vmscan.c 110.4 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 62 63 64
struct scan_control {
	/* Incremented by the number of inactive pages that were scanned */
	unsigned long nr_scanned;

65 66 67
	/* Number of pages freed so far during a call to shrink_zones() */
	unsigned long nr_reclaimed;

68 69 70
	/* How many pages shrink_list() should reclaim */
	unsigned long nr_to_reclaim;

71 72
	unsigned long hibernation_mode;

L
Linus Torvalds 已提交
73
	/* This context's GFP mask */
A
Al Viro 已提交
74
	gfp_t gfp_mask;
L
Linus Torvalds 已提交
75 76 77

	int may_writepage;

78 79
	/* Can mapped pages be reclaimed? */
	int may_unmap;
80

81 82 83
	/* Can pages be swapped as part of reclaim? */
	int may_swap;

A
Andy Whitcroft 已提交
84
	int order;
85

86 87 88
	/* Scan (total_size >> priority) pages at once */
	int priority;

89 90 91
	/* anon vs. file LRUs scanning "ratio" */
	int swappiness;

92 93 94 95 96
	/*
	 * 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;
97

98 99 100 101 102
	/*
	 * Nodemask of nodes allowed by the caller. If NULL, all nodes
	 * are scanned.
	 */
	nodemask_t	*nodemask;
L
Linus Torvalds 已提交
103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138
};

#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;
139
unsigned long vm_total_pages;	/* The total number of pages which the VM controls */
L
Linus Torvalds 已提交
140 141 142 143

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

A
Andrew Morton 已提交
144
#ifdef CONFIG_MEMCG
145 146
static bool global_reclaim(struct scan_control *sc)
{
147
	return !sc->target_mem_cgroup;
148
}
149
#else
150 151 152 153
static bool global_reclaim(struct scan_control *sc)
{
	return true;
}
154 155
#endif

156
static unsigned long zone_reclaimable_pages(struct zone *zone)
157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174
{
	int nr;

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

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

	return nr;
}

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

175
static unsigned long get_lru_size(struct lruvec *lruvec, enum lru_list lru)
176
{
177
	if (!mem_cgroup_disabled())
178
		return mem_cgroup_get_lru_size(lruvec, lru);
179

180
	return zone_page_state(lruvec_zone(lruvec), NR_LRU_BASE + lru);
181 182
}

L
Linus Torvalds 已提交
183
/*
G
Glauber Costa 已提交
184
 * Add a shrinker callback to be called from the vm.
L
Linus Torvalds 已提交
185
 */
G
Glauber Costa 已提交
186
int register_shrinker(struct shrinker *shrinker)
L
Linus Torvalds 已提交
187
{
G
Glauber Costa 已提交
188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204
	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;

205 206 207
	down_write(&shrinker_rwsem);
	list_add_tail(&shrinker->list, &shrinker_list);
	up_write(&shrinker_rwsem);
G
Glauber Costa 已提交
208
	return 0;
L
Linus Torvalds 已提交
209
}
210
EXPORT_SYMBOL(register_shrinker);
L
Linus Torvalds 已提交
211 212 213 214

/*
 * Remove one
 */
215
void unregister_shrinker(struct shrinker *shrinker)
L
Linus Torvalds 已提交
216 217 218 219
{
	down_write(&shrinker_rwsem);
	list_del(&shrinker->list);
	up_write(&shrinker_rwsem);
220
	kfree(shrinker->nr_deferred);
L
Linus Torvalds 已提交
221
}
222
EXPORT_SYMBOL(unregister_shrinker);
L
Linus Torvalds 已提交
223 224

#define SHRINK_BATCH 128
G
Glauber Costa 已提交
225 226 227 228 229 230 231 232

static unsigned long
shrink_slab_node(struct shrink_control *shrinkctl, struct shrinker *shrinker,
		 unsigned long nr_pages_scanned, unsigned long lru_pages)
{
	unsigned long freed = 0;
	unsigned long long delta;
	long total_scan;
233
	long freeable;
G
Glauber Costa 已提交
234 235 236 237 238 239
	long nr;
	long new_nr;
	int nid = shrinkctl->nid;
	long batch_size = shrinker->batch ? shrinker->batch
					  : SHRINK_BATCH;

240 241
	freeable = shrinker->count_objects(shrinker, shrinkctl);
	if (freeable == 0)
G
Glauber Costa 已提交
242 243 244 245 246 247 248 249 250 251 252
		return 0;

	/*
	 * copy the current shrinker scan count into a local variable
	 * and zero it so that other concurrent shrinker invocations
	 * don't also do this scanning work.
	 */
	nr = atomic_long_xchg(&shrinker->nr_deferred[nid], 0);

	total_scan = nr;
	delta = (4 * nr_pages_scanned) / shrinker->seeks;
253
	delta *= freeable;
G
Glauber Costa 已提交
254 255 256 257 258
	do_div(delta, lru_pages + 1);
	total_scan += delta;
	if (total_scan < 0) {
		printk(KERN_ERR
		"shrink_slab: %pF negative objects to delete nr=%ld\n",
D
Dave Chinner 已提交
259
		       shrinker->scan_objects, total_scan);
260
		total_scan = freeable;
G
Glauber Costa 已提交
261 262 263 264 265 266 267 268
	}

	/*
	 * 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 >>>
269
	 * freeable. This is bad for sustaining a working set in
G
Glauber Costa 已提交
270 271 272 273 274
	 * memory.
	 *
	 * Hence only allow the shrinker to scan the entire cache when
	 * a large delta change is calculated directly.
	 */
275 276
	if (delta < freeable / 4)
		total_scan = min(total_scan, freeable / 2);
G
Glauber Costa 已提交
277 278 279 280 281 282

	/*
	 * Avoid risking looping forever due to too large nr value:
	 * never try to free more than twice the estimate number of
	 * freeable entries.
	 */
283 284
	if (total_scan > freeable * 2)
		total_scan = freeable * 2;
G
Glauber Costa 已提交
285 286 287

	trace_mm_shrink_slab_start(shrinker, shrinkctl, nr,
				nr_pages_scanned, lru_pages,
288
				freeable, delta, total_scan);
G
Glauber Costa 已提交
289

290 291 292 293 294 295 296 297 298 299 300
	/*
	 * 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
301
	 * than the total number of objects on slab (freeable), we must be
302 303 304 305
	 * scanning at high prio and therefore should try to reclaim as much as
	 * possible.
	 */
	while (total_scan >= batch_size ||
306
	       total_scan >= freeable) {
D
Dave Chinner 已提交
307
		unsigned long ret;
308
		unsigned long nr_to_scan = min(batch_size, total_scan);
G
Glauber Costa 已提交
309

310
		shrinkctl->nr_to_scan = nr_to_scan;
D
Dave Chinner 已提交
311 312 313 314
		ret = shrinker->scan_objects(shrinker, shrinkctl);
		if (ret == SHRINK_STOP)
			break;
		freed += ret;
G
Glauber Costa 已提交
315

316 317
		count_vm_events(SLABS_SCANNED, nr_to_scan);
		total_scan -= nr_to_scan;
G
Glauber Costa 已提交
318 319 320 321 322 323 324 325 326 327 328 329 330 331 332

		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]);

333
	trace_mm_shrink_slab_end(shrinker, nid, freed, nr, new_nr, total_scan);
G
Glauber Costa 已提交
334
	return freed;
335 336
}

L
Linus Torvalds 已提交
337 338 339 340 341 342 343 344
/*
 * Call the shrink functions to age shrinkable caches
 *
 * Here we assume it costs one seek to replace a lru page and that it also
 * takes a seek to recreate a cache object.  With this in mind we age equal
 * percentages of the lru and ageable caches.  This should balance the seeks
 * generated by these structures.
 *
S
Simon Arlott 已提交
345
 * If the vm encountered mapped pages on the LRU it increase the pressure on
L
Linus Torvalds 已提交
346 347 348 349 350 351 352
 * slab to avoid swapping.
 *
 * We do weird things to avoid (scanned*seeks*entries) overflowing 32 bits.
 *
 * `lru_pages' represents the number of on-LRU pages in all the zones which
 * are eligible for the caller's allocation attempt.  It is used for balancing
 * slab reclaim versus page reclaim.
353 354
 *
 * Returns the number of slab objects which we shrunk.
L
Linus Torvalds 已提交
355
 */
D
Dave Chinner 已提交
356
unsigned long shrink_slab(struct shrink_control *shrinkctl,
357
			  unsigned long nr_pages_scanned,
358
			  unsigned long lru_pages)
L
Linus Torvalds 已提交
359 360
{
	struct shrinker *shrinker;
D
Dave Chinner 已提交
361
	unsigned long freed = 0;
L
Linus Torvalds 已提交
362

363 364
	if (nr_pages_scanned == 0)
		nr_pages_scanned = SWAP_CLUSTER_MAX;
L
Linus Torvalds 已提交
365

366
	if (!down_read_trylock(&shrinker_rwsem)) {
D
Dave Chinner 已提交
367 368 369 370 371 372 373
		/*
		 * 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;
374 375
		goto out;
	}
L
Linus Torvalds 已提交
376 377

	list_for_each_entry(shrinker, &shrinker_list, list) {
378 379
		if (!(shrinker->flags & SHRINKER_NUMA_AWARE)) {
			shrinkctl->nid = 0;
G
Glauber Costa 已提交
380
			freed += shrink_slab_node(shrinkctl, shrinker,
381 382 383 384 385 386 387 388
					nr_pages_scanned, lru_pages);
			continue;
		}

		for_each_node_mask(shrinkctl->nid, shrinkctl->nodes_to_scan) {
			if (node_online(shrinkctl->nid))
				freed += shrink_slab_node(shrinkctl, shrinker,
						nr_pages_scanned, lru_pages);
L
Linus Torvalds 已提交
389 390 391 392

		}
	}
	up_read(&shrinker_rwsem);
393 394
out:
	cond_resched();
D
Dave Chinner 已提交
395
	return freed;
L
Linus Torvalds 已提交
396 397 398 399
}

static inline int is_page_cache_freeable(struct page *page)
{
400 401 402 403 404
	/*
	 * 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.
	 */
405
	return page_count(page) - page_has_private(page) == 2;
L
Linus Torvalds 已提交
406 407
}

408 409
static int may_write_to_queue(struct backing_dev_info *bdi,
			      struct scan_control *sc)
L
Linus Torvalds 已提交
410
{
411
	if (current->flags & PF_SWAPWRITE)
L
Linus Torvalds 已提交
412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434
		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
435
	lock_page(page);
436 437
	if (page_mapping(page) == mapping)
		mapping_set_error(mapping, error);
L
Linus Torvalds 已提交
438 439 440
	unlock_page(page);
}

441 442 443 444 445 446 447 448 449 450 451 452
/* 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
453
/*
A
Andrew Morton 已提交
454 455
 * pageout is called by shrink_page_list() for each dirty page.
 * Calls ->writepage().
L
Linus Torvalds 已提交
456
 */
457
static pageout_t pageout(struct page *page, struct address_space *mapping,
458
			 struct scan_control *sc)
L
Linus Torvalds 已提交
459 460 461 462 463 464 465 466
{
	/*
	 * 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.
	 *
467
	 * If this process is currently in __generic_file_aio_write() against
L
Linus Torvalds 已提交
468 469 470 471 472 473 474 475 476 477 478 479 480 481 482
	 * 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.
		 */
483
		if (page_has_private(page)) {
L
Linus Torvalds 已提交
484 485
			if (try_to_free_buffers(page)) {
				ClearPageDirty(page);
486
				pr_info("%s: orphaned page\n", __func__);
L
Linus Torvalds 已提交
487 488 489 490 491 492 493
				return PAGE_CLEAN;
			}
		}
		return PAGE_KEEP;
	}
	if (mapping->a_ops->writepage == NULL)
		return PAGE_ACTIVATE;
494
	if (!may_write_to_queue(mapping->backing_dev_info, sc))
L
Linus Torvalds 已提交
495 496 497 498 499 500 501
		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,
502 503
			.range_start = 0,
			.range_end = LLONG_MAX,
L
Linus Torvalds 已提交
504 505 506 507 508 509 510
			.for_reclaim = 1,
		};

		SetPageReclaim(page);
		res = mapping->a_ops->writepage(page, &wbc);
		if (res < 0)
			handle_write_error(mapping, page, res);
511
		if (res == AOP_WRITEPAGE_ACTIVATE) {
L
Linus Torvalds 已提交
512 513 514
			ClearPageReclaim(page);
			return PAGE_ACTIVATE;
		}
515

L
Linus Torvalds 已提交
516 517 518 519
		if (!PageWriteback(page)) {
			/* synchronous write or broken a_ops? */
			ClearPageReclaim(page);
		}
M
Mel Gorman 已提交
520
		trace_mm_vmscan_writepage(page, trace_reclaim_flags(page));
521
		inc_zone_page_state(page, NR_VMSCAN_WRITE);
L
Linus Torvalds 已提交
522 523 524 525 526 527
		return PAGE_SUCCESS;
	}

	return PAGE_CLEAN;
}

528
/*
N
Nick Piggin 已提交
529 530
 * Same as remove_mapping, but if the page is removed from the mapping, it
 * gets returned with a refcount of 0.
531
 */
532 533
static int __remove_mapping(struct address_space *mapping, struct page *page,
			    bool reclaimed)
534
{
535 536
	BUG_ON(!PageLocked(page));
	BUG_ON(mapping != page_mapping(page));
537

N
Nick Piggin 已提交
538
	spin_lock_irq(&mapping->tree_lock);
539
	/*
N
Nick Piggin 已提交
540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562
	 * 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.
563
	 */
N
Nick Piggin 已提交
564
	if (!page_freeze_refs(page, 2))
565
		goto cannot_free;
N
Nick Piggin 已提交
566 567 568
	/* note: atomic_cmpxchg in page_freeze_refs provides the smp_rmb */
	if (unlikely(PageDirty(page))) {
		page_unfreeze_refs(page, 2);
569
		goto cannot_free;
N
Nick Piggin 已提交
570
	}
571 572 573 574

	if (PageSwapCache(page)) {
		swp_entry_t swap = { .val = page_private(page) };
		__delete_from_swap_cache(page);
N
Nick Piggin 已提交
575
		spin_unlock_irq(&mapping->tree_lock);
576
		swapcache_free(swap, page);
N
Nick Piggin 已提交
577
	} else {
578
		void (*freepage)(struct page *);
579
		void *shadow = NULL;
580 581

		freepage = mapping->a_ops->freepage;
582 583 584 585 586 587 588 589 590 591 592 593 594 595
		/*
		 * 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
596
		spin_unlock_irq(&mapping->tree_lock);
597
		mem_cgroup_uncharge_cache_page(page);
598 599 600

		if (freepage != NULL)
			freepage(page);
601 602 603 604 605
	}

	return 1;

cannot_free:
N
Nick Piggin 已提交
606
	spin_unlock_irq(&mapping->tree_lock);
607 608 609
	return 0;
}

N
Nick Piggin 已提交
610 611 612 613 614 615 616 617
/*
 * 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)
{
618
	if (__remove_mapping(mapping, page, false)) {
N
Nick Piggin 已提交
619 620 621 622 623 624 625 626 627 628 629
		/*
		 * 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
630 631 632 633 634 635 636 637 638 639 640
/**
 * 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)
{
641
	bool is_unevictable;
642
	int was_unevictable = PageUnevictable(page);
L
Lee Schermerhorn 已提交
643

644
	VM_BUG_ON_PAGE(PageLRU(page), page);
L
Lee Schermerhorn 已提交
645 646 647 648

redo:
	ClearPageUnevictable(page);

649
	if (page_evictable(page)) {
L
Lee Schermerhorn 已提交
650 651 652 653 654 655
		/*
		 * 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.
		 */
656
		is_unevictable = false;
657
		lru_cache_add(page);
L
Lee Schermerhorn 已提交
658 659 660 661 662
	} else {
		/*
		 * Put unevictable pages directly on zone's unevictable
		 * list.
		 */
663
		is_unevictable = true;
L
Lee Schermerhorn 已提交
664
		add_page_to_unevictable_list(page);
665
		/*
666 667 668
		 * 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
669
		 * isolation/check_move_unevictable_pages,
670
		 * we see PG_mlocked/AS_UNEVICTABLE cleared below and move
671 672
		 * the page back to the evictable list.
		 *
673
		 * The other side is TestClearPageMlocked() or shmem_lock().
674 675
		 */
		smp_mb();
L
Lee Schermerhorn 已提交
676 677 678 679 680 681 682
	}

	/*
	 * 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.
	 */
683
	if (is_unevictable && page_evictable(page)) {
L
Lee Schermerhorn 已提交
684 685 686 687 688 689 690 691 692 693
		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.
		 */
	}

694
	if (was_unevictable && !is_unevictable)
695
		count_vm_event(UNEVICTABLE_PGRESCUED);
696
	else if (!was_unevictable && is_unevictable)
697 698
		count_vm_event(UNEVICTABLE_PGCULLED);

L
Lee Schermerhorn 已提交
699 700 701
	put_page(page);		/* drop ref from isolate */
}

702 703 704
enum page_references {
	PAGEREF_RECLAIM,
	PAGEREF_RECLAIM_CLEAN,
705
	PAGEREF_KEEP,
706 707 708 709 710 711
	PAGEREF_ACTIVATE,
};

static enum page_references page_check_references(struct page *page,
						  struct scan_control *sc)
{
712
	int referenced_ptes, referenced_page;
713 714
	unsigned long vm_flags;

715 716
	referenced_ptes = page_referenced(page, 1, sc->target_mem_cgroup,
					  &vm_flags);
717
	referenced_page = TestClearPageReferenced(page);
718 719 720 721 722 723 724 725

	/*
	 * 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;

726
	if (referenced_ptes) {
727
		if (PageSwapBacked(page))
728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744
			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);

745
		if (referenced_page || referenced_ptes > 1)
746 747
			return PAGEREF_ACTIVATE;

748 749 750 751 752 753
		/*
		 * Activate file-backed executable pages after first usage.
		 */
		if (vm_flags & VM_EXEC)
			return PAGEREF_ACTIVATE;

754 755
		return PAGEREF_KEEP;
	}
756 757

	/* Reclaim if clean, defer dirty pages to writeback */
758
	if (referenced_page && !PageSwapBacked(page))
759 760 761
		return PAGEREF_RECLAIM_CLEAN;

	return PAGEREF_RECLAIM;
762 763
}

764 765 766 767
/* Check if a page is dirty or under writeback */
static void page_check_dirty_writeback(struct page *page,
				       bool *dirty, bool *writeback)
{
768 769
	struct address_space *mapping;

770 771 772 773 774 775 776 777 778 779 780 781 782
	/*
	 * 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);
783 784 785 786 787 788 789 790

	/* 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);
791 792
}

L
Linus Torvalds 已提交
793
/*
A
Andrew Morton 已提交
794
 * shrink_page_list() returns the number of reclaimed pages
L
Linus Torvalds 已提交
795
 */
A
Andrew Morton 已提交
796
static unsigned long shrink_page_list(struct list_head *page_list,
797
				      struct zone *zone,
798
				      struct scan_control *sc,
799
				      enum ttu_flags ttu_flags,
800
				      unsigned long *ret_nr_dirty,
801
				      unsigned long *ret_nr_unqueued_dirty,
802
				      unsigned long *ret_nr_congested,
803
				      unsigned long *ret_nr_writeback,
804
				      unsigned long *ret_nr_immediate,
805
				      bool force_reclaim)
L
Linus Torvalds 已提交
806 807
{
	LIST_HEAD(ret_pages);
808
	LIST_HEAD(free_pages);
L
Linus Torvalds 已提交
809
	int pgactivate = 0;
810
	unsigned long nr_unqueued_dirty = 0;
811 812
	unsigned long nr_dirty = 0;
	unsigned long nr_congested = 0;
813
	unsigned long nr_reclaimed = 0;
814
	unsigned long nr_writeback = 0;
815
	unsigned long nr_immediate = 0;
L
Linus Torvalds 已提交
816 817 818

	cond_resched();

819
	mem_cgroup_uncharge_start();
L
Linus Torvalds 已提交
820 821 822 823
	while (!list_empty(page_list)) {
		struct address_space *mapping;
		struct page *page;
		int may_enter_fs;
824
		enum page_references references = PAGEREF_RECLAIM_CLEAN;
825
		bool dirty, writeback;
L
Linus Torvalds 已提交
826 827 828 829 830 831

		cond_resched();

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

N
Nick Piggin 已提交
832
		if (!trylock_page(page))
L
Linus Torvalds 已提交
833 834
			goto keep;

835 836
		VM_BUG_ON_PAGE(PageActive(page), page);
		VM_BUG_ON_PAGE(page_zone(page) != zone, page);
L
Linus Torvalds 已提交
837 838

		sc->nr_scanned++;
839

840
		if (unlikely(!page_evictable(page)))
N
Nick Piggin 已提交
841
			goto cull_mlocked;
L
Lee Schermerhorn 已提交
842

843
		if (!sc->may_unmap && page_mapped(page))
844 845
			goto keep_locked;

L
Linus Torvalds 已提交
846 847 848 849
		/* Double the slab pressure for mapped and swapcache pages */
		if (page_mapped(page) || PageSwapCache(page))
			sc->nr_scanned++;

850 851 852
		may_enter_fs = (sc->gfp_mask & __GFP_FS) ||
			(PageSwapCache(page) && (sc->gfp_mask & __GFP_IO));

853 854 855 856 857 858 859 860 861 862 863 864 865
		/*
		 * 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++;

866 867 868 869 870 871
		/*
		 * 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.
		 */
872
		mapping = page_mapping(page);
873 874
		if ((mapping && bdi_write_congested(mapping->backing_dev_info)) ||
		    (writeback && PageReclaim(page)))
875 876
			nr_congested++;

877 878 879 880 881 882 883 884 885 886 887
		/*
		 * 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
888 889
		 *    note that the LRU is being scanned too quickly and the
		 *    caller can stall after page list has been processed.
890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913
		 *
		 * 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.
		 */
914
		if (PageWriteback(page)) {
915 916 917 918
			/* Case 1 above */
			if (current_is_kswapd() &&
			    PageReclaim(page) &&
			    zone_is_reclaim_writeback(zone)) {
919 920
				nr_immediate++;
				goto keep_locked;
921 922 923

			/* Case 2 above */
			} else if (global_reclaim(sc) ||
924 925 926 927 928 929 930 931 932 933 934 935 936
			    !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);
937
				nr_writeback++;
938

939
				goto keep_locked;
940 941 942 943

			/* Case 3 above */
			} else {
				wait_on_page_writeback(page);
944
			}
945
		}
L
Linus Torvalds 已提交
946

947 948 949
		if (!force_reclaim)
			references = page_check_references(page, sc);

950 951
		switch (references) {
		case PAGEREF_ACTIVATE:
L
Linus Torvalds 已提交
952
			goto activate_locked;
953 954
		case PAGEREF_KEEP:
			goto keep_locked;
955 956 957 958
		case PAGEREF_RECLAIM:
		case PAGEREF_RECLAIM_CLEAN:
			; /* try to reclaim the page below */
		}
L
Linus Torvalds 已提交
959 960 961 962 963

		/*
		 * Anonymous process memory has backing store?
		 * Try to allocate it some swap space here.
		 */
N
Nick Piggin 已提交
964
		if (PageAnon(page) && !PageSwapCache(page)) {
965 966
			if (!(sc->gfp_mask & __GFP_IO))
				goto keep_locked;
967
			if (!add_to_swap(page, page_list))
L
Linus Torvalds 已提交
968
				goto activate_locked;
969
			may_enter_fs = 1;
L
Linus Torvalds 已提交
970

971 972 973
			/* Adding to swap updated mapping */
			mapping = page_mapping(page);
		}
L
Linus Torvalds 已提交
974 975 976 977 978 979

		/*
		 * The page is mapped into the page tables of one or more
		 * processes. Try to unmap it here.
		 */
		if (page_mapped(page) && mapping) {
980
			switch (try_to_unmap(page, ttu_flags)) {
L
Linus Torvalds 已提交
981 982 983 984
			case SWAP_FAIL:
				goto activate_locked;
			case SWAP_AGAIN:
				goto keep_locked;
N
Nick Piggin 已提交
985 986
			case SWAP_MLOCK:
				goto cull_mlocked;
L
Linus Torvalds 已提交
987 988 989 990 991 992
			case SWAP_SUCCESS:
				; /* try to free the page below */
			}
		}

		if (PageDirty(page)) {
993 994
			/*
			 * Only kswapd can writeback filesystem pages to
995 996
			 * avoid risk of stack overflow but only writeback
			 * if many dirty pages have been encountered.
997
			 */
998
			if (page_is_file_cache(page) &&
999
					(!current_is_kswapd() ||
1000
					 !zone_is_reclaim_dirty(zone))) {
1001 1002 1003 1004 1005 1006 1007 1008 1009
				/*
				 * 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);

1010 1011 1012
				goto keep_locked;
			}

1013
			if (references == PAGEREF_RECLAIM_CLEAN)
L
Linus Torvalds 已提交
1014
				goto keep_locked;
1015
			if (!may_enter_fs)
L
Linus Torvalds 已提交
1016
				goto keep_locked;
1017
			if (!sc->may_writepage)
L
Linus Torvalds 已提交
1018 1019 1020
				goto keep_locked;

			/* Page is dirty, try to write it out here */
1021
			switch (pageout(page, mapping, sc)) {
L
Linus Torvalds 已提交
1022 1023 1024 1025 1026
			case PAGE_KEEP:
				goto keep_locked;
			case PAGE_ACTIVATE:
				goto activate_locked;
			case PAGE_SUCCESS:
1027
				if (PageWriteback(page))
1028
					goto keep;
1029
				if (PageDirty(page))
L
Linus Torvalds 已提交
1030
					goto keep;
1031

L
Linus Torvalds 已提交
1032 1033 1034 1035
				/*
				 * A synchronous write - probably a ramdisk.  Go
				 * ahead and try to reclaim the page.
				 */
N
Nick Piggin 已提交
1036
				if (!trylock_page(page))
L
Linus Torvalds 已提交
1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055
					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
1056
		 * will do this, as well as the blockdev mapping.
L
Linus Torvalds 已提交
1057 1058 1059 1060 1061 1062 1063 1064 1065 1066
		 * 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.
		 */
1067
		if (page_has_private(page)) {
L
Linus Torvalds 已提交
1068 1069
			if (!try_to_release_page(page, sc->gfp_mask))
				goto activate_locked;
N
Nick Piggin 已提交
1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085
			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
1086 1087
		}

1088
		if (!mapping || !__remove_mapping(mapping, page, true))
1089
			goto keep_locked;
L
Linus Torvalds 已提交
1090

N
Nick Piggin 已提交
1091 1092 1093 1094 1095 1096 1097 1098
		/*
		 * 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
1099
free_it:
1100
		nr_reclaimed++;
1101 1102 1103 1104 1105 1106

		/*
		 * 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
1107 1108
		continue;

N
Nick Piggin 已提交
1109
cull_mlocked:
1110 1111
		if (PageSwapCache(page))
			try_to_free_swap(page);
N
Nick Piggin 已提交
1112 1113 1114 1115
		unlock_page(page);
		putback_lru_page(page);
		continue;

L
Linus Torvalds 已提交
1116
activate_locked:
1117 1118
		/* Not a candidate for swapping, so reclaim swap space. */
		if (PageSwapCache(page) && vm_swap_full())
1119
			try_to_free_swap(page);
1120
		VM_BUG_ON_PAGE(PageActive(page), page);
L
Linus Torvalds 已提交
1121 1122 1123 1124 1125 1126
		SetPageActive(page);
		pgactivate++;
keep_locked:
		unlock_page(page);
keep:
		list_add(&page->lru, &ret_pages);
1127
		VM_BUG_ON_PAGE(PageLRU(page) || PageUnevictable(page), page);
L
Linus Torvalds 已提交
1128
	}
1129

1130
	free_hot_cold_page_list(&free_pages, true);
1131

L
Linus Torvalds 已提交
1132
	list_splice(&ret_pages, page_list);
1133
	count_vm_events(PGACTIVATE, pgactivate);
1134
	mem_cgroup_uncharge_end();
1135 1136
	*ret_nr_dirty += nr_dirty;
	*ret_nr_congested += nr_congested;
1137
	*ret_nr_unqueued_dirty += nr_unqueued_dirty;
1138
	*ret_nr_writeback += nr_writeback;
1139
	*ret_nr_immediate += nr_immediate;
1140
	return nr_reclaimed;
L
Linus Torvalds 已提交
1141 1142
}

1143 1144 1145 1146 1147 1148 1149 1150
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,
	};
1151
	unsigned long ret, dummy1, dummy2, dummy3, dummy4, dummy5;
1152 1153 1154 1155
	struct page *page, *next;
	LIST_HEAD(clean_pages);

	list_for_each_entry_safe(page, next, page_list, lru) {
1156 1157
		if (page_is_file_cache(page) && !PageDirty(page) &&
		    !isolated_balloon_page(page)) {
1158 1159 1160 1161 1162 1163
			ClearPageActive(page);
			list_move(&page->lru, &clean_pages);
		}
	}

	ret = shrink_page_list(&clean_pages, zone, &sc,
1164 1165
			TTU_UNMAP|TTU_IGNORE_ACCESS,
			&dummy1, &dummy2, &dummy3, &dummy4, &dummy5, true);
1166
	list_splice(&clean_pages, page_list);
1167
	mod_zone_page_state(zone, NR_ISOLATED_FILE, -ret);
1168 1169 1170
	return ret;
}

A
Andy Whitcroft 已提交
1171 1172 1173 1174 1175 1176 1177 1178 1179 1180
/*
 * 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.
 */
1181
int __isolate_lru_page(struct page *page, isolate_mode_t mode)
A
Andy Whitcroft 已提交
1182 1183 1184 1185 1186 1187 1188
{
	int ret = -EINVAL;

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

M
Minchan Kim 已提交
1189 1190
	/* Compaction should not handle unevictable pages but CMA can do so */
	if (PageUnevictable(page) && !(mode & ISOLATE_UNEVICTABLE))
L
Lee Schermerhorn 已提交
1191 1192
		return ret;

A
Andy Whitcroft 已提交
1193
	ret = -EBUSY;
K
KAMEZAWA Hiroyuki 已提交
1194

1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227
	/*
	 * 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;
		}
	}
1228

1229 1230 1231
	if ((mode & ISOLATE_UNMAPPED) && page_mapped(page))
		return ret;

A
Andy Whitcroft 已提交
1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244
	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
1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255
/*
 * 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.
1256
 * @lruvec:	The LRU vector to pull pages from.
L
Linus Torvalds 已提交
1257
 * @dst:	The temp list to put pages on to.
H
Hugh Dickins 已提交
1258
 * @nr_scanned:	The number of pages that were scanned.
1259
 * @sc:		The scan_control struct for this reclaim session
A
Andy Whitcroft 已提交
1260
 * @mode:	One of the LRU isolation modes
1261
 * @lru:	LRU list id for isolating
L
Linus Torvalds 已提交
1262 1263 1264
 *
 * returns how many pages were moved onto *@dst.
 */
1265
static unsigned long isolate_lru_pages(unsigned long nr_to_scan,
1266
		struct lruvec *lruvec, struct list_head *dst,
1267
		unsigned long *nr_scanned, struct scan_control *sc,
1268
		isolate_mode_t mode, enum lru_list lru)
L
Linus Torvalds 已提交
1269
{
H
Hugh Dickins 已提交
1270
	struct list_head *src = &lruvec->lists[lru];
1271
	unsigned long nr_taken = 0;
1272
	unsigned long scan;
L
Linus Torvalds 已提交
1273

1274
	for (scan = 0; scan < nr_to_scan && !list_empty(src); scan++) {
A
Andy Whitcroft 已提交
1275
		struct page *page;
1276
		int nr_pages;
A
Andy Whitcroft 已提交
1277

L
Linus Torvalds 已提交
1278 1279 1280
		page = lru_to_page(src);
		prefetchw_prev_lru_page(page, src, flags);

1281
		VM_BUG_ON_PAGE(!PageLRU(page), page);
N
Nick Piggin 已提交
1282

1283
		switch (__isolate_lru_page(page, mode)) {
A
Andy Whitcroft 已提交
1284
		case 0:
1285 1286
			nr_pages = hpage_nr_pages(page);
			mem_cgroup_update_lru_size(lruvec, lru, -nr_pages);
A
Andy Whitcroft 已提交
1287
			list_move(&page->lru, dst);
1288
			nr_taken += nr_pages;
A
Andy Whitcroft 已提交
1289 1290 1291 1292 1293 1294
			break;

		case -EBUSY:
			/* else it is being freed elsewhere */
			list_move(&page->lru, src);
			continue;
1295

A
Andy Whitcroft 已提交
1296 1297 1298
		default:
			BUG();
		}
L
Linus Torvalds 已提交
1299 1300
	}

H
Hugh Dickins 已提交
1301
	*nr_scanned = scan;
H
Hugh Dickins 已提交
1302 1303
	trace_mm_vmscan_lru_isolate(sc->order, nr_to_scan, scan,
				    nr_taken, mode, is_file_lru(lru));
L
Linus Torvalds 已提交
1304 1305 1306
	return nr_taken;
}

1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317
/**
 * 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
1318 1319 1320
 * 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.
1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335
 *
 * 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;

1336
	VM_BUG_ON_PAGE(!page_count(page), page);
1337

1338 1339
	if (PageLRU(page)) {
		struct zone *zone = page_zone(page);
1340
		struct lruvec *lruvec;
1341 1342

		spin_lock_irq(&zone->lru_lock);
1343
		lruvec = mem_cgroup_page_lruvec(page, zone);
1344
		if (PageLRU(page)) {
L
Lee Schermerhorn 已提交
1345
			int lru = page_lru(page);
1346
			get_page(page);
1347
			ClearPageLRU(page);
1348 1349
			del_page_from_lru_list(page, lruvec, lru);
			ret = 0;
1350 1351 1352 1353 1354 1355
		}
		spin_unlock_irq(&zone->lru_lock);
	}
	return ret;
}

1356
/*
F
Fengguang Wu 已提交
1357 1358 1359 1360 1361
 * 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.
1362 1363 1364 1365 1366 1367 1368 1369 1370
 */
static int too_many_isolated(struct zone *zone, int file,
		struct scan_control *sc)
{
	unsigned long inactive, isolated;

	if (current_is_kswapd())
		return 0;

1371
	if (!global_reclaim(sc))
1372 1373 1374 1375 1376 1377 1378 1379 1380 1381
		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);
	}

1382 1383 1384 1385 1386 1387 1388 1389
	/*
	 * 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;

1390 1391 1392
	return isolated > inactive;
}

1393
static noinline_for_stack void
H
Hugh Dickins 已提交
1394
putback_inactive_pages(struct lruvec *lruvec, struct list_head *page_list)
1395
{
1396 1397
	struct zone_reclaim_stat *reclaim_stat = &lruvec->reclaim_stat;
	struct zone *zone = lruvec_zone(lruvec);
1398
	LIST_HEAD(pages_to_free);
1399 1400 1401 1402 1403

	/*
	 * Put back any unfreeable pages.
	 */
	while (!list_empty(page_list)) {
1404
		struct page *page = lru_to_page(page_list);
1405
		int lru;
1406

1407
		VM_BUG_ON_PAGE(PageLRU(page), page);
1408
		list_del(&page->lru);
1409
		if (unlikely(!page_evictable(page))) {
1410 1411 1412 1413 1414
			spin_unlock_irq(&zone->lru_lock);
			putback_lru_page(page);
			spin_lock_irq(&zone->lru_lock);
			continue;
		}
1415 1416 1417

		lruvec = mem_cgroup_page_lruvec(page, zone);

1418
		SetPageLRU(page);
1419
		lru = page_lru(page);
1420 1421
		add_page_to_lru_list(page, lruvec, lru);

1422 1423
		if (is_active_lru(lru)) {
			int file = is_file_lru(lru);
1424 1425
			int numpages = hpage_nr_pages(page);
			reclaim_stat->recent_rotated[file] += numpages;
1426
		}
1427 1428 1429
		if (put_page_testzero(page)) {
			__ClearPageLRU(page);
			__ClearPageActive(page);
1430
			del_page_from_lru_list(page, lruvec, lru);
1431 1432 1433 1434 1435 1436 1437

			if (unlikely(PageCompound(page))) {
				spin_unlock_irq(&zone->lru_lock);
				(*get_compound_page_dtor(page))(page);
				spin_lock_irq(&zone->lru_lock);
			} else
				list_add(&page->lru, &pages_to_free);
1438 1439 1440
		}
	}

1441 1442 1443 1444
	/*
	 * To save our caller's stack, now use input list for pages to free.
	 */
	list_splice(&pages_to_free, page_list);
1445 1446
}

1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459
/*
 * 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
1460
/*
A
Andrew Morton 已提交
1461 1462
 * shrink_inactive_list() is a helper for shrink_zone().  It returns the number
 * of reclaimed pages
L
Linus Torvalds 已提交
1463
 */
1464
static noinline_for_stack unsigned long
1465
shrink_inactive_list(unsigned long nr_to_scan, struct lruvec *lruvec,
1466
		     struct scan_control *sc, enum lru_list lru)
L
Linus Torvalds 已提交
1467 1468
{
	LIST_HEAD(page_list);
1469
	unsigned long nr_scanned;
1470
	unsigned long nr_reclaimed = 0;
1471
	unsigned long nr_taken;
1472 1473
	unsigned long nr_dirty = 0;
	unsigned long nr_congested = 0;
1474
	unsigned long nr_unqueued_dirty = 0;
1475
	unsigned long nr_writeback = 0;
1476
	unsigned long nr_immediate = 0;
1477
	isolate_mode_t isolate_mode = 0;
1478
	int file = is_file_lru(lru);
1479 1480
	struct zone *zone = lruvec_zone(lruvec);
	struct zone_reclaim_stat *reclaim_stat = &lruvec->reclaim_stat;
1481

1482
	while (unlikely(too_many_isolated(zone, file, sc))) {
1483
		congestion_wait(BLK_RW_ASYNC, HZ/10);
1484 1485 1486 1487 1488 1489

		/* 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
1490
	lru_add_drain();
1491 1492

	if (!sc->may_unmap)
1493
		isolate_mode |= ISOLATE_UNMAPPED;
1494
	if (!sc->may_writepage)
1495
		isolate_mode |= ISOLATE_CLEAN;
1496

L
Linus Torvalds 已提交
1497
	spin_lock_irq(&zone->lru_lock);
1498

1499 1500
	nr_taken = isolate_lru_pages(nr_to_scan, lruvec, &page_list,
				     &nr_scanned, sc, isolate_mode, lru);
1501 1502 1503 1504

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

1505
	if (global_reclaim(sc)) {
1506 1507
		zone->pages_scanned += nr_scanned;
		if (current_is_kswapd())
H
Hugh Dickins 已提交
1508
			__count_zone_vm_events(PGSCAN_KSWAPD, zone, nr_scanned);
1509
		else
H
Hugh Dickins 已提交
1510
			__count_zone_vm_events(PGSCAN_DIRECT, zone, nr_scanned);
1511
	}
1512
	spin_unlock_irq(&zone->lru_lock);
1513

1514
	if (nr_taken == 0)
1515
		return 0;
A
Andy Whitcroft 已提交
1516

1517
	nr_reclaimed = shrink_page_list(&page_list, zone, sc, TTU_UNMAP,
1518 1519 1520
				&nr_dirty, &nr_unqueued_dirty, &nr_congested,
				&nr_writeback, &nr_immediate,
				false);
1521

1522 1523
	spin_lock_irq(&zone->lru_lock);

1524
	reclaim_stat->recent_scanned[file] += nr_taken;
1525

Y
Ying Han 已提交
1526 1527 1528 1529 1530 1531 1532 1533
	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
1534

1535
	putback_inactive_pages(lruvec, &page_list);
1536

1537
	__mod_zone_page_state(zone, NR_ISOLATED_ANON + file, -nr_taken);
1538 1539 1540

	spin_unlock_irq(&zone->lru_lock);

1541
	free_hot_cold_page_list(&page_list, true);
1542

1543 1544 1545 1546 1547 1548 1549 1550 1551 1552
	/*
	 * 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.
	 *
1553 1554 1555
	 * 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.
1556
	 */
1557
	if (nr_writeback && nr_writeback == nr_taken)
1558
		zone_set_flag(zone, ZONE_WRITEBACK);
1559

1560
	/*
1561 1562
	 * memcg will stall in page writeback so only consider forcibly
	 * stalling for global reclaim
1563
	 */
1564
	if (global_reclaim(sc)) {
1565 1566 1567 1568 1569 1570 1571
		/*
		 * Tag a zone as congested if all the dirty pages scanned were
		 * backed by a congested BDI and wait_iff_congested will stall.
		 */
		if (nr_dirty && nr_dirty == nr_congested)
			zone_set_flag(zone, ZONE_CONGESTED);

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

		/*
1582 1583 1584
		 * 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
1585 1586
		 * they are written so also forcibly stall.
		 */
1587
		if (nr_immediate && current_may_throttle())
1588
			congestion_wait(BLK_RW_ASYNC, HZ/10);
1589
	}
1590

1591 1592 1593 1594 1595
	/*
	 * 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.
	 */
1596 1597
	if (!sc->hibernation_mode && !current_is_kswapd() &&
	    current_may_throttle())
1598 1599
		wait_iff_congested(zone, BLK_RW_ASYNC, HZ/10);

1600 1601 1602
	trace_mm_vmscan_lru_shrink_inactive(zone->zone_pgdat->node_id,
		zone_idx(zone),
		nr_scanned, nr_reclaimed,
1603
		sc->priority,
M
Mel Gorman 已提交
1604
		trace_shrink_flags(file));
1605
	return nr_reclaimed;
L
Linus Torvalds 已提交
1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622 1623 1624
}

/*
 * 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.
 */
1625

1626
static void move_active_pages_to_lru(struct lruvec *lruvec,
1627
				     struct list_head *list,
1628
				     struct list_head *pages_to_free,
1629 1630
				     enum lru_list lru)
{
1631
	struct zone *zone = lruvec_zone(lruvec);
1632 1633
	unsigned long pgmoved = 0;
	struct page *page;
1634
	int nr_pages;
1635 1636 1637

	while (!list_empty(list)) {
		page = lru_to_page(list);
1638
		lruvec = mem_cgroup_page_lruvec(page, zone);
1639

1640
		VM_BUG_ON_PAGE(PageLRU(page), page);
1641 1642
		SetPageLRU(page);

1643 1644
		nr_pages = hpage_nr_pages(page);
		mem_cgroup_update_lru_size(lruvec, lru, nr_pages);
1645
		list_move(&page->lru, &lruvec->lists[lru]);
1646
		pgmoved += nr_pages;
1647

1648 1649 1650
		if (put_page_testzero(page)) {
			__ClearPageLRU(page);
			__ClearPageActive(page);
1651
			del_page_from_lru_list(page, lruvec, lru);
1652 1653 1654 1655 1656 1657 1658

			if (unlikely(PageCompound(page))) {
				spin_unlock_irq(&zone->lru_lock);
				(*get_compound_page_dtor(page))(page);
				spin_lock_irq(&zone->lru_lock);
			} else
				list_add(&page->lru, pages_to_free);
1659 1660 1661 1662 1663 1664
		}
	}
	__mod_zone_page_state(zone, NR_LRU_BASE + lru, pgmoved);
	if (!is_active_lru(lru))
		__count_vm_events(PGDEACTIVATE, pgmoved);
}
1665

H
Hugh Dickins 已提交
1666
static void shrink_active_list(unsigned long nr_to_scan,
1667
			       struct lruvec *lruvec,
1668
			       struct scan_control *sc,
1669
			       enum lru_list lru)
L
Linus Torvalds 已提交
1670
{
1671
	unsigned long nr_taken;
H
Hugh Dickins 已提交
1672
	unsigned long nr_scanned;
1673
	unsigned long vm_flags;
L
Linus Torvalds 已提交
1674
	LIST_HEAD(l_hold);	/* The pages which were snipped off */
1675
	LIST_HEAD(l_active);
1676
	LIST_HEAD(l_inactive);
L
Linus Torvalds 已提交
1677
	struct page *page;
1678
	struct zone_reclaim_stat *reclaim_stat = &lruvec->reclaim_stat;
1679
	unsigned long nr_rotated = 0;
1680
	isolate_mode_t isolate_mode = 0;
1681
	int file = is_file_lru(lru);
1682
	struct zone *zone = lruvec_zone(lruvec);
L
Linus Torvalds 已提交
1683 1684

	lru_add_drain();
1685 1686

	if (!sc->may_unmap)
1687
		isolate_mode |= ISOLATE_UNMAPPED;
1688
	if (!sc->may_writepage)
1689
		isolate_mode |= ISOLATE_CLEAN;
1690

L
Linus Torvalds 已提交
1691
	spin_lock_irq(&zone->lru_lock);
1692

1693 1694
	nr_taken = isolate_lru_pages(nr_to_scan, lruvec, &l_hold,
				     &nr_scanned, sc, isolate_mode, lru);
1695
	if (global_reclaim(sc))
H
Hugh Dickins 已提交
1696
		zone->pages_scanned += nr_scanned;
1697

1698
	reclaim_stat->recent_scanned[file] += nr_taken;
1699

H
Hugh Dickins 已提交
1700
	__count_zone_vm_events(PGREFILL, zone, nr_scanned);
1701
	__mod_zone_page_state(zone, NR_LRU_BASE + lru, -nr_taken);
K
KOSAKI Motohiro 已提交
1702
	__mod_zone_page_state(zone, NR_ISOLATED_ANON + file, nr_taken);
L
Linus Torvalds 已提交
1703 1704 1705 1706 1707 1708
	spin_unlock_irq(&zone->lru_lock);

	while (!list_empty(&l_hold)) {
		cond_resched();
		page = lru_to_page(&l_hold);
		list_del(&page->lru);
1709

1710
		if (unlikely(!page_evictable(page))) {
L
Lee Schermerhorn 已提交
1711 1712 1713 1714
			putback_lru_page(page);
			continue;
		}

1715 1716 1717 1718 1719 1720 1721 1722
		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);
			}
		}

1723 1724
		if (page_referenced(page, 0, sc->target_mem_cgroup,
				    &vm_flags)) {
1725
			nr_rotated += hpage_nr_pages(page);
1726 1727 1728 1729 1730 1731 1732 1733 1734
			/*
			 * 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.
			 */
1735
			if ((vm_flags & VM_EXEC) && page_is_file_cache(page)) {
1736 1737 1738 1739
				list_add(&page->lru, &l_active);
				continue;
			}
		}
1740

1741
		ClearPageActive(page);	/* we are de-activating */
L
Linus Torvalds 已提交
1742 1743 1744
		list_add(&page->lru, &l_inactive);
	}

1745
	/*
1746
	 * Move pages back to the lru list.
1747
	 */
1748
	spin_lock_irq(&zone->lru_lock);
1749
	/*
1750 1751 1752 1753
	 * Count referenced pages from currently used mappings as rotated,
	 * even though only some of them are actually re-activated.  This
	 * helps balance scan pressure between file and anonymous pages in
	 * get_scan_ratio.
1754
	 */
1755
	reclaim_stat->recent_rotated[file] += nr_rotated;
1756

1757 1758
	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
1759
	__mod_zone_page_state(zone, NR_ISOLATED_ANON + file, -nr_taken);
1760
	spin_unlock_irq(&zone->lru_lock);
1761

1762
	free_hot_cold_page_list(&l_hold, true);
L
Linus Torvalds 已提交
1763 1764
}

1765
#ifdef CONFIG_SWAP
1766
static int inactive_anon_is_low_global(struct zone *zone)
1767 1768 1769 1770 1771 1772 1773 1774 1775 1776 1777 1778
{
	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;
}

1779 1780
/**
 * inactive_anon_is_low - check if anonymous pages need to be deactivated
1781
 * @lruvec: LRU vector to check
1782 1783 1784 1785
 *
 * Returns true if the zone does not have enough inactive anon pages,
 * meaning some active anon pages need to be deactivated.
 */
1786
static int inactive_anon_is_low(struct lruvec *lruvec)
1787
{
1788 1789 1790 1791 1792 1793 1794
	/*
	 * If we don't have swap space, anonymous page deactivation
	 * is pointless.
	 */
	if (!total_swap_pages)
		return 0;

1795
	if (!mem_cgroup_disabled())
1796
		return mem_cgroup_inactive_anon_is_low(lruvec);
1797

1798
	return inactive_anon_is_low_global(lruvec_zone(lruvec));
1799
}
1800
#else
1801
static inline int inactive_anon_is_low(struct lruvec *lruvec)
1802 1803 1804 1805
{
	return 0;
}
#endif
1806

1807 1808
/**
 * inactive_file_is_low - check if file pages need to be deactivated
1809
 * @lruvec: LRU vector to check
1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820
 *
 * 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.
 */
1821
static int inactive_file_is_low(struct lruvec *lruvec)
1822
{
1823 1824 1825 1826 1827
	unsigned long inactive;
	unsigned long active;

	inactive = get_lru_size(lruvec, LRU_INACTIVE_FILE);
	active = get_lru_size(lruvec, LRU_ACTIVE_FILE);
1828

1829
	return active > inactive;
1830 1831
}

H
Hugh Dickins 已提交
1832
static int inactive_list_is_low(struct lruvec *lruvec, enum lru_list lru)
1833
{
H
Hugh Dickins 已提交
1834
	if (is_file_lru(lru))
1835
		return inactive_file_is_low(lruvec);
1836
	else
1837
		return inactive_anon_is_low(lruvec);
1838 1839
}

1840
static unsigned long shrink_list(enum lru_list lru, unsigned long nr_to_scan,
1841
				 struct lruvec *lruvec, struct scan_control *sc)
1842
{
1843
	if (is_active_lru(lru)) {
H
Hugh Dickins 已提交
1844
		if (inactive_list_is_low(lruvec, lru))
1845
			shrink_active_list(nr_to_scan, lruvec, sc, lru);
1846 1847 1848
		return 0;
	}

1849
	return shrink_inactive_list(nr_to_scan, lruvec, sc, lru);
1850 1851
}

1852 1853 1854 1855 1856 1857 1858
enum scan_balance {
	SCAN_EQUAL,
	SCAN_FRACT,
	SCAN_ANON,
	SCAN_FILE,
};

1859 1860 1861 1862 1863 1864
/*
 * 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
1865 1866
 * 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
1867
 */
1868
static void get_scan_count(struct lruvec *lruvec, struct scan_control *sc,
1869
			   unsigned long *nr)
1870
{
1871 1872 1873 1874
	struct zone_reclaim_stat *reclaim_stat = &lruvec->reclaim_stat;
	u64 fraction[2];
	u64 denominator = 0;	/* gcc */
	struct zone *zone = lruvec_zone(lruvec);
1875
	unsigned long anon_prio, file_prio;
1876
	enum scan_balance scan_balance;
1877
	unsigned long anon, file;
1878
	bool force_scan = false;
1879
	unsigned long ap, fp;
H
Hugh Dickins 已提交
1880
	enum lru_list lru;
1881 1882
	bool some_scanned;
	int pass;
1883

1884 1885 1886 1887 1888 1889 1890 1891 1892 1893
	/*
	 * 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.
	 */
1894
	if (current_is_kswapd() && !zone_reclaimable(zone))
1895
		force_scan = true;
1896
	if (!global_reclaim(sc))
1897
		force_scan = true;
1898 1899

	/* If we have no swap space, do not bother scanning anon pages. */
1900
	if (!sc->may_swap || (get_nr_swap_pages() <= 0)) {
1901
		scan_balance = SCAN_FILE;
1902 1903
		goto out;
	}
1904

1905 1906 1907 1908 1909 1910 1911
	/*
	 * 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.
	 */
1912
	if (!global_reclaim(sc) && !sc->swappiness) {
1913
		scan_balance = SCAN_FILE;
1914 1915 1916 1917 1918 1919 1920 1921
		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).
	 */
1922
	if (!sc->priority && sc->swappiness) {
1923
		scan_balance = SCAN_EQUAL;
1924 1925 1926
		goto out;
	}

1927 1928 1929 1930
	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);
1931

1932 1933 1934 1935 1936 1937 1938 1939 1940 1941 1942 1943 1944 1945 1946 1947 1948 1949
	/*
	 * Prevent the reclaimer from falling into the cache trap: as
	 * cache pages start out inactive, every cache fault will tip
	 * the scan balance towards the file LRU.  And as the file LRU
	 * shrinks, so does the window for rotation from references.
	 * This means we have a runaway feedback loop where a tiny
	 * thrashing file LRU becomes infinitely more attractive than
	 * anon pages.  Try to detect this based on file LRU size.
	 */
	if (global_reclaim(sc)) {
		unsigned long free = zone_page_state(zone, NR_FREE_PAGES);

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

1950 1951 1952 1953 1954
	/*
	 * There is enough inactive page cache, do not reclaim
	 * anything from the anonymous working set right now.
	 */
	if (!inactive_file_is_low(lruvec)) {
1955
		scan_balance = SCAN_FILE;
1956 1957 1958
		goto out;
	}

1959 1960
	scan_balance = SCAN_FRACT;

1961 1962 1963 1964
	/*
	 * With swappiness at 100, anonymous and file have the same priority.
	 * This scanning priority is essentially the inverse of IO cost.
	 */
1965
	anon_prio = sc->swappiness;
H
Hugh Dickins 已提交
1966
	file_prio = 200 - anon_prio;
1967

1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978
	/*
	 * 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]
	 */
1979
	spin_lock_irq(&zone->lru_lock);
1980 1981 1982
	if (unlikely(reclaim_stat->recent_scanned[0] > anon / 4)) {
		reclaim_stat->recent_scanned[0] /= 2;
		reclaim_stat->recent_rotated[0] /= 2;
1983 1984
	}

1985 1986 1987
	if (unlikely(reclaim_stat->recent_scanned[1] > file / 4)) {
		reclaim_stat->recent_scanned[1] /= 2;
		reclaim_stat->recent_rotated[1] /= 2;
1988 1989 1990
	}

	/*
1991 1992 1993
	 * 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.
1994
	 */
1995
	ap = anon_prio * (reclaim_stat->recent_scanned[0] + 1);
1996
	ap /= reclaim_stat->recent_rotated[0] + 1;
1997

1998
	fp = file_prio * (reclaim_stat->recent_scanned[1] + 1);
1999
	fp /= reclaim_stat->recent_rotated[1] + 1;
2000
	spin_unlock_irq(&zone->lru_lock);
2001

2002 2003 2004 2005
	fraction[0] = ap;
	fraction[1] = fp;
	denominator = ap + fp + 1;
out:
2006 2007 2008 2009 2010 2011 2012
	some_scanned = false;
	/* Only use force_scan on second pass. */
	for (pass = 0; !some_scanned && pass < 2; pass++) {
		for_each_evictable_lru(lru) {
			int file = is_file_lru(lru);
			unsigned long size;
			unsigned long scan;
2013

2014 2015
			size = get_lru_size(lruvec, lru);
			scan = size >> sc->priority;
2016

2017 2018
			if (!scan && pass && force_scan)
				scan = min(size, SWAP_CLUSTER_MAX);
2019

2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042
			switch (scan_balance) {
			case SCAN_EQUAL:
				/* Scan lists relative to size */
				break;
			case SCAN_FRACT:
				/*
				 * Scan types proportional to swappiness and
				 * their relative recent reclaim efficiency.
				 */
				scan = div64_u64(scan * fraction[file],
							denominator);
				break;
			case SCAN_FILE:
			case SCAN_ANON:
				/* Scan one type exclusively */
				if ((scan_balance == SCAN_FILE) != file)
					scan = 0;
				break;
			default:
				/* Look ma, no brain */
				BUG();
			}
			nr[lru] = scan;
2043
			/*
2044 2045
			 * Skip the second pass and don't force_scan,
			 * if we found something to scan.
2046
			 */
2047
			some_scanned |= !!scan;
2048
		}
2049
	}
2050
}
2051

2052 2053 2054 2055 2056 2057
/*
 * This is a basic per-zone page freer.  Used by both kswapd and direct reclaim.
 */
static void shrink_lruvec(struct lruvec *lruvec, struct scan_control *sc)
{
	unsigned long nr[NR_LRU_LISTS];
2058
	unsigned long targets[NR_LRU_LISTS];
2059 2060 2061 2062 2063
	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;
2064
	bool scan_adjusted;
2065 2066 2067

	get_scan_count(lruvec, sc, nr);

2068 2069 2070
	/* Record the original scan target for proportional adjustments later */
	memcpy(targets, nr, sizeof(nr));

2071 2072 2073 2074 2075 2076 2077 2078 2079 2080 2081 2082 2083 2084
	/*
	 * 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);

2085 2086 2087
	blk_start_plug(&plug);
	while (nr[LRU_INACTIVE_ANON] || nr[LRU_ACTIVE_FILE] ||
					nr[LRU_INACTIVE_FILE]) {
2088 2089 2090
		unsigned long nr_anon, nr_file, percentage;
		unsigned long nr_scanned;

2091 2092 2093 2094 2095 2096 2097 2098 2099
		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);
			}
		}
2100 2101 2102 2103 2104 2105

		if (nr_reclaimed < nr_to_reclaim || scan_adjusted)
			continue;

		/*
		 * For kswapd and memcg, reclaim at least the number of pages
2106
		 * requested. Ensure that the anon and file LRUs are scanned
2107 2108 2109 2110 2111 2112 2113
		 * 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];

2114 2115 2116 2117 2118 2119 2120 2121 2122
		/*
		 * 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;

2123 2124 2125 2126 2127 2128 2129 2130 2131 2132 2133 2134 2135 2136 2137 2138 2139 2140 2141 2142 2143 2144 2145 2146 2147 2148 2149 2150 2151 2152 2153
		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;
2154 2155 2156 2157 2158 2159 2160 2161 2162 2163 2164 2165 2166 2167 2168
	}
	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
2169
/* Use reclaim/compaction for costly allocs or under memory pressure */
2170
static bool in_reclaim_compaction(struct scan_control *sc)
M
Mel Gorman 已提交
2171
{
2172
	if (IS_ENABLED(CONFIG_COMPACTION) && sc->order &&
M
Mel Gorman 已提交
2173
			(sc->order > PAGE_ALLOC_COSTLY_ORDER ||
2174
			 sc->priority < DEF_PRIORITY - 2))
M
Mel Gorman 已提交
2175 2176 2177 2178 2179
		return true;

	return false;
}

2180
/*
M
Mel Gorman 已提交
2181 2182 2183 2184 2185
 * 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.
2186
 */
2187
static inline bool should_continue_reclaim(struct zone *zone,
2188 2189 2190 2191 2192 2193 2194 2195
					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 */
2196
	if (!in_reclaim_compaction(sc))
2197 2198
		return false;

2199 2200 2201 2202 2203 2204 2205 2206 2207 2208 2209 2210 2211 2212 2213 2214 2215 2216 2217 2218 2219 2220
	/* 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;
	}
2221 2222 2223 2224 2225 2226

	/*
	 * If we have not reclaimed enough pages for compaction and the
	 * inactive lists are large enough, continue reclaiming
	 */
	pages_for_compaction = (2UL << sc->order);
2227
	inactive_lru_pages = zone_page_state(zone, NR_INACTIVE_FILE);
2228
	if (get_nr_swap_pages() > 0)
2229
		inactive_lru_pages += zone_page_state(zone, NR_INACTIVE_ANON);
2230 2231 2232 2233 2234
	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 */
2235
	switch (compaction_suitable(zone, sc->order)) {
2236 2237 2238 2239 2240 2241 2242 2243
	case COMPACT_PARTIAL:
	case COMPACT_CONTINUE:
		return false;
	default:
		return true;
	}
}

2244
static void shrink_zone(struct zone *zone, struct scan_control *sc)
L
Linus Torvalds 已提交
2245
{
2246
	unsigned long nr_reclaimed, nr_scanned;
L
Linus Torvalds 已提交
2247

2248 2249 2250 2251 2252 2253
	do {
		struct mem_cgroup *root = sc->target_mem_cgroup;
		struct mem_cgroup_reclaim_cookie reclaim = {
			.zone = zone,
			.priority = sc->priority,
		};
2254
		struct mem_cgroup *memcg;
2255

2256 2257
		nr_reclaimed = sc->nr_reclaimed;
		nr_scanned = sc->nr_scanned;
L
Linus Torvalds 已提交
2258

2259 2260
		memcg = mem_cgroup_iter(root, NULL, &reclaim);
		do {
2261
			struct lruvec *lruvec;
2262

2263
			lruvec = mem_cgroup_zone_lruvec(zone, memcg);
2264

2265
			sc->swappiness = mem_cgroup_swappiness(memcg);
2266
			shrink_lruvec(lruvec, sc);
2267

2268
			/*
2269 2270
			 * Direct reclaim and kswapd have to scan all memory
			 * cgroups to fulfill the overall scan target for the
2271
			 * zone.
2272 2273 2274 2275 2276
			 *
			 * 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.
2277
			 */
2278 2279
			if (!global_reclaim(sc) &&
					sc->nr_reclaimed >= sc->nr_to_reclaim) {
2280 2281 2282
				mem_cgroup_iter_break(root, memcg);
				break;
			}
2283 2284
			memcg = mem_cgroup_iter(root, memcg, &reclaim);
		} while (memcg);
2285 2286 2287 2288 2289

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

2290 2291
	} while (should_continue_reclaim(zone, sc->nr_reclaimed - nr_reclaimed,
					 sc->nr_scanned - nr_scanned, sc));
2292 2293
}

2294 2295 2296 2297 2298 2299 2300 2301 2302 2303 2304 2305 2306 2307 2308 2309
/* Returns true if compaction should go ahead for a high-order request */
static inline bool compaction_ready(struct zone *zone, struct scan_control *sc)
{
	unsigned long balance_gap, watermark;
	bool watermark_ok;

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

	/*
	 * Compaction takes time to run and there are potentially other
	 * callers using the pages just freed. Continue reclaiming until
	 * there is a buffer of free pages available to give compaction
	 * a reasonable chance of completing and allocating the page
	 */
2310 2311
	balance_gap = min(low_wmark_pages(zone), DIV_ROUND_UP(
			zone->managed_pages, KSWAPD_ZONE_BALANCE_GAP_RATIO));
2312 2313 2314 2315 2316 2317 2318
	watermark = high_wmark_pages(zone) + balance_gap + (2UL << sc->order);
	watermark_ok = zone_watermark_ok_safe(zone, 0, watermark, 0, 0);

	/*
	 * If compaction is deferred, reclaim up to a point where
	 * compaction will have a chance of success when re-enabled
	 */
2319
	if (compaction_deferred(zone, sc->order))
2320 2321 2322 2323 2324 2325 2326 2327 2328
		return watermark_ok;

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

	return watermark_ok;
}

L
Linus Torvalds 已提交
2329 2330 2331 2332 2333
/*
 * 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.
 *
2334 2335
 * We reclaim from a zone even if that zone is over high_wmark_pages(zone).
 * Because:
L
Linus Torvalds 已提交
2336 2337
 * a) The caller may be trying to free *extra* pages to satisfy a higher-order
 *    allocation or
2338 2339 2340
 * 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
2341 2342 2343
 *
 * If a zone is deemed to be full of pinned pages then just give it a light
 * scan then give up on it.
2344 2345
 *
 * This function returns true if a zone is being reclaimed for a costly
2346
 * high-order allocation and compaction is ready to begin. This indicates to
2347 2348
 * the caller that it should consider retrying the allocation instead of
 * further reclaim.
L
Linus Torvalds 已提交
2349
 */
2350
static bool shrink_zones(struct zonelist *zonelist, struct scan_control *sc)
L
Linus Torvalds 已提交
2351
{
2352
	struct zoneref *z;
2353
	struct zone *zone;
2354 2355
	unsigned long nr_soft_reclaimed;
	unsigned long nr_soft_scanned;
2356
	unsigned long lru_pages = 0;
2357
	bool aborted_reclaim = false;
2358
	struct reclaim_state *reclaim_state = current->reclaim_state;
2359
	gfp_t orig_mask;
2360 2361 2362
	struct shrink_control shrink = {
		.gfp_mask = sc->gfp_mask,
	};
2363
	enum zone_type requested_highidx = gfp_zone(sc->gfp_mask);
2364

2365 2366 2367 2368 2369
	/*
	 * 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
	 */
2370
	orig_mask = sc->gfp_mask;
2371 2372 2373
	if (buffer_heads_over_limit)
		sc->gfp_mask |= __GFP_HIGHMEM;

2374
	nodes_clear(shrink.nodes_to_scan);
2375

2376 2377
	for_each_zone_zonelist_nodemask(zone, z, zonelist,
					gfp_zone(sc->gfp_mask), sc->nodemask) {
2378
		if (!populated_zone(zone))
L
Linus Torvalds 已提交
2379
			continue;
2380 2381 2382 2383
		/*
		 * Take care memory controller reclaiming has small influence
		 * to global LRU.
		 */
2384
		if (global_reclaim(sc)) {
2385 2386
			if (!cpuset_zone_allowed_hardwall(zone, GFP_KERNEL))
				continue;
2387 2388

			lru_pages += zone_reclaimable_pages(zone);
2389
			node_set(zone_to_nid(zone), shrink.nodes_to_scan);
2390

2391 2392
			if (sc->priority != DEF_PRIORITY &&
			    !zone_reclaimable(zone))
2393
				continue;	/* Let kswapd poll it */
2394
			if (IS_ENABLED(CONFIG_COMPACTION)) {
2395
				/*
2396 2397 2398 2399 2400
				 * 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
2401 2402
				 * noticeable problem, like transparent huge
				 * page allocations.
2403
				 */
2404 2405
				if ((zonelist_zone_idx(z) <= requested_highidx)
				    && compaction_ready(zone, sc)) {
2406
					aborted_reclaim = true;
2407
					continue;
2408
				}
2409
			}
2410 2411 2412 2413 2414 2415 2416 2417 2418 2419 2420 2421
			/*
			 * 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;
2422
			/* need some check for avoid more shrink_zone() */
2423
		}
2424

2425
		shrink_zone(zone, sc);
L
Linus Torvalds 已提交
2426
	}
2427

2428 2429 2430 2431 2432 2433 2434
	/*
	 * Don't shrink slabs when reclaiming memory from over limit cgroups
	 * but do shrink slab at least once when aborting reclaim for
	 * compaction to avoid unevenly scanning file/anon LRU pages over slab
	 * pages.
	 */
	if (global_reclaim(sc)) {
2435
		shrink_slab(&shrink, sc->nr_scanned, lru_pages);
2436 2437 2438 2439 2440 2441
		if (reclaim_state) {
			sc->nr_reclaimed += reclaim_state->reclaimed_slab;
			reclaim_state->reclaimed_slab = 0;
		}
	}

2442 2443 2444 2445 2446 2447
	/*
	 * Restore to original mask to avoid the impact on the caller if we
	 * promoted it to __GFP_HIGHMEM.
	 */
	sc->gfp_mask = orig_mask;

2448
	return aborted_reclaim;
2449 2450
}

2451
/* All zones in zonelist are unreclaimable? */
2452 2453 2454 2455 2456 2457 2458 2459 2460 2461 2462 2463
static bool all_unreclaimable(struct zonelist *zonelist,
		struct scan_control *sc)
{
	struct zoneref *z;
	struct zone *zone;

	for_each_zone_zonelist_nodemask(zone, z, zonelist,
			gfp_zone(sc->gfp_mask), sc->nodemask) {
		if (!populated_zone(zone))
			continue;
		if (!cpuset_zone_allowed_hardwall(zone, GFP_KERNEL))
			continue;
2464
		if (zone_reclaimable(zone))
2465
			return false;
2466 2467
	}

2468
	return true;
L
Linus Torvalds 已提交
2469
}
2470

L
Linus Torvalds 已提交
2471 2472 2473 2474 2475 2476 2477 2478
/*
 * 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
2479 2480 2481 2482
 * 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.
2483 2484 2485
 *
 * returns:	0, if no pages reclaimed
 * 		else, the number of pages reclaimed
L
Linus Torvalds 已提交
2486
 */
2487
static unsigned long do_try_to_free_pages(struct zonelist *zonelist,
2488
					  struct scan_control *sc)
L
Linus Torvalds 已提交
2489
{
2490
	unsigned long total_scanned = 0;
2491
	unsigned long writeback_threshold;
2492
	bool aborted_reclaim;
L
Linus Torvalds 已提交
2493

2494 2495
	delayacct_freepages_start();

2496
	if (global_reclaim(sc))
2497
		count_vm_event(ALLOCSTALL);
L
Linus Torvalds 已提交
2498

2499
	do {
2500 2501
		vmpressure_prio(sc->gfp_mask, sc->target_mem_cgroup,
				sc->priority);
2502
		sc->nr_scanned = 0;
2503
		aborted_reclaim = shrink_zones(zonelist, sc);
2504

2505
		total_scanned += sc->nr_scanned;
2506
		if (sc->nr_reclaimed >= sc->nr_to_reclaim)
L
Linus Torvalds 已提交
2507 2508
			goto out;

2509 2510 2511 2512 2513 2514 2515
		/*
		 * 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
2516 2517 2518 2519 2520 2521 2522
		/*
		 * 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.
		 */
2523 2524
		writeback_threshold = sc->nr_to_reclaim + sc->nr_to_reclaim / 2;
		if (total_scanned > writeback_threshold) {
2525 2526
			wakeup_flusher_threads(laptop_mode ? 0 : total_scanned,
						WB_REASON_TRY_TO_FREE_PAGES);
2527
			sc->may_writepage = 1;
L
Linus Torvalds 已提交
2528
		}
2529
	} while (--sc->priority >= 0 && !aborted_reclaim);
2530

L
Linus Torvalds 已提交
2531
out:
2532 2533
	delayacct_freepages_end();

2534 2535 2536
	if (sc->nr_reclaimed)
		return sc->nr_reclaimed;

2537 2538 2539 2540 2541 2542 2543 2544
	/*
	 * As hibernation is going on, kswapd is freezed so that it can't mark
	 * the zone into all_unreclaimable. Thus bypassing all_unreclaimable
	 * check.
	 */
	if (oom_killer_disabled)
		return 0;

2545 2546
	/* Aborted reclaim to try compaction? don't OOM, then */
	if (aborted_reclaim)
2547 2548
		return 1;

2549
	/* top priority shrink_zones still had more to do? don't OOM, then */
2550
	if (global_reclaim(sc) && !all_unreclaimable(zonelist, sc))
2551 2552 2553
		return 1;

	return 0;
L
Linus Torvalds 已提交
2554 2555
}

2556 2557 2558 2559 2560 2561 2562 2563 2564 2565
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];
2566 2567 2568
		if (!populated_zone(zone))
			continue;

2569 2570 2571 2572
		pfmemalloc_reserve += min_wmark_pages(zone);
		free_pages += zone_page_state(zone, NR_FREE_PAGES);
	}

2573 2574 2575 2576
	/* If there are no reserves (unexpected config) then do not throttle */
	if (!pfmemalloc_reserve)
		return true;

2577 2578 2579 2580 2581 2582 2583 2584 2585 2586 2587 2588 2589 2590 2591 2592
	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
2593 2594 2595 2596
 * 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.
2597
 */
2598
static bool throttle_direct_reclaim(gfp_t gfp_mask, struct zonelist *zonelist,
2599 2600
					nodemask_t *nodemask)
{
2601
	struct zoneref *z;
2602
	struct zone *zone;
2603
	pg_data_t *pgdat = NULL;
2604 2605 2606 2607 2608 2609 2610 2611 2612

	/*
	 * 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)
2613 2614 2615 2616 2617 2618 2619 2620
		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;
2621

2622 2623 2624 2625 2626 2627 2628 2629 2630 2631 2632 2633 2634 2635 2636 2637 2638 2639 2640 2641 2642 2643 2644 2645 2646 2647 2648 2649
	/*
	 * 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)
2650
		goto out;
2651

2652 2653 2654
	/* Account for the throttling */
	count_vm_event(PGSCAN_DIRECT_THROTTLE);

2655 2656 2657 2658 2659 2660 2661 2662 2663 2664 2665
	/*
	 * 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);
2666 2667

		goto check_pending;
2668 2669 2670 2671 2672
	}

	/* Throttle until kswapd wakes the process */
	wait_event_killable(zone->zone_pgdat->pfmemalloc_wait,
		pfmemalloc_watermark_ok(pgdat));
2673 2674 2675 2676 2677 2678 2679

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

out:
	return false;
2680 2681
}

2682
unsigned long try_to_free_pages(struct zonelist *zonelist, int order,
2683
				gfp_t gfp_mask, nodemask_t *nodemask)
2684
{
2685
	unsigned long nr_reclaimed;
2686
	struct scan_control sc = {
2687
		.gfp_mask = (gfp_mask = memalloc_noio_flags(gfp_mask)),
2688
		.may_writepage = !laptop_mode,
2689
		.nr_to_reclaim = SWAP_CLUSTER_MAX,
2690
		.may_unmap = 1,
2691
		.may_swap = 1,
2692
		.order = order,
2693
		.priority = DEF_PRIORITY,
2694
		.target_mem_cgroup = NULL,
2695
		.nodemask = nodemask,
2696 2697
	};

2698
	/*
2699 2700 2701
	 * 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.
2702
	 */
2703
	if (throttle_direct_reclaim(gfp_mask, zonelist, nodemask))
2704 2705
		return 1;

2706 2707 2708 2709
	trace_mm_vmscan_direct_reclaim_begin(order,
				sc.may_writepage,
				gfp_mask);

2710
	nr_reclaimed = do_try_to_free_pages(zonelist, &sc);
2711 2712 2713 2714

	trace_mm_vmscan_direct_reclaim_end(nr_reclaimed);

	return nr_reclaimed;
2715 2716
}

A
Andrew Morton 已提交
2717
#ifdef CONFIG_MEMCG
2718

2719
unsigned long mem_cgroup_shrink_node_zone(struct mem_cgroup *memcg,
2720
						gfp_t gfp_mask, bool noswap,
2721 2722
						struct zone *zone,
						unsigned long *nr_scanned)
2723 2724
{
	struct scan_control sc = {
2725
		.nr_scanned = 0,
2726
		.nr_to_reclaim = SWAP_CLUSTER_MAX,
2727 2728 2729 2730
		.may_writepage = !laptop_mode,
		.may_unmap = 1,
		.may_swap = !noswap,
		.order = 0,
2731
		.priority = 0,
2732
		.swappiness = mem_cgroup_swappiness(memcg),
2733
		.target_mem_cgroup = memcg,
2734
	};
2735
	struct lruvec *lruvec = mem_cgroup_zone_lruvec(zone, memcg);
2736

2737 2738
	sc.gfp_mask = (gfp_mask & GFP_RECLAIM_MASK) |
			(GFP_HIGHUSER_MOVABLE & ~GFP_RECLAIM_MASK);
2739

2740
	trace_mm_vmscan_memcg_softlimit_reclaim_begin(sc.order,
2741 2742 2743
						      sc.may_writepage,
						      sc.gfp_mask);

2744 2745 2746 2747 2748 2749 2750
	/*
	 * 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.
	 */
2751
	shrink_lruvec(lruvec, &sc);
2752 2753 2754

	trace_mm_vmscan_memcg_softlimit_reclaim_end(sc.nr_reclaimed);

2755
	*nr_scanned = sc.nr_scanned;
2756 2757 2758
	return sc.nr_reclaimed;
}

2759
unsigned long try_to_free_mem_cgroup_pages(struct mem_cgroup *memcg,
K
KOSAKI Motohiro 已提交
2760
					   gfp_t gfp_mask,
2761
					   bool noswap)
2762
{
2763
	struct zonelist *zonelist;
2764
	unsigned long nr_reclaimed;
2765
	int nid;
2766 2767
	struct scan_control sc = {
		.may_writepage = !laptop_mode,
2768
		.may_unmap = 1,
2769
		.may_swap = !noswap,
2770
		.nr_to_reclaim = SWAP_CLUSTER_MAX,
2771
		.order = 0,
2772
		.priority = DEF_PRIORITY,
2773
		.target_mem_cgroup = memcg,
2774
		.nodemask = NULL, /* we don't care the placement */
2775 2776 2777
		.gfp_mask = (gfp_mask & GFP_RECLAIM_MASK) |
				(GFP_HIGHUSER_MOVABLE & ~GFP_RECLAIM_MASK),
	};
2778

2779 2780 2781 2782 2783
	/*
	 * 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.
	 */
2784
	nid = mem_cgroup_select_victim_node(memcg);
2785 2786

	zonelist = NODE_DATA(nid)->node_zonelists;
2787 2788 2789 2790 2791

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

2792
	nr_reclaimed = do_try_to_free_pages(zonelist, &sc);
2793 2794 2795 2796

	trace_mm_vmscan_memcg_reclaim_end(nr_reclaimed);

	return nr_reclaimed;
2797 2798 2799
}
#endif

2800
static void age_active_anon(struct zone *zone, struct scan_control *sc)
2801
{
2802
	struct mem_cgroup *memcg;
2803

2804 2805 2806 2807 2808
	if (!total_swap_pages)
		return;

	memcg = mem_cgroup_iter(NULL, NULL, NULL);
	do {
2809
		struct lruvec *lruvec = mem_cgroup_zone_lruvec(zone, memcg);
2810

2811
		if (inactive_anon_is_low(lruvec))
2812
			shrink_active_list(SWAP_CLUSTER_MAX, lruvec,
2813
					   sc, LRU_ACTIVE_ANON);
2814 2815 2816

		memcg = mem_cgroup_iter(NULL, memcg, NULL);
	} while (memcg);
2817 2818
}

2819 2820 2821 2822 2823 2824 2825
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;

2826 2827
	if (IS_ENABLED(CONFIG_COMPACTION) && order &&
	    !compaction_suitable(zone, order))
2828 2829 2830 2831 2832
		return false;

	return true;
}

2833
/*
2834 2835 2836 2837 2838 2839 2840 2841 2842 2843
 * 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.
2844 2845 2846 2847
 * 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
2848
 *     percentage of the middle zones. For example, on 32-bit x86, highmem
2849 2850 2851 2852
 *     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.
 */
2853
static bool pgdat_balanced(pg_data_t *pgdat, int order, int classzone_idx)
2854
{
2855
	unsigned long managed_pages = 0;
2856
	unsigned long balanced_pages = 0;
2857 2858
	int i;

2859 2860 2861
	/* Check the watermark levels */
	for (i = 0; i <= classzone_idx; i++) {
		struct zone *zone = pgdat->node_zones + i;
2862

2863 2864 2865
		if (!populated_zone(zone))
			continue;

2866
		managed_pages += zone->managed_pages;
2867 2868 2869 2870 2871 2872 2873 2874

		/*
		 * 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!
		 */
2875
		if (!zone_reclaimable(zone)) {
2876
			balanced_pages += zone->managed_pages;
2877 2878 2879 2880
			continue;
		}

		if (zone_balanced(zone, order, 0, i))
2881
			balanced_pages += zone->managed_pages;
2882 2883 2884 2885 2886
		else if (!order)
			return false;
	}

	if (order)
2887
		return balanced_pages >= (managed_pages >> 2);
2888 2889
	else
		return true;
2890 2891
}

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

	/*
	 * There is a potential race between when kswapd checks its watermarks
	 * and a process gets throttled. There is also a potential race if
	 * processes get throttled, kswapd wakes, a large process exits therby
	 * balancing the zones that causes kswapd to miss a wakeup. If kswapd
	 * is going to sleep, no process should be sleeping on pfmemalloc_wait
	 * so wake them now if necessary. If necessary, processes will wake
	 * kswapd and get throttled again
	 */
	if (waitqueue_active(&pgdat->pfmemalloc_wait)) {
		wake_up(&pgdat->pfmemalloc_wait);
		return false;
	}
2918

2919
	return pgdat_balanced(pgdat, order, classzone_idx);
2920 2921
}

2922 2923 2924
/*
 * kswapd shrinks the zone by the number of pages required to reach
 * the high watermark.
2925 2926
 *
 * Returns true if kswapd scanned at least the requested number of pages to
2927 2928
 * 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.
2929
 */
2930
static bool kswapd_shrink_zone(struct zone *zone,
2931
			       int classzone_idx,
2932
			       struct scan_control *sc,
2933 2934
			       unsigned long lru_pages,
			       unsigned long *nr_attempted)
2935
{
2936 2937
	int testorder = sc->order;
	unsigned long balance_gap;
2938 2939 2940 2941
	struct reclaim_state *reclaim_state = current->reclaim_state;
	struct shrink_control shrink = {
		.gfp_mask = sc->gfp_mask,
	};
2942
	bool lowmem_pressure;
2943 2944 2945

	/* Reclaim above the high watermark. */
	sc->nr_to_reclaim = max(SWAP_CLUSTER_MAX, high_wmark_pages(zone));
2946 2947 2948 2949 2950 2951 2952 2953 2954 2955 2956 2957 2958 2959 2960 2961 2962 2963

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

	/*
	 * We put equal pressure on every zone, unless one zone has way too
	 * many pages free already. The "too many pages" is defined as the
	 * high wmark plus a "gap" where the gap is either the low
	 * watermark or 1% of the zone, whichever is smaller.
	 */
2964 2965
	balance_gap = min(low_wmark_pages(zone), DIV_ROUND_UP(
			zone->managed_pages, KSWAPD_ZONE_BALANCE_GAP_RATIO));
2966 2967 2968 2969 2970 2971 2972 2973 2974 2975

	/*
	 * 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;

2976
	shrink_zone(zone, sc);
D
Dave Chinner 已提交
2977 2978
	nodes_clear(shrink.nodes_to_scan);
	node_set(zone_to_nid(zone), shrink.nodes_to_scan);
2979 2980

	reclaim_state->reclaimed_slab = 0;
2981
	shrink_slab(&shrink, sc->nr_scanned, lru_pages);
2982 2983
	sc->nr_reclaimed += reclaim_state->reclaimed_slab;

2984 2985 2986
	/* Account for the number of pages attempted to reclaim */
	*nr_attempted += sc->nr_to_reclaim;

2987 2988
	zone_clear_flag(zone, ZONE_WRITEBACK);

2989 2990 2991 2992 2993 2994
	/*
	 * 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.
	 */
2995
	if (zone_reclaimable(zone) &&
2996 2997 2998 2999 3000
	    zone_balanced(zone, testorder, 0, classzone_idx)) {
		zone_clear_flag(zone, ZONE_CONGESTED);
		zone_clear_flag(zone, ZONE_TAIL_LRU_DIRTY);
	}

3001
	return sc->nr_scanned >= sc->nr_to_reclaim;
3002 3003
}

L
Linus Torvalds 已提交
3004 3005
/*
 * For kswapd, balance_pgdat() will work across all this node's zones until
3006
 * they are all at high_wmark_pages(zone).
L
Linus Torvalds 已提交
3007
 *
3008
 * Returns the final order kswapd was reclaiming at
L
Linus Torvalds 已提交
3009 3010 3011 3012 3013 3014 3015 3016 3017 3018
 *
 * 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
3019 3020 3021 3022 3023
 * 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
3024
 */
3025
static unsigned long balance_pgdat(pg_data_t *pgdat, int order,
3026
							int *classzone_idx)
L
Linus Torvalds 已提交
3027 3028
{
	int i;
3029
	int end_zone = 0;	/* Inclusive.  0 = ZONE_DMA */
3030 3031
	unsigned long nr_soft_reclaimed;
	unsigned long nr_soft_scanned;
3032 3033
	struct scan_control sc = {
		.gfp_mask = GFP_KERNEL,
3034
		.priority = DEF_PRIORITY,
3035
		.may_unmap = 1,
3036
		.may_swap = 1,
3037
		.may_writepage = !laptop_mode,
A
Andy Whitcroft 已提交
3038
		.order = order,
3039
		.target_mem_cgroup = NULL,
3040
	};
3041
	count_vm_event(PAGEOUTRUN);
L
Linus Torvalds 已提交
3042

3043
	do {
L
Linus Torvalds 已提交
3044
		unsigned long lru_pages = 0;
3045
		unsigned long nr_attempted = 0;
3046
		bool raise_priority = true;
3047
		bool pgdat_needs_compaction = (order > 0);
3048 3049

		sc.nr_reclaimed = 0;
L
Linus Torvalds 已提交
3050

3051 3052 3053 3054 3055 3056
		/*
		 * 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
3057

3058 3059
			if (!populated_zone(zone))
				continue;
L
Linus Torvalds 已提交
3060

3061 3062
			if (sc.priority != DEF_PRIORITY &&
			    !zone_reclaimable(zone))
3063
				continue;
L
Linus Torvalds 已提交
3064

3065 3066 3067 3068
			/*
			 * Do some background aging of the anon list, to give
			 * pages a chance to be referenced before reclaiming.
			 */
3069
			age_active_anon(zone, &sc);
3070

3071 3072 3073 3074 3075 3076 3077 3078 3079 3080 3081
			/*
			 * 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;
			}

3082
			if (!zone_balanced(zone, order, 0, 0)) {
3083
				end_zone = i;
A
Andrew Morton 已提交
3084
				break;
3085
			} else {
3086 3087 3088 3089
				/*
				 * If balanced, clear the dirty and congested
				 * flags
				 */
3090
				zone_clear_flag(zone, ZONE_CONGESTED);
3091
				zone_clear_flag(zone, ZONE_TAIL_LRU_DIRTY);
L
Linus Torvalds 已提交
3092 3093
			}
		}
3094

3095
		if (i < 0)
A
Andrew Morton 已提交
3096 3097
			goto out;

L
Linus Torvalds 已提交
3098 3099 3100
		for (i = 0; i <= end_zone; i++) {
			struct zone *zone = pgdat->node_zones + i;

3101 3102 3103
			if (!populated_zone(zone))
				continue;

3104
			lru_pages += zone_reclaimable_pages(zone);
3105 3106 3107 3108 3109 3110 3111 3112 3113 3114 3115

			/*
			 * 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
3116 3117
		}

3118 3119 3120 3121 3122 3123 3124
		/*
		 * 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
3125 3126 3127 3128 3129 3130 3131 3132 3133 3134 3135 3136
		/*
		 * 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;

3137
			if (!populated_zone(zone))
L
Linus Torvalds 已提交
3138 3139
				continue;

3140 3141
			if (sc.priority != DEF_PRIORITY &&
			    !zone_reclaimable(zone))
L
Linus Torvalds 已提交
3142 3143 3144
				continue;

			sc.nr_scanned = 0;
3145

3146 3147 3148 3149 3150 3151 3152 3153 3154
			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;

3155
			/*
3156 3157 3158 3159
			 * 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.
3160
			 */
3161 3162 3163
			if (kswapd_shrink_zone(zone, end_zone, &sc,
					lru_pages, &nr_attempted))
				raise_priority = false;
L
Linus Torvalds 已提交
3164
		}
3165 3166 3167 3168 3169 3170 3171 3172 3173 3174

		/*
		 * 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
3175
		/*
3176 3177 3178 3179 3180 3181
		 * 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
3182
		 */
3183 3184
		if (order && sc.nr_reclaimed >= 2UL << order)
			order = sc.order = 0;
3185

3186 3187 3188
		/* Check if kswapd should be suspending */
		if (try_to_freeze() || kthread_should_stop())
			break;
3189

3190 3191 3192 3193 3194 3195 3196
		/*
		 * 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);

3197
		/*
3198 3199
		 * Raise priority if scanning rate is too low or there was no
		 * progress in reclaiming pages
3200
		 */
3201 3202
		if (raise_priority || !sc.nr_reclaimed)
			sc.priority--;
3203
	} while (sc.priority >= 1 &&
3204
		 !pgdat_balanced(pgdat, order, *classzone_idx));
L
Linus Torvalds 已提交
3205

3206
out:
3207
	/*
3208
	 * Return the order we were reclaiming at so prepare_kswapd_sleep()
3209 3210 3211 3212
	 * 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
	 */
3213
	*classzone_idx = end_zone;
3214
	return order;
L
Linus Torvalds 已提交
3215 3216
}

3217
static void kswapd_try_to_sleep(pg_data_t *pgdat, int order, int classzone_idx)
3218 3219 3220 3221 3222 3223 3224 3225 3226 3227
{
	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 */
3228
	if (prepare_kswapd_sleep(pgdat, order, remaining, classzone_idx)) {
3229 3230 3231 3232 3233 3234 3235 3236 3237
		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.
	 */
3238
	if (prepare_kswapd_sleep(pgdat, order, remaining, classzone_idx)) {
3239 3240 3241 3242 3243 3244 3245 3246 3247 3248 3249
		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);
3250

3251 3252 3253 3254 3255 3256 3257 3258
		/*
		 * 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);

3259 3260 3261
		if (!kthread_should_stop())
			schedule();

3262 3263 3264 3265 3266 3267 3268 3269 3270 3271
		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
3272 3273
/*
 * The background pageout daemon, started as a kernel thread
3274
 * from the init process.
L
Linus Torvalds 已提交
3275 3276 3277 3278 3279 3280 3281 3282 3283 3284 3285 3286
 *
 * 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)
{
3287
	unsigned long order, new_order;
3288
	unsigned balanced_order;
3289
	int classzone_idx, new_classzone_idx;
3290
	int balanced_classzone_idx;
L
Linus Torvalds 已提交
3291 3292
	pg_data_t *pgdat = (pg_data_t*)p;
	struct task_struct *tsk = current;
3293

L
Linus Torvalds 已提交
3294 3295 3296
	struct reclaim_state reclaim_state = {
		.reclaimed_slab = 0,
	};
3297
	const struct cpumask *cpumask = cpumask_of_node(pgdat->node_id);
L
Linus Torvalds 已提交
3298

3299 3300
	lockdep_set_current_reclaim_state(GFP_KERNEL);

R
Rusty Russell 已提交
3301
	if (!cpumask_empty(cpumask))
3302
		set_cpus_allowed_ptr(tsk, cpumask);
L
Linus Torvalds 已提交
3303 3304 3305 3306 3307 3308 3309 3310 3311 3312 3313 3314 3315 3316
	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).
	 */
3317
	tsk->flags |= PF_MEMALLOC | PF_SWAPWRITE | PF_KSWAPD;
3318
	set_freezable();
L
Linus Torvalds 已提交
3319

3320
	order = new_order = 0;
3321
	balanced_order = 0;
3322
	classzone_idx = new_classzone_idx = pgdat->nr_zones - 1;
3323
	balanced_classzone_idx = classzone_idx;
L
Linus Torvalds 已提交
3324
	for ( ; ; ) {
3325
		bool ret;
3326

3327 3328 3329 3330 3331
		/*
		 * 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
		 */
3332 3333
		if (balanced_classzone_idx >= new_classzone_idx &&
					balanced_order == new_order) {
3334 3335 3336 3337 3338 3339
			new_order = pgdat->kswapd_max_order;
			new_classzone_idx = pgdat->classzone_idx;
			pgdat->kswapd_max_order =  0;
			pgdat->classzone_idx = pgdat->nr_zones - 1;
		}

3340
		if (order < new_order || classzone_idx > new_classzone_idx) {
L
Linus Torvalds 已提交
3341 3342
			/*
			 * Don't sleep if someone wants a larger 'order'
3343
			 * allocation or has tigher zone constraints
L
Linus Torvalds 已提交
3344 3345
			 */
			order = new_order;
3346
			classzone_idx = new_classzone_idx;
L
Linus Torvalds 已提交
3347
		} else {
3348 3349
			kswapd_try_to_sleep(pgdat, balanced_order,
						balanced_classzone_idx);
L
Linus Torvalds 已提交
3350
			order = pgdat->kswapd_max_order;
3351
			classzone_idx = pgdat->classzone_idx;
3352 3353
			new_order = order;
			new_classzone_idx = classzone_idx;
3354
			pgdat->kswapd_max_order = 0;
3355
			pgdat->classzone_idx = pgdat->nr_zones - 1;
L
Linus Torvalds 已提交
3356 3357
		}

3358 3359 3360 3361 3362 3363 3364 3365
		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
		 */
3366 3367
		if (!ret) {
			trace_mm_vmscan_kswapd_wake(pgdat->node_id, order);
3368 3369 3370
			balanced_classzone_idx = classzone_idx;
			balanced_order = balance_pgdat(pgdat, order,
						&balanced_classzone_idx);
3371
		}
L
Linus Torvalds 已提交
3372
	}
3373

3374
	tsk->flags &= ~(PF_MEMALLOC | PF_SWAPWRITE | PF_KSWAPD);
3375
	current->reclaim_state = NULL;
3376 3377
	lockdep_clear_current_reclaim_state();

L
Linus Torvalds 已提交
3378 3379 3380 3381 3382 3383
	return 0;
}

/*
 * A zone is low on free memory, so wake its kswapd task to service it.
 */
3384
void wakeup_kswapd(struct zone *zone, int order, enum zone_type classzone_idx)
L
Linus Torvalds 已提交
3385 3386 3387
{
	pg_data_t *pgdat;

3388
	if (!populated_zone(zone))
L
Linus Torvalds 已提交
3389 3390
		return;

3391
	if (!cpuset_zone_allowed_hardwall(zone, GFP_KERNEL))
L
Linus Torvalds 已提交
3392
		return;
3393
	pgdat = zone->zone_pgdat;
3394
	if (pgdat->kswapd_max_order < order) {
L
Linus Torvalds 已提交
3395
		pgdat->kswapd_max_order = order;
3396 3397
		pgdat->classzone_idx = min(pgdat->classzone_idx, classzone_idx);
	}
3398
	if (!waitqueue_active(&pgdat->kswapd_wait))
L
Linus Torvalds 已提交
3399
		return;
3400
	if (zone_balanced(zone, order, 0, 0))
3401 3402 3403
		return;

	trace_mm_vmscan_wakeup_kswapd(pgdat->node_id, zone_idx(zone), order);
3404
	wake_up_interruptible(&pgdat->kswapd_wait);
L
Linus Torvalds 已提交
3405 3406
}

3407
#ifdef CONFIG_HIBERNATION
L
Linus Torvalds 已提交
3408
/*
3409
 * Try to free `nr_to_reclaim' of memory, system-wide, and return the number of
3410 3411 3412 3413 3414
 * 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
3415
 */
3416
unsigned long shrink_all_memory(unsigned long nr_to_reclaim)
L
Linus Torvalds 已提交
3417
{
3418 3419
	struct reclaim_state reclaim_state;
	struct scan_control sc = {
3420 3421 3422
		.gfp_mask = GFP_HIGHUSER_MOVABLE,
		.may_swap = 1,
		.may_unmap = 1,
3423
		.may_writepage = 1,
3424 3425 3426
		.nr_to_reclaim = nr_to_reclaim,
		.hibernation_mode = 1,
		.order = 0,
3427
		.priority = DEF_PRIORITY,
L
Linus Torvalds 已提交
3428
	};
3429
	struct zonelist *zonelist = node_zonelist(numa_node_id(), sc.gfp_mask);
3430 3431
	struct task_struct *p = current;
	unsigned long nr_reclaimed;
L
Linus Torvalds 已提交
3432

3433 3434 3435 3436
	p->flags |= PF_MEMALLOC;
	lockdep_set_current_reclaim_state(sc.gfp_mask);
	reclaim_state.reclaimed_slab = 0;
	p->reclaim_state = &reclaim_state;
3437

3438
	nr_reclaimed = do_try_to_free_pages(zonelist, &sc);
3439

3440 3441 3442
	p->reclaim_state = NULL;
	lockdep_clear_current_reclaim_state();
	p->flags &= ~PF_MEMALLOC;
3443

3444
	return nr_reclaimed;
L
Linus Torvalds 已提交
3445
}
3446
#endif /* CONFIG_HIBERNATION */
L
Linus Torvalds 已提交
3447 3448 3449 3450 3451

/* 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. */
3452 3453
static int cpu_callback(struct notifier_block *nfb, unsigned long action,
			void *hcpu)
L
Linus Torvalds 已提交
3454
{
3455
	int nid;
L
Linus Torvalds 已提交
3456

3457
	if (action == CPU_ONLINE || action == CPU_ONLINE_FROZEN) {
3458
		for_each_node_state(nid, N_MEMORY) {
3459
			pg_data_t *pgdat = NODE_DATA(nid);
3460 3461 3462
			const struct cpumask *mask;

			mask = cpumask_of_node(pgdat->node_id);
3463

3464
			if (cpumask_any_and(cpu_online_mask, mask) < nr_cpu_ids)
L
Linus Torvalds 已提交
3465
				/* One of our CPUs online: restore mask */
3466
				set_cpus_allowed_ptr(pgdat->kswapd, mask);
L
Linus Torvalds 已提交
3467 3468 3469 3470 3471
		}
	}
	return NOTIFY_OK;
}

3472 3473 3474 3475 3476 3477 3478 3479 3480 3481 3482 3483 3484 3485 3486 3487
/*
 * 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);
3488 3489
		pr_err("Failed to start kswapd on node %d\n", nid);
		ret = PTR_ERR(pgdat->kswapd);
3490
		pgdat->kswapd = NULL;
3491 3492 3493 3494
	}
	return ret;
}

3495
/*
3496
 * Called by memory hotplug when all memory in a node is offlined.  Caller must
3497
 * hold mem_hotplug_begin/end().
3498 3499 3500 3501 3502
 */
void kswapd_stop(int nid)
{
	struct task_struct *kswapd = NODE_DATA(nid)->kswapd;

3503
	if (kswapd) {
3504
		kthread_stop(kswapd);
3505 3506
		NODE_DATA(nid)->kswapd = NULL;
	}
3507 3508
}

L
Linus Torvalds 已提交
3509 3510
static int __init kswapd_init(void)
{
3511
	int nid;
3512

L
Linus Torvalds 已提交
3513
	swap_setup();
3514
	for_each_node_state(nid, N_MEMORY)
3515
 		kswapd_run(nid);
L
Linus Torvalds 已提交
3516 3517 3518 3519 3520
	hotcpu_notifier(cpu_callback, 0);
	return 0;
}

module_init(kswapd_init)
3521 3522 3523 3524 3525 3526 3527 3528 3529 3530

#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;

3531
#define RECLAIM_OFF 0
3532
#define RECLAIM_ZONE (1<<0)	/* Run shrink_inactive_list on the zone */
3533 3534 3535
#define RECLAIM_WRITE (1<<1)	/* Writeout pages during reclaim */
#define RECLAIM_SWAP (1<<2)	/* Swap pages out during reclaim */

3536 3537 3538 3539 3540 3541 3542
/*
 * 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

3543 3544 3545 3546 3547 3548
/*
 * Percentage of pages in a zone that must be unmapped for zone_reclaim to
 * occur.
 */
int sysctl_min_unmapped_ratio = 1;

3549 3550 3551 3552 3553 3554
/*
 * 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;

3555 3556 3557 3558 3559 3560 3561 3562 3563 3564 3565 3566 3567 3568 3569 3570 3571 3572 3573 3574 3575 3576 3577 3578 3579 3580 3581 3582 3583 3584 3585 3586 3587 3588 3589 3590 3591 3592 3593 3594 3595 3596
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;
}

3597 3598 3599
/*
 * Try to free up some pages from this zone through reclaim.
 */
3600
static int __zone_reclaim(struct zone *zone, gfp_t gfp_mask, unsigned int order)
3601
{
3602
	/* Minimum pages needed in order to stay on node */
3603
	const unsigned long nr_pages = 1 << order;
3604 3605
	struct task_struct *p = current;
	struct reclaim_state reclaim_state;
3606 3607
	struct scan_control sc = {
		.may_writepage = !!(zone_reclaim_mode & RECLAIM_WRITE),
3608
		.may_unmap = !!(zone_reclaim_mode & RECLAIM_SWAP),
3609
		.may_swap = 1,
3610
		.nr_to_reclaim = max(nr_pages, SWAP_CLUSTER_MAX),
3611
		.gfp_mask = (gfp_mask = memalloc_noio_flags(gfp_mask)),
3612
		.order = order,
3613
		.priority = ZONE_RECLAIM_PRIORITY,
3614
	};
3615 3616 3617
	struct shrink_control shrink = {
		.gfp_mask = sc.gfp_mask,
	};
3618
	unsigned long nr_slab_pages0, nr_slab_pages1;
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 3638
			shrink_zone(zone, &sc);
		} while (sc.nr_reclaimed < nr_pages && --sc.priority >= 0);
3639
	}
3640

3641 3642
	nr_slab_pages0 = zone_page_state(zone, NR_SLAB_RECLAIMABLE);
	if (nr_slab_pages0 > zone->min_slab_pages) {
3643
		/*
3644
		 * shrink_slab() does not currently allow us to determine how
3645 3646 3647 3648
		 * many pages were freed in this zone. So we take the current
		 * number of slab pages and shake the slab until it is reduced
		 * by the same nr_pages that we used for reclaiming unmapped
		 * pages.
3649
		 */
D
Dave Chinner 已提交
3650 3651
		nodes_clear(shrink.nodes_to_scan);
		node_set(zone_to_nid(zone), shrink.nodes_to_scan);
3652 3653 3654 3655
		for (;;) {
			unsigned long lru_pages = zone_reclaimable_pages(zone);

			/* No reclaimable slab or very low memory pressure */
3656
			if (!shrink_slab(&shrink, sc.nr_scanned, lru_pages))
3657 3658 3659 3660 3661 3662 3663 3664
				break;

			/* Freed enough memory */
			nr_slab_pages1 = zone_page_state(zone,
							NR_SLAB_RECLAIMABLE);
			if (nr_slab_pages1 + nr_pages <= nr_slab_pages0)
				break;
		}
3665 3666 3667 3668 3669

		/*
		 * Update nr_reclaimed by the number of slab pages we
		 * reclaimed from this zone.
		 */
3670 3671 3672
		nr_slab_pages1 = zone_page_state(zone, NR_SLAB_RECLAIMABLE);
		if (nr_slab_pages1 < nr_slab_pages0)
			sc.nr_reclaimed += nr_slab_pages0 - nr_slab_pages1;
3673 3674
	}

3675
	p->reclaim_state = NULL;
3676
	current->flags &= ~(PF_MEMALLOC | PF_SWAPWRITE);
3677
	lockdep_clear_current_reclaim_state();
3678
	return sc.nr_reclaimed >= nr_pages;
3679
}
3680 3681 3682 3683

int zone_reclaim(struct zone *zone, gfp_t gfp_mask, unsigned int order)
{
	int node_id;
3684
	int ret;
3685 3686

	/*
3687 3688
	 * Zone reclaim reclaims unmapped file backed pages and
	 * slab pages if we are over the defined limits.
3689
	 *
3690 3691 3692 3693 3694
	 * 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.
3695
	 */
3696 3697
	if (zone_pagecache_reclaimable(zone) <= zone->min_unmapped_pages &&
	    zone_page_state(zone, NR_SLAB_RECLAIMABLE) <= zone->min_slab_pages)
3698
		return ZONE_RECLAIM_FULL;
3699

3700
	if (!zone_reclaimable(zone))
3701
		return ZONE_RECLAIM_FULL;
3702

3703
	/*
3704
	 * Do not scan if the allocation should not be delayed.
3705
	 */
3706
	if (!(gfp_mask & __GFP_WAIT) || (current->flags & PF_MEMALLOC))
3707
		return ZONE_RECLAIM_NOSCAN;
3708 3709 3710 3711 3712 3713 3714

	/*
	 * 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.
	 */
3715
	node_id = zone_to_nid(zone);
3716
	if (node_state(node_id, N_CPU) && node_id != numa_node_id())
3717
		return ZONE_RECLAIM_NOSCAN;
3718 3719

	if (zone_test_and_set_flag(zone, ZONE_RECLAIM_LOCKED))
3720 3721
		return ZONE_RECLAIM_NOSCAN;

3722 3723 3724
	ret = __zone_reclaim(zone, gfp_mask, order);
	zone_clear_flag(zone, ZONE_RECLAIM_LOCKED);

3725 3726 3727
	if (!ret)
		count_vm_event(PGSCAN_ZONE_RECLAIM_FAILED);

3728
	return ret;
3729
}
3730
#endif
L
Lee Schermerhorn 已提交
3731 3732 3733 3734 3735 3736

/*
 * 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
3737
 * lists vs unevictable list.
L
Lee Schermerhorn 已提交
3738 3739
 *
 * Reasons page might not be evictable:
3740
 * (1) page's mapping marked unevictable
N
Nick Piggin 已提交
3741
 * (2) page is part of an mlocked VMA
3742
 *
L
Lee Schermerhorn 已提交
3743
 */
3744
int page_evictable(struct page *page)
L
Lee Schermerhorn 已提交
3745
{
3746
	return !mapping_unevictable(page_mapping(page)) && !PageMlocked(page);
L
Lee Schermerhorn 已提交
3747
}
3748

3749
#ifdef CONFIG_SHMEM
3750
/**
3751 3752 3753
 * 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
3754
 *
3755
 * Checks pages for evictability and moves them to the appropriate lru list.
3756 3757
 *
 * This function is only used for SysV IPC SHM_UNLOCK.
3758
 */
3759
void check_move_unevictable_pages(struct page **pages, int nr_pages)
3760
{
3761
	struct lruvec *lruvec;
3762 3763 3764 3765
	struct zone *zone = NULL;
	int pgscanned = 0;
	int pgrescued = 0;
	int i;
3766

3767 3768 3769
	for (i = 0; i < nr_pages; i++) {
		struct page *page = pages[i];
		struct zone *pagezone;
3770

3771 3772 3773 3774 3775 3776 3777 3778
		pgscanned++;
		pagezone = page_zone(page);
		if (pagezone != zone) {
			if (zone)
				spin_unlock_irq(&zone->lru_lock);
			zone = pagezone;
			spin_lock_irq(&zone->lru_lock);
		}
3779
		lruvec = mem_cgroup_page_lruvec(page, zone);
3780

3781 3782
		if (!PageLRU(page) || !PageUnevictable(page))
			continue;
3783

3784
		if (page_evictable(page)) {
3785 3786
			enum lru_list lru = page_lru_base_type(page);

3787
			VM_BUG_ON_PAGE(PageActive(page), page);
3788
			ClearPageUnevictable(page);
3789 3790
			del_page_from_lru_list(page, lruvec, LRU_UNEVICTABLE);
			add_page_to_lru_list(page, lruvec, lru);
3791
			pgrescued++;
3792
		}
3793
	}
3794

3795 3796 3797 3798
	if (zone) {
		__count_vm_events(UNEVICTABLE_PGRESCUED, pgrescued);
		__count_vm_events(UNEVICTABLE_PGSCANNED, pgscanned);
		spin_unlock_irq(&zone->lru_lock);
3799 3800
	}
}
3801
#endif /* CONFIG_SHMEM */
3802

3803
static void warn_scan_unevictable_pages(void)
3804
{
3805
	printk_once(KERN_WARNING
3806
		    "%s: The scan_unevictable_pages sysctl/node-interface has been "
3807
		    "disabled for lack of a legitimate use case.  If you have "
3808 3809
		    "one, please send an email to linux-mm@kvack.org.\n",
		    current->comm);
3810 3811 3812 3813 3814 3815 3816 3817 3818
}

/*
 * scan_unevictable_pages [vm] sysctl handler.  On demand re-scan of
 * all nodes' unevictable lists for evictable pages
 */
unsigned long scan_unevictable_pages;

int scan_unevictable_handler(struct ctl_table *table, int write,
3819
			   void __user *buffer,
3820 3821
			   size_t *length, loff_t *ppos)
{
3822
	warn_scan_unevictable_pages();
3823
	proc_doulongvec_minmax(table, write, buffer, length, ppos);
3824 3825 3826 3827
	scan_unevictable_pages = 0;
	return 0;
}

3828
#ifdef CONFIG_NUMA
3829 3830 3831 3832 3833
/*
 * per node 'scan_unevictable_pages' attribute.  On demand re-scan of
 * a specified node's per zone unevictable lists for evictable pages.
 */

3834 3835
static ssize_t read_scan_unevictable_node(struct device *dev,
					  struct device_attribute *attr,
3836 3837
					  char *buf)
{
3838
	warn_scan_unevictable_pages();
3839 3840 3841
	return sprintf(buf, "0\n");	/* always zero; should fit... */
}

3842 3843
static ssize_t write_scan_unevictable_node(struct device *dev,
					   struct device_attribute *attr,
3844 3845
					const char *buf, size_t count)
{
3846
	warn_scan_unevictable_pages();
3847 3848 3849 3850
	return 1;
}


3851
static DEVICE_ATTR(scan_unevictable_pages, S_IRUGO | S_IWUSR,
3852 3853 3854 3855 3856
			read_scan_unevictable_node,
			write_scan_unevictable_node);

int scan_unevictable_register_node(struct node *node)
{
3857
	return device_create_file(&node->dev, &dev_attr_scan_unevictable_pages);
3858 3859 3860 3861
}

void scan_unevictable_unregister_node(struct node *node)
{
3862
	device_remove_file(&node->dev, &dev_attr_scan_unevictable_pages);
3863
}
3864
#endif