page_alloc.c 170 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>
Linus Torvalds's avatar
Linus Torvalds committed
61 62

#include <asm/tlbflush.h>
63
#include <asm/div64.h>
Linus Torvalds's avatar
Linus Torvalds committed
64 65
#include "internal.h"

66 67 68 69 70
#ifdef CONFIG_USE_PERCPU_NUMA_NODE_ID
DEFINE_PER_CPU(int, numa_node);
EXPORT_PER_CPU_SYMBOL(numa_node);
#endif

71 72 73 74 75 76 77 78 79 80 81
#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
82
/*
83
 * Array of node states.
Linus Torvalds's avatar
Linus Torvalds committed
84
 */
85 86 87 88 89 90 91
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 } },
92 93 94
#endif
#ifdef CONFIG_MOVABLE_NODE
	[N_MEMORY] = { { [0] = 1UL } },
95 96 97 98 99 100
#endif
	[N_CPU] = { { [0] = 1UL } },
#endif	/* NUMA */
};
EXPORT_SYMBOL(node_states);

101
unsigned long totalram_pages __read_mostly;
102
unsigned long totalreserve_pages __read_mostly;
103 104 105 106 107 108 109 110
/*
 * 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;

111
int percpu_pagelist_fraction;
112
gfp_t gfp_allowed_mask __read_mostly = GFP_BOOT_MASK;
Linus Torvalds's avatar
Linus Torvalds committed
113

114 115 116 117 118 119 120 121 122
#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).
 */
123 124 125 126

static gfp_t saved_gfp_mask;

void pm_restore_gfp_mask(void)
127 128
{
	WARN_ON(!mutex_is_locked(&pm_mutex));
129 130 131 132
	if (saved_gfp_mask) {
		gfp_allowed_mask = saved_gfp_mask;
		saved_gfp_mask = 0;
	}
133 134
}

135
void pm_restrict_gfp_mask(void)
136 137
{
	WARN_ON(!mutex_is_locked(&pm_mutex));
138 139 140
	WARN_ON(saved_gfp_mask);
	saved_gfp_mask = gfp_allowed_mask;
	gfp_allowed_mask &= ~GFP_IOFS;
141
}
142 143 144 145 146 147 148

bool pm_suspended_storage(void)
{
	if ((gfp_allowed_mask & GFP_IOFS) == GFP_IOFS)
		return false;
	return true;
}
149 150
#endif /* CONFIG_PM_SLEEP */

151 152 153 154
#ifdef CONFIG_HUGETLB_PAGE_SIZE_VARIABLE
int pageblock_order __read_mostly;
#endif

155
static void __free_pages_ok(struct page *page, unsigned int order);
156

Linus Torvalds's avatar
Linus Torvalds committed
157 158 159 160 161 162 163
/*
 * 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
164 165 166
 *
 * 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
167
 */
168
int sysctl_lowmem_reserve_ratio[MAX_NR_ZONES-1] = {
169
#ifdef CONFIG_ZONE_DMA
170
	 256,
171
#endif
172
#ifdef CONFIG_ZONE_DMA32
173
	 256,
174
#endif
175
#ifdef CONFIG_HIGHMEM
Mel Gorman's avatar
Mel Gorman committed
176
	 32,
177
#endif
Mel Gorman's avatar
Mel Gorman committed
178
	 32,
179
};
Linus Torvalds's avatar
Linus Torvalds committed
180 181 182

EXPORT_SYMBOL(totalram_pages);

183
static char * const zone_names[MAX_NR_ZONES] = {
184
#ifdef CONFIG_ZONE_DMA
185
	 "DMA",
186
#endif
187
#ifdef CONFIG_ZONE_DMA32
188
	 "DMA32",
189
#endif
190
	 "Normal",
191
#ifdef CONFIG_HIGHMEM
Mel Gorman's avatar
Mel Gorman committed
192
	 "HighMem",
193
#endif
Mel Gorman's avatar
Mel Gorman committed
194
	 "Movable",
195 196
};

Linus Torvalds's avatar
Linus Torvalds committed
197 198
int min_free_kbytes = 1024;

199 200
static unsigned long __meminitdata nr_kernel_pages;
static unsigned long __meminitdata nr_all_pages;
201
static unsigned long __meminitdata dma_reserve;
Linus Torvalds's avatar
Linus Torvalds committed
202

Tejun Heo's avatar
Tejun Heo committed
203 204 205 206 207 208 209 210 211 212 213
#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 */
214

Miklos Szeredi's avatar
Miklos Szeredi committed
215 216
#if MAX_NUMNODES > 1
int nr_node_ids __read_mostly = MAX_NUMNODES;
217
int nr_online_nodes __read_mostly = 1;
Miklos Szeredi's avatar
Miklos Szeredi committed
218
EXPORT_SYMBOL(nr_node_ids);
219
EXPORT_SYMBOL(nr_online_nodes);
Miklos Szeredi's avatar
Miklos Szeredi committed
220 221
#endif

222 223
int page_group_by_mobility_disabled __read_mostly;

224
void set_pageblock_migratetype(struct page *page, int migratetype)
225
{
226 227 228 229

	if (unlikely(page_group_by_mobility_disabled))
		migratetype = MIGRATE_UNMOVABLE;

230 231 232 233
	set_pageblock_flags_group(page, (unsigned long)migratetype,
					PB_migrate, PB_migrate_end);
}

234 235
bool oom_killer_disabled __read_mostly;

Nick Piggin's avatar
Nick Piggin committed
236
#ifdef CONFIG_DEBUG_VM
237
static int page_outside_zone_boundaries(struct zone *zone, struct page *page)
Linus Torvalds's avatar
Linus Torvalds committed
238
{
239 240 241
	int ret = 0;
	unsigned seq;
	unsigned long pfn = page_to_pfn(page);
242

243 244 245 246 247 248 249 250 251
	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;
252 253 254 255
}

static int page_is_consistent(struct zone *zone, struct page *page)
{
256
	if (!pfn_valid_within(page_to_pfn(page)))
257
		return 0;
Linus Torvalds's avatar
Linus Torvalds committed
258
	if (zone != page_zone(page))
259 260 261 262 263 264 265 266 267 268
		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
269
		return 1;
270 271 272
	if (!page_is_consistent(zone, page))
		return 1;

Linus Torvalds's avatar
Linus Torvalds committed
273 274
	return 0;
}
Nick Piggin's avatar
Nick Piggin committed
275 276 277 278 279 280 281
#else
static inline int bad_range(struct zone *zone, struct page *page)
{
	return 0;
}
#endif

282
static void bad_page(struct page *page)
Linus Torvalds's avatar
Linus Torvalds committed
283
{
284 285 286 287
	static unsigned long resume;
	static unsigned long nr_shown;
	static unsigned long nr_unshown;

288 289
	/* Don't complain about poisoned pages */
	if (PageHWPoison(page)) {
290
		reset_page_mapcount(page); /* remove PageBuddy */
291 292 293
		return;
	}

294 295 296 297 298 299 300 301 302 303
	/*
	 * 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) {
304 305
			printk(KERN_ALERT
			      "BUG: Bad page state: %lu messages suppressed\n",
306 307 308 309 310 311 312 313
				nr_unshown);
			nr_unshown = 0;
		}
		nr_shown = 0;
	}
	if (nr_shown++ == 0)
		resume = jiffies + 60 * HZ;

314
	printk(KERN_ALERT "BUG: Bad page state in process %s  pfn:%05lx\n",
315
		current->comm, page_to_pfn(page));
316
	dump_page(page);
317

318
	print_modules();
Linus Torvalds's avatar
Linus Torvalds committed
319
	dump_stack();
320
out:
321
	/* Leave bad fields for debug, except PageBuddy could make trouble */
322
	reset_page_mapcount(page); /* remove PageBuddy */
323
	add_taint(TAINT_BAD_PAGE);
Linus Torvalds's avatar
Linus Torvalds committed
324 325 326 327 328 329 330 331 332
}

/*
 * 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".
 *
333 334
 * 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
335
 *
336 337 338
 * 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
339
 */
340 341 342

static void free_compound_page(struct page *page)
{
343
	__free_pages_ok(page, compound_order(page));
344 345
}

346
void prep_compound_page(struct page *page, unsigned long order)
347 348 349 350 351 352 353 354 355 356
{
	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);
357
		set_page_count(p, 0);
358 359 360 361
		p->first_page = page;
	}
}

362
/* update __split_huge_page_refcount if you change this function */
363
static int destroy_compound_page(struct page *page, unsigned long order)
Linus Torvalds's avatar
Linus Torvalds committed
364 365 366
{
	int i;
	int nr_pages = 1 << order;
367
	int bad = 0;
Linus Torvalds's avatar
Linus Torvalds committed
368

369
	if (unlikely(compound_order(page) != order)) {
370
		bad_page(page);
371 372
		bad++;
	}
Linus Torvalds's avatar
Linus Torvalds committed
373

374
	__ClearPageHead(page);
375

376 377
	for (i = 1; i < nr_pages; i++) {
		struct page *p = page + i;
Linus Torvalds's avatar
Linus Torvalds committed
378

379
		if (unlikely(!PageTail(p) || (p->first_page != page))) {
380
			bad_page(page);
381 382
			bad++;
		}
383
		__ClearPageTail(p);
Linus Torvalds's avatar
Linus Torvalds committed
384
	}
385 386

	return bad;
Linus Torvalds's avatar
Linus Torvalds committed
387 388
}

Nick Piggin's avatar
Nick Piggin committed
389 390 391 392
static inline void prep_zero_page(struct page *page, int order, gfp_t gfp_flags)
{
	int i;

393 394 395 396
	/*
	 * 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
397
	VM_BUG_ON((gfp_flags & __GFP_HIGHMEM) && in_interrupt());
Nick Piggin's avatar
Nick Piggin committed
398 399 400 401
	for (i = 0; i < (1 << order); i++)
		clear_highpage(page + i);
}

402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432
#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

433 434
static inline void set_page_order(struct page *page, int order)
{
435
	set_page_private(page, order);
436
	__SetPageBuddy(page);
Linus Torvalds's avatar
Linus Torvalds committed
437 438 439 440
}

static inline void rmv_page_order(struct page *page)
{
441
	__ClearPageBuddy(page);
442
	set_page_private(page, 0);
Linus Torvalds's avatar
Linus Torvalds committed
443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459
}

/*
 * 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
460
 * Assumption: *_mem_map is contiguous at least up to MAX_ORDER
Linus Torvalds's avatar
Linus Torvalds committed
461 462
 */
static inline unsigned long
463
__find_buddy_index(unsigned long page_idx, unsigned int order)
Linus Torvalds's avatar
Linus Torvalds committed
464
{
465
	return page_idx ^ (1 << order);
Linus Torvalds's avatar
Linus Torvalds committed
466 467 468 469 470
}

/*
 * 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
471
 * (a) the buddy is not in a hole &&
472
 * (b) the buddy is in the buddy system &&
473 474
 * (c) a page and its buddy have the same order &&
 * (d) a page and its buddy are in the same zone.
475
 *
Andrea Arcangeli's avatar
Andrea Arcangeli committed
476 477
 * 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
478
 *
479
 * For recording page's order, we use page_private(page).
Linus Torvalds's avatar
Linus Torvalds committed
480
 */
481 482
static inline int page_is_buddy(struct page *page, struct page *buddy,
								int order)
Linus Torvalds's avatar
Linus Torvalds committed
483
{
484
	if (!pfn_valid_within(page_to_pfn(buddy)))
Nick Piggin's avatar
Nick Piggin committed
485 486
		return 0;

487 488 489
	if (page_zone_id(page) != page_zone_id(buddy))
		return 0;

490 491 492 493 494
	if (page_is_guard(buddy) && page_order(buddy) == order) {
		VM_BUG_ON(page_count(buddy) != 0);
		return 1;
	}

495
	if (PageBuddy(buddy) && page_order(buddy) == order) {
496
		VM_BUG_ON(page_count(buddy) != 0);
497
		return 1;
498
	}
499
	return 0;
Linus Torvalds's avatar
Linus Torvalds committed
500 501 502 503 504 505 506 507 508 509 510 511 512 513 514
}

/*
 * 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
515
 * free pages of length of (1 << order) and marked with _mapcount -2. Page's
516
 * order is recorded in page_private(page) field.
Linus Torvalds's avatar
Linus Torvalds committed
517
 * So when we are allocating or freeing one, we can derive the state of the
518 519
 * 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
520
 * If a block is freed, and its buddy is also free, then this
521
 * triggers coalescing into a block of larger size.
Linus Torvalds's avatar
Linus Torvalds committed
522
 *
523
 * -- nyc
Linus Torvalds's avatar
Linus Torvalds committed
524 525
 */

Nick Piggin's avatar
Nick Piggin committed
526
static inline void __free_one_page(struct page *page,
527 528
		struct zone *zone, unsigned int order,
		int migratetype)
Linus Torvalds's avatar
Linus Torvalds committed
529 530
{
	unsigned long page_idx;
531
	unsigned long combined_idx;
532
	unsigned long uninitialized_var(buddy_idx);
533
	struct page *buddy;
Linus Torvalds's avatar
Linus Torvalds committed
534

535
	if (unlikely(PageCompound(page)))
536 537
		if (unlikely(destroy_compound_page(page, order)))
			return;
Linus Torvalds's avatar
Linus Torvalds committed
538

539 540
	VM_BUG_ON(migratetype == -1);

Linus Torvalds's avatar
Linus Torvalds committed
541 542
	page_idx = page_to_pfn(page) & ((1 << MAX_ORDER) - 1);

543
	VM_BUG_ON(page_idx & ((1 << order) - 1));
Nick Piggin's avatar
Nick Piggin committed
544
	VM_BUG_ON(bad_range(zone, page));
Linus Torvalds's avatar
Linus Torvalds committed
545 546

	while (order < MAX_ORDER-1) {
547 548
		buddy_idx = __find_buddy_index(page_idx, order);
		buddy = page + (buddy_idx - page_idx);
549
		if (!page_is_buddy(page, buddy, order))
550
			break;
551 552 553 554 555 556 557
		/*
		 * 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);
558 559
			__mod_zone_freepage_state(zone, 1 << order,
						  migratetype);
560 561 562 563 564
		} else {
			list_del(&buddy->lru);
			zone->free_area[order].nr_free--;
			rmv_page_order(buddy);
		}
565
		combined_idx = buddy_idx & page_idx;
Linus Torvalds's avatar
Linus Torvalds committed
566 567 568 569 570
		page = page + (combined_idx - page_idx);
		page_idx = combined_idx;
		order++;
	}
	set_page_order(page, order);
571 572 573 574 575 576 577 578 579

	/*
	 * 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
	 */
580
	if ((order < MAX_ORDER-2) && pfn_valid_within(page_to_pfn(buddy))) {
581
		struct page *higher_page, *higher_buddy;
582 583 584
		combined_idx = buddy_idx & page_idx;
		higher_page = page + (combined_idx - page_idx);
		buddy_idx = __find_buddy_index(combined_idx, order + 1);
585
		higher_buddy = higher_page + (buddy_idx - combined_idx);
586 587 588 589 590 591 592 593 594
		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
595 596 597
	zone->free_area[order].nr_free++;
}

598
static inline int free_pages_check(struct page *page)
Linus Torvalds's avatar
Linus Torvalds committed
599
{
Nick Piggin's avatar
Nick Piggin committed
600 601
	if (unlikely(page_mapcount(page) |
		(page->mapping != NULL)  |
602
		(atomic_read(&page->_count) != 0) |
603 604
		(page->flags & PAGE_FLAGS_CHECK_AT_FREE) |
		(mem_cgroup_bad_page_check(page)))) {
605
		bad_page(page);
606
		return 1;
607
	}
608
	reset_page_last_nid(page);
609 610 611
	if (page->flags & PAGE_FLAGS_CHECK_AT_PREP)
		page->flags &= ~PAGE_FLAGS_CHECK_AT_PREP;
	return 0;
Linus Torvalds's avatar
Linus Torvalds committed
612 613 614
}

/*
615
 * Frees a number of pages from the PCP lists
Linus Torvalds's avatar
Linus Torvalds committed
616
 * Assumes all pages on list are in same zone, and of same order.
617
 * count is the number of pages to free.
Linus Torvalds's avatar
Linus Torvalds committed
618 619 620 621 622 623 624
 *
 * 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.
 */
625 626
static void free_pcppages_bulk(struct zone *zone, int count,
					struct per_cpu_pages *pcp)
Linus Torvalds's avatar
Linus Torvalds committed
627
{
628
	int migratetype = 0;
629
	int batch_free = 0;
630
	int to_free = count;
631

Nick Piggin's avatar
Nick Piggin committed
632
	spin_lock(&zone->lock);
633
	zone->all_unreclaimable = 0;
Linus Torvalds's avatar
Linus Torvalds committed
634
	zone->pages_scanned = 0;
635

636
	while (to_free) {
Nick Piggin's avatar
Nick Piggin committed
637
		struct page *page;
638 639 640
		struct list_head *list;

		/*
641 642 643 644 645
		 * 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
646 647
		 */
		do {
648
			batch_free++;
649 650 651 652
			if (++migratetype == MIGRATE_PCPTYPES)
				migratetype = 0;
			list = &pcp->lists[migratetype];
		} while (list_empty(list));
Nick Piggin's avatar
Nick Piggin committed
653

654 655 656 657
		/* This is the only non-empty list. Free them all. */
		if (batch_free == MIGRATE_PCPTYPES)
			batch_free = to_free;

658
		do {
659 660
			int mt;	/* migratetype of the to-be-freed page */

661 662 663
			page = list_entry(list->prev, struct page, lru);
			/* must delete as __free_one_page list manipulates */
			list_del(&page->lru);
664
			mt = get_freepage_migratetype(page);
665
			/* MIGRATE_MOVABLE list may include MIGRATE_RESERVEs */
666 667
			__free_one_page(page, zone, 0, mt);
			trace_mm_page_pcpu_drain(page, 0, mt);
668 669 670 671 672
			if (likely(get_pageblock_migratetype(page) != MIGRATE_ISOLATE)) {
				__mod_zone_page_state(zone, NR_FREE_PAGES, 1);
				if (is_migrate_cma(mt))
					__mod_zone_page_state(zone, NR_FREE_CMA_PAGES, 1);
			}
673
		} while (--to_free && --batch_free && !list_empty(list));
Linus Torvalds's avatar
Linus Torvalds committed
674
	}
Nick Piggin's avatar
Nick Piggin committed
675
	spin_unlock(&zone->lock);
Linus Torvalds's avatar
Linus Torvalds committed
676 677
}

678 679
static void free_one_page(struct zone *zone, struct page *page, int order,
				int migratetype)
Linus Torvalds's avatar
Linus Torvalds committed
680
{
681
	spin_lock(&zone->lock);
682
	zone->all_unreclaimable = 0;
683
	zone->pages_scanned = 0;
684

685
	__free_one_page(page, zone, order, migratetype);
686
	if (unlikely(migratetype != MIGRATE_ISOLATE))
687
		__mod_zone_freepage_state(zone, 1 << order, migratetype);
688
	spin_unlock(&zone->lock);
Nick Piggin's avatar
Nick Piggin committed
689 690
}

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

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

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

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

714 715 716 717 718 719
	return true;
}

static void __free_pages_ok(struct page *page, unsigned int order)
{
	unsigned long flags;
720
	int migratetype;
721 722 723 724

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

Nick Piggin's avatar
Nick Piggin committed
725
	local_irq_save(flags);
726
	__count_vm_events(PGFREE, 1 << order);
727 728 729
	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
730
	local_irq_restore(flags);
Linus Torvalds's avatar
Linus Torvalds committed
731 732
}

733 734 735 736 737 738 739
/*
 * 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.
 */
740
void __meminit __free_pages_bootmem(struct page *page, unsigned int order)
741
{
742 743
	unsigned int nr_pages = 1 << order;
	unsigned int loop;
744

745 746 747 748 749 750 751 752
	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);
753
	}
754

755
	page_zone(page)->managed_pages += 1 << order;
756 757
	set_page_refcounted(page);
	__free_pages(page, order);
758 759
}

760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777
#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;
}
#endif
Linus Torvalds's avatar
Linus Torvalds committed
778 779 780 781 782 783 784 785 786 787 788 789 790

/*
 * 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.
 *
791
 * -- nyc
Linus Torvalds's avatar
Linus Torvalds committed
792
 */
Nick Piggin's avatar
Nick Piggin committed
793
static inline void expand(struct zone *zone, struct page *page,
794 795
	int low, int high, struct free_area *area,
	int migratetype)
Linus Torvalds's avatar
Linus Torvalds committed
796 797 798 799 800 801 802
{
	unsigned long size = 1 << high;

	while (high > low) {
		area--;
		high--;
		size >>= 1;
Nick Piggin's avatar
Nick Piggin committed
803
		VM_BUG_ON(bad_range(zone, &page[size]));
804 805 806 807 808 809 810 811 812 813 814 815 816

#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 */
817 818
			__mod_zone_freepage_state(zone, -(1 << high),
						  migratetype);
819 820 821
			continue;
		}
#endif
822
		list_add(&page[size].lru, &area->free_list[migratetype]);
Linus Torvalds's avatar
Linus Torvalds committed
823 824 825 826 827 828 829 830
		area->nr_free++;
		set_page_order(&page[size], high);
	}
}

/*
 * This page is about to be returned from the page allocator
 */
831
static inline int check_new_page(struct page *page)
Linus Torvalds's avatar
Linus Torvalds committed
832
{
Nick Piggin's avatar
Nick Piggin committed
833 834
	if (unlikely(page_mapcount(page) |
		(page->mapping != NULL)  |
835
		(atomic_read(&page->_count) != 0)  |
836 837
		(page->flags & PAGE_FLAGS_CHECK_AT_PREP) |
		(mem_cgroup_bad_page_check(page)))) {
838
		bad_page(page);
839
		return 1;
840
	}
841 842 843 844 845 846 847 848 849 850 851 852
	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;
	}
853

854
	set_page_private(page, 0);
855
	set_page_refcounted(page);
Nick Piggin's avatar
Nick Piggin committed
856 857

	arch_alloc_page(page, order);
Linus Torvalds's avatar
Linus Torvalds committed
858
	kernel_map_pages(page, 1 << order, 1);
Nick Piggin's avatar
Nick Piggin committed
859 860 861 862 863 864 865

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

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

866
	return 0;
Linus Torvalds's avatar
Linus Torvalds committed
867 868
}

869 870 871 872
/*
 * Go through the free lists for the given migratetype and remove
 * the smallest available page from the freelists
 */
873 874
static inline
struct page *__rmqueue_smallest(struct zone *zone, unsigned int order,
875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899
						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;
}


900 901 902 903
/*
 * This array describes the order lists are fallen back to when
 * the free lists for the desirable migrate type are depleted
 */
904 905 906 907 908 909 910 911 912
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
913 914
	[MIGRATE_RESERVE]     = { MIGRATE_RESERVE }, /* Never used */
	[MIGRATE_ISOLATE]     = { MIGRATE_RESERVE }, /* Never used */
915 916
};

917 918
/*
 * Move the free pages in a range to the free lists of the requested type.
919
 * Note that start_page and end_pages are not aligned on a pageblock
920 921
 * boundary. If alignment is required, use move_freepages_block()
 */
922
int move_freepages(struct zone *zone,
Adrian Bunk's avatar
Adrian Bunk committed
923 924
			  struct page *start_page, struct page *end_page,
			  int migratetype)
925 926 927
{
	struct page *page;
	unsigned long order;
928
	int pages_moved = 0;
929 930 931 932 933 934 935

#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
936
	 * grouping pages by mobility
937 938 939 940 941
	 */
	BUG_ON(page_zone(start_page) != page_zone(end_page));
#endif

	for (page = start_page; page <= end_page;) {
942 943 944
		/* Make sure we are not inadvertently changing nodes */
		VM_BUG_ON(page_to_nid(page) != zone_to_nid(zone));

945 946 947 948 949 950 951 952 953 954 955
		if (!pfn_valid_within(page_to_pfn(page))) {
			page++;
			continue;
		}

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

		order = page_order(page);