page_alloc.c 82.8 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
22
23
/*
 *  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>
#include <linux/bootmem.h>
#include <linux/compiler.h>
24
#include <linux/kernel.h>
Linus Torvalds's avatar
Linus Torvalds committed
25
26
27
28
29
30
31
32
33
34
#include <linux/module.h>
#include <linux/suspend.h>
#include <linux/pagevec.h>
#include <linux/blkdev.h>
#include <linux/slab.h>
#include <linux/notifier.h>
#include <linux/topology.h>
#include <linux/sysctl.h>
#include <linux/cpu.h>
#include <linux/cpuset.h>
35
#include <linux/memory_hotplug.h>
Linus Torvalds's avatar
Linus Torvalds committed
36
37
#include <linux/nodemask.h>
#include <linux/vmalloc.h>
38
#include <linux/mempolicy.h>
39
#include <linux/stop_machine.h>
40
41
#include <linux/sort.h>
#include <linux/pfn.h>
42
#include <linux/backing-dev.h>
Linus Torvalds's avatar
Linus Torvalds committed
43
44

#include <asm/tlbflush.h>
45
#include <asm/div64.h>
Linus Torvalds's avatar
Linus Torvalds committed
46
47
48
49
50
51
#include "internal.h"

/*
 * MCD - HACK: Find somewhere to initialize this EARLY, or make this
 * initializer cleaner
 */
52
nodemask_t node_online_map __read_mostly = { { [0] = 1UL } };
53
EXPORT_SYMBOL(node_online_map);
54
nodemask_t node_possible_map __read_mostly = NODE_MASK_ALL;
55
EXPORT_SYMBOL(node_possible_map);
56
unsigned long totalram_pages __read_mostly;
57
unsigned long totalreserve_pages __read_mostly;
Linus Torvalds's avatar
Linus Torvalds committed
58
long nr_swap_pages;
59
int percpu_pagelist_fraction;
Linus Torvalds's avatar
Linus Torvalds committed
60

61
static void __free_pages_ok(struct page *page, unsigned int order);
62

Linus Torvalds's avatar
Linus Torvalds committed
63
64
65
66
67
68
69
/*
 * 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
70
71
72
 *
 * 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
73
 */
74
75
int sysctl_lowmem_reserve_ratio[MAX_NR_ZONES-1] = {
	 256,
76
#ifdef CONFIG_ZONE_DMA32
77
	 256,
78
#endif
79
#ifdef CONFIG_HIGHMEM
80
	 32
81
#endif
82
};
Linus Torvalds's avatar
Linus Torvalds committed
83
84
85

EXPORT_SYMBOL(totalram_pages);

86
87
static char *zone_names[MAX_NR_ZONES] = {
	 "DMA",
88
#ifdef CONFIG_ZONE_DMA32
89
	 "DMA32",
90
#endif
91
	 "Normal",
92
#ifdef CONFIG_HIGHMEM
93
	 "HighMem"
94
#endif
95
96
};

Linus Torvalds's avatar
Linus Torvalds committed
97
98
int min_free_kbytes = 1024;

99
100
unsigned long __meminitdata nr_kernel_pages;
unsigned long __meminitdata nr_all_pages;
101
static unsigned long __initdata dma_reserve;
Linus Torvalds's avatar
Linus Torvalds committed
102

103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
#ifdef CONFIG_ARCH_POPULATES_NODE_MAP
  /*
   * MAX_ACTIVE_REGIONS determines the maxmimum number of distinct
   * ranges of memory (RAM) that may be registered with add_active_range().
   * Ranges passed to add_active_range() will be merged if possible
   * so the number of times add_active_range() can be called is
   * related to the number of nodes and the number of holes
   */
  #ifdef CONFIG_MAX_ACTIVE_REGIONS
    /* Allow an architecture to set MAX_ACTIVE_REGIONS to save memory */
    #define MAX_ACTIVE_REGIONS CONFIG_MAX_ACTIVE_REGIONS
  #else
    #if MAX_NUMNODES >= 32
      /* If there can be many nodes, allow up to 50 holes per node */
      #define MAX_ACTIVE_REGIONS (MAX_NUMNODES*50)
    #else
      /* By default, allow up to 256 distinct regions */
      #define MAX_ACTIVE_REGIONS 256
    #endif
  #endif

  struct node_active_region __initdata early_node_map[MAX_ACTIVE_REGIONS];
  int __initdata nr_nodemap_entries;
  unsigned long __initdata arch_zone_lowest_possible_pfn[MAX_NR_ZONES];
  unsigned long __initdata arch_zone_highest_possible_pfn[MAX_NR_ZONES];
128
129
130
131
#ifdef CONFIG_MEMORY_HOTPLUG_RESERVE
  unsigned long __initdata node_boundary_start_pfn[MAX_NUMNODES];
  unsigned long __initdata node_boundary_end_pfn[MAX_NUMNODES];
#endif /* CONFIG_MEMORY_HOTPLUG_RESERVE */
132
133
#endif /* CONFIG_ARCH_POPULATES_NODE_MAP */

Nick Piggin's avatar
Nick Piggin committed
134
#ifdef CONFIG_DEBUG_VM
135
static int page_outside_zone_boundaries(struct zone *zone, struct page *page)
Linus Torvalds's avatar
Linus Torvalds committed
136
{
137
138
139
	int ret = 0;
	unsigned seq;
	unsigned long pfn = page_to_pfn(page);
140

141
142
143
144
145
146
147
148
149
	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;
150
151
152
153
}

static int page_is_consistent(struct zone *zone, struct page *page)
{
Linus Torvalds's avatar
Linus Torvalds committed
154
155
#ifdef CONFIG_HOLES_IN_ZONE
	if (!pfn_valid(page_to_pfn(page)))
156
		return 0;
Linus Torvalds's avatar
Linus Torvalds committed
157
158
#endif
	if (zone != page_zone(page))
159
160
161
162
163
164
165
166
167
168
		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
169
		return 1;
170
171
172
	if (!page_is_consistent(zone, page))
		return 1;

Linus Torvalds's avatar
Linus Torvalds committed
173
174
	return 0;
}
Nick Piggin's avatar
Nick Piggin committed
175
176
177
178
179
180
181
#else
static inline int bad_range(struct zone *zone, struct page *page)
{
	return 0;
}
#endif

182
static void bad_page(struct page *page)
Linus Torvalds's avatar
Linus Torvalds committed
183
{
184
	printk(KERN_EMERG "Bad page state in process '%s'\n"
185
186
187
		KERN_EMERG "page:%p flags:0x%0*lx mapping:%p mapcount:%d count:%d\n"
		KERN_EMERG "Trying to fix it up, but a reboot is needed\n"
		KERN_EMERG "Backtrace:\n",
188
189
190
		current->comm, page, (int)(2*sizeof(unsigned long)),
		(unsigned long)page->flags, page->mapping,
		page_mapcount(page), page_count(page));
Linus Torvalds's avatar
Linus Torvalds committed
191
	dump_stack();
192
193
	page->flags &= ~(1 << PG_lru	|
			1 << PG_private |
Linus Torvalds's avatar
Linus Torvalds committed
194
195
196
			1 << PG_locked	|
			1 << PG_active	|
			1 << PG_dirty	|
197
198
			1 << PG_reclaim |
			1 << PG_slab    |
Linus Torvalds's avatar
Linus Torvalds committed
199
			1 << PG_swapcache |
200
201
			1 << PG_writeback |
			1 << PG_buddy );
Linus Torvalds's avatar
Linus Torvalds committed
202
203
204
	set_page_count(page, 0);
	reset_page_mapcount(page);
	page->mapping = NULL;
205
	add_taint(TAINT_BAD_PAGE);
Linus Torvalds's avatar
Linus Torvalds committed
206
207
208
209
210
211
212
213
214
215
216
217
}

/*
 * 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".
 *
 * All pages have PG_compound set.  All pages have their ->private pointing at
 * the head page (even the head page has this).
 *
218
219
220
 * 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
221
 */
222
223
224
225
226
227

static void free_compound_page(struct page *page)
{
	__free_pages_ok(page, (unsigned long)page[1].lru.prev);
}

Linus Torvalds's avatar
Linus Torvalds committed
228
229
230
231
232
static void prep_compound_page(struct page *page, unsigned long order)
{
	int i;
	int nr_pages = 1 << order;

233
	page[1].lru.next = (void *)free_compound_page;	/* set dtor */
234
	page[1].lru.prev = (void *)order;
Linus Torvalds's avatar
Linus Torvalds committed
235
236
237
	for (i = 0; i < nr_pages; i++) {
		struct page *p = page + i;

238
		__SetPageCompound(p);
239
		set_page_private(p, (unsigned long)page);
Linus Torvalds's avatar
Linus Torvalds committed
240
241
242
243
244
245
246
247
	}
}

static void destroy_compound_page(struct page *page, unsigned long order)
{
	int i;
	int nr_pages = 1 << order;

248
	if (unlikely((unsigned long)page[1].lru.prev != order))
249
		bad_page(page);
Linus Torvalds's avatar
Linus Torvalds committed
250
251
252
253

	for (i = 0; i < nr_pages; i++) {
		struct page *p = page + i;

254
255
256
		if (unlikely(!PageCompound(p) |
				(page_private(p) != (unsigned long)page)))
			bad_page(page);
257
		__ClearPageCompound(p);
Linus Torvalds's avatar
Linus Torvalds committed
258
259
260
	}
}

Nick Piggin's avatar
Nick Piggin committed
261
262
263
264
static inline void prep_zero_page(struct page *page, int order, gfp_t gfp_flags)
{
	int i;

Nick Piggin's avatar
Nick Piggin committed
265
	VM_BUG_ON((gfp_flags & (__GFP_WAIT | __GFP_HIGHMEM)) == __GFP_HIGHMEM);
266
267
268
269
	/*
	 * 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
270
	VM_BUG_ON((gfp_flags & __GFP_HIGHMEM) && in_interrupt());
Nick Piggin's avatar
Nick Piggin committed
271
272
273
274
	for (i = 0; i < (1 << order); i++)
		clear_highpage(page + i);
}

Linus Torvalds's avatar
Linus Torvalds committed
275
276
277
278
279
/*
 * function for dealing with page's order in buddy system.
 * zone->lock is already acquired when we use these.
 * So, we don't need atomic page->flags operations here.
 */
280
281
static inline unsigned long page_order(struct page *page)
{
282
	return page_private(page);
Linus Torvalds's avatar
Linus Torvalds committed
283
284
}

285
286
static inline void set_page_order(struct page *page, int order)
{
287
	set_page_private(page, order);
288
	__SetPageBuddy(page);
Linus Torvalds's avatar
Linus Torvalds committed
289
290
291
292
}

static inline void rmv_page_order(struct page *page)
{
293
	__ClearPageBuddy(page);
294
	set_page_private(page, 0);
Linus Torvalds's avatar
Linus Torvalds committed
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
}

/*
 * 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
312
 * Assumption: *_mem_map is contiguous at least up to MAX_ORDER
Linus Torvalds's avatar
Linus Torvalds committed
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
 */
static inline struct page *
__page_find_buddy(struct page *page, unsigned long page_idx, unsigned int order)
{
	unsigned long buddy_idx = page_idx ^ (1 << order);

	return page + (buddy_idx - page_idx);
}

static inline unsigned long
__find_combined_index(unsigned long page_idx, unsigned int order)
{
	return (page_idx & ~(1 << order));
}

/*
 * 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
331
 * (a) the buddy is not in a hole &&
332
 * (b) the buddy is in the buddy system &&
333
334
 * (c) a page and its buddy have the same order &&
 * (d) a page and its buddy are in the same zone.
335
336
337
 *
 * For recording whether a page is in the buddy system, we use PG_buddy.
 * Setting, clearing, and testing PG_buddy is serialized by zone->lock.
Linus Torvalds's avatar
Linus Torvalds committed
338
 *
339
 * For recording page's order, we use page_private(page).
Linus Torvalds's avatar
Linus Torvalds committed
340
 */
341
342
static inline int page_is_buddy(struct page *page, struct page *buddy,
								int order)
Linus Torvalds's avatar
Linus Torvalds committed
343
{
Nick Piggin's avatar
Nick Piggin committed
344
#ifdef CONFIG_HOLES_IN_ZONE
345
	if (!pfn_valid(page_to_pfn(buddy)))
Nick Piggin's avatar
Nick Piggin committed
346
347
348
		return 0;
#endif

349
350
351
352
353
	if (page_zone_id(page) != page_zone_id(buddy))
		return 0;

	if (PageBuddy(buddy) && page_order(buddy) == order) {
		BUG_ON(page_count(buddy) != 0);
354
		return 1;
355
	}
356
	return 0;
Linus Torvalds's avatar
Linus Torvalds committed
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
}

/*
 * 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
372
 * free pages of length of (1 << order) and marked with PG_buddy. Page's
373
 * order is recorded in page_private(page) field.
Linus Torvalds's avatar
Linus Torvalds committed
374
375
376
377
378
379
380
381
382
 * 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
383
static inline void __free_one_page(struct page *page,
Linus Torvalds's avatar
Linus Torvalds committed
384
385
386
387
388
		struct zone *zone, unsigned int order)
{
	unsigned long page_idx;
	int order_size = 1 << order;

389
	if (unlikely(PageCompound(page)))
Linus Torvalds's avatar
Linus Torvalds committed
390
391
392
393
		destroy_compound_page(page, order);

	page_idx = page_to_pfn(page) & ((1 << MAX_ORDER) - 1);

Nick Piggin's avatar
Nick Piggin committed
394
395
	VM_BUG_ON(page_idx & (order_size - 1));
	VM_BUG_ON(bad_range(zone, page));
Linus Torvalds's avatar
Linus Torvalds committed
396
397
398
399
400
401
402
403

	zone->free_pages += order_size;
	while (order < MAX_ORDER-1) {
		unsigned long combined_idx;
		struct free_area *area;
		struct page *buddy;

		buddy = __page_find_buddy(page, page_idx, order);
404
		if (!page_is_buddy(page, buddy, order))
Linus Torvalds's avatar
Linus Torvalds committed
405
			break;		/* Move the buddy up one level. */
Nick Piggin's avatar
Nick Piggin committed
406

Linus Torvalds's avatar
Linus Torvalds committed
407
408
409
410
		list_del(&buddy->lru);
		area = zone->free_area + order;
		area->nr_free--;
		rmv_page_order(buddy);
Nick Piggin's avatar
Nick Piggin committed
411
		combined_idx = __find_combined_index(page_idx, order);
Linus Torvalds's avatar
Linus Torvalds committed
412
413
414
415
416
417
418
419
420
		page = page + (combined_idx - page_idx);
		page_idx = combined_idx;
		order++;
	}
	set_page_order(page, order);
	list_add(&page->lru, &zone->free_area[order].free_list);
	zone->free_area[order].nr_free++;
}

421
static inline int free_pages_check(struct page *page)
Linus Torvalds's avatar
Linus Torvalds committed
422
{
Nick Piggin's avatar
Nick Piggin committed
423
424
425
	if (unlikely(page_mapcount(page) |
		(page->mapping != NULL)  |
		(page_count(page) != 0)  |
Linus Torvalds's avatar
Linus Torvalds committed
426
427
428
429
430
431
432
433
		(page->flags & (
			1 << PG_lru	|
			1 << PG_private |
			1 << PG_locked	|
			1 << PG_active	|
			1 << PG_reclaim	|
			1 << PG_slab	|
			1 << PG_swapcache |
Nick Piggin's avatar
Nick Piggin committed
434
			1 << PG_writeback |
435
436
			1 << PG_reserved |
			1 << PG_buddy ))))
437
		bad_page(page);
Linus Torvalds's avatar
Linus Torvalds committed
438
	if (PageDirty(page))
Nick Piggin's avatar
Nick Piggin committed
439
		__ClearPageDirty(page);
440
441
442
443
444
445
	/*
	 * For now, we report if PG_reserved was found set, but do not
	 * clear it, and do not free the page.  But we shall soon need
	 * to do more, for when the ZERO_PAGE count wraps negative.
	 */
	return PageReserved(page);
Linus Torvalds's avatar
Linus Torvalds committed
446
447
448
449
450
}

/*
 * Frees a list of pages. 
 * Assumes all pages on list are in same zone, and of same order.
451
 * count is the number of pages to free.
Linus Torvalds's avatar
Linus Torvalds committed
452
453
454
455
456
457
458
 *
 * 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.
 */
Nick Piggin's avatar
Nick Piggin committed
459
460
static void free_pages_bulk(struct zone *zone, int count,
					struct list_head *list, int order)
Linus Torvalds's avatar
Linus Torvalds committed
461
{
Nick Piggin's avatar
Nick Piggin committed
462
	spin_lock(&zone->lock);
Linus Torvalds's avatar
Linus Torvalds committed
463
464
	zone->all_unreclaimable = 0;
	zone->pages_scanned = 0;
Nick Piggin's avatar
Nick Piggin committed
465
466
467
	while (count--) {
		struct page *page;

Nick Piggin's avatar
Nick Piggin committed
468
		VM_BUG_ON(list_empty(list));
Linus Torvalds's avatar
Linus Torvalds committed
469
		page = list_entry(list->prev, struct page, lru);
Nick Piggin's avatar
Nick Piggin committed
470
		/* have to delete it as __free_one_page list manipulates */
Linus Torvalds's avatar
Linus Torvalds committed
471
		list_del(&page->lru);
Nick Piggin's avatar
Nick Piggin committed
472
		__free_one_page(page, zone, order);
Linus Torvalds's avatar
Linus Torvalds committed
473
	}
Nick Piggin's avatar
Nick Piggin committed
474
	spin_unlock(&zone->lock);
Linus Torvalds's avatar
Linus Torvalds committed
475
476
}

Nick Piggin's avatar
Nick Piggin committed
477
static void free_one_page(struct zone *zone, struct page *page, int order)
Linus Torvalds's avatar
Linus Torvalds committed
478
{
479
480
481
	spin_lock(&zone->lock);
	zone->all_unreclaimable = 0;
	zone->pages_scanned = 0;
482
	__free_one_page(page, zone, order);
483
	spin_unlock(&zone->lock);
Nick Piggin's avatar
Nick Piggin committed
484
485
486
487
488
}

static void __free_pages_ok(struct page *page, unsigned int order)
{
	unsigned long flags;
Linus Torvalds's avatar
Linus Torvalds committed
489
	int i;
490
	int reserved = 0;
Linus Torvalds's avatar
Linus Torvalds committed
491
492

	for (i = 0 ; i < (1 << order) ; ++i)
493
		reserved += free_pages_check(page + i);
494
495
496
	if (reserved)
		return;

Nick Piggin's avatar
Nick Piggin committed
497
498
	if (!PageHighMem(page))
		debug_check_no_locks_freed(page_address(page),PAGE_SIZE<<order);
Nick Piggin's avatar
Nick Piggin committed
499
	arch_free_page(page, order);
Nick Piggin's avatar
Nick Piggin committed
500
	kernel_map_pages(page, 1 << order, 0);
Nick Piggin's avatar
Nick Piggin committed
501

Nick Piggin's avatar
Nick Piggin committed
502
	local_irq_save(flags);
503
	__count_vm_events(PGFREE, 1 << order);
Nick Piggin's avatar
Nick Piggin committed
504
	free_one_page(page_zone(page), page, order);
Nick Piggin's avatar
Nick Piggin committed
505
	local_irq_restore(flags);
Linus Torvalds's avatar
Linus Torvalds committed
506
507
}

508
509
510
511
512
513
514
515
/*
 * permit the bootmem allocator to evade page validation on high-order frees
 */
void fastcall __init __free_pages_bootmem(struct page *page, unsigned int order)
{
	if (order == 0) {
		__ClearPageReserved(page);
		set_page_count(page, 0);
516
		set_page_refcounted(page);
Nick Piggin's avatar
Nick Piggin committed
517
		__free_page(page);
518
519
520
	} else {
		int loop;

Nick Piggin's avatar
Nick Piggin committed
521
		prefetchw(page);
522
523
524
		for (loop = 0; loop < BITS_PER_LONG; loop++) {
			struct page *p = &page[loop];

Nick Piggin's avatar
Nick Piggin committed
525
526
			if (loop + 1 < BITS_PER_LONG)
				prefetchw(p + 1);
527
528
529
530
			__ClearPageReserved(p);
			set_page_count(p, 0);
		}

531
		set_page_refcounted(page);
Nick Piggin's avatar
Nick Piggin committed
532
		__free_pages(page, order);
533
534
535
	}
}

Linus Torvalds's avatar
Linus Torvalds committed
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550

/*
 * 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
551
static inline void expand(struct zone *zone, struct page *page,
Linus Torvalds's avatar
Linus Torvalds committed
552
553
554
555
556
557
558
559
 	int low, int high, struct free_area *area)
{
	unsigned long size = 1 << high;

	while (high > low) {
		area--;
		high--;
		size >>= 1;
Nick Piggin's avatar
Nick Piggin committed
560
		VM_BUG_ON(bad_range(zone, &page[size]));
Linus Torvalds's avatar
Linus Torvalds committed
561
562
563
564
565
566
567
568
569
		list_add(&page[size].lru, &area->free_list);
		area->nr_free++;
		set_page_order(&page[size], high);
	}
}

/*
 * This page is about to be returned from the page allocator
 */
Nick Piggin's avatar
Nick Piggin committed
570
static int prep_new_page(struct page *page, int order, gfp_t gfp_flags)
Linus Torvalds's avatar
Linus Torvalds committed
571
{
Nick Piggin's avatar
Nick Piggin committed
572
573
574
	if (unlikely(page_mapcount(page) |
		(page->mapping != NULL)  |
		(page_count(page) != 0)  |
575
576
		(page->flags & (
			1 << PG_lru	|
Linus Torvalds's avatar
Linus Torvalds committed
577
578
579
580
581
			1 << PG_private	|
			1 << PG_locked	|
			1 << PG_active	|
			1 << PG_dirty	|
			1 << PG_reclaim	|
582
			1 << PG_slab    |
Linus Torvalds's avatar
Linus Torvalds committed
583
			1 << PG_swapcache |
Nick Piggin's avatar
Nick Piggin committed
584
			1 << PG_writeback |
585
586
			1 << PG_reserved |
			1 << PG_buddy ))))
587
		bad_page(page);
Linus Torvalds's avatar
Linus Torvalds committed
588

589
590
591
592
593
594
595
	/*
	 * For now, we report if PG_reserved was found set, but do not
	 * clear it, and do not allocate the page: as a safety net.
	 */
	if (PageReserved(page))
		return 1;

Linus Torvalds's avatar
Linus Torvalds committed
596
597
598
	page->flags &= ~(1 << PG_uptodate | 1 << PG_error |
			1 << PG_referenced | 1 << PG_arch_1 |
			1 << PG_checked | 1 << PG_mappedtodisk);
599
	set_page_private(page, 0);
600
	set_page_refcounted(page);
Linus Torvalds's avatar
Linus Torvalds committed
601
	kernel_map_pages(page, 1 << order, 1);
Nick Piggin's avatar
Nick Piggin committed
602
603
604
605
606
607
608

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

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

609
	return 0;
Linus Torvalds's avatar
Linus Torvalds committed
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
}

/* 
 * Do the hard work of removing an element from the buddy allocator.
 * Call me with the zone->lock already held.
 */
static struct page *__rmqueue(struct zone *zone, unsigned int order)
{
	struct free_area * area;
	unsigned int current_order;
	struct page *page;

	for (current_order = order; current_order < MAX_ORDER; ++current_order) {
		area = zone->free_area + current_order;
		if (list_empty(&area->free_list))
			continue;

		page = list_entry(area->free_list.next, struct page, lru);
		list_del(&page->lru);
		rmv_page_order(page);
		area->nr_free--;
		zone->free_pages -= 1UL << order;
Nick Piggin's avatar
Nick Piggin committed
632
633
		expand(zone, page, order, current_order, area);
		return page;
Linus Torvalds's avatar
Linus Torvalds committed
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
	}

	return NULL;
}

/* 
 * 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, 
			unsigned long count, struct list_head *list)
{
	int i;
	
Nick Piggin's avatar
Nick Piggin committed
649
	spin_lock(&zone->lock);
Linus Torvalds's avatar
Linus Torvalds committed
650
	for (i = 0; i < count; ++i) {
Nick Piggin's avatar
Nick Piggin committed
651
652
		struct page *page = __rmqueue(zone, order);
		if (unlikely(page == NULL))
Linus Torvalds's avatar
Linus Torvalds committed
653
654
655
			break;
		list_add_tail(&page->lru, list);
	}
Nick Piggin's avatar
Nick Piggin committed
656
	spin_unlock(&zone->lock);
Nick Piggin's avatar
Nick Piggin committed
657
	return i;
Linus Torvalds's avatar
Linus Torvalds committed
658
659
}

660
#ifdef CONFIG_NUMA
661
662
/*
 * Called from the slab reaper to drain pagesets on a particular node that
663
 * belongs to the currently executing processor.
664
665
 * Note that this function must be called with the thread pinned to
 * a single processor.
666
667
 */
void drain_node_pages(int nodeid)
668
{
669
670
	int i;
	enum zone_type z;
671
672
	unsigned long flags;

673
674
	for (z = 0; z < MAX_NR_ZONES; z++) {
		struct zone *zone = NODE_DATA(nodeid)->node_zones + z;
675
676
		struct per_cpu_pageset *pset;

677
678
679
		if (!populated_zone(zone))
			continue;

Nick Piggin's avatar
Nick Piggin committed
680
		pset = zone_pcp(zone, smp_processor_id());
681
682
683
684
		for (i = 0; i < ARRAY_SIZE(pset->pcp); i++) {
			struct per_cpu_pages *pcp;

			pcp = &pset->pcp[i];
685
686
687
688
689
690
			if (pcp->count) {
				local_irq_save(flags);
				free_pages_bulk(zone, pcp->count, &pcp->list, 0);
				pcp->count = 0;
				local_irq_restore(flags);
			}
691
692
693
694
695
		}
	}
}
#endif

Linus Torvalds's avatar
Linus Torvalds committed
696
697
698
#if defined(CONFIG_PM) || defined(CONFIG_HOTPLUG_CPU)
static void __drain_pages(unsigned int cpu)
{
Nick Piggin's avatar
Nick Piggin committed
699
	unsigned long flags;
Linus Torvalds's avatar
Linus Torvalds committed
700
701
702
703
704
705
	struct zone *zone;
	int i;

	for_each_zone(zone) {
		struct per_cpu_pageset *pset;

706
		pset = zone_pcp(zone, cpu);
Linus Torvalds's avatar
Linus Torvalds committed
707
708
709
710
		for (i = 0; i < ARRAY_SIZE(pset->pcp); i++) {
			struct per_cpu_pages *pcp;

			pcp = &pset->pcp[i];
Nick Piggin's avatar
Nick Piggin committed
711
			local_irq_save(flags);
Nick Piggin's avatar
Nick Piggin committed
712
713
			free_pages_bulk(zone, pcp->count, &pcp->list, 0);
			pcp->count = 0;
Nick Piggin's avatar
Nick Piggin committed
714
			local_irq_restore(flags);
Linus Torvalds's avatar
Linus Torvalds committed
715
716
717
718
719
720
721
722
723
		}
	}
}
#endif /* CONFIG_PM || CONFIG_HOTPLUG_CPU */

#ifdef CONFIG_PM

void mark_free_pages(struct zone *zone)
{
724
725
	unsigned long pfn, max_zone_pfn;
	unsigned long flags;
Linus Torvalds's avatar
Linus Torvalds committed
726
727
728
729
730
731
732
	int order;
	struct list_head *curr;

	if (!zone->spanned_pages)
		return;

	spin_lock_irqsave(&zone->lock, flags);
733
734
735
736
737
738
739
740
741

	max_zone_pfn = zone->zone_start_pfn + zone->spanned_pages;
	for (pfn = zone->zone_start_pfn; pfn < max_zone_pfn; pfn++)
		if (pfn_valid(pfn)) {
			struct page *page = pfn_to_page(pfn);

			if (!PageNosave(page))
				ClearPageNosaveFree(page);
		}
Linus Torvalds's avatar
Linus Torvalds committed
742
743
744

	for (order = MAX_ORDER - 1; order >= 0; --order)
		list_for_each(curr, &zone->free_area[order].free_list) {
745
			unsigned long i;
Linus Torvalds's avatar
Linus Torvalds committed
746

747
748
749
750
			pfn = page_to_pfn(list_entry(curr, struct page, lru));
			for (i = 0; i < (1UL << order); i++)
				SetPageNosaveFree(pfn_to_page(pfn + i));
		}
Linus Torvalds's avatar
Linus Torvalds committed
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778

	spin_unlock_irqrestore(&zone->lock, flags);
}

/*
 * Spill all of this CPU's per-cpu pages back into the buddy allocator.
 */
void drain_local_pages(void)
{
	unsigned long flags;

	local_irq_save(flags);	
	__drain_pages(smp_processor_id());
	local_irq_restore(flags);	
}
#endif /* CONFIG_PM */

/*
 * Free a 0-order page
 */
static void fastcall free_hot_cold_page(struct page *page, int cold)
{
	struct zone *zone = page_zone(page);
	struct per_cpu_pages *pcp;
	unsigned long flags;

	if (PageAnon(page))
		page->mapping = NULL;
779
	if (free_pages_check(page))
780
781
		return;

Nick Piggin's avatar
Nick Piggin committed
782
783
	if (!PageHighMem(page))
		debug_check_no_locks_freed(page_address(page), PAGE_SIZE);
Nick Piggin's avatar
Nick Piggin committed
784
	arch_free_page(page, 0);
785
786
	kernel_map_pages(page, 1, 0);

787
	pcp = &zone_pcp(zone, get_cpu())->pcp[cold];
Linus Torvalds's avatar
Linus Torvalds committed
788
	local_irq_save(flags);
789
	__count_vm_event(PGFREE);
Linus Torvalds's avatar
Linus Torvalds committed
790
791
	list_add(&page->lru, &pcp->list);
	pcp->count++;
Nick Piggin's avatar
Nick Piggin committed
792
793
794
795
	if (pcp->count >= pcp->high) {
		free_pages_bulk(zone, pcp->batch, &pcp->list, 0);
		pcp->count -= pcp->batch;
	}
Linus Torvalds's avatar
Linus Torvalds committed
796
797
798
799
800
801
802
803
804
805
806
807
808
809
	local_irq_restore(flags);
	put_cpu();
}

void fastcall free_hot_page(struct page *page)
{
	free_hot_cold_page(page, 0);
}
	
void fastcall free_cold_page(struct page *page)
{
	free_hot_cold_page(page, 1);
}

810
811
812
813
814
815
816
817
818
819
820
821
/*
 * split_page takes a non-compound higher-order page, and splits it into
 * n (1<<order) sub-pages: page[0..n]
 * Each sub-page must be freed individually.
 *
 * Note: this is probably too low level an operation for use in drivers.
 * Please consult with lkml before using this in your driver.
 */
void split_page(struct page *page, unsigned int order)
{
	int i;

Nick Piggin's avatar
Nick Piggin committed
822
823
	VM_BUG_ON(PageCompound(page));
	VM_BUG_ON(!page_count(page));
824
825
	for (i = 1; i < (1 << order); i++)
		set_page_refcounted(page + i);
826
827
}

Linus Torvalds's avatar
Linus Torvalds committed
828
829
830
831
832
/*
 * Really, prep_compound_page() should be called from __rmqueue_bulk().  But
 * we cheat by calling it from here, in the order > 0 path.  Saves a branch
 * or two.
 */
Nick Piggin's avatar
Nick Piggin committed
833
834
static struct page *buffered_rmqueue(struct zonelist *zonelist,
			struct zone *zone, int order, gfp_t gfp_flags)
Linus Torvalds's avatar
Linus Torvalds committed
835
836
{
	unsigned long flags;
837
	struct page *page;
Linus Torvalds's avatar
Linus Torvalds committed
838
	int cold = !!(gfp_flags & __GFP_COLD);
Nick Piggin's avatar
Nick Piggin committed
839
	int cpu;
Linus Torvalds's avatar
Linus Torvalds committed
840

841
again:
Nick Piggin's avatar
Nick Piggin committed
842
	cpu  = get_cpu();
Nick Piggin's avatar
Nick Piggin committed
843
	if (likely(order == 0)) {
Linus Torvalds's avatar
Linus Torvalds committed
844
845
		struct per_cpu_pages *pcp;

Nick Piggin's avatar
Nick Piggin committed
846
		pcp = &zone_pcp(zone, cpu)->pcp[cold];
Linus Torvalds's avatar
Linus Torvalds committed
847
		local_irq_save(flags);
Nick Piggin's avatar
Nick Piggin committed
848
		if (!pcp->count) {
849
			pcp->count = rmqueue_bulk(zone, 0,
Linus Torvalds's avatar
Linus Torvalds committed
850
						pcp->batch, &pcp->list);
Nick Piggin's avatar
Nick Piggin committed
851
852
			if (unlikely(!pcp->count))
				goto failed;
Linus Torvalds's avatar
Linus Torvalds committed
853
		}
Nick Piggin's avatar
Nick Piggin committed
854
855
856
		page = list_entry(pcp->list.next, struct page, lru);
		list_del(&page->lru);
		pcp->count--;
Rohit Seth's avatar
Rohit Seth committed
857
	} else {
Linus Torvalds's avatar
Linus Torvalds committed
858
859
		spin_lock_irqsave(&zone->lock, flags);
		page = __rmqueue(zone, order);
Nick Piggin's avatar
Nick Piggin committed
860
861
862
		spin_unlock(&zone->lock);
		if (!page)
			goto failed;
Linus Torvalds's avatar
Linus Torvalds committed
863
864
	}

865
	__count_zone_vm_events(PGALLOC, zone, 1 << order);
866
	zone_statistics(zonelist, zone);
Nick Piggin's avatar
Nick Piggin committed
867
868
	local_irq_restore(flags);
	put_cpu();
Linus Torvalds's avatar
Linus Torvalds committed
869

Nick Piggin's avatar
Nick Piggin committed
870
	VM_BUG_ON(bad_range(zone, page));
Nick Piggin's avatar
Nick Piggin committed
871
	if (prep_new_page(page, order, gfp_flags))
Nick Piggin's avatar
Nick Piggin committed
872
		goto again;
Linus Torvalds's avatar
Linus Torvalds committed
873
	return page;
Nick Piggin's avatar
Nick Piggin committed
874
875
876
877
878

failed:
	local_irq_restore(flags);
	put_cpu();
	return NULL;
Linus Torvalds's avatar
Linus Torvalds committed
879
880
}

Rohit Seth's avatar
Rohit Seth committed
881
#define ALLOC_NO_WATERMARKS	0x01 /* don't check watermarks at all */
882
883
884
885
886
887
#define ALLOC_WMARK_MIN		0x02 /* use pages_min watermark */
#define ALLOC_WMARK_LOW		0x04 /* use pages_low watermark */
#define ALLOC_WMARK_HIGH	0x08 /* use pages_high watermark */
#define ALLOC_HARDER		0x10 /* try to alloc harder */
#define ALLOC_HIGH		0x20 /* __GFP_HIGH set */
#define ALLOC_CPUSET		0x40 /* check for correct cpuset */
Rohit Seth's avatar
Rohit Seth committed
888

Linus Torvalds's avatar
Linus Torvalds committed
889
890
891
892
893
/*
 * Return 1 if free pages are above 'mark'. This takes into account the order
 * of the allocation.
 */
int zone_watermark_ok(struct zone *z, int order, unsigned long mark,
Rohit Seth's avatar
Rohit Seth committed
894
		      int classzone_idx, int alloc_flags)
Linus Torvalds's avatar
Linus Torvalds committed
895
896
{
	/* free_pages my go negative - that's OK */
897
898
	unsigned long min = mark;
	long free_pages = z->free_pages - (1 << order) + 1;
Linus Torvalds's avatar
Linus Torvalds committed
899
900
	int o;

Rohit Seth's avatar
Rohit Seth committed
901
	if (alloc_flags & ALLOC_HIGH)
Linus Torvalds's avatar
Linus Torvalds committed
902
		min -= min / 2;
Rohit Seth's avatar
Rohit Seth committed
903
	if (alloc_flags & ALLOC_HARDER)
Linus Torvalds's avatar
Linus Torvalds committed
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
		min -= min / 4;

	if (free_pages <= min + z->lowmem_reserve[classzone_idx])
		return 0;
	for (o = 0; o < order; o++) {
		/* At the next order, this order's pages become unavailable */
		free_pages -= z->free_area[o].nr_free << o;

		/* Require fewer higher order pages to be free */
		min >>= 1;

		if (free_pages <= min)
			return 0;
	}
	return 1;
}

Rohit Seth's avatar
Rohit Seth committed
921
/*
922
 * get_page_from_freelist goes through the zonelist trying to allocate
Rohit Seth's avatar
Rohit Seth committed
923
924
925
926
927
 * a page.
 */
static struct page *
get_page_from_freelist(gfp_t gfp_mask, unsigned int order,
		struct zonelist *zonelist, int alloc_flags)
Martin Hicks's avatar
Martin Hicks committed
928
{
Rohit Seth's avatar
Rohit Seth committed
929
930
931
	struct zone **z = zonelist->zones;
	struct page *page = NULL;
	int classzone_idx = zone_idx(*z);
932
	struct zone *zone;
Rohit Seth's avatar
Rohit Seth committed
933
934
935
936
937
938

	/*
	 * Go through the zonelist once, looking for a zone with enough free.
	 * See also cpuset_zone_allowed() comment in kernel/cpuset.c.
	 */
	do {
939
		zone = *z;
940
		if (unlikely(NUMA_BUILD && (gfp_mask & __GFP_THISNODE) &&
941
			zone->zone_pgdat != zonelist->zones[0]->zone_pgdat))
942
				break;
Rohit Seth's avatar
Rohit Seth committed
943
		if ((alloc_flags & ALLOC_CPUSET) &&
944
945
			!cpuset_zone_allowed(zone, gfp_mask))
				continue;
Rohit Seth's avatar
Rohit Seth committed
946
947

		if (!(alloc_flags & ALLOC_NO_WATERMARKS)) {
948
949
			unsigned long mark;
			if (alloc_flags & ALLOC_WMARK_MIN)
950
				mark = zone->pages_min;
951
			else if (alloc_flags & ALLOC_WMARK_LOW)
952
				mark = zone->pages_low;
953
			else
954
				mark = zone->pages_high;
955
956
			if (!zone_watermark_ok(zone, order, mark,
				    classzone_idx, alloc_flags)) {
957
				if (!zone_reclaim_mode ||
958
				    !zone_reclaim(zone, gfp_mask, order))
959
					continue;
960
			}
Rohit Seth's avatar
Rohit Seth committed
961
962
		}

963
		page = buffered_rmqueue(zonelist, zone, order, gfp_mask);
964
		if (page)
Rohit Seth's avatar
Rohit Seth committed
965
			break;
966

Rohit Seth's avatar
Rohit Seth committed
967
968
	} while (*(++z) != NULL);
	return page;
Martin Hicks's avatar
Martin Hicks committed
969
970
}

Linus Torvalds's avatar
Linus Torvalds committed
971
972
973
974
/*
 * This is the 'heart' of the zoned buddy allocator.
 */
struct page * fastcall
975
__alloc_pages(gfp_t gfp_mask, unsigned int order,
Linus Torvalds's avatar
Linus Torvalds committed
976
977
		struct zonelist *zonelist)
{
Al Viro's avatar
Al Viro committed
978
	const gfp_t wait = gfp_mask & __GFP_WAIT;
Rohit Seth's avatar
Rohit Seth committed
979
	struct zone **z;
Linus Torvalds's avatar
Linus Torvalds committed
980
981
982
983
	struct page *page;
	struct reclaim_state reclaim_state;
	struct task_struct *p = current;
	int do_retry;
Rohit Seth's avatar
Rohit Seth committed
984
	int alloc_flags;
Linus Torvalds's avatar
Linus Torvalds committed
985
986
987
988
	int did_some_progress;

	might_sleep_if(wait);

989
restart:
Rohit Seth's avatar
Rohit Seth committed
990
	z = zonelist->zones;  /* the list of zones suitable for gfp_mask */
Linus Torvalds's avatar
Linus Torvalds committed
991

Rohit Seth's avatar
Rohit Seth committed
992
	if (unlikely(*z == NULL)) {
Linus Torvalds's avatar
Linus Torvalds committed
993
994
995
		/* Should this ever happen?? */
		return NULL;
	}
996

Rohit Seth's avatar
Rohit Seth committed
997
	page = get_page_from_freelist(gfp_mask|__GFP_HARDWALL, order,
998
				zonelist, ALLOC_WMARK_LOW|ALLOC_CPUSET);
Rohit Seth's avatar
Rohit Seth committed
999
1000
	if (page)
		goto got_pg;
Linus Torvalds's avatar
Linus Torvalds committed
1001

1002
	for (z = zonelist->zones; *z; z++)
1003
		wakeup_kswapd(*z, order);
Linus Torvalds's avatar
Linus Torvalds committed
1004

1005
	/*
Rohit Seth's avatar
Rohit Seth committed
1006
1007
1008
1009
1010
1011
	 * OK, we're below the kswapd watermark and have kicked background
	 * reclaim. Now things get more complex, so set up alloc_flags according
	 * to how we want to proceed.
	 *
	 * The caller may dip into page reserves a bit more if the caller
	 * cannot run direct reclaim, or if the caller has realtime scheduling
Paul Jackson's avatar
Paul Jackson committed
1012
1013
	 * policy or is asking for __GFP_HIGH memory.  GFP_ATOMIC requests will
	 * set both ALLOC_HARDER (!wait) and ALLOC_HIGH (__GFP_HIGH).
1014
	 */
1015
	alloc_flags = ALLOC_WMARK_MIN;
Rohit Seth's avatar
Rohit Seth committed
1016
1017
1018
1019
	if ((unlikely(rt_task(p)) && !in_interrupt()) || !wait)
		alloc_flags |= ALLOC_HARDER;
	if (gfp_mask & __GFP_HIGH)
		alloc_flags |= ALLOC_HIGH;
1020
1021
	if (wait)
		alloc_flags |= ALLOC_CPUSET;
Linus Torvalds's avatar
Linus Torvalds committed
1022
1023
1024

	/*
	 * Go through the zonelist again. Let __GFP_HIGH and allocations
Rohit Seth's avatar
Rohit Seth committed
1025
	 * coming from realtime tasks go deeper into reserves.
Linus Torvalds's avatar
Linus Torvalds committed
1026
1027
1028
	 *
	 * This is the last chance, in general, before the goto nopage.
	 * Ignore cpuset if GFP_ATOMIC (!wait) rather than fail alloc.
1029
	 * See also cpuset_zone_allowed() comment in kernel/cpuset.c.
Linus Torvalds's avatar
Linus Torvalds committed
1030
	 */
Rohit Seth's avatar
Rohit Seth committed
1031
1032
1033
	page = get_page_from_freelist(gfp_mask, order, zonelist, alloc_flags);
	if (page)
		goto got_pg;
Linus Torvalds's avatar
Linus Torvalds committed
1034
1035

	/* This allocation should allow future memory freeing. */
1036
1037
1038
1039

	if (((p->flags & PF_MEMALLOC) || unlikely(test_thread_flag(TIF_MEMDIE)))
			&& !in_interrupt()) {
		if (!(gfp_mask & __GFP_NOMEMALLOC)) {
Kirill Korotaev's avatar
Kirill Korotaev committed
1040
nofail_alloc:
1041
			/* go through the zonelist yet again, ignoring mins */
Rohit Seth's avatar
Rohit Seth committed
1042
			page = get_page_from_freelist(gfp_mask, order,
1043
				zonelist, ALLOC_NO_WATERMARKS);
Rohit Seth's avatar
Rohit Seth committed
1044
1045
			if (page)
				goto got_pg;
Kirill Korotaev's avatar
Kirill Korotaev committed
1046
			if (gfp_mask & __GFP_NOFAIL) {
1047
				congestion_wait(WRITE, HZ/50);
Kirill Korotaev's avatar
Kirill Korotaev committed
1048
1049
				goto nofail_alloc;
			}
Linus Torvalds's avatar
Linus Torvalds committed
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
		}
		goto nopage;
	}

	/* Atomic allocations - we can't balance anything */
	if (!wait)
		goto nopage;

rebalance:
	cond_resched();

	/* We now go into synchronous reclaim */
1062
	cpuset_memory_pressure_bump();
Linus Torvalds's avatar
Linus Torvalds committed
1063
1064
1065
1066
	p->flags |= PF_MEMALLOC;
	reclaim_state.reclaimed_slab = 0;
	p->reclaim_state = &reclaim_state;

Rohit Seth's avatar
Rohit Seth committed
1067
	did_some_progress = try_to_free_pages(zonelist->zones, gfp_mask);
Linus Torvalds's avatar
Linus Torvalds committed
1068
1069
1070
1071
1072
1073
1074

	p->reclaim_state = NULL;
	p->flags &= ~PF_MEMALLOC;

	cond_resched();

	if (likely(did_some_progress)) {
Rohit Seth's avatar
Rohit Seth committed
1075
1076
1077
1078
		page = get_page_from_freelist(gfp_mask, order,
						zonelist, alloc_flags);
		if (page)
			goto got_pg;
Linus Torvalds's avatar
Linus Torvalds committed
1079
1080
1081
1082
1083
1084
1085
	} else if ((gfp_mask & __GFP_FS) && !(gfp_mask & __GFP_NORETRY)) {
		/*
		 * Go through the zonelist yet one more time, keep
		 * very high watermark here, this is only to catch
		 * a parallel oom killing, we must fail if we're still
		 * under heavy pressure.
		 */
Rohit Seth's avatar
Rohit Seth committed
1086
		page = get_page_from_freelist(gfp_mask|__GFP_HARDWALL, order,
1087
				zonelist, ALLOC_WMARK_HIGH|ALLOC_CPUSET);
Rohit Seth's avatar
Rohit Seth committed
1088
1089
		if (page)
			goto got_pg;
Linus Torvalds's avatar
Linus Torvalds committed
1090

1091
		out_of_memory(zonelist, gfp_mask, order);
Linus Torvalds's avatar
Linus Torvalds committed
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
		goto restart;
	}

	/*
	 * Don't let big-order allocations loop unless the caller explicitly
	 * requests that.  Wait for some write requests to complete then retry.
	 *
	 * In this implementation, __GFP_REPEAT means __GFP_NOFAIL for order
	 * <= 3, but that may not be true in other implementations.
	 */
	do_retry = 0;
	if (!(gfp_mask & __GFP_NORETRY)) {
		if ((order <= 3) || (gfp_mask & __GFP_REPEAT))
			do_retry = 1;
		if (gfp_mask & __GFP_NOFAIL)
			do_retry = 1;
	}
	if (do_retry) {
1110
		congestion_wait(WRITE, HZ/50);
Linus Torvalds's avatar
Linus Torvalds committed
1111
1112
1113
1114
1115
1116
1117
1118
1119
		goto rebalance;
	}

nopage:
	if (!(gfp_mask & __GFP_NOWARN) && printk_ratelimit()) {
		printk(KERN_WARNING "%s: page allocation failure."
			" order:%d, mode:0x%x\n",
			p->comm, order, gfp_mask);
		dump_stack();
Janet Morgan's avatar
Janet Morgan committed
1120
		show_mem();
Linus Torvalds's avatar
Linus Torvalds committed
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
	}
got_pg:
	return page;
}

EXPORT_SYMBOL(__alloc_pages);

/*
 * Common helper functions.
 */
1131
fastcall unsigned long __get_free_pages(gfp_t gfp_mask, unsigned int order)
Linus Torvalds's avatar
Linus Torvalds committed
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
{
	struct page *</