vmscan.c 101 KB
Newer Older
Linus Torvalds's avatar
Linus Torvalds committed
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>
Linus Torvalds's avatar
Linus Torvalds committed
17 18 19 20 21
#include <linux/kernel_stat.h>
#include <linux/swap.h>
#include <linux/pagemap.h>
#include <linux/init.h>
#include <linux/highmem.h>
22
#include <linux/vmstat.h>
Linus Torvalds's avatar
Linus Torvalds committed
23 24 25 26 27 28 29 30 31 32 33
#include <linux/file.h>
#include <linux/writeback.h>
#include <linux/blkdev.h>
#include <linux/buffer_head.h>	/* for try_to_release_page(),
					buffer_heads_over_limit */
#include <linux/mm_inline.h>
#include <linux/backing-dev.h>
#include <linux/rmap.h>
#include <linux/topology.h>
#include <linux/cpu.h>
#include <linux/cpuset.h>
34
#include <linux/compaction.h>
Linus Torvalds's avatar
Linus Torvalds committed
35 36
#include <linux/notifier.h>
#include <linux/rwsem.h>
37
#include <linux/delay.h>
38
#include <linux/kthread.h>
39
#include <linux/freezer.h>
40
#include <linux/memcontrol.h>
41
#include <linux/delayacct.h>
42
#include <linux/sysctl.h>
43
#include <linux/oom.h>
44
#include <linux/prefetch.h>
Linus Torvalds's avatar
Linus Torvalds committed
45 46 47 48 49 50

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

#include <linux/swapops.h>

51 52
#include "internal.h"

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

Linus Torvalds's avatar
Linus Torvalds committed
56 57 58 59
struct scan_control {
	/* Incremented by the number of inactive pages that were scanned */
	unsigned long nr_scanned;

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

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

66 67
	unsigned long hibernation_mode;

Linus Torvalds's avatar
Linus Torvalds committed
68
	/* This context's GFP mask */
Al Viro's avatar
Al Viro committed
69
	gfp_t gfp_mask;
Linus Torvalds's avatar
Linus Torvalds committed
70 71 72

	int may_writepage;

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

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

Andy Whitcroft's avatar
Andy Whitcroft committed
79
	int order;
80

81 82 83
	/* Scan (total_size >> priority) pages at once */
	int priority;

84 85 86 87 88
	/*
	 * The memory cgroup that hit its limit and as a result is the
	 * primary target of this reclaim invocation.
	 */
	struct mem_cgroup *target_mem_cgroup;
89

90 91 92 93 94
	/*
	 * Nodemask of nodes allowed by the caller. If NULL, all nodes
	 * are scanned.
	 */
	nodemask_t	*nodemask;
Linus Torvalds's avatar
Linus Torvalds committed
95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130
};

#define lru_to_page(_head) (list_entry((_head)->prev, struct page, lru))

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

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

/*
 * From 0 .. 100.  Higher means more swappy.
 */
int vm_swappiness = 60;
131
long vm_total_pages;	/* The total number of pages which the VM controls */
Linus Torvalds's avatar
Linus Torvalds committed
132 133 134 135

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

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

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

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

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

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

179 180 181 182 183 184 185 186
static inline int do_shrinker_shrink(struct shrinker *shrinker,
				     struct shrink_control *sc,
				     unsigned long nr_to_scan)
{
	sc->nr_to_scan = nr_to_scan;
	return (*shrinker->shrink)(shrinker, sc);
}

Linus Torvalds's avatar
Linus Torvalds committed
187 188 189 190 191 192 193 194 195
#define SHRINK_BATCH 128
/*
 * Call the shrink functions to age shrinkable caches
 *
 * Here we assume it costs one seek to replace a lru page and that it also
 * takes a seek to recreate a cache object.  With this in mind we age equal
 * percentages of the lru and ageable caches.  This should balance the seeks
 * generated by these structures.
 *
Simon Arlott's avatar
Simon Arlott committed
196
 * If the vm encountered mapped pages on the LRU it increase the pressure on
Linus Torvalds's avatar
Linus Torvalds committed
197 198 199 200 201 202 203
 * slab to avoid swapping.
 *
 * We do weird things to avoid (scanned*seeks*entries) overflowing 32 bits.
 *
 * `lru_pages' represents the number of on-LRU pages in all the zones which
 * are eligible for the caller's allocation attempt.  It is used for balancing
 * slab reclaim versus page reclaim.
204 205
 *
 * Returns the number of slab objects which we shrunk.
Linus Torvalds's avatar
Linus Torvalds committed
206
 */
207
unsigned long shrink_slab(struct shrink_control *shrink,
208
			  unsigned long nr_pages_scanned,
209
			  unsigned long lru_pages)
Linus Torvalds's avatar
Linus Torvalds committed
210 211
{
	struct shrinker *shrinker;
212
	unsigned long ret = 0;
Linus Torvalds's avatar
Linus Torvalds committed
213

214 215
	if (nr_pages_scanned == 0)
		nr_pages_scanned = SWAP_CLUSTER_MAX;
Linus Torvalds's avatar
Linus Torvalds committed
216

217 218 219 220 221
	if (!down_read_trylock(&shrinker_rwsem)) {
		/* Assume we'll be able to shrink next time */
		ret = 1;
		goto out;
	}
Linus Torvalds's avatar
Linus Torvalds committed
222 223 224

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

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

237 238 239 240 241
		/*
		 * copy the current shrinker scan count into a local variable
		 * and zero it so that other concurrent shrinker invocations
		 * don't also do this scanning work.
		 */
242
		nr = atomic_long_xchg(&shrinker->nr_in_batch, 0);
243 244

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

256 257 258 259 260 261 262 263 264 265 266 267 268 269 270
		/*
		 * We need to avoid excessive windup on filesystem shrinkers
		 * due to large numbers of GFP_NOFS allocations causing the
		 * shrinkers to return -1 all the time. This results in a large
		 * nr being built up so when a shrink that can do some work
		 * comes along it empties the entire cache due to nr >>>
		 * max_pass.  This is bad for sustaining a working set in
		 * memory.
		 *
		 * Hence only allow the shrinker to scan the entire cache when
		 * a large delta change is calculated directly.
		 */
		if (delta < max_pass / 4)
			total_scan = min(total_scan, max_pass / 2);

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

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

283
		while (total_scan >= batch_size) {
284
			int nr_before;
Linus Torvalds's avatar
Linus Torvalds committed
285

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

			cond_resched();
		}

299 300 301 302 303
		/*
		 * move the unused scan count back into the shrinker in a
		 * manner that handles concurrent updates. If we exhausted the
		 * scan, there is no need to do an update.
		 */
304 305 306 307 308
		if (total_scan > 0)
			new_nr = atomic_long_add_return(total_scan,
					&shrinker->nr_in_batch);
		else
			new_nr = atomic_long_read(&shrinker->nr_in_batch);
309 310

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

static inline int is_page_cache_freeable(struct page *page)
{
320 321 322 323 324
	/*
	 * A freeable page cache page is referenced only by the caller
	 * that isolated the page, the page cache radix tree and
	 * optional buffer heads at page->private.
	 */
325
	return page_count(page) - page_has_private(page) == 2;
Linus Torvalds's avatar
Linus Torvalds committed
326 327
}

328 329
static int may_write_to_queue(struct backing_dev_info *bdi,
			      struct scan_control *sc)
Linus Torvalds's avatar
Linus Torvalds committed
330
{
331
	if (current->flags & PF_SWAPWRITE)
Linus Torvalds's avatar
Linus Torvalds committed
332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354
		return 1;
	if (!bdi_write_congested(bdi))
		return 1;
	if (bdi == current->backing_dev_info)
		return 1;
	return 0;
}

/*
 * We detected a synchronous write error writing a page out.  Probably
 * -ENOSPC.  We need to propagate that into the address_space for a subsequent
 * fsync(), msync() or close().
 *
 * The tricky part is that after writepage we cannot touch the mapping: nothing
 * prevents it from being freed up.  But we have a ref on the page and once
 * that page is locked, the mapping is pinned.
 *
 * We're allowed to run sleeping lock_page() here because we know the caller has
 * __GFP_FS.
 */
static void handle_write_error(struct address_space *mapping,
				struct page *page, int error)
{
355
	lock_page(page);
356 357
	if (page_mapping(page) == mapping)
		mapping_set_error(mapping, error);
Linus Torvalds's avatar
Linus Torvalds committed
358 359 360
	unlock_page(page);
}

361 362 363 364 365 366 367 368 369 370 371 372
/* possible outcome of pageout() */
typedef enum {
	/* failed to write page out, page is locked */
	PAGE_KEEP,
	/* move page to the active list, page is locked */
	PAGE_ACTIVATE,
	/* page has been sent to the disk successfully, page is unlocked */
	PAGE_SUCCESS,
	/* page is clean and locked */
	PAGE_CLEAN,
} pageout_t;

Linus Torvalds's avatar
Linus Torvalds committed
373
/*
374 375
 * pageout is called by shrink_page_list() for each dirty page.
 * Calls ->writepage().
Linus Torvalds's avatar
Linus Torvalds committed
376
 */
377
static pageout_t pageout(struct page *page, struct address_space *mapping,
378
			 struct scan_control *sc)
Linus Torvalds's avatar
Linus Torvalds committed
379 380 381 382 383 384 385 386
{
	/*
	 * If the page is dirty, only perform writeback if that write
	 * will be non-blocking.  To prevent this allocation from being
	 * stalled by pagecache activity.  But note that there may be
	 * stalls if we need to run get_block().  We could test
	 * PagePrivate for that.
	 *
387
	 * If this process is currently in __generic_file_aio_write() against
Linus Torvalds's avatar
Linus Torvalds committed
388 389 390 391 392 393 394 395 396 397 398 399 400 401 402
	 * this page's queue, we can perform writeback even if that
	 * will block.
	 *
	 * If the page is swapcache, write it back even if that would
	 * block, for some throttling. This happens by accident, because
	 * swap_backing_dev_info is bust: it doesn't reflect the
	 * congestion state of the swapdevs.  Easy to fix, if needed.
	 */
	if (!is_page_cache_freeable(page))
		return PAGE_KEEP;
	if (!mapping) {
		/*
		 * Some data journaling orphaned pages can have
		 * page->mapping == NULL while being dirty with clean buffers.
		 */
403
		if (page_has_private(page)) {
Linus Torvalds's avatar
Linus Torvalds committed
404 405
			if (try_to_free_buffers(page)) {
				ClearPageDirty(page);
406
				printk("%s: orphaned page\n", __func__);
Linus Torvalds's avatar
Linus Torvalds committed
407 408 409 410 411 412 413
				return PAGE_CLEAN;
			}
		}
		return PAGE_KEEP;
	}
	if (mapping->a_ops->writepage == NULL)
		return PAGE_ACTIVATE;
414
	if (!may_write_to_queue(mapping->backing_dev_info, sc))
Linus Torvalds's avatar
Linus Torvalds committed
415 416 417 418 419 420 421
		return PAGE_KEEP;

	if (clear_page_dirty_for_io(page)) {
		int res;
		struct writeback_control wbc = {
			.sync_mode = WB_SYNC_NONE,
			.nr_to_write = SWAP_CLUSTER_MAX,
422 423
			.range_start = 0,
			.range_end = LLONG_MAX,
Linus Torvalds's avatar
Linus Torvalds committed
424 425 426 427 428 429 430
			.for_reclaim = 1,
		};

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

Linus Torvalds's avatar
Linus Torvalds committed
436 437 438 439
		if (!PageWriteback(page)) {
			/* synchronous write or broken a_ops? */
			ClearPageReclaim(page);
		}
440
		trace_mm_vmscan_writepage(page, trace_reclaim_flags(page));
441
		inc_zone_page_state(page, NR_VMSCAN_WRITE);
Linus Torvalds's avatar
Linus Torvalds committed
442 443 444 445 446 447
		return PAGE_SUCCESS;
	}

	return PAGE_CLEAN;
}

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

Nick Piggin's avatar
Nick Piggin committed
457
	spin_lock_irq(&mapping->tree_lock);
458
	/*
459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481
	 * The non racy check for a busy page.
	 *
	 * Must be careful with the order of the tests. When someone has
	 * a ref to the page, it may be possible that they dirty it then
	 * drop the reference. So if PageDirty is tested before page_count
	 * here, then the following race may occur:
	 *
	 * get_user_pages(&page);
	 * [user mapping goes away]
	 * write_to(page);
	 *				!PageDirty(page)    [good]
	 * SetPageDirty(page);
	 * put_page(page);
	 *				!page_count(page)   [good, discard it]
	 *
	 * [oops, our write_to data is lost]
	 *
	 * Reversing the order of the tests ensures such a situation cannot
	 * escape unnoticed. The smp_rmb is needed to ensure the page->flags
	 * load is not satisfied before that of page->_count.
	 *
	 * Note that if SetPageDirty is always performed via set_page_dirty,
	 * and thus under tree_lock, then this ordering is not required.
482
	 */
483
	if (!page_freeze_refs(page, 2))
484
		goto cannot_free;
485 486 487
	/* note: atomic_cmpxchg in page_freeze_refs provides the smp_rmb */
	if (unlikely(PageDirty(page))) {
		page_unfreeze_refs(page, 2);
488
		goto cannot_free;
489
	}
490 491 492 493

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

		freepage = mapping->a_ops->freepage;

501
		__delete_from_page_cache(page);
Nick Piggin's avatar
Nick Piggin committed
502
		spin_unlock_irq(&mapping->tree_lock);
503
		mem_cgroup_uncharge_cache_page(page);
504 505 506

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

	return 1;

cannot_free:
Nick Piggin's avatar
Nick Piggin committed
512
	spin_unlock_irq(&mapping->tree_lock);
513 514 515
	return 0;
}

516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535
/*
 * Attempt to detach a locked page from its ->mapping.  If it is dirty or if
 * someone else has a ref on the page, abort and return 0.  If it was
 * successfully detached, return 1.  Assumes the caller has a single ref on
 * this page.
 */
int remove_mapping(struct address_space *mapping, struct page *page)
{
	if (__remove_mapping(mapping, page)) {
		/*
		 * Unfreezing the refcount with 1 rather than 2 effectively
		 * drops the pagecache ref for us without requiring another
		 * atomic operation.
		 */
		page_unfreeze_refs(page, 1);
		return 1;
	}
	return 0;
}

536 537 538 539 540 541 542 543 544 545 546 547 548
/**
 * putback_lru_page - put previously isolated page onto appropriate LRU list
 * @page: page to be put back to appropriate lru list
 *
 * Add previously isolated @page to appropriate LRU list.
 * Page may still be unevictable for other reasons.
 *
 * lru_lock must not be held, interrupts must be enabled.
 */
void putback_lru_page(struct page *page)
{
	int lru;
	int active = !!TestClearPageActive(page);
549
	int was_unevictable = PageUnevictable(page);
550 551 552 553 554 555

	VM_BUG_ON(PageLRU(page));

redo:
	ClearPageUnevictable(page);

556
	if (page_evictable(page)) {
557 558 559 560 561 562
		/*
		 * For evictable pages, we can use the cache.
		 * In event of a race, worst case is we end up with an
		 * unevictable page on [in]active list.
		 * We know how to handle that.
		 */
563
		lru = active + page_lru_base_type(page);
564 565 566 567 568 569 570 571
		lru_cache_add_lru(page, lru);
	} else {
		/*
		 * Put unevictable pages directly on zone's unevictable
		 * list.
		 */
		lru = LRU_UNEVICTABLE;
		add_page_to_unevictable_list(page);
572
		/*
573 574 575
		 * When racing with an mlock or AS_UNEVICTABLE clearing
		 * (page is unlocked) make sure that if the other thread
		 * does not observe our setting of PG_lru and fails
576
		 * isolation/check_move_unevictable_pages,
577
		 * we see PG_mlocked/AS_UNEVICTABLE cleared below and move
578 579
		 * the page back to the evictable list.
		 *
580
		 * The other side is TestClearPageMlocked() or shmem_lock().
581 582
		 */
		smp_mb();
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)) {
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);

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
}

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

	cond_resched();

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

		cond_resched();

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

Nick Piggin's avatar
Nick Piggin committed
704
		if (!trylock_page(page))
Linus Torvalds's avatar
Linus Torvalds committed
705 706
			goto keep;

Nick Piggin's avatar
Nick Piggin committed
707
		VM_BUG_ON(PageActive(page));
708
		VM_BUG_ON(page_zone(page) != zone);
Linus Torvalds's avatar
Linus Torvalds committed
709 710

		sc->nr_scanned++;
711

712
		if (unlikely(!page_evictable(page)))
713
			goto cull_mlocked;
714

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

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

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

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

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

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

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

		mapping = page_mapping(page);

		/*
		 * The page is mapped into the page tables of one or more
		 * processes. Try to unmap it here.
		 */
		if (page_mapped(page) && mapping) {
795
			switch (try_to_unmap(page, ttu_flags)) {
Linus Torvalds's avatar
Linus Torvalds committed
796 797 798 799
			case SWAP_FAIL:
				goto activate_locked;
			case SWAP_AGAIN:
				goto keep_locked;
800 801
			case SWAP_MLOCK:
				goto cull_mlocked;
Linus Torvalds's avatar
Linus Torvalds committed
802 803 804 805 806 807
			case SWAP_SUCCESS:
				; /* try to free the page below */
			}
		}

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

810 811
			/*
			 * Only kswapd can writeback filesystem pages to
812 813
			 * avoid risk of stack overflow but do not writeback
			 * unless under significant pressure.
814
			 */
815
			if (page_is_file_cache(page) &&
816 817
					(!current_is_kswapd() ||
					 sc->priority >= DEF_PRIORITY - 2)) {
818 819 820 821 822 823 824 825 826
				/*
				 * Immediately reclaim when written back.
				 * Similar in principal to deactivate_page()
				 * except we already have the page isolated
				 * and know it's dirty
				 */
				inc_zone_page_state(page, NR_VMSCAN_IMMEDIATE);
				SetPageReclaim(page);

827 828 829
				goto keep_locked;
			}

830
			if (references == PAGEREF_RECLAIM_CLEAN)
Linus Torvalds's avatar
Linus Torvalds committed
831
				goto keep_locked;
832
			if (!may_enter_fs)
Linus Torvalds's avatar
Linus Torvalds committed
833
				goto keep_locked;
834
			if (!sc->may_writepage)
Linus Torvalds's avatar
Linus Torvalds committed
835 836 837
				goto keep_locked;

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

Linus Torvalds's avatar
Linus Torvalds committed
850 851 852 853
				/*
				 * A synchronous write - probably a ramdisk.  Go
				 * ahead and try to reclaim the page.
				 */
Nick Piggin's avatar
Nick Piggin committed
854
				if (!trylock_page(page))
Linus Torvalds's avatar
Linus Torvalds committed
855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873
					goto keep;
				if (PageDirty(page) || PageWriteback(page))
					goto keep_locked;
				mapping = page_mapping(page);
			case PAGE_CLEAN:
				; /* try to free the page below */
			}
		}

		/*
		 * If the page has buffers, try to free the buffer mappings
		 * associated with this page. If we succeed we try to free
		 * the page as well.
		 *
		 * We do this even if the page is PageDirty().
		 * try_to_release_page() does not perform I/O, but it is
		 * possible for a page to have PageDirty set, but it is actually
		 * clean (all its buffers are clean).  This happens if the
		 * buffers were written out directly, with submit_bh(). ext3
874
		 * will do this, as well as the blockdev mapping.
Linus Torvalds's avatar
Linus Torvalds committed
875 876 877 878 879 880 881 882 883 884
		 * try_to_release_page() will discover that cleanness and will
		 * drop the buffers and mark the page clean - it can be freed.
		 *
		 * Rarely, pages can have buffers and no ->mapping.  These are
		 * the pages which were not successfully invalidated in
		 * truncate_complete_page().  We try to drop those buffers here
		 * and if that worked, and the page is no longer mapped into
		 * process address space (page_count == 1) it can be freed.
		 * Otherwise, leave the page on the LRU so it is swappable.
		 */
885
		if (page_has_private(page)) {
Linus Torvalds's avatar
Linus Torvalds committed
886 887
			if (!try_to_release_page(page, sc->gfp_mask))
				goto activate_locked;
888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903
			if (!mapping && page_count(page) == 1) {
				unlock_page(page);
				if (put_page_testzero(page))
					goto free_it;
				else {
					/*
					 * rare race with speculative reference.
					 * the speculative reference will free
					 * this page shortly, so we may
					 * increment nr_reclaimed here (and
					 * leave it off the LRU).
					 */
					nr_reclaimed++;
					continue;
				}
			}
Linus Torvalds's avatar
Linus Torvalds committed
904 905
		}

906
		if (!mapping || !__remove_mapping(mapping, page))
907
			goto keep_locked;
Linus Torvalds's avatar
Linus Torvalds committed
908

Nick Piggin's avatar
Nick Piggin committed
909 910 911 912 913 914 915 916
		/*
		 * At this point, we have no other references and there is
		 * no way to pick any more up (removed from LRU, removed
		 * from pagecache). Can use non-atomic bitops now (and
		 * we obviously don't have to worry about waking up a process
		 * waiting on the page lock, because there are no references.
		 */
		__clear_page_locked(page);
917
free_it:
918
		nr_reclaimed++;
919 920 921 922 923 924

		/*
		 * Is there need to periodically free_page_list? It would
		 * appear not as the counts should be low
		 */
		list_add(&page->lru, &free_pages);
Linus Torvalds's avatar
Linus Torvalds committed
925 926
		continue;

927
cull_mlocked:
928 929
		if (PageSwapCache(page))
			try_to_free_swap(page);
930 931 932 933
		unlock_page(page);
		putback_lru_page(page);
		continue;

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

948 949 950 951 952 953
	/*
	 * Tag a zone as congested if all the dirty pages encountered were
	 * backed by a congested BDI. In this case, reclaimers should just
	 * back off and wait for congestion to clear because further reclaim
	 * will encounter the same problem
	 */
954
	if (nr_dirty && nr_dirty == nr_congested && global_reclaim(sc))
955
		zone_set_flag(zone, ZONE_CONGESTED);
956

957
	free_hot_cold_page_list(&free_pages, 1);
958

Linus Torvalds's avatar
Linus Torvalds committed
959
	list_splice(&ret_pages, page_list);
960
	count_vm_events(PGACTIVATE, pgactivate);
961
	mem_cgroup_uncharge_end();
962 963
	*ret_nr_dirty += nr_dirty;
	*ret_nr_writeback += nr_writeback;
964
	return nr_reclaimed;
Linus Torvalds's avatar
Linus Torvalds committed
965 966
}

967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993
unsigned long reclaim_clean_pages_from_list(struct zone *zone,
					    struct list_head *page_list)
{
	struct scan_control sc = {
		.gfp_mask = GFP_KERNEL,
		.priority = DEF_PRIORITY,
		.may_unmap = 1,
	};
	unsigned long ret, dummy1, dummy2;
	struct page *page, *next;
	LIST_HEAD(clean_pages);

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

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

Andy Whitcroft's avatar
Andy Whitcroft committed
994 995 996 997 998 999 1000 1001 1002 1003
/*
 * Attempt to remove the specified page from its LRU.  Only take this page
 * if it is of the appropriate PageActive status.  Pages which are being
 * freed elsewhere are also ignored.
 *
 * page:	page to consider
 * mode:	one of the LRU isolation modes defined above
 *
 * returns 0 on success, -ve errno on failure.
 */
1004
int __isolate_lru_page(struct page *page, isolate_mode_t mode)
Andy Whitcroft's avatar
Andy Whitcroft committed
1005 1006 1007 1008 1009 1010 1011
{
	int ret = -EINVAL;

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

Minchan Kim's avatar
Minchan Kim committed
1012 1013
	/* Compaction should not handle unevictable pages but CMA can do so */
	if (PageUnevictable(page) && !(mode & ISOLATE_UNEVICTABLE))
1014 1015
		return ret;

Andy Whitcroft's avatar
Andy Whitcroft committed
1016
	ret = -EBUSY;
KAMEZAWA Hiroyuki's avatar
KAMEZAWA Hiroyuki committed
1017

1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050
	/*
	 * To minimise LRU disruption, the caller can indicate that it only
	 * wants to isolate pages it will be able to operate on without
	 * blocking - clean pages for the most part.
	 *
	 * ISOLATE_CLEAN means that only clean pages should be isolated. This
	 * is used by reclaim when it is cannot write to backing storage
	 *
	 * ISOLATE_ASYNC_MIGRATE is used to indicate that it only wants to pages
	 * that it is possible to migrate without blocking
	 */
	if (mode & (ISOLATE_CLEAN|ISOLATE_ASYNC_MIGRATE)) {
		/* All the caller can do on PageWriteback is block */
		if (PageWriteback(page))
			return ret;

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

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

			/*
			 * Only pages without mappings or that have a
			 * ->migratepage callback are possible to migrate
			 * without blocking
			 */
			mapping = page_mapping(page);
			if (mapping && !mapping->a_ops->migratepage)
				return ret;
		}
	}
1051

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

Andy Whitcroft's avatar
Andy Whitcroft committed
1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067
	if (likely(get_page_unless_zero(page))) {
		/*
		 * Be careful not to clear PageLRU until after we're
		 * sure the page is not being freed elsewhere -- the
		 * page release code relies on it.
		 */
		ClearPageLRU(page);
		ret = 0;
	}

	return ret;
}

Linus Torvalds's avatar
Linus Torvalds committed
1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078
/*
 * zone->lru_lock is heavily contended.  Some of the functions that
 * shrink the lists perform better by taking out a batch of pages
 * and working on them outside the LRU lock.
 *
 * For pagecache intensive workloads, this function is the hottest
 * spot in the kernel (apart from copy_*_user functions).
 *
 * Appropriate locks must be held before calling this function.
 *
 * @nr_to_scan:	The number of pages to look through on the list.
1079
 * @lruvec:	The LRU vector to pull pages from.
Linus Torvalds's avatar
Linus Torvalds committed
1080
 * @dst:	The temp list to put pages on to.
Hugh Dickins's avatar
Hugh Dickins committed
1081
 * @nr_scanned:	The number of pages that were scanned.
1082
 * @sc:		The scan_control struct for this reclaim session
Andy Whitcroft's avatar
Andy Whitcroft committed
1083
 * @mode:	One of the LRU isolation modes
1084
 * @lru:	LRU list id for isolating
Linus Torvalds's avatar
Linus Torvalds committed
1085 1086 1087
 *
 * returns how many pages were moved onto *@dst.
 */
1088
static unsigned long isolate_lru_pages(unsigned long nr_to_scan,
1089
		struct lruvec *lruvec, struct list_head *dst,
1090
		unsigned long *nr_scanned, struct scan_control *sc,
1091
		isolate_mode_t mode, enum lru_list lru)
Linus Torvalds's avatar
Linus Torvalds committed
1092
{
1093
	struct list_head *src = &lruvec->lists[lru];
1094
	unsigned long nr_taken = 0;
1095
	unsigned long scan;
Linus Torvalds's avatar
Linus Torvalds committed
1096

1097
	for (scan = 0; scan < nr_to_scan && !list_empty(src); scan++) {
Andy Whitcroft's avatar
Andy Whitcroft committed
1098
		struct page *page;
1099
		int nr_pages;
Andy Whitcroft's avatar
Andy Whitcroft committed
1100

Linus Torvalds's avatar
Linus Torvalds committed
1101 1102 1103
		page = lru_to_page(src);
		prefetchw_prev_lru_page(page, src, flags);

Nick Piggin's avatar
Nick Piggin committed
1104
		VM_BUG_ON(!PageLRU(page));
Nick Piggin's avatar
Nick Piggin committed
1105

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

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

Andy Whitcroft's avatar
Andy Whitcroft committed
1119 1120 1121
		default:
			BUG();
		}
Linus Torvalds's avatar
Linus Torvalds committed
1122 1123
	}

Hugh Dickins's avatar
Hugh Dickins committed
1124
	*nr_scanned = scan;
1125 1126
	trace_mm_vmscan_lru_isolate(sc->order, nr_to_scan, scan,
				    nr_taken, mode, is_file_lru(lru));
Linus Torvalds's avatar
Linus Torvalds committed
1127 1128 1129
	return nr_taken;
}

1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140
/**
 * isolate_lru_page - tries to isolate a page from its LRU list
 * @page: page to isolate from its LRU list
 *
 * Isolates a @page from an LRU list, clears PageLRU and adjusts the
 * vmstat statistic corresponding to whatever LRU list the page was on.
 *
 * Returns 0 if the page was removed from an LRU list.
 * Returns -EBUSY if the page was not on an LRU list.
 *
 * The returned page will have PageLRU() cleared.  If it was found on
1141 1142 1143
 * the active list, it will have PageActive set.  If it was found on
 * the unevictable list, it will have the PageUnevictable bit set. That flag
 * may need to be cleared by the caller before letting the page go.
1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158
 *
 * The vmstat statistic corresponding to the list on which the page was
 * found will be decremented.
 *
 * Restrictions:
 * (1) Must be called with an elevated refcount on the page. This is a
 *     fundamentnal difference from isolate_lru_pages (which is called
 *     without a stable reference).
 * (2) the lru_lock must not be held.
 * (3) interrupts must be enabled.
 */
int isolate_lru_page(struct page *page)
{
	int ret = -EBUSY;

1159 1160
	VM_BUG_ON(!page_count(page));

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

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

1179
/*
1180 1181 1182 1183 1184
 * 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.
1185 1186 1187 1188 1189 1190 1191 1192 1193
 */
static int too_many_isolated(struct zone *zone, int file,
		struct scan_control *sc)
{
	unsigned long inactive, isolated;

	if (current_is_kswapd())
		return 0;

1194
	if (!global_reclaim(sc))
1195 1196 1197 1198 1199 1200 1201 1202 1203 1204
		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);
	}

1205 1206 1207 1208 1209 1210 1211 1212
	/*
	 * 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;

1213 1214 1215
	return isolated > inactive;
}

1216
static noinline_for_stack void
1217
putback_inactive_pages(struct lruvec *lruvec, struct list_head *page_list)
1218
{
1219 1220
	struct zone_reclaim_stat *reclaim_stat = &lruvec->reclaim_stat;
	struct zone *zone = lruvec_zone(lruvec);
1221
	LIST_HEAD(pages_to_free);
1222 1223 1224 1225 1226

	/*
	 * Put back any unfreeable pages.
	 */
	while (!list_empty(page_list)) {
1227
		struct page *page = lru_to_page(page_list);
1228
		int lru;
1229

1230 1231
		VM_BUG_ON(PageLRU(page));
		list_del(&page->lru);
1232
		if (unlikely(!page_evictable(page))) {
1233 1234 1235 1236 1237
			spin_unlock_irq(&zone->lru_lock);
			putback_lru_page(page);
			spin_lock_irq(&zone->lru_lock);
			continue;
		}
1238 1239 1240

		lruvec = mem_cgroup_page_lruvec(page, zone);

1241
		SetPageLRU(page);
1242
		lru = page_lru(page);
1243 1244
		add_page_to_lru_list(page, lruvec, lru);

1245 1246
		if (is_active_lru(lru)) {
			int file = is_file_lru(lru);
1247 1248
			int numpages = hpage_nr_pages(page);
			reclaim_stat->recent_rotated[file] += numpages;
1249
		}
1250 1251 1252
		if (put_page_testzero(page)) {
			__ClearPageLRU(page);
			__ClearPageActive(page);
1253
			del_page_from_lru_list(page, lruvec, lru);
1254 1255 1256 1257 1258 1259 1260

			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);
1261 1262 1263
		}
	}

1264 1265 1266 1267
	/*
	 * To save our caller's stack, now use input list for pages to free.
	 */
	list_splice(&pages_to_free, page_list);
1268 1269
}

Linus Torvalds's avatar
Linus Torvalds committed
1270
/*
1271 1272
 * shrink_inactive_list() is a helper for shrink_zone().  It returns the number
 * of reclaimed pages
Linus Torvalds's avatar
Linus Torvalds committed
1273
 */
1274
static noinline_for_stack unsigned long
1275
shrink_inactive_list(unsigned long nr_to_scan, struct lruvec *lruvec,
1276
		     struct scan_control *sc, enum lru_list lru)
Linus Torvalds's avatar
Linus Torvalds committed
1277 1278
{
	LIST_HEAD(page_list);
1279
	unsigned long nr_scanned;
1280
	unsigned long nr_reclaimed = 0;
1281
	unsigned long nr_taken;
1282 1283
	unsigned long nr_dirty = 0;
	unsigned long nr_writeback = 0;
1284
	isolate_mode_t isolate_mode = 0;
1285
	int file = is_file_lru(lru);
1286 1287
	struct zone *zone = lruvec_zone(lruvec);
	struct zone_reclaim_stat *reclaim_stat = &lruvec->reclaim_stat;
1288

1289
	while (unlikely(too_many_isolated(zone, file, sc))) {
1290
		congestion_wait(BLK_RW_ASYNC, HZ/10);
1291 1292 1293 1294 1295 1296

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

Linus Torvalds's avatar
Linus Torvalds committed
1297
	lru_add_drain();
1298 1299

	if (!sc->may_unmap)
1300
		isolate_mode |= ISOLATE_UNMAPPED;
1301
	if (!sc->may_writepage)
1302
		isolate_mode |= ISOLATE_CLEAN;
1303

Linus Torvalds's avatar
Linus Torvalds committed
1304
	spin_lock_irq(&zone->lru_lock);
1305

1306 1307
	nr_taken = isolate_lru_pages(nr_to_scan, lruvec, &page_list,
				     &nr_scanned, sc, isolate_mode, lru);
1308 1309 1310 1311

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

1312
	if (global_reclaim(sc)) {
1313 1314
		zone->pages_scanned += nr_scanned;
		if (current_is_kswapd())
1315
			__count_zone_vm_events(PGSCAN_KSWAPD, zone, nr_scanned);
1316
		else
1317
			__count_zone_vm_events(PGSCAN_DIRECT, zone, nr_scanned);
1318
	}
1319
	spin_unlock_irq(&zone->lru_lock);
1320

1321
	if (nr_taken == 0)
1322
		return 0;
Andy Whitcroft's avatar
Andy Whitcroft committed
1323

1324 1325
	nr_reclaimed = shrink_page_list(&page_list, zone, sc, TTU_UNMAP,
					&nr_dirty, &nr_writeback, false);
1326

1327 1328
	spin_lock_irq(&zone->lru_lock);

1329
	reclaim_stat->recent_scanned[file] += nr_taken;
1330

1331 1332 1333 1334 1335 1336 1337 1338
	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);
	}
Nick Piggin's avatar
Nick Piggin committed
1339

1340
	putback_inactive_pages(lruvec, &page_list);
1341

1342
	__mod_zone_page_state(zone, NR_ISOLATED_ANON + file, -nr_taken);
1343 1344 1345 1346

	spin_unlock_irq(&zone->lru_lock);

	free_hot_cold_page_list(&page_list, 1);
1347

1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370
	/*
	 * If reclaim is isolating dirty pages under writeback, it implies
	 * that the long-lived page allocation rate is exceeding the page
	 * laundering rate. Either the global limits are not being effective
	 * at throttling processes due to the page distribution throughout
	 * zones or there is heavy usage of a slow backing device. The
	 * only option is to throttle from reclaim context which is not ideal
	 * as there is no guarantee the dirtying process is throttled in the
	 * same way balance_dirty_pages() manages.
	 *
	 * This scales the number of dirty pages that must be under writeback
	 * before throttling depending on priority. It is a simple backoff
	 * function that has the most effect in the range DEF_PRIORITY to
	 * DEF_PRIORITY-2 which is the priority reclaim is considered to be
	 * in trouble and reclaim is considered to be in trouble.
	 *
	 * DEF_PRIORITY   100% isolated pages must be PageWriteback to throttle
	 * DEF_PRIORITY-1  50% must be PageWriteback
	 * DEF_PRIORITY-2  25% must be PageWriteback, kswapd in trouble
	 * ...
	 * DEF_PRIORITY-6 For SWAP_CLUSTER_MAX isolated pages, throttle if any
	 *                     isolated page is PageWriteback
	 */
1371 1372
	if (nr_writeback && nr_writeback >=
			(nr_taken >> (DEF_PRIORITY - sc->priority)))
1373 1374
		wait_iff_congested(zone, BLK_RW_ASYNC, HZ/10);

1375 1376 1377
	trace_mm_vmscan_lru_shrink_inactive(zone->zone_pgdat->node_id,
		zone_idx(zone),
		nr_scanned, nr_reclaimed,
1378
		sc->priority,
1379
		trace_shrink_flags(file));
1380
	return nr_reclaimed;
Linus Torvalds's avatar
Linus Torvalds committed
1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399
}

/*
 * 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.
 */
1400

1401
static void move_active_pages_to_lru(struct lruvec *lruvec,
1402
				     struct list_head *list,
1403
				     struct list_head *pages_to_free,
1404 1405
				     enum lru_list lru)
{
1406
	struct zone *zone = lruvec_zone(lruvec);
1407 1408
	unsigned long pgmoved = 0;
	struct page *page;
1409
	int nr_pages;
1410 1411 1412

	while (!list_empty(list)) {
		page = lru_to_page(list);
1413
		lruvec = mem_cgroup_page_lruvec(page, zone);
1414 1415 1416 1417

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

1418 1419
		nr_pages = hpage_nr_pages(page);
		mem_cgroup_update_lru_size(lruvec, lru, nr_pages);
1420
		list_move(&page->lru, &lruvec->lists[lru]);
1421
		pgmoved += nr_pages;
1422

1423 1424 1425
		if (put_page_testzero(page)) {
			__ClearPageLRU(page);
			__ClearPageActive(page);
1426
			del_page_from_lru_list(page, lruvec, lru);
1427 1428 1429 1430 1431 1432 1433

			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);
1434 1435 1436 1437 1438 1439
		}
	}
	__mod_zone_page_state(zone, NR_LRU_BASE + lru, pgmoved);
	if (!is_active_lru(lru))
		__count_vm_events(PGDEACTIVATE, pgmoved);
}
1440

Hugh Dickins's avatar
Hugh Dickins committed
1441
static void shrink_active_list(unsigned long nr_to_scan,
1442
			       struct lruvec *lruvec,
1443
			       struct scan_control *sc,
1444
			       enum lru_list lru)
Linus Torvalds's avatar
Linus Torvalds committed
1445
{
1446
	unsigned long nr_taken;
Hugh Dickins's avatar
Hugh Dickins committed
1447
	unsigned long nr_scanned;
1448
	unsigned long vm_flags;
Linus Torvalds's avatar
Linus Torvalds committed
1449
	LIST_HEAD(l_hold);	/* The pages which were snipped off */
1450
	LIST_HEAD(l_active);
1451
	LIST_HEAD(l_inactive);
Linus Torvalds's avatar
Linus Torvalds committed
1452
	struct page *page;
1453
	struct zone_reclaim_stat *reclaim_stat = &lruvec->reclaim_stat;
1454
	unsigned long nr_rotated = 0;
1455
	isolate_mode_t isolate_mode = 0;
1456
	int file = is_file_lru(lru);
1457
	struct zone *zone = lruvec_zone(lruvec);
Linus Torvalds's avatar
Linus Torvalds committed
1458 1459

	lru_add_drain();
1460 1461

	if (!sc->may_unmap)
1462
		isolate_mode |= ISOLATE_UNMAPPED;
1463
	if (!sc->may_writepage)
1464
		isolate_mode |= ISOLATE_CLEAN;
1465

Linus Torvalds's avatar
Linus Torvalds committed
1466
	spin_lock_irq(&zone->lru_lock);
1467

1468 1469
	nr_taken = isolate_lru_pages(nr_to_scan, lruvec, &l_hold,
				     &nr_scanned, sc, isolate_mode, lru);
1470
	if (global_reclaim(sc))
Hugh Dickins's avatar
Hugh Dickins committed
1471
		zone->pages_scanned += nr_scanned;
1472

1473
	reclaim_stat->recent_scanned[file] += nr_taken;
1474

Hugh Dickins's avatar
Hugh Dickins committed
1475
	__count_zone_vm_events(PGREFILL, zone, nr_scanned);
1476
	__mod_zone_page_state(zone, NR_LRU_BASE + lru, -nr_taken);
1477
	__mod_zone_page_state(zone, NR_ISOLATED_ANON + file, nr_taken);
Linus Torvalds's avatar
Linus Torvalds committed
1478 1479 1480 1481 1482 1483
	spin_unlock_irq(&zone->lru_lock);

	while (!list_empty(&l_hold)) {
		cond_resched();
		page = lru_to_page(&l_hold);
		list_del(&page->lru);
1484

1485
		if (unlikely(!page_evictable(page))) {
1486 1487 1488 1489
			putback_lru_page(page);
			continue;
		}

1490 1491 1492 1493 1494 1495 1496 1497
		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);
			}
		}

1498 1499
		if (page_referenced(page, 0, sc->target_mem_cgroup,
				    &vm_flags)) {
1500
			nr_rotated += hpage_nr_pages(page);
1501 1502 1503 1504 1505 1506 1507 1508 1509
			/*
			 * 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.
			 */
1510
			if ((vm_flags & VM_EXEC) && page_is_file_cache(page)) {
1511 1512 1513 1514
				list_add(&page->lru, &l_active);
				continue;
			}
		}
1515

1516
		ClearPageActive(page);	/* we are de-activating */
Linus Torvalds's avatar
Linus Torvalds committed
1517 1518 1519
		list_add(&page->lru, &l_inactive);
	}

1520
	/*
1521
	 * Move pages back to the lru list.
1522
	 */
1523
	spin_lock_irq(&zone->lru_lock);
1524
	/*
1525 1526 1527 1528
	 * 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.
1529
	 */
1530
	reclaim_stat->recent_rotated[file] += nr_rotated;
1531

1532 1533
	move_active_pages_to_lru(lruvec, &l_active, &l_hold, lru);
	move_active_pages_to_lru(lruvec, &l_inactive, &l_hold, lru - LRU_ACTIVE);
1534
	__mod_zone_page_state(zone, NR_ISOLATED_ANON + file, -nr_taken);
1535
	spin_unlock_irq(&zone->lru_lock);
1536 1537

	free_hot_cold_page_list(&l_hold, 1);
Linus Torvalds's avatar
Linus Torvalds committed
1538 1539
}

1540
#ifdef CONFIG_SWAP
1541
static int inactive_anon_is_low_global(struct zone *zone)
1542 1543 1544 1545 1546 1547 1548 1549 1550 1551 1552 1553
{
	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;
}

1554 1555
/**
 * inactive_anon_is_low - check if anonymous pages need to be deactivated
1556
 * @lruvec: LRU vector to check
1557 1558 1559 1560
 *
 * Returns true if the zone does not have enough inactive anon pages,
 * meaning some active anon pages need to be deactivated.
 */
1561
static int inactive_anon_is_low(struct lruvec *lruvec)
1562
{
1563 1564 1565 1566 1567 1568 1569
	/*
	 * If we don't have swap space, anonymous page deactivation
	 * is pointless.
	 */
	if (!total_swap_pages)
		return 0;

1570
	if (!mem_cgroup_disabled())
1571
		return mem_cgroup_inactive_anon_is_low(lruvec);
1572

1573
	return inactive_anon_is_low_global(lruvec_zone(lruvec));
1574
}
1575
#else
1576
static inline int inactive_anon_is_low(struct lruvec *lruvec)
1577 1578 1579 1580
{
	return 0;
}
#endif
1581

1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593
static int inactive_file_is_low_global(struct zone *zone)
{
	unsigned long active, inactive;

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

	return (active > inactive);
}

/**
 * inactive_file_is_low - check if file pages need to be deactivated
1594
 * @lruvec: LRU vector to check
1595 1596 1597 1598 1599 1600 1601 1602 1603 1604 1605
 *
 * 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.
 */
1606
static int inactive_file_is_low(struct lruvec *lruvec)
1607
{
1608
	if (!mem_cgroup_disabled())
1609
		return mem_cgroup_inactive_file_is_low(lruvec);
1610

1611
	return inactive_file_is_low_global(lruvec_zone(lruvec));
1612 1613
}

1614
static int inactive_list_is_low(struct lruvec *lruvec, enum lru_list lru)
1615
{
1616
	if (is_file_lru(lru))
1617
		return inactive_file_is_low(lruvec);
1618
	else
1619
		return inactive_anon_is_low(lruvec);
1620 1621
}

1622
static unsigned long shrink_list(enum lru_list lru, unsigned long nr_to_scan,
1623
				 struct lruvec *lruvec, struct scan_control *sc)
1624
{
1625
	if (is_active_lru(lru)) {
1626
		if (inactive_list_is_low(lruvec, lru))
1627
			shrink_active_list(nr_to_scan, lruvec, sc, lru);
1628 1629 1630
		return 0;
	}

1631
	return shrink_inactive_list(nr_to_scan, lruvec, sc, lru);
1632 1633
}

1634
static int vmscan_swappiness(struct scan_control *sc)
1635
{
1636
	if (global_reclaim(sc))
1637
		return vm_swappiness;
1638
	return mem_cgroup_swappiness(sc->target_mem_cgroup);
1639 1640
}

1641 1642 1643 1644 1645 1646 1647
enum scan_balance {
	SCAN_EQUAL,
	SCAN_FRACT,
	SCAN_ANON,
	SCAN_FILE,
};

1648 1649 1650 1651 1652 1653
/*
 * 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.
 *
1654 1655
 * 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
1656
 */
1657
static void get_scan_count(struct lruvec *lruvec, struct scan_control *sc,
1658
			   unsigned long *nr)
1659
{
1660 1661 1662 1663
	struct zone_reclaim_stat *reclaim_stat = &lruvec->reclaim_stat;
	u64 fraction[2];
	u64 denominator = 0;	/* gcc */
	struct zone *zone = lruvec_zone(lruvec);
1664
	unsigned long anon_prio, file_prio;
1665 1666 1667
	enum scan_balance scan_balance;
	unsigned long anon, file, free;
	bool force_scan = false;
1668
	unsigned long ap, fp;
Hugh Dickins's avatar
Hugh Dickins committed
1669
	enum lru_list lru;
1670

1671 1672 1673 1674 1675 1676 1677 1678 1679 1680
	/*
	 * 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.
	 */
1681
	if (current_is_kswapd() && zone->all_unreclaimable)
1682
		force_scan = true;
1683
	if (!global_reclaim(sc))
1684
		force_scan = true;
1685 1686 1687

	/* If we have no swap space, do not bother scanning anon pages. */
	if (!sc->may_swap || (nr_swap_pages <= 0)) {
1688
		scan_balance = SCAN_FILE;
1689 1690
		goto out;
	}
1691

1692 1693 1694 1695 1696 1697 1698 1699
	/*
	 * 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)) {
1700
		scan_balance = SCAN_FILE;
1701 1702 1703 1704 1705 1706 1707 1708 1709
		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)) {
1710
		scan_balance = SCAN_EQUAL;
1711 1712 1713
		goto out;
	}

1714 1715 1716 1717
	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);
1718

1719 1720 1721 1722 1723 1724
	/*
	 * 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.
	 */
1725
	if (global_reclaim(sc)) {
1726
		free = zone_page_state(zone, NR_FREE_PAGES);
1727
		if (unlikely(file + free <= high_wmark_pages(zone))) {
1728
			scan_balance = SCAN_ANON;
1729
			goto out;
1730
		}
1731 1732
	}

1733 1734 1735 1736 1737
	/*
	 * There is enough inactive page cache, do not reclaim
	 * anything from the anonymous working set right now.
	 */
	if (!inactive_file_is_low(lruvec)) {
1738
		scan_balance = SCAN_FILE;
1739 1740 1741
		goto out;
	}

1742 1743
	scan_balance = SCAN_FRACT;

1744 1745 1746 1747
	/*
	 * With swappiness at 100, anonymous and file have the same priority.
	 * This scanning priority is essentially the inverse of IO cost.
	 */
1748
	anon_prio = vmscan_swappiness(sc);
1749
	file_prio = 200 - anon_prio;
1750

1751 1752 1753 1754 1755 1756 1757 1758 1759 1760 1761
	/*
	 * 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]
	 */
1762
	spin_lock_irq(&zone->lru_lock);
1763 1764 1765
	if (unlikely(reclaim_stat->recent_scanned[0] > anon / 4)) {
		reclaim_stat->recent_scanned[0] /= 2;
		reclaim_stat->recent_rotated[0] /= 2;
1766 1767
	}

1768 1769 1770
	if (unlikely(reclaim_stat->recent_scanned[1] > file / 4)) {
		reclaim_stat->recent_scanned[1] /= 2;
		reclaim_stat->recent_rotated[1] /= 2;
1771 1772 1773
	}

	/*
1774 1775 1776
	 * 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.
1777
	 */
1778
	ap = anon_prio * (reclaim_stat->recent_scanned[0] + 1);
1779
	ap /= reclaim_stat->recent_rotated[0] + 1;
1780

1781
	fp = file_prio * (reclaim_stat->recent_scanned[1] + 1);
1782
	fp /= reclaim_stat->recent_rotated[1] + 1;
1783
	spin_unlock_irq(&zone->lru_lock);
1784

1785 1786 1787 1788
	fraction[0] = ap;
	fraction[1] = fp;
	denominator = ap + fp + 1;
out:
Hugh Dickins's avatar
Hugh Dickins committed
1789 1790
	for_each_evictable_lru(lru) {
		int file = is_file_lru(lru);
1791
		unsigned long size;
1792
		unsigned long scan;
1793

1794
		size = get_lru_size(lruvec, lru);
1795
		scan = size >> sc->priority;
1796

1797 1798
		if (!scan && force_scan)
			scan = min(size, SWAP_CLUSTER_MAX);
1799 1800 1801 1802 1803 1804 1805 1806 1807 1808 1809 1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820

		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();
		}
Hugh Dickins's avatar
Hugh Dickins committed
1821
		nr[lru] = scan;
1822
	}
1823
}
1824

1825 1826 1827 1828 1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866 1867 1868 1869 1870 1871 1872 1873 1874 1875 1876
/*
 * 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];
	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;

	get_scan_count(lruvec, sc, nr);

	blk_start_plug(&plug);
	while (nr[LRU_INACTIVE_ANON] || nr[LRU_ACTIVE_FILE] ||
					nr[LRU_INACTIVE_FILE]) {
		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);
			}
		}
		/*
		 * On large memory systems, scan >> priority can become
		 * really large. This is fine for the starting priority;
		 * we want to put equal scanning pressure on each zone.
		 * However, if the VM has a harder time of freeing pages,
		 * with multiple processes reclaiming pages, the total
		 * freeing target can get unreasonably large.
		 */
		if (nr_reclaimed >= nr_to_reclaim &&
		    sc->priority < DEF_PRIORITY)
			break;
	}
	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);
}

1877
/* Use reclaim/compaction for costly allocs or under memory pressure */
1878
static bool in_reclaim_compaction(struct scan_control *sc)
1879
{
1880
	if (IS_ENABLED(CONFIG_COMPACTION) && sc->order &&
1881
			(sc->order > PAGE_ALLOC_COSTLY_ORDER ||
1882
			 sc->priority < DEF_PRIORITY - 2))
1883 1884 1885 1886 1887
		return true;

	return false;
}

1888
/*
1889 1890 1891 1892 1893
 * 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.
1894
 */
1895
static inline bool should_continue_reclaim(struct zone *zone,
1896 1897 1898 1899 1900 1901 1902 1903
					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 */
1904
	if (!in_reclaim_compaction(sc))
1905 1906
		return false;

1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928
	/* 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;
	}
1929 1930 1931 1932 1933 1934

	/*
	 * If we have not reclaimed enough pages for compaction and the
	 * inactive lists are large enough, continue reclaiming
	 */
	pages_for_compaction = (2UL << sc->order);
1935
	inactive_lru_pages = zone_page_state(zone, NR_INACTIVE_FILE);
1936
	if (nr_swap_pages > 0)
1937
		inactive_lru_pages += zone_page_state(zone, NR_INACTIVE_ANON);
1938 1939 1940 1941 1942
	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 */
1943
	switch (compaction_suitable(zone, sc->order)) {
1944 1945 1946 1947 1948 1949 1950 1951
	case COMPACT_PARTIAL:
	case COMPACT_CONTINUE:
		return false;
	default:
		return true;
	}
}

1952
static void shrink_zone(struct zone *zone, struct scan_control *sc)
Linus Torvalds's avatar
Linus Torvalds committed
1953
{
1954
	unsigned long nr_reclaimed, nr_scanned;
Linus Torvalds's avatar
Linus Torvalds committed
1955

1956 1957 1958 1959 1960 1961 1962
	do {
		struct mem_cgroup *root = sc->target_mem_cgroup;
		struct mem_cgroup_reclaim_cookie reclaim = {
			.zone = zone,
			.priority = sc->priority,
		};
		struct mem_cgroup *memcg;
1963

1964 1965
		nr_reclaimed = sc->nr_reclaimed;
		nr_scanned = sc->nr_scanned;
Linus Torvalds's avatar
Linus Torvalds committed
1966

1967 1968 1969
		memcg = mem_cgroup_iter(root, NULL, &reclaim);
		do {
			struct lruvec *lruvec;
1970

1971
			lruvec = mem_cgroup_zone_lruvec(zone, memcg);
1972

1973
			shrink_lruvec(lruvec, sc);
1974

1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994
			/*
			 * Limit reclaim has historically picked one
			 * memcg and scanned it with decreasing
			 * priority levels until nr_to_reclaim had
			 * been reclaimed.  This priority cycle is
			 * thus over after a single memcg.
			 *
			 * Direct reclaim and kswapd, on the other
			 * hand, have to scan all memory cgroups to
			 * fulfill the overall scan target for the
			 * zone.
			 */
			if (!global_reclaim(sc)) {
				mem_cgroup_iter_break(root, memcg);
				break;
			}
			memcg = mem_cgroup_iter(root, memcg, &reclaim);
		} while (memcg);
	} while (should_continue_reclaim(zone, sc->nr_reclaimed - nr_reclaimed,
					 sc->nr_scanned - nr_scanned, sc));
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
/* Returns true if compaction should go ahead for a high-order request */
static inline bool compaction_ready(struct zone *zone, struct scan_control *sc)
{
	unsigned long balance_gap, watermark;
	bool watermark_ok;

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

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

	/*
	 * If compaction is deferred, reclaim up to a point where
	 * compaction will have a chance of success when re-enabled
	 */
2023
	if (compaction_deferred(zone, sc->order))
2024 2025 2026 2027 2028 2029 2030 2031 2032
		return watermark_ok;

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

	return watermark_ok;
}

Linus Torvalds's avatar
Linus Torvalds committed
2033 2034 2035 2036 2037
/*
 * 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.
 *
2038 2039
 * We reclaim from a zone even if that zone is over high_wmark_pages(zone).
 * Because:
Linus Torvalds's avatar
Linus Torvalds committed
2040 2041
 * a) The caller may be trying to free *extra* pages to satisfy a higher-order
 *    allocation or
2042 2043 2044
 * 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.
Linus Torvalds's avatar
Linus Torvalds committed
2045 2046 2047
 *
 * If a zone is deemed to be full of pinned pages then just give it a light
 * scan then give up on it.
2048 2049
 *
 * This function returns true if a zone is being reclaimed for a costly
2050
 * high-order allocation and compaction is ready to begin. This indicates to
2051 2052
 * the caller that it should consider retrying the allocation instead of
 * further reclaim.
Linus Torvalds's avatar
Linus Torvalds committed
2053
 */
2054
static bool shrink_zones(struct zonelist *zonelist, struct scan_control *sc)
Linus Torvalds's avatar
Linus Torvalds committed
2055
{
2056
	struct zoneref *z;
2057
	struct zone *zone;
2058 2059
	unsigned long nr_soft_reclaimed;
	unsigned long nr_soft_scanned;
2060
	bool aborted_reclaim = false;
2061

2062 2063 2064 2065 2066 2067 2068 2069
	/*
	 * 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;

2070 2071
	for_each_zone_zonelist_nodemask(zone, z, zonelist,
					gfp_zone(sc->gfp_mask), sc->nodemask) {
2072
		if (!populated_zone(zone))
Linus Torvalds's avatar
Linus Torvalds committed
2073
			continue;
2074 2075 2076 2077
		/*
		 * Take care memory controller reclaiming has small influence
		 * to global LRU.
		 */
2078
		if (global_reclaim(sc)) {
2079 2080
			if (!cpuset_zone_allowed_hardwall(zone, GFP_KERNEL))
				continue;
2081 2082
			if (zone->all_unreclaimable &&
					sc->priority != DEF_PRIORITY)
2083
				continue;	/* Let kswapd poll it */
2084
			if (IS_ENABLED(CONFIG_COMPACTION)) {
2085
				/*
2086 2087 2088 2089 2090
				 * 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
2091 2092
				 * noticeable problem, like transparent huge
				 * page allocations.
2093
				 */
2094
				if (compaction_ready(zone, sc)) {
2095
					aborted_reclaim = true;
2096
					continue;
2097
				}
2098
			}
2099 2100 2101 2102 2103 2104 2105 2106 2107 2108 2109 2110 2111
			/*
			 * 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() */
2112
		}
2113

2114
		shrink_zone(zone, sc);
Linus Torvalds's avatar
Linus Torvalds committed
2115
	}
2116

2117
	return aborted_reclaim;
2118 2119 2120 2121 2122 2123 2124
}

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

2125
/* All zones in zonelist are unreclaimable? */
2126 2127 2128 2129 2130 2131 2132 2133 2134 2135 2136 2137
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;
2138 2139
		if (!zone->all_unreclaimable)
			return false;
2140 2141
	}

2142
	return true;
Linus Torvalds's avatar
Linus Torvalds committed
2143
}
2144

Linus Torvalds's avatar
Linus Torvalds committed
2145 2146 2147 2148 2149 2150 2151 2152
/*
 * 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
2153 2154 2155 2156
 * 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.
2157 2158 2159
 *
 * returns:	0, if no pages reclaimed
 * 		else, the number of pages reclaimed
Linus Torvalds's avatar
Linus Torvalds committed
2160
 */
2161
static unsigned long do_try_to_free_pages(struct zonelist *zonelist,
2162 2163
					struct scan_control *sc,
					struct shrink_control *shrink)
Linus Torvalds's avatar
Linus Torvalds committed
2164
{
2165
	unsigned long total_scanned = 0;
Linus Torvalds's avatar
Linus Torvalds committed
2166
	struct reclaim_state *reclaim_state = current->reclaim_state;
2167
	struct zoneref *z;
2168
	struct zone *zone;
2169
	unsigned long writeback_threshold;
2170
	bool aborted_reclaim;
Linus Torvalds's avatar
Linus Torvalds committed
2171

2172 2173
	delayacct_freepages_start();

2174
	if (global_reclaim(sc))
2175
		count_vm_event(ALLOCSTALL);
Linus Torvalds's avatar
Linus Torvalds committed
2176

2177
	do {
2178
		sc->nr_scanned = 0;
2179
		aborted_reclaim = shrink_zones(zonelist, sc);
2180

2181 2182 2183 2184
		/*
		 * Don't shrink slabs when reclaiming memory from
		 * over limit cgroups
		 */
2185
		if (global_reclaim(sc)) {
2186
			unsigned long lru_pages = 0;
2187 2188
			for_each_zone_zonelist(zone, z, zonelist,
					gfp_zone(sc->gfp_mask)) {
2189 2190 2191 2192 2193 2194
				if (!cpuset_zone_allowed_hardwall(zone, GFP_KERNEL))
					continue;

				lru_pages += zone_reclaimable_pages(zone);
			}

2195
			shrink_slab(shrink, sc->nr_scanned, lru_pages);
2196
			if (reclaim_state) {
2197
				sc->nr_reclaimed += reclaim_state->reclaimed_slab;
2198 2199
				reclaim_state->reclaimed_slab = 0;
			}
Linus Torvalds's avatar
Linus Torvalds committed
2200
		}
2201
		total_scanned += sc->nr_scanned;
2202
		if (sc->nr_reclaimed >= sc->nr_to_reclaim)
Linus Torvalds's avatar
Linus Torvalds committed
2203 2204 2205 2206 2207 2208 2209 2210 2211
			goto out;

		/*
		 * Try to write back as many pages as we just scanned.  This
		 * tends to cause slow streaming writers to write data to the
		 * disk smoothly, at the dirtying rate, which is nice.   But
		 * that's undesirable in laptop mode, where we *want* lumpy
		 * writeout.  So in laptop mode, write out the whole world.
		 */
2212 2213
		writeback_threshold = sc->nr_to_reclaim + sc->nr_to_reclaim / 2;
		if (total_scanned > writeback_threshold) {
2214 2215
			wakeup_flusher_threads(laptop_mode ? 0 : total_scanned,
						WB_REASON_TRY_TO_FREE_PAGES);
2216
			sc->may_writepage = 1;
Linus Torvalds's avatar
Linus Torvalds committed
2217 2218 2219
		}

		/* Take a nap, wait for some writeback to complete */
2220
		if (!sc->hibernation_mode && sc->nr_scanned &&
2221
		    sc->priority < DEF_PRIORITY - 2) {
2222 2223 2224
			struct zone *preferred_zone;

			first_zones_zonelist(zonelist, gfp_zone(sc->gfp_mask),
2225 2226
						&cpuset_current_mems_allowed,
						&preferred_zone);
2227 2228
			wait_iff_congested(preferred_zone, BLK_RW_ASYNC, HZ/10);
		}
2229
	} while (--sc->priority >= 0);
2230

Linus Torvalds's avatar
Linus Torvalds committed
2231
out:
2232 2233
	delayacct_freepages_end();

2234 2235 2236
	if (sc->nr_reclaimed)
		return sc->nr_reclaimed;

2237 2238 2239 2240 2241 2242 2243 2244
	/*
	 * 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;

2245 2246
	/* Aborted reclaim to try compaction? don't OOM, then */
	if (aborted_reclaim)
2247 2248
		return 1;

2249
	/* top priority shrink_zones still had more to do? don't OOM, then */
2250
	if (global_reclaim(sc) && !all_unreclaimable(zonelist, sc))
2251 2252 2253
		return 1;

	return 0;
Linus Torvalds's avatar
Linus Torvalds committed
2254 2255
}

2256 2257 2258 2259 2260 2261 2262 2263 2264 2265 2266 2267 2268 2269 2270 2271 2272 2273 2274 2275 2276 2277 2278 2279 2280 2281 2282 2283 2284 2285
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
2286 2287 2288 2289
 * 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.
2290
 */
2291
static bool throttle_direct_reclaim(gfp_t gfp_mask, struct zonelist *zonelist,
2292 2293 2294 2295 2296 2297 2298 2299 2300 2301 2302 2303 2304 2305
					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)
2306 2307 2308 2309 2310 2311 2312 2313
		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;
2314 2315 2316 2317 2318

	/* Check if the pfmemalloc reserves are ok */
	first_zones_zonelist(zonelist, high_zoneidx, NULL, &zone);
	pgdat = zone->zone_pgdat;
	if (pfmemalloc_watermark_ok(pgdat))
2319
		goto out;
2320

2321 2322 2323
	/* Account for the throttling */
	count_vm_event(PGSCAN_DIRECT_THROTTLE);

2324 2325 2326 2327 2328 2329 2330 2331 2332 2333 2334
	/*
	 * 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);
2335 2336

		goto check_pending;
2337 2338 2339 2340 2341
	}

	/* Throttle until kswapd wakes the process */
	wait_event_killable(zone->zone_pgdat->pfmemalloc_wait,
		pfmemalloc_watermark_ok(pgdat));
2342 2343 2344 2345 2346 2347 2348

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

out:
	return false;
2349 2350
}

2351
unsigned long try_to_free_pages(struct zonelist *zonelist, int order,
2352
				gfp_t gfp_mask, nodemask_t *nodemask)
2353
{
2354
	unsigned long nr_reclaimed;
2355 2356 2357
	struct scan_control sc = {
		.gfp_mask = gfp_mask,
		.may_writepage = !laptop_mode,
2358
		.nr_to_reclaim = SWAP_CLUSTER_MAX,
2359
		.may_unmap = 1,
2360
		.may_swap = 1,
2361
		.order = order,
2362
		.priority = DEF_PRIORITY,
2363
		.target_mem_cgroup = NULL,
2364
		.nodemask = nodemask,
2365
	};
2366 2367 2368
	struct shrink_control shrink = {
		.gfp_mask = sc.gfp_mask,
	};
2369

2370
	/*
2371 2372 2373
	 * 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.
2374
	 */
2375
	if (throttle_direct_reclaim(gfp_mask, zonelist, nodemask))
2376 2377
		return 1;

2378 2379 2380 2381
	trace_mm_vmscan_direct_reclaim_begin(order,
				sc.may_writepage,
				gfp_mask);

2382
	nr_reclaimed = do_try_to_free_pages(zonelist, &sc, &shrink);
2383 2384 2385 2386

	trace_mm_vmscan_direct_reclaim_end(nr_reclaimed);

	return nr_reclaimed;
2387 2388
}

2389
#ifdef CONFIG_MEMCG
2390

2391
unsigned long mem_cgroup_shrink_node_zone(struct mem_cgroup *memcg,
2392
						gfp_t gfp_mask, bool noswap,
2393 2394
						struct zone *zone,
						unsigned long *nr_scanned)
2395 2396
{
	struct scan_control sc = {
2397
		.nr_scanned = 0,
2398
		.nr_to_reclaim = SWAP_CLUSTER_MAX,
2399 2400 2401 2402
		.may_writepage = !laptop_mode,
		.may_unmap = 1,
		.may_swap = !noswap,
		.order = 0,
2403
		.priority = 0,
2404
		.target_mem_cgroup = memcg,
2405
	};
2406
	struct lruvec *lruvec = mem_cgroup_zone_lruvec(zone, memcg);
2407

2408 2409
	sc.gfp_mask = (gfp_mask & GFP_RECLAIM_MASK) |
			(GFP_HIGHUSER_MOVABLE & ~GFP_RECLAIM_MASK);
2410

2411
	trace_mm_vmscan_memcg_softlimit_reclaim_begin(sc.order,
2412 2413 2414
						      sc.may_writepage,
						      sc.gfp_mask);

2415 2416 2417 2418 2419 2420 2421
	/*
	 * 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.
	 */
2422
	shrink_lruvec(lruvec, &sc);
2423 2424 2425

	trace_mm_vmscan_memcg_softlimit_reclaim_end(sc.nr_reclaimed);

2426
	*nr_scanned = sc.nr_scanned;
2427 2428 2429
	return sc.nr_reclaimed;
}

2430
unsigned long try_to_free_mem_cgroup_pages(struct mem_cgroup *memcg,
KOSAKI Motohiro's avatar
KOSAKI Motohiro committed
2431
					   gfp_t gfp_mask,
2432
					   bool noswap)
2433
{
2434
	struct zonelist *zonelist;
2435
	unsigned long nr_reclaimed;
2436
	int nid;
2437 2438
	struct scan_control sc = {
		.may_writepage = !laptop_mode,
2439
		.may_unmap = 1,
2440
		.may_swap = !noswap,
2441
		.nr_to_reclaim = SWAP_CLUSTER_MAX,
2442
		.order = 0,
2443
		.priority = DEF_PRIORITY,
2444
		.target_mem_cgroup = memcg,
2445
		.nodemask = NULL, /* we don't care the placement */
2446 2447 2448 2449 2450
		.gfp_mask = (gfp_mask & GFP_RECLAIM_MASK) |
				(GFP_HIGHUSER_MOVABLE & ~GFP_RECLAIM_MASK),
	};
	struct shrink_control shrink = {
		.gfp_mask = sc.gfp_mask,
2451 2452
	};

2453 2454 2455 2456 2457
	/*
	 * 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.
	 */
2458
	nid = mem_cgroup_select_victim_node(memcg);
2459 2460

	zonelist = NODE_DATA(nid)->node_zonelists;
2461 2462 2463 2464 2465

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

2466
	nr_reclaimed = do_try_to_free_pages(zonelist, &sc, &shrink);
2467 2468 2469 2470

	trace_mm_vmscan_memcg_reclaim_end(nr_reclaimed);

	return nr_reclaimed;
2471 2472 2473
}
#endif

2474
static void age_active_anon(struct zone *zone, struct scan_control *sc)
2475
{
2476
	struct mem_cgroup *memcg;
2477

2478 2479 2480 2481 2482
	if (!total_swap_pages)
		return;

	memcg = mem_cgroup_iter(NULL, NULL, NULL);
	do {
2483
		struct lruvec *lruvec = mem_cgroup_zone_lruvec(zone, memcg);
2484

2485
		if (inactive_anon_is_low(lruvec))
2486
			shrink_active_list(SWAP_CLUSTER_MAX, lruvec,
2487
					   sc, LRU_ACTIVE_ANON);
2488 2489 2490

		memcg = mem_cgroup_iter(NULL, memcg, NULL);
	} while (memcg);
2491 2492
}

2493 2494 2495 2496 2497 2498 2499
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;

2500 2501
	if (IS_ENABLED(CONFIG_COMPACTION) && order &&
	    !compaction_suitable(zone, order))
2502 2503 2504 2505 2506
		return false;

	return true;
}

2507
/*
2508 2509 2510 2511 2512 2513 2514 2515 2516 2517
 * 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.
2518 2519 2520 2521
 * 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
Lucas De Marchi's avatar
Lucas De Marchi committed
2522
 *     percentage of the middle zones. For example, on 32-bit x86, highmem
2523 2524 2525 2526
 *     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.
 */
2527
static bool pgdat_balanced(pg_data_t *pgdat, int order, int classzone_idx)
2528 2529
{
	unsigned long present_pages = 0;
2530
	unsigned long balanced_pages = 0;
2531 2532
	int i;

2533 2534 2535
	/* Check the watermark levels */
	for (i = 0; i <= classzone_idx; i++) {
		struct zone *zone = pgdat->node_zones + i;
2536

2537 2538 2539 2540 2541 2542 2543 2544 2545 2546 2547 2548 2549 2550 2551 2552 2553 2554 2555 2556 2557 2558 2559 2560 2561 2562 2563
		if (!populated_zone(zone))
			continue;

		present_pages += zone->present_pages;

		/*
		 * 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) {
			balanced_pages += zone->present_pages;
			continue;
		}

		if (zone_balanced(zone, order, 0, i))
			balanced_pages += zone->present_pages;
		else if (!order)
			return false;
	}

	if (order)
		return balanced_pages >= (present_pages >> 2);
	else
		return true;
2564 2565
}

2566 2567 2568 2569 2570 2571 2572
/*
 * 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,
2573
					int classzone_idx)
2574 2575 2576
{
	/* If a direct reclaimer woke kswapd within HZ/10, it's premature */
	if (remaining)
2577 2578 2579 2580 2581 2582 2583 2584 2585 2586 2587 2588 2589 2590 2591
		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;
	}
2592

2593
	return pgdat_balanced(pgdat, order, classzone_idx);
2594 2595
}

Linus Torvalds's avatar
Linus Torvalds committed
2596 2597
/*
 * For kswapd, balance_pgdat() will work across all this node's zones until
2598
 * they are all at high_wmark_pages(zone).
Linus Torvalds's avatar
Linus Torvalds committed
2599
 *
2600
 * Returns the final order kswapd was reclaiming at
Linus Torvalds's avatar
Linus Torvalds committed
2601 2602 2603 2604 2605 2606 2607 2608 2609 2610
 *
 * 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
2611 2612 2613 2614 2615
 * 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.
Linus Torvalds's avatar
Linus Torvalds committed
2616
 */
2617
static unsigned long balance_pgdat(pg_data_t *pgdat, int order,
2618
							int *classzone_idx)
Linus Torvalds's avatar
Linus Torvalds committed
2619
{
2620
	struct zone *unbalanced_zone;
Linus Torvalds's avatar
Linus Torvalds committed
2621
	int i;
2622
	int end_zone = 0;	/* Inclusive.  0 = ZONE_DMA */
2623
	unsigned long total_scanned;
Linus Torvalds's avatar
Linus Torvalds committed
2624
	struct reclaim_state *reclaim_state = current->reclaim_state;
2625 2626
	unsigned long nr_soft_reclaimed;
	unsigned long nr_soft_scanned;
2627 2628
	struct scan_control sc = {
		.gfp_mask = GFP_KERNEL,
2629
		.may_unmap = 1,
2630
		.may_swap = 1,
2631 2632 2633 2634 2635
		/*
		 * kswapd doesn't want to be bailed out while reclaim. because
		 * we want to put equal scanning pressure on each zone.
		 */
		.nr_to_reclaim = ULONG_MAX,
Andy Whitcroft's avatar
Andy Whitcroft committed
2636
		.order = order,
2637
		.target_mem_cgroup = NULL,
2638
	};
2639 2640 2641
	struct shrink_control shrink = {
		.gfp_mask = sc.gfp_mask,
	};
Linus Torvalds's avatar
Linus Torvalds committed
2642 2643
loop_again:
	total_scanned = 0;
2644
	sc.priority = DEF_PRIORITY;
2645
	sc.nr_reclaimed = 0;
Christoph Lameter's avatar
Christoph Lameter committed
2646
	sc.may_writepage = !laptop_mode;
2647
	count_vm_event(PAGEOUTRUN);
Linus Torvalds's avatar
Linus Torvalds committed
2648

2649
	do {
Linus Torvalds's avatar
Linus Torvalds committed
2650
		unsigned long lru_pages = 0;
2651
		int has_under_min_watermark_zone = 0;
Linus Torvalds's avatar
Linus Torvalds committed
2652

2653
		unbalanced_zone = NULL;
Linus Torvalds's avatar
Linus Torvalds committed
2654

2655 2656 2657 2658 2659 2660
		/*
		 * 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;
Linus Torvalds's avatar
Linus Torvalds committed
2661

2662 2663
			if (!populated_zone(zone))
				continue;
Linus Torvalds's avatar
Linus Torvalds committed
2664

2665 2666
			if (zone->all_unreclaimable &&
			    sc.priority != DEF_PRIORITY)
2667
				continue;
Linus Torvalds's avatar
Linus Torvalds committed
2668

2669 2670 2671 2672
			/*
			 * Do some background aging of the anon list, to give
			 * pages a chance to be referenced before reclaiming.
			 */
2673
			age_active_anon(zone, &sc);
2674

2675 2676 2677 2678 2679 2680 2681 2682 2683 2684 2685
			/*
			 * 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;
			}

2686
			if (!zone_balanced(zone, order, 0, 0)) {
2687
				end_zone = i;
2688
				break;
2689 2690 2691
			} else {
				/* If balanced, clear the congested flag */
				zone_clear_flag(zone, ZONE_CONGESTED);
Linus Torvalds's avatar
Linus Torvalds committed
2692 2693
			}
		}
2694 2695 2696
		if (i < 0)
			goto out;

Linus Torvalds's avatar
Linus Torvalds committed
2697 2698 2699
		for (i = 0; i <= end_zone; i++) {
			struct zone *zone = pgdat->node_zones + i;

2700
			lru_pages += zone_reclaimable_pages(zone);
Linus Torvalds's avatar
Linus Torvalds committed
2701 2702 2703 2704 2705 2706 2707 2708 2709 2710 2711 2712 2713
		}

		/*
		 * 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;
2714
			int nr_slab, testorder;
2715
			unsigned long balance_gap;
Linus Torvalds's avatar
Linus Torvalds committed
2716

2717
			if (!populated_zone(zone))
Linus Torvalds's avatar
Linus Torvalds committed
2718 2719
				continue;

2720 2721
			if (zone->all_unreclaimable &&
			    sc.priority != DEF_PRIORITY)
Linus Torvalds's avatar
Linus Torvalds committed
2722 2723 2724
				continue;

			sc.nr_scanned = 0;
2725

2726
			nr_soft_scanned = 0;
2727 2728 2729
			/*
			 * Call soft limit reclaim before calling shrink_zone.
			 */
2730 2731 2732 2733 2734
			nr_soft_reclaimed = mem_cgroup_soft_limit_reclaim(zone,
							order, sc.gfp_mask,
							&nr_soft_scanned);
			sc.nr_reclaimed += nr_soft_reclaimed;
			total_scanned += nr_soft_scanned;
2735

2736
			/*
2737 2738 2739 2740 2741 2742
			 * 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.
2743
			 */
2744 2745 2746 2747
			balance_gap = min(low_wmark_pages(zone),
				(zone->present_pages +
					KSWAPD_ZONE_BALANCE_GAP_RATIO-1) /
				KSWAPD_ZONE_BALANCE_GAP_RATIO);
2748 2749 2750 2751 2752 2753 2754 2755
			/*
			 * Kswapd reclaims only single pages with compaction
			 * enabled. Trying too hard to reclaim until contiguous
			 * free pages have become available can hurt performance
			 * by evicting too much useful data from memory.
			 * Do not reclaim more than needed for compaction.
			 */
			testorder = order;
2756
			if (IS_ENABLED(CONFIG_COMPACTION) && order &&
2757 2758 2759 2760
					compaction_suitable(zone, order) !=
						COMPACT_SKIPPED)
				testorder = 0;

2761
			if ((buffer_heads_over_limit && is_highmem_idx(i)) ||
2762 2763
			    !zone_balanced(zone, testorder,
					   balance_gap, end_zone)) {
2764
				shrink_zone(zone, &sc);
2765

2766 2767 2768 2769 2770 2771 2772 2773 2774
				reclaim_state->reclaimed_slab = 0;
				nr_slab = shrink_slab(&shrink, sc.nr_scanned, lru_pages);
				sc.nr_reclaimed += reclaim_state->reclaimed_slab;
				total_scanned += sc.nr_scanned;

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

Linus Torvalds's avatar
Linus Torvalds committed
2775 2776 2777 2778 2779 2780
			/*
			 * If we've done a decent amount of scanning and
			 * the reclaim ratio is low, start doing writepage
			 * even in laptop mode
			 */
			if (total_scanned > SWAP_CLUSTER_MAX * 2 &&
2781
			    total_scanned > sc.nr_reclaimed + sc.nr_reclaimed / 2)
Linus Torvalds's avatar
Linus Torvalds committed
2782
				sc.may_writepage = 1;
2783

2784 2785 2786
			if (zone->all_unreclaimable) {
				if (end_zone && end_zone == i)
					end_zone--;
2787
				continue;
2788
			}
2789

2790
			if (!zone_balanced(zone, testorder, 0, end_zone)) {
2791
				unbalanced_zone = zone;
2792 2793 2794 2795 2796
				/*
				 * We are still under min water mark.  This
				 * means that we have a GFP_ATOMIC allocation
				 * failure risk. Hurry up!
				 */
2797
				if (!zone_watermark_ok_safe(zone, order,
2798 2799
					    min_wmark_pages(zone), end_zone, 0))
					has_under_min_watermark_zone = 1;
2800 2801 2802 2803 2804 2805
			} else {
				/*
				 * If a zone reaches its high watermark,
				 * consider it to be no longer congested. It's
				 * possible there are dirty pages backed by
				 * congested BDIs but as pressure is relieved,
2806
				 * speculatively avoid congestion waits
2807 2808
				 */
				zone_clear_flag(zone, ZONE_CONGESTED);
2809
			}
2810

Linus Torvalds's avatar
Linus Torvalds committed
2811
		}
2812 2813 2814 2815 2816 2817 2818 2819 2820 2821

		/*
		 * 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);

2822
		if (pgdat_balanced(pgdat, order, *classzone_idx))
Linus Torvalds's avatar
Linus Torvalds committed
2823 2824 2825 2826 2827
			break;		/* kswapd: all done */
		/*
		 * OK, kswapd is getting into trouble.  Take a nap, then take
		 * another pass across the zones.
		 */
2828
		if (total_scanned && (sc.priority < DEF_PRIORITY - 2)) {
2829 2830
			if (has_under_min_watermark_zone)
				count_vm_event(KSWAPD_SKIP_CONGESTION_WAIT);
2831
			else if (unbalanced_zone)
2832
				wait_iff_congested(unbalanced_zone, BLK_RW_ASYNC, HZ/10);
2833
		}
Linus Torvalds's avatar
Linus Torvalds committed
2834 2835 2836 2837 2838 2839 2840

		/*
		 * We do this so kswapd doesn't build up large priorities for
		 * example when it is freeing in parallel with allocators. It
		 * matches the direct reclaim path behaviour in terms of impact
		 * on zone->*_priority.
		 */
2841
		if (sc.nr_reclaimed >= SWAP_CLUSTER_MAX)
Linus Torvalds's avatar
Linus Torvalds committed
2842
			break;
2843
	} while (--sc.priority >= 0);
Linus Torvalds's avatar
Linus Torvalds committed
2844
out:
2845

2846
	if (!pgdat_balanced(pgdat, order, *classzone_idx)) {
Linus Torvalds's avatar
Linus Torvalds committed
2847
		cond_resched();
2848 2849 2850

		try_to_freeze();

2851 2852 2853 2854 2855 2856 2857 2858 2859 2860 2861 2862 2863 2864 2865 2866 2867
		/*
		 * Fragmentation may mean that the system cannot be
		 * rebalanced for high-order allocations in all zones.
		 * At this point, if nr_reclaimed < SWAP_CLUSTER_MAX,
		 * it means the zones have been fully scanned and are still
		 * not balanced. For high-order allocations, there is
		 * little point trying all over again as kswapd may
		 * infinite loop.
		 *
		 * Instead, recheck all watermarks at order-0 as they
		 * are the most important. If watermarks are ok, kswapd will go
		 * back to sleep. High-order users can still perform direct
		 * reclaim if they wish.
		 */
		if (sc.nr_reclaimed < SWAP_CLUSTER_MAX)
			order = sc.order = 0;

Linus Torvalds's avatar
Linus Torvalds committed
2868 2869 2870
		goto loop_again;
	}

2871 2872 2873 2874 2875 2876 2877 2878 2879
	/*
	 * If kswapd was reclaiming at a higher order, it has the option of
	 * sleeping without all zones being balanced. Before it does, it must
	 * ensure that the watermarks for order-0 on *all* zones are met and
	 * that the congestion flags are cleared. The congestion flag must
	 * be cleared as kswapd is the only mechanism that clears the flag
	 * and it is potentially going to sleep here.
	 */
	if (order) {
2880 2881
		int zones_need_compaction = 1;

2882 2883 2884 2885 2886 2887
		for (i = 0; i <= end_zone; i++) {
			struct zone *zone = pgdat->node_zones + i;

			if (!populated_zone(zone))
				continue;

2888 2889 2890 2891
			/* Check if the memory needs to be defragmented. */
			if (zone_watermark_ok(zone, order,
				    low_wmark_pages(zone), *classzone_idx, 0))
				zones_need_compaction = 0;
2892
		}
2893 2894 2895

		if (zones_need_compaction)
			compact_pgdat(pgdat, order);
2896 2897
	}

2898
	/*
2899
	 * Return the order we were reclaiming at so prepare_kswapd_sleep()
2900 2901 2902 2903
	 * 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
	 */
2904
	*classzone_idx = end_zone;
2905
	return order;
Linus Torvalds's avatar
Linus Torvalds committed
2906 2907
}

2908
static void kswapd_try_to_sleep(pg_data_t *pgdat, int order, int classzone_idx)
2909 2910 2911 2912 2913 2914 2915 2916 2917 2918
{
	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 */
2919
	if (prepare_kswapd_sleep(pgdat, order, remaining, classzone_idx)) {
2920 2921 2922 2923 2924 2925 2926 2927 2928
		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.
	 */
2929
	if (prepare_kswapd_sleep(pgdat, order, remaining, classzone_idx)) {
2930 2931 2932 2933 2934 2935 2936 2937 2938 2939 2940
		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);
2941

2942 2943 2944 2945 2946 2947 2948 2949
		/*
		 * 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);

2950 2951 2952
		if (!kthread_should_stop())
			schedule();

2953 2954 2955 2956 2957 2958 2959 2960 2961 2962
		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);
}

Linus Torvalds's avatar
Linus Torvalds committed
2963 2964
/*
 * The background pageout daemon, started as a kernel thread
2965
 * from the init process.
Linus Torvalds's avatar
Linus Torvalds committed
2966 2967 2968 2969 2970 2971 2972 2973 2974 2975 2976 2977
 *
 * 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)
{
2978
	unsigned long order, new_order;
2979
	unsigned balanced_order;
2980
	int classzone_idx, new_classzone_idx;
2981
	int balanced_classzone_idx;
Linus Torvalds's avatar
Linus Torvalds committed
2982 2983
	pg_data_t *pgdat = (pg_data_t*)p;
	struct task_struct *tsk = current;
2984

Linus Torvalds's avatar
Linus Torvalds committed
2985 2986 2987
	struct reclaim_state reclaim_state = {
		.reclaimed_slab = 0,
	};
2988
	const struct cpumask *cpumask = cpumask_of_node(pgdat->node_id);
Linus Torvalds's avatar
Linus Torvalds committed
2989

2990 2991
	lockdep_set_current_reclaim_state(GFP_KERNEL);

Rusty Russell's avatar
Rusty Russell committed
2992
	if (!cpumask_empty(cpumask))
2993
		set_cpus_allowed_ptr(tsk, cpumask);
Linus Torvalds's avatar
Linus Torvalds committed
2994 2995 2996 2997 2998 2999 3000 3001 3002 3003 3004 3005 3006 3007
	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).
	 */
3008
	tsk->flags |= PF_MEMALLOC | PF_SWAPWRITE | PF_KSWAPD;
3009
	set_freezable();
Linus Torvalds's avatar
Linus Torvalds committed
3010

3011
	order = new_order = 0;
3012
	balanced_order = 0;
3013
	classzone_idx = new_classzone_idx = pgdat->nr_zones - 1;
3014
	balanced_classzone_idx = classzone_idx;
Linus Torvalds's avatar
Linus Torvalds committed
3015
	for ( ; ; ) {
3016
		bool ret;
3017

3018 3019 3020 3021 3022
		/*
		 * 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
		 */
3023 3024
		if (balanced_classzone_idx >= new_classzone_idx &&
					balanced_order == new_order) {
3025 3026 3027 3028 3029 3030
			new_order = pgdat->kswapd_max_order;
			new_classzone_idx = pgdat->classzone_idx;
			pgdat->kswapd_max_order =  0;
			pgdat->classzone_idx = pgdat->nr_zones - 1;
		}

3031
		if (order < new_order || classzone_idx > new_classzone_idx) {
Linus Torvalds's avatar
Linus Torvalds committed
3032 3033
			/*
			 * Don't sleep if someone wants a larger 'order'
3034
			 * allocation or has tigher zone constraints
Linus Torvalds's avatar
Linus Torvalds committed
3035 3036
			 */
			order = new_order;
3037
			classzone_idx = new_classzone_idx;
Linus Torvalds's avatar
Linus Torvalds committed
3038
		} else {
3039 3040
			kswapd_try_to_sleep(pgdat, balanced_order,
						balanced_classzone_idx);
Linus Torvalds's avatar
Linus Torvalds committed
3041
			order = pgdat->kswapd_max_order;
3042
			classzone_idx = pgdat->classzone_idx;
3043 3044
			new_order = order;
			new_classzone_idx = classzone_idx;
3045
			pgdat->kswapd_max_order = 0;
3046
			pgdat->classzone_idx = pgdat->nr_zones - 1;
Linus Torvalds's avatar
Linus Torvalds committed
3047 3048
		}

3049 3050 3051 3052 3053 3054 3055 3056
		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
		 */
3057 3058
		if (!ret) {
			trace_mm_vmscan_kswapd_wake(pgdat->node_id, order);
3059 3060 3061
			balanced_classzone_idx = classzone_idx;
			balanced_order = balance_pgdat(pgdat, order,
						&balanced_classzone_idx);
3062
		}
Linus Torvalds's avatar
Linus Torvalds committed
3063
	}
3064 3065

	current->reclaim_state = NULL;
Linus Torvalds's avatar
Linus Torvalds committed
3066 3067 3068 3069 3070 3071
	return 0;
}

/*
 * A zone is low on free memory, so wake its kswapd task to service it.
 */
3072
void wakeup_kswapd(struct zone *zone, int order, enum zone_type classzone_idx)
Linus Torvalds's avatar
Linus Torvalds committed
3073 3074 3075
{
	pg_data_t *pgdat;

3076
	if (!populated_zone(zone))
Linus Torvalds's avatar
Linus Torvalds committed
3077 3078
		return;

3079
	if (!cpuset_zone_allowed_hardwall(zone, GFP_KERNEL))
Linus Torvalds's avatar
Linus Torvalds committed
3080
		return;
3081
	pgdat = zone->zone_pgdat;
3082
	if (pgdat->kswapd_max_order < order) {
Linus Torvalds's avatar
Linus Torvalds committed
3083
		pgdat->kswapd_max_order = order;
3084 3085
		pgdat->classzone_idx = min(pgdat->classzone_idx, classzone_idx);
	}
3086
	if (!waitqueue_active(&pgdat->kswapd_wait))
Linus Torvalds's avatar
Linus Torvalds committed
3087
		return;
3088 3089 3090 3091
	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);
3092
	wake_up_interruptible(&pgdat->kswapd_wait);
Linus Torvalds's avatar
Linus Torvalds committed
3093 3094
}

3095 3096 3097 3098 3099 3100 3101 3102
/*
 * 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)
3103
{
3104 3105 3106 3107 3108 3109 3110 3111 3112 3113 3114 3115 3116 3117 3118 3119 3120 3121 3122 3123 3124 3125 3126 3127
	int nr;

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

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

	return nr;
}

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

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

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

	return nr;
3128 3129
}

3130
#ifdef CONFIG_HIBERNATION
Linus Torvalds's avatar
Linus Torvalds committed
3131
/*
3132
 * Try to free `nr_to_reclaim' of memory, system-wide, and return the number of
3133 3134 3135 3136 3137
 * 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
Linus Torvalds's avatar
Linus Torvalds committed
3138
 */
3139
unsigned long shrink_all_memory(unsigned long nr_to_reclaim)
Linus Torvalds's avatar
Linus Torvalds committed
3140
{
3141 3142
	struct reclaim_state reclaim_state;
	struct scan_control sc = {
3143 3144 3145
		.gfp_mask = GFP_HIGHUSER_MOVABLE,
		.may_swap = 1,
		.may_unmap = 1,
3146
		.may_writepage = 1,
3147 3148 3149
		.nr_to_reclaim = nr_to_reclaim,
		.hibernation_mode = 1,
		.order = 0,
3150
		.priority = DEF_PRIORITY,
Linus Torvalds's avatar
Linus Torvalds committed
3151
	};
3152 3153 3154 3155
	struct shrink_control shrink = {
		.gfp_mask = sc.gfp_mask,
	};
	struct zonelist *zonelist = node_zonelist(numa_node_id(), sc.gfp_mask);
3156 3157
	struct task_struct *p = current;
	unsigned long nr_reclaimed;
Linus Torvalds's avatar
Linus Torvalds committed
3158

3159 3160 3161 3162
	p->flags |= PF_MEMALLOC;
	lockdep_set_current_reclaim_state(sc.gfp_mask);
	reclaim_state.reclaimed_slab = 0;
	p->reclaim_state = &reclaim_state;
3163

3164
	nr_reclaimed = do_try_to_free_pages(zonelist, &sc, &shrink);
3165

3166 3167 3168
	p->reclaim_state = NULL;
	lockdep_clear_current_reclaim_state();
	p->flags &= ~PF_MEMALLOC;
3169

3170
	return nr_reclaimed;
Linus Torvalds's avatar
Linus Torvalds committed
3171
}
3172
#endif /* CONFIG_HIBERNATION */
Linus Torvalds's avatar
Linus Torvalds committed
3173 3174 3175 3176 3177

/* 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. */
3178 3179
static int cpu_callback(struct notifier_block *nfb, unsigned long action,
			void *hcpu)
Linus Torvalds's avatar
Linus Torvalds committed
3180
{
3181
	int nid;
Linus Torvalds's avatar
Linus Torvalds committed
3182

3183
	if (action == CPU_ONLINE || action == CPU_ONLINE_FROZEN) {
3184
		for_each_node_state(nid, N_MEMORY) {
3185
			pg_data_t *pgdat = NODE_DATA(nid);
3186 3187 3188
			const struct cpumask *mask;

			mask = cpumask_of_node(pgdat->node_id);
3189

3190
			if (cpumask_any_and(cpu_online_mask, mask) < nr_cpu_ids)
Linus Torvalds's avatar
Linus Torvalds committed
3191
				/* One of our CPUs online: restore mask */
3192
				set_cpus_allowed_ptr(pgdat->kswapd, mask);
Linus Torvalds's avatar
Linus Torvalds committed
3193 3194 3195 3196 3197
		}
	}
	return NOTIFY_OK;
}

3198 3199 3200 3201 3202 3203 3204 3205 3206 3207 3208 3209 3210 3211 3212 3213
/*
 * 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);
3214
		pgdat->kswapd = NULL;
3215 3216
		pr_err("Failed to start kswapd on node %d\n", nid);
		ret = PTR_ERR(pgdat->kswapd);
3217 3218 3219 3220
	}
	return ret;
}

3221
/*
3222 3223
 * Called by memory hotplug when all memory in a node is offlined.  Caller must
 * hold lock_memory_hotplug().
3224 3225 3226 3227 3228
 */
void kswapd_stop(int nid)
{
	struct task_struct *kswapd = NODE_DATA(nid)->kswapd;

3229
	if (kswapd) {
3230
		kthread_stop(kswapd);
3231 3232
		NODE_DATA(nid)->kswapd = NULL;
	}
3233 3234
}

Linus Torvalds's avatar
Linus Torvalds committed
3235 3236
static int __init kswapd_init(void)
{
3237
	int nid;
3238

Linus Torvalds's avatar
Linus Torvalds committed
3239
	swap_setup();
3240
	for_each_node_state(nid, N_MEMORY)
3241
 		kswapd_run(nid);
Linus Torvalds's avatar
Linus Torvalds committed
3242 3243 3244 3245 3246
	hotcpu_notifier(cpu_callback, 0);
	return 0;
}

module_init(kswapd_init)
3247 3248 3249 3250 3251 3252 3253 3254 3255 3256

#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;

3257
#define RECLAIM_OFF 0
3258
#define RECLAIM_ZONE (1<<0)	/* Run shrink_inactive_list on the zone */
3259 3260 3261
#define RECLAIM_WRITE (1<<1)	/* Writeout pages during reclaim */
#define RECLAIM_SWAP (1<<2)	/* Swap pages out during reclaim */

3262 3263 3264 3265 3266 3267 3268
/*
 * 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

3269 3270 3271 3272 3273 3274
/*
 * Percentage of pages in a zone that must be unmapped for zone_reclaim to
 * occur.
 */
int sysctl_min_unmapped_ratio = 1;

3275 3276 3277 3278 3279 3280
/*
 * 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;

3281 3282 3283 3284 3285 3286 3287 3288 3289 3290 3291 3292 3293 3294 3295 3296 3297 3298 3299 3300 3301 3302 3303 3304 3305 3306 3307 3308 3309 3310 3311 3312 3313 3314 3315 3316 3317 3318 3319 3320 3321 3322
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;
}

3323 3324 3325
/*
 * Try to free up some pages from this zone through reclaim.
 */
3326
static int __zone_reclaim(struct zone *zone, gfp_t gfp_mask, unsigned int order)
3327
{
3328
	/* Minimum pages needed in order to stay on node */
3329
	const unsigned long nr_pages = 1 << order;
3330 3331
	struct task_struct *p = current;
	struct reclaim_state reclaim_state;
3332 3333
	struct scan_control sc = {
		.may_writepage = !!(zone_reclaim_mode & RECLAIM_WRITE),
3334
		.may_unmap = !!(zone_reclaim_mode & RECLAIM_SWAP),
3335
		.may_swap = 1,
3336 3337
		.nr_to_reclaim = max_t(unsigned long, nr_pages,
				       SWAP_CLUSTER_MAX),
3338
		.gfp_mask = gfp_mask,
3339
		.order = order,
3340
		.priority = ZONE_RECLAIM_PRIORITY,
3341
	};
3342 3343 3344
	struct shrink_control shrink = {
		.gfp_mask = sc.gfp_mask,
	};
3345
	unsigned long nr_slab_pages0, nr_slab_pages1;
3346 3347

	cond_resched();
3348 3349 3350 3351 3352 3353
	/*
	 * 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;
3354
	lockdep_set_current_reclaim_state(gfp_mask);
3355 3356
	reclaim_state.reclaimed_slab = 0;
	p->reclaim_state = &reclaim_state;
3357

3358
	if (zone_pagecache_reclaimable(zone) > zone->min_unmapped_pages) {
3359 3360 3361 3362 3363
		/*
		 * Free memory by calling shrink zone with increasing
		 * priorities until we have enough memory freed.
		 */
		do {
3364 3365
			shrink_zone(zone, &sc);
		} while (sc.nr_reclaimed < nr_pages && --sc.priority >= 0);
3366
	}
3367

3368 3369
	nr_slab_pages0 = zone_page_state(zone, NR_SLAB_RECLAIMABLE);
	if (nr_slab_pages0 > zone->min_slab_pages) {
3370
		/*
3371
		 * shrink_slab() does not currently allow us to determine how
3372 3373 3374 3375
		 * 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.
3376
		 *
3377 3378
		 * Note that shrink_slab will free memory on all zones and may
		 * take a long time.
3379
		 */
3380 3381 3382 3383
		for (;;) {
			unsigned long lru_pages = zone_reclaimable_pages(zone);

			/* No reclaimable slab or very low memory pressure */
3384
			if (!shrink_slab(&shrink, sc.nr_scanned, lru_pages))
3385 3386 3387 3388 3389 3390 3391 3392
				break;

			/* Freed enough memory */
			nr_slab_pages1 = zone_page_state(zone,
							NR_SLAB_RECLAIMABLE);
			if (nr_slab_pages1 + nr_pages <= nr_slab_pages0)
				break;
		}
3393 3394 3395 3396 3397

		/*
		 * Update nr_reclaimed by the number of slab pages we
		 * reclaimed from this zone.
		 */
3398 3399 3400
		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;
3401 3402
	}

3403
	p->reclaim_state = NULL;
3404
	current->flags &= ~(PF_MEMALLOC | PF_SWAPWRITE);
3405
	lockdep_clear_current_reclaim_state();
3406
	return sc.nr_reclaimed >= nr_pages;
3407
}
3408 3409 3410 3411

int zone_reclaim(struct zone *zone, gfp_t gfp_mask, unsigned int order)
{
	int node_id;
3412
	int ret;
3413 3414

	/*
3415 3416
	 * Zone reclaim reclaims unmapped file backed pages and
	 * slab pages if we are over the defined limits.
3417
	 *
3418 3419 3420 3421 3422
	 * 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.
3423
	 */
3424 3425
	if (zone_pagecache_reclaimable(zone) <= zone->min_unmapped_pages &&
	    zone_page_state(zone, NR_SLAB_RECLAIMABLE) <= zone->min_slab_pages)
3426
		return ZONE_RECLAIM_FULL;
3427

3428
	if (zone->all_unreclaimable)
3429
		return ZONE_RECLAIM_FULL;
3430

3431
	/*
3432
	 * Do not scan if the allocation should not be delayed.
3433
	 */
3434
	if (!(gfp_mask & __GFP_WAIT) || (current->flags & PF_MEMALLOC))
3435
		return ZONE_RECLAIM_NOSCAN;
3436 3437 3438 3439 3440 3441 3442

	/*
	 * 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.
	 */
3443
	node_id = zone_to_nid(zone);
3444
	if (node_state(node_id, N_CPU) && node_id != numa_node_id())
3445
		return ZONE_RECLAIM_NOSCAN;
3446 3447

	if (zone_test_and_set_flag(zone, ZONE_RECLAIM_LOCKED))
3448 3449
		return ZONE_RECLAIM_NOSCAN;

3450 3451 3452
	ret = __zone_reclaim(zone, gfp_mask, order);
	zone_clear_flag(zone, ZONE_RECLAIM_LOCKED);

3453 3454 3455
	if (!ret)
		count_vm_event(PGSCAN_ZONE_RECLAIM_FAILED);

3456
	return ret;
3457
}
3458
#endif
3459 3460 3461 3462 3463 3464

/*
 * 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
3465
 * lists vs unevictable list.
3466 3467
 *
 * Reasons page might not be evictable:
3468
 * (1) page's mapping marked unevictable
3469
 * (2) page is part of an mlocked VMA
3470
 *
3471
 */
3472
int page_evictable(struct page *page)
3473
{
3474
	return !mapping_unevictable(page_mapping(page)) && !PageMlocked(page);
3475
}
3476

3477
#ifdef CONFIG_SHMEM
3478
/**
3479 3480 3481
 * 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
3482
 *
3483
 * Checks pages for evictability and moves them to the appropriate lru list.
3484 3485
 *
 * This function is only used for SysV IPC SHM_UNLOCK.
3486
 */
3487
void check_move_unevictable_pages(struct page **pages, int nr_pages)
3488
{
3489
	struct lruvec *lruvec;
3490 3491 3492 3493
	struct zone *zone = NULL;
	int pgscanned = 0;
	int pgrescued = 0;
	int i;
3494

3495 3496 3497
	for (i = 0; i < nr_pages; i++) {
		struct page *page = pages[i];
		struct zone *pagezone;
3498

3499 3500 3501 3502 3503 3504 3505 3506
		pgscanned++;
		pagezone = page_zone(page);
		if (pagezone != zone) {
			if (zone)
				spin_unlock_irq(&zone->lru_lock);
			zone = pagezone;
			spin_lock_irq(&zone->lru_lock);
		}
3507
		lruvec = mem_cgroup_page_lruvec(page, zone);
3508

3509 3510
		if (!PageLRU(page) || !PageUnevictable(page))
			continue;
3511

3512
		if (page_evictable(page)) {
3513 3514 3515 3516
			enum lru_list lru = page_lru_base_type(page);

			VM_BUG_ON(PageActive(page));
			ClearPageUnevictable(page);
3517 3518
			del_page_from_lru_list(page, lruvec, LRU_UNEVICTABLE);
			add_page_to_lru_list(page, lruvec, lru);
3519
			pgrescued++;
3520
		}
3521
	}
3522

3523 3524 3525 3526
	if (zone) {
		__count_vm_events(UNEVICTABLE_PGRESCUED, pgrescued);
		__count_vm_events(UNEVICTABLE_PGSCANNED, pgscanned);
		spin_unlock_irq(&zone->lru_lock);
3527 3528
	}
}
3529
#endif /* CONFIG_SHMEM */
3530

3531
static void warn_scan_unevictable_pages(void)
3532
{
3533
	printk_once(KERN_WARNING
3534
		    "%s: The scan_unevictable_pages sysctl/node-interface has been "
3535
		    "disabled for lack of a legitimate use case.  If you have "
3536 3537
		    "one, please send an email to linux-mm@kvack.org.\n",
		    current->comm);
3538 3539 3540 3541 3542 3543 3544 3545 3546
}

/*
 * 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,
3547
			   void __user *buffer,
3548 3549
			   size_t *length, loff_t *ppos)
{
3550
	warn_scan_unevictable_pages();
3551
	proc_doulongvec_minmax(table, write, buffer, length, ppos);
3552 3553 3554 3555
	scan_unevictable_pages = 0;
	return 0;
}

3556
#ifdef CONFIG_NUMA
3557 3558 3559 3560 3561
/*
 * per node 'scan_unevictable_pages' attribute.  On demand re-scan of
 * a specified node's per zone unevictable lists for evictable pages.
 */

3562 3563
static ssize_t read_scan_unevictable_node(struct device *dev,
					  struct device_attribute *attr,
3564 3565
					  char *buf)
{
3566
	warn_scan_unevictable_pages();
3567 3568 3569
	return sprintf(buf, "0\n");	/* always zero; should fit... */
}

3570 3571
static ssize_t write_scan_unevictable_node(struct device *dev,
					   struct device_attribute *attr,
3572 3573
					const char *buf, size_t count)
{
3574
	warn_scan_unevictable_pages();
3575 3576 3577 3578
	return 1;
}


3579
static DEVICE_ATTR(scan_unevictable_pages, S_IRUGO | S_IWUSR,
3580 3581 3582 3583 3584
			read_scan_unevictable_node,
			write_scan_unevictable_node);

int scan_unevictable_register_node(struct node *node)
{
3585
	return device_create_file(&node->dev, &dev_attr_scan_unevictable_pages);
3586 3587 3588 3589
}

void scan_unevictable_unregister_node(struct node *node)
{
3590
	device_remove_file(&node->dev, &dev_attr_scan_unevictable_pages);
3591
}
3592
#endif