page_alloc.c 183 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/mm_inline.h>
60
#include <linux/migrate.h>
61
#include <linux/page-debug-flags.h>
62
#include <linux/hugetlb.h>
63
#include <linux/sched/rt.h>
Linus Torvalds's avatar
Linus Torvalds committed
64

65
#include <asm/sections.h>
Linus Torvalds's avatar
Linus Torvalds committed
66
#include <asm/tlbflush.h>
67
#include <asm/div64.h>
Linus Torvalds's avatar
Linus Torvalds committed
68 69
#include "internal.h"

70 71 72
/* prevent >1 _updater_ of zone percpu pageset ->high and ->batch fields */
static DEFINE_MUTEX(pcp_batch_high_lock);

73 74 75 76 77
#ifdef CONFIG_USE_PERCPU_NUMA_NODE_ID
DEFINE_PER_CPU(int, numa_node);
EXPORT_PER_CPU_SYMBOL(numa_node);
#endif

78 79 80 81 82 83 84 85 86 87 88
#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
89
/*
90
 * Array of node states.
Linus Torvalds's avatar
Linus Torvalds committed
91
 */
92 93 94 95 96 97 98
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 } },
99 100 101
#endif
#ifdef CONFIG_MOVABLE_NODE
	[N_MEMORY] = { { [0] = 1UL } },
102 103 104 105 106 107
#endif
	[N_CPU] = { { [0] = 1UL } },
#endif	/* NUMA */
};
EXPORT_SYMBOL(node_states);

108 109 110
/* Protect totalram_pages and zone->managed_pages */
static DEFINE_SPINLOCK(managed_page_count_lock);

111
unsigned long totalram_pages __read_mostly;
112
unsigned long totalreserve_pages __read_mostly;
113 114 115 116 117 118 119 120
/*
 * 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;

121
int percpu_pagelist_fraction;
122
gfp_t gfp_allowed_mask __read_mostly = GFP_BOOT_MASK;
Linus Torvalds's avatar
Linus Torvalds committed
123

124 125 126 127 128 129 130 131 132
#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).
 */
133 134 135 136

static gfp_t saved_gfp_mask;

void pm_restore_gfp_mask(void)
137 138
{
	WARN_ON(!mutex_is_locked(&pm_mutex));
139 140 141 142
	if (saved_gfp_mask) {
		gfp_allowed_mask = saved_gfp_mask;
		saved_gfp_mask = 0;
	}
143 144
}

145
void pm_restrict_gfp_mask(void)
146 147
{
	WARN_ON(!mutex_is_locked(&pm_mutex));
148 149 150
	WARN_ON(saved_gfp_mask);
	saved_gfp_mask = gfp_allowed_mask;
	gfp_allowed_mask &= ~GFP_IOFS;
151
}
152 153 154 155 156 157 158

bool pm_suspended_storage(void)
{
	if ((gfp_allowed_mask & GFP_IOFS) == GFP_IOFS)
		return false;
	return true;
}
159 160
#endif /* CONFIG_PM_SLEEP */

161 162 163 164
#ifdef CONFIG_HUGETLB_PAGE_SIZE_VARIABLE
int pageblock_order __read_mostly;
#endif

165
static void __free_pages_ok(struct page *page, unsigned int order);
166

Linus Torvalds's avatar
Linus Torvalds committed
167 168 169 170 171 172 173
/*
 * 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
174 175 176
 *
 * 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
177
 */
178
int sysctl_lowmem_reserve_ratio[MAX_NR_ZONES-1] = {
179
#ifdef CONFIG_ZONE_DMA
180
	 256,
181
#endif
182
#ifdef CONFIG_ZONE_DMA32
183
	 256,
184
#endif
185
#ifdef CONFIG_HIGHMEM
Mel Gorman's avatar
Mel Gorman committed
186
	 32,
187
#endif
Mel Gorman's avatar
Mel Gorman committed
188
	 32,
189
};
Linus Torvalds's avatar
Linus Torvalds committed
190 191 192

EXPORT_SYMBOL(totalram_pages);

193
static char * const zone_names[MAX_NR_ZONES] = {
194
#ifdef CONFIG_ZONE_DMA
195
	 "DMA",
196
#endif
197
#ifdef CONFIG_ZONE_DMA32
198
	 "DMA32",
199
#endif
200
	 "Normal",
201
#ifdef CONFIG_HIGHMEM
Mel Gorman's avatar
Mel Gorman committed
202
	 "HighMem",
203
#endif
Mel Gorman's avatar
Mel Gorman committed
204
	 "Movable",
205 206
};

Linus Torvalds's avatar
Linus Torvalds committed
207
int min_free_kbytes = 1024;
208
int user_min_free_kbytes = -1;
Linus Torvalds's avatar
Linus Torvalds committed
209

210 211
static unsigned long __meminitdata nr_kernel_pages;
static unsigned long __meminitdata nr_all_pages;
212
static unsigned long __meminitdata dma_reserve;
Linus Torvalds's avatar
Linus Torvalds committed
213

Tejun Heo's avatar
Tejun Heo committed
214 215 216 217 218 219 220 221 222 223 224
#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 */
225

Miklos Szeredi's avatar
Miklos Szeredi committed
226 227
#if MAX_NUMNODES > 1
int nr_node_ids __read_mostly = MAX_NUMNODES;
228
int nr_online_nodes __read_mostly = 1;
Miklos Szeredi's avatar
Miklos Szeredi committed
229
EXPORT_SYMBOL(nr_node_ids);
230
EXPORT_SYMBOL(nr_online_nodes);
Miklos Szeredi's avatar
Miklos Szeredi committed
231 232
#endif

233 234
int page_group_by_mobility_disabled __read_mostly;

235
void set_pageblock_migratetype(struct page *page, int migratetype)
236
{
237 238
	if (unlikely(page_group_by_mobility_disabled &&
		     migratetype < MIGRATE_PCPTYPES))
239 240
		migratetype = MIGRATE_UNMOVABLE;

241 242 243 244
	set_pageblock_flags_group(page, (unsigned long)migratetype,
					PB_migrate, PB_migrate_end);
}

245 246
bool oom_killer_disabled __read_mostly;

Nick Piggin's avatar
Nick Piggin committed
247
#ifdef CONFIG_DEBUG_VM
248
static int page_outside_zone_boundaries(struct zone *zone, struct page *page)
Linus Torvalds's avatar
Linus Torvalds committed
249
{
250 251 252
	int ret = 0;
	unsigned seq;
	unsigned long pfn = page_to_pfn(page);
253
	unsigned long sp, start_pfn;
254

255 256
	do {
		seq = zone_span_seqbegin(zone);
257 258
		start_pfn = zone->zone_start_pfn;
		sp = zone->spanned_pages;
259
		if (!zone_spans_pfn(zone, pfn))
260 261 262
			ret = 1;
	} while (zone_span_seqretry(zone, seq));

263
	if (ret)
264 265 266
		pr_err("page 0x%lx outside node %d zone %s [ 0x%lx - 0x%lx ]\n",
			pfn, zone_to_nid(zone), zone->name,
			start_pfn, start_pfn + sp);
267

268
	return ret;
269 270 271 272
}

static int page_is_consistent(struct zone *zone, struct page *page)
{
273
	if (!pfn_valid_within(page_to_pfn(page)))
274
		return 0;
Linus Torvalds's avatar
Linus Torvalds committed
275
	if (zone != page_zone(page))
276 277 278 279 280 281 282 283 284 285
		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
286
		return 1;
287 288 289
	if (!page_is_consistent(zone, page))
		return 1;

Linus Torvalds's avatar
Linus Torvalds committed
290 291
	return 0;
}
Nick Piggin's avatar
Nick Piggin committed
292 293 294 295 296 297 298
#else
static inline int bad_range(struct zone *zone, struct page *page)
{
	return 0;
}
#endif

299 300
static void bad_page(struct page *page, const char *reason,
		unsigned long bad_flags)
Linus Torvalds's avatar
Linus Torvalds committed
301
{
302 303 304 305
	static unsigned long resume;
	static unsigned long nr_shown;
	static unsigned long nr_unshown;

306 307
	/* Don't complain about poisoned pages */
	if (PageHWPoison(page)) {
308
		page_mapcount_reset(page); /* remove PageBuddy */
309 310 311
		return;
	}

312 313 314 315 316 317 318 319 320 321
	/*
	 * 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) {
322 323
			printk(KERN_ALERT
			      "BUG: Bad page state: %lu messages suppressed\n",
324 325 326 327 328 329 330 331
				nr_unshown);
			nr_unshown = 0;
		}
		nr_shown = 0;
	}
	if (nr_shown++ == 0)
		resume = jiffies + 60 * HZ;

332
	printk(KERN_ALERT "BUG: Bad page state in process %s  pfn:%05lx\n",
333
		current->comm, page_to_pfn(page));
334
	dump_page_badflags(page, reason, bad_flags);
335

336
	print_modules();
Linus Torvalds's avatar
Linus Torvalds committed
337
	dump_stack();
338
out:
339
	/* Leave bad fields for debug, except PageBuddy could make trouble */
340
	page_mapcount_reset(page); /* remove PageBuddy */
341
	add_taint(TAINT_BAD_PAGE, LOCKDEP_NOW_UNRELIABLE);
Linus Torvalds's avatar
Linus Torvalds committed
342 343 344 345 346 347 348 349 350
}

/*
 * 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".
 *
351 352
 * 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
353
 *
354 355 356
 * 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
357
 */
358 359 360

static void free_compound_page(struct page *page)
{
361
	__free_pages_ok(page, compound_order(page));
362 363
}

364
void prep_compound_page(struct page *page, unsigned long order)
365 366 367 368 369 370 371 372 373
{
	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;
374
		set_page_count(p, 0);
375
		p->first_page = page;
David Rientjes's avatar
David Rientjes committed
376 377 378
		/* Make sure p->first_page is always valid for PageTail() */
		smp_wmb();
		__SetPageTail(p);
379 380 381
	}
}

382
/* update __split_huge_page_refcount if you change this function */
383
static int destroy_compound_page(struct page *page, unsigned long order)
Linus Torvalds's avatar
Linus Torvalds committed
384 385 386
{
	int i;
	int nr_pages = 1 << order;
387
	int bad = 0;
Linus Torvalds's avatar
Linus Torvalds committed
388

389
	if (unlikely(compound_order(page) != order)) {
390
		bad_page(page, "wrong compound order", 0);
391 392
		bad++;
	}
Linus Torvalds's avatar
Linus Torvalds committed
393

394
	__ClearPageHead(page);
395

396 397
	for (i = 1; i < nr_pages; i++) {
		struct page *p = page + i;
Linus Torvalds's avatar
Linus Torvalds committed
398

399 400 401 402 403
		if (unlikely(!PageTail(p))) {
			bad_page(page, "PageTail not set", 0);
			bad++;
		} else if (unlikely(p->first_page != page)) {
			bad_page(page, "first_page not consistent", 0);
404 405
			bad++;
		}
406
		__ClearPageTail(p);
Linus Torvalds's avatar
Linus Torvalds committed
407
	}
408 409

	return bad;
Linus Torvalds's avatar
Linus Torvalds committed
410 411
}

Nick Piggin's avatar
Nick Piggin committed
412 413 414 415
static inline void prep_zero_page(struct page *page, int order, gfp_t gfp_flags)
{
	int i;

416 417 418 419
	/*
	 * 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
420
	VM_BUG_ON((gfp_flags & __GFP_HIGHMEM) && in_interrupt());
Nick Piggin's avatar
Nick Piggin committed
421 422 423 424
	for (i = 0; i < (1 << order); i++)
		clear_highpage(page + i);
}

425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455
#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

456 457
static inline void set_page_order(struct page *page, int order)
{
458
	set_page_private(page, order);
459
	__SetPageBuddy(page);
Linus Torvalds's avatar
Linus Torvalds committed
460 461 462 463
}

static inline void rmv_page_order(struct page *page)
{
464
	__ClearPageBuddy(page);
465
	set_page_private(page, 0);
Linus Torvalds's avatar
Linus Torvalds committed
466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482
}

/*
 * 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
483
 * Assumption: *_mem_map is contiguous at least up to MAX_ORDER
Linus Torvalds's avatar
Linus Torvalds committed
484 485
 */
static inline unsigned long
486
__find_buddy_index(unsigned long page_idx, unsigned int order)
Linus Torvalds's avatar
Linus Torvalds committed
487
{
488
	return page_idx ^ (1 << order);
Linus Torvalds's avatar
Linus Torvalds committed
489 490 491 492 493
}

/*
 * 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
494
 * (a) the buddy is not in a hole &&
495
 * (b) the buddy is in the buddy system &&
496 497
 * (c) a page and its buddy have the same order &&
 * (d) a page and its buddy are in the same zone.
498
 *
499 500 501 502
 * For recording whether a page is in the buddy system, we set ->_mapcount
 * PAGE_BUDDY_MAPCOUNT_VALUE.
 * Setting, clearing, and testing _mapcount PAGE_BUDDY_MAPCOUNT_VALUE is
 * serialized by zone->lock.
Linus Torvalds's avatar
Linus Torvalds committed
503
 *
504
 * For recording page's order, we use page_private(page).
Linus Torvalds's avatar
Linus Torvalds committed
505
 */
506 507
static inline int page_is_buddy(struct page *page, struct page *buddy,
								int order)
Linus Torvalds's avatar
Linus Torvalds committed
508
{
509
	if (!pfn_valid_within(page_to_pfn(buddy)))
Nick Piggin's avatar
Nick Piggin committed
510 511
		return 0;

512
	if (page_is_guard(buddy) && page_order(buddy) == order) {
513
		VM_BUG_ON_PAGE(page_count(buddy) != 0, buddy);
514 515 516 517

		if (page_zone_id(page) != page_zone_id(buddy))
			return 0;

518 519 520
		return 1;
	}

521
	if (PageBuddy(buddy) && page_order(buddy) == order) {
522
		VM_BUG_ON_PAGE(page_count(buddy) != 0, buddy);
523 524 525 526 527 528 529 530 531

		/*
		 * zone check is done late to avoid uselessly
		 * calculating zone/node ids for pages that could
		 * never merge.
		 */
		if (page_zone_id(page) != page_zone_id(buddy))
			return 0;

532
		return 1;
533
	}
534
	return 0;
Linus Torvalds's avatar
Linus Torvalds committed
535 536 537 538 539 540 541 542 543 544 545 546 547 548 549
}

/*
 * 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
550 551 552
 * free pages of length of (1 << order) and marked with _mapcount
 * PAGE_BUDDY_MAPCOUNT_VALUE. Page's order is recorded in page_private(page)
 * field.
Linus Torvalds's avatar
Linus Torvalds committed
553
 * So when we are allocating or freeing one, we can derive the state of the
554 555
 * 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
556
 * If a block is freed, and its buddy is also free, then this
557
 * triggers coalescing into a block of larger size.
Linus Torvalds's avatar
Linus Torvalds committed
558
 *
559
 * -- nyc
Linus Torvalds's avatar
Linus Torvalds committed
560 561
 */

Nick Piggin's avatar
Nick Piggin committed
562
static inline void __free_one_page(struct page *page,
563 564
		struct zone *zone, unsigned int order,
		int migratetype)
Linus Torvalds's avatar
Linus Torvalds committed
565 566
{
	unsigned long page_idx;
567
	unsigned long combined_idx;
568
	unsigned long uninitialized_var(buddy_idx);
569
	struct page *buddy;
Linus Torvalds's avatar
Linus Torvalds committed
570

571 572
	VM_BUG_ON(!zone_is_initialized(zone));

573
	if (unlikely(PageCompound(page)))
574 575
		if (unlikely(destroy_compound_page(page, order)))
			return;
Linus Torvalds's avatar
Linus Torvalds committed
576

577 578
	VM_BUG_ON(migratetype == -1);

Linus Torvalds's avatar
Linus Torvalds committed
579 580
	page_idx = page_to_pfn(page) & ((1 << MAX_ORDER) - 1);

581 582
	VM_BUG_ON_PAGE(page_idx & ((1 << order) - 1), page);
	VM_BUG_ON_PAGE(bad_range(zone, page), page);
Linus Torvalds's avatar
Linus Torvalds committed
583 584

	while (order < MAX_ORDER-1) {
585 586
		buddy_idx = __find_buddy_index(page_idx, order);
		buddy = page + (buddy_idx - page_idx);
587
		if (!page_is_buddy(page, buddy, order))
588
			break;
589 590 591 592 593 594 595
		/*
		 * 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);
596 597
			__mod_zone_freepage_state(zone, 1 << order,
						  migratetype);
598 599 600 601 602
		} else {
			list_del(&buddy->lru);
			zone->free_area[order].nr_free--;
			rmv_page_order(buddy);
		}
603
		combined_idx = buddy_idx & page_idx;
Linus Torvalds's avatar
Linus Torvalds committed
604 605 606 607 608
		page = page + (combined_idx - page_idx);
		page_idx = combined_idx;
		order++;
	}
	set_page_order(page, order);
609 610 611 612 613 614 615 616 617

	/*
	 * 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
	 */
618
	if ((order < MAX_ORDER-2) && pfn_valid_within(page_to_pfn(buddy))) {
619
		struct page *higher_page, *higher_buddy;
620 621 622
		combined_idx = buddy_idx & page_idx;
		higher_page = page + (combined_idx - page_idx);
		buddy_idx = __find_buddy_index(combined_idx, order + 1);
623
		higher_buddy = higher_page + (buddy_idx - combined_idx);
624 625 626 627 628 629 630 631 632
		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
633 634 635
	zone->free_area[order].nr_free++;
}

636
static inline int free_pages_check(struct page *page)
Linus Torvalds's avatar
Linus Torvalds committed
637
{
638
	const char *bad_reason = NULL;
639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654
	unsigned long bad_flags = 0;

	if (unlikely(page_mapcount(page)))
		bad_reason = "nonzero mapcount";
	if (unlikely(page->mapping != NULL))
		bad_reason = "non-NULL mapping";
	if (unlikely(atomic_read(&page->_count) != 0))
		bad_reason = "nonzero _count";
	if (unlikely(page->flags & PAGE_FLAGS_CHECK_AT_FREE)) {
		bad_reason = "PAGE_FLAGS_CHECK_AT_FREE flag(s) set";
		bad_flags = PAGE_FLAGS_CHECK_AT_FREE;
	}
	if (unlikely(mem_cgroup_bad_page_check(page)))
		bad_reason = "cgroup check failed";
	if (unlikely(bad_reason)) {
		bad_page(page, bad_reason, bad_flags);
655
		return 1;
656
	}
657
	page_cpupid_reset_last(page);
658 659 660
	if (page->flags & PAGE_FLAGS_CHECK_AT_PREP)
		page->flags &= ~PAGE_FLAGS_CHECK_AT_PREP;
	return 0;
Linus Torvalds's avatar
Linus Torvalds committed
661 662 663
}

/*
664
 * Frees a number of pages from the PCP lists
Linus Torvalds's avatar
Linus Torvalds committed
665
 * Assumes all pages on list are in same zone, and of same order.
666
 * count is the number of pages to free.
Linus Torvalds's avatar
Linus Torvalds committed
667 668 669 670 671 672 673
 *
 * 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.
 */
674 675
static void free_pcppages_bulk(struct zone *zone, int count,
					struct per_cpu_pages *pcp)
Linus Torvalds's avatar
Linus Torvalds committed
676
{
677
	int migratetype = 0;
678
	int batch_free = 0;
679
	int to_free = count;
680

Nick Piggin's avatar
Nick Piggin committed
681
	spin_lock(&zone->lock);
Linus Torvalds's avatar
Linus Torvalds committed
682
	zone->pages_scanned = 0;
683

684
	while (to_free) {
Nick Piggin's avatar
Nick Piggin committed
685
		struct page *page;
686 687 688
		struct list_head *list;

		/*
689 690 691 692 693
		 * 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
694 695
		 */
		do {
696
			batch_free++;
697 698 699 700
			if (++migratetype == MIGRATE_PCPTYPES)
				migratetype = 0;
			list = &pcp->lists[migratetype];
		} while (list_empty(list));
Nick Piggin's avatar
Nick Piggin committed
701

702 703 704 705
		/* This is the only non-empty list. Free them all. */
		if (batch_free == MIGRATE_PCPTYPES)
			batch_free = to_free;

706
		do {
707 708
			int mt;	/* migratetype of the to-be-freed page */

709 710 711
			page = list_entry(list->prev, struct page, lru);
			/* must delete as __free_one_page list manipulates */
			list_del(&page->lru);
712
			mt = get_freepage_migratetype(page);
713
			/* MIGRATE_MOVABLE list may include MIGRATE_RESERVEs */
714 715
			__free_one_page(page, zone, 0, mt);
			trace_mm_page_pcpu_drain(page, 0, mt);
716
			if (likely(!is_migrate_isolate_page(page))) {
717 718 719 720
				__mod_zone_page_state(zone, NR_FREE_PAGES, 1);
				if (is_migrate_cma(mt))
					__mod_zone_page_state(zone, NR_FREE_CMA_PAGES, 1);
			}
721
		} while (--to_free && --batch_free && !list_empty(list));
Linus Torvalds's avatar
Linus Torvalds committed
722
	}
Nick Piggin's avatar
Nick Piggin committed
723
	spin_unlock(&zone->lock);
Linus Torvalds's avatar
Linus Torvalds committed
724 725
}

726 727
static void free_one_page(struct zone *zone, struct page *page, int order,
				int migratetype)
Linus Torvalds's avatar
Linus Torvalds committed
728
{
729 730
	spin_lock(&zone->lock);
	zone->pages_scanned = 0;
731

732
	__free_one_page(page, zone, order, migratetype);
733
	if (unlikely(!is_migrate_isolate(migratetype)))
734
		__mod_zone_freepage_state(zone, 1 << order, migratetype);
735
	spin_unlock(&zone->lock);
Nick Piggin's avatar
Nick Piggin committed
736 737
}

738
static bool free_pages_prepare(struct page *page, unsigned int order)
Nick Piggin's avatar
Nick Piggin committed
739
{
Linus Torvalds's avatar
Linus Torvalds committed
740
	int i;
741
	int bad = 0;
Linus Torvalds's avatar
Linus Torvalds committed
742

743
	trace_mm_page_free(page, order);
744 745
	kmemcheck_free_shadow(page, order);

Andrea Arcangeli's avatar
Andrea Arcangeli committed
746 747 748 749
	if (PageAnon(page))
		page->mapping = NULL;
	for (i = 0; i < (1 << order); i++)
		bad += free_pages_check(page + i);
750
	if (bad)
751
		return false;
752

753
	if (!PageHighMem(page)) {
754 755
		debug_check_no_locks_freed(page_address(page),
					   PAGE_SIZE << order);
756 757 758
		debug_check_no_obj_freed(page_address(page),
					   PAGE_SIZE << order);
	}
Nick Piggin's avatar
Nick Piggin committed
759
	arch_free_page(page, order);
Nick Piggin's avatar
Nick Piggin committed
760
	kernel_map_pages(page, 1 << order, 0);
Nick Piggin's avatar
Nick Piggin committed
761

762 763 764 765 766 767
	return true;
}

static void __free_pages_ok(struct page *page, unsigned int order)
{
	unsigned long flags;
768
	int migratetype;
769 770 771 772

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

Nick Piggin's avatar
Nick Piggin committed
773
	local_irq_save(flags);
774
	__count_vm_events(PGFREE, 1 << order);
775 776 777
	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
778
	local_irq_restore(flags);
Linus Torvalds's avatar
Linus Torvalds committed
779 780
}

781
void __init __free_pages_bootmem(struct page *page, unsigned int order)
782
{
783
	unsigned int nr_pages = 1 << order;
784
	struct page *p = page;
785
	unsigned int loop;
786

787 788 789
	prefetchw(p);
	for (loop = 0; loop < (nr_pages - 1); loop++, p++) {
		prefetchw(p + 1);
790 791
		__ClearPageReserved(p);
		set_page_count(p, 0);
792
	}
793 794
	__ClearPageReserved(p);
	set_page_count(p, 0);
795

796
	page_zone(page)->managed_pages += nr_pages;
797 798
	set_page_refcounted(page);
	__free_pages(page, order);
799 800
}

801
#ifdef CONFIG_CMA
802
/* Free whole pageblock and set its migration type to MIGRATE_CMA. */
803 804 805 806 807 808 809 810 811 812 813 814 815
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);
816
	adjust_managed_page_count(page, pageblock_nr_pages);
817 818
}
#endif
Linus Torvalds's avatar
Linus Torvalds committed
819 820 821 822 823 824 825 826 827 828 829 830 831

/*
 * 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.
 *
832
 * -- nyc
Linus Torvalds's avatar
Linus Torvalds committed
833
 */
Nick Piggin's avatar
Nick Piggin committed
834
static inline void expand(struct zone *zone, struct page *page,
835 836
	int low, int high, struct free_area *area,
	int migratetype)
Linus Torvalds's avatar
Linus Torvalds committed
837 838 839 840 841 842 843
{
	unsigned long size = 1 << high;

	while (high > low) {
		area--;
		high--;
		size >>= 1;
844
		VM_BUG_ON_PAGE(bad_range(zone, &page[size]), &page[size]);
845 846 847 848 849 850 851 852 853 854 855 856 857

#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 */
858 859
			__mod_zone_freepage_state(zone, -(1 << high),
						  migratetype);
860 861 862
			continue;
		}
#endif
863
		list_add(&page[size].lru, &area->free_list[migratetype]);
Linus Torvalds's avatar
Linus Torvalds committed
864 865 866 867 868 869 870 871
		area->nr_free++;
		set_page_order(&page[size], high);
	}
}

/*
 * This page is about to be returned from the page allocator
 */
872
static inline int check_new_page(struct page *page)
Linus Torvalds's avatar
Linus Torvalds committed
873
{
874
	const char *bad_reason = NULL;
875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890
	unsigned long bad_flags = 0;

	if (unlikely(page_mapcount(page)))
		bad_reason = "nonzero mapcount";
	if (unlikely(page->mapping != NULL))
		bad_reason = "non-NULL mapping";
	if (unlikely(atomic_read(&page->_count) != 0))
		bad_reason = "nonzero _count";
	if (unlikely(page->flags & PAGE_FLAGS_CHECK_AT_PREP)) {
		bad_reason = "PAGE_FLAGS_CHECK_AT_PREP flag set";
		bad_flags = PAGE_FLAGS_CHECK_AT_PREP;
	}
	if (unlikely(mem_cgroup_bad_page_check(page)))
		bad_reason = "cgroup check failed";
	if (unlikely(bad_reason)) {
		bad_page(page, bad_reason, bad_flags);
891
		return 1;
892
	}
893 894 895 896 897 898 899 900 901 902 903 904
	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;
	}
905

906
	set_page_private(page, 0);
907
	set_page_refcounted(page);
Nick Piggin's avatar
Nick Piggin committed
908 909

	arch_alloc_page(page, order);
Linus Torvalds's avatar
Linus Torvalds committed
910
	kernel_map_pages(page, 1 << order, 1);
Nick Piggin's avatar
Nick Piggin committed
911 912 913 914 915 916 917

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

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

918
	return 0;
Linus Torvalds's avatar
Linus Torvalds committed
919 920
}

921 922 923 924
/*
 * Go through the free lists for the given migratetype and remove
 * the smallest available page from the freelists
 */
925 926
static inline
struct page *__rmqueue_smallest(struct zone *zone, unsigned int order,
927 928 929
						int migratetype)
{
	unsigned int current_order;
930
	struct free_area *area;