsigned long nr_to_scan,
H
Hugh Dickins 已提交
1150 1151 1152
		struct mem_cgroup_zone *mz, struct list_head *dst,
		unsigned long *nr_scanned, int order, isolate_mode_t mode,
		int active, int file)
L
Linus Torvalds 已提交
1153
{
H
Hugh Dickins 已提交
1154 1155
	struct lruvec *lruvec;
	struct list_head *src;
1156
	unsigned long nr_taken = 0;
1157 1158 1159
	unsigned long nr_lumpy_taken = 0;
	unsigned long nr_lumpy_dirty = 0;
	unsigned long nr_lumpy_failed = 0;
1160
	unsigned long scan;
H
Hugh Dickins 已提交
1161 1162 1163 1164 1165 1166 1167 1168
	int lru = LRU_BASE;

	lruvec = mem_cgroup_zone_lruvec(mz->zone, mz->mem_cgroup);
	if (active)
		lru += LRU_ACTIVE;
	if (file)
		lru += LRU_FILE;
	src = &lruvec->lists[lru];
L
Linus Torvalds 已提交
1169

1170
	for (scan = 0; scan < nr_to_scan && !list_empty(src); scan++) {
A
Andy Whitcroft 已提交
1171 1172 1173 1174 1175 1176
		struct page *page;
		unsigned long pfn;
		unsigned long end_pfn;
		unsigned long page_pfn;
		int zone_id;

L
Linus Torvalds 已提交
1177 1178 1179
		page = lru_to_page(src);
		prefetchw_prev_lru_page(page, src, flags);

N
Nick Piggin 已提交
1180
		VM_BUG_ON(!PageLRU(page));
N
Nick Piggin 已提交
1181

1182
		switch (__isolate_lru_page(page, mode, file)) {
A
Andy Whitcroft 已提交
1183
		case 0:
1184
			mem_cgroup_lru_del(page);
A
Andy Whitcroft 已提交
1185
			list_move(&page->lru, dst);
1186
			nr_taken += hpage_nr_pages(page);
A
Andy Whitcroft 已提交
1187 1188 1189 1190 1191 1192
			break;

		case -EBUSY:
			/* else it is being freed elsewhere */
			list_move(&page->lru, src);
			continue;
1193

A
Andy Whitcroft 已提交
1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205
		default:
			BUG();
		}

		if (!order)
			continue;

		/*
		 * Attempt to take all pages in the order aligned region
		 * surrounding the tag page.  Only take those pages of
		 * the same active state as that tag page.  We may safely
		 * round the target page pfn down to the requested order
L
Lucas De Marchi 已提交
1206
		 * as the mem_map is guaranteed valid out to MAX_ORDER,
A
Andy Whitcroft 已提交
1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225
		 * where that page is in a different zone we will detect
		 * it from its zone id and abort this block scan.
		 */
		zone_id = page_zone_id(page);
		page_pfn = page_to_pfn(page);
		pfn = page_pfn & ~((1 << order) - 1);
		end_pfn = pfn + (1 << order);
		for (; pfn < end_pfn; pfn++) {
			struct page *cursor_page;

			/* The target page is in the block, ignore it. */
			if (unlikely(pfn == page_pfn))
				continue;

			/* Avoid holes within the zone. */
			if (unlikely(!pfn_valid_within(pfn)))
				break;

			cursor_page = pfn_to_page(pfn);
1226

A
Andy Whitcroft 已提交
1227 1228
			/* Check that we have not crossed a zone boundary. */
			if (unlikely(page_zone_id(cursor_page) != zone_id))
1229
				break;
1230 1231 1232 1233 1234 1235

			/*
			 * If we don't have enough swap space, reclaiming of
			 * anon page which don't already have a swap slot is
			 * pointless.
			 */
1236
			if (nr_swap_pages <= 0 && PageSwapBacked(cursor_page) &&
1237 1238
			    !PageSwapCache(cursor_page))
				break;
1239

1240
			if (__isolate_lru_page(cursor_page, mode, file) == 0) {
1241 1242
				unsigned int isolated_pages;

1243
				mem_cgroup_lru_del(cursor_page);
A
Andy Whitcroft 已提交
1244
				list_move(&cursor_page->lru, dst);
1245 1246 1247
				isolated_pages = hpage_nr_pages(cursor_page);
				nr_taken += isolated_pages;
				nr_lumpy_taken += isolated_pages;
1248
				if (PageDirty(cursor_page))
1249
					nr_lumpy_dirty += isolated_pages;
A
Andy Whitcroft 已提交
1250
				scan++;
1251
				pfn += isolated_pages - 1;
1252
			} else {
1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266
				/*
				 * Check if the page is freed already.
				 *
				 * We can't use page_count() as that
				 * requires compound_head and we don't
				 * have a pin on the page here. If a
				 * page is tail, we may or may not
				 * have isolated the head, so assume
				 * it's not free, it'd be tricky to
				 * track the head status without a
				 * page pin.
				 */
				if (!PageTail(cursor_page) &&
				    !atomic_read(&cursor_page->_count))
1267 1268
					continue;
				break;
A
Andy Whitcroft 已提交
1269 1270
			}
		}
1271 1272 1273 1274

		/* If we break out of the loop above, lumpy reclaim failed */
		if (pfn < end_pfn)
			nr_lumpy_failed++;
L
Linus Torvalds 已提交
1275 1276
	}

H
Hugh Dickins 已提交
1277
	*nr_scanned = scan;
1278 1279 1280 1281 1282

	trace_mm_vmscan_lru_isolate(order,
			nr_to_scan, scan,
			nr_taken,
			nr_lumpy_taken, nr_lumpy_dirty, nr_lumpy_failed,
1283
			mode, file);
L
Linus Torvalds 已提交
1284 1285 1286
	return nr_taken;
}

1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297
/**
 * isolate_lru_page - tries to isolate a page from its LRU list
 * @page: page to isolate from its LRU list
 *
 * Isolates a @page from an LRU list, clears PageLRU and adjusts the
 * vmstat statistic corresponding to whatever LRU list the page was on.
 *
 * Returns 0 if the page was removed from an LRU list.
 * Returns -EBUSY if the page was not on an LRU list.
 *
 * The returned page will have PageLRU() cleared.  If it was found on
L
Lee Schermerhorn 已提交
1298 1299 1300
 * the active list, it will have PageActive set.  If it was found on
 * the unevictable list, it will have the PageUnevictable bit set. That flag
 * may need to be cleared by the caller before letting the page go.
1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315
 *
 * The vmstat statistic corresponding to the list on which the page was
 * found will be decremented.
 *
 * Restrictions:
 * (1) Must be called with an elevated refcount on the page. This is a
 *     fundamentnal difference from isolate_lru_pages (which is called
 *     without a stable reference).
 * (2) the lru_lock must not be held.
 * (3) interrupts must be enabled.
 */
int isolate_lru_page(struct page *page)
{
	int ret = -EBUSY;

1316 1317
	VM_BUG_ON(!page_count(page));

1318 1319 1320 1321
	if (PageLRU(page)) {
		struct zone *zone = page_zone(page);

		spin_lock_irq(&zone->lru_lock);
1322
		if (PageLRU(page)) {
L
Lee Schermerhorn 已提交
1323
			int lru = page_lru(page);
1324
			ret = 0;
1325
			get_page(page);
1326
			ClearPageLRU(page);
1327 1328

			del_page_from_lru_list(zone, page, lru);
1329 1330 1331 1332 1333 1334
		}
		spin_unlock_irq(&zone->lru_lock);
	}
	return ret;
}

1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345
/*
 * Are there way too many processes in the direct reclaim path already?
 */
static int too_many_isolated(struct zone *zone, int file,
		struct scan_control *sc)
{
	unsigned long inactive, isolated;

	if (current_is_kswapd())
		return 0;

1346
	if (!global_reclaim(sc))
1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359
		return 0;

	if (file) {
		inactive = zone_page_state(zone, NR_INACTIVE_FILE);
		isolated = zone_page_state(zone, NR_ISOLATED_FILE);
	} else {
		inactive = zone_page_state(zone, NR_INACTIVE_ANON);
		isolated = zone_page_state(zone, NR_ISOLATED_ANON);
	}

	return isolated > inactive;
}

1360
static noinline_for_stack void
1361 1362
putback_inactive_pages(struct mem_cgroup_zone *mz,
		       struct list_head *page_list)
1363
{
1364
	struct zone_reclaim_stat *reclaim_stat = get_reclaim_stat(mz);
1365 1366
	struct zone *zone = mz->zone;
	LIST_HEAD(pages_to_free);
1367 1368 1369 1370 1371

	/*
	 * Put back any unfreeable pages.
	 */
	while (!list_empty(page_list)) {
1372
		struct page *page = lru_to_page(page_list);
1373
		int lru;
1374

1375 1376 1377 1378 1379 1380 1381 1382
		VM_BUG_ON(PageLRU(page));
		list_del(&page->lru);
		if (unlikely(!page_evictable(page, NULL))) {
			spin_unlock_irq(&zone->lru_lock);
			putback_lru_page(page);
			spin_lock_irq(&zone->lru_lock);
			continue;
		}
1383
		SetPageLRU(page);
1384
		lru = page_lru(page);
1385
		add_page_to_lru_list(zone, page, lru);
1386 1387
		if (is_active_lru(lru)) {
			int file = is_file_lru(lru);
1388 1389
			int numpages = hpage_nr_pages(page);
			reclaim_stat->recent_rotated[file] += numpages;
1390
		}
1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401
		if (put_page_testzero(page)) {
			__ClearPageLRU(page);
			__ClearPageActive(page);
			del_page_from_lru_list(zone, page, lru);

			if (unlikely(PageCompound(page))) {
				spin_unlock_irq(&zone->lru_lock);
				(*get_compound_page_dtor(page))(page);
				spin_lock_irq(&zone->lru_lock);
			} else
				list_add(&page->lru, &pages_to_free);
1402 1403 1404
		}
	}

1405 1406 1407 1408
	/*
	 * To save our caller's stack, now use input list for pages to free.
	 */
	list_splice(&pages_to_free, page_list);
1409 1410
}

1411 1412
static noinline_for_stack void
update_isolated_counts(struct mem_cgroup_zone *mz,
1413
		       struct list_head *page_list,
1414
		       unsigned long *nr_anon,
1415
		       unsigned long *nr_file)
1416
{
1417
	struct zone_reclaim_stat *reclaim_stat = get_reclaim_stat(mz);
1418
	struct zone *zone = mz->zone;
1419
	unsigned int count[NR_LRU_LISTS] = { 0, };
1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436
	unsigned long nr_active = 0;
	struct page *page;
	int lru;

	/*
	 * Count pages and clear active flags
	 */
	list_for_each_entry(page, page_list, lru) {
		int numpages = hpage_nr_pages(page);
		lru = page_lru_base_type(page);
		if (PageActive(page)) {
			lru += LRU_ACTIVE;
			ClearPageActive(page);
			nr_active += numpages;
		}
		count[lru] += numpages;
	}
1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455

	__count_vm_events(PGDEACTIVATE, nr_active);

	__mod_zone_page_state(zone, NR_ACTIVE_FILE,
			      -count[LRU_ACTIVE_FILE]);
	__mod_zone_page_state(zone, NR_INACTIVE_FILE,
			      -count[LRU_INACTIVE_FILE]);
	__mod_zone_page_state(zone, NR_ACTIVE_ANON,
			      -count[LRU_ACTIVE_ANON]);
	__mod_zone_page_state(zone, NR_INACTIVE_ANON,
			      -count[LRU_INACTIVE_ANON]);

	*nr_anon = count[LRU_ACTIVE_ANON] + count[LRU_INACTIVE_ANON];
	*nr_file = count[LRU_ACTIVE_FILE] + count[LRU_INACTIVE_FILE];

	reclaim_stat->recent_scanned[0] += *nr_anon;
	reclaim_stat->recent_scanned[1] += *nr_file;
}

1456
/*
1457
 * Returns true if a direct reclaim should wait on pages under writeback.
1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475
 *
 * If we are direct reclaiming for contiguous pages and we do not reclaim
 * everything in the list, try again and wait for writeback IO to complete.
 * This will stall high-order allocations noticeably. Only do that when really
 * need to free the pages under high memory pressure.
 */
static inline bool should_reclaim_stall(unsigned long nr_taken,
					unsigned long nr_freed,
					int priority,
					struct scan_control *sc)
{
	int lumpy_stall_priority;

	/* kswapd should not stall on sync IO */
	if (current_is_kswapd())
		return false;

	/* Only stall on lumpy reclaim */
1476
	if (sc->reclaim_mode & RECLAIM_MODE_SINGLE)
1477 1478
		return false;

1479
	/* If we have reclaimed everything on the isolated list, no stall */
1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496
	if (nr_freed == nr_taken)
		return false;

	/*
	 * For high-order allocations, there are two stall thresholds.
	 * High-cost allocations stall immediately where as lower
	 * order allocations such as stacks require the scanning
	 * priority to be much higher before stalling.
	 */
	if (sc->order > PAGE_ALLOC_COSTLY_ORDER)
		lumpy_stall_priority = DEF_PRIORITY;
	else
		lumpy_stall_priority = DEF_PRIORITY / 3;

	return priority <= lumpy_stall_priority;
}

L
Linus Torvalds 已提交
1497
/*
A
Andrew Morton 已提交
1498 1499
 * shrink_inactive_list() is a helper for shrink_zone().  It returns the number
 * of reclaimed pages
L
Linus Torvalds 已提交
1500
 */
1501
static noinline_for_stack unsigned long
1502 1503
shrink_inactive_list(unsigned long nr_to_scan, struct mem_cgroup_zone *mz,
		     struct scan_control *sc, int priority, int file)
L
Linus Torvalds 已提交
1504 1505
{
	LIST_HEAD(page_list);
1506
	unsigned long nr_scanned;
1507
	unsigned long nr_reclaimed = 0;
1508 1509 1510
	unsigned long nr_taken;
	unsigned long nr_anon;
	unsigned long nr_file;
1511 1512
	unsigned long nr_dirty = 0;
	unsigned long nr_writeback = 0;
1513
	isolate_mode_t reclaim_mode = ISOLATE_INACTIVE;
1514
	struct zone *zone = mz->zone;
1515

1516
	while (unlikely(too_many_isolated(zone, file, sc))) {
1517
		congestion_wait(BLK_RW_ASYNC, HZ/10);
1518 1519 1520 1521 1522 1523

		/* We are about to die and free our memory. Return now. */
		if (fatal_signal_pending(current))
			return SWAP_CLUSTER_MAX;
	}

1524
	set_reclaim_mode(priority, sc, false);
1525 1526 1527
	if (sc->reclaim_mode & RECLAIM_MODE_LUMPYRECLAIM)
		reclaim_mode |= ISOLATE_ACTIVE;

L
Linus Torvalds 已提交
1528
	lru_add_drain();
1529 1530 1531 1532 1533 1534

	if (!sc->may_unmap)
		reclaim_mode |= ISOLATE_UNMAPPED;
	if (!sc->may_writepage)
		reclaim_mode |= ISOLATE_CLEAN;

L
Linus Torvalds 已提交
1535
	spin_lock_irq(&zone->lru_lock);
1536

H
Hugh Dickins 已提交
1537 1538 1539
	nr_taken = isolate_lru_pages(nr_to_scan, mz, &page_list,
				     &nr_scanned, sc->order,
				     reclaim_mode, 0, file);
1540
	if (global_reclaim(sc)) {
1541 1542 1543 1544 1545 1546 1547 1548
		zone->pages_scanned += nr_scanned;
		if (current_is_kswapd())
			__count_zone_vm_events(PGSCAN_KSWAPD, zone,
					       nr_scanned);
		else
			__count_zone_vm_events(PGSCAN_DIRECT, zone,
					       nr_scanned);
	}
1549

1550 1551 1552 1553
	if (nr_taken == 0) {
		spin_unlock_irq(&zone->lru_lock);
		return 0;
	}
A
Andy Whitcroft 已提交
1554

1555 1556 1557 1558
	update_isolated_counts(mz, &page_list, &nr_anon, &nr_file);

	__mod_zone_page_state(zone, NR_ISOLATED_ANON, nr_anon);
	__mod_zone_page_state(zone, NR_ISOLATED_FILE, nr_file);
L
Linus Torvalds 已提交
1559

1560
	spin_unlock_irq(&zone->lru_lock);
1561

1562
	nr_reclaimed = shrink_page_list(&page_list, mz, sc, priority,
1563
						&nr_dirty, &nr_writeback);
1564

1565 1566
	/* Check if we should syncronously wait for writeback */
	if (should_reclaim_stall(nr_taken, nr_reclaimed, priority, sc)) {
1567
		set_reclaim_mode(priority, sc, true);
1568
		nr_reclaimed += shrink_page_list(&page_list, mz, sc,
1569
					priority, &nr_dirty, &nr_writeback);
1570
	}
1571

1572 1573
	spin_lock_irq(&zone->lru_lock);

1574 1575 1576
	if (current_is_kswapd())
		__count_vm_events(KSWAPD_STEAL, nr_reclaimed);
	__count_zone_vm_events(PGSTEAL, zone, nr_reclaimed);
N
Nick Piggin 已提交
1577

1578 1579 1580 1581 1582 1583 1584 1585
	putback_inactive_pages(mz, &page_list);

	__mod_zone_page_state(zone, NR_ISOLATED_ANON, -nr_anon);
	__mod_zone_page_state(zone, NR_ISOLATED_FILE, -nr_file);

	spin_unlock_irq(&zone->lru_lock);

	free_hot_cold_page_list(&page_list, 1);
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
	/*
	 * If reclaim is isolating dirty pages under writeback, it implies
	 * that the long-lived page allocation rate is exceeding the page
	 * laundering rate. Either the global limits are not being effective
	 * at throttling processes due to the page distribution throughout
	 * zones or there is heavy usage of a slow backing device. The
	 * only option is to throttle from reclaim context which is not ideal
	 * as there is no guarantee the dirtying process is throttled in the
	 * same way balance_dirty_pages() manages.
	 *
	 * This scales the number of dirty pages that must be under writeback
	 * before throttling depending on priority. It is a simple backoff
	 * function that has the most effect in the range DEF_PRIORITY to
	 * DEF_PRIORITY-2 which is the priority reclaim is considered to be
	 * in trouble and reclaim is considered to be in trouble.
	 *
	 * DEF_PRIORITY   100% isolated pages must be PageWriteback to throttle
	 * DEF_PRIORITY-1  50% must be PageWriteback
	 * DEF_PRIORITY-2  25% must be PageWriteback, kswapd in trouble
	 * ...
	 * DEF_PRIORITY-6 For SWAP_CLUSTER_MAX isolated pages, throttle if any
	 *                     isolated page is PageWriteback
	 */
	if (nr_writeback && nr_writeback >= (nr_taken >> (DEF_PRIORITY-priority)))
		wait_iff_congested(zone, BLK_RW_ASYNC, HZ/10);

1613 1614 1615 1616
	trace_mm_vmscan_lru_shrink_inactive(zone->zone_pgdat->node_id,
		zone_idx(zone),
		nr_scanned, nr_reclaimed,
		priority,
1617
		trace_shrink_flags(file, sc->reclaim_mode));
1618
	return nr_reclaimed;
L
Linus Torvalds 已提交
1619 1620 1621 1622 1623 1624 1625 1626 1627 1628 1629 1630 1631 1632 1633 1634 1635 1636 1637
}

/*
 * This moves pages from the active list to the inactive list.
 *
 * We move them the other way if the page is referenced by one or more
 * processes, from rmap.
 *
 * If the pages are mostly unmapped, the processing is fast and it is
 * appropriate to hold zone->lru_lock across the whole operation.  But if
 * the pages are mapped, the processing is slow (page_referenced()) so we
 * should drop zone->lru_lock around each page.  It's impossible to balance
 * this, so instead we remove the pages from the LRU while processing them.
 * It is safe to rely on PG_active against the non-LRU pages in here because
 * nobody will play with that bit on a non-LRU page.
 *
 * The downside is that we have to touch page->_count against each page.
 * But we had to alter page->flags anyway.
 */
1638

1639 1640
static void move_active_pages_to_lru(struct zone *zone,
				     struct list_head *list,
1641
				     struct list_head *pages_to_free,
1642 1643 1644 1645 1646
				     enum lru_list lru)
{
	unsigned long pgmoved = 0;
	struct page *page;

1647 1648 1649 1650 1651 1652 1653 1654 1655 1656 1657
	if (buffer_heads_over_limit) {
		spin_unlock_irq(&zone->lru_lock);
		list_for_each_entry(page, list, lru) {
			if (page_has_private(page) && trylock_page(page)) {
				if (page_has_private(page))
					try_to_release_page(page, 0);
				unlock_page(page);
			}
		}
		spin_lock_irq(&zone->lru_lock);
	}
1658 1659

	while (!list_empty(list)) {
1660 1661
		struct lruvec *lruvec;

1662 1663 1664 1665 1666
		page = lru_to_page(list);

		VM_BUG_ON(PageLRU(page));
		SetPageLRU(page);

1667 1668
		lruvec = mem_cgroup_lru_add_list(zone, page, lru);
		list_move(&page->lru, &lruvec->lists[lru]);
1669
		pgmoved += hpage_nr_pages(page);
1670

1671 1672 1673 1674 1675 1676 1677 1678 1679 1680 1681
		if (put_page_testzero(page)) {
			__ClearPageLRU(page);
			__ClearPageActive(page);
			del_page_from_lru_list(zone, page, lru);

			if (unlikely(PageCompound(page))) {
				spin_unlock_irq(&zone->lru_lock);
				(*get_compound_page_dtor(page))(page);
				spin_lock_irq(&zone->lru_lock);
			} else
				list_add(&page->lru, pages_to_free);
1682 1683 1684 1685 1686 1687
		}
	}
	__mod_zone_page_state(zone, NR_LRU_BASE + lru, pgmoved);
	if (!is_active_lru(lru))
		__count_vm_events(PGDEACTIVATE, pgmoved);
}
1688

H
Hugh Dickins 已提交
1689
static void shrink_active_list(unsigned long nr_to_scan,
1690 1691 1692
			       struct mem_cgroup_zone *mz,
			       struct scan_control *sc,
			       int priority, int file)
L
Linus Torvalds 已提交
1693
{
1694
	unsigned long nr_taken;
H
Hugh Dickins 已提交
1695
	unsigned long nr_scanned;
1696
	unsigned long vm_flags;
L
Linus Torvalds 已提交
1697
	LIST_HEAD(l_hold);	/* The pages which were snipped off */
1698
	LIST_HEAD(l_active);
1699
	LIST_HEAD(l_inactive);
L
Linus Torvalds 已提交
1700
	struct page *page;
1701
	struct zone_reclaim_stat *reclaim_stat = get_reclaim_stat(mz);
1702
	unsigned long nr_rotated = 0;
1703
	isolate_mode_t reclaim_mode = ISOLATE_ACTIVE;
1704
	struct zone *zone = mz->zone;
L
Linus Torvalds 已提交
1705 1706

	lru_add_drain();
1707 1708 1709 1710 1711 1712

	if (!sc->may_unmap)
		reclaim_mode |= ISOLATE_UNMAPPED;
	if (!sc->may_writepage)
		reclaim_mode |= ISOLATE_CLEAN;

L
Linus Torvalds 已提交
1713
	spin_lock_irq(&zone->lru_lock);
1714

H
Hugh Dickins 已提交
1715 1716 1717
	nr_taken = isolate_lru_pages(nr_to_scan, mz, &l_hold,
				     &nr_scanned, sc->order,
				     reclaim_mode, 1, file);
1718
	if (global_reclaim(sc))
H
Hugh Dickins 已提交
1719
		zone->pages_scanned += nr_scanned;
1720

1721
	reclaim_stat->recent_scanned[file] += nr_taken;
1722

H
Hugh Dickins 已提交
1723
	__count_zone_vm_events(PGREFILL, zone, nr_scanned);
1724
	if (file)
1725
		__mod_zone_page_state(zone, NR_ACTIVE_FILE, -nr_taken);
1726
	else
1727
		__mod_zone_page_state(zone, NR_ACTIVE_ANON, -nr_taken);
K
KOSAKI Motohiro 已提交
1728
	__mod_zone_page_state(zone, NR_ISOLATED_ANON + file, nr_taken);
L
Linus Torvalds 已提交
1729 1730 1731 1732 1733 1734
	spin_unlock_irq(&zone->lru_lock);

	while (!list_empty(&l_hold)) {
		cond_resched();
		page = lru_to_page(&l_hold);
		list_del(&page->lru);
1735

L
Lee Schermerhorn 已提交
1736 1737 1738 1739 1740
		if (unlikely(!page_evictable(page, NULL))) {
			putback_lru_page(page);
			continue;
		}

1741
		if (page_referenced(page, 0, mz->mem_cgroup, &vm_flags)) {
1742
			nr_rotated += hpage_nr_pages(page);
1743 1744 1745 1746 1747 1748 1749 1750 1751
			/*
			 * Identify referenced, file-backed active pages and
			 * give them one more trip around the active list. So
			 * that executable code get better chances to stay in
			 * memory under moderate memory pressure.  Anon pages
			 * are not likely to be evicted by use-once streaming
			 * IO, plus JVM can create lots of anon VM_EXEC pages,
			 * so we ignore them here.
			 */
1752
			if ((vm_flags & VM_EXEC) && page_is_file_cache(page)) {
1753 1754 1755 1756
				list_add(&page->lru, &l_active);
				continue;
			}
		}
1757

1758
		ClearPageActive(page);	/* we are de-activating */
L
Linus Torvalds 已提交
1759 1760 1761
		list_add(&page->lru, &l_inactive);
	}

1762
	/*
1763
	 * Move pages back to the lru list.
1764
	 */
1765
	spin_lock_irq(&zone->lru_lock);
1766
	/*
1767 1768 1769 1770
	 * Count referenced pages from currently used mappings as rotated,
	 * even though only some of them are actually re-activated.  This
	 * helps balance scan pressure between file and anonymous pages in
	 * get_scan_ratio.
1771
	 */
1772
	reclaim_stat->recent_rotated[file] += nr_rotated;
1773

1774
	move_active_pages_to_lru(zone, &l_active, &l_hold,
1775
						LRU_ACTIVE + file * LRU_FILE);
1776
	move_active_pages_to_lru(zone, &l_inactive, &l_hold,
1777
						LRU_BASE   + file * LRU_FILE);
K
KOSAKI Motohiro 已提交
1778
	__mod_zone_page_state(zone, NR_ISOLATED_ANON + file, -nr_taken);
1779
	spin_unlock_irq(&zone->lru_lock);
1780 1781

	free_hot_cold_page_list(&l_hold, 1);
L
Linus Torvalds 已提交
1782 1783
}

1784
#ifdef CONFIG_SWAP
1785
static int inactive_anon_is_low_global(struct zone *zone)
1786 1787 1788 1789 1790 1791 1792 1793 1794 1795 1796 1797
{
	unsigned long active, inactive;

	active = zone_page_state(zone, NR_ACTIVE_ANON);
	inactive = zone_page_state(zone, NR_INACTIVE_ANON);

	if (inactive * zone->inactive_ratio < active)
		return 1;

	return 0;
}

1798 1799 1800 1801 1802 1803 1804 1805
/**
 * inactive_anon_is_low - check if anonymous pages need to be deactivated
 * @zone: zone to check
 * @sc:   scan control of this context
 *
 * Returns true if the zone does not have enough inactive anon pages,
 * meaning some active anon pages need to be deactivated.
 */
1806
static int inactive_anon_is_low(struct mem_cgroup_zone *mz)
1807
{
1808 1809 1810 1811 1812 1813 1814
	/*
	 * If we don't have swap space, anonymous page deactivation
	 * is pointless.
	 */
	if (!total_swap_pages)
		return 0;

1815 1816 1817 1818 1819
	if (!scanning_global_lru(mz))
		return mem_cgroup_inactive_anon_is_low(mz->mem_cgroup,
						       mz->zone);

	return inactive_anon_is_low_global(mz->zone);
1820
}
1821
#else
1822
static inline int inactive_anon_is_low(struct mem_cgroup_zone *mz)
1823 1824 1825 1826
{
	return 0;
}
#endif
1827

1828 1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839
static int inactive_file_is_low_global(struct zone *zone)
{
	unsigned long active, inactive;

	active = zone_page_state(zone, NR_ACTIVE_FILE);
	inactive = zone_page_state(zone, NR_INACTIVE_FILE);

	return (active > inactive);
}

/**
 * inactive_file_is_low - check if file pages need to be deactivated
1840
 * @mz: memory cgroup and zone to check
1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851
 *
 * When the system is doing streaming IO, memory pressure here
 * ensures that active file pages get deactivated, until more
 * than half of the file pages are on the inactive list.
 *
 * Once we get to that situation, protect the system's working
 * set from being evicted by disabling active file page aging.
 *
 * This uses a different ratio than the anonymous pages, because
 * the page cache uses a use-once replacement algorithm.
 */
1852
static int inactive_file_is_low(struct mem_cgroup_zone *mz)
1853
{
1854 1855 1856
	if (!scanning_global_lru(mz))
		return mem_cgroup_inactive_file_is_low(mz->mem_cgroup,
						       mz->zone);
1857

1858
	return inactive_file_is_low_global(mz->zone);
1859 1860
}

1861
static int inactive_list_is_low(struct mem_cgroup_zone *mz, int file)
1862 1863
{
	if (file)
1864
		return inactive_file_is_low(mz);
1865
	else
1866
		return inactive_anon_is_low(mz);
1867 1868
}

1869
static unsigned long shrink_list(enum lru_list lru, unsigned long nr_to_scan,
1870 1871
				 struct mem_cgroup_zone *mz,
				 struct scan_control *sc, int priority)
1872
{
1873 1874
	int file = is_file_lru(lru);

1875
	if (is_active_lru(lru)) {
1876 1877
		if (inactive_list_is_low(mz, file))
			shrink_active_list(nr_to_scan, mz, sc, priority, file);
1878 1879 1880
		return 0;
	}

1881
	return shrink_inactive_list(nr_to_scan, mz, sc, priority, file);
1882 1883
}

1884 1885
static int vmscan_swappiness(struct mem_cgroup_zone *mz,
			     struct scan_control *sc)
1886
{
1887
	if (global_reclaim(sc))
1888
		return vm_swappiness;
1889
	return mem_cgroup_swappiness(mz->mem_cgroup);
1890 1891
}

1892 1893 1894 1895 1896 1897
/*
 * Determine how aggressively the anon and file LRU lists should be
 * scanned.  The relative value of each set of LRU lists is determined
 * by looking at the fraction of the pages scanned we did rotate back
 * onto the active list instead of evict.
 *
1898
 * nr[0] = anon pages to scan; nr[1] = file pages to scan
1899
 */
1900 1901
static void get_scan_count(struct mem_cgroup_zone *mz, struct scan_control *sc,
			   unsigned long *nr, int priority)
1902 1903 1904 1905
{
	unsigned long anon, file, free;
	unsigned long anon_prio, file_prio;
	unsigned long ap, fp;
1906
	struct zone_reclaim_stat *reclaim_stat = get_reclaim_stat(mz);
1907
	u64 fraction[2], denominator;
H
Hugh Dickins 已提交
1908
	enum lru_list lru;
1909
	int noswap = 0;
1910
	bool force_scan = false;
1911

1912 1913 1914 1915 1916 1917 1918 1919 1920 1921
	/*
	 * If the zone or memcg is small, nr[l] can be 0.  This
	 * results in no scanning on this priority and a potential
	 * priority drop.  Global direct reclaim can go to the next
	 * zone and tends to have no problems. Global kswapd is for
	 * zone balancing and it needs to scan a minimum amount. When
	 * reclaiming for a memcg, a priority drop can cause high
	 * latencies, so it's better to scan a minimum amount there as
	 * well.
	 */
1922
	if (current_is_kswapd() && mz->zone->all_unreclaimable)
1923
		force_scan = true;
1924
	if (!global_reclaim(sc))
1925
		force_scan = true;
1926 1927 1928 1929 1930 1931 1932 1933 1934

	/* If we have no swap space, do not bother scanning anon pages. */
	if (!sc->may_swap || (nr_swap_pages <= 0)) {
		noswap = 1;
		fraction[0] = 0;
		fraction[1] = 1;
		denominator = 1;
		goto out;
	}
1935

1936 1937 1938 1939
	anon  = zone_nr_lru_pages(mz, LRU_ACTIVE_ANON) +
		zone_nr_lru_pages(mz, LRU_INACTIVE_ANON);
	file  = zone_nr_lru_pages(mz, LRU_ACTIVE_FILE) +
		zone_nr_lru_pages(mz, LRU_INACTIVE_FILE);
1940

1941
	if (global_reclaim(sc)) {
1942
		free  = zone_page_state(mz->zone, NR_FREE_PAGES);
1943 1944
		/* If we have very few page cache pages,
		   force-scan anon pages. */
1945
		if (unlikely(file + free <= high_wmark_pages(mz->zone))) {
1946 1947 1948 1949
			fraction[0] = 1;
			fraction[1] = 0;
			denominator = 1;
			goto out;
1950
		}
1951 1952
	}

1953 1954 1955 1956
	/*
	 * With swappiness at 100, anonymous and file have the same priority.
	 * This scanning priority is essentially the inverse of IO cost.
	 */
1957 1958
	anon_prio = vmscan_swappiness(mz, sc);
	file_prio = 200 - vmscan_swappiness(mz, sc);
1959

1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970
	/*
	 * OK, so we have swap space and a fair amount of page cache
	 * pages.  We use the recently rotated / recently scanned
	 * ratios to determine how valuable each cache is.
	 *
	 * Because workloads change over time (and to avoid overflow)
	 * we keep these statistics as a floating average, which ends
	 * up weighing recent references more than old ones.
	 *
	 * anon in [0], file in [1]
	 */
1971
	spin_lock_irq(&mz->zone->lru_lock);
1972 1973 1974
	if (unlikely(reclaim_stat->recent_scanned[0] > anon / 4)) {
		reclaim_stat->recent_scanned[0] /= 2;
		reclaim_stat->recent_rotated[0] /= 2;
1975 1976
	}

1977 1978 1979
	if (unlikely(reclaim_stat->recent_scanned[1] > file / 4)) {
		reclaim_stat->recent_scanned[1] /= 2;
		reclaim_stat->recent_rotated[1] /= 2;
1980 1981 1982
	}

	/*
1983 1984 1985
	 * The amount of pressure on anon vs file pages is inversely
	 * proportional to the fraction of recently scanned pages on
	 * each list that were recently referenced and in active use.
1986
	 */
1987 1988
	ap = (anon_prio + 1) * (reclaim_stat->recent_scanned[0] + 1);
	ap /= reclaim_stat->recent_rotated[0] + 1;
1989

1990 1991
	fp = (file_prio + 1) * (reclaim_stat->recent_scanned[1] + 1);
	fp /= reclaim_stat->recent_rotated[1] + 1;
1992
	spin_unlock_irq(&mz->zone->lru_lock);
1993

1994 1995 1996 1997
	fraction[0] = ap;
	fraction[1] = fp;
	denominator = ap + fp + 1;
out:
H
Hugh Dickins 已提交
1998 1999
	for_each_evictable_lru(lru) {
		int file = is_file_lru(lru);
2000
		unsigned long scan;
2001

H
Hugh Dickins 已提交
2002
		scan = zone_nr_lru_pages(mz, lru);
2003 2004
		if (priority || noswap) {
			scan >>= priority;
2005 2006
			if (!scan && force_scan)
				scan = SWAP_CLUSTER_MAX;
2007 2008
			scan = div64_u64(scan * fraction[file], denominator);
		}
H
Hugh Dickins 已提交
2009
		nr[lru] = scan;
2010
	}
2011
}
2012

2013 2014 2015 2016 2017 2018 2019
/*
 * Reclaim/compaction depends on a number of pages being freed. To avoid
 * disruption to the system, a small number of order-0 pages continue to be
 * rotated and reclaimed in the normal fashion. However, by the time we get
 * back to the allocator and call try_to_compact_zone(), we ensure that
 * there are enough free pages for it to be likely successful
 */
2020
static inline bool should_continue_reclaim(struct mem_cgroup_zone *mz,
2021 2022 2023 2024 2025 2026 2027 2028
					unsigned long nr_reclaimed,
					unsigned long nr_scanned,
					struct scan_control *sc)
{
	unsigned long pages_for_compaction;
	unsigned long inactive_lru_pages;

	/* If not in reclaim/compaction mode, stop */
2029
	if (!(sc->reclaim_mode & RECLAIM_MODE_COMPACTION))
2030 2031
		return false;

2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046 2047 2048 2049 2050 2051 2052 2053
	/* Consider stopping depending on scan and reclaim activity */
	if (sc->gfp_mask & __GFP_REPEAT) {
		/*
		 * For __GFP_REPEAT allocations, stop reclaiming if the
		 * full LRU list has been scanned and we are still failing
		 * to reclaim pages. This full LRU scan is potentially
		 * expensive but a __GFP_REPEAT caller really wants to succeed
		 */
		if (!nr_reclaimed && !nr_scanned)
			return false;
	} else {
		/*
		 * For non-__GFP_REPEAT allocations which can presumably
		 * fail without consequence, stop if we failed to reclaim
		 * any pages from the last SWAP_CLUSTER_MAX number of
		 * pages that were scanned. This will return to the
		 * caller faster at the risk reclaim/compaction and
		 * the resulting allocation attempt fails
		 */
		if (!nr_reclaimed)
			return false;
	}
2054 2055 2056 2057 2058 2059

	/*
	 * If we have not reclaimed enough pages for compaction and the
	 * inactive lists are large enough, continue reclaiming
	 */
	pages_for_compaction = (2UL << sc->order);
2060
	inactive_lru_pages = zone_nr_lru_pages(mz, LRU_INACTIVE_FILE);
2061
	if (nr_swap_pages > 0)
2062
		inactive_lru_pages += zone_nr_lru_pages(mz, LRU_INACTIVE_ANON);
2063 2064 2065 2066 2067
	if (sc->nr_reclaimed < pages_for_compaction &&
			inactive_lru_pages > pages_for_compaction)
		return true;

	/* If compaction would go ahead or the allocation would succeed, stop */
2068
	switch (compaction_suitable(mz->zone, sc->order)) {
2069 2070 2071 2072 2073 2074 2075 2076
	case COMPACT_PARTIAL:
	case COMPACT_CONTINUE:
		return false;
	default:
		return true;
	}
}

L
Linus Torvalds 已提交
2077 2078 2079
/*
 * This is a basic per-zone page freer.  Used by both kswapd and direct reclaim.
 */
2080 2081
static void shrink_mem_cgroup_zone(int priority, struct mem_cgroup_zone *mz,
				   struct scan_control *sc)
L
Linus Torvalds 已提交
2082
{
2083
	unsigned long nr[NR_LRU_LISTS];
2084
	unsigned long nr_to_scan;
H
Hugh Dickins 已提交
2085
	enum lru_list lru;
2086
	unsigned long nr_reclaimed, nr_scanned;
2087
	unsigned long nr_to_reclaim = sc->nr_to_reclaim;
2088
	struct blk_plug plug;
2089

2090 2091
restart:
	nr_reclaimed = 0;
2092
	nr_scanned = sc->nr_scanned;
2093
	get_scan_count(mz, sc, nr, priority);
L
Linus Torvalds 已提交
2094

2095
	blk_start_plug(&plug);
2096 2097
	while (nr[LRU_INACTIVE_ANON] || nr[LRU_ACTIVE_FILE] ||
					nr[LRU_INACTIVE_FILE]) {
H
Hugh Dickins 已提交
2098 2099
		for_each_evictable_lru(lru) {
			if (nr[lru]) {
K
KOSAKI Motohiro 已提交
2100
				nr_to_scan = min_t(unsigned long,
H
Hugh Dickins 已提交
2101 2102
						   nr[lru], SWAP_CLUSTER_MAX);
				nr[lru] -= nr_to_scan;
L
Linus Torvalds 已提交
2103

H
Hugh Dickins 已提交
2104
				nr_reclaimed += shrink_list(lru, nr_to_scan,
2105
							    mz, sc, priority);
2106
			}
L
Linus Torvalds 已提交
2107
		}
2108 2109 2110 2111 2112 2113 2114 2115
		/*
		 * On large memory systems, scan >> priority can become
		 * really large. This is fine for the starting priority;
		 * we want to put equal scanning pressure on each zone.
		 * However, if the VM has a harder time of freeing pages,
		 * with multiple processes reclaiming pages, the total
		 * freeing target can get unreasonably large.
		 */
2116
		if (nr_reclaimed >= nr_to_reclaim && priority < DEF_PRIORITY)
2117
			break;
L
Linus Torvalds 已提交
2118
	}
2119
	blk_finish_plug(&plug);
2120
	sc->nr_reclaimed += nr_reclaimed;
2121

2122 2123 2124 2125
	/*
	 * Even if we did not try to evict anon pages at all, we want to
	 * rebalance the anon lru active/inactive ratio.
	 */
2126 2127
	if (inactive_anon_is_low(mz))
		shrink_active_list(SWAP_CLUSTER_MAX, mz, sc, priority, 0);
2128

2129
	/* reclaim/compaction might need reclaim to continue */
2130
	if (should_continue_reclaim(mz, nr_reclaimed,
2131 2132 2133
					sc->nr_scanned - nr_scanned, sc))
		goto restart;

2134
	throttle_vm_writeout(sc->gfp_mask);
L
Linus Torvalds 已提交
2135 2136
}

2137 2138 2139
static void shrink_zone(int priority, struct zone *zone,
			struct scan_control *sc)
{
2140 2141
	struct mem_cgroup *root = sc->target_mem_cgroup;
	struct mem_cgroup_reclaim_cookie reclaim = {
2142
		.zone = zone,
2143
		.priority = priority,
2144
	};
2145 2146 2147 2148 2149 2150 2151 2152
	struct mem_cgroup *memcg;

	memcg = mem_cgroup_iter(root, NULL, &reclaim);
	do {
		struct mem_cgroup_zone mz = {
			.mem_cgroup = memcg,
			.zone = zone,
		};
2153

2154 2155 2156 2157 2158 2159
		shrink_mem_cgroup_zone(priority, &mz, sc);
		/*
		 * Limit reclaim has historically picked one memcg and
		 * scanned it with decreasing priority levels until
		 * nr_to_reclaim had been reclaimed.  This priority
		 * cycle is thus over after a single memcg.
2160 2161 2162 2163
		 *
		 * Direct reclaim and kswapd, on the other hand, have
		 * to scan all memory cgroups to fulfill the overall
		 * scan target for the zone.
2164 2165 2166 2167 2168 2169 2170
		 */
		if (!global_reclaim(sc)) {
			mem_cgroup_iter_break(root, memcg);
			break;
		}
		memcg = mem_cgroup_iter(root, memcg, &reclaim);
	} while (memcg);
2171 2172
}

2173 2174 2175 2176 2177 2178 2179 2180 2181 2182 2183 2184 2185 2186 2187 2188 2189 2190 2191 2192 2193 2194 2195 2196 2197 2198 2199 2200 2201 2202 2203 2204 2205 2206 2207 2208
/* Returns true if compaction should go ahead for a high-order request */
static inline bool compaction_ready(struct zone *zone, struct scan_control *sc)
{
	unsigned long balance_gap, watermark;
	bool watermark_ok;

	/* Do not consider compaction for orders reclaim is meant to satisfy */
	if (sc->order <= PAGE_ALLOC_COSTLY_ORDER)
		return false;

	/*
	 * Compaction takes time to run and there are potentially other
	 * callers using the pages just freed. Continue reclaiming until
	 * there is a buffer of free pages available to give compaction
	 * a reasonable chance of completing and allocating the page
	 */
	balance_gap = min(low_wmark_pages(zone),
		(zone->present_pages + KSWAPD_ZONE_BALANCE_GAP_RATIO-1) /
			KSWAPD_ZONE_BALANCE_GAP_RATIO);
	watermark = high_wmark_pages(zone) + balance_gap + (2UL << sc->order);
	watermark_ok = zone_watermark_ok_safe(zone, 0, watermark, 0, 0);

	/*
	 * If compaction is deferred, reclaim up to a point where
	 * compaction will have a chance of success when re-enabled
	 */
	if (compaction_deferred(zone))
		return watermark_ok;

	/* If compaction is not ready to start, keep reclaiming */
	if (!compaction_suitable(zone, sc->order))
		return false;

	return watermark_ok;
}

L
Linus Torvalds 已提交
2209 2210 2211 2212 2213
/*
 * This is the direct reclaim path, for page-allocating processes.  We only
 * try to reclaim pages from zones which will satisfy the caller's allocation
 * request.
 *
2214 2215
 * We reclaim from a zone even if that zone is over high_wmark_pages(zone).
 * Because:
L
Linus Torvalds 已提交
2216 2217
 * a) The caller may be trying to free *extra* pages to satisfy a higher-order
 *    allocation or
2218 2219 2220
 * b) The target zone may be at high_wmark_pages(zone) but the lower zones
 *    must go *over* high_wmark_pages(zone) to satisfy the `incremental min'
 *    zone defense algorithm.
L
Linus Torvalds 已提交
2221 2222 2223
 *
 * If a zone is deemed to be full of pinned pages then just give it a light
 * scan then give up on it.
2224 2225
 *
 * This function returns true if a zone is being reclaimed for a costly
2226
 * high-order allocation and compaction is ready to begin. This indicates to
2227 2228
 * the caller that it should consider retrying the allocation instead of
 * further reclaim.
L
Linus Torvalds 已提交
2229
 */
2230
static bool shrink_zones(int priority, struct zonelist *zonelist,
2231
					struct scan_control *sc)
L
Linus Torvalds 已提交
2232
{
2233
	struct zoneref *z;
2234
	struct zone *zone;
2235 2236
	unsigned long nr_soft_reclaimed;
	unsigned long nr_soft_scanned;
2237
	bool aborted_reclaim = false;
2238

2239 2240
	for_each_zone_zonelist_nodemask(zone, z, zonelist,
					gfp_zone(sc->gfp_mask), sc->nodemask) {
2241
		if (!populated_zone(zone))
L
Linus Torvalds 已提交
2242
			continue;
2243 2244 2245 2246
		/*
		 * Take care memory controller reclaiming has small influence
		 * to global LRU.
		 */
2247
		if (global_reclaim(sc)) {
2248 2249
			if (!cpuset_zone_allowed_hardwall(zone, GFP_KERNEL))
				continue;
2250
			if (zone->all_unreclaimable && priority != DEF_PRIORITY)
2251
				continue;	/* Let kswapd poll it */
2252 2253
			if (COMPACTION_BUILD) {
				/*
2254 2255 2256 2257 2258 2259 2260
				 * If we already have plenty of memory free for
				 * compaction in this zone, don't free any more.
				 * Even though compaction is invoked for any
				 * non-zero order, only frequent costly order
				 * reclamation is disruptive enough to become a
				 * noticable problem, like transparent huge page
				 * allocations.
2261
				 */
2262
				if (compaction_ready(zone, sc)) {
2263
					aborted_reclaim = true;
2264
					continue;
2265
				}
2266
			}
2267 2268 2269 2270 2271 2272 2273 2274 2275 2276 2277 2278 2279
			/*
			 * This steals pages from memory cgroups over softlimit
			 * and returns the number of reclaimed pages and
			 * scanned pages. This works for global memory pressure
			 * and balancing, not for a memcg's limit.
			 */
			nr_soft_scanned = 0;
			nr_soft_reclaimed = mem_cgroup_soft_limit_reclaim(zone,
						sc->order, sc->gfp_mask,
						&nr_soft_scanned);
			sc->nr_reclaimed += nr_soft_reclaimed;
			sc->nr_scanned += nr_soft_scanned;
			/* need some check for avoid more shrink_zone() */
2280
		}
2281

2282
		shrink_zone(priority, zone, sc);
L
Linus Torvalds 已提交
2283
	}
2284

2285
	return aborted_reclaim;
2286 2287 2288 2289 2290 2291 2292
}

static bool zone_reclaimable(struct zone *zone)
{
	return zone->pages_scanned < zone_reclaimable_pages(zone) * 6;
}

2293
/* All zones in zonelist are unreclaimable? */
2294 2295 2296 2297 2298 2299 2300 2301 2302 2303 2304 2305
static bool all_unreclaimable(struct zonelist *zonelist,
		struct scan_control *sc)
{
	struct zoneref *z;
	struct zone *zone;

	for_each_zone_zonelist_nodemask(zone, z, zonelist,
			gfp_zone(sc->gfp_mask), sc->nodemask) {
		if (!populated_zone(zone))
			continue;
		if (!cpuset_zone_allowed_hardwall(zone, GFP_KERNEL))
			continue;
2306 2307
		if (!zone->all_unreclaimable)
			return false;
2308 2309
	}

2310
	return true;
L
Linus Torvalds 已提交
2311
}
2312

L
Linus Torvalds 已提交
2313 2314 2315 2316 2317 2318 2319 2320
/*
 * This is the main entry point to direct page reclaim.
 *
 * If a full scan of the inactive list fails to free enough memory then we
 * are "out of memory" and something needs to be killed.
 *
 * If the caller is !__GFP_FS then the probability of a failure is reasonably
 * high - the zone may be full of dirty or under-writeback pages, which this
2321 2322 2323 2324
 * caller can't do much about.  We kick the writeback threads and take explicit
 * naps in the hope that some of these pages can be written.  But if the
 * allocating task holds filesystem locks which prevent writeout this might not
 * work, and the allocation attempt will fail.
2325 2326 2327
 *
 * returns:	0, if no pages reclaimed
 * 		else, the number of pages reclaimed
L
Linus Torvalds 已提交
2328
 */
2329
static unsigned long do_try_to_free_pages(struct zonelist *zonelist,
2330 2331
					struct scan_control *sc,
					struct shrink_control *shrink)
L
Linus Torvalds 已提交
2332 2333
{
	int priority;
2334
	unsigned long total_scanned = 0;
L
Linus Torvalds 已提交
2335
	struct reclaim_state *reclaim_state = current->reclaim_state;
2336
	struct zoneref *z;
2337
	struct zone *zone;
2338
	unsigned long writeback_threshold;
2339
	bool aborted_reclaim;
L
Linus Torvalds 已提交
2340

2341
	get_mems_allowed();
2342 2343
	delayacct_freepages_start();

2344
	if (global_reclaim(sc))
2345
		count_vm_event(ALLOCSTALL);
L
Linus Torvalds 已提交
2346 2347

	for (priority = DEF_PRIORITY; priority >= 0; priority--) {
2348
		sc->nr_scanned = 0;
2349
		if (!priority)
2350
			disable_swap_token(sc->target_mem_cgroup);
2351
		aborted_reclaim = shrink_zones(priority, zonelist, sc);
2352

2353 2354 2355 2356
		/*
		 * Don't shrink slabs when reclaiming memory from
		 * over limit cgroups
		 */
2357
		if (global_reclaim(sc)) {
2358
			unsigned long lru_pages = 0;
2359 2360
			for_each_zone_zonelist(zone, z, zonelist,
					gfp_zone(sc->gfp_mask)) {
2361 2362 2363 2364 2365 2366
				if (!cpuset_zone_allowed_hardwall(zone, GFP_KERNEL))
					continue;

				lru_pages += zone_reclaimable_pages(zone);
			}

2367
			shrink_slab(shrink, sc->nr_scanned, lru_pages);
2368
			if (reclaim_state) {
2369
				sc->nr_reclaimed += reclaim_state->reclaimed_slab;
2370 2371
				reclaim_state->reclaimed_slab = 0;
			}
L
Linus Torvalds 已提交
2372
		}
2373
		total_scanned += sc->nr_scanned;
2374
		if (sc->nr_reclaimed >= sc->nr_to_reclaim)
L
Linus Torvalds 已提交
2375 2376 2377 2378 2379 2380 2381 2382 2383
			goto out;

		/*
		 * Try to write back as many pages as we just scanned.  This
		 * tends to cause slow streaming writers to write data to the
		 * disk smoothly, at the dirtying rate, which is nice.   But
		 * that's undesirable in laptop mode, where we *want* lumpy
		 * writeout.  So in laptop mode, write out the whole world.
		 */
2384 2385
		writeback_threshold = sc->nr_to_reclaim + sc->nr_to_reclaim / 2;
		if (total_scanned > writeback_threshold) {
2386 2387
			wakeup_flusher_threads(laptop_mode ? 0 : total_scanned,
						WB_REASON_TRY_TO_FREE_PAGES);
2388
			sc->may_writepage = 1;
L
Linus Torvalds 已提交
2389 2390 2391
		}

		/* Take a nap, wait for some writeback to complete */
2392
		if (!sc->hibernation_mode && sc->nr_scanned &&
2393 2394 2395 2396
		    priority < DEF_PRIORITY - 2) {
			struct zone *preferred_zone;

			first_zones_zonelist(zonelist, gfp_zone(sc->gfp_mask),
2397 2398
						&cpuset_current_mems_allowed,
						&preferred_zone);
2399 2400
			wait_iff_congested(preferred_zone, BLK_RW_ASYNC, HZ/10);
		}
L
Linus Torvalds 已提交
2401
	}
2402

L
Linus Torvalds 已提交
2403
out:
2404
	delayacct_freepages_end();
2405
	put_mems_allowed();
2406

2407 2408 2409
	if (sc->nr_reclaimed)
		return sc->nr_reclaimed;

2410 2411 2412 2413 2414 2415 2416 2417
	/*
	 * As hibernation is going on, kswapd is freezed so that it can't mark
	 * the zone into all_unreclaimable. Thus bypassing all_unreclaimable
	 * check.
	 */
	if (oom_killer_disabled)
		return 0;

2418 2419
	/* Aborted reclaim to try compaction? don't OOM, then */
	if (aborted_reclaim)
2420 2421
		return 1;

2422
	/* top priority shrink_zones still had more to do? don't OOM, then */
2423
	if (global_reclaim(sc) && !all_unreclaimable(zonelist, sc))
2424 2425 2426
		return 1;

	return 0;
L
Linus Torvalds 已提交
2427 2428
}

2429
unsigned long try_to_free_pages(struct zonelist *zonelist, int order,
2430
				gfp_t gfp_mask, nodemask_t *nodemask)
2431
{
2432
	unsigned long nr_reclaimed;
2433 2434 2435
	struct scan_control sc = {
		.gfp_mask = gfp_mask,
		.may_writepage = !laptop_mode,
2436
		.nr_to_reclaim = SWAP_CLUSTER_MAX,
2437
		.may_unmap = 1,
2438
		.may_swap = 1,
2439
		.order = order,
2440
		.target_mem_cgroup = NULL,
2441
		.nodemask = nodemask,
2442
	};
2443 2444 2445
	struct shrink_control shrink = {
		.gfp_mask = sc.gfp_mask,
	};
2446

2447 2448 2449 2450
	trace_mm_vmscan_direct_reclaim_begin(order,
				sc.may_writepage,
				gfp_mask);

2451
	nr_reclaimed = do_try_to_free_pages(zonelist, &sc, &shrink);
2452 2453 2454 2455

	trace_mm_vmscan_direct_reclaim_end(nr_reclaimed);

	return nr_reclaimed;
2456 2457
}

2458
#ifdef CONFIG_CGROUP_MEM_RES_CTLR
2459

2460
unsigned long mem_cgroup_shrink_node_zone(struct mem_cgroup *memcg,
2461
						gfp_t gfp_mask, bool noswap,
2462 2463
						struct zone *zone,
						unsigned long *nr_scanned)
2464 2465
{
	struct scan_control sc = {
2466
		.nr_scanned = 0,
2467
		.nr_to_reclaim = SWAP_CLUSTER_MAX,
2468 2469 2470 2471
		.may_writepage = !laptop_mode,
		.may_unmap = 1,
		.may_swap = !noswap,
		.order = 0,
2472
		.target_mem_cgroup = memcg,
2473
	};
2474
	struct mem_cgroup_zone mz = {
2475
		.mem_cgroup = memcg,
2476 2477
		.zone = zone,
	};
2478

2479 2480
	sc.gfp_mask = (gfp_mask & GFP_RECLAIM_MASK) |
			(GFP_HIGHUSER_MOVABLE & ~GFP_RECLAIM_MASK);
2481 2482 2483 2484 2485

	trace_mm_vmscan_memcg_softlimit_reclaim_begin(0,
						      sc.may_writepage,
						      sc.gfp_mask);

2486 2487 2488 2489 2490 2491 2492
	/*
	 * NOTE: Although we can get the priority field, using it
	 * here is not a good idea, since it limits the pages we can scan.
	 * if we don't reclaim here, the shrink_zone from balance_pgdat
	 * will pick up pages from other mem cgroup's as well. We hack
	 * the priority and make it zero.
	 */
2493
	shrink_mem_cgroup_zone(0, &mz, &sc);
2494 2495 2496

	trace_mm_vmscan_memcg_softlimit_reclaim_end(sc.nr_reclaimed);

2497
	*nr_scanned = sc.nr_scanned;
2498 2499 2500
	return sc.nr_reclaimed;
}

2501
unsigned long try_to_free_mem_cgroup_pages(struct mem_cgroup *memcg,
K
KOSAKI Motohiro 已提交
2502
					   gfp_t gfp_mask,
2503
					   bool noswap)
2504
{
2505
	struct zonelist *zonelist;
2506
	unsigned long nr_reclaimed;
2507
	int nid;
2508 2509
	struct scan_control sc = {
		.may_writepage = !laptop_mode,
2510
		.may_unmap = 1,
2511
		.may_swap = !noswap,
2512
		.nr_to_reclaim = SWAP_CLUSTER_MAX,
2513
		.order = 0,
2514
		.target_mem_cgroup = memcg,
2515
		.nodemask = NULL, /* we don't care the placement */
2516 2517 2518 2519 2520
		.gfp_mask = (gfp_mask & GFP_RECLAIM_MASK) |
				(GFP_HIGHUSER_MOVABLE & ~GFP_RECLAIM_MASK),
	};
	struct shrink_control shrink = {
		.gfp_mask = sc.gfp_mask,
2521 2522
	};

2523 2524 2525 2526 2527
	/*
	 * Unlike direct reclaim via alloc_pages(), memcg's reclaim doesn't
	 * take care of from where we get pages. So the node where we start the
	 * scan does not need to be the current node.
	 */
2528
	nid = mem_cgroup_select_victim_node(memcg);
2529 2530

	zonelist = NODE_DATA(nid)->node_zonelists;
2531 2532 2533 2534 2535

	trace_mm_vmscan_memcg_reclaim_begin(0,
					    sc.may_writepage,
					    sc.gfp_mask);

2536
	nr_reclaimed = do_try_to_free_pages(zonelist, &sc, &shrink);
2537 2538 2539 2540

	trace_mm_vmscan_memcg_reclaim_end(nr_reclaimed);

	return nr_reclaimed;
2541 2542 2543
}
#endif

2544 2545 2546
static void age_active_anon(struct zone *zone, struct scan_control *sc,
			    int priority)
{
2547
	struct mem_cgroup *memcg;
2548

2549 2550 2551 2552 2553 2554 2555 2556 2557 2558 2559 2560 2561 2562 2563 2564
	if (!total_swap_pages)
		return;

	memcg = mem_cgroup_iter(NULL, NULL, NULL);
	do {
		struct mem_cgroup_zone mz = {
			.mem_cgroup = memcg,
			.zone = zone,
		};

		if (inactive_anon_is_low(&mz))
			shrink_active_list(SWAP_CLUSTER_MAX, &mz,
					   sc, priority, 0);

		memcg = mem_cgroup_iter(NULL, memcg, NULL);
	} while (memcg);
2565 2566
}

2567 2568 2569 2570 2571 2572 2573 2574 2575 2576 2577
/*
 * pgdat_balanced is used when checking if a node is balanced for high-order
 * allocations. Only zones that meet watermarks and are in a zone allowed
 * by the callers classzone_idx are added to balanced_pages. The total of
 * balanced pages must be at least 25% of the zones allowed by classzone_idx
 * for the node to be considered balanced. Forcing all zones to be balanced
 * for high orders can cause excessive reclaim when there are imbalanced zones.
 * The choice of 25% is due to
 *   o a 16M DMA zone that is balanced will not balance a zone on any
 *     reasonable sized machine
 *   o On all other machines, the top zone must be at least a reasonable
L
Lucas De Marchi 已提交
2578
 *     percentage of the middle zones. For example, on 32-bit x86, highmem
2579 2580 2581 2582 2583 2584 2585 2586 2587 2588 2589 2590 2591
 *     would need to be at least 256M for it to be balance a whole node.
 *     Similarly, on x86-64 the Normal zone would need to be at least 1G
 *     to balance a node on its own. These seemed like reasonable ratios.
 */
static bool pgdat_balanced(pg_data_t *pgdat, unsigned long balanced_pages,
						int classzone_idx)
{
	unsigned long present_pages = 0;
	int i;

	for (i = 0; i <= classzone_idx; i++)
		present_pages += pgdat->node_zones[i].present_pages;

S
Shaohua Li 已提交
2592 2593
	/* A special case here: if zone has no page, we think it's balanced */
	return balanced_pages >= (present_pages >> 2);
2594 2595
}

2596
/* is kswapd sleeping prematurely? */
2597 2598
static bool sleeping_prematurely(pg_data_t *pgdat, int order, long remaining,
					int classzone_idx)
2599
{
2600
	int i;
2601 2602
	unsigned long balanced = 0;
	bool all_zones_ok = true;
2603 2604 2605

	/* If a direct reclaimer woke kswapd within HZ/10, it's premature */
	if (remaining)
2606
		return true;
2607

2608
	/* Check the watermark levels */
2609
	for (i = 0; i <= classzone_idx; i++) {
2610 2611 2612 2613 2614
		struct zone *zone = pgdat->node_zones + i;

		if (!populated_zone(zone))
			continue;

2615 2616 2617 2618 2619 2620 2621 2622
		/*
		 * balance_pgdat() skips over all_unreclaimable after
		 * DEF_PRIORITY. Effectively, it considers them balanced so
		 * they must be considered balanced here as well if kswapd
		 * is to sleep
		 */
		if (zone->all_unreclaimable) {
			balanced += zone->present_pages;
2623
			continue;
2624
		}
2625

2626
		if (!zone_watermark_ok_safe(zone, order, high_wmark_pages(zone),
2627
							i, 0))
2628 2629 2630
			all_zones_ok = false;
		else
			balanced += zone->present_pages;
2631
	}
2632

2633 2634 2635 2636 2637 2638
	/*
	 * For high-order requests, the balanced zones must contain at least
	 * 25% of the nodes pages for kswapd to sleep. For order-0, all zones
	 * must be balanced
	 */
	if (order)
2639
		return !pgdat_balanced(pgdat, balanced, classzone_idx);
2640 2641
	else
		return !all_zones_ok;
2642 2643
}

L
Linus Torvalds 已提交
2644 2645
/*
 * For kswapd, balance_pgdat() will work across all this node's zones until
2646
 * they are all at high_wmark_pages(zone).
L
Linus Torvalds 已提交
2647
 *
2648
 * Returns the final order kswapd was reclaiming at
L
Linus Torvalds 已提交
2649 2650 2651 2652 2653 2654 2655 2656 2657 2658
 *
 * There is special handling here for zones which are full of pinned pages.
 * This can happen if the pages are all mlocked, or if they are all used by
 * device drivers (say, ZONE_DMA).  Or if they are all in use by hugetlb.
 * What we do is to detect the case where all pages in the zone have been
 * scanned twice and there has been zero successful reclaim.  Mark the zone as
 * dead and from now on, only perform a short scan.  Basically we're polling
 * the zone for when the problem goes away.
 *
 * kswapd scans the zones in the highmem->normal->dma direction.  It skips
2659 2660 2661 2662 2663
 * zones which have free_pages > high_wmark_pages(zone), but once a zone is
 * found to have free_pages <= high_wmark_pages(zone), we scan that zone and the
 * lower zones regardless of the number of free pages in the lower zones. This
 * interoperates with the page allocator fallback scheme to ensure that aging
 * of pages is balanced across the zones.
L
Linus Torvalds 已提交
2664
 */
2665
static unsigned long balance_pgdat(pg_data_t *pgdat, int order,
2666
							int *classzone_idx)
L
Linus Torvalds 已提交
2667 2668
{
	int all_zones_ok;
2669
	unsigned long balanced;
L
Linus Torvalds 已提交
2670 2671
	int priority;
	int i;
2672
	int end_zone = 0;	/* Inclusive.  0 = ZONE_DMA */
2673
	unsigned long total_scanned;
L
Linus Torvalds 已提交
2674
	struct reclaim_state *reclaim_state = current->reclaim_state;
2675 2676
	unsigned long nr_soft_reclaimed;
	unsigned long nr_soft_scanned;
2677 2678
	struct scan_control sc = {
		.gfp_mask = GFP_KERNEL,
2679
		.may_unmap = 1,
2680
		.may_swap = 1,
2681 2682 2683 2684 2685
		/*
		 * kswapd doesn't want to be bailed out while reclaim. because
		 * we want to put equal scanning pressure on each zone.
		 */
		.nr_to_reclaim = ULONG_MAX,
A
Andy Whitcroft 已提交
2686
		.order = order,
2687
		.target_mem_cgroup = NULL,
2688
	};
2689 2690 2691
	struct shrink_control shrink = {
		.gfp_mask = sc.gfp_mask,
	};
L
Linus Torvalds 已提交
2692 2693
loop_again:
	total_scanned = 0;
2694
	sc.nr_reclaimed = 0;
C
Christoph Lameter 已提交
2695
	sc.may_writepage = !laptop_mode;
2696
	count_vm_event(PAGEOUTRUN);
L
Linus Torvalds 已提交
2697 2698 2699

	for (priority = DEF_PRIORITY; priority >= 0; priority--) {
		unsigned long lru_pages = 0;
2700
		int has_under_min_watermark_zone = 0;
L
Linus Torvalds 已提交
2701

2702 2703
		/* The swap token gets in the way of swapout... */
		if (!priority)
2704
			disable_swap_token(NULL);
2705

L
Linus Torvalds 已提交
2706
		all_zones_ok = 1;
2707
		balanced = 0;
L
Linus Torvalds 已提交
2708

2709 2710 2711 2712 2713 2714
		/*
		 * Scan in the highmem->dma direction for the highest
		 * zone which needs scanning
		 */
		for (i = pgdat->nr_zones - 1; i >= 0; i--) {
			struct zone *zone = pgdat->node_zones + i;
L
Linus Torvalds 已提交
2715

2716 2717
			if (!populated_zone(zone))
				continue;
L
Linus Torvalds 已提交
2718

2719
			if (zone->all_unreclaimable && priority != DEF_PRIORITY)
2720
				continue;
L
Linus Torvalds 已提交
2721

2722 2723 2724 2725
			/*
			 * Do some background aging of the anon list, to give
			 * pages a chance to be referenced before reclaiming.
			 */
2726
			age_active_anon(zone, &sc, priority);
2727

2728
			if (!zone_watermark_ok_safe(zone, order,
2729
					high_wmark_pages(zone), 0, 0)) {
2730
				end_zone = i;
A
Andrew Morton 已提交
2731
				break;
2732 2733 2734
			} else {
				/* If balanced, clear the congested flag */
				zone_clear_flag(zone, ZONE_CONGESTED);
L
Linus Torvalds 已提交
2735 2736
			}
		}
A
Andrew Morton 已提交
2737 2738 2739
		if (i < 0)
			goto out;

L
Linus Torvalds 已提交
2740 2741 2742
		for (i = 0; i <= end_zone; i++) {
			struct zone *zone = pgdat->node_zones + i;

2743
			lru_pages += zone_reclaimable_pages(zone);
L
Linus Torvalds 已提交
2744 2745 2746 2747 2748 2749 2750 2751 2752 2753 2754 2755 2756
		}

		/*
		 * Now scan the zone in the dma->highmem direction, stopping
		 * at the last zone which needs scanning.
		 *
		 * We do this because the page allocator works in the opposite
		 * direction.  This prevents the page allocator from allocating
		 * pages behind kswapd's direction of progress, which would
		 * cause too much scanning of the lower zones.
		 */
		for (i = 0; i <= end_zone; i++) {
			struct zone *zone = pgdat->node_zones + i;
2757
			int nr_slab;
2758
			unsigned long balance_gap;
L
Linus Torvalds 已提交
2759

2760
			if (!populated_zone(zone))
L
Linus Torvalds 已提交
2761 2762
				continue;

2763
			if (zone->all_unreclaimable && priority != DEF_PRIORITY)
L
Linus Torvalds 已提交
2764 2765 2766
				continue;

			sc.nr_scanned = 0;
2767

2768
			nr_soft_scanned = 0;
2769 2770 2771
			/*
			 * Call soft limit reclaim before calling shrink_zone.
			 */
2772 2773 2774 2775 2776
			nr_soft_reclaimed = mem_cgroup_soft_limit_reclaim(zone,
							order, sc.gfp_mask,
							&nr_soft_scanned);
			sc.nr_reclaimed += nr_soft_reclaimed;
			total_scanned += nr_soft_scanned;
2777

2778
			/*
2779 2780 2781 2782 2783 2784
			 * We put equal pressure on every zone, unless
			 * one zone has way too many pages free
			 * already. The "too many pages" is defined
			 * as the high wmark plus a "gap" where the
			 * gap is either the low watermark or 1%
			 * of the zone, whichever is smaller.
2785
			 */
2786 2787 2788 2789
			balance_gap = min(low_wmark_pages(zone),
				(zone->present_pages +
					KSWAPD_ZONE_BALANCE_GAP_RATIO-1) /
				KSWAPD_ZONE_BALANCE_GAP_RATIO);
2790
			if (!zone_watermark_ok_safe(zone, order,
2791
					high_wmark_pages(zone) + balance_gap,
2792
					end_zone, 0)) {
2793
				shrink_zone(priority, zone, &sc);
2794

2795 2796 2797 2798 2799 2800 2801 2802 2803
				reclaim_state->reclaimed_slab = 0;
				nr_slab = shrink_slab(&shrink, sc.nr_scanned, lru_pages);
				sc.nr_reclaimed += reclaim_state->reclaimed_slab;
				total_scanned += sc.nr_scanned;

				if (nr_slab == 0 && !zone_reclaimable(zone))
					zone->all_unreclaimable = 1;
			}

L
Linus Torvalds 已提交
2804 2805 2806 2807 2808 2809
			/*
			 * If we've done a decent amount of scanning and
			 * the reclaim ratio is low, start doing writepage
			 * even in laptop mode
			 */
			if (total_scanned > SWAP_CLUSTER_MAX * 2 &&
2810
			    total_scanned > sc.nr_reclaimed + sc.nr_reclaimed / 2)
L
Linus Torvalds 已提交
2811
				sc.may_writepage = 1;
2812

2813 2814 2815
			if (zone->all_unreclaimable) {
				if (end_zone && end_zone == i)
					end_zone--;
2816
				continue;
2817
			}
2818

2819
			if (!zone_watermark_ok_safe(zone, order,
2820 2821 2822 2823 2824 2825 2826
					high_wmark_pages(zone), end_zone, 0)) {
				all_zones_ok = 0;
				/*
				 * We are still under min water mark.  This
				 * means that we have a GFP_ATOMIC allocation
				 * failure risk. Hurry up!
				 */
2827
				if (!zone_watermark_ok_safe(zone, order,
2828 2829
					    min_wmark_pages(zone), end_zone, 0))
					has_under_min_watermark_zone = 1;
2830 2831 2832 2833 2834 2835 2836 2837 2838
			} else {
				/*
				 * If a zone reaches its high watermark,
				 * consider it to be no longer congested. It's
				 * possible there are dirty pages backed by
				 * congested BDIs but as pressure is relieved,
				 * spectulatively avoid congestion waits
				 */
				zone_clear_flag(zone, ZONE_CONGESTED);
2839
				if (i <= *classzone_idx)
2840
					balanced += zone->present_pages;
2841
			}
2842

L
Linus Torvalds 已提交
2843
		}
2844
		if (all_zones_ok || (order && pgdat_balanced(pgdat, balanced, *classzone_idx)))
L
Linus Torvalds 已提交
2845 2846 2847 2848 2849
			break;		/* kswapd: all done */
		/*
		 * OK, kswapd is getting into trouble.  Take a nap, then take
		 * another pass across the zones.
		 */
2850 2851 2852 2853 2854 2855
		if (total_scanned && (priority < DEF_PRIORITY - 2)) {
			if (has_under_min_watermark_zone)
				count_vm_event(KSWAPD_SKIP_CONGESTION_WAIT);
			else
				congestion_wait(BLK_RW_ASYNC, HZ/10);
		}
L
Linus Torvalds 已提交
2856 2857 2858 2859 2860 2861 2862

		/*
		 * We do this so kswapd doesn't build up large priorities for
		 * example when it is freeing in parallel with allocators. It
		 * matches the direct reclaim path behaviour in terms of impact
		 * on zone->*_priority.
		 */
2863
		if (sc.nr_reclaimed >= SWAP_CLUSTER_MAX)
L
Linus Torvalds 已提交
2864 2865 2866
			break;
	}
out:
2867 2868 2869

	/*
	 * order-0: All zones must meet high watermark for a balanced node
2870 2871
	 * high-order: Balanced zones must make up at least 25% of the node
	 *             for the node to be balanced
2872
	 */
2873
	if (!(all_zones_ok || (order && pgdat_balanced(pgdat, balanced, *classzone_idx)))) {
L
Linus Torvalds 已提交
2874
		cond_resched();
2875 2876 2877

		try_to_freeze();

2878 2879 2880 2881 2882 2883 2884 2885 2886 2887 2888 2889 2890 2891 2892 2893 2894
		/*
		 * Fragmentation may mean that the system cannot be
		 * rebalanced for high-order allocations in all zones.
		 * At this point, if nr_reclaimed < SWAP_CLUSTER_MAX,
		 * it means the zones have been fully scanned and are still
		 * not balanced. For high-order allocations, there is
		 * little point trying all over again as kswapd may
		 * infinite loop.
		 *
		 * Instead, recheck all watermarks at order-0 as they
		 * are the most important. If watermarks are ok, kswapd will go
		 * back to sleep. High-order users can still perform direct
		 * reclaim if they wish.
		 */
		if (sc.nr_reclaimed < SWAP_CLUSTER_MAX)
			order = sc.order = 0;

L
Linus Torvalds 已提交
2895 2896 2897
		goto loop_again;
	}

2898 2899 2900 2901 2902 2903 2904 2905 2906 2907 2908 2909 2910 2911 2912 2913 2914 2915 2916 2917 2918 2919 2920 2921 2922 2923 2924
	/*
	 * If kswapd was reclaiming at a higher order, it has the option of
	 * sleeping without all zones being balanced. Before it does, it must
	 * ensure that the watermarks for order-0 on *all* zones are met and
	 * that the congestion flags are cleared. The congestion flag must
	 * be cleared as kswapd is the only mechanism that clears the flag
	 * and it is potentially going to sleep here.
	 */
	if (order) {
		for (i = 0; i <= end_zone; i++) {
			struct zone *zone = pgdat->node_zones + i;

			if (!populated_zone(zone))
				continue;

			if (zone->all_unreclaimable && priority != DEF_PRIORITY)
				continue;

			/* Confirm the zone is balanced for order-0 */
			if (!zone_watermark_ok(zone, 0,
					high_wmark_pages(zone), 0, 0)) {
				order = sc.order = 0;
				goto loop_again;
			}

			/* If balanced, clear the congested flag */
			zone_clear_flag(zone, ZONE_CONGESTED);
2925 2926
			if (i <= *classzone_idx)
				balanced += zone->present_pages;
2927 2928 2929
		}
	}

2930 2931 2932 2933 2934 2935
	/*
	 * Return the order we were reclaiming at so sleeping_prematurely()
	 * makes a decision on the order we were last reclaiming at. However,
	 * if another caller entered the allocator slow path while kswapd
	 * was awake, order will remain at the higher level
	 */
2936
	*classzone_idx = end_zone;
2937
	return order;
L
Linus Torvalds 已提交
2938 2939
}

2940
static void kswapd_try_to_sleep(pg_data_t *pgdat, int order, int classzone_idx)
2941 2942 2943 2944 2945 2946 2947 2948 2949 2950
{
	long remaining = 0;
	DEFINE_WAIT(wait);

	if (freezing(current) || kthread_should_stop())
		return;

	prepare_to_wait(&pgdat->kswapd_wait, &wait, TASK_INTERRUPTIBLE);

	/* Try to sleep for a short interval */
2951
	if (!sleeping_prematurely(pgdat, order, remaining, classzone_idx)) {
2952 2953 2954 2955 2956 2957 2958 2959 2960
		remaining = schedule_timeout(HZ/10);
		finish_wait(&pgdat->kswapd_wait, &wait);
		prepare_to_wait(&pgdat->kswapd_wait, &wait, TASK_INTERRUPTIBLE);
	}

	/*
	 * After a short sleep, check if it was a premature sleep. If not, then
	 * go fully to sleep until explicitly woken up.
	 */
2961
	if (!sleeping_prematurely(pgdat, order, remaining, classzone_idx)) {
2962 2963 2964 2965 2966 2967 2968 2969 2970 2971 2972 2973 2974 2975 2976 2977 2978 2979 2980 2981 2982 2983
		trace_mm_vmscan_kswapd_sleep(pgdat->node_id);

		/*
		 * vmstat counters are not perfectly accurate and the estimated
		 * value for counters such as NR_FREE_PAGES can deviate from the
		 * true value by nr_online_cpus * threshold. To avoid the zone
		 * watermarks being breached while under pressure, we reduce the
		 * per-cpu vmstat threshold while kswapd is awake and restore
		 * them before going back to sleep.
		 */
		set_pgdat_percpu_threshold(pgdat, calculate_normal_threshold);
		schedule();
		set_pgdat_percpu_threshold(pgdat, calculate_pressure_threshold);
	} else {
		if (remaining)
			count_vm_event(KSWAPD_LOW_WMARK_HIT_QUICKLY);
		else
			count_vm_event(KSWAPD_HIGH_WMARK_HIT_QUICKLY);
	}
	finish_wait(&pgdat->kswapd_wait, &wait);
}

L
Linus Torvalds 已提交
2984 2985
/*
 * The background pageout daemon, started as a kernel thread
2986
 * from the init process.
L
Linus Torvalds 已提交
2987 2988 2989 2990 2991 2992 2993 2994 2995 2996 2997 2998
 *
 * This basically trickles out pages so that we have _some_
 * free memory available even if there is no other activity
 * that frees anything up. This is needed for things like routing
 * etc, where we otherwise might have all activity going on in
 * asynchronous contexts that cannot page things out.
 *
 * If there are applications that are active memory-allocators
 * (most normal use), this basically shouldn't matter.
 */
static int kswapd(void *p)
{
2999
	unsigned long order, new_order;
3000
	unsigned balanced_order;
3001
	int classzone_idx, new_classzone_idx;
3002
	int balanced_classzone_idx;
L
Linus Torvalds 已提交
3003 3004
	pg_data_t *pgdat = (pg_data_t*)p;
	struct task_struct *tsk = current;
3005

L
Linus Torvalds 已提交
3006 3007 3008
	struct reclaim_state reclaim_state = {
		.reclaimed_slab = 0,
	};
3009
	const struct cpumask *cpumask = cpumask_of_node(pgdat->node_id);
L
Linus Torvalds 已提交
3010

3011 3012
	lockdep_set_current_reclaim_state(GFP_KERNEL);

R
Rusty Russell 已提交
3013
	if (!cpumask_empty(cpumask))
3014
		set_cpus_allowed_ptr(tsk, cpumask);
L
Linus Torvalds 已提交
3015 3016 3017 3018 3019 3020 3021 3022 3023 3024 3025 3026 3027 3028
	current->reclaim_state = &reclaim_state;

	/*
	 * Tell the memory management that we're a "memory allocator",
	 * and that if we need more memory we should get access to it
	 * regardless (see "__alloc_pages()"). "kswapd" should
	 * never get caught in the normal page freeing logic.
	 *
	 * (Kswapd normally doesn't need memory anyway, but sometimes
	 * you need a small amount of memory in order to be able to
	 * page out something else, and this flag essentially protects
	 * us from recursively trying to free more memory as we're
	 * trying to free the first piece of memory in the first place).
	 */
3029
	tsk->flags |= PF_MEMALLOC | PF_SWAPWRITE | PF_KSWAPD;
3030
	set_freezable();
L
Linus Torvalds 已提交
3031

3032
	order = new_order = 0;
3033
	balanced_order = 0;
3034
	classzone_idx = new_classzone_idx = pgdat->nr_zones - 1;
3035
	balanced_classzone_idx = classzone_idx;
L
Linus Torvalds 已提交
3036
	for ( ; ; ) {
3037
		int ret;
3038

3039 3040 3041 3042 3043
		/*
		 * If the last balance_pgdat was unsuccessful it's unlikely a
		 * new request of a similar or harder type will succeed soon
		 * so consider going to sleep on the basis we reclaimed at
		 */
3044 3045
		if (balanced_classzone_idx >= new_classzone_idx &&
					balanced_order == new_order) {
3046 3047 3048 3049 3050 3051
			new_order = pgdat->kswapd_max_order;
			new_classzone_idx = pgdat->classzone_idx;
			pgdat->kswapd_max_order =  0;
			pgdat->classzone_idx = pgdat->nr_zones - 1;
		}

3052
		if (order < new_order || classzone_idx > new_classzone_idx) {
L
Linus Torvalds 已提交
3053 3054
			/*
			 * Don't sleep if someone wants a larger 'order'
3055
			 * allocation or has tigher zone constraints
L
Linus Torvalds 已提交
3056 3057
			 */
			order = new_order;
3058
			classzone_idx = new_classzone_idx;
L
Linus Torvalds 已提交
3059
		} else {
3060 3061
			kswapd_try_to_sleep(pgdat, balanced_order,
						balanced_classzone_idx);
L
Linus Torvalds 已提交
3062
			order = pgdat->kswapd_max_order;
3063
			classzone_idx = pgdat->classzone_idx;
3064 3065
			new_order = order;
			new_classzone_idx = classzone_idx;
3066
			pgdat->kswapd_max_order = 0;
3067
			pgdat->classzone_idx = pgdat->nr_zones - 1;
L
Linus Torvalds 已提交
3068 3069
		}

3070 3071 3072 3073 3074 3075 3076 3077
		ret = try_to_freeze();
		if (kthread_should_stop())
			break;

		/*
		 * We can speed up thawing tasks if we don't call balance_pgdat
		 * after returning from the refrigerator
		 */
3078 3079
		if (!ret) {
			trace_mm_vmscan_kswapd_wake(pgdat->node_id, order);
3080 3081 3082
			balanced_classzone_idx = classzone_idx;
			balanced_order = balance_pgdat(pgdat, order,
						&balanced_classzone_idx);
3083
		}
L
Linus Torvalds 已提交
3084 3085 3086 3087 3088 3089 3090
	}
	return 0;
}

/*
 * A zone is low on free memory, so wake its kswapd task to service it.
 */
3091
void wakeup_kswapd(struct zone *zone, int order, enum zone_type classzone_idx)
L
Linus Torvalds 已提交
3092 3093 3094
{
	pg_data_t *pgdat;

3095
	if (!populated_zone(zone))
L
Linus Torvalds 已提交
3096 3097
		return;

3098
	if (!cpuset_zone_allowed_hardwall(zone, GFP_KERNEL))
L
Linus Torvalds 已提交
3099
		return;
3100
	pgdat = zone->zone_pgdat;
3101
	if (pgdat->kswapd_max_order < order) {
L
Linus Torvalds 已提交
3102
		pgdat->kswapd_max_order = order;
3103 3104
		pgdat->classzone_idx = min(pgdat->classzone_idx, classzone_idx);
	}
3105
	if (!waitqueue_active(&pgdat->kswapd_wait))
L
Linus Torvalds 已提交
3106
		return;
3107 3108 3109 3110
	if (zone_watermark_ok_safe(zone, order, low_wmark_pages(zone), 0, 0))
		return;

	trace_mm_vmscan_wakeup_kswapd(pgdat->node_id, zone_idx(zone), order);
3111
	wake_up_interruptible(&pgdat->kswapd_wait);
L
Linus Torvalds 已提交
3112 3113
}

3114 3115 3116 3117 3118 3119 3120 3121
/*
 * The reclaimable count would be mostly accurate.
 * The less reclaimable pages may be
 * - mlocked pages, which will be moved to unevictable list when encountered
 * - mapped pages, which may require several travels to be reclaimed
 * - dirty pages, which is not "instantly" reclaimable
 */
unsigned long global_reclaimable_pages(void)
3122
{
3123 3124 3125 3126 3127 3128 3129 3130 3131 3132 3133 3134 3135 3136 3137 3138 3139 3140 3141 3142 3143 3144 3145 3146
	int nr;

	nr = global_page_state(NR_ACTIVE_FILE) +
	     global_page_state(NR_INACTIVE_FILE);

	if (nr_swap_pages > 0)
		nr += global_page_state(NR_ACTIVE_ANON) +
		      global_page_state(NR_INACTIVE_ANON);

	return nr;
}

unsigned long zone_reclaimable_pages(struct zone *zone)
{
	int nr;

	nr = zone_page_state(zone, NR_ACTIVE_FILE) +
	     zone_page_state(zone, NR_INACTIVE_FILE);

	if (nr_swap_pages > 0)
		nr += zone_page_state(zone, NR_ACTIVE_ANON) +
		      zone_page_state(zone, NR_INACTIVE_ANON);

	return nr;
3147 3148
}

3149
#ifdef CONFIG_HIBERNATION
L
Linus Torvalds 已提交
3150
/*
3151
 * Try to free `nr_to_reclaim' of memory, system-wide, and return the number of
3152 3153 3154 3155 3156
 * freed pages.
 *
 * Rather than trying to age LRUs the aim is to preserve the overall
 * LRU order by reclaiming preferentially
 * inactive > active > active referenced > active mapped
L
Linus Torvalds 已提交
3157
 */
3158
unsigned long shrink_all_memory(unsigned long nr_to_reclaim)
L
Linus Torvalds 已提交
3159
{
3160 3161
	struct reclaim_state reclaim_state;
	struct scan_control sc = {
3162 3163 3164
		.gfp_mask = GFP_HIGHUSER_MOVABLE,
		.may_swap = 1,
		.may_unmap = 1,
3165
		.may_writepage = 1,
3166 3167 3168
		.nr_to_reclaim = nr_to_reclaim,
		.hibernation_mode = 1,
		.order = 0,
L
Linus Torvalds 已提交
3169
	};
3170 3171 3172 3173
	struct shrink_control shrink = {
		.gfp_mask = sc.gfp_mask,
	};
	struct zonelist *zonelist = node_zonelist(numa_node_id(), sc.gfp_mask);
3174 3175
	struct task_struct *p = current;
	unsigned long nr_reclaimed;
L
Linus Torvalds 已提交
3176

3177 3178 3179 3180
	p->flags |= PF_MEMALLOC;
	lockdep_set_current_reclaim_state(sc.gfp_mask);
	reclaim_state.reclaimed_slab = 0;
	p->reclaim_state = &reclaim_state;
3181

3182
	nr_reclaimed = do_try_to_free_pages(zonelist, &sc, &shrink);
3183

3184 3185 3186
	p->reclaim_state = NULL;
	lockdep_clear_current_reclaim_state();
	p->flags &= ~PF_MEMALLOC;
3187

3188
	return nr_reclaimed;
L
Linus Torvalds 已提交
3189
}
3190
#endif /* CONFIG_HIBERNATION */
L
Linus Torvalds 已提交
3191 3192 3193 3194 3195

/* It's optimal to keep kswapds on the same CPUs as their memory, but
   not required for correctness.  So if the last cpu in a node goes
   away, we get changed to run anywhere: as the first one comes back,
   restore their cpu bindings. */
3196
static int __devinit cpu_callback(struct notifier_block *nfb,
3197
				  unsigned long action, void *hcpu)
L
Linus Torvalds 已提交
3198
{
3199
	int nid;
L
Linus Torvalds 已提交
3200

3201
	if (action == CPU_ONLINE || action == CPU_ONLINE_FROZEN) {
3202
		for_each_node_state(nid, N_HIGH_MEMORY) {
3203
			pg_data_t *pgdat = NODE_DATA(nid);
3204 3205 3206
			const struct cpumask *mask;

			mask = cpumask_of_node(pgdat->node_id);
3207

3208
			if (cpumask_any_and(cpu_online_mask, mask) < nr_cpu_ids)
L
Linus Torvalds 已提交
3209
				/* One of our CPUs online: restore mask */
3210
				set_cpus_allowed_ptr(pgdat->kswapd, mask);
L
Linus Torvalds 已提交
3211 3212 3213 3214 3215
		}
	}
	return NOTIFY_OK;
}

3216 3217 3218 3219 3220 3221 3222 3223 3224 3225 3226 3227 3228 3229 3230 3231 3232 3233 3234 3235 3236 3237
/*
 * This kswapd start function will be called by init and node-hot-add.
 * On node-hot-add, kswapd will moved to proper cpus if cpus are hot-added.
 */
int kswapd_run(int nid)
{
	pg_data_t *pgdat = NODE_DATA(nid);
	int ret = 0;

	if (pgdat->kswapd)
		return 0;

	pgdat->kswapd = kthread_run(kswapd, pgdat, "kswapd%d", nid);
	if (IS_ERR(pgdat->kswapd)) {
		/* failure at boot is fatal */
		BUG_ON(system_state == SYSTEM_BOOTING);
		printk("Failed to start kswapd on node %d\n",nid);
		ret = -1;
	}
	return ret;
}

3238 3239 3240 3241 3242 3243 3244 3245 3246 3247 3248
/*
 * Called by memory hotplug when all memory in a node is offlined.
 */
void kswapd_stop(int nid)
{
	struct task_struct *kswapd = NODE_DATA(nid)->kswapd;

	if (kswapd)
		kthread_stop(kswapd);
}

L
Linus Torvalds 已提交
3249 3250
static int __init kswapd_init(void)
{
3251
	int nid;
3252

L
Linus Torvalds 已提交
3253
	swap_setup();
3254
	for_each_node_state(nid, N_HIGH_MEMORY)
3255
 		kswapd_run(nid);
L
Linus Torvalds 已提交
3256 3257 3258 3259 3260
	hotcpu_notifier(cpu_callback, 0);
	return 0;
}

module_init(kswapd_init)
3261 3262 3263 3264 3265 3266 3267 3268 3269 3270

#ifdef CONFIG_NUMA
/*
 * Zone reclaim mode
 *
 * If non-zero call zone_reclaim when the number of free pages falls below
 * the watermarks.
 */
int zone_reclaim_mode __read_mostly;

3271
#define RECLAIM_OFF 0
3272
#define RECLAIM_ZONE (1<<0)	/* Run shrink_inactive_list on the zone */
3273 3274 3275
#define RECLAIM_WRITE (1<<1)	/* Writeout pages during reclaim */
#define RECLAIM_SWAP (1<<2)	/* Swap pages out during reclaim */

3276 3277 3278 3279 3280 3281 3282
/*
 * Priority for ZONE_RECLAIM. This determines the fraction of pages
 * of a node considered for each zone_reclaim. 4 scans 1/16th of
 * a zone.
 */
#define ZONE_RECLAIM_PRIORITY 4

3283 3284 3285 3286 3287 3288
/*
 * Percentage of pages in a zone that must be unmapped for zone_reclaim to
 * occur.
 */
int sysctl_min_unmapped_ratio = 1;

3289 3290 3291 3292 3293 3294
/*
 * If the number of slab pages in a zone grows beyond this percentage then
 * slab reclaim needs to occur.
 */
int sysctl_min_slab_ratio = 5;

3295 3296 3297 3298 3299 3300 3301 3302 3303 3304 3305 3306 3307 3308 3309 3310 3311 3312 3313 3314 3315 3316 3317 3318 3319 3320 3321 3322 3323 3324 3325 3326 3327 3328 3329 3330 3331 3332 3333 3334 3335 3336
static inline unsigned long zone_unmapped_file_pages(struct zone *zone)
{
	unsigned long file_mapped = zone_page_state(zone, NR_FILE_MAPPED);
	unsigned long file_lru = zone_page_state(zone, NR_INACTIVE_FILE) +
		zone_page_state(zone, NR_ACTIVE_FILE);

	/*
	 * It's possible for there to be more file mapped pages than
	 * accounted for by the pages on the file LRU lists because
	 * tmpfs pages accounted for as ANON can also be FILE_MAPPED
	 */
	return (file_lru > file_mapped) ? (file_lru - file_mapped) : 0;
}

/* Work out how many page cache pages we can reclaim in this reclaim_mode */
static long zone_pagecache_reclaimable(struct zone *zone)
{
	long nr_pagecache_reclaimable;
	long delta = 0;

	/*
	 * If RECLAIM_SWAP is set, then all file pages are considered
	 * potentially reclaimable. Otherwise, we have to worry about
	 * pages like swapcache and zone_unmapped_file_pages() provides
	 * a better estimate
	 */
	if (zone_reclaim_mode & RECLAIM_SWAP)
		nr_pagecache_reclaimable = zone_page_state(zone, NR_FILE_PAGES);
	else
		nr_pagecache_reclaimable = zone_unmapped_file_pages(zone);

	/* If we can't clean pages, remove dirty pages from consideration */
	if (!(zone_reclaim_mode & RECLAIM_WRITE))
		delta += zone_page_state(zone, NR_FILE_DIRTY);

	/* Watch for any possible underflows due to delta */
	if (unlikely(delta > nr_pagecache_reclaimable))
		delta = nr_pagecache_reclaimable;

	return nr_pagecache_reclaimable - delta;
}

3337 3338 3339
/*
 * Try to free up some pages from this zone through reclaim.
 */
3340
static int __zone_reclaim(struct zone *zone, gfp_t gfp_mask, unsigned int order)
3341
{
3342
	/* Minimum pages needed in order to stay on node */
3343
	const unsigned long nr_pages = 1 << order;
3344 3345
	struct task_struct *p = current;
	struct reclaim_state reclaim_state;
3346
	int priority;
3347 3348
	struct scan_control sc = {
		.may_writepage = !!(zone_reclaim_mode & RECLAIM_WRITE),
3349
		.may_unmap = !!(zone_reclaim_mode & RECLAIM_SWAP),
3350
		.may_swap = 1,
3351 3352
		.nr_to_reclaim = max_t(unsigned long, nr_pages,
				       SWAP_CLUSTER_MAX),
3353
		.gfp_mask = gfp_mask,
3354
		.order = order,
3355
	};
3356 3357 3358
	struct shrink_control shrink = {
		.gfp_mask = sc.gfp_mask,
	};
3359
	unsigned long nr_slab_pages0, nr_slab_pages1;
3360 3361

	cond_resched();
3362 3363 3364 3365 3366 3367
	/*
	 * We need to be able to allocate from the reserves for RECLAIM_SWAP
	 * and we also need to be able to write out pages for RECLAIM_WRITE
	 * and RECLAIM_SWAP.
	 */
	p->flags |= PF_MEMALLOC | PF_SWAPWRITE;
3368
	lockdep_set_current_reclaim_state(gfp_mask);
3369 3370
	reclaim_state.reclaimed_slab = 0;
	p->reclaim_state = &reclaim_state;
3371

3372
	if (zone_pagecache_reclaimable(zone) > zone->min_unmapped_pages) {
3373 3374 3375 3376 3377 3378
		/*
		 * Free memory by calling shrink zone with increasing
		 * priorities until we have enough memory freed.
		 */
		priority = ZONE_RECLAIM_PRIORITY;
		do {
3379
			shrink_zone(priority, zone, &sc);
3380
			priority--;
3381
		} while (priority >= 0 && sc.nr_reclaimed < nr_pages);
3382
	}
3383

3384 3385
	nr_slab_pages0 = zone_page_state(zone, NR_SLAB_RECLAIMABLE);
	if (nr_slab_pages0 > zone->min_slab_pages) {
3386
		/*
3387
		 * shrink_slab() does not currently allow us to determine how
3388 3389 3390 3391
		 * many pages were freed in this zone. So we take the current
		 * number of slab pages and shake the slab until it is reduced
		 * by the same nr_pages that we used for reclaiming unmapped
		 * pages.
3392
		 *
3393 3394
		 * Note that shrink_slab will free memory on all zones and may
		 * take a long time.
3395
		 */
3396 3397 3398 3399
		for (;;) {
			unsigned long lru_pages = zone_reclaimable_pages(zone);

			/* No reclaimable slab or very low memory pressure */
3400
			if (!shrink_slab(&shrink, sc.nr_scanned, lru_pages))
3401 3402 3403 3404 3405 3406 3407 3408
				break;

			/* Freed enough memory */
			nr_slab_pages1 = zone_page_state(zone,
							NR_SLAB_RECLAIMABLE);
			if (nr_slab_pages1 + nr_pages <= nr_slab_pages0)
				break;
		}
3409 3410 3411 3412 3413

		/*
		 * Update nr_reclaimed by the number of slab pages we
		 * reclaimed from this zone.
		 */
3414 3415 3416
		nr_slab_pages1 = zone_page_state(zone, NR_SLAB_RECLAIMABLE);
		if (nr_slab_pages1 < nr_slab_pages0)
			sc.nr_reclaimed += nr_slab_pages0 - nr_slab_pages1;
3417 3418
	}

3419
	p->reclaim_state = NULL;
3420
	current->flags &= ~(PF_MEMALLOC | PF_SWAPWRITE);
3421
	lockdep_clear_current_reclaim_state();
3422
	return sc.nr_reclaimed >= nr_pages;
3423
}
3424 3425 3426 3427

int zone_reclaim(struct zone *zone, gfp_t gfp_mask, unsigned int order)
{
	int node_id;
3428
	int ret;
3429 3430

	/*
3431 3432
	 * Zone reclaim reclaims unmapped file backed pages and
	 * slab pages if we are over the defined limits.
3433
	 *
3434 3435 3436 3437 3438
	 * A small portion of unmapped file backed pages is needed for
	 * file I/O otherwise pages read by file I/O will be immediately
	 * thrown out if the zone is overallocated. So we do not reclaim
	 * if less than a specified percentage of the zone is used by
	 * unmapped file backed pages.
3439
	 */
3440 3441
	if (zone_pagecache_reclaimable(zone) <= zone->min_unmapped_pages &&
	    zone_page_state(zone, NR_SLAB_RECLAIMABLE) <= zone->min_slab_pages)
3442
		return ZONE_RECLAIM_FULL;
3443

3444
	if (zone->all_unreclaimable)
3445
		return ZONE_RECLAIM_FULL;
3446

3447
	/*
3448
	 * Do not scan if the allocation should not be delayed.
3449
	 */
3450
	if (!(gfp_mask & __GFP_WAIT) || (current->flags & PF_MEMALLOC))
3451
		return ZONE_RECLAIM_NOSCAN;
3452 3453 3454 3455 3456 3457 3458

	/*
	 * Only run zone reclaim on the local zone or on zones that do not
	 * have associated processors. This will favor the local processor
	 * over remote processors and spread off node memory allocations
	 * as wide as possible.
	 */
3459
	node_id = zone_to_nid(zone);
3460
	if (node_state(node_id, N_CPU) && node_id != numa_node_id())
3461
		return ZONE_RECLAIM_NOSCAN;
3462 3463

	if (zone_test_and_set_flag(zone, ZONE_RECLAIM_LOCKED))
3464 3465
		return ZONE_RECLAIM_NOSCAN;

3466 3467 3468
	ret = __zone_reclaim(zone, gfp_mask, order);
	zone_clear_flag(zone, ZONE_RECLAIM_LOCKED);

3469 3470 3471
	if (!ret)
		count_vm_event(PGSCAN_ZONE_RECLAIM_FAILED);

3472
	return ret;
3473
}
3474
#endif
L
Lee Schermerhorn 已提交
3475 3476 3477 3478 3479 3480 3481

/*
 * page_evictable - test whether a page is evictable
 * @page: the page to test
 * @vma: the VMA in which the page is or will be mapped, may be NULL
 *
 * Test whether page is evictable--i.e., should be placed on active/inactive
N
Nick Piggin 已提交
3482 3483
 * lists vs unevictable list.  The vma argument is !NULL when called from the
 * fault path to determine how to instantate a new page.
L
Lee Schermerhorn 已提交
3484 3485
 *
 * Reasons page might not be evictable:
3486
 * (1) page's mapping marked unevictable
N
Nick Piggin 已提交
3487
 * (2) page is part of an mlocked VMA
3488
 *
L
Lee Schermerhorn 已提交
3489 3490 3491 3492
 */
int page_evictable(struct page *page, struct vm_area_struct *vma)
{

3493 3494 3495
	if (mapping_unevictable(page_mapping(page)))
		return 0;

N
Nick Piggin 已提交
3496 3497
	if (PageMlocked(page) || (vma && is_mlocked_vma(vma, page)))
		return 0;
L
Lee Schermerhorn 已提交
3498 3499 3500

	return 1;
}
3501

3502
#ifdef CONFIG_SHMEM
3503 3504 3505 3506 3507 3508 3509 3510 3511 3512
/**
 * check_move_unevictable_page - check page for evictability and move to appropriate zone lru list
 * @page: page to check evictability and move to appropriate lru list
 * @zone: zone page is in
 *
 * Checks a page for evictability and moves the page to the appropriate
 * zone lru list.
 *
 * Restrictions: zone->lru_lock must be held, page must be on LRU and must
 * have PageUnevictable set.
3513 3514
 *
 * This function is only used for SysV IPC SHM_UNLOCK.
3515 3516 3517
 */
static void check_move_unevictable_page(struct page *page, struct zone *zone)
{
3518
	struct lruvec *lruvec;
3519

3520
	VM_BUG_ON(PageActive(page));
3521 3522 3523
retry:
	ClearPageUnevictable(page);
	if (page_evictable(page, NULL)) {
3524
		enum lru_list l = page_lru_base_type(page);
3525

3526
		__dec_zone_state(zone, NR_UNEVICTABLE);
3527 3528 3529
		lruvec = mem_cgroup_lru_move_lists(zone, page,
						   LRU_UNEVICTABLE, l);
		list_move(&page->lru, &lruvec->lists[l]);
3530 3531 3532 3533 3534 3535 3536
		__inc_zone_state(zone, NR_INACTIVE_ANON + l);
		__count_vm_event(UNEVICTABLE_PGRESCUED);
	} else {
		/*
		 * rotate unevictable list
		 */
		SetPageUnevictable(page);
3537 3538 3539
		lruvec = mem_cgroup_lru_move_lists(zone, page, LRU_UNEVICTABLE,
						   LRU_UNEVICTABLE);
		list_move(&page->lru, &lruvec->lists[LRU_UNEVICTABLE]);
3540 3541 3542 3543 3544 3545 3546 3547 3548 3549 3550
		if (page_evictable(page, NULL))
			goto retry;
	}
}

/**
 * scan_mapping_unevictable_pages - scan an address space for evictable pages
 * @mapping: struct address_space to scan for evictable pages
 *
 * Scan all pages in mapping.  Check unevictable pages for
 * evictability and move them to the appropriate zone lru list.
3551 3552
 *
 * This function is only used for SysV IPC SHM_UNLOCK.
3553 3554 3555 3556 3557 3558 3559 3560 3561 3562 3563 3564 3565 3566 3567 3568 3569 3570 3571 3572 3573 3574 3575 3576 3577 3578 3579 3580 3581 3582 3583 3584 3585 3586 3587 3588 3589 3590 3591 3592 3593 3594 3595 3596 3597
 */
void scan_mapping_unevictable_pages(struct address_space *mapping)
{
	pgoff_t next = 0;
	pgoff_t end   = (i_size_read(mapping->host) + PAGE_CACHE_SIZE - 1) >>
			 PAGE_CACHE_SHIFT;
	struct zone *zone;
	struct pagevec pvec;

	if (mapping->nrpages == 0)
		return;

	pagevec_init(&pvec, 0);
	while (next < end &&
		pagevec_lookup(&pvec, mapping, next, PAGEVEC_SIZE)) {
		int i;
		int pg_scanned = 0;

		zone = NULL;

		for (i = 0; i < pagevec_count(&pvec); i++) {
			struct page *page = pvec.pages[i];
			pgoff_t page_index = page->index;
			struct zone *pagezone = page_zone(page);

			pg_scanned++;
			if (page_index > next)
				next = page_index;
			next++;

			if (pagezone != zone) {
				if (zone)
					spin_unlock_irq(&zone->lru_lock);
				zone = pagezone;
				spin_lock_irq(&zone->lru_lock);
			}

			if (PageLRU(page) && PageUnevictable(page))
				check_move_unevictable_page(page, zone);
		}
		if (zone)
			spin_unlock_irq(&zone->lru_lock);
		pagevec_release(&pvec);

		count_vm_events(UNEVICTABLE_PGSCANNED, pg_scanned);
3598
		cond_resched();
3599 3600
	}
}
3601 3602 3603 3604 3605
#else
void scan_mapping_unevictable_pages(struct address_space *mapping)
{
}
#endif /* CONFIG_SHMEM */
3606

3607
static void warn_scan_unevictable_pages(void)
3608
{
3609
	printk_once(KERN_WARNING
3610
		    "%s: The scan_unevictable_pages sysctl/node-interface has been "
3611
		    "disabled for lack of a legitimate use case.  If you have "
3612 3613
		    "one, please send an email to linux-mm@kvack.org.\n",
		    current->comm);
3614 3615 3616 3617 3618 3619 3620 3621 3622
}

/*
 * scan_unevictable_pages [vm] sysctl handler.  On demand re-scan of
 * all nodes' unevictable lists for evictable pages
 */
unsigned long scan_unevictable_pages;

int scan_unevictable_handler(struct ctl_table *table, int write,
3623
			   void __user *buffer,
3624 3625
			   size_t *length, loff_t *ppos)
{
3626
	warn_scan_unevictable_pages();
3627
	proc_doulongvec_minmax(table, write, buffer, length, ppos);
3628 3629 3630 3631
	scan_unevictable_pages = 0;
	return 0;
}

3632
#ifdef CONFIG_NUMA
3633 3634 3635 3636 3637
/*
 * per node 'scan_unevictable_pages' attribute.  On demand re-scan of
 * a specified node's per zone unevictable lists for evictable pages.
 */

3638 3639
static ssize_t read_scan_unevictable_node(struct device *dev,
					  struct device_attribute *attr,
3640 3641
					  char *buf)
{
3642
	warn_scan_unevictable_pages();
3643 3644 3645
	return sprintf(buf, "0\n");	/* always zero; should fit... */
}

3646 3647
static ssize_t write_scan_unevictable_node(struct device *dev,
					   struct device_attribute *attr,
3648 3649
					const char *buf, size_t count)
{
3650
	warn_scan_unevictable_pages();
3651 3652 3653 3654
	return 1;
}


3655
static DEVICE_ATTR(scan_unevictable_pages, S_IRUGO | S_IWUSR,
3656 3657 3658 3659 3660
			read_scan_unevictable_node,
			write_scan_unevictable_node);

int scan_unevictable_register_node(struct node *node)
{
3661
	return device_create_file(&node->dev, &dev_attr_scan_unevictable_pages);
3662 3663 3664 3665
}

void scan_unevictable_unregister_node(struct node *node)
{
3666
	device_remove_file(&node->dev, &dev_attr_scan_unevictable_pages);
3667
}
3668
#endif