vmscan.c 95.2 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
#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/backing-dev.h>
#include <linux/rmap.h>
#include <linux/topology.h>
#include <linux/cpu.h>
#include <linux/cpuset.h>
34
#include <linux/compaction.h>
L
Linus Torvalds 已提交
35 36
#include <linux/notifier.h>
#include <linux/rwsem.h>
37
#include <linux/delay.h>
38
#include <linux/kthread.h>
39
#include <linux/freezer.h>
40
#include <linux/memcontrol.h>
41
#include <linux/delayacct.h>
42
#include <linux/sysctl.h>
43
#include <linux/oom.h>
44
#include <linux/prefetch.h>
L
Linus Torvalds 已提交
45 46 47 48 49 50

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

#include <linux/swapops.h>

51 52
#include "internal.h"

53 54 55
#define CREATE_TRACE_POINTS
#include <trace/events/vmscan.h>

L
Linus Torvalds 已提交
56 57 58 59
struct scan_control {
	/* Incremented by the number of inactive pages that were scanned */
	unsigned long nr_scanned;

60 61 62
	/* Number of pages freed so far during a call to shrink_zones() */
	unsigned long nr_reclaimed;

63 64 65
	/* How many pages shrink_list() should reclaim */
	unsigned long nr_to_reclaim;

66 67
	unsigned long hibernation_mode;

L
Linus Torvalds 已提交
68
	/* This context's GFP mask */
A
Al Viro 已提交
69
	gfp_t gfp_mask;
L
Linus Torvalds 已提交
70 71 72

	int may_writepage;

73 74
	/* Can mapped pages be reclaimed? */
	int may_unmap;
75

76 77 78
	/* Can pages be swapped as part of reclaim? */
	int may_swap;

A
Andy Whitcroft 已提交
79
	int order;
80

81 82 83 84 85
	/*
	 * The memory cgroup that hit its limit and as a result is the
	 * primary target of this reclaim invocation.
	 */
	struct mem_cgroup *target_mem_cgroup;
86

87 88 89 90 91
	/*
	 * Nodemask of nodes allowed by the caller. If NULL, all nodes
	 * are scanned.
	 */
	nodemask_t	*nodemask;
L
Linus Torvalds 已提交
92 93
};

94 95 96 97 98
struct mem_cgroup_zone {
	struct mem_cgroup *mem_cgroup;
	struct zone *zone;
};

L
Linus Torvalds 已提交
99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132
#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;
133
long vm_total_pages;	/* The total number of pages which the VM controls */
L
Linus Torvalds 已提交
134 135 136 137

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

138
#ifdef CONFIG_CGROUP_MEM_RES_CTLR
139 140
static bool global_reclaim(struct scan_control *sc)
{
141
	return !sc->target_mem_cgroup;
142
}
143
#else
144 145 146 147
static bool global_reclaim(struct scan_control *sc)
{
	return true;
}
148 149
#endif

150
static struct zone_reclaim_stat *get_reclaim_stat(struct mem_cgroup_zone *mz)
151
{
152
	return &mem_cgroup_zone_lruvec(mz->zone, mz->mem_cgroup)->reclaim_stat;
153 154
}

155 156
static unsigned long zone_nr_lru_pages(struct mem_cgroup_zone *mz,
				       enum lru_list lru)
157
{
158
	if (!mem_cgroup_disabled())
159 160 161 162
		return mem_cgroup_zone_nr_lru_pages(mz->mem_cgroup,
						    zone_to_nid(mz->zone),
						    zone_idx(mz->zone),
						    BIT(lru));
163

164
	return zone_page_state(mz->zone, NR_LRU_BASE + lru);
165 166 167
}


L
Linus Torvalds 已提交
168 169 170
/*
 * Add a shrinker callback to be called from the vm
 */
171
void register_shrinker(struct shrinker *shrinker)
L
Linus Torvalds 已提交
172
{
173
	atomic_long_set(&shrinker->nr_in_batch, 0);
174 175 176
	down_write(&shrinker_rwsem);
	list_add_tail(&shrinker->list, &shrinker_list);
	up_write(&shrinker_rwsem);
L
Linus Torvalds 已提交
177
}
178
EXPORT_SYMBOL(register_shrinker);
L
Linus Torvalds 已提交
179 180 181 182

/*
 * Remove one
 */
183
void unregister_shrinker(struct shrinker *shrinker)
L
Linus Torvalds 已提交
184 185 186 187 188
{
	down_write(&shrinker_rwsem);
	list_del(&shrinker->list);
	up_write(&shrinker_rwsem);
}
189
EXPORT_SYMBOL(unregister_shrinker);
L
Linus Torvalds 已提交
190

191 192 193 194 195 196 197 198
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
199 200 201 202 203 204 205 206 207
#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
208
 * If the vm encountered mapped pages on the LRU it increase the pressure on
L
Linus Torvalds 已提交
209 210 211 212 213 214 215
 * 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.
216 217
 *
 * Returns the number of slab objects which we shrunk.
L
Linus Torvalds 已提交
218
 */
219
unsigned long shrink_slab(struct shrink_control *shrink,
220
			  unsigned long nr_pages_scanned,
221
			  unsigned long lru_pages)
L
Linus Torvalds 已提交
222 223
{
	struct shrinker *shrinker;
224
	unsigned long ret = 0;
L
Linus Torvalds 已提交
225

226 227
	if (nr_pages_scanned == 0)
		nr_pages_scanned = SWAP_CLUSTER_MAX;
L
Linus Torvalds 已提交
228

229 230 231 232 233
	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
234 235 236

	list_for_each_entry(shrinker, &shrinker_list, list) {
		unsigned long long delta;
237 238
		long total_scan;
		long max_pass;
239
		int shrink_ret = 0;
240 241
		long nr;
		long new_nr;
242 243
		long batch_size = shrinker->batch ? shrinker->batch
						  : SHRINK_BATCH;
L
Linus Torvalds 已提交
244

245 246 247 248
		max_pass = do_shrinker_shrink(shrinker, shrink, 0);
		if (max_pass <= 0)
			continue;

249 250 251 252 253
		/*
		 * copy the current shrinker scan count into a local variable
		 * and zero it so that other concurrent shrinker invocations
		 * don't also do this scanning work.
		 */
254
		nr = atomic_long_xchg(&shrinker->nr_in_batch, 0);
255 256

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

268 269 270 271 272 273 274 275 276 277 278 279 280 281 282
		/*
		 * We need to avoid excessive windup on filesystem shrinkers
		 * due to large numbers of GFP_NOFS allocations causing the
		 * shrinkers to return -1 all the time. This results in a large
		 * nr being built up so when a shrink that can do some work
		 * comes along it empties the entire cache due to nr >>>
		 * max_pass.  This is bad for sustaining a working set in
		 * memory.
		 *
		 * Hence only allow the shrinker to scan the entire cache when
		 * a large delta change is calculated directly.
		 */
		if (delta < max_pass / 4)
			total_scan = min(total_scan, max_pass / 2);

283 284 285 286 287
		/*
		 * Avoid risking looping forever due to too large nr value:
		 * never try to free more than twice the estimate number of
		 * freeable entries.
		 */
288 289
		if (total_scan > max_pass * 2)
			total_scan = max_pass * 2;
L
Linus Torvalds 已提交
290

291
		trace_mm_shrink_slab_start(shrinker, shrink, nr,
292 293 294
					nr_pages_scanned, lru_pages,
					max_pass, delta, total_scan);

295
		while (total_scan >= batch_size) {
296
			int nr_before;
L
Linus Torvalds 已提交
297

298 299
			nr_before = do_shrinker_shrink(shrinker, shrink, 0);
			shrink_ret = do_shrinker_shrink(shrinker, shrink,
300
							batch_size);
L
Linus Torvalds 已提交
301 302
			if (shrink_ret == -1)
				break;
303 304
			if (shrink_ret < nr_before)
				ret += nr_before - shrink_ret;
305 306
			count_vm_events(SLABS_SCANNED, batch_size);
			total_scan -= batch_size;
L
Linus Torvalds 已提交
307 308 309 310

			cond_resched();
		}

311 312 313 314 315
		/*
		 * move the unused scan count back into the shrinker in a
		 * manner that handles concurrent updates. If we exhausted the
		 * scan, there is no need to do an update.
		 */
316 317 318 319 320
		if (total_scan > 0)
			new_nr = atomic_long_add_return(total_scan,
					&shrinker->nr_in_batch);
		else
			new_nr = atomic_long_read(&shrinker->nr_in_batch);
321 322

		trace_mm_shrink_slab_end(shrinker, shrink_ret, nr, new_nr);
L
Linus Torvalds 已提交
323 324
	}
	up_read(&shrinker_rwsem);
325 326
out:
	cond_resched();
327
	return ret;
L
Linus Torvalds 已提交
328 329 330 331
}

static inline int is_page_cache_freeable(struct page *page)
{
332 333 334 335 336
	/*
	 * 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.
	 */
337
	return page_count(page) - page_has_private(page) == 2;
L
Linus Torvalds 已提交
338 339
}

340 341
static int may_write_to_queue(struct backing_dev_info *bdi,
			      struct scan_control *sc)
L
Linus Torvalds 已提交
342
{
343
	if (current->flags & PF_SWAPWRITE)
L
Linus Torvalds 已提交
344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366
		return 1;
	if (!bdi_write_congested(bdi))
		return 1;
	if (bdi == current->backing_dev_info)
		return 1;
	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
367
	lock_page(page);
368 369
	if (page_mapping(page) == mapping)
		mapping_set_error(mapping, error);
L
Linus Torvalds 已提交
370 371 372
	unlock_page(page);
}

373 374 375 376 377 378 379 380 381 382 383 384
/* 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
385
/*
A
Andrew Morton 已提交
386 387
 * pageout is called by shrink_page_list() for each dirty page.
 * Calls ->writepage().
L
Linus Torvalds 已提交
388
 */
389
static pageout_t pageout(struct page *page, struct address_space *mapping,
390
			 struct scan_control *sc)
L
Linus Torvalds 已提交
391 392 393 394 395 396 397 398
{
	/*
	 * 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.
	 *
399
	 * If this process is currently in __generic_file_aio_write() against
L
Linus Torvalds 已提交
400 401 402 403 404 405 406 407 408 409 410 411 412 413 414
	 * 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.
		 */
415
		if (page_has_private(page)) {
L
Linus Torvalds 已提交
416 417
			if (try_to_free_buffers(page)) {
				ClearPageDirty(page);
418
				printk("%s: orphaned page\n", __func__);
L
Linus Torvalds 已提交
419 420 421 422 423 424 425
				return PAGE_CLEAN;
			}
		}
		return PAGE_KEEP;
	}
	if (mapping->a_ops->writepage == NULL)
		return PAGE_ACTIVATE;
426
	if (!may_write_to_queue(mapping->backing_dev_info, sc))
L
Linus Torvalds 已提交
427 428 429 430 431 432 433
		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,
434 435
			.range_start = 0,
			.range_end = LLONG_MAX,
L
Linus Torvalds 已提交
436 437 438 439 440 441 442
			.for_reclaim = 1,
		};

		SetPageReclaim(page);
		res = mapping->a_ops->writepage(page, &wbc);
		if (res < 0)
			handle_write_error(mapping, page, res);
443
		if (res == AOP_WRITEPAGE_ACTIVATE) {
L
Linus Torvalds 已提交
444 445 446
			ClearPageReclaim(page);
			return PAGE_ACTIVATE;
		}
447

L
Linus Torvalds 已提交
448 449 450 451
		if (!PageWriteback(page)) {
			/* synchronous write or broken a_ops? */
			ClearPageReclaim(page);
		}
M
Mel Gorman 已提交
452
		trace_mm_vmscan_writepage(page, trace_reclaim_flags(page));
453
		inc_zone_page_state(page, NR_VMSCAN_WRITE);
L
Linus Torvalds 已提交
454 455 456 457 458 459
		return PAGE_SUCCESS;
	}

	return PAGE_CLEAN;
}

460
/*
N
Nick Piggin 已提交
461 462
 * Same as remove_mapping, but if the page is removed from the mapping, it
 * gets returned with a refcount of 0.
463
 */
N
Nick Piggin 已提交
464
static int __remove_mapping(struct address_space *mapping, struct page *page)
465
{
466 467
	BUG_ON(!PageLocked(page));
	BUG_ON(mapping != page_mapping(page));
468

N
Nick Piggin 已提交
469
	spin_lock_irq(&mapping->tree_lock);
470
	/*
N
Nick Piggin 已提交
471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493
	 * 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.
494
	 */
N
Nick Piggin 已提交
495
	if (!page_freeze_refs(page, 2))
496
		goto cannot_free;
N
Nick Piggin 已提交
497 498 499
	/* note: atomic_cmpxchg in page_freeze_refs provides the smp_rmb */
	if (unlikely(PageDirty(page))) {
		page_unfreeze_refs(page, 2);
500
		goto cannot_free;
N
Nick Piggin 已提交
501
	}
502 503 504 505

	if (PageSwapCache(page)) {
		swp_entry_t swap = { .val = page_private(page) };
		__delete_from_swap_cache(page);
N
Nick Piggin 已提交
506
		spin_unlock_irq(&mapping->tree_lock);
507
		swapcache_free(swap, page);
N
Nick Piggin 已提交
508
	} else {
509 510 511 512
		void (*freepage)(struct page *);

		freepage = mapping->a_ops->freepage;

513
		__delete_from_page_cache(page);
N
Nick Piggin 已提交
514
		spin_unlock_irq(&mapping->tree_lock);
515
		mem_cgroup_uncharge_cache_page(page);
516 517 518

		if (freepage != NULL)
			freepage(page);
519 520 521 522 523
	}

	return 1;

cannot_free:
N
Nick Piggin 已提交
524
	spin_unlock_irq(&mapping->tree_lock);
525 526 527
	return 0;
}

N
Nick Piggin 已提交
528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547
/*
 * 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
548 549 550 551 552 553 554 555 556 557 558 559 560
/**
 * 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);
561
	int was_unevictable = PageUnevictable(page);
L
Lee Schermerhorn 已提交
562 563 564 565 566 567 568 569 570 571 572 573 574

	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.
		 */
575
		lru = active + page_lru_base_type(page);
L
Lee Schermerhorn 已提交
576 577 578 579 580 581 582 583
		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);
584
		/*
585 586 587
		 * When racing with an mlock or AS_UNEVICTABLE clearing
		 * (page is unlocked) make sure that if the other thread
		 * does not observe our setting of PG_lru and fails
588
		 * isolation/check_move_unevictable_pages,
589
		 * we see PG_mlocked/AS_UNEVICTABLE cleared below and move
590 591
		 * the page back to the evictable list.
		 *
592
		 * The other side is TestClearPageMlocked() or shmem_lock().
593 594
		 */
		smp_mb();
L
Lee Schermerhorn 已提交
595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612
	}

	/*
	 * 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.
		 */
	}

613 614 615 616 617
	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
618 619 620
	put_page(page);		/* drop ref from isolate */
}

621 622 623
enum page_references {
	PAGEREF_RECLAIM,
	PAGEREF_RECLAIM_CLEAN,
624
	PAGEREF_KEEP,
625 626 627 628
	PAGEREF_ACTIVATE,
};

static enum page_references page_check_references(struct page *page,
629
						  struct mem_cgroup_zone *mz,
630 631
						  struct scan_control *sc)
{
632
	int referenced_ptes, referenced_page;
633 634
	unsigned long vm_flags;

635 636
	referenced_ptes = page_referenced(page, 1, sc->target_mem_cgroup,
					  &vm_flags);
637
	referenced_page = TestClearPageReferenced(page);
638 639 640 641 642 643 644 645

	/*
	 * 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;

646
	if (referenced_ptes) {
647
		if (PageSwapBacked(page))
648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664
			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);

665
		if (referenced_page || referenced_ptes > 1)
666 667
			return PAGEREF_ACTIVATE;

668 669 670 671 672 673
		/*
		 * Activate file-backed executable pages after first usage.
		 */
		if (vm_flags & VM_EXEC)
			return PAGEREF_ACTIVATE;

674 675
		return PAGEREF_KEEP;
	}
676 677

	/* Reclaim if clean, defer dirty pages to writeback */
678
	if (referenced_page && !PageSwapBacked(page))
679 680 681
		return PAGEREF_RECLAIM_CLEAN;

	return PAGEREF_RECLAIM;
682 683
}

L
Linus Torvalds 已提交
684
/*
A
Andrew Morton 已提交
685
 * shrink_page_list() returns the number of reclaimed pages
L
Linus Torvalds 已提交
686
 */
A
Andrew Morton 已提交
687
static unsigned long shrink_page_list(struct list_head *page_list,
688
				      struct mem_cgroup_zone *mz,
689
				      struct scan_control *sc,
690 691 692
				      int priority,
				      unsigned long *ret_nr_dirty,
				      unsigned long *ret_nr_writeback)
L
Linus Torvalds 已提交
693 694
{
	LIST_HEAD(ret_pages);
695
	LIST_HEAD(free_pages);
L
Linus Torvalds 已提交
696
	int pgactivate = 0;
697 698
	unsigned long nr_dirty = 0;
	unsigned long nr_congested = 0;
699
	unsigned long nr_reclaimed = 0;
700
	unsigned long nr_writeback = 0;
L
Linus Torvalds 已提交
701 702 703 704

	cond_resched();

	while (!list_empty(page_list)) {
705
		enum page_references references;
L
Linus Torvalds 已提交
706 707 708 709 710 711 712 713 714
		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
715
		if (!trylock_page(page))
L
Linus Torvalds 已提交
716 717
			goto keep;

N
Nick Piggin 已提交
718
		VM_BUG_ON(PageActive(page));
719
		VM_BUG_ON(page_zone(page) != mz->zone);
L
Linus Torvalds 已提交
720 721

		sc->nr_scanned++;
722

N
Nick Piggin 已提交
723 724
		if (unlikely(!page_evictable(page, NULL)))
			goto cull_mlocked;
L
Lee Schermerhorn 已提交
725

726
		if (!sc->may_unmap && page_mapped(page))
727 728
			goto keep_locked;

L
Linus Torvalds 已提交
729 730 731 732
		/* Double the slab pressure for mapped and swapcache pages */
		if (page_mapped(page) || PageSwapCache(page))
			sc->nr_scanned++;

733 734 735 736
		may_enter_fs = (sc->gfp_mask & __GFP_FS) ||
			(PageSwapCache(page) && (sc->gfp_mask & __GFP_IO));

		if (PageWriteback(page)) {
737
			nr_writeback++;
738 739
			unlock_page(page);
			goto keep;
740
		}
L
Linus Torvalds 已提交
741

742
		references = page_check_references(page, mz, sc);
743 744
		switch (references) {
		case PAGEREF_ACTIVATE:
L
Linus Torvalds 已提交
745
			goto activate_locked;
746 747
		case PAGEREF_KEEP:
			goto keep_locked;
748 749 750 751
		case PAGEREF_RECLAIM:
		case PAGEREF_RECLAIM_CLEAN:
			; /* try to reclaim the page below */
		}
L
Linus Torvalds 已提交
752 753 754 755 756

		/*
		 * Anonymous process memory has backing store?
		 * Try to allocate it some swap space here.
		 */
N
Nick Piggin 已提交
757
		if (PageAnon(page) && !PageSwapCache(page)) {
758 759
			if (!(sc->gfp_mask & __GFP_IO))
				goto keep_locked;
760
			if (!add_to_swap(page))
L
Linus Torvalds 已提交
761
				goto activate_locked;
762
			may_enter_fs = 1;
N
Nick Piggin 已提交
763
		}
L
Linus Torvalds 已提交
764 765 766 767 768 769 770 771

		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) {
772
			switch (try_to_unmap(page, TTU_UNMAP)) {
L
Linus Torvalds 已提交
773 774 775 776
			case SWAP_FAIL:
				goto activate_locked;
			case SWAP_AGAIN:
				goto keep_locked;
N
Nick Piggin 已提交
777 778
			case SWAP_MLOCK:
				goto cull_mlocked;
L
Linus Torvalds 已提交
779 780 781 782 783 784
			case SWAP_SUCCESS:
				; /* try to free the page below */
			}
		}

		if (PageDirty(page)) {
785 786
			nr_dirty++;

787 788
			/*
			 * Only kswapd can writeback filesystem pages to
789 790
			 * avoid risk of stack overflow but do not writeback
			 * unless under significant pressure.
791
			 */
792 793
			if (page_is_file_cache(page) &&
					(!current_is_kswapd() || priority >= DEF_PRIORITY - 2)) {
794 795 796 797 798 799 800 801 802
				/*
				 * Immediately reclaim when written back.
				 * Similar in principal to deactivate_page()
				 * except we already have the page isolated
				 * and know it's dirty
				 */
				inc_zone_page_state(page, NR_VMSCAN_IMMEDIATE);
				SetPageReclaim(page);

803 804 805
				goto keep_locked;
			}

806
			if (references == PAGEREF_RECLAIM_CLEAN)
L
Linus Torvalds 已提交
807
				goto keep_locked;
808
			if (!may_enter_fs)
L
Linus Torvalds 已提交
809
				goto keep_locked;
810
			if (!sc->may_writepage)
L
Linus Torvalds 已提交
811 812 813
				goto keep_locked;

			/* Page is dirty, try to write it out here */
814
			switch (pageout(page, mapping, sc)) {
L
Linus Torvalds 已提交
815
			case PAGE_KEEP:
816
				nr_congested++;
L
Linus Torvalds 已提交
817 818 819 820
				goto keep_locked;
			case PAGE_ACTIVATE:
				goto activate_locked;
			case PAGE_SUCCESS:
821
				if (PageWriteback(page))
822
					goto keep;
823
				if (PageDirty(page))
L
Linus Torvalds 已提交
824
					goto keep;
825

L
Linus Torvalds 已提交
826 827 828 829
				/*
				 * A synchronous write - probably a ramdisk.  Go
				 * ahead and try to reclaim the page.
				 */
N
Nick Piggin 已提交
830
				if (!trylock_page(page))
L
Linus Torvalds 已提交
831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849
					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
850
		 * will do this, as well as the blockdev mapping.
L
Linus Torvalds 已提交
851 852 853 854 855 856 857 858 859 860
		 * 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.
		 */
861
		if (page_has_private(page)) {
L
Linus Torvalds 已提交
862 863
			if (!try_to_release_page(page, sc->gfp_mask))
				goto activate_locked;
N
Nick Piggin 已提交
864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879
			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
880 881
		}

N
Nick Piggin 已提交
882
		if (!mapping || !__remove_mapping(mapping, page))
883
			goto keep_locked;
L
Linus Torvalds 已提交
884

N
Nick Piggin 已提交
885 886 887 888 889 890 891 892
		/*
		 * 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
893
free_it:
894
		nr_reclaimed++;
895 896 897 898 899 900

		/*
		 * 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
901 902
		continue;

N
Nick Piggin 已提交
903
cull_mlocked:
904 905
		if (PageSwapCache(page))
			try_to_free_swap(page);
N
Nick Piggin 已提交
906 907 908 909
		unlock_page(page);
		putback_lru_page(page);
		continue;

L
Linus Torvalds 已提交
910
activate_locked:
911 912
		/* Not a candidate for swapping, so reclaim swap space. */
		if (PageSwapCache(page) && vm_swap_full())
913
			try_to_free_swap(page);
L
Lee Schermerhorn 已提交
914
		VM_BUG_ON(PageActive(page));
L
Linus Torvalds 已提交
915 916 917 918 919 920
		SetPageActive(page);
		pgactivate++;
keep_locked:
		unlock_page(page);
keep:
		list_add(&page->lru, &ret_pages);
N
Nick Piggin 已提交
921
		VM_BUG_ON(PageLRU(page) || PageUnevictable(page));
L
Linus Torvalds 已提交
922
	}
923

924 925 926 927 928 929
	/*
	 * 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
	 */
930
	if (nr_dirty && nr_dirty == nr_congested && global_reclaim(sc))
931
		zone_set_flag(mz->zone, ZONE_CONGESTED);
932

933
	free_hot_cold_page_list(&free_pages, 1);
934

L
Linus Torvalds 已提交
935
	list_splice(&ret_pages, page_list);
936
	count_vm_events(PGACTIVATE, pgactivate);
937 938
	*ret_nr_dirty += nr_dirty;
	*ret_nr_writeback += nr_writeback;
939
	return nr_reclaimed;
L
Linus Torvalds 已提交
940 941
}

A
Andy Whitcroft 已提交
942 943 944 945 946 947 948 949 950 951
/*
 * 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.
 */
952
int __isolate_lru_page(struct page *page, isolate_mode_t mode)
A
Andy Whitcroft 已提交
953 954 955 956 957 958 959
{
	int ret = -EINVAL;

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

M
Mel Gorman 已提交
960
	/* Do not give back unevictable pages for compaction */
L
Lee Schermerhorn 已提交
961 962 963
	if (PageUnevictable(page))
		return ret;

A
Andy Whitcroft 已提交
964
	ret = -EBUSY;
K
KAMEZAWA Hiroyuki 已提交
965

966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998
	/*
	 * To minimise LRU disruption, the caller can indicate that it only
	 * wants to isolate pages it will be able to operate on without
	 * blocking - clean pages for the most part.
	 *
	 * ISOLATE_CLEAN means that only clean pages should be isolated. This
	 * is used by reclaim when it is cannot write to backing storage
	 *
	 * ISOLATE_ASYNC_MIGRATE is used to indicate that it only wants to pages
	 * that it is possible to migrate without blocking
	 */
	if (mode & (ISOLATE_CLEAN|ISOLATE_ASYNC_MIGRATE)) {
		/* All the caller can do on PageWriteback is block */
		if (PageWriteback(page))
			return ret;

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

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

			/*
			 * Only pages without mappings or that have a
			 * ->migratepage callback are possible to migrate
			 * without blocking
			 */
			mapping = page_mapping(page);
			if (mapping && !mapping->a_ops->migratepage)
				return ret;
		}
	}
999

1000 1001 1002
	if ((mode & ISOLATE_UNMAPPED) && page_mapped(page))
		return ret;

A
Andy Whitcroft 已提交
1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015
	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
1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026
/*
 * zone->lru_lock is heavily contended.  Some of the functions that
 * shrink the lists perform better by taking out a batch of pages
 * and working on them outside the LRU lock.
 *
 * For pagecache intensive workloads, this function is the hottest
 * spot in the kernel (apart from copy_*_user functions).
 *
 * Appropriate locks must be held before calling this function.
 *
 * @nr_to_scan:	The number of pages to look through on the list.
H
Hugh Dickins 已提交
1027
 * @mz:		The mem_cgroup_zone to pull pages from.
L
Linus Torvalds 已提交
1028
 * @dst:	The temp list to put pages on to.
H
Hugh Dickins 已提交
1029
 * @nr_scanned:	The number of pages that were scanned.
1030
 * @sc:		The scan_control struct for this reclaim session
A
Andy Whitcroft 已提交
1031
 * @mode:	One of the LRU isolation modes
1032
 * @lru:	LRU list id for isolating
L
Linus Torvalds 已提交
1033 1034 1035
 *
 * returns how many pages were moved onto *@dst.
 */
1036
static unsigned long isolate_lru_pages(unsigned long nr_to_scan,
H
Hugh Dickins 已提交
1037
		struct mem_cgroup_zone *mz, struct list_head *dst,
1038
		unsigned long *nr_scanned, struct scan_control *sc,
1039
		isolate_mode_t mode, enum lru_list lru)
L
Linus Torvalds 已提交
1040
{
H
Hugh Dickins 已提交
1041 1042
	struct lruvec *lruvec;
	struct list_head *src;
1043
	unsigned long nr_taken = 0;
1044
	unsigned long scan;
1045
	int file = is_file_lru(lru);
H
Hugh Dickins 已提交
1046 1047 1048

	lruvec = mem_cgroup_zone_lruvec(mz->zone, mz->mem_cgroup);
	src = &lruvec->lists[lru];
L
Linus Torvalds 已提交
1049

1050
	for (scan = 0; scan < nr_to_scan && !list_empty(src); scan++) {
A
Andy Whitcroft 已提交
1051 1052
		struct page *page;

L
Linus Torvalds 已提交
1053 1054 1055
		page = lru_to_page(src);
		prefetchw_prev_lru_page(page, src, flags);

N
Nick Piggin 已提交
1056
		VM_BUG_ON(!PageLRU(page));
N
Nick Piggin 已提交
1057

1058
		switch (__isolate_lru_page(page, mode)) {
A
Andy Whitcroft 已提交
1059
		case 0:
1060
			mem_cgroup_lru_del_list(page, lru);
A
Andy Whitcroft 已提交
1061
			list_move(&page->lru, dst);
1062
			nr_taken += hpage_nr_pages(page);
A
Andy Whitcroft 已提交
1063 1064 1065 1066 1067 1068
			break;

		case -EBUSY:
			/* else it is being freed elsewhere */
			list_move(&page->lru, src);
			continue;
1069

A
Andy Whitcroft 已提交
1070 1071 1072
		default:
			BUG();
		}
L
Linus Torvalds 已提交
1073 1074
	}

H
Hugh Dickins 已提交
1075
	*nr_scanned = scan;
1076

1077
	trace_mm_vmscan_lru_isolate(sc->order,
1078 1079
			nr_to_scan, scan,
			nr_taken,
1080
			mode, file);
L
Linus Torvalds 已提交
1081 1082 1083
	return nr_taken;
}

1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094
/**
 * 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
1095 1096 1097
 * 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.
1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112
 *
 * 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;

1113 1114
	VM_BUG_ON(!page_count(page));

1115 1116 1117 1118
	if (PageLRU(page)) {
		struct zone *zone = page_zone(page);

		spin_lock_irq(&zone->lru_lock);
1119
		if (PageLRU(page)) {
L
Lee Schermerhorn 已提交
1120
			int lru = page_lru(page);
1121
			ret = 0;
1122
			get_page(page);
1123
			ClearPageLRU(page);
1124 1125

			del_page_from_lru_list(zone, page, lru);
1126 1127 1128 1129 1130 1131
		}
		spin_unlock_irq(&zone->lru_lock);
	}
	return ret;
}

1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142
/*
 * 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;

1143
	if (!global_reclaim(sc))
1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156
		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;
}

1157
static noinline_for_stack void
1158 1159
putback_inactive_pages(struct mem_cgroup_zone *mz,
		       struct list_head *page_list)
1160
{
1161
	struct zone_reclaim_stat *reclaim_stat = get_reclaim_stat(mz);
1162 1163
	struct zone *zone = mz->zone;
	LIST_HEAD(pages_to_free);
1164 1165 1166 1167 1168

	/*
	 * Put back any unfreeable pages.
	 */
	while (!list_empty(page_list)) {
1169
		struct page *page = lru_to_page(page_list);
1170
		int lru;
1171

1172 1173 1174 1175 1176 1177 1178 1179
		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;
		}
1180
		SetPageLRU(page);
1181
		lru = page_lru(page);
1182
		add_page_to_lru_list(zone, page, lru);
1183 1184
		if (is_active_lru(lru)) {
			int file = is_file_lru(lru);
1185 1186
			int numpages = hpage_nr_pages(page);
			reclaim_stat->recent_rotated[file] += numpages;
1187
		}
1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198
		if (put_page_testzero(page)) {
			__ClearPageLRU(page);
			__ClearPageActive(page);
			del_page_from_lru_list(zone, page, lru);

			if (unlikely(PageCompound(page))) {
				spin_unlock_irq(&zone->lru_lock);
				(*get_compound_page_dtor(page))(page);
				spin_lock_irq(&zone->lru_lock);
			} else
				list_add(&page->lru, &pages_to_free);
1199 1200 1201
		}
	}

1202 1203 1204 1205
	/*
	 * To save our caller's stack, now use input list for pages to free.
	 */
	list_splice(&pages_to_free, page_list);
1206 1207
}

1208 1209
static noinline_for_stack void
update_isolated_counts(struct mem_cgroup_zone *mz,
1210
		       struct list_head *page_list,
1211
		       unsigned long *nr_anon,
1212
		       unsigned long *nr_file)
1213
{
1214
	struct zone *zone = mz->zone;
1215
	unsigned int count[NR_LRU_LISTS] = { 0, };
1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232
	unsigned long nr_active = 0;
	struct page *page;
	int lru;

	/*
	 * Count pages and clear active flags
	 */
	list_for_each_entry(page, page_list, lru) {
		int numpages = hpage_nr_pages(page);
		lru = page_lru_base_type(page);
		if (PageActive(page)) {
			lru += LRU_ACTIVE;
			ClearPageActive(page);
			nr_active += numpages;
		}
		count[lru] += numpages;
	}
1233

1234
	preempt_disable();
1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248
	__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];

1249 1250 1251
	__mod_zone_page_state(zone, NR_ISOLATED_ANON, *nr_anon);
	__mod_zone_page_state(zone, NR_ISOLATED_FILE, *nr_file);
	preempt_enable();
1252 1253
}

L
Linus Torvalds 已提交
1254
/*
A
Andrew Morton 已提交
1255 1256
 * shrink_inactive_list() is a helper for shrink_zone().  It returns the number
 * of reclaimed pages
L
Linus Torvalds 已提交
1257
 */
1258
static noinline_for_stack unsigned long
1259
shrink_inactive_list(unsigned long nr_to_scan, struct mem_cgroup_zone *mz,
1260
		     struct scan_control *sc, int priority, enum lru_list lru)
L
Linus Torvalds 已提交
1261 1262
{
	LIST_HEAD(page_list);
1263
	unsigned long nr_scanned;
1264
	unsigned long nr_reclaimed = 0;
1265 1266 1267
	unsigned long nr_taken;
	unsigned long nr_anon;
	unsigned long nr_file;
1268 1269
	unsigned long nr_dirty = 0;
	unsigned long nr_writeback = 0;
1270
	isolate_mode_t isolate_mode = 0;
1271
	int file = is_file_lru(lru);
1272
	struct zone *zone = mz->zone;
1273
	struct zone_reclaim_stat *reclaim_stat = get_reclaim_stat(mz);
1274

1275
	while (unlikely(too_many_isolated(zone, file, sc))) {
1276
		congestion_wait(BLK_RW_ASYNC, HZ/10);
1277 1278 1279 1280 1281 1282

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

L
Linus Torvalds 已提交
1283
	lru_add_drain();
1284 1285

	if (!sc->may_unmap)
1286
		isolate_mode |= ISOLATE_UNMAPPED;
1287
	if (!sc->may_writepage)
1288
		isolate_mode |= ISOLATE_CLEAN;
1289

L
Linus Torvalds 已提交
1290
	spin_lock_irq(&zone->lru_lock);
1291

1292
	nr_taken = isolate_lru_pages(nr_to_scan, mz, &page_list, &nr_scanned,
1293
				     sc, isolate_mode, lru);
1294
	if (global_reclaim(sc)) {
1295 1296 1297 1298 1299 1300 1301 1302
		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);
	}
1303
	spin_unlock_irq(&zone->lru_lock);
1304

1305
	if (nr_taken == 0)
1306
		return 0;
A
Andy Whitcroft 已提交
1307

1308 1309
	update_isolated_counts(mz, &page_list, &nr_anon, &nr_file);

1310
	nr_reclaimed = shrink_page_list(&page_list, mz, sc, priority,
1311
						&nr_dirty, &nr_writeback);
1312

1313 1314
	spin_lock_irq(&zone->lru_lock);

1315 1316 1317
	reclaim_stat->recent_scanned[0] += nr_anon;
	reclaim_stat->recent_scanned[1] += nr_file;

Y
Ying Han 已提交
1318 1319 1320 1321 1322 1323 1324 1325
	if (global_reclaim(sc)) {
		if (current_is_kswapd())
			__count_zone_vm_events(PGSTEAL_KSWAPD, zone,
					       nr_reclaimed);
		else
			__count_zone_vm_events(PGSTEAL_DIRECT, zone,
					       nr_reclaimed);
	}
N
Nick Piggin 已提交
1326

1327 1328 1329 1330 1331 1332 1333 1334
	putback_inactive_pages(mz, &page_list);

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

	spin_unlock_irq(&zone->lru_lock);

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

1362 1363 1364 1365
	trace_mm_vmscan_lru_shrink_inactive(zone->zone_pgdat->node_id,
		zone_idx(zone),
		nr_scanned, nr_reclaimed,
		priority,
M
Mel Gorman 已提交
1366
		trace_shrink_flags(file));
1367
	return nr_reclaimed;
L
Linus Torvalds 已提交
1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386
}

/*
 * 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.
 */
1387

1388 1389
static void move_active_pages_to_lru(struct zone *zone,
				     struct list_head *list,
1390
				     struct list_head *pages_to_free,
1391 1392 1393 1394 1395 1396
				     enum lru_list lru)
{
	unsigned long pgmoved = 0;
	struct page *page;

	while (!list_empty(list)) {
1397 1398
		struct lruvec *lruvec;

1399 1400 1401 1402 1403
		page = lru_to_page(list);

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

1404 1405
		lruvec = mem_cgroup_lru_add_list(zone, page, lru);
		list_move(&page->lru, &lruvec->lists[lru]);
1406
		pgmoved += hpage_nr_pages(page);
1407

1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418
		if (put_page_testzero(page)) {
			__ClearPageLRU(page);
			__ClearPageActive(page);
			del_page_from_lru_list(zone, page, lru);

			if (unlikely(PageCompound(page))) {
				spin_unlock_irq(&zone->lru_lock);
				(*get_compound_page_dtor(page))(page);
				spin_lock_irq(&zone->lru_lock);
			} else
				list_add(&page->lru, pages_to_free);
1419 1420 1421 1422 1423 1424
		}
	}
	__mod_zone_page_state(zone, NR_LRU_BASE + lru, pgmoved);
	if (!is_active_lru(lru))
		__count_vm_events(PGDEACTIVATE, pgmoved);
}
1425

H
Hugh Dickins 已提交
1426
static void shrink_active_list(unsigned long nr_to_scan,
1427 1428
			       struct mem_cgroup_zone *mz,
			       struct scan_control *sc,
1429
			       int priority, enum lru_list lru)
L
Linus Torvalds 已提交
1430
{
1431
	unsigned long nr_taken;
H
Hugh Dickins 已提交
1432
	unsigned long nr_scanned;
1433
	unsigned long vm_flags;
L
Linus Torvalds 已提交
1434
	LIST_HEAD(l_hold);	/* The pages which were snipped off */
1435
	LIST_HEAD(l_active);
1436
	LIST_HEAD(l_inactive);
L
Linus Torvalds 已提交
1437
	struct page *page;
1438
	struct zone_reclaim_stat *reclaim_stat = get_reclaim_stat(mz);
1439
	unsigned long nr_rotated = 0;
1440
	isolate_mode_t isolate_mode = 0;
1441
	int file = is_file_lru(lru);
1442
	struct zone *zone = mz->zone;
L
Linus Torvalds 已提交
1443 1444

	lru_add_drain();
1445 1446

	if (!sc->may_unmap)
1447
		isolate_mode |= ISOLATE_UNMAPPED;
1448
	if (!sc->may_writepage)
1449
		isolate_mode |= ISOLATE_CLEAN;
1450

L
Linus Torvalds 已提交
1451
	spin_lock_irq(&zone->lru_lock);
1452

1453
	nr_taken = isolate_lru_pages(nr_to_scan, mz, &l_hold, &nr_scanned, sc,
1454
				     isolate_mode, lru);
1455
	if (global_reclaim(sc))
H
Hugh Dickins 已提交
1456
		zone->pages_scanned += nr_scanned;
1457

1458
	reclaim_stat->recent_scanned[file] += nr_taken;
1459

H
Hugh Dickins 已提交
1460
	__count_zone_vm_events(PGREFILL, zone, nr_scanned);
1461
	__mod_zone_page_state(zone, NR_LRU_BASE + lru, -nr_taken);
K
KOSAKI Motohiro 已提交
1462
	__mod_zone_page_state(zone, NR_ISOLATED_ANON + file, nr_taken);
L
Linus Torvalds 已提交
1463 1464 1465 1466 1467 1468
	spin_unlock_irq(&zone->lru_lock);

	while (!list_empty(&l_hold)) {
		cond_resched();
		page = lru_to_page(&l_hold);
		list_del(&page->lru);
1469

L
Lee Schermerhorn 已提交
1470 1471 1472 1473 1474
		if (unlikely(!page_evictable(page, NULL))) {
			putback_lru_page(page);
			continue;
		}

1475 1476 1477 1478 1479 1480 1481 1482
		if (unlikely(buffer_heads_over_limit)) {
			if (page_has_private(page) && trylock_page(page)) {
				if (page_has_private(page))
					try_to_release_page(page, 0);
				unlock_page(page);
			}
		}

1483 1484
		if (page_referenced(page, 0, sc->target_mem_cgroup,
				    &vm_flags)) {
1485
			nr_rotated += hpage_nr_pages(page);
1486 1487 1488 1489 1490 1491 1492 1493 1494
			/*
			 * 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.
			 */
1495
			if ((vm_flags & VM_EXEC) && page_is_file_cache(page)) {
1496 1497 1498 1499
				list_add(&page->lru, &l_active);
				continue;
			}
		}
1500

1501
		ClearPageActive(page);	/* we are de-activating */
L
Linus Torvalds 已提交
1502 1503 1504
		list_add(&page->lru, &l_inactive);
	}

1505
	/*
1506
	 * Move pages back to the lru list.
1507
	 */
1508
	spin_lock_irq(&zone->lru_lock);
1509
	/*
1510 1511 1512 1513
	 * 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.
1514
	 */
1515
	reclaim_stat->recent_rotated[file] += nr_rotated;
1516

1517 1518
	move_active_pages_to_lru(zone, &l_active, &l_hold, lru);
	move_active_pages_to_lru(zone, &l_inactive, &l_hold, lru - LRU_ACTIVE);
K
KOSAKI Motohiro 已提交
1519
	__mod_zone_page_state(zone, NR_ISOLATED_ANON + file, -nr_taken);
1520
	spin_unlock_irq(&zone->lru_lock);
1521 1522

	free_hot_cold_page_list(&l_hold, 1);
L
Linus Torvalds 已提交
1523 1524
}

1525
#ifdef CONFIG_SWAP
1526
static int inactive_anon_is_low_global(struct zone *zone)
1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538
{
	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;
}

1539 1540 1541 1542 1543 1544 1545 1546
/**
 * 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.
 */
1547
static int inactive_anon_is_low(struct mem_cgroup_zone *mz)
1548
{
1549 1550 1551 1552 1553 1554 1555
	/*
	 * If we don't have swap space, anonymous page deactivation
	 * is pointless.
	 */
	if (!total_swap_pages)
		return 0;

1556
	if (!mem_cgroup_disabled())
1557 1558 1559 1560
		return mem_cgroup_inactive_anon_is_low(mz->mem_cgroup,
						       mz->zone);

	return inactive_anon_is_low_global(mz->zone);
1561
}
1562
#else
1563
static inline int inactive_anon_is_low(struct mem_cgroup_zone *mz)
1564 1565 1566 1567
{
	return 0;
}
#endif
1568

1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580
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
1581
 * @mz: memory cgroup and zone to check
1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592
 *
 * 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.
 */
1593
static int inactive_file_is_low(struct mem_cgroup_zone *mz)
1594
{
1595
	if (!mem_cgroup_disabled())
1596 1597
		return mem_cgroup_inactive_file_is_low(mz->mem_cgroup,
						       mz->zone);
1598

1599
	return inactive_file_is_low_global(mz->zone);
1600 1601
}

1602
static int inactive_list_is_low(struct mem_cgroup_zone *mz, int file)
1603 1604
{
	if (file)
1605
		return inactive_file_is_low(mz);
1606
	else
1607
		return inactive_anon_is_low(mz);
1608 1609
}

1610
static unsigned long shrink_list(enum lru_list lru, unsigned long nr_to_scan,
1611 1612
				 struct mem_cgroup_zone *mz,
				 struct scan_control *sc, int priority)
1613
{
1614 1615
	int file = is_file_lru(lru);

1616
	if (is_active_lru(lru)) {
1617
		if (inactive_list_is_low(mz, file))
1618
			shrink_active_list(nr_to_scan, mz, sc, priority, lru);
1619 1620 1621
		return 0;
	}

1622
	return shrink_inactive_list(nr_to_scan, mz, sc, priority, lru);
1623 1624
}

1625 1626
static int vmscan_swappiness(struct mem_cgroup_zone *mz,
			     struct scan_control *sc)
1627
{
1628
	if (global_reclaim(sc))
1629
		return vm_swappiness;
1630
	return mem_cgroup_swappiness(mz->mem_cgroup);
1631 1632
}

1633 1634 1635 1636 1637 1638
/*
 * 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.
 *
1639
 * nr[0] = anon pages to scan; nr[1] = file pages to scan
1640
 */
1641 1642
static void get_scan_count(struct mem_cgroup_zone *mz, struct scan_control *sc,
			   unsigned long *nr, int priority)
1643 1644 1645 1646
{
	unsigned long anon, file, free;
	unsigned long anon_prio, file_prio;
	unsigned long ap, fp;
1647
	struct zone_reclaim_stat *reclaim_stat = get_reclaim_stat(mz);
1648
	u64 fraction[2], denominator;
H
Hugh Dickins 已提交
1649
	enum lru_list lru;
1650
	int noswap = 0;
1651
	bool force_scan = false;
1652

1653 1654 1655 1656 1657 1658 1659 1660 1661 1662
	/*
	 * If the zone or memcg is small, nr[l] can be 0.  This
	 * results in no scanning on this priority and a potential
	 * priority drop.  Global direct reclaim can go to the next
	 * zone and tends to have no problems. Global kswapd is for
	 * zone balancing and it needs to scan a minimum amount. When
	 * reclaiming for a memcg, a priority drop can cause high
	 * latencies, so it's better to scan a minimum amount there as
	 * well.
	 */
1663
	if (current_is_kswapd() && mz->zone->all_unreclaimable)
1664
		force_scan = true;
1665
	if (!global_reclaim(sc))
1666
		force_scan = true;
1667 1668 1669 1670 1671 1672 1673 1674 1675

	/* 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;
	}
1676

1677 1678 1679 1680
	anon  = zone_nr_lru_pages(mz, LRU_ACTIVE_ANON) +
		zone_nr_lru_pages(mz, LRU_INACTIVE_ANON);
	file  = zone_nr_lru_pages(mz, LRU_ACTIVE_FILE) +
		zone_nr_lru_pages(mz, LRU_INACTIVE_FILE);
1681

1682
	if (global_reclaim(sc)) {
1683
		free  = zone_page_state(mz->zone, NR_FREE_PAGES);
1684 1685
		/* If we have very few page cache pages,
		   force-scan anon pages. */
1686
		if (unlikely(file + free <= high_wmark_pages(mz->zone))) {
1687 1688 1689 1690
			fraction[0] = 1;
			fraction[1] = 0;
			denominator = 1;
			goto out;
1691
		}
1692 1693
	}

1694 1695 1696 1697
	/*
	 * With swappiness at 100, anonymous and file have the same priority.
	 * This scanning priority is essentially the inverse of IO cost.
	 */
1698 1699
	anon_prio = vmscan_swappiness(mz, sc);
	file_prio = 200 - vmscan_swappiness(mz, sc);
1700

1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711
	/*
	 * 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]
	 */
1712
	spin_lock_irq(&mz->zone->lru_lock);
1713 1714 1715
	if (unlikely(reclaim_stat->recent_scanned[0] > anon / 4)) {
		reclaim_stat->recent_scanned[0] /= 2;
		reclaim_stat->recent_rotated[0] /= 2;
1716 1717
	}

1718 1719 1720
	if (unlikely(reclaim_stat->recent_scanned[1] > file / 4)) {
		reclaim_stat->recent_scanned[1] /= 2;
		reclaim_stat->recent_rotated[1] /= 2;
1721 1722 1723
	}

	/*
1724 1725 1726
	 * 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.
1727
	 */
1728
	ap = anon_prio * (reclaim_stat->recent_scanned[0] + 1);
1729
	ap /= reclaim_stat->recent_rotated[0] + 1;
1730

1731
	fp = file_prio * (reclaim_stat->recent_scanned[1] + 1);
1732
	fp /= reclaim_stat->recent_rotated[1] + 1;
1733
	spin_unlock_irq(&mz->zone->lru_lock);
1734

1735 1736 1737 1738
	fraction[0] = ap;
	fraction[1] = fp;
	denominator = ap + fp + 1;
out:
H
Hugh Dickins 已提交
1739 1740
	for_each_evictable_lru(lru) {
		int file = is_file_lru(lru);
1741
		unsigned long scan;
1742

H
Hugh Dickins 已提交
1743
		scan = zone_nr_lru_pages(mz, lru);
1744
		if (priority || noswap || !vmscan_swappiness(mz, sc)) {
1745
			scan >>= priority;
1746 1747
			if (!scan && force_scan)
				scan = SWAP_CLUSTER_MAX;
1748 1749
			scan = div64_u64(scan * fraction[file], denominator);
		}
H
Hugh Dickins 已提交
1750
		nr[lru] = scan;
1751
	}
1752
}
1753

M
Mel Gorman 已提交
1754 1755 1756 1757 1758 1759 1760 1761 1762 1763 1764
/* Use reclaim/compaction for costly allocs or under memory pressure */
static bool in_reclaim_compaction(int priority, struct scan_control *sc)
{
	if (COMPACTION_BUILD && sc->order &&
			(sc->order > PAGE_ALLOC_COSTLY_ORDER ||
			 priority < DEF_PRIORITY - 2))
		return true;

	return false;
}

1765
/*
M
Mel Gorman 已提交
1766 1767 1768 1769 1770
 * Reclaim/compaction is used for high-order allocation requests. It reclaims
 * order-0 pages before compacting the zone. should_continue_reclaim() returns
 * true if more pages should be reclaimed such that when the page allocator
 * calls try_to_compact_zone() that it will have enough free pages to succeed.
 * It will give up earlier than that if there is difficulty reclaiming pages.
1771
 */
1772
static inline bool should_continue_reclaim(struct mem_cgroup_zone *mz,
1773 1774
					unsigned long nr_reclaimed,
					unsigned long nr_scanned,
M
Mel Gorman 已提交
1775
					int priority,
1776 1777 1778 1779 1780 1781
					struct scan_control *sc)
{
	unsigned long pages_for_compaction;
	unsigned long inactive_lru_pages;

	/* If not in reclaim/compaction mode, stop */
M
Mel Gorman 已提交
1782
	if (!in_reclaim_compaction(priority, sc))
1783 1784
		return false;

1785 1786 1787 1788 1789 1790 1791 1792 1793 1794 1795 1796 1797 1798 1799 1800 1801 1802 1803 1804 1805 1806
	/* 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;
	}
1807 1808 1809 1810 1811 1812

	/*
	 * If we have not reclaimed enough pages for compaction and the
	 * inactive lists are large enough, continue reclaiming
	 */
	pages_for_compaction = (2UL << sc->order);
1813
	inactive_lru_pages = zone_nr_lru_pages(mz, LRU_INACTIVE_FILE);
1814
	if (nr_swap_pages > 0)
1815
		inactive_lru_pages += zone_nr_lru_pages(mz, LRU_INACTIVE_ANON);
1816 1817 1818 1819 1820
	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 */
1821
	switch (compaction_suitable(mz->zone, sc->order)) {
1822 1823 1824 1825 1826 1827 1828 1829
	case COMPACT_PARTIAL:
	case COMPACT_CONTINUE:
		return false;
	default:
		return true;
	}
}

L
Linus Torvalds 已提交
1830 1831 1832
/*
 * This is a basic per-zone page freer.  Used by both kswapd and direct reclaim.
 */
1833 1834
static void shrink_mem_cgroup_zone(int priority, struct mem_cgroup_zone *mz,
				   struct scan_control *sc)
L
Linus Torvalds 已提交
1835
{
1836
	unsigned long nr[NR_LRU_LISTS];
1837
	unsigned long nr_to_scan;
H
Hugh Dickins 已提交
1838
	enum lru_list lru;
1839
	unsigned long nr_reclaimed, nr_scanned;
1840
	unsigned long nr_to_reclaim = sc->nr_to_reclaim;
1841
	struct blk_plug plug;
1842

1843 1844
restart:
	nr_reclaimed = 0;
1845
	nr_scanned = sc->nr_scanned;
1846
	get_scan_count(mz, sc, nr, priority);
L
Linus Torvalds 已提交
1847

1848
	blk_start_plug(&plug);
1849 1850
	while (nr[LRU_INACTIVE_ANON] || nr[LRU_ACTIVE_FILE] ||
					nr[LRU_INACTIVE_FILE]) {
H
Hugh Dickins 已提交
1851 1852
		for_each_evictable_lru(lru) {
			if (nr[lru]) {
K
KOSAKI Motohiro 已提交
1853
				nr_to_scan = min_t(unsigned long,
H
Hugh Dickins 已提交
1854 1855
						   nr[lru], SWAP_CLUSTER_MAX);
				nr[lru] -= nr_to_scan;
L
Linus Torvalds 已提交
1856

H
Hugh Dickins 已提交
1857
				nr_reclaimed += shrink_list(lru, nr_to_scan,
1858
							    mz, sc, priority);
1859
			}
L
Linus Torvalds 已提交
1860
		}
1861 1862 1863 1864 1865 1866 1867 1868
		/*
		 * 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.
		 */
1869
		if (nr_reclaimed >= nr_to_reclaim && priority < DEF_PRIORITY)
1870
			break;
L
Linus Torvalds 已提交
1871
	}
1872
	blk_finish_plug(&plug);
1873
	sc->nr_reclaimed += nr_reclaimed;
1874

1875 1876 1877 1878
	/*
	 * Even if we did not try to evict anon pages at all, we want to
	 * rebalance the anon lru active/inactive ratio.
	 */
1879
	if (inactive_anon_is_low(mz))
1880 1881
		shrink_active_list(SWAP_CLUSTER_MAX, mz,
				   sc, priority, LRU_ACTIVE_ANON);
1882

1883
	/* reclaim/compaction might need reclaim to continue */
1884
	if (should_continue_reclaim(mz, nr_reclaimed,
M
Mel Gorman 已提交
1885 1886
					sc->nr_scanned - nr_scanned,
					priority, sc))
1887 1888
		goto restart;

1889
	throttle_vm_writeout(sc->gfp_mask);
L
Linus Torvalds 已提交
1890 1891
}

1892 1893 1894
static void shrink_zone(int priority, struct zone *zone,
			struct scan_control *sc)
{
1895 1896
	struct mem_cgroup *root = sc->target_mem_cgroup;
	struct mem_cgroup_reclaim_cookie reclaim = {
1897
		.zone = zone,
1898
		.priority = priority,
1899
	};
1900 1901 1902 1903 1904 1905 1906 1907
	struct mem_cgroup *memcg;

	memcg = mem_cgroup_iter(root, NULL, &reclaim);
	do {
		struct mem_cgroup_zone mz = {
			.mem_cgroup = memcg,
			.zone = zone,
		};
1908

1909 1910 1911 1912 1913 1914
		shrink_mem_cgroup_zone(priority, &mz, sc);
		/*
		 * Limit reclaim has historically picked one memcg and
		 * scanned it with decreasing priority levels until
		 * nr_to_reclaim had been reclaimed.  This priority
		 * cycle is thus over after a single memcg.
1915 1916 1917 1918
		 *
		 * Direct reclaim and kswapd, on the other hand, have
		 * to scan all memory cgroups to fulfill the overall
		 * scan target for the zone.
1919 1920 1921 1922 1923 1924 1925
		 */
		if (!global_reclaim(sc)) {
			mem_cgroup_iter_break(root, memcg);
			break;
		}
		memcg = mem_cgroup_iter(root, memcg, &reclaim);
	} while (memcg);
1926 1927
}

1928 1929 1930 1931 1932 1933 1934 1935 1936 1937 1938 1939 1940 1941 1942 1943 1944 1945 1946 1947 1948 1949 1950 1951 1952 1953
/* Returns true if compaction should go ahead for a high-order request */
static inline bool compaction_ready(struct zone *zone, struct scan_control *sc)
{
	unsigned long balance_gap, watermark;
	bool watermark_ok;

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

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

	/*
	 * If compaction is deferred, reclaim up to a point where
	 * compaction will have a chance of success when re-enabled
	 */
1954
	if (compaction_deferred(zone, sc->order))
1955 1956 1957 1958 1959 1960 1961 1962 1963
		return watermark_ok;

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

	return watermark_ok;
}

L
Linus Torvalds 已提交
1964 1965 1966 1967 1968
/*
 * 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.
 *
1969 1970
 * We reclaim from a zone even if that zone is over high_wmark_pages(zone).
 * Because:
L
Linus Torvalds 已提交
1971 1972
 * a) The caller may be trying to free *extra* pages to satisfy a higher-order
 *    allocation or
1973 1974 1975
 * 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
1976 1977 1978
 *
 * If a zone is deemed to be full of pinned pages then just give it a light
 * scan then give up on it.
1979 1980
 *
 * This function returns true if a zone is being reclaimed for a costly
1981
 * high-order allocation and compaction is ready to begin. This indicates to
1982 1983
 * the caller that it should consider retrying the allocation instead of
 * further reclaim.
L
Linus Torvalds 已提交
1984
 */
1985
static bool shrink_zones(int priority, struct zonelist *zonelist,
1986
					struct scan_control *sc)
L
Linus Torvalds 已提交
1987
{
1988
	struct zoneref *z;
1989
	struct zone *zone;
1990 1991
	unsigned long nr_soft_reclaimed;
	unsigned long nr_soft_scanned;
1992
	bool aborted_reclaim = false;
1993

1994 1995 1996 1997 1998 1999 2000 2001
	/*
	 * If the number of buffer_heads in the machine exceeds the maximum
	 * allowed level, force direct reclaim to scan the highmem zone as
	 * highmem pages could be pinning lowmem pages storing buffer_heads
	 */
	if (buffer_heads_over_limit)
		sc->gfp_mask |= __GFP_HIGHMEM;

2002 2003
	for_each_zone_zonelist_nodemask(zone, z, zonelist,
					gfp_zone(sc->gfp_mask), sc->nodemask) {
2004
		if (!populated_zone(zone))
L
Linus Torvalds 已提交
2005
			continue;
2006 2007 2008 2009
		/*
		 * Take care memory controller reclaiming has small influence
		 * to global LRU.
		 */
2010
		if (global_reclaim(sc)) {
2011 2012
			if (!cpuset_zone_allowed_hardwall(zone, GFP_KERNEL))
				continue;
2013
			if (zone->all_unreclaimable && priority != DEF_PRIORITY)
2014
				continue;	/* Let kswapd poll it */
2015 2016
			if (COMPACTION_BUILD) {
				/*
2017 2018 2019 2020 2021
				 * If we already have plenty of memory free for
				 * compaction in this zone, don't free any more.
				 * Even though compaction is invoked for any
				 * non-zero order, only frequent costly order
				 * reclamation is disruptive enough to become a
2022 2023
				 * noticeable problem, like transparent huge
				 * page allocations.
2024
				 */
2025
				if (compaction_ready(zone, sc)) {
2026
					aborted_reclaim = true;
2027
					continue;
2028
				}
2029
			}
2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042
			/*
			 * 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() */
2043
		}
2044

2045
		shrink_zone(priority, zone, sc);
L
Linus Torvalds 已提交
2046
	}
2047

2048
	return aborted_reclaim;
2049 2050 2051 2052 2053 2054 2055
}

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

2056
/* All zones in zonelist are unreclaimable? */
2057 2058 2059 2060 2061 2062 2063 2064 2065 2066 2067 2068
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;
2069 2070
		if (!zone->all_unreclaimable)
			return false;
2071 2072
	}

2073
	return true;
L
Linus Torvalds 已提交
2074
}
2075

L
Linus Torvalds 已提交
2076 2077 2078 2079 2080 2081 2082 2083
/*
 * 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
2084 2085 2086 2087
 * 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.
2088 2089 2090
 *
 * returns:	0, if no pages reclaimed
 * 		else, the number of pages reclaimed
L
Linus Torvalds 已提交
2091
 */
2092
static unsigned long do_try_to_free_pages(struct zonelist *zonelist,
2093 2094
					struct scan_control *sc,
					struct shrink_control *shrink)
L
Linus Torvalds 已提交
2095 2096
{
	int priority;
2097
	unsigned long total_scanned = 0;
L
Linus Torvalds 已提交
2098
	struct reclaim_state *reclaim_state = current->reclaim_state;
2099
	struct zoneref *z;
2100
	struct zone *zone;
2101
	unsigned long writeback_threshold;
2102
	bool aborted_reclaim;
L
Linus Torvalds 已提交
2103

2104 2105
	delayacct_freepages_start();

2106
	if (global_reclaim(sc))
2107
		count_vm_event(ALLOCSTALL);
L
Linus Torvalds 已提交
2108 2109

	for (priority = DEF_PRIORITY; priority >= 0; priority--) {
2110
		sc->nr_scanned = 0;
2111
		aborted_reclaim = shrink_zones(priority, zonelist, sc);
2112

2113 2114 2115 2116
		/*
		 * Don't shrink slabs when reclaiming memory from
		 * over limit cgroups
		 */
2117
		if (global_reclaim(sc)) {
2118
			unsigned long lru_pages = 0;
2119 2120
			for_each_zone_zonelist(zone, z, zonelist,
					gfp_zone(sc->gfp_mask)) {
2121 2122 2123 2124 2125 2126
				if (!cpuset_zone_allowed_hardwall(zone, GFP_KERNEL))
					continue;

				lru_pages += zone_reclaimable_pages(zone);
			}

2127
			shrink_slab(shrink, sc->nr_scanned, lru_pages);
2128
			if (reclaim_state) {
2129
				sc->nr_reclaimed += reclaim_state->reclaimed_slab;
2130 2131
				reclaim_state->reclaimed_slab = 0;
			}
L
Linus Torvalds 已提交
2132
		}
2133
		total_scanned += sc->nr_scanned;
2134
		if (sc->nr_reclaimed >= sc->nr_to_reclaim)
L
Linus Torvalds 已提交
2135 2136 2137 2138 2139 2140 2141 2142 2143
			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.
		 */
2144 2145
		writeback_threshold = sc->nr_to_reclaim + sc->nr_to_reclaim / 2;
		if (total_scanned > writeback_threshold) {
2146 2147
			wakeup_flusher_threads(laptop_mode ? 0 : total_scanned,
						WB_REASON_TRY_TO_FREE_PAGES);
2148
			sc->may_writepage = 1;
L
Linus Torvalds 已提交
2149 2150 2151
		}

		/* Take a nap, wait for some writeback to complete */
2152
		if (!sc->hibernation_mode && sc->nr_scanned &&
2153 2154 2155 2156
		    priority < DEF_PRIORITY - 2) {
			struct zone *preferred_zone;

			first_zones_zonelist(zonelist, gfp_zone(sc->gfp_mask),
2157 2158
						&cpuset_current_mems_allowed,
						&preferred_zone);
2159 2160
			wait_iff_congested(preferred_zone, BLK_RW_ASYNC, HZ/10);
		}
L
Linus Torvalds 已提交
2161
	}
2162

L
Linus Torvalds 已提交
2163
out:
2164 2165
	delayacct_freepages_end();

2166 2167 2168
	if (sc->nr_reclaimed)
		return sc->nr_reclaimed;

2169 2170 2171 2172 2173 2174 2175 2176
	/*
	 * 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;

2177 2178
	/* Aborted reclaim to try compaction? don't OOM, then */
	if (aborted_reclaim)
2179 2180
		return 1;

2181
	/* top priority shrink_zones still had more to do? don't OOM, then */
2182
	if (global_reclaim(sc) && !all_unreclaimable(zonelist, sc))
2183 2184 2185
		return 1;

	return 0;
L
Linus Torvalds 已提交
2186 2187
}

2188
unsigned long try_to_free_pages(struct zonelist *zonelist, int order,
2189
				gfp_t gfp_mask, nodemask_t *nodemask)
2190
{
2191
	unsigned long nr_reclaimed;
2192 2193 2194
	struct scan_control sc = {
		.gfp_mask = gfp_mask,
		.may_writepage = !laptop_mode,
2195
		.nr_to_reclaim = SWAP_CLUSTER_MAX,
2196
		.may_unmap = 1,
2197
		.may_swap = 1,
2198
		.order = order,
2199
		.target_mem_cgroup = NULL,
2200
		.nodemask = nodemask,
2201
	};
2202 2203 2204
	struct shrink_control shrink = {
		.gfp_mask = sc.gfp_mask,
	};
2205

2206 2207 2208 2209
	trace_mm_vmscan_direct_reclaim_begin(order,
				sc.may_writepage,
				gfp_mask);

2210
	nr_reclaimed = do_try_to_free_pages(zonelist, &sc, &shrink);
2211 2212 2213 2214

	trace_mm_vmscan_direct_reclaim_end(nr_reclaimed);

	return nr_reclaimed;
2215 2216
}

2217
#ifdef CONFIG_CGROUP_MEM_RES_CTLR
2218

2219
unsigned long mem_cgroup_shrink_node_zone(struct mem_cgroup *memcg,
2220
						gfp_t gfp_mask, bool noswap,
2221 2222
						struct zone *zone,
						unsigned long *nr_scanned)
2223 2224
{
	struct scan_control sc = {
2225
		.nr_scanned = 0,
2226
		.nr_to_reclaim = SWAP_CLUSTER_MAX,
2227 2228 2229 2230
		.may_writepage = !laptop_mode,
		.may_unmap = 1,
		.may_swap = !noswap,
		.order = 0,
2231
		.target_mem_cgroup = memcg,
2232
	};
2233
	struct mem_cgroup_zone mz = {
2234
		.mem_cgroup = memcg,
2235 2236
		.zone = zone,
	};
2237

2238 2239
	sc.gfp_mask = (gfp_mask & GFP_RECLAIM_MASK) |
			(GFP_HIGHUSER_MOVABLE & ~GFP_RECLAIM_MASK);
2240 2241 2242 2243 2244

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

2245 2246 2247 2248 2249 2250 2251
	/*
	 * 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.
	 */
2252
	shrink_mem_cgroup_zone(0, &mz, &sc);
2253 2254 2255

	trace_mm_vmscan_memcg_softlimit_reclaim_end(sc.nr_reclaimed);

2256
	*nr_scanned = sc.nr_scanned;
2257 2258 2259
	return sc.nr_reclaimed;
}

2260
unsigned long try_to_free_mem_cgroup_pages(struct mem_cgroup *memcg,
K
KOSAKI Motohiro 已提交
2261
					   gfp_t gfp_mask,
2262
					   bool noswap)
2263
{
2264
	struct zonelist *zonelist;
2265
	unsigned long nr_reclaimed;
2266
	int nid;
2267 2268
	struct scan_control sc = {
		.may_writepage = !laptop_mode,
2269
		.may_unmap = 1,
2270
		.may_swap = !noswap,
2271
		.nr_to_reclaim = SWAP_CLUSTER_MAX,
2272
		.order = 0,
2273
		.target_mem_cgroup = memcg,
2274
		.nodemask = NULL, /* we don't care the placement */
2275 2276 2277 2278 2279
		.gfp_mask = (gfp_mask & GFP_RECLAIM_MASK) |
				(GFP_HIGHUSER_MOVABLE & ~GFP_RECLAIM_MASK),
	};
	struct shrink_control shrink = {
		.gfp_mask = sc.gfp_mask,
2280 2281
	};

2282 2283 2284 2285 2286
	/*
	 * 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.
	 */
2287
	nid = mem_cgroup_select_victim_node(memcg);
2288 2289

	zonelist = NODE_DATA(nid)->node_zonelists;
2290 2291 2292 2293 2294

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

2295
	nr_reclaimed = do_try_to_free_pages(zonelist, &sc, &shrink);
2296 2297 2298 2299

	trace_mm_vmscan_memcg_reclaim_end(nr_reclaimed);

	return nr_reclaimed;
2300 2301 2302
}
#endif

2303 2304 2305
static void age_active_anon(struct zone *zone, struct scan_control *sc,
			    int priority)
{
2306
	struct mem_cgroup *memcg;
2307

2308 2309 2310 2311 2312 2313 2314 2315 2316 2317 2318 2319
	if (!total_swap_pages)
		return;

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

		if (inactive_anon_is_low(&mz))
			shrink_active_list(SWAP_CLUSTER_MAX, &mz,
2320
					   sc, priority, LRU_ACTIVE_ANON);
2321 2322 2323

		memcg = mem_cgroup_iter(NULL, memcg, NULL);
	} while (memcg);
2324 2325
}

2326 2327 2328 2329 2330 2331 2332 2333 2334 2335 2336
/*
 * 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
2337
 *     percentage of the middle zones. For example, on 32-bit x86, highmem
2338 2339 2340 2341 2342 2343 2344 2345 2346 2347 2348 2349 2350
 *     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
2351 2352
	/* A special case here: if zone has no page, we think it's balanced */
	return balanced_pages >= (present_pages >> 2);
2353 2354
}

2355
/* is kswapd sleeping prematurely? */
2356 2357
static bool sleeping_prematurely(pg_data_t *pgdat, int order, long remaining,
					int classzone_idx)
2358
{
2359
	int i;
2360 2361
	unsigned long balanced = 0;
	bool all_zones_ok = true;
2362 2363 2364

	/* If a direct reclaimer woke kswapd within HZ/10, it's premature */
	if (remaining)
2365
		return true;
2366

2367
	/* Check the watermark levels */
2368
	for (i = 0; i <= classzone_idx; i++) {
2369 2370 2371 2372 2373
		struct zone *zone = pgdat->node_zones + i;

		if (!populated_zone(zone))
			continue;

2374 2375 2376 2377 2378 2379 2380 2381
		/*
		 * 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;
2382
			continue;
2383
		}
2384

2385
		if (!zone_watermark_ok_safe(zone, order, high_wmark_pages(zone),
2386
							i, 0))
2387 2388 2389
			all_zones_ok = false;
		else
			balanced += zone->present_pages;
2390
	}
2391

2392 2393 2394 2395 2396 2397
	/*
	 * 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)
2398
		return !pgdat_balanced(pgdat, balanced, classzone_idx);
2399 2400
	else
		return !all_zones_ok;
2401 2402
}

L
Linus Torvalds 已提交
2403 2404
/*
 * For kswapd, balance_pgdat() will work across all this node's zones until
2405
 * they are all at high_wmark_pages(zone).
L
Linus Torvalds 已提交
2406
 *
2407
 * Returns the final order kswapd was reclaiming at
L
Linus Torvalds 已提交
2408 2409 2410 2411 2412 2413 2414 2415 2416 2417
 *
 * 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
2418 2419 2420 2421 2422
 * 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
2423
 */
2424
static unsigned long balance_pgdat(pg_data_t *pgdat, int order,
2425
							int *classzone_idx)
L
Linus Torvalds 已提交
2426 2427
{
	int all_zones_ok;
2428
	unsigned long balanced;
L
Linus Torvalds 已提交
2429 2430
	int priority;
	int i;
2431
	int end_zone = 0;	/* Inclusive.  0 = ZONE_DMA */
2432
	unsigned long total_scanned;
L
Linus Torvalds 已提交
2433
	struct reclaim_state *reclaim_state = current->reclaim_state;
2434 2435
	unsigned long nr_soft_reclaimed;
	unsigned long nr_soft_scanned;
2436 2437
	struct scan_control sc = {
		.gfp_mask = GFP_KERNEL,
2438
		.may_unmap = 1,
2439
		.may_swap = 1,
2440 2441 2442 2443 2444
		/*
		 * kswapd doesn't want to be bailed out while reclaim. because
		 * we want to put equal scanning pressure on each zone.
		 */
		.nr_to_reclaim = ULONG_MAX,
A
Andy Whitcroft 已提交
2445
		.order = order,
2446
		.target_mem_cgroup = NULL,
2447
	};
2448 2449 2450
	struct shrink_control shrink = {
		.gfp_mask = sc.gfp_mask,
	};
L
Linus Torvalds 已提交
2451 2452
loop_again:
	total_scanned = 0;
2453
	sc.nr_reclaimed = 0;
C
Christoph Lameter 已提交
2454
	sc.may_writepage = !laptop_mode;
2455
	count_vm_event(PAGEOUTRUN);
L
Linus Torvalds 已提交
2456 2457 2458

	for (priority = DEF_PRIORITY; priority >= 0; priority--) {
		unsigned long lru_pages = 0;
2459
		int has_under_min_watermark_zone = 0;
L
Linus Torvalds 已提交
2460 2461

		all_zones_ok = 1;
2462
		balanced = 0;
L
Linus Torvalds 已提交
2463

2464 2465 2466 2467 2468 2469
		/*
		 * 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
2470

2471 2472
			if (!populated_zone(zone))
				continue;
L
Linus Torvalds 已提交
2473

2474
			if (zone->all_unreclaimable && priority != DEF_PRIORITY)
2475
				continue;
L
Linus Torvalds 已提交
2476

2477 2478 2479 2480
			/*
			 * Do some background aging of the anon list, to give
			 * pages a chance to be referenced before reclaiming.
			 */
2481
			age_active_anon(zone, &sc, priority);
2482

2483 2484 2485 2486 2487 2488 2489 2490 2491 2492 2493
			/*
			 * If the number of buffer_heads in the machine
			 * exceeds the maximum allowed level and this node
			 * has a highmem zone, force kswapd to reclaim from
			 * it to relieve lowmem pressure.
			 */
			if (buffer_heads_over_limit && is_highmem_idx(i)) {
				end_zone = i;
				break;
			}

2494
			if (!zone_watermark_ok_safe(zone, order,
2495
					high_wmark_pages(zone), 0, 0)) {
2496
				end_zone = i;
A
Andrew Morton 已提交
2497
				break;
2498 2499 2500
			} else {
				/* If balanced, clear the congested flag */
				zone_clear_flag(zone, ZONE_CONGESTED);
L
Linus Torvalds 已提交
2501 2502
			}
		}
A
Andrew Morton 已提交
2503 2504 2505
		if (i < 0)
			goto out;

L
Linus Torvalds 已提交
2506 2507 2508
		for (i = 0; i <= end_zone; i++) {
			struct zone *zone = pgdat->node_zones + i;

2509
			lru_pages += zone_reclaimable_pages(zone);
L
Linus Torvalds 已提交
2510 2511 2512 2513 2514 2515 2516 2517 2518 2519 2520 2521 2522
		}

		/*
		 * 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;
2523
			int nr_slab, testorder;
2524
			unsigned long balance_gap;
L
Linus Torvalds 已提交
2525

2526
			if (!populated_zone(zone))
L
Linus Torvalds 已提交
2527 2528
				continue;

2529
			if (zone->all_unreclaimable && priority != DEF_PRIORITY)
L
Linus Torvalds 已提交
2530 2531 2532
				continue;

			sc.nr_scanned = 0;
2533

2534
			nr_soft_scanned = 0;
2535 2536 2537
			/*
			 * Call soft limit reclaim before calling shrink_zone.
			 */
2538 2539 2540 2541 2542
			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;
2543

2544
			/*
2545 2546 2547 2548 2549 2550
			 * 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.
2551
			 */
2552 2553 2554 2555
			balance_gap = min(low_wmark_pages(zone),
				(zone->present_pages +
					KSWAPD_ZONE_BALANCE_GAP_RATIO-1) /
				KSWAPD_ZONE_BALANCE_GAP_RATIO);
2556 2557 2558 2559 2560 2561 2562 2563 2564 2565 2566 2567 2568
			/*
			 * Kswapd reclaims only single pages with compaction
			 * enabled. Trying too hard to reclaim until contiguous
			 * free pages have become available can hurt performance
			 * by evicting too much useful data from memory.
			 * Do not reclaim more than needed for compaction.
			 */
			testorder = order;
			if (COMPACTION_BUILD && order &&
					compaction_suitable(zone, order) !=
						COMPACT_SKIPPED)
				testorder = 0;

2569
			if ((buffer_heads_over_limit && is_highmem_idx(i)) ||
2570
				    !zone_watermark_ok_safe(zone, testorder,
2571
					high_wmark_pages(zone) + balance_gap,
2572
					end_zone, 0)) {
2573
				shrink_zone(priority, zone, &sc);
2574

2575 2576 2577 2578 2579 2580 2581 2582 2583
				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
2584 2585 2586 2587 2588 2589
			/*
			 * 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 &&
2590
			    total_scanned > sc.nr_reclaimed + sc.nr_reclaimed / 2)
L
Linus Torvalds 已提交
2591
				sc.may_writepage = 1;
2592

2593 2594 2595
			if (zone->all_unreclaimable) {
				if (end_zone && end_zone == i)
					end_zone--;
2596
				continue;
2597
			}
2598

2599
			if (!zone_watermark_ok_safe(zone, testorder,
2600 2601 2602 2603 2604 2605 2606
					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!
				 */
2607
				if (!zone_watermark_ok_safe(zone, order,
2608 2609
					    min_wmark_pages(zone), end_zone, 0))
					has_under_min_watermark_zone = 1;
2610 2611 2612 2613 2614 2615 2616 2617 2618
			} 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);
2619
				if (i <= *classzone_idx)
2620
					balanced += zone->present_pages;
2621
			}
2622

L
Linus Torvalds 已提交
2623
		}
2624
		if (all_zones_ok || (order && pgdat_balanced(pgdat, balanced, *classzone_idx)))
L
Linus Torvalds 已提交
2625 2626 2627 2628 2629
			break;		/* kswapd: all done */
		/*
		 * OK, kswapd is getting into trouble.  Take a nap, then take
		 * another pass across the zones.
		 */
2630 2631 2632 2633 2634 2635
		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
2636 2637 2638 2639 2640 2641 2642

		/*
		 * 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.
		 */
2643
		if (sc.nr_reclaimed >= SWAP_CLUSTER_MAX)
L
Linus Torvalds 已提交
2644 2645 2646
			break;
	}
out:
2647 2648 2649

	/*
	 * order-0: All zones must meet high watermark for a balanced node
2650 2651
	 * high-order: Balanced zones must make up at least 25% of the node
	 *             for the node to be balanced
2652
	 */
2653
	if (!(all_zones_ok || (order && pgdat_balanced(pgdat, balanced, *classzone_idx)))) {
L
Linus Torvalds 已提交
2654
		cond_resched();
2655 2656 2657

		try_to_freeze();

2658 2659 2660 2661 2662 2663 2664 2665 2666 2667 2668 2669 2670 2671 2672 2673 2674
		/*
		 * 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
2675 2676 2677
		goto loop_again;
	}

2678 2679 2680 2681 2682 2683 2684 2685 2686
	/*
	 * 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) {
2687 2688
		int zones_need_compaction = 1;

2689 2690 2691 2692 2693 2694 2695 2696 2697
		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;

2698
			/* Would compaction fail due to lack of free memory? */
2699 2700
			if (COMPACTION_BUILD &&
			    compaction_suitable(zone, order) == COMPACT_SKIPPED)
2701 2702
				goto loop_again;

2703 2704 2705 2706 2707 2708 2709
			/* 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;
			}

2710 2711 2712 2713 2714
			/* Check if the memory needs to be defragmented. */
			if (zone_watermark_ok(zone, order,
				    low_wmark_pages(zone), *classzone_idx, 0))
				zones_need_compaction = 0;

2715 2716 2717
			/* If balanced, clear the congested flag */
			zone_clear_flag(zone, ZONE_CONGESTED);
		}
2718 2719 2720

		if (zones_need_compaction)
			compact_pgdat(pgdat, order);
2721 2722
	}

2723 2724 2725 2726 2727 2728
	/*
	 * 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
	 */
2729
	*classzone_idx = end_zone;
2730
	return order;
L
Linus Torvalds 已提交
2731 2732
}

2733
static void kswapd_try_to_sleep(pg_data_t *pgdat, int order, int classzone_idx)
2734 2735 2736 2737 2738 2739 2740 2741 2742 2743
{
	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 */
2744
	if (!sleeping_prematurely(pgdat, order, remaining, classzone_idx)) {
2745 2746 2747 2748 2749 2750 2751 2752 2753
		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.
	 */
2754
	if (!sleeping_prematurely(pgdat, order, remaining, classzone_idx)) {
2755 2756 2757 2758 2759 2760 2761 2762 2763 2764 2765 2766 2767 2768 2769 2770 2771 2772 2773 2774 2775 2776
		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
2777 2778
/*
 * The background pageout daemon, started as a kernel thread
2779
 * from the init process.
L
Linus Torvalds 已提交
2780 2781 2782 2783 2784 2785 2786 2787 2788 2789 2790 2791
 *
 * 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)
{
2792
	unsigned long order, new_order;
2793
	unsigned balanced_order;
2794
	int classzone_idx, new_classzone_idx;
2795
	int balanced_classzone_idx;
L
Linus Torvalds 已提交
2796 2797
	pg_data_t *pgdat = (pg_data_t*)p;
	struct task_struct *tsk = current;
2798

L
Linus Torvalds 已提交
2799 2800 2801
	struct reclaim_state reclaim_state = {
		.reclaimed_slab = 0,
	};
2802
	const struct cpumask *cpumask = cpumask_of_node(pgdat->node_id);
L
Linus Torvalds 已提交
2803

2804 2805
	lockdep_set_current_reclaim_state(GFP_KERNEL);

R
Rusty Russell 已提交
2806
	if (!cpumask_empty(cpumask))
2807
		set_cpus_allowed_ptr(tsk, cpumask);
L
Linus Torvalds 已提交
2808 2809 2810 2811 2812 2813 2814 2815 2816 2817 2818 2819 2820 2821
	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).
	 */
2822
	tsk->flags |= PF_MEMALLOC | PF_SWAPWRITE | PF_KSWAPD;
2823
	set_freezable();
L
Linus Torvalds 已提交
2824

2825
	order = new_order = 0;
2826
	balanced_order = 0;
2827
	classzone_idx = new_classzone_idx = pgdat->nr_zones - 1;
2828
	balanced_classzone_idx = classzone_idx;
L
Linus Torvalds 已提交
2829
	for ( ; ; ) {
2830
		int ret;
2831

2832 2833 2834 2835 2836
		/*
		 * 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
		 */
2837 2838
		if (balanced_classzone_idx >= new_classzone_idx &&
					balanced_order == new_order) {
2839 2840 2841 2842 2843 2844
			new_order = pgdat->kswapd_max_order;
			new_classzone_idx = pgdat->classzone_idx;
			pgdat->kswapd_max_order =  0;
			pgdat->classzone_idx = pgdat->nr_zones - 1;
		}

2845
		if (order < new_order || classzone_idx > new_classzone_idx) {
L
Linus Torvalds 已提交
2846 2847
			/*
			 * Don't sleep if someone wants a larger 'order'
2848
			 * allocation or has tigher zone constraints
L
Linus Torvalds 已提交
2849 2850
			 */
			order = new_order;
2851
			classzone_idx = new_classzone_idx;
L
Linus Torvalds 已提交
2852
		} else {
2853 2854
			kswapd_try_to_sleep(pgdat, balanced_order,
						balanced_classzone_idx);
L
Linus Torvalds 已提交
2855
			order = pgdat->kswapd_max_order;
2856
			classzone_idx = pgdat->classzone_idx;
2857 2858
			new_order = order;
			new_classzone_idx = classzone_idx;
2859
			pgdat->kswapd_max_order = 0;
2860
			pgdat->classzone_idx = pgdat->nr_zones - 1;
L
Linus Torvalds 已提交
2861 2862
		}

2863 2864 2865 2866 2867 2868 2869 2870
		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
		 */
2871 2872
		if (!ret) {
			trace_mm_vmscan_kswapd_wake(pgdat->node_id, order);
2873 2874 2875
			balanced_classzone_idx = classzone_idx;
			balanced_order = balance_pgdat(pgdat, order,
						&balanced_classzone_idx);
2876
		}
L
Linus Torvalds 已提交
2877 2878 2879 2880 2881 2882 2883
	}
	return 0;
}

/*
 * A zone is low on free memory, so wake its kswapd task to service it.
 */
2884
void wakeup_kswapd(struct zone *zone, int order, enum zone_type classzone_idx)
L
Linus Torvalds 已提交
2885 2886 2887
{
	pg_data_t *pgdat;

2888
	if (!populated_zone(zone))
L
Linus Torvalds 已提交
2889 2890
		return;

2891
	if (!cpuset_zone_allowed_hardwall(zone, GFP_KERNEL))
L
Linus Torvalds 已提交
2892
		return;
2893
	pgdat = zone->zone_pgdat;
2894
	if (pgdat->kswapd_max_order < order) {
L
Linus Torvalds 已提交
2895
		pgdat->kswapd_max_order = order;
2896 2897
		pgdat->classzone_idx = min(pgdat->classzone_idx, classzone_idx);
	}
2898
	if (!waitqueue_active(&pgdat->kswapd_wait))
L
Linus Torvalds 已提交
2899
		return;
2900 2901 2902 2903
	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);
2904
	wake_up_interruptible(&pgdat->kswapd_wait);
L
Linus Torvalds 已提交
2905 2906
}

2907 2908 2909 2910 2911 2912 2913 2914
/*
 * 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)
2915
{
2916 2917 2918 2919 2920 2921 2922 2923 2924 2925 2926 2927 2928 2929 2930 2931 2932 2933 2934 2935 2936 2937 2938 2939
	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;
2940 2941
}

2942
#ifdef CONFIG_HIBERNATION
L
Linus Torvalds 已提交
2943
/*
2944
 * Try to free `nr_to_reclaim' of memory, system-wide, and return the number of
2945 2946 2947 2948 2949
 * 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
2950
 */
2951
unsigned long shrink_all_memory(unsigned long nr_to_reclaim)
L
Linus Torvalds 已提交
2952
{
2953 2954
	struct reclaim_state reclaim_state;
	struct scan_control sc = {
2955 2956 2957
		.gfp_mask = GFP_HIGHUSER_MOVABLE,
		.may_swap = 1,
		.may_unmap = 1,
2958
		.may_writepage = 1,
2959 2960 2961
		.nr_to_reclaim = nr_to_reclaim,
		.hibernation_mode = 1,
		.order = 0,
L
Linus Torvalds 已提交
2962
	};
2963 2964 2965 2966
	struct shrink_control shrink = {
		.gfp_mask = sc.gfp_mask,
	};
	struct zonelist *zonelist = node_zonelist(numa_node_id(), sc.gfp_mask);
2967 2968
	struct task_struct *p = current;
	unsigned long nr_reclaimed;
L
Linus Torvalds 已提交
2969

2970 2971 2972 2973
	p->flags |= PF_MEMALLOC;
	lockdep_set_current_reclaim_state(sc.gfp_mask);
	reclaim_state.reclaimed_slab = 0;
	p->reclaim_state = &reclaim_state;
2974

2975
	nr_reclaimed = do_try_to_free_pages(zonelist, &sc, &shrink);
2976

2977 2978 2979
	p->reclaim_state = NULL;
	lockdep_clear_current_reclaim_state();
	p->flags &= ~PF_MEMALLOC;
2980

2981
	return nr_reclaimed;
L
Linus Torvalds 已提交
2982
}
2983
#endif /* CONFIG_HIBERNATION */
L
Linus Torvalds 已提交
2984 2985 2986 2987 2988

/* 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. */
2989
static int __devinit cpu_callback(struct notifier_block *nfb,
2990
				  unsigned long action, void *hcpu)
L
Linus Torvalds 已提交
2991
{
2992
	int nid;
L
Linus Torvalds 已提交
2993

2994
	if (action == CPU_ONLINE || action == CPU_ONLINE_FROZEN) {
2995
		for_each_node_state(nid, N_HIGH_MEMORY) {
2996
			pg_data_t *pgdat = NODE_DATA(nid);
2997 2998 2999
			const struct cpumask *mask;

			mask = cpumask_of_node(pgdat->node_id);
3000

3001
			if (cpumask_any_and(cpu_online_mask, mask) < nr_cpu_ids)
L
Linus Torvalds 已提交
3002
				/* One of our CPUs online: restore mask */
3003
				set_cpus_allowed_ptr(pgdat->kswapd, mask);
L
Linus Torvalds 已提交
3004 3005 3006 3007 3008
		}
	}
	return NOTIFY_OK;
}

3009 3010 3011 3012 3013 3014 3015 3016 3017 3018 3019 3020 3021 3022 3023 3024 3025 3026 3027 3028 3029 3030
/*
 * 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;
}

3031 3032 3033 3034 3035 3036 3037 3038 3039 3040 3041
/*
 * 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
3042 3043
static int __init kswapd_init(void)
{
3044
	int nid;
3045

L
Linus Torvalds 已提交
3046
	swap_setup();
3047
	for_each_node_state(nid, N_HIGH_MEMORY)
3048
 		kswapd_run(nid);
L
Linus Torvalds 已提交
3049 3050 3051 3052 3053
	hotcpu_notifier(cpu_callback, 0);
	return 0;
}

module_init(kswapd_init)
3054 3055 3056 3057 3058 3059 3060 3061 3062 3063

#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;

3064
#define RECLAIM_OFF 0
3065
#define RECLAIM_ZONE (1<<0)	/* Run shrink_inactive_list on the zone */
3066 3067 3068
#define RECLAIM_WRITE (1<<1)	/* Writeout pages during reclaim */
#define RECLAIM_SWAP (1<<2)	/* Swap pages out during reclaim */

3069 3070 3071 3072 3073 3074 3075
/*
 * 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

3076 3077 3078 3079 3080 3081
/*
 * Percentage of pages in a zone that must be unmapped for zone_reclaim to
 * occur.
 */
int sysctl_min_unmapped_ratio = 1;

3082 3083 3084 3085 3086 3087
/*
 * 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;

3088 3089 3090 3091 3092 3093 3094 3095 3096 3097 3098 3099 3100 3101 3102 3103 3104 3105 3106 3107 3108 3109 3110 3111 3112 3113 3114 3115 3116 3117 3118 3119 3120 3121 3122 3123 3124 3125 3126 3127 3128 3129
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;
}

3130 3131 3132
/*
 * Try to free up some pages from this zone through reclaim.
 */
3133
static int __zone_reclaim(struct zone *zone, gfp_t gfp_mask, unsigned int order)
3134
{
3135
	/* Minimum pages needed in order to stay on node */
3136
	const unsigned long nr_pages = 1 << order;
3137 3138
	struct task_struct *p = current;
	struct reclaim_state reclaim_state;
3139
	int priority;
3140 3141
	struct scan_control sc = {
		.may_writepage = !!(zone_reclaim_mode & RECLAIM_WRITE),
3142
		.may_unmap = !!(zone_reclaim_mode & RECLAIM_SWAP),
3143
		.may_swap = 1,
3144 3145
		.nr_to_reclaim = max_t(unsigned long, nr_pages,
				       SWAP_CLUSTER_MAX),
3146
		.gfp_mask = gfp_mask,
3147
		.order = order,
3148
	};
3149 3150 3151
	struct shrink_control shrink = {
		.gfp_mask = sc.gfp_mask,
	};
3152
	unsigned long nr_slab_pages0, nr_slab_pages1;
3153 3154

	cond_resched();
3155 3156 3157 3158 3159 3160
	/*
	 * 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;
3161
	lockdep_set_current_reclaim_state(gfp_mask);
3162 3163
	reclaim_state.reclaimed_slab = 0;
	p->reclaim_state = &reclaim_state;
3164

3165
	if (zone_pagecache_reclaimable(zone) > zone->min_unmapped_pages) {
3166 3167 3168 3169 3170 3171
		/*
		 * Free memory by calling shrink zone with increasing
		 * priorities until we have enough memory freed.
		 */
		priority = ZONE_RECLAIM_PRIORITY;
		do {
3172
			shrink_zone(priority, zone, &sc);
3173
			priority--;
3174
		} while (priority >= 0 && sc.nr_reclaimed < nr_pages);
3175
	}
3176

3177 3178
	nr_slab_pages0 = zone_page_state(zone, NR_SLAB_RECLAIMABLE);
	if (nr_slab_pages0 > zone->min_slab_pages) {
3179
		/*
3180
		 * shrink_slab() does not currently allow us to determine how
3181 3182 3183 3184
		 * 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.
3185
		 *
3186 3187
		 * Note that shrink_slab will free memory on all zones and may
		 * take a long time.
3188
		 */
3189 3190 3191 3192
		for (;;) {
			unsigned long lru_pages = zone_reclaimable_pages(zone);

			/* No reclaimable slab or very low memory pressure */
3193
			if (!shrink_slab(&shrink, sc.nr_scanned, lru_pages))
3194 3195 3196 3197 3198 3199 3200 3201
				break;

			/* Freed enough memory */
			nr_slab_pages1 = zone_page_state(zone,
							NR_SLAB_RECLAIMABLE);
			if (nr_slab_pages1 + nr_pages <= nr_slab_pages0)
				break;
		}
3202 3203 3204 3205 3206

		/*
		 * Update nr_reclaimed by the number of slab pages we
		 * reclaimed from this zone.
		 */
3207 3208 3209
		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;
3210 3211
	}

3212
	p->reclaim_state = NULL;
3213
	current->flags &= ~(PF_MEMALLOC | PF_SWAPWRITE);
3214
	lockdep_clear_current_reclaim_state();
3215
	return sc.nr_reclaimed >= nr_pages;
3216
}
3217 3218 3219 3220

int zone_reclaim(struct zone *zone, gfp_t gfp_mask, unsigned int order)
{
	int node_id;
3221
	int ret;
3222 3223

	/*
3224 3225
	 * Zone reclaim reclaims unmapped file backed pages and
	 * slab pages if we are over the defined limits.
3226
	 *
3227 3228 3229 3230 3231
	 * 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.
3232
	 */
3233 3234
	if (zone_pagecache_reclaimable(zone) <= zone->min_unmapped_pages &&
	    zone_page_state(zone, NR_SLAB_RECLAIMABLE) <= zone->min_slab_pages)
3235
		return ZONE_RECLAIM_FULL;
3236

3237
	if (zone->all_unreclaimable)
3238
		return ZONE_RECLAIM_FULL;
3239

3240
	/*
3241
	 * Do not scan if the allocation should not be delayed.
3242
	 */
3243
	if (!(gfp_mask & __GFP_WAIT) || (current->flags & PF_MEMALLOC))
3244
		return ZONE_RECLAIM_NOSCAN;
3245 3246 3247 3248 3249 3250 3251

	/*
	 * 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.
	 */
3252
	node_id = zone_to_nid(zone);
3253
	if (node_state(node_id, N_CPU) && node_id != numa_node_id())
3254
		return ZONE_RECLAIM_NOSCAN;
3255 3256

	if (zone_test_and_set_flag(zone, ZONE_RECLAIM_LOCKED))
3257 3258
		return ZONE_RECLAIM_NOSCAN;

3259 3260 3261
	ret = __zone_reclaim(zone, gfp_mask, order);
	zone_clear_flag(zone, ZONE_RECLAIM_LOCKED);

3262 3263 3264
	if (!ret)
		count_vm_event(PGSCAN_ZONE_RECLAIM_FAILED);

3265
	return ret;
3266
}
3267
#endif
L
Lee Schermerhorn 已提交
3268 3269 3270 3271 3272 3273 3274

/*
 * 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
3275 3276
 * 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
3277 3278
 *
 * Reasons page might not be evictable:
3279
 * (1) page's mapping marked unevictable
N
Nick Piggin 已提交
3280
 * (2) page is part of an mlocked VMA
3281
 *
L
Lee Schermerhorn 已提交
3282 3283 3284 3285
 */
int page_evictable(struct page *page, struct vm_area_struct *vma)
{

3286 3287 3288
	if (mapping_unevictable(page_mapping(page)))
		return 0;

3289
	if (PageMlocked(page) || (vma && mlocked_vma_newpage(vma, page)))
N
Nick Piggin 已提交
3290
		return 0;
L
Lee Schermerhorn 已提交
3291 3292 3293

	return 1;
}
3294

3295
#ifdef CONFIG_SHMEM
3296
/**
3297 3298 3299
 * check_move_unevictable_pages - check pages for evictability and move to appropriate zone lru list
 * @pages:	array of pages to check
 * @nr_pages:	number of pages to check
3300
 *
3301
 * Checks pages for evictability and moves them to the appropriate lru list.
3302 3303
 *
 * This function is only used for SysV IPC SHM_UNLOCK.
3304
 */
3305
void check_move_unevictable_pages(struct page **pages, int nr_pages)
3306
{
3307
	struct lruvec *lruvec;
3308 3309 3310 3311
	struct zone *zone = NULL;
	int pgscanned = 0;
	int pgrescued = 0;
	int i;
3312

3313 3314 3315
	for (i = 0; i < nr_pages; i++) {
		struct page *page = pages[i];
		struct zone *pagezone;
3316

3317 3318 3319 3320 3321 3322 3323 3324
		pgscanned++;
		pagezone = page_zone(page);
		if (pagezone != zone) {
			if (zone)
				spin_unlock_irq(&zone->lru_lock);
			zone = pagezone;
			spin_lock_irq(&zone->lru_lock);
		}
3325

3326 3327
		if (!PageLRU(page) || !PageUnevictable(page))
			continue;
3328

3329 3330 3331 3332 3333 3334 3335 3336 3337 3338 3339
		if (page_evictable(page, NULL)) {
			enum lru_list lru = page_lru_base_type(page);

			VM_BUG_ON(PageActive(page));
			ClearPageUnevictable(page);
			__dec_zone_state(zone, NR_UNEVICTABLE);
			lruvec = mem_cgroup_lru_move_lists(zone, page,
						LRU_UNEVICTABLE, lru);
			list_move(&page->lru, &lruvec->lists[lru]);
			__inc_zone_state(zone, NR_INACTIVE_ANON + lru);
			pgrescued++;
3340
		}
3341
	}
3342

3343 3344 3345 3346
	if (zone) {
		__count_vm_events(UNEVICTABLE_PGRESCUED, pgrescued);
		__count_vm_events(UNEVICTABLE_PGSCANNED, pgscanned);
		spin_unlock_irq(&zone->lru_lock);
3347 3348
	}
}
3349
#endif /* CONFIG_SHMEM */
3350

3351
static void warn_scan_unevictable_pages(void)
3352
{
3353
	printk_once(KERN_WARNING
3354
		    "%s: The scan_unevictable_pages sysctl/node-interface has been "
3355
		    "disabled for lack of a legitimate use case.  If you have "
3356 3357
		    "one, please send an email to linux-mm@kvack.org.\n",
		    current->comm);
3358 3359 3360 3361 3362 3363 3364 3365 3366
}

/*
 * 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,
3367
			   void __user *buffer,
3368 3369
			   size_t *length, loff_t *ppos)
{
3370
	warn_scan_unevictable_pages();
3371
	proc_doulongvec_minmax(table, write, buffer, length, ppos);
3372 3373 3374 3375
	scan_unevictable_pages = 0;
	return 0;
}

3376
#ifdef CONFIG_NUMA
3377 3378 3379 3380 3381
/*
 * per node 'scan_unevictable_pages' attribute.  On demand re-scan of
 * a specified node's per zone unevictable lists for evictable pages.
 */

3382 3383
static ssize_t read_scan_unevictable_node(struct device *dev,
					  struct device_attribute *attr,
3384 3385
					  char *buf)
{
3386
	warn_scan_unevictable_pages();
3387 3388 3389
	return sprintf(buf, "0\n");	/* always zero; should fit... */
}

3390 3391
static ssize_t write_scan_unevictable_node(struct device *dev,
					   struct device_attribute *attr,
3392 3393
					const char *buf, size_t count)
{
3394
	warn_scan_unevictable_pages();
3395 3396 3397 3398
	return 1;
}


3399
static DEVICE_ATTR(scan_unevictable_pages, S_IRUGO | S_IWUSR,
3400 3401 3402 3403 3404
			read_scan_unevictable_node,
			write_scan_unevictable_node);

int scan_unevictable_register_node(struct node *node)
{
3405
	return device_create_file(&node->dev, &dev_attr_scan_unevictable_pages);
3406 3407 3408 3409
}

void scan_unevictable_unregister_node(struct node *node)
{
3410
	device_remove_file(&node->dev, &dev_attr_scan_unevictable_pages);
3411
}
3412
#endif