page_alloc.c 174 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/compaction.h>
55
#include <trace/events/kmem.h>
56
#include <linux/ftrace_event.h>
57
#include <linux/memcontrol.h>
58
#include <linux/prefetch.h>
59
#include <linux/migrate.h>
60
#include <linux/page-debug-flags.h>
61
#include <linux/hugetlb.h>
62
#include <linux/sched/rt.h>
Linus Torvalds's avatar
Linus Torvalds committed
63
64

#include <asm/tlbflush.h>
65
#include <asm/div64.h>
Linus Torvalds's avatar
Linus Torvalds committed
66
67
#include "internal.h"

68
69
70
/* prevent >1 _updater_ of zone percpu pageset ->high and ->batch fields */
static DEFINE_MUTEX(pcp_batch_high_lock);

71
72
73
74
75
#ifdef CONFIG_USE_PERCPU_NUMA_NODE_ID
DEFINE_PER_CPU(int, numa_node);
EXPORT_PER_CPU_SYMBOL(numa_node);
#endif

76
77
78
79
80
81
82
83
84
85
86
#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
87
/*
88
 * Array of node states.
Linus Torvalds's avatar
Linus Torvalds committed
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 } },
97
98
99
#endif
#ifdef CONFIG_MOVABLE_NODE
	[N_MEMORY] = { { [0] = 1UL } },
100
101
102
103
104
105
#endif
	[N_CPU] = { { [0] = 1UL } },
#endif	/* NUMA */
};
EXPORT_SYMBOL(node_states);

106
107
108
/* Protect totalram_pages and zone->managed_pages */
static DEFINE_SPINLOCK(managed_page_count_lock);

109
unsigned long totalram_pages __read_mostly;
110
unsigned long totalreserve_pages __read_mostly;
111
112
113
114
115
116
117
118
/*
 * 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;

119
int percpu_pagelist_fraction;
120
gfp_t gfp_allowed_mask __read_mostly = GFP_BOOT_MASK;
Linus Torvalds's avatar
Linus Torvalds committed
121

122
123
124
125
126
127
128
129
130
#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).
 */
131
132
133
134

static gfp_t saved_gfp_mask;

void pm_restore_gfp_mask(void)
135
136
{
	WARN_ON(!mutex_is_locked(&pm_mutex));
137
138
139
140
	if (saved_gfp_mask) {
		gfp_allowed_mask = saved_gfp_mask;
		saved_gfp_mask = 0;
	}
141
142
}

143
void pm_restrict_gfp_mask(void)
144
145
{
	WARN_ON(!mutex_is_locked(&pm_mutex));
146
147
148
	WARN_ON(saved_gfp_mask);
	saved_gfp_mask = gfp_allowed_mask;
	gfp_allowed_mask &= ~GFP_IOFS;
149
}
150
151
152
153
154
155
156

bool pm_suspended_storage(void)
{
	if ((gfp_allowed_mask & GFP_IOFS) == GFP_IOFS)
		return false;
	return true;
}
157
158
#endif /* CONFIG_PM_SLEEP */

159
160
161
162
#ifdef CONFIG_HUGETLB_PAGE_SIZE_VARIABLE
int pageblock_order __read_mostly;
#endif

163
static void __free_pages_ok(struct page *page, unsigned int order);
164

Linus Torvalds's avatar
Linus Torvalds committed
165
166
167
168
169
170
171
/*
 * 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
172
173
174
 *
 * 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
175
 */
176
int sysctl_lowmem_reserve_ratio[MAX_NR_ZONES-1] = {
177
#ifdef CONFIG_ZONE_DMA
178
	 256,
179
#endif
180
#ifdef CONFIG_ZONE_DMA32
181
	 256,
182
#endif
183
#ifdef CONFIG_HIGHMEM
Mel Gorman's avatar
Mel Gorman committed
184
	 32,
185
#endif
Mel Gorman's avatar
Mel Gorman committed
186
	 32,
187
};
Linus Torvalds's avatar
Linus Torvalds committed
188
189
190

EXPORT_SYMBOL(totalram_pages);

191
static char * const zone_names[MAX_NR_ZONES] = {
192
#ifdef CONFIG_ZONE_DMA
193
	 "DMA",
194
#endif
195
#ifdef CONFIG_ZONE_DMA32
196
	 "DMA32",
197
#endif
198
	 "Normal",
199
#ifdef CONFIG_HIGHMEM
Mel Gorman's avatar
Mel Gorman committed
200
	 "HighMem",
201
#endif
Mel Gorman's avatar
Mel Gorman committed
202
	 "Movable",
203
204
};

Linus Torvalds's avatar
Linus Torvalds committed
205
206
int min_free_kbytes = 1024;

207
208
static unsigned long __meminitdata nr_kernel_pages;
static unsigned long __meminitdata nr_all_pages;
209
static unsigned long __meminitdata dma_reserve;
Linus Torvalds's avatar
Linus Torvalds committed
210

Tejun Heo's avatar
Tejun Heo committed
211
212
213
214
215
216
217
218
219
220
221
#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 */
222

Miklos Szeredi's avatar
Miklos Szeredi committed
223
224
#if MAX_NUMNODES > 1
int nr_node_ids __read_mostly = MAX_NUMNODES;
225
int nr_online_nodes __read_mostly = 1;
Miklos Szeredi's avatar
Miklos Szeredi committed
226
EXPORT_SYMBOL(nr_node_ids);
227
EXPORT_SYMBOL(nr_online_nodes);
Miklos Szeredi's avatar
Miklos Szeredi committed
228
229
#endif

230
231
int page_group_by_mobility_disabled __read_mostly;

232
void set_pageblock_migratetype(struct page *page, int migratetype)
233
{
234
235
236
237

	if (unlikely(page_group_by_mobility_disabled))
		migratetype = MIGRATE_UNMOVABLE;

238
239
240
241
	set_pageblock_flags_group(page, (unsigned long)migratetype,
					PB_migrate, PB_migrate_end);
}

242
243
bool oom_killer_disabled __read_mostly;

Nick Piggin's avatar
Nick Piggin committed
244
#ifdef CONFIG_DEBUG_VM
245
static int page_outside_zone_boundaries(struct zone *zone, struct page *page)
Linus Torvalds's avatar
Linus Torvalds committed
246
{
247
248
249
	int ret = 0;
	unsigned seq;
	unsigned long pfn = page_to_pfn(page);
250
	unsigned long sp, start_pfn;
251

252
253
	do {
		seq = zone_span_seqbegin(zone);
254
255
		start_pfn = zone->zone_start_pfn;
		sp = zone->spanned_pages;
256
		if (!zone_spans_pfn(zone, pfn))
257
258
259
			ret = 1;
	} while (zone_span_seqretry(zone, seq));

260
261
262
263
	if (ret)
		pr_err("page %lu outside zone [ %lu - %lu ]\n",
			pfn, start_pfn, start_pfn + sp);

264
	return ret;
265
266
267
268
}

static int page_is_consistent(struct zone *zone, struct page *page)
{
269
	if (!pfn_valid_within(page_to_pfn(page)))
270
		return 0;
Linus Torvalds's avatar
Linus Torvalds committed
271
	if (zone != page_zone(page))
272
273
274
275
276
277
278
279
280
281
		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
282
		return 1;
283
284
285
	if (!page_is_consistent(zone, page))
		return 1;

Linus Torvalds's avatar
Linus Torvalds committed
286
287
	return 0;
}
Nick Piggin's avatar
Nick Piggin committed
288
289
290
291
292
293
294
#else
static inline int bad_range(struct zone *zone, struct page *page)
{
	return 0;
}
#endif

295
static void bad_page(struct page *page)
Linus Torvalds's avatar
Linus Torvalds committed
296
{
297
298
299
300
	static unsigned long resume;
	static unsigned long nr_shown;
	static unsigned long nr_unshown;

301
302
	/* Don't complain about poisoned pages */
	if (PageHWPoison(page)) {
303
		page_mapcount_reset(page); /* remove PageBuddy */
304
305
306
		return;
	}

307
308
309
310
311
312
313
314
315
316
	/*
	 * 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) {
317
318
			printk(KERN_ALERT
			      "BUG: Bad page state: %lu messages suppressed\n",
319
320
321
322
323
324
325
326
				nr_unshown);
			nr_unshown = 0;
		}
		nr_shown = 0;
	}
	if (nr_shown++ == 0)
		resume = jiffies + 60 * HZ;

327
	printk(KERN_ALERT "BUG: Bad page state in process %s  pfn:%05lx\n",
328
		current->comm, page_to_pfn(page));
329
	dump_page(page);
330

331
	print_modules();
Linus Torvalds's avatar
Linus Torvalds committed
332
	dump_stack();
333
out:
334
	/* Leave bad fields for debug, except PageBuddy could make trouble */
335
	page_mapcount_reset(page); /* remove PageBuddy */
336
	add_taint(TAINT_BAD_PAGE, LOCKDEP_NOW_UNRELIABLE);
Linus Torvalds's avatar
Linus Torvalds committed
337
338
339
340
341
342
343
344
345
}

/*
 * 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".
 *
346
347
 * 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
348
 *
349
350
351
 * 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
352
 */
353
354
355

static void free_compound_page(struct page *page)
{
356
	__free_pages_ok(page, compound_order(page));
357
358
}

359
void prep_compound_page(struct page *page, unsigned long order)
360
361
362
363
364
365
366
367
368
369
{
	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);
370
		set_page_count(p, 0);
371
372
373
374
		p->first_page = page;
	}
}

375
/* update __split_huge_page_refcount if you change this function */
376
static int destroy_compound_page(struct page *page, unsigned long order)
Linus Torvalds's avatar
Linus Torvalds committed
377
378
379
{
	int i;
	int nr_pages = 1 << order;
380
	int bad = 0;
Linus Torvalds's avatar
Linus Torvalds committed
381

382
	if (unlikely(compound_order(page) != order)) {
383
		bad_page(page);
384
385
		bad++;
	}
Linus Torvalds's avatar
Linus Torvalds committed
386

387
	__ClearPageHead(page);
388

389
390
	for (i = 1; i < nr_pages; i++) {
		struct page *p = page + i;
Linus Torvalds's avatar
Linus Torvalds committed
391

392
		if (unlikely(!PageTail(p) || (p->first_page != page))) {
393
			bad_page(page);
394
395
			bad++;
		}
396
		__ClearPageTail(p);
Linus Torvalds's avatar
Linus Torvalds committed
397
	}
398
399

	return bad;
Linus Torvalds's avatar
Linus Torvalds committed
400
401
}

Nick Piggin's avatar
Nick Piggin committed
402
403
404
405
static inline void prep_zero_page(struct page *page, int order, gfp_t gfp_flags)
{
	int i;

406
407
408
409
	/*
	 * 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
410
	VM_BUG_ON((gfp_flags & __GFP_HIGHMEM) && in_interrupt());
Nick Piggin's avatar
Nick Piggin committed
411
412
413
414
	for (i = 0; i < (1 << order); i++)
		clear_highpage(page + i);
}

415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
#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

446
447
static inline void set_page_order(struct page *page, int order)
{
448
	set_page_private(page, order);
449
	__SetPageBuddy(page);
Linus Torvalds's avatar
Linus Torvalds committed
450
451
452
453
}

static inline void rmv_page_order(struct page *page)
{
454
	__ClearPageBuddy(page);
455
	set_page_private(page, 0);
Linus Torvalds's avatar
Linus Torvalds committed
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
}

/*
 * 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
473
 * Assumption: *_mem_map is contiguous at least up to MAX_ORDER
Linus Torvalds's avatar
Linus Torvalds committed
474
475
 */
static inline unsigned long
476
__find_buddy_index(unsigned long page_idx, unsigned int order)
Linus Torvalds's avatar
Linus Torvalds committed
477
{
478
	return page_idx ^ (1 << order);
Linus Torvalds's avatar
Linus Torvalds committed
479
480
481
482
483
}

/*
 * 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
484
 * (a) the buddy is not in a hole &&
485
 * (b) the buddy is in the buddy system &&
486
487
 * (c) a page and its buddy have the same order &&
 * (d) a page and its buddy are in the same zone.
488
 *
Andrea Arcangeli's avatar
Andrea Arcangeli committed
489
490
 * 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
491
 *
492
 * For recording page's order, we use page_private(page).
Linus Torvalds's avatar
Linus Torvalds committed
493
 */
494
495
static inline int page_is_buddy(struct page *page, struct page *buddy,
								int order)
Linus Torvalds's avatar
Linus Torvalds committed
496
{
497
	if (!pfn_valid_within(page_to_pfn(buddy)))
Nick Piggin's avatar
Nick Piggin committed
498
499
		return 0;

500
501
502
	if (page_zone_id(page) != page_zone_id(buddy))
		return 0;

503
504
505
506
507
	if (page_is_guard(buddy) && page_order(buddy) == order) {
		VM_BUG_ON(page_count(buddy) != 0);
		return 1;
	}

508
	if (PageBuddy(buddy) && page_order(buddy) == order) {
509
		VM_BUG_ON(page_count(buddy) != 0);
510
		return 1;
511
	}
512
	return 0;
Linus Torvalds's avatar
Linus Torvalds committed
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
}

/*
 * 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
528
 * free pages of length of (1 << order) and marked with _mapcount -2. Page's
529
 * order is recorded in page_private(page) field.
Linus Torvalds's avatar
Linus Torvalds committed
530
 * So when we are allocating or freeing one, we can derive the state of the
531
532
 * 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
533
 * If a block is freed, and its buddy is also free, then this
534
 * triggers coalescing into a block of larger size.
Linus Torvalds's avatar
Linus Torvalds committed
535
 *
536
 * -- nyc
Linus Torvalds's avatar
Linus Torvalds committed
537
538
 */

Nick Piggin's avatar
Nick Piggin committed
539
static inline void __free_one_page(struct page *page,
540
541
		struct zone *zone, unsigned int order,
		int migratetype)
Linus Torvalds's avatar
Linus Torvalds committed
542
543
{
	unsigned long page_idx;
544
	unsigned long combined_idx;
545
	unsigned long uninitialized_var(buddy_idx);
546
	struct page *buddy;
Linus Torvalds's avatar
Linus Torvalds committed
547

548
549
	VM_BUG_ON(!zone_is_initialized(zone));

550
	if (unlikely(PageCompound(page)))
551
552
		if (unlikely(destroy_compound_page(page, order)))
			return;
Linus Torvalds's avatar
Linus Torvalds committed
553

554
555
	VM_BUG_ON(migratetype == -1);

Linus Torvalds's avatar
Linus Torvalds committed
556
557
	page_idx = page_to_pfn(page) & ((1 << MAX_ORDER) - 1);

558
	VM_BUG_ON(page_idx & ((1 << order) - 1));
Nick Piggin's avatar
Nick Piggin committed
559
	VM_BUG_ON(bad_range(zone, page));
Linus Torvalds's avatar
Linus Torvalds committed
560
561

	while (order < MAX_ORDER-1) {
562
563
		buddy_idx = __find_buddy_index(page_idx, order);
		buddy = page + (buddy_idx - page_idx);
564
		if (!page_is_buddy(page, buddy, order))
565
			break;
566
567
568
569
570
571
572
		/*
		 * 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);
573
574
			__mod_zone_freepage_state(zone, 1 << order,
						  migratetype);
575
576
577
578
579
		} else {
			list_del(&buddy->lru);
			zone->free_area[order].nr_free--;
			rmv_page_order(buddy);
		}
580
		combined_idx = buddy_idx & page_idx;
Linus Torvalds's avatar
Linus Torvalds committed
581
582
583
584
585
		page = page + (combined_idx - page_idx);
		page_idx = combined_idx;
		order++;
	}
	set_page_order(page, order);
586
587
588
589
590
591
592
593
594

	/*
	 * 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
	 */
595
	if ((order < MAX_ORDER-2) && pfn_valid_within(page_to_pfn(buddy))) {
596
		struct page *higher_page, *higher_buddy;
597
598
599
		combined_idx = buddy_idx & page_idx;
		higher_page = page + (combined_idx - page_idx);
		buddy_idx = __find_buddy_index(combined_idx, order + 1);
600
		higher_buddy = higher_page + (buddy_idx - combined_idx);
601
602
603
604
605
606
607
608
609
		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
610
611
612
	zone->free_area[order].nr_free++;
}

613
static inline int free_pages_check(struct page *page)
Linus Torvalds's avatar
Linus Torvalds committed
614
{
Nick Piggin's avatar
Nick Piggin committed
615
616
	if (unlikely(page_mapcount(page) |
		(page->mapping != NULL)  |
617
		(atomic_read(&page->_count) != 0) |
618
619
		(page->flags & PAGE_FLAGS_CHECK_AT_FREE) |
		(mem_cgroup_bad_page_check(page)))) {
620
		bad_page(page);
621
		return 1;
622
	}
623
	page_nid_reset_last(page);
624
625
626
	if (page->flags & PAGE_FLAGS_CHECK_AT_PREP)
		page->flags &= ~PAGE_FLAGS_CHECK_AT_PREP;
	return 0;
Linus Torvalds's avatar
Linus Torvalds committed
627
628
629
}

/*
630
 * Frees a number of pages from the PCP lists
Linus Torvalds's avatar
Linus Torvalds committed
631
 * Assumes all pages on list are in same zone, and of same order.
632
 * count is the number of pages to free.
Linus Torvalds's avatar
Linus Torvalds committed
633
634
635
636
637
638
639
 *
 * 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.
 */
640
641
static void free_pcppages_bulk(struct zone *zone, int count,
					struct per_cpu_pages *pcp)
Linus Torvalds's avatar
Linus Torvalds committed
642
{
643
	int migratetype = 0;
644
	int batch_free = 0;
645
	int to_free = count;
646

Nick Piggin's avatar
Nick Piggin committed
647
	spin_lock(&zone->lock);
648
	zone->all_unreclaimable = 0;
Linus Torvalds's avatar
Linus Torvalds committed
649
	zone->pages_scanned = 0;
650

651
	while (to_free) {
Nick Piggin's avatar
Nick Piggin committed
652
		struct page *page;
653
654
655
		struct list_head *list;

		/*
656
657
658
659
660
		 * 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
661
662
		 */
		do {
663
			batch_free++;
664
665
666
667
			if (++migratetype == MIGRATE_PCPTYPES)
				migratetype = 0;
			list = &pcp->lists[migratetype];
		} while (list_empty(list));
Nick Piggin's avatar
Nick Piggin committed
668

669
670
671
672
		/* This is the only non-empty list. Free them all. */
		if (batch_free == MIGRATE_PCPTYPES)
			batch_free = to_free;

673
		do {
674
675
			int mt;	/* migratetype of the to-be-freed page */

676
677
678
			page = list_entry(list->prev, struct page, lru);
			/* must delete as __free_one_page list manipulates */
			list_del(&page->lru);
679
			mt = get_freepage_migratetype(page);
680
			/* MIGRATE_MOVABLE list may include MIGRATE_RESERVEs */
681
682
			__free_one_page(page, zone, 0, mt);
			trace_mm_page_pcpu_drain(page, 0, mt);
683
			if (likely(!is_migrate_isolate_page(page))) {
684
685
686
687
				__mod_zone_page_state(zone, NR_FREE_PAGES, 1);
				if (is_migrate_cma(mt))
					__mod_zone_page_state(zone, NR_FREE_CMA_PAGES, 1);
			}
688
		} while (--to_free && --batch_free && !list_empty(list));
Linus Torvalds's avatar
Linus Torvalds committed
689
	}
Nick Piggin's avatar
Nick Piggin committed
690
	spin_unlock(&zone->lock);
Linus Torvalds's avatar
Linus Torvalds committed
691
692
}

693
694
static void free_one_page(struct zone *zone, struct page *page, int order,
				int migratetype)
Linus Torvalds's avatar
Linus Torvalds committed
695
{
696
	spin_lock(&zone->lock);
697
	zone->all_unreclaimable = 0;
698
	zone->pages_scanned = 0;
699

700
	__free_one_page(page, zone, order, migratetype);
701
	if (unlikely(!is_migrate_isolate(migratetype)))
702
		__mod_zone_freepage_state(zone, 1 << order, migratetype);
703
	spin_unlock(&zone->lock);
Nick Piggin's avatar
Nick Piggin committed
704
705
}

706
static bool free_pages_prepare(struct page *page, unsigned int order)
Nick Piggin's avatar
Nick Piggin committed
707
{
Linus Torvalds's avatar
Linus Torvalds committed
708
	int i;
709
	int bad = 0;
Linus Torvalds's avatar
Linus Torvalds committed
710

711
	trace_mm_page_free(page, order);
712
713
	kmemcheck_free_shadow(page, order);

Andrea Arcangeli's avatar
Andrea Arcangeli committed
714
715
716
717
	if (PageAnon(page))
		page->mapping = NULL;
	for (i = 0; i < (1 << order); i++)
		bad += free_pages_check(page + i);
718
	if (bad)
719
		return false;
720

721
	if (!PageHighMem(page)) {
Nick Piggin's avatar
Nick Piggin committed
722
		debug_check_no_locks_freed(page_address(page),PAGE_SIZE<<order);
723
724
725
		debug_check_no_obj_freed(page_address(page),
					   PAGE_SIZE << order);
	}
Nick Piggin's avatar
Nick Piggin committed
726
	arch_free_page(page, order);
Nick Piggin's avatar
Nick Piggin committed
727
	kernel_map_pages(page, 1 << order, 0);
Nick Piggin's avatar
Nick Piggin committed
728

729
730
731
732
733
734
	return true;
}

static void __free_pages_ok(struct page *page, unsigned int order)
{
	unsigned long flags;
735
	int migratetype;
736
737
738
739

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

Nick Piggin's avatar
Nick Piggin committed
740
	local_irq_save(flags);
741
	__count_vm_events(PGFREE, 1 << order);
742
743
744
	migratetype = get_pageblock_migratetype(page);
	set_freepage_migratetype(page, migratetype);
	free_one_page(page_zone(page), page, order, migratetype);
Nick Piggin's avatar
Nick Piggin committed
745
	local_irq_restore(flags);
Linus Torvalds's avatar
Linus Torvalds committed
746
747
}

748
749
750
751
752
753
754
/*
 * Read access to zone->managed_pages is safe because it's unsigned long,
 * but we still need to serialize writers. Currently all callers of
 * __free_pages_bootmem() except put_page_bootmem() should only be used
 * at boot time. So for shorter boot time, we shift the burden to
 * put_page_bootmem() to serialize writers.
 */
755
void __meminit __free_pages_bootmem(struct page *page, unsigned int order)
756
{
757
758
	unsigned int nr_pages = 1 << order;
	unsigned int loop;
759

760
761
762
763
764
765
766
767
	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);
768
	}
769

770
	page_zone(page)->managed_pages += 1 << order;
771
772
	set_page_refcounted(page);
	__free_pages(page, order);
773
774
}

775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
#ifdef CONFIG_CMA
/* Free whole pageblock and set it's migration type to MIGRATE_CMA. */
void __init init_cma_reserved_pageblock(struct page *page)
{
	unsigned i = pageblock_nr_pages;
	struct page *p = page;

	do {
		__ClearPageReserved(p);
		set_page_count(p, 0);
	} while (++p, --i);

	set_page_refcounted(page);
	set_pageblock_migratetype(page, MIGRATE_CMA);
	__free_pages(page, pageblock_order);
	totalram_pages += pageblock_nr_pages;
791
792
793
794
#ifdef CONFIG_HIGHMEM
	if (PageHighMem(page))
		totalhigh_pages += pageblock_nr_pages;
#endif
795
796
}
#endif
Linus Torvalds's avatar
Linus Torvalds committed
797
798
799
800
801
802
803
804
805
806
807
808
809

/*
 * 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.
 *
810
 * -- nyc
Linus Torvalds's avatar
Linus Torvalds committed
811
 */
Nick Piggin's avatar
Nick Piggin committed
812
static inline void expand(struct zone *zone, struct page *page,
813
814
	int low, int high, struct free_area *area,
	int migratetype)
Linus Torvalds's avatar
Linus Torvalds committed
815
816
817
818
819
820
821
{
	unsigned long size = 1 << high;

	while (high > low) {
		area--;
		high--;
		size >>= 1;
Nick Piggin's avatar
Nick Piggin committed
822
		VM_BUG_ON(bad_range(zone, &page[size]));
823
824
825
826
827
828
829
830
831
832
833
834
835

#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 */
836
837
			__mod_zone_freepage_state(zone, -(1 << high),
						  migratetype);
838
839
840
			continue;
		}
#endif
841
		list_add(&page[size].lru, &area->free_list[migratetype]);
Linus Torvalds's avatar
Linus Torvalds committed
842
843
844
845
846
847
848
849
		area->nr_free++;
		set_page_order(&page[size], high);
	}
}

/*
 * This page is about to be returned from the page allocator
 */
850
static inline int check_new_page(struct page *page)
Linus Torvalds's avatar
Linus Torvalds committed
851
{
Nick Piggin's avatar
Nick Piggin committed
852
853
	if (unlikely(page_mapcount(page) |
		(page->mapping != NULL)  |
854
		(atomic_read(&page->_count) != 0)  |
855
856
		(page->flags & PAGE_FLAGS_CHECK_AT_PREP) |
		(mem_cgroup_bad_page_check(page)))) {
857
		bad_page(page);
858
		return 1;
859
	}
860
861
862
863
864
865
866
867
868
869
870
871
	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;
	}
872

873
	set_page_private(page, 0);
874
	set_page_refcounted(page);
Nick Piggin's avatar
Nick Piggin committed
875
876

	arch_alloc_page(page, order);
Linus Torvalds's avatar
Linus Torvalds committed
877
	kernel_map_pages(page, 1 << order, 1);
Nick Piggin's avatar
Nick Piggin committed
878
879
880
881
882
883
884

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

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

885
	return 0;
Linus Torvalds's avatar
Linus Torvalds committed
886
887
}

888
889
890
891
/*
 * Go through the free lists for the given migratetype and remove
 * the smallest available page from the freelists
 */
892
893
static inline
struct page *__rmqueue_smallest(struct zone *zone, unsigned int order,
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
						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;
}


919
920
921
922
/*
 * This array describes the order lists are fallen back to when
 * the free lists for the desirable migrate type are depleted
 */
923
924
925
926
927
928
929
930
931
static int fallbacks[MIGRATE_TYPES][4] = {
	[MIGRATE_UNMOVABLE]   = { MIGRATE_RECLAIMABLE, MIGRATE_MOVABLE,     MIGRATE_RESERVE },
	[MIGRATE_RECLAIMABLE] = { MIGRATE_UNMOVABLE,   MIGRATE_MOVABLE,     MIGRATE_RESERVE },
#ifdef CONFIG_CMA
	[MIGRATE_MOVABLE]     = { MIGRATE_CMA,         MIGRATE_RECLAIMABLE, MIGRATE_UNMOVABLE, MIGRATE_RESERVE },
	[MIGRATE_CMA]         = { MIGRATE_RESERVE }, /* Never used */
#else
	[MIGRATE_MOVABLE]     = { MIGRATE_RECLAIMABLE, MIGRATE_UNMOVABLE,   MIGRATE_RESERVE },
#endif
932
	[MIGRATE_RESERVE]     = { MIGRATE_RESERVE }, /* Never used */
933
#ifdef CONFIG_MEMORY_ISOLATION
934
	[MIGRATE_ISOLATE]     = { MIGRATE_RESERVE }, /* Never used */
935
#endif
936
937
};

938
939
/*
 * Move the free pages in a range to the free lists of the requested type.
940
 * Note that start_page and end_pages are not aligned on a pageblock
941
942
 * boundary. If alignment is required, use move_freepages_block()
 */
943
int move_freepages(struct zone *zone,
Adrian Bunk's avatar
Adrian Bunk committed
944
945
			  struct page *start_page, struct page *end_page,
			  int migratetype)
946
947
948
{
	struct page *page;
	unsigned long order;
949
	int pages_moved = 0;
950
951
952
953
954
955
956

#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
957
	 * grouping pages by mobility
958
959
960
961
962
	 */
	BUG_ON(page_zone(start_page) != page_zone(end_page));
#endif

	for (page = start_page; page <= end_page;) {
963
964
965
		/* Make sure we are not inadvertently changing nodes */
		VM_BUG_ON(page_to_nid(page) != zone_to_nid(zone));

966
967
968
969
970
971
972
973
974
975
976
		if (!pfn_valid_within(page_to_pfn(page))) {
			page++;
			continue;
		}

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

		order = page_order(page);
977
978
		list_move(&page->lru,
			  &zone->free_area[order].free_list[migratetype]);
979
		set_freepage_migratetype(page, migratetype);
980
		page += 1 << order;
981
		pages_moved += 1 << order;
982
983
	}

984
	return pages_moved;
985
986
}

987
int move_freepages_block(struct zone *zone, struct page *page,
988
				int migratetype)
989
990
991
992
993
{
	unsigned long start_pfn, end_pfn;
	struct page *start_page, *end_page;

	start_pfn = page_to_pfn(page);
994
	start_pfn = start_pfn & ~(pageblock_nr_pages-1);
995
	start_page = pfn_to_page(start_pfn);
996
997
	end_page = start_page + pageblock_nr_pages - 1;
	end_pfn = start_pfn + pageblock_nr_pages - 1;
998
999

	/* Do not cross zone boundaries */
1000
	if (!zone_spans_pfn(zone, start_pfn))
1001
		start_page = page;
1002
	if (!zone_spans_pfn(zone, end_pfn))
1003
1004
1005
1006
1007
		return 0;

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

1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
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;
	}
}

1019
/* Remove an element from the buddy allocator from the fallback list */