page_alloc.c 172 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 71 72
#ifdef CONFIG_USE_PERCPU_NUMA_NODE_ID
DEFINE_PER_CPU(int, numa_node);
EXPORT_PER_CPU_SYMBOL(numa_node);
#endif

73 74 75 76 77 78 79 80 81 82 83
#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
84
/*
85
 * Array of node states.
Linus Torvalds's avatar
Linus Torvalds committed
86
 */
87 88 89 90 91 92 93
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 } },
94 95 96
#endif
#ifdef CONFIG_MOVABLE_NODE
	[N_MEMORY] = { { [0] = 1UL } },
97 98 99 100 101 102
#endif
	[N_CPU] = { { [0] = 1UL } },
#endif	/* NUMA */
};
EXPORT_SYMBOL(node_states);

103
unsigned long totalram_pages __read_mostly;
104
unsigned long totalreserve_pages __read_mostly;
105 106 107 108 109 110 111 112
/*
 * 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;

113
int percpu_pagelist_fraction;
114
gfp_t gfp_allowed_mask __read_mostly = GFP_BOOT_MASK;
Linus Torvalds's avatar
Linus Torvalds committed
115

116 117 118 119 120 121 122 123 124
#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).
 */
125 126 127 128

static gfp_t saved_gfp_mask;

void pm_restore_gfp_mask(void)
129 130
{
	WARN_ON(!mutex_is_locked(&pm_mutex));
131 132 133 134
	if (saved_gfp_mask) {
		gfp_allowed_mask = saved_gfp_mask;
		saved_gfp_mask = 0;
	}
135 136
}

137
void pm_restrict_gfp_mask(void)
138 139
{
	WARN_ON(!mutex_is_locked(&pm_mutex));
140 141 142
	WARN_ON(saved_gfp_mask);
	saved_gfp_mask = gfp_allowed_mask;
	gfp_allowed_mask &= ~GFP_IOFS;
143
}
144 145 146 147 148 149 150

bool pm_suspended_storage(void)
{
	if ((gfp_allowed_mask & GFP_IOFS) == GFP_IOFS)
		return false;
	return true;
}
151 152
#endif /* CONFIG_PM_SLEEP */

153 154 155 156
#ifdef CONFIG_HUGETLB_PAGE_SIZE_VARIABLE
int pageblock_order __read_mostly;
#endif

157
static void __free_pages_ok(struct page *page, unsigned int order);
158

Linus Torvalds's avatar
Linus Torvalds committed
159 160 161 162 163 164 165
/*
 * 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
166 167 168
 *
 * 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
169
 */
170
int sysctl_lowmem_reserve_ratio[MAX_NR_ZONES-1] = {
171
#ifdef CONFIG_ZONE_DMA
172
	 256,
173
#endif
174
#ifdef CONFIG_ZONE_DMA32
175
	 256,
176
#endif
177
#ifdef CONFIG_HIGHMEM
Mel Gorman's avatar
Mel Gorman committed
178
	 32,
179
#endif
Mel Gorman's avatar
Mel Gorman committed
180
	 32,
181
};
Linus Torvalds's avatar
Linus Torvalds committed
182 183 184

EXPORT_SYMBOL(totalram_pages);

185
static char * const zone_names[MAX_NR_ZONES] = {
186
#ifdef CONFIG_ZONE_DMA
187
	 "DMA",
188
#endif
189
#ifdef CONFIG_ZONE_DMA32
190
	 "DMA32",
191
#endif
192
	 "Normal",
193
#ifdef CONFIG_HIGHMEM
Mel Gorman's avatar
Mel Gorman committed
194
	 "HighMem",
195
#endif
Mel Gorman's avatar
Mel Gorman committed
196
	 "Movable",
197 198
};

Linus Torvalds's avatar
Linus Torvalds committed
199 200
int min_free_kbytes = 1024;

201 202
static unsigned long __meminitdata nr_kernel_pages;
static unsigned long __meminitdata nr_all_pages;
203
static unsigned long __meminitdata dma_reserve;
Linus Torvalds's avatar
Linus Torvalds committed
204

Tejun Heo's avatar
Tejun Heo committed
205 206 207 208 209 210 211 212 213 214 215
#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 */
216

Miklos Szeredi's avatar
Miklos Szeredi committed
217 218
#if MAX_NUMNODES > 1
int nr_node_ids __read_mostly = MAX_NUMNODES;
219
int nr_online_nodes __read_mostly = 1;
Miklos Szeredi's avatar
Miklos Szeredi committed
220
EXPORT_SYMBOL(nr_node_ids);
221
EXPORT_SYMBOL(nr_online_nodes);
Miklos Szeredi's avatar
Miklos Szeredi committed
222 223
#endif

224 225
int page_group_by_mobility_disabled __read_mostly;

226
void set_pageblock_migratetype(struct page *page, int migratetype)
227
{
228 229 230 231

	if (unlikely(page_group_by_mobility_disabled))
		migratetype = MIGRATE_UNMOVABLE;

232 233 234 235
	set_pageblock_flags_group(page, (unsigned long)migratetype,
					PB_migrate, PB_migrate_end);
}

236 237
bool oom_killer_disabled __read_mostly;

Nick Piggin's avatar
Nick Piggin committed
238
#ifdef CONFIG_DEBUG_VM
239
static int page_outside_zone_boundaries(struct zone *zone, struct page *page)
Linus Torvalds's avatar
Linus Torvalds committed
240
{
241 242 243
	int ret = 0;
	unsigned seq;
	unsigned long pfn = page_to_pfn(page);
244
	unsigned long sp, start_pfn;
245

246 247
	do {
		seq = zone_span_seqbegin(zone);
248 249
		start_pfn = zone->zone_start_pfn;
		sp = zone->spanned_pages;
250
		if (!zone_spans_pfn(zone, pfn))
251 252 253
			ret = 1;
	} while (zone_span_seqretry(zone, seq));

254 255 256 257
	if (ret)
		pr_err("page %lu outside zone [ %lu - %lu ]\n",
			pfn, start_pfn, start_pfn + sp);

258
	return ret;
259 260 261 262
}

static int page_is_consistent(struct zone *zone, struct page *page)
{
263
	if (!pfn_valid_within(page_to_pfn(page)))
264
		return 0;
Linus Torvalds's avatar
Linus Torvalds committed
265
	if (zone != page_zone(page))
266 267 268 269 270 271 272 273 274 275
		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
276
		return 1;
277 278 279
	if (!page_is_consistent(zone, page))
		return 1;

Linus Torvalds's avatar
Linus Torvalds committed
280 281
	return 0;
}
Nick Piggin's avatar
Nick Piggin committed
282 283 284 285 286 287 288
#else
static inline int bad_range(struct zone *zone, struct page *page)
{
	return 0;
}
#endif

289
static void bad_page(struct page *page)
Linus Torvalds's avatar
Linus Torvalds committed
290
{
291 292 293 294
	static unsigned long resume;
	static unsigned long nr_shown;
	static unsigned long nr_unshown;

295 296
	/* Don't complain about poisoned pages */
	if (PageHWPoison(page)) {
297
		page_mapcount_reset(page); /* remove PageBuddy */
298 299 300
		return;
	}

301 302 303 304 305 306 307 308 309 310
	/*
	 * 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) {
311 312
			printk(KERN_ALERT
			      "BUG: Bad page state: %lu messages suppressed\n",
313 314 315 316 317 318 319 320
				nr_unshown);
			nr_unshown = 0;
		}
		nr_shown = 0;
	}
	if (nr_shown++ == 0)
		resume = jiffies + 60 * HZ;

321
	printk(KERN_ALERT "BUG: Bad page state in process %s  pfn:%05lx\n",
322
		current->comm, page_to_pfn(page));
323
	dump_page(page);
324

325
	print_modules();
Linus Torvalds's avatar
Linus Torvalds committed
326
	dump_stack();
327
out:
328
	/* Leave bad fields for debug, except PageBuddy could make trouble */
329
	page_mapcount_reset(page); /* remove PageBuddy */
330
	add_taint(TAINT_BAD_PAGE, LOCKDEP_NOW_UNRELIABLE);
Linus Torvalds's avatar
Linus Torvalds committed
331 332 333 334 335 336 337 338 339
}

/*
 * 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".
 *
340 341
 * 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
342
 *
343 344 345
 * 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
346
 */
347 348 349

static void free_compound_page(struct page *page)
{
350
	__free_pages_ok(page, compound_order(page));
351 352
}

353
void prep_compound_page(struct page *page, unsigned long order)
354 355 356 357 358 359 360 361 362 363
{
	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);
364
		set_page_count(p, 0);
365 366 367 368
		p->first_page = page;
	}
}

369
/* update __split_huge_page_refcount if you change this function */
370
static int destroy_compound_page(struct page *page, unsigned long order)
Linus Torvalds's avatar
Linus Torvalds committed
371 372 373
{
	int i;
	int nr_pages = 1 << order;
374
	int bad = 0;
Linus Torvalds's avatar
Linus Torvalds committed
375

376
	if (unlikely(compound_order(page) != order)) {
377
		bad_page(page);
378 379
		bad++;
	}
Linus Torvalds's avatar
Linus Torvalds committed
380

381
	__ClearPageHead(page);
382

383 384
	for (i = 1; i < nr_pages; i++) {
		struct page *p = page + i;
Linus Torvalds's avatar
Linus Torvalds committed
385

386
		if (unlikely(!PageTail(p) || (p->first_page != page))) {
387
			bad_page(page);
388 389
			bad++;
		}
390
		__ClearPageTail(p);
Linus Torvalds's avatar
Linus Torvalds committed
391
	}
392 393

	return bad;
Linus Torvalds's avatar
Linus Torvalds committed
394 395
}

Nick Piggin's avatar
Nick Piggin committed
396 397 398 399
static inline void prep_zero_page(struct page *page, int order, gfp_t gfp_flags)
{
	int i;

400 401 402 403
	/*
	 * 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
404
	VM_BUG_ON((gfp_flags & __GFP_HIGHMEM) && in_interrupt());
Nick Piggin's avatar
Nick Piggin committed
405 406 407 408
	for (i = 0; i < (1 << order); i++)
		clear_highpage(page + i);
}

409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439
#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

440 441
static inline void set_page_order(struct page *page, int order)
{
442
	set_page_private(page, order);
443
	__SetPageBuddy(page);
Linus Torvalds's avatar
Linus Torvalds committed
444 445 446 447
}

static inline void rmv_page_order(struct page *page)
{
448
	__ClearPageBuddy(page);
449
	set_page_private(page, 0);
Linus Torvalds's avatar
Linus Torvalds committed
450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466
}

/*
 * 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
467
 * Assumption: *_mem_map is contiguous at least up to MAX_ORDER
Linus Torvalds's avatar
Linus Torvalds committed
468 469
 */
static inline unsigned long
470
__find_buddy_index(unsigned long page_idx, unsigned int order)
Linus Torvalds's avatar
Linus Torvalds committed
471
{
472
	return page_idx ^ (1 << order);
Linus Torvalds's avatar
Linus Torvalds committed
473 474 475 476 477
}

/*
 * 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
478
 * (a) the buddy is not in a hole &&
479
 * (b) the buddy is in the buddy system &&
480 481
 * (c) a page and its buddy have the same order &&
 * (d) a page and its buddy are in the same zone.
482
 *
Andrea Arcangeli's avatar
Andrea Arcangeli committed
483 484
 * 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
485
 *
486
 * For recording page's order, we use page_private(page).
Linus Torvalds's avatar
Linus Torvalds committed
487
 */
488 489
static inline int page_is_buddy(struct page *page, struct page *buddy,
								int order)
Linus Torvalds's avatar
Linus Torvalds committed
490
{
491
	if (!pfn_valid_within(page_to_pfn(buddy)))
Nick Piggin's avatar
Nick Piggin committed
492 493
		return 0;

494 495 496
	if (page_zone_id(page) != page_zone_id(buddy))
		return 0;

497 498 499 500 501
	if (page_is_guard(buddy) && page_order(buddy) == order) {
		VM_BUG_ON(page_count(buddy) != 0);
		return 1;
	}

502
	if (PageBuddy(buddy) && page_order(buddy) == order) {
503
		VM_BUG_ON(page_count(buddy) != 0);
504
		return 1;
505
	}
506
	return 0;
Linus Torvalds's avatar
Linus Torvalds committed
507 508 509 510 511 512 513 514 515 516 517 518 519 520 521
}

/*
 * 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
522
 * free pages of length of (1 << order) and marked with _mapcount -2. Page's
523
 * order is recorded in page_private(page) field.
Linus Torvalds's avatar
Linus Torvalds committed
524
 * So when we are allocating or freeing one, we can derive the state of the
525 526
 * 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
527
 * If a block is freed, and its buddy is also free, then this
528
 * triggers coalescing into a block of larger size.
Linus Torvalds's avatar
Linus Torvalds committed
529
 *
530
 * -- nyc
Linus Torvalds's avatar
Linus Torvalds committed
531 532
 */

Nick Piggin's avatar
Nick Piggin committed
533
static inline void __free_one_page(struct page *page,
534 535
		struct zone *zone, unsigned int order,
		int migratetype)
Linus Torvalds's avatar
Linus Torvalds committed
536 537
{
	unsigned long page_idx;
538
	unsigned long combined_idx;
539
	unsigned long uninitialized_var(buddy_idx);
540
	struct page *buddy;
Linus Torvalds's avatar
Linus Torvalds committed
541

542 543
	VM_BUG_ON(!zone_is_initialized(zone));

544
	if (unlikely(PageCompound(page)))
545 546
		if (unlikely(destroy_compound_page(page, order)))
			return;
Linus Torvalds's avatar
Linus Torvalds committed
547

548 549
	VM_BUG_ON(migratetype == -1);

Linus Torvalds's avatar
Linus Torvalds committed
550 551
	page_idx = page_to_pfn(page) & ((1 << MAX_ORDER) - 1);

552
	VM_BUG_ON(page_idx & ((1 << order) - 1));
Nick Piggin's avatar
Nick Piggin committed
553
	VM_BUG_ON(bad_range(zone, page));
Linus Torvalds's avatar
Linus Torvalds committed
554 555

	while (order < MAX_ORDER-1) {
556 557
		buddy_idx = __find_buddy_index(page_idx, order);
		buddy = page + (buddy_idx - page_idx);
558
		if (!page_is_buddy(page, buddy, order))
559
			break;
560 561 562 563 564 565 566
		/*
		 * 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);
567 568
			__mod_zone_freepage_state(zone, 1 << order,
						  migratetype);
569 570 571 572 573
		} else {
			list_del(&buddy->lru);
			zone->free_area[order].nr_free--;
			rmv_page_order(buddy);
		}
574
		combined_idx = buddy_idx & page_idx;
Linus Torvalds's avatar
Linus Torvalds committed
575 576 577 578 579
		page = page + (combined_idx - page_idx);
		page_idx = combined_idx;
		order++;
	}
	set_page_order(page, order);
580 581 582 583 584 585 586 587 588

	/*
	 * 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
	 */
589
	if ((order < MAX_ORDER-2) && pfn_valid_within(page_to_pfn(buddy))) {
590
		struct page *higher_page, *higher_buddy;
591 592 593
		combined_idx = buddy_idx & page_idx;
		higher_page = page + (combined_idx - page_idx);
		buddy_idx = __find_buddy_index(combined_idx, order + 1);
594
		higher_buddy = higher_page + (buddy_idx - combined_idx);
595 596 597 598 599 600 601 602 603
		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
604 605 606
	zone->free_area[order].nr_free++;
}

607
static inline int free_pages_check(struct page *page)
Linus Torvalds's avatar
Linus Torvalds committed
608
{
Nick Piggin's avatar
Nick Piggin committed
609 610
	if (unlikely(page_mapcount(page) |
		(page->mapping != NULL)  |
611
		(atomic_read(&page->_count) != 0) |
612 613
		(page->flags & PAGE_FLAGS_CHECK_AT_FREE) |
		(mem_cgroup_bad_page_check(page)))) {
614
		bad_page(page);
615
		return 1;
616
	}
617
	page_nid_reset_last(page);
618 619 620
	if (page->flags & PAGE_FLAGS_CHECK_AT_PREP)
		page->flags &= ~PAGE_FLAGS_CHECK_AT_PREP;
	return 0;
Linus Torvalds's avatar
Linus Torvalds committed
621 622 623
}

/*
624
 * Frees a number of pages from the PCP lists
Linus Torvalds's avatar
Linus Torvalds committed
625
 * Assumes all pages on list are in same zone, and of same order.
626
 * count is the number of pages to free.
Linus Torvalds's avatar
Linus Torvalds committed
627 628 629 630 631 632 633
 *
 * 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.
 */
634 635
static void free_pcppages_bulk(struct zone *zone, int count,
					struct per_cpu_pages *pcp)
Linus Torvalds's avatar
Linus Torvalds committed
636
{
637
	int migratetype = 0;
638
	int batch_free = 0;
639
	int to_free = count;
640

Nick Piggin's avatar
Nick Piggin committed
641
	spin_lock(&zone->lock);
642
	zone->all_unreclaimable = 0;
Linus Torvalds's avatar
Linus Torvalds committed
643
	zone->pages_scanned = 0;
644

645
	while (to_free) {
Nick Piggin's avatar
Nick Piggin committed
646
		struct page *page;
647 648 649
		struct list_head *list;

		/*
650 651 652 653 654
		 * 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
655 656
		 */
		do {
657
			batch_free++;
658 659 660 661
			if (++migratetype == MIGRATE_PCPTYPES)
				migratetype = 0;
			list = &pcp->lists[migratetype];
		} while (list_empty(list));
Nick Piggin's avatar
Nick Piggin committed
662

663 664 665 666
		/* This is the only non-empty list. Free them all. */
		if (batch_free == MIGRATE_PCPTYPES)
			batch_free = to_free;

667
		do {
668 669
			int mt;	/* migratetype of the to-be-freed page */

670 671 672
			page = list_entry(list->prev, struct page, lru);
			/* must delete as __free_one_page list manipulates */
			list_del(&page->lru);
673
			mt = get_freepage_migratetype(page);
674
			/* MIGRATE_MOVABLE list may include MIGRATE_RESERVEs */
675 676
			__free_one_page(page, zone, 0, mt);
			trace_mm_page_pcpu_drain(page, 0, mt);
677
			if (likely(!is_migrate_isolate_page(page))) {
678 679 680 681
				__mod_zone_page_state(zone, NR_FREE_PAGES, 1);
				if (is_migrate_cma(mt))
					__mod_zone_page_state(zone, NR_FREE_CMA_PAGES, 1);
			}
682
		} while (--to_free && --batch_free && !list_empty(list));
Linus Torvalds's avatar
Linus Torvalds committed
683
	}
Nick Piggin's avatar
Nick Piggin committed
684
	spin_unlock(&zone->lock);
Linus Torvalds's avatar
Linus Torvalds committed
685 686
}

687 688
static void free_one_page(struct zone *zone, struct page *page, int order,
				int migratetype)
Linus Torvalds's avatar
Linus Torvalds committed
689
{
690
	spin_lock(&zone->lock);
691
	zone->all_unreclaimable = 0;
692
	zone->pages_scanned = 0;
693

694
	__free_one_page(page, zone, order, migratetype);
695
	if (unlikely(!is_migrate_isolate(migratetype)))
696
		__mod_zone_freepage_state(zone, 1 << order, migratetype);
697
	spin_unlock(&zone->lock);
Nick Piggin's avatar
Nick Piggin committed
698 699
}

700
static bool free_pages_prepare(struct page *page, unsigned int order)
Nick Piggin's avatar
Nick Piggin committed
701
{
Linus Torvalds's avatar
Linus Torvalds committed
702
	int i;
703
	int bad = 0;
Linus Torvalds's avatar
Linus Torvalds committed
704

705
	trace_mm_page_free(page, order);
706 707
	kmemcheck_free_shadow(page, order);

Andrea Arcangeli's avatar
Andrea Arcangeli committed
708 709 710 711
	if (PageAnon(page))
		page->mapping = NULL;
	for (i = 0; i < (1 << order); i++)
		bad += free_pages_check(page + i);
712
	if (bad)
713
		return false;
714

715
	if (!PageHighMem(page)) {
Nick Piggin's avatar
Nick Piggin committed
716
		debug_check_no_locks_freed(page_address(page),PAGE_SIZE<<order);
717 718 719
		debug_check_no_obj_freed(page_address(page),
					   PAGE_SIZE << order);
	}
Nick Piggin's avatar
Nick Piggin committed
720
	arch_free_page(page, order);
Nick Piggin's avatar
Nick Piggin committed
721
	kernel_map_pages(page, 1 << order, 0);
Nick Piggin's avatar
Nick Piggin committed
722

723 724 725 726 727 728
	return true;
}

static void __free_pages_ok(struct page *page, unsigned int order)
{
	unsigned long flags;
729
	int migratetype;
730 731 732 733

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

Nick Piggin's avatar
Nick Piggin committed
734
	local_irq_save(flags);
735
	__count_vm_events(PGFREE, 1 << order);
736 737 738
	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
739
	local_irq_restore(flags);
Linus Torvalds's avatar
Linus Torvalds committed
740 741
}

742 743 744 745 746 747 748
/*
 * 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.
 */
749
void __meminit __free_pages_bootmem(struct page *page, unsigned int order)
750
{
751 752
	unsigned int nr_pages = 1 << order;
	unsigned int loop;
753

754 755 756 757 758 759 760 761
	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);
762
	}
763

764
	page_zone(page)->managed_pages += 1 << order;
765 766
	set_page_refcounted(page);
	__free_pages(page, order);
767 768
}

769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784
#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;
785 786 787 788
#ifdef CONFIG_HIGHMEM
	if (PageHighMem(page))
		totalhigh_pages += pageblock_nr_pages;
#endif
789 790
}
#endif
Linus Torvalds's avatar
Linus Torvalds committed
791 792 793 794 795 796 797 798 799 800 801 802 803

/*
 * 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.
 *
804
 * -- nyc
Linus Torvalds's avatar
Linus Torvalds committed
805
 */
Nick Piggin's avatar
Nick Piggin committed
806
static inline void expand(struct zone *zone, struct page *page,
807 808
	int low, int high, struct free_area *area,
	int migratetype)
Linus Torvalds's avatar
Linus Torvalds committed
809 810 811 812 813 814 815
{
	unsigned long size = 1 << high;

	while (high > low) {
		area--;
		high--;
		size >>= 1;
Nick Piggin's avatar
Nick Piggin committed
816
		VM_BUG_ON(bad_range(zone, &page[size]));
817 818 819 820 821 822 823 824 825 826 827 828 829

#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 */
830 831
			__mod_zone_freepage_state(zone, -(1 << high),
						  migratetype);
832 833 834
			continue;
		}
#endif
835
		list_add(&page[size].lru, &area->free_list[migratetype]);
Linus Torvalds's avatar
Linus Torvalds committed
836 837 838 839 840 841 842 843
		area->nr_free++;
		set_page_order(&page[size], high);
	}
}

/*
 * This page is about to be returned from the page allocator
 */
844
static inline int check_new_page(struct page *page)
Linus Torvalds's avatar
Linus Torvalds committed
845
{
Nick Piggin's avatar
Nick Piggin committed
846 847
	if (unlikely(page_mapcount(page) |
		(page->mapping != NULL)  |
848
		(atomic_read(&page->_count) != 0)  |
849 850
		(page->flags & PAGE_FLAGS_CHECK_AT_PREP) |
		(mem_cgroup_bad_page_check(page)))) {
851
		bad_page(page);
852
		return 1;
853
	}
854 855 856 857 858 859 860 861 862 863 864 865
	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;
	}
866

867
	set_page_private(page, 0);
868
	set_page_refcounted(page);
Nick Piggin's avatar
Nick Piggin committed
869 870

	arch_alloc_page(page, order);
Linus Torvalds's avatar
Linus Torvalds committed
871
	kernel_map_pages(page, 1 << order, 1);
Nick Piggin's avatar
Nick Piggin committed
872 873 874 875 876 877 878

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

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

879
	return 0;
Linus Torvalds's avatar
Linus Torvalds committed
880 881
}

882 883 884 885
/*
 * Go through the free lists for the given migratetype and remove
 * the smallest available page from the freelists
 */
886 887
static inline
struct page *__rmqueue_smallest(struct zone *zone, unsigned int order,
888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912
						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;
}


913 914 915 916
/*
 * This array describes the order lists are fallen back to when
 * the free lists for the desirable migrate type are depleted
 */
917 918 919 920 921 922 923 924 925
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
926
	[MIGRATE_RESERVE]     = { MIGRATE_RESERVE }, /* Never used */
927
#ifdef CONFIG_MEMORY_ISOLATION
928
	[MIGRATE_ISOLATE]     = { MIGRATE_RESERVE }, /* Never used */
929
#endif
930 931
};

932 933
/*
 * Move the free pages in a range to the free lists of the requested type.
934
 * Note that start_page and end_pages are not aligned on a pageblock
935 936