vmscan.c 110.5 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
/*
 *  linux/mm/vmscan.c
 *
 *  Copyright (C) 1991, 1992, 1993, 1994  Linus Torvalds
 *
 *  Swap reorganised 29.12.95, Stephen Tweedie.
 *  kswapd added: 7.1.96  sct
 *  Removed kswapd_ctl limits, and swap out as many pages as needed
 *  to bring the system back to freepages.high: 2.4.97, Rik van Riel.
 *  Zone aware kswapd started 02/00, Kanoj Sarcar (kanoj@sgi.com).
 *  Multiqueue VM started 5.8.00, Rik van Riel.
 */

#include <linux/mm.h>
#include <linux/module.h>
16
#include <linux/gfp.h>
L
Linus Torvalds 已提交
17 18 19 20 21
#include <linux/kernel_stat.h>
#include <linux/swap.h>
#include <linux/pagemap.h>
#include <linux/init.h>
#include <linux/highmem.h>
22
#include <linux/vmpressure.h>
23
#include <linux/vmstat.h>
L
Linus Torvalds 已提交
24 25 26 27 28 29 30 31 32 33 34
#include <linux/file.h>
#include <linux/writeback.h>
#include <linux/blkdev.h>
#include <linux/buffer_head.h>	/* for try_to_release_page(),
					buffer_heads_over_limit */
#include <linux/mm_inline.h>
#include <linux/backing-dev.h>
#include <linux/rmap.h>
#include <linux/topology.h>
#include <linux/cpu.h>
#include <linux/cpuset.h>
35
#include <linux/compaction.h>
L
Linus Torvalds 已提交
36 37
#include <linux/notifier.h>
#include <linux/rwsem.h>
38
#include <linux/delay.h>
39
#include <linux/kthread.h>
40
#include <linux/freezer.h>
41
#include <linux/memcontrol.h>
42
#include <linux/delayacct.h>
43
#include <linux/sysctl.h>
44
#include <linux/oom.h>
45
#include <linux/prefetch.h>
L
Linus Torvalds 已提交
46 47 48 49 50

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

#include <linux/swapops.h>
51
#include <linux/balloon_compaction.h>
L
Linus Torvalds 已提交
52

53 54
#include "internal.h"

55 56 57
#define CREATE_TRACE_POINTS
#include <trace/events/vmscan.h>

L
Linus Torvalds 已提交
58 59 60 61
struct scan_control {
	/* Incremented by the number of inactive pages that were scanned */
	unsigned long nr_scanned;

62 63 64
	/* Number of pages freed so far during a call to shrink_zones() */
	unsigned long nr_reclaimed;

65 66 67
	/* How many pages shrink_list() should reclaim */
	unsigned long nr_to_reclaim;

68 69
	unsigned long hibernation_mode;

L
Linus Torvalds 已提交
70
	/* This context's GFP mask */
A
Al Viro 已提交
71
	gfp_t gfp_mask;
L
Linus Torvalds 已提交
72 73 74

	int may_writepage;

75 76
	/* Can mapped pages be reclaimed? */
	int may_unmap;
77

78 79 80
	/* Can pages be swapped as part of reclaim? */
	int may_swap;

A
Andy Whitcroft 已提交
81
	int order;
82

83 84 85
	/* Scan (total_size >> priority) pages at once */
	int priority;

86 87 88
	/* anon vs. file LRUs scanning "ratio" */
	int swappiness;

89 90 91 92 93
	/*
	 * 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;
94

95 96 97 98 99
	/*
	 * Nodemask of nodes allowed by the caller. If NULL, all nodes
	 * are scanned.
	 */
	nodemask_t	*nodemask;
L
Linus Torvalds 已提交
100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135
};

#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;
136
unsigned long vm_total_pages;	/* The total number of pages which the VM controls */
L
Linus Torvalds 已提交
137 138 139 140

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

A
Andrew Morton 已提交
141
#ifdef CONFIG_MEMCG
142 143
static bool global_reclaim(struct scan_control *sc)
{
144
	return !sc->target_mem_cgroup;
145
}
146
#else
147 148 149 150
static bool global_reclaim(struct scan_control *sc)
{
	return true;
}
151 152
#endif

153
static unsigned long zone_reclaimable_pages(struct zone *zone)
154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171
{
	int nr;

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

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

	return nr;
}

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

172
static unsigned long get_lru_size(struct lruvec *lruvec, enum lru_list lru)
173
{
174
	if (!mem_cgroup_disabled())
175
		return mem_cgroup_get_lru_size(lruvec, lru);
176

177
	return zone_page_state(lruvec_zone(lruvec), NR_LRU_BASE + lru);
178 179
}

L
Linus Torvalds 已提交
180
/*
G
Glauber Costa 已提交
181
 * Add a shrinker callback to be called from the vm.
L
Linus Torvalds 已提交
182
 */
G
Glauber Costa 已提交
183
int register_shrinker(struct shrinker *shrinker)
L
Linus Torvalds 已提交
184
{
G
Glauber Costa 已提交
185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201
	size_t size = sizeof(*shrinker->nr_deferred);

	/*
	 * If we only have one possible node in the system anyway, save
	 * ourselves the trouble and disable NUMA aware behavior. This way we
	 * will save memory and some small loop time later.
	 */
	if (nr_node_ids == 1)
		shrinker->flags &= ~SHRINKER_NUMA_AWARE;

	if (shrinker->flags & SHRINKER_NUMA_AWARE)
		size *= nr_node_ids;

	shrinker->nr_deferred = kzalloc(size, GFP_KERNEL);
	if (!shrinker->nr_deferred)
		return -ENOMEM;

202 203 204
	down_write(&shrinker_rwsem);
	list_add_tail(&shrinker->list, &shrinker_list);
	up_write(&shrinker_rwsem);
G
Glauber Costa 已提交
205
	return 0;
L
Linus Torvalds 已提交
206
}
207
EXPORT_SYMBOL(register_shrinker);
L
Linus Torvalds 已提交
208 209 210 211

/*
 * Remove one
 */
212
void unregister_shrinker(struct shrinker *shrinker)
L
Linus Torvalds 已提交
213 214 215 216
{
	down_write(&shrinker_rwsem);
	list_del(&shrinker->list);
	up_write(&shrinker_rwsem);
217
	kfree(shrinker->nr_deferred);
L
Linus Torvalds 已提交
218
}
219
EXPORT_SYMBOL(unregister_shrinker);
L
Linus Torvalds 已提交
220 221

#define SHRINK_BATCH 128
G
Glauber Costa 已提交
222 223 224 225 226 227 228 229

static unsigned long
shrink_slab_node(struct shrink_control *shrinkctl, struct shrinker *shrinker,
		 unsigned long nr_pages_scanned, unsigned long lru_pages)
{
	unsigned long freed = 0;
	unsigned long long delta;
	long total_scan;
230
	long freeable;
G
Glauber Costa 已提交
231 232 233 234 235 236
	long nr;
	long new_nr;
	int nid = shrinkctl->nid;
	long batch_size = shrinker->batch ? shrinker->batch
					  : SHRINK_BATCH;

237 238
	freeable = shrinker->count_objects(shrinker, shrinkctl);
	if (freeable == 0)
G
Glauber Costa 已提交
239 240 241 242 243 244 245 246 247 248 249
		return 0;

	/*
	 * 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.
	 */
	nr = atomic_long_xchg(&shrinker->nr_deferred[nid], 0);

	total_scan = nr;
	delta = (4 * nr_pages_scanned) / shrinker->seeks;
250
	delta *= freeable;
G
Glauber Costa 已提交
251 252 253 254 255
	do_div(delta, lru_pages + 1);
	total_scan += delta;
	if (total_scan < 0) {
		printk(KERN_ERR
		"shrink_slab: %pF negative objects to delete nr=%ld\n",
D
Dave Chinner 已提交
256
		       shrinker->scan_objects, total_scan);
257
		total_scan = freeable;
G
Glauber Costa 已提交
258 259 260 261 262 263 264 265
	}

	/*
	 * 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 >>>
266
	 * freeable. This is bad for sustaining a working set in
G
Glauber Costa 已提交
267 268 269 270 271
	 * memory.
	 *
	 * Hence only allow the shrinker to scan the entire cache when
	 * a large delta change is calculated directly.
	 */
272 273
	if (delta < freeable / 4)
		total_scan = min(total_scan, freeable / 2);
G
Glauber Costa 已提交
274 275 276 277 278 279

	/*
	 * Avoid risking looping forever due to too large nr value:
	 * never try to free more than twice the estimate number of
	 * freeable entries.
	 */
280 281
	if (total_scan > freeable * 2)
		total_scan = freeable * 2;
G
Glauber Costa 已提交
282 283 284

	trace_mm_shrink_slab_start(shrinker, shrinkctl, nr,
				nr_pages_scanned, lru_pages,
285
				freeable, delta, total_scan);
G
Glauber Costa 已提交
286

287 288 289 290 291 292 293 294 295 296 297
	/*
	 * Normally, we should not scan less than batch_size objects in one
	 * pass to avoid too frequent shrinker calls, but if the slab has less
	 * than batch_size objects in total and we are really tight on memory,
	 * we will try to reclaim all available objects, otherwise we can end
	 * up failing allocations although there are plenty of reclaimable
	 * objects spread over several slabs with usage less than the
	 * batch_size.
	 *
	 * We detect the "tight on memory" situations by looking at the total
	 * number of objects we want to scan (total_scan). If it is greater
298
	 * than the total number of objects on slab (freeable), we must be
299 300 301 302
	 * scanning at high prio and therefore should try to reclaim as much as
	 * possible.
	 */
	while (total_scan >= batch_size ||
303
	       total_scan >= freeable) {
D
Dave Chinner 已提交
304
		unsigned long ret;
305
		unsigned long nr_to_scan = min(batch_size, total_scan);
G
Glauber Costa 已提交
306

307
		shrinkctl->nr_to_scan = nr_to_scan;
D
Dave Chinner 已提交
308 309 310 311
		ret = shrinker->scan_objects(shrinker, shrinkctl);
		if (ret == SHRINK_STOP)
			break;
		freed += ret;
G
Glauber Costa 已提交
312

313 314
		count_vm_events(SLABS_SCANNED, nr_to_scan);
		total_scan -= nr_to_scan;
G
Glauber Costa 已提交
315 316 317 318 319 320 321 322 323 324 325 326 327 328 329

		cond_resched();
	}

	/*
	 * 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.
	 */
	if (total_scan > 0)
		new_nr = atomic_long_add_return(total_scan,
						&shrinker->nr_deferred[nid]);
	else
		new_nr = atomic_long_read(&shrinker->nr_deferred[nid]);

330
	trace_mm_shrink_slab_end(shrinker, nid, freed, nr, new_nr, total_scan);
G
Glauber Costa 已提交
331
	return freed;
332 333
}

L
Linus Torvalds 已提交
334 335 336 337 338 339 340 341
/*
 * Call the shrink functions to age shrinkable caches
 *
 * Here we assume it costs one seek to replace a lru page and that it also
 * takes a seek to recreate a cache object.  With this in mind we age equal
 * percentages of the lru and ageable caches.  This should balance the seeks
 * generated by these structures.
 *
S
Simon Arlott 已提交
342
 * If the vm encountered mapped pages on the LRU it increase the pressure on
L
Linus Torvalds 已提交
343 344 345 346 347 348 349
 * 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.
350 351
 *
 * Returns the number of slab objects which we shrunk.
L
Linus Torvalds 已提交
352
 */
D
Dave Chinner 已提交
353
unsigned long shrink_slab(struct shrink_control *shrinkctl,
354
			  unsigned long nr_pages_scanned,
355
			  unsigned long lru_pages)
L
Linus Torvalds 已提交
356 357
{
	struct shrinker *shrinker;
D
Dave Chinner 已提交
358
	unsigned long freed = 0;
L
Linus Torvalds 已提交
359

360 361
	if (nr_pages_scanned == 0)
		nr_pages_scanned = SWAP_CLUSTER_MAX;
L
Linus Torvalds 已提交
362

363
	if (!down_read_trylock(&shrinker_rwsem)) {
D
Dave Chinner 已提交
364 365 366 367 368 369 370
		/*
		 * If we would return 0, our callers would understand that we
		 * have nothing else to shrink and give up trying. By returning
		 * 1 we keep it going and assume we'll be able to shrink next
		 * time.
		 */
		freed = 1;
371 372
		goto out;
	}
L
Linus Torvalds 已提交
373 374

	list_for_each_entry(shrinker, &shrinker_list, list) {
375 376
		if (!(shrinker->flags & SHRINKER_NUMA_AWARE)) {
			shrinkctl->nid = 0;
G
Glauber Costa 已提交
377
			freed += shrink_slab_node(shrinkctl, shrinker,
378 379 380 381 382 383 384 385
					nr_pages_scanned, lru_pages);
			continue;
		}

		for_each_node_mask(shrinkctl->nid, shrinkctl->nodes_to_scan) {
			if (node_online(shrinkctl->nid))
				freed += shrink_slab_node(shrinkctl, shrinker,
						nr_pages_scanned, lru_pages);
L
Linus Torvalds 已提交
386 387 388 389

		}
	}
	up_read(&shrinker_rwsem);
390 391
out:
	cond_resched();
D
Dave Chinner 已提交
392
	return freed;
L
Linus Torvalds 已提交
393 394 395 396
}

static inline int is_page_cache_freeable(struct page *page)
{
397 398 399 400 401
	/*
	 * 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.
	 */
402
	return page_count(page) - page_has_private(page) == 2;
L
Linus Torvalds 已提交
403 404
}

405 406
static int may_write_to_queue(struct backing_dev_info *bdi,
			      struct scan_control *sc)
L
Linus Torvalds 已提交
407
{
408
	if (current->flags & PF_SWAPWRITE)
L
Linus Torvalds 已提交
409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431
		return 1;
	if (!bdi_write_congested(bdi))
		return 1;
	if (bdi == current->backing_dev_info)
		return 1;
	return 0;
}

/*
 * We detected a synchronous write error writing a page out.  Probably
 * -ENOSPC.  We need to propagate that into the address_space for a subsequent
 * fsync(), msync() or close().
 *
 * The tricky part is that after writepage we cannot touch the mapping: nothing
 * prevents it from being freed up.  But we have a ref on the page and once
 * that page is locked, the mapping is pinned.
 *
 * We're allowed to run sleeping lock_page() here because we know the caller has
 * __GFP_FS.
 */
static void handle_write_error(struct address_space *mapping,
				struct page *page, int error)
{
J
Jens Axboe 已提交
432
	lock_page(page);
433 434
	if (page_mapping(page) == mapping)
		mapping_set_error(mapping, error);
L
Linus Torvalds 已提交
435 436 437
	unlock_page(page);
}

438 439 440 441 442 443 444 445 446 447 448 449
/* possible outcome of pageout() */
typedef enum {
	/* failed to write page out, page is locked */
	PAGE_KEEP,
	/* move page to the active list, page is locked */
	PAGE_ACTIVATE,
	/* page has been sent to the disk successfully, page is unlocked */
	PAGE_SUCCESS,
	/* page is clean and locked */
	PAGE_CLEAN,
} pageout_t;

L
Linus Torvalds 已提交
450
/*
A
Andrew Morton 已提交
451 452
 * pageout is called by shrink_page_list() for each dirty page.
 * Calls ->writepage().
L
Linus Torvalds 已提交
453
 */
454
static pageout_t pageout(struct page *page, struct address_space *mapping,
455
			 struct scan_control *sc)
L
Linus Torvalds 已提交
456 457 458 459 460 461 462 463
{
	/*
	 * 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.
	 *
464
	 * If this process is currently in __generic_file_aio_write() against
L
Linus Torvalds 已提交
465 466 467 468 469 470 471 472 473 474 475 476 477 478 479
	 * 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.
		 */
480
		if (page_has_private(page)) {
L
Linus Torvalds 已提交
481 482
			if (try_to_free_buffers(page)) {
				ClearPageDirty(page);
483
				printk("%s: orphaned page\n", __func__);
L
Linus Torvalds 已提交
484 485 486 487 488 489 490
				return PAGE_CLEAN;
			}
		}
		return PAGE_KEEP;
	}
	if (mapping->a_ops->writepage == NULL)
		return PAGE_ACTIVATE;
491
	if (!may_write_to_queue(mapping->backing_dev_info, sc))
L
Linus Torvalds 已提交
492 493 494 495 496 497 498
		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,
499 500
			.range_start = 0,
			.range_end = LLONG_MAX,
L
Linus Torvalds 已提交
501 502 503 504 505 506 507
			.for_reclaim = 1,
		};

		SetPageReclaim(page);
		res = mapping->a_ops->writepage(page, &wbc);
		if (res < 0)
			handle_write_error(mapping, page, res);
508
		if (res == AOP_WRITEPAGE_ACTIVATE) {
L
Linus Torvalds 已提交
509 510 511
			ClearPageReclaim(page);
			return PAGE_ACTIVATE;
		}
512

L
Linus Torvalds 已提交
513 514 515 516
		if (!PageWriteback(page)) {
			/* synchronous write or broken a_ops? */
			ClearPageReclaim(page);
		}
M
Mel Gorman 已提交
517
		trace_mm_vmscan_writepage(page, trace_reclaim_flags(page));
518
		inc_zone_page_state(page, NR_VMSCAN_WRITE);
L
Linus Torvalds 已提交
519 520 521 522 523 524
		return PAGE_SUCCESS;
	}

	return PAGE_CLEAN;
}

525
/*
N
Nick Piggin 已提交
526 527
 * Same as remove_mapping, but if the page is removed from the mapping, it
 * gets returned with a refcount of 0.
528
 */
529 530
static int __remove_mapping(struct address_space *mapping, struct page *page,
			    bool reclaimed)
531
{
532 533
	BUG_ON(!PageLocked(page));
	BUG_ON(mapping != page_mapping(page));
534

N
Nick Piggin 已提交
535
	spin_lock_irq(&mapping->tree_lock);
536
	/*
N
Nick Piggin 已提交
537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559
	 * 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.
560
	 */
N
Nick Piggin 已提交
561
	if (!page_freeze_refs(page, 2))
562
		goto cannot_free;
N
Nick Piggin 已提交
563 564 565
	/* note: atomic_cmpxchg in page_freeze_refs provides the smp_rmb */
	if (unlikely(PageDirty(page))) {
		page_unfreeze_refs(page, 2);
566
		goto cannot_free;
N
Nick Piggin 已提交
567
	}
568 569 570 571

	if (PageSwapCache(page)) {
		swp_entry_t swap = { .val = page_private(page) };
		__delete_from_swap_cache(page);
N
Nick Piggin 已提交
572
		spin_unlock_irq(&mapping->tree_lock);
573
		swapcache_free(swap, page);
N
Nick Piggin 已提交
574
	} else {
575
		void (*freepage)(struct page *);
576
		void *shadow = NULL;
577 578

		freepage = mapping->a_ops->freepage;
579 580 581 582 583 584 585 586 587 588 589 590 591 592
		/*
		 * Remember a shadow entry for reclaimed file cache in
		 * order to detect refaults, thus thrashing, later on.
		 *
		 * But don't store shadows in an address space that is
		 * already exiting.  This is not just an optizimation,
		 * inode reclaim needs to empty out the radix tree or
		 * the nodes are lost.  Don't plant shadows behind its
		 * back.
		 */
		if (reclaimed && page_is_file_cache(page) &&
		    !mapping_exiting(mapping))
			shadow = workingset_eviction(mapping, page);
		__delete_from_page_cache(page, shadow);
N
Nick Piggin 已提交
593
		spin_unlock_irq(&mapping->tree_lock);
594
		mem_cgroup_uncharge_cache_page(page);
595 596 597

		if (freepage != NULL)
			freepage(page);
598 599 600 601 602
	}

	return 1;

cannot_free:
N
Nick Piggin 已提交
603
	spin_unlock_irq(&mapping->tree_lock);
604 605 606
	return 0;
}

N
Nick Piggin 已提交
607 608 609 610 611 612 613 614
/*
 * 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)
{
615
	if (__remove_mapping(mapping, page, false)) {
N
Nick Piggin 已提交
616 617 618 619 620 621 622 623 624 625 626
		/*
		 * Unfreezing the refcount with 1 rather than 2 effectively
		 * drops the pagecache ref for us without requiring another
		 * atomic operation.
		 */
		page_unfreeze_refs(page, 1);
		return 1;
	}
	return 0;
}

L
Lee Schermerhorn 已提交
627 628 629 630 631 632 633 634 635 636 637
/**
 * 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)
{
638
	bool is_unevictable;
639
	int was_unevictable = PageUnevictable(page);
L
Lee Schermerhorn 已提交
640

641
	VM_BUG_ON_PAGE(PageLRU(page), page);
L
Lee Schermerhorn 已提交
642 643 644 645

redo:
	ClearPageUnevictable(page);

646
	if (page_evictable(page)) {
L
Lee Schermerhorn 已提交
647 648 649 650 651 652
		/*
		 * 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.
		 */
653
		is_unevictable = false;
654
		lru_cache_add(page);
L
Lee Schermerhorn 已提交
655 656 657 658 659
	} else {
		/*
		 * Put unevictable pages directly on zone's unevictable
		 * list.
		 */
660
		is_unevictable = true;
L
Lee Schermerhorn 已提交
661
		add_page_to_unevictable_list(page);
662
		/*
663 664 665
		 * 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
666
		 * isolation/check_move_unevictable_pages,
667
		 * we see PG_mlocked/AS_UNEVICTABLE cleared below and move
668 669
		 * the page back to the evictable list.
		 *
670
		 * The other side is TestClearPageMlocked() or shmem_lock().
671 672
		 */
		smp_mb();
L
Lee Schermerhorn 已提交
673 674 675 676 677 678 679
	}

	/*
	 * 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.
	 */
680
	if (is_unevictable && page_evictable(page)) {
L
Lee Schermerhorn 已提交
681 682 683 684 685 686 687 688 689 690
		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.
		 */
	}

691
	if (was_unevictable && !is_unevictable)
692
		count_vm_event(UNEVICTABLE_PGRESCUED);
693
	else if (!was_unevictable && is_unevictable)
694 695
		count_vm_event(UNEVICTABLE_PGCULLED);

L
Lee Schermerhorn 已提交
696 697 698
	put_page(page);		/* drop ref from isolate */
}

699 700 701
enum page_references {
	PAGEREF_RECLAIM,
	PAGEREF_RECLAIM_CLEAN,
702
	PAGEREF_KEEP,
703 704 705 706 707 708
	PAGEREF_ACTIVATE,
};

static enum page_references page_check_references(struct page *page,
						  struct scan_control *sc)
{
709
	int referenced_ptes, referenced_page;
710 711
	unsigned long vm_flags;

712 713
	referenced_ptes = page_referenced(page, 1, sc->target_mem_cgroup,
					  &vm_flags);
714
	referenced_page = TestClearPageReferenced(page);
715 716 717 718 719 720 721 722

	/*
	 * 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;

723
	if (referenced_ptes) {
724
		if (PageSwapBacked(page))
725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741
			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);

742
		if (referenced_page || referenced_ptes > 1)
743 744
			return PAGEREF_ACTIVATE;

745 746 747 748 749 750
		/*
		 * Activate file-backed executable pages after first usage.
		 */
		if (vm_flags & VM_EXEC)
			return PAGEREF_ACTIVATE;

751 752
		return PAGEREF_KEEP;
	}
753 754

	/* Reclaim if clean, defer dirty pages to writeback */
755
	if (referenced_page && !PageSwapBacked(page))
756 757 758
		return PAGEREF_RECLAIM_CLEAN;

	return PAGEREF_RECLAIM;
759 760
}

761 762 763 764
/* Check if a page is dirty or under writeback */
static void page_check_dirty_writeback(struct page *page,
				       bool *dirty, bool *writeback)
{
765 766
	struct address_space *mapping;

767 768 769 770 771 772 773 774 775 776 777 778 779
	/*
	 * Anonymous pages are not handled by flushers and must be written
	 * from reclaim context. Do not stall reclaim based on them
	 */
	if (!page_is_file_cache(page)) {
		*dirty = false;
		*writeback = false;
		return;
	}

	/* By default assume that the page flags are accurate */
	*dirty = PageDirty(page);
	*writeback = PageWriteback(page);
780 781 782 783 784 785 786 787

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

	mapping = page_mapping(page);
	if (mapping && mapping->a_ops->is_dirty_writeback)
		mapping->a_ops->is_dirty_writeback(page, dirty, writeback);
788 789
}

L
Linus Torvalds 已提交
790
/*
A
Andrew Morton 已提交
791
 * shrink_page_list() returns the number of reclaimed pages
L
Linus Torvalds 已提交
792
 */
A
Andrew Morton 已提交
793
static unsigned long shrink_page_list(struct list_head *page_list,
794
				      struct zone *zone,
795
				      struct scan_control *sc,
796
				      enum ttu_flags ttu_flags,
797
				      unsigned long *ret_nr_dirty,
798
				      unsigned long *ret_nr_unqueued_dirty,
799
				      unsigned long *ret_nr_congested,
800
				      unsigned long *ret_nr_writeback,
801
				      unsigned long *ret_nr_immediate,
802
				      bool force_reclaim)
L
Linus Torvalds 已提交
803 804
{
	LIST_HEAD(ret_pages);
805
	LIST_HEAD(free_pages);
L
Linus Torvalds 已提交
806
	int pgactivate = 0;
807
	unsigned long nr_unqueued_dirty = 0;
808 809
	unsigned long nr_dirty = 0;
	unsigned long nr_congested = 0;
810
	unsigned long nr_reclaimed = 0;
811
	unsigned long nr_writeback = 0;
812
	unsigned long nr_immediate = 0;
L
Linus Torvalds 已提交
813 814 815

	cond_resched();

816
	mem_cgroup_uncharge_start();
L
Linus Torvalds 已提交
817 818 819 820
	while (!list_empty(page_list)) {
		struct address_space *mapping;
		struct page *page;
		int may_enter_fs;
821
		enum page_references references = PAGEREF_RECLAIM_CLEAN;
822
		bool dirty, writeback;
L
Linus Torvalds 已提交
823 824 825 826 827 828

		cond_resched();

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

N
Nick Piggin 已提交
829
		if (!trylock_page(page))
L
Linus Torvalds 已提交
830 831
			goto keep;

832 833
		VM_BUG_ON_PAGE(PageActive(page), page);
		VM_BUG_ON_PAGE(page_zone(page) != zone, page);
L
Linus Torvalds 已提交
834 835

		sc->nr_scanned++;
836

837
		if (unlikely(!page_evictable(page)))
N
Nick Piggin 已提交
838
			goto cull_mlocked;
L
Lee Schermerhorn 已提交
839

840
		if (!sc->may_unmap && page_mapped(page))
841 842
			goto keep_locked;

L
Linus Torvalds 已提交
843 844 845 846
		/* Double the slab pressure for mapped and swapcache pages */
		if (page_mapped(page) || PageSwapCache(page))
			sc->nr_scanned++;

847 848 849
		may_enter_fs = (sc->gfp_mask & __GFP_FS) ||
			(PageSwapCache(page) && (sc->gfp_mask & __GFP_IO));

850 851 852 853 854 855 856 857 858 859 860 861 862
		/*
		 * The number of dirty pages determines if a zone is marked
		 * reclaim_congested which affects wait_iff_congested. kswapd
		 * will stall and start writing pages if the tail of the LRU
		 * is all dirty unqueued pages.
		 */
		page_check_dirty_writeback(page, &dirty, &writeback);
		if (dirty || writeback)
			nr_dirty++;

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

863 864 865 866 867 868
		/*
		 * Treat this page as congested if the underlying BDI is or if
		 * pages are cycling through the LRU so quickly that the
		 * pages marked for immediate reclaim are making it to the
		 * end of the LRU a second time.
		 */
869
		mapping = page_mapping(page);
870 871
		if ((mapping && bdi_write_congested(mapping->backing_dev_info)) ||
		    (writeback && PageReclaim(page)))
872 873
			nr_congested++;

874 875 876 877 878 879 880 881 882 883 884
		/*
		 * If a page at the tail of the LRU is under writeback, there
		 * are three cases to consider.
		 *
		 * 1) If reclaim is encountering an excessive number of pages
		 *    under writeback and this page is both under writeback and
		 *    PageReclaim then it indicates that pages are being queued
		 *    for IO but are being recycled through the LRU before the
		 *    IO can complete. Waiting on the page itself risks an
		 *    indefinite stall if it is impossible to writeback the
		 *    page due to IO error or disconnected storage so instead
885 886
		 *    note that the LRU is being scanned too quickly and the
		 *    caller can stall after page list has been processed.
887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910
		 *
		 * 2) Global reclaim encounters a page, memcg encounters a
		 *    page that is not marked for immediate reclaim or
		 *    the caller does not have __GFP_IO. In this case mark
		 *    the page for immediate reclaim and continue scanning.
		 *
		 *    __GFP_IO is checked  because a loop driver thread might
		 *    enter reclaim, and deadlock if it waits on a page for
		 *    which it is needed to do the write (loop masks off
		 *    __GFP_IO|__GFP_FS for this reason); but more thought
		 *    would probably show more reasons.
		 *
		 *    Don't require __GFP_FS, since we're not going into the
		 *    FS, just waiting on its writeback completion. Worryingly,
		 *    ext4 gfs2 and xfs allocate pages with
		 *    grab_cache_page_write_begin(,,AOP_FLAG_NOFS), so testing
		 *    may_enter_fs here is liable to OOM on them.
		 *
		 * 3) memcg encounters a page that is not already marked
		 *    PageReclaim. memcg does not have any dirty pages
		 *    throttling so we could easily OOM just because too many
		 *    pages are in writeback and there is nothing else to
		 *    reclaim. Wait for the writeback to complete.
		 */
911
		if (PageWriteback(page)) {
912 913 914 915
			/* Case 1 above */
			if (current_is_kswapd() &&
			    PageReclaim(page) &&
			    zone_is_reclaim_writeback(zone)) {
916 917
				nr_immediate++;
				goto keep_locked;
918 919 920

			/* Case 2 above */
			} else if (global_reclaim(sc) ||
921 922 923 924 925 926 927 928 929 930 931 932 933
			    !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);
934
				nr_writeback++;
935

936
				goto keep_locked;
937 938 939 940

			/* Case 3 above */
			} else {
				wait_on_page_writeback(page);
941
			}
942
		}
L
Linus Torvalds 已提交
943

944 945 946
		if (!force_reclaim)
			references = page_check_references(page, sc);

947 948
		switch (references) {
		case PAGEREF_ACTIVATE:
L
Linus Torvalds 已提交
949
			goto activate_locked;
950 951
		case PAGEREF_KEEP:
			goto keep_locked;
952 953 954 955
		case PAGEREF_RECLAIM:
		case PAGEREF_RECLAIM_CLEAN:
			; /* try to reclaim the page below */
		}
L
Linus Torvalds 已提交
956 957 958 959 960

		/*
		 * Anonymous process memory has backing store?
		 * Try to allocate it some swap space here.
		 */
N
Nick Piggin 已提交
961
		if (PageAnon(page) && !PageSwapCache(page)) {
962 963
			if (!(sc->gfp_mask & __GFP_IO))
				goto keep_locked;
964
			if (!add_to_swap(page, page_list))
L
Linus Torvalds 已提交
965
				goto activate_locked;
966
			may_enter_fs = 1;
L
Linus Torvalds 已提交
967

968 969 970
			/* Adding to swap updated mapping */
			mapping = page_mapping(page);
		}
L
Linus Torvalds 已提交
971 972 973 974 975 976

		/*
		 * The page is mapped into the page tables of one or more
		 * processes. Try to unmap it here.
		 */
		if (page_mapped(page) && mapping) {
977
			switch (try_to_unmap(page, ttu_flags)) {
L
Linus Torvalds 已提交
978 979 980 981
			case SWAP_FAIL:
				goto activate_locked;
			case SWAP_AGAIN:
				goto keep_locked;
N
Nick Piggin 已提交
982 983
			case SWAP_MLOCK:
				goto cull_mlocked;
L
Linus Torvalds 已提交
984 985 986 987 988 989
			case SWAP_SUCCESS:
				; /* try to free the page below */
			}
		}

		if (PageDirty(page)) {
990 991
			/*
			 * Only kswapd can writeback filesystem pages to
992 993
			 * avoid risk of stack overflow but only writeback
			 * if many dirty pages have been encountered.
994
			 */
995
			if (page_is_file_cache(page) &&
996
					(!current_is_kswapd() ||
997
					 !zone_is_reclaim_dirty(zone))) {
998 999 1000 1001 1002 1003 1004 1005 1006
				/*
				 * 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);

1007 1008 1009
				goto keep_locked;
			}

1010
			if (references == PAGEREF_RECLAIM_CLEAN)
L
Linus Torvalds 已提交
1011
				goto keep_locked;
1012
			if (!may_enter_fs)
L
Linus Torvalds 已提交
1013
				goto keep_locked;
1014
			if (!sc->may_writepage)
L
Linus Torvalds 已提交
1015 1016 1017
				goto keep_locked;

			/* Page is dirty, try to write it out here */
1018
			switch (pageout(page, mapping, sc)) {
L
Linus Torvalds 已提交
1019 1020 1021 1022 1023
			case PAGE_KEEP:
				goto keep_locked;
			case PAGE_ACTIVATE:
				goto activate_locked;
			case PAGE_SUCCESS:
1024
				if (PageWriteback(page))
1025
					goto keep;
1026
				if (PageDirty(page))
L
Linus Torvalds 已提交
1027
					goto keep;
1028

L
Linus Torvalds 已提交
1029 1030 1031 1032
				/*
				 * A synchronous write - probably a ramdisk.  Go
				 * ahead and try to reclaim the page.
				 */
N
Nick Piggin 已提交
1033
				if (!trylock_page(page))
L
Linus Torvalds 已提交
1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052
					goto keep;
				if (PageDirty(page) || PageWriteback(page))
					goto keep_locked;
				mapping = page_mapping(page);
			case PAGE_CLEAN:
				; /* try to free the page below */
			}
		}

		/*
		 * If the page has buffers, try to free the buffer mappings
		 * associated with this page. If we succeed we try to free
		 * the page as well.
		 *
		 * We do this even if the page is PageDirty().
		 * try_to_release_page() does not perform I/O, but it is
		 * possible for a page to have PageDirty set, but it is actually
		 * clean (all its buffers are clean).  This happens if the
		 * buffers were written out directly, with submit_bh(). ext3
L
Lee Schermerhorn 已提交
1053
		 * will do this, as well as the blockdev mapping.
L
Linus Torvalds 已提交
1054 1055 1056 1057 1058 1059 1060 1061 1062 1063
		 * 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.
		 */
1064
		if (page_has_private(page)) {
L
Linus Torvalds 已提交
1065 1066
			if (!try_to_release_page(page, sc->gfp_mask))
				goto activate_locked;
N
Nick Piggin 已提交
1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082
			if (!mapping && page_count(page) == 1) {
				unlock_page(page);
				if (put_page_testzero(page))
					goto free_it;
				else {
					/*
					 * rare race with speculative reference.
					 * the speculative reference will free
					 * this page shortly, so we may
					 * increment nr_reclaimed here (and
					 * leave it off the LRU).
					 */
					nr_reclaimed++;
					continue;
				}
			}
L
Linus Torvalds 已提交
1083 1084
		}

1085
		if (!mapping || !__remove_mapping(mapping, page, true))
1086
			goto keep_locked;
L
Linus Torvalds 已提交
1087

N
Nick Piggin 已提交
1088 1089 1090 1091 1092 1093 1094 1095
		/*
		 * At this point, we have no other references and there is
		 * no way to pick any more up (removed from LRU, removed
		 * from pagecache). Can use non-atomic bitops now (and
		 * we obviously don't have to worry about waking up a process
		 * waiting on the page lock, because there are no references.
		 */
		__clear_page_locked(page);
N
Nick Piggin 已提交
1096
free_it:
1097
		nr_reclaimed++;
1098 1099 1100 1101 1102 1103

		/*
		 * Is there need to periodically free_page_list? It would
		 * appear not as the counts should be low
		 */
		list_add(&page->lru, &free_pages);
L
Linus Torvalds 已提交
1104 1105
		continue;

N
Nick Piggin 已提交
1106
cull_mlocked:
1107 1108
		if (PageSwapCache(page))
			try_to_free_swap(page);
N
Nick Piggin 已提交
1109 1110 1111 1112
		unlock_page(page);
		putback_lru_page(page);
		continue;

L
Linus Torvalds 已提交
1113
activate_locked:
1114 1115
		/* Not a candidate for swapping, so reclaim swap space. */
		if (PageSwapCache(page) && vm_swap_full())
1116
			try_to_free_swap(page);
1117
		VM_BUG_ON_PAGE(PageActive(page), page);
L
Linus Torvalds 已提交
1118 1119 1120 1121 1122 1123
		SetPageActive(page);
		pgactivate++;
keep_locked:
		unlock_page(page);
keep:
		list_add(&page->lru, &ret_pages);
1124
		VM_BUG_ON_PAGE(PageLRU(page) || PageUnevictable(page), page);
L
Linus Torvalds 已提交
1125
	}
1126

1127
	free_hot_cold_page_list(&free_pages, true);
1128

L
Linus Torvalds 已提交
1129
	list_splice(&ret_pages, page_list);
1130
	count_vm_events(PGACTIVATE, pgactivate);
1131
	mem_cgroup_uncharge_end();
1132 1133
	*ret_nr_dirty += nr_dirty;
	*ret_nr_congested += nr_congested;
1134
	*ret_nr_unqueued_dirty += nr_unqueued_dirty;
1135
	*ret_nr_writeback += nr_writeback;
1136
	*ret_nr_immediate += nr_immediate;
1137
	return nr_reclaimed;
L
Linus Torvalds 已提交
1138 1139
}

1140 1141 1142 1143 1144 1145 1146 1147
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,
	};
1148
	unsigned long ret, dummy1, dummy2, dummy3, dummy4, dummy5;
1149 1150 1151 1152
	struct page *page, *next;
	LIST_HEAD(clean_pages);

	list_for_each_entry_safe(page, next, page_list, lru) {
1153 1154
		if (page_is_file_cache(page) && !PageDirty(page) &&
		    !isolated_balloon_page(page)) {
1155 1156 1157 1158 1159 1160
			ClearPageActive(page);
			list_move(&page->lru, &clean_pages);
		}
	}

	ret = shrink_page_list(&clean_pages, zone, &sc,
1161 1162
			TTU_UNMAP|TTU_IGNORE_ACCESS,
			&dummy1, &dummy2, &dummy3, &dummy4, &dummy5, true);
1163
	list_splice(&clean_pages, page_list);
1164
	mod_zone_page_state(zone, NR_ISOLATED_FILE, -ret);
1165 1166 1167
	return ret;
}

A
Andy Whitcroft 已提交
1168 1169 1170 1171 1172 1173 1174 1175 1176 1177
/*
 * 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.
 */
1178
int __isolate_lru_page(struct page *page, isolate_mode_t mode)
A
Andy Whitcroft 已提交
1179 1180 1181 1182 1183 1184 1185
{
	int ret = -EINVAL;

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

M
Minchan Kim 已提交
1186 1187
	/* Compaction should not handle unevictable pages but CMA can do so */
	if (PageUnevictable(page) && !(mode & ISOLATE_UNEVICTABLE))
L
Lee Schermerhorn 已提交
1188 1189
		return ret;

A
Andy Whitcroft 已提交
1190
	ret = -EBUSY;
K
KAMEZAWA Hiroyuki 已提交
1191

1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224
	/*
	 * 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;
		}
	}
1225

1226 1227 1228
	if ((mode & ISOLATE_UNMAPPED) && page_mapped(page))
		return ret;

A
Andy Whitcroft 已提交
1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241
	if (likely(get_page_unless_zero(page))) {
		/*
		 * Be careful not to clear PageLRU until after we're
		 * sure the page is not being freed elsewhere -- the
		 * page release code relies on it.
		 */
		ClearPageLRU(page);
		ret = 0;
	}

	return ret;
}

L
Linus Torvalds 已提交
1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252
/*
 * 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.
1253
 * @lruvec:	The LRU vector to pull pages from.
L
Linus Torvalds 已提交
1254
 * @dst:	The temp list to put pages on to.
H
Hugh Dickins 已提交
1255
 * @nr_scanned:	The number of pages that were scanned.
1256
 * @sc:		The scan_control struct for this reclaim session
A
Andy Whitcroft 已提交
1257
 * @mode:	One of the LRU isolation modes
1258
 * @lru:	LRU list id for isolating
L
Linus Torvalds 已提交
1259 1260 1261
 *
 * returns how many pages were moved onto *@dst.
 */
1262
static unsigned long isolate_lru_pages(unsigned long nr_to_scan,
1263
		struct lruvec *lruvec, struct list_head *dst,
1264
		unsigned long *nr_scanned, struct scan_control *sc,
1265
		isolate_mode_t mode, enum lru_list lru)
L
Linus Torvalds 已提交
1266
{
H
Hugh Dickins 已提交
1267
	struct list_head *src = &lruvec->lists[lru];
1268
	unsigned long nr_taken = 0;
1269
	unsigned long scan;
L
Linus Torvalds 已提交
1270

1271
	for (scan = 0; scan < nr_to_scan && !list_empty(src); scan++) {
A
Andy Whitcroft 已提交
1272
		struct page *page;
1273
		int nr_pages;
A
Andy Whitcroft 已提交
1274

L
Linus Torvalds 已提交
1275 1276 1277
		page = lru_to_page(src);
		prefetchw_prev_lru_page(page, src, flags);

1278
		VM_BUG_ON_PAGE(!PageLRU(page), page);
N
Nick Piggin 已提交
1279

1280
		switch (__isolate_lru_page(page, mode)) {
A
Andy Whitcroft 已提交
1281
		case 0:
1282 1283
			nr_pages = hpage_nr_pages(page);
			mem_cgroup_update_lru_size(lruvec, lru, -nr_pages);
A
Andy Whitcroft 已提交
1284
			list_move(&page->lru, dst);
1285
			nr_taken += nr_pages;
A
Andy Whitcroft 已提交
1286 1287 1288 1289 1290 1291
			break;

		case -EBUSY:
			/* else it is being freed elsewhere */
			list_move(&page->lru, src);
			continue;
1292

A
Andy Whitcroft 已提交
1293 1294 1295
		default:
			BUG();
		}
L
Linus Torvalds 已提交
1296 1297
	}

H
Hugh Dickins 已提交
1298
	*nr_scanned = scan;
H
Hugh Dickins 已提交
1299 1300
	trace_mm_vmscan_lru_isolate(sc->order, nr_to_scan, scan,
				    nr_taken, mode, is_file_lru(lru));
L
Linus Torvalds 已提交
1301 1302 1303
	return nr_taken;
}

1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314
/**
 * isolate_lru_page - tries to isolate a page from its LRU list
 * @page: page to isolate from its LRU list
 *
 * Isolates a @page from an LRU list, clears PageLRU and adjusts the
 * vmstat statistic corresponding to whatever LRU list the page was on.
 *
 * Returns 0 if the page was removed from an LRU list.
 * Returns -EBUSY if the page was not on an LRU list.
 *
 * The returned page will have PageLRU() cleared.  If it was found on
L
Lee Schermerhorn 已提交
1315 1316 1317
 * 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.
1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332
 *
 * 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;

1333
	VM_BUG_ON_PAGE(!page_count(page), page);
1334

1335 1336
	if (PageLRU(page)) {
		struct zone *zone = page_zone(page);
1337
		struct lruvec *lruvec;
1338 1339

		spin_lock_irq(&zone->lru_lock);
1340
		lruvec = mem_cgroup_page_lruvec(page, zone);
1341
		if (PageLRU(page)) {
L
Lee Schermerhorn 已提交
1342
			int lru = page_lru(page);
1343
			get_page(page);
1344
			ClearPageLRU(page);
1345 1346
			del_page_from_lru_list(page, lruvec, lru);
			ret = 0;
1347 1348 1349 1350 1351 1352
		}
		spin_unlock_irq(&zone->lru_lock);
	}
	return ret;
}

1353
/*
F
Fengguang Wu 已提交
1354 1355 1356 1357 1358
 * 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.
1359 1360 1361 1362 1363 1364 1365 1366 1367
 */
static int too_many_isolated(struct zone *zone, int file,
		struct scan_control *sc)
{
	unsigned long inactive, isolated;

	if (current_is_kswapd())
		return 0;

1368
	if (!global_reclaim(sc))
1369 1370 1371 1372 1373 1374 1375 1376 1377 1378
		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);
	}

1379 1380 1381 1382 1383 1384 1385 1386
	/*
	 * 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;

1387 1388 1389
	return isolated > inactive;
}

1390
static noinline_for_stack void
H
Hugh Dickins 已提交
1391
putback_inactive_pages(struct lruvec *lruvec, struct list_head *page_list)
1392
{
1393 1394
	struct zone_reclaim_stat *reclaim_stat = &lruvec->reclaim_stat;
	struct zone *zone = lruvec_zone(lruvec);
1395
	LIST_HEAD(pages_to_free);
1396 1397 1398 1399 1400

	/*
	 * Put back any unfreeable pages.
	 */
	while (!list_empty(page_list)) {
1401
		struct page *page = lru_to_page(page_list);
1402
		int lru;
1403

1404
		VM_BUG_ON_PAGE(PageLRU(page), page);
1405
		list_del(&page->lru);
1406
		if (unlikely(!page_evictable(page))) {
1407 1408 1409 1410 1411
			spin_unlock_irq(&zone->lru_lock);
			putback_lru_page(page);
			spin_lock_irq(&zone->lru_lock);
			continue;
		}
1412 1413 1414

		lruvec = mem_cgroup_page_lruvec(page, zone);

1415
		SetPageLRU(page);
1416
		lru = page_lru(page);
1417 1418
		add_page_to_lru_list(page, lruvec, lru);

1419 1420
		if (is_active_lru(lru)) {
			int file = is_file_lru(lru);
1421 1422
			int numpages = hpage_nr_pages(page);
			reclaim_stat->recent_rotated[file] += numpages;
1423
		}
1424 1425 1426
		if (put_page_testzero(page)) {
			__ClearPageLRU(page);
			__ClearPageActive(page);
1427
			del_page_from_lru_list(page, lruvec, lru);
1428 1429 1430 1431 1432 1433 1434

			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);
1435 1436 1437
		}
	}

1438 1439 1440 1441
	/*
	 * To save our caller's stack, now use input list for pages to free.
	 */
	list_splice(&pages_to_free, page_list);
1442 1443
}

1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456
/*
 * If a kernel thread (such as nfsd for loop-back mounts) services
 * a backing device by writing to the page cache it sets PF_LESS_THROTTLE.
 * In that case we should only throttle if the backing device it is
 * writing to is congested.  In other cases it is safe to throttle.
 */
static int current_may_throttle(void)
{
	return !(current->flags & PF_LESS_THROTTLE) ||
		current->backing_dev_info == NULL ||
		bdi_write_congested(current->backing_dev_info);
}

L
Linus Torvalds 已提交
1457
/*
A
Andrew Morton 已提交
1458 1459
 * shrink_inactive_list() is a helper for shrink_zone().  It returns the number
 * of reclaimed pages
L
Linus Torvalds 已提交
1460
 */
1461
static noinline_for_stack unsigned long
1462
shrink_inactive_list(unsigned long nr_to_scan, struct lruvec *lruvec,
1463
		     struct scan_control *sc, enum lru_list lru)
L
Linus Torvalds 已提交
1464 1465
{
	LIST_HEAD(page_list);
1466
	unsigned long nr_scanned;
1467
	unsigned long nr_reclaimed = 0;
1468
	unsigned long nr_taken;
1469 1470
	unsigned long nr_dirty = 0;
	unsigned long nr_congested = 0;
1471
	unsigned long nr_unqueued_dirty = 0;
1472
	unsigned long nr_writeback = 0;
1473
	unsigned long nr_immediate = 0;
1474
	isolate_mode_t isolate_mode = 0;
1475
	int file = is_file_lru(lru);
1476 1477
	struct zone *zone = lruvec_zone(lruvec);
	struct zone_reclaim_stat *reclaim_stat = &lruvec->reclaim_stat;
1478

1479
	while (unlikely(too_many_isolated(zone, file, sc))) {
1480
		congestion_wait(BLK_RW_ASYNC, HZ/10);
1481 1482 1483 1484 1485 1486

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

L
Linus Torvalds 已提交
1487
	lru_add_drain();
1488 1489

	if (!sc->may_unmap)
1490
		isolate_mode |= ISOLATE_UNMAPPED;
1491
	if (!sc->may_writepage)
1492
		isolate_mode |= ISOLATE_CLEAN;
1493

L
Linus Torvalds 已提交
1494
	spin_lock_irq(&zone->lru_lock);
1495

1496 1497
	nr_taken = isolate_lru_pages(nr_to_scan, lruvec, &page_list,
				     &nr_scanned, sc, isolate_mode, lru);
1498 1499 1500 1501

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

1502
	if (global_reclaim(sc)) {
1503 1504
		zone->pages_scanned += nr_scanned;
		if (current_is_kswapd())
H
Hugh Dickins 已提交
1505
			__count_zone_vm_events(PGSCAN_KSWAPD, zone, nr_scanned);
1506
		else
H
Hugh Dickins 已提交
1507
			__count_zone_vm_events(PGSCAN_DIRECT, zone, nr_scanned);
1508
	}
1509
	spin_unlock_irq(&zone->lru_lock);
1510

1511
	if (nr_taken == 0)
1512
		return 0;
A
Andy Whitcroft 已提交
1513

1514
	nr_reclaimed = shrink_page_list(&page_list, zone, sc, TTU_UNMAP,
1515 1516 1517
				&nr_dirty, &nr_unqueued_dirty, &nr_congested,
				&nr_writeback, &nr_immediate,
				false);
1518

1519 1520
	spin_lock_irq(&zone->lru_lock);

1521
	reclaim_stat->recent_scanned[file] += nr_taken;
1522

Y
Ying Han 已提交
1523 1524 1525 1526 1527 1528 1529 1530
	if (global_reclaim(sc)) {
		if (current_is_kswapd())
			__count_zone_vm_events(PGSTEAL_KSWAPD, zone,
					       nr_reclaimed);
		else
			__count_zone_vm_events(PGSTEAL_DIRECT, zone,
					       nr_reclaimed);
	}
N
Nick Piggin 已提交
1531

1532
	putback_inactive_pages(lruvec, &page_list);
1533

1534
	__mod_zone_page_state(zone, NR_ISOLATED_ANON + file, -nr_taken);
1535 1536 1537

	spin_unlock_irq(&zone->lru_lock);

1538
	free_hot_cold_page_list(&page_list, true);
1539

1540 1541 1542 1543 1544 1545 1546 1547 1548 1549
	/*
	 * 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.
	 *
1550 1551 1552
	 * Once a zone is flagged ZONE_WRITEBACK, kswapd will count the number
	 * of pages under pages flagged for immediate reclaim and stall if any
	 * are encountered in the nr_immediate check below.
1553
	 */
1554
	if (nr_writeback && nr_writeback == nr_taken)
1555
		zone_set_flag(zone, ZONE_WRITEBACK);
1556

1557
	/*
1558 1559
	 * memcg will stall in page writeback so only consider forcibly
	 * stalling for global reclaim
1560
	 */
1561
	if (global_reclaim(sc)) {
1562 1563 1564 1565 1566 1567 1568
		/*
		 * Tag a zone as congested if all the dirty pages scanned were
		 * backed by a congested BDI and wait_iff_congested will stall.
		 */
		if (nr_dirty && nr_dirty == nr_congested)
			zone_set_flag(zone, ZONE_CONGESTED);

1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584
		/*
		 * If dirty pages are scanned that are not queued for IO, it
		 * implies that flushers are not keeping up. In this case, flag
		 * the zone ZONE_TAIL_LRU_DIRTY and kswapd will start writing
		 * pages from reclaim context. It will forcibly stall in the
		 * next check.
		 */
		if (nr_unqueued_dirty == nr_taken)
			zone_set_flag(zone, ZONE_TAIL_LRU_DIRTY);

		/*
		 * In addition, if kswapd scans pages marked marked for
		 * immediate reclaim and under writeback (nr_immediate), it
		 * implies that pages are cycling through the LRU faster than
		 * they are written so also forcibly stall.
		 */
1585 1586
		if ((nr_unqueued_dirty == nr_taken || nr_immediate) &&
		    current_may_throttle())
1587
			congestion_wait(BLK_RW_ASYNC, HZ/10);
1588
	}
1589

1590 1591 1592 1593 1594
	/*
	 * Stall direct reclaim for IO completions if underlying BDIs or zone
	 * is congested. Allow kswapd to continue until it starts encountering
	 * unqueued dirty pages or cycling through the LRU too quickly.
	 */
1595 1596
	if (!sc->hibernation_mode && !current_is_kswapd() &&
	    current_may_throttle())
1597 1598
		wait_iff_congested(zone, BLK_RW_ASYNC, HZ/10);

1599 1600 1601
	trace_mm_vmscan_lru_shrink_inactive(zone->zone_pgdat->node_id,
		zone_idx(zone),
		nr_scanned, nr_reclaimed,
1602
		sc->priority,
M
Mel Gorman 已提交
1603
		trace_shrink_flags(file));
1604
	return nr_reclaimed;
L
Linus Torvalds 已提交
1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622 1623
}

/*
 * 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.
 */
1624

1625
static void move_active_pages_to_lru(struct lruvec *lruvec,
1626
				     struct list_head *list,
1627
				     struct list_head *pages_to_free,
1628 1629
				     enum lru_list lru)
{
1630
	struct zone *zone = lruvec_zone(lruvec);
1631 1632
	unsigned long pgmoved = 0;
	struct page *page;
1633
	int nr_pages;
1634 1635 1636

	while (!list_empty(list)) {
		page = lru_to_page(list);
1637
		lruvec = mem_cgroup_page_lruvec(page, zone);
1638

1639
		VM_BUG_ON_PAGE(PageLRU(page), page);
1640 1641
		SetPageLRU(page);

1642 1643
		nr_pages = hpage_nr_pages(page);
		mem_cgroup_update_lru_size(lruvec, lru, nr_pages);
1644
		list_move(&page->lru, &lruvec->lists[lru]);
1645
		pgmoved += nr_pages;
1646

1647 1648 1649
		if (put_page_testzero(page)) {
			__ClearPageLRU(page);
			__ClearPageActive(page);
1650
			del_page_from_lru_list(page, lruvec, lru);
1651 1652 1653 1654 1655 1656 1657

			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);
1658 1659 1660 1661 1662 1663
		}
	}
	__mod_zone_page_state(zone, NR_LRU_BASE + lru, pgmoved);
	if (!is_active_lru(lru))
		__count_vm_events(PGDEACTIVATE, pgmoved);
}
1664

H
Hugh Dickins 已提交
1665
static void shrink_active_list(unsigned long nr_to_scan,
1666
			       struct lruvec *lruvec,
1667
			       struct scan_control *sc,
1668
			       enum lru_list lru)
L
Linus Torvalds 已提交
1669
{
1670
	unsigned long nr_taken;
H
Hugh Dickins 已提交
1671
	unsigned long nr_scanned;
1672
	unsigned long vm_flags;
L
Linus Torvalds 已提交
1673
	LIST_HEAD(l_hold);	/* The pages which were snipped off */
1674
	LIST_HEAD(l_active);
1675
	LIST_HEAD(l_inactive);
L
Linus Torvalds 已提交
1676
	struct page *page;
1677
	struct zone_reclaim_stat *reclaim_stat = &lruvec->reclaim_stat;
1678
	unsigned long nr_rotated = 0;
1679
	isolate_mode_t isolate_mode = 0;
1680
	int file = is_file_lru(lru);
1681
	struct zone *zone = lruvec_zone(lruvec);
L
Linus Torvalds 已提交
1682 1683

	lru_add_drain();
1684 1685

	if (!sc->may_unmap)
1686
		isolate_mode |= ISOLATE_UNMAPPED;
1687
	if (!sc->may_writepage)
1688
		isolate_mode |= ISOLATE_CLEAN;
1689

L
Linus Torvalds 已提交
1690
	spin_lock_irq(&zone->lru_lock);
1691

1692 1693
	nr_taken = isolate_lru_pages(nr_to_scan, lruvec, &l_hold,
				     &nr_scanned, sc, isolate_mode, lru);
1694
	if (global_reclaim(sc))
H
Hugh Dickins 已提交
1695
		zone->pages_scanned += nr_scanned;
1696

1697
	reclaim_stat->recent_scanned[file] += nr_taken;
1698

H
Hugh Dickins 已提交
1699
	__count_zone_vm_events(PGREFILL, zone, nr_scanned);
1700
	__mod_zone_page_state(zone, NR_LRU_BASE + lru, -nr_taken);
K
KOSAKI Motohiro 已提交
1701
	__mod_zone_page_state(zone, NR_ISOLATED_ANON + file, nr_taken);
L
Linus Torvalds 已提交
1702 1703 1704 1705 1706 1707
	spin_unlock_irq(&zone->lru_lock);

	while (!list_empty(&l_hold)) {
		cond_resched();
		page = lru_to_page(&l_hold);
		list_del(&page->lru);
1708

1709
		if (unlikely(!page_evictable(page))) {
L
Lee Schermerhorn 已提交
1710 1711 1712 1713
			putback_lru_page(page);
			continue;
		}

1714 1715 1716 1717 1718 1719 1720 1721
		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);
			}
		}

1722 1723
		if (page_referenced(page, 0, sc->target_mem_cgroup,
				    &vm_flags)) {
1724
			nr_rotated += hpage_nr_pages(page);
1725 1726 1727 1728 1729 1730 1731 1732 1733
			/*
			 * 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.
			 */
1734
			if ((vm_flags & VM_EXEC) && page_is_file_cache(page)) {
1735 1736 1737 1738
				list_add(&page->lru, &l_active);
				continue;
			}
		}
1739

1740
		ClearPageActive(page);	/* we are de-activating */
L
Linus Torvalds 已提交
1741 1742 1743
		list_add(&page->lru, &l_inactive);
	}

1744
	/*
1745
	 * Move pages back to the lru list.
1746
	 */
1747
	spin_lock_irq(&zone->lru_lock);
1748
	/*
1749 1750 1751 1752
	 * 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.
1753
	 */
1754
	reclaim_stat->recent_rotated[file] += nr_rotated;
1755

1756 1757
	move_active_pages_to_lru(lruvec, &l_active, &l_hold, lru);
	move_active_pages_to_lru(lruvec, &l_inactive, &l_hold, lru - LRU_ACTIVE);
K
KOSAKI Motohiro 已提交
1758
	__mod_zone_page_state(zone, NR_ISOLATED_ANON + file, -nr_taken);
1759
	spin_unlock_irq(&zone->lru_lock);
1760

1761
	free_hot_cold_page_list(&l_hold, true);
L
Linus Torvalds 已提交
1762 1763
}

1764
#ifdef CONFIG_SWAP
1765
static int inactive_anon_is_low_global(struct zone *zone)
1766 1767 1768 1769 1770 1771 1772 1773 1774 1775 1776 1777
{
	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;
}

1778 1779
/**
 * inactive_anon_is_low - check if anonymous pages need to be deactivated
1780
 * @lruvec: LRU vector to check
1781 1782 1783 1784
 *
 * Returns true if the zone does not have enough inactive anon pages,
 * meaning some active anon pages need to be deactivated.
 */
1785
static int inactive_anon_is_low(struct lruvec *lruvec)
1786
{
1787 1788 1789 1790 1791 1792 1793
	/*
	 * If we don't have swap space, anonymous page deactivation
	 * is pointless.
	 */
	if (!total_swap_pages)
		return 0;

1794
	if (!mem_cgroup_disabled())
1795
		return mem_cgroup_inactive_anon_is_low(lruvec);
1796

1797
	return inactive_anon_is_low_global(lruvec_zone(lruvec));
1798
}
1799
#else
1800
static inline int inactive_anon_is_low(struct lruvec *lruvec)
1801 1802 1803 1804
{
	return 0;
}
#endif
1805

1806 1807
/**
 * inactive_file_is_low - check if file pages need to be deactivated
1808
 * @lruvec: LRU vector to check
1809 1810 1811 1812 1813 1814 1815 1816 1817 1818 1819
 *
 * 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.
 */
1820
static int inactive_file_is_low(struct lruvec *lruvec)
1821
{
1822 1823 1824 1825 1826
	unsigned long inactive;
	unsigned long active;

	inactive = get_lru_size(lruvec, LRU_INACTIVE_FILE);
	active = get_lru_size(lruvec, LRU_ACTIVE_FILE);
1827

1828
	return active > inactive;
1829 1830
}

H
Hugh Dickins 已提交
1831
static int inactive_list_is_low(struct lruvec *lruvec, enum lru_list lru)
1832
{
H
Hugh Dickins 已提交
1833
	if (is_file_lru(lru))
1834
		return inactive_file_is_low(lruvec);
1835
	else
1836
		return inactive_anon_is_low(lruvec);
1837 1838
}

1839
static unsigned long shrink_list(enum lru_list lru, unsigned long nr_to_scan,
1840
				 struct lruvec *lruvec, struct scan_control *sc)
1841
{
1842
	if (is_active_lru(lru)) {
H
Hugh Dickins 已提交
1843
		if (inactive_list_is_low(lruvec, lru))
1844
			shrink_active_list(nr_to_scan, lruvec, sc, lru);
1845 1846 1847
		return 0;
	}

1848
	return shrink_inactive_list(nr_to_scan, lruvec, sc, lru);
1849 1850
}

1851 1852 1853 1854 1855 1856 1857
enum scan_balance {
	SCAN_EQUAL,
	SCAN_FRACT,
	SCAN_ANON,
	SCAN_FILE,
};

1858 1859 1860 1861 1862 1863
/*
 * Determine how aggressively the anon and file LRU lists should be
 * scanned.  The relative value of each set of LRU lists is determined
 * by looking at the fraction of the pages scanned we did rotate back
 * onto the active list instead of evict.
 *
W
Wanpeng Li 已提交
1864 1865
 * 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
1866
 */
1867
static void get_scan_count(struct lruvec *lruvec, struct scan_control *sc,
1868
			   unsigned long *nr)
1869
{
1870 1871 1872 1873
	struct zone_reclaim_stat *reclaim_stat = &lruvec->reclaim_stat;
	u64 fraction[2];
	u64 denominator = 0;	/* gcc */
	struct zone *zone = lruvec_zone(lruvec);
1874
	unsigned long anon_prio, file_prio;
1875
	enum scan_balance scan_balance;
1876
	unsigned long anon, file;
1877
	bool force_scan = false;
1878
	unsigned long ap, fp;
H
Hugh Dickins 已提交
1879
	enum lru_list lru;
1880 1881
	bool some_scanned;
	int pass;
1882

1883 1884 1885 1886 1887 1888 1889 1890 1891 1892
	/*
	 * 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.
	 */
1893
	if (current_is_kswapd() && !zone_reclaimable(zone))
1894
		force_scan = true;
1895
	if (!global_reclaim(sc))
1896
		force_scan = true;
1897 1898

	/* If we have no swap space, do not bother scanning anon pages. */
1899
	if (!sc->may_swap || (get_nr_swap_pages() <= 0)) {
1900
		scan_balance = SCAN_FILE;
1901 1902
		goto out;
	}
1903

1904 1905 1906 1907 1908 1909 1910
	/*
	 * 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.
	 */
1911
	if (!global_reclaim(sc) && !sc->swappiness) {
1912
		scan_balance = SCAN_FILE;
1913 1914 1915 1916 1917 1918 1919 1920
		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).
	 */
1921
	if (!sc->priority && sc->swappiness) {
1922
		scan_balance = SCAN_EQUAL;
1923 1924 1925
		goto out;
	}

1926 1927 1928 1929
	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);
1930

1931 1932 1933 1934 1935 1936 1937 1938 1939 1940 1941 1942 1943 1944 1945 1946 1947 1948
	/*
	 * Prevent the reclaimer from falling into the cache trap: as
	 * cache pages start out inactive, every cache fault will tip
	 * the scan balance towards the file LRU.  And as the file LRU
	 * shrinks, so does the window for rotation from references.
	 * This means we have a runaway feedback loop where a tiny
	 * thrashing file LRU becomes infinitely more attractive than
	 * anon pages.  Try to detect this based on file LRU size.
	 */
	if (global_reclaim(sc)) {
		unsigned long free = zone_page_state(zone, NR_FREE_PAGES);

		if (unlikely(file + free <= high_wmark_pages(zone))) {
			scan_balance = SCAN_ANON;
			goto out;
		}
	}

1949 1950 1951 1952 1953
	/*
	 * There is enough inactive page cache, do not reclaim
	 * anything from the anonymous working set right now.
	 */
	if (!inactive_file_is_low(lruvec)) {
1954
		scan_balance = SCAN_FILE;
1955 1956 1957
		goto out;
	}

1958 1959
	scan_balance = SCAN_FRACT;

1960 1961 1962 1963
	/*
	 * With swappiness at 100, anonymous and file have the same priority.
	 * This scanning priority is essentially the inverse of IO cost.
	 */
1964
	anon_prio = sc->swappiness;
H
Hugh Dickins 已提交
1965
	file_prio = 200 - anon_prio;
1966

1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977
	/*
	 * 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]
	 */
1978
	spin_lock_irq(&zone->lru_lock);
1979 1980 1981
	if (unlikely(reclaim_stat->recent_scanned[0] > anon / 4)) {
		reclaim_stat->recent_scanned[0] /= 2;
		reclaim_stat->recent_rotated[0] /= 2;
1982 1983
	}

1984 1985 1986
	if (unlikely(reclaim_stat->recent_scanned[1] > file / 4)) {
		reclaim_stat->recent_scanned[1] /= 2;
		reclaim_stat->recent_rotated[1] /= 2;
1987 1988 1989
	}

	/*
1990 1991 1992
	 * 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.
1993
	 */
1994
	ap = anon_prio * (reclaim_stat->recent_scanned[0] + 1);
1995
	ap /= reclaim_stat->recent_rotated[0] + 1;
1996

1997
	fp = file_prio * (reclaim_stat->recent_scanned[1] + 1);
1998
	fp /= reclaim_stat->recent_rotated[1] + 1;
1999
	spin_unlock_irq(&zone->lru_lock);
2000

2001 2002 2003 2004
	fraction[0] = ap;
	fraction[1] = fp;
	denominator = ap + fp + 1;
out:
2005 2006 2007 2008 2009 2010 2011
	some_scanned = false;
	/* Only use force_scan on second pass. */
	for (pass = 0; !some_scanned && pass < 2; pass++) {
		for_each_evictable_lru(lru) {
			int file = is_file_lru(lru);
			unsigned long size;
			unsigned long scan;
2012

2013 2014
			size = get_lru_size(lruvec, lru);
			scan = size >> sc->priority;
2015

2016 2017
			if (!scan && pass && force_scan)
				scan = min(size, SWAP_CLUSTER_MAX);
2018

2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041
			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();
			}
			nr[lru] = scan;
2042
			/*
2043 2044
			 * Skip the second pass and don't force_scan,
			 * if we found something to scan.
2045
			 */
2046
			some_scanned |= !!scan;
2047
		}
2048
	}
2049
}
2050

2051 2052 2053 2054 2055 2056
/*
 * 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];
2057
	unsigned long targets[NR_LRU_LISTS];
2058 2059 2060 2061 2062
	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;
2063
	bool scan_adjusted;
2064 2065 2066

	get_scan_count(lruvec, sc, nr);

2067 2068 2069
	/* Record the original scan target for proportional adjustments later */
	memcpy(targets, nr, sizeof(nr));

2070 2071 2072 2073 2074 2075 2076 2077 2078 2079 2080 2081 2082 2083
	/*
	 * Global reclaiming within direct reclaim at DEF_PRIORITY is a normal
	 * event that can occur when there is little memory pressure e.g.
	 * multiple streaming readers/writers. Hence, we do not abort scanning
	 * when the requested number of pages are reclaimed when scanning at
	 * DEF_PRIORITY on the assumption that the fact we are direct
	 * reclaiming implies that kswapd is not keeping up and it is best to
	 * do a batch of work at once. For memcg reclaim one check is made to
	 * abort proportional reclaim if either the file or anon lru has already
	 * dropped to zero at the first pass.
	 */
	scan_adjusted = (global_reclaim(sc) && !current_is_kswapd() &&
			 sc->priority == DEF_PRIORITY);

2084 2085 2086
	blk_start_plug(&plug);
	while (nr[LRU_INACTIVE_ANON] || nr[LRU_ACTIVE_FILE] ||
					nr[LRU_INACTIVE_FILE]) {
2087 2088 2089
		unsigned long nr_anon, nr_file, percentage;
		unsigned long nr_scanned;

2090 2091 2092 2093 2094 2095 2096 2097 2098
		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);
			}
		}
2099 2100 2101 2102 2103 2104

		if (nr_reclaimed < nr_to_reclaim || scan_adjusted)
			continue;

		/*
		 * For kswapd and memcg, reclaim at least the number of pages
2105
		 * requested. Ensure that the anon and file LRUs are scanned
2106 2107 2108 2109 2110 2111 2112
		 * proportionally what was requested by get_scan_count(). We
		 * stop reclaiming one LRU and reduce the amount scanning
		 * proportional to the original scan target.
		 */
		nr_file = nr[LRU_INACTIVE_FILE] + nr[LRU_ACTIVE_FILE];
		nr_anon = nr[LRU_INACTIVE_ANON] + nr[LRU_ACTIVE_ANON];

2113 2114 2115 2116 2117 2118 2119 2120 2121
		/*
		 * It's just vindictive to attack the larger once the smaller
		 * has gone to zero.  And given the way we stop scanning the
		 * smaller below, this makes sure that we only make one nudge
		 * towards proportionality once we've got nr_to_reclaim.
		 */
		if (!nr_file || !nr_anon)
			break;

2122 2123 2124 2125 2126 2127 2128 2129 2130 2131 2132 2133 2134 2135 2136 2137 2138 2139 2140 2141 2142 2143 2144 2145 2146 2147 2148 2149 2150 2151 2152
		if (nr_file > nr_anon) {
			unsigned long scan_target = targets[LRU_INACTIVE_ANON] +
						targets[LRU_ACTIVE_ANON] + 1;
			lru = LRU_BASE;
			percentage = nr_anon * 100 / scan_target;
		} else {
			unsigned long scan_target = targets[LRU_INACTIVE_FILE] +
						targets[LRU_ACTIVE_FILE] + 1;
			lru = LRU_FILE;
			percentage = nr_file * 100 / scan_target;
		}

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

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

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

		scan_adjusted = true;
2153 2154 2155 2156 2157 2158 2159 2160 2161 2162 2163 2164 2165 2166 2167
	}
	blk_finish_plug(&plug);
	sc->nr_reclaimed += nr_reclaimed;

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

	throttle_vm_writeout(sc->gfp_mask);
}

M
Mel Gorman 已提交
2168
/* Use reclaim/compaction for costly allocs or under memory pressure */
2169
static bool in_reclaim_compaction(struct scan_control *sc)
M
Mel Gorman 已提交
2170
{
2171
	if (IS_ENABLED(CONFIG_COMPACTION) && sc->order &&
M
Mel Gorman 已提交
2172
			(sc->order > PAGE_ALLOC_COSTLY_ORDER ||
2173
			 sc->priority < DEF_PRIORITY - 2))
M
Mel Gorman 已提交
2174 2175 2176 2177 2178
		return true;

	return false;
}

2179
/*
M
Mel Gorman 已提交
2180 2181 2182 2183 2184
 * 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.
2185
 */
2186
static inline bool should_continue_reclaim(struct zone *zone,
2187 2188 2189 2190 2191 2192 2193 2194
					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 */
2195
	if (!in_reclaim_compaction(sc))
2196 2197
		return false;

2198 2199 2200 2201 2202 2203 2204 2205 2206 2207 2208 2209 2210 2211 2212 2213 2214 2215 2216 2217 2218 2219
	/* 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;
	}
2220 2221 2222 2223 2224 2225

	/*
	 * If we have not reclaimed enough pages for compaction and the
	 * inactive lists are large enough, continue reclaiming
	 */
	pages_for_compaction = (2UL << sc->order);
2226
	inactive_lru_pages = zone_page_state(zone, NR_INACTIVE_FILE);
2227
	if (get_nr_swap_pages() > 0)
2228
		inactive_lru_pages += zone_page_state(zone, NR_INACTIVE_ANON);
2229 2230 2231 2232 2233
	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 */
2234
	switch (compaction_suitable(zone, sc->order)) {
2235 2236 2237 2238 2239 2240 2241 2242
	case COMPACT_PARTIAL:
	case COMPACT_CONTINUE:
		return false;
	default:
		return true;
	}
}

2243
static void shrink_zone(struct zone *zone, struct scan_control *sc)
L
Linus Torvalds 已提交
2244
{
2245
	unsigned long nr_reclaimed, nr_scanned;
L
Linus Torvalds 已提交
2246

2247 2248 2249 2250 2251 2252
	do {
		struct mem_cgroup *root = sc->target_mem_cgroup;
		struct mem_cgroup_reclaim_cookie reclaim = {
			.zone = zone,
			.priority = sc->priority,
		};
2253
		struct mem_cgroup *memcg;
2254

2255 2256
		nr_reclaimed = sc->nr_reclaimed;
		nr_scanned = sc->nr_scanned;
L
Linus Torvalds 已提交
2257

2258 2259
		memcg = mem_cgroup_iter(root, NULL, &reclaim);
		do {
2260
			struct lruvec *lruvec;
2261

2262
			lruvec = mem_cgroup_zone_lruvec(zone, memcg);
2263

2264
			sc->swappiness = mem_cgroup_swappiness(memcg);
2265
			shrink_lruvec(lruvec, sc);
2266

2267
			/*
2268 2269
			 * Direct reclaim and kswapd have to scan all memory
			 * cgroups to fulfill the overall scan target for the
2270
			 * zone.
2271 2272 2273 2274 2275
			 *
			 * Limit reclaim, on the other hand, only cares about
			 * nr_to_reclaim pages to be reclaimed and it will
			 * retry with decreasing priority if one round over the
			 * whole hierarchy is not sufficient.
2276
			 */
2277 2278
			if (!global_reclaim(sc) &&
					sc->nr_reclaimed >= sc->nr_to_reclaim) {
2279 2280 2281
				mem_cgroup_iter_break(root, memcg);
				break;
			}
2282 2283
			memcg = mem_cgroup_iter(root, memcg, &reclaim);
		} while (memcg);
2284 2285 2286 2287 2288

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

2289 2290
	} while (should_continue_reclaim(zone, sc->nr_reclaimed - nr_reclaimed,
					 sc->nr_scanned - nr_scanned, sc));
2291 2292
}

2293 2294 2295 2296 2297 2298 2299 2300 2301 2302 2303 2304 2305 2306 2307 2308
/* 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
	 */
2309 2310
	balance_gap = min(low_wmark_pages(zone), DIV_ROUND_UP(
			zone->managed_pages, KSWAPD_ZONE_BALANCE_GAP_RATIO));
2311 2312 2313 2314 2315 2316 2317
	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
	 */
2318
	if (compaction_deferred(zone, sc->order))
2319 2320 2321 2322 2323 2324 2325 2326 2327
		return watermark_ok;

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

	return watermark_ok;
}

L
Linus Torvalds 已提交
2328 2329 2330 2331 2332
/*
 * 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.
 *
2333 2334
 * We reclaim from a zone even if that zone is over high_wmark_pages(zone).
 * Because:
L
Linus Torvalds 已提交
2335 2336
 * a) The caller may be trying to free *extra* pages to satisfy a higher-order
 *    allocation or
2337 2338 2339
 * b) The target zone may be at high_wmark_pages(zone) but the lower zones
 *    must go *over* high_wmark_pages(zone) to satisfy the `incremental min'
 *    zone defense algorithm.
L
Linus Torvalds 已提交
2340 2341 2342
 *
 * If a zone is deemed to be full of pinned pages then just give it a light
 * scan then give up on it.
2343 2344
 *
 * This function returns true if a zone is being reclaimed for a costly
2345
 * high-order allocation and compaction is ready to begin. This indicates to
2346 2347
 * the caller that it should consider retrying the allocation instead of
 * further reclaim.
L
Linus Torvalds 已提交
2348
 */
2349
static bool shrink_zones(struct zonelist *zonelist, struct scan_control *sc)
L
Linus Torvalds 已提交
2350
{
2351
	struct zoneref *z;
2352
	struct zone *zone;
2353 2354
	unsigned long nr_soft_reclaimed;
	unsigned long nr_soft_scanned;
2355
	unsigned long lru_pages = 0;
2356
	bool aborted_reclaim = false;
2357
	struct reclaim_state *reclaim_state = current->reclaim_state;
2358
	gfp_t orig_mask;
2359 2360 2361
	struct shrink_control shrink = {
		.gfp_mask = sc->gfp_mask,
	};
2362
	enum zone_type requested_highidx = gfp_zone(sc->gfp_mask);
2363

2364 2365 2366 2367 2368
	/*
	 * 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
	 */
2369
	orig_mask = sc->gfp_mask;
2370 2371 2372
	if (buffer_heads_over_limit)
		sc->gfp_mask |= __GFP_HIGHMEM;

2373
	nodes_clear(shrink.nodes_to_scan);
2374

2375 2376
	for_each_zone_zonelist_nodemask(zone, z, zonelist,
					gfp_zone(sc->gfp_mask), sc->nodemask) {
2377
		if (!populated_zone(zone))
L
Linus Torvalds 已提交
2378
			continue;
2379 2380 2381 2382
		/*
		 * Take care memory controller reclaiming has small influence
		 * to global LRU.
		 */
2383
		if (global_reclaim(sc)) {
2384 2385
			if (!cpuset_zone_allowed_hardwall(zone, GFP_KERNEL))
				continue;
2386 2387

			lru_pages += zone_reclaimable_pages(zone);
2388
			node_set(zone_to_nid(zone), shrink.nodes_to_scan);
2389

2390 2391
			if (sc->priority != DEF_PRIORITY &&
			    !zone_reclaimable(zone))
2392
				continue;	/* Let kswapd poll it */
2393
			if (IS_ENABLED(CONFIG_COMPACTION)) {
2394
				/*
2395 2396 2397 2398 2399
				 * 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
2400 2401
				 * noticeable problem, like transparent huge
				 * page allocations.
2402
				 */
2403 2404
				if ((zonelist_zone_idx(z) <= requested_highidx)
				    && compaction_ready(zone, sc)) {
2405
					aborted_reclaim = true;
2406
					continue;
2407
				}
2408
			}
2409 2410 2411 2412 2413 2414 2415 2416 2417 2418 2419 2420
			/*
			 * 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;
2421
			/* need some check for avoid more shrink_zone() */
2422
		}
2423

2424
		shrink_zone(zone, sc);
L
Linus Torvalds 已提交
2425
	}
2426

2427 2428 2429 2430 2431 2432 2433
	/*
	 * Don't shrink slabs when reclaiming memory from over limit cgroups
	 * but do shrink slab at least once when aborting reclaim for
	 * compaction to avoid unevenly scanning file/anon LRU pages over slab
	 * pages.
	 */
	if (global_reclaim(sc)) {
2434
		shrink_slab(&shrink, sc->nr_scanned, lru_pages);
2435 2436 2437 2438 2439 2440
		if (reclaim_state) {
			sc->nr_reclaimed += reclaim_state->reclaimed_slab;
			reclaim_state->reclaimed_slab = 0;
		}
	}

2441 2442 2443 2444 2445 2446
	/*
	 * Restore to original mask to avoid the impact on the caller if we
	 * promoted it to __GFP_HIGHMEM.
	 */
	sc->gfp_mask = orig_mask;

2447
	return aborted_reclaim;
2448 2449
}

2450
/* All zones in zonelist are unreclaimable? */
2451 2452 2453 2454 2455 2456 2457 2458 2459 2460 2461 2462
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;
2463
		if (zone_reclaimable(zone))
2464
			return false;
2465 2466
	}

2467
	return true;
L
Linus Torvalds 已提交
2468
}
2469

L
Linus Torvalds 已提交
2470 2471 2472 2473 2474 2475 2476 2477
/*
 * 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
2478 2479 2480 2481
 * 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.
2482 2483 2484
 *
 * returns:	0, if no pages reclaimed
 * 		else, the number of pages reclaimed
L
Linus Torvalds 已提交
2485
 */
2486
static unsigned long do_try_to_free_pages(struct zonelist *zonelist,
2487
					  struct scan_control *sc)
L
Linus Torvalds 已提交
2488
{
2489
	unsigned long total_scanned = 0;
2490
	unsigned long writeback_threshold;
2491
	bool aborted_reclaim;
L
Linus Torvalds 已提交
2492

2493 2494
	delayacct_freepages_start();

2495
	if (global_reclaim(sc))
2496
		count_vm_event(ALLOCSTALL);
L
Linus Torvalds 已提交
2497

2498
	do {
2499 2500
		vmpressure_prio(sc->gfp_mask, sc->target_mem_cgroup,
				sc->priority);
2501
		sc->nr_scanned = 0;
2502
		aborted_reclaim = shrink_zones(zonelist, sc);
2503

2504
		total_scanned += sc->nr_scanned;
2505
		if (sc->nr_reclaimed >= sc->nr_to_reclaim)
L
Linus Torvalds 已提交
2506 2507
			goto out;

2508 2509 2510 2511 2512 2513 2514
		/*
		 * If we're getting trouble reclaiming, start doing
		 * writepage even in laptop mode.
		 */
		if (sc->priority < DEF_PRIORITY - 2)
			sc->may_writepage = 1;

L
Linus Torvalds 已提交
2515 2516 2517 2518 2519 2520 2521
		/*
		 * 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.
		 */
2522 2523
		writeback_threshold = sc->nr_to_reclaim + sc->nr_to_reclaim / 2;
		if (total_scanned > writeback_threshold) {
2524 2525
			wakeup_flusher_threads(laptop_mode ? 0 : total_scanned,
						WB_REASON_TRY_TO_FREE_PAGES);
2526
			sc->may_writepage = 1;
L
Linus Torvalds 已提交
2527
		}
2528
	} while (--sc->priority >= 0 && !aborted_reclaim);
2529

L
Linus Torvalds 已提交
2530
out:
2531 2532
	delayacct_freepages_end();

2533 2534 2535
	if (sc->nr_reclaimed)
		return sc->nr_reclaimed;

2536 2537 2538 2539 2540 2541 2542 2543
	/*
	 * 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;

2544 2545
	/* Aborted reclaim to try compaction? don't OOM, then */
	if (aborted_reclaim)
2546 2547
		return 1;

2548
	/* top priority shrink_zones still had more to do? don't OOM, then */
2549
	if (global_reclaim(sc) && !all_unreclaimable(zonelist, sc))
2550 2551 2552
		return 1;

	return 0;
L
Linus Torvalds 已提交
2553 2554
}

2555 2556 2557 2558 2559 2560 2561 2562 2563 2564
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];
2565 2566 2567
		if (!populated_zone(zone))
			continue;

2568 2569 2570 2571
		pfmemalloc_reserve += min_wmark_pages(zone);
		free_pages += zone_page_state(zone, NR_FREE_PAGES);
	}

2572 2573 2574 2575
	/* If there are no reserves (unexpected config) then do not throttle */
	if (!pfmemalloc_reserve)
		return true;

2576 2577 2578 2579 2580 2581 2582 2583 2584 2585 2586 2587 2588 2589 2590 2591
	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
2592 2593 2594 2595
 * 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.
2596
 */
2597
static bool throttle_direct_reclaim(gfp_t gfp_mask, struct zonelist *zonelist,
2598 2599
					nodemask_t *nodemask)
{
2600
	struct zoneref *z;
2601
	struct zone *zone;
2602
	pg_data_t *pgdat = NULL;
2603 2604 2605 2606 2607 2608 2609 2610 2611

	/*
	 * 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)
2612 2613 2614 2615 2616 2617 2618 2619
		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;
2620

2621 2622 2623 2624 2625 2626 2627 2628 2629 2630 2631 2632 2633 2634 2635 2636 2637 2638 2639 2640 2641 2642 2643 2644 2645 2646 2647 2648
	/*
	 * Check if the pfmemalloc reserves are ok by finding the first node
	 * with a usable ZONE_NORMAL or lower zone. The expectation is that
	 * GFP_KERNEL will be required for allocating network buffers when
	 * swapping over the network so ZONE_HIGHMEM is unusable.
	 *
	 * Throttling is based on the first usable node and throttled processes
	 * wait on a queue until kswapd makes progress and wakes them. There
	 * is an affinity then between processes waking up and where reclaim
	 * progress has been made assuming the process wakes on the same node.
	 * More importantly, processes running on remote nodes will not compete
	 * for remote pfmemalloc reserves and processes on different nodes
	 * should make reasonable progress.
	 */
	for_each_zone_zonelist_nodemask(zone, z, zonelist,
					gfp_mask, nodemask) {
		if (zone_idx(zone) > ZONE_NORMAL)
			continue;

		/* Throttle based on the first usable node */
		pgdat = zone->zone_pgdat;
		if (pfmemalloc_watermark_ok(pgdat))
			goto out;
		break;
	}

	/* If no zone was usable by the allocation flags then do not throttle */
	if (!pgdat)
2649
		goto out;
2650

2651 2652 2653
	/* Account for the throttling */
	count_vm_event(PGSCAN_DIRECT_THROTTLE);

2654 2655 2656 2657 2658 2659 2660 2661 2662 2663 2664
	/*
	 * 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);
2665 2666

		goto check_pending;
2667 2668 2669 2670 2671
	}

	/* Throttle until kswapd wakes the process */
	wait_event_killable(zone->zone_pgdat->pfmemalloc_wait,
		pfmemalloc_watermark_ok(pgdat));
2672 2673 2674 2675 2676 2677 2678

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

out:
	return false;
2679 2680
}

2681
unsigned long try_to_free_pages(struct zonelist *zonelist, int order,
2682
				gfp_t gfp_mask, nodemask_t *nodemask)
2683
{
2684
	unsigned long nr_reclaimed;
2685
	struct scan_control sc = {
2686
		.gfp_mask = (gfp_mask = memalloc_noio_flags(gfp_mask)),
2687
		.may_writepage = !laptop_mode,
2688
		.nr_to_reclaim = SWAP_CLUSTER_MAX,
2689
		.may_unmap = 1,
2690
		.may_swap = 1,
2691
		.order = order,
2692
		.priority = DEF_PRIORITY,
2693
		.target_mem_cgroup = NULL,
2694
		.nodemask = nodemask,
2695 2696
	};

2697
	/*
2698 2699 2700
	 * 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.
2701
	 */
2702
	if (throttle_direct_reclaim(gfp_mask, zonelist, nodemask))
2703 2704
		return 1;

2705 2706 2707 2708
	trace_mm_vmscan_direct_reclaim_begin(order,
				sc.may_writepage,
				gfp_mask);

2709
	nr_reclaimed = do_try_to_free_pages(zonelist, &sc);
2710 2711 2712 2713

	trace_mm_vmscan_direct_reclaim_end(nr_reclaimed);

	return nr_reclaimed;
2714 2715
}

A
Andrew Morton 已提交
2716
#ifdef CONFIG_MEMCG
2717

2718
unsigned long mem_cgroup_shrink_node_zone(struct mem_cgroup *memcg,
2719
						gfp_t gfp_mask, bool noswap,
2720 2721
						struct zone *zone,
						unsigned long *nr_scanned)
2722 2723
{
	struct scan_control sc = {
2724
		.nr_scanned = 0,
2725
		.nr_to_reclaim = SWAP_CLUSTER_MAX,
2726 2727 2728 2729
		.may_writepage = !laptop_mode,
		.may_unmap = 1,
		.may_swap = !noswap,
		.order = 0,
2730
		.priority = 0,
2731
		.swappiness = mem_cgroup_swappiness(memcg),
2732
		.target_mem_cgroup = memcg,
2733
	};
2734
	struct lruvec *lruvec = mem_cgroup_zone_lruvec(zone, memcg);
2735

2736 2737
	sc.gfp_mask = (gfp_mask & GFP_RECLAIM_MASK) |
			(GFP_HIGHUSER_MOVABLE & ~GFP_RECLAIM_MASK);
2738

2739
	trace_mm_vmscan_memcg_softlimit_reclaim_begin(sc.order,
2740 2741 2742
						      sc.may_writepage,
						      sc.gfp_mask);

2743 2744 2745 2746 2747 2748 2749
	/*
	 * 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.
	 */
2750
	shrink_lruvec(lruvec, &sc);
2751 2752 2753

	trace_mm_vmscan_memcg_softlimit_reclaim_end(sc.nr_reclaimed);

2754
	*nr_scanned = sc.nr_scanned;
2755 2756 2757
	return sc.nr_reclaimed;
}

2758
unsigned long try_to_free_mem_cgroup_pages(struct mem_cgroup *memcg,
K
KOSAKI Motohiro 已提交
2759
					   gfp_t gfp_mask,
2760
					   bool noswap)
2761
{
2762
	struct zonelist *zonelist;
2763
	unsigned long nr_reclaimed;
2764
	int nid;
2765 2766
	struct scan_control sc = {
		.may_writepage = !laptop_mode,
2767
		.may_unmap = 1,
2768
		.may_swap = !noswap,
2769
		.nr_to_reclaim = SWAP_CLUSTER_MAX,
2770
		.order = 0,
2771
		.priority = DEF_PRIORITY,
2772
		.target_mem_cgroup = memcg,
2773
		.nodemask = NULL, /* we don't care the placement */
2774 2775 2776
		.gfp_mask = (gfp_mask & GFP_RECLAIM_MASK) |
				(GFP_HIGHUSER_MOVABLE & ~GFP_RECLAIM_MASK),
	};
2777

2778 2779 2780 2781 2782
	/*
	 * 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.
	 */
2783
	nid = mem_cgroup_select_victim_node(memcg);
2784 2785

	zonelist = NODE_DATA(nid)->node_zonelists;
2786 2787 2788 2789 2790

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

2791
	nr_reclaimed = do_try_to_free_pages(zonelist, &sc);
2792 2793 2794 2795

	trace_mm_vmscan_memcg_reclaim_end(nr_reclaimed);

	return nr_reclaimed;
2796 2797 2798
}
#endif

2799
static void age_active_anon(struct zone *zone, struct scan_control *sc)
2800
{
2801
	struct mem_cgroup *memcg;
2802

2803 2804 2805 2806 2807
	if (!total_swap_pages)
		return;

	memcg = mem_cgroup_iter(NULL, NULL, NULL);
	do {
2808
		struct lruvec *lruvec = mem_cgroup_zone_lruvec(zone, memcg);
2809

2810
		if (inactive_anon_is_low(lruvec))
2811
			shrink_active_list(SWAP_CLUSTER_MAX, lruvec,
2812
					   sc, LRU_ACTIVE_ANON);
2813 2814 2815

		memcg = mem_cgroup_iter(NULL, memcg, NULL);
	} while (memcg);
2816 2817
}

2818 2819 2820 2821 2822 2823 2824
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;

2825 2826
	if (IS_ENABLED(CONFIG_COMPACTION) && order &&
	    !compaction_suitable(zone, order))
2827 2828 2829 2830 2831
		return false;

	return true;
}

2832
/*
2833 2834 2835 2836 2837 2838 2839 2840 2841 2842
 * 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.
2843 2844 2845 2846
 * The choice of 25% is due to
 *   o a 16M DMA zone that is balanced will not balance a zone on any
 *     reasonable sized machine
 *   o On all other machines, the top zone must be at least a reasonable
L
Lucas De Marchi 已提交
2847
 *     percentage of the middle zones. For example, on 32-bit x86, highmem
2848 2849 2850 2851
 *     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.
 */
2852
static bool pgdat_balanced(pg_data_t *pgdat, int order, int classzone_idx)
2853
{
2854
	unsigned long managed_pages = 0;
2855
	unsigned long balanced_pages = 0;
2856 2857
	int i;

2858 2859 2860
	/* Check the watermark levels */
	for (i = 0; i <= classzone_idx; i++) {
		struct zone *zone = pgdat->node_zones + i;
2861

2862 2863 2864
		if (!populated_zone(zone))
			continue;

2865
		managed_pages += zone->managed_pages;
2866 2867 2868 2869 2870 2871 2872 2873

		/*
		 * 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!
		 */
2874
		if (!zone_reclaimable(zone)) {
2875
			balanced_pages += zone->managed_pages;
2876 2877 2878 2879
			continue;
		}

		if (zone_balanced(zone, order, 0, i))
2880
			balanced_pages += zone->managed_pages;
2881 2882 2883 2884 2885
		else if (!order)
			return false;
	}

	if (order)
2886
		return balanced_pages >= (managed_pages >> 2);
2887 2888
	else
		return true;
2889 2890
}

2891 2892 2893 2894 2895 2896 2897
/*
 * 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,
2898
					int classzone_idx)
2899 2900 2901
{
	/* If a direct reclaimer woke kswapd within HZ/10, it's premature */
	if (remaining)
2902 2903 2904 2905 2906 2907 2908 2909 2910 2911 2912 2913 2914 2915 2916
		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;
	}
2917

2918
	return pgdat_balanced(pgdat, order, classzone_idx);
2919 2920
}

2921 2922 2923
/*
 * kswapd shrinks the zone by the number of pages required to reach
 * the high watermark.
2924 2925
 *
 * Returns true if kswapd scanned at least the requested number of pages to
2926 2927
 * reclaim or if the lack of progress was due to pages under writeback.
 * This is used to determine if the scanning priority needs to be raised.
2928
 */
2929
static bool kswapd_shrink_zone(struct zone *zone,
2930
			       int classzone_idx,
2931
			       struct scan_control *sc,
2932 2933
			       unsigned long lru_pages,
			       unsigned long *nr_attempted)
2934
{
2935 2936
	int testorder = sc->order;
	unsigned long balance_gap;
2937 2938 2939 2940
	struct reclaim_state *reclaim_state = current->reclaim_state;
	struct shrink_control shrink = {
		.gfp_mask = sc->gfp_mask,
	};
2941
	bool lowmem_pressure;
2942 2943 2944

	/* Reclaim above the high watermark. */
	sc->nr_to_reclaim = max(SWAP_CLUSTER_MAX, high_wmark_pages(zone));
2945 2946 2947 2948 2949 2950 2951 2952 2953 2954 2955 2956 2957 2958 2959 2960 2961 2962

	/*
	 * Kswapd reclaims only single pages with compaction enabled. Trying
	 * too hard to reclaim until contiguous free pages have become
	 * available can hurt performance by evicting too much useful data
	 * from memory. Do not reclaim more than needed for compaction.
	 */
	if (IS_ENABLED(CONFIG_COMPACTION) && sc->order &&
			compaction_suitable(zone, sc->order) !=
				COMPACT_SKIPPED)
		testorder = 0;

	/*
	 * We put equal pressure on every zone, unless one zone has way too
	 * many pages free already. The "too many pages" is defined as the
	 * high wmark plus a "gap" where the gap is either the low
	 * watermark or 1% of the zone, whichever is smaller.
	 */
2963 2964
	balance_gap = min(low_wmark_pages(zone), DIV_ROUND_UP(
			zone->managed_pages, KSWAPD_ZONE_BALANCE_GAP_RATIO));
2965 2966 2967 2968 2969 2970 2971 2972 2973 2974

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

2975
	shrink_zone(zone, sc);
D
Dave Chinner 已提交
2976 2977
	nodes_clear(shrink.nodes_to_scan);
	node_set(zone_to_nid(zone), shrink.nodes_to_scan);
2978 2979

	reclaim_state->reclaimed_slab = 0;
2980
	shrink_slab(&shrink, sc->nr_scanned, lru_pages);
2981 2982
	sc->nr_reclaimed += reclaim_state->reclaimed_slab;

2983 2984 2985
	/* Account for the number of pages attempted to reclaim */
	*nr_attempted += sc->nr_to_reclaim;

2986 2987
	zone_clear_flag(zone, ZONE_WRITEBACK);

2988 2989 2990 2991 2992 2993
	/*
	 * If a zone reaches its high watermark, consider it to be no longer
	 * congested. It's possible there are dirty pages backed by congested
	 * BDIs but as pressure is relieved, speculatively avoid congestion
	 * waits.
	 */
2994
	if (zone_reclaimable(zone) &&
2995 2996 2997 2998 2999
	    zone_balanced(zone, testorder, 0, classzone_idx)) {
		zone_clear_flag(zone, ZONE_CONGESTED);
		zone_clear_flag(zone, ZONE_TAIL_LRU_DIRTY);
	}

3000
	return sc->nr_scanned >= sc->nr_to_reclaim;
3001 3002
}

L
Linus Torvalds 已提交
3003 3004
/*
 * For kswapd, balance_pgdat() will work across all this node's zones until
3005
 * they are all at high_wmark_pages(zone).
L
Linus Torvalds 已提交
3006
 *
3007
 * Returns the final order kswapd was reclaiming at
L
Linus Torvalds 已提交
3008 3009 3010 3011 3012 3013 3014 3015 3016 3017
 *
 * 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
3018 3019 3020 3021 3022
 * zones which have free_pages > high_wmark_pages(zone), but once a zone is
 * found to have free_pages <= high_wmark_pages(zone), we scan that zone and the
 * lower zones regardless of the number of free pages in the lower zones. This
 * interoperates with the page allocator fallback scheme to ensure that aging
 * of pages is balanced across the zones.
L
Linus Torvalds 已提交
3023
 */
3024
static unsigned long balance_pgdat(pg_data_t *pgdat, int order,
3025
							int *classzone_idx)
L
Linus Torvalds 已提交
3026 3027
{
	int i;
3028
	int end_zone = 0;	/* Inclusive.  0 = ZONE_DMA */
3029 3030
	unsigned long nr_soft_reclaimed;
	unsigned long nr_soft_scanned;
3031 3032
	struct scan_control sc = {
		.gfp_mask = GFP_KERNEL,
3033
		.priority = DEF_PRIORITY,
3034
		.may_unmap = 1,
3035
		.may_swap = 1,
3036
		.may_writepage = !laptop_mode,
A
Andy Whitcroft 已提交
3037
		.order = order,
3038
		.target_mem_cgroup = NULL,
3039
	};
3040
	count_vm_event(PAGEOUTRUN);
L
Linus Torvalds 已提交
3041

3042
	do {
L
Linus Torvalds 已提交
3043
		unsigned long lru_pages = 0;
3044
		unsigned long nr_attempted = 0;
3045
		bool raise_priority = true;
3046
		bool pgdat_needs_compaction = (order > 0);
3047 3048

		sc.nr_reclaimed = 0;
L
Linus Torvalds 已提交
3049

3050 3051 3052 3053 3054 3055
		/*
		 * Scan in the highmem->dma direction for the highest
		 * zone which needs scanning
		 */
		for (i = pgdat->nr_zones - 1; i >= 0; i--) {
			struct zone *zone = pgdat->node_zones + i;
L
Linus Torvalds 已提交
3056

3057 3058
			if (!populated_zone(zone))
				continue;
L
Linus Torvalds 已提交
3059

3060 3061
			if (sc.priority != DEF_PRIORITY &&
			    !zone_reclaimable(zone))
3062
				continue;
L
Linus Torvalds 已提交
3063

3064 3065 3066 3067
			/*
			 * Do some background aging of the anon list, to give
			 * pages a chance to be referenced before reclaiming.
			 */
3068
			age_active_anon(zone, &sc);
3069

3070 3071 3072 3073 3074 3075 3076 3077 3078 3079 3080
			/*
			 * 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;
			}

3081
			if (!zone_balanced(zone, order, 0, 0)) {
3082
				end_zone = i;
A
Andrew Morton 已提交
3083
				break;
3084
			} else {
3085 3086 3087 3088
				/*
				 * If balanced, clear the dirty and congested
				 * flags
				 */
3089
				zone_clear_flag(zone, ZONE_CONGESTED);
3090
				zone_clear_flag(zone, ZONE_TAIL_LRU_DIRTY);
L
Linus Torvalds 已提交
3091 3092
			}
		}
3093

3094
		if (i < 0)
A
Andrew Morton 已提交
3095 3096
			goto out;

L
Linus Torvalds 已提交
3097 3098 3099
		for (i = 0; i <= end_zone; i++) {
			struct zone *zone = pgdat->node_zones + i;

3100 3101 3102
			if (!populated_zone(zone))
				continue;

3103
			lru_pages += zone_reclaimable_pages(zone);
3104 3105 3106 3107 3108 3109 3110 3111 3112 3113 3114

			/*
			 * If any zone is currently balanced then kswapd will
			 * not call compaction as it is expected that the
			 * necessary pages are already available.
			 */
			if (pgdat_needs_compaction &&
					zone_watermark_ok(zone, order,
						low_wmark_pages(zone),
						*classzone_idx, 0))
				pgdat_needs_compaction = false;
L
Linus Torvalds 已提交
3115 3116
		}

3117 3118 3119 3120 3121 3122 3123
		/*
		 * If we're getting trouble reclaiming, start doing writepage
		 * even in laptop mode.
		 */
		if (sc.priority < DEF_PRIORITY - 2)
			sc.may_writepage = 1;

L
Linus Torvalds 已提交
3124 3125 3126 3127 3128 3129 3130 3131 3132 3133 3134 3135
		/*
		 * 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;

3136
			if (!populated_zone(zone))
L
Linus Torvalds 已提交
3137 3138
				continue;

3139 3140
			if (sc.priority != DEF_PRIORITY &&
			    !zone_reclaimable(zone))
L
Linus Torvalds 已提交
3141 3142 3143
				continue;

			sc.nr_scanned = 0;
3144

3145 3146 3147 3148 3149 3150 3151 3152 3153
			nr_soft_scanned = 0;
			/*
			 * Call soft limit reclaim before calling shrink_zone.
			 */
			nr_soft_reclaimed = mem_cgroup_soft_limit_reclaim(zone,
							order, sc.gfp_mask,
							&nr_soft_scanned);
			sc.nr_reclaimed += nr_soft_reclaimed;

3154
			/*
3155 3156 3157 3158
			 * There should be no need to raise the scanning
			 * priority if enough pages are already being scanned
			 * that that high watermark would be met at 100%
			 * efficiency.
3159
			 */
3160 3161 3162
			if (kswapd_shrink_zone(zone, end_zone, &sc,
					lru_pages, &nr_attempted))
				raise_priority = false;
L
Linus Torvalds 已提交
3163
		}
3164 3165 3166 3167 3168 3169 3170 3171 3172 3173

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

L
Linus Torvalds 已提交
3174
		/*
3175 3176 3177 3178 3179 3180
		 * Fragmentation may mean that the system cannot be rebalanced
		 * for high-order allocations in all zones. If twice the
		 * allocation size has been reclaimed and the zones are still
		 * not balanced then recheck the watermarks at order-0 to
		 * prevent kswapd reclaiming excessively. Assume that a
		 * process requested a high-order can direct reclaim/compact.
L
Linus Torvalds 已提交
3181
		 */
3182 3183
		if (order && sc.nr_reclaimed >= 2UL << order)
			order = sc.order = 0;
3184

3185 3186 3187
		/* Check if kswapd should be suspending */
		if (try_to_freeze() || kthread_should_stop())
			break;
3188

3189 3190 3191 3192 3193 3194 3195
		/*
		 * Compact if necessary and kswapd is reclaiming at least the
		 * high watermark number of pages as requsted
		 */
		if (pgdat_needs_compaction && sc.nr_reclaimed > nr_attempted)
			compact_pgdat(pgdat, order);

3196
		/*
3197 3198
		 * Raise priority if scanning rate is too low or there was no
		 * progress in reclaiming pages
3199
		 */
3200 3201
		if (raise_priority || !sc.nr_reclaimed)
			sc.priority--;
3202
	} while (sc.priority >= 1 &&
3203
		 !pgdat_balanced(pgdat, order, *classzone_idx));
L
Linus Torvalds 已提交
3204

3205
out:
3206
	/*
3207
	 * Return the order we were reclaiming at so prepare_kswapd_sleep()
3208 3209 3210 3211
	 * 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
	 */
3212
	*classzone_idx = end_zone;
3213
	return order;
L
Linus Torvalds 已提交
3214 3215
}

3216
static void kswapd_try_to_sleep(pg_data_t *pgdat, int order, int classzone_idx)
3217 3218 3219 3220 3221 3222 3223 3224 3225 3226
{
	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 */
3227
	if (prepare_kswapd_sleep(pgdat, order, remaining, classzone_idx)) {
3228 3229 3230 3231 3232 3233 3234 3235 3236
		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.
	 */
3237
	if (prepare_kswapd_sleep(pgdat, order, remaining, classzone_idx)) {
3238 3239 3240 3241 3242 3243 3244 3245 3246 3247 3248
		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);
3249

3250 3251 3252 3253 3254 3255 3256 3257
		/*
		 * 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);

3258 3259 3260
		if (!kthread_should_stop())
			schedule();

3261 3262 3263 3264 3265 3266 3267 3268 3269 3270
		set_pgdat_percpu_threshold(pgdat, calculate_pressure_threshold);
	} else {
		if (remaining)
			count_vm_event(KSWAPD_LOW_WMARK_HIT_QUICKLY);
		else
			count_vm_event(KSWAPD_HIGH_WMARK_HIT_QUICKLY);
	}
	finish_wait(&pgdat->kswapd_wait, &wait);
}

L
Linus Torvalds 已提交
3271 3272
/*
 * The background pageout daemon, started as a kernel thread
3273
 * from the init process.
L
Linus Torvalds 已提交
3274 3275 3276 3277 3278 3279 3280 3281 3282 3283 3284 3285
 *
 * 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)
{
3286
	unsigned long order, new_order;
3287
	unsigned balanced_order;
3288
	int classzone_idx, new_classzone_idx;
3289
	int balanced_classzone_idx;
L
Linus Torvalds 已提交
3290 3291
	pg_data_t *pgdat = (pg_data_t*)p;
	struct task_struct *tsk = current;
3292

L
Linus Torvalds 已提交
3293 3294 3295
	struct reclaim_state reclaim_state = {
		.reclaimed_slab = 0,
	};
3296
	const struct cpumask *cpumask = cpumask_of_node(pgdat->node_id);
L
Linus Torvalds 已提交
3297

3298 3299
	lockdep_set_current_reclaim_state(GFP_KERNEL);

R
Rusty Russell 已提交
3300
	if (!cpumask_empty(cpumask))
3301
		set_cpus_allowed_ptr(tsk, cpumask);
L
Linus Torvalds 已提交
3302 3303 3304 3305 3306 3307 3308 3309 3310 3311 3312 3313 3314 3315
	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).
	 */
3316
	tsk->flags |= PF_MEMALLOC | PF_SWAPWRITE | PF_KSWAPD;
3317
	set_freezable();
L
Linus Torvalds 已提交
3318

3319
	order = new_order = 0;
3320
	balanced_order = 0;
3321
	classzone_idx = new_classzone_idx = pgdat->nr_zones - 1;
3322
	balanced_classzone_idx = classzone_idx;
L
Linus Torvalds 已提交
3323
	for ( ; ; ) {
3324
		bool ret;
3325

3326 3327 3328 3329 3330
		/*
		 * 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
		 */
3331 3332
		if (balanced_classzone_idx >= new_classzone_idx &&
					balanced_order == new_order) {
3333 3334 3335 3336 3337 3338
			new_order = pgdat->kswapd_max_order;
			new_classzone_idx = pgdat->classzone_idx;
			pgdat->kswapd_max_order =  0;
			pgdat->classzone_idx = pgdat->nr_zones - 1;
		}

3339
		if (order < new_order || classzone_idx > new_classzone_idx) {
L
Linus Torvalds 已提交
3340 3341
			/*
			 * Don't sleep if someone wants a larger 'order'
3342
			 * allocation or has tigher zone constraints
L
Linus Torvalds 已提交
3343 3344
			 */
			order = new_order;
3345
			classzone_idx = new_classzone_idx;
L
Linus Torvalds 已提交
3346
		} else {
3347 3348
			kswapd_try_to_sleep(pgdat, balanced_order,
						balanced_classzone_idx);
L
Linus Torvalds 已提交
3349
			order = pgdat->kswapd_max_order;
3350
			classzone_idx = pgdat->classzone_idx;
3351 3352
			new_order = order;
			new_classzone_idx = classzone_idx;
3353
			pgdat->kswapd_max_order = 0;
3354
			pgdat->classzone_idx = pgdat->nr_zones - 1;
L
Linus Torvalds 已提交
3355 3356
		}

3357 3358 3359 3360 3361 3362 3363 3364
		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
		 */
3365 3366
		if (!ret) {
			trace_mm_vmscan_kswapd_wake(pgdat->node_id, order);
3367 3368 3369
			balanced_classzone_idx = classzone_idx;
			balanced_order = balance_pgdat(pgdat, order,
						&balanced_classzone_idx);
3370
		}
L
Linus Torvalds 已提交
3371
	}
3372

3373
	tsk->flags &= ~(PF_MEMALLOC | PF_SWAPWRITE | PF_KSWAPD);
3374
	current->reclaim_state = NULL;
3375 3376
	lockdep_clear_current_reclaim_state();

L
Linus Torvalds 已提交
3377 3378 3379 3380 3381 3382
	return 0;
}

/*
 * A zone is low on free memory, so wake its kswapd task to service it.
 */
3383
void wakeup_kswapd(struct zone *zone, int order, enum zone_type classzone_idx)
L
Linus Torvalds 已提交
3384 3385 3386
{
	pg_data_t *pgdat;

3387
	if (!populated_zone(zone))
L
Linus Torvalds 已提交
3388 3389
		return;

3390
	if (!cpuset_zone_allowed_hardwall(zone, GFP_KERNEL))
L
Linus Torvalds 已提交
3391
		return;
3392
	pgdat = zone->zone_pgdat;
3393
	if (pgdat->kswapd_max_order < order) {
L
Linus Torvalds 已提交
3394
		pgdat->kswapd_max_order = order;
3395 3396
		pgdat->classzone_idx = min(pgdat->classzone_idx, classzone_idx);
	}
3397
	if (!waitqueue_active(&pgdat->kswapd_wait))
L
Linus Torvalds 已提交
3398
		return;
3399
	if (zone_balanced(zone, order, 0, 0))
3400 3401 3402
		return;

	trace_mm_vmscan_wakeup_kswapd(pgdat->node_id, zone_idx(zone), order);
3403
	wake_up_interruptible(&pgdat->kswapd_wait);
L
Linus Torvalds 已提交
3404 3405
}

3406
#ifdef CONFIG_HIBERNATION
L
Linus Torvalds 已提交
3407
/*
3408
 * Try to free `nr_to_reclaim' of memory, system-wide, and return the number of
3409 3410 3411 3412 3413
 * freed pages.
 *
 * Rather than trying to age LRUs the aim is to preserve the overall
 * LRU order by reclaiming preferentially
 * inactive > active > active referenced > active mapped
L
Linus Torvalds 已提交
3414
 */
3415
unsigned long shrink_all_memory(unsigned long nr_to_reclaim)
L
Linus Torvalds 已提交
3416
{
3417 3418
	struct reclaim_state reclaim_state;
	struct scan_control sc = {
3419 3420 3421
		.gfp_mask = GFP_HIGHUSER_MOVABLE,
		.may_swap = 1,
		.may_unmap = 1,
3422
		.may_writepage = 1,
3423 3424 3425
		.nr_to_reclaim = nr_to_reclaim,
		.hibernation_mode = 1,
		.order = 0,
3426
		.priority = DEF_PRIORITY,
L
Linus Torvalds 已提交
3427
	};
3428
	struct zonelist *zonelist = node_zonelist(numa_node_id(), sc.gfp_mask);
3429 3430
	struct task_struct *p = current;
	unsigned long nr_reclaimed;
L
Linus Torvalds 已提交
3431

3432 3433 3434 3435
	p->flags |= PF_MEMALLOC;
	lockdep_set_current_reclaim_state(sc.gfp_mask);
	reclaim_state.reclaimed_slab = 0;
	p->reclaim_state = &reclaim_state;
3436

3437
	nr_reclaimed = do_try_to_free_pages(zonelist, &sc);
3438

3439 3440 3441
	p->reclaim_state = NULL;
	lockdep_clear_current_reclaim_state();
	p->flags &= ~PF_MEMALLOC;
3442

3443
	return nr_reclaimed;
L
Linus Torvalds 已提交
3444
}
3445
#endif /* CONFIG_HIBERNATION */
L
Linus Torvalds 已提交
3446 3447 3448 3449 3450

/* 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. */
3451 3452
static int cpu_callback(struct notifier_block *nfb, unsigned long action,
			void *hcpu)
L
Linus Torvalds 已提交
3453
{
3454
	int nid;
L
Linus Torvalds 已提交
3455

3456
	if (action == CPU_ONLINE || action == CPU_ONLINE_FROZEN) {
3457
		for_each_node_state(nid, N_MEMORY) {
3458
			pg_data_t *pgdat = NODE_DATA(nid);
3459 3460 3461
			const struct cpumask *mask;

			mask = cpumask_of_node(pgdat->node_id);
3462

3463
			if (cpumask_any_and(cpu_online_mask, mask) < nr_cpu_ids)
L
Linus Torvalds 已提交
3464
				/* One of our CPUs online: restore mask */
3465
				set_cpus_allowed_ptr(pgdat->kswapd, mask);
L
Linus Torvalds 已提交
3466 3467 3468 3469 3470
		}
	}
	return NOTIFY_OK;
}

3471 3472 3473 3474 3475 3476 3477 3478 3479 3480 3481 3482 3483 3484 3485 3486
/*
 * 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);
3487 3488
		pr_err("Failed to start kswapd on node %d\n", nid);
		ret = PTR_ERR(pgdat->kswapd);
3489
		pgdat->kswapd = NULL;
3490 3491 3492 3493
	}
	return ret;
}

3494
/*
3495
 * Called by memory hotplug when all memory in a node is offlined.  Caller must
3496
 * hold mem_hotplug_begin/end().
3497 3498 3499 3500 3501
 */
void kswapd_stop(int nid)
{
	struct task_struct *kswapd = NODE_DATA(nid)->kswapd;

3502
	if (kswapd) {
3503
		kthread_stop(kswapd);
3504 3505
		NODE_DATA(nid)->kswapd = NULL;
	}
3506 3507
}

L
Linus Torvalds 已提交
3508 3509
static int __init kswapd_init(void)
{
3510
	int nid;
3511

L
Linus Torvalds 已提交
3512
	swap_setup();
3513
	for_each_node_state(nid, N_MEMORY)
3514
 		kswapd_run(nid);
L
Linus Torvalds 已提交
3515 3516 3517 3518 3519
	hotcpu_notifier(cpu_callback, 0);
	return 0;
}

module_init(kswapd_init)
3520 3521 3522 3523 3524 3525 3526 3527 3528 3529

#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;

3530
#define RECLAIM_OFF 0
3531
#define RECLAIM_ZONE (1<<0)	/* Run shrink_inactive_list on the zone */
3532 3533 3534
#define RECLAIM_WRITE (1<<1)	/* Writeout pages during reclaim */
#define RECLAIM_SWAP (1<<2)	/* Swap pages out during reclaim */

3535 3536 3537 3538 3539 3540 3541
/*
 * 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

3542 3543 3544 3545 3546 3547
/*
 * Percentage of pages in a zone that must be unmapped for zone_reclaim to
 * occur.
 */
int sysctl_min_unmapped_ratio = 1;

3548 3549 3550 3551 3552 3553
/*
 * 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;

3554 3555 3556 3557 3558 3559 3560 3561 3562 3563 3564 3565 3566 3567 3568 3569 3570 3571 3572 3573 3574 3575 3576 3577 3578 3579 3580 3581 3582 3583 3584 3585 3586 3587 3588 3589 3590 3591 3592 3593 3594 3595
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;
}

3596 3597 3598
/*
 * Try to free up some pages from this zone through reclaim.
 */
3599
static int __zone_reclaim(struct zone *zone, gfp_t gfp_mask, unsigned int order)
3600
{
3601
	/* Minimum pages needed in order to stay on node */
3602
	const unsigned long nr_pages = 1 << order;
3603 3604
	struct task_struct *p = current;
	struct reclaim_state reclaim_state;
3605 3606
	struct scan_control sc = {
		.may_writepage = !!(zone_reclaim_mode & RECLAIM_WRITE),
3607
		.may_unmap = !!(zone_reclaim_mode & RECLAIM_SWAP),
3608
		.may_swap = 1,
3609
		.nr_to_reclaim = max(nr_pages, SWAP_CLUSTER_MAX),
3610
		.gfp_mask = (gfp_mask = memalloc_noio_flags(gfp_mask)),
3611
		.order = order,
3612
		.priority = ZONE_RECLAIM_PRIORITY,
3613
	};
3614 3615 3616
	struct shrink_control shrink = {
		.gfp_mask = sc.gfp_mask,
	};
3617
	unsigned long nr_slab_pages0, nr_slab_pages1;
3618 3619

	cond_resched();
3620 3621 3622 3623 3624 3625
	/*
	 * 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;
3626
	lockdep_set_current_reclaim_state(gfp_mask);
3627 3628
	reclaim_state.reclaimed_slab = 0;
	p->reclaim_state = &reclaim_state;
3629

3630
	if (zone_pagecache_reclaimable(zone) > zone->min_unmapped_pages) {
3631 3632 3633 3634 3635
		/*
		 * Free memory by calling shrink zone with increasing
		 * priorities until we have enough memory freed.
		 */
		do {
3636 3637
			shrink_zone(zone, &sc);
		} while (sc.nr_reclaimed < nr_pages && --sc.priority >= 0);
3638
	}
3639

3640 3641
	nr_slab_pages0 = zone_page_state(zone, NR_SLAB_RECLAIMABLE);
	if (nr_slab_pages0 > zone->min_slab_pages) {
3642
		/*
3643
		 * shrink_slab() does not currently allow us to determine how
3644 3645 3646 3647
		 * 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.
3648
		 */
D
Dave Chinner 已提交
3649 3650
		nodes_clear(shrink.nodes_to_scan);
		node_set(zone_to_nid(zone), shrink.nodes_to_scan);
3651 3652 3653 3654
		for (;;) {
			unsigned long lru_pages = zone_reclaimable_pages(zone);

			/* No reclaimable slab or very low memory pressure */
3655
			if (!shrink_slab(&shrink, sc.nr_scanned, lru_pages))
3656 3657 3658 3659 3660 3661 3662 3663
				break;

			/* Freed enough memory */
			nr_slab_pages1 = zone_page_state(zone,
							NR_SLAB_RECLAIMABLE);
			if (nr_slab_pages1 + nr_pages <= nr_slab_pages0)
				break;
		}
3664 3665 3666 3667 3668

		/*
		 * Update nr_reclaimed by the number of slab pages we
		 * reclaimed from this zone.
		 */
3669 3670 3671
		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;
3672 3673
	}

3674
	p->reclaim_state = NULL;
3675
	current->flags &= ~(PF_MEMALLOC | PF_SWAPWRITE);
3676
	lockdep_clear_current_reclaim_state();
3677
	return sc.nr_reclaimed >= nr_pages;
3678
}
3679 3680 3681 3682

int zone_reclaim(struct zone *zone, gfp_t gfp_mask, unsigned int order)
{
	int node_id;
3683
	int ret;
3684 3685

	/*
3686 3687
	 * Zone reclaim reclaims unmapped file backed pages and
	 * slab pages if we are over the defined limits.
3688
	 *
3689 3690 3691 3692 3693
	 * 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.
3694
	 */
3695 3696
	if (zone_pagecache_reclaimable(zone) <= zone->min_unmapped_pages &&
	    zone_page_state(zone, NR_SLAB_RECLAIMABLE) <= zone->min_slab_pages)
3697
		return ZONE_RECLAIM_FULL;
3698

3699
	if (!zone_reclaimable(zone))
3700
		return ZONE_RECLAIM_FULL;
3701

3702
	/*
3703
	 * Do not scan if the allocation should not be delayed.
3704
	 */
3705
	if (!(gfp_mask & __GFP_WAIT) || (current->flags & PF_MEMALLOC))
3706
		return ZONE_RECLAIM_NOSCAN;
3707 3708 3709 3710 3711 3712 3713

	/*
	 * 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.
	 */
3714
	node_id = zone_to_nid(zone);
3715
	if (node_state(node_id, N_CPU) && node_id != numa_node_id())
3716
		return ZONE_RECLAIM_NOSCAN;
3717 3718

	if (zone_test_and_set_flag(zone, ZONE_RECLAIM_LOCKED))
3719 3720
		return ZONE_RECLAIM_NOSCAN;

3721 3722 3723
	ret = __zone_reclaim(zone, gfp_mask, order);
	zone_clear_flag(zone, ZONE_RECLAIM_LOCKED);

3724 3725 3726
	if (!ret)
		count_vm_event(PGSCAN_ZONE_RECLAIM_FAILED);

3727
	return ret;
3728
}
3729
#endif
L
Lee Schermerhorn 已提交
3730 3731 3732 3733 3734 3735

/*
 * 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
3736
 * lists vs unevictable list.
L
Lee Schermerhorn 已提交
3737 3738
 *
 * Reasons page might not be evictable:
3739
 * (1) page's mapping marked unevictable
N
Nick Piggin 已提交
3740
 * (2) page is part of an mlocked VMA
3741
 *
L
Lee Schermerhorn 已提交
3742
 */
3743
int page_evictable(struct page *page)
L
Lee Schermerhorn 已提交
3744
{
3745
	return !mapping_unevictable(page_mapping(page)) && !PageMlocked(page);
L
Lee Schermerhorn 已提交
3746
}
3747

3748
#ifdef CONFIG_SHMEM
3749
/**
3750 3751 3752
 * 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
3753
 *
3754
 * Checks pages for evictability and moves them to the appropriate lru list.
3755 3756
 *
 * This function is only used for SysV IPC SHM_UNLOCK.
3757
 */
3758
void check_move_unevictable_pages(struct page **pages, int nr_pages)
3759
{
3760
	struct lruvec *lruvec;
3761 3762 3763 3764
	struct zone *zone = NULL;
	int pgscanned = 0;
	int pgrescued = 0;
	int i;
3765

3766 3767 3768
	for (i = 0; i < nr_pages; i++) {
		struct page *page = pages[i];
		struct zone *pagezone;
3769

3770 3771 3772 3773 3774 3775 3776 3777
		pgscanned++;
		pagezone = page_zone(page);
		if (pagezone != zone) {
			if (zone)
				spin_unlock_irq(&zone->lru_lock);
			zone = pagezone;
			spin_lock_irq(&zone->lru_lock);
		}
3778
		lruvec = mem_cgroup_page_lruvec(page, zone);
3779

3780 3781
		if (!PageLRU(page) || !PageUnevictable(page))
			continue;
3782

3783
		if (page_evictable(page)) {
3784 3785
			enum lru_list lru = page_lru_base_type(page);

3786
			VM_BUG_ON_PAGE(PageActive(page), page);
3787
			ClearPageUnevictable(page);
3788 3789
			del_page_from_lru_list(page, lruvec, LRU_UNEVICTABLE);
			add_page_to_lru_list(page, lruvec, lru);
3790
			pgrescued++;
3791
		}
3792
	}
3793

3794 3795 3796 3797
	if (zone) {
		__count_vm_events(UNEVICTABLE_PGRESCUED, pgrescued);
		__count_vm_events(UNEVICTABLE_PGSCANNED, pgscanned);
		spin_unlock_irq(&zone->lru_lock);
3798 3799
	}
}
3800
#endif /* CONFIG_SHMEM */
3801

3802
static void warn_scan_unevictable_pages(void)
3803
{
3804
	printk_once(KERN_WARNING
3805
		    "%s: The scan_unevictable_pages sysctl/node-interface has been "
3806
		    "disabled for lack of a legitimate use case.  If you have "
3807 3808
		    "one, please send an email to linux-mm@kvack.org.\n",
		    current->comm);
3809 3810 3811 3812 3813 3814 3815 3816 3817
}

/*
 * 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,
3818
			   void __user *buffer,
3819 3820
			   size_t *length, loff_t *ppos)
{
3821
	warn_scan_unevictable_pages();
3822
	proc_doulongvec_minmax(table, write, buffer, length, ppos);
3823 3824 3825 3826
	scan_unevictable_pages = 0;
	return 0;
}

3827
#ifdef CONFIG_NUMA
3828 3829 3830 3831 3832
/*
 * per node 'scan_unevictable_pages' attribute.  On demand re-scan of
 * a specified node's per zone unevictable lists for evictable pages.
 */

3833 3834
static ssize_t read_scan_unevictable_node(struct device *dev,
					  struct device_attribute *attr,
3835 3836
					  char *buf)
{
3837
	warn_scan_unevictable_pages();
3838 3839 3840
	return sprintf(buf, "0\n");	/* always zero; should fit... */
}

3841 3842
static ssize_t write_scan_unevictable_node(struct device *dev,
					   struct device_attribute *attr,
3843 3844
					const char *buf, size_t count)
{
3845
	warn_scan_unevictable_pages();
3846 3847 3848 3849
	return 1;
}


3850
static DEVICE_ATTR(scan_unevictable_pages, S_IRUGO | S_IWUSR,
3851 3852 3853 3854 3855
			read_scan_unevictable_node,
			write_scan_unevictable_node);

int scan_unevictable_register_node(struct node *node)
{
3856
	return device_create_file(&node->dev, &dev_attr_scan_unevictable_pages);
3857 3858 3859 3860
}

void scan_unevictable_unregister_node(struct node *node)
{
3861
	device_remove_file(&node->dev, &dev_attr_scan_unevictable_pages);
3862
}
3863
#endif