vmscan.c 105.9 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/vmpressure.h>
23
#include <linux/vmstat.h>
L
Linus Torvalds 已提交
24 25 26 27 28 29 30 31 32 33 34
#include <linux/file.h>
#include <linux/writeback.h>
#include <linux/blkdev.h>
#include <linux/buffer_head.h>	/* for try_to_release_page(),
					buffer_heads_over_limit */
#include <linux/mm_inline.h>
#include <linux/backing-dev.h>
#include <linux/rmap.h>
#include <linux/topology.h>
#include <linux/cpu.h>
#include <linux/cpuset.h>
35
#include <linux/compaction.h>
L
Linus Torvalds 已提交
36 37
#include <linux/notifier.h>
#include <linux/rwsem.h>
38
#include <linux/delay.h>
39
#include <linux/kthread.h>
40
#include <linux/freezer.h>
41
#include <linux/memcontrol.h>
42
#include <linux/delayacct.h>
43
#include <linux/sysctl.h>
44
#include <linux/oom.h>
45
#include <linux/prefetch.h>
L
Linus Torvalds 已提交
46 47 48 49 50 51

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

#include <linux/swapops.h>

52 53
#include "internal.h"

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

L
Linus Torvalds 已提交
57 58 59 60
struct scan_control {
	/* Incremented by the number of inactive pages that were scanned */
	unsigned long nr_scanned;

61 62 63
	/* Number of pages freed so far during a call to shrink_zones() */
	unsigned long nr_reclaimed;

64 65 66
	/* How many pages shrink_list() should reclaim */
	unsigned long nr_to_reclaim;

67 68
	unsigned long hibernation_mode;

L
Linus Torvalds 已提交
69
	/* This context's GFP mask */
A
Al Viro 已提交
70
	gfp_t gfp_mask;
L
Linus Torvalds 已提交
71 72 73

	int may_writepage;

74 75
	/* Can mapped pages be reclaimed? */
	int may_unmap;
76

77 78 79
	/* Can pages be swapped as part of reclaim? */
	int may_swap;

A
Andy Whitcroft 已提交
80
	int order;
81

82 83 84
	/* Scan (total_size >> priority) pages at once */
	int priority;

85 86 87 88 89
	/*
	 * 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;
90

91 92 93 94 95
	/*
	 * Nodemask of nodes allowed by the caller. If NULL, all nodes
	 * are scanned.
	 */
	nodemask_t	*nodemask;
L
Linus Torvalds 已提交
96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131
};

#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;
132
unsigned long vm_total_pages;	/* The total number of pages which the VM controls */
L
Linus Torvalds 已提交
133 134 135 136

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

A
Andrew Morton 已提交
137
#ifdef CONFIG_MEMCG
138 139
static bool global_reclaim(struct scan_control *sc)
{
140
	return !sc->target_mem_cgroup;
141
}
142
#else
143 144 145 146
static bool global_reclaim(struct scan_control *sc)
{
	return true;
}
147 148
#endif

149
static unsigned long get_lru_size(struct lruvec *lruvec, enum lru_list lru)
150
{
151
	if (!mem_cgroup_disabled())
152
		return mem_cgroup_get_lru_size(lruvec, lru);
153

154
	return zone_page_state(lruvec_zone(lruvec), NR_LRU_BASE + lru);
155 156
}

L
Linus Torvalds 已提交
157 158 159
/*
 * Add a shrinker callback to be called from the vm
 */
160
void register_shrinker(struct shrinker *shrinker)
L
Linus Torvalds 已提交
161
{
162
	atomic_long_set(&shrinker->nr_in_batch, 0);
163 164 165
	down_write(&shrinker_rwsem);
	list_add_tail(&shrinker->list, &shrinker_list);
	up_write(&shrinker_rwsem);
L
Linus Torvalds 已提交
166
}
167
EXPORT_SYMBOL(register_shrinker);
L
Linus Torvalds 已提交
168 169 170 171

/*
 * Remove one
 */
172
void unregister_shrinker(struct shrinker *shrinker)
L
Linus Torvalds 已提交
173 174 175 176 177
{
	down_write(&shrinker_rwsem);
	list_del(&shrinker->list);
	up_write(&shrinker_rwsem);
}
178
EXPORT_SYMBOL(unregister_shrinker);
L
Linus Torvalds 已提交
179

180 181 182 183 184 185 186 187
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
188 189 190 191 192 193 194 195 196
#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
197
 * If the vm encountered mapped pages on the LRU it increase the pressure on
L
Linus Torvalds 已提交
198 199 200 201 202 203 204
 * 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.
205 206
 *
 * Returns the number of slab objects which we shrunk.
L
Linus Torvalds 已提交
207
 */
208
unsigned long shrink_slab(struct shrink_control *shrink,
209
			  unsigned long nr_pages_scanned,
210
			  unsigned long lru_pages)
L
Linus Torvalds 已提交
211 212
{
	struct shrinker *shrinker;
213
	unsigned long ret = 0;
L
Linus Torvalds 已提交
214

215 216
	if (nr_pages_scanned == 0)
		nr_pages_scanned = SWAP_CLUSTER_MAX;
L
Linus Torvalds 已提交
217

218 219 220 221 222
	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
223 224 225

	list_for_each_entry(shrinker, &shrinker_list, list) {
		unsigned long long delta;
226 227
		long total_scan;
		long max_pass;
228
		int shrink_ret = 0;
229 230
		long nr;
		long new_nr;
231 232
		long batch_size = shrinker->batch ? shrinker->batch
						  : SHRINK_BATCH;
L
Linus Torvalds 已提交
233

234 235 236 237
		max_pass = do_shrinker_shrink(shrinker, shrink, 0);
		if (max_pass <= 0)
			continue;

238 239 240 241 242
		/*
		 * 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.
		 */
243
		nr = atomic_long_xchg(&shrinker->nr_in_batch, 0);
244 245

		total_scan = nr;
246
		delta = (4 * nr_pages_scanned) / shrinker->seeks;
247
		delta *= max_pass;
L
Linus Torvalds 已提交
248
		do_div(delta, lru_pages + 1);
249 250
		total_scan += delta;
		if (total_scan < 0) {
251 252
			printk(KERN_ERR "shrink_slab: %pF negative objects to "
			       "delete nr=%ld\n",
253 254
			       shrinker->shrink, total_scan);
			total_scan = max_pass;
255 256
		}

257 258 259 260 261 262 263 264 265 266 267 268 269 270 271
		/*
		 * 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);

272 273 274 275 276
		/*
		 * Avoid risking looping forever due to too large nr value:
		 * never try to free more than twice the estimate number of
		 * freeable entries.
		 */
277 278
		if (total_scan > max_pass * 2)
			total_scan = max_pass * 2;
L
Linus Torvalds 已提交
279

280
		trace_mm_shrink_slab_start(shrinker, shrink, nr,
281 282 283
					nr_pages_scanned, lru_pages,
					max_pass, delta, total_scan);

284
		while (total_scan >= batch_size) {
285
			int nr_before;
L
Linus Torvalds 已提交
286

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

			cond_resched();
		}

300 301 302 303 304
		/*
		 * 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.
		 */
305 306 307 308 309
		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);
310 311

		trace_mm_shrink_slab_end(shrinker, shrink_ret, nr, new_nr);
L
Linus Torvalds 已提交
312 313
	}
	up_read(&shrinker_rwsem);
314 315
out:
	cond_resched();
316
	return ret;
L
Linus Torvalds 已提交
317 318 319 320
}

static inline int is_page_cache_freeable(struct page *page)
{
321 322 323 324 325
	/*
	 * 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.
	 */
326
	return page_count(page) - page_has_private(page) == 2;
L
Linus Torvalds 已提交
327 328
}

329 330
static int may_write_to_queue(struct backing_dev_info *bdi,
			      struct scan_control *sc)
L
Linus Torvalds 已提交
331
{
332
	if (current->flags & PF_SWAPWRITE)
L
Linus Torvalds 已提交
333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355
		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
356
	lock_page(page);
357 358
	if (page_mapping(page) == mapping)
		mapping_set_error(mapping, error);
L
Linus Torvalds 已提交
359 360 361
	unlock_page(page);
}

362 363 364 365 366 367 368 369 370 371 372 373
/* 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
374
/*
A
Andrew Morton 已提交
375 376
 * pageout is called by shrink_page_list() for each dirty page.
 * Calls ->writepage().
L
Linus Torvalds 已提交
377
 */
378
static pageout_t pageout(struct page *page, struct address_space *mapping,
379
			 struct scan_control *sc)
L
Linus Torvalds 已提交
380 381 382 383 384 385 386 387
{
	/*
	 * 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.
	 *
388
	 * If this process is currently in __generic_file_aio_write() against
L
Linus Torvalds 已提交
389 390 391 392 393 394 395 396 397 398 399 400 401 402 403
	 * 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.
		 */
404
		if (page_has_private(page)) {
L
Linus Torvalds 已提交
405 406
			if (try_to_free_buffers(page)) {
				ClearPageDirty(page);
407
				printk("%s: orphaned page\n", __func__);
L
Linus Torvalds 已提交
408 409 410 411 412 413 414
				return PAGE_CLEAN;
			}
		}
		return PAGE_KEEP;
	}
	if (mapping->a_ops->writepage == NULL)
		return PAGE_ACTIVATE;
415
	if (!may_write_to_queue(mapping->backing_dev_info, sc))
L
Linus Torvalds 已提交
416 417 418 419 420 421 422
		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,
423 424
			.range_start = 0,
			.range_end = LLONG_MAX,
L
Linus Torvalds 已提交
425 426 427 428 429 430 431
			.for_reclaim = 1,
		};

		SetPageReclaim(page);
		res = mapping->a_ops->writepage(page, &wbc);
		if (res < 0)
			handle_write_error(mapping, page, res);
432
		if (res == AOP_WRITEPAGE_ACTIVATE) {
L
Linus Torvalds 已提交
433 434 435
			ClearPageReclaim(page);
			return PAGE_ACTIVATE;
		}
436

L
Linus Torvalds 已提交
437 438 439 440
		if (!PageWriteback(page)) {
			/* synchronous write or broken a_ops? */
			ClearPageReclaim(page);
		}
M
Mel Gorman 已提交
441
		trace_mm_vmscan_writepage(page, trace_reclaim_flags(page));
442
		inc_zone_page_state(page, NR_VMSCAN_WRITE);
L
Linus Torvalds 已提交
443 444 445 446 447 448
		return PAGE_SUCCESS;
	}

	return PAGE_CLEAN;
}

449
/*
N
Nick Piggin 已提交
450 451
 * Same as remove_mapping, but if the page is removed from the mapping, it
 * gets returned with a refcount of 0.
452
 */
N
Nick Piggin 已提交
453
static int __remove_mapping(struct address_space *mapping, struct page *page)
454
{
455 456
	BUG_ON(!PageLocked(page));
	BUG_ON(mapping != page_mapping(page));
457

N
Nick Piggin 已提交
458
	spin_lock_irq(&mapping->tree_lock);
459
	/*
N
Nick Piggin 已提交
460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482
	 * 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.
483
	 */
N
Nick Piggin 已提交
484
	if (!page_freeze_refs(page, 2))
485
		goto cannot_free;
N
Nick Piggin 已提交
486 487 488
	/* note: atomic_cmpxchg in page_freeze_refs provides the smp_rmb */
	if (unlikely(PageDirty(page))) {
		page_unfreeze_refs(page, 2);
489
		goto cannot_free;
N
Nick Piggin 已提交
490
	}
491 492 493 494

	if (PageSwapCache(page)) {
		swp_entry_t swap = { .val = page_private(page) };
		__delete_from_swap_cache(page);
N
Nick Piggin 已提交
495
		spin_unlock_irq(&mapping->tree_lock);
496
		swapcache_free(swap, page);
N
Nick Piggin 已提交
497
	} else {
498 499 500 501
		void (*freepage)(struct page *);

		freepage = mapping->a_ops->freepage;

502
		__delete_from_page_cache(page);
N
Nick Piggin 已提交
503
		spin_unlock_irq(&mapping->tree_lock);
504
		mem_cgroup_uncharge_cache_page(page);
505 506 507

		if (freepage != NULL)
			freepage(page);
508 509 510 511 512
	}

	return 1;

cannot_free:
N
Nick Piggin 已提交
513
	spin_unlock_irq(&mapping->tree_lock);
514 515 516
	return 0;
}

N
Nick Piggin 已提交
517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536
/*
 * 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
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;
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.
		 */
563 564
		lru = page_lru_base_type(page);
		lru_cache_add(page);
L
Lee Schermerhorn 已提交
565 566 567 568 569 570 571
	} 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
}

671 672 673 674
/* Check if a page is dirty or under writeback */
static void page_check_dirty_writeback(struct page *page,
				       bool *dirty, bool *writeback)
{
675 676
	struct address_space *mapping;

677 678 679 680 681 682 683 684 685 686 687 688 689
	/*
	 * Anonymous pages are not handled by flushers and must be written
	 * from reclaim context. Do not stall reclaim based on them
	 */
	if (!page_is_file_cache(page)) {
		*dirty = false;
		*writeback = false;
		return;
	}

	/* By default assume that the page flags are accurate */
	*dirty = PageDirty(page);
	*writeback = PageWriteback(page);
690 691 692 693 694 695 696 697

	/* Verify dirty/writeback state if the filesystem supports it */
	if (!page_has_private(page))
		return;

	mapping = page_mapping(page);
	if (mapping && mapping->a_ops->is_dirty_writeback)
		mapping->a_ops->is_dirty_writeback(page, dirty, writeback);
698 699
}

L
Linus Torvalds 已提交
700
/*
A
Andrew Morton 已提交
701
 * shrink_page_list() returns the number of reclaimed pages
L
Linus Torvalds 已提交
702
 */
A
Andrew Morton 已提交
703
static unsigned long shrink_page_list(struct list_head *page_list,
704
				      struct zone *zone,
705
				      struct scan_control *sc,
706
				      enum ttu_flags ttu_flags,
707
				      unsigned long *ret_nr_dirty,
708
				      unsigned long *ret_nr_unqueued_dirty,
709
				      unsigned long *ret_nr_congested,
710
				      unsigned long *ret_nr_writeback,
711
				      unsigned long *ret_nr_immediate,
712
				      bool force_reclaim)
L
Linus Torvalds 已提交
713 714
{
	LIST_HEAD(ret_pages);
715
	LIST_HEAD(free_pages);
L
Linus Torvalds 已提交
716
	int pgactivate = 0;
717
	unsigned long nr_unqueued_dirty = 0;
718 719
	unsigned long nr_dirty = 0;
	unsigned long nr_congested = 0;
720
	unsigned long nr_reclaimed = 0;
721
	unsigned long nr_writeback = 0;
722
	unsigned long nr_immediate = 0;
L
Linus Torvalds 已提交
723 724 725

	cond_resched();

726
	mem_cgroup_uncharge_start();
L
Linus Torvalds 已提交
727 728 729 730
	while (!list_empty(page_list)) {
		struct address_space *mapping;
		struct page *page;
		int may_enter_fs;
731
		enum page_references references = PAGEREF_RECLAIM_CLEAN;
732
		bool dirty, writeback;
L
Linus Torvalds 已提交
733 734 735 736 737 738

		cond_resched();

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

N
Nick Piggin 已提交
739
		if (!trylock_page(page))
L
Linus Torvalds 已提交
740 741
			goto keep;

N
Nick Piggin 已提交
742
		VM_BUG_ON(PageActive(page));
743
		VM_BUG_ON(page_zone(page) != zone);
L
Linus Torvalds 已提交
744 745

		sc->nr_scanned++;
746

747
		if (unlikely(!page_evictable(page)))
N
Nick Piggin 已提交
748
			goto cull_mlocked;
L
Lee Schermerhorn 已提交
749

750
		if (!sc->may_unmap && page_mapped(page))
751 752
			goto keep_locked;

L
Linus Torvalds 已提交
753 754 755 756
		/* Double the slab pressure for mapped and swapcache pages */
		if (page_mapped(page) || PageSwapCache(page))
			sc->nr_scanned++;

757 758 759
		may_enter_fs = (sc->gfp_mask & __GFP_FS) ||
			(PageSwapCache(page) && (sc->gfp_mask & __GFP_IO));

760 761 762 763 764 765 766 767 768 769 770 771 772
		/*
		 * The number of dirty pages determines if a zone is marked
		 * reclaim_congested which affects wait_iff_congested. kswapd
		 * will stall and start writing pages if the tail of the LRU
		 * is all dirty unqueued pages.
		 */
		page_check_dirty_writeback(page, &dirty, &writeback);
		if (dirty || writeback)
			nr_dirty++;

		if (dirty && !writeback)
			nr_unqueued_dirty++;

773 774 775 776 777 778
		/*
		 * Treat this page as congested if the underlying BDI is or if
		 * pages are cycling through the LRU so quickly that the
		 * pages marked for immediate reclaim are making it to the
		 * end of the LRU a second time.
		 */
779
		mapping = page_mapping(page);
780 781
		if ((mapping && bdi_write_congested(mapping->backing_dev_info)) ||
		    (writeback && PageReclaim(page)))
782 783
			nr_congested++;

784 785 786 787 788 789 790 791 792 793 794
		/*
		 * If a page at the tail of the LRU is under writeback, there
		 * are three cases to consider.
		 *
		 * 1) If reclaim is encountering an excessive number of pages
		 *    under writeback and this page is both under writeback and
		 *    PageReclaim then it indicates that pages are being queued
		 *    for IO but are being recycled through the LRU before the
		 *    IO can complete. Waiting on the page itself risks an
		 *    indefinite stall if it is impossible to writeback the
		 *    page due to IO error or disconnected storage so instead
795 796
		 *    note that the LRU is being scanned too quickly and the
		 *    caller can stall after page list has been processed.
797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820
		 *
		 * 2) Global reclaim encounters a page, memcg encounters a
		 *    page that is not marked for immediate reclaim or
		 *    the caller does not have __GFP_IO. In this case mark
		 *    the page for immediate reclaim and continue scanning.
		 *
		 *    __GFP_IO is checked  because a loop driver 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.
		 *
		 *    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.
		 *
		 * 3) memcg encounters a page that is not already marked
		 *    PageReclaim. memcg does not have any dirty pages
		 *    throttling so we could easily OOM just because too many
		 *    pages are in writeback and there is nothing else to
		 *    reclaim. Wait for the writeback to complete.
		 */
821
		if (PageWriteback(page)) {
822 823 824 825
			/* Case 1 above */
			if (current_is_kswapd() &&
			    PageReclaim(page) &&
			    zone_is_reclaim_writeback(zone)) {
826 827
				nr_immediate++;
				goto keep_locked;
828 829 830

			/* Case 2 above */
			} else if (global_reclaim(sc) ||
831 832 833 834 835 836 837 838 839 840 841 842 843
			    !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);
844
				nr_writeback++;
845

846
				goto keep_locked;
847 848 849 850

			/* Case 3 above */
			} else {
				wait_on_page_writeback(page);
851
			}
852
		}
L
Linus Torvalds 已提交
853

854 855 856
		if (!force_reclaim)
			references = page_check_references(page, sc);

857 858
		switch (references) {
		case PAGEREF_ACTIVATE:
L
Linus Torvalds 已提交
859
			goto activate_locked;
860 861
		case PAGEREF_KEEP:
			goto keep_locked;
862 863 864 865
		case PAGEREF_RECLAIM:
		case PAGEREF_RECLAIM_CLEAN:
			; /* try to reclaim the page below */
		}
L
Linus Torvalds 已提交
866 867 868 869 870

		/*
		 * Anonymous process memory has backing store?
		 * Try to allocate it some swap space here.
		 */
N
Nick Piggin 已提交
871
		if (PageAnon(page) && !PageSwapCache(page)) {
872 873
			if (!(sc->gfp_mask & __GFP_IO))
				goto keep_locked;
874
			if (!add_to_swap(page, page_list))
L
Linus Torvalds 已提交
875
				goto activate_locked;
876
			may_enter_fs = 1;
L
Linus Torvalds 已提交
877

878 879 880
			/* Adding to swap updated mapping */
			mapping = page_mapping(page);
		}
L
Linus Torvalds 已提交
881 882 883 884 885 886

		/*
		 * The page is mapped into the page tables of one or more
		 * processes. Try to unmap it here.
		 */
		if (page_mapped(page) && mapping) {
887
			switch (try_to_unmap(page, ttu_flags)) {
L
Linus Torvalds 已提交
888 889 890 891
			case SWAP_FAIL:
				goto activate_locked;
			case SWAP_AGAIN:
				goto keep_locked;
N
Nick Piggin 已提交
892 893
			case SWAP_MLOCK:
				goto cull_mlocked;
L
Linus Torvalds 已提交
894 895 896 897 898 899
			case SWAP_SUCCESS:
				; /* try to free the page below */
			}
		}

		if (PageDirty(page)) {
900 901
			/*
			 * Only kswapd can writeback filesystem pages to
902 903
			 * avoid risk of stack overflow but only writeback
			 * if many dirty pages have been encountered.
904
			 */
905
			if (page_is_file_cache(page) &&
906
					(!current_is_kswapd() ||
907
					 !zone_is_reclaim_dirty(zone))) {
908 909 910 911 912 913 914 915 916
				/*
				 * 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);

917 918 919
				goto keep_locked;
			}

920
			if (references == PAGEREF_RECLAIM_CLEAN)
L
Linus Torvalds 已提交
921
				goto keep_locked;
922
			if (!may_enter_fs)
L
Linus Torvalds 已提交
923
				goto keep_locked;
924
			if (!sc->may_writepage)
L
Linus Torvalds 已提交
925 926 927
				goto keep_locked;

			/* Page is dirty, try to write it out here */
928
			switch (pageout(page, mapping, sc)) {
L
Linus Torvalds 已提交
929 930 931 932 933
			case PAGE_KEEP:
				goto keep_locked;
			case PAGE_ACTIVATE:
				goto activate_locked;
			case PAGE_SUCCESS:
934
				if (PageWriteback(page))
935
					goto keep;
936
				if (PageDirty(page))
L
Linus Torvalds 已提交
937
					goto keep;
938

L
Linus Torvalds 已提交
939 940 941 942
				/*
				 * A synchronous write - probably a ramdisk.  Go
				 * ahead and try to reclaim the page.
				 */
N
Nick Piggin 已提交
943
				if (!trylock_page(page))
L
Linus Torvalds 已提交
944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962
					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
963
		 * will do this, as well as the blockdev mapping.
L
Linus Torvalds 已提交
964 965 966 967 968 969 970 971 972 973
		 * 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.
		 */
974
		if (page_has_private(page)) {
L
Linus Torvalds 已提交
975 976
			if (!try_to_release_page(page, sc->gfp_mask))
				goto activate_locked;
N
Nick Piggin 已提交
977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992
			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
993 994
		}

N
Nick Piggin 已提交
995
		if (!mapping || !__remove_mapping(mapping, page))
996
			goto keep_locked;
L
Linus Torvalds 已提交
997

N
Nick Piggin 已提交
998 999 1000 1001 1002 1003 1004 1005
		/*
		 * 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
1006
free_it:
1007
		nr_reclaimed++;
1008 1009 1010 1011 1012 1013

		/*
		 * 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
1014 1015
		continue;

N
Nick Piggin 已提交
1016
cull_mlocked:
1017 1018
		if (PageSwapCache(page))
			try_to_free_swap(page);
N
Nick Piggin 已提交
1019 1020 1021 1022
		unlock_page(page);
		putback_lru_page(page);
		continue;

L
Linus Torvalds 已提交
1023
activate_locked:
1024 1025
		/* Not a candidate for swapping, so reclaim swap space. */
		if (PageSwapCache(page) && vm_swap_full())
1026
			try_to_free_swap(page);
L
Lee Schermerhorn 已提交
1027
		VM_BUG_ON(PageActive(page));
L
Linus Torvalds 已提交
1028 1029 1030 1031 1032 1033
		SetPageActive(page);
		pgactivate++;
keep_locked:
		unlock_page(page);
keep:
		list_add(&page->lru, &ret_pages);
N
Nick Piggin 已提交
1034
		VM_BUG_ON(PageLRU(page) || PageUnevictable(page));
L
Linus Torvalds 已提交
1035
	}
1036

1037
	free_hot_cold_page_list(&free_pages, 1);
1038

L
Linus Torvalds 已提交
1039
	list_splice(&ret_pages, page_list);
1040
	count_vm_events(PGACTIVATE, pgactivate);
1041
	mem_cgroup_uncharge_end();
1042 1043
	*ret_nr_dirty += nr_dirty;
	*ret_nr_congested += nr_congested;
1044
	*ret_nr_unqueued_dirty += nr_unqueued_dirty;
1045
	*ret_nr_writeback += nr_writeback;
1046
	*ret_nr_immediate += nr_immediate;
1047
	return nr_reclaimed;
L
Linus Torvalds 已提交
1048 1049
}

1050 1051 1052 1053 1054 1055 1056 1057
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,
	};
1058
	unsigned long ret, dummy1, dummy2, dummy3, dummy4, dummy5;
1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069
	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,
1070 1071
			TTU_UNMAP|TTU_IGNORE_ACCESS,
			&dummy1, &dummy2, &dummy3, &dummy4, &dummy5, true);
1072 1073 1074 1075 1076
	list_splice(&clean_pages, page_list);
	__mod_zone_page_state(zone, NR_ISOLATED_FILE, -ret);
	return ret;
}

A
Andy Whitcroft 已提交
1077 1078 1079 1080 1081 1082 1083 1084 1085 1086
/*
 * 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.
 */
1087
int __isolate_lru_page(struct page *page, isolate_mode_t mode)
A
Andy Whitcroft 已提交
1088 1089 1090 1091 1092 1093 1094
{
	int ret = -EINVAL;

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

M
Minchan Kim 已提交
1095 1096
	/* Compaction should not handle unevictable pages but CMA can do so */
	if (PageUnevictable(page) && !(mode & ISOLATE_UNEVICTABLE))
L
Lee Schermerhorn 已提交
1097 1098
		return ret;

A
Andy Whitcroft 已提交
1099
	ret = -EBUSY;
K
KAMEZAWA Hiroyuki 已提交
1100

1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133
	/*
	 * 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;
		}
	}
1134

1135 1136 1137
	if ((mode & ISOLATE_UNMAPPED) && page_mapped(page))
		return ret;

A
Andy Whitcroft 已提交
1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150
	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
1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161
/*
 * 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.
1162
 * @lruvec:	The LRU vector to pull pages from.
L
Linus Torvalds 已提交
1163
 * @dst:	The temp list to put pages on to.
H
Hugh Dickins 已提交
1164
 * @nr_scanned:	The number of pages that were scanned.
1165
 * @sc:		The scan_control struct for this reclaim session
A
Andy Whitcroft 已提交
1166
 * @mode:	One of the LRU isolation modes
1167
 * @lru:	LRU list id for isolating
L
Linus Torvalds 已提交
1168 1169 1170
 *
 * returns how many pages were moved onto *@dst.
 */
1171
static unsigned long isolate_lru_pages(unsigned long nr_to_scan,
1172
		struct lruvec *lruvec, struct list_head *dst,
1173
		unsigned long *nr_scanned, struct scan_control *sc,
1174
		isolate_mode_t mode, enum lru_list lru)
L
Linus Torvalds 已提交
1175
{
H
Hugh Dickins 已提交
1176
	struct list_head *src = &lruvec->lists[lru];
1177
	unsigned long nr_taken = 0;
1178
	unsigned long scan;
L
Linus Torvalds 已提交
1179

1180
	for (scan = 0; scan < nr_to_scan && !list_empty(src); scan++) {
A
Andy Whitcroft 已提交
1181
		struct page *page;
1182
		int nr_pages;
A
Andy Whitcroft 已提交
1183

L
Linus Torvalds 已提交
1184 1185 1186
		page = lru_to_page(src);
		prefetchw_prev_lru_page(page, src, flags);

N
Nick Piggin 已提交
1187
		VM_BUG_ON(!PageLRU(page));
N
Nick Piggin 已提交
1188

1189
		switch (__isolate_lru_page(page, mode)) {
A
Andy Whitcroft 已提交
1190
		case 0:
1191 1192
			nr_pages = hpage_nr_pages(page);
			mem_cgroup_update_lru_size(lruvec, lru, -nr_pages);
A
Andy Whitcroft 已提交
1193
			list_move(&page->lru, dst);
1194
			nr_taken += nr_pages;
A
Andy Whitcroft 已提交
1195 1196 1197 1198 1199 1200
			break;

		case -EBUSY:
			/* else it is being freed elsewhere */
			list_move(&page->lru, src);
			continue;
1201

A
Andy Whitcroft 已提交
1202 1203 1204
		default:
			BUG();
		}
L
Linus Torvalds 已提交
1205 1206
	}

H
Hugh Dickins 已提交
1207
	*nr_scanned = scan;
H
Hugh Dickins 已提交
1208 1209
	trace_mm_vmscan_lru_isolate(sc->order, nr_to_scan, scan,
				    nr_taken, mode, is_file_lru(lru));
L
Linus Torvalds 已提交
1210 1211 1212
	return nr_taken;
}

1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223
/**
 * 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
1224 1225 1226
 * 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.
1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241
 *
 * 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;

1242 1243
	VM_BUG_ON(!page_count(page));

1244 1245
	if (PageLRU(page)) {
		struct zone *zone = page_zone(page);
1246
		struct lruvec *lruvec;
1247 1248

		spin_lock_irq(&zone->lru_lock);
1249
		lruvec = mem_cgroup_page_lruvec(page, zone);
1250
		if (PageLRU(page)) {
L
Lee Schermerhorn 已提交
1251
			int lru = page_lru(page);
1252
			get_page(page);
1253
			ClearPageLRU(page);
1254 1255
			del_page_from_lru_list(page, lruvec, lru);
			ret = 0;
1256 1257 1258 1259 1260 1261
		}
		spin_unlock_irq(&zone->lru_lock);
	}
	return ret;
}

1262
/*
F
Fengguang Wu 已提交
1263 1264 1265 1266 1267
 * A direct reclaimer may isolate SWAP_CLUSTER_MAX pages from the LRU list and
 * then get resheduled. When there are massive number of tasks doing page
 * allocation, such sleeping direct reclaimers may keep piling up on each CPU,
 * the LRU list will go small and be scanned faster than necessary, leading to
 * unnecessary swapping, thrashing and OOM.
1268 1269 1270 1271 1272 1273 1274 1275 1276
 */
static int too_many_isolated(struct zone *zone, int file,
		struct scan_control *sc)
{
	unsigned long inactive, isolated;

	if (current_is_kswapd())
		return 0;

1277
	if (!global_reclaim(sc))
1278 1279 1280 1281 1282 1283 1284 1285 1286 1287
		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);
	}

1288 1289 1290 1291 1292 1293 1294 1295
	/*
	 * GFP_NOIO/GFP_NOFS callers are allowed to isolate more pages, so they
	 * won't get blocked by normal direct-reclaimers, forming a circular
	 * deadlock.
	 */
	if ((sc->gfp_mask & GFP_IOFS) == GFP_IOFS)
		inactive >>= 3;

1296 1297 1298
	return isolated > inactive;
}

1299
static noinline_for_stack void
H
Hugh Dickins 已提交
1300
putback_inactive_pages(struct lruvec *lruvec, struct list_head *page_list)
1301
{
1302 1303
	struct zone_reclaim_stat *reclaim_stat = &lruvec->reclaim_stat;
	struct zone *zone = lruvec_zone(lruvec);
1304
	LIST_HEAD(pages_to_free);
1305 1306 1307 1308 1309

	/*
	 * Put back any unfreeable pages.
	 */
	while (!list_empty(page_list)) {
1310
		struct page *page = lru_to_page(page_list);
1311
		int lru;
1312

1313 1314
		VM_BUG_ON(PageLRU(page));
		list_del(&page->lru);
1315
		if (unlikely(!page_evictable(page))) {
1316 1317 1318 1319 1320
			spin_unlock_irq(&zone->lru_lock);
			putback_lru_page(page);
			spin_lock_irq(&zone->lru_lock);
			continue;
		}
1321 1322 1323

		lruvec = mem_cgroup_page_lruvec(page, zone);

1324
		SetPageLRU(page);
1325
		lru = page_lru(page);
1326 1327
		add_page_to_lru_list(page, lruvec, lru);

1328 1329
		if (is_active_lru(lru)) {
			int file = is_file_lru(lru);
1330 1331
			int numpages = hpage_nr_pages(page);
			reclaim_stat->recent_rotated[file] += numpages;
1332
		}
1333 1334 1335
		if (put_page_testzero(page)) {
			__ClearPageLRU(page);
			__ClearPageActive(page);
1336
			del_page_from_lru_list(page, lruvec, lru);
1337 1338 1339 1340 1341 1342 1343

			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);
1344 1345 1346
		}
	}

1347 1348 1349 1350
	/*
	 * To save our caller's stack, now use input list for pages to free.
	 */
	list_splice(&pages_to_free, page_list);
1351 1352
}

L
Linus Torvalds 已提交
1353
/*
A
Andrew Morton 已提交
1354 1355
 * shrink_inactive_list() is a helper for shrink_zone().  It returns the number
 * of reclaimed pages
L
Linus Torvalds 已提交
1356
 */
1357
static noinline_for_stack unsigned long
1358
shrink_inactive_list(unsigned long nr_to_scan, struct lruvec *lruvec,
1359
		     struct scan_control *sc, enum lru_list lru)
L
Linus Torvalds 已提交
1360 1361
{
	LIST_HEAD(page_list);
1362
	unsigned long nr_scanned;
1363
	unsigned long nr_reclaimed = 0;
1364
	unsigned long nr_taken;
1365 1366
	unsigned long nr_dirty = 0;
	unsigned long nr_congested = 0;
1367
	unsigned long nr_unqueued_dirty = 0;
1368
	unsigned long nr_writeback = 0;
1369
	unsigned long nr_immediate = 0;
1370
	isolate_mode_t isolate_mode = 0;
1371
	int file = is_file_lru(lru);
1372 1373
	struct zone *zone = lruvec_zone(lruvec);
	struct zone_reclaim_stat *reclaim_stat = &lruvec->reclaim_stat;
1374

1375
	while (unlikely(too_many_isolated(zone, file, sc))) {
1376
		congestion_wait(BLK_RW_ASYNC, HZ/10);
1377 1378 1379 1380 1381 1382

		/* 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
1383
	lru_add_drain();
1384 1385

	if (!sc->may_unmap)
1386
		isolate_mode |= ISOLATE_UNMAPPED;
1387
	if (!sc->may_writepage)
1388
		isolate_mode |= ISOLATE_CLEAN;
1389

L
Linus Torvalds 已提交
1390
	spin_lock_irq(&zone->lru_lock);
1391

1392 1393
	nr_taken = isolate_lru_pages(nr_to_scan, lruvec, &page_list,
				     &nr_scanned, sc, isolate_mode, lru);
1394 1395 1396 1397

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

1398
	if (global_reclaim(sc)) {
1399 1400
		zone->pages_scanned += nr_scanned;
		if (current_is_kswapd())
H
Hugh Dickins 已提交
1401
			__count_zone_vm_events(PGSCAN_KSWAPD, zone, nr_scanned);
1402
		else
H
Hugh Dickins 已提交
1403
			__count_zone_vm_events(PGSCAN_DIRECT, zone, nr_scanned);
1404
	}
1405
	spin_unlock_irq(&zone->lru_lock);
1406

1407
	if (nr_taken == 0)
1408
		return 0;
A
Andy Whitcroft 已提交
1409

1410
	nr_reclaimed = shrink_page_list(&page_list, zone, sc, TTU_UNMAP,
1411 1412 1413
				&nr_dirty, &nr_unqueued_dirty, &nr_congested,
				&nr_writeback, &nr_immediate,
				false);
1414

1415 1416
	spin_lock_irq(&zone->lru_lock);

1417
	reclaim_stat->recent_scanned[file] += nr_taken;
1418

Y
Ying Han 已提交
1419 1420 1421 1422 1423 1424 1425 1426
	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
1427

1428
	putback_inactive_pages(lruvec, &page_list);
1429

1430
	__mod_zone_page_state(zone, NR_ISOLATED_ANON + file, -nr_taken);
1431 1432 1433 1434

	spin_unlock_irq(&zone->lru_lock);

	free_hot_cold_page_list(&page_list, 1);
1435

1436 1437 1438 1439 1440 1441 1442 1443 1444 1445
	/*
	 * 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.
	 *
1446 1447 1448
	 * Once a zone is flagged ZONE_WRITEBACK, kswapd will count the number
	 * of pages under pages flagged for immediate reclaim and stall if any
	 * are encountered in the nr_immediate check below.
1449
	 */
1450
	if (nr_writeback && nr_writeback == nr_taken)
1451
		zone_set_flag(zone, ZONE_WRITEBACK);
1452

1453
	/*
1454 1455
	 * memcg will stall in page writeback so only consider forcibly
	 * stalling for global reclaim
1456
	 */
1457
	if (global_reclaim(sc)) {
1458 1459 1460 1461 1462 1463 1464
		/*
		 * Tag a zone as congested if all the dirty pages scanned were
		 * backed by a congested BDI and wait_iff_congested will stall.
		 */
		if (nr_dirty && nr_dirty == nr_congested)
			zone_set_flag(zone, ZONE_CONGESTED);

1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482
		/*
		 * If dirty pages are scanned that are not queued for IO, it
		 * implies that flushers are not keeping up. In this case, flag
		 * the zone ZONE_TAIL_LRU_DIRTY and kswapd will start writing
		 * pages from reclaim context. It will forcibly stall in the
		 * next check.
		 */
		if (nr_unqueued_dirty == nr_taken)
			zone_set_flag(zone, ZONE_TAIL_LRU_DIRTY);

		/*
		 * In addition, if kswapd scans pages marked marked for
		 * immediate reclaim and under writeback (nr_immediate), it
		 * implies that pages are cycling through the LRU faster than
		 * they are written so also forcibly stall.
		 */
		if (nr_unqueued_dirty == nr_taken || nr_immediate)
			congestion_wait(BLK_RW_ASYNC, HZ/10);
1483
	}
1484

1485 1486 1487 1488 1489 1490 1491 1492
	/*
	 * Stall direct reclaim for IO completions if underlying BDIs or zone
	 * is congested. Allow kswapd to continue until it starts encountering
	 * unqueued dirty pages or cycling through the LRU too quickly.
	 */
	if (!sc->hibernation_mode && !current_is_kswapd())
		wait_iff_congested(zone, BLK_RW_ASYNC, HZ/10);

1493 1494 1495
	trace_mm_vmscan_lru_shrink_inactive(zone->zone_pgdat->node_id,
		zone_idx(zone),
		nr_scanned, nr_reclaimed,
1496
		sc->priority,
M
Mel Gorman 已提交
1497
		trace_shrink_flags(file));
1498
	return nr_reclaimed;
L
Linus Torvalds 已提交
1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516 1517
}

/*
 * 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.
 */
1518

1519
static void move_active_pages_to_lru(struct lruvec *lruvec,
1520
				     struct list_head *list,
1521
				     struct list_head *pages_to_free,
1522 1523
				     enum lru_list lru)
{
1524
	struct zone *zone = lruvec_zone(lruvec);
1525 1526
	unsigned long pgmoved = 0;
	struct page *page;
1527
	int nr_pages;
1528 1529 1530

	while (!list_empty(list)) {
		page = lru_to_page(list);
1531
		lruvec = mem_cgroup_page_lruvec(page, zone);
1532 1533 1534 1535

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

1536 1537
		nr_pages = hpage_nr_pages(page);
		mem_cgroup_update_lru_size(lruvec, lru, nr_pages);
1538
		list_move(&page->lru, &lruvec->lists[lru]);
1539
		pgmoved += nr_pages;
1540

1541 1542 1543
		if (put_page_testzero(page)) {
			__ClearPageLRU(page);
			__ClearPageActive(page);
1544
			del_page_from_lru_list(page, lruvec, lru);
1545 1546 1547 1548 1549 1550 1551

			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);
1552 1553 1554 1555 1556 1557
		}
	}
	__mod_zone_page_state(zone, NR_LRU_BASE + lru, pgmoved);
	if (!is_active_lru(lru))
		__count_vm_events(PGDEACTIVATE, pgmoved);
}
1558

H
Hugh Dickins 已提交
1559
static void shrink_active_list(unsigned long nr_to_scan,
1560
			       struct lruvec *lruvec,
1561
			       struct scan_control *sc,
1562
			       enum lru_list lru)
L
Linus Torvalds 已提交
1563
{
1564
	unsigned long nr_taken;
H
Hugh Dickins 已提交
1565
	unsigned long nr_scanned;
1566
	unsigned long vm_flags;
L
Linus Torvalds 已提交
1567
	LIST_HEAD(l_hold);	/* The pages which were snipped off */
1568
	LIST_HEAD(l_active);
1569
	LIST_HEAD(l_inactive);
L
Linus Torvalds 已提交
1570
	struct page *page;
1571
	struct zone_reclaim_stat *reclaim_stat = &lruvec->reclaim_stat;
1572
	unsigned long nr_rotated = 0;
1573
	isolate_mode_t isolate_mode = 0;
1574
	int file = is_file_lru(lru);
1575
	struct zone *zone = lruvec_zone(lruvec);
L
Linus Torvalds 已提交
1576 1577

	lru_add_drain();
1578 1579

	if (!sc->may_unmap)
1580
		isolate_mode |= ISOLATE_UNMAPPED;
1581
	if (!sc->may_writepage)
1582
		isolate_mode |= ISOLATE_CLEAN;
1583

L
Linus Torvalds 已提交
1584
	spin_lock_irq(&zone->lru_lock);
1585

1586 1587
	nr_taken = isolate_lru_pages(nr_to_scan, lruvec, &l_hold,
				     &nr_scanned, sc, isolate_mode, lru);
1588
	if (global_reclaim(sc))
H
Hugh Dickins 已提交
1589
		zone->pages_scanned += nr_scanned;
1590

1591
	reclaim_stat->recent_scanned[file] += nr_taken;
1592

H
Hugh Dickins 已提交
1593
	__count_zone_vm_events(PGREFILL, zone, nr_scanned);
1594
	__mod_zone_page_state(zone, NR_LRU_BASE + lru, -nr_taken);
K
KOSAKI Motohiro 已提交
1595
	__mod_zone_page_state(zone, NR_ISOLATED_ANON + file, nr_taken);
L
Linus Torvalds 已提交
1596 1597 1598 1599 1600 1601
	spin_unlock_irq(&zone->lru_lock);

	while (!list_empty(&l_hold)) {
		cond_resched();
		page = lru_to_page(&l_hold);
		list_del(&page->lru);
1602

1603
		if (unlikely(!page_evictable(page))) {
L
Lee Schermerhorn 已提交
1604 1605 1606 1607
			putback_lru_page(page);
			continue;
		}

1608 1609 1610 1611 1612 1613 1614 1615
		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);
			}
		}

1616 1617
		if (page_referenced(page, 0, sc->target_mem_cgroup,
				    &vm_flags)) {
1618
			nr_rotated += hpage_nr_pages(page);
1619 1620 1621 1622 1623 1624 1625 1626 1627
			/*
			 * 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.
			 */
1628
			if ((vm_flags & VM_EXEC) && page_is_file_cache(page)) {
1629 1630 1631 1632
				list_add(&page->lru, &l_active);
				continue;
			}
		}
1633

1634
		ClearPageActive(page);	/* we are de-activating */
L
Linus Torvalds 已提交
1635 1636 1637
		list_add(&page->lru, &l_inactive);
	}

1638
	/*
1639
	 * Move pages back to the lru list.
1640
	 */
1641
	spin_lock_irq(&zone->lru_lock);
1642
	/*
1643 1644 1645 1646
	 * 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.
1647
	 */
1648
	reclaim_stat->recent_rotated[file] += nr_rotated;
1649

1650 1651
	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
1652
	__mod_zone_page_state(zone, NR_ISOLATED_ANON + file, -nr_taken);
1653
	spin_unlock_irq(&zone->lru_lock);
1654 1655

	free_hot_cold_page_list(&l_hold, 1);
L
Linus Torvalds 已提交
1656 1657
}

1658
#ifdef CONFIG_SWAP
1659
static int inactive_anon_is_low_global(struct zone *zone)
1660 1661 1662 1663 1664 1665 1666 1667 1668 1669 1670 1671
{
	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;
}

1672 1673
/**
 * inactive_anon_is_low - check if anonymous pages need to be deactivated
1674
 * @lruvec: LRU vector to check
1675 1676 1677 1678
 *
 * Returns true if the zone does not have enough inactive anon pages,
 * meaning some active anon pages need to be deactivated.
 */
1679
static int inactive_anon_is_low(struct lruvec *lruvec)
1680
{
1681 1682 1683 1684 1685 1686 1687
	/*
	 * If we don't have swap space, anonymous page deactivation
	 * is pointless.
	 */
	if (!total_swap_pages)
		return 0;

1688
	if (!mem_cgroup_disabled())
1689
		return mem_cgroup_inactive_anon_is_low(lruvec);
1690

1691
	return inactive_anon_is_low_global(lruvec_zone(lruvec));
1692
}
1693
#else
1694
static inline int inactive_anon_is_low(struct lruvec *lruvec)
1695 1696 1697 1698
{
	return 0;
}
#endif
1699

1700 1701
/**
 * inactive_file_is_low - check if file pages need to be deactivated
1702
 * @lruvec: LRU vector to check
1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713
 *
 * 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.
 */
1714
static int inactive_file_is_low(struct lruvec *lruvec)
1715
{
1716 1717 1718 1719 1720
	unsigned long inactive;
	unsigned long active;

	inactive = get_lru_size(lruvec, LRU_INACTIVE_FILE);
	active = get_lru_size(lruvec, LRU_ACTIVE_FILE);
1721

1722
	return active > inactive;
1723 1724
}

H
Hugh Dickins 已提交
1725
static int inactive_list_is_low(struct lruvec *lruvec, enum lru_list lru)
1726
{
H
Hugh Dickins 已提交
1727
	if (is_file_lru(lru))
1728
		return inactive_file_is_low(lruvec);
1729
	else
1730
		return inactive_anon_is_low(lruvec);
1731 1732
}

1733
static unsigned long shrink_list(enum lru_list lru, unsigned long nr_to_scan,
1734
				 struct lruvec *lruvec, struct scan_control *sc)
1735
{
1736
	if (is_active_lru(lru)) {
H
Hugh Dickins 已提交
1737
		if (inactive_list_is_low(lruvec, lru))
1738
			shrink_active_list(nr_to_scan, lruvec, sc, lru);
1739 1740 1741
		return 0;
	}

1742
	return shrink_inactive_list(nr_to_scan, lruvec, sc, lru);
1743 1744
}

1745
static int vmscan_swappiness(struct scan_control *sc)
1746
{
1747
	if (global_reclaim(sc))
1748
		return vm_swappiness;
1749
	return mem_cgroup_swappiness(sc->target_mem_cgroup);
1750 1751
}

1752 1753 1754 1755 1756 1757 1758
enum scan_balance {
	SCAN_EQUAL,
	SCAN_FRACT,
	SCAN_ANON,
	SCAN_FILE,
};

1759 1760 1761 1762 1763 1764
/*
 * 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
1765 1766
 * 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
1767
 */
1768
static void get_scan_count(struct lruvec *lruvec, struct scan_control *sc,
1769
			   unsigned long *nr)
1770
{
1771 1772 1773 1774
	struct zone_reclaim_stat *reclaim_stat = &lruvec->reclaim_stat;
	u64 fraction[2];
	u64 denominator = 0;	/* gcc */
	struct zone *zone = lruvec_zone(lruvec);
1775
	unsigned long anon_prio, file_prio;
1776 1777 1778
	enum scan_balance scan_balance;
	unsigned long anon, file, free;
	bool force_scan = false;
1779
	unsigned long ap, fp;
H
Hugh Dickins 已提交
1780
	enum lru_list lru;
1781

1782 1783 1784 1785 1786 1787 1788 1789 1790 1791
	/*
	 * 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.
	 */
1792
	if (current_is_kswapd() && zone->all_unreclaimable)
1793
		force_scan = true;
1794
	if (!global_reclaim(sc))
1795
		force_scan = true;
1796 1797

	/* If we have no swap space, do not bother scanning anon pages. */
1798
	if (!sc->may_swap || (get_nr_swap_pages() <= 0)) {
1799
		scan_balance = SCAN_FILE;
1800 1801
		goto out;
	}
1802

1803 1804 1805 1806 1807 1808 1809 1810
	/*
	 * Global reclaim will swap to prevent OOM even with no
	 * swappiness, but memcg users want to use this knob to
	 * disable swapping for individual groups completely when
	 * using the memory controller's swap limit feature would be
	 * too expensive.
	 */
	if (!global_reclaim(sc) && !vmscan_swappiness(sc)) {
1811
		scan_balance = SCAN_FILE;
1812 1813 1814 1815 1816 1817 1818 1819 1820
		goto out;
	}

	/*
	 * Do not apply any pressure balancing cleverness when the
	 * system is close to OOM, scan both anon and file equally
	 * (unless the swappiness setting disagrees with swapping).
	 */
	if (!sc->priority && vmscan_swappiness(sc)) {
1821
		scan_balance = SCAN_EQUAL;
1822 1823 1824
		goto out;
	}

1825 1826 1827 1828
	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);
1829

1830 1831 1832 1833 1834 1835
	/*
	 * If it's foreseeable that reclaiming the file cache won't be
	 * enough to get the zone back into a desirable shape, we have
	 * to swap.  Better start now and leave the - probably heavily
	 * thrashing - remaining file pages alone.
	 */
1836
	if (global_reclaim(sc)) {
1837
		free = zone_page_state(zone, NR_FREE_PAGES);
1838
		if (unlikely(file + free <= high_wmark_pages(zone))) {
1839
			scan_balance = SCAN_ANON;
1840
			goto out;
1841
		}
1842 1843
	}

1844 1845 1846 1847 1848
	/*
	 * There is enough inactive page cache, do not reclaim
	 * anything from the anonymous working set right now.
	 */
	if (!inactive_file_is_low(lruvec)) {
1849
		scan_balance = SCAN_FILE;
1850 1851 1852
		goto out;
	}

1853 1854
	scan_balance = SCAN_FRACT;

1855 1856 1857 1858
	/*
	 * With swappiness at 100, anonymous and file have the same priority.
	 * This scanning priority is essentially the inverse of IO cost.
	 */
1859
	anon_prio = vmscan_swappiness(sc);
H
Hugh Dickins 已提交
1860
	file_prio = 200 - anon_prio;
1861

1862 1863 1864 1865 1866 1867 1868 1869 1870 1871 1872
	/*
	 * 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]
	 */
1873
	spin_lock_irq(&zone->lru_lock);
1874 1875 1876
	if (unlikely(reclaim_stat->recent_scanned[0] > anon / 4)) {
		reclaim_stat->recent_scanned[0] /= 2;
		reclaim_stat->recent_rotated[0] /= 2;
1877 1878
	}

1879 1880 1881
	if (unlikely(reclaim_stat->recent_scanned[1] > file / 4)) {
		reclaim_stat->recent_scanned[1] /= 2;
		reclaim_stat->recent_rotated[1] /= 2;
1882 1883 1884
	}

	/*
1885 1886 1887
	 * 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.
1888
	 */
1889
	ap = anon_prio * (reclaim_stat->recent_scanned[0] + 1);
1890
	ap /= reclaim_stat->recent_rotated[0] + 1;
1891

1892
	fp = file_prio * (reclaim_stat->recent_scanned[1] + 1);
1893
	fp /= reclaim_stat->recent_rotated[1] + 1;
1894
	spin_unlock_irq(&zone->lru_lock);
1895

1896 1897 1898 1899
	fraction[0] = ap;
	fraction[1] = fp;
	denominator = ap + fp + 1;
out:
H
Hugh Dickins 已提交
1900 1901
	for_each_evictable_lru(lru) {
		int file = is_file_lru(lru);
1902
		unsigned long size;
1903
		unsigned long scan;
1904

1905
		size = get_lru_size(lruvec, lru);
1906
		scan = size >> sc->priority;
1907

1908 1909
		if (!scan && force_scan)
			scan = min(size, SWAP_CLUSTER_MAX);
1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931

		switch (scan_balance) {
		case SCAN_EQUAL:
			/* Scan lists relative to size */
			break;
		case SCAN_FRACT:
			/*
			 * Scan types proportional to swappiness and
			 * their relative recent reclaim efficiency.
			 */
			scan = div64_u64(scan * fraction[file], denominator);
			break;
		case SCAN_FILE:
		case SCAN_ANON:
			/* Scan one type exclusively */
			if ((scan_balance == SCAN_FILE) != file)
				scan = 0;
			break;
		default:
			/* Look ma, no brain */
			BUG();
		}
H
Hugh Dickins 已提交
1932
		nr[lru] = scan;
1933
	}
1934
}
1935

1936 1937 1938 1939 1940 1941
/*
 * This is a basic per-zone page freer.  Used by both kswapd and direct reclaim.
 */
static void shrink_lruvec(struct lruvec *lruvec, struct scan_control *sc)
{
	unsigned long nr[NR_LRU_LISTS];
1942
	unsigned long targets[NR_LRU_LISTS];
1943 1944 1945 1946 1947
	unsigned long nr_to_scan;
	enum lru_list lru;
	unsigned long nr_reclaimed = 0;
	unsigned long nr_to_reclaim = sc->nr_to_reclaim;
	struct blk_plug plug;
1948
	bool scan_adjusted = false;
1949 1950 1951

	get_scan_count(lruvec, sc, nr);

1952 1953 1954
	/* Record the original scan target for proportional adjustments later */
	memcpy(targets, nr, sizeof(nr));

1955 1956 1957
	blk_start_plug(&plug);
	while (nr[LRU_INACTIVE_ANON] || nr[LRU_ACTIVE_FILE] ||
					nr[LRU_INACTIVE_FILE]) {
1958 1959 1960
		unsigned long nr_anon, nr_file, percentage;
		unsigned long nr_scanned;

1961 1962 1963 1964 1965 1966 1967 1968 1969
		for_each_evictable_lru(lru) {
			if (nr[lru]) {
				nr_to_scan = min(nr[lru], SWAP_CLUSTER_MAX);
				nr[lru] -= nr_to_scan;

				nr_reclaimed += shrink_list(lru, nr_to_scan,
							    lruvec, sc);
			}
		}
1970 1971 1972 1973

		if (nr_reclaimed < nr_to_reclaim || scan_adjusted)
			continue;

1974
		/*
1975 1976 1977 1978
		 * For global direct reclaim, reclaim only the number of pages
		 * requested. Less care is taken to scan proportionally as it
		 * is more important to minimise direct reclaim stall latency
		 * than it is to properly age the LRU lists.
1979
		 */
1980
		if (global_reclaim(sc) && !current_is_kswapd())
1981
			break;
1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023

		/*
		 * For kswapd and memcg, reclaim at least the number of pages
		 * requested. Ensure that the anon and file LRUs shrink
		 * proportionally what was requested by get_scan_count(). We
		 * stop reclaiming one LRU and reduce the amount scanning
		 * proportional to the original scan target.
		 */
		nr_file = nr[LRU_INACTIVE_FILE] + nr[LRU_ACTIVE_FILE];
		nr_anon = nr[LRU_INACTIVE_ANON] + nr[LRU_ACTIVE_ANON];

		if (nr_file > nr_anon) {
			unsigned long scan_target = targets[LRU_INACTIVE_ANON] +
						targets[LRU_ACTIVE_ANON] + 1;
			lru = LRU_BASE;
			percentage = nr_anon * 100 / scan_target;
		} else {
			unsigned long scan_target = targets[LRU_INACTIVE_FILE] +
						targets[LRU_ACTIVE_FILE] + 1;
			lru = LRU_FILE;
			percentage = nr_file * 100 / scan_target;
		}

		/* Stop scanning the smaller of the LRU */
		nr[lru] = 0;
		nr[lru + LRU_ACTIVE] = 0;

		/*
		 * Recalculate the other LRU scan count based on its original
		 * scan target and the percentage scanning already complete
		 */
		lru = (lru == LRU_FILE) ? LRU_BASE : LRU_FILE;
		nr_scanned = targets[lru] - nr[lru];
		nr[lru] = targets[lru] * (100 - percentage) / 100;
		nr[lru] -= min(nr[lru], nr_scanned);

		lru += LRU_ACTIVE;
		nr_scanned = targets[lru] - nr[lru];
		nr[lru] = targets[lru] * (100 - percentage) / 100;
		nr[lru] -= min(nr[lru], nr_scanned);

		scan_adjusted = true;
2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037 2038
	}
	blk_finish_plug(&plug);
	sc->nr_reclaimed += nr_reclaimed;

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

	throttle_vm_writeout(sc->gfp_mask);
}

M
Mel Gorman 已提交
2039
/* Use reclaim/compaction for costly allocs or under memory pressure */
2040
static bool in_reclaim_compaction(struct scan_control *sc)
M
Mel Gorman 已提交
2041
{
2042
	if (IS_ENABLED(CONFIG_COMPACTION) && sc->order &&
M
Mel Gorman 已提交
2043
			(sc->order > PAGE_ALLOC_COSTLY_ORDER ||
2044
			 sc->priority < DEF_PRIORITY - 2))
M
Mel Gorman 已提交
2045 2046 2047 2048 2049
		return true;

	return false;
}

2050
/*
M
Mel Gorman 已提交
2051 2052 2053 2054 2055
 * 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.
2056
 */
2057
static inline bool should_continue_reclaim(struct zone *zone,
2058 2059 2060 2061 2062 2063 2064 2065
					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 */
2066
	if (!in_reclaim_compaction(sc))
2067 2068
		return false;

2069 2070 2071 2072 2073 2074 2075 2076 2077 2078 2079 2080 2081 2082 2083 2084 2085 2086 2087 2088 2089 2090
	/* 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;
	}
2091 2092 2093 2094 2095 2096

	/*
	 * If we have not reclaimed enough pages for compaction and the
	 * inactive lists are large enough, continue reclaiming
	 */
	pages_for_compaction = (2UL << sc->order);
2097
	inactive_lru_pages = zone_page_state(zone, NR_INACTIVE_FILE);
2098
	if (get_nr_swap_pages() > 0)
2099
		inactive_lru_pages += zone_page_state(zone, NR_INACTIVE_ANON);
2100 2101 2102 2103 2104
	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 */
2105
	switch (compaction_suitable(zone, sc->order)) {
2106 2107 2108 2109 2110 2111 2112 2113
	case COMPACT_PARTIAL:
	case COMPACT_CONTINUE:
		return false;
	default:
		return true;
	}
}

2114
static void shrink_zone(struct zone *zone, struct scan_control *sc)
L
Linus Torvalds 已提交
2115
{
2116
	unsigned long nr_reclaimed, nr_scanned;
L
Linus Torvalds 已提交
2117

2118 2119 2120 2121 2122 2123 2124
	do {
		struct mem_cgroup *root = sc->target_mem_cgroup;
		struct mem_cgroup_reclaim_cookie reclaim = {
			.zone = zone,
			.priority = sc->priority,
		};
		struct mem_cgroup *memcg;
2125

2126 2127
		nr_reclaimed = sc->nr_reclaimed;
		nr_scanned = sc->nr_scanned;
L
Linus Torvalds 已提交
2128

2129 2130 2131
		memcg = mem_cgroup_iter(root, NULL, &reclaim);
		do {
			struct lruvec *lruvec;
2132

2133
			lruvec = mem_cgroup_zone_lruvec(zone, memcg);
2134

2135
			shrink_lruvec(lruvec, sc);
2136

2137
			/*
2138 2139
			 * Direct reclaim and kswapd have to scan all memory
			 * cgroups to fulfill the overall scan target for the
2140
			 * zone.
2141 2142 2143 2144 2145
			 *
			 * Limit reclaim, on the other hand, only cares about
			 * nr_to_reclaim pages to be reclaimed and it will
			 * retry with decreasing priority if one round over the
			 * whole hierarchy is not sufficient.
2146
			 */
2147 2148
			if (!global_reclaim(sc) &&
					sc->nr_reclaimed >= sc->nr_to_reclaim) {
2149 2150 2151 2152 2153
				mem_cgroup_iter_break(root, memcg);
				break;
			}
			memcg = mem_cgroup_iter(root, memcg, &reclaim);
		} while (memcg);
2154 2155 2156 2157 2158

		vmpressure(sc->gfp_mask, sc->target_mem_cgroup,
			   sc->nr_scanned - nr_scanned,
			   sc->nr_reclaimed - nr_reclaimed);

2159 2160
	} while (should_continue_reclaim(zone, sc->nr_reclaimed - nr_reclaimed,
					 sc->nr_scanned - nr_scanned, sc));
2161 2162
}

2163 2164 2165 2166 2167 2168 2169 2170 2171 2172 2173 2174 2175 2176 2177 2178 2179
/* 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),
2180
		(zone->managed_pages + KSWAPD_ZONE_BALANCE_GAP_RATIO-1) /
2181 2182 2183 2184 2185 2186 2187 2188
			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
	 */
2189
	if (compaction_deferred(zone, sc->order))
2190 2191 2192 2193 2194 2195 2196 2197 2198
		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
2199 2200 2201 2202 2203
/*
 * 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.
 *
2204 2205
 * We reclaim from a zone even if that zone is over high_wmark_pages(zone).
 * Because:
L
Linus Torvalds 已提交
2206 2207
 * a) The caller may be trying to free *extra* pages to satisfy a higher-order
 *    allocation or
2208 2209 2210
 * 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
2211 2212 2213
 *
 * If a zone is deemed to be full of pinned pages then just give it a light
 * scan then give up on it.
2214 2215
 *
 * This function returns true if a zone is being reclaimed for a costly
2216
 * high-order allocation and compaction is ready to begin. This indicates to
2217 2218
 * the caller that it should consider retrying the allocation instead of
 * further reclaim.
L
Linus Torvalds 已提交
2219
 */
2220
static bool shrink_zones(struct zonelist *zonelist, struct scan_control *sc)
L
Linus Torvalds 已提交
2221
{
2222
	struct zoneref *z;
2223
	struct zone *zone;
2224 2225
	unsigned long nr_soft_reclaimed;
	unsigned long nr_soft_scanned;
2226
	bool aborted_reclaim = false;
2227

2228 2229 2230 2231 2232 2233 2234 2235
	/*
	 * 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;

2236 2237
	for_each_zone_zonelist_nodemask(zone, z, zonelist,
					gfp_zone(sc->gfp_mask), sc->nodemask) {
2238
		if (!populated_zone(zone))
L
Linus Torvalds 已提交
2239
			continue;
2240 2241 2242 2243
		/*
		 * Take care memory controller reclaiming has small influence
		 * to global LRU.
		 */
2244
		if (global_reclaim(sc)) {
2245 2246
			if (!cpuset_zone_allowed_hardwall(zone, GFP_KERNEL))
				continue;
2247 2248
			if (zone->all_unreclaimable &&
					sc->priority != DEF_PRIORITY)
2249
				continue;	/* Let kswapd poll it */
2250
			if (IS_ENABLED(CONFIG_COMPACTION)) {
2251
				/*
2252 2253 2254 2255 2256
				 * 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
2257 2258
				 * noticeable problem, like transparent huge
				 * page allocations.
2259
				 */
2260
				if (compaction_ready(zone, sc)) {
2261
					aborted_reclaim = true;
2262
					continue;
2263
				}
2264
			}
2265 2266 2267 2268 2269 2270 2271 2272 2273 2274 2275 2276 2277
			/*
			 * 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() */
2278
		}
2279

2280
		shrink_zone(zone, sc);
L
Linus Torvalds 已提交
2281
	}
2282

2283
	return aborted_reclaim;
2284 2285 2286 2287 2288 2289 2290
}

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

2291
/* All zones in zonelist are unreclaimable? */
2292 2293 2294 2295 2296 2297 2298 2299 2300 2301 2302 2303
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;
2304 2305
		if (!zone->all_unreclaimable)
			return false;
2306 2307
	}

2308
	return true;
L
Linus Torvalds 已提交
2309
}
2310

L
Linus Torvalds 已提交
2311 2312 2313 2314 2315 2316 2317 2318
/*
 * 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
2319 2320 2321 2322
 * 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.
2323 2324 2325
 *
 * returns:	0, if no pages reclaimed
 * 		else, the number of pages reclaimed
L
Linus Torvalds 已提交
2326
 */
2327
static unsigned long do_try_to_free_pages(struct zonelist *zonelist,
2328 2329
					struct scan_control *sc,
					struct shrink_control *shrink)
L
Linus Torvalds 已提交
2330
{
2331
	unsigned long total_scanned = 0;
L
Linus Torvalds 已提交
2332
	struct reclaim_state *reclaim_state = current->reclaim_state;
2333
	struct zoneref *z;
2334
	struct zone *zone;
2335
	unsigned long writeback_threshold;
2336
	bool aborted_reclaim;
L
Linus Torvalds 已提交
2337

2338 2339
	delayacct_freepages_start();

2340
	if (global_reclaim(sc))
2341
		count_vm_event(ALLOCSTALL);
L
Linus Torvalds 已提交
2342

2343
	do {
2344 2345
		vmpressure_prio(sc->gfp_mask, sc->target_mem_cgroup,
				sc->priority);
2346
		sc->nr_scanned = 0;
2347
		aborted_reclaim = shrink_zones(zonelist, sc);
2348

2349
		/*
2350 2351 2352 2353
		 * Don't shrink slabs when reclaiming memory from over limit
		 * cgroups but do shrink slab at least once when aborting
		 * reclaim for compaction to avoid unevenly scanning file/anon
		 * LRU pages over slab pages.
2354
		 */
2355
		if (global_reclaim(sc)) {
2356
			unsigned long lru_pages = 0;
2357 2358
			for_each_zone_zonelist(zone, z, zonelist,
					gfp_zone(sc->gfp_mask)) {
2359 2360 2361 2362 2363 2364
				if (!cpuset_zone_allowed_hardwall(zone, GFP_KERNEL))
					continue;

				lru_pages += zone_reclaimable_pages(zone);
			}

2365
			shrink_slab(shrink, sc->nr_scanned, lru_pages);
2366
			if (reclaim_state) {
2367
				sc->nr_reclaimed += reclaim_state->reclaimed_slab;
2368 2369
				reclaim_state->reclaimed_slab = 0;
			}
L
Linus Torvalds 已提交
2370
		}
2371
		total_scanned += sc->nr_scanned;
2372
		if (sc->nr_reclaimed >= sc->nr_to_reclaim)
L
Linus Torvalds 已提交
2373 2374
			goto out;

2375 2376 2377 2378 2379 2380 2381
		/*
		 * If we're getting trouble reclaiming, start doing
		 * writepage even in laptop mode.
		 */
		if (sc->priority < DEF_PRIORITY - 2)
			sc->may_writepage = 1;

L
Linus Torvalds 已提交
2382 2383 2384 2385 2386 2387 2388
		/*
		 * 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.
		 */
2389 2390
		writeback_threshold = sc->nr_to_reclaim + sc->nr_to_reclaim / 2;
		if (total_scanned > writeback_threshold) {
2391 2392
			wakeup_flusher_threads(laptop_mode ? 0 : total_scanned,
						WB_REASON_TRY_TO_FREE_PAGES);
2393
			sc->may_writepage = 1;
L
Linus Torvalds 已提交
2394
		}
2395
	} while (--sc->priority >= 0 && !aborted_reclaim);
2396

L
Linus Torvalds 已提交
2397
out:
2398 2399
	delayacct_freepages_end();

2400 2401 2402
	if (sc->nr_reclaimed)
		return sc->nr_reclaimed;

2403 2404 2405 2406 2407 2408 2409 2410
	/*
	 * 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;

2411 2412
	/* Aborted reclaim to try compaction? don't OOM, then */
	if (aborted_reclaim)
2413 2414
		return 1;

2415
	/* top priority shrink_zones still had more to do? don't OOM, then */
2416
	if (global_reclaim(sc) && !all_unreclaimable(zonelist, sc))
2417 2418 2419
		return 1;

	return 0;
L
Linus Torvalds 已提交
2420 2421
}

2422 2423 2424 2425 2426 2427 2428 2429 2430 2431 2432 2433 2434 2435 2436 2437 2438 2439 2440 2441 2442 2443 2444 2445 2446 2447 2448 2449 2450 2451
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
2452 2453 2454 2455
 * 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.
2456
 */
2457
static bool throttle_direct_reclaim(gfp_t gfp_mask, struct zonelist *zonelist,
2458 2459 2460 2461 2462 2463 2464 2465 2466 2467 2468 2469 2470 2471
					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)
2472 2473 2474 2475 2476 2477 2478 2479
		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;
2480 2481 2482 2483 2484

	/* Check if the pfmemalloc reserves are ok */
	first_zones_zonelist(zonelist, high_zoneidx, NULL, &zone);
	pgdat = zone->zone_pgdat;
	if (pfmemalloc_watermark_ok(pgdat))
2485
		goto out;
2486

2487 2488 2489
	/* Account for the throttling */
	count_vm_event(PGSCAN_DIRECT_THROTTLE);

2490 2491 2492 2493 2494 2495 2496 2497 2498 2499 2500
	/*
	 * 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);
2501 2502

		goto check_pending;
2503 2504 2505 2506 2507
	}

	/* Throttle until kswapd wakes the process */
	wait_event_killable(zone->zone_pgdat->pfmemalloc_wait,
		pfmemalloc_watermark_ok(pgdat));
2508 2509 2510 2511 2512 2513 2514

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

out:
	return false;
2515 2516
}

2517
unsigned long try_to_free_pages(struct zonelist *zonelist, int order,
2518
				gfp_t gfp_mask, nodemask_t *nodemask)
2519
{
2520
	unsigned long nr_reclaimed;
2521
	struct scan_control sc = {
2522
		.gfp_mask = (gfp_mask = memalloc_noio_flags(gfp_mask)),
2523
		.may_writepage = !laptop_mode,
2524
		.nr_to_reclaim = SWAP_CLUSTER_MAX,
2525
		.may_unmap = 1,
2526
		.may_swap = 1,
2527
		.order = order,
2528
		.priority = DEF_PRIORITY,
2529
		.target_mem_cgroup = NULL,
2530
		.nodemask = nodemask,
2531
	};
2532 2533 2534
	struct shrink_control shrink = {
		.gfp_mask = sc.gfp_mask,
	};
2535

2536
	/*
2537 2538 2539
	 * 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.
2540
	 */
2541
	if (throttle_direct_reclaim(gfp_mask, zonelist, nodemask))
2542 2543
		return 1;

2544 2545 2546 2547
	trace_mm_vmscan_direct_reclaim_begin(order,
				sc.may_writepage,
				gfp_mask);

2548
	nr_reclaimed = do_try_to_free_pages(zonelist, &sc, &shrink);
2549 2550 2551 2552

	trace_mm_vmscan_direct_reclaim_end(nr_reclaimed);

	return nr_reclaimed;
2553 2554
}

A
Andrew Morton 已提交
2555
#ifdef CONFIG_MEMCG
2556

2557
unsigned long mem_cgroup_shrink_node_zone(struct mem_cgroup *memcg,
2558
						gfp_t gfp_mask, bool noswap,
2559 2560
						struct zone *zone,
						unsigned long *nr_scanned)
2561 2562
{
	struct scan_control sc = {
2563
		.nr_scanned = 0,
2564
		.nr_to_reclaim = SWAP_CLUSTER_MAX,
2565 2566 2567 2568
		.may_writepage = !laptop_mode,
		.may_unmap = 1,
		.may_swap = !noswap,
		.order = 0,
2569
		.priority = 0,
2570
		.target_mem_cgroup = memcg,
2571
	};
2572
	struct lruvec *lruvec = mem_cgroup_zone_lruvec(zone, memcg);
2573

2574 2575
	sc.gfp_mask = (gfp_mask & GFP_RECLAIM_MASK) |
			(GFP_HIGHUSER_MOVABLE & ~GFP_RECLAIM_MASK);
2576

2577
	trace_mm_vmscan_memcg_softlimit_reclaim_begin(sc.order,
2578 2579 2580
						      sc.may_writepage,
						      sc.gfp_mask);

2581 2582 2583 2584 2585 2586 2587
	/*
	 * 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.
	 */
2588
	shrink_lruvec(lruvec, &sc);
2589 2590 2591

	trace_mm_vmscan_memcg_softlimit_reclaim_end(sc.nr_reclaimed);

2592
	*nr_scanned = sc.nr_scanned;
2593 2594 2595
	return sc.nr_reclaimed;
}

2596
unsigned long try_to_free_mem_cgroup_pages(struct mem_cgroup *memcg,
K
KOSAKI Motohiro 已提交
2597
					   gfp_t gfp_mask,
2598
					   bool noswap)
2599
{
2600
	struct zonelist *zonelist;
2601
	unsigned long nr_reclaimed;
2602
	int nid;
2603 2604
	struct scan_control sc = {
		.may_writepage = !laptop_mode,
2605
		.may_unmap = 1,
2606
		.may_swap = !noswap,
2607
		.nr_to_reclaim = SWAP_CLUSTER_MAX,
2608
		.order = 0,
2609
		.priority = DEF_PRIORITY,
2610
		.target_mem_cgroup = memcg,
2611
		.nodemask = NULL, /* we don't care the placement */
2612 2613 2614 2615 2616
		.gfp_mask = (gfp_mask & GFP_RECLAIM_MASK) |
				(GFP_HIGHUSER_MOVABLE & ~GFP_RECLAIM_MASK),
	};
	struct shrink_control shrink = {
		.gfp_mask = sc.gfp_mask,
2617 2618
	};

2619 2620 2621 2622 2623
	/*
	 * 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.
	 */
2624
	nid = mem_cgroup_select_victim_node(memcg);
2625 2626

	zonelist = NODE_DATA(nid)->node_zonelists;
2627 2628 2629 2630 2631

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

2632
	nr_reclaimed = do_try_to_free_pages(zonelist, &sc, &shrink);
2633 2634 2635 2636

	trace_mm_vmscan_memcg_reclaim_end(nr_reclaimed);

	return nr_reclaimed;
2637 2638 2639
}
#endif

2640
static void age_active_anon(struct zone *zone, struct scan_control *sc)
2641
{
2642
	struct mem_cgroup *memcg;
2643

2644 2645 2646 2647 2648
	if (!total_swap_pages)
		return;

	memcg = mem_cgroup_iter(NULL, NULL, NULL);
	do {
2649
		struct lruvec *lruvec = mem_cgroup_zone_lruvec(zone, memcg);
2650

2651
		if (inactive_anon_is_low(lruvec))
2652
			shrink_active_list(SWAP_CLUSTER_MAX, lruvec,
2653
					   sc, LRU_ACTIVE_ANON);
2654 2655 2656

		memcg = mem_cgroup_iter(NULL, memcg, NULL);
	} while (memcg);
2657 2658
}

2659 2660 2661 2662 2663 2664 2665
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;

2666 2667
	if (IS_ENABLED(CONFIG_COMPACTION) && order &&
	    !compaction_suitable(zone, order))
2668 2669 2670 2671 2672
		return false;

	return true;
}

2673
/*
2674 2675 2676 2677 2678 2679 2680 2681 2682 2683
 * pgdat_balanced() is used when checking if a node is balanced.
 *
 * For order-0, all zones must be 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.
2684 2685 2686 2687
 * 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
2688
 *     percentage of the middle zones. For example, on 32-bit x86, highmem
2689 2690 2691 2692
 *     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.
 */
2693
static bool pgdat_balanced(pg_data_t *pgdat, int order, int classzone_idx)
2694
{
2695
	unsigned long managed_pages = 0;
2696
	unsigned long balanced_pages = 0;
2697 2698
	int i;

2699 2700 2701
	/* Check the watermark levels */
	for (i = 0; i <= classzone_idx; i++) {
		struct zone *zone = pgdat->node_zones + i;
2702

2703 2704 2705
		if (!populated_zone(zone))
			continue;

2706
		managed_pages += zone->managed_pages;
2707 2708 2709 2710 2711 2712 2713 2714 2715

		/*
		 * A special case here:
		 *
		 * balance_pgdat() skips over all_unreclaimable after
		 * DEF_PRIORITY. Effectively, it considers them balanced so
		 * they must be considered balanced here as well!
		 */
		if (zone->all_unreclaimable) {
2716
			balanced_pages += zone->managed_pages;
2717 2718 2719 2720
			continue;
		}

		if (zone_balanced(zone, order, 0, i))
2721
			balanced_pages += zone->managed_pages;
2722 2723 2724 2725 2726
		else if (!order)
			return false;
	}

	if (order)
2727
		return balanced_pages >= (managed_pages >> 2);
2728 2729
	else
		return true;
2730 2731
}

2732 2733 2734 2735 2736 2737 2738
/*
 * 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,
2739
					int classzone_idx)
2740 2741 2742
{
	/* If a direct reclaimer woke kswapd within HZ/10, it's premature */
	if (remaining)
2743 2744 2745 2746 2747 2748 2749 2750 2751 2752 2753 2754 2755 2756 2757
		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;
	}
2758

2759
	return pgdat_balanced(pgdat, order, classzone_idx);
2760 2761
}

2762 2763 2764
/*
 * kswapd shrinks the zone by the number of pages required to reach
 * the high watermark.
2765 2766
 *
 * Returns true if kswapd scanned at least the requested number of pages to
2767 2768
 * reclaim or if the lack of progress was due to pages under writeback.
 * This is used to determine if the scanning priority needs to be raised.
2769
 */
2770
static bool kswapd_shrink_zone(struct zone *zone,
2771
			       int classzone_idx,
2772
			       struct scan_control *sc,
2773 2774
			       unsigned long lru_pages,
			       unsigned long *nr_attempted)
2775 2776
{
	unsigned long nr_slab;
2777 2778
	int testorder = sc->order;
	unsigned long balance_gap;
2779 2780 2781 2782
	struct reclaim_state *reclaim_state = current->reclaim_state;
	struct shrink_control shrink = {
		.gfp_mask = sc->gfp_mask,
	};
2783
	bool lowmem_pressure;
2784 2785 2786

	/* Reclaim above the high watermark. */
	sc->nr_to_reclaim = max(SWAP_CLUSTER_MAX, high_wmark_pages(zone));
2787 2788 2789 2790 2791 2792 2793 2794 2795 2796 2797 2798 2799 2800 2801 2802 2803 2804 2805 2806 2807 2808 2809 2810 2811 2812 2813 2814 2815 2816 2817

	/*
	 * 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.
	 */
	if (IS_ENABLED(CONFIG_COMPACTION) && sc->order &&
			compaction_suitable(zone, sc->order) !=
				COMPACT_SKIPPED)
		testorder = 0;

	/*
	 * 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.
	 */
	balance_gap = min(low_wmark_pages(zone),
		(zone->managed_pages + KSWAPD_ZONE_BALANCE_GAP_RATIO-1) /
		KSWAPD_ZONE_BALANCE_GAP_RATIO);

	/*
	 * If there is no low memory pressure or the zone is balanced then no
	 * reclaim is necessary
	 */
	lowmem_pressure = (buffer_heads_over_limit && is_highmem(zone));
	if (!lowmem_pressure && zone_balanced(zone, testorder,
						balance_gap, classzone_idx))
		return true;

2818 2819 2820 2821 2822 2823
	shrink_zone(zone, sc);

	reclaim_state->reclaimed_slab = 0;
	nr_slab = shrink_slab(&shrink, sc->nr_scanned, lru_pages);
	sc->nr_reclaimed += reclaim_state->reclaimed_slab;

2824 2825 2826
	/* Account for the number of pages attempted to reclaim */
	*nr_attempted += sc->nr_to_reclaim;

2827 2828
	if (nr_slab == 0 && !zone_reclaimable(zone))
		zone->all_unreclaimable = 1;
2829

2830 2831
	zone_clear_flag(zone, ZONE_WRITEBACK);

2832 2833 2834 2835 2836 2837 2838 2839 2840 2841 2842 2843
	/*
	 * 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, speculatively avoid congestion
	 * waits.
	 */
	if (!zone->all_unreclaimable &&
	    zone_balanced(zone, testorder, 0, classzone_idx)) {
		zone_clear_flag(zone, ZONE_CONGESTED);
		zone_clear_flag(zone, ZONE_TAIL_LRU_DIRTY);
	}

2844
	return sc->nr_scanned >= sc->nr_to_reclaim;
2845 2846
}

L
Linus Torvalds 已提交
2847 2848
/*
 * For kswapd, balance_pgdat() will work across all this node's zones until
2849
 * they are all at high_wmark_pages(zone).
L
Linus Torvalds 已提交
2850
 *
2851
 * Returns the final order kswapd was reclaiming at
L
Linus Torvalds 已提交
2852 2853 2854 2855 2856 2857 2858 2859 2860 2861
 *
 * 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
2862 2863 2864 2865 2866
 * 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
2867
 */
2868
static unsigned long balance_pgdat(pg_data_t *pgdat, int order,
2869
							int *classzone_idx)
L
Linus Torvalds 已提交
2870 2871
{
	int i;
2872
	int end_zone = 0;	/* Inclusive.  0 = ZONE_DMA */
2873 2874
	unsigned long nr_soft_reclaimed;
	unsigned long nr_soft_scanned;
2875 2876
	struct scan_control sc = {
		.gfp_mask = GFP_KERNEL,
2877
		.priority = DEF_PRIORITY,
2878
		.may_unmap = 1,
2879
		.may_swap = 1,
2880
		.may_writepage = !laptop_mode,
A
Andy Whitcroft 已提交
2881
		.order = order,
2882
		.target_mem_cgroup = NULL,
2883
	};
2884
	count_vm_event(PAGEOUTRUN);
L
Linus Torvalds 已提交
2885

2886
	do {
L
Linus Torvalds 已提交
2887
		unsigned long lru_pages = 0;
2888
		unsigned long nr_attempted = 0;
2889
		bool raise_priority = true;
2890
		bool pgdat_needs_compaction = (order > 0);
2891 2892

		sc.nr_reclaimed = 0;
L
Linus Torvalds 已提交
2893

2894 2895 2896 2897 2898 2899
		/*
		 * 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
2900

2901 2902
			if (!populated_zone(zone))
				continue;
L
Linus Torvalds 已提交
2903

2904 2905
			if (zone->all_unreclaimable &&
			    sc.priority != DEF_PRIORITY)
2906
				continue;
L
Linus Torvalds 已提交
2907

2908 2909 2910 2911
			/*
			 * Do some background aging of the anon list, to give
			 * pages a chance to be referenced before reclaiming.
			 */
2912
			age_active_anon(zone, &sc);
2913

2914 2915 2916 2917 2918 2919 2920 2921 2922 2923 2924
			/*
			 * 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;
			}

2925
			if (!zone_balanced(zone, order, 0, 0)) {
2926
				end_zone = i;
A
Andrew Morton 已提交
2927
				break;
2928
			} else {
2929 2930 2931 2932
				/*
				 * If balanced, clear the dirty and congested
				 * flags
				 */
2933
				zone_clear_flag(zone, ZONE_CONGESTED);
2934
				zone_clear_flag(zone, ZONE_TAIL_LRU_DIRTY);
L
Linus Torvalds 已提交
2935 2936
			}
		}
2937

2938
		if (i < 0)
A
Andrew Morton 已提交
2939 2940
			goto out;

L
Linus Torvalds 已提交
2941 2942 2943
		for (i = 0; i <= end_zone; i++) {
			struct zone *zone = pgdat->node_zones + i;

2944 2945 2946
			if (!populated_zone(zone))
				continue;

2947
			lru_pages += zone_reclaimable_pages(zone);
2948 2949 2950 2951 2952 2953 2954 2955 2956 2957 2958

			/*
			 * If any zone is currently balanced then kswapd will
			 * not call compaction as it is expected that the
			 * necessary pages are already available.
			 */
			if (pgdat_needs_compaction &&
					zone_watermark_ok(zone, order,
						low_wmark_pages(zone),
						*classzone_idx, 0))
				pgdat_needs_compaction = false;
L
Linus Torvalds 已提交
2959 2960
		}

2961 2962 2963 2964 2965 2966 2967
		/*
		 * If we're getting trouble reclaiming, start doing writepage
		 * even in laptop mode.
		 */
		if (sc.priority < DEF_PRIORITY - 2)
			sc.may_writepage = 1;

L
Linus Torvalds 已提交
2968 2969 2970 2971 2972 2973 2974 2975 2976 2977 2978 2979
		/*
		 * 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;

2980
			if (!populated_zone(zone))
L
Linus Torvalds 已提交
2981 2982
				continue;

2983 2984
			if (zone->all_unreclaimable &&
			    sc.priority != DEF_PRIORITY)
L
Linus Torvalds 已提交
2985 2986 2987
				continue;

			sc.nr_scanned = 0;
2988

2989
			nr_soft_scanned = 0;
2990 2991 2992
			/*
			 * Call soft limit reclaim before calling shrink_zone.
			 */
2993 2994 2995 2996
			nr_soft_reclaimed = mem_cgroup_soft_limit_reclaim(zone,
							order, sc.gfp_mask,
							&nr_soft_scanned);
			sc.nr_reclaimed += nr_soft_reclaimed;
2997

2998
			/*
2999 3000 3001 3002
			 * There should be no need to raise the scanning
			 * priority if enough pages are already being scanned
			 * that that high watermark would be met at 100%
			 * efficiency.
3003
			 */
3004 3005 3006
			if (kswapd_shrink_zone(zone, end_zone, &sc,
					lru_pages, &nr_attempted))
				raise_priority = false;
L
Linus Torvalds 已提交
3007
		}
3008 3009 3010 3011 3012 3013 3014 3015 3016 3017

		/*
		 * 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);

L
Linus Torvalds 已提交
3018
		/*
3019 3020 3021 3022 3023 3024
		 * Fragmentation may mean that the system cannot be rebalanced
		 * for high-order allocations in all zones. If twice the
		 * allocation size has been reclaimed and the zones are still
		 * not balanced then recheck the watermarks at order-0 to
		 * prevent kswapd reclaiming excessively. Assume that a
		 * process requested a high-order can direct reclaim/compact.
L
Linus Torvalds 已提交
3025
		 */
3026 3027
		if (order && sc.nr_reclaimed >= 2UL << order)
			order = sc.order = 0;
3028

3029 3030 3031
		/* Check if kswapd should be suspending */
		if (try_to_freeze() || kthread_should_stop())
			break;
3032

3033 3034 3035 3036 3037 3038 3039
		/*
		 * Compact if necessary and kswapd is reclaiming at least the
		 * high watermark number of pages as requsted
		 */
		if (pgdat_needs_compaction && sc.nr_reclaimed > nr_attempted)
			compact_pgdat(pgdat, order);

3040
		/*
3041 3042
		 * Raise priority if scanning rate is too low or there was no
		 * progress in reclaiming pages
3043
		 */
3044 3045
		if (raise_priority || !sc.nr_reclaimed)
			sc.priority--;
3046
	} while (sc.priority >= 1 &&
3047
		 !pgdat_balanced(pgdat, order, *classzone_idx));
L
Linus Torvalds 已提交
3048

3049
out:
3050
	/*
3051
	 * Return the order we were reclaiming at so prepare_kswapd_sleep()
3052 3053 3054 3055
	 * 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
	 */
3056
	*classzone_idx = end_zone;
3057
	return order;
L
Linus Torvalds 已提交
3058 3059
}

3060
static void kswapd_try_to_sleep(pg_data_t *pgdat, int order, int classzone_idx)
3061 3062 3063 3064 3065 3066 3067 3068 3069 3070
{
	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 */
3071
	if (prepare_kswapd_sleep(pgdat, order, remaining, classzone_idx)) {
3072 3073 3074 3075 3076 3077 3078 3079 3080
		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.
	 */
3081
	if (prepare_kswapd_sleep(pgdat, order, remaining, classzone_idx)) {
3082 3083 3084 3085 3086 3087 3088 3089 3090 3091 3092
		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);
3093

3094 3095 3096 3097 3098 3099 3100 3101
		/*
		 * 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);

3102 3103 3104
		if (!kthread_should_stop())
			schedule();

3105 3106 3107 3108 3109 3110 3111 3112 3113 3114
		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
3115 3116
/*
 * The background pageout daemon, started as a kernel thread
3117
 * from the init process.
L
Linus Torvalds 已提交
3118 3119 3120 3121 3122 3123 3124 3125 3126 3127 3128 3129
 *
 * 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)
{
3130
	unsigned long order, new_order;
3131
	unsigned balanced_order;
3132
	int classzone_idx, new_classzone_idx;
3133
	int balanced_classzone_idx;
L
Linus Torvalds 已提交
3134 3135
	pg_data_t *pgdat = (pg_data_t*)p;
	struct task_struct *tsk = current;
3136

L
Linus Torvalds 已提交
3137 3138 3139
	struct reclaim_state reclaim_state = {
		.reclaimed_slab = 0,
	};
3140
	const struct cpumask *cpumask = cpumask_of_node(pgdat->node_id);
L
Linus Torvalds 已提交
3141

3142 3143
	lockdep_set_current_reclaim_state(GFP_KERNEL);

R
Rusty Russell 已提交
3144
	if (!cpumask_empty(cpumask))
3145
		set_cpus_allowed_ptr(tsk, cpumask);
L
Linus Torvalds 已提交
3146 3147 3148 3149 3150 3151 3152 3153 3154 3155 3156 3157 3158 3159
	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).
	 */
3160
	tsk->flags |= PF_MEMALLOC | PF_SWAPWRITE | PF_KSWAPD;
3161
	set_freezable();
L
Linus Torvalds 已提交
3162

3163
	order = new_order = 0;
3164
	balanced_order = 0;
3165
	classzone_idx = new_classzone_idx = pgdat->nr_zones - 1;
3166
	balanced_classzone_idx = classzone_idx;
L
Linus Torvalds 已提交
3167
	for ( ; ; ) {
3168
		bool ret;
3169

3170 3171 3172 3173 3174
		/*
		 * 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
		 */
3175 3176
		if (balanced_classzone_idx >= new_classzone_idx &&
					balanced_order == new_order) {
3177 3178 3179 3180 3181 3182
			new_order = pgdat->kswapd_max_order;
			new_classzone_idx = pgdat->classzone_idx;
			pgdat->kswapd_max_order =  0;
			pgdat->classzone_idx = pgdat->nr_zones - 1;
		}

3183
		if (order < new_order || classzone_idx > new_classzone_idx) {
L
Linus Torvalds 已提交
3184 3185
			/*
			 * Don't sleep if someone wants a larger 'order'
3186
			 * allocation or has tigher zone constraints
L
Linus Torvalds 已提交
3187 3188
			 */
			order = new_order;
3189
			classzone_idx = new_classzone_idx;
L
Linus Torvalds 已提交
3190
		} else {
3191 3192
			kswapd_try_to_sleep(pgdat, balanced_order,
						balanced_classzone_idx);
L
Linus Torvalds 已提交
3193
			order = pgdat->kswapd_max_order;
3194
			classzone_idx = pgdat->classzone_idx;
3195 3196
			new_order = order;
			new_classzone_idx = classzone_idx;
3197
			pgdat->kswapd_max_order = 0;
3198
			pgdat->classzone_idx = pgdat->nr_zones - 1;
L
Linus Torvalds 已提交
3199 3200
		}

3201 3202 3203 3204 3205 3206 3207 3208
		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
		 */
3209 3210
		if (!ret) {
			trace_mm_vmscan_kswapd_wake(pgdat->node_id, order);
3211 3212 3213
			balanced_classzone_idx = classzone_idx;
			balanced_order = balance_pgdat(pgdat, order,
						&balanced_classzone_idx);
3214
		}
L
Linus Torvalds 已提交
3215
	}
3216 3217

	current->reclaim_state = NULL;
L
Linus Torvalds 已提交
3218 3219 3220 3221 3222 3223
	return 0;
}

/*
 * A zone is low on free memory, so wake its kswapd task to service it.
 */
3224
void wakeup_kswapd(struct zone *zone, int order, enum zone_type classzone_idx)
L
Linus Torvalds 已提交
3225 3226 3227
{
	pg_data_t *pgdat;

3228
	if (!populated_zone(zone))
L
Linus Torvalds 已提交
3229 3230
		return;

3231
	if (!cpuset_zone_allowed_hardwall(zone, GFP_KERNEL))
L
Linus Torvalds 已提交
3232
		return;
3233
	pgdat = zone->zone_pgdat;
3234
	if (pgdat->kswapd_max_order < order) {
L
Linus Torvalds 已提交
3235
		pgdat->kswapd_max_order = order;
3236 3237
		pgdat->classzone_idx = min(pgdat->classzone_idx, classzone_idx);
	}
3238
	if (!waitqueue_active(&pgdat->kswapd_wait))
L
Linus Torvalds 已提交
3239
		return;
3240 3241 3242 3243
	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);
3244
	wake_up_interruptible(&pgdat->kswapd_wait);
L
Linus Torvalds 已提交
3245 3246
}

3247 3248 3249 3250 3251 3252 3253 3254
/*
 * 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)
3255
{
3256 3257 3258 3259 3260
	int nr;

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

3261
	if (get_nr_swap_pages() > 0)
3262 3263 3264 3265 3266 3267 3268 3269 3270 3271 3272 3273 3274
		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);

3275
	if (get_nr_swap_pages() > 0)
3276 3277 3278 3279
		nr += zone_page_state(zone, NR_ACTIVE_ANON) +
		      zone_page_state(zone, NR_INACTIVE_ANON);

	return nr;
3280 3281
}

3282
#ifdef CONFIG_HIBERNATION
L
Linus Torvalds 已提交
3283
/*
3284
 * Try to free `nr_to_reclaim' of memory, system-wide, and return the number of
3285 3286 3287 3288 3289
 * 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
3290
 */
3291
unsigned long shrink_all_memory(unsigned long nr_to_reclaim)
L
Linus Torvalds 已提交
3292
{
3293 3294
	struct reclaim_state reclaim_state;
	struct scan_control sc = {
3295 3296 3297
		.gfp_mask = GFP_HIGHUSER_MOVABLE,
		.may_swap = 1,
		.may_unmap = 1,
3298
		.may_writepage = 1,
3299 3300 3301
		.nr_to_reclaim = nr_to_reclaim,
		.hibernation_mode = 1,
		.order = 0,
3302
		.priority = DEF_PRIORITY,
L
Linus Torvalds 已提交
3303
	};
3304 3305 3306 3307
	struct shrink_control shrink = {
		.gfp_mask = sc.gfp_mask,
	};
	struct zonelist *zonelist = node_zonelist(numa_node_id(), sc.gfp_mask);
3308 3309
	struct task_struct *p = current;
	unsigned long nr_reclaimed;
L
Linus Torvalds 已提交
3310

3311 3312 3313 3314
	p->flags |= PF_MEMALLOC;
	lockdep_set_current_reclaim_state(sc.gfp_mask);
	reclaim_state.reclaimed_slab = 0;
	p->reclaim_state = &reclaim_state;
3315

3316
	nr_reclaimed = do_try_to_free_pages(zonelist, &sc, &shrink);
3317

3318 3319 3320
	p->reclaim_state = NULL;
	lockdep_clear_current_reclaim_state();
	p->flags &= ~PF_MEMALLOC;
3321

3322
	return nr_reclaimed;
L
Linus Torvalds 已提交
3323
}
3324
#endif /* CONFIG_HIBERNATION */
L
Linus Torvalds 已提交
3325 3326 3327 3328 3329

/* 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. */
3330 3331
static int cpu_callback(struct notifier_block *nfb, unsigned long action,
			void *hcpu)
L
Linus Torvalds 已提交
3332
{
3333
	int nid;
L
Linus Torvalds 已提交
3334

3335
	if (action == CPU_ONLINE || action == CPU_ONLINE_FROZEN) {
3336
		for_each_node_state(nid, N_MEMORY) {
3337
			pg_data_t *pgdat = NODE_DATA(nid);
3338 3339 3340
			const struct cpumask *mask;

			mask = cpumask_of_node(pgdat->node_id);
3341

3342
			if (cpumask_any_and(cpu_online_mask, mask) < nr_cpu_ids)
L
Linus Torvalds 已提交
3343
				/* One of our CPUs online: restore mask */
3344
				set_cpus_allowed_ptr(pgdat->kswapd, mask);
L
Linus Torvalds 已提交
3345 3346 3347 3348 3349
		}
	}
	return NOTIFY_OK;
}

3350 3351 3352 3353 3354 3355 3356 3357 3358 3359 3360 3361 3362 3363 3364 3365
/*
 * 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);
3366 3367
		pr_err("Failed to start kswapd on node %d\n", nid);
		ret = PTR_ERR(pgdat->kswapd);
3368
		pgdat->kswapd = NULL;
3369 3370 3371 3372
	}
	return ret;
}

3373
/*
3374 3375
 * Called by memory hotplug when all memory in a node is offlined.  Caller must
 * hold lock_memory_hotplug().
3376 3377 3378 3379 3380
 */
void kswapd_stop(int nid)
{
	struct task_struct *kswapd = NODE_DATA(nid)->kswapd;

3381
	if (kswapd) {
3382
		kthread_stop(kswapd);
3383 3384
		NODE_DATA(nid)->kswapd = NULL;
	}
3385 3386
}

L
Linus Torvalds 已提交
3387 3388
static int __init kswapd_init(void)
{
3389
	int nid;
3390

L
Linus Torvalds 已提交
3391
	swap_setup();
3392
	for_each_node_state(nid, N_MEMORY)
3393
 		kswapd_run(nid);
L
Linus Torvalds 已提交
3394 3395 3396 3397 3398
	hotcpu_notifier(cpu_callback, 0);
	return 0;
}

module_init(kswapd_init)
3399 3400 3401 3402 3403 3404 3405 3406 3407 3408

#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;

3409
#define RECLAIM_OFF 0
3410
#define RECLAIM_ZONE (1<<0)	/* Run shrink_inactive_list on the zone */
3411 3412 3413
#define RECLAIM_WRITE (1<<1)	/* Writeout pages during reclaim */
#define RECLAIM_SWAP (1<<2)	/* Swap pages out during reclaim */

3414 3415 3416 3417 3418 3419 3420
/*
 * 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

3421 3422 3423 3424 3425 3426
/*
 * Percentage of pages in a zone that must be unmapped for zone_reclaim to
 * occur.
 */
int sysctl_min_unmapped_ratio = 1;

3427 3428 3429 3430 3431 3432
/*
 * 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;

3433 3434 3435 3436 3437 3438 3439 3440 3441 3442 3443 3444 3445 3446 3447 3448 3449 3450 3451 3452 3453 3454 3455 3456 3457 3458 3459 3460 3461 3462 3463 3464 3465 3466 3467 3468 3469 3470 3471 3472 3473 3474
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;
}

3475 3476 3477
/*
 * Try to free up some pages from this zone through reclaim.
 */
3478
static int __zone_reclaim(struct zone *zone, gfp_t gfp_mask, unsigned int order)
3479
{
3480
	/* Minimum pages needed in order to stay on node */
3481
	const unsigned long nr_pages = 1 << order;
3482 3483
	struct task_struct *p = current;
	struct reclaim_state reclaim_state;
3484 3485
	struct scan_control sc = {
		.may_writepage = !!(zone_reclaim_mode & RECLAIM_WRITE),
3486
		.may_unmap = !!(zone_reclaim_mode & RECLAIM_SWAP),
3487
		.may_swap = 1,
3488
		.nr_to_reclaim = max(nr_pages, SWAP_CLUSTER_MAX),
3489
		.gfp_mask = (gfp_mask = memalloc_noio_flags(gfp_mask)),
3490
		.order = order,
3491
		.priority = ZONE_RECLAIM_PRIORITY,
3492
	};
3493 3494 3495
	struct shrink_control shrink = {
		.gfp_mask = sc.gfp_mask,
	};
3496
	unsigned long nr_slab_pages0, nr_slab_pages1;
3497 3498

	cond_resched();
3499 3500 3501 3502 3503 3504
	/*
	 * 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;
3505
	lockdep_set_current_reclaim_state(gfp_mask);
3506 3507
	reclaim_state.reclaimed_slab = 0;
	p->reclaim_state = &reclaim_state;
3508

3509
	if (zone_pagecache_reclaimable(zone) > zone->min_unmapped_pages) {
3510 3511 3512 3513 3514
		/*
		 * Free memory by calling shrink zone with increasing
		 * priorities until we have enough memory freed.
		 */
		do {
3515 3516
			shrink_zone(zone, &sc);
		} while (sc.nr_reclaimed < nr_pages && --sc.priority >= 0);
3517
	}
3518

3519 3520
	nr_slab_pages0 = zone_page_state(zone, NR_SLAB_RECLAIMABLE);
	if (nr_slab_pages0 > zone->min_slab_pages) {
3521
		/*
3522
		 * shrink_slab() does not currently allow us to determine how
3523 3524 3525 3526
		 * 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.
3527
		 *
3528 3529
		 * Note that shrink_slab will free memory on all zones and may
		 * take a long time.
3530
		 */
3531 3532 3533 3534
		for (;;) {
			unsigned long lru_pages = zone_reclaimable_pages(zone);

			/* No reclaimable slab or very low memory pressure */
3535
			if (!shrink_slab(&shrink, sc.nr_scanned, lru_pages))
3536 3537 3538 3539 3540 3541 3542 3543
				break;

			/* Freed enough memory */
			nr_slab_pages1 = zone_page_state(zone,
							NR_SLAB_RECLAIMABLE);
			if (nr_slab_pages1 + nr_pages <= nr_slab_pages0)
				break;
		}
3544 3545 3546 3547 3548

		/*
		 * Update nr_reclaimed by the number of slab pages we
		 * reclaimed from this zone.
		 */
3549 3550 3551
		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;
3552 3553
	}

3554
	p->reclaim_state = NULL;
3555
	current->flags &= ~(PF_MEMALLOC | PF_SWAPWRITE);
3556
	lockdep_clear_current_reclaim_state();
3557
	return sc.nr_reclaimed >= nr_pages;
3558
}
3559 3560 3561 3562

int zone_reclaim(struct zone *zone, gfp_t gfp_mask, unsigned int order)
{
	int node_id;
3563
	int ret;
3564 3565

	/*
3566 3567
	 * Zone reclaim reclaims unmapped file backed pages and
	 * slab pages if we are over the defined limits.
3568
	 *
3569 3570 3571 3572 3573
	 * 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.
3574
	 */
3575 3576
	if (zone_pagecache_reclaimable(zone) <= zone->min_unmapped_pages &&
	    zone_page_state(zone, NR_SLAB_RECLAIMABLE) <= zone->min_slab_pages)
3577
		return ZONE_RECLAIM_FULL;
3578

3579
	if (zone->all_unreclaimable)
3580
		return ZONE_RECLAIM_FULL;
3581

3582
	/*
3583
	 * Do not scan if the allocation should not be delayed.
3584
	 */
3585
	if (!(gfp_mask & __GFP_WAIT) || (current->flags & PF_MEMALLOC))
3586
		return ZONE_RECLAIM_NOSCAN;
3587 3588 3589 3590 3591 3592 3593

	/*
	 * 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.
	 */
3594
	node_id = zone_to_nid(zone);
3595
	if (node_state(node_id, N_CPU) && node_id != numa_node_id())
3596
		return ZONE_RECLAIM_NOSCAN;
3597 3598

	if (zone_test_and_set_flag(zone, ZONE_RECLAIM_LOCKED))
3599 3600
		return ZONE_RECLAIM_NOSCAN;

3601 3602 3603
	ret = __zone_reclaim(zone, gfp_mask, order);
	zone_clear_flag(zone, ZONE_RECLAIM_LOCKED);

3604 3605 3606
	if (!ret)
		count_vm_event(PGSCAN_ZONE_RECLAIM_FAILED);

3607
	return ret;
3608
}
3609
#endif
L
Lee Schermerhorn 已提交
3610 3611 3612 3613 3614 3615

/*
 * 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
3616
 * lists vs unevictable list.
L
Lee Schermerhorn 已提交
3617 3618
 *
 * Reasons page might not be evictable:
3619
 * (1) page's mapping marked unevictable
N
Nick Piggin 已提交
3620
 * (2) page is part of an mlocked VMA
3621
 *
L
Lee Schermerhorn 已提交
3622
 */
3623
int page_evictable(struct page *page)
L
Lee Schermerhorn 已提交
3624
{
3625
	return !mapping_unevictable(page_mapping(page)) && !PageMlocked(page);
L
Lee Schermerhorn 已提交
3626
}
3627

3628
#ifdef CONFIG_SHMEM
3629
/**
3630 3631 3632
 * 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
3633
 *
3634
 * Checks pages for evictability and moves them to the appropriate lru list.
3635 3636
 *
 * This function is only used for SysV IPC SHM_UNLOCK.
3637
 */
3638
void check_move_unevictable_pages(struct page **pages, int nr_pages)
3639
{
3640
	struct lruvec *lruvec;
3641 3642 3643 3644
	struct zone *zone = NULL;
	int pgscanned = 0;
	int pgrescued = 0;
	int i;
3645

3646 3647 3648
	for (i = 0; i < nr_pages; i++) {
		struct page *page = pages[i];
		struct zone *pagezone;
3649

3650 3651 3652 3653 3654 3655 3656 3657
		pgscanned++;
		pagezone = page_zone(page);
		if (pagezone != zone) {
			if (zone)
				spin_unlock_irq(&zone->lru_lock);
			zone = pagezone;
			spin_lock_irq(&zone->lru_lock);
		}
3658
		lruvec = mem_cgroup_page_lruvec(page, zone);
3659

3660 3661
		if (!PageLRU(page) || !PageUnevictable(page))
			continue;
3662

3663
		if (page_evictable(page)) {
3664 3665 3666 3667
			enum lru_list lru = page_lru_base_type(page);

			VM_BUG_ON(PageActive(page));
			ClearPageUnevictable(page);
3668 3669
			del_page_from_lru_list(page, lruvec, LRU_UNEVICTABLE);
			add_page_to_lru_list(page, lruvec, lru);
3670
			pgrescued++;
3671
		}
3672
	}
3673

3674 3675 3676 3677
	if (zone) {
		__count_vm_events(UNEVICTABLE_PGRESCUED, pgrescued);
		__count_vm_events(UNEVICTABLE_PGSCANNED, pgscanned);
		spin_unlock_irq(&zone->lru_lock);
3678 3679
	}
}
3680
#endif /* CONFIG_SHMEM */
3681

3682
static void warn_scan_unevictable_pages(void)
3683
{
3684
	printk_once(KERN_WARNING
3685
		    "%s: The scan_unevictable_pages sysctl/node-interface has been "
3686
		    "disabled for lack of a legitimate use case.  If you have "
3687 3688
		    "one, please send an email to linux-mm@kvack.org.\n",
		    current->comm);
3689 3690 3691 3692 3693 3694 3695 3696 3697
}

/*
 * 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,
3698
			   void __user *buffer,
3699 3700
			   size_t *length, loff_t *ppos)
{
3701
	warn_scan_unevictable_pages();
3702
	proc_doulongvec_minmax(table, write, buffer, length, ppos);
3703 3704 3705 3706
	scan_unevictable_pages = 0;
	return 0;
}

3707
#ifdef CONFIG_NUMA
3708 3709 3710 3711 3712
/*
 * per node 'scan_unevictable_pages' attribute.  On demand re-scan of
 * a specified node's per zone unevictable lists for evictable pages.
 */

3713 3714
static ssize_t read_scan_unevictable_node(struct device *dev,
					  struct device_attribute *attr,
3715 3716
					  char *buf)
{
3717
	warn_scan_unevictable_pages();
3718 3719 3720
	return sprintf(buf, "0\n");	/* always zero; should fit... */
}

3721 3722
static ssize_t write_scan_unevictable_node(struct device *dev,
					   struct device_attribute *attr,
3723 3724
					const char *buf, size_t count)
{
3725
	warn_scan_unevictable_pages();
3726 3727 3728 3729
	return 1;
}


3730
static DEVICE_ATTR(scan_unevictable_pages, S_IRUGO | S_IWUSR,
3731 3732 3733 3734 3735
			read_scan_unevictable_node,
			write_scan_unevictable_node);

int scan_unevictable_register_node(struct node *node)
{
3736
	return device_create_file(&node->dev, &dev_attr_scan_unevictable_pages);
3737 3738 3739 3740
}

void scan_unevictable_unregister_node(struct node *node)
{
3741
	device_remove_file(&node->dev, &dev_attr_scan_unevictable_pages);
3742
}
3743
#endif