vmscan.c 107.6 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13
/*
 *  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.
 */

14 15
#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt

L
Linus Torvalds 已提交
16 17
#include <linux/mm.h>
#include <linux/module.h>
18
#include <linux/gfp.h>
L
Linus Torvalds 已提交
19 20 21 22 23
#include <linux/kernel_stat.h>
#include <linux/swap.h>
#include <linux/pagemap.h>
#include <linux/init.h>
#include <linux/highmem.h>
24
#include <linux/vmpressure.h>
25
#include <linux/vmstat.h>
L
Linus Torvalds 已提交
26 27 28 29 30 31 32 33 34 35 36
#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>
37
#include <linux/compaction.h>
L
Linus Torvalds 已提交
38 39
#include <linux/notifier.h>
#include <linux/rwsem.h>
40
#include <linux/delay.h>
41
#include <linux/kthread.h>
42
#include <linux/freezer.h>
43
#include <linux/memcontrol.h>
44
#include <linux/delayacct.h>
45
#include <linux/sysctl.h>
46
#include <linux/oom.h>
47
#include <linux/prefetch.h>
48
#include <linux/printk.h>
L
Linus Torvalds 已提交
49 50 51 52 53

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

#include <linux/swapops.h>
54
#include <linux/balloon_compaction.h>
L
Linus Torvalds 已提交
55

56 57
#include "internal.h"

58 59 60
#define CREATE_TRACE_POINTS
#include <trace/events/vmscan.h>

L
Linus Torvalds 已提交
61
struct scan_control {
62 63 64
	/* How many pages shrink_list() should reclaim */
	unsigned long nr_to_reclaim;

L
Linus Torvalds 已提交
65
	/* This context's GFP mask */
A
Al Viro 已提交
66
	gfp_t gfp_mask;
L
Linus Torvalds 已提交
67

68
	/* Allocation order */
A
Andy Whitcroft 已提交
69
	int order;
70

71 72 73 74 75
	/*
	 * Nodemask of nodes allowed by the caller. If NULL, all nodes
	 * are scanned.
	 */
	nodemask_t	*nodemask;
76

77 78 79 80 81
	/*
	 * 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;
82

83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103
	/* Scan (total_size >> priority) pages at once */
	int priority;

	unsigned int may_writepage:1;

	/* Can mapped pages be reclaimed? */
	unsigned int may_unmap:1;

	/* Can pages be swapped as part of reclaim? */
	unsigned int may_swap:1;

	unsigned int hibernation_mode:1;

	/* One of the zones is ready for compaction */
	unsigned int compaction_ready:1;

	/* Incremented by the number of inactive pages that were scanned */
	unsigned long nr_scanned;

	/* Number of pages freed so far during a call to shrink_zones() */
	unsigned long nr_reclaimed;
L
Linus Torvalds 已提交
104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139
};

#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;
140 141 142 143 144
/*
 * The total number of pages which are beyond the high watermark within all
 * zones.
 */
unsigned long vm_total_pages;
L
Linus Torvalds 已提交
145 146 147 148

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

A
Andrew Morton 已提交
149
#ifdef CONFIG_MEMCG
150 151
static bool global_reclaim(struct scan_control *sc)
{
152
	return !sc->target_mem_cgroup;
153
}
154
#else
155 156 157 158
static bool global_reclaim(struct scan_control *sc)
{
	return true;
}
159 160
#endif

161
static unsigned long zone_reclaimable_pages(struct zone *zone)
162 163 164 165 166 167 168 169 170 171 172 173 174 175 176
{
	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)
{
177 178
	return zone_page_state(zone, NR_PAGES_SCANNED) <
		zone_reclaimable_pages(zone) * 6;
179 180
}

181
static unsigned long get_lru_size(struct lruvec *lruvec, enum lru_list lru)
182
{
183
	if (!mem_cgroup_disabled())
184
		return mem_cgroup_get_lru_size(lruvec, lru);
185

186
	return zone_page_state(lruvec_zone(lruvec), NR_LRU_BASE + lru);
187 188
}

L
Linus Torvalds 已提交
189
/*
G
Glauber Costa 已提交
190
 * Add a shrinker callback to be called from the vm.
L
Linus Torvalds 已提交
191
 */
G
Glauber Costa 已提交
192
int register_shrinker(struct shrinker *shrinker)
L
Linus Torvalds 已提交
193
{
G
Glauber Costa 已提交
194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210
	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;

211 212 213
	down_write(&shrinker_rwsem);
	list_add_tail(&shrinker->list, &shrinker_list);
	up_write(&shrinker_rwsem);
G
Glauber Costa 已提交
214
	return 0;
L
Linus Torvalds 已提交
215
}
216
EXPORT_SYMBOL(register_shrinker);
L
Linus Torvalds 已提交
217 218 219 220

/*
 * Remove one
 */
221
void unregister_shrinker(struct shrinker *shrinker)
L
Linus Torvalds 已提交
222 223 224 225
{
	down_write(&shrinker_rwsem);
	list_del(&shrinker->list);
	up_write(&shrinker_rwsem);
226
	kfree(shrinker->nr_deferred);
L
Linus Torvalds 已提交
227
}
228
EXPORT_SYMBOL(unregister_shrinker);
L
Linus Torvalds 已提交
229 230

#define SHRINK_BATCH 128
G
Glauber Costa 已提交
231

232 233 234 235
static unsigned long shrink_slabs(struct shrink_control *shrinkctl,
				  struct shrinker *shrinker,
				  unsigned long nr_scanned,
				  unsigned long nr_eligible)
G
Glauber Costa 已提交
236 237 238 239
{
	unsigned long freed = 0;
	unsigned long long delta;
	long total_scan;
240
	long freeable;
G
Glauber Costa 已提交
241 242 243 244 245 246
	long nr;
	long new_nr;
	int nid = shrinkctl->nid;
	long batch_size = shrinker->batch ? shrinker->batch
					  : SHRINK_BATCH;

247 248
	freeable = shrinker->count_objects(shrinker, shrinkctl);
	if (freeable == 0)
G
Glauber Costa 已提交
249 250 251 252 253 254 255 256 257 258
		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;
259
	delta = (4 * nr_scanned) / shrinker->seeks;
260
	delta *= freeable;
261
	do_div(delta, nr_eligible + 1);
G
Glauber Costa 已提交
262 263
	total_scan += delta;
	if (total_scan < 0) {
264
		pr_err("shrink_slab: %pF negative objects to delete nr=%ld\n",
D
Dave Chinner 已提交
265
		       shrinker->scan_objects, total_scan);
266
		total_scan = freeable;
G
Glauber Costa 已提交
267 268 269 270 271 272 273 274
	}

	/*
	 * 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 >>>
275
	 * freeable. This is bad for sustaining a working set in
G
Glauber Costa 已提交
276 277 278 279 280
	 * memory.
	 *
	 * Hence only allow the shrinker to scan the entire cache when
	 * a large delta change is calculated directly.
	 */
281 282
	if (delta < freeable / 4)
		total_scan = min(total_scan, freeable / 2);
G
Glauber Costa 已提交
283 284 285 286 287 288

	/*
	 * Avoid risking looping forever due to too large nr value:
	 * never try to free more than twice the estimate number of
	 * freeable entries.
	 */
289 290
	if (total_scan > freeable * 2)
		total_scan = freeable * 2;
G
Glauber Costa 已提交
291 292

	trace_mm_shrink_slab_start(shrinker, shrinkctl, nr,
293 294
				   nr_scanned, nr_eligible,
				   freeable, delta, total_scan);
G
Glauber Costa 已提交
295

296 297 298 299 300 301 302 303 304 305 306
	/*
	 * 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
307
	 * than the total number of objects on slab (freeable), we must be
308 309 310 311
	 * scanning at high prio and therefore should try to reclaim as much as
	 * possible.
	 */
	while (total_scan >= batch_size ||
312
	       total_scan >= freeable) {
D
Dave Chinner 已提交
313
		unsigned long ret;
314
		unsigned long nr_to_scan = min(batch_size, total_scan);
G
Glauber Costa 已提交
315

316
		shrinkctl->nr_to_scan = nr_to_scan;
D
Dave Chinner 已提交
317 318 319 320
		ret = shrinker->scan_objects(shrinker, shrinkctl);
		if (ret == SHRINK_STOP)
			break;
		freed += ret;
G
Glauber Costa 已提交
321

322 323
		count_vm_events(SLABS_SCANNED, nr_to_scan);
		total_scan -= nr_to_scan;
G
Glauber Costa 已提交
324 325 326 327 328 329 330 331 332 333 334 335 336 337 338

		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]);

339
	trace_mm_shrink_slab_end(shrinker, nid, freed, nr, new_nr, total_scan);
G
Glauber Costa 已提交
340
	return freed;
341 342
}

343 344 345 346 347 348
/**
 * shrink_node_slabs - shrink slab caches of a given node
 * @gfp_mask: allocation context
 * @nid: node whose slab caches to target
 * @nr_scanned: pressure numerator
 * @nr_eligible: pressure denominator
L
Linus Torvalds 已提交
349
 *
350
 * Call the shrink functions to age shrinkable caches.
L
Linus Torvalds 已提交
351
 *
352 353
 * @nid is passed along to shrinkers with SHRINKER_NUMA_AWARE set,
 * unaware shrinkers will receive a node id of 0 instead.
L
Linus Torvalds 已提交
354
 *
355 356 357 358 359 360 361
 * @nr_scanned and @nr_eligible form a ratio that indicate how much of
 * the available objects should be scanned.  Page reclaim for example
 * passes the number of pages scanned and the number of pages on the
 * LRU lists that it considered on @nid, plus a bias in @nr_scanned
 * when it encountered mapped pages.  The ratio is further biased by
 * the ->seeks setting of the shrink function, which indicates the
 * cost to recreate an object relative to that of an LRU page.
362
 *
363
 * Returns the number of reclaimed slab objects.
L
Linus Torvalds 已提交
364
 */
365 366 367
unsigned long shrink_node_slabs(gfp_t gfp_mask, int nid,
				unsigned long nr_scanned,
				unsigned long nr_eligible)
L
Linus Torvalds 已提交
368 369
{
	struct shrinker *shrinker;
D
Dave Chinner 已提交
370
	unsigned long freed = 0;
L
Linus Torvalds 已提交
371

372 373
	if (nr_scanned == 0)
		nr_scanned = SWAP_CLUSTER_MAX;
L
Linus Torvalds 已提交
374

375
	if (!down_read_trylock(&shrinker_rwsem)) {
D
Dave Chinner 已提交
376 377 378 379 380 381 382
		/*
		 * 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;
383 384
		goto out;
	}
L
Linus Torvalds 已提交
385 386

	list_for_each_entry(shrinker, &shrinker_list, list) {
387 388 389 390
		struct shrink_control sc = {
			.gfp_mask = gfp_mask,
			.nid = nid,
		};
391

392 393
		if (!(shrinker->flags & SHRINKER_NUMA_AWARE))
			sc.nid = 0;
L
Linus Torvalds 已提交
394

395
		freed += shrink_slabs(&sc, shrinker, nr_scanned, nr_eligible);
L
Linus Torvalds 已提交
396
	}
397

L
Linus Torvalds 已提交
398
	up_read(&shrinker_rwsem);
399 400
out:
	cond_resched();
D
Dave Chinner 已提交
401
	return freed;
L
Linus Torvalds 已提交
402 403 404 405
}

static inline int is_page_cache_freeable(struct page *page)
{
406 407 408 409 410
	/*
	 * 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.
	 */
411
	return page_count(page) - page_has_private(page) == 2;
L
Linus Torvalds 已提交
412 413
}

414 415
static int may_write_to_queue(struct backing_dev_info *bdi,
			      struct scan_control *sc)
L
Linus Torvalds 已提交
416
{
417
	if (current->flags & PF_SWAPWRITE)
L
Linus Torvalds 已提交
418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440
		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
441
	lock_page(page);
442 443
	if (page_mapping(page) == mapping)
		mapping_set_error(mapping, error);
L
Linus Torvalds 已提交
444 445 446
	unlock_page(page);
}

447 448 449 450 451 452 453 454 455 456 457 458
/* 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
459
/*
A
Andrew Morton 已提交
460 461
 * pageout is called by shrink_page_list() for each dirty page.
 * Calls ->writepage().
L
Linus Torvalds 已提交
462
 */
463
static pageout_t pageout(struct page *page, struct address_space *mapping,
464
			 struct scan_control *sc)
L
Linus Torvalds 已提交
465 466 467 468 469 470 471 472
{
	/*
	 * 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.
	 *
473
	 * If this process is currently in __generic_file_write_iter() against
L
Linus Torvalds 已提交
474 475 476 477 478 479 480 481 482 483 484 485 486 487 488
	 * 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.
		 */
489
		if (page_has_private(page)) {
L
Linus Torvalds 已提交
490 491
			if (try_to_free_buffers(page)) {
				ClearPageDirty(page);
492
				pr_info("%s: orphaned page\n", __func__);
L
Linus Torvalds 已提交
493 494 495 496 497 498 499
				return PAGE_CLEAN;
			}
		}
		return PAGE_KEEP;
	}
	if (mapping->a_ops->writepage == NULL)
		return PAGE_ACTIVATE;
500
	if (!may_write_to_queue(mapping->backing_dev_info, sc))
L
Linus Torvalds 已提交
501 502 503 504 505 506 507
		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,
508 509
			.range_start = 0,
			.range_end = LLONG_MAX,
L
Linus Torvalds 已提交
510 511 512 513 514 515 516
			.for_reclaim = 1,
		};

		SetPageReclaim(page);
		res = mapping->a_ops->writepage(page, &wbc);
		if (res < 0)
			handle_write_error(mapping, page, res);
517
		if (res == AOP_WRITEPAGE_ACTIVATE) {
L
Linus Torvalds 已提交
518 519 520
			ClearPageReclaim(page);
			return PAGE_ACTIVATE;
		}
521

L
Linus Torvalds 已提交
522 523 524 525
		if (!PageWriteback(page)) {
			/* synchronous write or broken a_ops? */
			ClearPageReclaim(page);
		}
M
Mel Gorman 已提交
526
		trace_mm_vmscan_writepage(page, trace_reclaim_flags(page));
527
		inc_zone_page_state(page, NR_VMSCAN_WRITE);
L
Linus Torvalds 已提交
528 529 530 531 532 533
		return PAGE_SUCCESS;
	}

	return PAGE_CLEAN;
}

534
/*
N
Nick Piggin 已提交
535 536
 * Same as remove_mapping, but if the page is removed from the mapping, it
 * gets returned with a refcount of 0.
537
 */
538 539
static int __remove_mapping(struct address_space *mapping, struct page *page,
			    bool reclaimed)
540
{
541 542
	BUG_ON(!PageLocked(page));
	BUG_ON(mapping != page_mapping(page));
543

N
Nick Piggin 已提交
544
	spin_lock_irq(&mapping->tree_lock);
545
	/*
N
Nick Piggin 已提交
546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568
	 * 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.
569
	 */
N
Nick Piggin 已提交
570
	if (!page_freeze_refs(page, 2))
571
		goto cannot_free;
N
Nick Piggin 已提交
572 573 574
	/* note: atomic_cmpxchg in page_freeze_refs provides the smp_rmb */
	if (unlikely(PageDirty(page))) {
		page_unfreeze_refs(page, 2);
575
		goto cannot_free;
N
Nick Piggin 已提交
576
	}
577 578 579

	if (PageSwapCache(page)) {
		swp_entry_t swap = { .val = page_private(page) };
580
		mem_cgroup_swapout(page, swap);
581
		__delete_from_swap_cache(page);
N
Nick Piggin 已提交
582
		spin_unlock_irq(&mapping->tree_lock);
583
		swapcache_free(swap);
N
Nick Piggin 已提交
584
	} else {
585
		void (*freepage)(struct page *);
586
		void *shadow = NULL;
587 588

		freepage = mapping->a_ops->freepage;
589 590 591 592 593 594 595 596 597 598 599 600 601 602
		/*
		 * 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
603
		spin_unlock_irq(&mapping->tree_lock);
604 605 606

		if (freepage != NULL)
			freepage(page);
607 608 609 610 611
	}

	return 1;

cannot_free:
N
Nick Piggin 已提交
612
	spin_unlock_irq(&mapping->tree_lock);
613 614 615
	return 0;
}

N
Nick Piggin 已提交
616 617 618 619 620 621 622 623
/*
 * 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)
{
624
	if (__remove_mapping(mapping, page, false)) {
N
Nick Piggin 已提交
625 626 627 628 629 630 631 632 633 634 635
		/*
		 * 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
636 637 638 639 640 641 642 643 644 645 646
/**
 * 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)
{
647
	bool is_unevictable;
648
	int was_unevictable = PageUnevictable(page);
L
Lee Schermerhorn 已提交
649

650
	VM_BUG_ON_PAGE(PageLRU(page), page);
L
Lee Schermerhorn 已提交
651 652 653 654

redo:
	ClearPageUnevictable(page);

655
	if (page_evictable(page)) {
L
Lee Schermerhorn 已提交
656 657 658 659 660 661
		/*
		 * 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.
		 */
662
		is_unevictable = false;
663
		lru_cache_add(page);
L
Lee Schermerhorn 已提交
664 665 666 667 668
	} else {
		/*
		 * Put unevictable pages directly on zone's unevictable
		 * list.
		 */
669
		is_unevictable = true;
L
Lee Schermerhorn 已提交
670
		add_page_to_unevictable_list(page);
671
		/*
672 673 674
		 * 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
675
		 * isolation/check_move_unevictable_pages,
676
		 * we see PG_mlocked/AS_UNEVICTABLE cleared below and move
677 678
		 * the page back to the evictable list.
		 *
679
		 * The other side is TestClearPageMlocked() or shmem_lock().
680 681
		 */
		smp_mb();
L
Lee Schermerhorn 已提交
682 683 684 685 686 687 688
	}

	/*
	 * 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.
	 */
689
	if (is_unevictable && page_evictable(page)) {
L
Lee Schermerhorn 已提交
690 691 692 693 694 695 696 697 698 699
		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.
		 */
	}

700
	if (was_unevictable && !is_unevictable)
701
		count_vm_event(UNEVICTABLE_PGRESCUED);
702
	else if (!was_unevictable && is_unevictable)
703 704
		count_vm_event(UNEVICTABLE_PGCULLED);

L
Lee Schermerhorn 已提交
705 706 707
	put_page(page);		/* drop ref from isolate */
}

708 709 710
enum page_references {
	PAGEREF_RECLAIM,
	PAGEREF_RECLAIM_CLEAN,
711
	PAGEREF_KEEP,
712 713 714 715 716 717
	PAGEREF_ACTIVATE,
};

static enum page_references page_check_references(struct page *page,
						  struct scan_control *sc)
{
718
	int referenced_ptes, referenced_page;
719 720
	unsigned long vm_flags;

721 722
	referenced_ptes = page_referenced(page, 1, sc->target_mem_cgroup,
					  &vm_flags);
723
	referenced_page = TestClearPageReferenced(page);
724 725 726 727 728 729 730 731

	/*
	 * 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;

732
	if (referenced_ptes) {
733
		if (PageSwapBacked(page))
734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750
			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);

751
		if (referenced_page || referenced_ptes > 1)
752 753
			return PAGEREF_ACTIVATE;

754 755 756 757 758 759
		/*
		 * Activate file-backed executable pages after first usage.
		 */
		if (vm_flags & VM_EXEC)
			return PAGEREF_ACTIVATE;

760 761
		return PAGEREF_KEEP;
	}
762 763

	/* Reclaim if clean, defer dirty pages to writeback */
764
	if (referenced_page && !PageSwapBacked(page))
765 766 767
		return PAGEREF_RECLAIM_CLEAN;

	return PAGEREF_RECLAIM;
768 769
}

770 771 772 773
/* Check if a page is dirty or under writeback */
static void page_check_dirty_writeback(struct page *page,
				       bool *dirty, bool *writeback)
{
774 775
	struct address_space *mapping;

776 777 778 779 780 781 782 783 784 785 786 787 788
	/*
	 * 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);
789 790 791 792 793 794 795 796

	/* 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);
797 798
}

L
Linus Torvalds 已提交
799
/*
A
Andrew Morton 已提交
800
 * shrink_page_list() returns the number of reclaimed pages
L
Linus Torvalds 已提交
801
 */
A
Andrew Morton 已提交
802
static unsigned long shrink_page_list(struct list_head *page_list,
803
				      struct zone *zone,
804
				      struct scan_control *sc,
805
				      enum ttu_flags ttu_flags,
806
				      unsigned long *ret_nr_dirty,
807
				      unsigned long *ret_nr_unqueued_dirty,
808
				      unsigned long *ret_nr_congested,
809
				      unsigned long *ret_nr_writeback,
810
				      unsigned long *ret_nr_immediate,
811
				      bool force_reclaim)
L
Linus Torvalds 已提交
812 813
{
	LIST_HEAD(ret_pages);
814
	LIST_HEAD(free_pages);
L
Linus Torvalds 已提交
815
	int pgactivate = 0;
816
	unsigned long nr_unqueued_dirty = 0;
817 818
	unsigned long nr_dirty = 0;
	unsigned long nr_congested = 0;
819
	unsigned long nr_reclaimed = 0;
820
	unsigned long nr_writeback = 0;
821
	unsigned long nr_immediate = 0;
L
Linus Torvalds 已提交
822 823 824 825 826 827 828

	cond_resched();

	while (!list_empty(page_list)) {
		struct address_space *mapping;
		struct page *page;
		int may_enter_fs;
829
		enum page_references references = PAGEREF_RECLAIM_CLEAN;
830
		bool dirty, writeback;
L
Linus Torvalds 已提交
831 832 833 834 835 836

		cond_resched();

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

N
Nick Piggin 已提交
837
		if (!trylock_page(page))
L
Linus Torvalds 已提交
838 839
			goto keep;

840 841
		VM_BUG_ON_PAGE(PageActive(page), page);
		VM_BUG_ON_PAGE(page_zone(page) != zone, page);
L
Linus Torvalds 已提交
842 843

		sc->nr_scanned++;
844

845
		if (unlikely(!page_evictable(page)))
N
Nick Piggin 已提交
846
			goto cull_mlocked;
L
Lee Schermerhorn 已提交
847

848
		if (!sc->may_unmap && page_mapped(page))
849 850
			goto keep_locked;

L
Linus Torvalds 已提交
851 852 853 854
		/* Double the slab pressure for mapped and swapcache pages */
		if (page_mapped(page) || PageSwapCache(page))
			sc->nr_scanned++;

855 856 857
		may_enter_fs = (sc->gfp_mask & __GFP_FS) ||
			(PageSwapCache(page) && (sc->gfp_mask & __GFP_IO));

858 859 860 861 862 863 864 865 866 867 868 869 870
		/*
		 * 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++;

871 872 873 874 875 876
		/*
		 * 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.
		 */
877
		mapping = page_mapping(page);
878 879
		if (((dirty || writeback) && mapping &&
		     bdi_write_congested(mapping->backing_dev_info)) ||
880
		    (writeback && PageReclaim(page)))
881 882
			nr_congested++;

883 884 885 886 887 888 889 890 891 892 893
		/*
		 * 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
894 895
		 *    note that the LRU is being scanned too quickly and the
		 *    caller can stall after page list has been processed.
896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919
		 *
		 * 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.
		 */
920
		if (PageWriteback(page)) {
921 922 923
			/* Case 1 above */
			if (current_is_kswapd() &&
			    PageReclaim(page) &&
J
Johannes Weiner 已提交
924
			    test_bit(ZONE_WRITEBACK, &zone->flags)) {
925 926
				nr_immediate++;
				goto keep_locked;
927 928 929

			/* Case 2 above */
			} else if (global_reclaim(sc) ||
930 931 932 933 934 935 936 937 938 939 940 941 942
			    !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);
943
				nr_writeback++;
944

945
				goto keep_locked;
946 947 948 949

			/* Case 3 above */
			} else {
				wait_on_page_writeback(page);
950
			}
951
		}
L
Linus Torvalds 已提交
952

953 954 955
		if (!force_reclaim)
			references = page_check_references(page, sc);

956 957
		switch (references) {
		case PAGEREF_ACTIVATE:
L
Linus Torvalds 已提交
958
			goto activate_locked;
959 960
		case PAGEREF_KEEP:
			goto keep_locked;
961 962 963 964
		case PAGEREF_RECLAIM:
		case PAGEREF_RECLAIM_CLEAN:
			; /* try to reclaim the page below */
		}
L
Linus Torvalds 已提交
965 966 967 968 969

		/*
		 * Anonymous process memory has backing store?
		 * Try to allocate it some swap space here.
		 */
N
Nick Piggin 已提交
970
		if (PageAnon(page) && !PageSwapCache(page)) {
971 972
			if (!(sc->gfp_mask & __GFP_IO))
				goto keep_locked;
973
			if (!add_to_swap(page, page_list))
L
Linus Torvalds 已提交
974
				goto activate_locked;
975
			may_enter_fs = 1;
L
Linus Torvalds 已提交
976

977 978 979
			/* Adding to swap updated mapping */
			mapping = page_mapping(page);
		}
L
Linus Torvalds 已提交
980 981 982 983 984 985

		/*
		 * The page is mapped into the page tables of one or more
		 * processes. Try to unmap it here.
		 */
		if (page_mapped(page) && mapping) {
986
			switch (try_to_unmap(page, ttu_flags)) {
L
Linus Torvalds 已提交
987 988 989 990
			case SWAP_FAIL:
				goto activate_locked;
			case SWAP_AGAIN:
				goto keep_locked;
N
Nick Piggin 已提交
991 992
			case SWAP_MLOCK:
				goto cull_mlocked;
L
Linus Torvalds 已提交
993 994 995 996 997 998
			case SWAP_SUCCESS:
				; /* try to free the page below */
			}
		}

		if (PageDirty(page)) {
999 1000
			/*
			 * Only kswapd can writeback filesystem pages to
1001 1002
			 * avoid risk of stack overflow but only writeback
			 * if many dirty pages have been encountered.
1003
			 */
1004
			if (page_is_file_cache(page) &&
1005
					(!current_is_kswapd() ||
J
Johannes Weiner 已提交
1006
					 !test_bit(ZONE_DIRTY, &zone->flags))) {
1007 1008 1009 1010 1011 1012 1013 1014 1015
				/*
				 * 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);

1016 1017 1018
				goto keep_locked;
			}

1019
			if (references == PAGEREF_RECLAIM_CLEAN)
L
Linus Torvalds 已提交
1020
				goto keep_locked;
1021
			if (!may_enter_fs)
L
Linus Torvalds 已提交
1022
				goto keep_locked;
1023
			if (!sc->may_writepage)
L
Linus Torvalds 已提交
1024 1025 1026
				goto keep_locked;

			/* Page is dirty, try to write it out here */
1027
			switch (pageout(page, mapping, sc)) {
L
Linus Torvalds 已提交
1028 1029 1030 1031 1032
			case PAGE_KEEP:
				goto keep_locked;
			case PAGE_ACTIVATE:
				goto activate_locked;
			case PAGE_SUCCESS:
1033
				if (PageWriteback(page))
1034
					goto keep;
1035
				if (PageDirty(page))
L
Linus Torvalds 已提交
1036
					goto keep;
1037

L
Linus Torvalds 已提交
1038 1039 1040 1041
				/*
				 * A synchronous write - probably a ramdisk.  Go
				 * ahead and try to reclaim the page.
				 */
N
Nick Piggin 已提交
1042
				if (!trylock_page(page))
L
Linus Torvalds 已提交
1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061
					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
1062
		 * will do this, as well as the blockdev mapping.
L
Linus Torvalds 已提交
1063 1064 1065 1066 1067 1068 1069 1070 1071 1072
		 * 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.
		 */
1073
		if (page_has_private(page)) {
L
Linus Torvalds 已提交
1074 1075
			if (!try_to_release_page(page, sc->gfp_mask))
				goto activate_locked;
N
Nick Piggin 已提交
1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091
			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
1092 1093
		}

1094
		if (!mapping || !__remove_mapping(mapping, page, true))
1095
			goto keep_locked;
L
Linus Torvalds 已提交
1096

N
Nick Piggin 已提交
1097 1098 1099 1100 1101 1102 1103 1104
		/*
		 * 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
1105
free_it:
1106
		nr_reclaimed++;
1107 1108 1109 1110 1111 1112

		/*
		 * 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
1113 1114
		continue;

N
Nick Piggin 已提交
1115
cull_mlocked:
1116 1117
		if (PageSwapCache(page))
			try_to_free_swap(page);
N
Nick Piggin 已提交
1118 1119 1120 1121
		unlock_page(page);
		putback_lru_page(page);
		continue;

L
Linus Torvalds 已提交
1122
activate_locked:
1123 1124
		/* Not a candidate for swapping, so reclaim swap space. */
		if (PageSwapCache(page) && vm_swap_full())
1125
			try_to_free_swap(page);
1126
		VM_BUG_ON_PAGE(PageActive(page), page);
L
Linus Torvalds 已提交
1127 1128 1129 1130 1131 1132
		SetPageActive(page);
		pgactivate++;
keep_locked:
		unlock_page(page);
keep:
		list_add(&page->lru, &ret_pages);
1133
		VM_BUG_ON_PAGE(PageLRU(page) || PageUnevictable(page), page);
L
Linus Torvalds 已提交
1134
	}
1135

1136
	mem_cgroup_uncharge_list(&free_pages);
1137
	free_hot_cold_page_list(&free_pages, true);
1138

L
Linus Torvalds 已提交
1139
	list_splice(&ret_pages, page_list);
1140
	count_vm_events(PGACTIVATE, pgactivate);
1141

1142 1143
	*ret_nr_dirty += nr_dirty;
	*ret_nr_congested += nr_congested;
1144
	*ret_nr_unqueued_dirty += nr_unqueued_dirty;
1145
	*ret_nr_writeback += nr_writeback;
1146
	*ret_nr_immediate += nr_immediate;
1147
	return nr_reclaimed;
L
Linus Torvalds 已提交
1148 1149
}

1150 1151 1152 1153 1154 1155 1156 1157
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,
	};
1158
	unsigned long ret, dummy1, dummy2, dummy3, dummy4, dummy5;
1159 1160 1161 1162
	struct page *page, *next;
	LIST_HEAD(clean_pages);

	list_for_each_entry_safe(page, next, page_list, lru) {
1163 1164
		if (page_is_file_cache(page) && !PageDirty(page) &&
		    !isolated_balloon_page(page)) {
1165 1166 1167 1168 1169 1170
			ClearPageActive(page);
			list_move(&page->lru, &clean_pages);
		}
	}

	ret = shrink_page_list(&clean_pages, zone, &sc,
1171 1172
			TTU_UNMAP|TTU_IGNORE_ACCESS,
			&dummy1, &dummy2, &dummy3, &dummy4, &dummy5, true);
1173
	list_splice(&clean_pages, page_list);
1174
	mod_zone_page_state(zone, NR_ISOLATED_FILE, -ret);
1175 1176 1177
	return ret;
}

A
Andy Whitcroft 已提交
1178 1179 1180 1181 1182 1183 1184 1185 1186 1187
/*
 * 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.
 */
1188
int __isolate_lru_page(struct page *page, isolate_mode_t mode)
A
Andy Whitcroft 已提交
1189 1190 1191 1192 1193 1194 1195
{
	int ret = -EINVAL;

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

M
Minchan Kim 已提交
1196 1197
	/* Compaction should not handle unevictable pages but CMA can do so */
	if (PageUnevictable(page) && !(mode & ISOLATE_UNEVICTABLE))
L
Lee Schermerhorn 已提交
1198 1199
		return ret;

A
Andy Whitcroft 已提交
1200
	ret = -EBUSY;
K
KAMEZAWA Hiroyuki 已提交
1201

1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234
	/*
	 * 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;
		}
	}
1235

1236 1237 1238
	if ((mode & ISOLATE_UNMAPPED) && page_mapped(page))
		return ret;

A
Andy Whitcroft 已提交
1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251
	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
1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262
/*
 * 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.
1263
 * @lruvec:	The LRU vector to pull pages from.
L
Linus Torvalds 已提交
1264
 * @dst:	The temp list to put pages on to.
H
Hugh Dickins 已提交
1265
 * @nr_scanned:	The number of pages that were scanned.
1266
 * @sc:		The scan_control struct for this reclaim session
A
Andy Whitcroft 已提交
1267
 * @mode:	One of the LRU isolation modes
1268
 * @lru:	LRU list id for isolating
L
Linus Torvalds 已提交
1269 1270 1271
 *
 * returns how many pages were moved onto *@dst.
 */
1272
static unsigned long isolate_lru_pages(unsigned long nr_to_scan,
1273
		struct lruvec *lruvec, struct list_head *dst,
1274
		unsigned long *nr_scanned, struct scan_control *sc,
1275
		isolate_mode_t mode, enum lru_list lru)
L
Linus Torvalds 已提交
1276
{
H
Hugh Dickins 已提交
1277
	struct list_head *src = &lruvec->lists[lru];
1278
	unsigned long nr_taken = 0;
1279
	unsigned long scan;
L
Linus Torvalds 已提交
1280

1281
	for (scan = 0; scan < nr_to_scan && !list_empty(src); scan++) {
A
Andy Whitcroft 已提交
1282
		struct page *page;
1283
		int nr_pages;
A
Andy Whitcroft 已提交
1284

L
Linus Torvalds 已提交
1285 1286 1287
		page = lru_to_page(src);
		prefetchw_prev_lru_page(page, src, flags);

1288
		VM_BUG_ON_PAGE(!PageLRU(page), page);
N
Nick Piggin 已提交
1289

1290
		switch (__isolate_lru_page(page, mode)) {
A
Andy Whitcroft 已提交
1291
		case 0:
1292 1293
			nr_pages = hpage_nr_pages(page);
			mem_cgroup_update_lru_size(lruvec, lru, -nr_pages);
A
Andy Whitcroft 已提交
1294
			list_move(&page->lru, dst);
1295
			nr_taken += nr_pages;
A
Andy Whitcroft 已提交
1296 1297 1298 1299 1300 1301
			break;

		case -EBUSY:
			/* else it is being freed elsewhere */
			list_move(&page->lru, src);
			continue;
1302

A
Andy Whitcroft 已提交
1303 1304 1305
		default:
			BUG();
		}
L
Linus Torvalds 已提交
1306 1307
	}

H
Hugh Dickins 已提交
1308
	*nr_scanned = scan;
H
Hugh Dickins 已提交
1309 1310
	trace_mm_vmscan_lru_isolate(sc->order, nr_to_scan, scan,
				    nr_taken, mode, is_file_lru(lru));
L
Linus Torvalds 已提交
1311 1312 1313
	return nr_taken;
}

1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324
/**
 * 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
1325 1326 1327
 * 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.
1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342
 *
 * 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;

1343
	VM_BUG_ON_PAGE(!page_count(page), page);
1344

1345 1346
	if (PageLRU(page)) {
		struct zone *zone = page_zone(page);
1347
		struct lruvec *lruvec;
1348 1349

		spin_lock_irq(&zone->lru_lock);
1350
		lruvec = mem_cgroup_page_lruvec(page, zone);
1351
		if (PageLRU(page)) {
L
Lee Schermerhorn 已提交
1352
			int lru = page_lru(page);
1353
			get_page(page);
1354
			ClearPageLRU(page);
1355 1356
			del_page_from_lru_list(page, lruvec, lru);
			ret = 0;
1357 1358 1359 1360 1361 1362
		}
		spin_unlock_irq(&zone->lru_lock);
	}
	return ret;
}

1363
/*
F
Fengguang Wu 已提交
1364 1365 1366 1367 1368
 * 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.
1369 1370 1371 1372 1373 1374 1375 1376 1377
 */
static int too_many_isolated(struct zone *zone, int file,
		struct scan_control *sc)
{
	unsigned long inactive, isolated;

	if (current_is_kswapd())
		return 0;

1378
	if (!global_reclaim(sc))
1379 1380 1381 1382 1383 1384 1385 1386 1387 1388
		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);
	}

1389 1390 1391 1392 1393 1394 1395 1396
	/*
	 * 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;

1397 1398 1399
	return isolated > inactive;
}

1400
static noinline_for_stack void
H
Hugh Dickins 已提交
1401
putback_inactive_pages(struct lruvec *lruvec, struct list_head *page_list)
1402
{
1403 1404
	struct zone_reclaim_stat *reclaim_stat = &lruvec->reclaim_stat;
	struct zone *zone = lruvec_zone(lruvec);
1405
	LIST_HEAD(pages_to_free);
1406 1407 1408 1409 1410

	/*
	 * Put back any unfreeable pages.
	 */
	while (!list_empty(page_list)) {
1411
		struct page *page = lru_to_page(page_list);
1412
		int lru;
1413

1414
		VM_BUG_ON_PAGE(PageLRU(page), page);
1415
		list_del(&page->lru);
1416
		if (unlikely(!page_evictable(page))) {
1417 1418 1419 1420 1421
			spin_unlock_irq(&zone->lru_lock);
			putback_lru_page(page);
			spin_lock_irq(&zone->lru_lock);
			continue;
		}
1422 1423 1424

		lruvec = mem_cgroup_page_lruvec(page, zone);

1425
		SetPageLRU(page);
1426
		lru = page_lru(page);
1427 1428
		add_page_to_lru_list(page, lruvec, lru);

1429 1430
		if (is_active_lru(lru)) {
			int file = is_file_lru(lru);
1431 1432
			int numpages = hpage_nr_pages(page);
			reclaim_stat->recent_rotated[file] += numpages;
1433
		}
1434 1435 1436
		if (put_page_testzero(page)) {
			__ClearPageLRU(page);
			__ClearPageActive(page);
1437
			del_page_from_lru_list(page, lruvec, lru);
1438 1439 1440

			if (unlikely(PageCompound(page))) {
				spin_unlock_irq(&zone->lru_lock);
1441
				mem_cgroup_uncharge(page);
1442 1443 1444 1445
				(*get_compound_page_dtor(page))(page);
				spin_lock_irq(&zone->lru_lock);
			} else
				list_add(&page->lru, &pages_to_free);
1446 1447 1448
		}
	}

1449 1450 1451 1452
	/*
	 * To save our caller's stack, now use input list for pages to free.
	 */
	list_splice(&pages_to_free, page_list);
1453 1454
}

1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467
/*
 * 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
1468
/*
A
Andrew Morton 已提交
1469 1470
 * shrink_inactive_list() is a helper for shrink_zone().  It returns the number
 * of reclaimed pages
L
Linus Torvalds 已提交
1471
 */
1472
static noinline_for_stack unsigned long
1473
shrink_inactive_list(unsigned long nr_to_scan, struct lruvec *lruvec,
1474
		     struct scan_control *sc, enum lru_list lru)
L
Linus Torvalds 已提交
1475 1476
{
	LIST_HEAD(page_list);
1477
	unsigned long nr_scanned;
1478
	unsigned long nr_reclaimed = 0;
1479
	unsigned long nr_taken;
1480 1481
	unsigned long nr_dirty = 0;
	unsigned long nr_congested = 0;
1482
	unsigned long nr_unqueued_dirty = 0;
1483
	unsigned long nr_writeback = 0;
1484
	unsigned long nr_immediate = 0;
1485
	isolate_mode_t isolate_mode = 0;
1486
	int file = is_file_lru(lru);
1487 1488
	struct zone *zone = lruvec_zone(lruvec);
	struct zone_reclaim_stat *reclaim_stat = &lruvec->reclaim_stat;
1489

1490
	while (unlikely(too_many_isolated(zone, file, sc))) {
1491
		congestion_wait(BLK_RW_ASYNC, HZ/10);
1492 1493 1494 1495 1496 1497

		/* 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
1498
	lru_add_drain();
1499 1500

	if (!sc->may_unmap)
1501
		isolate_mode |= ISOLATE_UNMAPPED;
1502
	if (!sc->may_writepage)
1503
		isolate_mode |= ISOLATE_CLEAN;
1504

L
Linus Torvalds 已提交
1505
	spin_lock_irq(&zone->lru_lock);
1506

1507 1508
	nr_taken = isolate_lru_pages(nr_to_scan, lruvec, &page_list,
				     &nr_scanned, sc, isolate_mode, lru);
1509 1510 1511 1512

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

1513
	if (global_reclaim(sc)) {
1514
		__mod_zone_page_state(zone, NR_PAGES_SCANNED, nr_scanned);
1515
		if (current_is_kswapd())
H
Hugh Dickins 已提交
1516
			__count_zone_vm_events(PGSCAN_KSWAPD, zone, nr_scanned);
1517
		else
H
Hugh Dickins 已提交
1518
			__count_zone_vm_events(PGSCAN_DIRECT, zone, nr_scanned);
1519
	}
1520
	spin_unlock_irq(&zone->lru_lock);
1521

1522
	if (nr_taken == 0)
1523
		return 0;
A
Andy Whitcroft 已提交
1524

1525
	nr_reclaimed = shrink_page_list(&page_list, zone, sc, TTU_UNMAP,
1526 1527 1528
				&nr_dirty, &nr_unqueued_dirty, &nr_congested,
				&nr_writeback, &nr_immediate,
				false);
1529

1530 1531
	spin_lock_irq(&zone->lru_lock);

1532
	reclaim_stat->recent_scanned[file] += nr_taken;
1533

Y
Ying Han 已提交
1534 1535 1536 1537 1538 1539 1540 1541
	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
1542

1543
	putback_inactive_pages(lruvec, &page_list);
1544

1545
	__mod_zone_page_state(zone, NR_ISOLATED_ANON + file, -nr_taken);
1546 1547 1548

	spin_unlock_irq(&zone->lru_lock);

1549
	mem_cgroup_uncharge_list(&page_list);
1550
	free_hot_cold_page_list(&page_list, true);
1551

1552 1553 1554 1555 1556 1557 1558 1559 1560 1561
	/*
	 * 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.
	 *
1562 1563 1564
	 * 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.
1565
	 */
1566
	if (nr_writeback && nr_writeback == nr_taken)
J
Johannes Weiner 已提交
1567
		set_bit(ZONE_WRITEBACK, &zone->flags);
1568

1569
	/*
1570 1571
	 * memcg will stall in page writeback so only consider forcibly
	 * stalling for global reclaim
1572
	 */
1573
	if (global_reclaim(sc)) {
1574 1575 1576 1577 1578
		/*
		 * 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)
J
Johannes Weiner 已提交
1579
			set_bit(ZONE_CONGESTED, &zone->flags);
1580

1581 1582 1583
		/*
		 * If dirty pages are scanned that are not queued for IO, it
		 * implies that flushers are not keeping up. In this case, flag
J
Johannes Weiner 已提交
1584 1585
		 * the zone ZONE_DIRTY and kswapd will start writing pages from
		 * reclaim context.
1586 1587
		 */
		if (nr_unqueued_dirty == nr_taken)
J
Johannes Weiner 已提交
1588
			set_bit(ZONE_DIRTY, &zone->flags);
1589 1590

		/*
1591 1592 1593
		 * 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
1594 1595
		 * they are written so also forcibly stall.
		 */
1596
		if (nr_immediate && current_may_throttle())
1597
			congestion_wait(BLK_RW_ASYNC, HZ/10);
1598
	}
1599

1600 1601 1602 1603 1604
	/*
	 * 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.
	 */
1605 1606
	if (!sc->hibernation_mode && !current_is_kswapd() &&
	    current_may_throttle())
1607 1608
		wait_iff_congested(zone, BLK_RW_ASYNC, HZ/10);

1609 1610 1611
	trace_mm_vmscan_lru_shrink_inactive(zone->zone_pgdat->node_id,
		zone_idx(zone),
		nr_scanned, nr_reclaimed,
1612
		sc->priority,
M
Mel Gorman 已提交
1613
		trace_shrink_flags(file));
1614
	return nr_reclaimed;
L
Linus Torvalds 已提交
1615 1616 1617 1618 1619 1620 1621 1622 1623 1624 1625 1626 1627 1628 1629 1630 1631 1632 1633
}

/*
 * 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.
 */
1634

1635
static void move_active_pages_to_lru(struct lruvec *lruvec,
1636
				     struct list_head *list,
1637
				     struct list_head *pages_to_free,
1638 1639
				     enum lru_list lru)
{
1640
	struct zone *zone = lruvec_zone(lruvec);
1641 1642
	unsigned long pgmoved = 0;
	struct page *page;
1643
	int nr_pages;
1644 1645 1646

	while (!list_empty(list)) {
		page = lru_to_page(list);
1647
		lruvec = mem_cgroup_page_lruvec(page, zone);
1648

1649
		VM_BUG_ON_PAGE(PageLRU(page), page);
1650 1651
		SetPageLRU(page);

1652 1653
		nr_pages = hpage_nr_pages(page);
		mem_cgroup_update_lru_size(lruvec, lru, nr_pages);
1654
		list_move(&page->lru, &lruvec->lists[lru]);
1655
		pgmoved += nr_pages;
1656

1657 1658 1659
		if (put_page_testzero(page)) {
			__ClearPageLRU(page);
			__ClearPageActive(page);
1660
			del_page_from_lru_list(page, lruvec, lru);
1661 1662 1663

			if (unlikely(PageCompound(page))) {
				spin_unlock_irq(&zone->lru_lock);
1664
				mem_cgroup_uncharge(page);
1665 1666 1667 1668
				(*get_compound_page_dtor(page))(page);
				spin_lock_irq(&zone->lru_lock);
			} else
				list_add(&page->lru, pages_to_free);
1669 1670 1671 1672 1673 1674
		}
	}
	__mod_zone_page_state(zone, NR_LRU_BASE + lru, pgmoved);
	if (!is_active_lru(lru))
		__count_vm_events(PGDEACTIVATE, pgmoved);
}
1675

H
Hugh Dickins 已提交
1676
static void shrink_active_list(unsigned long nr_to_scan,
1677
			       struct lruvec *lruvec,
1678
			       struct scan_control *sc,
1679
			       enum lru_list lru)
L
Linus Torvalds 已提交
1680
{
1681
	unsigned long nr_taken;
H
Hugh Dickins 已提交
1682
	unsigned long nr_scanned;
1683
	unsigned long vm_flags;
L
Linus Torvalds 已提交
1684
	LIST_HEAD(l_hold);	/* The pages which were snipped off */
1685
	LIST_HEAD(l_active);
1686
	LIST_HEAD(l_inactive);
L
Linus Torvalds 已提交
1687
	struct page *page;
1688
	struct zone_reclaim_stat *reclaim_stat = &lruvec->reclaim_stat;
1689
	unsigned long nr_rotated = 0;
1690
	isolate_mode_t isolate_mode = 0;
1691
	int file = is_file_lru(lru);
1692
	struct zone *zone = lruvec_zone(lruvec);
L
Linus Torvalds 已提交
1693 1694

	lru_add_drain();
1695 1696

	if (!sc->may_unmap)
1697
		isolate_mode |= ISOLATE_UNMAPPED;
1698
	if (!sc->may_writepage)
1699
		isolate_mode |= ISOLATE_CLEAN;
1700

L
Linus Torvalds 已提交
1701
	spin_lock_irq(&zone->lru_lock);
1702

1703 1704
	nr_taken = isolate_lru_pages(nr_to_scan, lruvec, &l_hold,
				     &nr_scanned, sc, isolate_mode, lru);
1705
	if (global_reclaim(sc))
1706
		__mod_zone_page_state(zone, NR_PAGES_SCANNED, nr_scanned);
1707

1708
	reclaim_stat->recent_scanned[file] += nr_taken;
1709

H
Hugh Dickins 已提交
1710
	__count_zone_vm_events(PGREFILL, zone, nr_scanned);
1711
	__mod_zone_page_state(zone, NR_LRU_BASE + lru, -nr_taken);
K
KOSAKI Motohiro 已提交
1712
	__mod_zone_page_state(zone, NR_ISOLATED_ANON + file, nr_taken);
L
Linus Torvalds 已提交
1713 1714 1715 1716 1717 1718
	spin_unlock_irq(&zone->lru_lock);

	while (!list_empty(&l_hold)) {
		cond_resched();
		page = lru_to_page(&l_hold);
		list_del(&page->lru);
1719

1720
		if (unlikely(!page_evictable(page))) {
L
Lee Schermerhorn 已提交
1721 1722 1723 1724
			putback_lru_page(page);
			continue;
		}

1725 1726 1727 1728 1729 1730 1731 1732
		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);
			}
		}

1733 1734
		if (page_referenced(page, 0, sc->target_mem_cgroup,
				    &vm_flags)) {
1735
			nr_rotated += hpage_nr_pages(page);
1736 1737 1738 1739 1740 1741 1742 1743 1744
			/*
			 * 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.
			 */
1745
			if ((vm_flags & VM_EXEC) && page_is_file_cache(page)) {
1746 1747 1748 1749
				list_add(&page->lru, &l_active);
				continue;
			}
		}
1750

1751
		ClearPageActive(page);	/* we are de-activating */
L
Linus Torvalds 已提交
1752 1753 1754
		list_add(&page->lru, &l_inactive);
	}

1755
	/*
1756
	 * Move pages back to the lru list.
1757
	 */
1758
	spin_lock_irq(&zone->lru_lock);
1759
	/*
1760 1761 1762
	 * 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
1763
	 * get_scan_count.
1764
	 */
1765
	reclaim_stat->recent_rotated[file] += nr_rotated;
1766

1767 1768
	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
1769
	__mod_zone_page_state(zone, NR_ISOLATED_ANON + file, -nr_taken);
1770
	spin_unlock_irq(&zone->lru_lock);
1771

1772
	mem_cgroup_uncharge_list(&l_hold);
1773
	free_hot_cold_page_list(&l_hold, true);
L
Linus Torvalds 已提交
1774 1775
}

1776
#ifdef CONFIG_SWAP
1777
static int inactive_anon_is_low_global(struct zone *zone)
1778 1779 1780 1781 1782 1783 1784 1785 1786 1787 1788 1789
{
	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;
}

1790 1791
/**
 * inactive_anon_is_low - check if anonymous pages need to be deactivated
1792
 * @lruvec: LRU vector to check
1793 1794 1795 1796
 *
 * Returns true if the zone does not have enough inactive anon pages,
 * meaning some active anon pages need to be deactivated.
 */
1797
static int inactive_anon_is_low(struct lruvec *lruvec)
1798
{
1799 1800 1801 1802 1803 1804 1805
	/*
	 * If we don't have swap space, anonymous page deactivation
	 * is pointless.
	 */
	if (!total_swap_pages)
		return 0;

1806
	if (!mem_cgroup_disabled())
1807
		return mem_cgroup_inactive_anon_is_low(lruvec);
1808

1809
	return inactive_anon_is_low_global(lruvec_zone(lruvec));
1810
}
1811
#else
1812
static inline int inactive_anon_is_low(struct lruvec *lruvec)
1813 1814 1815 1816
{
	return 0;
}
#endif
1817

1818 1819
/**
 * inactive_file_is_low - check if file pages need to be deactivated
1820
 * @lruvec: LRU vector to check
1821 1822 1823 1824 1825 1826 1827 1828 1829 1830 1831
 *
 * 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.
 */
1832
static int inactive_file_is_low(struct lruvec *lruvec)
1833
{
1834 1835 1836 1837 1838
	unsigned long inactive;
	unsigned long active;

	inactive = get_lru_size(lruvec, LRU_INACTIVE_FILE);
	active = get_lru_size(lruvec, LRU_ACTIVE_FILE);
1839

1840
	return active > inactive;
1841 1842
}

H
Hugh Dickins 已提交
1843
static int inactive_list_is_low(struct lruvec *lruvec, enum lru_list lru)
1844
{
H
Hugh Dickins 已提交
1845
	if (is_file_lru(lru))
1846
		return inactive_file_is_low(lruvec);
1847
	else
1848
		return inactive_anon_is_low(lruvec);
1849 1850
}

1851
static unsigned long shrink_list(enum lru_list lru, unsigned long nr_to_scan,
1852
				 struct lruvec *lruvec, struct scan_control *sc)
1853
{
1854
	if (is_active_lru(lru)) {
H
Hugh Dickins 已提交
1855
		if (inactive_list_is_low(lruvec, lru))
1856
			shrink_active_list(nr_to_scan, lruvec, sc, lru);
1857 1858 1859
		return 0;
	}

1860
	return shrink_inactive_list(nr_to_scan, lruvec, sc, lru);
1861 1862
}

1863 1864 1865 1866 1867 1868 1869
enum scan_balance {
	SCAN_EQUAL,
	SCAN_FRACT,
	SCAN_ANON,
	SCAN_FILE,
};

1870 1871 1872 1873 1874 1875
/*
 * 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
1876 1877
 * 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
1878
 */
1879
static void get_scan_count(struct lruvec *lruvec, int swappiness,
1880 1881
			   struct scan_control *sc, unsigned long *nr,
			   unsigned long *lru_pages)
1882
{
1883 1884 1885 1886
	struct zone_reclaim_stat *reclaim_stat = &lruvec->reclaim_stat;
	u64 fraction[2];
	u64 denominator = 0;	/* gcc */
	struct zone *zone = lruvec_zone(lruvec);
1887
	unsigned long anon_prio, file_prio;
1888
	enum scan_balance scan_balance;
1889
	unsigned long anon, file;
1890
	bool force_scan = false;
1891
	unsigned long ap, fp;
H
Hugh Dickins 已提交
1892
	enum lru_list lru;
1893 1894
	bool some_scanned;
	int pass;
1895

1896 1897 1898 1899 1900 1901 1902 1903 1904 1905
	/*
	 * 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.
	 */
1906
	if (current_is_kswapd() && !zone_reclaimable(zone))
1907
		force_scan = true;
1908
	if (!global_reclaim(sc))
1909
		force_scan = true;
1910 1911

	/* If we have no swap space, do not bother scanning anon pages. */
1912
	if (!sc->may_swap || (get_nr_swap_pages() <= 0)) {
1913
		scan_balance = SCAN_FILE;
1914 1915
		goto out;
	}
1916

1917 1918 1919 1920 1921 1922 1923
	/*
	 * 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.
	 */
1924
	if (!global_reclaim(sc) && !swappiness) {
1925
		scan_balance = SCAN_FILE;
1926 1927 1928 1929 1930 1931 1932 1933
		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).
	 */
1934
	if (!sc->priority && swappiness) {
1935
		scan_balance = SCAN_EQUAL;
1936 1937 1938
		goto out;
	}

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)) {
1949 1950 1951 1952 1953 1954
		unsigned long zonefile;
		unsigned long zonefree;

		zonefree = zone_page_state(zone, NR_FREE_PAGES);
		zonefile = zone_page_state(zone, NR_ACTIVE_FILE) +
			   zone_page_state(zone, NR_INACTIVE_FILE);
1955

1956
		if (unlikely(zonefile + zonefree <= high_wmark_pages(zone))) {
1957 1958 1959 1960 1961
			scan_balance = SCAN_ANON;
			goto out;
		}
	}

1962 1963 1964 1965 1966
	/*
	 * There is enough inactive page cache, do not reclaim
	 * anything from the anonymous working set right now.
	 */
	if (!inactive_file_is_low(lruvec)) {
1967
		scan_balance = SCAN_FILE;
1968 1969 1970
		goto out;
	}

1971 1972
	scan_balance = SCAN_FRACT;

1973 1974 1975 1976
	/*
	 * With swappiness at 100, anonymous and file have the same priority.
	 * This scanning priority is essentially the inverse of IO cost.
	 */
1977
	anon_prio = swappiness;
H
Hugh Dickins 已提交
1978
	file_prio = 200 - anon_prio;
1979

1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990
	/*
	 * 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]
	 */
1991 1992 1993 1994 1995 1996

	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);

1997
	spin_lock_irq(&zone->lru_lock);
1998 1999 2000
	if (unlikely(reclaim_stat->recent_scanned[0] > anon / 4)) {
		reclaim_stat->recent_scanned[0] /= 2;
		reclaim_stat->recent_rotated[0] /= 2;
2001 2002
	}

2003 2004 2005
	if (unlikely(reclaim_stat->recent_scanned[1] > file / 4)) {
		reclaim_stat->recent_scanned[1] /= 2;
		reclaim_stat->recent_rotated[1] /= 2;
2006 2007 2008
	}

	/*
2009 2010 2011
	 * 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.
2012
	 */
2013
	ap = anon_prio * (reclaim_stat->recent_scanned[0] + 1);
2014
	ap /= reclaim_stat->recent_rotated[0] + 1;
2015

2016
	fp = file_prio * (reclaim_stat->recent_scanned[1] + 1);
2017
	fp /= reclaim_stat->recent_rotated[1] + 1;
2018
	spin_unlock_irq(&zone->lru_lock);
2019

2020 2021 2022 2023
	fraction[0] = ap;
	fraction[1] = fp;
	denominator = ap + fp + 1;
out:
2024 2025 2026
	some_scanned = false;
	/* Only use force_scan on second pass. */
	for (pass = 0; !some_scanned && pass < 2; pass++) {
2027
		*lru_pages = 0;
2028 2029 2030 2031
		for_each_evictable_lru(lru) {
			int file = is_file_lru(lru);
			unsigned long size;
			unsigned long scan;
2032

2033 2034
			size = get_lru_size(lruvec, lru);
			scan = size >> sc->priority;
2035

2036 2037
			if (!scan && pass && force_scan)
				scan = min(size, SWAP_CLUSTER_MAX);
2038

2039 2040 2041 2042 2043 2044 2045 2046 2047 2048 2049 2050 2051 2052 2053
			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 */
2054 2055
				if ((scan_balance == SCAN_FILE) != file) {
					size = 0;
2056
					scan = 0;
2057
				}
2058 2059 2060 2061 2062
				break;
			default:
				/* Look ma, no brain */
				BUG();
			}
2063 2064

			*lru_pages += size;
2065
			nr[lru] = scan;
2066

2067
			/*
2068 2069
			 * Skip the second pass and don't force_scan,
			 * if we found something to scan.
2070
			 */
2071
			some_scanned |= !!scan;
2072
		}
2073
	}
2074
}
2075

2076 2077 2078
/*
 * This is a basic per-zone page freer.  Used by both kswapd and direct reclaim.
 */
2079
static void shrink_lruvec(struct lruvec *lruvec, int swappiness,
2080
			  struct scan_control *sc, unsigned long *lru_pages)
2081 2082
{
	unsigned long nr[NR_LRU_LISTS];
2083
	unsigned long targets[NR_LRU_LISTS];
2084 2085 2086 2087 2088
	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;
2089
	bool scan_adjusted;
2090

2091
	get_scan_count(lruvec, swappiness, sc, nr, lru_pages);
2092

2093 2094 2095
	/* Record the original scan target for proportional adjustments later */
	memcpy(targets, nr, sizeof(nr));

2096 2097 2098 2099 2100 2101 2102 2103 2104 2105 2106 2107 2108 2109
	/*
	 * 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);

2110 2111 2112
	blk_start_plug(&plug);
	while (nr[LRU_INACTIVE_ANON] || nr[LRU_ACTIVE_FILE] ||
					nr[LRU_INACTIVE_FILE]) {
2113 2114 2115
		unsigned long nr_anon, nr_file, percentage;
		unsigned long nr_scanned;

2116 2117 2118 2119 2120 2121 2122 2123 2124
		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);
			}
		}
2125 2126 2127 2128 2129 2130

		if (nr_reclaimed < nr_to_reclaim || scan_adjusted)
			continue;

		/*
		 * For kswapd and memcg, reclaim at least the number of pages
2131
		 * requested. Ensure that the anon and file LRUs are scanned
2132 2133 2134 2135 2136 2137 2138
		 * 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];

2139 2140 2141 2142 2143 2144 2145 2146 2147
		/*
		 * 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;

2148 2149 2150 2151 2152 2153 2154 2155 2156 2157 2158 2159 2160 2161 2162 2163 2164 2165 2166 2167 2168 2169 2170 2171 2172 2173 2174 2175 2176 2177 2178
		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;
2179 2180 2181 2182 2183 2184 2185 2186 2187 2188 2189 2190 2191 2192 2193
	}
	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
2194
/* Use reclaim/compaction for costly allocs or under memory pressure */
2195
static bool in_reclaim_compaction(struct scan_control *sc)
M
Mel Gorman 已提交
2196
{
2197
	if (IS_ENABLED(CONFIG_COMPACTION) && sc->order &&
M
Mel Gorman 已提交
2198
			(sc->order > PAGE_ALLOC_COSTLY_ORDER ||
2199
			 sc->priority < DEF_PRIORITY - 2))
M
Mel Gorman 已提交
2200 2201 2202 2203 2204
		return true;

	return false;
}

2205
/*
M
Mel Gorman 已提交
2206 2207 2208 2209 2210
 * 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.
2211
 */
2212
static inline bool should_continue_reclaim(struct zone *zone,
2213 2214 2215 2216 2217 2218 2219 2220
					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 */
2221
	if (!in_reclaim_compaction(sc))
2222 2223
		return false;

2224 2225 2226 2227 2228 2229 2230 2231 2232 2233 2234 2235 2236 2237 2238 2239 2240 2241 2242 2243 2244 2245
	/* 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;
	}
2246 2247 2248 2249 2250 2251

	/*
	 * If we have not reclaimed enough pages for compaction and the
	 * inactive lists are large enough, continue reclaiming
	 */
	pages_for_compaction = (2UL << sc->order);
2252
	inactive_lru_pages = zone_page_state(zone, NR_INACTIVE_FILE);
2253
	if (get_nr_swap_pages() > 0)
2254
		inactive_lru_pages += zone_page_state(zone, NR_INACTIVE_ANON);
2255 2256 2257 2258 2259
	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 */
2260
	switch (compaction_suitable(zone, sc->order, 0, 0)) {
2261 2262 2263 2264 2265 2266 2267 2268
	case COMPACT_PARTIAL:
	case COMPACT_CONTINUE:
		return false;
	default:
		return true;
	}
}

2269 2270
static bool shrink_zone(struct zone *zone, struct scan_control *sc,
			bool is_classzone)
L
Linus Torvalds 已提交
2271
{
2272
	unsigned long nr_reclaimed, nr_scanned;
2273
	bool reclaimable = false;
L
Linus Torvalds 已提交
2274

2275 2276 2277 2278 2279 2280
	do {
		struct mem_cgroup *root = sc->target_mem_cgroup;
		struct mem_cgroup_reclaim_cookie reclaim = {
			.zone = zone,
			.priority = sc->priority,
		};
2281
		unsigned long zone_lru_pages = 0;
2282
		struct mem_cgroup *memcg;
2283

2284 2285
		nr_reclaimed = sc->nr_reclaimed;
		nr_scanned = sc->nr_scanned;
L
Linus Torvalds 已提交
2286

2287 2288
		memcg = mem_cgroup_iter(root, NULL, &reclaim);
		do {
2289
			unsigned long lru_pages;
2290
			struct lruvec *lruvec;
2291
			int swappiness;
2292

2293
			lruvec = mem_cgroup_zone_lruvec(zone, memcg);
2294
			swappiness = mem_cgroup_swappiness(memcg);
2295

2296 2297
			shrink_lruvec(lruvec, swappiness, sc, &lru_pages);
			zone_lru_pages += lru_pages;
2298

2299
			/*
2300 2301
			 * Direct reclaim and kswapd have to scan all memory
			 * cgroups to fulfill the overall scan target for the
2302
			 * zone.
2303 2304 2305 2306 2307
			 *
			 * 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.
2308
			 */
2309 2310
			if (!global_reclaim(sc) &&
					sc->nr_reclaimed >= sc->nr_to_reclaim) {
2311 2312 2313
				mem_cgroup_iter_break(root, memcg);
				break;
			}
2314 2315
			memcg = mem_cgroup_iter(root, memcg, &reclaim);
		} while (memcg);
2316

2317 2318 2319 2320 2321 2322 2323 2324 2325 2326 2327 2328 2329 2330 2331 2332 2333 2334 2335
		/*
		 * Shrink the slab caches in the same proportion that
		 * the eligible LRU pages were scanned.
		 */
		if (global_reclaim(sc) && is_classzone) {
			struct reclaim_state *reclaim_state;

			shrink_node_slabs(sc->gfp_mask, zone_to_nid(zone),
					  sc->nr_scanned - nr_scanned,
					  zone_lru_pages);

			reclaim_state = current->reclaim_state;
			if (reclaim_state) {
				sc->nr_reclaimed +=
					reclaim_state->reclaimed_slab;
				reclaim_state->reclaimed_slab = 0;
			}
		}

2336 2337 2338 2339
		vmpressure(sc->gfp_mask, sc->target_mem_cgroup,
			   sc->nr_scanned - nr_scanned,
			   sc->nr_reclaimed - nr_reclaimed);

2340 2341 2342
		if (sc->nr_reclaimed - nr_reclaimed)
			reclaimable = true;

2343 2344
	} while (should_continue_reclaim(zone, sc->nr_reclaimed - nr_reclaimed,
					 sc->nr_scanned - nr_scanned, sc));
2345 2346

	return reclaimable;
2347 2348
}

2349 2350 2351 2352
/*
 * Returns true if compaction should go ahead for a high-order request, or
 * the high-order allocation would succeed without compaction.
 */
2353
static inline bool compaction_ready(struct zone *zone, int order)
2354 2355 2356 2357 2358 2359 2360 2361 2362 2363
{
	unsigned long balance_gap, watermark;
	bool watermark_ok;

	/*
	 * 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
	 */
2364 2365
	balance_gap = min(low_wmark_pages(zone), DIV_ROUND_UP(
			zone->managed_pages, KSWAPD_ZONE_BALANCE_GAP_RATIO));
2366
	watermark = high_wmark_pages(zone) + balance_gap + (2UL << order);
2367 2368 2369 2370 2371 2372
	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
	 */
2373
	if (compaction_deferred(zone, order))
2374 2375
		return watermark_ok;

2376 2377 2378 2379
	/*
	 * If compaction is not ready to start and allocation is not likely
	 * to succeed without it, then keep reclaiming.
	 */
2380
	if (compaction_suitable(zone, order, 0, 0) == COMPACT_SKIPPED)
2381 2382 2383 2384 2385
		return false;

	return watermark_ok;
}

L
Linus Torvalds 已提交
2386 2387 2388 2389 2390
/*
 * 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.
 *
2391 2392
 * We reclaim from a zone even if that zone is over high_wmark_pages(zone).
 * Because:
L
Linus Torvalds 已提交
2393 2394
 * a) The caller may be trying to free *extra* pages to satisfy a higher-order
 *    allocation or
2395 2396 2397
 * 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
2398 2399 2400
 *
 * If a zone is deemed to be full of pinned pages then just give it a light
 * scan then give up on it.
2401 2402
 *
 * Returns true if a zone was reclaimable.
L
Linus Torvalds 已提交
2403
 */
2404
static bool shrink_zones(struct zonelist *zonelist, struct scan_control *sc)
L
Linus Torvalds 已提交
2405
{
2406
	struct zoneref *z;
2407
	struct zone *zone;
2408 2409
	unsigned long nr_soft_reclaimed;
	unsigned long nr_soft_scanned;
2410
	gfp_t orig_mask;
2411
	enum zone_type requested_highidx = gfp_zone(sc->gfp_mask);
2412
	bool reclaimable = false;
2413

2414 2415 2416 2417 2418
	/*
	 * 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
	 */
2419
	orig_mask = sc->gfp_mask;
2420 2421 2422
	if (buffer_heads_over_limit)
		sc->gfp_mask |= __GFP_HIGHMEM;

2423
	for_each_zone_zonelist_nodemask(zone, z, zonelist,
2424 2425 2426
					requested_highidx, sc->nodemask) {
		enum zone_type classzone_idx;

2427
		if (!populated_zone(zone))
L
Linus Torvalds 已提交
2428
			continue;
2429 2430 2431 2432 2433 2434

		classzone_idx = requested_highidx;
		while (!populated_zone(zone->zone_pgdat->node_zones +
							classzone_idx))
			classzone_idx--;

2435 2436 2437 2438
		/*
		 * Take care memory controller reclaiming has small influence
		 * to global LRU.
		 */
2439
		if (global_reclaim(sc)) {
2440 2441
			if (!cpuset_zone_allowed(zone,
						 GFP_KERNEL | __GFP_HARDWALL))
2442
				continue;
2443

2444 2445
			if (sc->priority != DEF_PRIORITY &&
			    !zone_reclaimable(zone))
2446
				continue;	/* Let kswapd poll it */
2447 2448 2449 2450 2451 2452 2453 2454 2455 2456 2457 2458 2459 2460 2461 2462

			/*
			 * 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
			 * noticeable problem, like transparent huge
			 * page allocations.
			 */
			if (IS_ENABLED(CONFIG_COMPACTION) &&
			    sc->order > PAGE_ALLOC_COSTLY_ORDER &&
			    zonelist_zone_idx(z) <= requested_highidx &&
			    compaction_ready(zone, sc->order)) {
				sc->compaction_ready = true;
				continue;
2463
			}
2464

2465 2466 2467 2468 2469 2470 2471 2472 2473 2474 2475 2476
			/*
			 * 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;
2477 2478
			if (nr_soft_reclaimed)
				reclaimable = true;
2479
			/* need some check for avoid more shrink_zone() */
2480
		}
2481

2482
		if (shrink_zone(zone, sc, zone_idx(zone) == classzone_idx))
2483 2484 2485 2486 2487
			reclaimable = true;

		if (global_reclaim(sc) &&
		    !reclaimable && zone_reclaimable(zone))
			reclaimable = true;
L
Linus Torvalds 已提交
2488
	}
2489

2490 2491 2492 2493 2494
	/*
	 * Restore to original mask to avoid the impact on the caller if we
	 * promoted it to __GFP_HIGHMEM.
	 */
	sc->gfp_mask = orig_mask;
2495

2496
	return reclaimable;
L
Linus Torvalds 已提交
2497
}
2498

L
Linus Torvalds 已提交
2499 2500 2501 2502 2503 2504 2505 2506
/*
 * 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
2507 2508 2509 2510
 * 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.
2511 2512 2513
 *
 * returns:	0, if no pages reclaimed
 * 		else, the number of pages reclaimed
L
Linus Torvalds 已提交
2514
 */
2515
static unsigned long do_try_to_free_pages(struct zonelist *zonelist,
2516
					  struct scan_control *sc)
L
Linus Torvalds 已提交
2517
{
2518
	unsigned long total_scanned = 0;
2519
	unsigned long writeback_threshold;
2520
	bool zones_reclaimable;
L
Linus Torvalds 已提交
2521

2522 2523
	delayacct_freepages_start();

2524
	if (global_reclaim(sc))
2525
		count_vm_event(ALLOCSTALL);
L
Linus Torvalds 已提交
2526

2527
	do {
2528 2529
		vmpressure_prio(sc->gfp_mask, sc->target_mem_cgroup,
				sc->priority);
2530
		sc->nr_scanned = 0;
2531
		zones_reclaimable = shrink_zones(zonelist, sc);
2532

2533
		total_scanned += sc->nr_scanned;
2534
		if (sc->nr_reclaimed >= sc->nr_to_reclaim)
2535 2536 2537 2538
			break;

		if (sc->compaction_ready)
			break;
L
Linus Torvalds 已提交
2539

2540 2541 2542 2543 2544 2545 2546
		/*
		 * 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
2547 2548 2549 2550 2551 2552 2553
		/*
		 * 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.
		 */
2554 2555
		writeback_threshold = sc->nr_to_reclaim + sc->nr_to_reclaim / 2;
		if (total_scanned > writeback_threshold) {
2556 2557
			wakeup_flusher_threads(laptop_mode ? 0 : total_scanned,
						WB_REASON_TRY_TO_FREE_PAGES);
2558
			sc->may_writepage = 1;
L
Linus Torvalds 已提交
2559
		}
2560
	} while (--sc->priority >= 0);
2561

2562 2563
	delayacct_freepages_end();

2564 2565 2566
	if (sc->nr_reclaimed)
		return sc->nr_reclaimed;

2567
	/* Aborted reclaim to try compaction? don't OOM, then */
2568
	if (sc->compaction_ready)
2569 2570
		return 1;

2571 2572
	/* Any of the zones still reclaimable?  Don't OOM. */
	if (zones_reclaimable)
2573 2574 2575
		return 1;

	return 0;
L
Linus Torvalds 已提交
2576 2577
}

2578 2579 2580 2581 2582 2583 2584 2585 2586 2587
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];
2588 2589 2590
		if (!populated_zone(zone))
			continue;

2591 2592 2593 2594
		pfmemalloc_reserve += min_wmark_pages(zone);
		free_pages += zone_page_state(zone, NR_FREE_PAGES);
	}

2595 2596 2597 2598
	/* If there are no reserves (unexpected config) then do not throttle */
	if (!pfmemalloc_reserve)
		return true;

2599 2600 2601 2602 2603 2604 2605 2606 2607 2608 2609 2610 2611 2612 2613 2614
	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
2615 2616 2617 2618
 * 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.
2619
 */
2620
static bool throttle_direct_reclaim(gfp_t gfp_mask, struct zonelist *zonelist,
2621 2622
					nodemask_t *nodemask)
{
2623
	struct zoneref *z;
2624
	struct zone *zone;
2625
	pg_data_t *pgdat = NULL;
2626 2627 2628 2629 2630 2631 2632 2633 2634

	/*
	 * 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)
2635 2636 2637 2638 2639 2640 2641 2642
		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;
2643

2644 2645 2646 2647 2648 2649 2650 2651 2652 2653 2654 2655 2656 2657 2658 2659 2660 2661 2662 2663 2664 2665 2666 2667 2668 2669 2670 2671
	/*
	 * 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)
2672
		goto out;
2673

2674 2675 2676
	/* Account for the throttling */
	count_vm_event(PGSCAN_DIRECT_THROTTLE);

2677 2678 2679 2680 2681 2682 2683 2684 2685 2686 2687
	/*
	 * 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);
2688 2689

		goto check_pending;
2690 2691 2692 2693 2694
	}

	/* Throttle until kswapd wakes the process */
	wait_event_killable(zone->zone_pgdat->pfmemalloc_wait,
		pfmemalloc_watermark_ok(pgdat));
2695 2696 2697 2698 2699 2700 2701

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

out:
	return false;
2702 2703
}

2704
unsigned long try_to_free_pages(struct zonelist *zonelist, int order,
2705
				gfp_t gfp_mask, nodemask_t *nodemask)
2706
{
2707
	unsigned long nr_reclaimed;
2708
	struct scan_control sc = {
2709
		.nr_to_reclaim = SWAP_CLUSTER_MAX,
2710
		.gfp_mask = (gfp_mask = memalloc_noio_flags(gfp_mask)),
2711 2712 2713
		.order = order,
		.nodemask = nodemask,
		.priority = DEF_PRIORITY,
2714
		.may_writepage = !laptop_mode,
2715
		.may_unmap = 1,
2716
		.may_swap = 1,
2717 2718
	};

2719
	/*
2720 2721 2722
	 * 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.
2723
	 */
2724
	if (throttle_direct_reclaim(gfp_mask, zonelist, nodemask))
2725 2726
		return 1;

2727 2728 2729 2730
	trace_mm_vmscan_direct_reclaim_begin(order,
				sc.may_writepage,
				gfp_mask);

2731
	nr_reclaimed = do_try_to_free_pages(zonelist, &sc);
2732 2733 2734 2735

	trace_mm_vmscan_direct_reclaim_end(nr_reclaimed);

	return nr_reclaimed;
2736 2737
}

A
Andrew Morton 已提交
2738
#ifdef CONFIG_MEMCG
2739

2740
unsigned long mem_cgroup_shrink_node_zone(struct mem_cgroup *memcg,
2741
						gfp_t gfp_mask, bool noswap,
2742 2743
						struct zone *zone,
						unsigned long *nr_scanned)
2744 2745
{
	struct scan_control sc = {
2746
		.nr_to_reclaim = SWAP_CLUSTER_MAX,
2747
		.target_mem_cgroup = memcg,
2748 2749 2750 2751
		.may_writepage = !laptop_mode,
		.may_unmap = 1,
		.may_swap = !noswap,
	};
2752
	struct lruvec *lruvec = mem_cgroup_zone_lruvec(zone, memcg);
2753
	int swappiness = mem_cgroup_swappiness(memcg);
2754
	unsigned long lru_pages;
2755

2756 2757
	sc.gfp_mask = (gfp_mask & GFP_RECLAIM_MASK) |
			(GFP_HIGHUSER_MOVABLE & ~GFP_RECLAIM_MASK);
2758

2759
	trace_mm_vmscan_memcg_softlimit_reclaim_begin(sc.order,
2760 2761 2762
						      sc.may_writepage,
						      sc.gfp_mask);

2763 2764 2765 2766 2767 2768 2769
	/*
	 * 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.
	 */
2770
	shrink_lruvec(lruvec, swappiness, &sc, &lru_pages);
2771 2772 2773

	trace_mm_vmscan_memcg_softlimit_reclaim_end(sc.nr_reclaimed);

2774
	*nr_scanned = sc.nr_scanned;
2775 2776 2777
	return sc.nr_reclaimed;
}

2778
unsigned long try_to_free_mem_cgroup_pages(struct mem_cgroup *memcg,
2779
					   unsigned long nr_pages,
K
KOSAKI Motohiro 已提交
2780
					   gfp_t gfp_mask,
2781
					   bool may_swap)
2782
{
2783
	struct zonelist *zonelist;
2784
	unsigned long nr_reclaimed;
2785
	int nid;
2786
	struct scan_control sc = {
2787
		.nr_to_reclaim = max(nr_pages, SWAP_CLUSTER_MAX),
2788 2789
		.gfp_mask = (gfp_mask & GFP_RECLAIM_MASK) |
				(GFP_HIGHUSER_MOVABLE & ~GFP_RECLAIM_MASK),
2790 2791 2792 2793
		.target_mem_cgroup = memcg,
		.priority = DEF_PRIORITY,
		.may_writepage = !laptop_mode,
		.may_unmap = 1,
2794
		.may_swap = may_swap,
2795
	};
2796

2797 2798 2799 2800 2801
	/*
	 * 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.
	 */
2802
	nid = mem_cgroup_select_victim_node(memcg);
2803 2804

	zonelist = NODE_DATA(nid)->node_zonelists;
2805 2806 2807 2808 2809

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

2810
	nr_reclaimed = do_try_to_free_pages(zonelist, &sc);
2811 2812 2813 2814

	trace_mm_vmscan_memcg_reclaim_end(nr_reclaimed);

	return nr_reclaimed;
2815 2816 2817
}
#endif

2818
static void age_active_anon(struct zone *zone, struct scan_control *sc)
2819
{
2820
	struct mem_cgroup *memcg;
2821

2822 2823 2824 2825 2826
	if (!total_swap_pages)
		return;

	memcg = mem_cgroup_iter(NULL, NULL, NULL);
	do {
2827
		struct lruvec *lruvec = mem_cgroup_zone_lruvec(zone, memcg);
2828

2829
		if (inactive_anon_is_low(lruvec))
2830
			shrink_active_list(SWAP_CLUSTER_MAX, lruvec,
2831
					   sc, LRU_ACTIVE_ANON);
2832 2833 2834

		memcg = mem_cgroup_iter(NULL, memcg, NULL);
	} while (memcg);
2835 2836
}

2837 2838 2839 2840 2841 2842 2843
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;

2844 2845
	if (IS_ENABLED(CONFIG_COMPACTION) && order && compaction_suitable(zone,
				order, 0, classzone_idx) == COMPACT_SKIPPED)
2846 2847 2848 2849 2850
		return false;

	return true;
}

2851
/*
2852 2853 2854 2855 2856 2857 2858 2859 2860 2861
 * 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.
2862 2863 2864 2865
 * 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
2866
 *     percentage of the middle zones. For example, on 32-bit x86, highmem
2867 2868 2869 2870
 *     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.
 */
2871
static bool pgdat_balanced(pg_data_t *pgdat, int order, int classzone_idx)
2872
{
2873
	unsigned long managed_pages = 0;
2874
	unsigned long balanced_pages = 0;
2875 2876
	int i;

2877 2878 2879
	/* Check the watermark levels */
	for (i = 0; i <= classzone_idx; i++) {
		struct zone *zone = pgdat->node_zones + i;
2880

2881 2882 2883
		if (!populated_zone(zone))
			continue;

2884
		managed_pages += zone->managed_pages;
2885 2886 2887 2888 2889 2890 2891 2892

		/*
		 * 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!
		 */
2893
		if (!zone_reclaimable(zone)) {
2894
			balanced_pages += zone->managed_pages;
2895 2896 2897 2898
			continue;
		}

		if (zone_balanced(zone, order, 0, i))
2899
			balanced_pages += zone->managed_pages;
2900 2901 2902 2903 2904
		else if (!order)
			return false;
	}

	if (order)
2905
		return balanced_pages >= (managed_pages >> 2);
2906 2907
	else
		return true;
2908 2909
}

2910 2911 2912 2913 2914 2915 2916
/*
 * 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,
2917
					int classzone_idx)
2918 2919 2920
{
	/* If a direct reclaimer woke kswapd within HZ/10, it's premature */
	if (remaining)
2921 2922 2923
		return false;

	/*
2924 2925 2926 2927 2928 2929 2930 2931 2932 2933 2934
	 * The throttled processes are normally woken up in balance_pgdat() as
	 * soon as pfmemalloc_watermark_ok() is true. But there is a potential
	 * race between when kswapd checks the watermarks and a process gets
	 * throttled. There is also a potential race if processes get
	 * throttled, kswapd wakes, a large process exits thereby balancing the
	 * zones, which causes kswapd to exit balance_pgdat() before reaching
	 * the wake up checks. If kswapd is going to sleep, no process should
	 * be sleeping on pfmemalloc_wait, so wake them now if necessary. If
	 * the wake up is premature, processes will wake kswapd and get
	 * throttled again. The difference from wake ups in balance_pgdat() is
	 * that here we are under prepare_to_wait().
2935
	 */
2936 2937
	if (waitqueue_active(&pgdat->pfmemalloc_wait))
		wake_up_all(&pgdat->pfmemalloc_wait);
2938

2939
	return pgdat_balanced(pgdat, order, classzone_idx);
2940 2941
}

2942 2943 2944
/*
 * kswapd shrinks the zone by the number of pages required to reach
 * the high watermark.
2945 2946
 *
 * Returns true if kswapd scanned at least the requested number of pages to
2947 2948
 * 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.
2949
 */
2950
static bool kswapd_shrink_zone(struct zone *zone,
2951
			       int classzone_idx,
2952
			       struct scan_control *sc,
2953
			       unsigned long *nr_attempted)
2954
{
2955 2956 2957
	int testorder = sc->order;
	unsigned long balance_gap;
	bool lowmem_pressure;
2958 2959 2960

	/* Reclaim above the high watermark. */
	sc->nr_to_reclaim = max(SWAP_CLUSTER_MAX, high_wmark_pages(zone));
2961 2962 2963 2964 2965 2966 2967 2968

	/*
	 * 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 &&
2969 2970
			compaction_suitable(zone, sc->order, 0, classzone_idx)
							!= COMPACT_SKIPPED)
2971 2972 2973 2974 2975 2976 2977 2978
		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.
	 */
2979 2980
	balance_gap = min(low_wmark_pages(zone), DIV_ROUND_UP(
			zone->managed_pages, KSWAPD_ZONE_BALANCE_GAP_RATIO));
2981 2982 2983 2984 2985 2986 2987 2988 2989 2990

	/*
	 * 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;

2991
	shrink_zone(zone, sc, zone_idx(zone) == classzone_idx);
2992

2993 2994 2995
	/* Account for the number of pages attempted to reclaim */
	*nr_attempted += sc->nr_to_reclaim;

J
Johannes Weiner 已提交
2996
	clear_bit(ZONE_WRITEBACK, &zone->flags);
2997

2998 2999 3000 3001 3002 3003
	/*
	 * 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.
	 */
3004
	if (zone_reclaimable(zone) &&
3005
	    zone_balanced(zone, testorder, 0, classzone_idx)) {
J
Johannes Weiner 已提交
3006 3007
		clear_bit(ZONE_CONGESTED, &zone->flags);
		clear_bit(ZONE_DIRTY, &zone->flags);
3008 3009
	}

3010
	return sc->nr_scanned >= sc->nr_to_reclaim;
3011 3012
}

L
Linus Torvalds 已提交
3013 3014
/*
 * For kswapd, balance_pgdat() will work across all this node's zones until
3015
 * they are all at high_wmark_pages(zone).
L
Linus Torvalds 已提交
3016
 *
3017
 * Returns the final order kswapd was reclaiming at
L
Linus Torvalds 已提交
3018 3019 3020 3021 3022 3023 3024 3025 3026 3027
 *
 * 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
3028 3029 3030 3031 3032
 * 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
3033
 */
3034
static unsigned long balance_pgdat(pg_data_t *pgdat, int order,
3035
							int *classzone_idx)
L
Linus Torvalds 已提交
3036 3037
{
	int i;
3038
	int end_zone = 0;	/* Inclusive.  0 = ZONE_DMA */
3039 3040
	unsigned long nr_soft_reclaimed;
	unsigned long nr_soft_scanned;
3041 3042
	struct scan_control sc = {
		.gfp_mask = GFP_KERNEL,
3043
		.order = order,
3044
		.priority = DEF_PRIORITY,
3045
		.may_writepage = !laptop_mode,
3046
		.may_unmap = 1,
3047
		.may_swap = 1,
3048
	};
3049
	count_vm_event(PAGEOUTRUN);
L
Linus Torvalds 已提交
3050

3051
	do {
3052
		unsigned long nr_attempted = 0;
3053
		bool raise_priority = true;
3054
		bool pgdat_needs_compaction = (order > 0);
3055 3056

		sc.nr_reclaimed = 0;
L
Linus Torvalds 已提交
3057

3058 3059 3060 3061 3062 3063
		/*
		 * 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
3064

3065 3066
			if (!populated_zone(zone))
				continue;
L
Linus Torvalds 已提交
3067

3068 3069
			if (sc.priority != DEF_PRIORITY &&
			    !zone_reclaimable(zone))
3070
				continue;
L
Linus Torvalds 已提交
3071

3072 3073 3074 3075
			/*
			 * Do some background aging of the anon list, to give
			 * pages a chance to be referenced before reclaiming.
			 */
3076
			age_active_anon(zone, &sc);
3077

3078 3079 3080 3081 3082 3083 3084 3085 3086 3087 3088
			/*
			 * 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;
			}

3089
			if (!zone_balanced(zone, order, 0, 0)) {
3090
				end_zone = i;
A
Andrew Morton 已提交
3091
				break;
3092
			} else {
3093 3094 3095 3096
				/*
				 * If balanced, clear the dirty and congested
				 * flags
				 */
J
Johannes Weiner 已提交
3097 3098
				clear_bit(ZONE_CONGESTED, &zone->flags);
				clear_bit(ZONE_DIRTY, &zone->flags);
L
Linus Torvalds 已提交
3099 3100
			}
		}
3101

3102
		if (i < 0)
A
Andrew Morton 已提交
3103 3104
			goto out;

L
Linus Torvalds 已提交
3105 3106 3107
		for (i = 0; i <= end_zone; i++) {
			struct zone *zone = pgdat->node_zones + i;

3108 3109 3110 3111 3112 3113 3114 3115 3116 3117 3118 3119 3120
			if (!populated_zone(zone))
				continue;

			/*
			 * 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
3121 3122
		}

3123 3124 3125 3126 3127 3128 3129
		/*
		 * 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
3130 3131 3132 3133 3134 3135 3136 3137 3138 3139 3140 3141
		/*
		 * 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;

3142
			if (!populated_zone(zone))
L
Linus Torvalds 已提交
3143 3144
				continue;

3145 3146
			if (sc.priority != DEF_PRIORITY &&
			    !zone_reclaimable(zone))
L
Linus Torvalds 已提交
3147 3148 3149
				continue;

			sc.nr_scanned = 0;
3150

3151 3152 3153 3154 3155 3156 3157 3158 3159
			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;

3160
			/*
3161 3162 3163 3164
			 * 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.
3165
			 */
3166 3167
			if (kswapd_shrink_zone(zone, end_zone,
					       &sc, &nr_attempted))
3168
				raise_priority = false;
L
Linus Torvalds 已提交
3169
		}
3170 3171 3172 3173 3174 3175 3176 3177 3178 3179

		/*
		 * 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
3180
		/*
3181 3182 3183 3184 3185 3186
		 * 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
3187
		 */
3188 3189
		if (order && sc.nr_reclaimed >= 2UL << order)
			order = sc.order = 0;
3190

3191 3192 3193
		/* Check if kswapd should be suspending */
		if (try_to_freeze() || kthread_should_stop())
			break;
3194

3195 3196 3197 3198 3199 3200 3201
		/*
		 * 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);

3202
		/*
3203 3204
		 * Raise priority if scanning rate is too low or there was no
		 * progress in reclaiming pages
3205
		 */
3206 3207
		if (raise_priority || !sc.nr_reclaimed)
			sc.priority--;
3208
	} while (sc.priority >= 1 &&
3209
		 !pgdat_balanced(pgdat, order, *classzone_idx));
L
Linus Torvalds 已提交
3210

3211
out:
3212
	/*
3213
	 * Return the order we were reclaiming at so prepare_kswapd_sleep()
3214 3215 3216 3217
	 * 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
	 */
3218
	*classzone_idx = end_zone;
3219
	return order;
L
Linus Torvalds 已提交
3220 3221
}

3222
static void kswapd_try_to_sleep(pg_data_t *pgdat, int order, int classzone_idx)
3223 3224 3225 3226 3227 3228 3229 3230 3231 3232
{
	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 */
3233
	if (prepare_kswapd_sleep(pgdat, order, remaining, classzone_idx)) {
3234 3235 3236 3237 3238 3239 3240 3241 3242
		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.
	 */
3243
	if (prepare_kswapd_sleep(pgdat, order, remaining, classzone_idx)) {
3244 3245 3246 3247 3248 3249 3250 3251 3252 3253 3254
		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);
3255

3256 3257 3258 3259 3260 3261 3262 3263
		/*
		 * 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);

3264 3265 3266
		if (!kthread_should_stop())
			schedule();

3267 3268 3269 3270 3271 3272 3273 3274 3275 3276
		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
3277 3278
/*
 * The background pageout daemon, started as a kernel thread
3279
 * from the init process.
L
Linus Torvalds 已提交
3280 3281 3282 3283 3284 3285 3286 3287 3288 3289 3290 3291
 *
 * 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)
{
3292
	unsigned long order, new_order;
3293
	unsigned balanced_order;
3294
	int classzone_idx, new_classzone_idx;
3295
	int balanced_classzone_idx;
L
Linus Torvalds 已提交
3296 3297
	pg_data_t *pgdat = (pg_data_t*)p;
	struct task_struct *tsk = current;
3298

L
Linus Torvalds 已提交
3299 3300 3301
	struct reclaim_state reclaim_state = {
		.reclaimed_slab = 0,
	};
3302
	const struct cpumask *cpumask = cpumask_of_node(pgdat->node_id);
L
Linus Torvalds 已提交
3303

3304 3305
	lockdep_set_current_reclaim_state(GFP_KERNEL);

R
Rusty Russell 已提交
3306
	if (!cpumask_empty(cpumask))
3307
		set_cpus_allowed_ptr(tsk, cpumask);
L
Linus Torvalds 已提交
3308 3309 3310 3311 3312 3313 3314 3315 3316 3317 3318 3319 3320 3321
	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).
	 */
3322
	tsk->flags |= PF_MEMALLOC | PF_SWAPWRITE | PF_KSWAPD;
3323
	set_freezable();
L
Linus Torvalds 已提交
3324

3325
	order = new_order = 0;
3326
	balanced_order = 0;
3327
	classzone_idx = new_classzone_idx = pgdat->nr_zones - 1;
3328
	balanced_classzone_idx = classzone_idx;
L
Linus Torvalds 已提交
3329
	for ( ; ; ) {
3330
		bool ret;
3331

3332 3333 3334 3335 3336
		/*
		 * 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
		 */
3337 3338
		if (balanced_classzone_idx >= new_classzone_idx &&
					balanced_order == new_order) {
3339 3340 3341 3342 3343 3344
			new_order = pgdat->kswapd_max_order;
			new_classzone_idx = pgdat->classzone_idx;
			pgdat->kswapd_max_order =  0;
			pgdat->classzone_idx = pgdat->nr_zones - 1;
		}

3345
		if (order < new_order || classzone_idx > new_classzone_idx) {
L
Linus Torvalds 已提交
3346 3347
			/*
			 * Don't sleep if someone wants a larger 'order'
3348
			 * allocation or has tigher zone constraints
L
Linus Torvalds 已提交
3349 3350
			 */
			order = new_order;
3351
			classzone_idx = new_classzone_idx;
L
Linus Torvalds 已提交
3352
		} else {
3353 3354
			kswapd_try_to_sleep(pgdat, balanced_order,
						balanced_classzone_idx);
L
Linus Torvalds 已提交
3355
			order = pgdat->kswapd_max_order;
3356
			classzone_idx = pgdat->classzone_idx;
3357 3358
			new_order = order;
			new_classzone_idx = classzone_idx;
3359
			pgdat->kswapd_max_order = 0;
3360
			pgdat->classzone_idx = pgdat->nr_zones - 1;
L
Linus Torvalds 已提交
3361 3362
		}

3363 3364 3365 3366 3367 3368 3369 3370
		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
		 */
3371 3372
		if (!ret) {
			trace_mm_vmscan_kswapd_wake(pgdat->node_id, order);
3373 3374 3375
			balanced_classzone_idx = classzone_idx;
			balanced_order = balance_pgdat(pgdat, order,
						&balanced_classzone_idx);
3376
		}
L
Linus Torvalds 已提交
3377
	}
3378

3379
	tsk->flags &= ~(PF_MEMALLOC | PF_SWAPWRITE | PF_KSWAPD);
3380
	current->reclaim_state = NULL;
3381 3382
	lockdep_clear_current_reclaim_state();

L
Linus Torvalds 已提交
3383 3384 3385 3386 3387 3388
	return 0;
}

/*
 * A zone is low on free memory, so wake its kswapd task to service it.
 */
3389
void wakeup_kswapd(struct zone *zone, int order, enum zone_type classzone_idx)
L
Linus Torvalds 已提交
3390 3391 3392
{
	pg_data_t *pgdat;

3393
	if (!populated_zone(zone))
L
Linus Torvalds 已提交
3394 3395
		return;

3396
	if (!cpuset_zone_allowed(zone, GFP_KERNEL | __GFP_HARDWALL))
L
Linus Torvalds 已提交
3397
		return;
3398
	pgdat = zone->zone_pgdat;
3399
	if (pgdat->kswapd_max_order < order) {
L
Linus Torvalds 已提交
3400
		pgdat->kswapd_max_order = order;
3401 3402
		pgdat->classzone_idx = min(pgdat->classzone_idx, classzone_idx);
	}
3403
	if (!waitqueue_active(&pgdat->kswapd_wait))
L
Linus Torvalds 已提交
3404
		return;
3405
	if (zone_balanced(zone, order, 0, 0))
3406 3407 3408
		return;

	trace_mm_vmscan_wakeup_kswapd(pgdat->node_id, zone_idx(zone), order);
3409
	wake_up_interruptible(&pgdat->kswapd_wait);
L
Linus Torvalds 已提交
3410 3411
}

3412
#ifdef CONFIG_HIBERNATION
L
Linus Torvalds 已提交
3413
/*
3414
 * Try to free `nr_to_reclaim' of memory, system-wide, and return the number of
3415 3416 3417 3418 3419
 * 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
3420
 */
3421
unsigned long shrink_all_memory(unsigned long nr_to_reclaim)
L
Linus Torvalds 已提交
3422
{
3423 3424
	struct reclaim_state reclaim_state;
	struct scan_control sc = {
3425
		.nr_to_reclaim = nr_to_reclaim,
3426
		.gfp_mask = GFP_HIGHUSER_MOVABLE,
3427
		.priority = DEF_PRIORITY,
3428
		.may_writepage = 1,
3429 3430
		.may_unmap = 1,
		.may_swap = 1,
3431
		.hibernation_mode = 1,
L
Linus Torvalds 已提交
3432
	};
3433
	struct zonelist *zonelist = node_zonelist(numa_node_id(), sc.gfp_mask);
3434 3435
	struct task_struct *p = current;
	unsigned long nr_reclaimed;
L
Linus Torvalds 已提交
3436

3437 3438 3439 3440
	p->flags |= PF_MEMALLOC;
	lockdep_set_current_reclaim_state(sc.gfp_mask);
	reclaim_state.reclaimed_slab = 0;
	p->reclaim_state = &reclaim_state;
3441

3442
	nr_reclaimed = do_try_to_free_pages(zonelist, &sc);
3443

3444 3445 3446
	p->reclaim_state = NULL;
	lockdep_clear_current_reclaim_state();
	p->flags &= ~PF_MEMALLOC;
3447

3448
	return nr_reclaimed;
L
Linus Torvalds 已提交
3449
}
3450
#endif /* CONFIG_HIBERNATION */
L
Linus Torvalds 已提交
3451 3452 3453 3454 3455

/* 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. */
3456 3457
static int cpu_callback(struct notifier_block *nfb, unsigned long action,
			void *hcpu)
L
Linus Torvalds 已提交
3458
{
3459
	int nid;
L
Linus Torvalds 已提交
3460

3461
	if (action == CPU_ONLINE || action == CPU_ONLINE_FROZEN) {
3462
		for_each_node_state(nid, N_MEMORY) {
3463
			pg_data_t *pgdat = NODE_DATA(nid);
3464 3465 3466
			const struct cpumask *mask;

			mask = cpumask_of_node(pgdat->node_id);
3467

3468
			if (cpumask_any_and(cpu_online_mask, mask) < nr_cpu_ids)
L
Linus Torvalds 已提交
3469
				/* One of our CPUs online: restore mask */
3470
				set_cpus_allowed_ptr(pgdat->kswapd, mask);
L
Linus Torvalds 已提交
3471 3472 3473 3474 3475
		}
	}
	return NOTIFY_OK;
}

3476 3477 3478 3479 3480 3481 3482 3483 3484 3485 3486 3487 3488 3489 3490 3491
/*
 * 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);
3492 3493
		pr_err("Failed to start kswapd on node %d\n", nid);
		ret = PTR_ERR(pgdat->kswapd);
3494
		pgdat->kswapd = NULL;
3495 3496 3497 3498
	}
	return ret;
}

3499
/*
3500
 * Called by memory hotplug when all memory in a node is offlined.  Caller must
3501
 * hold mem_hotplug_begin/end().
3502 3503 3504 3505 3506
 */
void kswapd_stop(int nid)
{
	struct task_struct *kswapd = NODE_DATA(nid)->kswapd;

3507
	if (kswapd) {
3508
		kthread_stop(kswapd);
3509 3510
		NODE_DATA(nid)->kswapd = NULL;
	}
3511 3512
}

L
Linus Torvalds 已提交
3513 3514
static int __init kswapd_init(void)
{
3515
	int nid;
3516

L
Linus Torvalds 已提交
3517
	swap_setup();
3518
	for_each_node_state(nid, N_MEMORY)
3519
 		kswapd_run(nid);
L
Linus Torvalds 已提交
3520 3521 3522 3523 3524
	hotcpu_notifier(cpu_callback, 0);
	return 0;
}

module_init(kswapd_init)
3525 3526 3527 3528 3529 3530 3531 3532 3533 3534

#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;

3535
#define RECLAIM_OFF 0
3536
#define RECLAIM_ZONE (1<<0)	/* Run shrink_inactive_list on the zone */
3537 3538 3539
#define RECLAIM_WRITE (1<<1)	/* Writeout pages during reclaim */
#define RECLAIM_SWAP (1<<2)	/* Swap pages out during reclaim */

3540 3541 3542 3543 3544 3545 3546
/*
 * 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

3547 3548 3549 3550 3551 3552
/*
 * Percentage of pages in a zone that must be unmapped for zone_reclaim to
 * occur.
 */
int sysctl_min_unmapped_ratio = 1;

3553 3554 3555 3556 3557 3558
/*
 * 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;

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 3596 3597 3598 3599 3600
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;
}

3601 3602 3603
/*
 * Try to free up some pages from this zone through reclaim.
 */
3604
static int __zone_reclaim(struct zone *zone, gfp_t gfp_mask, unsigned int order)
3605
{
3606
	/* Minimum pages needed in order to stay on node */
3607
	const unsigned long nr_pages = 1 << order;
3608 3609
	struct task_struct *p = current;
	struct reclaim_state reclaim_state;
3610
	struct scan_control sc = {
3611
		.nr_to_reclaim = max(nr_pages, SWAP_CLUSTER_MAX),
3612
		.gfp_mask = (gfp_mask = memalloc_noio_flags(gfp_mask)),
3613
		.order = order,
3614
		.priority = ZONE_RECLAIM_PRIORITY,
3615 3616 3617
		.may_writepage = !!(zone_reclaim_mode & RECLAIM_WRITE),
		.may_unmap = !!(zone_reclaim_mode & RECLAIM_SWAP),
		.may_swap = 1,
3618
	};
3619 3620

	cond_resched();
3621 3622 3623 3624 3625 3626
	/*
	 * 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;
3627
	lockdep_set_current_reclaim_state(gfp_mask);
3628 3629
	reclaim_state.reclaimed_slab = 0;
	p->reclaim_state = &reclaim_state;
3630

3631
	if (zone_pagecache_reclaimable(zone) > zone->min_unmapped_pages) {
3632 3633 3634 3635 3636
		/*
		 * Free memory by calling shrink zone with increasing
		 * priorities until we have enough memory freed.
		 */
		do {
3637
			shrink_zone(zone, &sc, true);
3638
		} while (sc.nr_reclaimed < nr_pages && --sc.priority >= 0);
3639
	}
3640

3641
	p->reclaim_state = NULL;
3642
	current->flags &= ~(PF_MEMALLOC | PF_SWAPWRITE);
3643
	lockdep_clear_current_reclaim_state();
3644
	return sc.nr_reclaimed >= nr_pages;
3645
}
3646 3647 3648 3649

int zone_reclaim(struct zone *zone, gfp_t gfp_mask, unsigned int order)
{
	int node_id;
3650
	int ret;
3651 3652

	/*
3653 3654
	 * Zone reclaim reclaims unmapped file backed pages and
	 * slab pages if we are over the defined limits.
3655
	 *
3656 3657 3658 3659 3660
	 * 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.
3661
	 */
3662 3663
	if (zone_pagecache_reclaimable(zone) <= zone->min_unmapped_pages &&
	    zone_page_state(zone, NR_SLAB_RECLAIMABLE) <= zone->min_slab_pages)
3664
		return ZONE_RECLAIM_FULL;
3665

3666
	if (!zone_reclaimable(zone))
3667
		return ZONE_RECLAIM_FULL;
3668

3669
	/*
3670
	 * Do not scan if the allocation should not be delayed.
3671
	 */
3672
	if (!(gfp_mask & __GFP_WAIT) || (current->flags & PF_MEMALLOC))
3673
		return ZONE_RECLAIM_NOSCAN;
3674 3675 3676 3677 3678 3679 3680

	/*
	 * 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.
	 */
3681
	node_id = zone_to_nid(zone);
3682
	if (node_state(node_id, N_CPU) && node_id != numa_node_id())
3683
		return ZONE_RECLAIM_NOSCAN;
3684

J
Johannes Weiner 已提交
3685
	if (test_and_set_bit(ZONE_RECLAIM_LOCKED, &zone->flags))
3686 3687
		return ZONE_RECLAIM_NOSCAN;

3688
	ret = __zone_reclaim(zone, gfp_mask, order);
J
Johannes Weiner 已提交
3689
	clear_bit(ZONE_RECLAIM_LOCKED, &zone->flags);
3690

3691 3692 3693
	if (!ret)
		count_vm_event(PGSCAN_ZONE_RECLAIM_FAILED);

3694
	return ret;
3695
}
3696
#endif
L
Lee Schermerhorn 已提交
3697 3698 3699 3700 3701 3702

/*
 * 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
3703
 * lists vs unevictable list.
L
Lee Schermerhorn 已提交
3704 3705
 *
 * Reasons page might not be evictable:
3706
 * (1) page's mapping marked unevictable
N
Nick Piggin 已提交
3707
 * (2) page is part of an mlocked VMA
3708
 *
L
Lee Schermerhorn 已提交
3709
 */
3710
int page_evictable(struct page *page)
L
Lee Schermerhorn 已提交
3711
{
3712
	return !mapping_unevictable(page_mapping(page)) && !PageMlocked(page);
L
Lee Schermerhorn 已提交
3713
}
3714

3715
#ifdef CONFIG_SHMEM
3716
/**
3717 3718 3719
 * 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
3720
 *
3721
 * Checks pages for evictability and moves them to the appropriate lru list.
3722 3723
 *
 * This function is only used for SysV IPC SHM_UNLOCK.
3724
 */
3725
void check_move_unevictable_pages(struct page **pages, int nr_pages)
3726
{
3727
	struct lruvec *lruvec;
3728 3729 3730 3731
	struct zone *zone = NULL;
	int pgscanned = 0;
	int pgrescued = 0;
	int i;
3732

3733 3734 3735
	for (i = 0; i < nr_pages; i++) {
		struct page *page = pages[i];
		struct zone *pagezone;
3736

3737 3738 3739 3740 3741 3742 3743 3744
		pgscanned++;
		pagezone = page_zone(page);
		if (pagezone != zone) {
			if (zone)
				spin_unlock_irq(&zone->lru_lock);
			zone = pagezone;
			spin_lock_irq(&zone->lru_lock);
		}
3745
		lruvec = mem_cgroup_page_lruvec(page, zone);
3746

3747 3748
		if (!PageLRU(page) || !PageUnevictable(page))
			continue;
3749

3750
		if (page_evictable(page)) {
3751 3752
			enum lru_list lru = page_lru_base_type(page);

3753
			VM_BUG_ON_PAGE(PageActive(page), page);
3754
			ClearPageUnevictable(page);
3755 3756
			del_page_from_lru_list(page, lruvec, LRU_UNEVICTABLE);
			add_page_to_lru_list(page, lruvec, lru);
3757
			pgrescued++;
3758
		}
3759
	}
3760

3761 3762 3763 3764
	if (zone) {
		__count_vm_events(UNEVICTABLE_PGRESCUED, pgrescued);
		__count_vm_events(UNEVICTABLE_PGSCANNED, pgscanned);
		spin_unlock_irq(&zone->lru_lock);
3765 3766
	}
}
3767
#endif /* CONFIG_SHMEM */