page_alloc.c 162 KB
Newer Older
Linus Torvalds's avatar
Linus Torvalds committed
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
/*
 *  linux/mm/page_alloc.c
 *
 *  Manages the free list, the system allocates free pages here.
 *  Note that kmalloc() lives in slab.c
 *
 *  Copyright (C) 1991, 1992, 1993, 1994  Linus Torvalds
 *  Swap reorganised 29.12.95, Stephen Tweedie
 *  Support of BIGMEM added by Gerhard Wichert, Siemens AG, July 1999
 *  Reshaped it to be a zoned allocator, Ingo Molnar, Red Hat, 1999
 *  Discontiguous memory support, Kanoj Sarcar, SGI, Nov 1999
 *  Zone balancing, Kanoj Sarcar, SGI, Jan 2000
 *  Per cpu hot/cold page lists, bulk allocation, Martin J. Bligh, Sept 2002
 *          (lots of bits borrowed from Ingo Molnar & Andrew Morton)
 */

#include <linux/stddef.h>
#include <linux/mm.h>
#include <linux/swap.h>
#include <linux/interrupt.h>
#include <linux/pagemap.h>
22
#include <linux/jiffies.h>
Linus Torvalds's avatar
Linus Torvalds committed
23
#include <linux/bootmem.h>
24
#include <linux/memblock.h>
Linus Torvalds's avatar
Linus Torvalds committed
25
#include <linux/compiler.h>
26
#include <linux/kernel.h>
27
#include <linux/kmemcheck.h>
Linus Torvalds's avatar
Linus Torvalds committed
28
29
30
31
32
#include <linux/module.h>
#include <linux/suspend.h>
#include <linux/pagevec.h>
#include <linux/blkdev.h>
#include <linux/slab.h>
33
#include <linux/ratelimit.h>
34
#include <linux/oom.h>
Linus Torvalds's avatar
Linus Torvalds committed
35
36
37
38
39
#include <linux/notifier.h>
#include <linux/topology.h>
#include <linux/sysctl.h>
#include <linux/cpu.h>
#include <linux/cpuset.h>
40
#include <linux/memory_hotplug.h>
Linus Torvalds's avatar
Linus Torvalds committed
41
42
#include <linux/nodemask.h>
#include <linux/vmalloc.h>
43
#include <linux/vmstat.h>
44
#include <linux/mempolicy.h>
45
#include <linux/stop_machine.h>
46
47
#include <linux/sort.h>
#include <linux/pfn.h>
48
#include <linux/backing-dev.h>
49
#include <linux/fault-inject.h>
50
#include <linux/page-isolation.h>
51
#include <linux/page_cgroup.h>
52
#include <linux/debugobjects.h>
53
#include <linux/kmemleak.h>
54
#include <linux/memory.h>
55
#include <linux/compaction.h>
56
#include <trace/events/kmem.h>
57
#include <linux/ftrace_event.h>
58
#include <linux/memcontrol.h>
59
#include <linux/prefetch.h>
60
#include <linux/migrate.h>
61
#include <linux/page-debug-flags.h>
Linus Torvalds's avatar
Linus Torvalds committed
62
63

#include <asm/tlbflush.h>
64
#include <asm/div64.h>
Linus Torvalds's avatar
Linus Torvalds committed
65
66
#include "internal.h"

67
68
69
70
71
#ifdef CONFIG_USE_PERCPU_NUMA_NODE_ID
DEFINE_PER_CPU(int, numa_node);
EXPORT_PER_CPU_SYMBOL(numa_node);
#endif

72
73
74
75
76
77
78
79
80
81
82
#ifdef CONFIG_HAVE_MEMORYLESS_NODES
/*
 * N.B., Do NOT reference the '_numa_mem_' per cpu variable directly.
 * It will not be defined when CONFIG_HAVE_MEMORYLESS_NODES is not defined.
 * Use the accessor functions set_numa_mem(), numa_mem_id() and cpu_to_mem()
 * defined in <linux/topology.h>.
 */
DEFINE_PER_CPU(int, _numa_mem_);		/* Kernel "local memory" node */
EXPORT_PER_CPU_SYMBOL(_numa_mem_);
#endif

Linus Torvalds's avatar
Linus Torvalds committed
83
/*
84
 * Array of node states.
Linus Torvalds's avatar
Linus Torvalds committed
85
 */
86
87
88
89
90
91
92
93
94
95
96
97
98
nodemask_t node_states[NR_NODE_STATES] __read_mostly = {
	[N_POSSIBLE] = NODE_MASK_ALL,
	[N_ONLINE] = { { [0] = 1UL } },
#ifndef CONFIG_NUMA
	[N_NORMAL_MEMORY] = { { [0] = 1UL } },
#ifdef CONFIG_HIGHMEM
	[N_HIGH_MEMORY] = { { [0] = 1UL } },
#endif
	[N_CPU] = { { [0] = 1UL } },
#endif	/* NUMA */
};
EXPORT_SYMBOL(node_states);

99
unsigned long totalram_pages __read_mostly;
100
unsigned long totalreserve_pages __read_mostly;
101
102
103
104
105
106
107
108
/*
 * When calculating the number of globally allowed dirty pages, there
 * is a certain number of per-zone reserves that should not be
 * considered dirtyable memory.  This is the sum of those reserves
 * over all existing zones that contribute dirtyable memory.
 */
unsigned long dirty_balance_reserve __read_mostly;

109
int percpu_pagelist_fraction;
110
gfp_t gfp_allowed_mask __read_mostly = GFP_BOOT_MASK;
Linus Torvalds's avatar
Linus Torvalds committed
111

112
113
114
115
116
117
118
119
120
#ifdef CONFIG_PM_SLEEP
/*
 * The following functions are used by the suspend/hibernate code to temporarily
 * change gfp_allowed_mask in order to avoid using I/O during memory allocations
 * while devices are suspended.  To avoid races with the suspend/hibernate code,
 * they should always be called with pm_mutex held (gfp_allowed_mask also should
 * only be modified with pm_mutex held, unless the suspend/hibernate code is
 * guaranteed not to run in parallel with that modification).
 */
121
122
123
124

static gfp_t saved_gfp_mask;

void pm_restore_gfp_mask(void)
125
126
{
	WARN_ON(!mutex_is_locked(&pm_mutex));
127
128
129
130
	if (saved_gfp_mask) {
		gfp_allowed_mask = saved_gfp_mask;
		saved_gfp_mask = 0;
	}
131
132
}

133
void pm_restrict_gfp_mask(void)
134
135
{
	WARN_ON(!mutex_is_locked(&pm_mutex));
136
137
138
	WARN_ON(saved_gfp_mask);
	saved_gfp_mask = gfp_allowed_mask;
	gfp_allowed_mask &= ~GFP_IOFS;
139
}
140
141
142
143
144
145
146

bool pm_suspended_storage(void)
{
	if ((gfp_allowed_mask & GFP_IOFS) == GFP_IOFS)
		return false;
	return true;
}
147
148
#endif /* CONFIG_PM_SLEEP */

149
150
151
152
#ifdef CONFIG_HUGETLB_PAGE_SIZE_VARIABLE
int pageblock_order __read_mostly;
#endif

153
static void __free_pages_ok(struct page *page, unsigned int order);
154

Linus Torvalds's avatar
Linus Torvalds committed
155
156
157
158
159
160
161
/*
 * results with 256, 32 in the lowmem_reserve sysctl:
 *	1G machine -> (16M dma, 800M-16M normal, 1G-800M high)
 *	1G machine -> (16M dma, 784M normal, 224M high)
 *	NORMAL allocation will leave 784M/256 of ram reserved in the ZONE_DMA
 *	HIGHMEM allocation will leave 224M/32 of ram reserved in ZONE_NORMAL
 *	HIGHMEM allocation will (224M+784M)/256 of ram reserved in ZONE_DMA
162
163
164
 *
 * TBD: should special case ZONE_DMA32 machines here - in those we normally
 * don't need any ZONE_NORMAL reservation
Linus Torvalds's avatar
Linus Torvalds committed
165
 */
166
int sysctl_lowmem_reserve_ratio[MAX_NR_ZONES-1] = {
167
#ifdef CONFIG_ZONE_DMA
168
	 256,
169
#endif
170
#ifdef CONFIG_ZONE_DMA32
171
	 256,
172
#endif
173
#ifdef CONFIG_HIGHMEM
Mel Gorman's avatar
Mel Gorman committed
174
	 32,
175
#endif
Mel Gorman's avatar
Mel Gorman committed
176
	 32,
177
};
Linus Torvalds's avatar
Linus Torvalds committed
178
179
180

EXPORT_SYMBOL(totalram_pages);

181
static char * const zone_names[MAX_NR_ZONES] = {
182
#ifdef CONFIG_ZONE_DMA
183
	 "DMA",
184
#endif
185
#ifdef CONFIG_ZONE_DMA32
186
	 "DMA32",
187
#endif
188
	 "Normal",
189
#ifdef CONFIG_HIGHMEM
Mel Gorman's avatar
Mel Gorman committed
190
	 "HighMem",
191
#endif
Mel Gorman's avatar
Mel Gorman committed
192
	 "Movable",
193
194
};

Linus Torvalds's avatar
Linus Torvalds committed
195
196
int min_free_kbytes = 1024;

197
198
static unsigned long __meminitdata nr_kernel_pages;
static unsigned long __meminitdata nr_all_pages;
199
static unsigned long __meminitdata dma_reserve;
Linus Torvalds's avatar
Linus Torvalds committed
200

Tejun Heo's avatar
Tejun Heo committed
201
202
203
204
205
206
207
208
209
210
211
#ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
static unsigned long __meminitdata arch_zone_lowest_possible_pfn[MAX_NR_ZONES];
static unsigned long __meminitdata arch_zone_highest_possible_pfn[MAX_NR_ZONES];
static unsigned long __initdata required_kernelcore;
static unsigned long __initdata required_movablecore;
static unsigned long __meminitdata zone_movable_pfn[MAX_NUMNODES];

/* movable_zone is the "real" zone pages in ZONE_MOVABLE are taken from */
int movable_zone;
EXPORT_SYMBOL(movable_zone);
#endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
212

Miklos Szeredi's avatar
Miklos Szeredi committed
213
214
#if MAX_NUMNODES > 1
int nr_node_ids __read_mostly = MAX_NUMNODES;
215
int nr_online_nodes __read_mostly = 1;
Miklos Szeredi's avatar
Miklos Szeredi committed
216
EXPORT_SYMBOL(nr_node_ids);
217
EXPORT_SYMBOL(nr_online_nodes);
Miklos Szeredi's avatar
Miklos Szeredi committed
218
219
#endif

220
221
int page_group_by_mobility_disabled __read_mostly;

222
223
static void set_pageblock_migratetype(struct page *page, int migratetype)
{
224
225
226
227

	if (unlikely(page_group_by_mobility_disabled))
		migratetype = MIGRATE_UNMOVABLE;

228
229
230
231
	set_pageblock_flags_group(page, (unsigned long)migratetype,
					PB_migrate, PB_migrate_end);
}

232
233
bool oom_killer_disabled __read_mostly;

Nick Piggin's avatar
Nick Piggin committed
234
#ifdef CONFIG_DEBUG_VM
235
static int page_outside_zone_boundaries(struct zone *zone, struct page *page)
Linus Torvalds's avatar
Linus Torvalds committed
236
{
237
238
239
	int ret = 0;
	unsigned seq;
	unsigned long pfn = page_to_pfn(page);
240

241
242
243
244
245
246
247
248
249
	do {
		seq = zone_span_seqbegin(zone);
		if (pfn >= zone->zone_start_pfn + zone->spanned_pages)
			ret = 1;
		else if (pfn < zone->zone_start_pfn)
			ret = 1;
	} while (zone_span_seqretry(zone, seq));

	return ret;
250
251
252
253
}

static int page_is_consistent(struct zone *zone, struct page *page)
{
254
	if (!pfn_valid_within(page_to_pfn(page)))
255
		return 0;
Linus Torvalds's avatar
Linus Torvalds committed
256
	if (zone != page_zone(page))
257
258
259
260
261
262
263
264
265
266
		return 0;

	return 1;
}
/*
 * Temporary debugging check for pages not lying within a given zone.
 */
static int bad_range(struct zone *zone, struct page *page)
{
	if (page_outside_zone_boundaries(zone, page))
Linus Torvalds's avatar
Linus Torvalds committed
267
		return 1;
268
269
270
	if (!page_is_consistent(zone, page))
		return 1;

Linus Torvalds's avatar
Linus Torvalds committed
271
272
	return 0;
}
Nick Piggin's avatar
Nick Piggin committed
273
274
275
276
277
278
279
#else
static inline int bad_range(struct zone *zone, struct page *page)
{
	return 0;
}
#endif

280
static void bad_page(struct page *page)
Linus Torvalds's avatar
Linus Torvalds committed
281
{
282
283
284
285
	static unsigned long resume;
	static unsigned long nr_shown;
	static unsigned long nr_unshown;

286
287
	/* Don't complain about poisoned pages */
	if (PageHWPoison(page)) {
288
		reset_page_mapcount(page); /* remove PageBuddy */
289
290
291
		return;
	}

292
293
294
295
296
297
298
299
300
301
	/*
	 * Allow a burst of 60 reports, then keep quiet for that minute;
	 * or allow a steady drip of one report per second.
	 */
	if (nr_shown == 60) {
		if (time_before(jiffies, resume)) {
			nr_unshown++;
			goto out;
		}
		if (nr_unshown) {
302
303
			printk(KERN_ALERT
			      "BUG: Bad page state: %lu messages suppressed\n",
304
305
306
307
308
309
310
311
				nr_unshown);
			nr_unshown = 0;
		}
		nr_shown = 0;
	}
	if (nr_shown++ == 0)
		resume = jiffies + 60 * HZ;

312
	printk(KERN_ALERT "BUG: Bad page state in process %s  pfn:%05lx\n",
313
		current->comm, page_to_pfn(page));
314
	dump_page(page);
315

316
	print_modules();
Linus Torvalds's avatar
Linus Torvalds committed
317
	dump_stack();
318
out:
319
	/* Leave bad fields for debug, except PageBuddy could make trouble */
320
	reset_page_mapcount(page); /* remove PageBuddy */
321
	add_taint(TAINT_BAD_PAGE);
Linus Torvalds's avatar
Linus Torvalds committed
322
323
324
325
326
327
328
329
330
}

/*
 * Higher-order pages are called "compound pages".  They are structured thusly:
 *
 * The first PAGE_SIZE page is called the "head page".
 *
 * The remaining PAGE_SIZE pages are called "tail pages".
 *
331
332
 * All pages have PG_compound set.  All tail pages have their ->first_page
 * pointing at the head page.
Linus Torvalds's avatar
Linus Torvalds committed
333
 *
334
335
336
 * The first tail page's ->lru.next holds the address of the compound page's
 * put_page() function.  Its ->lru.prev holds the order of allocation.
 * This usage means that zero-order pages may not be compound.
Linus Torvalds's avatar
Linus Torvalds committed
337
 */
338
339
340

static void free_compound_page(struct page *page)
{
341
	__free_pages_ok(page, compound_order(page));
342
343
}

344
void prep_compound_page(struct page *page, unsigned long order)
345
346
347
348
349
350
351
352
353
354
{
	int i;
	int nr_pages = 1 << order;

	set_compound_page_dtor(page, free_compound_page);
	set_compound_order(page, order);
	__SetPageHead(page);
	for (i = 1; i < nr_pages; i++) {
		struct page *p = page + i;
		__SetPageTail(p);
355
		set_page_count(p, 0);
356
357
358
359
		p->first_page = page;
	}
}

360
/* update __split_huge_page_refcount if you change this function */
361
static int destroy_compound_page(struct page *page, unsigned long order)
Linus Torvalds's avatar
Linus Torvalds committed
362
363
364
{
	int i;
	int nr_pages = 1 << order;
365
	int bad = 0;
Linus Torvalds's avatar
Linus Torvalds committed
366

367
368
	if (unlikely(compound_order(page) != order) ||
	    unlikely(!PageHead(page))) {
369
		bad_page(page);
370
371
		bad++;
	}
Linus Torvalds's avatar
Linus Torvalds committed
372

373
	__ClearPageHead(page);
374

375
376
	for (i = 1; i < nr_pages; i++) {
		struct page *p = page + i;
Linus Torvalds's avatar
Linus Torvalds committed
377

378
		if (unlikely(!PageTail(p) || (p->first_page != page))) {
379
			bad_page(page);
380
381
			bad++;
		}
382
		__ClearPageTail(p);
Linus Torvalds's avatar
Linus Torvalds committed
383
	}
384
385

	return bad;
Linus Torvalds's avatar
Linus Torvalds committed
386
387
}

Nick Piggin's avatar
Nick Piggin committed
388
389
390
391
static inline void prep_zero_page(struct page *page, int order, gfp_t gfp_flags)
{
	int i;

392
393
394
395
	/*
	 * clear_highpage() will use KM_USER0, so it's a bug to use __GFP_ZERO
	 * and __GFP_HIGHMEM from hard or soft interrupt context.
	 */
Nick Piggin's avatar
Nick Piggin committed
396
	VM_BUG_ON((gfp_flags & __GFP_HIGHMEM) && in_interrupt());
Nick Piggin's avatar
Nick Piggin committed
397
398
399
400
	for (i = 0; i < (1 << order); i++)
		clear_highpage(page + i);
}

401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
#ifdef CONFIG_DEBUG_PAGEALLOC
unsigned int _debug_guardpage_minorder;

static int __init debug_guardpage_minorder_setup(char *buf)
{
	unsigned long res;

	if (kstrtoul(buf, 10, &res) < 0 ||  res > MAX_ORDER / 2) {
		printk(KERN_ERR "Bad debug_guardpage_minorder value\n");
		return 0;
	}
	_debug_guardpage_minorder = res;
	printk(KERN_INFO "Setting debug_guardpage_minorder to %lu\n", res);
	return 0;
}
__setup("debug_guardpage_minorder=", debug_guardpage_minorder_setup);

static inline void set_page_guard_flag(struct page *page)
{
	__set_bit(PAGE_DEBUG_FLAG_GUARD, &page->debug_flags);
}

static inline void clear_page_guard_flag(struct page *page)
{
	__clear_bit(PAGE_DEBUG_FLAG_GUARD, &page->debug_flags);
}
#else
static inline void set_page_guard_flag(struct page *page) { }
static inline void clear_page_guard_flag(struct page *page) { }
#endif

432
433
static inline void set_page_order(struct page *page, int order)
{
434
	set_page_private(page, order);
435
	__SetPageBuddy(page);
Linus Torvalds's avatar
Linus Torvalds committed
436
437
438
439
}

static inline void rmv_page_order(struct page *page)
{
440
	__ClearPageBuddy(page);
441
	set_page_private(page, 0);
Linus Torvalds's avatar
Linus Torvalds committed
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
}

/*
 * Locate the struct page for both the matching buddy in our
 * pair (buddy1) and the combined O(n+1) page they form (page).
 *
 * 1) Any buddy B1 will have an order O twin B2 which satisfies
 * the following equation:
 *     B2 = B1 ^ (1 << O)
 * For example, if the starting buddy (buddy2) is #8 its order
 * 1 buddy is #10:
 *     B2 = 8 ^ (1 << 1) = 8 ^ 2 = 10
 *
 * 2) Any buddy B will have an order O+1 parent P which
 * satisfies the following equation:
 *     P = B & ~(1 << O)
 *
Andreas Mohr's avatar
Andreas Mohr committed
459
 * Assumption: *_mem_map is contiguous at least up to MAX_ORDER
Linus Torvalds's avatar
Linus Torvalds committed
460
461
 */
static inline unsigned long
462
__find_buddy_index(unsigned long page_idx, unsigned int order)
Linus Torvalds's avatar
Linus Torvalds committed
463
{
464
	return page_idx ^ (1 << order);
Linus Torvalds's avatar
Linus Torvalds committed
465
466
467
468
469
}

/*
 * This function checks whether a page is free && is the buddy
 * we can do coalesce a page and its buddy if
Nick Piggin's avatar
Nick Piggin committed
470
 * (a) the buddy is not in a hole &&
471
 * (b) the buddy is in the buddy system &&
472
473
 * (c) a page and its buddy have the same order &&
 * (d) a page and its buddy are in the same zone.
474
 *
Andrea Arcangeli's avatar
Andrea Arcangeli committed
475
476
 * For recording whether a page is in the buddy system, we set ->_mapcount -2.
 * Setting, clearing, and testing _mapcount -2 is serialized by zone->lock.
Linus Torvalds's avatar
Linus Torvalds committed
477
 *
478
 * For recording page's order, we use page_private(page).
Linus Torvalds's avatar
Linus Torvalds committed
479
 */
480
481
static inline int page_is_buddy(struct page *page, struct page *buddy,
								int order)
Linus Torvalds's avatar
Linus Torvalds committed
482
{
483
	if (!pfn_valid_within(page_to_pfn(buddy)))
Nick Piggin's avatar
Nick Piggin committed
484
485
		return 0;

486
487
488
	if (page_zone_id(page) != page_zone_id(buddy))
		return 0;

489
490
491
492
493
	if (page_is_guard(buddy) && page_order(buddy) == order) {
		VM_BUG_ON(page_count(buddy) != 0);
		return 1;
	}

494
	if (PageBuddy(buddy) && page_order(buddy) == order) {
495
		VM_BUG_ON(page_count(buddy) != 0);
496
		return 1;
497
	}
498
	return 0;
Linus Torvalds's avatar
Linus Torvalds committed
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
}

/*
 * Freeing function for a buddy system allocator.
 *
 * The concept of a buddy system is to maintain direct-mapped table
 * (containing bit values) for memory blocks of various "orders".
 * The bottom level table contains the map for the smallest allocatable
 * units of memory (here, pages), and each level above it describes
 * pairs of units from the levels below, hence, "buddies".
 * At a high level, all that happens here is marking the table entry
 * at the bottom level available, and propagating the changes upward
 * as necessary, plus some accounting needed to play nicely with other
 * parts of the VM system.
 * At each level, we keep a list of pages, which are heads of continuous
Andrea Arcangeli's avatar
Andrea Arcangeli committed
514
 * free pages of length of (1 << order) and marked with _mapcount -2. Page's
515
 * order is recorded in page_private(page) field.
Linus Torvalds's avatar
Linus Torvalds committed
516
 * So when we are allocating or freeing one, we can derive the state of the
517
518
 * other.  That is, if we allocate a small block, and both were
 * free, the remainder of the region must be split into blocks.
Linus Torvalds's avatar
Linus Torvalds committed
519
 * If a block is freed, and its buddy is also free, then this
520
 * triggers coalescing into a block of larger size.
Linus Torvalds's avatar
Linus Torvalds committed
521
522
523
524
 *
 * -- wli
 */

Nick Piggin's avatar
Nick Piggin committed
525
static inline void __free_one_page(struct page *page,
526
527
		struct zone *zone, unsigned int order,
		int migratetype)
Linus Torvalds's avatar
Linus Torvalds committed
528
529
{
	unsigned long page_idx;
530
	unsigned long combined_idx;
531
	unsigned long uninitialized_var(buddy_idx);
532
	struct page *buddy;
Linus Torvalds's avatar
Linus Torvalds committed
533

534
	if (unlikely(PageCompound(page)))
535
536
		if (unlikely(destroy_compound_page(page, order)))
			return;
Linus Torvalds's avatar
Linus Torvalds committed
537

538
539
	VM_BUG_ON(migratetype == -1);

Linus Torvalds's avatar
Linus Torvalds committed
540
541
	page_idx = page_to_pfn(page) & ((1 << MAX_ORDER) - 1);

542
	VM_BUG_ON(page_idx & ((1 << order) - 1));
Nick Piggin's avatar
Nick Piggin committed
543
	VM_BUG_ON(bad_range(zone, page));
Linus Torvalds's avatar
Linus Torvalds committed
544
545

	while (order < MAX_ORDER-1) {
546
547
		buddy_idx = __find_buddy_index(page_idx, order);
		buddy = page + (buddy_idx - page_idx);
548
		if (!page_is_buddy(page, buddy, order))
549
			break;
550
551
552
553
554
555
556
557
558
559
560
561
562
		/*
		 * Our buddy is free or it is CONFIG_DEBUG_PAGEALLOC guard page,
		 * merge with it and move up one order.
		 */
		if (page_is_guard(buddy)) {
			clear_page_guard_flag(buddy);
			set_page_private(page, 0);
			__mod_zone_page_state(zone, NR_FREE_PAGES, 1 << order);
		} else {
			list_del(&buddy->lru);
			zone->free_area[order].nr_free--;
			rmv_page_order(buddy);
		}
563
		combined_idx = buddy_idx & page_idx;
Linus Torvalds's avatar
Linus Torvalds committed
564
565
566
567
568
		page = page + (combined_idx - page_idx);
		page_idx = combined_idx;
		order++;
	}
	set_page_order(page, order);
569
570
571
572
573
574
575
576
577

	/*
	 * If this is not the largest possible page, check if the buddy
	 * of the next-highest order is free. If it is, it's possible
	 * that pages are being freed that will coalesce soon. In case,
	 * that is happening, add the free page to the tail of the list
	 * so it's less likely to be used soon and more likely to be merged
	 * as a higher order page
	 */
578
	if ((order < MAX_ORDER-2) && pfn_valid_within(page_to_pfn(buddy))) {
579
		struct page *higher_page, *higher_buddy;
580
581
582
583
		combined_idx = buddy_idx & page_idx;
		higher_page = page + (combined_idx - page_idx);
		buddy_idx = __find_buddy_index(combined_idx, order + 1);
		higher_buddy = page + (buddy_idx - combined_idx);
584
585
586
587
588
589
590
591
592
		if (page_is_buddy(higher_page, higher_buddy, order + 1)) {
			list_add_tail(&page->lru,
				&zone->free_area[order].free_list[migratetype]);
			goto out;
		}
	}

	list_add(&page->lru, &zone->free_area[order].free_list[migratetype]);
out:
Linus Torvalds's avatar
Linus Torvalds committed
593
594
595
	zone->free_area[order].nr_free++;
}

596
597
598
599
600
601
602
603
604
605
606
/*
 * free_page_mlock() -- clean up attempts to free and mlocked() page.
 * Page should not be on lru, so no need to fix that up.
 * free_pages_check() will verify...
 */
static inline void free_page_mlock(struct page *page)
{
	__dec_zone_page_state(page, NR_MLOCK);
	__count_vm_event(UNEVICTABLE_MLOCKFREED);
}

607
static inline int free_pages_check(struct page *page)
Linus Torvalds's avatar
Linus Torvalds committed
608
{
Nick Piggin's avatar
Nick Piggin committed
609
610
	if (unlikely(page_mapcount(page) |
		(page->mapping != NULL)  |
611
		(atomic_read(&page->_count) != 0) |
612
613
		(page->flags & PAGE_FLAGS_CHECK_AT_FREE) |
		(mem_cgroup_bad_page_check(page)))) {
614
		bad_page(page);
615
		return 1;
616
	}
617
618
619
	if (page->flags & PAGE_FLAGS_CHECK_AT_PREP)
		page->flags &= ~PAGE_FLAGS_CHECK_AT_PREP;
	return 0;
Linus Torvalds's avatar
Linus Torvalds committed
620
621
622
}

/*
623
 * Frees a number of pages from the PCP lists
Linus Torvalds's avatar
Linus Torvalds committed
624
 * Assumes all pages on list are in same zone, and of same order.
625
 * count is the number of pages to free.
Linus Torvalds's avatar
Linus Torvalds committed
626
627
628
629
630
631
632
 *
 * If the zone was previously in an "all pages pinned" state then look to
 * see if this freeing clears that state.
 *
 * And clear the zone's pages_scanned counter, to hold off the "all pages are
 * pinned" detection logic.
 */
633
634
static void free_pcppages_bulk(struct zone *zone, int count,
					struct per_cpu_pages *pcp)
Linus Torvalds's avatar
Linus Torvalds committed
635
{
636
	int migratetype = 0;
637
	int batch_free = 0;
638
	int to_free = count;
639

Nick Piggin's avatar
Nick Piggin committed
640
	spin_lock(&zone->lock);
641
	zone->all_unreclaimable = 0;
Linus Torvalds's avatar
Linus Torvalds committed
642
	zone->pages_scanned = 0;
643

644
	while (to_free) {
Nick Piggin's avatar
Nick Piggin committed
645
		struct page *page;
646
647
648
		struct list_head *list;

		/*
649
650
651
652
653
		 * Remove pages from lists in a round-robin fashion. A
		 * batch_free count is maintained that is incremented when an
		 * empty list is encountered.  This is so more pages are freed
		 * off fuller lists instead of spinning excessively around empty
		 * lists
654
655
		 */
		do {
656
			batch_free++;
657
658
659
660
			if (++migratetype == MIGRATE_PCPTYPES)
				migratetype = 0;
			list = &pcp->lists[migratetype];
		} while (list_empty(list));
Nick Piggin's avatar
Nick Piggin committed
661

662
663
664
665
		/* This is the only non-empty list. Free them all. */
		if (batch_free == MIGRATE_PCPTYPES)
			batch_free = to_free;

666
667
668
669
		do {
			page = list_entry(list->prev, struct page, lru);
			/* must delete as __free_one_page list manipulates */
			list_del(&page->lru);
670
671
672
			/* MIGRATE_MOVABLE list may include MIGRATE_RESERVEs */
			__free_one_page(page, zone, 0, page_private(page));
			trace_mm_page_pcpu_drain(page, 0, page_private(page));
673
		} while (--to_free && --batch_free && !list_empty(list));
Linus Torvalds's avatar
Linus Torvalds committed
674
	}
675
	__mod_zone_page_state(zone, NR_FREE_PAGES, count);
Nick Piggin's avatar
Nick Piggin committed
676
	spin_unlock(&zone->lock);
Linus Torvalds's avatar
Linus Torvalds committed
677
678
}

679
680
static void free_one_page(struct zone *zone, struct page *page, int order,
				int migratetype)
Linus Torvalds's avatar
Linus Torvalds committed
681
{
682
	spin_lock(&zone->lock);
683
	zone->all_unreclaimable = 0;
684
	zone->pages_scanned = 0;
685

686
	__free_one_page(page, zone, order, migratetype);
687
	__mod_zone_page_state(zone, NR_FREE_PAGES, 1 << order);
688
	spin_unlock(&zone->lock);
Nick Piggin's avatar
Nick Piggin committed
689
690
}

691
static bool free_pages_prepare(struct page *page, unsigned int order)
Nick Piggin's avatar
Nick Piggin committed
692
{
Linus Torvalds's avatar
Linus Torvalds committed
693
	int i;
694
	int bad = 0;
Linus Torvalds's avatar
Linus Torvalds committed
695

696
	trace_mm_page_free(page, order);
697
698
	kmemcheck_free_shadow(page, order);

Andrea Arcangeli's avatar
Andrea Arcangeli committed
699
700
701
702
	if (PageAnon(page))
		page->mapping = NULL;
	for (i = 0; i < (1 << order); i++)
		bad += free_pages_check(page + i);
703
	if (bad)
704
		return false;
705

706
	if (!PageHighMem(page)) {
Nick Piggin's avatar
Nick Piggin committed
707
		debug_check_no_locks_freed(page_address(page),PAGE_SIZE<<order);
708
709
710
		debug_check_no_obj_freed(page_address(page),
					   PAGE_SIZE << order);
	}
Nick Piggin's avatar
Nick Piggin committed
711
	arch_free_page(page, order);
Nick Piggin's avatar
Nick Piggin committed
712
	kernel_map_pages(page, 1 << order, 0);
Nick Piggin's avatar
Nick Piggin committed
713

714
715
716
717
718
719
720
721
722
723
724
	return true;
}

static void __free_pages_ok(struct page *page, unsigned int order)
{
	unsigned long flags;
	int wasMlocked = __TestClearPageMlocked(page);

	if (!free_pages_prepare(page, order))
		return;

Nick Piggin's avatar
Nick Piggin committed
725
	local_irq_save(flags);
726
	if (unlikely(wasMlocked))
727
		free_page_mlock(page);
728
	__count_vm_events(PGFREE, 1 << order);
729
730
	free_one_page(page_zone(page), page, order,
					get_pageblock_migratetype(page));
Nick Piggin's avatar
Nick Piggin committed
731
	local_irq_restore(flags);
Linus Torvalds's avatar
Linus Torvalds committed
732
733
}

734
void __meminit __free_pages_bootmem(struct page *page, unsigned int order)
735
{
736
737
	unsigned int nr_pages = 1 << order;
	unsigned int loop;
738

739
740
741
742
743
744
745
746
	prefetchw(page);
	for (loop = 0; loop < nr_pages; loop++) {
		struct page *p = &page[loop];

		if (loop + 1 < nr_pages)
			prefetchw(p + 1);
		__ClearPageReserved(p);
		set_page_count(p, 0);
747
	}
748
749
750

	set_page_refcounted(page);
	__free_pages(page, order);
751
752
}

Linus Torvalds's avatar
Linus Torvalds committed
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767

/*
 * The order of subdivision here is critical for the IO subsystem.
 * Please do not alter this order without good reasons and regression
 * testing. Specifically, as large blocks of memory are subdivided,
 * the order in which smaller blocks are delivered depends on the order
 * they're subdivided in this function. This is the primary factor
 * influencing the order in which pages are delivered to the IO
 * subsystem according to empirical testing, and this is also justified
 * by considering the behavior of a buddy system containing a single
 * large block of memory acted on by a series of small allocations.
 * This behavior is a critical factor in sglist merging's success.
 *
 * -- wli
 */
Nick Piggin's avatar
Nick Piggin committed
768
static inline void expand(struct zone *zone, struct page *page,
769
770
	int low, int high, struct free_area *area,
	int migratetype)
Linus Torvalds's avatar
Linus Torvalds committed
771
772
773
774
775
776
777
{
	unsigned long size = 1 << high;

	while (high > low) {
		area--;
		high--;
		size >>= 1;
Nick Piggin's avatar
Nick Piggin committed
778
		VM_BUG_ON(bad_range(zone, &page[size]));
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795

#ifdef CONFIG_DEBUG_PAGEALLOC
		if (high < debug_guardpage_minorder()) {
			/*
			 * Mark as guard pages (or page), that will allow to
			 * merge back to allocator when buddy will be freed.
			 * Corresponding page table entries will not be touched,
			 * pages will stay not present in virtual address space
			 */
			INIT_LIST_HEAD(&page[size].lru);
			set_page_guard_flag(&page[size]);
			set_page_private(&page[size], high);
			/* Guard pages are not available for any usage */
			__mod_zone_page_state(zone, NR_FREE_PAGES, -(1 << high));
			continue;
		}
#endif
796
		list_add(&page[size].lru, &area->free_list[migratetype]);
Linus Torvalds's avatar
Linus Torvalds committed
797
798
799
800
801
802
803
804
		area->nr_free++;
		set_page_order(&page[size], high);
	}
}

/*
 * This page is about to be returned from the page allocator
 */
805
static inline int check_new_page(struct page *page)
Linus Torvalds's avatar
Linus Torvalds committed
806
{
Nick Piggin's avatar
Nick Piggin committed
807
808
	if (unlikely(page_mapcount(page) |
		(page->mapping != NULL)  |
809
		(atomic_read(&page->_count) != 0)  |
810
811
		(page->flags & PAGE_FLAGS_CHECK_AT_PREP) |
		(mem_cgroup_bad_page_check(page)))) {
812
		bad_page(page);
813
		return 1;
814
	}
815
816
817
818
819
820
821
822
823
824
825
826
	return 0;
}

static int prep_new_page(struct page *page, int order, gfp_t gfp_flags)
{
	int i;

	for (i = 0; i < (1 << order); i++) {
		struct page *p = page + i;
		if (unlikely(check_new_page(p)))
			return 1;
	}
827

828
	set_page_private(page, 0);
829
	set_page_refcounted(page);
Nick Piggin's avatar
Nick Piggin committed
830
831

	arch_alloc_page(page, order);
Linus Torvalds's avatar
Linus Torvalds committed
832
	kernel_map_pages(page, 1 << order, 1);
Nick Piggin's avatar
Nick Piggin committed
833
834
835
836
837
838
839

	if (gfp_flags & __GFP_ZERO)
		prep_zero_page(page, order, gfp_flags);

	if (order && (gfp_flags & __GFP_COMP))
		prep_compound_page(page, order);

840
	return 0;
Linus Torvalds's avatar
Linus Torvalds committed
841
842
}

843
844
845
846
/*
 * Go through the free lists for the given migratetype and remove
 * the smallest available page from the freelists
 */
847
848
static inline
struct page *__rmqueue_smallest(struct zone *zone, unsigned int order,
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
						int migratetype)
{
	unsigned int current_order;
	struct free_area * area;
	struct page *page;

	/* Find a page of the appropriate size in the preferred list */
	for (current_order = order; current_order < MAX_ORDER; ++current_order) {
		area = &(zone->free_area[current_order]);
		if (list_empty(&area->free_list[migratetype]))
			continue;

		page = list_entry(area->free_list[migratetype].next,
							struct page, lru);
		list_del(&page->lru);
		rmv_page_order(page);
		area->nr_free--;
		expand(zone, page, order, current_order, area, migratetype);
		return page;
	}

	return NULL;
}


874
875
876
877
/*
 * This array describes the order lists are fallen back to when
 * the free lists for the desirable migrate type are depleted
 */
878
static int fallbacks[MIGRATE_TYPES][3] = {
879
880
881
	[MIGRATE_UNMOVABLE]   = { MIGRATE_RECLAIMABLE, MIGRATE_MOVABLE,   MIGRATE_RESERVE },
	[MIGRATE_RECLAIMABLE] = { MIGRATE_UNMOVABLE,   MIGRATE_MOVABLE,   MIGRATE_RESERVE },
	[MIGRATE_MOVABLE]     = { MIGRATE_RECLAIMABLE, MIGRATE_UNMOVABLE, MIGRATE_RESERVE },
882
883
	[MIGRATE_RESERVE]     = { MIGRATE_RESERVE }, /* Never used */
	[MIGRATE_ISOLATE]     = { MIGRATE_RESERVE }, /* Never used */
884
885
};

886
887
/*
 * Move the free pages in a range to the free lists of the requested type.
888
 * Note that start_page and end_pages are not aligned on a pageblock
889
890
 * boundary. If alignment is required, use move_freepages_block()
 */
Adrian Bunk's avatar
Adrian Bunk committed
891
892
893
static int move_freepages(struct zone *zone,
			  struct page *start_page, struct page *end_page,
			  int migratetype)
894
895
896
{
	struct page *page;
	unsigned long order;
897
	int pages_moved = 0;
898
899
900
901
902
903
904

#ifndef CONFIG_HOLES_IN_ZONE
	/*
	 * page_zone is not safe to call in this context when
	 * CONFIG_HOLES_IN_ZONE is set. This bug check is probably redundant
	 * anyway as we check zone boundaries in move_freepages_block().
	 * Remove at a later date when no bug reports exist related to
Mel Gorman's avatar
Mel Gorman committed
905
	 * grouping pages by mobility
906
907
908
909
910
	 */
	BUG_ON(page_zone(start_page) != page_zone(end_page));
#endif

	for (page = start_page; page <= end_page;) {
911
912
913
		/* Make sure we are not inadvertently changing nodes */
		VM_BUG_ON(page_to_nid(page) != zone_to_nid(zone));

914
915
916
917
918
919
920
921
922
923
924
		if (!pfn_valid_within(page_to_pfn(page))) {
			page++;
			continue;
		}

		if (!PageBuddy(page)) {
			page++;
			continue;
		}

		order = page_order(page);
925
926
		list_move(&page->lru,
			  &zone->free_area[order].free_list[migratetype]);
927
		page += 1 << order;
928
		pages_moved += 1 << order;
929
930
	}

931
	return pages_moved;
932
933
}

Adrian Bunk's avatar
Adrian Bunk committed
934
935
static int move_freepages_block(struct zone *zone, struct page *page,
				int migratetype)
936
937
938
939
940
{
	unsigned long start_pfn, end_pfn;
	struct page *start_page, *end_page;

	start_pfn = page_to_pfn(page);
941
	start_pfn = start_pfn & ~(pageblock_nr_pages-1);
942
	start_page = pfn_to_page(start_pfn);
943
944
	end_page = start_page + pageblock_nr_pages - 1;
	end_pfn = start_pfn + pageblock_nr_pages - 1;
945
946
947
948
949
950
951
952
953
954

	/* Do not cross zone boundaries */
	if (start_pfn < zone->zone_start_pfn)
		start_page = page;
	if (end_pfn >= zone->zone_start_pfn + zone->spanned_pages)
		return 0;

	return move_freepages(zone, start_page, end_page, migratetype);
}

955
956
957
958
959
960
961
962
963
964
965
static void change_pageblock_range(struct page *pageblock_page,
					int start_order, int migratetype)
{
	int nr_pageblocks = 1 << (start_order - pageblock_order);

	while (nr_pageblocks--) {
		set_pageblock_migratetype(pageblock_page, migratetype);
		pageblock_page += pageblock_nr_pages;
	}
}

966
/* Remove an element from the buddy allocator from the fallback list */
967
968
static inline struct page *
__rmqueue_fallback(struct zone *zone, int order, int start_migratetype)
969
970
971
972
973
974
975
976
977
{
	struct free_area * area;
	int current_order;
	struct page *page;
	int migratetype, i;

	/* Find the largest possible block of pages in the other list */
	for (current_order = MAX_ORDER-1; current_order >= order;
						--current_order) {
978
		for (i = 0;; i++) {
979
980
			migratetype = fallbacks[start_migratetype][i];

981
982
			/* MIGRATE_RESERVE handled later if necessary */
			if (migratetype == MIGRATE_RESERVE)
983
				break;
984

985
986
987
988
989
990
991
992
993
			area = &(zone->free_area[current_order]);
			if (list_empty(&area->free_list[migratetype]))
				continue;

			page = list_entry(area->free_list[migratetype].next,
					struct page, lru);
			area->nr_free--;

			/*
994
			 * If breaking a large block of pages, move all free
995
996
			 * pages to the preferred allocation list. If falling
			 * back for a reclaimable kernel allocation, be more
Lucas De Marchi's avatar
Lucas De Marchi committed
997
			 * aggressive about taking ownership of free pages
998
			 */
999
			if (unlikely(current_order >= (pageblock_order >> 1)) ||
1000
1001
					start_migratetype == MIGRATE_RECLAIMABLE ||
					page_group_by_mobility_disabled) {
1002
1003
1004
1005
1006
				unsigned long pages;
				pages = move_freepages_block(zone, page,
								start_migratetype);

				/* Claim the whole block if over half of it is free */
1007
1008
				if (pages >= (1 << (pageblock_order-1)) ||
						page_group_by_mobility_disabled)
1009
1010
1011
					set_pageblock_migratetype(page,
								start_migratetype);

1012
				migratetype = start_migratetype;
1013
			}
1014
1015
1016
1017
1018

			/* Remove the page from the freelists */
			list_del(&page->lru);
			rmv_page_order(page);

1019
1020
1021
			/* Take ownership for orders >= pageblock_order */
			if (current_order >= pageblock_order)
				change_pageblock_range(page, current_order,
1022
1023
1024
							start_migratetype);

			expand(zone, page, order, current_order, area, migratetype);
1025
1026
1027
1028

			trace_mm_page_alloc_extfrag(page, order, current_order,
				start_migratetype, migratetype);

1029
1030
1031
1032
			return page;
		}
	}

1033
	return NULL;
1034
1035
}

1036
/*
Linus Torvalds's avatar
Linus Torvalds committed
1037
1038
1039
 * Do the hard work of removing an element from the buddy allocator.
 * Call me with the zone->lock already held.
 */
1040
1041
static struct page *__rmqueue(struct zone *zone, unsigned int order,
						int migratetype)
Linus Torvalds's avatar
Linus Torvalds committed
1042
1043
1044
{
	struct page *page;

1045
retry_reserve:
1046
	page = __rmqueue_smallest(zone, order, migratetype);
1047

1048
	if (unlikely(!page) && migratetype != MIGRATE_RESERVE) {
1049
		page = __rmqueue_fallback(zone, order, migratetype);
1050

1051
1052