vmscan.c 99.5 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
/*
 *  linux/mm/vmscan.c
 *
 *  Copyright (C) 1991, 1992, 1993, 1994  Linus Torvalds
 *
 *  Swap reorganised 29.12.95, Stephen Tweedie.
 *  kswapd added: 7.1.96  sct
 *  Removed kswapd_ctl limits, and swap out as many pages as needed
 *  to bring the system back to freepages.high: 2.4.97, Rik van Riel.
 *  Zone aware kswapd started 02/00, Kanoj Sarcar (kanoj@sgi.com).
 *  Multiqueue VM started 5.8.00, Rik van Riel.
 */

#include <linux/mm.h>
#include <linux/module.h>
16
#include <linux/gfp.h>
L
Linus Torvalds 已提交
17 18 19 20 21
#include <linux/kernel_stat.h>
#include <linux/swap.h>
#include <linux/pagemap.h>
#include <linux/init.h>
#include <linux/highmem.h>
22
#include <linux/vmstat.h>
L
Linus Torvalds 已提交
23 24 25 26 27 28 29 30 31 32 33
#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
	/* Scan (total_size >> priority) pages at once */
	int priority;

84 85 86 87 88
	/*
	 * 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;
89

90 91 92 93 94
	/*
	 * Nodemask of nodes allowed by the caller. If NULL, all nodes
	 * are scanned.
	 */
	nodemask_t	*nodemask;
L
Linus Torvalds 已提交
95 96 97 98 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
};

#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;
131
long vm_total_pages;	/* The total number of pages which the VM controls */
L
Linus Torvalds 已提交
132 133 134 135

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

A
Andrew Morton 已提交
136
#ifdef CONFIG_MEMCG
137 138
static bool global_reclaim(struct scan_control *sc)
{
139
	return !sc->target_mem_cgroup;
140
}
141
#else
142 143 144 145
static bool global_reclaim(struct scan_control *sc)
{
	return true;
}
146 147
#endif

148
static unsigned long get_lru_size(struct lruvec *lruvec, enum lru_list lru)
149
{
150
	if (!mem_cgroup_disabled())
151
		return mem_cgroup_get_lru_size(lruvec, lru);
152

153
	return zone_page_state(lruvec_zone(lruvec), NR_LRU_BASE + lru);
154 155
}

L
Linus Torvalds 已提交
156 157 158
/*
 * Add a shrinker callback to be called from the vm
 */
159
void register_shrinker(struct shrinker *shrinker)
L
Linus Torvalds 已提交
160
{
161
	atomic_long_set(&shrinker->nr_in_batch, 0);
162 163 164
	down_write(&shrinker_rwsem);
	list_add_tail(&shrinker->list, &shrinker_list);
	up_write(&shrinker_rwsem);
L
Linus Torvalds 已提交
165
}
166
EXPORT_SYMBOL(register_shrinker);
L
Linus Torvalds 已提交
167 168 169 170

/*
 * Remove one
 */
171
void unregister_shrinker(struct shrinker *shrinker)
L
Linus Torvalds 已提交
172 173 174 175 176
{
	down_write(&shrinker_rwsem);
	list_del(&shrinker->list);
	up_write(&shrinker_rwsem);
}
177
EXPORT_SYMBOL(unregister_shrinker);
L
Linus Torvalds 已提交
178

179 180 181 182 183 184 185 186
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
187 188 189 190 191 192 193 194 195
#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
196
 * If the vm encountered mapped pages on the LRU it increase the pressure on
L
Linus Torvalds 已提交
197 198 199 200 201 202 203
 * 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.
204 205
 *
 * Returns the number of slab objects which we shrunk.
L
Linus Torvalds 已提交
206
 */
207
unsigned long shrink_slab(struct shrink_control *shrink,
208
			  unsigned long nr_pages_scanned,
209
			  unsigned long lru_pages)
L
Linus Torvalds 已提交
210 211
{
	struct shrinker *shrinker;
212
	unsigned long ret = 0;
L
Linus Torvalds 已提交
213

214 215
	if (nr_pages_scanned == 0)
		nr_pages_scanned = SWAP_CLUSTER_MAX;
L
Linus Torvalds 已提交
216

217 218 219 220 221
	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
222 223 224

	list_for_each_entry(shrinker, &shrinker_list, list) {
		unsigned long long delta;
225 226
		long total_scan;
		long max_pass;
227
		int shrink_ret = 0;
228 229
		long nr;
		long new_nr;
230 231
		long batch_size = shrinker->batch ? shrinker->batch
						  : SHRINK_BATCH;
L
Linus Torvalds 已提交
232

233 234 235 236
		max_pass = do_shrinker_shrink(shrinker, shrink, 0);
		if (max_pass <= 0)
			continue;

237 238 239 240 241
		/*
		 * 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.
		 */
242
		nr = atomic_long_xchg(&shrinker->nr_in_batch, 0);
243 244

		total_scan = nr;
245
		delta = (4 * nr_pages_scanned) / shrinker->seeks;
246
		delta *= max_pass;
L
Linus Torvalds 已提交
247
		do_div(delta, lru_pages + 1);
248 249
		total_scan += delta;
		if (total_scan < 0) {
250 251
			printk(KERN_ERR "shrink_slab: %pF negative objects to "
			       "delete nr=%ld\n",
252 253
			       shrinker->shrink, total_scan);
			total_scan = max_pass;
254 255
		}

256 257 258 259 260 261 262 263 264 265 266 267 268 269 270
		/*
		 * 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);

271 272 273 274 275
		/*
		 * Avoid risking looping forever due to too large nr value:
		 * never try to free more than twice the estimate number of
		 * freeable entries.
		 */
276 277
		if (total_scan > max_pass * 2)
			total_scan = max_pass * 2;
L
Linus Torvalds 已提交
278

279
		trace_mm_shrink_slab_start(shrinker, shrink, nr,
280 281 282
					nr_pages_scanned, lru_pages,
					max_pass, delta, total_scan);

283
		while (total_scan >= batch_size) {
284
			int nr_before;
L
Linus Torvalds 已提交
285

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

			cond_resched();
		}

299 300 301 302 303
		/*
		 * 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.
		 */
304 305 306 307 308
		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);
309 310

		trace_mm_shrink_slab_end(shrinker, shrink_ret, nr, new_nr);
L
Linus Torvalds 已提交
311 312
	}
	up_read(&shrinker_rwsem);
313 314
out:
	cond_resched();
315
	return ret;
L
Linus Torvalds 已提交
316 317 318 319
}

static inline int is_page_cache_freeable(struct page *page)
{
320 321 322 323 324
	/*
	 * 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.
	 */
325
	return page_count(page) - page_has_private(page) == 2;
L
Linus Torvalds 已提交
326 327
}

328 329
static int may_write_to_queue(struct backing_dev_info *bdi,
			      struct scan_control *sc)
L
Linus Torvalds 已提交
330
{
331
	if (current->flags & PF_SWAPWRITE)
L
Linus Torvalds 已提交
332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354
		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
355
	lock_page(page);
356 357
	if (page_mapping(page) == mapping)
		mapping_set_error(mapping, error);
L
Linus Torvalds 已提交
358 359 360
	unlock_page(page);
}

361 362 363 364 365 366 367 368 369 370 371 372
/* 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
373
/*
A
Andrew Morton 已提交
374 375
 * pageout is called by shrink_page_list() for each dirty page.
 * Calls ->writepage().
L
Linus Torvalds 已提交
376
 */
377
static pageout_t pageout(struct page *page, struct address_space *mapping,
378
			 struct scan_control *sc)
L
Linus Torvalds 已提交
379 380 381 382 383 384 385 386
{
	/*
	 * 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.
	 *
387
	 * If this process is currently in __generic_file_aio_write() against
L
Linus Torvalds 已提交
388 389 390 391 392 393 394 395 396 397 398 399 400 401 402
	 * 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.
		 */
403
		if (page_has_private(page)) {
L
Linus Torvalds 已提交
404 405
			if (try_to_free_buffers(page)) {
				ClearPageDirty(page);
406
				printk("%s: orphaned page\n", __func__);
L
Linus Torvalds 已提交
407 408 409 410 411 412 413
				return PAGE_CLEAN;
			}
		}
		return PAGE_KEEP;
	}
	if (mapping->a_ops->writepage == NULL)
		return PAGE_ACTIVATE;
414
	if (!may_write_to_queue(mapping->backing_dev_info, sc))
L
Linus Torvalds 已提交
415 416 417 418 419 420 421
		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,
422 423
			.range_start = 0,
			.range_end = LLONG_MAX,
L
Linus Torvalds 已提交
424 425 426 427 428 429 430
			.for_reclaim = 1,
		};

		SetPageReclaim(page);
		res = mapping->a_ops->writepage(page, &wbc);
		if (res < 0)
			handle_write_error(mapping, page, res);
431
		if (res == AOP_WRITEPAGE_ACTIVATE) {
L
Linus Torvalds 已提交
432 433 434
			ClearPageReclaim(page);
			return PAGE_ACTIVATE;
		}
435

L
Linus Torvalds 已提交
436 437 438 439
		if (!PageWriteback(page)) {
			/* synchronous write or broken a_ops? */
			ClearPageReclaim(page);
		}
M
Mel Gorman 已提交
440
		trace_mm_vmscan_writepage(page, trace_reclaim_flags(page));
441
		inc_zone_page_state(page, NR_VMSCAN_WRITE);
L
Linus Torvalds 已提交
442 443 444 445 446 447
		return PAGE_SUCCESS;
	}

	return PAGE_CLEAN;
}

448
/*
N
Nick Piggin 已提交
449 450
 * Same as remove_mapping, but if the page is removed from the mapping, it
 * gets returned with a refcount of 0.
451
 */
N
Nick Piggin 已提交
452
static int __remove_mapping(struct address_space *mapping, struct page *page)
453
{
454 455
	BUG_ON(!PageLocked(page));
	BUG_ON(mapping != page_mapping(page));
456

N
Nick Piggin 已提交
457
	spin_lock_irq(&mapping->tree_lock);
458
	/*
N
Nick Piggin 已提交
459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481
	 * 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.
482
	 */
N
Nick Piggin 已提交
483
	if (!page_freeze_refs(page, 2))
484
		goto cannot_free;
N
Nick Piggin 已提交
485 486 487
	/* note: atomic_cmpxchg in page_freeze_refs provides the smp_rmb */
	if (unlikely(PageDirty(page))) {
		page_unfreeze_refs(page, 2);
488
		goto cannot_free;
N
Nick Piggin 已提交
489
	}
490 491 492 493

	if (PageSwapCache(page)) {
		swp_entry_t swap = { .val = page_private(page) };
		__delete_from_swap_cache(page);
N
Nick Piggin 已提交
494
		spin_unlock_irq(&mapping->tree_lock);
495
		swapcache_free(swap, page);
N
Nick Piggin 已提交
496
	} else {
497 498 499 500
		void (*freepage)(struct page *);

		freepage = mapping->a_ops->freepage;

501
		__delete_from_page_cache(page);
N
Nick Piggin 已提交
502
		spin_unlock_irq(&mapping->tree_lock);
503
		mem_cgroup_uncharge_cache_page(page);
504 505 506

		if (freepage != NULL)
			freepage(page);
507 508 509 510 511
	}

	return 1;

cannot_free:
N
Nick Piggin 已提交
512
	spin_unlock_irq(&mapping->tree_lock);
513 514 515
	return 0;
}

N
Nick Piggin 已提交
516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535
/*
 * 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
536 537 538 539 540 541 542 543 544 545 546 547 548
/**
 * 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);
549
	int was_unevictable = PageUnevictable(page);
L
Lee Schermerhorn 已提交
550 551 552 553 554 555

	VM_BUG_ON(PageLRU(page));

redo:
	ClearPageUnevictable(page);

556
	if (page_evictable(page)) {
L
Lee Schermerhorn 已提交
557 558 559 560 561 562
		/*
		 * 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.
		 */
563
		lru = active + page_lru_base_type(page);
L
Lee Schermerhorn 已提交
564 565 566 567 568 569 570 571
		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);
572
		/*
573 574 575
		 * 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
576
		 * isolation/check_move_unevictable_pages,
577
		 * we see PG_mlocked/AS_UNEVICTABLE cleared below and move
578 579
		 * the page back to the evictable list.
		 *
580
		 * The other side is TestClearPageMlocked() or shmem_lock().
581 582
		 */
		smp_mb();
L
Lee Schermerhorn 已提交
583 584 585 586 587 588 589
	}

	/*
	 * 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.
	 */
590
	if (lru == LRU_UNEVICTABLE && page_evictable(page)) {
L
Lee Schermerhorn 已提交
591 592 593 594 595 596 597 598 599 600
		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.
		 */
	}

601 602 603 604 605
	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
606 607 608
	put_page(page);		/* drop ref from isolate */
}

609 610 611
enum page_references {
	PAGEREF_RECLAIM,
	PAGEREF_RECLAIM_CLEAN,
612
	PAGEREF_KEEP,
613 614 615 616 617 618
	PAGEREF_ACTIVATE,
};

static enum page_references page_check_references(struct page *page,
						  struct scan_control *sc)
{
619
	int referenced_ptes, referenced_page;
620 621
	unsigned long vm_flags;

622 623
	referenced_ptes = page_referenced(page, 1, sc->target_mem_cgroup,
					  &vm_flags);
624
	referenced_page = TestClearPageReferenced(page);
625 626 627 628 629 630 631 632

	/*
	 * 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;

633
	if (referenced_ptes) {
634
		if (PageSwapBacked(page))
635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651
			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);

652
		if (referenced_page || referenced_ptes > 1)
653 654
			return PAGEREF_ACTIVATE;

655 656 657 658 659 660
		/*
		 * Activate file-backed executable pages after first usage.
		 */
		if (vm_flags & VM_EXEC)
			return PAGEREF_ACTIVATE;

661 662
		return PAGEREF_KEEP;
	}
663 664

	/* Reclaim if clean, defer dirty pages to writeback */
665
	if (referenced_page && !PageSwapBacked(page))
666 667 668
		return PAGEREF_RECLAIM_CLEAN;

	return PAGEREF_RECLAIM;
669 670
}

L
Linus Torvalds 已提交
671
/*
A
Andrew Morton 已提交
672
 * shrink_page_list() returns the number of reclaimed pages
L
Linus Torvalds 已提交
673
 */
A
Andrew Morton 已提交
674
static unsigned long shrink_page_list(struct list_head *page_list,
675
				      struct zone *zone,
676
				      struct scan_control *sc,
677
				      enum ttu_flags ttu_flags,
678
				      unsigned long *ret_nr_dirty,
679 680
				      unsigned long *ret_nr_writeback,
				      bool force_reclaim)
L
Linus Torvalds 已提交
681 682
{
	LIST_HEAD(ret_pages);
683
	LIST_HEAD(free_pages);
L
Linus Torvalds 已提交
684
	int pgactivate = 0;
685 686
	unsigned long nr_dirty = 0;
	unsigned long nr_congested = 0;
687
	unsigned long nr_reclaimed = 0;
688
	unsigned long nr_writeback = 0;
L
Linus Torvalds 已提交
689 690 691

	cond_resched();

692
	mem_cgroup_uncharge_start();
L
Linus Torvalds 已提交
693 694 695 696
	while (!list_empty(page_list)) {
		struct address_space *mapping;
		struct page *page;
		int may_enter_fs;
697
		enum page_references references = PAGEREF_RECLAIM_CLEAN;
L
Linus Torvalds 已提交
698 699 700 701 702 703

		cond_resched();

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

N
Nick Piggin 已提交
704
		if (!trylock_page(page))
L
Linus Torvalds 已提交
705 706
			goto keep;

N
Nick Piggin 已提交
707
		VM_BUG_ON(PageActive(page));
708
		VM_BUG_ON(page_zone(page) != zone);
L
Linus Torvalds 已提交
709 710

		sc->nr_scanned++;
711

712
		if (unlikely(!page_evictable(page)))
N
Nick Piggin 已提交
713
			goto cull_mlocked;
L
Lee Schermerhorn 已提交
714

715
		if (!sc->may_unmap && page_mapped(page))
716 717
			goto keep_locked;

L
Linus Torvalds 已提交
718 719 720 721
		/* Double the slab pressure for mapped and swapcache pages */
		if (page_mapped(page) || PageSwapCache(page))
			sc->nr_scanned++;

722 723 724 725
		may_enter_fs = (sc->gfp_mask & __GFP_FS) ||
			(PageSwapCache(page) && (sc->gfp_mask & __GFP_IO));

		if (PageWriteback(page)) {
726 727 728
			/*
			 * memcg doesn't have any dirty pages throttling so we
			 * could easily OOM just because too many pages are in
729
			 * writeback and there is nothing else to reclaim.
730
			 *
731
			 * Check __GFP_IO, certainly because a loop driver
732 733 734 735
			 * thread might enter reclaim, and deadlock if it waits
			 * on a page for which it is needed to do the write
			 * (loop masks off __GFP_IO|__GFP_FS for this reason);
			 * but more thought would probably show more reasons.
736 737 738 739 740 741
			 *
			 * Don't require __GFP_FS, since we're not going into
			 * the FS, just waiting on its writeback completion.
			 * Worryingly, ext4 gfs2 and xfs allocate pages with
			 * grab_cache_page_write_begin(,,AOP_FLAG_NOFS), so
			 * testing may_enter_fs here is liable to OOM on them.
742
			 */
743 744 745 746 747 748 749 750 751 752 753 754 755 756
			if (global_reclaim(sc) ||
			    !PageReclaim(page) || !(sc->gfp_mask & __GFP_IO)) {
				/*
				 * This is slightly racy - end_page_writeback()
				 * might have just cleared PageReclaim, then
				 * setting PageReclaim here end up interpreted
				 * as PageReadahead - but that does not matter
				 * enough to care.  What we do want is for this
				 * page to have PageReclaim set next time memcg
				 * reclaim reaches the tests above, so it will
				 * then wait_on_page_writeback() to avoid OOM;
				 * and it's also appropriate in global reclaim.
				 */
				SetPageReclaim(page);
757
				nr_writeback++;
758
				goto keep_locked;
759
			}
760
			wait_on_page_writeback(page);
761
		}
L
Linus Torvalds 已提交
762

763 764 765
		if (!force_reclaim)
			references = page_check_references(page, sc);

766 767
		switch (references) {
		case PAGEREF_ACTIVATE:
L
Linus Torvalds 已提交
768
			goto activate_locked;
769 770
		case PAGEREF_KEEP:
			goto keep_locked;
771 772 773 774
		case PAGEREF_RECLAIM:
		case PAGEREF_RECLAIM_CLEAN:
			; /* try to reclaim the page below */
		}
L
Linus Torvalds 已提交
775 776 777 778 779

		/*
		 * Anonymous process memory has backing store?
		 * Try to allocate it some swap space here.
		 */
N
Nick Piggin 已提交
780
		if (PageAnon(page) && !PageSwapCache(page)) {
781 782
			if (!(sc->gfp_mask & __GFP_IO))
				goto keep_locked;
783
			if (!add_to_swap(page))
L
Linus Torvalds 已提交
784
				goto activate_locked;
785
			may_enter_fs = 1;
N
Nick Piggin 已提交
786
		}
L
Linus Torvalds 已提交
787 788 789 790 791 792 793 794

		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) {
795
			switch (try_to_unmap(page, ttu_flags)) {
L
Linus Torvalds 已提交
796 797 798 799
			case SWAP_FAIL:
				goto activate_locked;
			case SWAP_AGAIN:
				goto keep_locked;
N
Nick Piggin 已提交
800 801
			case SWAP_MLOCK:
				goto cull_mlocked;
L
Linus Torvalds 已提交
802 803 804 805 806 807
			case SWAP_SUCCESS:
				; /* try to free the page below */
			}
		}

		if (PageDirty(page)) {
808 809
			nr_dirty++;

810 811
			/*
			 * Only kswapd can writeback filesystem pages to
812 813
			 * avoid risk of stack overflow but do not writeback
			 * unless under significant pressure.
814
			 */
815
			if (page_is_file_cache(page) &&
816 817
					(!current_is_kswapd() ||
					 sc->priority >= DEF_PRIORITY - 2)) {
818 819 820 821 822 823 824 825 826
				/*
				 * 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);

827 828 829
				goto keep_locked;
			}

830
			if (references == PAGEREF_RECLAIM_CLEAN)
L
Linus Torvalds 已提交
831
				goto keep_locked;
832
			if (!may_enter_fs)
L
Linus Torvalds 已提交
833
				goto keep_locked;
834
			if (!sc->may_writepage)
L
Linus Torvalds 已提交
835 836 837
				goto keep_locked;

			/* Page is dirty, try to write it out here */
838
			switch (pageout(page, mapping, sc)) {
L
Linus Torvalds 已提交
839
			case PAGE_KEEP:
840
				nr_congested++;
L
Linus Torvalds 已提交
841 842 843 844
				goto keep_locked;
			case PAGE_ACTIVATE:
				goto activate_locked;
			case PAGE_SUCCESS:
845
				if (PageWriteback(page))
846
					goto keep;
847
				if (PageDirty(page))
L
Linus Torvalds 已提交
848
					goto keep;
849

L
Linus Torvalds 已提交
850 851 852 853
				/*
				 * A synchronous write - probably a ramdisk.  Go
				 * ahead and try to reclaim the page.
				 */
N
Nick Piggin 已提交
854
				if (!trylock_page(page))
L
Linus Torvalds 已提交
855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873
					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
874
		 * will do this, as well as the blockdev mapping.
L
Linus Torvalds 已提交
875 876 877 878 879 880 881 882 883 884
		 * 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.
		 */
885
		if (page_has_private(page)) {
L
Linus Torvalds 已提交
886 887
			if (!try_to_release_page(page, sc->gfp_mask))
				goto activate_locked;
N
Nick Piggin 已提交
888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903
			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
904 905
		}

N
Nick Piggin 已提交
906
		if (!mapping || !__remove_mapping(mapping, page))
907
			goto keep_locked;
L
Linus Torvalds 已提交
908

N
Nick Piggin 已提交
909 910 911 912 913 914 915 916
		/*
		 * 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
917
free_it:
918
		nr_reclaimed++;
919 920 921 922 923 924

		/*
		 * 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
925 926
		continue;

N
Nick Piggin 已提交
927
cull_mlocked:
928 929
		if (PageSwapCache(page))
			try_to_free_swap(page);
N
Nick Piggin 已提交
930 931 932 933
		unlock_page(page);
		putback_lru_page(page);
		continue;

L
Linus Torvalds 已提交
934
activate_locked:
935 936
		/* Not a candidate for swapping, so reclaim swap space. */
		if (PageSwapCache(page) && vm_swap_full())
937
			try_to_free_swap(page);
L
Lee Schermerhorn 已提交
938
		VM_BUG_ON(PageActive(page));
L
Linus Torvalds 已提交
939 940 941 942 943 944
		SetPageActive(page);
		pgactivate++;
keep_locked:
		unlock_page(page);
keep:
		list_add(&page->lru, &ret_pages);
N
Nick Piggin 已提交
945
		VM_BUG_ON(PageLRU(page) || PageUnevictable(page));
L
Linus Torvalds 已提交
946
	}
947

948 949 950 951 952 953
	/*
	 * 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
	 */
954
	if (nr_dirty && nr_dirty == nr_congested && global_reclaim(sc))
955
		zone_set_flag(zone, ZONE_CONGESTED);
956

957
	free_hot_cold_page_list(&free_pages, 1);
958

L
Linus Torvalds 已提交
959
	list_splice(&ret_pages, page_list);
960
	count_vm_events(PGACTIVATE, pgactivate);
961
	mem_cgroup_uncharge_end();
962 963
	*ret_nr_dirty += nr_dirty;
	*ret_nr_writeback += nr_writeback;
964
	return nr_reclaimed;
L
Linus Torvalds 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
unsigned long reclaim_clean_pages_from_list(struct zone *zone,
					    struct list_head *page_list)
{
	struct scan_control sc = {
		.gfp_mask = GFP_KERNEL,
		.priority = DEF_PRIORITY,
		.may_unmap = 1,
	};
	unsigned long ret, dummy1, dummy2;
	struct page *page, *next;
	LIST_HEAD(clean_pages);

	list_for_each_entry_safe(page, next, page_list, lru) {
		if (page_is_file_cache(page) && !PageDirty(page)) {
			ClearPageActive(page);
			list_move(&page->lru, &clean_pages);
		}
	}

	ret = shrink_page_list(&clean_pages, zone, &sc,
				TTU_UNMAP|TTU_IGNORE_ACCESS,
				&dummy1, &dummy2, true);
	list_splice(&clean_pages, page_list);
	__mod_zone_page_state(zone, NR_ISOLATED_FILE, -ret);
	return ret;
}

A
Andy Whitcroft 已提交
994 995 996 997 998 999 1000 1001 1002 1003
/*
 * 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.
 */
1004
int __isolate_lru_page(struct page *page, isolate_mode_t mode)
A
Andy Whitcroft 已提交
1005 1006 1007 1008 1009 1010 1011
{
	int ret = -EINVAL;

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

M
Minchan Kim 已提交
1012 1013
	/* Compaction should not handle unevictable pages but CMA can do so */
	if (PageUnevictable(page) && !(mode & ISOLATE_UNEVICTABLE))
L
Lee Schermerhorn 已提交
1014 1015
		return ret;

A
Andy Whitcroft 已提交
1016
	ret = -EBUSY;
K
KAMEZAWA Hiroyuki 已提交
1017

1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050
	/*
	 * 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;
		}
	}
1051

1052 1053 1054
	if ((mode & ISOLATE_UNMAPPED) && page_mapped(page))
		return ret;

A
Andy Whitcroft 已提交
1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067
	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
1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078
/*
 * 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.
1079
 * @lruvec:	The LRU vector to pull pages from.
L
Linus Torvalds 已提交
1080
 * @dst:	The temp list to put pages on to.
H
Hugh Dickins 已提交
1081
 * @nr_scanned:	The number of pages that were scanned.
1082
 * @sc:		The scan_control struct for this reclaim session
A
Andy Whitcroft 已提交
1083
 * @mode:	One of the LRU isolation modes
1084
 * @lru:	LRU list id for isolating
L
Linus Torvalds 已提交
1085 1086 1087
 *
 * returns how many pages were moved onto *@dst.
 */
1088
static unsigned long isolate_lru_pages(unsigned long nr_to_scan,
1089
		struct lruvec *lruvec, struct list_head *dst,
1090
		unsigned long *nr_scanned, struct scan_control *sc,
1091
		isolate_mode_t mode, enum lru_list lru)
L
Linus Torvalds 已提交
1092
{
H
Hugh Dickins 已提交
1093
	struct list_head *src = &lruvec->lists[lru];
1094
	unsigned long nr_taken = 0;
1095
	unsigned long scan;
L
Linus Torvalds 已提交
1096

1097
	for (scan = 0; scan < nr_to_scan && !list_empty(src); scan++) {
A
Andy Whitcroft 已提交
1098
		struct page *page;
1099
		int nr_pages;
A
Andy Whitcroft 已提交
1100

L
Linus Torvalds 已提交
1101 1102 1103
		page = lru_to_page(src);
		prefetchw_prev_lru_page(page, src, flags);

N
Nick Piggin 已提交
1104
		VM_BUG_ON(!PageLRU(page));
N
Nick Piggin 已提交
1105

1106
		switch (__isolate_lru_page(page, mode)) {
A
Andy Whitcroft 已提交
1107
		case 0:
1108 1109
			nr_pages = hpage_nr_pages(page);
			mem_cgroup_update_lru_size(lruvec, lru, -nr_pages);
A
Andy Whitcroft 已提交
1110
			list_move(&page->lru, dst);
1111
			nr_taken += nr_pages;
A
Andy Whitcroft 已提交
1112 1113 1114 1115 1116 1117
			break;

		case -EBUSY:
			/* else it is being freed elsewhere */
			list_move(&page->lru, src);
			continue;
1118

A
Andy Whitcroft 已提交
1119 1120 1121
		default:
			BUG();
		}
L
Linus Torvalds 已提交
1122 1123
	}

H
Hugh Dickins 已提交
1124
	*nr_scanned = scan;
H
Hugh Dickins 已提交
1125 1126
	trace_mm_vmscan_lru_isolate(sc->order, nr_to_scan, scan,
				    nr_taken, mode, is_file_lru(lru));
L
Linus Torvalds 已提交
1127 1128 1129
	return nr_taken;
}

1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140
/**
 * 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
1141 1142 1143
 * 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.
1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158
 *
 * 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;

1159 1160
	VM_BUG_ON(!page_count(page));

1161 1162
	if (PageLRU(page)) {
		struct zone *zone = page_zone(page);
1163
		struct lruvec *lruvec;
1164 1165

		spin_lock_irq(&zone->lru_lock);
1166
		lruvec = mem_cgroup_page_lruvec(page, zone);
1167
		if (PageLRU(page)) {
L
Lee Schermerhorn 已提交
1168
			int lru = page_lru(page);
1169
			get_page(page);
1170
			ClearPageLRU(page);
1171 1172
			del_page_from_lru_list(page, lruvec, lru);
			ret = 0;
1173 1174 1175 1176 1177 1178
		}
		spin_unlock_irq(&zone->lru_lock);
	}
	return ret;
}

1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189
/*
 * 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;

1190
	if (!global_reclaim(sc))
1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203
		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;
}

1204
static noinline_for_stack void
H
Hugh Dickins 已提交
1205
putback_inactive_pages(struct lruvec *lruvec, struct list_head *page_list)
1206
{
1207 1208
	struct zone_reclaim_stat *reclaim_stat = &lruvec->reclaim_stat;
	struct zone *zone = lruvec_zone(lruvec);
1209
	LIST_HEAD(pages_to_free);
1210 1211 1212 1213 1214

	/*
	 * Put back any unfreeable pages.
	 */
	while (!list_empty(page_list)) {
1215
		struct page *page = lru_to_page(page_list);
1216
		int lru;
1217

1218 1219
		VM_BUG_ON(PageLRU(page));
		list_del(&page->lru);
1220
		if (unlikely(!page_evictable(page))) {
1221 1222 1223 1224 1225
			spin_unlock_irq(&zone->lru_lock);
			putback_lru_page(page);
			spin_lock_irq(&zone->lru_lock);
			continue;
		}
1226 1227 1228

		lruvec = mem_cgroup_page_lruvec(page, zone);

1229
		SetPageLRU(page);
1230
		lru = page_lru(page);
1231 1232
		add_page_to_lru_list(page, lruvec, lru);

1233 1234
		if (is_active_lru(lru)) {
			int file = is_file_lru(lru);
1235 1236
			int numpages = hpage_nr_pages(page);
			reclaim_stat->recent_rotated[file] += numpages;
1237
		}
1238 1239 1240
		if (put_page_testzero(page)) {
			__ClearPageLRU(page);
			__ClearPageActive(page);
1241
			del_page_from_lru_list(page, lruvec, lru);
1242 1243 1244 1245 1246 1247 1248

			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);
1249 1250 1251
		}
	}

1252 1253 1254 1255
	/*
	 * To save our caller's stack, now use input list for pages to free.
	 */
	list_splice(&pages_to_free, page_list);
1256 1257
}

L
Linus Torvalds 已提交
1258
/*
A
Andrew Morton 已提交
1259 1260
 * shrink_inactive_list() is a helper for shrink_zone().  It returns the number
 * of reclaimed pages
L
Linus Torvalds 已提交
1261
 */
1262
static noinline_for_stack unsigned long
1263
shrink_inactive_list(unsigned long nr_to_scan, struct lruvec *lruvec,
1264
		     struct scan_control *sc, enum lru_list lru)
L
Linus Torvalds 已提交
1265 1266
{
	LIST_HEAD(page_list);
1267
	unsigned long nr_scanned;
1268
	unsigned long nr_reclaimed = 0;
1269
	unsigned long nr_taken;
1270 1271
	unsigned long nr_dirty = 0;
	unsigned long nr_writeback = 0;
1272
	isolate_mode_t isolate_mode = 0;
1273
	int file = is_file_lru(lru);
1274 1275
	struct zone *zone = lruvec_zone(lruvec);
	struct zone_reclaim_stat *reclaim_stat = &lruvec->reclaim_stat;
1276

1277
	while (unlikely(too_many_isolated(zone, file, sc))) {
1278
		congestion_wait(BLK_RW_ASYNC, HZ/10);
1279 1280 1281 1282 1283 1284

		/* 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
1285
	lru_add_drain();
1286 1287

	if (!sc->may_unmap)
1288
		isolate_mode |= ISOLATE_UNMAPPED;
1289
	if (!sc->may_writepage)
1290
		isolate_mode |= ISOLATE_CLEAN;
1291

L
Linus Torvalds 已提交
1292
	spin_lock_irq(&zone->lru_lock);
1293

1294 1295
	nr_taken = isolate_lru_pages(nr_to_scan, lruvec, &page_list,
				     &nr_scanned, sc, isolate_mode, lru);
1296 1297 1298 1299

	__mod_zone_page_state(zone, NR_LRU_BASE + lru, -nr_taken);
	__mod_zone_page_state(zone, NR_ISOLATED_ANON + file, nr_taken);

1300
	if (global_reclaim(sc)) {
1301 1302
		zone->pages_scanned += nr_scanned;
		if (current_is_kswapd())
H
Hugh Dickins 已提交
1303
			__count_zone_vm_events(PGSCAN_KSWAPD, zone, nr_scanned);
1304
		else
H
Hugh Dickins 已提交
1305
			__count_zone_vm_events(PGSCAN_DIRECT, zone, nr_scanned);
1306
	}
1307
	spin_unlock_irq(&zone->lru_lock);
1308

1309
	if (nr_taken == 0)
1310
		return 0;
A
Andy Whitcroft 已提交
1311

1312 1313
	nr_reclaimed = shrink_page_list(&page_list, zone, sc, TTU_UNMAP,
					&nr_dirty, &nr_writeback, false);
1314

1315 1316
	spin_lock_irq(&zone->lru_lock);

1317
	reclaim_stat->recent_scanned[file] += nr_taken;
1318

Y
Ying Han 已提交
1319 1320 1321 1322 1323 1324 1325 1326
	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
1327

1328
	putback_inactive_pages(lruvec, &page_list);
1329

1330
	__mod_zone_page_state(zone, NR_ISOLATED_ANON + file, -nr_taken);
1331 1332 1333 1334

	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
	/*
	 * 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
	 */
1359 1360
	if (nr_writeback && nr_writeback >=
			(nr_taken >> (DEF_PRIORITY - sc->priority)))
1361 1362
		wait_iff_congested(zone, BLK_RW_ASYNC, HZ/10);

1363 1364 1365
	trace_mm_vmscan_lru_shrink_inactive(zone->zone_pgdat->node_id,
		zone_idx(zone),
		nr_scanned, nr_reclaimed,
1366
		sc->priority,
M
Mel Gorman 已提交
1367
		trace_shrink_flags(file));
1368
	return nr_reclaimed;
L
Linus Torvalds 已提交
1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387
}

/*
 * 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.
 */
1388

1389
static void move_active_pages_to_lru(struct lruvec *lruvec,
1390
				     struct list_head *list,
1391
				     struct list_head *pages_to_free,
1392 1393
				     enum lru_list lru)
{
1394
	struct zone *zone = lruvec_zone(lruvec);
1395 1396
	unsigned long pgmoved = 0;
	struct page *page;
1397
	int nr_pages;
1398 1399 1400

	while (!list_empty(list)) {
		page = lru_to_page(list);
1401
		lruvec = mem_cgroup_page_lruvec(page, zone);
1402 1403 1404 1405

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

1406 1407
		nr_pages = hpage_nr_pages(page);
		mem_cgroup_update_lru_size(lruvec, lru, nr_pages);
1408
		list_move(&page->lru, &lruvec->lists[lru]);
1409
		pgmoved += nr_pages;
1410

1411 1412 1413
		if (put_page_testzero(page)) {
			__ClearPageLRU(page);
			__ClearPageActive(page);
1414
			del_page_from_lru_list(page, lruvec, lru);
1415 1416 1417 1418 1419 1420 1421

			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);
1422 1423 1424 1425 1426 1427
		}
	}
	__mod_zone_page_state(zone, NR_LRU_BASE + lru, pgmoved);
	if (!is_active_lru(lru))
		__count_vm_events(PGDEACTIVATE, pgmoved);
}
1428

H
Hugh Dickins 已提交
1429
static void shrink_active_list(unsigned long nr_to_scan,
1430
			       struct lruvec *lruvec,
1431
			       struct scan_control *sc,
1432
			       enum lru_list lru)
L
Linus Torvalds 已提交
1433
{
1434
	unsigned long nr_taken;
H
Hugh Dickins 已提交
1435
	unsigned long nr_scanned;
1436
	unsigned long vm_flags;
L
Linus Torvalds 已提交
1437
	LIST_HEAD(l_hold);	/* The pages which were snipped off */
1438
	LIST_HEAD(l_active);
1439
	LIST_HEAD(l_inactive);
L
Linus Torvalds 已提交
1440
	struct page *page;
1441
	struct zone_reclaim_stat *reclaim_stat = &lruvec->reclaim_stat;
1442
	unsigned long nr_rotated = 0;
1443
	isolate_mode_t isolate_mode = 0;
1444
	int file = is_file_lru(lru);
1445
	struct zone *zone = lruvec_zone(lruvec);
L
Linus Torvalds 已提交
1446 1447

	lru_add_drain();
1448 1449

	if (!sc->may_unmap)
1450
		isolate_mode |= ISOLATE_UNMAPPED;
1451
	if (!sc->may_writepage)
1452
		isolate_mode |= ISOLATE_CLEAN;
1453

L
Linus Torvalds 已提交
1454
	spin_lock_irq(&zone->lru_lock);
1455

1456 1457
	nr_taken = isolate_lru_pages(nr_to_scan, lruvec, &l_hold,
				     &nr_scanned, sc, isolate_mode, lru);
1458
	if (global_reclaim(sc))
H
Hugh Dickins 已提交
1459
		zone->pages_scanned += nr_scanned;
1460

1461
	reclaim_stat->recent_scanned[file] += nr_taken;
1462

H
Hugh Dickins 已提交
1463
	__count_zone_vm_events(PGREFILL, zone, nr_scanned);
1464
	__mod_zone_page_state(zone, NR_LRU_BASE + lru, -nr_taken);
K
KOSAKI Motohiro 已提交
1465
	__mod_zone_page_state(zone, NR_ISOLATED_ANON + file, nr_taken);
L
Linus Torvalds 已提交
1466 1467 1468 1469 1470 1471
	spin_unlock_irq(&zone->lru_lock);

	while (!list_empty(&l_hold)) {
		cond_resched();
		page = lru_to_page(&l_hold);
		list_del(&page->lru);
1472

1473
		if (unlikely(!page_evictable(page))) {
L
Lee Schermerhorn 已提交
1474 1475 1476 1477
			putback_lru_page(page);
			continue;
		}

1478 1479 1480 1481 1482 1483 1484 1485
		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);
			}
		}

1486 1487
		if (page_referenced(page, 0, sc->target_mem_cgroup,
				    &vm_flags)) {
1488
			nr_rotated += hpage_nr_pages(page);
1489 1490 1491 1492 1493 1494 1495 1496 1497
			/*
			 * 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.
			 */
1498
			if ((vm_flags & VM_EXEC) && page_is_file_cache(page)) {
1499 1500 1501 1502
				list_add(&page->lru, &l_active);
				continue;
			}
		}
1503

1504
		ClearPageActive(page);	/* we are de-activating */
L
Linus Torvalds 已提交
1505 1506 1507
		list_add(&page->lru, &l_inactive);
	}

1508
	/*
1509
	 * Move pages back to the lru list.
1510
	 */
1511
	spin_lock_irq(&zone->lru_lock);
1512
	/*
1513 1514 1515 1516
	 * 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.
1517
	 */
1518
	reclaim_stat->recent_rotated[file] += nr_rotated;
1519

1520 1521
	move_active_pages_to_lru(lruvec, &l_active, &l_hold, lru);
	move_active_pages_to_lru(lruvec, &l_inactive, &l_hold, lru - LRU_ACTIVE);
K
KOSAKI Motohiro 已提交
1522
	__mod_zone_page_state(zone, NR_ISOLATED_ANON + file, -nr_taken);
1523
	spin_unlock_irq(&zone->lru_lock);
1524 1525

	free_hot_cold_page_list(&l_hold, 1);
L
Linus Torvalds 已提交
1526 1527
}

1528
#ifdef CONFIG_SWAP
1529
static int inactive_anon_is_low_global(struct zone *zone)
1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541
{
	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;
}

1542 1543
/**
 * inactive_anon_is_low - check if anonymous pages need to be deactivated
1544
 * @lruvec: LRU vector to check
1545 1546 1547 1548
 *
 * Returns true if the zone does not have enough inactive anon pages,
 * meaning some active anon pages need to be deactivated.
 */
1549
static int inactive_anon_is_low(struct lruvec *lruvec)
1550
{
1551 1552 1553 1554 1555 1556 1557
	/*
	 * If we don't have swap space, anonymous page deactivation
	 * is pointless.
	 */
	if (!total_swap_pages)
		return 0;

1558
	if (!mem_cgroup_disabled())
1559
		return mem_cgroup_inactive_anon_is_low(lruvec);
1560

1561
	return inactive_anon_is_low_global(lruvec_zone(lruvec));
1562
}
1563
#else
1564
static inline int inactive_anon_is_low(struct lruvec *lruvec)
1565 1566 1567 1568
{
	return 0;
}
#endif
1569

1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581
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
1582
 * @lruvec: LRU vector to check
1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593
 *
 * 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.
 */
1594
static int inactive_file_is_low(struct lruvec *lruvec)
1595
{
1596
	if (!mem_cgroup_disabled())
1597
		return mem_cgroup_inactive_file_is_low(lruvec);
1598

1599
	return inactive_file_is_low_global(lruvec_zone(lruvec));
1600 1601
}

H
Hugh Dickins 已提交
1602
static int inactive_list_is_low(struct lruvec *lruvec, enum lru_list lru)
1603
{
H
Hugh Dickins 已提交
1604
	if (is_file_lru(lru))
1605
		return inactive_file_is_low(lruvec);
1606
	else
1607
		return inactive_anon_is_low(lruvec);
1608 1609
}

1610
static unsigned long shrink_list(enum lru_list lru, unsigned long nr_to_scan,
1611
				 struct lruvec *lruvec, struct scan_control *sc)
1612
{
1613
	if (is_active_lru(lru)) {
H
Hugh Dickins 已提交
1614
		if (inactive_list_is_low(lruvec, lru))
1615
			shrink_active_list(nr_to_scan, lruvec, sc, lru);
1616 1617 1618
		return 0;
	}

1619
	return shrink_inactive_list(nr_to_scan, lruvec, sc, lru);
1620 1621
}

1622
static int vmscan_swappiness(struct scan_control *sc)
1623
{
1624
	if (global_reclaim(sc))
1625
		return vm_swappiness;
1626
	return mem_cgroup_swappiness(sc->target_mem_cgroup);
1627 1628
}

1629 1630 1631 1632 1633 1634
/*
 * 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.
 *
W
Wanpeng Li 已提交
1635 1636
 * nr[0] = anon inactive pages to scan; nr[1] = anon active pages to scan
 * nr[2] = file inactive pages to scan; nr[3] = file active pages to scan
1637
 */
1638
static void get_scan_count(struct lruvec *lruvec, struct scan_control *sc,
1639
			   unsigned long *nr)
1640 1641 1642 1643
{
	unsigned long anon, file, free;
	unsigned long anon_prio, file_prio;
	unsigned long ap, fp;
1644
	struct zone_reclaim_stat *reclaim_stat = &lruvec->reclaim_stat;
1645
	u64 fraction[2], denominator;
H
Hugh Dickins 已提交
1646
	enum lru_list lru;
1647
	int noswap = 0;
1648
	bool force_scan = false;
1649
	struct zone *zone = lruvec_zone(lruvec);
1650

1651 1652 1653 1654 1655 1656 1657 1658 1659 1660
	/*
	 * 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.
	 */
1661
	if (current_is_kswapd() && zone->all_unreclaimable)
1662
		force_scan = true;
1663
	if (!global_reclaim(sc))
1664
		force_scan = true;
1665 1666 1667 1668 1669 1670 1671 1672 1673

	/* 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;
	}
1674

1675 1676 1677 1678
	anon  = get_lru_size(lruvec, LRU_ACTIVE_ANON) +
		get_lru_size(lruvec, LRU_INACTIVE_ANON);
	file  = get_lru_size(lruvec, LRU_ACTIVE_FILE) +
		get_lru_size(lruvec, LRU_INACTIVE_FILE);
1679

1680
	if (global_reclaim(sc)) {
1681
		free  = zone_page_state(zone, NR_FREE_PAGES);
1682 1683
		/* If we have very few page cache pages,
		   force-scan anon pages. */
1684
		if (unlikely(file + free <= high_wmark_pages(zone))) {
1685 1686 1687 1688
			fraction[0] = 1;
			fraction[1] = 0;
			denominator = 1;
			goto out;
1689
		}
1690 1691
	}

1692 1693 1694 1695
	/*
	 * With swappiness at 100, anonymous and file have the same priority.
	 * This scanning priority is essentially the inverse of IO cost.
	 */
1696
	anon_prio = vmscan_swappiness(sc);
H
Hugh Dickins 已提交
1697
	file_prio = 200 - anon_prio;
1698

1699 1700 1701 1702 1703 1704 1705 1706 1707 1708 1709
	/*
	 * 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]
	 */
1710
	spin_lock_irq(&zone->lru_lock);
1711 1712 1713
	if (unlikely(reclaim_stat->recent_scanned[0] > anon / 4)) {
		reclaim_stat->recent_scanned[0] /= 2;
		reclaim_stat->recent_rotated[0] /= 2;
1714 1715
	}

1716 1717 1718
	if (unlikely(reclaim_stat->recent_scanned[1] > file / 4)) {
		reclaim_stat->recent_scanned[1] /= 2;
		reclaim_stat->recent_rotated[1] /= 2;
1719 1720 1721
	}

	/*
1722 1723 1724
	 * 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.
1725
	 */
1726
	ap = anon_prio * (reclaim_stat->recent_scanned[0] + 1);
1727
	ap /= reclaim_stat->recent_rotated[0] + 1;
1728

1729
	fp = file_prio * (reclaim_stat->recent_scanned[1] + 1);
1730
	fp /= reclaim_stat->recent_rotated[1] + 1;
1731
	spin_unlock_irq(&zone->lru_lock);
1732

1733 1734 1735 1736
	fraction[0] = ap;
	fraction[1] = fp;
	denominator = ap + fp + 1;
out:
H
Hugh Dickins 已提交
1737 1738
	for_each_evictable_lru(lru) {
		int file = is_file_lru(lru);
1739
		unsigned long scan;
1740

1741
		scan = get_lru_size(lruvec, lru);
1742 1743
		if (sc->priority || noswap || !vmscan_swappiness(sc)) {
			scan >>= sc->priority;
1744 1745
			if (!scan && force_scan)
				scan = SWAP_CLUSTER_MAX;
1746 1747
			scan = div64_u64(scan * fraction[file], denominator);
		}
H
Hugh Dickins 已提交
1748
		nr[lru] = scan;
1749
	}
1750
}
1751

M
Mel Gorman 已提交
1752
/* Use reclaim/compaction for costly allocs or under memory pressure */
1753
static bool in_reclaim_compaction(struct scan_control *sc)
M
Mel Gorman 已提交
1754
{
1755
	if (IS_ENABLED(CONFIG_COMPACTION) && sc->order &&
M
Mel Gorman 已提交
1756
			(sc->order > PAGE_ALLOC_COSTLY_ORDER ||
1757
			 sc->priority < DEF_PRIORITY - 2))
M
Mel Gorman 已提交
1758 1759 1760 1761 1762
		return true;

	return false;
}

1763
/*
M
Mel Gorman 已提交
1764 1765 1766 1767 1768
 * 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.
1769
 */
1770
static inline bool should_continue_reclaim(struct lruvec *lruvec,
1771 1772 1773 1774 1775 1776 1777 1778
					unsigned long nr_reclaimed,
					unsigned long nr_scanned,
					struct scan_control *sc)
{
	unsigned long pages_for_compaction;
	unsigned long inactive_lru_pages;

	/* If not in reclaim/compaction mode, stop */
1779
	if (!in_reclaim_compaction(sc))
1780 1781
		return false;

1782 1783 1784 1785 1786 1787 1788 1789 1790 1791 1792 1793 1794 1795 1796 1797 1798 1799 1800 1801 1802 1803
	/* 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;
	}
1804 1805 1806 1807 1808 1809

	/*
	 * If we have not reclaimed enough pages for compaction and the
	 * inactive lists are large enough, continue reclaiming
	 */
	pages_for_compaction = (2UL << sc->order);
1810
	inactive_lru_pages = get_lru_size(lruvec, LRU_INACTIVE_FILE);
1811
	if (nr_swap_pages > 0)
1812
		inactive_lru_pages += get_lru_size(lruvec, LRU_INACTIVE_ANON);
1813 1814 1815 1816 1817
	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 */
1818
	switch (compaction_suitable(lruvec_zone(lruvec), sc->order)) {
1819 1820 1821 1822 1823 1824 1825 1826
	case COMPACT_PARTIAL:
	case COMPACT_CONTINUE:
		return false;
	default:
		return true;
	}
}

L
Linus Torvalds 已提交
1827 1828 1829
/*
 * This is a basic per-zone page freer.  Used by both kswapd and direct reclaim.
 */
1830
static void shrink_lruvec(struct lruvec *lruvec, struct scan_control *sc)
L
Linus Torvalds 已提交
1831
{
1832
	unsigned long nr[NR_LRU_LISTS];
1833
	unsigned long nr_to_scan;
H
Hugh Dickins 已提交
1834
	enum lru_list lru;
1835
	unsigned long nr_reclaimed, nr_scanned;
1836
	unsigned long nr_to_reclaim = sc->nr_to_reclaim;
1837
	struct blk_plug plug;
1838

1839 1840
restart:
	nr_reclaimed = 0;
1841
	nr_scanned = sc->nr_scanned;
1842
	get_scan_count(lruvec, sc, nr);
L
Linus Torvalds 已提交
1843

1844
	blk_start_plug(&plug);
1845 1846
	while (nr[LRU_INACTIVE_ANON] || nr[LRU_ACTIVE_FILE] ||
					nr[LRU_INACTIVE_FILE]) {
H
Hugh Dickins 已提交
1847 1848
		for_each_evictable_lru(lru) {
			if (nr[lru]) {
K
KOSAKI Motohiro 已提交
1849
				nr_to_scan = min_t(unsigned long,
H
Hugh Dickins 已提交
1850 1851
						   nr[lru], SWAP_CLUSTER_MAX);
				nr[lru] -= nr_to_scan;
L
Linus Torvalds 已提交
1852

H
Hugh Dickins 已提交
1853
				nr_reclaimed += shrink_list(lru, nr_to_scan,
1854
							    lruvec, sc);
1855
			}
L
Linus Torvalds 已提交
1856
		}
1857 1858 1859 1860 1861 1862 1863 1864
		/*
		 * 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.
		 */
1865 1866
		if (nr_reclaimed >= nr_to_reclaim &&
		    sc->priority < DEF_PRIORITY)
1867
			break;
L
Linus Torvalds 已提交
1868
	}
1869
	blk_finish_plug(&plug);
1870
	sc->nr_reclaimed += nr_reclaimed;
1871

1872 1873 1874 1875
	/*
	 * Even if we did not try to evict anon pages at all, we want to
	 * rebalance the anon lru active/inactive ratio.
	 */
1876
	if (inactive_anon_is_low(lruvec))
1877
		shrink_active_list(SWAP_CLUSTER_MAX, lruvec,
1878
				   sc, LRU_ACTIVE_ANON);
1879

1880
	/* reclaim/compaction might need reclaim to continue */
1881
	if (should_continue_reclaim(lruvec, nr_reclaimed,
1882
				    sc->nr_scanned - nr_scanned, sc))
1883 1884
		goto restart;

1885
	throttle_vm_writeout(sc->gfp_mask);
L
Linus Torvalds 已提交
1886 1887
}

1888
static void shrink_zone(struct zone *zone, struct scan_control *sc)
1889
{
1890 1891
	struct mem_cgroup *root = sc->target_mem_cgroup;
	struct mem_cgroup_reclaim_cookie reclaim = {
1892
		.zone = zone,
1893
		.priority = sc->priority,
1894
	};
1895 1896 1897 1898
	struct mem_cgroup *memcg;

	memcg = mem_cgroup_iter(root, NULL, &reclaim);
	do {
1899 1900 1901
		struct lruvec *lruvec = mem_cgroup_zone_lruvec(zone, memcg);

		shrink_lruvec(lruvec, sc);
1902

1903 1904 1905 1906 1907
		/*
		 * 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.
1908 1909 1910 1911
		 *
		 * Direct reclaim and kswapd, on the other hand, have
		 * to scan all memory cgroups to fulfill the overall
		 * scan target for the zone.
1912 1913 1914 1915 1916 1917 1918
		 */
		if (!global_reclaim(sc)) {
			mem_cgroup_iter_break(root, memcg);
			break;
		}
		memcg = mem_cgroup_iter(root, memcg, &reclaim);
	} while (memcg);
1919 1920
}

1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931 1932 1933 1934 1935 1936 1937 1938 1939 1940 1941 1942 1943 1944 1945 1946
/* 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
	 */
1947
	if (compaction_deferred(zone, sc->order))
1948 1949 1950 1951 1952 1953 1954 1955 1956
		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
1957 1958 1959 1960 1961
/*
 * 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.
 *
1962 1963
 * We reclaim from a zone even if that zone is over high_wmark_pages(zone).
 * Because:
L
Linus Torvalds 已提交
1964 1965
 * a) The caller may be trying to free *extra* pages to satisfy a higher-order
 *    allocation or
1966 1967 1968
 * 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
1969 1970 1971
 *
 * If a zone is deemed to be full of pinned pages then just give it a light
 * scan then give up on it.
1972 1973
 *
 * This function returns true if a zone is being reclaimed for a costly
1974
 * high-order allocation and compaction is ready to begin. This indicates to
1975 1976
 * the caller that it should consider retrying the allocation instead of
 * further reclaim.
L
Linus Torvalds 已提交
1977
 */
1978
static bool shrink_zones(struct zonelist *zonelist, struct scan_control *sc)
L
Linus Torvalds 已提交
1979
{
1980
	struct zoneref *z;
1981
	struct zone *zone;
1982 1983
	unsigned long nr_soft_reclaimed;
	unsigned long nr_soft_scanned;
1984
	bool aborted_reclaim = false;
1985

1986 1987 1988 1989 1990 1991 1992 1993
	/*
	 * 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;

1994 1995
	for_each_zone_zonelist_nodemask(zone, z, zonelist,
					gfp_zone(sc->gfp_mask), sc->nodemask) {
1996
		if (!populated_zone(zone))
L
Linus Torvalds 已提交
1997
			continue;
1998 1999 2000 2001
		/*
		 * Take care memory controller reclaiming has small influence
		 * to global LRU.
		 */
2002
		if (global_reclaim(sc)) {
2003 2004
			if (!cpuset_zone_allowed_hardwall(zone, GFP_KERNEL))
				continue;
2005 2006
			if (zone->all_unreclaimable &&
					sc->priority != DEF_PRIORITY)
2007
				continue;	/* Let kswapd poll it */
2008
			if (IS_ENABLED(CONFIG_COMPACTION)) {
2009
				/*
2010 2011 2012 2013 2014
				 * 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
2015 2016
				 * noticeable problem, like transparent huge
				 * page allocations.
2017
				 */
2018
				if (compaction_ready(zone, sc)) {
2019
					aborted_reclaim = true;
2020
					continue;
2021
				}
2022
			}
2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035
			/*
			 * 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() */
2036
		}
2037

2038
		shrink_zone(zone, sc);
L
Linus Torvalds 已提交
2039
	}
2040

2041
	return aborted_reclaim;
2042 2043 2044 2045 2046 2047 2048
}

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

2049
/* All zones in zonelist are unreclaimable? */
2050 2051 2052 2053 2054 2055 2056 2057 2058 2059 2060 2061
static bool all_unreclaimable(struct zonelist *zonelist,
		struct scan_control *sc)
{
	struct zoneref *z;
	struct zone *zone;

	for_each_zone_zonelist_nodemask(zone, z, zonelist,
			gfp_zone(sc->gfp_mask), sc->nodemask) {
		if (!populated_zone(zone))
			continue;
		if (!cpuset_zone_allowed_hardwall(zone, GFP_KERNEL))
			continue;
2062 2063
		if (!zone->all_unreclaimable)
			return false;
2064 2065
	}

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

L
Linus Torvalds 已提交
2069 2070 2071 2072 2073 2074 2075 2076
/*
 * This is the main entry point to direct page reclaim.
 *
 * If a full scan of the inactive list fails to free enough memory then we
 * are "out of memory" and something needs to be killed.
 *
 * If the caller is !__GFP_FS then the probability of a failure is reasonably
 * high - the zone may be full of dirty or under-writeback pages, which this
2077 2078 2079 2080
 * caller can't do much about.  We kick the writeback threads and take explicit
 * naps in the hope that some of these pages can be written.  But if the
 * allocating task holds filesystem locks which prevent writeout this might not
 * work, and the allocation attempt will fail.
2081 2082 2083
 *
 * returns:	0, if no pages reclaimed
 * 		else, the number of pages reclaimed
L
Linus Torvalds 已提交
2084
 */
2085
static unsigned long do_try_to_free_pages(struct zonelist *zonelist,
2086 2087
					struct scan_control *sc,
					struct shrink_control *shrink)
L
Linus Torvalds 已提交
2088
{
2089
	unsigned long total_scanned = 0;
L
Linus Torvalds 已提交
2090
	struct reclaim_state *reclaim_state = current->reclaim_state;
2091
	struct zoneref *z;
2092
	struct zone *zone;
2093
	unsigned long writeback_threshold;
2094
	bool aborted_reclaim;
L
Linus Torvalds 已提交
2095

2096 2097
	delayacct_freepages_start();

2098
	if (global_reclaim(sc))
2099
		count_vm_event(ALLOCSTALL);
L
Linus Torvalds 已提交
2100

2101
	do {
2102
		sc->nr_scanned = 0;
2103
		aborted_reclaim = shrink_zones(zonelist, sc);
2104

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

				lru_pages += zone_reclaimable_pages(zone);
			}

2119
			shrink_slab(shrink, sc->nr_scanned, lru_pages);
2120
			if (reclaim_state) {
2121
				sc->nr_reclaimed += reclaim_state->reclaimed_slab;
2122 2123
				reclaim_state->reclaimed_slab = 0;
			}
L
Linus Torvalds 已提交
2124
		}
2125
		total_scanned += sc->nr_scanned;
2126
		if (sc->nr_reclaimed >= sc->nr_to_reclaim)
L
Linus Torvalds 已提交
2127 2128 2129 2130 2131 2132 2133 2134 2135
			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.
		 */
2136 2137
		writeback_threshold = sc->nr_to_reclaim + sc->nr_to_reclaim / 2;
		if (total_scanned > writeback_threshold) {
2138 2139
			wakeup_flusher_threads(laptop_mode ? 0 : total_scanned,
						WB_REASON_TRY_TO_FREE_PAGES);
2140
			sc->may_writepage = 1;
L
Linus Torvalds 已提交
2141 2142 2143
		}

		/* Take a nap, wait for some writeback to complete */
2144
		if (!sc->hibernation_mode && sc->nr_scanned &&
2145
		    sc->priority < DEF_PRIORITY - 2) {
2146 2147 2148
			struct zone *preferred_zone;

			first_zones_zonelist(zonelist, gfp_zone(sc->gfp_mask),
2149 2150
						&cpuset_current_mems_allowed,
						&preferred_zone);
2151 2152
			wait_iff_congested(preferred_zone, BLK_RW_ASYNC, HZ/10);
		}
2153
	} while (--sc->priority >= 0);
2154

L
Linus Torvalds 已提交
2155
out:
2156 2157
	delayacct_freepages_end();

2158 2159 2160
	if (sc->nr_reclaimed)
		return sc->nr_reclaimed;

2161 2162 2163 2164 2165 2166 2167 2168
	/*
	 * 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;

2169 2170
	/* Aborted reclaim to try compaction? don't OOM, then */
	if (aborted_reclaim)
2171 2172
		return 1;

2173
	/* top priority shrink_zones still had more to do? don't OOM, then */
2174
	if (global_reclaim(sc) && !all_unreclaimable(zonelist, sc))
2175 2176 2177
		return 1;

	return 0;
L
Linus Torvalds 已提交
2178 2179
}

2180 2181 2182 2183 2184 2185 2186 2187 2188 2189 2190 2191 2192 2193 2194 2195 2196 2197 2198 2199 2200 2201 2202 2203 2204 2205 2206 2207 2208 2209
static bool pfmemalloc_watermark_ok(pg_data_t *pgdat)
{
	struct zone *zone;
	unsigned long pfmemalloc_reserve = 0;
	unsigned long free_pages = 0;
	int i;
	bool wmark_ok;

	for (i = 0; i <= ZONE_NORMAL; i++) {
		zone = &pgdat->node_zones[i];
		pfmemalloc_reserve += min_wmark_pages(zone);
		free_pages += zone_page_state(zone, NR_FREE_PAGES);
	}

	wmark_ok = free_pages > pfmemalloc_reserve / 2;

	/* kswapd must be awake if processes are being throttled */
	if (!wmark_ok && waitqueue_active(&pgdat->kswapd_wait)) {
		pgdat->classzone_idx = min(pgdat->classzone_idx,
						(enum zone_type)ZONE_NORMAL);
		wake_up_interruptible(&pgdat->kswapd_wait);
	}

	return wmark_ok;
}

/*
 * Throttle direct reclaimers if backing storage is backed by the network
 * and the PFMEMALLOC reserve for the preferred node is getting dangerously
 * depleted. kswapd will continue to make progress and wake the processes
2210 2211 2212 2213
 * when the low watermark is reached.
 *
 * Returns true if a fatal signal was delivered during throttling. If this
 * happens, the page allocator should not consider triggering the OOM killer.
2214
 */
2215
static bool throttle_direct_reclaim(gfp_t gfp_mask, struct zonelist *zonelist,
2216 2217 2218 2219 2220 2221 2222 2223 2224 2225 2226 2227 2228 2229
					nodemask_t *nodemask)
{
	struct zone *zone;
	int high_zoneidx = gfp_zone(gfp_mask);
	pg_data_t *pgdat;

	/*
	 * Kernel threads should not be throttled as they may be indirectly
	 * responsible for cleaning pages necessary for reclaim to make forward
	 * progress. kjournald for example may enter direct reclaim while
	 * committing a transaction where throttling it could forcing other
	 * processes to block on log_wait_commit().
	 */
	if (current->flags & PF_KTHREAD)
2230 2231 2232 2233 2234 2235 2236 2237
		goto out;

	/*
	 * If a fatal signal is pending, this process should not throttle.
	 * It should return quickly so it can exit and free its memory
	 */
	if (fatal_signal_pending(current))
		goto out;
2238 2239 2240 2241 2242

	/* Check if the pfmemalloc reserves are ok */
	first_zones_zonelist(zonelist, high_zoneidx, NULL, &zone);
	pgdat = zone->zone_pgdat;
	if (pfmemalloc_watermark_ok(pgdat))
2243
		goto out;
2244

2245 2246 2247
	/* Account for the throttling */
	count_vm_event(PGSCAN_DIRECT_THROTTLE);

2248 2249 2250 2251 2252 2253 2254 2255 2256 2257 2258
	/*
	 * If the caller cannot enter the filesystem, it's possible that it
	 * is due to the caller holding an FS lock or performing a journal
	 * transaction in the case of a filesystem like ext[3|4]. In this case,
	 * it is not safe to block on pfmemalloc_wait as kswapd could be
	 * blocked waiting on the same lock. Instead, throttle for up to a
	 * second before continuing.
	 */
	if (!(gfp_mask & __GFP_FS)) {
		wait_event_interruptible_timeout(pgdat->pfmemalloc_wait,
			pfmemalloc_watermark_ok(pgdat), HZ);
2259 2260

		goto check_pending;
2261 2262 2263 2264 2265
	}

	/* Throttle until kswapd wakes the process */
	wait_event_killable(zone->zone_pgdat->pfmemalloc_wait,
		pfmemalloc_watermark_ok(pgdat));
2266 2267 2268 2269 2270 2271 2272

check_pending:
	if (fatal_signal_pending(current))
		return true;

out:
	return false;
2273 2274
}

2275
unsigned long try_to_free_pages(struct zonelist *zonelist, int order,
2276
				gfp_t gfp_mask, nodemask_t *nodemask)
2277
{
2278
	unsigned long nr_reclaimed;
2279 2280 2281
	struct scan_control sc = {
		.gfp_mask = gfp_mask,
		.may_writepage = !laptop_mode,
2282
		.nr_to_reclaim = SWAP_CLUSTER_MAX,
2283
		.may_unmap = 1,
2284
		.may_swap = 1,
2285
		.order = order,
2286
		.priority = DEF_PRIORITY,
2287
		.target_mem_cgroup = NULL,
2288
		.nodemask = nodemask,
2289
	};
2290 2291 2292
	struct shrink_control shrink = {
		.gfp_mask = sc.gfp_mask,
	};
2293

2294
	/*
2295 2296 2297
	 * Do not enter reclaim if fatal signal was delivered while throttled.
	 * 1 is returned so that the page allocator does not OOM kill at this
	 * point.
2298
	 */
2299
	if (throttle_direct_reclaim(gfp_mask, zonelist, nodemask))
2300 2301
		return 1;

2302 2303 2304 2305
	trace_mm_vmscan_direct_reclaim_begin(order,
				sc.may_writepage,
				gfp_mask);

2306
	nr_reclaimed = do_try_to_free_pages(zonelist, &sc, &shrink);
2307 2308 2309 2310

	trace_mm_vmscan_direct_reclaim_end(nr_reclaimed);

	return nr_reclaimed;
2311 2312
}

A
Andrew Morton 已提交
2313
#ifdef CONFIG_MEMCG
2314

2315
unsigned long mem_cgroup_shrink_node_zone(struct mem_cgroup *memcg,
2316
						gfp_t gfp_mask, bool noswap,
2317 2318
						struct zone *zone,
						unsigned long *nr_scanned)
2319 2320
{
	struct scan_control sc = {
2321
		.nr_scanned = 0,
2322
		.nr_to_reclaim = SWAP_CLUSTER_MAX,
2323 2324 2325 2326
		.may_writepage = !laptop_mode,
		.may_unmap = 1,
		.may_swap = !noswap,
		.order = 0,
2327
		.priority = 0,
2328
		.target_mem_cgroup = memcg,
2329
	};
2330
	struct lruvec *lruvec = mem_cgroup_zone_lruvec(zone, memcg);
2331

2332 2333
	sc.gfp_mask = (gfp_mask & GFP_RECLAIM_MASK) |
			(GFP_HIGHUSER_MOVABLE & ~GFP_RECLAIM_MASK);
2334

2335
	trace_mm_vmscan_memcg_softlimit_reclaim_begin(sc.order,
2336 2337 2338
						      sc.may_writepage,
						      sc.gfp_mask);

2339 2340 2341 2342 2343 2344 2345
	/*
	 * 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.
	 */
2346
	shrink_lruvec(lruvec, &sc);
2347 2348 2349

	trace_mm_vmscan_memcg_softlimit_reclaim_end(sc.nr_reclaimed);

2350
	*nr_scanned = sc.nr_scanned;
2351 2352 2353
	return sc.nr_reclaimed;
}

2354
unsigned long try_to_free_mem_cgroup_pages(struct mem_cgroup *memcg,
K
KOSAKI Motohiro 已提交
2355
					   gfp_t gfp_mask,
2356
					   bool noswap)
2357
{
2358
	struct zonelist *zonelist;
2359
	unsigned long nr_reclaimed;
2360
	int nid;
2361 2362
	struct scan_control sc = {
		.may_writepage = !laptop_mode,
2363
		.may_unmap = 1,
2364
		.may_swap = !noswap,
2365
		.nr_to_reclaim = SWAP_CLUSTER_MAX,
2366
		.order = 0,
2367
		.priority = DEF_PRIORITY,
2368
		.target_mem_cgroup = memcg,
2369
		.nodemask = NULL, /* we don't care the placement */
2370 2371 2372 2373 2374
		.gfp_mask = (gfp_mask & GFP_RECLAIM_MASK) |
				(GFP_HIGHUSER_MOVABLE & ~GFP_RECLAIM_MASK),
	};
	struct shrink_control shrink = {
		.gfp_mask = sc.gfp_mask,
2375 2376
	};

2377 2378 2379 2380 2381
	/*
	 * 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.
	 */
2382
	nid = mem_cgroup_select_victim_node(memcg);
2383 2384

	zonelist = NODE_DATA(nid)->node_zonelists;
2385 2386 2387 2388 2389

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

2390
	nr_reclaimed = do_try_to_free_pages(zonelist, &sc, &shrink);
2391 2392 2393 2394

	trace_mm_vmscan_memcg_reclaim_end(nr_reclaimed);

	return nr_reclaimed;
2395 2396 2397
}
#endif

2398
static void age_active_anon(struct zone *zone, struct scan_control *sc)
2399
{
2400
	struct mem_cgroup *memcg;
2401

2402 2403 2404 2405 2406
	if (!total_swap_pages)
		return;

	memcg = mem_cgroup_iter(NULL, NULL, NULL);
	do {
2407
		struct lruvec *lruvec = mem_cgroup_zone_lruvec(zone, memcg);
2408

2409
		if (inactive_anon_is_low(lruvec))
2410
			shrink_active_list(SWAP_CLUSTER_MAX, lruvec,
2411
					   sc, LRU_ACTIVE_ANON);
2412 2413 2414

		memcg = mem_cgroup_iter(NULL, memcg, NULL);
	} while (memcg);
2415 2416
}

2417 2418 2419 2420 2421 2422 2423
static bool zone_balanced(struct zone *zone, int order,
			  unsigned long balance_gap, int classzone_idx)
{
	if (!zone_watermark_ok_safe(zone, order, high_wmark_pages(zone) +
				    balance_gap, classzone_idx, 0))
		return false;

2424 2425
	if (IS_ENABLED(CONFIG_COMPACTION) && order &&
	    !compaction_suitable(zone, order))
2426 2427 2428 2429 2430
		return false;

	return true;
}

2431 2432 2433 2434 2435 2436 2437 2438 2439 2440 2441
/*
 * 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
2442
 *     percentage of the middle zones. For example, on 32-bit x86, highmem
2443 2444 2445 2446 2447 2448 2449 2450 2451 2452 2453 2454 2455
 *     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
2456 2457
	/* A special case here: if zone has no page, we think it's balanced */
	return balanced_pages >= (present_pages >> 2);
2458 2459
}

2460 2461 2462 2463 2464 2465 2466
/*
 * Prepare kswapd for sleeping. This verifies that there are no processes
 * waiting in throttle_direct_reclaim() and that watermarks have been met.
 *
 * Returns true if kswapd is ready to sleep
 */
static bool prepare_kswapd_sleep(pg_data_t *pgdat, int order, long remaining,
2467
					int classzone_idx)
2468
{
2469
	int i;
2470 2471
	unsigned long balanced = 0;
	bool all_zones_ok = true;
2472 2473 2474

	/* If a direct reclaimer woke kswapd within HZ/10, it's premature */
	if (remaining)
2475 2476 2477 2478 2479 2480 2481 2482 2483 2484 2485 2486 2487 2488 2489
		return false;

	/*
	 * There is a potential race between when kswapd checks its watermarks
	 * and a process gets throttled. There is also a potential race if
	 * processes get throttled, kswapd wakes, a large process exits therby
	 * balancing the zones that causes kswapd to miss a wakeup. If kswapd
	 * is going to sleep, no process should be sleeping on pfmemalloc_wait
	 * so wake them now if necessary. If necessary, processes will wake
	 * kswapd and get throttled again
	 */
	if (waitqueue_active(&pgdat->pfmemalloc_wait)) {
		wake_up(&pgdat->pfmemalloc_wait);
		return false;
	}
2490

2491
	/* Check the watermark levels */
2492
	for (i = 0; i <= classzone_idx; i++) {
2493 2494 2495 2496 2497
		struct zone *zone = pgdat->node_zones + i;

		if (!populated_zone(zone))
			continue;

2498 2499 2500 2501 2502 2503 2504 2505
		/*
		 * 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;
2506
			continue;
2507
		}
2508

2509
		if (!zone_balanced(zone, order, 0, i))
2510 2511 2512
			all_zones_ok = false;
		else
			balanced += zone->present_pages;
2513
	}
2514

2515 2516 2517 2518 2519 2520
	/*
	 * 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)
2521
		return pgdat_balanced(pgdat, balanced, classzone_idx);
2522
	else
2523
		return all_zones_ok;
2524 2525
}

L
Linus Torvalds 已提交
2526 2527
/*
 * For kswapd, balance_pgdat() will work across all this node's zones until
2528
 * they are all at high_wmark_pages(zone).
L
Linus Torvalds 已提交
2529
 *
2530
 * Returns the final order kswapd was reclaiming at
L
Linus Torvalds 已提交
2531 2532 2533 2534 2535 2536 2537 2538 2539 2540
 *
 * 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
2541 2542 2543 2544 2545
 * 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
2546
 */
2547
static unsigned long balance_pgdat(pg_data_t *pgdat, int order,
2548
							int *classzone_idx)
L
Linus Torvalds 已提交
2549 2550
{
	int all_zones_ok;
2551
	unsigned long balanced;
L
Linus Torvalds 已提交
2552
	int i;
2553
	int end_zone = 0;	/* Inclusive.  0 = ZONE_DMA */
2554
	unsigned long total_scanned;
L
Linus Torvalds 已提交
2555
	struct reclaim_state *reclaim_state = current->reclaim_state;
2556 2557
	unsigned long nr_soft_reclaimed;
	unsigned long nr_soft_scanned;
2558 2559
	struct scan_control sc = {
		.gfp_mask = GFP_KERNEL,
2560
		.may_unmap = 1,
2561
		.may_swap = 1,
2562 2563 2564 2565 2566
		/*
		 * 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
2567
		.order = order,
2568
		.target_mem_cgroup = NULL,
2569
	};
2570 2571 2572
	struct shrink_control shrink = {
		.gfp_mask = sc.gfp_mask,
	};
L
Linus Torvalds 已提交
2573 2574
loop_again:
	total_scanned = 0;
2575
	sc.priority = DEF_PRIORITY;
2576
	sc.nr_reclaimed = 0;
C
Christoph Lameter 已提交
2577
	sc.may_writepage = !laptop_mode;
2578
	count_vm_event(PAGEOUTRUN);
L
Linus Torvalds 已提交
2579

2580
	do {
L
Linus Torvalds 已提交
2581
		unsigned long lru_pages = 0;
2582
		int has_under_min_watermark_zone = 0;
L
Linus Torvalds 已提交
2583 2584

		all_zones_ok = 1;
2585
		balanced = 0;
L
Linus Torvalds 已提交
2586

2587 2588 2589 2590 2591 2592
		/*
		 * 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
2593

2594 2595
			if (!populated_zone(zone))
				continue;
L
Linus Torvalds 已提交
2596

2597 2598
			if (zone->all_unreclaimable &&
			    sc.priority != DEF_PRIORITY)
2599
				continue;
L
Linus Torvalds 已提交
2600

2601 2602 2603 2604
			/*
			 * Do some background aging of the anon list, to give
			 * pages a chance to be referenced before reclaiming.
			 */
2605
			age_active_anon(zone, &sc);
2606

2607 2608 2609 2610 2611 2612 2613 2614 2615 2616 2617
			/*
			 * 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;
			}

2618
			if (!zone_balanced(zone, order, 0, 0)) {
2619
				end_zone = i;
A
Andrew Morton 已提交
2620
				break;
2621 2622 2623
			} else {
				/* If balanced, clear the congested flag */
				zone_clear_flag(zone, ZONE_CONGESTED);
L
Linus Torvalds 已提交
2624 2625
			}
		}
A
Andrew Morton 已提交
2626 2627 2628
		if (i < 0)
			goto out;

L
Linus Torvalds 已提交
2629 2630 2631
		for (i = 0; i <= end_zone; i++) {
			struct zone *zone = pgdat->node_zones + i;

2632
			lru_pages += zone_reclaimable_pages(zone);
L
Linus Torvalds 已提交
2633 2634 2635 2636 2637 2638 2639 2640 2641 2642 2643 2644 2645
		}

		/*
		 * 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;
2646
			int nr_slab, testorder;
2647
			unsigned long balance_gap;
L
Linus Torvalds 已提交
2648

2649
			if (!populated_zone(zone))
L
Linus Torvalds 已提交
2650 2651
				continue;

2652 2653
			if (zone->all_unreclaimable &&
			    sc.priority != DEF_PRIORITY)
L
Linus Torvalds 已提交
2654 2655 2656
				continue;

			sc.nr_scanned = 0;
2657

2658
			nr_soft_scanned = 0;
2659 2660 2661
			/*
			 * Call soft limit reclaim before calling shrink_zone.
			 */
2662 2663 2664 2665 2666
			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;
2667

2668
			/*
2669 2670 2671 2672 2673 2674
			 * 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.
2675
			 */
2676 2677 2678 2679
			balance_gap = min(low_wmark_pages(zone),
				(zone->present_pages +
					KSWAPD_ZONE_BALANCE_GAP_RATIO-1) /
				KSWAPD_ZONE_BALANCE_GAP_RATIO);
2680 2681 2682 2683 2684 2685 2686 2687
			/*
			 * 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;
2688
			if (IS_ENABLED(CONFIG_COMPACTION) && order &&
2689 2690 2691 2692
					compaction_suitable(zone, order) !=
						COMPACT_SKIPPED)
				testorder = 0;

2693
			if ((buffer_heads_over_limit && is_highmem_idx(i)) ||
2694 2695
			    !zone_balanced(zone, testorder,
					   balance_gap, end_zone)) {
2696
				shrink_zone(zone, &sc);
2697

2698 2699 2700 2701 2702 2703 2704 2705 2706
				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
2707 2708 2709 2710 2711 2712
			/*
			 * 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 &&
2713
			    total_scanned > sc.nr_reclaimed + sc.nr_reclaimed / 2)
L
Linus Torvalds 已提交
2714
				sc.may_writepage = 1;
2715

2716 2717 2718
			if (zone->all_unreclaimable) {
				if (end_zone && end_zone == i)
					end_zone--;
2719
				continue;
2720
			}
2721

2722
			if (!zone_balanced(zone, testorder, 0, end_zone)) {
2723 2724 2725 2726 2727 2728
				all_zones_ok = 0;
				/*
				 * We are still under min water mark.  This
				 * means that we have a GFP_ATOMIC allocation
				 * failure risk. Hurry up!
				 */
2729
				if (!zone_watermark_ok_safe(zone, order,
2730 2731
					    min_wmark_pages(zone), end_zone, 0))
					has_under_min_watermark_zone = 1;
2732 2733 2734 2735 2736 2737
			} 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,
2738
				 * speculatively avoid congestion waits
2739 2740
				 */
				zone_clear_flag(zone, ZONE_CONGESTED);
2741
				if (i <= *classzone_idx)
2742
					balanced += zone->present_pages;
2743
			}
2744

L
Linus Torvalds 已提交
2745
		}
2746 2747 2748 2749 2750 2751 2752 2753 2754 2755

		/*
		 * If the low watermark is met there is no need for processes
		 * to be throttled on pfmemalloc_wait as they should not be
		 * able to safely make forward progress. Wake them
		 */
		if (waitqueue_active(&pgdat->pfmemalloc_wait) &&
				pfmemalloc_watermark_ok(pgdat))
			wake_up(&pgdat->pfmemalloc_wait);

2756
		if (all_zones_ok || (order && pgdat_balanced(pgdat, balanced, *classzone_idx)))
L
Linus Torvalds 已提交
2757 2758 2759 2760 2761
			break;		/* kswapd: all done */
		/*
		 * OK, kswapd is getting into trouble.  Take a nap, then take
		 * another pass across the zones.
		 */
2762
		if (total_scanned && (sc.priority < DEF_PRIORITY - 2)) {
2763 2764 2765 2766 2767
			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
2768 2769 2770 2771 2772 2773 2774

		/*
		 * 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.
		 */
2775
		if (sc.nr_reclaimed >= SWAP_CLUSTER_MAX)
L
Linus Torvalds 已提交
2776
			break;
2777
	} while (--sc.priority >= 0);
L
Linus Torvalds 已提交
2778
out:
2779 2780 2781

	/*
	 * order-0: All zones must meet high watermark for a balanced node
2782 2783
	 * high-order: Balanced zones must make up at least 25% of the node
	 *             for the node to be balanced
2784
	 */
2785
	if (!(all_zones_ok || (order && pgdat_balanced(pgdat, balanced, *classzone_idx)))) {
L
Linus Torvalds 已提交
2786
		cond_resched();
2787 2788 2789

		try_to_freeze();

2790 2791 2792 2793 2794 2795 2796 2797 2798 2799 2800 2801 2802 2803 2804 2805 2806
		/*
		 * 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
2807 2808 2809
		goto loop_again;
	}

2810 2811 2812 2813 2814 2815 2816 2817 2818
	/*
	 * 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) {
2819 2820
		int zones_need_compaction = 1;

2821 2822 2823 2824 2825 2826
		for (i = 0; i <= end_zone; i++) {
			struct zone *zone = pgdat->node_zones + i;

			if (!populated_zone(zone))
				continue;

2827 2828 2829 2830
			/* 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;
2831
		}
2832 2833 2834

		if (zones_need_compaction)
			compact_pgdat(pgdat, order);
2835 2836
	}

2837
	/*
2838
	 * Return the order we were reclaiming at so prepare_kswapd_sleep()
2839 2840 2841 2842
	 * 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
	 */
2843
	*classzone_idx = end_zone;
2844
	return order;
L
Linus Torvalds 已提交
2845 2846
}

2847
static void kswapd_try_to_sleep(pg_data_t *pgdat, int order, int classzone_idx)
2848 2849 2850 2851 2852 2853 2854 2855 2856 2857
{
	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 */
2858
	if (prepare_kswapd_sleep(pgdat, order, remaining, classzone_idx)) {
2859 2860 2861 2862 2863 2864 2865 2866 2867
		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.
	 */
2868
	if (prepare_kswapd_sleep(pgdat, order, remaining, classzone_idx)) {
2869 2870 2871 2872 2873 2874 2875 2876 2877 2878 2879
		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);
2880

2881 2882 2883 2884 2885 2886 2887 2888
		/*
		 * Compaction records what page blocks it recently failed to
		 * isolate pages from and skips them in the future scanning.
		 * When kswapd is going to sleep, it is reasonable to assume
		 * that pages and compaction may succeed so reset the cache.
		 */
		reset_isolation_suitable(pgdat);

2889 2890 2891
		if (!kthread_should_stop())
			schedule();

2892 2893 2894 2895 2896 2897 2898 2899 2900 2901
		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
2902 2903
/*
 * The background pageout daemon, started as a kernel thread
2904
 * from the init process.
L
Linus Torvalds 已提交
2905 2906 2907 2908 2909 2910 2911 2912 2913 2914 2915 2916
 *
 * 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)
{
2917
	unsigned long order, new_order;
2918
	unsigned balanced_order;
2919
	int classzone_idx, new_classzone_idx;
2920
	int balanced_classzone_idx;
L
Linus Torvalds 已提交
2921 2922
	pg_data_t *pgdat = (pg_data_t*)p;
	struct task_struct *tsk = current;
2923

L
Linus Torvalds 已提交
2924 2925 2926
	struct reclaim_state reclaim_state = {
		.reclaimed_slab = 0,
	};
2927
	const struct cpumask *cpumask = cpumask_of_node(pgdat->node_id);
L
Linus Torvalds 已提交
2928

2929 2930
	lockdep_set_current_reclaim_state(GFP_KERNEL);

R
Rusty Russell 已提交
2931
	if (!cpumask_empty(cpumask))
2932
		set_cpus_allowed_ptr(tsk, cpumask);
L
Linus Torvalds 已提交
2933 2934 2935 2936 2937 2938 2939 2940 2941 2942 2943 2944 2945 2946
	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).
	 */
2947
	tsk->flags |= PF_MEMALLOC | PF_SWAPWRITE | PF_KSWAPD;
2948
	set_freezable();
L
Linus Torvalds 已提交
2949

2950
	order = new_order = 0;
2951
	balanced_order = 0;
2952
	classzone_idx = new_classzone_idx = pgdat->nr_zones - 1;
2953
	balanced_classzone_idx = classzone_idx;
L
Linus Torvalds 已提交
2954
	for ( ; ; ) {
2955
		int ret;
2956

2957 2958 2959 2960 2961
		/*
		 * 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
		 */
2962 2963
		if (balanced_classzone_idx >= new_classzone_idx &&
					balanced_order == new_order) {
2964 2965 2966 2967 2968 2969
			new_order = pgdat->kswapd_max_order;
			new_classzone_idx = pgdat->classzone_idx;
			pgdat->kswapd_max_order =  0;
			pgdat->classzone_idx = pgdat->nr_zones - 1;
		}

2970
		if (order < new_order || classzone_idx > new_classzone_idx) {
L
Linus Torvalds 已提交
2971 2972
			/*
			 * Don't sleep if someone wants a larger 'order'
2973
			 * allocation or has tigher zone constraints
L
Linus Torvalds 已提交
2974 2975
			 */
			order = new_order;
2976
			classzone_idx = new_classzone_idx;
L
Linus Torvalds 已提交
2977
		} else {
2978 2979
			kswapd_try_to_sleep(pgdat, balanced_order,
						balanced_classzone_idx);
L
Linus Torvalds 已提交
2980
			order = pgdat->kswapd_max_order;
2981
			classzone_idx = pgdat->classzone_idx;
2982 2983
			new_order = order;
			new_classzone_idx = classzone_idx;
2984
			pgdat->kswapd_max_order = 0;
2985
			pgdat->classzone_idx = pgdat->nr_zones - 1;
L
Linus Torvalds 已提交
2986 2987
		}

2988 2989 2990 2991 2992 2993 2994 2995
		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
		 */
2996 2997
		if (!ret) {
			trace_mm_vmscan_kswapd_wake(pgdat->node_id, order);
2998 2999 3000
			balanced_classzone_idx = classzone_idx;
			balanced_order = balance_pgdat(pgdat, order,
						&balanced_classzone_idx);
3001
		}
L
Linus Torvalds 已提交
3002
	}
3003 3004

	current->reclaim_state = NULL;
L
Linus Torvalds 已提交
3005 3006 3007 3008 3009 3010
	return 0;
}

/*
 * A zone is low on free memory, so wake its kswapd task to service it.
 */
3011
void wakeup_kswapd(struct zone *zone, int order, enum zone_type classzone_idx)
L
Linus Torvalds 已提交
3012 3013 3014
{
	pg_data_t *pgdat;

3015
	if (!populated_zone(zone))
L
Linus Torvalds 已提交
3016 3017
		return;

3018
	if (!cpuset_zone_allowed_hardwall(zone, GFP_KERNEL))
L
Linus Torvalds 已提交
3019
		return;
3020
	pgdat = zone->zone_pgdat;
3021
	if (pgdat->kswapd_max_order < order) {
L
Linus Torvalds 已提交
3022
		pgdat->kswapd_max_order = order;
3023 3024
		pgdat->classzone_idx = min(pgdat->classzone_idx, classzone_idx);
	}
3025
	if (!waitqueue_active(&pgdat->kswapd_wait))
L
Linus Torvalds 已提交
3026
		return;
3027 3028 3029 3030
	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);
3031
	wake_up_interruptible(&pgdat->kswapd_wait);
L
Linus Torvalds 已提交
3032 3033
}

3034 3035 3036 3037 3038 3039 3040 3041
/*
 * 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)
3042
{
3043 3044 3045 3046 3047 3048 3049 3050 3051 3052 3053 3054 3055 3056 3057 3058 3059 3060 3061 3062 3063 3064 3065 3066
	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;
3067 3068
}

3069
#ifdef CONFIG_HIBERNATION
L
Linus Torvalds 已提交
3070
/*
3071
 * Try to free `nr_to_reclaim' of memory, system-wide, and return the number of
3072 3073 3074 3075 3076
 * 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
3077
 */
3078
unsigned long shrink_all_memory(unsigned long nr_to_reclaim)
L
Linus Torvalds 已提交
3079
{
3080 3081
	struct reclaim_state reclaim_state;
	struct scan_control sc = {
3082 3083 3084
		.gfp_mask = GFP_HIGHUSER_MOVABLE,
		.may_swap = 1,
		.may_unmap = 1,
3085
		.may_writepage = 1,
3086 3087 3088
		.nr_to_reclaim = nr_to_reclaim,
		.hibernation_mode = 1,
		.order = 0,
3089
		.priority = DEF_PRIORITY,
L
Linus Torvalds 已提交
3090
	};
3091 3092 3093 3094
	struct shrink_control shrink = {
		.gfp_mask = sc.gfp_mask,
	};
	struct zonelist *zonelist = node_zonelist(numa_node_id(), sc.gfp_mask);
3095 3096
	struct task_struct *p = current;
	unsigned long nr_reclaimed;
L
Linus Torvalds 已提交
3097

3098 3099 3100 3101
	p->flags |= PF_MEMALLOC;
	lockdep_set_current_reclaim_state(sc.gfp_mask);
	reclaim_state.reclaimed_slab = 0;
	p->reclaim_state = &reclaim_state;
3102

3103
	nr_reclaimed = do_try_to_free_pages(zonelist, &sc, &shrink);
3104

3105 3106 3107
	p->reclaim_state = NULL;
	lockdep_clear_current_reclaim_state();
	p->flags &= ~PF_MEMALLOC;
3108

3109
	return nr_reclaimed;
L
Linus Torvalds 已提交
3110
}
3111
#endif /* CONFIG_HIBERNATION */
L
Linus Torvalds 已提交
3112 3113 3114 3115 3116

/* 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. */
3117
static int __devinit cpu_callback(struct notifier_block *nfb,
3118
				  unsigned long action, void *hcpu)
L
Linus Torvalds 已提交
3119
{
3120
	int nid;
L
Linus Torvalds 已提交
3121

3122
	if (action == CPU_ONLINE || action == CPU_ONLINE_FROZEN) {
3123
		for_each_node_state(nid, N_HIGH_MEMORY) {
3124
			pg_data_t *pgdat = NODE_DATA(nid);
3125 3126 3127
			const struct cpumask *mask;

			mask = cpumask_of_node(pgdat->node_id);
3128

3129
			if (cpumask_any_and(cpu_online_mask, mask) < nr_cpu_ids)
L
Linus Torvalds 已提交
3130
				/* One of our CPUs online: restore mask */
3131
				set_cpus_allowed_ptr(pgdat->kswapd, mask);
L
Linus Torvalds 已提交
3132 3133 3134 3135 3136
		}
	}
	return NOTIFY_OK;
}

3137 3138 3139 3140 3141 3142 3143 3144 3145 3146 3147 3148 3149 3150 3151 3152
/*
 * 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);
3153
		pgdat->kswapd = NULL;
3154 3155
		pr_err("Failed to start kswapd on node %d\n", nid);
		ret = PTR_ERR(pgdat->kswapd);
3156 3157 3158 3159
	}
	return ret;
}

3160
/*
3161 3162
 * Called by memory hotplug when all memory in a node is offlined.  Caller must
 * hold lock_memory_hotplug().
3163 3164 3165 3166 3167
 */
void kswapd_stop(int nid)
{
	struct task_struct *kswapd = NODE_DATA(nid)->kswapd;

3168
	if (kswapd) {
3169
		kthread_stop(kswapd);
3170 3171
		NODE_DATA(nid)->kswapd = NULL;
	}
3172 3173
}

L
Linus Torvalds 已提交
3174 3175
static int __init kswapd_init(void)
{
3176
	int nid;
3177

L
Linus Torvalds 已提交
3178
	swap_setup();
3179
	for_each_node_state(nid, N_HIGH_MEMORY)
3180
 		kswapd_run(nid);
L
Linus Torvalds 已提交
3181 3182 3183 3184 3185
	hotcpu_notifier(cpu_callback, 0);
	return 0;
}

module_init(kswapd_init)
3186 3187 3188 3189 3190 3191 3192 3193 3194 3195

#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;

3196
#define RECLAIM_OFF 0
3197
#define RECLAIM_ZONE (1<<0)	/* Run shrink_inactive_list on the zone */
3198 3199 3200
#define RECLAIM_WRITE (1<<1)	/* Writeout pages during reclaim */
#define RECLAIM_SWAP (1<<2)	/* Swap pages out during reclaim */

3201 3202 3203 3204 3205 3206 3207
/*
 * 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

3208 3209 3210 3211 3212 3213
/*
 * Percentage of pages in a zone that must be unmapped for zone_reclaim to
 * occur.
 */
int sysctl_min_unmapped_ratio = 1;

3214 3215 3216 3217 3218 3219
/*
 * 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;

3220 3221 3222 3223 3224 3225 3226 3227 3228 3229 3230 3231 3232 3233 3234 3235 3236 3237 3238 3239 3240 3241 3242 3243 3244 3245 3246 3247 3248 3249 3250 3251 3252 3253 3254 3255 3256 3257 3258 3259 3260 3261
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;
}

3262 3263 3264
/*
 * Try to free up some pages from this zone through reclaim.
 */
3265
static int __zone_reclaim(struct zone *zone, gfp_t gfp_mask, unsigned int order)
3266
{
3267
	/* Minimum pages needed in order to stay on node */
3268
	const unsigned long nr_pages = 1 << order;
3269 3270
	struct task_struct *p = current;
	struct reclaim_state reclaim_state;
3271 3272
	struct scan_control sc = {
		.may_writepage = !!(zone_reclaim_mode & RECLAIM_WRITE),
3273
		.may_unmap = !!(zone_reclaim_mode & RECLAIM_SWAP),
3274
		.may_swap = 1,
3275 3276
		.nr_to_reclaim = max_t(unsigned long, nr_pages,
				       SWAP_CLUSTER_MAX),
3277
		.gfp_mask = gfp_mask,
3278
		.order = order,
3279
		.priority = ZONE_RECLAIM_PRIORITY,
3280
	};
3281 3282 3283
	struct shrink_control shrink = {
		.gfp_mask = sc.gfp_mask,
	};
3284
	unsigned long nr_slab_pages0, nr_slab_pages1;
3285 3286

	cond_resched();
3287 3288 3289 3290 3291 3292
	/*
	 * 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;
3293
	lockdep_set_current_reclaim_state(gfp_mask);
3294 3295
	reclaim_state.reclaimed_slab = 0;
	p->reclaim_state = &reclaim_state;
3296

3297
	if (zone_pagecache_reclaimable(zone) > zone->min_unmapped_pages) {
3298 3299 3300 3301 3302
		/*
		 * Free memory by calling shrink zone with increasing
		 * priorities until we have enough memory freed.
		 */
		do {
3303 3304
			shrink_zone(zone, &sc);
		} while (sc.nr_reclaimed < nr_pages && --sc.priority >= 0);
3305
	}
3306

3307 3308
	nr_slab_pages0 = zone_page_state(zone, NR_SLAB_RECLAIMABLE);
	if (nr_slab_pages0 > zone->min_slab_pages) {
3309
		/*
3310
		 * shrink_slab() does not currently allow us to determine how
3311 3312 3313 3314
		 * 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.
3315
		 *
3316 3317
		 * Note that shrink_slab will free memory on all zones and may
		 * take a long time.
3318
		 */
3319 3320 3321 3322
		for (;;) {
			unsigned long lru_pages = zone_reclaimable_pages(zone);

			/* No reclaimable slab or very low memory pressure */
3323
			if (!shrink_slab(&shrink, sc.nr_scanned, lru_pages))
3324 3325 3326 3327 3328 3329 3330 3331
				break;

			/* Freed enough memory */
			nr_slab_pages1 = zone_page_state(zone,
							NR_SLAB_RECLAIMABLE);
			if (nr_slab_pages1 + nr_pages <= nr_slab_pages0)
				break;
		}
3332 3333 3334 3335 3336

		/*
		 * Update nr_reclaimed by the number of slab pages we
		 * reclaimed from this zone.
		 */
3337 3338 3339
		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;
3340 3341
	}

3342
	p->reclaim_state = NULL;
3343
	current->flags &= ~(PF_MEMALLOC | PF_SWAPWRITE);
3344
	lockdep_clear_current_reclaim_state();
3345
	return sc.nr_reclaimed >= nr_pages;
3346
}
3347 3348 3349 3350

int zone_reclaim(struct zone *zone, gfp_t gfp_mask, unsigned int order)
{
	int node_id;
3351
	int ret;
3352 3353

	/*
3354 3355
	 * Zone reclaim reclaims unmapped file backed pages and
	 * slab pages if we are over the defined limits.
3356
	 *
3357 3358 3359 3360 3361
	 * 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.
3362
	 */
3363 3364
	if (zone_pagecache_reclaimable(zone) <= zone->min_unmapped_pages &&
	    zone_page_state(zone, NR_SLAB_RECLAIMABLE) <= zone->min_slab_pages)
3365
		return ZONE_RECLAIM_FULL;
3366

3367
	if (zone->all_unreclaimable)
3368
		return ZONE_RECLAIM_FULL;
3369

3370
	/*
3371
	 * Do not scan if the allocation should not be delayed.
3372
	 */
3373
	if (!(gfp_mask & __GFP_WAIT) || (current->flags & PF_MEMALLOC))
3374
		return ZONE_RECLAIM_NOSCAN;
3375 3376 3377 3378 3379 3380 3381

	/*
	 * 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.
	 */
3382
	node_id = zone_to_nid(zone);
3383
	if (node_state(node_id, N_CPU) && node_id != numa_node_id())
3384
		return ZONE_RECLAIM_NOSCAN;
3385 3386

	if (zone_test_and_set_flag(zone, ZONE_RECLAIM_LOCKED))
3387 3388
		return ZONE_RECLAIM_NOSCAN;

3389 3390 3391
	ret = __zone_reclaim(zone, gfp_mask, order);
	zone_clear_flag(zone, ZONE_RECLAIM_LOCKED);

3392 3393 3394
	if (!ret)
		count_vm_event(PGSCAN_ZONE_RECLAIM_FAILED);

3395
	return ret;
3396
}
3397
#endif
L
Lee Schermerhorn 已提交
3398 3399 3400 3401 3402 3403

/*
 * page_evictable - test whether a page is evictable
 * @page: the page to test
 *
 * Test whether page is evictable--i.e., should be placed on active/inactive
3404
 * lists vs unevictable list.
L
Lee Schermerhorn 已提交
3405 3406
 *
 * Reasons page might not be evictable:
3407
 * (1) page's mapping marked unevictable
N
Nick Piggin 已提交
3408
 * (2) page is part of an mlocked VMA
3409
 *
L
Lee Schermerhorn 已提交
3410
 */
3411
int page_evictable(struct page *page)
L
Lee Schermerhorn 已提交
3412
{
3413
	return !mapping_unevictable(page_mapping(page)) && !PageMlocked(page);
L
Lee Schermerhorn 已提交
3414
}
3415

3416
#ifdef CONFIG_SHMEM
3417
/**
3418 3419 3420
 * 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
3421
 *
3422
 * Checks pages for evictability and moves them to the appropriate lru list.
3423 3424
 *
 * This function is only used for SysV IPC SHM_UNLOCK.
3425
 */
3426
void check_move_unevictable_pages(struct page **pages, int nr_pages)
3427
{
3428
	struct lruvec *lruvec;
3429 3430 3431 3432
	struct zone *zone = NULL;
	int pgscanned = 0;
	int pgrescued = 0;
	int i;
3433

3434 3435 3436
	for (i = 0; i < nr_pages; i++) {
		struct page *page = pages[i];
		struct zone *pagezone;
3437

3438 3439 3440 3441 3442 3443 3444 3445
		pgscanned++;
		pagezone = page_zone(page);
		if (pagezone != zone) {
			if (zone)
				spin_unlock_irq(&zone->lru_lock);
			zone = pagezone;
			spin_lock_irq(&zone->lru_lock);
		}
3446
		lruvec = mem_cgroup_page_lruvec(page, zone);
3447

3448 3449
		if (!PageLRU(page) || !PageUnevictable(page))
			continue;
3450

3451
		if (page_evictable(page)) {
3452 3453 3454 3455
			enum lru_list lru = page_lru_base_type(page);

			VM_BUG_ON(PageActive(page));
			ClearPageUnevictable(page);
3456 3457
			del_page_from_lru_list(page, lruvec, LRU_UNEVICTABLE);
			add_page_to_lru_list(page, lruvec, lru);
3458
			pgrescued++;
3459
		}
3460
	}
3461

3462 3463 3464 3465
	if (zone) {
		__count_vm_events(UNEVICTABLE_PGRESCUED, pgrescued);
		__count_vm_events(UNEVICTABLE_PGSCANNED, pgscanned);
		spin_unlock_irq(&zone->lru_lock);
3466 3467
	}
}
3468
#endif /* CONFIG_SHMEM */
3469

3470
static void warn_scan_unevictable_pages(void)
3471
{
3472
	printk_once(KERN_WARNING
3473
		    "%s: The scan_unevictable_pages sysctl/node-interface has been "
3474
		    "disabled for lack of a legitimate use case.  If you have "
3475 3476
		    "one, please send an email to linux-mm@kvack.org.\n",
		    current->comm);
3477 3478 3479 3480 3481 3482 3483 3484 3485
}

/*
 * 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,
3486
			   void __user *buffer,
3487 3488
			   size_t *length, loff_t *ppos)
{
3489
	warn_scan_unevictable_pages();
3490
	proc_doulongvec_minmax(table, write, buffer, length, ppos);
3491 3492 3493 3494
	scan_unevictable_pages = 0;
	return 0;
}

3495
#ifdef CONFIG_NUMA
3496 3497 3498 3499 3500
/*
 * per node 'scan_unevictable_pages' attribute.  On demand re-scan of
 * a specified node's per zone unevictable lists for evictable pages.
 */

3501 3502
static ssize_t read_scan_unevictable_node(struct device *dev,
					  struct device_attribute *attr,
3503 3504
					  char *buf)
{
3505
	warn_scan_unevictable_pages();
3506 3507 3508
	return sprintf(buf, "0\n");	/* always zero; should fit... */
}

3509 3510
static ssize_t write_scan_unevictable_node(struct device *dev,
					   struct device_attribute *attr,
3511 3512
					const char *buf, size_t count)
{
3513
	warn_scan_unevictable_pages();
3514 3515 3516 3517
	return 1;
}


3518
static DEVICE_ATTR(scan_unevictable_pages, S_IRUGO | S_IWUSR,
3519 3520 3521 3522 3523
			read_scan_unevictable_node,
			write_scan_unevictable_node);

int scan_unevictable_register_node(struct node *node)
{
3524
	return device_create_file(&node->dev, &dev_attr_scan_unevictable_pages);
3525 3526 3527 3528
}

void scan_unevictable_unregister_node(struct node *node)
{
3529
	device_remove_file(&node->dev, &dev_attr_scan_unevictable_pages);
3530
}
3531
#endif