page_alloc.c 150 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>
Linus Torvalds's avatar
Linus Torvalds committed
60
61

#include <asm/tlbflush.h>
62
#include <asm/div64.h>
Linus Torvalds's avatar
Linus Torvalds committed
63
64
#include "internal.h"

65
66
67
68
69
#ifdef CONFIG_USE_PERCPU_NUMA_NODE_ID
DEFINE_PER_CPU(int, numa_node);
EXPORT_PER_CPU_SYMBOL(numa_node);
#endif

70
71
72
73
74
75
76
77
78
79
80
#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
81
/*
82
 * Array of node states.
Linus Torvalds's avatar
Linus Torvalds committed
83
 */
84
85
86
87
88
89
90
91
92
93
94
95
96
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);

97
unsigned long totalram_pages __read_mostly;
98
unsigned long totalreserve_pages __read_mostly;
99
int percpu_pagelist_fraction;
100
gfp_t gfp_allowed_mask __read_mostly = GFP_BOOT_MASK;
Linus Torvalds's avatar
Linus Torvalds committed
101

102
103
104
105
106
107
108
109
110
#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).
 */
111
112
113
114

static gfp_t saved_gfp_mask;

void pm_restore_gfp_mask(void)
115
116
{
	WARN_ON(!mutex_is_locked(&pm_mutex));
117
118
119
120
	if (saved_gfp_mask) {
		gfp_allowed_mask = saved_gfp_mask;
		saved_gfp_mask = 0;
	}
121
122
}

123
void pm_restrict_gfp_mask(void)
124
125
{
	WARN_ON(!mutex_is_locked(&pm_mutex));
126
127
128
	WARN_ON(saved_gfp_mask);
	saved_gfp_mask = gfp_allowed_mask;
	gfp_allowed_mask &= ~GFP_IOFS;
129
130
131
}
#endif /* CONFIG_PM_SLEEP */

132
133
134
135
#ifdef CONFIG_HUGETLB_PAGE_SIZE_VARIABLE
int pageblock_order __read_mostly;
#endif

136
static void __free_pages_ok(struct page *page, unsigned int order);
137

Linus Torvalds's avatar
Linus Torvalds committed
138
139
140
141
142
143
144
/*
 * 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
145
146
147
 *
 * 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
148
 */
149
int sysctl_lowmem_reserve_ratio[MAX_NR_ZONES-1] = {
150
#ifdef CONFIG_ZONE_DMA
151
	 256,
152
#endif
153
#ifdef CONFIG_ZONE_DMA32
154
	 256,
155
#endif
156
#ifdef CONFIG_HIGHMEM
Mel Gorman's avatar
Mel Gorman committed
157
	 32,
158
#endif
Mel Gorman's avatar
Mel Gorman committed
159
	 32,
160
};
Linus Torvalds's avatar
Linus Torvalds committed
161
162
163

EXPORT_SYMBOL(totalram_pages);

164
static char * const zone_names[MAX_NR_ZONES] = {
165
#ifdef CONFIG_ZONE_DMA
166
	 "DMA",
167
#endif
168
#ifdef CONFIG_ZONE_DMA32
169
	 "DMA32",
170
#endif
171
	 "Normal",
172
#ifdef CONFIG_HIGHMEM
Mel Gorman's avatar
Mel Gorman committed
173
	 "HighMem",
174
#endif
Mel Gorman's avatar
Mel Gorman committed
175
	 "Movable",
176
177
};

Linus Torvalds's avatar
Linus Torvalds committed
178
179
int min_free_kbytes = 1024;

180
181
static unsigned long __meminitdata nr_kernel_pages;
static unsigned long __meminitdata nr_all_pages;
182
static unsigned long __meminitdata dma_reserve;
Linus Torvalds's avatar
Linus Torvalds committed
183

Tejun Heo's avatar
Tejun Heo committed
184
185
186
187
188
189
190
191
192
193
194
#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 */
195

Miklos Szeredi's avatar
Miklos Szeredi committed
196
197
#if MAX_NUMNODES > 1
int nr_node_ids __read_mostly = MAX_NUMNODES;
198
int nr_online_nodes __read_mostly = 1;
Miklos Szeredi's avatar
Miklos Szeredi committed
199
EXPORT_SYMBOL(nr_node_ids);
200
EXPORT_SYMBOL(nr_online_nodes);
Miklos Szeredi's avatar
Miklos Szeredi committed
201
202
#endif

203
204
int page_group_by_mobility_disabled __read_mostly;

205
206
static void set_pageblock_migratetype(struct page *page, int migratetype)
{
207
208
209
210

	if (unlikely(page_group_by_mobility_disabled))
		migratetype = MIGRATE_UNMOVABLE;

211
212
213
214
	set_pageblock_flags_group(page, (unsigned long)migratetype,
					PB_migrate, PB_migrate_end);
}

215
216
bool oom_killer_disabled __read_mostly;

Nick Piggin's avatar
Nick Piggin committed
217
#ifdef CONFIG_DEBUG_VM
218
static int page_outside_zone_boundaries(struct zone *zone, struct page *page)
Linus Torvalds's avatar
Linus Torvalds committed
219
{
220
221
222
	int ret = 0;
	unsigned seq;
	unsigned long pfn = page_to_pfn(page);
223

224
225
226
227
228
229
230
231
232
	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;
233
234
235
236
}

static int page_is_consistent(struct zone *zone, struct page *page)
{
237
	if (!pfn_valid_within(page_to_pfn(page)))
238
		return 0;
Linus Torvalds's avatar
Linus Torvalds committed
239
	if (zone != page_zone(page))
240
241
242
243
244
245
246
247
248
249
		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
250
		return 1;
251
252
253
	if (!page_is_consistent(zone, page))
		return 1;

Linus Torvalds's avatar
Linus Torvalds committed
254
255
	return 0;
}
Nick Piggin's avatar
Nick Piggin committed
256
257
258
259
260
261
262
#else
static inline int bad_range(struct zone *zone, struct page *page)
{
	return 0;
}
#endif

263
static void bad_page(struct page *page)
Linus Torvalds's avatar
Linus Torvalds committed
264
{
265
266
267
268
	static unsigned long resume;
	static unsigned long nr_shown;
	static unsigned long nr_unshown;

269
270
	/* Don't complain about poisoned pages */
	if (PageHWPoison(page)) {
271
		reset_page_mapcount(page); /* remove PageBuddy */
272
273
274
		return;
	}

275
276
277
278
279
280
281
282
283
284
	/*
	 * 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) {
285
286
			printk(KERN_ALERT
			      "BUG: Bad page state: %lu messages suppressed\n",
287
288
289
290
291
292
293
294
				nr_unshown);
			nr_unshown = 0;
		}
		nr_shown = 0;
	}
	if (nr_shown++ == 0)
		resume = jiffies + 60 * HZ;

295
	printk(KERN_ALERT "BUG: Bad page state in process %s  pfn:%05lx\n",
296
		current->comm, page_to_pfn(page));
297
	dump_page(page);
298

299
	print_modules();
Linus Torvalds's avatar
Linus Torvalds committed
300
	dump_stack();
301
out:
302
	/* Leave bad fields for debug, except PageBuddy could make trouble */
303
	reset_page_mapcount(page); /* remove PageBuddy */
304
	add_taint(TAINT_BAD_PAGE);
Linus Torvalds's avatar
Linus Torvalds committed
305
306
307
308
309
310
311
312
313
}

/*
 * 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".
 *
314
315
 * 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
316
 *
317
318
319
 * 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
320
 */
321
322
323

static void free_compound_page(struct page *page)
{
324
	__free_pages_ok(page, compound_order(page));
325
326
}

327
void prep_compound_page(struct page *page, unsigned long order)
328
329
330
331
332
333
334
335
336
337
{
	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);
338
		set_page_count(p, 0);
339
340
341
342
		p->first_page = page;
	}
}

343
/* update __split_huge_page_refcount if you change this function */
344
static int destroy_compound_page(struct page *page, unsigned long order)
Linus Torvalds's avatar
Linus Torvalds committed
345
346
347
{
	int i;
	int nr_pages = 1 << order;
348
	int bad = 0;
Linus Torvalds's avatar
Linus Torvalds committed
349

350
351
	if (unlikely(compound_order(page) != order) ||
	    unlikely(!PageHead(page))) {
352
		bad_page(page);
353
354
		bad++;
	}
Linus Torvalds's avatar
Linus Torvalds committed
355

356
	__ClearPageHead(page);
357

358
359
	for (i = 1; i < nr_pages; i++) {
		struct page *p = page + i;
Linus Torvalds's avatar
Linus Torvalds committed
360

361
		if (unlikely(!PageTail(p) || (p->first_page != page))) {
362
			bad_page(page);
363
364
			bad++;
		}
365
		__ClearPageTail(p);
Linus Torvalds's avatar
Linus Torvalds committed
366
	}
367
368

	return bad;
Linus Torvalds's avatar
Linus Torvalds committed
369
370
}

Nick Piggin's avatar
Nick Piggin committed
371
372
373
374
static inline void prep_zero_page(struct page *page, int order, gfp_t gfp_flags)
{
	int i;

375
376
377
378
	/*
	 * 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
379
	VM_BUG_ON((gfp_flags & __GFP_HIGHMEM) && in_interrupt());
Nick Piggin's avatar
Nick Piggin committed
380
381
382
383
	for (i = 0; i < (1 << order); i++)
		clear_highpage(page + i);
}

384
385
static inline void set_page_order(struct page *page, int order)
{
386
	set_page_private(page, order);
387
	__SetPageBuddy(page);
Linus Torvalds's avatar
Linus Torvalds committed
388
389
390
391
}

static inline void rmv_page_order(struct page *page)
{
392
	__ClearPageBuddy(page);
393
	set_page_private(page, 0);
Linus Torvalds's avatar
Linus Torvalds committed
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
}

/*
 * 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
411
 * Assumption: *_mem_map is contiguous at least up to MAX_ORDER
Linus Torvalds's avatar
Linus Torvalds committed
412
413
 */
static inline unsigned long
414
__find_buddy_index(unsigned long page_idx, unsigned int order)
Linus Torvalds's avatar
Linus Torvalds committed
415
{
416
	return page_idx ^ (1 << order);
Linus Torvalds's avatar
Linus Torvalds committed
417
418
419
420
421
}

/*
 * 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
422
 * (a) the buddy is not in a hole &&
423
 * (b) the buddy is in the buddy system &&
424
425
 * (c) a page and its buddy have the same order &&
 * (d) a page and its buddy are in the same zone.
426
 *
Andrea Arcangeli's avatar
Andrea Arcangeli committed
427
428
 * 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
429
 *
430
 * For recording page's order, we use page_private(page).
Linus Torvalds's avatar
Linus Torvalds committed
431
 */
432
433
static inline int page_is_buddy(struct page *page, struct page *buddy,
								int order)
Linus Torvalds's avatar
Linus Torvalds committed
434
{
435
	if (!pfn_valid_within(page_to_pfn(buddy)))
Nick Piggin's avatar
Nick Piggin committed
436
437
		return 0;

438
439
440
441
	if (page_zone_id(page) != page_zone_id(buddy))
		return 0;

	if (PageBuddy(buddy) && page_order(buddy) == order) {
442
		VM_BUG_ON(page_count(buddy) != 0);
443
		return 1;
444
	}
445
	return 0;
Linus Torvalds's avatar
Linus Torvalds committed
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
}

/*
 * 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
461
 * free pages of length of (1 << order) and marked with _mapcount -2. Page's
462
 * order is recorded in page_private(page) field.
Linus Torvalds's avatar
Linus Torvalds committed
463
464
465
466
467
468
469
470
471
 * So when we are allocating or freeing one, we can derive the state of the
 * other.  That is, if we allocate a small block, and both were   
 * free, the remainder of the region must be split into blocks.   
 * If a block is freed, and its buddy is also free, then this
 * triggers coalescing into a block of larger size.            
 *
 * -- wli
 */

Nick Piggin's avatar
Nick Piggin committed
472
static inline void __free_one_page(struct page *page,
473
474
		struct zone *zone, unsigned int order,
		int migratetype)
Linus Torvalds's avatar
Linus Torvalds committed
475
476
{
	unsigned long page_idx;
477
	unsigned long combined_idx;
478
	unsigned long uninitialized_var(buddy_idx);
479
	struct page *buddy;
Linus Torvalds's avatar
Linus Torvalds committed
480

481
	if (unlikely(PageCompound(page)))
482
483
		if (unlikely(destroy_compound_page(page, order)))
			return;
Linus Torvalds's avatar
Linus Torvalds committed
484

485
486
	VM_BUG_ON(migratetype == -1);

Linus Torvalds's avatar
Linus Torvalds committed
487
488
	page_idx = page_to_pfn(page) & ((1 << MAX_ORDER) - 1);

489
	VM_BUG_ON(page_idx & ((1 << order) - 1));
Nick Piggin's avatar
Nick Piggin committed
490
	VM_BUG_ON(bad_range(zone, page));
Linus Torvalds's avatar
Linus Torvalds committed
491
492

	while (order < MAX_ORDER-1) {
493
494
		buddy_idx = __find_buddy_index(page_idx, order);
		buddy = page + (buddy_idx - page_idx);
495
		if (!page_is_buddy(page, buddy, order))
496
			break;
Nick Piggin's avatar
Nick Piggin committed
497

498
		/* Our buddy is free, merge with it and move up one order. */
Linus Torvalds's avatar
Linus Torvalds committed
499
		list_del(&buddy->lru);
500
		zone->free_area[order].nr_free--;
Linus Torvalds's avatar
Linus Torvalds committed
501
		rmv_page_order(buddy);
502
		combined_idx = buddy_idx & page_idx;
Linus Torvalds's avatar
Linus Torvalds committed
503
504
505
506
507
		page = page + (combined_idx - page_idx);
		page_idx = combined_idx;
		order++;
	}
	set_page_order(page, order);
508
509
510
511
512
513
514
515
516

	/*
	 * 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
	 */
517
	if ((order < MAX_ORDER-2) && pfn_valid_within(page_to_pfn(buddy))) {
518
		struct page *higher_page, *higher_buddy;
519
520
521
522
		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);
523
524
525
526
527
528
529
530
531
		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
532
533
534
	zone->free_area[order].nr_free++;
}

535
536
537
538
539
540
541
542
543
544
545
/*
 * 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);
}

546
static inline int free_pages_check(struct page *page)
Linus Torvalds's avatar
Linus Torvalds committed
547
{
Nick Piggin's avatar
Nick Piggin committed
548
549
	if (unlikely(page_mapcount(page) |
		(page->mapping != NULL)  |
550
		(atomic_read(&page->_count) != 0) |
551
552
		(page->flags & PAGE_FLAGS_CHECK_AT_FREE) |
		(mem_cgroup_bad_page_check(page)))) {
553
		bad_page(page);
554
		return 1;
555
	}
556
557
558
	if (page->flags & PAGE_FLAGS_CHECK_AT_PREP)
		page->flags &= ~PAGE_FLAGS_CHECK_AT_PREP;
	return 0;
Linus Torvalds's avatar
Linus Torvalds committed
559
560
561
}

/*
562
 * Frees a number of pages from the PCP lists
Linus Torvalds's avatar
Linus Torvalds committed
563
 * Assumes all pages on list are in same zone, and of same order.
564
 * count is the number of pages to free.
Linus Torvalds's avatar
Linus Torvalds committed
565
566
567
568
569
570
571
 *
 * 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.
 */
572
573
static void free_pcppages_bulk(struct zone *zone, int count,
					struct per_cpu_pages *pcp)
Linus Torvalds's avatar
Linus Torvalds committed
574
{
575
	int migratetype = 0;
576
	int batch_free = 0;
577
	int to_free = count;
578

Nick Piggin's avatar
Nick Piggin committed
579
	spin_lock(&zone->lock);
580
	zone->all_unreclaimable = 0;
Linus Torvalds's avatar
Linus Torvalds committed
581
	zone->pages_scanned = 0;
582

583
	while (to_free) {
Nick Piggin's avatar
Nick Piggin committed
584
		struct page *page;
585
586
587
		struct list_head *list;

		/*
588
589
590
591
592
		 * 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
593
594
		 */
		do {
595
			batch_free++;
596
597
598
599
			if (++migratetype == MIGRATE_PCPTYPES)
				migratetype = 0;
			list = &pcp->lists[migratetype];
		} while (list_empty(list));
Nick Piggin's avatar
Nick Piggin committed
600

601
602
603
604
		/* This is the only non-empty list. Free them all. */
		if (batch_free == MIGRATE_PCPTYPES)
			batch_free = to_free;

605
606
607
608
		do {
			page = list_entry(list->prev, struct page, lru);
			/* must delete as __free_one_page list manipulates */
			list_del(&page->lru);
609
610
611
			/* 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));
612
		} while (--to_free && --batch_free && !list_empty(list));
Linus Torvalds's avatar
Linus Torvalds committed
613
	}
614
	__mod_zone_page_state(zone, NR_FREE_PAGES, count);
Nick Piggin's avatar
Nick Piggin committed
615
	spin_unlock(&zone->lock);
Linus Torvalds's avatar
Linus Torvalds committed
616
617
}

618
619
static void free_one_page(struct zone *zone, struct page *page, int order,
				int migratetype)
Linus Torvalds's avatar
Linus Torvalds committed
620
{
621
	spin_lock(&zone->lock);
622
	zone->all_unreclaimable = 0;
623
	zone->pages_scanned = 0;
624

625
	__free_one_page(page, zone, order, migratetype);
626
	__mod_zone_page_state(zone, NR_FREE_PAGES, 1 << order);
627
	spin_unlock(&zone->lock);
Nick Piggin's avatar
Nick Piggin committed
628
629
}

630
static bool free_pages_prepare(struct page *page, unsigned int order)
Nick Piggin's avatar
Nick Piggin committed
631
{
Linus Torvalds's avatar
Linus Torvalds committed
632
	int i;
633
	int bad = 0;
Linus Torvalds's avatar
Linus Torvalds committed
634

635
	trace_mm_page_free_direct(page, order);
636
637
	kmemcheck_free_shadow(page, order);

Andrea Arcangeli's avatar
Andrea Arcangeli committed
638
639
640
641
	if (PageAnon(page))
		page->mapping = NULL;
	for (i = 0; i < (1 << order); i++)
		bad += free_pages_check(page + i);
642
	if (bad)
643
		return false;
644

645
	if (!PageHighMem(page)) {
Nick Piggin's avatar
Nick Piggin committed
646
		debug_check_no_locks_freed(page_address(page),PAGE_SIZE<<order);
647
648
649
		debug_check_no_obj_freed(page_address(page),
					   PAGE_SIZE << order);
	}
Nick Piggin's avatar
Nick Piggin committed
650
	arch_free_page(page, order);
Nick Piggin's avatar
Nick Piggin committed
651
	kernel_map_pages(page, 1 << order, 0);
Nick Piggin's avatar
Nick Piggin committed
652

653
654
655
656
657
658
659
660
661
662
663
	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
664
	local_irq_save(flags);
665
	if (unlikely(wasMlocked))
666
		free_page_mlock(page);
667
	__count_vm_events(PGFREE, 1 << order);
668
669
	free_one_page(page_zone(page), page, order,
					get_pageblock_migratetype(page));
Nick Piggin's avatar
Nick Piggin committed
670
	local_irq_restore(flags);
Linus Torvalds's avatar
Linus Torvalds committed
671
672
}

673
674
675
/*
 * permit the bootmem allocator to evade page validation on high-order frees
 */
676
void __meminit __free_pages_bootmem(struct page *page, unsigned int order)
677
678
679
680
{
	if (order == 0) {
		__ClearPageReserved(page);
		set_page_count(page, 0);
681
		set_page_refcounted(page);
Nick Piggin's avatar
Nick Piggin committed
682
		__free_page(page);
683
684
685
	} else {
		int loop;

Nick Piggin's avatar
Nick Piggin committed
686
		prefetchw(page);
687
		for (loop = 0; loop < (1 << order); loop++) {
688
689
			struct page *p = &page[loop];

690
			if (loop + 1 < (1 << order))
Nick Piggin's avatar
Nick Piggin committed
691
				prefetchw(p + 1);
692
693
694
695
			__ClearPageReserved(p);
			set_page_count(p, 0);
		}

696
		set_page_refcounted(page);
Nick Piggin's avatar
Nick Piggin committed
697
		__free_pages(page, order);
698
699
700
	}
}

Linus Torvalds's avatar
Linus Torvalds committed
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715

/*
 * 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
716
static inline void expand(struct zone *zone, struct page *page,
717
718
	int low, int high, struct free_area *area,
	int migratetype)
Linus Torvalds's avatar
Linus Torvalds committed
719
720
721
722
723
724
725
{
	unsigned long size = 1 << high;

	while (high > low) {
		area--;
		high--;
		size >>= 1;
Nick Piggin's avatar
Nick Piggin committed
726
		VM_BUG_ON(bad_range(zone, &page[size]));
727
		list_add(&page[size].lru, &area->free_list[migratetype]);
Linus Torvalds's avatar
Linus Torvalds committed
728
729
730
731
732
733
734
735
		area->nr_free++;
		set_page_order(&page[size], high);
	}
}

/*
 * This page is about to be returned from the page allocator
 */
736
static inline int check_new_page(struct page *page)
Linus Torvalds's avatar
Linus Torvalds committed
737
{
Nick Piggin's avatar
Nick Piggin committed
738
739
	if (unlikely(page_mapcount(page) |
		(page->mapping != NULL)  |
740
		(atomic_read(&page->_count) != 0)  |
741
742
		(page->flags & PAGE_FLAGS_CHECK_AT_PREP) |
		(mem_cgroup_bad_page_check(page)))) {
743
		bad_page(page);
744
		return 1;
745
	}
746
747
748
749
750
751
752
753
754
755
756
757
	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;
	}
758

759
	set_page_private(page, 0);
760
	set_page_refcounted(page);
Nick Piggin's avatar
Nick Piggin committed
761
762

	arch_alloc_page(page, order);
Linus Torvalds's avatar
Linus Torvalds committed
763
	kernel_map_pages(page, 1 << order, 1);
Nick Piggin's avatar
Nick Piggin committed
764
765
766
767
768
769
770

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

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

771
	return 0;
Linus Torvalds's avatar
Linus Torvalds committed
772
773
}

774
775
776
777
/*
 * Go through the free lists for the given migratetype and remove
 * the smallest available page from the freelists
 */
778
779
static inline
struct page *__rmqueue_smallest(struct zone *zone, unsigned int order,
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
						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;
}


805
806
807
808
809
/*
 * This array describes the order lists are fallen back to when
 * the free lists for the desirable migrate type are depleted
 */
static int fallbacks[MIGRATE_TYPES][MIGRATE_TYPES-1] = {
810
811
812
813
	[MIGRATE_UNMOVABLE]   = { MIGRATE_RECLAIMABLE, MIGRATE_MOVABLE,   MIGRATE_RESERVE },
	[MIGRATE_RECLAIMABLE] = { MIGRATE_UNMOVABLE,   MIGRATE_MOVABLE,   MIGRATE_RESERVE },
	[MIGRATE_MOVABLE]     = { MIGRATE_RECLAIMABLE, MIGRATE_UNMOVABLE, MIGRATE_RESERVE },
	[MIGRATE_RESERVE]     = { MIGRATE_RESERVE,     MIGRATE_RESERVE,   MIGRATE_RESERVE }, /* Never used */
814
815
};

816
817
/*
 * Move the free pages in a range to the free lists of the requested type.
818
 * Note that start_page and end_pages are not aligned on a pageblock
819
820
 * boundary. If alignment is required, use move_freepages_block()
 */
Adrian Bunk's avatar
Adrian Bunk committed
821
822
823
static int move_freepages(struct zone *zone,
			  struct page *start_page, struct page *end_page,
			  int migratetype)
824
825
826
{
	struct page *page;
	unsigned long order;
827
	int pages_moved = 0;
828
829
830
831
832
833
834

#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
835
	 * grouping pages by mobility
836
837
838
839
840
	 */
	BUG_ON(page_zone(start_page) != page_zone(end_page));
#endif

	for (page = start_page; page <= end_page;) {
841
842
843
		/* Make sure we are not inadvertently changing nodes */
		VM_BUG_ON(page_to_nid(page) != zone_to_nid(zone));

844
845
846
847
848
849
850
851
852
853
854
		if (!pfn_valid_within(page_to_pfn(page))) {
			page++;
			continue;
		}

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

		order = page_order(page);
855
856
		list_move(&page->lru,
			  &zone->free_area[order].free_list[migratetype]);
857
		page += 1 << order;
858
		pages_moved += 1 << order;
859
860
	}

861
	return pages_moved;
862
863
}

Adrian Bunk's avatar
Adrian Bunk committed
864
865
static int move_freepages_block(struct zone *zone, struct page *page,
				int migratetype)
866
867
868
869
870
{
	unsigned long start_pfn, end_pfn;
	struct page *start_page, *end_page;

	start_pfn = page_to_pfn(page);
871
	start_pfn = start_pfn & ~(pageblock_nr_pages-1);
872
	start_page = pfn_to_page(start_pfn);
873
874
	end_page = start_page + pageblock_nr_pages - 1;
	end_pfn = start_pfn + pageblock_nr_pages - 1;
875
876
877
878
879
880
881
882
883
884

	/* 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);
}

885
886
887
888
889
890
891
892
893
894
895
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;
	}
}

896
/* Remove an element from the buddy allocator from the fallback list */
897
898
static inline struct page *
__rmqueue_fallback(struct zone *zone, int order, int start_migratetype)
899
900
901
902
903
904
905
906
907
908
909
910
{
	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) {
		for (i = 0; i < MIGRATE_TYPES - 1; i++) {
			migratetype = fallbacks[start_migratetype][i];

911
912
913
			/* MIGRATE_RESERVE handled later if necessary */
			if (migratetype == MIGRATE_RESERVE)
				continue;
914

915
916
917
918
919
920
921
922
923
			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--;

			/*
924
			 * If breaking a large block of pages, move all free
925
926
			 * 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
927
			 * aggressive about taking ownership of free pages
928
			 */
929
			if (unlikely(current_order >= (pageblock_order >> 1)) ||
930
931
					start_migratetype == MIGRATE_RECLAIMABLE ||
					page_group_by_mobility_disabled) {
932
933
934
935
936
				unsigned long pages;
				pages = move_freepages_block(zone, page,
								start_migratetype);

				/* Claim the whole block if over half of it is free */
937
938
				if (pages >= (1 << (pageblock_order-1)) ||
						page_group_by_mobility_disabled)
939
940
941
					set_pageblock_migratetype(page,
								start_migratetype);

942
				migratetype = start_migratetype;
943
			}
944
945
946
947
948

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

949
950
951
			/* Take ownership for orders >= pageblock_order */
			if (current_order >= pageblock_order)
				change_pageblock_range(page, current_order,
952
953
954
							start_migratetype);

			expand(zone, page, order, current_order, area, migratetype);
955
956
957
958

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

959
960
961
962
			return page;
		}
	}

963
	return NULL;
964
965
}

966
/*
Linus Torvalds's avatar
Linus Torvalds committed
967
968
969
 * Do the hard work of removing an element from the buddy allocator.
 * Call me with the zone->lock already held.
 */
970
971
static struct page *__rmqueue(struct zone *zone, unsigned int order,
						int migratetype)
Linus Torvalds's avatar
Linus Torvalds committed
972
973
974
{
	struct page *page;

975
retry_reserve:
976
	page = __rmqueue_smallest(zone, order, migratetype);
977

978
	if (unlikely(!page) && migratetype != MIGRATE_RESERVE) {
979
		page = __rmqueue_fallback(zone, order, migratetype);
980

981
982
983
984
985
986
987
988
989
990
991
		/*
		 * Use MIGRATE_RESERVE rather than fail an allocation. goto
		 * is used because __rmqueue_smallest is an inline function
		 * and we want just one call site
		 */
		if (!page) {
			migratetype = MIGRATE_RESERVE;
			goto retry_reserve;
		}
	}

992
	trace_mm_page_alloc_zone_locked(page, order, migratetype);
993
	return page;
Linus Torvalds's avatar
Linus Torvalds committed
994
995
996
997
998
999
1000
1001
}

/* 
 * Obtain a specified number of elements from the buddy allocator, all under
 * a single hold of the lock, for efficiency.  Add them to the supplied list.
 * Returns the number of new pages which were placed at *list.
 */
static int rmqueue_bulk(struct zone *zone, unsigned int order, 
1002
			unsigned long count, struct list_head *list,
1003
			int migratetype, int cold)
Linus Torvalds's avatar
Linus Torvalds committed
1004
1005
1006
{
	int i;
	
Nick Piggin's avatar
Nick Piggin committed
1007
	spin_lock(&zone->lock);
Linus Torvalds's avatar
Linus Torvalds committed
1008
	for (i = 0; i < count; ++i) {
1009
		struct page *page = __rmqueue(zone, order, migratetype);
Nick Piggin's avatar
Nick Piggin committed
1010
		if (unlikely(page == NULL))
Linus Torvalds's avatar
Linus Torvalds committed
1011
			break;
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021

		/*
		 * Split buddy pages returned by expand() are received here
		 * in physical page order. The page is added to the callers and
		 * list and the list head then moves forward. From the callers
		 * perspective, the linked list is ordered by page number in
		 * some conditions. This is useful for IO devices that can
		 * merge IO requests if the physical pages are ordered
		 * properly.
		 */
1022
1023
1024
1025
		if (likely(cold == 0))
			list_add(&page->lru, list);
		else
			list_add_tail(&page->lru, list);