vmscan.c 95.6 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;

98
	int swappiness;
99

A
Andy Whitcroft 已提交
100
	int order;
101

102
	/*
103 104
	 * Intend to reclaim enough continuous memory rather than reclaim
	 * enough amount of memory. i.e, mode for high order allocation.
105
	 */
106
	reclaim_mode_t reclaim_mode;
107

108 109 110
	/* Which cgroup do we reclaim from */
	struct mem_cgroup *mem_cgroup;

111 112 113 114 115
	/*
	 * Nodemask of nodes allowed by the caller. If NULL, all nodes
	 * are scanned.
	 */
	nodemask_t	*nodemask;
L
Linus Torvalds 已提交
116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151
};

#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;
152
long vm_total_pages;	/* The total number of pages which the VM controls */
L
Linus Torvalds 已提交
153 154 155 156

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

157
#ifdef CONFIG_CGROUP_MEM_RES_CTLR
158
#define scanning_global_lru(sc)	(!(sc)->mem_cgroup)
159
#else
160
#define scanning_global_lru(sc)	(1)
161 162
#endif

163 164 165
static struct zone_reclaim_stat *get_reclaim_stat(struct zone *zone,
						  struct scan_control *sc)
{
166
	if (!scanning_global_lru(sc))
K
KOSAKI Motohiro 已提交
167 168
		return mem_cgroup_get_reclaim_stat(sc->mem_cgroup, zone);

169 170 171
	return &zone->reclaim_stat;
}

172 173
static unsigned long zone_nr_lru_pages(struct zone *zone,
				struct scan_control *sc, enum lru_list lru)
174
{
175
	if (!scanning_global_lru(sc))
176
		return mem_cgroup_zone_nr_lru_pages(sc->mem_cgroup, zone, lru);
177

178 179 180 181
	return zone_page_state(zone, NR_LRU_BASE + lru);
}


L
Linus Torvalds 已提交
182 183 184
/*
 * Add a shrinker callback to be called from the vm
 */
185
void register_shrinker(struct shrinker *shrinker)
L
Linus Torvalds 已提交
186
{
187 188 189 190
	shrinker->nr = 0;
	down_write(&shrinker_rwsem);
	list_add_tail(&shrinker->list, &shrinker_list);
	up_write(&shrinker_rwsem);
L
Linus Torvalds 已提交
191
}
192
EXPORT_SYMBOL(register_shrinker);
L
Linus Torvalds 已提交
193 194 195 196

/*
 * Remove one
 */
197
void unregister_shrinker(struct shrinker *shrinker)
L
Linus Torvalds 已提交
198 199 200 201 202
{
	down_write(&shrinker_rwsem);
	list_del(&shrinker->list);
	up_write(&shrinker_rwsem);
}
203
EXPORT_SYMBOL(unregister_shrinker);
L
Linus Torvalds 已提交
204

205 206 207 208 209 210 211 212
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
213 214 215 216 217 218 219 220 221
#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
222
 * If the vm encountered mapped pages on the LRU it increase the pressure on
L
Linus Torvalds 已提交
223 224 225 226 227 228 229
 * 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.
230 231
 *
 * Returns the number of slab objects which we shrunk.
L
Linus Torvalds 已提交
232
 */
233
unsigned long shrink_slab(struct shrink_control *shrink,
234
			  unsigned long nr_pages_scanned,
235
			  unsigned long lru_pages)
L
Linus Torvalds 已提交
236 237
{
	struct shrinker *shrinker;
238
	unsigned long ret = 0;
L
Linus Torvalds 已提交
239

240 241
	if (nr_pages_scanned == 0)
		nr_pages_scanned = SWAP_CLUSTER_MAX;
L
Linus Torvalds 已提交
242

243 244 245 246 247
	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
248 249 250 251

	list_for_each_entry(shrinker, &shrinker_list, list) {
		unsigned long long delta;
		unsigned long total_scan;
252
		unsigned long max_pass;
253
		int shrink_ret = 0;
L
Linus Torvalds 已提交
254

255 256
		max_pass = do_shrinker_shrink(shrinker, shrink, 0);
		delta = (4 * nr_pages_scanned) / shrinker->seeks;
257
		delta *= max_pass;
L
Linus Torvalds 已提交
258 259
		do_div(delta, lru_pages + 1);
		shrinker->nr += delta;
260
		if (shrinker->nr < 0) {
261 262 263
			printk(KERN_ERR "shrink_slab: %pF negative objects to "
			       "delete nr=%ld\n",
			       shrinker->shrink, shrinker->nr);
264 265 266 267 268 269 270 271 272 273
			shrinker->nr = max_pass;
		}

		/*
		 * Avoid risking looping forever due to too large nr value:
		 * never try to free more than twice the estimate number of
		 * freeable entries.
		 */
		if (shrinker->nr > max_pass * 2)
			shrinker->nr = max_pass * 2;
L
Linus Torvalds 已提交
274 275 276 277

		total_scan = shrinker->nr;
		shrinker->nr = 0;

278 279 280 281
		trace_mm_shrink_slab_start(shrinker, shrink, total_scan,
					nr_pages_scanned, lru_pages,
					max_pass, delta, total_scan);

L
Linus Torvalds 已提交
282 283
		while (total_scan >= SHRINK_BATCH) {
			long this_scan = SHRINK_BATCH;
284
			int nr_before;
L
Linus Torvalds 已提交
285

286 287 288
			nr_before = do_shrinker_shrink(shrinker, shrink, 0);
			shrink_ret = do_shrinker_shrink(shrinker, shrink,
							this_scan);
L
Linus Torvalds 已提交
289 290
			if (shrink_ret == -1)
				break;
291 292
			if (shrink_ret < nr_before)
				ret += nr_before - shrink_ret;
293
			count_vm_events(SLABS_SCANNED, this_scan);
L
Linus Torvalds 已提交
294 295 296 297 298 299
			total_scan -= this_scan;

			cond_resched();
		}

		shrinker->nr += total_scan;
300 301
		trace_mm_shrink_slab_end(shrinker, shrink_ret, total_scan,
					 shrinker->nr);
L
Linus Torvalds 已提交
302 303
	}
	up_read(&shrinker_rwsem);
304 305
out:
	cond_resched();
306
	return ret;
L
Linus Torvalds 已提交
307 308
}

309
static void set_reclaim_mode(int priority, struct scan_control *sc,
310 311
				   bool sync)
{
312
	reclaim_mode_t syncmode = sync ? RECLAIM_MODE_SYNC : RECLAIM_MODE_ASYNC;
313 314

	/*
315 316 317
	 * 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.
318
	 */
319
	if (COMPACTION_BUILD)
320
		sc->reclaim_mode = RECLAIM_MODE_COMPACTION;
321
	else
322
		sc->reclaim_mode = RECLAIM_MODE_LUMPYRECLAIM;
323 324

	/*
325 326 327
	 * Avoid using lumpy reclaim or reclaim/compaction if possible by
	 * restricting when its set to either costly allocations or when
	 * under memory pressure
328 329
	 */
	if (sc->order > PAGE_ALLOC_COSTLY_ORDER)
330
		sc->reclaim_mode |= syncmode;
331
	else if (sc->order && priority < DEF_PRIORITY - 2)
332
		sc->reclaim_mode |= syncmode;
333
	else
334
		sc->reclaim_mode = RECLAIM_MODE_SINGLE | RECLAIM_MODE_ASYNC;
335 336
}

337
static void reset_reclaim_mode(struct scan_control *sc)
338
{
339
	sc->reclaim_mode = RECLAIM_MODE_SINGLE | RECLAIM_MODE_ASYNC;
340 341
}

L
Linus Torvalds 已提交
342 343
static inline int is_page_cache_freeable(struct page *page)
{
344 345 346 347 348
	/*
	 * 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.
	 */
349
	return page_count(page) - page_has_private(page) == 2;
L
Linus Torvalds 已提交
350 351
}

352 353
static int may_write_to_queue(struct backing_dev_info *bdi,
			      struct scan_control *sc)
L
Linus Torvalds 已提交
354
{
355
	if (current->flags & PF_SWAPWRITE)
L
Linus Torvalds 已提交
356 357 358 359 360
		return 1;
	if (!bdi_write_congested(bdi))
		return 1;
	if (bdi == current->backing_dev_info)
		return 1;
361 362 363 364

	/* 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
365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382
	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
383
	lock_page(page);
384 385
	if (page_mapping(page) == mapping)
		mapping_set_error(mapping, error);
L
Linus Torvalds 已提交
386 387 388
	unlock_page(page);
}

389 390 391 392 393 394 395 396 397 398 399 400
/* 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
401
/*
A
Andrew Morton 已提交
402 403
 * pageout is called by shrink_page_list() for each dirty page.
 * Calls ->writepage().
L
Linus Torvalds 已提交
404
 */
405
static pageout_t pageout(struct page *page, struct address_space *mapping,
406
			 struct scan_control *sc)
L
Linus Torvalds 已提交
407 408 409 410 411 412 413 414
{
	/*
	 * 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.
	 *
415
	 * If this process is currently in __generic_file_aio_write() against
L
Linus Torvalds 已提交
416 417 418 419 420 421 422 423 424 425 426 427 428 429 430
	 * 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.
		 */
431
		if (page_has_private(page)) {
L
Linus Torvalds 已提交
432 433
			if (try_to_free_buffers(page)) {
				ClearPageDirty(page);
434
				printk("%s: orphaned page\n", __func__);
L
Linus Torvalds 已提交
435 436 437 438 439 440 441
				return PAGE_CLEAN;
			}
		}
		return PAGE_KEEP;
	}
	if (mapping->a_ops->writepage == NULL)
		return PAGE_ACTIVATE;
442
	if (!may_write_to_queue(mapping->backing_dev_info, sc))
L
Linus Torvalds 已提交
443 444 445 446 447 448 449
		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,
450 451
			.range_start = 0,
			.range_end = LLONG_MAX,
L
Linus Torvalds 已提交
452 453 454 455 456 457 458
			.for_reclaim = 1,
		};

		SetPageReclaim(page);
		res = mapping->a_ops->writepage(page, &wbc);
		if (res < 0)
			handle_write_error(mapping, page, res);
459
		if (res == AOP_WRITEPAGE_ACTIVATE) {
L
Linus Torvalds 已提交
460 461 462
			ClearPageReclaim(page);
			return PAGE_ACTIVATE;
		}
463 464 465 466 467 468

		/*
		 * Wait on writeback if requested to. This happens when
		 * direct reclaiming a large contiguous area and the
		 * first attempt to free a range of pages fails.
		 */
469
		if (PageWriteback(page) &&
470
		    (sc->reclaim_mode & RECLAIM_MODE_SYNC))
471 472
			wait_on_page_writeback(page);

L
Linus Torvalds 已提交
473 474 475 476
		if (!PageWriteback(page)) {
			/* synchronous write or broken a_ops? */
			ClearPageReclaim(page);
		}
477
		trace_mm_vmscan_writepage(page,
478
			trace_reclaim_flags(page, sc->reclaim_mode));
479
		inc_zone_page_state(page, NR_VMSCAN_WRITE);
L
Linus Torvalds 已提交
480 481 482 483 484 485
		return PAGE_SUCCESS;
	}

	return PAGE_CLEAN;
}

486
/*
N
Nick Piggin 已提交
487 488
 * Same as remove_mapping, but if the page is removed from the mapping, it
 * gets returned with a refcount of 0.
489
 */
N
Nick Piggin 已提交
490
static int __remove_mapping(struct address_space *mapping, struct page *page)
491
{
492 493
	BUG_ON(!PageLocked(page));
	BUG_ON(mapping != page_mapping(page));
494

N
Nick Piggin 已提交
495
	spin_lock_irq(&mapping->tree_lock);
496
	/*
N
Nick Piggin 已提交
497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519
	 * 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.
520
	 */
N
Nick Piggin 已提交
521
	if (!page_freeze_refs(page, 2))
522
		goto cannot_free;
N
Nick Piggin 已提交
523 524 525
	/* note: atomic_cmpxchg in page_freeze_refs provides the smp_rmb */
	if (unlikely(PageDirty(page))) {
		page_unfreeze_refs(page, 2);
526
		goto cannot_free;
N
Nick Piggin 已提交
527
	}
528 529 530 531

	if (PageSwapCache(page)) {
		swp_entry_t swap = { .val = page_private(page) };
		__delete_from_swap_cache(page);
N
Nick Piggin 已提交
532
		spin_unlock_irq(&mapping->tree_lock);
533
		swapcache_free(swap, page);
N
Nick Piggin 已提交
534
	} else {
535 536 537 538
		void (*freepage)(struct page *);

		freepage = mapping->a_ops->freepage;

539
		__delete_from_page_cache(page);
N
Nick Piggin 已提交
540
		spin_unlock_irq(&mapping->tree_lock);
541
		mem_cgroup_uncharge_cache_page(page);
542 543 544

		if (freepage != NULL)
			freepage(page);
545 546 547 548 549
	}

	return 1;

cannot_free:
N
Nick Piggin 已提交
550
	spin_unlock_irq(&mapping->tree_lock);
551 552 553
	return 0;
}

N
Nick Piggin 已提交
554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573
/*
 * 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
574 575 576 577 578 579 580 581 582 583 584 585 586
/**
 * 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);
587
	int was_unevictable = PageUnevictable(page);
L
Lee Schermerhorn 已提交
588 589 590 591 592 593 594 595 596 597 598 599 600

	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.
		 */
601
		lru = active + page_lru_base_type(page);
L
Lee Schermerhorn 已提交
602 603 604 605 606 607 608 609
		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);
610 611 612 613 614 615 616 617 618 619
		/*
		 * When racing with an mlock clearing (page is
		 * unlocked), make sure that if the other thread does
		 * not observe our setting of PG_lru and fails
		 * isolation, we see PG_mlocked cleared below and move
		 * the page back to the evictable list.
		 *
		 * The other side is TestClearPageMlocked().
		 */
		smp_mb();
L
Lee Schermerhorn 已提交
620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637
	}

	/*
	 * 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.
		 */
	}

638 639 640 641 642
	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
643 644 645
	put_page(page);		/* drop ref from isolate */
}

646 647 648
enum page_references {
	PAGEREF_RECLAIM,
	PAGEREF_RECLAIM_CLEAN,
649
	PAGEREF_KEEP,
650 651 652 653 654 655
	PAGEREF_ACTIVATE,
};

static enum page_references page_check_references(struct page *page,
						  struct scan_control *sc)
{
656
	int referenced_ptes, referenced_page;
657 658
	unsigned long vm_flags;

659 660
	referenced_ptes = page_referenced(page, 1, sc->mem_cgroup, &vm_flags);
	referenced_page = TestClearPageReferenced(page);
661 662

	/* Lumpy reclaim - ignore references */
663
	if (sc->reclaim_mode & RECLAIM_MODE_LUMPYRECLAIM)
664 665 666 667 668 669 670 671 672
		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;

673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696
	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);

		if (referenced_page)
			return PAGEREF_ACTIVATE;

		return PAGEREF_KEEP;
	}
697 698

	/* Reclaim if clean, defer dirty pages to writeback */
699
	if (referenced_page && !PageSwapBacked(page))
700 701 702
		return PAGEREF_RECLAIM_CLEAN;

	return PAGEREF_RECLAIM;
703 704
}

705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722
static noinline_for_stack void free_page_list(struct list_head *free_pages)
{
	struct pagevec freed_pvec;
	struct page *page, *tmp;

	pagevec_init(&freed_pvec, 1);

	list_for_each_entry_safe(page, tmp, free_pages, lru) {
		list_del(&page->lru);
		if (!pagevec_add(&freed_pvec, page)) {
			__pagevec_free(&freed_pvec);
			pagevec_reinit(&freed_pvec);
		}
	}

	pagevec_free(&freed_pvec);
}

L
Linus Torvalds 已提交
723
/*
A
Andrew Morton 已提交
724
 * shrink_page_list() returns the number of reclaimed pages
L
Linus Torvalds 已提交
725
 */
A
Andrew Morton 已提交
726
static unsigned long shrink_page_list(struct list_head *page_list,
727
				      struct zone *zone,
728
				      struct scan_control *sc)
L
Linus Torvalds 已提交
729 730
{
	LIST_HEAD(ret_pages);
731
	LIST_HEAD(free_pages);
L
Linus Torvalds 已提交
732
	int pgactivate = 0;
733 734
	unsigned long nr_dirty = 0;
	unsigned long nr_congested = 0;
735
	unsigned long nr_reclaimed = 0;
L
Linus Torvalds 已提交
736 737 738 739

	cond_resched();

	while (!list_empty(page_list)) {
740
		enum page_references references;
L
Linus Torvalds 已提交
741 742 743 744 745 746 747 748 749
		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
750
		if (!trylock_page(page))
L
Linus Torvalds 已提交
751 752
			goto keep;

N
Nick Piggin 已提交
753
		VM_BUG_ON(PageActive(page));
754
		VM_BUG_ON(page_zone(page) != zone);
L
Linus Torvalds 已提交
755 756

		sc->nr_scanned++;
757

N
Nick Piggin 已提交
758 759
		if (unlikely(!page_evictable(page, NULL)))
			goto cull_mlocked;
L
Lee Schermerhorn 已提交
760

761
		if (!sc->may_unmap && page_mapped(page))
762 763
			goto keep_locked;

L
Linus Torvalds 已提交
764 765 766 767
		/* Double the slab pressure for mapped and swapcache pages */
		if (page_mapped(page) || PageSwapCache(page))
			sc->nr_scanned++;

768 769 770 771 772 773 774 775 776 777 778 779
		may_enter_fs = (sc->gfp_mask & __GFP_FS) ||
			(PageSwapCache(page) && (sc->gfp_mask & __GFP_IO));

		if (PageWriteback(page)) {
			/*
			 * Synchronous reclaim is performed in two passes,
			 * first an asynchronous pass over the list to
			 * start parallel writeback, and a second synchronous
			 * pass to wait for the IO to complete.  Wait here
			 * for any page for which writeback has already
			 * started.
			 */
780
			if ((sc->reclaim_mode & RECLAIM_MODE_SYNC) &&
781
			    may_enter_fs)
782
				wait_on_page_writeback(page);
783 784 785 786
			else {
				unlock_page(page);
				goto keep_lumpy;
			}
787
		}
L
Linus Torvalds 已提交
788

789 790 791
		references = page_check_references(page, sc);
		switch (references) {
		case PAGEREF_ACTIVATE:
L
Linus Torvalds 已提交
792
			goto activate_locked;
793 794
		case PAGEREF_KEEP:
			goto keep_locked;
795 796 797 798
		case PAGEREF_RECLAIM:
		case PAGEREF_RECLAIM_CLEAN:
			; /* try to reclaim the page below */
		}
L
Linus Torvalds 已提交
799 800 801 802 803

		/*
		 * Anonymous process memory has backing store?
		 * Try to allocate it some swap space here.
		 */
N
Nick Piggin 已提交
804
		if (PageAnon(page) && !PageSwapCache(page)) {
805 806
			if (!(sc->gfp_mask & __GFP_IO))
				goto keep_locked;
807
			if (!add_to_swap(page))
L
Linus Torvalds 已提交
808
				goto activate_locked;
809
			may_enter_fs = 1;
N
Nick Piggin 已提交
810
		}
L
Linus Torvalds 已提交
811 812 813 814 815 816 817 818

		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) {
819
			switch (try_to_unmap(page, TTU_UNMAP)) {
L
Linus Torvalds 已提交
820 821 822 823
			case SWAP_FAIL:
				goto activate_locked;
			case SWAP_AGAIN:
				goto keep_locked;
N
Nick Piggin 已提交
824 825
			case SWAP_MLOCK:
				goto cull_mlocked;
L
Linus Torvalds 已提交
826 827 828 829 830 831
			case SWAP_SUCCESS:
				; /* try to free the page below */
			}
		}

		if (PageDirty(page)) {
832 833
			nr_dirty++;

834
			if (references == PAGEREF_RECLAIM_CLEAN)
L
Linus Torvalds 已提交
835
				goto keep_locked;
836
			if (!may_enter_fs)
L
Linus Torvalds 已提交
837
				goto keep_locked;
838
			if (!sc->may_writepage)
L
Linus Torvalds 已提交
839 840 841
				goto keep_locked;

			/* Page is dirty, try to write it out here */
842
			switch (pageout(page, mapping, sc)) {
L
Linus Torvalds 已提交
843
			case PAGE_KEEP:
844
				nr_congested++;
L
Linus Torvalds 已提交
845 846 847 848
				goto keep_locked;
			case PAGE_ACTIVATE:
				goto activate_locked;
			case PAGE_SUCCESS:
849 850 851
				if (PageWriteback(page))
					goto keep_lumpy;
				if (PageDirty(page))
L
Linus Torvalds 已提交
852
					goto keep;
853

L
Linus Torvalds 已提交
854 855 856 857
				/*
				 * A synchronous write - probably a ramdisk.  Go
				 * ahead and try to reclaim the page.
				 */
N
Nick Piggin 已提交
858
				if (!trylock_page(page))
L
Linus Torvalds 已提交
859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877
					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
878
		 * will do this, as well as the blockdev mapping.
L
Linus Torvalds 已提交
879 880 881 882 883 884 885 886 887 888
		 * 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.
		 */
889
		if (page_has_private(page)) {
L
Linus Torvalds 已提交
890 891
			if (!try_to_release_page(page, sc->gfp_mask))
				goto activate_locked;
N
Nick Piggin 已提交
892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907
			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
908 909
		}

N
Nick Piggin 已提交
910
		if (!mapping || !__remove_mapping(mapping, page))
911
			goto keep_locked;
L
Linus Torvalds 已提交
912

N
Nick Piggin 已提交
913 914 915 916 917 918 919 920
		/*
		 * 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
921
free_it:
922
		nr_reclaimed++;
923 924 925 926 927 928

		/*
		 * 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
929 930
		continue;

N
Nick Piggin 已提交
931
cull_mlocked:
932 933
		if (PageSwapCache(page))
			try_to_free_swap(page);
N
Nick Piggin 已提交
934 935
		unlock_page(page);
		putback_lru_page(page);
936
		reset_reclaim_mode(sc);
N
Nick Piggin 已提交
937 938
		continue;

L
Linus Torvalds 已提交
939
activate_locked:
940 941
		/* Not a candidate for swapping, so reclaim swap space. */
		if (PageSwapCache(page) && vm_swap_full())
942
			try_to_free_swap(page);
L
Lee Schermerhorn 已提交
943
		VM_BUG_ON(PageActive(page));
L
Linus Torvalds 已提交
944 945 946 947 948
		SetPageActive(page);
		pgactivate++;
keep_locked:
		unlock_page(page);
keep:
949
		reset_reclaim_mode(sc);
950
keep_lumpy:
L
Linus Torvalds 已提交
951
		list_add(&page->lru, &ret_pages);
N
Nick Piggin 已提交
952
		VM_BUG_ON(PageLRU(page) || PageUnevictable(page));
L
Linus Torvalds 已提交
953
	}
954

955 956 957 958 959 960
	/*
	 * 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
	 */
K
KAMEZAWA Hiroyuki 已提交
961
	if (nr_dirty && nr_dirty == nr_congested && scanning_global_lru(sc))
962 963
		zone_set_flag(zone, ZONE_CONGESTED);

964 965
	free_page_list(&free_pages);

L
Linus Torvalds 已提交
966
	list_splice(&ret_pages, page_list);
967
	count_vm_events(PGACTIVATE, pgactivate);
968
	return nr_reclaimed;
L
Linus Torvalds 已提交
969 970
}

A
Andy Whitcroft 已提交
971 972 973 974 975 976 977 978 979 980
/*
 * 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.
 */
981
int __isolate_lru_page(struct page *page, int mode, int file)
A
Andy Whitcroft 已提交
982 983 984 985 986 987 988 989 990 991 992 993 994 995 996
{
	int ret = -EINVAL;

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

	/*
	 * When checking the active state, we need to be sure we are
	 * dealing with comparible boolean values.  Take the logical not
	 * of each.
	 */
	if (mode != ISOLATE_BOTH && (!PageActive(page) != !mode))
		return ret;

997
	if (mode != ISOLATE_BOTH && page_is_file_cache(page) != file)
998 999
		return ret;

L
Lee Schermerhorn 已提交
1000 1001 1002 1003 1004 1005 1006 1007
	/*
	 * 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
1008
	ret = -EBUSY;
K
KAMEZAWA Hiroyuki 已提交
1009

A
Andy Whitcroft 已提交
1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022
	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
1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036
/*
 * 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.
 * @src:	The LRU list to pull pages off.
 * @dst:	The temp list to put pages on to.
 * @scanned:	The number of pages that were scanned.
A
Andy Whitcroft 已提交
1037 1038
 * @order:	The caller's attempted allocation order
 * @mode:	One of the LRU isolation modes
1039
 * @file:	True [1] if isolating file [!anon] pages
L
Linus Torvalds 已提交
1040 1041 1042
 *
 * returns how many pages were moved onto *@dst.
 */
1043 1044
static unsigned long isolate_lru_pages(unsigned long nr_to_scan,
		struct list_head *src, struct list_head *dst,
1045
		unsigned long *scanned, int order, int mode, int file)
L
Linus Torvalds 已提交
1046
{
1047
	unsigned long nr_taken = 0;
1048 1049 1050
	unsigned long nr_lumpy_taken = 0;
	unsigned long nr_lumpy_dirty = 0;
	unsigned long nr_lumpy_failed = 0;
1051
	unsigned long scan;
L
Linus Torvalds 已提交
1052

1053
	for (scan = 0; scan < nr_to_scan && !list_empty(src); scan++) {
A
Andy Whitcroft 已提交
1054 1055 1056 1057 1058 1059
		struct page *page;
		unsigned long pfn;
		unsigned long end_pfn;
		unsigned long page_pfn;
		int zone_id;

L
Linus Torvalds 已提交
1060 1061 1062
		page = lru_to_page(src);
		prefetchw_prev_lru_page(page, src, flags);

N
Nick Piggin 已提交
1063
		VM_BUG_ON(!PageLRU(page));
N
Nick Piggin 已提交
1064

1065
		switch (__isolate_lru_page(page, mode, file)) {
A
Andy Whitcroft 已提交
1066 1067
		case 0:
			list_move(&page->lru, dst);
1068
			mem_cgroup_del_lru(page);
1069
			nr_taken += hpage_nr_pages(page);
A
Andy Whitcroft 已提交
1070 1071 1072 1073 1074
			break;

		case -EBUSY:
			/* else it is being freed elsewhere */
			list_move(&page->lru, src);
1075
			mem_cgroup_rotate_lru_list(page, page_lru(page));
A
Andy Whitcroft 已提交
1076
			continue;
1077

A
Andy Whitcroft 已提交
1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089
		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
1090
		 * as the mem_map is guaranteed valid out to MAX_ORDER,
A
Andy Whitcroft 已提交
1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109
		 * 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);
1110

A
Andy Whitcroft 已提交
1111 1112
			/* Check that we have not crossed a zone boundary. */
			if (unlikely(page_zone_id(cursor_page) != zone_id))
1113
				break;
1114 1115 1116 1117 1118 1119 1120

			/*
			 * If we don't have enough swap space, reclaiming of
			 * anon page which don't already have a swap slot is
			 * pointless.
			 */
			if (nr_swap_pages <= 0 && PageAnon(cursor_page) &&
1121 1122
			    !PageSwapCache(cursor_page))
				break;
1123

1124
			if (__isolate_lru_page(cursor_page, mode, file) == 0) {
A
Andy Whitcroft 已提交
1125
				list_move(&cursor_page->lru, dst);
1126
				mem_cgroup_del_lru(cursor_page);
1127
				nr_taken += hpage_nr_pages(page);
1128 1129 1130
				nr_lumpy_taken++;
				if (PageDirty(cursor_page))
					nr_lumpy_dirty++;
A
Andy Whitcroft 已提交
1131
				scan++;
1132
			} else {
1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146
				/*
				 * 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))
1147 1148
					continue;
				break;
A
Andy Whitcroft 已提交
1149 1150
			}
		}
1151 1152 1153 1154

		/* If we break out of the loop above, lumpy reclaim failed */
		if (pfn < end_pfn)
			nr_lumpy_failed++;
L
Linus Torvalds 已提交
1155 1156 1157
	}

	*scanned = scan;
1158 1159 1160 1161 1162 1163

	trace_mm_vmscan_lru_isolate(order,
			nr_to_scan, scan,
			nr_taken,
			nr_lumpy_taken, nr_lumpy_dirty, nr_lumpy_failed,
			mode);
L
Linus Torvalds 已提交
1164 1165 1166
	return nr_taken;
}

1167 1168 1169 1170
static unsigned long isolate_pages_global(unsigned long nr,
					struct list_head *dst,
					unsigned long *scanned, int order,
					int mode, struct zone *z,
1171
					int active, int file)
1172
{
1173
	int lru = LRU_BASE;
1174
	if (active)
1175 1176 1177 1178
		lru += LRU_ACTIVE;
	if (file)
		lru += LRU_FILE;
	return isolate_lru_pages(nr, &z->lru[lru].list, dst, scanned, order,
1179
								mode, file);
1180 1181
}

A
Andy Whitcroft 已提交
1182 1183 1184 1185
/*
 * clear_active_flags() is a helper for shrink_active_list(), clearing
 * any active bits from the pages in the list.
 */
1186 1187
static unsigned long clear_active_flags(struct list_head *page_list,
					unsigned int *count)
A
Andy Whitcroft 已提交
1188 1189
{
	int nr_active = 0;
1190
	int lru;
A
Andy Whitcroft 已提交
1191 1192
	struct page *page;

1193
	list_for_each_entry(page, page_list, lru) {
1194
		int numpages = hpage_nr_pages(page);
1195
		lru = page_lru_base_type(page);
A
Andy Whitcroft 已提交
1196
		if (PageActive(page)) {
1197
			lru += LRU_ACTIVE;
A
Andy Whitcroft 已提交
1198
			ClearPageActive(page);
1199
			nr_active += numpages;
A
Andy Whitcroft 已提交
1200
		}
1201
		if (count)
1202
			count[lru] += numpages;
1203
	}
A
Andy Whitcroft 已提交
1204 1205 1206 1207

	return nr_active;
}

1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218
/**
 * 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
1219 1220 1221
 * 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.
1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236
 *
 * 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;

1237 1238
	VM_BUG_ON(!page_count(page));

1239 1240 1241 1242
	if (PageLRU(page)) {
		struct zone *zone = page_zone(page);

		spin_lock_irq(&zone->lru_lock);
1243
		if (PageLRU(page)) {
L
Lee Schermerhorn 已提交
1244
			int lru = page_lru(page);
1245
			ret = 0;
1246
			get_page(page);
1247
			ClearPageLRU(page);
1248 1249

			del_page_from_lru_list(zone, page, lru);
1250 1251 1252 1253 1254 1255
		}
		spin_unlock_irq(&zone->lru_lock);
	}
	return ret;
}

1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280
/*
 * 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;

	if (!scanning_global_lru(sc))
		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;
}

1281 1282 1283 1284
/*
 * TODO: Try merging with migrations version of putback_lru_pages
 */
static noinline_for_stack void
1285
putback_lru_pages(struct zone *zone, struct scan_control *sc,
1286 1287 1288 1289 1290
				unsigned long nr_anon, unsigned long nr_file,
				struct list_head *page_list)
{
	struct page *page;
	struct pagevec pvec;
1291
	struct zone_reclaim_stat *reclaim_stat = get_reclaim_stat(zone, sc);
1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309

	pagevec_init(&pvec, 1);

	/*
	 * Put back any unfreeable pages.
	 */
	spin_lock(&zone->lru_lock);
	while (!list_empty(page_list)) {
		int lru;
		page = lru_to_page(page_list);
		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;
		}
1310
		SetPageLRU(page);
1311
		lru = page_lru(page);
1312
		add_page_to_lru_list(zone, page, lru);
1313 1314
		if (is_active_lru(lru)) {
			int file = is_file_lru(lru);
1315 1316
			int numpages = hpage_nr_pages(page);
			reclaim_stat->recent_rotated[file] += numpages;
1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330
		}
		if (!pagevec_add(&pvec, page)) {
			spin_unlock_irq(&zone->lru_lock);
			__pagevec_release(&pvec);
			spin_lock_irq(&zone->lru_lock);
		}
	}
	__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);
	pagevec_release(&pvec);
}

1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361
static noinline_for_stack void update_isolated_counts(struct zone *zone,
					struct scan_control *sc,
					unsigned long *nr_anon,
					unsigned long *nr_file,
					struct list_head *isolated_list)
{
	unsigned long nr_active;
	unsigned int count[NR_LRU_LISTS] = { 0, };
	struct zone_reclaim_stat *reclaim_stat = get_reclaim_stat(zone, sc);

	nr_active = clear_active_flags(isolated_list, count);
	__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];
	__mod_zone_page_state(zone, NR_ISOLATED_ANON, *nr_anon);
	__mod_zone_page_state(zone, NR_ISOLATED_FILE, *nr_file);

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

1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381
/*
 * Returns true if the caller should wait to clean dirty/writeback pages.
 *
 * 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 */
1382
	if (sc->reclaim_mode & RECLAIM_MODE_SINGLE)
1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402
		return false;

	/* If we have relaimed everything on the isolated list, no stall */
	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
1403
/*
A
Andrew Morton 已提交
1404 1405
 * shrink_inactive_list() is a helper for shrink_zone().  It returns the number
 * of reclaimed pages
L
Linus Torvalds 已提交
1406
 */
1407 1408 1409
static noinline_for_stack unsigned long
shrink_inactive_list(unsigned long nr_to_scan, struct zone *zone,
			struct scan_control *sc, int priority, int file)
L
Linus Torvalds 已提交
1410 1411
{
	LIST_HEAD(page_list);
1412
	unsigned long nr_scanned;
1413
	unsigned long nr_reclaimed = 0;
1414 1415 1416
	unsigned long nr_taken;
	unsigned long nr_anon;
	unsigned long nr_file;
1417

1418
	while (unlikely(too_many_isolated(zone, file, sc))) {
1419
		congestion_wait(BLK_RW_ASYNC, HZ/10);
1420 1421 1422 1423 1424 1425

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

1426
	set_reclaim_mode(priority, sc, false);
L
Linus Torvalds 已提交
1427 1428
	lru_add_drain();
	spin_lock_irq(&zone->lru_lock);
1429

1430 1431 1432
	if (scanning_global_lru(sc)) {
		nr_taken = isolate_pages_global(nr_to_scan,
			&page_list, &nr_scanned, sc->order,
1433
			sc->reclaim_mode & RECLAIM_MODE_LUMPYRECLAIM ?
1434
					ISOLATE_BOTH : ISOLATE_INACTIVE,
1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445
			zone, 0, file);
		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);
	} else {
		nr_taken = mem_cgroup_isolate_pages(nr_to_scan,
			&page_list, &nr_scanned, sc->order,
1446
			sc->reclaim_mode & RECLAIM_MODE_LUMPYRECLAIM ?
1447
					ISOLATE_BOTH : ISOLATE_INACTIVE,
1448 1449 1450 1451 1452 1453 1454
			zone, sc->mem_cgroup,
			0, file);
		/*
		 * mem_cgroup_isolate_pages() keeps track of
		 * scanned pages on its own.
		 */
	}
1455

1456 1457 1458 1459
	if (nr_taken == 0) {
		spin_unlock_irq(&zone->lru_lock);
		return 0;
	}
A
Andy Whitcroft 已提交
1460

1461
	update_isolated_counts(zone, sc, &nr_anon, &nr_file, &page_list);
L
Linus Torvalds 已提交
1462

1463
	spin_unlock_irq(&zone->lru_lock);
1464

1465
	nr_reclaimed = shrink_page_list(&page_list, zone, sc);
1466

1467 1468
	/* Check if we should syncronously wait for writeback */
	if (should_reclaim_stall(nr_taken, nr_reclaimed, priority, sc)) {
1469
		set_reclaim_mode(priority, sc, true);
1470
		nr_reclaimed += shrink_page_list(&page_list, zone, sc);
1471
	}
1472

1473 1474 1475 1476
	local_irq_disable();
	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
1477

1478
	putback_lru_pages(zone, sc, nr_anon, nr_file, &page_list);
1479 1480 1481 1482 1483

	trace_mm_vmscan_lru_shrink_inactive(zone->zone_pgdat->node_id,
		zone_idx(zone),
		nr_scanned, nr_reclaimed,
		priority,
1484
		trace_shrink_flags(file, sc->reclaim_mode));
1485
	return nr_reclaimed;
L
Linus Torvalds 已提交
1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504
}

/*
 * 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.
 */
1505

1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523
static void move_active_pages_to_lru(struct zone *zone,
				     struct list_head *list,
				     enum lru_list lru)
{
	unsigned long pgmoved = 0;
	struct pagevec pvec;
	struct page *page;

	pagevec_init(&pvec, 1);

	while (!list_empty(list)) {
		page = lru_to_page(list);

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

		list_move(&page->lru, &zone->lru[lru].list);
		mem_cgroup_add_lru_list(page, lru);
1524
		pgmoved += hpage_nr_pages(page);
1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537

		if (!pagevec_add(&pvec, page) || list_empty(list)) {
			spin_unlock_irq(&zone->lru_lock);
			if (buffer_heads_over_limit)
				pagevec_strip(&pvec);
			__pagevec_release(&pvec);
			spin_lock_irq(&zone->lru_lock);
		}
	}
	__mod_zone_page_state(zone, NR_LRU_BASE + lru, pgmoved);
	if (!is_active_lru(lru))
		__count_vm_events(PGDEACTIVATE, pgmoved);
}
1538

A
Andrew Morton 已提交
1539
static void shrink_active_list(unsigned long nr_pages, struct zone *zone,
1540
			struct scan_control *sc, int priority, int file)
L
Linus Torvalds 已提交
1541
{
1542
	unsigned long nr_taken;
1543
	unsigned long pgscanned;
1544
	unsigned long vm_flags;
L
Linus Torvalds 已提交
1545
	LIST_HEAD(l_hold);	/* The pages which were snipped off */
1546
	LIST_HEAD(l_active);
1547
	LIST_HEAD(l_inactive);
L
Linus Torvalds 已提交
1548
	struct page *page;
1549
	struct zone_reclaim_stat *reclaim_stat = get_reclaim_stat(zone, sc);
1550
	unsigned long nr_rotated = 0;
L
Linus Torvalds 已提交
1551 1552 1553

	lru_add_drain();
	spin_lock_irq(&zone->lru_lock);
1554
	if (scanning_global_lru(sc)) {
1555 1556 1557 1558
		nr_taken = isolate_pages_global(nr_pages, &l_hold,
						&pgscanned, sc->order,
						ISOLATE_ACTIVE, zone,
						1, file);
1559
		zone->pages_scanned += pgscanned;
1560 1561 1562 1563 1564 1565 1566 1567 1568
	} else {
		nr_taken = mem_cgroup_isolate_pages(nr_pages, &l_hold,
						&pgscanned, sc->order,
						ISOLATE_ACTIVE, zone,
						sc->mem_cgroup, 1, file);
		/*
		 * mem_cgroup_isolate_pages() keeps track of
		 * scanned pages on its own.
		 */
1569
	}
1570

1571
	reclaim_stat->recent_scanned[file] += nr_taken;
1572

1573
	__count_zone_vm_events(PGREFILL, zone, pgscanned);
1574
	if (file)
1575
		__mod_zone_page_state(zone, NR_ACTIVE_FILE, -nr_taken);
1576
	else
1577
		__mod_zone_page_state(zone, NR_ACTIVE_ANON, -nr_taken);
K
KOSAKI Motohiro 已提交
1578
	__mod_zone_page_state(zone, NR_ISOLATED_ANON + file, nr_taken);
L
Linus Torvalds 已提交
1579 1580 1581 1582 1583 1584
	spin_unlock_irq(&zone->lru_lock);

	while (!list_empty(&l_hold)) {
		cond_resched();
		page = lru_to_page(&l_hold);
		list_del(&page->lru);
1585

L
Lee Schermerhorn 已提交
1586 1587 1588 1589 1590
		if (unlikely(!page_evictable(page, NULL))) {
			putback_lru_page(page);
			continue;
		}

1591
		if (page_referenced(page, 0, sc->mem_cgroup, &vm_flags)) {
1592
			nr_rotated += hpage_nr_pages(page);
1593 1594 1595 1596 1597 1598 1599 1600 1601
			/*
			 * 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.
			 */
1602
			if ((vm_flags & VM_EXEC) && page_is_file_cache(page)) {
1603 1604 1605 1606
				list_add(&page->lru, &l_active);
				continue;
			}
		}
1607

1608
		ClearPageActive(page);	/* we are de-activating */
L
Linus Torvalds 已提交
1609 1610 1611
		list_add(&page->lru, &l_inactive);
	}

1612
	/*
1613
	 * Move pages back to the lru list.
1614
	 */
1615
	spin_lock_irq(&zone->lru_lock);
1616
	/*
1617 1618 1619 1620
	 * 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.
1621
	 */
1622
	reclaim_stat->recent_rotated[file] += nr_rotated;
1623

1624 1625 1626 1627
	move_active_pages_to_lru(zone, &l_active,
						LRU_ACTIVE + file * LRU_FILE);
	move_active_pages_to_lru(zone, &l_inactive,
						LRU_BASE   + file * LRU_FILE);
K
KOSAKI Motohiro 已提交
1628
	__mod_zone_page_state(zone, NR_ISOLATED_ANON + file, -nr_taken);
1629
	spin_unlock_irq(&zone->lru_lock);
L
Linus Torvalds 已提交
1630 1631
}

1632
#ifdef CONFIG_SWAP
1633
static int inactive_anon_is_low_global(struct zone *zone)
1634 1635 1636 1637 1638 1639 1640 1641 1642 1643 1644 1645
{
	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;
}

1646 1647 1648 1649 1650 1651 1652 1653 1654 1655 1656 1657
/**
 * 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.
 */
static int inactive_anon_is_low(struct zone *zone, struct scan_control *sc)
{
	int low;

1658 1659 1660 1661 1662 1663 1664
	/*
	 * If we don't have swap space, anonymous page deactivation
	 * is pointless.
	 */
	if (!total_swap_pages)
		return 0;

1665
	if (scanning_global_lru(sc))
1666 1667
		low = inactive_anon_is_low_global(zone);
	else
1668
		low = mem_cgroup_inactive_anon_is_low(sc->mem_cgroup);
1669 1670
	return low;
}
1671 1672 1673 1674 1675 1676 1677
#else
static inline int inactive_anon_is_low(struct zone *zone,
					struct scan_control *sc)
{
	return 0;
}
#endif
1678

1679 1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692 1693 1694 1695 1696 1697 1698 1699 1700 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714
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
 * @zone: zone to check
 * @sc:   scan control of this context
 *
 * 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.
 */
static int inactive_file_is_low(struct zone *zone, struct scan_control *sc)
{
	int low;

	if (scanning_global_lru(sc))
		low = inactive_file_is_low_global(zone);
	else
		low = mem_cgroup_inactive_file_is_low(sc->mem_cgroup);
	return low;
}

1715 1716 1717 1718 1719 1720 1721 1722 1723
static int inactive_list_is_low(struct zone *zone, struct scan_control *sc,
				int file)
{
	if (file)
		return inactive_file_is_low(zone, sc);
	else
		return inactive_anon_is_low(zone, sc);
}

1724
static unsigned long shrink_list(enum lru_list lru, unsigned long nr_to_scan,
1725 1726
	struct zone *zone, struct scan_control *sc, int priority)
{
1727 1728
	int file = is_file_lru(lru);

1729 1730 1731
	if (is_active_lru(lru)) {
		if (inactive_list_is_low(zone, sc, file))
		    shrink_active_list(nr_to_scan, zone, sc, priority, file);
1732 1733 1734
		return 0;
	}

R
Rik van Riel 已提交
1735
	return shrink_inactive_list(nr_to_scan, zone, sc, priority, file);
1736 1737 1738 1739 1740 1741 1742 1743
}

/*
 * 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.
 *
1744
 * nr[0] = anon pages to scan; nr[1] = file pages to scan
1745
 */
1746 1747
static void get_scan_count(struct zone *zone, struct scan_control *sc,
					unsigned long *nr, int priority)
1748 1749 1750 1751
{
	unsigned long anon, file, free;
	unsigned long anon_prio, file_prio;
	unsigned long ap, fp;
1752
	struct zone_reclaim_stat *reclaim_stat = get_reclaim_stat(zone, sc);
1753 1754 1755
	u64 fraction[2], denominator;
	enum lru_list l;
	int noswap = 0;
1756 1757 1758 1759 1760 1761 1762 1763 1764 1765 1766 1767 1768 1769 1770 1771
	int force_scan = 0;


	anon  = zone_nr_lru_pages(zone, sc, LRU_ACTIVE_ANON) +
		zone_nr_lru_pages(zone, sc, LRU_INACTIVE_ANON);
	file  = zone_nr_lru_pages(zone, sc, LRU_ACTIVE_FILE) +
		zone_nr_lru_pages(zone, sc, LRU_INACTIVE_FILE);

	if (((anon + file) >> priority) < SWAP_CLUSTER_MAX) {
		/* kswapd does zone balancing and need to scan this zone */
		if (scanning_global_lru(sc) && current_is_kswapd())
			force_scan = 1;
		/* memcg may have small limit and need to avoid priority drop */
		if (!scanning_global_lru(sc))
			force_scan = 1;
	}
1772 1773 1774 1775 1776 1777 1778 1779 1780

	/* 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;
	}
1781

1782
	if (scanning_global_lru(sc)) {
1783 1784 1785
		free  = zone_page_state(zone, NR_FREE_PAGES);
		/* If we have very few page cache pages,
		   force-scan anon pages. */
1786
		if (unlikely(file + free <= high_wmark_pages(zone))) {
1787 1788 1789 1790
			fraction[0] = 1;
			fraction[1] = 0;
			denominator = 1;
			goto out;
1791
		}
1792 1793
	}

1794 1795 1796 1797 1798 1799 1800
	/*
	 * With swappiness at 100, anonymous and file have the same priority.
	 * This scanning priority is essentially the inverse of IO cost.
	 */
	anon_prio = sc->swappiness;
	file_prio = 200 - sc->swappiness;

1801 1802 1803 1804 1805 1806 1807 1808 1809 1810 1811
	/*
	 * 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]
	 */
1812
	spin_lock_irq(&zone->lru_lock);
1813 1814 1815
	if (unlikely(reclaim_stat->recent_scanned[0] > anon / 4)) {
		reclaim_stat->recent_scanned[0] /= 2;
		reclaim_stat->recent_rotated[0] /= 2;
1816 1817
	}

1818 1819 1820
	if (unlikely(reclaim_stat->recent_scanned[1] > file / 4)) {
		reclaim_stat->recent_scanned[1] /= 2;
		reclaim_stat->recent_rotated[1] /= 2;
1821 1822 1823
	}

	/*
1824 1825 1826
	 * 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.
1827
	 */
1828 1829
	ap = (anon_prio + 1) * (reclaim_stat->recent_scanned[0] + 1);
	ap /= reclaim_stat->recent_rotated[0] + 1;
1830

1831 1832
	fp = (file_prio + 1) * (reclaim_stat->recent_scanned[1] + 1);
	fp /= reclaim_stat->recent_rotated[1] + 1;
1833
	spin_unlock_irq(&zone->lru_lock);
1834

1835 1836 1837 1838 1839 1840 1841
	fraction[0] = ap;
	fraction[1] = fp;
	denominator = ap + fp + 1;
out:
	for_each_evictable_lru(l) {
		int file = is_file_lru(l);
		unsigned long scan;
1842

1843 1844 1845 1846 1847
		scan = zone_nr_lru_pages(zone, sc, l);
		if (priority || noswap) {
			scan >>= priority;
			scan = div64_u64(scan * fraction[file], denominator);
		}
1848 1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 1864

		/*
		 * If zone is small or memcg is small, nr[l] can be 0.
		 * This results no-scan on this priority and priority drop down.
		 * For global direct reclaim, it can visit next zone and tend
		 * not to have problems. For global kswapd, it's for zone
		 * balancing and it need to scan a small amounts. When using
		 * memcg, priority drop can cause big latency. So, it's better
		 * to scan small amount. See may_noscan above.
		 */
		if (!scan && force_scan) {
			if (file)
				scan = SWAP_CLUSTER_MAX;
			else if (!noswap)
				scan = SWAP_CLUSTER_MAX;
		}
		nr[l] = scan;
1865
	}
1866
}
1867

1868 1869 1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880 1881 1882 1883
/*
 * 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
 */
static inline bool should_continue_reclaim(struct zone *zone,
					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 */
1884
	if (!(sc->reclaim_mode & RECLAIM_MODE_COMPACTION))
1885 1886
		return false;

1887 1888 1889 1890 1891 1892 1893 1894 1895 1896 1897 1898 1899 1900 1901 1902 1903 1904 1905 1906 1907 1908
	/* 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;
	}
1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930

	/*
	 * If we have not reclaimed enough pages for compaction and the
	 * inactive lists are large enough, continue reclaiming
	 */
	pages_for_compaction = (2UL << sc->order);
	inactive_lru_pages = zone_nr_lru_pages(zone, sc, LRU_INACTIVE_ANON) +
				zone_nr_lru_pages(zone, sc, LRU_INACTIVE_FILE);
	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 */
	switch (compaction_suitable(zone, sc->order)) {
	case COMPACT_PARTIAL:
	case COMPACT_CONTINUE:
		return false;
	default:
		return true;
	}
}

L
Linus Torvalds 已提交
1931 1932 1933
/*
 * This is a basic per-zone page freer.  Used by both kswapd and direct reclaim.
 */
1934
static void shrink_zone(int priority, struct zone *zone,
1935
				struct scan_control *sc)
L
Linus Torvalds 已提交
1936
{
1937
	unsigned long nr[NR_LRU_LISTS];
1938
	unsigned long nr_to_scan;
1939
	enum lru_list l;
1940
	unsigned long nr_reclaimed, nr_scanned;
1941
	unsigned long nr_to_reclaim = sc->nr_to_reclaim;
1942

1943 1944
restart:
	nr_reclaimed = 0;
1945
	nr_scanned = sc->nr_scanned;
1946
	get_scan_count(zone, sc, nr, priority);
L
Linus Torvalds 已提交
1947

1948 1949
	while (nr[LRU_INACTIVE_ANON] || nr[LRU_ACTIVE_FILE] ||
					nr[LRU_INACTIVE_FILE]) {
L
Lee Schermerhorn 已提交
1950
		for_each_evictable_lru(l) {
1951
			if (nr[l]) {
K
KOSAKI Motohiro 已提交
1952 1953
				nr_to_scan = min_t(unsigned long,
						   nr[l], SWAP_CLUSTER_MAX);
1954
				nr[l] -= nr_to_scan;
L
Linus Torvalds 已提交
1955

1956 1957
				nr_reclaimed += shrink_list(l, nr_to_scan,
							    zone, sc, priority);
1958
			}
L
Linus Torvalds 已提交
1959
		}
1960 1961 1962 1963 1964 1965 1966 1967
		/*
		 * 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.
		 */
1968
		if (nr_reclaimed >= nr_to_reclaim && priority < DEF_PRIORITY)
1969
			break;
L
Linus Torvalds 已提交
1970
	}
1971
	sc->nr_reclaimed += nr_reclaimed;
1972

1973 1974 1975 1976
	/*
	 * Even if we did not try to evict anon pages at all, we want to
	 * rebalance the anon lru active/inactive ratio.
	 */
1977
	if (inactive_anon_is_low(zone, sc))
1978 1979
		shrink_active_list(SWAP_CLUSTER_MAX, zone, sc, priority, 0);

1980 1981 1982 1983 1984
	/* reclaim/compaction might need reclaim to continue */
	if (should_continue_reclaim(zone, nr_reclaimed,
					sc->nr_scanned - nr_scanned, sc))
		goto restart;

1985
	throttle_vm_writeout(sc->gfp_mask);
L
Linus Torvalds 已提交
1986 1987 1988 1989 1990 1991 1992
}

/*
 * 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.
 *
1993 1994
 * We reclaim from a zone even if that zone is over high_wmark_pages(zone).
 * Because:
L
Linus Torvalds 已提交
1995 1996
 * a) The caller may be trying to free *extra* pages to satisfy a higher-order
 *    allocation or
1997 1998 1999
 * 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
2000 2001 2002 2003
 *
 * If a zone is deemed to be full of pinned pages then just give it a light
 * scan then give up on it.
 */
2004
static void shrink_zones(int priority, struct zonelist *zonelist,
2005
					struct scan_control *sc)
L
Linus Torvalds 已提交
2006
{
2007
	struct zoneref *z;
2008
	struct zone *zone;
2009 2010
	unsigned long nr_soft_reclaimed;
	unsigned long nr_soft_scanned;
2011

2012 2013
	for_each_zone_zonelist_nodemask(zone, z, zonelist,
					gfp_zone(sc->gfp_mask), sc->nodemask) {
2014
		if (!populated_zone(zone))
L
Linus Torvalds 已提交
2015
			continue;
2016 2017 2018 2019
		/*
		 * Take care memory controller reclaiming has small influence
		 * to global LRU.
		 */
2020
		if (scanning_global_lru(sc)) {
2021 2022
			if (!cpuset_zone_allowed_hardwall(zone, GFP_KERNEL))
				continue;
2023
			if (zone->all_unreclaimable && priority != DEF_PRIORITY)
2024
				continue;	/* Let kswapd poll it */
2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037
			/*
			 * 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() */
2038
		}
2039

2040
		shrink_zone(priority, zone, sc);
L
Linus Torvalds 已提交
2041
	}
2042 2043 2044 2045 2046 2047 2048
}

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

2049
/* All zones in zonelist are unreclaimable? */
2050 2051 2052 2053 2054 2055 2056 2057 2058 2059 2060 2061
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;
2062 2063
		if (!zone->all_unreclaimable)
			return false;
2064 2065
	}

2066
	return true;
L
Linus Torvalds 已提交
2067
}
2068

L
Linus Torvalds 已提交
2069 2070 2071 2072 2073 2074 2075 2076
/*
 * 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
2077 2078 2079 2080
 * 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.
2081 2082 2083
 *
 * returns:	0, if no pages reclaimed
 * 		else, the number of pages reclaimed
L
Linus Torvalds 已提交
2084
 */
2085
static unsigned long do_try_to_free_pages(struct zonelist *zonelist,
2086 2087
					struct scan_control *sc,
					struct shrink_control *shrink)
L
Linus Torvalds 已提交
2088 2089
{
	int priority;
2090
	unsigned long total_scanned = 0;
L
Linus Torvalds 已提交
2091
	struct reclaim_state *reclaim_state = current->reclaim_state;
2092
	struct zoneref *z;
2093
	struct zone *zone;
2094
	unsigned long writeback_threshold;
L
Linus Torvalds 已提交
2095

2096
	get_mems_allowed();
2097 2098
	delayacct_freepages_start();

2099
	if (scanning_global_lru(sc))
2100
		count_vm_event(ALLOCSTALL);
L
Linus Torvalds 已提交
2101 2102

	for (priority = DEF_PRIORITY; priority >= 0; priority--) {
2103
		sc->nr_scanned = 0;
2104
		if (!priority)
2105
			disable_swap_token(sc->mem_cgroup);
2106
		shrink_zones(priority, zonelist, sc);
2107 2108 2109 2110
		/*
		 * Don't shrink slabs when reclaiming memory from
		 * over limit cgroups
		 */
2111
		if (scanning_global_lru(sc)) {
2112
			unsigned long lru_pages = 0;
2113 2114
			for_each_zone_zonelist(zone, z, zonelist,
					gfp_zone(sc->gfp_mask)) {
2115 2116 2117 2118 2119 2120
				if (!cpuset_zone_allowed_hardwall(zone, GFP_KERNEL))
					continue;

				lru_pages += zone_reclaimable_pages(zone);
			}

2121
			shrink_slab(shrink, sc->nr_scanned, lru_pages);
2122
			if (reclaim_state) {
2123
				sc->nr_reclaimed += reclaim_state->reclaimed_slab;
2124 2125
				reclaim_state->reclaimed_slab = 0;
			}
L
Linus Torvalds 已提交
2126
		}
2127
		total_scanned += sc->nr_scanned;
2128
		if (sc->nr_reclaimed >= sc->nr_to_reclaim)
L
Linus Torvalds 已提交
2129 2130 2131 2132 2133 2134 2135 2136 2137
			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.
		 */
2138 2139
		writeback_threshold = sc->nr_to_reclaim + sc->nr_to_reclaim / 2;
		if (total_scanned > writeback_threshold) {
2140
			wakeup_flusher_threads(laptop_mode ? 0 : total_scanned);
2141
			sc->may_writepage = 1;
L
Linus Torvalds 已提交
2142 2143 2144
		}

		/* Take a nap, wait for some writeback to complete */
2145
		if (!sc->hibernation_mode && sc->nr_scanned &&
2146 2147 2148 2149
		    priority < DEF_PRIORITY - 2) {
			struct zone *preferred_zone;

			first_zones_zonelist(zonelist, gfp_zone(sc->gfp_mask),
2150 2151
						&cpuset_current_mems_allowed,
						&preferred_zone);
2152 2153
			wait_iff_congested(preferred_zone, BLK_RW_ASYNC, HZ/10);
		}
L
Linus Torvalds 已提交
2154
	}
2155

L
Linus Torvalds 已提交
2156
out:
2157
	delayacct_freepages_end();
2158
	put_mems_allowed();
2159

2160 2161 2162
	if (sc->nr_reclaimed)
		return sc->nr_reclaimed;

2163 2164 2165 2166 2167 2168 2169 2170
	/*
	 * 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;

2171
	/* top priority shrink_zones still had more to do? don't OOM, then */
2172
	if (scanning_global_lru(sc) && !all_unreclaimable(zonelist, sc))
2173 2174 2175
		return 1;

	return 0;
L
Linus Torvalds 已提交
2176 2177
}

2178
unsigned long try_to_free_pages(struct zonelist *zonelist, int order,
2179
				gfp_t gfp_mask, nodemask_t *nodemask)
2180
{
2181
	unsigned long nr_reclaimed;
2182 2183 2184
	struct scan_control sc = {
		.gfp_mask = gfp_mask,
		.may_writepage = !laptop_mode,
2185
		.nr_to_reclaim = SWAP_CLUSTER_MAX,
2186
		.may_unmap = 1,
2187
		.may_swap = 1,
2188 2189 2190
		.swappiness = vm_swappiness,
		.order = order,
		.mem_cgroup = NULL,
2191
		.nodemask = nodemask,
2192
	};
2193 2194 2195
	struct shrink_control shrink = {
		.gfp_mask = sc.gfp_mask,
	};
2196

2197 2198 2199 2200
	trace_mm_vmscan_direct_reclaim_begin(order,
				sc.may_writepage,
				gfp_mask);

2201
	nr_reclaimed = do_try_to_free_pages(zonelist, &sc, &shrink);
2202 2203 2204 2205

	trace_mm_vmscan_direct_reclaim_end(nr_reclaimed);

	return nr_reclaimed;
2206 2207
}

2208
#ifdef CONFIG_CGROUP_MEM_RES_CTLR
2209

2210 2211 2212
unsigned long mem_cgroup_shrink_node_zone(struct mem_cgroup *mem,
						gfp_t gfp_mask, bool noswap,
						unsigned int swappiness,
2213 2214
						struct zone *zone,
						unsigned long *nr_scanned)
2215 2216
{
	struct scan_control sc = {
2217
		.nr_scanned = 0,
2218
		.nr_to_reclaim = SWAP_CLUSTER_MAX,
2219 2220 2221 2222 2223 2224 2225
		.may_writepage = !laptop_mode,
		.may_unmap = 1,
		.may_swap = !noswap,
		.swappiness = swappiness,
		.order = 0,
		.mem_cgroup = mem,
	};
2226

2227 2228
	sc.gfp_mask = (gfp_mask & GFP_RECLAIM_MASK) |
			(GFP_HIGHUSER_MOVABLE & ~GFP_RECLAIM_MASK);
2229 2230 2231 2232 2233

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

2234 2235 2236 2237 2238 2239 2240 2241
	/*
	 * 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.
	 */
	shrink_zone(0, zone, &sc);
2242 2243 2244

	trace_mm_vmscan_memcg_softlimit_reclaim_end(sc.nr_reclaimed);

2245
	*nr_scanned = sc.nr_scanned;
2246 2247 2248
	return sc.nr_reclaimed;
}

2249
unsigned long try_to_free_mem_cgroup_pages(struct mem_cgroup *mem_cont,
K
KOSAKI Motohiro 已提交
2250 2251 2252
					   gfp_t gfp_mask,
					   bool noswap,
					   unsigned int swappiness)
2253
{
2254
	struct zonelist *zonelist;
2255
	unsigned long nr_reclaimed;
2256
	int nid;
2257 2258
	struct scan_control sc = {
		.may_writepage = !laptop_mode,
2259
		.may_unmap = 1,
2260
		.may_swap = !noswap,
2261
		.nr_to_reclaim = SWAP_CLUSTER_MAX,
K
KOSAKI Motohiro 已提交
2262
		.swappiness = swappiness,
2263 2264
		.order = 0,
		.mem_cgroup = mem_cont,
2265
		.nodemask = NULL, /* we don't care the placement */
2266 2267 2268 2269 2270
		.gfp_mask = (gfp_mask & GFP_RECLAIM_MASK) |
				(GFP_HIGHUSER_MOVABLE & ~GFP_RECLAIM_MASK),
	};
	struct shrink_control shrink = {
		.gfp_mask = sc.gfp_mask,
2271 2272
	};

2273 2274 2275 2276 2277 2278 2279 2280
	/*
	 * 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.
	 */
	nid = mem_cgroup_select_victim_node(mem_cont);

	zonelist = NODE_DATA(nid)->node_zonelists;
2281 2282 2283 2284 2285

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

2286
	nr_reclaimed = do_try_to_free_pages(zonelist, &sc, &shrink);
2287 2288 2289 2290

	trace_mm_vmscan_memcg_reclaim_end(nr_reclaimed);

	return nr_reclaimed;
2291 2292 2293
}
#endif

2294 2295 2296 2297 2298 2299 2300 2301 2302 2303 2304
/*
 * 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
2305
 *     percentage of the middle zones. For example, on 32-bit x86, highmem
2306 2307 2308 2309 2310 2311 2312 2313 2314 2315 2316 2317 2318
 *     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
2319 2320
	/* A special case here: if zone has no page, we think it's balanced */
	return balanced_pages >= (present_pages >> 2);
2321 2322
}

2323
/* is kswapd sleeping prematurely? */
2324 2325
static bool sleeping_prematurely(pg_data_t *pgdat, int order, long remaining,
					int classzone_idx)
2326
{
2327
	int i;
2328 2329
	unsigned long balanced = 0;
	bool all_zones_ok = true;
2330 2331 2332

	/* If a direct reclaimer woke kswapd within HZ/10, it's premature */
	if (remaining)
2333
		return true;
2334

2335
	/* Check the watermark levels */
2336
	for (i = 0; i <= classzone_idx; i++) {
2337 2338 2339 2340 2341
		struct zone *zone = pgdat->node_zones + i;

		if (!populated_zone(zone))
			continue;

2342 2343 2344 2345 2346 2347 2348 2349
		/*
		 * 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;
2350
			continue;
2351
		}
2352

2353
		if (!zone_watermark_ok_safe(zone, order, high_wmark_pages(zone),
2354
							i, 0))
2355 2356 2357
			all_zones_ok = false;
		else
			balanced += zone->present_pages;
2358
	}
2359

2360 2361 2362 2363 2364 2365
	/*
	 * 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)
2366
		return !pgdat_balanced(pgdat, balanced, classzone_idx);
2367 2368
	else
		return !all_zones_ok;
2369 2370
}

L
Linus Torvalds 已提交
2371 2372
/*
 * For kswapd, balance_pgdat() will work across all this node's zones until
2373
 * they are all at high_wmark_pages(zone).
L
Linus Torvalds 已提交
2374
 *
2375
 * Returns the final order kswapd was reclaiming at
L
Linus Torvalds 已提交
2376 2377 2378 2379 2380 2381 2382 2383 2384 2385
 *
 * 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
2386 2387 2388 2389 2390
 * 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
2391
 */
2392
static unsigned long balance_pgdat(pg_data_t *pgdat, int order,
2393
							int *classzone_idx)
L
Linus Torvalds 已提交
2394 2395
{
	int all_zones_ok;
2396
	unsigned long balanced;
L
Linus Torvalds 已提交
2397 2398
	int priority;
	int i;
2399
	int end_zone = 0;	/* Inclusive.  0 = ZONE_DMA */
2400
	unsigned long total_scanned;
L
Linus Torvalds 已提交
2401
	struct reclaim_state *reclaim_state = current->reclaim_state;
2402 2403
	unsigned long nr_soft_reclaimed;
	unsigned long nr_soft_scanned;
2404 2405
	struct scan_control sc = {
		.gfp_mask = GFP_KERNEL,
2406
		.may_unmap = 1,
2407
		.may_swap = 1,
2408 2409 2410 2411 2412
		/*
		 * 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,
2413
		.swappiness = vm_swappiness,
A
Andy Whitcroft 已提交
2414
		.order = order,
2415
		.mem_cgroup = NULL,
2416
	};
2417 2418 2419
	struct shrink_control shrink = {
		.gfp_mask = sc.gfp_mask,
	};
L
Linus Torvalds 已提交
2420 2421
loop_again:
	total_scanned = 0;
2422
	sc.nr_reclaimed = 0;
C
Christoph Lameter 已提交
2423
	sc.may_writepage = !laptop_mode;
2424
	count_vm_event(PAGEOUTRUN);
L
Linus Torvalds 已提交
2425 2426 2427

	for (priority = DEF_PRIORITY; priority >= 0; priority--) {
		unsigned long lru_pages = 0;
2428
		int has_under_min_watermark_zone = 0;
L
Linus Torvalds 已提交
2429

2430 2431
		/* The swap token gets in the way of swapout... */
		if (!priority)
2432
			disable_swap_token(NULL);
2433

L
Linus Torvalds 已提交
2434
		all_zones_ok = 1;
2435
		balanced = 0;
L
Linus Torvalds 已提交
2436

2437 2438 2439 2440 2441 2442
		/*
		 * 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
2443

2444 2445
			if (!populated_zone(zone))
				continue;
L
Linus Torvalds 已提交
2446

2447
			if (zone->all_unreclaimable && priority != DEF_PRIORITY)
2448
				continue;
L
Linus Torvalds 已提交
2449

2450 2451 2452 2453
			/*
			 * Do some background aging of the anon list, to give
			 * pages a chance to be referenced before reclaiming.
			 */
2454
			if (inactive_anon_is_low(zone, &sc))
2455 2456 2457
				shrink_active_list(SWAP_CLUSTER_MAX, zone,
							&sc, priority, 0);

2458
			if (!zone_watermark_ok_safe(zone, order,
2459
					high_wmark_pages(zone), 0, 0)) {
2460
				end_zone = i;
A
Andrew Morton 已提交
2461
				break;
L
Linus Torvalds 已提交
2462 2463
			}
		}
A
Andrew Morton 已提交
2464 2465 2466
		if (i < 0)
			goto out;

L
Linus Torvalds 已提交
2467 2468 2469
		for (i = 0; i <= end_zone; i++) {
			struct zone *zone = pgdat->node_zones + i;

2470
			lru_pages += zone_reclaimable_pages(zone);
L
Linus Torvalds 已提交
2471 2472 2473 2474 2475 2476 2477 2478 2479 2480 2481 2482 2483
		}

		/*
		 * 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;
2484
			int nr_slab;
2485
			unsigned long balance_gap;
L
Linus Torvalds 已提交
2486

2487
			if (!populated_zone(zone))
L
Linus Torvalds 已提交
2488 2489
				continue;

2490
			if (zone->all_unreclaimable && priority != DEF_PRIORITY)
L
Linus Torvalds 已提交
2491 2492 2493
				continue;

			sc.nr_scanned = 0;
2494

2495
			nr_soft_scanned = 0;
2496 2497 2498
			/*
			 * Call soft limit reclaim before calling shrink_zone.
			 */
2499 2500 2501 2502 2503
			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;
2504

2505
			/*
2506 2507 2508 2509 2510 2511
			 * 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.
2512
			 */
2513 2514 2515 2516
			balance_gap = min(low_wmark_pages(zone),
				(zone->present_pages +
					KSWAPD_ZONE_BALANCE_GAP_RATIO-1) /
				KSWAPD_ZONE_BALANCE_GAP_RATIO);
2517
			if (!zone_watermark_ok_safe(zone, order,
2518
					high_wmark_pages(zone) + balance_gap,
2519
					end_zone, 0)) {
2520
				shrink_zone(priority, zone, &sc);
2521

2522 2523 2524 2525 2526 2527 2528 2529 2530
				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
2531 2532 2533 2534 2535 2536
			/*
			 * 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 &&
2537
			    total_scanned > sc.nr_reclaimed + sc.nr_reclaimed / 2)
L
Linus Torvalds 已提交
2538
				sc.may_writepage = 1;
2539

2540 2541 2542
			if (zone->all_unreclaimable) {
				if (end_zone && end_zone == i)
					end_zone--;
2543
				continue;
2544
			}
2545

2546
			if (!zone_watermark_ok_safe(zone, order,
2547 2548 2549 2550 2551 2552 2553
					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!
				 */
2554
				if (!zone_watermark_ok_safe(zone, order,
2555 2556
					    min_wmark_pages(zone), end_zone, 0))
					has_under_min_watermark_zone = 1;
2557 2558 2559 2560 2561 2562 2563 2564 2565
			} 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);
2566
				if (i <= *classzone_idx)
2567
					balanced += zone->present_pages;
2568
			}
2569

L
Linus Torvalds 已提交
2570
		}
2571
		if (all_zones_ok || (order && pgdat_balanced(pgdat, balanced, *classzone_idx)))
L
Linus Torvalds 已提交
2572 2573 2574 2575 2576
			break;		/* kswapd: all done */
		/*
		 * OK, kswapd is getting into trouble.  Take a nap, then take
		 * another pass across the zones.
		 */
2577 2578 2579 2580 2581 2582
		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
2583 2584 2585 2586 2587 2588 2589

		/*
		 * 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.
		 */
2590
		if (sc.nr_reclaimed >= SWAP_CLUSTER_MAX)
L
Linus Torvalds 已提交
2591 2592 2593
			break;
	}
out:
2594 2595 2596

	/*
	 * order-0: All zones must meet high watermark for a balanced node
2597 2598
	 * high-order: Balanced zones must make up at least 25% of the node
	 *             for the node to be balanced
2599
	 */
2600
	if (!(all_zones_ok || (order && pgdat_balanced(pgdat, balanced, *classzone_idx)))) {
L
Linus Torvalds 已提交
2601
		cond_resched();
2602 2603 2604

		try_to_freeze();

2605 2606 2607 2608 2609 2610 2611 2612 2613 2614 2615 2616 2617 2618 2619 2620 2621
		/*
		 * 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
2622 2623 2624
		goto loop_again;
	}

2625 2626 2627 2628 2629 2630 2631 2632 2633 2634 2635 2636 2637 2638 2639 2640 2641 2642 2643 2644 2645 2646 2647 2648 2649 2650 2651 2652 2653 2654
	/*
	 * 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);
		}
	}

2655 2656 2657 2658 2659 2660
	/*
	 * 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
	 */
2661
	*classzone_idx = end_zone;
2662
	return order;
L
Linus Torvalds 已提交
2663 2664
}

2665
static void kswapd_try_to_sleep(pg_data_t *pgdat, int order, int classzone_idx)
2666 2667 2668 2669 2670 2671 2672 2673 2674 2675
{
	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 */
2676
	if (!sleeping_prematurely(pgdat, order, remaining, classzone_idx)) {
2677 2678 2679 2680 2681 2682 2683 2684 2685
		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.
	 */
2686
	if (!sleeping_prematurely(pgdat, order, remaining, classzone_idx)) {
2687 2688 2689 2690 2691 2692 2693 2694 2695 2696 2697 2698 2699 2700 2701 2702 2703 2704 2705 2706 2707 2708
		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
2709 2710
/*
 * The background pageout daemon, started as a kernel thread
2711
 * from the init process.
L
Linus Torvalds 已提交
2712 2713 2714 2715 2716 2717 2718 2719 2720 2721 2722 2723
 *
 * 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)
{
2724 2725
	unsigned long order, new_order;
	int classzone_idx, new_classzone_idx;
L
Linus Torvalds 已提交
2726 2727
	pg_data_t *pgdat = (pg_data_t*)p;
	struct task_struct *tsk = current;
2728

L
Linus Torvalds 已提交
2729 2730 2731
	struct reclaim_state reclaim_state = {
		.reclaimed_slab = 0,
	};
2732
	const struct cpumask *cpumask = cpumask_of_node(pgdat->node_id);
L
Linus Torvalds 已提交
2733

2734 2735
	lockdep_set_current_reclaim_state(GFP_KERNEL);

R
Rusty Russell 已提交
2736
	if (!cpumask_empty(cpumask))
2737
		set_cpus_allowed_ptr(tsk, cpumask);
L
Linus Torvalds 已提交
2738 2739 2740 2741 2742 2743 2744 2745 2746 2747 2748 2749 2750 2751
	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).
	 */
2752
	tsk->flags |= PF_MEMALLOC | PF_SWAPWRITE | PF_KSWAPD;
2753
	set_freezable();
L
Linus Torvalds 已提交
2754

2755 2756
	order = new_order = 0;
	classzone_idx = new_classzone_idx = pgdat->nr_zones - 1;
L
Linus Torvalds 已提交
2757
	for ( ; ; ) {
2758
		int ret;
2759

2760 2761 2762 2763 2764 2765 2766 2767 2768 2769 2770 2771
		/*
		 * 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
		 */
		if (classzone_idx >= new_classzone_idx && order == new_order) {
			new_order = pgdat->kswapd_max_order;
			new_classzone_idx = pgdat->classzone_idx;
			pgdat->kswapd_max_order =  0;
			pgdat->classzone_idx = pgdat->nr_zones - 1;
		}

2772
		if (order < new_order || classzone_idx > new_classzone_idx) {
L
Linus Torvalds 已提交
2773 2774
			/*
			 * Don't sleep if someone wants a larger 'order'
2775
			 * allocation or has tigher zone constraints
L
Linus Torvalds 已提交
2776 2777
			 */
			order = new_order;
2778
			classzone_idx = new_classzone_idx;
L
Linus Torvalds 已提交
2779
		} else {
2780
			kswapd_try_to_sleep(pgdat, order, classzone_idx);
L
Linus Torvalds 已提交
2781
			order = pgdat->kswapd_max_order;
2782
			classzone_idx = pgdat->classzone_idx;
2783
			pgdat->kswapd_max_order = 0;
2784
			pgdat->classzone_idx = pgdat->nr_zones - 1;
L
Linus Torvalds 已提交
2785 2786
		}

2787 2788 2789 2790 2791 2792 2793 2794
		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
		 */
2795 2796
		if (!ret) {
			trace_mm_vmscan_kswapd_wake(pgdat->node_id, order);
2797
			order = balance_pgdat(pgdat, order, &classzone_idx);
2798
		}
L
Linus Torvalds 已提交
2799 2800 2801 2802 2803 2804 2805
	}
	return 0;
}

/*
 * A zone is low on free memory, so wake its kswapd task to service it.
 */
2806
void wakeup_kswapd(struct zone *zone, int order, enum zone_type classzone_idx)
L
Linus Torvalds 已提交
2807 2808 2809
{
	pg_data_t *pgdat;

2810
	if (!populated_zone(zone))
L
Linus Torvalds 已提交
2811 2812
		return;

2813
	if (!cpuset_zone_allowed_hardwall(zone, GFP_KERNEL))
L
Linus Torvalds 已提交
2814
		return;
2815
	pgdat = zone->zone_pgdat;
2816
	if (pgdat->kswapd_max_order < order) {
L
Linus Torvalds 已提交
2817
		pgdat->kswapd_max_order = order;
2818 2819
		pgdat->classzone_idx = min(pgdat->classzone_idx, classzone_idx);
	}
2820
	if (!waitqueue_active(&pgdat->kswapd_wait))
L
Linus Torvalds 已提交
2821
		return;
2822 2823 2824 2825
	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);
2826
	wake_up_interruptible(&pgdat->kswapd_wait);
L
Linus Torvalds 已提交
2827 2828
}

2829 2830 2831 2832 2833 2834 2835 2836
/*
 * 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)
2837
{
2838 2839 2840 2841 2842 2843 2844 2845 2846 2847 2848 2849 2850 2851 2852 2853 2854 2855 2856 2857 2858 2859 2860 2861
	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;
2862 2863
}

2864
#ifdef CONFIG_HIBERNATION
L
Linus Torvalds 已提交
2865
/*
2866
 * Try to free `nr_to_reclaim' of memory, system-wide, and return the number of
2867 2868 2869 2870 2871
 * 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
2872
 */
2873
unsigned long shrink_all_memory(unsigned long nr_to_reclaim)
L
Linus Torvalds 已提交
2874
{
2875 2876
	struct reclaim_state reclaim_state;
	struct scan_control sc = {
2877 2878 2879
		.gfp_mask = GFP_HIGHUSER_MOVABLE,
		.may_swap = 1,
		.may_unmap = 1,
2880
		.may_writepage = 1,
2881 2882 2883 2884
		.nr_to_reclaim = nr_to_reclaim,
		.hibernation_mode = 1,
		.swappiness = vm_swappiness,
		.order = 0,
L
Linus Torvalds 已提交
2885
	};
2886 2887 2888 2889
	struct shrink_control shrink = {
		.gfp_mask = sc.gfp_mask,
	};
	struct zonelist *zonelist = node_zonelist(numa_node_id(), sc.gfp_mask);
2890 2891
	struct task_struct *p = current;
	unsigned long nr_reclaimed;
L
Linus Torvalds 已提交
2892

2893 2894 2895 2896
	p->flags |= PF_MEMALLOC;
	lockdep_set_current_reclaim_state(sc.gfp_mask);
	reclaim_state.reclaimed_slab = 0;
	p->reclaim_state = &reclaim_state;
2897

2898
	nr_reclaimed = do_try_to_free_pages(zonelist, &sc, &shrink);
2899

2900 2901 2902
	p->reclaim_state = NULL;
	lockdep_clear_current_reclaim_state();
	p->flags &= ~PF_MEMALLOC;
2903

2904
	return nr_reclaimed;
L
Linus Torvalds 已提交
2905
}
2906
#endif /* CONFIG_HIBERNATION */
L
Linus Torvalds 已提交
2907 2908 2909 2910 2911

/* 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. */
2912
static int __devinit cpu_callback(struct notifier_block *nfb,
2913
				  unsigned long action, void *hcpu)
L
Linus Torvalds 已提交
2914
{
2915
	int nid;
L
Linus Torvalds 已提交
2916

2917
	if (action == CPU_ONLINE || action == CPU_ONLINE_FROZEN) {
2918
		for_each_node_state(nid, N_HIGH_MEMORY) {
2919
			pg_data_t *pgdat = NODE_DATA(nid);
2920 2921 2922
			const struct cpumask *mask;

			mask = cpumask_of_node(pgdat->node_id);
2923

2924
			if (cpumask_any_and(cpu_online_mask, mask) < nr_cpu_ids)
L
Linus Torvalds 已提交
2925
				/* One of our CPUs online: restore mask */
2926
				set_cpus_allowed_ptr(pgdat->kswapd, mask);
L
Linus Torvalds 已提交
2927 2928 2929 2930 2931
		}
	}
	return NOTIFY_OK;
}

2932 2933 2934 2935 2936 2937 2938 2939 2940 2941 2942 2943 2944 2945 2946 2947 2948 2949 2950 2951 2952 2953
/*
 * 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;
}

2954 2955 2956 2957 2958 2959 2960 2961 2962 2963 2964
/*
 * 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
2965 2966
static int __init kswapd_init(void)
{
2967
	int nid;
2968

L
Linus Torvalds 已提交
2969
	swap_setup();
2970
	for_each_node_state(nid, N_HIGH_MEMORY)
2971
 		kswapd_run(nid);
L
Linus Torvalds 已提交
2972 2973 2974 2975 2976
	hotcpu_notifier(cpu_callback, 0);
	return 0;
}

module_init(kswapd_init)
2977 2978 2979 2980 2981 2982 2983 2984 2985 2986

#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;

2987
#define RECLAIM_OFF 0
2988
#define RECLAIM_ZONE (1<<0)	/* Run shrink_inactive_list on the zone */
2989 2990 2991
#define RECLAIM_WRITE (1<<1)	/* Writeout pages during reclaim */
#define RECLAIM_SWAP (1<<2)	/* Swap pages out during reclaim */

2992 2993 2994 2995 2996 2997 2998
/*
 * 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

2999 3000 3001 3002 3003 3004
/*
 * Percentage of pages in a zone that must be unmapped for zone_reclaim to
 * occur.
 */
int sysctl_min_unmapped_ratio = 1;

3005 3006 3007 3008 3009 3010
/*
 * 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;

3011 3012 3013 3014 3015 3016 3017 3018 3019 3020 3021 3022 3023 3024 3025 3026 3027 3028 3029 3030 3031 3032 3033 3034 3035 3036 3037 3038 3039 3040 3041 3042 3043 3044 3045 3046 3047 3048 3049 3050 3051 3052
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;
}

3053 3054 3055
/*
 * Try to free up some pages from this zone through reclaim.
 */
3056
static int __zone_reclaim(struct zone *zone, gfp_t gfp_mask, unsigned int order)
3057
{
3058
	/* Minimum pages needed in order to stay on node */
3059
	const unsigned long nr_pages = 1 << order;
3060 3061
	struct task_struct *p = current;
	struct reclaim_state reclaim_state;
3062
	int priority;
3063 3064
	struct scan_control sc = {
		.may_writepage = !!(zone_reclaim_mode & RECLAIM_WRITE),
3065
		.may_unmap = !!(zone_reclaim_mode & RECLAIM_SWAP),
3066
		.may_swap = 1,
3067 3068
		.nr_to_reclaim = max_t(unsigned long, nr_pages,
				       SWAP_CLUSTER_MAX),
3069
		.gfp_mask = gfp_mask,
3070
		.swappiness = vm_swappiness,
3071
		.order = order,
3072
	};
3073 3074 3075
	struct shrink_control shrink = {
		.gfp_mask = sc.gfp_mask,
	};
3076
	unsigned long nr_slab_pages0, nr_slab_pages1;
3077 3078

	cond_resched();
3079 3080 3081 3082 3083 3084
	/*
	 * 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;
3085
	lockdep_set_current_reclaim_state(gfp_mask);
3086 3087
	reclaim_state.reclaimed_slab = 0;
	p->reclaim_state = &reclaim_state;
3088

3089
	if (zone_pagecache_reclaimable(zone) > zone->min_unmapped_pages) {
3090 3091 3092 3093 3094 3095
		/*
		 * Free memory by calling shrink zone with increasing
		 * priorities until we have enough memory freed.
		 */
		priority = ZONE_RECLAIM_PRIORITY;
		do {
3096
			shrink_zone(priority, zone, &sc);
3097
			priority--;
3098
		} while (priority >= 0 && sc.nr_reclaimed < nr_pages);
3099
	}
3100

3101 3102
	nr_slab_pages0 = zone_page_state(zone, NR_SLAB_RECLAIMABLE);
	if (nr_slab_pages0 > zone->min_slab_pages) {
3103
		/*
3104
		 * shrink_slab() does not currently allow us to determine how
3105 3106 3107 3108
		 * 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.
3109
		 *
3110 3111
		 * Note that shrink_slab will free memory on all zones and may
		 * take a long time.
3112
		 */
3113 3114 3115 3116
		for (;;) {
			unsigned long lru_pages = zone_reclaimable_pages(zone);

			/* No reclaimable slab or very low memory pressure */
3117
			if (!shrink_slab(&shrink, sc.nr_scanned, lru_pages))
3118 3119 3120 3121 3122 3123 3124 3125
				break;

			/* Freed enough memory */
			nr_slab_pages1 = zone_page_state(zone,
							NR_SLAB_RECLAIMABLE);
			if (nr_slab_pages1 + nr_pages <= nr_slab_pages0)
				break;
		}
3126 3127 3128 3129 3130

		/*
		 * Update nr_reclaimed by the number of slab pages we
		 * reclaimed from this zone.
		 */
3131 3132 3133
		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;
3134 3135
	}

3136
	p->reclaim_state = NULL;
3137
	current->flags &= ~(PF_MEMALLOC | PF_SWAPWRITE);
3138
	lockdep_clear_current_reclaim_state();
3139
	return sc.nr_reclaimed >= nr_pages;
3140
}
3141 3142 3143 3144

int zone_reclaim(struct zone *zone, gfp_t gfp_mask, unsigned int order)
{
	int node_id;
3145
	int ret;
3146 3147

	/*
3148 3149
	 * Zone reclaim reclaims unmapped file backed pages and
	 * slab pages if we are over the defined limits.
3150
	 *
3151 3152 3153 3154 3155
	 * 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.
3156
	 */
3157 3158
	if (zone_pagecache_reclaimable(zone) <= zone->min_unmapped_pages &&
	    zone_page_state(zone, NR_SLAB_RECLAIMABLE) <= zone->min_slab_pages)
3159
		return ZONE_RECLAIM_FULL;
3160

3161
	if (zone->all_unreclaimable)
3162
		return ZONE_RECLAIM_FULL;
3163

3164
	/*
3165
	 * Do not scan if the allocation should not be delayed.
3166
	 */
3167
	if (!(gfp_mask & __GFP_WAIT) || (current->flags & PF_MEMALLOC))
3168
		return ZONE_RECLAIM_NOSCAN;
3169 3170 3171 3172 3173 3174 3175

	/*
	 * 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.
	 */
3176
	node_id = zone_to_nid(zone);
3177
	if (node_state(node_id, N_CPU) && node_id != numa_node_id())
3178
		return ZONE_RECLAIM_NOSCAN;
3179 3180

	if (zone_test_and_set_flag(zone, ZONE_RECLAIM_LOCKED))
3181 3182
		return ZONE_RECLAIM_NOSCAN;

3183 3184 3185
	ret = __zone_reclaim(zone, gfp_mask, order);
	zone_clear_flag(zone, ZONE_RECLAIM_LOCKED);

3186 3187 3188
	if (!ret)
		count_vm_event(PGSCAN_ZONE_RECLAIM_FAILED);

3189
	return ret;
3190
}
3191
#endif
L
Lee Schermerhorn 已提交
3192 3193 3194 3195 3196 3197 3198

/*
 * 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
3199 3200
 * 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
3201 3202
 *
 * Reasons page might not be evictable:
3203
 * (1) page's mapping marked unevictable
N
Nick Piggin 已提交
3204
 * (2) page is part of an mlocked VMA
3205
 *
L
Lee Schermerhorn 已提交
3206 3207 3208 3209
 */
int page_evictable(struct page *page, struct vm_area_struct *vma)
{

3210 3211 3212
	if (mapping_unevictable(page_mapping(page)))
		return 0;

N
Nick Piggin 已提交
3213 3214
	if (PageMlocked(page) || (vma && is_mlocked_vma(vma, page)))
		return 0;
L
Lee Schermerhorn 已提交
3215 3216 3217

	return 1;
}
3218 3219 3220 3221 3222 3223 3224 3225 3226 3227 3228 3229 3230 3231 3232 3233 3234 3235 3236

/**
 * 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.
 */
static void check_move_unevictable_page(struct page *page, struct zone *zone)
{
	VM_BUG_ON(PageActive(page));

retry:
	ClearPageUnevictable(page);
	if (page_evictable(page, NULL)) {
3237
		enum lru_list l = page_lru_base_type(page);
3238

3239 3240
		__dec_zone_state(zone, NR_UNEVICTABLE);
		list_move(&page->lru, &zone->lru[l].list);
K
KAMEZAWA Hiroyuki 已提交
3241
		mem_cgroup_move_lists(page, LRU_UNEVICTABLE, l);
3242 3243 3244 3245 3246 3247 3248 3249
		__inc_zone_state(zone, NR_INACTIVE_ANON + l);
		__count_vm_event(UNEVICTABLE_PGRESCUED);
	} else {
		/*
		 * rotate unevictable list
		 */
		SetPageUnevictable(page);
		list_move(&page->lru, &zone->lru[LRU_UNEVICTABLE].list);
K
KAMEZAWA Hiroyuki 已提交
3250
		mem_cgroup_rotate_lru_list(page, LRU_UNEVICTABLE);
3251 3252 3253 3254 3255 3256 3257 3258 3259 3260 3261 3262 3263 3264 3265 3266 3267 3268 3269 3270 3271 3272 3273 3274 3275 3276 3277 3278 3279 3280 3281 3282 3283 3284 3285 3286 3287 3288 3289 3290 3291 3292 3293 3294 3295 3296 3297 3298 3299 3300 3301 3302 3303 3304 3305 3306 3307 3308 3309
		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.
 */
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);
	}

}
3310 3311 3312 3313 3314 3315 3316 3317 3318 3319 3320 3321

/**
 * scan_zone_unevictable_pages - check unevictable list for evictable pages
 * @zone - zone of which to scan the unevictable list
 *
 * Scan @zone's unevictable LRU lists to check for pages that have become
 * evictable.  Move those that have to @zone's inactive list where they
 * become candidates for reclaim, unless shrink_inactive_zone() decides
 * to reactivate them.  Pages that are still unevictable are rotated
 * back onto @zone's unevictable list.
 */
#define SCAN_UNEVICTABLE_BATCH_SIZE 16UL /* arbitrary lock hold batch size */
3322
static void scan_zone_unevictable_pages(struct zone *zone)
3323 3324 3325 3326 3327 3328 3329 3330 3331 3332 3333 3334 3335 3336 3337 3338 3339 3340 3341 3342 3343 3344 3345 3346 3347 3348 3349 3350 3351 3352 3353 3354 3355 3356 3357 3358 3359 3360 3361 3362 3363
{
	struct list_head *l_unevictable = &zone->lru[LRU_UNEVICTABLE].list;
	unsigned long scan;
	unsigned long nr_to_scan = zone_page_state(zone, NR_UNEVICTABLE);

	while (nr_to_scan > 0) {
		unsigned long batch_size = min(nr_to_scan,
						SCAN_UNEVICTABLE_BATCH_SIZE);

		spin_lock_irq(&zone->lru_lock);
		for (scan = 0;  scan < batch_size; scan++) {
			struct page *page = lru_to_page(l_unevictable);

			if (!trylock_page(page))
				continue;

			prefetchw_prev_lru_page(page, l_unevictable, flags);

			if (likely(PageLRU(page) && PageUnevictable(page)))
				check_move_unevictable_page(page, zone);

			unlock_page(page);
		}
		spin_unlock_irq(&zone->lru_lock);

		nr_to_scan -= batch_size;
	}
}


/**
 * scan_all_zones_unevictable_pages - scan all unevictable lists for evictable pages
 *
 * A really big hammer:  scan all zones' unevictable LRU lists to check for
 * pages that have become evictable.  Move those back to the zones'
 * inactive list where they become candidates for reclaim.
 * This occurs when, e.g., we have unswappable pages on the unevictable lists,
 * and we add swap to the system.  As such, it runs in the context of a task
 * that has possibly/probably made some previously unevictable pages
 * evictable.
 */
3364
static void scan_all_zones_unevictable_pages(void)
3365 3366 3367 3368 3369 3370 3371 3372 3373 3374 3375 3376 3377 3378 3379
{
	struct zone *zone;

	for_each_zone(zone) {
		scan_zone_unevictable_pages(zone);
	}
}

/*
 * 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,
3380
			   void __user *buffer,
3381 3382
			   size_t *length, loff_t *ppos)
{
3383
	proc_doulongvec_minmax(table, write, buffer, length, ppos);
3384 3385 3386 3387 3388 3389 3390 3391

	if (write && *(unsigned long *)table->data)
		scan_all_zones_unevictable_pages();

	scan_unevictable_pages = 0;
	return 0;
}

3392
#ifdef CONFIG_NUMA
3393 3394 3395 3396 3397 3398 3399 3400 3401 3402 3403 3404 3405 3406 3407 3408 3409 3410 3411 3412 3413 3414 3415 3416 3417 3418 3419 3420 3421 3422 3423 3424 3425 3426 3427 3428 3429 3430 3431 3432 3433 3434 3435 3436 3437 3438
/*
 * per node 'scan_unevictable_pages' attribute.  On demand re-scan of
 * a specified node's per zone unevictable lists for evictable pages.
 */

static ssize_t read_scan_unevictable_node(struct sys_device *dev,
					  struct sysdev_attribute *attr,
					  char *buf)
{
	return sprintf(buf, "0\n");	/* always zero; should fit... */
}

static ssize_t write_scan_unevictable_node(struct sys_device *dev,
					   struct sysdev_attribute *attr,
					const char *buf, size_t count)
{
	struct zone *node_zones = NODE_DATA(dev->id)->node_zones;
	struct zone *zone;
	unsigned long res;
	unsigned long req = strict_strtoul(buf, 10, &res);

	if (!req)
		return 1;	/* zero is no-op */

	for (zone = node_zones; zone - node_zones < MAX_NR_ZONES; ++zone) {
		if (!populated_zone(zone))
			continue;
		scan_zone_unevictable_pages(zone);
	}
	return 1;
}


static SYSDEV_ATTR(scan_unevictable_pages, S_IRUGO | S_IWUSR,
			read_scan_unevictable_node,
			write_scan_unevictable_node);

int scan_unevictable_register_node(struct node *node)
{
	return sysdev_create_file(&node->sysdev, &attr_scan_unevictable_pages);
}

void scan_unevictable_unregister_node(struct node *node)
{
	sysdev_remove_file(&node->sysdev, &attr_scan_unevictable_pages);
}
3439
#endif