page_alloc.c 182.5 KB
Newer Older
L
Linus Torvalds 已提交
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>
L
Linus Torvalds 已提交
23
#include <linux/bootmem.h>
24
#include <linux/memblock.h>
L
Linus Torvalds 已提交
25
#include <linux/compiler.h>
26
#include <linux/kernel.h>
27
#include <linux/kmemcheck.h>
L
Linus Torvalds 已提交
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>
L
Linus Torvalds 已提交
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>
L
Linus Torvalds 已提交
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>
K
KAMEZAWA Hiroyuki 已提交
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>
L
Linus Torvalds 已提交
64

65
#include <asm/sections.h>
L
Linus Torvalds 已提交
66
#include <asm/tlbflush.h>
67
#include <asm/div64.h>
L
Linus Torvalds 已提交
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

L
Linus Torvalds 已提交
89
/*
90
 * Array of node states.
L
Linus Torvalds 已提交
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;
L
Linus Torvalds 已提交
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

L
Linus Torvalds 已提交
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
A
Andi Kleen 已提交
174 175 176
 *
 * TBD: should special case ZONE_DMA32 machines here - in those we normally
 * don't need any ZONE_NORMAL reservation
L
Linus Torvalds 已提交
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
M
Mel Gorman 已提交
186
	 32,
187
#endif
M
Mel Gorman 已提交
188
	 32,
189
};
L
Linus Torvalds 已提交
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
M
Mel Gorman 已提交
202
	 "HighMem",
203
#endif
M
Mel Gorman 已提交
204
	 "Movable",
205 206
};

L
Linus Torvalds 已提交
207
int min_free_kbytes = 1024;
208
int user_min_free_kbytes = -1;
L
Linus Torvalds 已提交
209

210 211
static unsigned long __meminitdata nr_kernel_pages;
static unsigned long __meminitdata nr_all_pages;
212
static unsigned long __meminitdata dma_reserve;
L
Linus Torvalds 已提交
213

T
Tejun Heo 已提交
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

M
Miklos Szeredi 已提交
226 227
#if MAX_NUMNODES > 1
int nr_node_ids __read_mostly = MAX_NUMNODES;
228
int nr_online_nodes __read_mostly = 1;
M
Miklos Szeredi 已提交
229
EXPORT_SYMBOL(nr_node_ids);
230
EXPORT_SYMBOL(nr_online_nodes);
M
Miklos Szeredi 已提交
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;

N
Nick Piggin 已提交
247
#ifdef CONFIG_DEBUG_VM
248
static int page_outside_zone_boundaries(struct zone *zone, struct page *page)
L
Linus Torvalds 已提交
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;
L
Linus Torvalds 已提交
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))
L
Linus Torvalds 已提交
286
		return 1;
287 288 289
	if (!page_is_consistent(zone, page))
		return 1;

L
Linus Torvalds 已提交
290 291
	return 0;
}
N
Nick Piggin 已提交
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)
L
Linus Torvalds 已提交
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();
L
Linus Torvalds 已提交
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);
L
Linus Torvalds 已提交
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.
L
Linus Torvalds 已提交
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.
L
Linus Torvalds 已提交
357
 */
358 359 360

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

A
Andi Kleen 已提交
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;
D
David Rientjes 已提交
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)
L
Linus Torvalds 已提交
384 385 386
{
	int i;
	int nr_pages = 1 << order;
387
	int bad = 0;
L
Linus Torvalds 已提交
388

389
	if (unlikely(compound_order(page) != order)) {
390
		bad_page(page, "wrong compound order", 0);
391 392
		bad++;
	}
L
Linus Torvalds 已提交
393

394
	__ClearPageHead(page);
395

396 397
	for (i = 1; i < nr_pages; i++) {
		struct page *p = page + i;
L
Linus Torvalds 已提交
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);
L
Linus Torvalds 已提交
407
	}
408 409

	return bad;
L
Linus Torvalds 已提交
410 411
}

N
Nick Piggin 已提交
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.
	 */
N
Nick Piggin 已提交
420
	VM_BUG_ON((gfp_flags & __GFP_HIGHMEM) && in_interrupt());
N
Nick Piggin 已提交
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)
{
H
Hugh Dickins 已提交
458
	set_page_private(page, order);
459
	__SetPageBuddy(page);
L
Linus Torvalds 已提交
460 461 462 463
}

static inline void rmv_page_order(struct page *page)
{
464
	__ClearPageBuddy(page);
H
Hugh Dickins 已提交
465
	set_page_private(page, 0);
L
Linus Torvalds 已提交
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)
 *
A
Andreas Mohr 已提交
483
 * Assumption: *_mem_map is contiguous at least up to MAX_ORDER
L
Linus Torvalds 已提交
484 485
 */
static inline unsigned long
486
__find_buddy_index(unsigned long page_idx, unsigned int order)
L
Linus Torvalds 已提交
487
{
488
	return page_idx ^ (1 << order);
L
Linus Torvalds 已提交
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
N
Nick Piggin 已提交
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.
L
Linus Torvalds 已提交
503
 *
504
 * For recording page's order, we use page_private(page).
L
Linus Torvalds 已提交
505
 */
506 507
static inline int page_is_buddy(struct page *page, struct page *buddy,
								int order)
L
Linus Torvalds 已提交
508
{
509
	if (!pfn_valid_within(page_to_pfn(buddy)))
N
Nick Piggin 已提交
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;
L
Linus Torvalds 已提交
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.
L
Linus Torvalds 已提交
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.
L
Linus Torvalds 已提交
556
 * If a block is freed, and its buddy is also free, then this
557
 * triggers coalescing into a block of larger size.
L
Linus Torvalds 已提交
558
 *
559
 * -- nyc
L
Linus Torvalds 已提交
560 561
 */

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

571 572
	VM_BUG_ON(!zone_is_initialized(zone));

N
Nick Piggin 已提交
573
	if (unlikely(PageCompound(page)))
574 575
		if (unlikely(destroy_compound_page(page, order)))
			return;
L
Linus Torvalds 已提交
576

577 578
	VM_BUG_ON(migratetype == -1);

L
Linus Torvalds 已提交
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);
L
Linus Torvalds 已提交
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;
L
Linus Torvalds 已提交
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:
L
Linus Torvalds 已提交
633 634 635
	zone->free_area[order].nr_free++;
}

N
Nick Piggin 已提交
636
static inline int free_pages_check(struct page *page)
L
Linus Torvalds 已提交
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;
L
Linus Torvalds 已提交
661 662 663
}

/*
664
 * Frees a number of pages from the PCP lists
L
Linus Torvalds 已提交
665
 * Assumes all pages on list are in same zone, and of same order.
666
 * count is the number of pages to free.
L
Linus Torvalds 已提交
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)
L
Linus Torvalds 已提交
676
{
677
	int migratetype = 0;
678
	int batch_free = 0;
679
	int to_free = count;
680

N
Nick Piggin 已提交
681
	spin_lock(&zone->lock);
L
Linus Torvalds 已提交
682
	zone->pages_scanned = 0;
683

684
	while (to_free) {
N
Nick Piggin 已提交
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));
N
Nick Piggin 已提交
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));
L
Linus Torvalds 已提交
722
	}
N
Nick Piggin 已提交
723
	spin_unlock(&zone->lock);
L
Linus Torvalds 已提交
724 725
}

726 727
static void free_one_page(struct zone *zone, struct page *page, int order,
				int migratetype)
L
Linus Torvalds 已提交
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);
N
Nick Piggin 已提交
736 737
}

738
static bool free_pages_prepare(struct page *page, unsigned int order)
N
Nick Piggin 已提交
739
{
L
Linus Torvalds 已提交
740
	int i;
741
	int bad = 0;
L
Linus Torvalds 已提交
742

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

A
Andrea Arcangeli 已提交
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);
	}
N
Nick Piggin 已提交
759
	arch_free_page(page, order);
N
Nick Piggin 已提交
760
	kernel_map_pages(page, 1 << order, 0);
N
Nick Piggin 已提交
761

762 763 764 765 766 767
	return true;
}

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

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

N
Nick Piggin 已提交
773
	local_irq_save(flags);
774
	__count_vm_events(PGFREE, 1 << order);
M
Minchan Kim 已提交
775 776 777
	migratetype = get_pageblock_migratetype(page);
	set_freepage_migratetype(page, migratetype);
	free_one_page(page_zone(page), page, order, migratetype);
N
Nick Piggin 已提交
778
	local_irq_restore(flags);
L
Linus Torvalds 已提交
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
L
Linus Torvalds 已提交
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
L
Linus Torvalds 已提交
833
 */
N
Nick Piggin 已提交
834
static inline void expand(struct zone *zone, struct page *page,
835 836
	int low, int high, struct free_area *area,
	int migratetype)
L
Linus Torvalds 已提交
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]);
L
Linus Torvalds 已提交
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)
L
Linus Torvalds 已提交
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

H
Hugh Dickins 已提交
906
	set_page_private(page, 0);
907
	set_page_refcounted(page);
N
Nick Piggin 已提交
908 909

	arch_alloc_page(page, order);
L
Linus Torvalds 已提交
910
	kernel_map_pages(page, 1 << order, 1);
N
Nick Piggin 已提交
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;
L
Linus Torvalds 已提交
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;
931 932 933 934 935 936 937 938 939 940 941 942 943 944
	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);
945
		set_freepage_migratetype(page, migratetype);
946 947 948 949 950 951 952
		return page;
	}

	return NULL;
}


953 954 955 956
/*
 * This array describes the order lists are fallen back to when
 * the free lists for the desirable migrate type are depleted
 */
957 958 959 960 961 962 963 964 965
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
966
	[MIGRATE_RESERVE]     = { MIGRATE_RESERVE }, /* Never used */
967
#ifdef CONFIG_MEMORY_ISOLATION
968
	[MIGRATE_ISOLATE]     = { MIGRATE_RESERVE }, /* Never used */
969
#endif
970 971
};

972 973
/*
 * Move the free pages in a range to the free lists of the requested type.
974
 * Note that start_page and end_pages are not aligned on a pageblock
975 976
 * boundary. If alignment is required, use move_freepages_block()
 */
977
int move_freepages(struct zone *zone,
A
Adrian Bunk 已提交
978 979
			  struct page *start_page, struct page *end_page,
			  int migratetype)
980 981 982
{
	struct page *page;
	unsigned long order;
983
	int pages_moved = 0;
984 985 986 987 988 989 990

#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
M
Mel Gorman 已提交
991
	 * grouping pages by mobility
992 993 994 995 996
	 */
	BUG_ON(page_zone(start_page) != page_zone(end_page));
#endif

	for (page = start_page; page <= end_page;) {
997
		/* Make sure we are not inadvertently changing nodes */
998
		VM_BUG_ON_PAGE(page_to_nid(page) != zone_to_nid(zone), page);
999

1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010
		if (!pfn_valid_within(page_to_pfn(page))) {
			page++;
			continue;
		}

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

		order = page_order(page);
1011 1012
		list_move(&page->lru,
			  &zone->free_area[order].free_list[migratetype]);
M
Minchan Kim 已提交
1013
		set_freepage_migratetype(page, migratetype);
1014
		page += 1 << order;
1015
		pages_moved += 1 << order;
1016 1017
	}

1018
	return pages_moved;
1019 1020
}

1021
int move_freepages_block(struct zone *zone, struct page *page,
1022
				int migratetype)
1023 1024 1025 1026 1027
{
	unsigned long start_pfn, end_pfn;
	struct page *start_page, *end_page;

	start_pfn = page_to_pfn(page);
1028
	start_pfn = start_pfn & ~(pageblock_nr_pages-1);
1029
	start_page = pfn_to_page(start_pfn);
1030 1031
	end_page = start_page + pageblock_nr_pages - 1;
	end_pfn = start_pfn + pageblock_nr_pages - 1;
1032 1033

	/* Do not cross zone boundaries */
1034
	if (!zone_spans_pfn(zone, start_pfn))
1035
		start_page = page;
1036
	if (!zone_spans_pfn(zone, end_pfn))
1037 1038 1039 1040 1041
		return 0;

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

1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052
static void change_pageblock_range(struct page *pageblock_page,
					int start_order, int migratetype)
{
	int nr_pageblocks = 1 << (start_order - pageblock_order);

	while (nr_pageblocks--) {
		set_pageblock_migratetype(pageblock_page, migratetype);
		pageblock_page += pageblock_nr_pages;
	}
}

1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069
/*
 * If breaking a large block of pages, move all free pages to the preferred
 * allocation list. If falling back for a reclaimable kernel allocation, be
 * more aggressive about taking ownership of free pages.
 *
 * On the other hand, never change migration type of MIGRATE_CMA pageblocks
 * nor move CMA pages to different free lists. We don't want unmovable pages
 * to be allocated from MIGRATE_CMA areas.
 *
 * Returns the new migratetype of the pageblock (or the same old migratetype
 * if it was unchanged).
 */
static int try_to_steal_freepages(struct zone *zone, struct page *page,
				  int start_type, int fallback_type)
{
	int current_order = page_order(page);

1070 1071
	/*
	 * When borrowing from MIGRATE_CMA, we need to release the excess
1072 1073 1074
	 * buddy pages to CMA itself. We also ensure the freepage_migratetype
	 * is set to CMA so it is returned to the correct freelist in case
	 * the page ends up being not actually allocated from the pcp lists.
1075
	 */
1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104
	if (is_migrate_cma(fallback_type))
		return fallback_type;

	/* Take ownership for orders >= pageblock_order */
	if (current_order >= pageblock_order) {
		change_pageblock_range(page, current_order, start_type);
		return start_type;
	}

	if (current_order >= pageblock_order / 2 ||
	    start_type == MIGRATE_RECLAIMABLE ||
	    page_group_by_mobility_disabled) {
		int pages;

		pages = move_freepages_block(zone, page, start_type);

		/* Claim the whole block if over half of it is free */
		if (pages >= (1 << (pageblock_order-1)) ||
				page_group_by_mobility_disabled) {

			set_pageblock_migratetype(page, start_type);
			return start_type;
		}

	}

	return fallback_type;
}

1105
/* Remove an element from the buddy allocator from the fallback list */
1106 1107
static inline struct page *
__rmqueue_fallback(struct zone *zone, int order, int start_migratetype)
1108
{
1109
	struct free_area *area;
1110 1111
	int current_order;
	struct page *page;
1112
	int migratetype, new_type, i;
1113 1114 1115 1116

	/* Find the largest possible block of pages in the other list */
	for (current_order = MAX_ORDER-1; current_order >= order;
						--current_order) {
1117
		for (i = 0;; i++) {
1118 1119
			migratetype = fallbacks[start_migratetype][i];

1120 1121
			/* MIGRATE_RESERVE handled later if necessary */
			if (migratetype == MIGRATE_RESERVE)
1122
				break;
M
Mel Gorman 已提交
1123

1124 1125 1126 1127 1128 1129 1130 1131
			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);
			area->nr_free--;

1132 1133 1134
			new_type = try_to_steal_freepages(zone, page,
							  start_migratetype,
							  migratetype);
1135 1136 1137 1138 1139

			/* Remove the page from the freelists */
			list_del(&page->lru);
			rmv_page_order(page);

1140
			expand(zone, page, order, current_order, area,
1141
			       new_type);
1142 1143 1144 1145 1146 1147
			/* The freepage_migratetype may differ from pageblock's
			 * migratetype depending on the decisions in
			 * try_to_steal_freepages. This is OK as long as it does
			 * not differ for MIGRATE_CMA type.
			 */
			set_freepage_migratetype(page, new_type);
1148

1149 1150
			trace_mm_page_alloc_extfrag(page, order, current_order,
				start_migratetype, migratetype, new_type);
1151

1152 1153 1154 1155
			return page;
		}
	}

1156
	return NULL;
1157 1158
}

1159
/*
L
Linus Torvalds 已提交
1160 1161 1162
 * Do the hard work of removing an element from the buddy allocator.
 * Call me with the zone->lock already held.
 */
1163 1164
static struct page *__rmqueue(struct zone *zone, unsigned int order,
						int migratetype)
L
Linus Torvalds 已提交
1165 1166 1167
{
	struct page *page;

1168
retry_reserve:
1169
	page = __rmqueue_smallest(zone, order, migratetype);
1170

1171
	if (unlikely(!page) && migratetype != MIGRATE_RESERVE) {
1172
		page = __rmqueue_fallback(zone, order, migratetype);
1173

1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184
		/*
		 * Use MIGRATE_RESERVE rather than fail an allocation. goto
		 * is used because __rmqueue_smallest is an inline function
		 * and we want just one call site
		 */
		if (!page) {
			migratetype = MIGRATE_RESERVE;
			goto retry_reserve;
		}
	}

1185
	trace_mm_page_alloc_zone_locked(page, order, migratetype);
1186
	return page;
L
Linus Torvalds 已提交
1187 1188
}

1189
/*
L
Linus Torvalds 已提交
1190 1191 1192 1193
 * Obtain a specified number of elements from the buddy allocator, all under
 * a single hold of the lock, for efficiency.  Add them to the supplied list.
 * Returns the number of new pages which were placed at *list.
 */
1194
static int rmqueue_bulk(struct zone *zone, unsigned int order,
1195
			unsigned long count, struct list_head *list,
1196
			int migratetype, int cold)
L
Linus Torvalds 已提交
1197
{
1198
	int i;
1199

N
Nick Piggin 已提交
1200
	spin_lock(&zone->lock);
L
Linus Torvalds 已提交
1201
	for (i = 0; i < count; ++i) {
1202
		struct page *page = __rmqueue(zone, order, migratetype);
N
Nick Piggin 已提交
1203
		if (unlikely(page == NULL))
L
Linus Torvalds 已提交
1204
			break;
1205 1206 1207 1208 1209 1210 1211 1212 1213 1214

		/*
		 * Split buddy pages returned by expand() are received here
		 * in physical page order. The page is added to the callers and
		 * list and the list head then moves forward. From the callers
		 * perspective, the linked list is ordered by page number in
		 * some conditions. This is useful for IO devices that can
		 * merge IO requests if the physical pages are ordered
		 * properly.
		 */
1215 1216 1217 1218
		if (likely(cold == 0))
			list_add(&page->lru, list);
		else
			list_add_tail(&page->lru, list);
1219
		list = &page->lru;
1220
		if (is_migrate_cma(get_freepage_migratetype(page)))
1221 1222
			__mod_zone_page_state(zone, NR_FREE_CMA_PAGES,
					      -(1 << order));
L
Linus Torvalds 已提交
1223
	}
1224
	__mod_zone_page_state(zone, NR_FREE_PAGES, -(i << order));
N
Nick Piggin 已提交
1225
	spin_unlock(&zone->lock);
N
Nick Piggin 已提交
1226
	return i;
L
Linus Torvalds 已提交
1227 1228
}

1229
#ifdef CONFIG_NUMA
1230
/*
1231 1232 1233 1234
 * Called from the vmstat counter updater to drain pagesets of this
 * currently executing processor on remote nodes after they have
 * expired.
 *
1235 1236
 * Note that this function must be called with the thread pinned to
 * a single processor.
1237
 */
1238
void drain_zone_pages(struct zone *zone, struct per_cpu_pages *pcp)
1239 1240
{
	unsigned long flags;
1241
	int to_drain;
1242
	unsigned long batch;
1243

1244
	local_irq_save(flags);
1245 1246 1247
	batch = ACCESS_ONCE(pcp->batch);
	if (pcp->count >= batch)
		to_drain = batch;
1248 1249
	else
		to_drain = pcp->count;
1250 1251 1252 1253
	if (to_drain > 0) {
		free_pcppages_bulk(zone, to_drain, pcp);
		pcp->count -= to_drain;
	}
1254
	local_irq_restore(flags);
1255 1256 1257
}
#endif

1258 1259 1260 1261 1262 1263 1264 1265
/*
 * Drain pages of the indicated processor.
 *
 * The processor must either be the current processor and the
 * thread pinned to the current processor or a processor that
 * is not online.
 */
static void drain_pages(unsigned int cpu)
L
Linus Torvalds 已提交
1266
{
N
Nick Piggin 已提交
1267
	unsigned long flags;
L
Linus Torvalds 已提交
1268 1269
	struct zone *zone;

1270
	for_each_populated_zone(zone) {
L
Linus Torvalds 已提交
1271
		struct per_cpu_pageset *pset;
1272
		struct per_cpu_pages *pcp;
L
Linus Torvalds 已提交
1273

1274 1275
		local_irq_save(flags);
		pset = per_cpu_ptr(zone->pageset, cpu);
1276 1277

		pcp = &pset->pcp;
1278 1279 1280 1281
		if (pcp->count) {
			free_pcppages_bulk(zone, pcp->count, pcp);
			pcp->count = 0;
		}
1282
		local_irq_restore(flags);
L
Linus Torvalds 已提交
1283 1284 1285
	}
}

1286 1287 1288 1289 1290 1291 1292 1293 1294
/*
 * Spill all of this CPU's per-cpu pages back into the buddy allocator.
 */
void drain_local_pages(void *arg)
{
	drain_pages(smp_processor_id());
}

/*
1295 1296 1297 1298 1299 1300 1301
 * Spill all the per-cpu pages from all CPUs back into the buddy allocator.
 *
 * Note that this code is protected against sending an IPI to an offline
 * CPU but does not guarantee sending an IPI to newly hotplugged CPUs:
 * on_each_cpu_mask() blocks hotplug and won't talk to offlined CPUs but
 * nothing keeps CPUs from showing up after we populated the cpumask and
 * before the call to on_each_cpu_mask().
1302 1303 1304
 */
void drain_all_pages(void)
{
1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335
	int cpu;
	struct per_cpu_pageset *pcp;
	struct zone *zone;

	/*
	 * Allocate in the BSS so we wont require allocation in
	 * direct reclaim path for CONFIG_CPUMASK_OFFSTACK=y
	 */
	static cpumask_t cpus_with_pcps;

	/*
	 * We don't care about racing with CPU hotplug event
	 * as offline notification will cause the notified
	 * cpu to drain that CPU pcps and on_each_cpu_mask
	 * disables preemption as part of its processing
	 */
	for_each_online_cpu(cpu) {
		bool has_pcps = false;
		for_each_populated_zone(zone) {
			pcp = per_cpu_ptr(zone->pageset, cpu);
			if (pcp->pcp.count) {
				has_pcps = true;
				break;
			}
		}
		if (has_pcps)
			cpumask_set_cpu(cpu, &cpus_with_pcps);
		else
			cpumask_clear_cpu(cpu, &cpus_with_pcps);
	}
	on_each_cpu_mask(&cpus_with_pcps, drain_local_pages, NULL, 1);
1336 1337
}

1338
#ifdef CONFIG_HIBERNATION
L
Linus Torvalds 已提交
1339 1340 1341

void mark_free_pages(struct zone *zone)
{
1342 1343
	unsigned long pfn, max_zone_pfn;
	unsigned long flags;
1344
	int order, t;
L
Linus Torvalds 已提交
1345 1346
	struct list_head *curr;

1347
	if (zone_is_empty(zone))
L
Linus Torvalds 已提交
1348 1349 1350
		return;

	spin_lock_irqsave(&zone->lock, flags);
1351

1352
	max_zone_pfn = zone_end_pfn(zone);
1353 1354 1355 1356
	for (pfn = zone->zone_start_pfn; pfn < max_zone_pfn; pfn++)
		if (pfn_valid(pfn)) {
			struct page *page = pfn_to_page(pfn);

1357 1358
			if (!swsusp_page_is_forbidden(page))
				swsusp_unset_page_free(page);
1359
		}
L
Linus Torvalds 已提交
1360

1361 1362
	for_each_migratetype_order(order, t) {
		list_for_each(curr, &zone->free_area[order].free_list[t]) {
1363
			unsigned long i;
L
Linus Torvalds 已提交
1364

1365 1366
			pfn = page_to_pfn(list_entry(curr, struct page, lru));
			for (i = 0; i < (1UL << order); i++)
1367
				swsusp_set_page_free(pfn_to_page(pfn + i));
1368
		}
1369
	}
L
Linus Torvalds 已提交
1370 1371
	spin_unlock_irqrestore(&zone->lock, flags);
}
1372
#endif /* CONFIG_PM */
L
Linus Torvalds 已提交
1373 1374 1375

/*
 * Free a 0-order page
L
Li Hong 已提交
1376
 * cold == 1 ? free a cold page : free a hot page
L
Linus Torvalds 已提交
1377
 */
L
Li Hong 已提交
1378
void free_hot_cold_page(struct page *page, int cold)
L
Linus Torvalds 已提交
1379 1380 1381 1382
{
	struct zone *zone = page_zone(page);
	struct per_cpu_pages *pcp;
	unsigned long flags;
1383
	int migratetype;
L
Linus Torvalds 已提交
1384

1385
	if (!free_pages_prepare(page, 0))
1386 1387
		return;

1388
	migratetype = get_pageblock_migratetype(page);
1389
	set_freepage_migratetype(page, migratetype);
L
Linus Torvalds 已提交
1390
	local_irq_save(flags);
1391
	__count_vm_event(PGFREE);
1392

1393 1394 1395 1396 1397 1398 1399 1400
	/*
	 * We only track unmovable, reclaimable and movable on pcp lists.
	 * Free ISOLATE pages back to the allocator because they are being
	 * offlined but treat RESERVE as movable pages so we can get those
	 * areas back if necessary. Otherwise, we may have to free
	 * excessively into the page allocator
	 */
	if (migratetype >= MIGRATE_PCPTYPES) {
1401
		if (unlikely(is_migrate_isolate(migratetype))) {
1402 1403 1404 1405 1406 1407
			free_one_page(zone, page, 0, migratetype);
			goto out;
		}
		migratetype = MIGRATE_MOVABLE;
	}

1408
	pcp = &this_cpu_ptr(zone->pageset)->pcp;
1409
	if (cold)
1410
		list_add_tail(&page->lru, &pcp->lists[migratetype]);
1411
	else
1412
		list_add(&page->lru, &pcp->lists[migratetype]);
L
Linus Torvalds 已提交
1413
	pcp->count++;
N
Nick Piggin 已提交
1414
	if (pcp->count >= pcp->high) {
1415 1416 1417
		unsigned long batch = ACCESS_ONCE(pcp->batch);
		free_pcppages_bulk(zone, batch, pcp);
		pcp->count -= batch;
N
Nick Piggin 已提交
1418
	}
1419 1420

out:
L
Linus Torvalds 已提交
1421 1422 1423
	local_irq_restore(flags);
}

1424 1425 1426 1427 1428 1429 1430 1431
/*
 * Free a list of 0-order pages
 */
void free_hot_cold_page_list(struct list_head *list, int cold)
{
	struct page *page, *next;

	list_for_each_entry_safe(page, next, list, lru) {
1432
		trace_mm_page_free_batched(page, cold);
1433 1434 1435 1436
		free_hot_cold_page(page, cold);
	}
}

N
Nick Piggin 已提交
1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448
/*
 * split_page takes a non-compound higher-order page, and splits it into
 * n (1<<order) sub-pages: page[0..n]
 * Each sub-page must be freed individually.
 *
 * Note: this is probably too low level an operation for use in drivers.
 * Please consult with lkml before using this in your driver.
 */
void split_page(struct page *page, unsigned int order)
{
	int i;

1449 1450
	VM_BUG_ON_PAGE(PageCompound(page), page);
	VM_BUG_ON_PAGE(!page_count(page), page);
1451 1452 1453 1454 1455 1456 1457 1458 1459 1460

#ifdef CONFIG_KMEMCHECK
	/*
	 * Split shadow pages too, because free(page[0]) would
	 * otherwise free the whole shadow.
	 */
	if (kmemcheck_page_is_tracked(page))
		split_page(virt_to_page(page[0].shadow), order);
#endif

1461 1462
	for (i = 1; i < (1 << order); i++)
		set_page_refcounted(page + i);
N
Nick Piggin 已提交
1463
}
K
K. Y. Srinivasan 已提交
1464
EXPORT_SYMBOL_GPL(split_page);
N
Nick Piggin 已提交
1465

1466
static int __isolate_free_page(struct page *page, unsigned int order)
1467 1468 1469
{
	unsigned long watermark;
	struct zone *zone;
1470
	int mt;
1471 1472 1473 1474

	BUG_ON(!PageBuddy(page));

	zone = page_zone(page);
1475
	mt = get_pageblock_migratetype(page);
1476

1477
	if (!is_migrate_isolate(mt)) {
1478 1479 1480 1481 1482
		/* Obey watermarks as if the page was being allocated */
		watermark = low_wmark_pages(zone) + (1 << order);
		if (!zone_watermark_ok(zone, 0, watermark, 0, 0))
			return 0;

1483
		__mod_zone_freepage_state(zone, -(1UL << order), mt);
1484
	}
1485 1486 1487 1488 1489

	/* Remove page from free list */
	list_del(&page->lru);
	zone->free_area[order].nr_free--;
	rmv_page_order(page);
1490

1491
	/* Set the pageblock if the isolated page is at least a pageblock */
1492 1493
	if (order >= pageblock_order - 1) {
		struct page *endpage = page + (1 << order) - 1;
1494 1495
		for (; page < endpage; page += pageblock_nr_pages) {
			int mt = get_pageblock_migratetype(page);
1496
			if (!is_migrate_isolate(mt) && !is_migrate_cma(mt))
1497 1498 1499
				set_pageblock_migratetype(page,
							  MIGRATE_MOVABLE);
		}
1500 1501
	}

1502
	return 1UL << order;
1503 1504 1505 1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518 1519 1520 1521
}

/*
 * Similar to split_page except the page is already free. As this is only
 * being used for migration, the migratetype of the block also changes.
 * As this is called with interrupts disabled, the caller is responsible
 * for calling arch_alloc_page() and kernel_map_page() after interrupts
 * are enabled.
 *
 * Note: this is probably too low level an operation for use in drivers.
 * Please consult with lkml before using this in your driver.
 */
int split_free_page(struct page *page)
{
	unsigned int order;
	int nr_pages;

	order = page_order(page);

1522
	nr_pages = __isolate_free_page(page, order);
1523 1524 1525 1526 1527 1528 1529
	if (!nr_pages)
		return 0;

	/* Split into individual pages */
	set_page_refcounted(page);
	split_page(page, order);
	return nr_pages;
1530 1531
}

L
Linus Torvalds 已提交
1532 1533 1534 1535 1536
/*
 * Really, prep_compound_page() should be called from __rmqueue_bulk().  But
 * we cheat by calling it from here, in the order > 0 path.  Saves a branch
 * or two.
 */
1537 1538
static inline
struct page *buffered_rmqueue(struct zone *preferred_zone,
1539 1540
			struct zone *zone, int order, gfp_t gfp_flags,
			int migratetype)
L
Linus Torvalds 已提交
1541 1542
{
	unsigned long flags;
1543
	struct page *page;
L
Linus Torvalds 已提交
1544 1545
	int cold = !!(gfp_flags & __GFP_COLD);

1546
again:
N
Nick Piggin 已提交
1547
	if (likely(order == 0)) {
L
Linus Torvalds 已提交
1548
		struct per_cpu_pages *pcp;
1549
		struct list_head *list;
L
Linus Torvalds 已提交
1550 1551

		local_irq_save(flags);
1552 1553
		pcp = &this_cpu_ptr(zone->pageset)->pcp;
		list = &pcp->lists[migratetype];
1554
		if (list_empty(list)) {
1555
			pcp->count += rmqueue_bulk(zone, 0,
1556
					pcp->batch, list,
1557
					migratetype, cold);
1558
			if (unlikely(list_empty(list)))
1559
				goto failed;
1560
		}
1561

1562 1563 1564 1565 1566
		if (cold)
			page = list_entry(list->prev, struct page, lru);
		else
			page = list_entry(list->next, struct page, lru);

1567 1568
		list_del(&page->lru);
		pcp->count--;
R
Rohit Seth 已提交
1569
	} else {
1570 1571 1572 1573 1574 1575 1576 1577
		if (unlikely(gfp_flags & __GFP_NOFAIL)) {
			/*
			 * __GFP_NOFAIL is not to be used in new code.
			 *
			 * All __GFP_NOFAIL callers should be fixed so that they
			 * properly detect and handle allocation failures.
			 *
			 * We most definitely don't want callers attempting to
1578
			 * allocate greater than order-1 page units with
1579 1580
			 * __GFP_NOFAIL.
			 */
1581
			WARN_ON_ONCE(order > 1);
1582
		}
L
Linus Torvalds 已提交
1583
		spin_lock_irqsave(&zone->lock, flags);
1584
		page = __rmqueue(zone, order, migratetype);
N
Nick Piggin 已提交
1585 1586 1587
		spin_unlock(&zone->lock);
		if (!page)
			goto failed;
1588
		__mod_zone_freepage_state(zone, -(1 << order),
1589
					  get_freepage_migratetype(page));
L
Linus Torvalds 已提交
1590 1591
	}

1592
	__mod_zone_page_state(zone, NR_ALLOC_BATCH, -(1 << order));
1593

1594
	__count_zone_vm_events(PGALLOC, zone, 1 << order);
A
Andi Kleen 已提交
1595
	zone_statistics(preferred_zone, zone, gfp_flags);
N
Nick Piggin 已提交
1596
	local_irq_restore(flags);
L
Linus Torvalds 已提交
1597

1598
	VM_BUG_ON_PAGE(bad_range(zone, page), page);
N
Nick Piggin 已提交
1599
	if (prep_new_page(page, order, gfp_flags))
N
Nick Piggin 已提交
1600
		goto again;
L
Linus Torvalds 已提交
1601
	return page;
N
Nick Piggin 已提交
1602 1603 1604 1605

failed:
	local_irq_restore(flags);
	return NULL;
L
Linus Torvalds 已提交
1606 1607
}

1608 1609
#ifdef CONFIG_FAIL_PAGE_ALLOC

1610
static struct {
1611 1612 1613 1614
	struct fault_attr attr;

	u32 ignore_gfp_highmem;
	u32 ignore_gfp_wait;
1615
	u32 min_order;
1616 1617
} fail_page_alloc = {
	.attr = FAULT_ATTR_INITIALIZER,
1618 1619
	.ignore_gfp_wait = 1,
	.ignore_gfp_highmem = 1,
1620
	.min_order = 1,
1621 1622 1623 1624 1625 1626 1627 1628
};

static int __init setup_fail_page_alloc(char *str)
{
	return setup_fault_attr(&fail_page_alloc.attr, str);
}
__setup("fail_page_alloc=", setup_fail_page_alloc);

1629
static bool should_fail_alloc_page(gfp_t gfp_mask, unsigned int order)
1630
{
1631
	if (order < fail_page_alloc.min_order)
1632
		return false;
1633
	if (gfp_mask & __GFP_NOFAIL)
1634
		return false;
1635
	if (fail_page_alloc.ignore_gfp_highmem && (gfp_mask & __GFP_HIGHMEM))
1636
		return false;
1637
	if (fail_page_alloc.ignore_gfp_wait && (gfp_mask & __GFP_WAIT))
1638
		return false;
1639 1640 1641 1642 1643 1644 1645 1646

	return should_fail(&fail_page_alloc.attr, 1 << order);
}

#ifdef CONFIG_FAULT_INJECTION_DEBUG_FS

static int __init fail_page_alloc_debugfs(void)
{
A
Al Viro 已提交
1647
	umode_t mode = S_IFREG | S_IRUSR | S_IWUSR;
1648 1649
	struct dentry *dir;

1650 1651 1652 1653
	dir = fault_create_debugfs_attr("fail_page_alloc", NULL,
					&fail_page_alloc.attr);
	if (IS_ERR(dir))
		return PTR_ERR(dir);
1654

1655 1656 1657 1658 1659 1660 1661 1662 1663 1664 1665 1666
	if (!debugfs_create_bool("ignore-gfp-wait", mode, dir,
				&fail_page_alloc.ignore_gfp_wait))
		goto fail;
	if (!debugfs_create_bool("ignore-gfp-highmem", mode, dir,
				&fail_page_alloc.ignore_gfp_highmem))
		goto fail;
	if (!debugfs_create_u32("min-order", mode, dir,
				&fail_page_alloc.min_order))
		goto fail;

	return 0;
fail:
1667
	debugfs_remove_recursive(dir);
1668

1669
	return -ENOMEM;
1670 1671 1672 1673 1674 1675 1676 1677
}

late_initcall(fail_page_alloc_debugfs);

#endif /* CONFIG_FAULT_INJECTION_DEBUG_FS */

#else /* CONFIG_FAIL_PAGE_ALLOC */

1678
static inline bool should_fail_alloc_page(gfp_t gfp_mask, unsigned int order)
1679
{
1680
	return false;
1681 1682 1683 1684
}

#endif /* CONFIG_FAIL_PAGE_ALLOC */

L
Linus Torvalds 已提交
1685
/*
1686
 * Return true if free pages are above 'mark'. This takes into account the order
L
Linus Torvalds 已提交
1687 1688
 * of the allocation.
 */
1689 1690
static bool __zone_watermark_ok(struct zone *z, int order, unsigned long mark,
		      int classzone_idx, int alloc_flags, long free_pages)
L
Linus Torvalds 已提交
1691 1692
{
	/* free_pages my go negative - that's OK */
1693
	long min = mark;
1694
	long lowmem_reserve = z->lowmem_reserve[classzone_idx];
L
Linus Torvalds 已提交
1695
	int o;
1696
	long free_cma = 0;
L
Linus Torvalds 已提交
1697

1698
	free_pages -= (1 << order) - 1;
R
Rohit Seth 已提交
1699
	if (alloc_flags & ALLOC_HIGH)
L
Linus Torvalds 已提交
1700
		min -= min / 2;
R
Rohit Seth 已提交
1701
	if (alloc_flags & ALLOC_HARDER)
L
Linus Torvalds 已提交
1702
		min -= min / 4;
1703 1704 1705
#ifdef CONFIG_CMA
	/* If allocation can't use CMA areas don't use free CMA pages */
	if (!(alloc_flags & ALLOC_CMA))
1706
		free_cma = zone_page_state(z, NR_FREE_CMA_PAGES);
1707
#endif
1708 1709

	if (free_pages - free_cma <= min + lowmem_reserve)
1710
		return false;
L
Linus Torvalds 已提交
1711 1712 1713 1714 1715 1716 1717 1718
	for (o = 0; o < order; o++) {
		/* At the next order, this order's pages become unavailable */
		free_pages -= z->free_area[o].nr_free << o;

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

		if (free_pages <= min)
1719
			return false;
L
Linus Torvalds 已提交
1720
	}
1721 1722 1723 1724 1725 1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736 1737 1738 1739 1740
	return true;
}

bool zone_watermark_ok(struct zone *z, int order, unsigned long mark,
		      int classzone_idx, int alloc_flags)
{
	return __zone_watermark_ok(z, order, mark, classzone_idx, alloc_flags,
					zone_page_state(z, NR_FREE_PAGES));
}

bool zone_watermark_ok_safe(struct zone *z, int order, unsigned long mark,
		      int classzone_idx, int alloc_flags)
{
	long free_pages = zone_page_state(z, NR_FREE_PAGES);

	if (z->percpu_drift_mark && free_pages < z->percpu_drift_mark)
		free_pages = zone_page_state_snapshot(z, NR_FREE_PAGES);

	return __zone_watermark_ok(z, order, mark, classzone_idx, alloc_flags,
								free_pages);
L
Linus Torvalds 已提交
1741 1742
}

1743 1744 1745 1746 1747 1748
#ifdef CONFIG_NUMA
/*
 * zlc_setup - Setup for "zonelist cache".  Uses cached zone data to
 * skip over zones that are not allowed by the cpuset, or that have
 * been recently (in last second) found to be nearly full.  See further
 * comments in mmzone.h.  Reduces cache footprint of zonelist scans
S
Simon Arlott 已提交
1749
 * that have to skip over a lot of full or unallowed zones.
1750
 *
1751
 * If the zonelist cache is present in the passed zonelist, then
1752
 * returns a pointer to the allowed node mask (either the current
1753
 * tasks mems_allowed, or node_states[N_MEMORY].)
1754 1755 1756 1757 1758 1759 1760 1761 1762 1763 1764 1765 1766 1767 1768 1769 1770 1771 1772 1773 1774
 *
 * If the zonelist cache is not available for this zonelist, does
 * nothing and returns NULL.
 *
 * If the fullzones BITMAP in the zonelist cache is stale (more than
 * a second since last zap'd) then we zap it out (clear its bits.)
 *
 * We hold off even calling zlc_setup, until after we've checked the
 * first zone in the zonelist, on the theory that most allocations will
 * be satisfied from that first zone, so best to examine that zone as
 * quickly as we can.
 */
static nodemask_t *zlc_setup(struct zonelist *zonelist, int alloc_flags)
{
	struct zonelist_cache *zlc;	/* cached zonelist speedup info */
	nodemask_t *allowednodes;	/* zonelist_cache approximation */

	zlc = zonelist->zlcache_ptr;
	if (!zlc)
		return NULL;

S
S.Caglar Onur 已提交
1775
	if (time_after(jiffies, zlc->last_full_zap + HZ)) {
1776 1777 1778 1779 1780 1781
		bitmap_zero(zlc->fullzones, MAX_ZONES_PER_ZONELIST);
		zlc->last_full_zap = jiffies;
	}

	allowednodes = !in_interrupt() && (alloc_flags & ALLOC_CPUSET) ?
					&cpuset_current_mems_allowed :
1782
					&node_states[N_MEMORY];
1783 1784 1785 1786 1787 1788 1789 1790 1791 1792 1793 1794 1795 1796 1797 1798 1799 1800 1801 1802 1803 1804 1805 1806 1807
	return allowednodes;
}

/*
 * Given 'z' scanning a zonelist, run a couple of quick checks to see
 * if it is worth looking at further for free memory:
 *  1) Check that the zone isn't thought to be full (doesn't have its
 *     bit set in the zonelist_cache fullzones BITMAP).
 *  2) Check that the zones node (obtained from the zonelist_cache
 *     z_to_n[] mapping) is allowed in the passed in allowednodes mask.
 * Return true (non-zero) if zone is worth looking at further, or
 * else return false (zero) if it is not.
 *
 * This check -ignores- the distinction between various watermarks,
 * such as GFP_HIGH, GFP_ATOMIC, PF_MEMALLOC, ...  If a zone is
 * found to be full for any variation of these watermarks, it will
 * be considered full for up to one second by all requests, unless
 * we are so low on memory on all allowed nodes that we are forced
 * into the second scan of the zonelist.
 *
 * In the second scan we ignore this zonelist cache and exactly
 * apply the watermarks to all zones, even it is slower to do so.
 * We are low on memory in the second scan, and should leave no stone
 * unturned looking for a free page.
 */
1808
static int zlc_zone_worth_trying(struct zonelist *zonelist, struct zoneref *z,
1809 1810 1811 1812 1813 1814 1815 1816 1817 1818
						nodemask_t *allowednodes)
{
	struct zonelist_cache *zlc;	/* cached zonelist speedup info */
	int i;				/* index of *z in zonelist zones */
	int n;				/* node that zone *z is on */

	zlc = zonelist->zlcache_ptr;
	if (!zlc)
		return 1;

1819
	i = z - zonelist->_zonerefs;
1820 1821 1822 1823 1824 1825 1826 1827 1828 1829 1830
	n = zlc->z_to_n[i];

	/* This zone is worth trying if it is allowed but not full */
	return node_isset(n, *allowednodes) && !test_bit(i, zlc->fullzones);
}

/*
 * Given 'z' scanning a zonelist, set the corresponding bit in
 * zlc->fullzones, so that subsequent attempts to allocate a page
 * from that zone don't waste time re-examining it.
 */
1831
static void zlc_mark_zone_full(struct zonelist *zonelist, struct zoneref *z)
1832 1833 1834 1835 1836 1837 1838 1839
{
	struct zonelist_cache *zlc;	/* cached zonelist speedup info */
	int i;				/* index of *z in zonelist zones */

	zlc = zonelist->zlcache_ptr;
	if (!zlc)
		return;

1840
	i = z - zonelist->_zonerefs;
1841 1842 1843 1844

	set_bit(i, zlc->fullzones);
}

1845 1846 1847 1848 1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859
/*
 * clear all zones full, called after direct reclaim makes progress so that
 * a zone that was recently full is not skipped over for up to a second
 */
static void zlc_clear_zones_full(struct zonelist *zonelist)
{
	struct zonelist_cache *zlc;	/* cached zonelist speedup info */

	zlc = zonelist->zlcache_ptr;
	if (!zlc)
		return;

	bitmap_zero(zlc->fullzones, MAX_ZONES_PER_ZONELIST);
}

1860 1861
static bool zone_local(struct zone *local_zone, struct zone *zone)
{
1862
	return local_zone->node == zone->node;
1863 1864
}

1865 1866
static bool zone_allows_reclaim(struct zone *local_zone, struct zone *zone)
{
1867 1868
	return node_distance(zone_to_nid(local_zone), zone_to_nid(zone)) <
				RECLAIM_DISTANCE;
1869 1870
}

1871 1872 1873 1874 1875 1876 1877
#else	/* CONFIG_NUMA */

static nodemask_t *zlc_setup(struct zonelist *zonelist, int alloc_flags)
{
	return NULL;
}

1878
static int zlc_zone_worth_trying(struct zonelist *zonelist, struct zoneref *z,
1879 1880 1881 1882 1883
				nodemask_t *allowednodes)
{
	return 1;
}

1884
static void zlc_mark_zone_full(struct zonelist *zonelist, struct zoneref *z)
1885 1886
{
}
1887 1888 1889 1890

static void zlc_clear_zones_full(struct zonelist *zonelist)
{
}
1891

1892 1893 1894 1895 1896
static bool zone_local(struct zone *local_zone, struct zone *zone)
{
	return true;
}

1897 1898 1899 1900 1901
static bool zone_allows_reclaim(struct zone *local_zone, struct zone *zone)
{
	return true;
}

1902 1903
#endif	/* CONFIG_NUMA */

R
Rohit Seth 已提交
1904
/*
1905
 * get_page_from_freelist goes through the zonelist trying to allocate
R
Rohit Seth 已提交
1906 1907 1908
 * a page.
 */
static struct page *
1909
get_page_from_freelist(gfp_t gfp_mask, nodemask_t *nodemask, unsigned int order,
1910
		struct zonelist *zonelist, int high_zoneidx, int alloc_flags,
1911
		struct zone *preferred_zone, int migratetype)
M
Martin Hicks 已提交
1912
{
1913
	struct zoneref *z;
R
Rohit Seth 已提交
1914
	struct page *page = NULL;
1915
	int classzone_idx;
1916
	struct zone *zone;
1917 1918 1919
	nodemask_t *allowednodes = NULL;/* zonelist_cache approximation */
	int zlc_active = 0;		/* set if using zonelist_cache */
	int did_zlc_setup = 0;		/* just call zlc_setup() one time */
1920

1921
	classzone_idx = zone_idx(preferred_zone);
1922
zonelist_scan:
R
Rohit Seth 已提交
1923
	/*
1924
	 * Scan zonelist, looking for a zone with enough free.
1925
	 * See also __cpuset_node_allowed_softwall() comment in kernel/cpuset.c.
R
Rohit Seth 已提交
1926
	 */
1927 1928
	for_each_zone_zonelist_nodemask(zone, z, zonelist,
						high_zoneidx, nodemask) {
1929 1930
		unsigned long mark;

1931
		if (IS_ENABLED(CONFIG_NUMA) && zlc_active &&
1932 1933
			!zlc_zone_worth_trying(zonelist, z, allowednodes))
				continue;
1934 1935
		if (cpusets_enabled() &&
			(alloc_flags & ALLOC_CPUSET) &&
1936
			!cpuset_zone_allowed_softwall(zone, gfp_mask))
1937
				continue;
1938
		BUILD_BUG_ON(ALLOC_NO_WATERMARKS < NR_WMARK);
1939
		if (unlikely(alloc_flags & ALLOC_NO_WATERMARKS))
1940
			goto try_this_zone;
1941 1942 1943 1944 1945 1946
		/*
		 * Distribute pages in proportion to the individual
		 * zone size to ensure fair page aging.  The zone a
		 * page was allocated in should have no effect on the
		 * time the page has in memory before being reclaimed.
		 */
1947
		if (alloc_flags & ALLOC_FAIR) {
1948
			if (!zone_local(preferred_zone, zone))
1949
				continue;
1950 1951
			if (zone_page_state(zone, NR_ALLOC_BATCH) <= 0)
				continue;
1952
		}
1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980
		/*
		 * When allocating a page cache page for writing, we
		 * want to get it from a zone that is within its dirty
		 * limit, such that no single zone holds more than its
		 * proportional share of globally allowed dirty pages.
		 * The dirty limits take into account the zone's
		 * lowmem reserves and high watermark so that kswapd
		 * should be able to balance it without having to
		 * write pages from its LRU list.
		 *
		 * This may look like it could increase pressure on
		 * lower zones by failing allocations in higher zones
		 * before they are full.  But the pages that do spill
		 * over are limited as the lower zones are protected
		 * by this very same mechanism.  It should not become
		 * a practical burden to them.
		 *
		 * XXX: For now, allow allocations to potentially
		 * exceed the per-zone dirty limit in the slowpath
		 * (ALLOC_WMARK_LOW unset) before going into reclaim,
		 * which is important when on a NUMA setup the allowed
		 * zones are together not big enough to reach the
		 * global limit.  The proper fix for these situations
		 * will require awareness of zones in the
		 * dirty-throttling and the flusher threads.
		 */
		if ((alloc_flags & ALLOC_WMARK_LOW) &&
		    (gfp_mask & __GFP_WRITE) && !zone_dirty_ok(zone))
1981
			continue;
R
Rohit Seth 已提交
1982

1983 1984 1985
		mark = zone->watermark[alloc_flags & ALLOC_WMARK_MASK];
		if (!zone_watermark_ok(zone, order, mark,
				       classzone_idx, alloc_flags)) {
1986 1987
			int ret;

1988 1989
			if (IS_ENABLED(CONFIG_NUMA) &&
					!did_zlc_setup && nr_online_nodes > 1) {
1990 1991 1992 1993 1994 1995 1996 1997 1998 1999
				/*
				 * we do zlc_setup if there are multiple nodes
				 * and before considering the first zone allowed
				 * by the cpuset.
				 */
				allowednodes = zlc_setup(zonelist, alloc_flags);
				zlc_active = 1;
				did_zlc_setup = 1;
			}

2000 2001
			if (zone_reclaim_mode == 0 ||
			    !zone_allows_reclaim(preferred_zone, zone))
2002 2003
				goto this_zone_full;

2004 2005 2006 2007
			/*
			 * As we may have just activated ZLC, check if the first
			 * eligible zone has failed zone_reclaim recently.
			 */
2008
			if (IS_ENABLED(CONFIG_NUMA) && zlc_active &&
2009 2010 2011
				!zlc_zone_worth_trying(zonelist, z, allowednodes))
				continue;

2012 2013 2014 2015
			ret = zone_reclaim(zone, gfp_mask, order);
			switch (ret) {
			case ZONE_RECLAIM_NOSCAN:
				/* did not scan */
2016
				continue;
2017 2018
			case ZONE_RECLAIM_FULL:
				/* scanned but unreclaimable */
2019
				continue;
2020 2021
			default:
				/* did we reclaim enough */
2022
				if (zone_watermark_ok(zone, order, mark,
2023
						classzone_idx, alloc_flags))
2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036
					goto try_this_zone;

				/*
				 * Failed to reclaim enough to meet watermark.
				 * Only mark the zone full if checking the min
				 * watermark or if we failed to reclaim just
				 * 1<<order pages or else the page allocator
				 * fastpath will prematurely mark zones full
				 * when the watermark is between the low and
				 * min watermarks.
				 */
				if (((alloc_flags & ALLOC_WMARK_MASK) == ALLOC_WMARK_MIN) ||
				    ret == ZONE_RECLAIM_SOME)
2037
					goto this_zone_full;
2038 2039

				continue;
2040
			}
R
Rohit Seth 已提交
2041 2042
		}

2043
try_this_zone:
2044 2045
		page = buffered_rmqueue(preferred_zone, zone, order,
						gfp_mask, migratetype);
2046
		if (page)
R
Rohit Seth 已提交
2047
			break;
2048
this_zone_full:
2049
		if (IS_ENABLED(CONFIG_NUMA) && zlc_active)
2050
			zlc_mark_zone_full(zonelist, z);
2051
	}
2052

2053
	if (unlikely(IS_ENABLED(CONFIG_NUMA) && page == NULL && zlc_active)) {
2054 2055 2056 2057
		/* Disable zlc cache for second zonelist scan */
		zlc_active = 0;
		goto zonelist_scan;
	}
2058 2059 2060 2061 2062 2063 2064 2065 2066 2067 2068

	if (page)
		/*
		 * page->pfmemalloc is set when ALLOC_NO_WATERMARKS was
		 * necessary to allocate the page. The expectation is
		 * that the caller is taking steps that will free more
		 * memory. The caller should avoid the page being used
		 * for !PFMEMALLOC purposes.
		 */
		page->pfmemalloc = !!(alloc_flags & ALLOC_NO_WATERMARKS);

R
Rohit Seth 已提交
2069
	return page;
M
Martin Hicks 已提交
2070 2071
}

2072 2073 2074 2075 2076 2077 2078 2079 2080 2081 2082 2083 2084 2085
/*
 * Large machines with many possible nodes should not always dump per-node
 * meminfo in irq context.
 */
static inline bool should_suppress_show_mem(void)
{
	bool ret = false;

#if NODES_SHIFT > 8
	ret = in_interrupt();
#endif
	return ret;
}

2086 2087 2088 2089 2090 2091 2092 2093
static DEFINE_RATELIMIT_STATE(nopage_rs,
		DEFAULT_RATELIMIT_INTERVAL,
		DEFAULT_RATELIMIT_BURST);

void warn_alloc_failed(gfp_t gfp_mask, int order, const char *fmt, ...)
{
	unsigned int filter = SHOW_MEM_FILTER_NODES;

2094 2095
	if ((gfp_mask & __GFP_NOWARN) || !__ratelimit(&nopage_rs) ||
	    debug_guardpage_minorder() > 0)
2096 2097 2098 2099 2100 2101 2102 2103 2104 2105 2106 2107 2108 2109 2110
		return;

	/*
	 * This documents exceptions given to allocations in certain
	 * contexts that are allowed to allocate outside current's set
	 * of allowed nodes.
	 */
	if (!(gfp_mask & __GFP_NOMEMALLOC))
		if (test_thread_flag(TIF_MEMDIE) ||
		    (current->flags & (PF_MEMALLOC | PF_EXITING)))
			filter &= ~SHOW_MEM_FILTER_NODES;
	if (in_interrupt() || !(gfp_mask & __GFP_WAIT))
		filter &= ~SHOW_MEM_FILTER_NODES;

	if (fmt) {
J
Joe Perches 已提交
2111 2112 2113
		struct va_format vaf;
		va_list args;

2114
		va_start(args, fmt);
J
Joe Perches 已提交
2115 2116 2117 2118 2119 2120

		vaf.fmt = fmt;
		vaf.va = &args;

		pr_warn("%pV", &vaf);

2121 2122 2123
		va_end(args);
	}

J
Joe Perches 已提交
2124 2125
	pr_warn("%s: page allocation failure: order:%d, mode:0x%x\n",
		current->comm, order, gfp_mask);
2126 2127 2128 2129 2130 2131

	dump_stack();
	if (!should_suppress_show_mem())
		show_mem(filter);
}

2132 2133
static inline int
should_alloc_retry(gfp_t gfp_mask, unsigned int order,
2134
				unsigned long did_some_progress,
2135
				unsigned long pages_reclaimed)
L
Linus Torvalds 已提交
2136
{
2137 2138 2139
	/* Do not loop if specifically requested */
	if (gfp_mask & __GFP_NORETRY)
		return 0;
L
Linus Torvalds 已提交
2140

2141 2142 2143 2144 2145 2146 2147 2148 2149 2150 2151 2152
	/* Always retry if specifically requested */
	if (gfp_mask & __GFP_NOFAIL)
		return 1;

	/*
	 * Suspend converts GFP_KERNEL to __GFP_WAIT which can prevent reclaim
	 * making forward progress without invoking OOM. Suspend also disables
	 * storage devices so kswapd will not help. Bail if we are suspending.
	 */
	if (!did_some_progress && pm_suspended_storage())
		return 0;

2153 2154 2155 2156 2157 2158 2159 2160 2161 2162 2163 2164 2165 2166 2167 2168 2169
	/*
	 * In this implementation, order <= PAGE_ALLOC_COSTLY_ORDER
	 * means __GFP_NOFAIL, but that may not be true in other
	 * implementations.
	 */
	if (order <= PAGE_ALLOC_COSTLY_ORDER)
		return 1;

	/*
	 * For order > PAGE_ALLOC_COSTLY_ORDER, if __GFP_REPEAT is
	 * specified, then we retry until we no longer reclaim any pages
	 * (above), or we've reclaimed an order of pages at least as
	 * large as the allocation's order. In both cases, if the
	 * allocation still fails, we stop retrying.
	 */
	if (gfp_mask & __GFP_REPEAT && pages_reclaimed < (1 << order))
		return 1;
2170

2171 2172
	return 0;
}
2173

2174 2175 2176
static inline struct page *
__alloc_pages_may_oom(gfp_t gfp_mask, unsigned int order,
	struct zonelist *zonelist, enum zone_type high_zoneidx,
2177 2178
	nodemask_t *nodemask, struct zone *preferred_zone,
	int migratetype)
2179 2180 2181 2182
{
	struct page *page;

	/* Acquire the OOM killer lock for the zones in zonelist */
2183
	if (!try_set_zonelist_oom(zonelist, gfp_mask)) {
2184
		schedule_timeout_uninterruptible(1);
L
Linus Torvalds 已提交
2185 2186
		return NULL;
	}
2187

2188 2189 2190 2191 2192 2193 2194
	/*
	 * Go through the zonelist yet one more time, keep very high watermark
	 * here, this is only to catch a parallel oom killing, we must fail if
	 * we're still under heavy pressure.
	 */
	page = get_page_from_freelist(gfp_mask|__GFP_HARDWALL, nodemask,
		order, zonelist, high_zoneidx,
2195
		ALLOC_WMARK_HIGH|ALLOC_CPUSET,
2196
		preferred_zone, migratetype);
R
Rohit Seth 已提交
2197
	if (page)
2198 2199
		goto out;

2200 2201 2202 2203
	if (!(gfp_mask & __GFP_NOFAIL)) {
		/* The OOM killer will not help higher order allocs */
		if (order > PAGE_ALLOC_COSTLY_ORDER)
			goto out;
2204 2205 2206
		/* The OOM killer does not needlessly kill tasks for lowmem */
		if (high_zoneidx < ZONE_NORMAL)
			goto out;
2207 2208 2209 2210 2211 2212 2213 2214 2215 2216
		/*
		 * GFP_THISNODE contains __GFP_NORETRY and we never hit this.
		 * Sanity check for bare calls of __GFP_THISNODE, not real OOM.
		 * The caller should handle page allocation failure by itself if
		 * it specifies __GFP_THISNODE.
		 * Note: Hugepage uses it but will hit PAGE_ALLOC_COSTLY_ORDER.
		 */
		if (gfp_mask & __GFP_THISNODE)
			goto out;
	}
2217
	/* Exhausted what can be done so it's blamo time */
2218
	out_of_memory(zonelist, gfp_mask, order, nodemask, false);
2219 2220 2221 2222 2223 2224

out:
	clear_zonelist_oom(zonelist, gfp_mask);
	return page;
}

2225 2226 2227 2228 2229 2230
#ifdef CONFIG_COMPACTION
/* Try memory compaction for high-order allocations before reclaim */
static struct page *
__alloc_pages_direct_compact(gfp_t gfp_mask, unsigned int order,
	struct zonelist *zonelist, enum zone_type high_zoneidx,
	nodemask_t *nodemask, int alloc_flags, struct zone *preferred_zone,
2231
	int migratetype, enum migrate_mode mode,
2232
	bool *contended_compaction, bool *deferred_compaction,
2233
	unsigned long *did_some_progress)
2234
{
2235
	if (!order)
2236 2237
		return NULL;

2238
	if (compaction_deferred(preferred_zone, order)) {
2239 2240 2241 2242
		*deferred_compaction = true;
		return NULL;
	}

2243
	current->flags |= PF_MEMALLOC;
2244
	*did_some_progress = try_to_compact_pages(zonelist, order, gfp_mask,
2245
						nodemask, mode,
2246
						contended_compaction);
2247
	current->flags &= ~PF_MEMALLOC;
2248

2249
	if (*did_some_progress != COMPACT_SKIPPED) {
2250 2251
		struct page *page;

2252 2253 2254 2255 2256 2257
		/* Page migration frees to the PCP lists but we want merging */
		drain_pages(get_cpu());
		put_cpu();

		page = get_page_from_freelist(gfp_mask, nodemask,
				order, zonelist, high_zoneidx,
2258 2259
				alloc_flags & ~ALLOC_NO_WATERMARKS,
				preferred_zone, migratetype);
2260
		if (page) {
2261
			preferred_zone->compact_blockskip_flush = false;
2262
			compaction_defer_reset(preferred_zone, order, true);
2263 2264 2265 2266 2267 2268 2269 2270 2271 2272
			count_vm_event(COMPACTSUCCESS);
			return page;
		}

		/*
		 * It's bad if compaction run occurs and fails.
		 * The most likely reason is that pages exist,
		 * but not enough to satisfy watermarks.
		 */
		count_vm_event(COMPACTFAIL);
2273 2274 2275 2276 2277

		/*
		 * As async compaction considers a subset of pageblocks, only
		 * defer if the failure was a sync compaction failure.
		 */
2278
		if (mode != MIGRATE_ASYNC)
2279
			defer_compaction(preferred_zone, order);
2280 2281 2282 2283 2284 2285 2286 2287 2288 2289 2290

		cond_resched();
	}

	return NULL;
}
#else
static inline struct page *
__alloc_pages_direct_compact(gfp_t gfp_mask, unsigned int order,
	struct zonelist *zonelist, enum zone_type high_zoneidx,
	nodemask_t *nodemask, int alloc_flags, struct zone *preferred_zone,
2291 2292
	int migratetype, enum migrate_mode mode, bool *contended_compaction,
	bool *deferred_compaction, unsigned long *did_some_progress)
2293 2294 2295 2296 2297
{
	return NULL;
}
#endif /* CONFIG_COMPACTION */

2298 2299 2300 2301
/* Perform direct synchronous page reclaim */
static int
__perform_reclaim(gfp_t gfp_mask, unsigned int order, struct zonelist *zonelist,
		  nodemask_t *nodemask)
2302 2303
{
	struct reclaim_state reclaim_state;
2304
	int progress;
2305 2306 2307 2308 2309

	cond_resched();

	/* We now go into synchronous reclaim */
	cpuset_memory_pressure_bump();
2310
	current->flags |= PF_MEMALLOC;
2311 2312
	lockdep_set_current_reclaim_state(gfp_mask);
	reclaim_state.reclaimed_slab = 0;
2313
	current->reclaim_state = &reclaim_state;
2314

2315
	progress = try_to_free_pages(zonelist, order, gfp_mask, nodemask);
2316

2317
	current->reclaim_state = NULL;
2318
	lockdep_clear_current_reclaim_state();
2319
	current->flags &= ~PF_MEMALLOC;
2320 2321 2322

	cond_resched();

2323 2324 2325 2326 2327 2328 2329 2330 2331 2332 2333 2334 2335 2336 2337
	return progress;
}

/* The really slow allocator path where we enter direct reclaim */
static inline struct page *
__alloc_pages_direct_reclaim(gfp_t gfp_mask, unsigned int order,
	struct zonelist *zonelist, enum zone_type high_zoneidx,
	nodemask_t *nodemask, int alloc_flags, struct zone *preferred_zone,
	int migratetype, unsigned long *did_some_progress)
{
	struct page *page = NULL;
	bool drained = false;

	*did_some_progress = __perform_reclaim(gfp_mask, order, zonelist,
					       nodemask);
2338 2339
	if (unlikely(!(*did_some_progress)))
		return NULL;
2340

2341
	/* After successful reclaim, reconsider all zones for allocation */
2342
	if (IS_ENABLED(CONFIG_NUMA))
2343 2344
		zlc_clear_zones_full(zonelist);

2345 2346
retry:
	page = get_page_from_freelist(gfp_mask, nodemask, order,
2347
					zonelist, high_zoneidx,
2348 2349
					alloc_flags & ~ALLOC_NO_WATERMARKS,
					preferred_zone, migratetype);
2350 2351 2352 2353 2354 2355 2356 2357 2358 2359 2360

	/*
	 * If an allocation failed after direct reclaim, it could be because
	 * pages are pinned on the per-cpu lists. Drain them and try again
	 */
	if (!page && !drained) {
		drain_all_pages();
		drained = true;
		goto retry;
	}

2361 2362 2363
	return page;
}

L
Linus Torvalds 已提交
2364
/*
2365 2366
 * This is called in the allocator slow-path if the allocation request is of
 * sufficient urgency to ignore watermarks and take other desperate measures
L
Linus Torvalds 已提交
2367
 */
2368 2369 2370
static inline struct page *
__alloc_pages_high_priority(gfp_t gfp_mask, unsigned int order,
	struct zonelist *zonelist, enum zone_type high_zoneidx,
2371 2372
	nodemask_t *nodemask, struct zone *preferred_zone,
	int migratetype)
2373 2374 2375 2376 2377
{
	struct page *page;

	do {
		page = get_page_from_freelist(gfp_mask, nodemask, order,
2378
			zonelist, high_zoneidx, ALLOC_NO_WATERMARKS,
2379
			preferred_zone, migratetype);
2380 2381

		if (!page && gfp_mask & __GFP_NOFAIL)
2382
			wait_iff_congested(preferred_zone, BLK_RW_ASYNC, HZ/50);
2383 2384 2385 2386 2387
	} while (!page && (gfp_mask & __GFP_NOFAIL));

	return page;
}

2388 2389 2390
static void reset_alloc_batches(struct zonelist *zonelist,
				enum zone_type high_zoneidx,
				struct zone *preferred_zone)
L
Linus Torvalds 已提交
2391
{
2392 2393
	struct zoneref *z;
	struct zone *zone;
L
Linus Torvalds 已提交
2394

2395 2396 2397
	for_each_zone_zonelist(zone, z, zonelist, high_zoneidx) {
		/*
		 * Only reset the batches of zones that were actually
2398 2399
		 * considered in the fairness pass, we don't want to
		 * trash fairness information for zones that are not
2400 2401
		 * actually part of this zonelist's round-robin cycle.
		 */
2402
		if (!zone_local(preferred_zone, zone))
2403 2404
			continue;
		mod_zone_page_state(zone, NR_ALLOC_BATCH,
2405 2406
			high_wmark_pages(zone) - low_wmark_pages(zone) -
			atomic_long_read(&zone->vm_stat[NR_ALLOC_BATCH]));
2407
	}
2408
}
2409

2410 2411 2412 2413 2414 2415 2416 2417 2418 2419 2420 2421
static void wake_all_kswapds(unsigned int order,
			     struct zonelist *zonelist,
			     enum zone_type high_zoneidx,
			     struct zone *preferred_zone)
{
	struct zoneref *z;
	struct zone *zone;

	for_each_zone_zonelist(zone, z, zonelist, high_zoneidx)
		wakeup_kswapd(zone, order, zone_idx(preferred_zone));
}

2422 2423 2424 2425 2426
static inline int
gfp_to_alloc_flags(gfp_t gfp_mask)
{
	int alloc_flags = ALLOC_WMARK_MIN | ALLOC_CPUSET;
	const gfp_t wait = gfp_mask & __GFP_WAIT;
L
Linus Torvalds 已提交
2427

2428
	/* __GFP_HIGH is assumed to be the same as ALLOC_HIGH to save a branch. */
2429
	BUILD_BUG_ON(__GFP_HIGH != (__force gfp_t) ALLOC_HIGH);
2430

2431 2432 2433 2434 2435 2436
	/*
	 * The caller may dip into page reserves a bit more if the caller
	 * cannot run direct reclaim, or if the caller has realtime scheduling
	 * policy or is asking for __GFP_HIGH memory.  GFP_ATOMIC requests will
	 * set both ALLOC_HARDER (!wait) and ALLOC_HIGH (__GFP_HIGH).
	 */
2437
	alloc_flags |= (__force int) (gfp_mask & __GFP_HIGH);
L
Linus Torvalds 已提交
2438

2439
	if (!wait) {
2440 2441 2442 2443 2444 2445
		/*
		 * Not worth trying to allocate harder for
		 * __GFP_NOMEMALLOC even if it can't schedule.
		 */
		if  (!(gfp_mask & __GFP_NOMEMALLOC))
			alloc_flags |= ALLOC_HARDER;
2446
		/*
2447 2448
		 * Ignore cpuset if GFP_ATOMIC (!wait) rather than fail alloc.
		 * See also cpuset_zone_allowed() comment in kernel/cpuset.c.
2449
		 */
2450
		alloc_flags &= ~ALLOC_CPUSET;
2451
	} else if (unlikely(rt_task(current)) && !in_interrupt())
2452 2453
		alloc_flags |= ALLOC_HARDER;

2454 2455 2456
	if (likely(!(gfp_mask & __GFP_NOMEMALLOC))) {
		if (gfp_mask & __GFP_MEMALLOC)
			alloc_flags |= ALLOC_NO_WATERMARKS;
2457 2458 2459 2460 2461
		else if (in_serving_softirq() && (current->flags & PF_MEMALLOC))
			alloc_flags |= ALLOC_NO_WATERMARKS;
		else if (!in_interrupt() &&
				((current->flags & PF_MEMALLOC) ||
				 unlikely(test_thread_flag(TIF_MEMDIE))))
2462
			alloc_flags |= ALLOC_NO_WATERMARKS;
L
Linus Torvalds 已提交
2463
	}
2464 2465 2466 2467
#ifdef CONFIG_CMA
	if (allocflags_to_migratetype(gfp_mask) == MIGRATE_MOVABLE)
		alloc_flags |= ALLOC_CMA;
#endif
2468 2469 2470
	return alloc_flags;
}

2471 2472
bool gfp_pfmemalloc_allowed(gfp_t gfp_mask)
{
2473
	return !!(gfp_to_alloc_flags(gfp_mask) & ALLOC_NO_WATERMARKS);
2474 2475
}

2476 2477 2478
static inline struct page *
__alloc_pages_slowpath(gfp_t gfp_mask, unsigned int order,
	struct zonelist *zonelist, enum zone_type high_zoneidx,
2479 2480
	nodemask_t *nodemask, struct zone *preferred_zone,
	int migratetype)
2481 2482 2483 2484 2485 2486
{
	const gfp_t wait = gfp_mask & __GFP_WAIT;
	struct page *page = NULL;
	int alloc_flags;
	unsigned long pages_reclaimed = 0;
	unsigned long did_some_progress;
2487
	enum migrate_mode migration_mode = MIGRATE_ASYNC;
2488
	bool deferred_compaction = false;
2489
	bool contended_compaction = false;
L
Linus Torvalds 已提交
2490

2491 2492 2493 2494 2495 2496
	/*
	 * In the slowpath, we sanity check order to avoid ever trying to
	 * reclaim >= MAX_ORDER areas which will never succeed. Callers may
	 * be using allocators in order of preference for an area that is
	 * too large.
	 */
2497 2498
	if (order >= MAX_ORDER) {
		WARN_ON_ONCE(!(gfp_mask & __GFP_NOWARN));
2499
		return NULL;
2500
	}
L
Linus Torvalds 已提交
2501

2502 2503 2504 2505 2506 2507 2508 2509
	/*
	 * GFP_THISNODE (meaning __GFP_THISNODE, __GFP_NORETRY and
	 * __GFP_NOWARN set) should not cause reclaim since the subsystem
	 * (f.e. slab) using GFP_THISNODE may choose to trigger reclaim
	 * using a larger set of nodes after it has established that the
	 * allowed per node queues are empty and that nodes are
	 * over allocated.
	 */
2510 2511
	if (IS_ENABLED(CONFIG_NUMA) &&
	    (gfp_mask & GFP_THISNODE) == GFP_THISNODE)
2512 2513
		goto nopage;

2514
restart:
2515 2516
	if (!(gfp_mask & __GFP_NO_KSWAPD))
		wake_all_kswapds(order, zonelist, high_zoneidx, preferred_zone);
L
Linus Torvalds 已提交
2517

2518
	/*
R
Rohit Seth 已提交
2519 2520 2521
	 * OK, we're below the kswapd watermark and have kicked background
	 * reclaim. Now things get more complex, so set up alloc_flags according
	 * to how we want to proceed.
2522
	 */
2523
	alloc_flags = gfp_to_alloc_flags(gfp_mask);
L
Linus Torvalds 已提交
2524

2525 2526 2527 2528 2529 2530 2531 2532
	/*
	 * Find the true preferred zone if the allocation is unconstrained by
	 * cpusets.
	 */
	if (!(alloc_flags & ALLOC_CPUSET) && !nodemask)
		first_zones_zonelist(zonelist, high_zoneidx, NULL,
					&preferred_zone);

2533
rebalance:
2534
	/* This is the last chance, in general, before the goto nopage. */
2535
	page = get_page_from_freelist(gfp_mask, nodemask, order, zonelist,
2536 2537
			high_zoneidx, alloc_flags & ~ALLOC_NO_WATERMARKS,
			preferred_zone, migratetype);
R
Rohit Seth 已提交
2538 2539
	if (page)
		goto got_pg;
L
Linus Torvalds 已提交
2540

2541
	/* Allocate without watermarks if the context allows */
2542
	if (alloc_flags & ALLOC_NO_WATERMARKS) {
2543 2544 2545 2546 2547 2548 2549
		/*
		 * Ignore mempolicies if ALLOC_NO_WATERMARKS on the grounds
		 * the allocation is high priority and these type of
		 * allocations are system rather than user orientated
		 */
		zonelist = node_zonelist(numa_node_id(), gfp_mask);

2550 2551 2552
		page = __alloc_pages_high_priority(gfp_mask, order,
				zonelist, high_zoneidx, nodemask,
				preferred_zone, migratetype);
2553
		if (page) {
2554
			goto got_pg;
2555
		}
L
Linus Torvalds 已提交
2556 2557 2558
	}

	/* Atomic allocations - we can't balance anything */
2559 2560 2561 2562 2563 2564 2565
	if (!wait) {
		/*
		 * All existing users of the deprecated __GFP_NOFAIL are
		 * blockable, so warn of any new users that actually allow this
		 * type of allocation to fail.
		 */
		WARN_ON_ONCE(gfp_mask & __GFP_NOFAIL);
L
Linus Torvalds 已提交
2566
		goto nopage;
2567
	}
L
Linus Torvalds 已提交
2568

2569
	/* Avoid recursion of direct reclaim */
2570
	if (current->flags & PF_MEMALLOC)
2571 2572
		goto nopage;

2573 2574 2575 2576
	/* Avoid allocations with no watermarks from looping endlessly */
	if (test_thread_flag(TIF_MEMDIE) && !(gfp_mask & __GFP_NOFAIL))
		goto nopage;

2577 2578 2579 2580
	/*
	 * Try direct compaction. The first pass is asynchronous. Subsequent
	 * attempts after direct reclaim are synchronous
	 */
2581 2582 2583 2584
	page = __alloc_pages_direct_compact(gfp_mask, order, zonelist,
					high_zoneidx, nodemask, alloc_flags,
					preferred_zone, migratetype,
					migration_mode, &contended_compaction,
2585 2586
					&deferred_compaction,
					&did_some_progress);
2587 2588
	if (page)
		goto got_pg;
2589 2590 2591 2592 2593 2594 2595 2596

	/*
	 * It can become very expensive to allocate transparent hugepages at
	 * fault, so use asynchronous memory compaction for THP unless it is
	 * khugepaged trying to collapse.
	 */
	if (!(gfp_mask & __GFP_NO_KSWAPD) || (current->flags & PF_KTHREAD))
		migration_mode = MIGRATE_SYNC_LIGHT;
2597

2598 2599 2600 2601 2602 2603 2604
	/*
	 * If compaction is deferred for high-order allocations, it is because
	 * sync compaction recently failed. In this is the case and the caller
	 * requested a movable allocation that does not heavily disrupt the
	 * system then fail the allocation instead of entering direct reclaim.
	 */
	if ((deferred_compaction || contended_compaction) &&
2605
						(gfp_mask & __GFP_NO_KSWAPD))
2606
		goto nopage;
2607

2608 2609 2610 2611
	/* Try direct reclaim and then allocating */
	page = __alloc_pages_direct_reclaim(gfp_mask, order,
					zonelist, high_zoneidx,
					nodemask,
2612
					alloc_flags, preferred_zone,
2613
					migratetype, &did_some_progress);
2614 2615
	if (page)
		goto got_pg;
L
Linus Torvalds 已提交
2616

2617
	/*
2618 2619
	 * If we failed to make any progress reclaiming, then we are
	 * running out of options and have to consider going OOM
2620
	 */
2621
	if (!did_some_progress) {
2622
		if (oom_gfp_allowed(gfp_mask)) {
2623 2624
			if (oom_killer_disabled)
				goto nopage;
2625 2626 2627 2628
			/* Coredumps can quickly deplete all memory reserves */
			if ((current->flags & PF_DUMPCORE) &&
			    !(gfp_mask & __GFP_NOFAIL))
				goto nopage;
2629 2630
			page = __alloc_pages_may_oom(gfp_mask, order,
					zonelist, high_zoneidx,
2631 2632
					nodemask, preferred_zone,
					migratetype);
2633 2634
			if (page)
				goto got_pg;
L
Linus Torvalds 已提交
2635

2636 2637 2638 2639 2640 2641 2642 2643 2644 2645 2646 2647 2648 2649 2650 2651 2652
			if (!(gfp_mask & __GFP_NOFAIL)) {
				/*
				 * The oom killer is not called for high-order
				 * allocations that may fail, so if no progress
				 * is being made, there are no other options and
				 * retrying is unlikely to help.
				 */
				if (order > PAGE_ALLOC_COSTLY_ORDER)
					goto nopage;
				/*
				 * The oom killer is not called for lowmem
				 * allocations to prevent needlessly killing
				 * innocent tasks.
				 */
				if (high_zoneidx < ZONE_NORMAL)
					goto nopage;
			}
2653

2654 2655
			goto restart;
		}
L
Linus Torvalds 已提交
2656 2657
	}

2658
	/* Check if we should retry the allocation */
2659
	pages_reclaimed += did_some_progress;
2660 2661
	if (should_alloc_retry(gfp_mask, order, did_some_progress,
						pages_reclaimed)) {
2662
		/* Wait for some write requests to complete then retry */
2663
		wait_iff_congested(preferred_zone, BLK_RW_ASYNC, HZ/50);
L
Linus Torvalds 已提交
2664
		goto rebalance;
2665 2666 2667 2668 2669 2670
	} else {
		/*
		 * High-order allocations do not necessarily loop after
		 * direct reclaim and reclaim/compaction depends on compaction
		 * being called after reclaim so call directly if necessary
		 */
2671 2672 2673 2674
		page = __alloc_pages_direct_compact(gfp_mask, order, zonelist,
					high_zoneidx, nodemask, alloc_flags,
					preferred_zone, migratetype,
					migration_mode, &contended_compaction,
2675 2676
					&deferred_compaction,
					&did_some_progress);
2677 2678
		if (page)
			goto got_pg;
L
Linus Torvalds 已提交
2679 2680 2681
	}

nopage:
2682
	warn_alloc_failed(gfp_mask, order, NULL);
2683
	return page;
L
Linus Torvalds 已提交
2684
got_pg:
2685 2686
	if (kmemcheck_enabled)
		kmemcheck_pagealloc_alloc(page, order, gfp_mask);
2687

2688
	return page;
L
Linus Torvalds 已提交
2689
}
2690 2691 2692 2693 2694 2695 2696 2697 2698

/*
 * This is the 'heart' of the zoned buddy allocator.
 */
struct page *
__alloc_pages_nodemask(gfp_t gfp_mask, unsigned int order,
			struct zonelist *zonelist, nodemask_t *nodemask)
{
	enum zone_type high_zoneidx = gfp_zone(gfp_mask);
2699
	struct zone *preferred_zone;
2700
	struct page *page = NULL;
2701
	int migratetype = allocflags_to_migratetype(gfp_mask);
2702
	unsigned int cpuset_mems_cookie;
2703
	int alloc_flags = ALLOC_WMARK_LOW|ALLOC_CPUSET|ALLOC_FAIR;
2704

2705 2706
	gfp_mask &= gfp_allowed_mask;

2707 2708 2709 2710 2711 2712 2713 2714 2715 2716 2717 2718 2719 2720 2721
	lockdep_trace_alloc(gfp_mask);

	might_sleep_if(gfp_mask & __GFP_WAIT);

	if (should_fail_alloc_page(gfp_mask, order))
		return NULL;

	/*
	 * Check the zones suitable for the gfp_mask contain at least one
	 * valid zone. It's possible to have an empty zonelist as a result
	 * of GFP_THISNODE and a memoryless node
	 */
	if (unlikely(!zonelist->_zonerefs->zone))
		return NULL;

2722
retry_cpuset:
2723
	cpuset_mems_cookie = read_mems_allowed_begin();
2724

2725
	/* The preferred zone is used for statistics later */
2726 2727 2728
	first_zones_zonelist(zonelist, high_zoneidx,
				nodemask ? : &cpuset_current_mems_allowed,
				&preferred_zone);
2729 2730
	if (!preferred_zone)
		goto out;
2731

2732 2733 2734 2735
#ifdef CONFIG_CMA
	if (allocflags_to_migratetype(gfp_mask) == MIGRATE_MOVABLE)
		alloc_flags |= ALLOC_CMA;
#endif
2736
retry:
2737
	/* First allocation attempt */
2738
	page = get_page_from_freelist(gfp_mask|__GFP_HARDWALL, nodemask, order,
2739
			zonelist, high_zoneidx, alloc_flags,
2740
			preferred_zone, migratetype);
2741
	if (unlikely(!page)) {
2742 2743 2744 2745 2746 2747 2748 2749 2750 2751 2752 2753 2754 2755 2756 2757
		/*
		 * The first pass makes sure allocations are spread
		 * fairly within the local node.  However, the local
		 * node might have free pages left after the fairness
		 * batches are exhausted, and remote zones haven't
		 * even been considered yet.  Try once more without
		 * fairness, and include remote zones now, before
		 * entering the slowpath and waking kswapd: prefer
		 * spilling to a remote zone over swapping locally.
		 */
		if (alloc_flags & ALLOC_FAIR) {
			reset_alloc_batches(zonelist, high_zoneidx,
					    preferred_zone);
			alloc_flags &= ~ALLOC_FAIR;
			goto retry;
		}
2758 2759 2760 2761 2762 2763
		/*
		 * Runtime PM, block IO and its error handling path
		 * can deadlock because I/O on the device might not
		 * complete.
		 */
		gfp_mask = memalloc_noio_flags(gfp_mask);
2764
		page = __alloc_pages_slowpath(gfp_mask, order,
2765
				zonelist, high_zoneidx, nodemask,
2766
				preferred_zone, migratetype);
2767
	}
2768

2769
	trace_mm_page_alloc(page, order, gfp_mask, migratetype);
2770 2771 2772 2773 2774 2775 2776 2777

out:
	/*
	 * When updating a task's mems_allowed, it is possible to race with
	 * parallel threads in such a way that an allocation can fail while
	 * the mask is being updated. If a page allocation is about to fail,
	 * check if the cpuset changed during allocation and if so, retry.
	 */
2778
	if (unlikely(!page && read_mems_allowed_retry(cpuset_mems_cookie)))
2779 2780
		goto retry_cpuset;

2781
	return page;
L
Linus Torvalds 已提交
2782
}
2783
EXPORT_SYMBOL(__alloc_pages_nodemask);
L
Linus Torvalds 已提交
2784 2785 2786 2787

/*
 * Common helper functions.
 */
H
Harvey Harrison 已提交
2788
unsigned long __get_free_pages(gfp_t gfp_mask, unsigned int order)
L
Linus Torvalds 已提交
2789
{
2790 2791 2792 2793 2794 2795 2796 2797
	struct page *page;

	/*
	 * __get_free_pages() returns a 32-bit address, which cannot represent
	 * a highmem page
	 */
	VM_BUG_ON((gfp_mask & __GFP_HIGHMEM) != 0);

L
Linus Torvalds 已提交
2798 2799 2800 2801 2802 2803 2804
	page = alloc_pages(gfp_mask, order);
	if (!page)
		return 0;
	return (unsigned long) page_address(page);
}
EXPORT_SYMBOL(__get_free_pages);

H
Harvey Harrison 已提交
2805
unsigned long get_zeroed_page(gfp_t gfp_mask)
L
Linus Torvalds 已提交
2806
{
2807
	return __get_free_pages(gfp_mask | __GFP_ZERO, 0);
L
Linus Torvalds 已提交
2808 2809 2810
}
EXPORT_SYMBOL(get_zeroed_page);

H
Harvey Harrison 已提交
2811
void __free_pages(struct page *page, unsigned int order)
L
Linus Torvalds 已提交
2812
{
N
Nick Piggin 已提交
2813
	if (put_page_testzero(page)) {
L
Linus Torvalds 已提交
2814
		if (order == 0)
L
Li Hong 已提交
2815
			free_hot_cold_page(page, 0);
L
Linus Torvalds 已提交
2816 2817 2818 2819 2820 2821 2822
		else
			__free_pages_ok(page, order);
	}
}

EXPORT_SYMBOL(__free_pages);

H
Harvey Harrison 已提交
2823
void free_pages(unsigned long addr, unsigned int order)
L
Linus Torvalds 已提交
2824 2825
{
	if (addr != 0) {
N
Nick Piggin 已提交
2826
		VM_BUG_ON(!virt_addr_valid((void *)addr));
L
Linus Torvalds 已提交
2827 2828 2829 2830 2831 2832
		__free_pages(virt_to_page((void *)addr), order);
	}
}

EXPORT_SYMBOL(free_pages);

2833
/*
V
Vladimir Davydov 已提交
2834 2835
 * alloc_kmem_pages charges newly allocated pages to the kmem resource counter
 * of the current memory cgroup.
2836
 *
V
Vladimir Davydov 已提交
2837 2838 2839 2840 2841 2842 2843 2844 2845 2846 2847 2848 2849 2850 2851 2852 2853 2854 2855 2856 2857 2858 2859 2860 2861 2862 2863 2864 2865 2866
 * It should be used when the caller would like to use kmalloc, but since the
 * allocation is large, it has to fall back to the page allocator.
 */
struct page *alloc_kmem_pages(gfp_t gfp_mask, unsigned int order)
{
	struct page *page;
	struct mem_cgroup *memcg = NULL;

	if (!memcg_kmem_newpage_charge(gfp_mask, &memcg, order))
		return NULL;
	page = alloc_pages(gfp_mask, order);
	memcg_kmem_commit_charge(page, memcg, order);
	return page;
}

struct page *alloc_kmem_pages_node(int nid, gfp_t gfp_mask, unsigned int order)
{
	struct page *page;
	struct mem_cgroup *memcg = NULL;

	if (!memcg_kmem_newpage_charge(gfp_mask, &memcg, order))
		return NULL;
	page = alloc_pages_node(nid, gfp_mask, order);
	memcg_kmem_commit_charge(page, memcg, order);
	return page;
}

/*
 * __free_kmem_pages and free_kmem_pages will free pages allocated with
 * alloc_kmem_pages.
2867
 */
V
Vladimir Davydov 已提交
2868
void __free_kmem_pages(struct page *page, unsigned int order)
2869 2870 2871 2872 2873
{
	memcg_kmem_uncharge_pages(page, order);
	__free_pages(page, order);
}

V
Vladimir Davydov 已提交
2874
void free_kmem_pages(unsigned long addr, unsigned int order)
2875 2876 2877
{
	if (addr != 0) {
		VM_BUG_ON(!virt_addr_valid((void *)addr));
V
Vladimir Davydov 已提交
2878
		__free_kmem_pages(virt_to_page((void *)addr), order);
2879 2880 2881
	}
}

A
Andi Kleen 已提交
2882 2883 2884 2885 2886 2887 2888 2889 2890 2891 2892 2893 2894 2895 2896
static void *make_alloc_exact(unsigned long addr, unsigned order, size_t size)
{
	if (addr) {
		unsigned long alloc_end = addr + (PAGE_SIZE << order);
		unsigned long used = addr + PAGE_ALIGN(size);

		split_page(virt_to_page((void *)addr), order);
		while (used < alloc_end) {
			free_page(used);
			used += PAGE_SIZE;
		}
	}
	return (void *)addr;
}

2897 2898 2899 2900 2901 2902 2903 2904 2905 2906 2907 2908 2909 2910 2911 2912 2913 2914 2915
/**
 * alloc_pages_exact - allocate an exact number physically-contiguous pages.
 * @size: the number of bytes to allocate
 * @gfp_mask: GFP flags for the allocation
 *
 * This function is similar to alloc_pages(), except that it allocates the
 * minimum number of pages to satisfy the request.  alloc_pages() can only
 * allocate memory in power-of-two pages.
 *
 * This function is also limited by MAX_ORDER.
 *
 * Memory allocated by this function must be released by free_pages_exact().
 */
void *alloc_pages_exact(size_t size, gfp_t gfp_mask)
{
	unsigned int order = get_order(size);
	unsigned long addr;

	addr = __get_free_pages(gfp_mask, order);
A
Andi Kleen 已提交
2916
	return make_alloc_exact(addr, order, size);
2917 2918 2919
}
EXPORT_SYMBOL(alloc_pages_exact);

A
Andi Kleen 已提交
2920 2921 2922
/**
 * alloc_pages_exact_nid - allocate an exact number of physically-contiguous
 *			   pages on a node.
2923
 * @nid: the preferred node ID where memory should be allocated
A
Andi Kleen 已提交
2924 2925 2926 2927 2928 2929 2930 2931 2932 2933 2934 2935 2936 2937 2938 2939 2940 2941
 * @size: the number of bytes to allocate
 * @gfp_mask: GFP flags for the allocation
 *
 * Like alloc_pages_exact(), but try to allocate on node nid first before falling
 * back.
 * Note this is not alloc_pages_exact_node() which allocates on a specific node,
 * but is not exact.
 */
void *alloc_pages_exact_nid(int nid, size_t size, gfp_t gfp_mask)
{
	unsigned order = get_order(size);
	struct page *p = alloc_pages_node(nid, gfp_mask, order);
	if (!p)
		return NULL;
	return make_alloc_exact((unsigned long)page_address(p), order, size);
}
EXPORT_SYMBOL(alloc_pages_exact_nid);

2942 2943 2944 2945 2946 2947 2948 2949 2950 2951 2952 2953 2954 2955 2956 2957 2958 2959 2960
/**
 * free_pages_exact - release memory allocated via alloc_pages_exact()
 * @virt: the value returned by alloc_pages_exact.
 * @size: size of allocation, same value as passed to alloc_pages_exact().
 *
 * Release the memory allocated by a previous call to alloc_pages_exact.
 */
void free_pages_exact(void *virt, size_t size)
{
	unsigned long addr = (unsigned long)virt;
	unsigned long end = addr + PAGE_ALIGN(size);

	while (addr < end) {
		free_page(addr);
		addr += PAGE_SIZE;
	}
}
EXPORT_SYMBOL(free_pages_exact);

2961 2962 2963 2964 2965 2966 2967
/**
 * nr_free_zone_pages - count number of pages beyond high watermark
 * @offset: The zone index of the highest zone
 *
 * nr_free_zone_pages() counts the number of counts pages which are beyond the
 * high watermark within all zones at or below a given zone index.  For each
 * zone, the number of pages is calculated as:
2968
 *     managed_pages - high_pages
2969
 */
2970
static unsigned long nr_free_zone_pages(int offset)
L
Linus Torvalds 已提交
2971
{
2972
	struct zoneref *z;
2973 2974
	struct zone *zone;

2975
	/* Just pick one node, since fallback list is circular */
2976
	unsigned long sum = 0;
L
Linus Torvalds 已提交
2977

2978
	struct zonelist *zonelist = node_zonelist(numa_node_id(), GFP_KERNEL);
L
Linus Torvalds 已提交
2979

2980
	for_each_zone_zonelist(zone, z, zonelist, offset) {
2981
		unsigned long size = zone->managed_pages;
2982
		unsigned long high = high_wmark_pages(zone);
2983 2984
		if (size > high)
			sum += size - high;
L
Linus Torvalds 已提交
2985 2986 2987 2988 2989
	}

	return sum;
}

2990 2991 2992 2993 2994
/**
 * nr_free_buffer_pages - count number of pages beyond high watermark
 *
 * nr_free_buffer_pages() counts the number of pages which are beyond the high
 * watermark within ZONE_DMA and ZONE_NORMAL.
L
Linus Torvalds 已提交
2995
 */
2996
unsigned long nr_free_buffer_pages(void)
L
Linus Torvalds 已提交
2997
{
A
Al Viro 已提交
2998
	return nr_free_zone_pages(gfp_zone(GFP_USER));
L
Linus Torvalds 已提交
2999
}
3000
EXPORT_SYMBOL_GPL(nr_free_buffer_pages);
L
Linus Torvalds 已提交
3001

3002 3003 3004 3005 3006
/**
 * nr_free_pagecache_pages - count number of pages beyond high watermark
 *
 * nr_free_pagecache_pages() counts the number of pages which are beyond the
 * high watermark within all zones.
L
Linus Torvalds 已提交
3007
 */
3008
unsigned long nr_free_pagecache_pages(void)
L
Linus Torvalds 已提交
3009
{
M
Mel Gorman 已提交
3010
	return nr_free_zone_pages(gfp_zone(GFP_HIGHUSER_MOVABLE));
L
Linus Torvalds 已提交
3011
}
3012 3013

static inline void show_node(struct zone *zone)
L
Linus Torvalds 已提交
3014
{
3015
	if (IS_ENABLED(CONFIG_NUMA))
3016
		printk("Node %d ", zone_to_nid(zone));
L
Linus Torvalds 已提交
3017 3018 3019 3020 3021 3022
}

void si_meminfo(struct sysinfo *val)
{
	val->totalram = totalram_pages;
	val->sharedram = 0;
3023
	val->freeram = global_page_state(NR_FREE_PAGES);
L
Linus Torvalds 已提交
3024 3025 3026 3027 3028 3029 3030 3031 3032 3033 3034
	val->bufferram = nr_blockdev_pages();
	val->totalhigh = totalhigh_pages;
	val->freehigh = nr_free_highpages();
	val->mem_unit = PAGE_SIZE;
}

EXPORT_SYMBOL(si_meminfo);

#ifdef CONFIG_NUMA
void si_meminfo_node(struct sysinfo *val, int nid)
{
3035 3036
	int zone_type;		/* needs to be signed */
	unsigned long managed_pages = 0;
L
Linus Torvalds 已提交
3037 3038
	pg_data_t *pgdat = NODE_DATA(nid);

3039 3040 3041
	for (zone_type = 0; zone_type < MAX_NR_ZONES; zone_type++)
		managed_pages += pgdat->node_zones[zone_type].managed_pages;
	val->totalram = managed_pages;
3042
	val->freeram = node_page_state(nid, NR_FREE_PAGES);
3043
#ifdef CONFIG_HIGHMEM
3044
	val->totalhigh = pgdat->node_zones[ZONE_HIGHMEM].managed_pages;
3045 3046
	val->freehigh = zone_page_state(&pgdat->node_zones[ZONE_HIGHMEM],
			NR_FREE_PAGES);
3047 3048 3049 3050
#else
	val->totalhigh = 0;
	val->freehigh = 0;
#endif
L
Linus Torvalds 已提交
3051 3052 3053 3054
	val->mem_unit = PAGE_SIZE;
}
#endif

3055
/*
3056 3057
 * Determine whether the node should be displayed or not, depending on whether
 * SHOW_MEM_FILTER_NODES was passed to show_free_areas().
3058
 */
3059
bool skip_free_areas_node(unsigned int flags, int nid)
3060 3061
{
	bool ret = false;
3062
	unsigned int cpuset_mems_cookie;
3063 3064 3065 3066

	if (!(flags & SHOW_MEM_FILTER_NODES))
		goto out;

3067
	do {
3068
		cpuset_mems_cookie = read_mems_allowed_begin();
3069
		ret = !node_isset(nid, cpuset_current_mems_allowed);
3070
	} while (read_mems_allowed_retry(cpuset_mems_cookie));
3071 3072 3073 3074
out:
	return ret;
}

L
Linus Torvalds 已提交
3075 3076
#define K(x) ((x) << (PAGE_SHIFT-10))

3077 3078 3079 3080 3081 3082 3083 3084 3085 3086
static void show_migration_types(unsigned char type)
{
	static const char types[MIGRATE_TYPES] = {
		[MIGRATE_UNMOVABLE]	= 'U',
		[MIGRATE_RECLAIMABLE]	= 'E',
		[MIGRATE_MOVABLE]	= 'M',
		[MIGRATE_RESERVE]	= 'R',
#ifdef CONFIG_CMA
		[MIGRATE_CMA]		= 'C',
#endif
3087
#ifdef CONFIG_MEMORY_ISOLATION
3088
		[MIGRATE_ISOLATE]	= 'I',
3089
#endif
3090 3091 3092 3093 3094 3095 3096 3097 3098 3099 3100 3101 3102 3103
	};
	char tmp[MIGRATE_TYPES + 1];
	char *p = tmp;
	int i;

	for (i = 0; i < MIGRATE_TYPES; i++) {
		if (type & (1 << i))
			*p++ = types[i];
	}

	*p = '\0';
	printk("(%s) ", tmp);
}

L
Linus Torvalds 已提交
3104 3105 3106 3107
/*
 * Show free area list (used inside shift_scroll-lock stuff)
 * We also calculate the percentage fragmentation. We do this by counting the
 * memory on each free list with the exception of the first item on the list.
3108 3109
 * Suppresses nodes that are not allowed by current's cpuset if
 * SHOW_MEM_FILTER_NODES is passed.
L
Linus Torvalds 已提交
3110
 */
3111
void show_free_areas(unsigned int filter)
L
Linus Torvalds 已提交
3112
{
3113
	int cpu;
L
Linus Torvalds 已提交
3114 3115
	struct zone *zone;

3116
	for_each_populated_zone(zone) {
3117
		if (skip_free_areas_node(filter, zone_to_nid(zone)))
3118
			continue;
3119 3120
		show_node(zone);
		printk("%s per-cpu:\n", zone->name);
L
Linus Torvalds 已提交
3121

3122
		for_each_online_cpu(cpu) {
L
Linus Torvalds 已提交
3123 3124
			struct per_cpu_pageset *pageset;

3125
			pageset = per_cpu_ptr(zone->pageset, cpu);
L
Linus Torvalds 已提交
3126

3127 3128 3129
			printk("CPU %4d: hi:%5d, btch:%4d usd:%4d\n",
			       cpu, pageset->pcp.high,
			       pageset->pcp.batch, pageset->pcp.count);
L
Linus Torvalds 已提交
3130 3131 3132
		}
	}

K
KOSAKI Motohiro 已提交
3133 3134
	printk("active_anon:%lu inactive_anon:%lu isolated_anon:%lu\n"
		" active_file:%lu inactive_file:%lu isolated_file:%lu\n"
L
Lee Schermerhorn 已提交
3135
		" unevictable:%lu"
3136
		" dirty:%lu writeback:%lu unstable:%lu\n"
3137
		" free:%lu slab_reclaimable:%lu slab_unreclaimable:%lu\n"
3138 3139
		" mapped:%lu shmem:%lu pagetables:%lu bounce:%lu\n"
		" free_cma:%lu\n",
3140 3141
		global_page_state(NR_ACTIVE_ANON),
		global_page_state(NR_INACTIVE_ANON),
K
KOSAKI Motohiro 已提交
3142 3143
		global_page_state(NR_ISOLATED_ANON),
		global_page_state(NR_ACTIVE_FILE),
3144
		global_page_state(NR_INACTIVE_FILE),
K
KOSAKI Motohiro 已提交
3145
		global_page_state(NR_ISOLATED_FILE),
L
Lee Schermerhorn 已提交
3146
		global_page_state(NR_UNEVICTABLE),
3147
		global_page_state(NR_FILE_DIRTY),
3148
		global_page_state(NR_WRITEBACK),
3149
		global_page_state(NR_UNSTABLE_NFS),
3150
		global_page_state(NR_FREE_PAGES),
3151 3152
		global_page_state(NR_SLAB_RECLAIMABLE),
		global_page_state(NR_SLAB_UNRECLAIMABLE),
3153
		global_page_state(NR_FILE_MAPPED),
3154
		global_page_state(NR_SHMEM),
3155
		global_page_state(NR_PAGETABLE),
3156 3157
		global_page_state(NR_BOUNCE),
		global_page_state(NR_FREE_CMA_PAGES));
L
Linus Torvalds 已提交
3158

3159
	for_each_populated_zone(zone) {
L
Linus Torvalds 已提交
3160 3161
		int i;

3162
		if (skip_free_areas_node(filter, zone_to_nid(zone)))
3163
			continue;
L
Linus Torvalds 已提交
3164 3165 3166 3167 3168 3169
		show_node(zone);
		printk("%s"
			" free:%lukB"
			" min:%lukB"
			" low:%lukB"
			" high:%lukB"
3170 3171 3172 3173
			" active_anon:%lukB"
			" inactive_anon:%lukB"
			" active_file:%lukB"
			" inactive_file:%lukB"
L
Lee Schermerhorn 已提交
3174
			" unevictable:%lukB"
K
KOSAKI Motohiro 已提交
3175 3176
			" isolated(anon):%lukB"
			" isolated(file):%lukB"
L
Linus Torvalds 已提交
3177
			" present:%lukB"
3178
			" managed:%lukB"
3179 3180 3181 3182
			" mlocked:%lukB"
			" dirty:%lukB"
			" writeback:%lukB"
			" mapped:%lukB"
3183
			" shmem:%lukB"
3184 3185
			" slab_reclaimable:%lukB"
			" slab_unreclaimable:%lukB"
3186
			" kernel_stack:%lukB"
3187 3188 3189
			" pagetables:%lukB"
			" unstable:%lukB"
			" bounce:%lukB"
3190
			" free_cma:%lukB"
3191
			" writeback_tmp:%lukB"
L
Linus Torvalds 已提交
3192 3193 3194 3195
			" pages_scanned:%lu"
			" all_unreclaimable? %s"
			"\n",
			zone->name,
3196
			K(zone_page_state(zone, NR_FREE_PAGES)),
3197 3198 3199
			K(min_wmark_pages(zone)),
			K(low_wmark_pages(zone)),
			K(high_wmark_pages(zone)),
3200 3201 3202 3203
			K(zone_page_state(zone, NR_ACTIVE_ANON)),
			K(zone_page_state(zone, NR_INACTIVE_ANON)),
			K(zone_page_state(zone, NR_ACTIVE_FILE)),
			K(zone_page_state(zone, NR_INACTIVE_FILE)),
L
Lee Schermerhorn 已提交
3204
			K(zone_page_state(zone, NR_UNEVICTABLE)),
K
KOSAKI Motohiro 已提交
3205 3206
			K(zone_page_state(zone, NR_ISOLATED_ANON)),
			K(zone_page_state(zone, NR_ISOLATED_FILE)),
L
Linus Torvalds 已提交
3207
			K(zone->present_pages),
3208
			K(zone->managed_pages),
3209 3210 3211 3212
			K(zone_page_state(zone, NR_MLOCK)),
			K(zone_page_state(zone, NR_FILE_DIRTY)),
			K(zone_page_state(zone, NR_WRITEBACK)),
			K(zone_page_state(zone, NR_FILE_MAPPED)),
3213
			K(zone_page_state(zone, NR_SHMEM)),
3214 3215
			K(zone_page_state(zone, NR_SLAB_RECLAIMABLE)),
			K(zone_page_state(zone, NR_SLAB_UNRECLAIMABLE)),
3216 3217
			zone_page_state(zone, NR_KERNEL_STACK) *
				THREAD_SIZE / 1024,
3218 3219 3220
			K(zone_page_state(zone, NR_PAGETABLE)),
			K(zone_page_state(zone, NR_UNSTABLE_NFS)),
			K(zone_page_state(zone, NR_BOUNCE)),
3221
			K(zone_page_state(zone, NR_FREE_CMA_PAGES)),
3222
			K(zone_page_state(zone, NR_WRITEBACK_TEMP)),
L
Linus Torvalds 已提交
3223
			zone->pages_scanned,
3224
			(!zone_reclaimable(zone) ? "yes" : "no")
L
Linus Torvalds 已提交
3225 3226 3227 3228 3229 3230 3231
			);
		printk("lowmem_reserve[]:");
		for (i = 0; i < MAX_NR_ZONES; i++)
			printk(" %lu", zone->lowmem_reserve[i]);
		printk("\n");
	}

3232
	for_each_populated_zone(zone) {
3233
		unsigned long nr[MAX_ORDER], flags, order, total = 0;
3234
		unsigned char types[MAX_ORDER];
L
Linus Torvalds 已提交
3235

3236
		if (skip_free_areas_node(filter, zone_to_nid(zone)))
3237
			continue;
L
Linus Torvalds 已提交
3238 3239 3240 3241 3242
		show_node(zone);
		printk("%s: ", zone->name);

		spin_lock_irqsave(&zone->lock, flags);
		for (order = 0; order < MAX_ORDER; order++) {
3243 3244 3245 3246
			struct free_area *area = &zone->free_area[order];
			int type;

			nr[order] = area->nr_free;
3247
			total += nr[order] << order;
3248 3249 3250 3251 3252 3253

			types[order] = 0;
			for (type = 0; type < MIGRATE_TYPES; type++) {
				if (!list_empty(&area->free_list[type]))
					types[order] |= 1 << type;
			}
L
Linus Torvalds 已提交
3254 3255
		}
		spin_unlock_irqrestore(&zone->lock, flags);
3256
		for (order = 0; order < MAX_ORDER; order++) {
3257
			printk("%lu*%lukB ", nr[order], K(1UL) << order);
3258 3259 3260
			if (nr[order])
				show_migration_types(types[order]);
		}
L
Linus Torvalds 已提交
3261 3262 3263
		printk("= %lukB\n", K(total));
	}

3264 3265
	hugetlb_show_meminfo();

3266 3267
	printk("%ld total pagecache pages\n", global_page_state(NR_FILE_PAGES));

L
Linus Torvalds 已提交
3268 3269 3270
	show_swap_cache_info();
}

3271 3272 3273 3274 3275 3276
static void zoneref_set_zone(struct zone *zone, struct zoneref *zoneref)
{
	zoneref->zone = zone;
	zoneref->zone_idx = zone_idx(zone);
}

L
Linus Torvalds 已提交
3277 3278
/*
 * Builds allocation fallback zone lists.
3279 3280
 *
 * Add all populated zones of a node to the zonelist.
L
Linus Torvalds 已提交
3281
 */
3282
static int build_zonelists_node(pg_data_t *pgdat, struct zonelist *zonelist,
3283
				int nr_zones)
L
Linus Torvalds 已提交
3284
{
3285
	struct zone *zone;
3286
	enum zone_type zone_type = MAX_NR_ZONES;
3287 3288

	do {
3289
		zone_type--;
3290
		zone = pgdat->node_zones + zone_type;
3291
		if (populated_zone(zone)) {
3292 3293
			zoneref_set_zone(zone,
				&zonelist->_zonerefs[nr_zones++]);
3294
			check_highest_zone(zone_type);
L
Linus Torvalds 已提交
3295
		}
3296
	} while (zone_type);
3297

3298
	return nr_zones;
L
Linus Torvalds 已提交
3299 3300
}

3301 3302 3303 3304 3305 3306 3307 3308 3309 3310 3311 3312 3313 3314 3315 3316 3317 3318 3319 3320 3321

/*
 *  zonelist_order:
 *  0 = automatic detection of better ordering.
 *  1 = order by ([node] distance, -zonetype)
 *  2 = order by (-zonetype, [node] distance)
 *
 *  If not NUMA, ZONELIST_ORDER_ZONE and ZONELIST_ORDER_NODE will create
 *  the same zonelist. So only NUMA can configure this param.
 */
#define ZONELIST_ORDER_DEFAULT  0
#define ZONELIST_ORDER_NODE     1
#define ZONELIST_ORDER_ZONE     2

/* zonelist order in the kernel.
 * set_zonelist_order() will set this to NODE or ZONE.
 */
static int current_zonelist_order = ZONELIST_ORDER_DEFAULT;
static char zonelist_order_name[3][8] = {"Default", "Node", "Zone"};


L
Linus Torvalds 已提交
3322
#ifdef CONFIG_NUMA
3323 3324 3325 3326 3327 3328 3329 3330 3331 3332 3333 3334 3335 3336 3337 3338 3339 3340 3341 3342 3343 3344 3345 3346 3347 3348 3349 3350 3351 3352 3353 3354 3355
/* The value user specified ....changed by config */
static int user_zonelist_order = ZONELIST_ORDER_DEFAULT;
/* string for sysctl */
#define NUMA_ZONELIST_ORDER_LEN	16
char numa_zonelist_order[16] = "default";

/*
 * interface for configure zonelist ordering.
 * command line option "numa_zonelist_order"
 *	= "[dD]efault	- default, automatic configuration.
 *	= "[nN]ode 	- order by node locality, then by zone within node
 *	= "[zZ]one      - order by zone, then by locality within zone
 */

static int __parse_numa_zonelist_order(char *s)
{
	if (*s == 'd' || *s == 'D') {
		user_zonelist_order = ZONELIST_ORDER_DEFAULT;
	} else if (*s == 'n' || *s == 'N') {
		user_zonelist_order = ZONELIST_ORDER_NODE;
	} else if (*s == 'z' || *s == 'Z') {
		user_zonelist_order = ZONELIST_ORDER_ZONE;
	} else {
		printk(KERN_WARNING
			"Ignoring invalid numa_zonelist_order value:  "
			"%s\n", s);
		return -EINVAL;
	}
	return 0;
}

static __init int setup_numa_zonelist_order(char *s)
{
3356 3357 3358 3359 3360 3361 3362 3363 3364 3365
	int ret;

	if (!s)
		return 0;

	ret = __parse_numa_zonelist_order(s);
	if (ret == 0)
		strlcpy(numa_zonelist_order, s, NUMA_ZONELIST_ORDER_LEN);

	return ret;
3366 3367 3368 3369 3370 3371 3372
}
early_param("numa_zonelist_order", setup_numa_zonelist_order);

/*
 * sysctl handler for numa_zonelist_order
 */
int numa_zonelist_order_handler(ctl_table *table, int write,
3373
		void __user *buffer, size_t *length,
3374 3375 3376 3377
		loff_t *ppos)
{
	char saved_string[NUMA_ZONELIST_ORDER_LEN];
	int ret;
3378
	static DEFINE_MUTEX(zl_order_mutex);
3379

3380
	mutex_lock(&zl_order_mutex);
3381 3382 3383 3384 3385 3386 3387
	if (write) {
		if (strlen((char *)table->data) >= NUMA_ZONELIST_ORDER_LEN) {
			ret = -EINVAL;
			goto out;
		}
		strcpy(saved_string, (char *)table->data);
	}
3388
	ret = proc_dostring(table, write, buffer, length, ppos);
3389
	if (ret)
3390
		goto out;
3391 3392
	if (write) {
		int oldval = user_zonelist_order;
3393 3394 3395

		ret = __parse_numa_zonelist_order((char *)table->data);
		if (ret) {
3396 3397 3398
			/*
			 * bogus value.  restore saved string
			 */
3399
			strncpy((char *)table->data, saved_string,
3400 3401
				NUMA_ZONELIST_ORDER_LEN);
			user_zonelist_order = oldval;
3402 3403
		} else if (oldval != user_zonelist_order) {
			mutex_lock(&zonelists_mutex);
3404
			build_all_zonelists(NULL, NULL);
3405 3406
			mutex_unlock(&zonelists_mutex);
		}
3407
	}
3408 3409 3410
out:
	mutex_unlock(&zl_order_mutex);
	return ret;
3411 3412 3413
}


3414
#define MAX_NODE_LOAD (nr_online_nodes)
3415 3416
static int node_load[MAX_NUMNODES];

L
Linus Torvalds 已提交
3417
/**
3418
 * find_next_best_node - find the next node that should appear in a given node's fallback list
L
Linus Torvalds 已提交
3419 3420 3421 3422 3423 3424 3425 3426 3427 3428 3429 3430
 * @node: node whose fallback list we're appending
 * @used_node_mask: nodemask_t of already used nodes
 *
 * We use a number of factors to determine which is the next node that should
 * appear on a given node's fallback list.  The node should not have appeared
 * already in @node's fallback list, and it should be the next closest node
 * according to the distance array (which contains arbitrary distance values
 * from each node to each node in the system), and should also prefer nodes
 * with no CPUs, since presumably they'll have very little allocation pressure
 * on them otherwise.
 * It returns -1 if no node is found.
 */
3431
static int find_next_best_node(int node, nodemask_t *used_node_mask)
L
Linus Torvalds 已提交
3432
{
3433
	int n, val;
L
Linus Torvalds 已提交
3434
	int min_val = INT_MAX;
D
David Rientjes 已提交
3435
	int best_node = NUMA_NO_NODE;
3436
	const struct cpumask *tmp = cpumask_of_node(0);
L
Linus Torvalds 已提交
3437

3438 3439 3440 3441 3442
	/* Use the local node if we haven't already */
	if (!node_isset(node, *used_node_mask)) {
		node_set(node, *used_node_mask);
		return node;
	}
L
Linus Torvalds 已提交
3443

3444
	for_each_node_state(n, N_MEMORY) {
L
Linus Torvalds 已提交
3445 3446 3447 3448 3449 3450 3451 3452

		/* Don't want a node to appear more than once */
		if (node_isset(n, *used_node_mask))
			continue;

		/* Use the distance array to find the distance */
		val = node_distance(node, n);

3453 3454 3455
		/* Penalize nodes under us ("prefer the next node") */
		val += (n < node);

L
Linus Torvalds 已提交
3456
		/* Give preference to headless and unused nodes */
3457 3458
		tmp = cpumask_of_node(n);
		if (!cpumask_empty(tmp))
L
Linus Torvalds 已提交
3459 3460 3461 3462 3463 3464 3465 3466 3467 3468 3469 3470 3471 3472 3473 3474 3475 3476
			val += PENALTY_FOR_NODE_WITH_CPUS;

		/* Slight preference for less loaded node */
		val *= (MAX_NODE_LOAD*MAX_NUMNODES);
		val += node_load[n];

		if (val < min_val) {
			min_val = val;
			best_node = n;
		}
	}

	if (best_node >= 0)
		node_set(best_node, *used_node_mask);

	return best_node;
}

3477 3478 3479 3480 3481 3482 3483

/*
 * Build zonelists ordered by node and zones within node.
 * This results in maximum locality--normal zone overflows into local
 * DMA zone, if any--but risks exhausting DMA zone.
 */
static void build_zonelists_in_node_order(pg_data_t *pgdat, int node)
L
Linus Torvalds 已提交
3484
{
3485
	int j;
L
Linus Torvalds 已提交
3486
	struct zonelist *zonelist;
3487

3488
	zonelist = &pgdat->node_zonelists[0];
3489
	for (j = 0; zonelist->_zonerefs[j].zone != NULL; j++)
3490
		;
3491
	j = build_zonelists_node(NODE_DATA(node), zonelist, j);
3492 3493
	zonelist->_zonerefs[j].zone = NULL;
	zonelist->_zonerefs[j].zone_idx = 0;
3494 3495
}

3496 3497 3498 3499 3500 3501 3502 3503
/*
 * Build gfp_thisnode zonelists
 */
static void build_thisnode_zonelists(pg_data_t *pgdat)
{
	int j;
	struct zonelist *zonelist;

3504
	zonelist = &pgdat->node_zonelists[1];
3505
	j = build_zonelists_node(pgdat, zonelist, 0);
3506 3507
	zonelist->_zonerefs[j].zone = NULL;
	zonelist->_zonerefs[j].zone_idx = 0;
3508 3509
}

3510 3511 3512 3513 3514 3515 3516 3517 3518 3519 3520 3521 3522 3523 3524
/*
 * Build zonelists ordered by zone and nodes within zones.
 * This results in conserving DMA zone[s] until all Normal memory is
 * exhausted, but results in overflowing to remote node while memory
 * may still exist in local DMA zone.
 */
static int node_order[MAX_NUMNODES];

static void build_zonelists_in_zone_order(pg_data_t *pgdat, int nr_nodes)
{
	int pos, j, node;
	int zone_type;		/* needs to be signed */
	struct zone *z;
	struct zonelist *zonelist;

3525 3526 3527 3528 3529 3530 3531
	zonelist = &pgdat->node_zonelists[0];
	pos = 0;
	for (zone_type = MAX_NR_ZONES - 1; zone_type >= 0; zone_type--) {
		for (j = 0; j < nr_nodes; j++) {
			node = node_order[j];
			z = &NODE_DATA(node)->node_zones[zone_type];
			if (populated_zone(z)) {
3532 3533
				zoneref_set_zone(z,
					&zonelist->_zonerefs[pos++]);
3534
				check_highest_zone(zone_type);
3535 3536 3537
			}
		}
	}
3538 3539
	zonelist->_zonerefs[pos].zone = NULL;
	zonelist->_zonerefs[pos].zone_idx = 0;
3540 3541 3542 3543 3544
}

static int default_zonelist_order(void)
{
	int nid, zone_type;
3545
	unsigned long low_kmem_size, total_size;
3546 3547 3548
	struct zone *z;
	int average_size;
	/*
3549
	 * ZONE_DMA and ZONE_DMA32 can be very small area in the system.
3550 3551
	 * If they are really small and used heavily, the system can fall
	 * into OOM very easily.
3552
	 * This function detect ZONE_DMA/DMA32 size and configures zone order.
3553 3554 3555 3556 3557 3558 3559 3560 3561
	 */
	/* Is there ZONE_NORMAL ? (ex. ppc has only DMA zone..) */
	low_kmem_size = 0;
	total_size = 0;
	for_each_online_node(nid) {
		for (zone_type = 0; zone_type < MAX_NR_ZONES; zone_type++) {
			z = &NODE_DATA(nid)->node_zones[zone_type];
			if (populated_zone(z)) {
				if (zone_type < ZONE_NORMAL)
3562 3563
					low_kmem_size += z->managed_pages;
				total_size += z->managed_pages;
3564 3565 3566 3567 3568 3569 3570 3571 3572
			} else if (zone_type == ZONE_NORMAL) {
				/*
				 * If any node has only lowmem, then node order
				 * is preferred to allow kernel allocations
				 * locally; otherwise, they can easily infringe
				 * on other nodes when there is an abundance of
				 * lowmem available to allocate from.
				 */
				return ZONELIST_ORDER_NODE;
3573 3574 3575 3576 3577 3578 3579 3580
			}
		}
	}
	if (!low_kmem_size ||  /* there are no DMA area. */
	    low_kmem_size > total_size/2) /* DMA/DMA32 is big. */
		return ZONELIST_ORDER_NODE;
	/*
	 * look into each node's config.
3581 3582 3583
	 * If there is a node whose DMA/DMA32 memory is very big area on
	 * local memory, NODE_ORDER may be suitable.
	 */
3584
	average_size = total_size /
3585
				(nodes_weight(node_states[N_MEMORY]) + 1);
3586 3587 3588 3589 3590 3591 3592 3593 3594 3595 3596 3597 3598 3599 3600 3601 3602 3603 3604 3605 3606 3607 3608 3609 3610 3611 3612 3613 3614 3615 3616
	for_each_online_node(nid) {
		low_kmem_size = 0;
		total_size = 0;
		for (zone_type = 0; zone_type < MAX_NR_ZONES; zone_type++) {
			z = &NODE_DATA(nid)->node_zones[zone_type];
			if (populated_zone(z)) {
				if (zone_type < ZONE_NORMAL)
					low_kmem_size += z->present_pages;
				total_size += z->present_pages;
			}
		}
		if (low_kmem_size &&
		    total_size > average_size && /* ignore small node */
		    low_kmem_size > total_size * 70/100)
			return ZONELIST_ORDER_NODE;
	}
	return ZONELIST_ORDER_ZONE;
}

static void set_zonelist_order(void)
{
	if (user_zonelist_order == ZONELIST_ORDER_DEFAULT)
		current_zonelist_order = default_zonelist_order();
	else
		current_zonelist_order = user_zonelist_order;
}

static void build_zonelists(pg_data_t *pgdat)
{
	int j, node, load;
	enum zone_type i;
L
Linus Torvalds 已提交
3617
	nodemask_t used_mask;
3618 3619 3620
	int local_node, prev_node;
	struct zonelist *zonelist;
	int order = current_zonelist_order;
L
Linus Torvalds 已提交
3621 3622

	/* initialize zonelists */
3623
	for (i = 0; i < MAX_ZONELISTS; i++) {
L
Linus Torvalds 已提交
3624
		zonelist = pgdat->node_zonelists + i;
3625 3626
		zonelist->_zonerefs[0].zone = NULL;
		zonelist->_zonerefs[0].zone_idx = 0;
L
Linus Torvalds 已提交
3627 3628 3629 3630
	}

	/* NUMA-aware ordering of nodes */
	local_node = pgdat->node_id;
3631
	load = nr_online_nodes;
L
Linus Torvalds 已提交
3632 3633
	prev_node = local_node;
	nodes_clear(used_mask);
3634 3635 3636 3637

	memset(node_order, 0, sizeof(node_order));
	j = 0;

L
Linus Torvalds 已提交
3638 3639 3640 3641 3642 3643
	while ((node = find_next_best_node(local_node, &used_mask)) >= 0) {
		/*
		 * We don't want to pressure a particular node.
		 * So adding penalty to the first node in same
		 * distance group to make it round-robin.
		 */
3644 3645
		if (node_distance(local_node, node) !=
		    node_distance(local_node, prev_node))
3646 3647
			node_load[node] = load;

L
Linus Torvalds 已提交
3648 3649
		prev_node = node;
		load--;
3650 3651 3652 3653 3654
		if (order == ZONELIST_ORDER_NODE)
			build_zonelists_in_node_order(pgdat, node);
		else
			node_order[j++] = node;	/* remember order */
	}
L
Linus Torvalds 已提交
3655

3656 3657 3658
	if (order == ZONELIST_ORDER_ZONE) {
		/* calculate node order -- i.e., DMA last! */
		build_zonelists_in_zone_order(pgdat, j);
L
Linus Torvalds 已提交
3659
	}
3660 3661

	build_thisnode_zonelists(pgdat);
L
Linus Torvalds 已提交
3662 3663
}

3664
/* Construct the zonelist performance cache - see further mmzone.h */
3665
static void build_zonelist_cache(pg_data_t *pgdat)
3666
{
3667 3668
	struct zonelist *zonelist;
	struct zonelist_cache *zlc;
3669
	struct zoneref *z;
3670

3671 3672 3673
	zonelist = &pgdat->node_zonelists[0];
	zonelist->zlcache_ptr = zlc = &zonelist->zlcache;
	bitmap_zero(zlc->fullzones, MAX_ZONES_PER_ZONELIST);
3674 3675
	for (z = zonelist->_zonerefs; z->zone; z++)
		zlc->z_to_n[z - zonelist->_zonerefs] = zonelist_node_idx(z);
3676 3677
}

3678 3679 3680 3681 3682 3683 3684 3685 3686 3687 3688 3689 3690 3691 3692 3693 3694 3695
#ifdef CONFIG_HAVE_MEMORYLESS_NODES
/*
 * Return node id of node used for "local" allocations.
 * I.e., first node id of first zone in arg node's generic zonelist.
 * Used for initializing percpu 'numa_mem', which is used primarily
 * for kernel allocations, so use GFP_KERNEL flags to locate zonelist.
 */
int local_memory_node(int node)
{
	struct zone *zone;

	(void)first_zones_zonelist(node_zonelist(node, GFP_KERNEL),
				   gfp_zone(GFP_KERNEL),
				   NULL,
				   &zone);
	return zone->node;
}
#endif
3696

L
Linus Torvalds 已提交
3697 3698
#else	/* CONFIG_NUMA */

3699 3700 3701 3702 3703 3704
static void set_zonelist_order(void)
{
	current_zonelist_order = ZONELIST_ORDER_ZONE;
}

static void build_zonelists(pg_data_t *pgdat)
L
Linus Torvalds 已提交
3705
{
3706
	int node, local_node;
3707 3708
	enum zone_type j;
	struct zonelist *zonelist;
L
Linus Torvalds 已提交
3709 3710 3711

	local_node = pgdat->node_id;

3712
	zonelist = &pgdat->node_zonelists[0];
3713
	j = build_zonelists_node(pgdat, zonelist, 0);
L
Linus Torvalds 已提交
3714

3715 3716 3717 3718 3719 3720 3721 3722 3723 3724 3725
	/*
	 * Now we build the zonelist so that it contains the zones
	 * of all the other nodes.
	 * We don't want to pressure a particular node, so when
	 * building the zones for node N, we make sure that the
	 * zones coming right after the local ones are those from
	 * node N+1 (modulo N)
	 */
	for (node = local_node + 1; node < MAX_NUMNODES; node++) {
		if (!node_online(node))
			continue;
3726
		j = build_zonelists_node(NODE_DATA(node), zonelist, j);
L
Linus Torvalds 已提交
3727
	}
3728 3729 3730
	for (node = 0; node < local_node; node++) {
		if (!node_online(node))
			continue;
3731
		j = build_zonelists_node(NODE_DATA(node), zonelist, j);
3732 3733
	}

3734 3735
	zonelist->_zonerefs[j].zone = NULL;
	zonelist->_zonerefs[j].zone_idx = 0;
L
Linus Torvalds 已提交
3736 3737
}

3738
/* non-NUMA variant of zonelist performance cache - just NULL zlcache_ptr */
3739
static void build_zonelist_cache(pg_data_t *pgdat)
3740
{
3741
	pgdat->node_zonelists[0].zlcache_ptr = NULL;
3742 3743
}

L
Linus Torvalds 已提交
3744 3745
#endif	/* CONFIG_NUMA */

3746 3747 3748 3749 3750 3751 3752 3753 3754 3755 3756 3757 3758 3759 3760 3761 3762
/*
 * Boot pageset table. One per cpu which is going to be used for all
 * zones and all nodes. The parameters will be set in such a way
 * that an item put on a list will immediately be handed over to
 * the buddy list. This is safe since pageset manipulation is done
 * with interrupts disabled.
 *
 * The boot_pagesets must be kept even after bootup is complete for
 * unused processors and/or zones. They do play a role for bootstrapping
 * hotplugged processors.
 *
 * zoneinfo_show() and maybe other functions do
 * not check if the processor is online before following the pageset pointer.
 * Other parts of the kernel may not check if the zone is available.
 */
static void setup_pageset(struct per_cpu_pageset *p, unsigned long batch);
static DEFINE_PER_CPU(struct per_cpu_pageset, boot_pageset);
3763
static void setup_zone_pageset(struct zone *zone);
3764

3765 3766 3767 3768 3769 3770
/*
 * Global mutex to protect against size modification of zonelists
 * as well as to serialize pageset setup for the new populated zone.
 */
DEFINE_MUTEX(zonelists_mutex);

3771
/* return values int ....just for stop_machine() */
3772
static int __build_all_zonelists(void *data)
L
Linus Torvalds 已提交
3773
{
3774
	int nid;
3775
	int cpu;
3776
	pg_data_t *self = data;
3777

3778 3779 3780
#ifdef CONFIG_NUMA
	memset(node_load, 0, sizeof(node_load));
#endif
3781 3782 3783 3784 3785 3786

	if (self && !node_online(self->node_id)) {
		build_zonelists(self);
		build_zonelist_cache(self);
	}

3787
	for_each_online_node(nid) {
3788 3789 3790 3791
		pg_data_t *pgdat = NODE_DATA(nid);

		build_zonelists(pgdat);
		build_zonelist_cache(pgdat);
3792
	}
3793 3794 3795 3796 3797 3798 3799 3800 3801 3802 3803 3804 3805 3806

	/*
	 * Initialize the boot_pagesets that are going to be used
	 * for bootstrapping processors. The real pagesets for
	 * each zone will be allocated later when the per cpu
	 * allocator is available.
	 *
	 * boot_pagesets are used also for bootstrapping offline
	 * cpus if the system is already booted because the pagesets
	 * are needed to initialize allocators on a specific cpu too.
	 * F.e. the percpu allocator needs the page allocator which
	 * needs the percpu allocator in order to allocate its pagesets
	 * (a chicken-egg dilemma).
	 */
3807
	for_each_possible_cpu(cpu) {
3808 3809
		setup_pageset(&per_cpu(boot_pageset, cpu), 0);

3810 3811 3812 3813 3814 3815 3816 3817 3818 3819 3820 3821 3822 3823
#ifdef CONFIG_HAVE_MEMORYLESS_NODES
		/*
		 * We now know the "local memory node" for each node--
		 * i.e., the node of the first zone in the generic zonelist.
		 * Set up numa_mem percpu variable for on-line cpus.  During
		 * boot, only the boot cpu should be on-line;  we'll init the
		 * secondary cpus' numa_mem as they come on-line.  During
		 * node/memory hotplug, we'll fixup all on-line cpus.
		 */
		if (cpu_online(cpu))
			set_cpu_numa_mem(cpu, local_memory_node(cpu_to_node(cpu)));
#endif
	}

3824 3825 3826
	return 0;
}

3827 3828 3829 3830
/*
 * Called with zonelists_mutex held always
 * unless system_state == SYSTEM_BOOTING.
 */
3831
void __ref build_all_zonelists(pg_data_t *pgdat, struct zone *zone)
3832
{
3833 3834
	set_zonelist_order();

3835
	if (system_state == SYSTEM_BOOTING) {
3836
		__build_all_zonelists(NULL);
3837
		mminit_verify_zonelist();
3838 3839
		cpuset_init_current_mems_allowed();
	} else {
3840
#ifdef CONFIG_MEMORY_HOTPLUG
3841 3842
		if (zone)
			setup_zone_pageset(zone);
3843
#endif
3844 3845
		/* we have to stop all cpus to guarantee there is no user
		   of zonelist */
3846
		stop_machine(__build_all_zonelists, pgdat, NULL);
3847 3848
		/* cpuset refresh routine should be here */
	}
3849
	vm_total_pages = nr_free_pagecache_pages();
3850 3851 3852 3853 3854 3855 3856
	/*
	 * Disable grouping by mobility if the number of pages in the
	 * system is too low to allow the mechanism to work. It would be
	 * more accurate, but expensive to check per-zone. This check is
	 * made on memory-hotadd so a system can start with mobility
	 * disabled and enable it later
	 */
3857
	if (vm_total_pages < (pageblock_nr_pages * MIGRATE_TYPES))
3858 3859 3860 3861 3862 3863
		page_group_by_mobility_disabled = 1;
	else
		page_group_by_mobility_disabled = 0;

	printk("Built %i zonelists in %s order, mobility grouping %s.  "
		"Total pages: %ld\n",
3864
			nr_online_nodes,
3865
			zonelist_order_name[current_zonelist_order],
3866
			page_group_by_mobility_disabled ? "off" : "on",
3867 3868 3869 3870
			vm_total_pages);
#ifdef CONFIG_NUMA
	printk("Policy zone: %s\n", zone_names[policy_zone]);
#endif
L
Linus Torvalds 已提交
3871 3872 3873 3874 3875 3876 3877 3878 3879 3880 3881 3882 3883 3884 3885
}

/*
 * Helper functions to size the waitqueue hash table.
 * Essentially these want to choose hash table sizes sufficiently
 * large so that collisions trying to wait on pages are rare.
 * But in fact, the number of active page waitqueues on typical
 * systems is ridiculously low, less than 200. So this is even
 * conservative, even though it seems large.
 *
 * The constant PAGES_PER_WAITQUEUE specifies the ratio of pages to
 * waitqueues, i.e. the size of the waitq table given the number of pages.
 */
#define PAGES_PER_WAITQUEUE	256

3886
#ifndef CONFIG_MEMORY_HOTPLUG
3887
static inline unsigned long wait_table_hash_nr_entries(unsigned long pages)
L
Linus Torvalds 已提交
3888 3889 3890 3891 3892 3893 3894 3895 3896 3897 3898 3899 3900 3901 3902 3903 3904
{
	unsigned long size = 1;

	pages /= PAGES_PER_WAITQUEUE;

	while (size < pages)
		size <<= 1;

	/*
	 * Once we have dozens or even hundreds of threads sleeping
	 * on IO we've got bigger problems than wait queue collision.
	 * Limit the size of the wait table to a reasonable size.
	 */
	size = min(size, 4096UL);

	return max(size, 4UL);
}
3905 3906 3907 3908 3909 3910 3911 3912 3913 3914 3915 3916 3917 3918 3919 3920 3921 3922 3923 3924 3925 3926 3927
#else
/*
 * A zone's size might be changed by hot-add, so it is not possible to determine
 * a suitable size for its wait_table.  So we use the maximum size now.
 *
 * The max wait table size = 4096 x sizeof(wait_queue_head_t).   ie:
 *
 *    i386 (preemption config)    : 4096 x 16 = 64Kbyte.
 *    ia64, x86-64 (no preemption): 4096 x 20 = 80Kbyte.
 *    ia64, x86-64 (preemption)   : 4096 x 24 = 96Kbyte.
 *
 * The maximum entries are prepared when a zone's memory is (512K + 256) pages
 * or more by the traditional way. (See above).  It equals:
 *
 *    i386, x86-64, powerpc(4K page size) : =  ( 2G + 1M)byte.
 *    ia64(16K page size)                 : =  ( 8G + 4M)byte.
 *    powerpc (64K page size)             : =  (32G +16M)byte.
 */
static inline unsigned long wait_table_hash_nr_entries(unsigned long pages)
{
	return 4096UL;
}
#endif
L
Linus Torvalds 已提交
3928 3929 3930 3931 3932 3933 3934 3935 3936 3937 3938

/*
 * This is an integer logarithm so that shifts can be used later
 * to extract the more random high bits from the multiplicative
 * hash function before the remainder is taken.
 */
static inline unsigned long wait_table_bits(unsigned long size)
{
	return ffz(~size);
}

3939 3940 3941 3942 3943 3944 3945 3946 3947 3948 3949 3950 3951 3952
/*
 * Check if a pageblock contains reserved pages
 */
static int pageblock_is_reserved(unsigned long start_pfn, unsigned long end_pfn)
{
	unsigned long pfn;

	for (pfn = start_pfn; pfn < end_pfn; pfn++) {
		if (!pfn_valid_within(pfn) || PageReserved(pfn_to_page(pfn)))
			return 1;
	}
	return 0;
}

3953
/*
3954
 * Mark a number of pageblocks as MIGRATE_RESERVE. The number
3955 3956
 * of blocks reserved is based on min_wmark_pages(zone). The memory within
 * the reserve will tend to store contiguous free pages. Setting min_free_kbytes
3957 3958 3959 3960 3961
 * higher will lead to a bigger reserve which will get freed as contiguous
 * blocks as reclaim kicks in
 */
static void setup_zone_migrate_reserve(struct zone *zone)
{
3962
	unsigned long start_pfn, pfn, end_pfn, block_end_pfn;
3963
	struct page *page;
3964 3965
	unsigned long block_migratetype;
	int reserve;
3966
	int old_reserve;
3967

3968 3969 3970 3971 3972 3973
	/*
	 * Get the start pfn, end pfn and the number of blocks to reserve
	 * We have to be careful to be aligned to pageblock_nr_pages to
	 * make sure that we always check pfn_valid for the first page in
	 * the block.
	 */
3974
	start_pfn = zone->zone_start_pfn;
3975
	end_pfn = zone_end_pfn(zone);
3976
	start_pfn = roundup(start_pfn, pageblock_nr_pages);
3977
	reserve = roundup(min_wmark_pages(zone), pageblock_nr_pages) >>
3978
							pageblock_order;
3979

3980 3981 3982 3983 3984 3985 3986 3987
	/*
	 * Reserve blocks are generally in place to help high-order atomic
	 * allocations that are short-lived. A min_free_kbytes value that
	 * would result in more than 2 reserve blocks for atomic allocations
	 * is assumed to be in place to help anti-fragmentation for the
	 * future allocation of hugepages at runtime.
	 */
	reserve = min(2, reserve);
3988 3989 3990 3991 3992 3993
	old_reserve = zone->nr_migrate_reserve_block;

	/* When memory hot-add, we almost always need to do nothing */
	if (reserve == old_reserve)
		return;
	zone->nr_migrate_reserve_block = reserve;
3994

3995
	for (pfn = start_pfn; pfn < end_pfn; pfn += pageblock_nr_pages) {
3996 3997 3998 3999
		if (!pfn_valid(pfn))
			continue;
		page = pfn_to_page(pfn);

4000 4001 4002 4003
		/* Watch out for overlapping nodes */
		if (page_to_nid(page) != zone_to_nid(zone))
			continue;

4004 4005
		block_migratetype = get_pageblock_migratetype(page);

4006 4007 4008 4009 4010 4011 4012 4013 4014
		/* Only test what is necessary when the reserves are not met */
		if (reserve > 0) {
			/*
			 * Blocks with reserved pages will never free, skip
			 * them.
			 */
			block_end_pfn = min(pfn + pageblock_nr_pages, end_pfn);
			if (pageblock_is_reserved(pfn, block_end_pfn))
				continue;
4015

4016 4017 4018 4019 4020 4021 4022 4023 4024 4025 4026 4027 4028 4029 4030
			/* If this block is reserved, account for it */
			if (block_migratetype == MIGRATE_RESERVE) {
				reserve--;
				continue;
			}

			/* Suitable for reserving if this block is movable */
			if (block_migratetype == MIGRATE_MOVABLE) {
				set_pageblock_migratetype(page,
							MIGRATE_RESERVE);
				move_freepages_block(zone, page,
							MIGRATE_RESERVE);
				reserve--;
				continue;
			}
4031 4032 4033 4034 4035 4036
		} else if (!old_reserve) {
			/*
			 * At boot time we don't need to scan the whole zone
			 * for turning off MIGRATE_RESERVE.
			 */
			break;
4037 4038 4039 4040 4041 4042 4043 4044 4045 4046 4047 4048
		}

		/*
		 * If the reserve is met and this is a previous reserved block,
		 * take it back
		 */
		if (block_migratetype == MIGRATE_RESERVE) {
			set_pageblock_migratetype(page, MIGRATE_MOVABLE);
			move_freepages_block(zone, page, MIGRATE_MOVABLE);
		}
	}
}
M
Mel Gorman 已提交
4049

L
Linus Torvalds 已提交
4050 4051 4052 4053 4054
/*
 * Initially all pages are reserved - free ones are freed
 * up by free_all_bootmem() once the early boot process is
 * done. Non-atomic initialization, single-pass.
 */
4055
void __meminit memmap_init_zone(unsigned long size, int nid, unsigned long zone,
D
Dave Hansen 已提交
4056
		unsigned long start_pfn, enum memmap_context context)
L
Linus Torvalds 已提交
4057 4058
{
	struct page *page;
A
Andy Whitcroft 已提交
4059 4060
	unsigned long end_pfn = start_pfn + size;
	unsigned long pfn;
4061
	struct zone *z;
L
Linus Torvalds 已提交
4062

4063 4064 4065
	if (highest_memmap_pfn < end_pfn - 1)
		highest_memmap_pfn = end_pfn - 1;

4066
	z = &NODE_DATA(nid)->node_zones[zone];
4067
	for (pfn = start_pfn; pfn < end_pfn; pfn++) {
D
Dave Hansen 已提交
4068 4069 4070 4071 4072 4073 4074 4075 4076 4077 4078
		/*
		 * There can be holes in boot-time mem_map[]s
		 * handed to this function.  They do not
		 * exist on hotplugged memory.
		 */
		if (context == MEMMAP_EARLY) {
			if (!early_pfn_valid(pfn))
				continue;
			if (!early_pfn_in_nid(pfn, nid))
				continue;
		}
A
Andy Whitcroft 已提交
4079 4080
		page = pfn_to_page(pfn);
		set_page_links(page, zone, nid, pfn);
4081
		mminit_verify_page_links(page, zone, nid, pfn);
4082
		init_page_count(page);
4083
		page_mapcount_reset(page);
4084
		page_cpupid_reset_last(page);
L
Linus Torvalds 已提交
4085
		SetPageReserved(page);
4086 4087 4088 4089 4090
		/*
		 * Mark the block movable so that blocks are reserved for
		 * movable at startup. This will force kernel allocations
		 * to reserve their blocks rather than leaking throughout
		 * the address space during boot when many long-lived
4091 4092 4093
		 * kernel allocations are made. Later some blocks near
		 * the start are marked MIGRATE_RESERVE by
		 * setup_zone_migrate_reserve()
4094 4095 4096 4097 4098
		 *
		 * bitmap is created for zone's valid pfn range. but memmap
		 * can be created for invalid pages (for alignment)
		 * check here not to call set_pageblock_migratetype() against
		 * pfn out of zone.
4099
		 */
4100
		if ((z->zone_start_pfn <= pfn)
4101
		    && (pfn < zone_end_pfn(z))
4102
		    && !(pfn & (pageblock_nr_pages - 1)))
4103
			set_pageblock_migratetype(page, MIGRATE_MOVABLE);
4104

L
Linus Torvalds 已提交
4105 4106 4107 4108
		INIT_LIST_HEAD(&page->lru);
#ifdef WANT_PAGE_VIRTUAL
		/* The shift won't overflow because ZONE_NORMAL is below 4G. */
		if (!is_highmem_idx(zone))
4109
			set_page_address(page, __va(pfn << PAGE_SHIFT));
L
Linus Torvalds 已提交
4110 4111 4112 4113
#endif
	}
}

4114
static void __meminit zone_init_free_lists(struct zone *zone)
L
Linus Torvalds 已提交
4115
{
4116 4117 4118
	int order, t;
	for_each_migratetype_order(order, t) {
		INIT_LIST_HEAD(&zone->free_area[order].free_list[t]);
L
Linus Torvalds 已提交
4119 4120 4121 4122 4123 4124
		zone->free_area[order].nr_free = 0;
	}
}

#ifndef __HAVE_ARCH_MEMMAP_INIT
#define memmap_init(size, nid, zone, start_pfn) \
D
Dave Hansen 已提交
4125
	memmap_init_zone((size), (nid), (zone), (start_pfn), MEMMAP_EARLY)
L
Linus Torvalds 已提交
4126 4127
#endif

4128
static int __meminit zone_batchsize(struct zone *zone)
4129
{
4130
#ifdef CONFIG_MMU
4131 4132 4133 4134
	int batch;

	/*
	 * The per-cpu-pages pools are set to around 1000th of the
4135
	 * size of the zone.  But no more than 1/2 of a meg.
4136 4137 4138
	 *
	 * OK, so we don't know how big the cache is.  So guess.
	 */
4139
	batch = zone->managed_pages / 1024;
4140 4141
	if (batch * PAGE_SIZE > 512 * 1024)
		batch = (512 * 1024) / PAGE_SIZE;
4142 4143 4144 4145 4146
	batch /= 4;		/* We effectively *= 4 below */
	if (batch < 1)
		batch = 1;

	/*
4147 4148 4149
	 * Clamp the batch to a 2^n - 1 value. Having a power
	 * of 2 value was found to be more likely to have
	 * suboptimal cache aliasing properties in some cases.
4150
	 *
4151 4152 4153 4154
	 * For example if 2 tasks are alternately allocating
	 * batches of pages, one task can end up with a lot
	 * of pages of one half of the possible page colors
	 * and the other with pages of the other colors.
4155
	 */
4156
	batch = rounddown_pow_of_two(batch + batch/2) - 1;
4157

4158
	return batch;
4159 4160 4161 4162 4163 4164 4165 4166 4167 4168 4169 4170 4171 4172 4173 4174 4175

#else
	/* The deferral and batching of frees should be suppressed under NOMMU
	 * conditions.
	 *
	 * The problem is that NOMMU needs to be able to allocate large chunks
	 * of contiguous memory as there's no hardware page translation to
	 * assemble apparent contiguous memory from discontiguous pages.
	 *
	 * Queueing large contiguous runs of pages for batching, however,
	 * causes the pages to actually be freed in smaller chunks.  As there
	 * can be a significant delay between the individual batches being
	 * recycled, this leads to the once large chunks of space being
	 * fragmented and becoming unavailable for high-order allocations.
	 */
	return 0;
#endif
4176 4177
}

4178 4179 4180 4181 4182 4183 4184 4185 4186 4187 4188 4189 4190 4191 4192 4193 4194 4195 4196 4197 4198 4199 4200 4201 4202 4203 4204
/*
 * pcp->high and pcp->batch values are related and dependent on one another:
 * ->batch must never be higher then ->high.
 * The following function updates them in a safe manner without read side
 * locking.
 *
 * Any new users of pcp->batch and pcp->high should ensure they can cope with
 * those fields changing asynchronously (acording the the above rule).
 *
 * mutex_is_locked(&pcp_batch_high_lock) required when calling this function
 * outside of boot time (or some other assurance that no concurrent updaters
 * exist).
 */
static void pageset_update(struct per_cpu_pages *pcp, unsigned long high,
		unsigned long batch)
{
       /* start with a fail safe value for batch */
	pcp->batch = 1;
	smp_wmb();

       /* Update high, then batch, in order */
	pcp->high = high;
	smp_wmb();

	pcp->batch = batch;
}

4205
/* a companion to pageset_set_high() */
4206 4207
static void pageset_set_batch(struct per_cpu_pageset *p, unsigned long batch)
{
4208
	pageset_update(&p->pcp, 6 * batch, max(1UL, 1 * batch));
4209 4210
}

4211
static void pageset_init(struct per_cpu_pageset *p)
4212 4213
{
	struct per_cpu_pages *pcp;
4214
	int migratetype;
4215

4216 4217
	memset(p, 0, sizeof(*p));

4218
	pcp = &p->pcp;
4219
	pcp->count = 0;
4220 4221
	for (migratetype = 0; migratetype < MIGRATE_PCPTYPES; migratetype++)
		INIT_LIST_HEAD(&pcp->lists[migratetype]);
4222 4223
}

4224 4225 4226 4227 4228 4229
static void setup_pageset(struct per_cpu_pageset *p, unsigned long batch)
{
	pageset_init(p);
	pageset_set_batch(p, batch);
}

4230
/*
4231
 * pageset_set_high() sets the high water mark for hot per_cpu_pagelist
4232 4233
 * to the value high for the pageset p.
 */
4234
static void pageset_set_high(struct per_cpu_pageset *p,
4235 4236
				unsigned long high)
{
4237 4238 4239
	unsigned long batch = max(1UL, high / 4);
	if ((high / 4) > (PAGE_SHIFT * 8))
		batch = PAGE_SHIFT * 8;
4240

4241
	pageset_update(&p->pcp, high, batch);
4242 4243
}

4244 4245
static void __meminit pageset_set_high_and_batch(struct zone *zone,
		struct per_cpu_pageset *pcp)
4246 4247
{
	if (percpu_pagelist_fraction)
4248
		pageset_set_high(pcp,
4249 4250 4251 4252 4253 4254
			(zone->managed_pages /
				percpu_pagelist_fraction));
	else
		pageset_set_batch(pcp, zone_batchsize(zone));
}

4255 4256 4257 4258 4259 4260 4261 4262
static void __meminit zone_pageset_init(struct zone *zone, int cpu)
{
	struct per_cpu_pageset *pcp = per_cpu_ptr(zone->pageset, cpu);

	pageset_init(pcp);
	pageset_set_high_and_batch(zone, pcp);
}

4263
static void __meminit setup_zone_pageset(struct zone *zone)
4264 4265 4266
{
	int cpu;
	zone->pageset = alloc_percpu(struct per_cpu_pageset);
4267 4268
	for_each_possible_cpu(cpu)
		zone_pageset_init(zone, cpu);
4269 4270
}

4271
/*
4272 4273
 * Allocate per cpu pagesets and initialize them.
 * Before this call only boot pagesets were available.
4274
 */
4275
void __init setup_per_cpu_pageset(void)
4276
{
4277
	struct zone *zone;
4278

4279 4280
	for_each_populated_zone(zone)
		setup_zone_pageset(zone);
4281 4282
}

S
Sam Ravnborg 已提交
4283
static noinline __init_refok
4284
int zone_wait_table_init(struct zone *zone, unsigned long zone_size_pages)
4285 4286
{
	int i;
4287
	size_t alloc_size;
4288 4289 4290 4291 4292

	/*
	 * The per-page waitqueue mechanism uses hashed waitqueues
	 * per zone.
	 */
4293 4294 4295 4296
	zone->wait_table_hash_nr_entries =
		 wait_table_hash_nr_entries(zone_size_pages);
	zone->wait_table_bits =
		wait_table_bits(zone->wait_table_hash_nr_entries);
4297 4298 4299
	alloc_size = zone->wait_table_hash_nr_entries
					* sizeof(wait_queue_head_t);

4300
	if (!slab_is_available()) {
4301
		zone->wait_table = (wait_queue_head_t *)
4302 4303
			memblock_virt_alloc_node_nopanic(
				alloc_size, zone->zone_pgdat->node_id);
4304 4305 4306 4307 4308 4309 4310 4311 4312 4313 4314
	} else {
		/*
		 * This case means that a zone whose size was 0 gets new memory
		 * via memory hot-add.
		 * But it may be the case that a new node was hot-added.  In
		 * this case vmalloc() will not be able to use this new node's
		 * memory - this wait_table must be initialized to use this new
		 * node itself as well.
		 * To use this new node's memory, further consideration will be
		 * necessary.
		 */
4315
		zone->wait_table = vmalloc(alloc_size);
4316 4317 4318
	}
	if (!zone->wait_table)
		return -ENOMEM;
4319

4320
	for (i = 0; i < zone->wait_table_hash_nr_entries; ++i)
4321
		init_waitqueue_head(zone->wait_table + i);
4322 4323

	return 0;
4324 4325
}

4326
static __meminit void zone_pcp_init(struct zone *zone)
4327
{
4328 4329 4330 4331 4332 4333
	/*
	 * per cpu subsystem is not up at this point. The following code
	 * relies on the ability of the linker to provide the
	 * offset of a (static) per cpu variable into the per cpu area.
	 */
	zone->pageset = &boot_pageset;
4334

4335
	if (populated_zone(zone))
4336 4337 4338
		printk(KERN_DEBUG "  %s zone: %lu pages, LIFO batch:%u\n",
			zone->name, zone->present_pages,
					 zone_batchsize(zone));
4339 4340
}

4341
int __meminit init_currently_empty_zone(struct zone *zone,
4342
					unsigned long zone_start_pfn,
D
Dave Hansen 已提交
4343 4344
					unsigned long size,
					enum memmap_context context)
4345 4346
{
	struct pglist_data *pgdat = zone->zone_pgdat;
4347 4348 4349 4350
	int ret;
	ret = zone_wait_table_init(zone, size);
	if (ret)
		return ret;
4351 4352 4353 4354
	pgdat->nr_zones = zone_idx(zone) + 1;

	zone->zone_start_pfn = zone_start_pfn;

4355 4356 4357 4358 4359 4360
	mminit_dprintk(MMINIT_TRACE, "memmap_init",
			"Initialising map node %d zone %lu pfns %lu -> %lu\n",
			pgdat->node_id,
			(unsigned long)zone_idx(zone),
			zone_start_pfn, (zone_start_pfn + size));

4361
	zone_init_free_lists(zone);
4362 4363

	return 0;
4364 4365
}

T
Tejun Heo 已提交
4366
#ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
4367 4368 4369 4370 4371 4372 4373
#ifndef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID
/*
 * Required by SPARSEMEM. Given a PFN, return what node the PFN is on.
 * Architectures may implement their own version but if add_active_range()
 * was used and there are no special requirements, this is a convenient
 * alternative
 */
4374
int __meminit __early_pfn_to_nid(unsigned long pfn)
4375
{
4376
	unsigned long start_pfn, end_pfn;
4377
	int nid;
4378 4379 4380 4381 4382 4383 4384 4385 4386
	/*
	 * NOTE: The following SMP-unsafe globals are only used early in boot
	 * when the kernel is running single-threaded.
	 */
	static unsigned long __meminitdata last_start_pfn, last_end_pfn;
	static int __meminitdata last_nid;

	if (last_start_pfn <= pfn && pfn < last_end_pfn)
		return last_nid;
4387

4388 4389 4390 4391 4392 4393 4394 4395
	nid = memblock_search_pfn_nid(pfn, &start_pfn, &end_pfn);
	if (nid != -1) {
		last_start_pfn = start_pfn;
		last_end_pfn = end_pfn;
		last_nid = nid;
	}

	return nid;
4396 4397 4398
}
#endif /* CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID */

4399 4400
int __meminit early_pfn_to_nid(unsigned long pfn)
{
4401 4402 4403 4404 4405 4406 4407
	int nid;

	nid = __early_pfn_to_nid(pfn);
	if (nid >= 0)
		return nid;
	/* just returns 0 */
	return 0;
4408 4409
}

4410 4411 4412 4413 4414 4415 4416 4417 4418 4419 4420
#ifdef CONFIG_NODES_SPAN_OTHER_NODES
bool __meminit early_pfn_in_nid(unsigned long pfn, int node)
{
	int nid;

	nid = __early_pfn_to_nid(pfn);
	if (nid >= 0 && nid != node)
		return false;
	return true;
}
#endif
4421

4422
/**
4423
 * free_bootmem_with_active_regions - Call memblock_free_early_nid for each active range
4424
 * @nid: The node to free memory on. If MAX_NUMNODES, all nodes are freed.
4425
 * @max_low_pfn: The highest PFN that will be passed to memblock_free_early_nid
4426 4427 4428
 *
 * If an architecture guarantees that all ranges registered with
 * add_active_ranges() contain no holes and may be freed, this
4429 4430
 * this function may be used instead of calling memblock_free_early_nid()
 * manually.
4431
 */
4432
void __init free_bootmem_with_active_regions(int nid, unsigned long max_low_pfn)
4433
{
4434 4435
	unsigned long start_pfn, end_pfn;
	int i, this_nid;
4436

4437 4438 4439
	for_each_mem_pfn_range(i, nid, &start_pfn, &end_pfn, &this_nid) {
		start_pfn = min(start_pfn, max_low_pfn);
		end_pfn = min(end_pfn, max_low_pfn);
4440

4441
		if (start_pfn < end_pfn)
4442 4443 4444
			memblock_free_early_nid(PFN_PHYS(start_pfn),
					(end_pfn - start_pfn) << PAGE_SHIFT,
					this_nid);
4445 4446 4447
	}
}

4448 4449
/**
 * sparse_memory_present_with_active_regions - Call memory_present for each active range
4450
 * @nid: The node to call memory_present for. If MAX_NUMNODES, all nodes will be used.
4451 4452 4453
 *
 * If an architecture guarantees that all ranges registered with
 * add_active_ranges() contain no holes and may be freed, this
4454
 * function may be used instead of calling memory_present() manually.
4455 4456 4457
 */
void __init sparse_memory_present_with_active_regions(int nid)
{
4458 4459
	unsigned long start_pfn, end_pfn;
	int i, this_nid;
4460

4461 4462
	for_each_mem_pfn_range(i, nid, &start_pfn, &end_pfn, &this_nid)
		memory_present(this_nid, start_pfn, end_pfn);
4463 4464 4465 4466
}

/**
 * get_pfn_range_for_nid - Return the start and end page frames for a node
4467 4468 4469
 * @nid: The nid to return the range for. If MAX_NUMNODES, the min and max PFN are returned.
 * @start_pfn: Passed by reference. On return, it will have the node start_pfn.
 * @end_pfn: Passed by reference. On return, it will have the node end_pfn.
4470 4471 4472 4473
 *
 * It returns the start and end page frame of a node based on information
 * provided by an arch calling add_active_range(). If called for a node
 * with no available memory, a warning is printed and the start and end
4474
 * PFNs will be 0.
4475
 */
4476
void __meminit get_pfn_range_for_nid(unsigned int nid,
4477 4478
			unsigned long *start_pfn, unsigned long *end_pfn)
{
4479
	unsigned long this_start_pfn, this_end_pfn;
4480
	int i;
4481

4482 4483 4484
	*start_pfn = -1UL;
	*end_pfn = 0;

4485 4486 4487
	for_each_mem_pfn_range(i, nid, &this_start_pfn, &this_end_pfn, NULL) {
		*start_pfn = min(*start_pfn, this_start_pfn);
		*end_pfn = max(*end_pfn, this_end_pfn);
4488 4489
	}

4490
	if (*start_pfn == -1UL)
4491 4492 4493
		*start_pfn = 0;
}

M
Mel Gorman 已提交
4494 4495 4496 4497 4498
/*
 * This finds a zone that can be used for ZONE_MOVABLE pages. The
 * assumption is made that zones within a node are ordered in monotonic
 * increasing memory addresses so that the "highest" populated zone is used
 */
A
Adrian Bunk 已提交
4499
static void __init find_usable_zone_for_movable(void)
M
Mel Gorman 已提交
4500 4501 4502 4503 4504 4505 4506 4507 4508 4509 4510 4511 4512 4513 4514 4515 4516
{
	int zone_index;
	for (zone_index = MAX_NR_ZONES - 1; zone_index >= 0; zone_index--) {
		if (zone_index == ZONE_MOVABLE)
			continue;

		if (arch_zone_highest_possible_pfn[zone_index] >
				arch_zone_lowest_possible_pfn[zone_index])
			break;
	}

	VM_BUG_ON(zone_index == -1);
	movable_zone = zone_index;
}

/*
 * The zone ranges provided by the architecture do not include ZONE_MOVABLE
L
Lucas De Marchi 已提交
4517
 * because it is sized independent of architecture. Unlike the other zones,
M
Mel Gorman 已提交
4518 4519 4520 4521 4522 4523 4524
 * the starting point for ZONE_MOVABLE is not fixed. It may be different
 * in each node depending on the size of each node and how evenly kernelcore
 * is distributed. This helper function adjusts the zone ranges
 * provided by the architecture for a given node by using the end of the
 * highest usable zone for ZONE_MOVABLE. This preserves the assumption that
 * zones within a node are in order of monotonic increases memory addresses
 */
A
Adrian Bunk 已提交
4525
static void __meminit adjust_zone_range_for_zone_movable(int nid,
M
Mel Gorman 已提交
4526 4527 4528 4529 4530 4531 4532 4533 4534 4535 4536 4537 4538 4539 4540 4541 4542 4543 4544 4545 4546 4547 4548 4549 4550
					unsigned long zone_type,
					unsigned long node_start_pfn,
					unsigned long node_end_pfn,
					unsigned long *zone_start_pfn,
					unsigned long *zone_end_pfn)
{
	/* Only adjust if ZONE_MOVABLE is on this node */
	if (zone_movable_pfn[nid]) {
		/* Size ZONE_MOVABLE */
		if (zone_type == ZONE_MOVABLE) {
			*zone_start_pfn = zone_movable_pfn[nid];
			*zone_end_pfn = min(node_end_pfn,
				arch_zone_highest_possible_pfn[movable_zone]);

		/* Adjust for ZONE_MOVABLE starting within this range */
		} else if (*zone_start_pfn < zone_movable_pfn[nid] &&
				*zone_end_pfn > zone_movable_pfn[nid]) {
			*zone_end_pfn = zone_movable_pfn[nid];

		/* Check if this whole range is within ZONE_MOVABLE */
		} else if (*zone_start_pfn >= zone_movable_pfn[nid])
			*zone_start_pfn = *zone_end_pfn;
	}
}

4551 4552 4553 4554
/*
 * Return the number of pages a zone spans in a node, including holes
 * present_pages = zone_spanned_pages_in_node() - zone_absent_pages_in_node()
 */
P
Paul Mundt 已提交
4555
static unsigned long __meminit zone_spanned_pages_in_node(int nid,
4556
					unsigned long zone_type,
4557 4558
					unsigned long node_start_pfn,
					unsigned long node_end_pfn,
4559 4560 4561 4562
					unsigned long *ignored)
{
	unsigned long zone_start_pfn, zone_end_pfn;

4563
	/* Get the start and end of the zone */
4564 4565
	zone_start_pfn = arch_zone_lowest_possible_pfn[zone_type];
	zone_end_pfn = arch_zone_highest_possible_pfn[zone_type];
M
Mel Gorman 已提交
4566 4567 4568
	adjust_zone_range_for_zone_movable(nid, zone_type,
				node_start_pfn, node_end_pfn,
				&zone_start_pfn, &zone_end_pfn);
4569 4570 4571 4572 4573 4574 4575 4576 4577 4578 4579 4580 4581 4582 4583

	/* Check that this node has pages within the zone's required range */
	if (zone_end_pfn < node_start_pfn || zone_start_pfn > node_end_pfn)
		return 0;

	/* Move the zone boundaries inside the node if necessary */
	zone_end_pfn = min(zone_end_pfn, node_end_pfn);
	zone_start_pfn = max(zone_start_pfn, node_start_pfn);

	/* Return the spanned pages */
	return zone_end_pfn - zone_start_pfn;
}

/*
 * Return the number of holes in a range on a node. If nid is MAX_NUMNODES,
4584
 * then all holes in the requested range will be accounted for.
4585
 */
4586
unsigned long __meminit __absent_pages_in_range(int nid,
4587 4588 4589
				unsigned long range_start_pfn,
				unsigned long range_end_pfn)
{
4590 4591 4592
	unsigned long nr_absent = range_end_pfn - range_start_pfn;
	unsigned long start_pfn, end_pfn;
	int i;
4593

4594 4595 4596 4597
	for_each_mem_pfn_range(i, nid, &start_pfn, &end_pfn, NULL) {
		start_pfn = clamp(start_pfn, range_start_pfn, range_end_pfn);
		end_pfn = clamp(end_pfn, range_start_pfn, range_end_pfn);
		nr_absent -= end_pfn - start_pfn;
4598
	}
4599
	return nr_absent;
4600 4601 4602 4603 4604 4605 4606
}

/**
 * absent_pages_in_range - Return number of page frames in holes within a range
 * @start_pfn: The start PFN to start searching for holes
 * @end_pfn: The end PFN to stop searching for holes
 *
4607
 * It returns the number of pages frames in memory holes within a range.
4608 4609 4610 4611 4612 4613 4614 4615
 */
unsigned long __init absent_pages_in_range(unsigned long start_pfn,
							unsigned long end_pfn)
{
	return __absent_pages_in_range(MAX_NUMNODES, start_pfn, end_pfn);
}

/* Return the number of page frames in holes in a zone on a node */
P
Paul Mundt 已提交
4616
static unsigned long __meminit zone_absent_pages_in_node(int nid,
4617
					unsigned long zone_type,
4618 4619
					unsigned long node_start_pfn,
					unsigned long node_end_pfn,
4620 4621
					unsigned long *ignored)
{
4622 4623
	unsigned long zone_low = arch_zone_lowest_possible_pfn[zone_type];
	unsigned long zone_high = arch_zone_highest_possible_pfn[zone_type];
4624 4625
	unsigned long zone_start_pfn, zone_end_pfn;

4626 4627
	zone_start_pfn = clamp(node_start_pfn, zone_low, zone_high);
	zone_end_pfn = clamp(node_end_pfn, zone_low, zone_high);
4628

M
Mel Gorman 已提交
4629 4630 4631
	adjust_zone_range_for_zone_movable(nid, zone_type,
			node_start_pfn, node_end_pfn,
			&zone_start_pfn, &zone_end_pfn);
4632
	return __absent_pages_in_range(nid, zone_start_pfn, zone_end_pfn);
4633
}
4634

T
Tejun Heo 已提交
4635
#else /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
P
Paul Mundt 已提交
4636
static inline unsigned long __meminit zone_spanned_pages_in_node(int nid,
4637
					unsigned long zone_type,
4638 4639
					unsigned long node_start_pfn,
					unsigned long node_end_pfn,
4640 4641 4642 4643 4644
					unsigned long *zones_size)
{
	return zones_size[zone_type];
}

P
Paul Mundt 已提交
4645
static inline unsigned long __meminit zone_absent_pages_in_node(int nid,
4646
						unsigned long zone_type,
4647 4648
						unsigned long node_start_pfn,
						unsigned long node_end_pfn,
4649 4650 4651 4652 4653 4654 4655
						unsigned long *zholes_size)
{
	if (!zholes_size)
		return 0;

	return zholes_size[zone_type];
}
4656

T
Tejun Heo 已提交
4657
#endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
4658

4659
static void __meminit calculate_node_totalpages(struct pglist_data *pgdat,
4660 4661 4662 4663
						unsigned long node_start_pfn,
						unsigned long node_end_pfn,
						unsigned long *zones_size,
						unsigned long *zholes_size)
4664 4665 4666 4667 4668 4669
{
	unsigned long realtotalpages, totalpages = 0;
	enum zone_type i;

	for (i = 0; i < MAX_NR_ZONES; i++)
		totalpages += zone_spanned_pages_in_node(pgdat->node_id, i,
4670 4671 4672
							 node_start_pfn,
							 node_end_pfn,
							 zones_size);
4673 4674 4675 4676 4677 4678
	pgdat->node_spanned_pages = totalpages;

	realtotalpages = totalpages;
	for (i = 0; i < MAX_NR_ZONES; i++)
		realtotalpages -=
			zone_absent_pages_in_node(pgdat->node_id, i,
4679 4680
						  node_start_pfn, node_end_pfn,
						  zholes_size);
4681 4682 4683 4684 4685
	pgdat->node_present_pages = realtotalpages;
	printk(KERN_DEBUG "On node %d totalpages: %lu\n", pgdat->node_id,
							realtotalpages);
}

4686 4687 4688
#ifndef CONFIG_SPARSEMEM
/*
 * Calculate the size of the zone->blockflags rounded to an unsigned long
4689 4690
 * Start by making sure zonesize is a multiple of pageblock_order by rounding
 * up. Then use 1 NR_PAGEBLOCK_BITS worth of bits per pageblock, finally
4691 4692 4693
 * round what is now in bits to nearest long in bits, then return it in
 * bytes.
 */
4694
static unsigned long __init usemap_size(unsigned long zone_start_pfn, unsigned long zonesize)
4695 4696 4697
{
	unsigned long usemapsize;

4698
	zonesize += zone_start_pfn & (pageblock_nr_pages-1);
4699 4700
	usemapsize = roundup(zonesize, pageblock_nr_pages);
	usemapsize = usemapsize >> pageblock_order;
4701 4702 4703 4704 4705 4706 4707
	usemapsize *= NR_PAGEBLOCK_BITS;
	usemapsize = roundup(usemapsize, 8 * sizeof(unsigned long));

	return usemapsize / 8;
}

static void __init setup_usemap(struct pglist_data *pgdat,
4708 4709 4710
				struct zone *zone,
				unsigned long zone_start_pfn,
				unsigned long zonesize)
4711
{
4712
	unsigned long usemapsize = usemap_size(zone_start_pfn, zonesize);
4713
	zone->pageblock_flags = NULL;
4714
	if (usemapsize)
4715 4716 4717
		zone->pageblock_flags =
			memblock_virt_alloc_node_nopanic(usemapsize,
							 pgdat->node_id);
4718 4719
}
#else
4720 4721
static inline void setup_usemap(struct pglist_data *pgdat, struct zone *zone,
				unsigned long zone_start_pfn, unsigned long zonesize) {}
4722 4723
#endif /* CONFIG_SPARSEMEM */

4724
#ifdef CONFIG_HUGETLB_PAGE_SIZE_VARIABLE
4725

4726
/* Initialise the number of pages represented by NR_PAGEBLOCK_BITS */
4727
void __paginginit set_pageblock_order(void)
4728
{
4729 4730
	unsigned int order;

4731 4732 4733 4734
	/* Check that pageblock_nr_pages has not already been setup */
	if (pageblock_order)
		return;

4735 4736 4737 4738 4739
	if (HPAGE_SHIFT > PAGE_SHIFT)
		order = HUGETLB_PAGE_ORDER;
	else
		order = MAX_ORDER - 1;

4740 4741
	/*
	 * Assume the largest contiguous order of interest is a huge page.
4742 4743
	 * This value may be variable depending on boot parameters on IA64 and
	 * powerpc.
4744 4745 4746 4747 4748
	 */
	pageblock_order = order;
}
#else /* CONFIG_HUGETLB_PAGE_SIZE_VARIABLE */

4749 4750
/*
 * When CONFIG_HUGETLB_PAGE_SIZE_VARIABLE is not set, set_pageblock_order()
4751 4752 4753
 * is unused as pageblock_order is set at compile-time. See
 * include/linux/pageblock-flags.h for the values of pageblock_order based on
 * the kernel config
4754
 */
4755
void __paginginit set_pageblock_order(void)
4756 4757
{
}
4758 4759 4760

#endif /* CONFIG_HUGETLB_PAGE_SIZE_VARIABLE */

4761 4762 4763 4764 4765 4766 4767 4768 4769 4770 4771 4772 4773 4774 4775 4776 4777 4778 4779 4780
static unsigned long __paginginit calc_memmap_size(unsigned long spanned_pages,
						   unsigned long present_pages)
{
	unsigned long pages = spanned_pages;

	/*
	 * Provide a more accurate estimation if there are holes within
	 * the zone and SPARSEMEM is in use. If there are holes within the
	 * zone, each populated memory region may cost us one or two extra
	 * memmap pages due to alignment because memmap pages for each
	 * populated regions may not naturally algined on page boundary.
	 * So the (present_pages >> 4) heuristic is a tradeoff for that.
	 */
	if (spanned_pages > present_pages + (present_pages >> 4) &&
	    IS_ENABLED(CONFIG_SPARSEMEM))
		pages = present_pages;

	return PAGE_ALIGN(pages * sizeof(struct page)) >> PAGE_SHIFT;
}

L
Linus Torvalds 已提交
4781 4782 4783 4784 4785
/*
 * Set up the zone data structures:
 *   - mark all pages reserved
 *   - mark all memory queues empty
 *   - clear the memory bitmaps
4786 4787
 *
 * NOTE: pgdat should get zeroed by caller.
L
Linus Torvalds 已提交
4788
 */
4789
static void __paginginit free_area_init_core(struct pglist_data *pgdat,
4790
		unsigned long node_start_pfn, unsigned long node_end_pfn,
L
Linus Torvalds 已提交
4791 4792
		unsigned long *zones_size, unsigned long *zholes_size)
{
4793
	enum zone_type j;
4794
	int nid = pgdat->node_id;
L
Linus Torvalds 已提交
4795
	unsigned long zone_start_pfn = pgdat->node_start_pfn;
4796
	int ret;
L
Linus Torvalds 已提交
4797

4798
	pgdat_resize_init(pgdat);
4799 4800 4801 4802 4803
#ifdef CONFIG_NUMA_BALANCING
	spin_lock_init(&pgdat->numabalancing_migrate_lock);
	pgdat->numabalancing_migrate_nr_pages = 0;
	pgdat->numabalancing_migrate_next_window = jiffies;
#endif
L
Linus Torvalds 已提交
4804
	init_waitqueue_head(&pgdat->kswapd_wait);
4805
	init_waitqueue_head(&pgdat->pfmemalloc_wait);
4806
	pgdat_page_cgroup_init(pgdat);
4807

L
Linus Torvalds 已提交
4808 4809
	for (j = 0; j < MAX_NR_ZONES; j++) {
		struct zone *zone = pgdat->node_zones + j;
4810
		unsigned long size, realsize, freesize, memmap_pages;
L
Linus Torvalds 已提交
4811

4812 4813
		size = zone_spanned_pages_in_node(nid, j, node_start_pfn,
						  node_end_pfn, zones_size);
4814
		realsize = freesize = size - zone_absent_pages_in_node(nid, j,
4815 4816
								node_start_pfn,
								node_end_pfn,
4817
								zholes_size);
L
Linus Torvalds 已提交
4818

4819
		/*
4820
		 * Adjust freesize so that it accounts for how much memory
4821 4822 4823
		 * is used by this zone for memmap. This affects the watermark
		 * and per-cpu initialisations
		 */
4824
		memmap_pages = calc_memmap_size(size, realsize);
4825 4826
		if (freesize >= memmap_pages) {
			freesize -= memmap_pages;
4827 4828 4829 4830
			if (memmap_pages)
				printk(KERN_DEBUG
				       "  %s zone: %lu pages used for memmap\n",
				       zone_names[j], memmap_pages);
4831 4832
		} else
			printk(KERN_WARNING
4833 4834
				"  %s zone: %lu pages exceeds freesize %lu\n",
				zone_names[j], memmap_pages, freesize);
4835

4836
		/* Account for reserved pages */
4837 4838
		if (j == 0 && freesize > dma_reserve) {
			freesize -= dma_reserve;
Y
Yinghai Lu 已提交
4839
			printk(KERN_DEBUG "  %s zone: %lu pages reserved\n",
4840
					zone_names[0], dma_reserve);
4841 4842
		}

4843
		if (!is_highmem_idx(j))
4844
			nr_kernel_pages += freesize;
4845 4846 4847
		/* Charge for highmem memmap if there are enough kernel pages */
		else if (nr_kernel_pages > memmap_pages * 2)
			nr_kernel_pages -= memmap_pages;
4848
		nr_all_pages += freesize;
L
Linus Torvalds 已提交
4849 4850

		zone->spanned_pages = size;
4851
		zone->present_pages = realsize;
4852 4853 4854 4855 4856 4857
		/*
		 * Set an approximate value for lowmem here, it will be adjusted
		 * when the bootmem allocator frees pages into the buddy system.
		 * And all highmem pages will be managed by the buddy system.
		 */
		zone->managed_pages = is_highmem_idx(j) ? realsize : freesize;
4858
#ifdef CONFIG_NUMA
4859
		zone->node = nid;
4860
		zone->min_unmapped_pages = (freesize*sysctl_min_unmapped_ratio)
4861
						/ 100;
4862
		zone->min_slab_pages = (freesize * sysctl_min_slab_ratio) / 100;
4863
#endif
L
Linus Torvalds 已提交
4864 4865 4866
		zone->name = zone_names[j];
		spin_lock_init(&zone->lock);
		spin_lock_init(&zone->lru_lock);
4867
		zone_seqlock_init(zone);
L
Linus Torvalds 已提交
4868
		zone->zone_pgdat = pgdat;
4869
		zone_pcp_init(zone);
4870 4871 4872 4873

		/* For bootup, initialized properly in watermark setup */
		mod_zone_page_state(zone, NR_ALLOC_BATCH, zone->managed_pages);

4874
		lruvec_init(&zone->lruvec);
L
Linus Torvalds 已提交
4875 4876 4877
		if (!size)
			continue;

4878
		set_pageblock_order();
4879
		setup_usemap(pgdat, zone, zone_start_pfn, size);
D
Dave Hansen 已提交
4880 4881
		ret = init_currently_empty_zone(zone, zone_start_pfn,
						size, MEMMAP_EARLY);
4882
		BUG_ON(ret);
4883
		memmap_init(size, nid, j, zone_start_pfn);
L
Linus Torvalds 已提交
4884 4885 4886 4887
		zone_start_pfn += size;
	}
}

S
Sam Ravnborg 已提交
4888
static void __init_refok alloc_node_mem_map(struct pglist_data *pgdat)
L
Linus Torvalds 已提交
4889 4890 4891 4892 4893
{
	/* Skip empty nodes */
	if (!pgdat->node_spanned_pages)
		return;

A
Andy Whitcroft 已提交
4894
#ifdef CONFIG_FLAT_NODE_MEM_MAP
L
Linus Torvalds 已提交
4895 4896
	/* ia64 gets its own node_mem_map, before this, without bootmem */
	if (!pgdat->node_mem_map) {
4897
		unsigned long size, start, end;
A
Andy Whitcroft 已提交
4898 4899
		struct page *map;

4900 4901 4902 4903 4904 4905
		/*
		 * The zone's endpoints aren't required to be MAX_ORDER
		 * aligned but the node_mem_map endpoints must be in order
		 * for the buddy allocator to function correctly.
		 */
		start = pgdat->node_start_pfn & ~(MAX_ORDER_NR_PAGES - 1);
4906
		end = pgdat_end_pfn(pgdat);
4907 4908
		end = ALIGN(end, MAX_ORDER_NR_PAGES);
		size =  (end - start) * sizeof(struct page);
4909 4910
		map = alloc_remap(pgdat->node_id, size);
		if (!map)
4911 4912
			map = memblock_virt_alloc_node_nopanic(size,
							       pgdat->node_id);
4913
		pgdat->node_mem_map = map + (pgdat->node_start_pfn - start);
L
Linus Torvalds 已提交
4914
	}
4915
#ifndef CONFIG_NEED_MULTIPLE_NODES
L
Linus Torvalds 已提交
4916 4917 4918
	/*
	 * With no DISCONTIG, the global mem_map is just set as node 0's
	 */
4919
	if (pgdat == NODE_DATA(0)) {
L
Linus Torvalds 已提交
4920
		mem_map = NODE_DATA(0)->node_mem_map;
T
Tejun Heo 已提交
4921
#ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
4922
		if (page_to_pfn(mem_map) != pgdat->node_start_pfn)
4923
			mem_map -= (pgdat->node_start_pfn - ARCH_PFN_OFFSET);
T
Tejun Heo 已提交
4924
#endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
4925
	}
L
Linus Torvalds 已提交
4926
#endif
A
Andy Whitcroft 已提交
4927
#endif /* CONFIG_FLAT_NODE_MEM_MAP */
L
Linus Torvalds 已提交
4928 4929
}

4930 4931
void __paginginit free_area_init_node(int nid, unsigned long *zones_size,
		unsigned long node_start_pfn, unsigned long *zholes_size)
L
Linus Torvalds 已提交
4932
{
4933
	pg_data_t *pgdat = NODE_DATA(nid);
4934 4935
	unsigned long start_pfn = 0;
	unsigned long end_pfn = 0;
4936

4937
	/* pg_data_t should be reset to zero when it's allocated */
4938
	WARN_ON(pgdat->nr_zones || pgdat->classzone_idx);
4939

L
Linus Torvalds 已提交
4940 4941
	pgdat->node_id = nid;
	pgdat->node_start_pfn = node_start_pfn;
4942 4943 4944 4945 4946
#ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
	get_pfn_range_for_nid(nid, &start_pfn, &end_pfn);
#endif
	calculate_node_totalpages(pgdat, start_pfn, end_pfn,
				  zones_size, zholes_size);
L
Linus Torvalds 已提交
4947 4948

	alloc_node_mem_map(pgdat);
4949 4950 4951 4952 4953
#ifdef CONFIG_FLAT_NODE_MEM_MAP
	printk(KERN_DEBUG "free_area_init_node: node %d, pgdat %08lx, node_mem_map %08lx\n",
		nid, (unsigned long)pgdat,
		(unsigned long)pgdat->node_mem_map);
#endif
L
Linus Torvalds 已提交
4954

4955 4956
	free_area_init_core(pgdat, start_pfn, end_pfn,
			    zones_size, zholes_size);
L
Linus Torvalds 已提交
4957 4958
}

T
Tejun Heo 已提交
4959
#ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
M
Miklos Szeredi 已提交
4960 4961 4962 4963 4964

#if MAX_NUMNODES > 1
/*
 * Figure out the number of possible node ids.
 */
4965
void __init setup_nr_node_ids(void)
M
Miklos Szeredi 已提交
4966 4967 4968 4969 4970 4971 4972 4973 4974 4975
{
	unsigned int node;
	unsigned int highest = 0;

	for_each_node_mask(node, node_possible_map)
		highest = node;
	nr_node_ids = highest + 1;
}
#endif

4976 4977 4978 4979 4980 4981 4982 4983 4984 4985 4986 4987 4988 4989 4990 4991 4992 4993 4994 4995 4996 4997
/**
 * node_map_pfn_alignment - determine the maximum internode alignment
 *
 * This function should be called after node map is populated and sorted.
 * It calculates the maximum power of two alignment which can distinguish
 * all the nodes.
 *
 * For example, if all nodes are 1GiB and aligned to 1GiB, the return value
 * would indicate 1GiB alignment with (1 << (30 - PAGE_SHIFT)).  If the
 * nodes are shifted by 256MiB, 256MiB.  Note that if only the last node is
 * shifted, 1GiB is enough and this function will indicate so.
 *
 * This is used to test whether pfn -> nid mapping of the chosen memory
 * model has fine enough granularity to avoid incorrect mapping for the
 * populated node map.
 *
 * Returns the determined alignment in pfn's.  0 if there is no alignment
 * requirement (single node).
 */
unsigned long __init node_map_pfn_alignment(void)
{
	unsigned long accl_mask = 0, last_end = 0;
4998
	unsigned long start, end, mask;
4999
	int last_nid = -1;
5000
	int i, nid;
5001

5002
	for_each_mem_pfn_range(i, MAX_NUMNODES, &start, &end, &nid) {
5003 5004 5005 5006 5007 5008 5009 5010 5011 5012 5013 5014 5015 5016 5017 5018 5019 5020 5021 5022 5023 5024 5025
		if (!start || last_nid < 0 || last_nid == nid) {
			last_nid = nid;
			last_end = end;
			continue;
		}

		/*
		 * Start with a mask granular enough to pin-point to the
		 * start pfn and tick off bits one-by-one until it becomes
		 * too coarse to separate the current node from the last.
		 */
		mask = ~((1 << __ffs(start)) - 1);
		while (mask && last_end <= (start & (mask << 1)))
			mask <<= 1;

		/* accumulate all internode masks */
		accl_mask |= mask;
	}

	/* convert mask to number of pages */
	return ~accl_mask + 1;
}

5026
/* Find the lowest pfn for a node */
A
Adrian Bunk 已提交
5027
static unsigned long __init find_min_pfn_for_node(int nid)
5028
{
5029
	unsigned long min_pfn = ULONG_MAX;
5030 5031
	unsigned long start_pfn;
	int i;
5032

5033 5034
	for_each_mem_pfn_range(i, nid, &start_pfn, NULL, NULL)
		min_pfn = min(min_pfn, start_pfn);
5035

5036 5037
	if (min_pfn == ULONG_MAX) {
		printk(KERN_WARNING
5038
			"Could not find start_pfn for node %d\n", nid);
5039 5040 5041 5042
		return 0;
	}

	return min_pfn;
5043 5044 5045 5046 5047 5048
}

/**
 * find_min_pfn_with_active_regions - Find the minimum PFN registered
 *
 * It returns the minimum PFN based on information provided via
5049
 * add_active_range().
5050 5051 5052 5053 5054 5055
 */
unsigned long __init find_min_pfn_with_active_regions(void)
{
	return find_min_pfn_for_node(MAX_NUMNODES);
}

5056 5057 5058
/*
 * early_calculate_totalpages()
 * Sum pages in active regions for movable zone.
5059
 * Populate N_MEMORY for calculating usable_nodes.
5060
 */
A
Adrian Bunk 已提交
5061
static unsigned long __init early_calculate_totalpages(void)
5062 5063
{
	unsigned long totalpages = 0;
5064 5065 5066 5067 5068
	unsigned long start_pfn, end_pfn;
	int i, nid;

	for_each_mem_pfn_range(i, MAX_NUMNODES, &start_pfn, &end_pfn, &nid) {
		unsigned long pages = end_pfn - start_pfn;
5069

5070 5071
		totalpages += pages;
		if (pages)
5072
			node_set_state(nid, N_MEMORY);
5073
	}
5074
	return totalpages;
5075 5076
}

M
Mel Gorman 已提交
5077 5078 5079 5080 5081 5082
/*
 * Find the PFN the Movable zone begins in each node. Kernel memory
 * is spread evenly between nodes as long as the nodes have enough
 * memory. When they don't, some nodes will have more kernelcore than
 * others
 */
5083
static void __init find_zone_movable_pfns_for_nodes(void)
M
Mel Gorman 已提交
5084 5085 5086 5087
{
	int i, nid;
	unsigned long usable_startpfn;
	unsigned long kernelcore_node, kernelcore_remaining;
5088
	/* save the state before borrow the nodemask */
5089
	nodemask_t saved_node_state = node_states[N_MEMORY];
5090
	unsigned long totalpages = early_calculate_totalpages();
5091
	int usable_nodes = nodes_weight(node_states[N_MEMORY]);
E
Emil Medve 已提交
5092
	struct memblock_region *r;
5093 5094 5095 5096 5097 5098 5099 5100 5101

	/* Need to find movable_zone earlier when movable_node is specified. */
	find_usable_zone_for_movable();

	/*
	 * If movable_node is specified, ignore kernelcore and movablecore
	 * options.
	 */
	if (movable_node_is_enabled()) {
E
Emil Medve 已提交
5102 5103
		for_each_memblock(memory, r) {
			if (!memblock_is_hotpluggable(r))
5104 5105
				continue;

E
Emil Medve 已提交
5106
			nid = r->nid;
5107

E
Emil Medve 已提交
5108
			usable_startpfn = PFN_DOWN(r->base);
5109 5110 5111 5112 5113 5114 5115
			zone_movable_pfn[nid] = zone_movable_pfn[nid] ?
				min(usable_startpfn, zone_movable_pfn[nid]) :
				usable_startpfn;
		}

		goto out2;
	}
M
Mel Gorman 已提交
5116

5117
	/*
5118
	 * If movablecore=nn[KMG] was specified, calculate what size of
5119 5120 5121 5122 5123 5124 5125 5126 5127 5128 5129 5130 5131 5132 5133 5134 5135 5136 5137 5138
	 * kernelcore that corresponds so that memory usable for
	 * any allocation type is evenly spread. If both kernelcore
	 * and movablecore are specified, then the value of kernelcore
	 * will be used for required_kernelcore if it's greater than
	 * what movablecore would have allowed.
	 */
	if (required_movablecore) {
		unsigned long corepages;

		/*
		 * Round-up so that ZONE_MOVABLE is at least as large as what
		 * was requested by the user
		 */
		required_movablecore =
			roundup(required_movablecore, MAX_ORDER_NR_PAGES);
		corepages = totalpages - required_movablecore;

		required_kernelcore = max(required_kernelcore, corepages);
	}

5139 5140
	/* If kernelcore was not specified, there is no ZONE_MOVABLE */
	if (!required_kernelcore)
5141
		goto out;
M
Mel Gorman 已提交
5142 5143 5144 5145 5146 5147 5148

	/* usable_startpfn is the lowest possible pfn ZONE_MOVABLE can be at */
	usable_startpfn = arch_zone_lowest_possible_pfn[movable_zone];

restart:
	/* Spread kernelcore memory as evenly as possible throughout nodes */
	kernelcore_node = required_kernelcore / usable_nodes;
5149
	for_each_node_state(nid, N_MEMORY) {
5150 5151
		unsigned long start_pfn, end_pfn;

M
Mel Gorman 已提交
5152 5153 5154 5155 5156 5157 5158 5159 5160 5161 5162 5163 5164 5165 5166 5167
		/*
		 * Recalculate kernelcore_node if the division per node
		 * now exceeds what is necessary to satisfy the requested
		 * amount of memory for the kernel
		 */
		if (required_kernelcore < kernelcore_node)
			kernelcore_node = required_kernelcore / usable_nodes;

		/*
		 * As the map is walked, we track how much memory is usable
		 * by the kernel using kernelcore_remaining. When it is
		 * 0, the rest of the node is usable by ZONE_MOVABLE
		 */
		kernelcore_remaining = kernelcore_node;

		/* Go through each range of PFNs within this node */
5168
		for_each_mem_pfn_range(i, nid, &start_pfn, &end_pfn, NULL) {
M
Mel Gorman 已提交
5169 5170
			unsigned long size_pages;

5171
			start_pfn = max(start_pfn, zone_movable_pfn[nid]);
M
Mel Gorman 已提交
5172 5173 5174 5175 5176 5177 5178 5179 5180 5181 5182 5183 5184 5185 5186 5187 5188 5189 5190 5191 5192 5193 5194 5195 5196 5197 5198 5199 5200 5201 5202 5203 5204 5205 5206 5207 5208 5209 5210 5211 5212 5213
			if (start_pfn >= end_pfn)
				continue;

			/* Account for what is only usable for kernelcore */
			if (start_pfn < usable_startpfn) {
				unsigned long kernel_pages;
				kernel_pages = min(end_pfn, usable_startpfn)
								- start_pfn;

				kernelcore_remaining -= min(kernel_pages,
							kernelcore_remaining);
				required_kernelcore -= min(kernel_pages,
							required_kernelcore);

				/* Continue if range is now fully accounted */
				if (end_pfn <= usable_startpfn) {

					/*
					 * Push zone_movable_pfn to the end so
					 * that if we have to rebalance
					 * kernelcore across nodes, we will
					 * not double account here
					 */
					zone_movable_pfn[nid] = end_pfn;
					continue;
				}
				start_pfn = usable_startpfn;
			}

			/*
			 * The usable PFN range for ZONE_MOVABLE is from
			 * start_pfn->end_pfn. Calculate size_pages as the
			 * number of pages used as kernelcore
			 */
			size_pages = end_pfn - start_pfn;
			if (size_pages > kernelcore_remaining)
				size_pages = kernelcore_remaining;
			zone_movable_pfn[nid] = start_pfn + size_pages;

			/*
			 * Some kernelcore has been met, update counts and
			 * break if the kernelcore for this node has been
5214
			 * satisfied
M
Mel Gorman 已提交
5215 5216 5217 5218 5219 5220 5221 5222 5223 5224 5225 5226 5227
			 */
			required_kernelcore -= min(required_kernelcore,
								size_pages);
			kernelcore_remaining -= size_pages;
			if (!kernelcore_remaining)
				break;
		}
	}

	/*
	 * If there is still required_kernelcore, we do another pass with one
	 * less node in the count. This will push zone_movable_pfn[nid] further
	 * along on the nodes that still have memory until kernelcore is
5228
	 * satisfied
M
Mel Gorman 已提交
5229 5230 5231 5232 5233
	 */
	usable_nodes--;
	if (usable_nodes && required_kernelcore > usable_nodes)
		goto restart;

5234
out2:
M
Mel Gorman 已提交
5235 5236 5237 5238
	/* Align start of ZONE_MOVABLE on all nids to MAX_ORDER_NR_PAGES */
	for (nid = 0; nid < MAX_NUMNODES; nid++)
		zone_movable_pfn[nid] =
			roundup(zone_movable_pfn[nid], MAX_ORDER_NR_PAGES);
5239

5240
out:
5241
	/* restore the node_state */
5242
	node_states[N_MEMORY] = saved_node_state;
M
Mel Gorman 已提交
5243 5244
}

5245 5246
/* Any regular or high memory on that node ? */
static void check_for_memory(pg_data_t *pgdat, int nid)
5247 5248 5249
{
	enum zone_type zone_type;

5250 5251 5252 5253
	if (N_MEMORY == N_NORMAL_MEMORY)
		return;

	for (zone_type = 0; zone_type <= ZONE_MOVABLE - 1; zone_type++) {
5254
		struct zone *zone = &pgdat->node_zones[zone_type];
5255
		if (populated_zone(zone)) {
5256 5257 5258 5259
			node_set_state(nid, N_HIGH_MEMORY);
			if (N_NORMAL_MEMORY != N_HIGH_MEMORY &&
			    zone_type <= ZONE_NORMAL)
				node_set_state(nid, N_NORMAL_MEMORY);
5260 5261
			break;
		}
5262 5263 5264
	}
}

5265 5266
/**
 * free_area_init_nodes - Initialise all pg_data_t and zone data
5267
 * @max_zone_pfn: an array of max PFNs for each zone
5268 5269 5270 5271 5272 5273 5274 5275 5276 5277 5278 5279
 *
 * This will call free_area_init_node() for each active node in the system.
 * Using the page ranges provided by add_active_range(), the size of each
 * zone in each node and their holes is calculated. If the maximum PFN
 * between two adjacent zones match, it is assumed that the zone is empty.
 * For example, if arch_max_dma_pfn == arch_max_dma32_pfn, it is assumed
 * that arch_max_dma32_pfn has no pages. It is also assumed that a zone
 * starts where the previous one ended. For example, ZONE_DMA32 starts
 * at arch_max_dma_pfn.
 */
void __init free_area_init_nodes(unsigned long *max_zone_pfn)
{
5280 5281
	unsigned long start_pfn, end_pfn;
	int i, nid;
5282

5283 5284 5285 5286 5287 5288 5289 5290
	/* Record where the zone boundaries are */
	memset(arch_zone_lowest_possible_pfn, 0,
				sizeof(arch_zone_lowest_possible_pfn));
	memset(arch_zone_highest_possible_pfn, 0,
				sizeof(arch_zone_highest_possible_pfn));
	arch_zone_lowest_possible_pfn[0] = find_min_pfn_with_active_regions();
	arch_zone_highest_possible_pfn[0] = max_zone_pfn[0];
	for (i = 1; i < MAX_NR_ZONES; i++) {
M
Mel Gorman 已提交
5291 5292
		if (i == ZONE_MOVABLE)
			continue;
5293 5294 5295 5296 5297
		arch_zone_lowest_possible_pfn[i] =
			arch_zone_highest_possible_pfn[i-1];
		arch_zone_highest_possible_pfn[i] =
			max(max_zone_pfn[i], arch_zone_lowest_possible_pfn[i]);
	}
M
Mel Gorman 已提交
5298 5299 5300 5301 5302
	arch_zone_lowest_possible_pfn[ZONE_MOVABLE] = 0;
	arch_zone_highest_possible_pfn[ZONE_MOVABLE] = 0;

	/* Find the PFNs that ZONE_MOVABLE begins at in each node */
	memset(zone_movable_pfn, 0, sizeof(zone_movable_pfn));
5303
	find_zone_movable_pfns_for_nodes();
5304 5305

	/* Print out the zone ranges */
5306
	printk("Zone ranges:\n");
M
Mel Gorman 已提交
5307 5308 5309
	for (i = 0; i < MAX_NR_ZONES; i++) {
		if (i == ZONE_MOVABLE)
			continue;
5310
		printk(KERN_CONT "  %-8s ", zone_names[i]);
5311 5312
		if (arch_zone_lowest_possible_pfn[i] ==
				arch_zone_highest_possible_pfn[i])
5313
			printk(KERN_CONT "empty\n");
5314
		else
5315 5316 5317 5318
			printk(KERN_CONT "[mem %0#10lx-%0#10lx]\n",
				arch_zone_lowest_possible_pfn[i] << PAGE_SHIFT,
				(arch_zone_highest_possible_pfn[i]
					<< PAGE_SHIFT) - 1);
M
Mel Gorman 已提交
5319 5320 5321
	}

	/* Print out the PFNs ZONE_MOVABLE begins at in each node */
5322
	printk("Movable zone start for each node\n");
M
Mel Gorman 已提交
5323 5324
	for (i = 0; i < MAX_NUMNODES; i++) {
		if (zone_movable_pfn[i])
5325 5326
			printk("  Node %d: %#010lx\n", i,
			       zone_movable_pfn[i] << PAGE_SHIFT);
M
Mel Gorman 已提交
5327
	}
5328

5329
	/* Print out the early node map */
5330
	printk("Early memory node ranges\n");
5331
	for_each_mem_pfn_range(i, MAX_NUMNODES, &start_pfn, &end_pfn, &nid)
5332 5333
		printk("  node %3d: [mem %#010lx-%#010lx]\n", nid,
		       start_pfn << PAGE_SHIFT, (end_pfn << PAGE_SHIFT) - 1);
5334 5335

	/* Initialise every node */
5336
	mminit_verify_pageflags_layout();
5337
	setup_nr_node_ids();
5338 5339
	for_each_online_node(nid) {
		pg_data_t *pgdat = NODE_DATA(nid);
5340
		free_area_init_node(nid, NULL,
5341
				find_min_pfn_for_node(nid), NULL);
5342 5343 5344

		/* Any memory on that node */
		if (pgdat->node_present_pages)
5345 5346
			node_set_state(nid, N_MEMORY);
		check_for_memory(pgdat, nid);
5347 5348
	}
}
M
Mel Gorman 已提交
5349

5350
static int __init cmdline_parse_core(char *p, unsigned long *core)
M
Mel Gorman 已提交
5351 5352 5353 5354 5355 5356
{
	unsigned long long coremem;
	if (!p)
		return -EINVAL;

	coremem = memparse(p, &p);
5357
	*core = coremem >> PAGE_SHIFT;
M
Mel Gorman 已提交
5358

5359
	/* Paranoid check that UL is enough for the coremem value */
M
Mel Gorman 已提交
5360 5361 5362 5363
	WARN_ON((coremem >> PAGE_SHIFT) > ULONG_MAX);

	return 0;
}
M
Mel Gorman 已提交
5364

5365 5366 5367 5368 5369 5370 5371 5372 5373 5374 5375 5376 5377 5378 5379 5380 5381 5382
/*
 * kernelcore=size sets the amount of memory for use for allocations that
 * cannot be reclaimed or migrated.
 */
static int __init cmdline_parse_kernelcore(char *p)
{
	return cmdline_parse_core(p, &required_kernelcore);
}

/*
 * movablecore=size sets the amount of memory for use for allocations that
 * can be reclaimed or migrated.
 */
static int __init cmdline_parse_movablecore(char *p)
{
	return cmdline_parse_core(p, &required_movablecore);
}

M
Mel Gorman 已提交
5383
early_param("kernelcore", cmdline_parse_kernelcore);
5384
early_param("movablecore", cmdline_parse_movablecore);
M
Mel Gorman 已提交
5385

T
Tejun Heo 已提交
5386
#endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
5387

5388 5389 5390 5391 5392
void adjust_managed_page_count(struct page *page, long count)
{
	spin_lock(&managed_page_count_lock);
	page_zone(page)->managed_pages += count;
	totalram_pages += count;
5393 5394 5395 5396
#ifdef CONFIG_HIGHMEM
	if (PageHighMem(page))
		totalhigh_pages += count;
#endif
5397 5398
	spin_unlock(&managed_page_count_lock);
}
5399
EXPORT_SYMBOL(adjust_managed_page_count);
5400

5401
unsigned long free_reserved_area(void *start, void *end, int poison, char *s)
5402
{
5403 5404
	void *pos;
	unsigned long pages = 0;
5405

5406 5407 5408
	start = (void *)PAGE_ALIGN((unsigned long)start);
	end = (void *)((unsigned long)end & PAGE_MASK);
	for (pos = start; pos < end; pos += PAGE_SIZE, pages++) {
5409
		if ((unsigned int)poison <= 0xFF)
5410 5411
			memset(pos, poison, PAGE_SIZE);
		free_reserved_page(virt_to_page(pos));
5412 5413 5414
	}

	if (pages && s)
5415
		pr_info("Freeing %s memory: %ldK (%p - %p)\n",
5416 5417 5418 5419
			s, pages << (PAGE_SHIFT - 10), start, end);

	return pages;
}
5420
EXPORT_SYMBOL(free_reserved_area);
5421

5422 5423 5424 5425 5426
#ifdef	CONFIG_HIGHMEM
void free_highmem_page(struct page *page)
{
	__free_reserved_page(page);
	totalram_pages++;
5427
	page_zone(page)->managed_pages++;
5428 5429 5430 5431
	totalhigh_pages++;
}
#endif

5432 5433 5434 5435 5436 5437 5438 5439 5440 5441 5442 5443 5444 5445 5446 5447 5448 5449 5450 5451 5452 5453

void __init mem_init_print_info(const char *str)
{
	unsigned long physpages, codesize, datasize, rosize, bss_size;
	unsigned long init_code_size, init_data_size;

	physpages = get_num_physpages();
	codesize = _etext - _stext;
	datasize = _edata - _sdata;
	rosize = __end_rodata - __start_rodata;
	bss_size = __bss_stop - __bss_start;
	init_data_size = __init_end - __init_begin;
	init_code_size = _einittext - _sinittext;

	/*
	 * Detect special cases and adjust section sizes accordingly:
	 * 1) .init.* may be embedded into .data sections
	 * 2) .init.text.* may be out of [__init_begin, __init_end],
	 *    please refer to arch/tile/kernel/vmlinux.lds.S.
	 * 3) .rodata.* may be embedded into .text or .data sections.
	 */
#define adj_init_size(start, end, size, pos, adj) \
5454 5455 5456 5457
	do { \
		if (start <= pos && pos < end && size > adj) \
			size -= adj; \
	} while (0)
5458 5459 5460 5461 5462 5463 5464 5465 5466 5467 5468 5469 5470 5471 5472 5473 5474 5475 5476 5477 5478 5479 5480 5481 5482 5483 5484

	adj_init_size(__init_begin, __init_end, init_data_size,
		     _sinittext, init_code_size);
	adj_init_size(_stext, _etext, codesize, _sinittext, init_code_size);
	adj_init_size(_sdata, _edata, datasize, __init_begin, init_data_size);
	adj_init_size(_stext, _etext, codesize, __start_rodata, rosize);
	adj_init_size(_sdata, _edata, datasize, __start_rodata, rosize);

#undef	adj_init_size

	printk("Memory: %luK/%luK available "
	       "(%luK kernel code, %luK rwdata, %luK rodata, "
	       "%luK init, %luK bss, %luK reserved"
#ifdef	CONFIG_HIGHMEM
	       ", %luK highmem"
#endif
	       "%s%s)\n",
	       nr_free_pages() << (PAGE_SHIFT-10), physpages << (PAGE_SHIFT-10),
	       codesize >> 10, datasize >> 10, rosize >> 10,
	       (init_data_size + init_code_size) >> 10, bss_size >> 10,
	       (physpages - totalram_pages) << (PAGE_SHIFT-10),
#ifdef	CONFIG_HIGHMEM
	       totalhigh_pages << (PAGE_SHIFT-10),
#endif
	       str ? ", " : "", str ? str : "");
}

5485
/**
5486 5487
 * set_dma_reserve - set the specified number of pages reserved in the first zone
 * @new_dma_reserve: The number of pages to mark reserved
5488 5489 5490 5491
 *
 * The per-cpu batchsize and zone watermarks are determined by present_pages.
 * In the DMA zone, a significant percentage may be consumed by kernel image
 * and other unfreeable allocations which can skew the watermarks badly. This
5492 5493 5494
 * function may optionally be used to account for unfreeable pages in the
 * first zone (e.g., ZONE_DMA). The effect will be lower watermarks and
 * smaller per-cpu batchsize.
5495 5496 5497 5498 5499 5500
 */
void __init set_dma_reserve(unsigned long new_dma_reserve)
{
	dma_reserve = new_dma_reserve;
}

L
Linus Torvalds 已提交
5501 5502
void __init free_area_init(unsigned long *zones_size)
{
5503
	free_area_init_node(0, zones_size,
L
Linus Torvalds 已提交
5504 5505 5506 5507 5508 5509 5510 5511
			__pa(PAGE_OFFSET) >> PAGE_SHIFT, NULL);
}

static int page_alloc_cpu_notify(struct notifier_block *self,
				 unsigned long action, void *hcpu)
{
	int cpu = (unsigned long)hcpu;

5512
	if (action == CPU_DEAD || action == CPU_DEAD_FROZEN) {
5513
		lru_add_drain_cpu(cpu);
5514 5515 5516 5517 5518 5519 5520 5521
		drain_pages(cpu);

		/*
		 * Spill the event counters of the dead processor
		 * into the current processors event counters.
		 * This artificially elevates the count of the current
		 * processor.
		 */
5522
		vm_events_fold_cpu(cpu);
5523 5524 5525 5526 5527 5528 5529 5530

		/*
		 * Zero the differential counters of the dead processor
		 * so that the vm statistics are consistent.
		 *
		 * This is only okay since the processor is dead and cannot
		 * race with what we are doing.
		 */
5531
		cpu_vm_stats_fold(cpu);
L
Linus Torvalds 已提交
5532 5533 5534 5535 5536 5537 5538 5539 5540
	}
	return NOTIFY_OK;
}

void __init page_alloc_init(void)
{
	hotcpu_notifier(page_alloc_cpu_notify, 0);
}

5541 5542 5543 5544 5545 5546 5547 5548
/*
 * calculate_totalreserve_pages - called when sysctl_lower_zone_reserve_ratio
 *	or min_free_kbytes changes.
 */
static void calculate_totalreserve_pages(void)
{
	struct pglist_data *pgdat;
	unsigned long reserve_pages = 0;
5549
	enum zone_type i, j;
5550 5551 5552 5553 5554 5555 5556 5557 5558 5559 5560 5561

	for_each_online_pgdat(pgdat) {
		for (i = 0; i < MAX_NR_ZONES; i++) {
			struct zone *zone = pgdat->node_zones + i;
			unsigned long max = 0;

			/* Find valid and maximum lowmem_reserve in the zone */
			for (j = i; j < MAX_NR_ZONES; j++) {
				if (zone->lowmem_reserve[j] > max)
					max = zone->lowmem_reserve[j];
			}

5562 5563
			/* we treat the high watermark as reserved pages. */
			max += high_wmark_pages(zone);
5564

5565 5566
			if (max > zone->managed_pages)
				max = zone->managed_pages;
5567
			reserve_pages += max;
5568 5569 5570 5571 5572 5573 5574 5575 5576 5577
			/*
			 * Lowmem reserves are not available to
			 * GFP_HIGHUSER page cache allocations and
			 * kswapd tries to balance zones to their high
			 * watermark.  As a result, neither should be
			 * regarded as dirtyable memory, to prevent a
			 * situation where reclaim has to clean pages
			 * in order to balance the zones.
			 */
			zone->dirty_balance_reserve = max;
5578 5579
		}
	}
5580
	dirty_balance_reserve = reserve_pages;
5581 5582 5583
	totalreserve_pages = reserve_pages;
}

L
Linus Torvalds 已提交
5584 5585 5586 5587 5588 5589 5590 5591 5592
/*
 * setup_per_zone_lowmem_reserve - called whenever
 *	sysctl_lower_zone_reserve_ratio changes.  Ensures that each zone
 *	has a correct pages reserved value, so an adequate number of
 *	pages are left in the zone after a successful __alloc_pages().
 */
static void setup_per_zone_lowmem_reserve(void)
{
	struct pglist_data *pgdat;
5593
	enum zone_type j, idx;
L
Linus Torvalds 已提交
5594

5595
	for_each_online_pgdat(pgdat) {
L
Linus Torvalds 已提交
5596 5597
		for (j = 0; j < MAX_NR_ZONES; j++) {
			struct zone *zone = pgdat->node_zones + j;
5598
			unsigned long managed_pages = zone->managed_pages;
L
Linus Torvalds 已提交
5599 5600 5601

			zone->lowmem_reserve[j] = 0;

5602 5603
			idx = j;
			while (idx) {
L
Linus Torvalds 已提交
5604 5605
				struct zone *lower_zone;

5606 5607
				idx--;

L
Linus Torvalds 已提交
5608 5609 5610 5611
				if (sysctl_lowmem_reserve_ratio[idx] < 1)
					sysctl_lowmem_reserve_ratio[idx] = 1;

				lower_zone = pgdat->node_zones + idx;
5612
				lower_zone->lowmem_reserve[j] = managed_pages /
L
Linus Torvalds 已提交
5613
					sysctl_lowmem_reserve_ratio[idx];
5614
				managed_pages += lower_zone->managed_pages;
L
Linus Torvalds 已提交
5615 5616 5617
			}
		}
	}
5618 5619 5620

	/* update totalreserve_pages */
	calculate_totalreserve_pages();
L
Linus Torvalds 已提交
5621 5622
}

5623
static void __setup_per_zone_wmarks(void)
L
Linus Torvalds 已提交
5624 5625 5626 5627 5628 5629 5630 5631 5632
{
	unsigned long pages_min = min_free_kbytes >> (PAGE_SHIFT - 10);
	unsigned long lowmem_pages = 0;
	struct zone *zone;
	unsigned long flags;

	/* Calculate total number of !ZONE_HIGHMEM pages */
	for_each_zone(zone) {
		if (!is_highmem(zone))
5633
			lowmem_pages += zone->managed_pages;
L
Linus Torvalds 已提交
5634 5635 5636
	}

	for_each_zone(zone) {
5637 5638
		u64 tmp;

5639
		spin_lock_irqsave(&zone->lock, flags);
5640
		tmp = (u64)pages_min * zone->managed_pages;
5641
		do_div(tmp, lowmem_pages);
L
Linus Torvalds 已提交
5642 5643
		if (is_highmem(zone)) {
			/*
N
Nick Piggin 已提交
5644 5645 5646 5647
			 * __GFP_HIGH and PF_MEMALLOC allocations usually don't
			 * need highmem pages, so cap pages_min to a small
			 * value here.
			 *
5648
			 * The WMARK_HIGH-WMARK_LOW and (WMARK_LOW-WMARK_MIN)
N
Nick Piggin 已提交
5649 5650
			 * deltas controls asynch page reclaim, and so should
			 * not be capped for highmem.
L
Linus Torvalds 已提交
5651
			 */
5652
			unsigned long min_pages;
L
Linus Torvalds 已提交
5653

5654
			min_pages = zone->managed_pages / 1024;
5655
			min_pages = clamp(min_pages, SWAP_CLUSTER_MAX, 128UL);
5656
			zone->watermark[WMARK_MIN] = min_pages;
L
Linus Torvalds 已提交
5657
		} else {
N
Nick Piggin 已提交
5658 5659
			/*
			 * If it's a lowmem zone, reserve a number of pages
L
Linus Torvalds 已提交
5660 5661
			 * proportionate to the zone's size.
			 */
5662
			zone->watermark[WMARK_MIN] = tmp;
L
Linus Torvalds 已提交
5663 5664
		}

5665 5666
		zone->watermark[WMARK_LOW]  = min_wmark_pages(zone) + (tmp >> 2);
		zone->watermark[WMARK_HIGH] = min_wmark_pages(zone) + (tmp >> 1);
5667

5668 5669 5670 5671 5672
		__mod_zone_page_state(zone, NR_ALLOC_BATCH,
				      high_wmark_pages(zone) -
				      low_wmark_pages(zone) -
				      zone_page_state(zone, NR_ALLOC_BATCH));

5673
		setup_zone_migrate_reserve(zone);
5674
		spin_unlock_irqrestore(&zone->lock, flags);
L
Linus Torvalds 已提交
5675
	}
5676 5677 5678

	/* update totalreserve_pages */
	calculate_totalreserve_pages();
L
Linus Torvalds 已提交
5679 5680
}

5681 5682 5683 5684 5685 5686 5687 5688 5689 5690 5691 5692 5693 5694
/**
 * setup_per_zone_wmarks - called when min_free_kbytes changes
 * or when memory is hot-{added|removed}
 *
 * Ensures that the watermark[min,low,high] values for each zone are set
 * correctly with respect to min_free_kbytes.
 */
void setup_per_zone_wmarks(void)
{
	mutex_lock(&zonelists_mutex);
	__setup_per_zone_wmarks();
	mutex_unlock(&zonelists_mutex);
}

5695
/*
5696 5697 5698 5699 5700 5701 5702 5703 5704 5705 5706 5707 5708 5709 5710 5711 5712 5713 5714 5715
 * The inactive anon list should be small enough that the VM never has to
 * do too much work, but large enough that each inactive page has a chance
 * to be referenced again before it is swapped out.
 *
 * The inactive_anon ratio is the target ratio of ACTIVE_ANON to
 * INACTIVE_ANON pages on this zone's LRU, maintained by the
 * pageout code. A zone->inactive_ratio of 3 means 3:1 or 25% of
 * the anonymous pages are kept on the inactive list.
 *
 * total     target    max
 * memory    ratio     inactive anon
 * -------------------------------------
 *   10MB       1         5MB
 *  100MB       1        50MB
 *    1GB       3       250MB
 *   10GB      10       0.9GB
 *  100GB      31         3GB
 *    1TB     101        10GB
 *   10TB     320        32GB
 */
5716
static void __meminit calculate_zone_inactive_ratio(struct zone *zone)
5717
{
5718
	unsigned int gb, ratio;
5719

5720
	/* Zone size in gigabytes */
5721
	gb = zone->managed_pages >> (30 - PAGE_SHIFT);
5722
	if (gb)
5723
		ratio = int_sqrt(10 * gb);
5724 5725
	else
		ratio = 1;
5726

5727 5728
	zone->inactive_ratio = ratio;
}
5729

5730
static void __meminit setup_per_zone_inactive_ratio(void)
5731 5732 5733 5734 5735
{
	struct zone *zone;

	for_each_zone(zone)
		calculate_zone_inactive_ratio(zone);
5736 5737
}

L
Linus Torvalds 已提交
5738 5739 5740 5741 5742 5743 5744
/*
 * Initialise min_free_kbytes.
 *
 * For small machines we want it small (128k min).  For large machines
 * we want it large (64MB max).  But it is not linear, because network
 * bandwidth does not increase linearly with machine size.  We use
 *
5745
 *	min_free_kbytes = 4 * sqrt(lowmem_kbytes), for better accuracy:
L
Linus Torvalds 已提交
5746 5747 5748 5749 5750 5751 5752 5753 5754 5755 5756 5757 5758 5759 5760 5761
 *	min_free_kbytes = sqrt(lowmem_kbytes * 16)
 *
 * which yields
 *
 * 16MB:	512k
 * 32MB:	724k
 * 64MB:	1024k
 * 128MB:	1448k
 * 256MB:	2048k
 * 512MB:	2896k
 * 1024MB:	4096k
 * 2048MB:	5792k
 * 4096MB:	8192k
 * 8192MB:	11584k
 * 16384MB:	16384k
 */
5762
int __meminit init_per_zone_wmark_min(void)
L
Linus Torvalds 已提交
5763 5764
{
	unsigned long lowmem_kbytes;
5765
	int new_min_free_kbytes;
L
Linus Torvalds 已提交
5766 5767

	lowmem_kbytes = nr_free_buffer_pages() * (PAGE_SIZE >> 10);
5768 5769 5770 5771 5772 5773 5774 5775 5776 5777 5778 5779
	new_min_free_kbytes = int_sqrt(lowmem_kbytes * 16);

	if (new_min_free_kbytes > user_min_free_kbytes) {
		min_free_kbytes = new_min_free_kbytes;
		if (min_free_kbytes < 128)
			min_free_kbytes = 128;
		if (min_free_kbytes > 65536)
			min_free_kbytes = 65536;
	} else {
		pr_warn("min_free_kbytes is not updated to %d because user defined value %d is preferred\n",
				new_min_free_kbytes, user_min_free_kbytes);
	}
5780
	setup_per_zone_wmarks();
5781
	refresh_zone_stat_thresholds();
L
Linus Torvalds 已提交
5782
	setup_per_zone_lowmem_reserve();
5783
	setup_per_zone_inactive_ratio();
L
Linus Torvalds 已提交
5784 5785
	return 0;
}
5786
module_init(init_per_zone_wmark_min)
L
Linus Torvalds 已提交
5787 5788

/*
5789
 * min_free_kbytes_sysctl_handler - just a wrapper around proc_dointvec() so
L
Linus Torvalds 已提交
5790 5791 5792
 *	that we can call two helper functions whenever min_free_kbytes
 *	changes.
 */
5793
int min_free_kbytes_sysctl_handler(ctl_table *table, int write,
5794
	void __user *buffer, size_t *length, loff_t *ppos)
L
Linus Torvalds 已提交
5795
{
5796 5797 5798 5799 5800 5801
	int rc;

	rc = proc_dointvec_minmax(table, write, buffer, length, ppos);
	if (rc)
		return rc;

5802 5803
	if (write) {
		user_min_free_kbytes = min_free_kbytes;
5804
		setup_per_zone_wmarks();
5805
	}
L
Linus Torvalds 已提交
5806 5807 5808
	return 0;
}

5809 5810
#ifdef CONFIG_NUMA
int sysctl_min_unmapped_ratio_sysctl_handler(ctl_table *table, int write,
5811
	void __user *buffer, size_t *length, loff_t *ppos)
5812 5813 5814 5815
{
	struct zone *zone;
	int rc;

5816
	rc = proc_dointvec_minmax(table, write, buffer, length, ppos);
5817 5818 5819 5820
	if (rc)
		return rc;

	for_each_zone(zone)
5821
		zone->min_unmapped_pages = (zone->managed_pages *
5822 5823 5824
				sysctl_min_unmapped_ratio) / 100;
	return 0;
}
5825 5826

int sysctl_min_slab_ratio_sysctl_handler(ctl_table *table, int write,
5827
	void __user *buffer, size_t *length, loff_t *ppos)
5828 5829 5830 5831
{
	struct zone *zone;
	int rc;

5832
	rc = proc_dointvec_minmax(table, write, buffer, length, ppos);
5833 5834 5835 5836
	if (rc)
		return rc;

	for_each_zone(zone)
5837
		zone->min_slab_pages = (zone->managed_pages *
5838 5839 5840
				sysctl_min_slab_ratio) / 100;
	return 0;
}
5841 5842
#endif

L
Linus Torvalds 已提交
5843 5844 5845 5846 5847 5848
/*
 * lowmem_reserve_ratio_sysctl_handler - just a wrapper around
 *	proc_dointvec() so that we can call setup_per_zone_lowmem_reserve()
 *	whenever sysctl_lowmem_reserve_ratio changes.
 *
 * The reserve ratio obviously has absolutely no relation with the
5849
 * minimum watermarks. The lowmem reserve ratio can only make sense
L
Linus Torvalds 已提交
5850 5851 5852
 * if in function of the boot time zone sizes.
 */
int lowmem_reserve_ratio_sysctl_handler(ctl_table *table, int write,
5853
	void __user *buffer, size_t *length, loff_t *ppos)
L
Linus Torvalds 已提交
5854
{
5855
	proc_dointvec_minmax(table, write, buffer, length, ppos);
L
Linus Torvalds 已提交
5856 5857 5858 5859
	setup_per_zone_lowmem_reserve();
	return 0;
}

5860 5861
/*
 * percpu_pagelist_fraction - changes the pcp->high for each zone on each
5862 5863
 * cpu.  It is the fraction of total pages in each zone that a hot per cpu
 * pagelist can have before it gets flushed back to buddy allocator.
5864 5865
 */
int percpu_pagelist_fraction_sysctl_handler(ctl_table *table, int write,
5866
	void __user *buffer, size_t *length, loff_t *ppos)
5867 5868 5869 5870 5871
{
	struct zone *zone;
	unsigned int cpu;
	int ret;

5872
	ret = proc_dointvec_minmax(table, write, buffer, length, ppos);
5873
	if (!write || (ret < 0))
5874
		return ret;
5875 5876

	mutex_lock(&pcp_batch_high_lock);
5877
	for_each_populated_zone(zone) {
5878 5879 5880
		unsigned long  high;
		high = zone->managed_pages / percpu_pagelist_fraction;
		for_each_possible_cpu(cpu)
5881 5882
			pageset_set_high(per_cpu_ptr(zone->pageset, cpu),
					 high);
5883
	}
5884
	mutex_unlock(&pcp_batch_high_lock);
5885 5886 5887
	return 0;
}

5888
int hashdist = HASHDIST_DEFAULT;
L
Linus Torvalds 已提交
5889 5890 5891 5892 5893 5894 5895 5896 5897 5898 5899 5900 5901 5902 5903 5904 5905 5906 5907 5908 5909 5910 5911 5912 5913

#ifdef CONFIG_NUMA
static int __init set_hashdist(char *str)
{
	if (!str)
		return 0;
	hashdist = simple_strtoul(str, &str, 0);
	return 1;
}
__setup("hashdist=", set_hashdist);
#endif

/*
 * allocate a large system hash table from bootmem
 * - it is assumed that the hash table must contain an exact power-of-2
 *   quantity of entries
 * - limit is the number of hash buckets, not the total allocation size
 */
void *__init alloc_large_system_hash(const char *tablename,
				     unsigned long bucketsize,
				     unsigned long numentries,
				     int scale,
				     int flags,
				     unsigned int *_hash_shift,
				     unsigned int *_hash_mask,
5914 5915
				     unsigned long low_limit,
				     unsigned long high_limit)
L
Linus Torvalds 已提交
5916
{
5917
	unsigned long long max = high_limit;
L
Linus Torvalds 已提交
5918 5919 5920 5921 5922 5923
	unsigned long log2qty, size;
	void *table = NULL;

	/* allow the kernel cmdline to have a say */
	if (!numentries) {
		/* round applicable memory size up to nearest megabyte */
A
Andrew Morton 已提交
5924
		numentries = nr_kernel_pages;
5925 5926 5927 5928

		/* It isn't necessary when PAGE_SIZE >= 1MB */
		if (PAGE_SHIFT < 20)
			numentries = round_up(numentries, (1<<20)/PAGE_SIZE);
L
Linus Torvalds 已提交
5929 5930 5931 5932 5933 5934

		/* limit to 1 bucket per 2^scale bytes of low memory */
		if (scale > PAGE_SHIFT)
			numentries >>= (scale - PAGE_SHIFT);
		else
			numentries <<= (PAGE_SHIFT - scale);
5935 5936

		/* Make sure we've got at least a 0-order allocation.. */
5937 5938 5939 5940 5941 5942 5943 5944
		if (unlikely(flags & HASH_SMALL)) {
			/* Makes no sense without HASH_EARLY */
			WARN_ON(!(flags & HASH_EARLY));
			if (!(numentries >> *_hash_shift)) {
				numentries = 1UL << *_hash_shift;
				BUG_ON(!numentries);
			}
		} else if (unlikely((numentries * bucketsize) < PAGE_SIZE))
5945
			numentries = PAGE_SIZE / bucketsize;
L
Linus Torvalds 已提交
5946
	}
5947
	numentries = roundup_pow_of_two(numentries);
L
Linus Torvalds 已提交
5948 5949 5950 5951 5952 5953

	/* limit allocation size to 1/16 total memory by default */
	if (max == 0) {
		max = ((unsigned long long)nr_all_pages << PAGE_SHIFT) >> 4;
		do_div(max, bucketsize);
	}
5954
	max = min(max, 0x80000000ULL);
L
Linus Torvalds 已提交
5955

5956 5957
	if (numentries < low_limit)
		numentries = low_limit;
L
Linus Torvalds 已提交
5958 5959 5960
	if (numentries > max)
		numentries = max;

5961
	log2qty = ilog2(numentries);
L
Linus Torvalds 已提交
5962 5963 5964 5965

	do {
		size = bucketsize << log2qty;
		if (flags & HASH_EARLY)
5966
			table = memblock_virt_alloc_nopanic(size, 0);
L
Linus Torvalds 已提交
5967 5968 5969
		else if (hashdist)
			table = __vmalloc(size, GFP_ATOMIC, PAGE_KERNEL);
		else {
5970 5971
			/*
			 * If bucketsize is not a power-of-two, we may free
5972 5973
			 * some pages at the end of hash table which
			 * alloc_pages_exact() automatically does
5974
			 */
5975
			if (get_order(size) < MAX_ORDER) {
5976
				table = alloc_pages_exact(size, GFP_ATOMIC);
5977 5978
				kmemleak_alloc(table, size, 1, GFP_ATOMIC);
			}
L
Linus Torvalds 已提交
5979 5980 5981 5982 5983 5984
		}
	} while (!table && size > PAGE_SIZE && --log2qty);

	if (!table)
		panic("Failed to allocate %s hash table\n", tablename);

5985
	printk(KERN_INFO "%s hash table entries: %ld (order: %d, %lu bytes)\n",
L
Linus Torvalds 已提交
5986
	       tablename,
5987
	       (1UL << log2qty),
5988
	       ilog2(size) - PAGE_SHIFT,
L
Linus Torvalds 已提交
5989 5990 5991 5992 5993 5994 5995 5996 5997
	       size);

	if (_hash_shift)
		*_hash_shift = log2qty;
	if (_hash_mask)
		*_hash_mask = (1 << log2qty) - 1;

	return table;
}
5998

5999 6000 6001 6002 6003 6004 6005 6006 6007 6008 6009 6010 6011 6012 6013
/* Return a pointer to the bitmap storing bits affecting a block of pages */
static inline unsigned long *get_pageblock_bitmap(struct zone *zone,
							unsigned long pfn)
{
#ifdef CONFIG_SPARSEMEM
	return __pfn_to_section(pfn)->pageblock_flags;
#else
	return zone->pageblock_flags;
#endif /* CONFIG_SPARSEMEM */
}

static inline int pfn_to_bitidx(struct zone *zone, unsigned long pfn)
{
#ifdef CONFIG_SPARSEMEM
	pfn &= (PAGES_PER_SECTION-1);
6014
	return (pfn >> pageblock_order) * NR_PAGEBLOCK_BITS;
6015
#else
6016
	pfn = pfn - round_down(zone->zone_start_pfn, pageblock_nr_pages);
6017
	return (pfn >> pageblock_order) * NR_PAGEBLOCK_BITS;
6018 6019 6020 6021
#endif /* CONFIG_SPARSEMEM */
}

/**
6022
 * get_pageblock_flags_group - Return the requested group of flags for the pageblock_nr_pages block of pages
6023 6024 6025 6026 6027 6028 6029 6030 6031 6032 6033 6034 6035 6036 6037 6038 6039 6040 6041 6042 6043 6044
 * @page: The page within the block of interest
 * @start_bitidx: The first bit of interest to retrieve
 * @end_bitidx: The last bit of interest
 * returns pageblock_bits flags
 */
unsigned long get_pageblock_flags_group(struct page *page,
					int start_bitidx, int end_bitidx)
{
	struct zone *zone;
	unsigned long *bitmap;
	unsigned long pfn, bitidx;
	unsigned long flags = 0;
	unsigned long value = 1;

	zone = page_zone(page);
	pfn = page_to_pfn(page);
	bitmap = get_pageblock_bitmap(zone, pfn);
	bitidx = pfn_to_bitidx(zone, pfn);

	for (; start_bitidx <= end_bitidx; start_bitidx++, value <<= 1)
		if (test_bit(bitidx + start_bitidx, bitmap))
			flags |= value;
6045

6046 6047 6048 6049
	return flags;
}

/**
6050
 * set_pageblock_flags_group - Set the requested group of flags for a pageblock_nr_pages block of pages
6051 6052 6053 6054 6055 6056 6057 6058 6059 6060 6061 6062 6063 6064 6065 6066 6067
 * @page: The page within the block of interest
 * @start_bitidx: The first bit of interest
 * @end_bitidx: The last bit of interest
 * @flags: The flags to set
 */
void set_pageblock_flags_group(struct page *page, unsigned long flags,
					int start_bitidx, int end_bitidx)
{
	struct zone *zone;
	unsigned long *bitmap;
	unsigned long pfn, bitidx;
	unsigned long value = 1;

	zone = page_zone(page);
	pfn = page_to_pfn(page);
	bitmap = get_pageblock_bitmap(zone, pfn);
	bitidx = pfn_to_bitidx(zone, pfn);
6068
	VM_BUG_ON_PAGE(!zone_spans_pfn(zone, pfn), page);
6069 6070 6071 6072 6073 6074 6075

	for (; start_bitidx <= end_bitidx; start_bitidx++, value <<= 1)
		if (flags & value)
			__set_bit(bitidx + start_bitidx, bitmap);
		else
			__clear_bit(bitidx + start_bitidx, bitmap);
}
K
KAMEZAWA Hiroyuki 已提交
6076 6077

/*
6078 6079 6080
 * This function checks whether pageblock includes unmovable pages or not.
 * If @count is not zero, it is okay to include less @count unmovable pages
 *
6081
 * PageLRU check without isolation or lru_lock could race so that
6082 6083
 * MIGRATE_MOVABLE block might include unmovable pages. It means you can't
 * expect this function should be exact.
K
KAMEZAWA Hiroyuki 已提交
6084
 */
6085 6086
bool has_unmovable_pages(struct zone *zone, struct page *page, int count,
			 bool skip_hwpoisoned_pages)
6087 6088
{
	unsigned long pfn, iter, found;
6089 6090
	int mt;

6091 6092
	/*
	 * For avoiding noise data, lru_add_drain_all() should be called
6093
	 * If ZONE_MOVABLE, the zone never contains unmovable pages
6094 6095
	 */
	if (zone_idx(zone) == ZONE_MOVABLE)
6096
		return false;
6097 6098
	mt = get_pageblock_migratetype(page);
	if (mt == MIGRATE_MOVABLE || is_migrate_cma(mt))
6099
		return false;
6100 6101 6102 6103 6104

	pfn = page_to_pfn(page);
	for (found = 0, iter = 0; iter < pageblock_nr_pages; iter++) {
		unsigned long check = pfn + iter;

6105
		if (!pfn_valid_within(check))
6106
			continue;
6107

6108
		page = pfn_to_page(check);
6109 6110 6111 6112 6113 6114 6115 6116 6117 6118 6119

		/*
		 * Hugepages are not in LRU lists, but they're movable.
		 * We need not scan over tail pages bacause we don't
		 * handle each tail page individually in migration.
		 */
		if (PageHuge(page)) {
			iter = round_up(iter + 1, 1<<compound_order(page)) - 1;
			continue;
		}

6120 6121 6122 6123 6124 6125 6126
		/*
		 * We can't use page_count without pin a page
		 * because another CPU can free compound page.
		 * This check already skips compound tails of THP
		 * because their page->_count is zero at all time.
		 */
		if (!atomic_read(&page->_count)) {
6127 6128 6129 6130
			if (PageBuddy(page))
				iter += (1 << page_order(page)) - 1;
			continue;
		}
6131

6132 6133 6134 6135 6136 6137 6138
		/*
		 * The HWPoisoned page may be not in buddy system, and
		 * page_count() is not 0.
		 */
		if (skip_hwpoisoned_pages && PageHWPoison(page))
			continue;

6139 6140 6141 6142 6143 6144 6145 6146 6147 6148 6149 6150 6151 6152 6153 6154
		if (!PageLRU(page))
			found++;
		/*
		 * If there are RECLAIMABLE pages, we need to check it.
		 * But now, memory offline itself doesn't call shrink_slab()
		 * and it still to be fixed.
		 */
		/*
		 * If the page is not RAM, page_count()should be 0.
		 * we don't need more check. This is an _used_ not-movable page.
		 *
		 * The problematic thing here is PG_reserved pages. PG_reserved
		 * is set to both of a memory hole page and a _used_ kernel
		 * page at boot.
		 */
		if (found > count)
6155
			return true;
6156
	}
6157
	return false;
6158 6159 6160 6161
}

bool is_pageblock_removable_nolock(struct page *page)
{
6162 6163
	struct zone *zone;
	unsigned long pfn;
6164 6165 6166 6167 6168

	/*
	 * We have to be careful here because we are iterating over memory
	 * sections which are not zone aware so we might end up outside of
	 * the zone but still within the section.
6169 6170
	 * We have to take care about the node as well. If the node is offline
	 * its NODE_DATA will be NULL - see page_zone.
6171
	 */
6172 6173 6174 6175 6176
	if (!node_online(page_to_nid(page)))
		return false;

	zone = page_zone(page);
	pfn = page_to_pfn(page);
6177
	if (!zone_spans_pfn(zone, pfn))
6178 6179
		return false;

6180
	return !has_unmovable_pages(zone, page, 0, true);
K
KAMEZAWA Hiroyuki 已提交
6181
}
K
KAMEZAWA Hiroyuki 已提交
6182

6183 6184 6185 6186 6187 6188 6189 6190 6191 6192 6193 6194 6195 6196 6197
#ifdef CONFIG_CMA

static unsigned long pfn_max_align_down(unsigned long pfn)
{
	return pfn & ~(max_t(unsigned long, MAX_ORDER_NR_PAGES,
			     pageblock_nr_pages) - 1);
}

static unsigned long pfn_max_align_up(unsigned long pfn)
{
	return ALIGN(pfn, max_t(unsigned long, MAX_ORDER_NR_PAGES,
				pageblock_nr_pages));
}

/* [start, end) must belong to a single zone. */
6198 6199
static int __alloc_contig_migrate_range(struct compact_control *cc,
					unsigned long start, unsigned long end)
6200 6201
{
	/* This function is based on compact_zone() from compaction.c. */
6202
	unsigned long nr_reclaimed;
6203 6204 6205 6206
	unsigned long pfn = start;
	unsigned int tries = 0;
	int ret = 0;

6207
	migrate_prep();
6208

6209
	while (pfn < end || !list_empty(&cc->migratepages)) {
6210 6211 6212 6213 6214
		if (fatal_signal_pending(current)) {
			ret = -EINTR;
			break;
		}

6215 6216 6217
		if (list_empty(&cc->migratepages)) {
			cc->nr_migratepages = 0;
			pfn = isolate_migratepages_range(cc->zone, cc,
M
Minchan Kim 已提交
6218
							 pfn, end, true);
6219 6220 6221 6222 6223 6224 6225 6226 6227 6228
			if (!pfn) {
				ret = -EINTR;
				break;
			}
			tries = 0;
		} else if (++tries == 5) {
			ret = ret < 0 ? ret : -EBUSY;
			break;
		}

6229 6230 6231
		nr_reclaimed = reclaim_clean_pages_from_list(cc->zone,
							&cc->migratepages);
		cc->nr_migratepages -= nr_reclaimed;
6232

6233
		ret = migrate_pages(&cc->migratepages, alloc_migrate_target,
6234
				    NULL, 0, cc->mode, MR_CMA);
6235
	}
6236 6237 6238 6239 6240
	if (ret < 0) {
		putback_movable_pages(&cc->migratepages);
		return ret;
	}
	return 0;
6241 6242 6243 6244 6245 6246
}

/**
 * alloc_contig_range() -- tries to allocate given range of pages
 * @start:	start PFN to allocate
 * @end:	one-past-the-last PFN to allocate
6247 6248 6249 6250
 * @migratetype:	migratetype of the underlaying pageblocks (either
 *			#MIGRATE_MOVABLE or #MIGRATE_CMA).  All pageblocks
 *			in range must have the same migratetype and it must
 *			be either of the two.
6251 6252 6253 6254 6255 6256 6257 6258 6259 6260 6261 6262
 *
 * The PFN range does not have to be pageblock or MAX_ORDER_NR_PAGES
 * aligned, however it's the caller's responsibility to guarantee that
 * we are the only thread that changes migrate type of pageblocks the
 * pages fall in.
 *
 * The PFN range must belong to a single zone.
 *
 * Returns zero on success or negative error code.  On success all
 * pages which PFN is in [start, end) are allocated for the caller and
 * need to be freed with free_contig_range().
 */
6263 6264
int alloc_contig_range(unsigned long start, unsigned long end,
		       unsigned migratetype)
6265 6266 6267 6268
{
	unsigned long outer_start, outer_end;
	int ret = 0, order;

6269 6270 6271 6272
	struct compact_control cc = {
		.nr_migratepages = 0,
		.order = -1,
		.zone = page_zone(pfn_to_page(start)),
6273
		.mode = MIGRATE_SYNC,
6274 6275 6276 6277
		.ignore_skip_hint = true,
	};
	INIT_LIST_HEAD(&cc.migratepages);

6278 6279 6280 6281 6282 6283 6284 6285 6286 6287 6288 6289 6290 6291 6292 6293 6294 6295 6296 6297 6298 6299 6300 6301 6302
	/*
	 * What we do here is we mark all pageblocks in range as
	 * MIGRATE_ISOLATE.  Because pageblock and max order pages may
	 * have different sizes, and due to the way page allocator
	 * work, we align the range to biggest of the two pages so
	 * that page allocator won't try to merge buddies from
	 * different pageblocks and change MIGRATE_ISOLATE to some
	 * other migration type.
	 *
	 * Once the pageblocks are marked as MIGRATE_ISOLATE, we
	 * migrate the pages from an unaligned range (ie. pages that
	 * we are interested in).  This will put all the pages in
	 * range back to page allocator as MIGRATE_ISOLATE.
	 *
	 * When this is done, we take the pages in range from page
	 * allocator removing them from the buddy system.  This way
	 * page allocator will never consider using them.
	 *
	 * This lets us mark the pageblocks back as
	 * MIGRATE_CMA/MIGRATE_MOVABLE so that free pages in the
	 * aligned range but not in the unaligned, original range are
	 * put back to page allocator so that buddy can use them.
	 */

	ret = start_isolate_page_range(pfn_max_align_down(start),
6303 6304
				       pfn_max_align_up(end), migratetype,
				       false);
6305
	if (ret)
6306
		return ret;
6307

6308
	ret = __alloc_contig_migrate_range(&cc, start, end);
6309 6310 6311 6312 6313 6314 6315 6316 6317 6318 6319 6320 6321 6322 6323 6324 6325 6326 6327 6328 6329 6330 6331 6332 6333 6334 6335 6336 6337 6338 6339 6340 6341 6342
	if (ret)
		goto done;

	/*
	 * Pages from [start, end) are within a MAX_ORDER_NR_PAGES
	 * aligned blocks that are marked as MIGRATE_ISOLATE.  What's
	 * more, all pages in [start, end) are free in page allocator.
	 * What we are going to do is to allocate all pages from
	 * [start, end) (that is remove them from page allocator).
	 *
	 * The only problem is that pages at the beginning and at the
	 * end of interesting range may be not aligned with pages that
	 * page allocator holds, ie. they can be part of higher order
	 * pages.  Because of this, we reserve the bigger range and
	 * once this is done free the pages we are not interested in.
	 *
	 * We don't have to hold zone->lock here because the pages are
	 * isolated thus they won't get removed from buddy.
	 */

	lru_add_drain_all();
	drain_all_pages();

	order = 0;
	outer_start = start;
	while (!PageBuddy(pfn_to_page(outer_start))) {
		if (++order >= MAX_ORDER) {
			ret = -EBUSY;
			goto done;
		}
		outer_start &= ~0UL << order;
	}

	/* Make sure the range is really isolated. */
6343
	if (test_pages_isolated(outer_start, end, false)) {
6344 6345 6346 6347 6348 6349
		pr_warn("alloc_contig_range test_pages_isolated(%lx, %lx) failed\n",
		       outer_start, end);
		ret = -EBUSY;
		goto done;
	}

6350 6351

	/* Grab isolated pages from freelists. */
6352
	outer_end = isolate_freepages_range(&cc, outer_start, end);
6353 6354 6355 6356 6357 6358 6359 6360 6361 6362 6363 6364 6365
	if (!outer_end) {
		ret = -EBUSY;
		goto done;
	}

	/* Free head and tail (if any) */
	if (start != outer_start)
		free_contig_range(outer_start, start - outer_start);
	if (end != outer_end)
		free_contig_range(end, outer_end - end);

done:
	undo_isolate_page_range(pfn_max_align_down(start),
6366
				pfn_max_align_up(end), migratetype);
6367 6368 6369 6370 6371
	return ret;
}

void free_contig_range(unsigned long pfn, unsigned nr_pages)
{
6372 6373 6374 6375 6376 6377 6378 6379 6380
	unsigned int count = 0;

	for (; nr_pages--; pfn++) {
		struct page *page = pfn_to_page(pfn);

		count += page_count(page) != 1;
		__free_page(page);
	}
	WARN(count != 0, "%d pages are still in use!\n", count);
6381 6382 6383
}
#endif

6384
#ifdef CONFIG_MEMORY_HOTPLUG
6385 6386 6387 6388
/*
 * The zone indicated has a new number of managed_pages; batch sizes and percpu
 * page high values need to be recalulated.
 */
6389 6390
void __meminit zone_pcp_update(struct zone *zone)
{
6391
	unsigned cpu;
6392
	mutex_lock(&pcp_batch_high_lock);
6393
	for_each_possible_cpu(cpu)
6394 6395
		pageset_set_high_and_batch(zone,
				per_cpu_ptr(zone->pageset, cpu));
6396
	mutex_unlock(&pcp_batch_high_lock);
6397 6398 6399
}
#endif

6400 6401 6402
void zone_pcp_reset(struct zone *zone)
{
	unsigned long flags;
6403 6404
	int cpu;
	struct per_cpu_pageset *pset;
6405 6406 6407 6408

	/* avoid races with drain_pages()  */
	local_irq_save(flags);
	if (zone->pageset != &boot_pageset) {
6409 6410 6411 6412
		for_each_online_cpu(cpu) {
			pset = per_cpu_ptr(zone->pageset, cpu);
			drain_zonestat(zone, pset);
		}
6413 6414 6415 6416 6417 6418
		free_percpu(zone->pageset);
		zone->pageset = &boot_pageset;
	}
	local_irq_restore(flags);
}

6419
#ifdef CONFIG_MEMORY_HOTREMOVE
K
KAMEZAWA Hiroyuki 已提交
6420 6421 6422 6423 6424 6425 6426 6427 6428 6429 6430 6431 6432 6433 6434 6435 6436 6437 6438 6439 6440 6441 6442 6443 6444 6445
/*
 * All pages in the range must be isolated before calling this.
 */
void
__offline_isolated_pages(unsigned long start_pfn, unsigned long end_pfn)
{
	struct page *page;
	struct zone *zone;
	int order, i;
	unsigned long pfn;
	unsigned long flags;
	/* find the first valid pfn */
	for (pfn = start_pfn; pfn < end_pfn; pfn++)
		if (pfn_valid(pfn))
			break;
	if (pfn == end_pfn)
		return;
	zone = page_zone(pfn_to_page(pfn));
	spin_lock_irqsave(&zone->lock, flags);
	pfn = start_pfn;
	while (pfn < end_pfn) {
		if (!pfn_valid(pfn)) {
			pfn++;
			continue;
		}
		page = pfn_to_page(pfn);
6446 6447 6448 6449 6450 6451 6452 6453 6454 6455
		/*
		 * The HWPoisoned page may be not in buddy system, and
		 * page_count() is not 0.
		 */
		if (unlikely(!PageBuddy(page) && PageHWPoison(page))) {
			pfn++;
			SetPageReserved(page);
			continue;
		}

K
KAMEZAWA Hiroyuki 已提交
6456 6457 6458 6459 6460 6461 6462 6463 6464 6465 6466 6467 6468 6469 6470 6471 6472
		BUG_ON(page_count(page));
		BUG_ON(!PageBuddy(page));
		order = page_order(page);
#ifdef CONFIG_DEBUG_VM
		printk(KERN_INFO "remove from free list %lx %d %lx\n",
		       pfn, 1 << order, end_pfn);
#endif
		list_del(&page->lru);
		rmv_page_order(page);
		zone->free_area[order].nr_free--;
		for (i = 0; i < (1 << order); i++)
			SetPageReserved((page+i));
		pfn += (1 << order);
	}
	spin_unlock_irqrestore(&zone->lock, flags);
}
#endif
6473 6474 6475 6476 6477 6478 6479 6480 6481 6482 6483 6484 6485 6486 6487 6488 6489 6490 6491 6492 6493

#ifdef CONFIG_MEMORY_FAILURE
bool is_free_buddy_page(struct page *page)
{
	struct zone *zone = page_zone(page);
	unsigned long pfn = page_to_pfn(page);
	unsigned long flags;
	int order;

	spin_lock_irqsave(&zone->lock, flags);
	for (order = 0; order < MAX_ORDER; order++) {
		struct page *page_head = page - (pfn & ((1 << order) - 1));

		if (PageBuddy(page_head) && page_order(page_head) >= order)
			break;
	}
	spin_unlock_irqrestore(&zone->lock, flags);

	return order < MAX_ORDER;
}
#endif
6494

A
Andrew Morton 已提交
6495
static const struct trace_print_flags pageflag_names[] = {
6496 6497 6498 6499 6500 6501 6502 6503 6504 6505 6506 6507 6508 6509 6510 6511 6512 6513 6514 6515 6516 6517 6518 6519 6520 6521 6522 6523 6524 6525 6526 6527 6528
	{1UL << PG_locked,		"locked"	},
	{1UL << PG_error,		"error"		},
	{1UL << PG_referenced,		"referenced"	},
	{1UL << PG_uptodate,		"uptodate"	},
	{1UL << PG_dirty,		"dirty"		},
	{1UL << PG_lru,			"lru"		},
	{1UL << PG_active,		"active"	},
	{1UL << PG_slab,		"slab"		},
	{1UL << PG_owner_priv_1,	"owner_priv_1"	},
	{1UL << PG_arch_1,		"arch_1"	},
	{1UL << PG_reserved,		"reserved"	},
	{1UL << PG_private,		"private"	},
	{1UL << PG_private_2,		"private_2"	},
	{1UL << PG_writeback,		"writeback"	},
#ifdef CONFIG_PAGEFLAGS_EXTENDED
	{1UL << PG_head,		"head"		},
	{1UL << PG_tail,		"tail"		},
#else
	{1UL << PG_compound,		"compound"	},
#endif
	{1UL << PG_swapcache,		"swapcache"	},
	{1UL << PG_mappedtodisk,	"mappedtodisk"	},
	{1UL << PG_reclaim,		"reclaim"	},
	{1UL << PG_swapbacked,		"swapbacked"	},
	{1UL << PG_unevictable,		"unevictable"	},
#ifdef CONFIG_MMU
	{1UL << PG_mlocked,		"mlocked"	},
#endif
#ifdef CONFIG_ARCH_USES_PG_UNCACHED
	{1UL << PG_uncached,		"uncached"	},
#endif
#ifdef CONFIG_MEMORY_FAILURE
	{1UL << PG_hwpoison,		"hwpoison"	},
6529 6530 6531
#endif
#ifdef CONFIG_TRANSPARENT_HUGEPAGE
	{1UL << PG_compound_lock,	"compound_lock"	},
6532 6533 6534 6535 6536 6537 6538 6539 6540
#endif
};

static void dump_page_flags(unsigned long flags)
{
	const char *delim = "";
	unsigned long mask;
	int i;

A
Andrew Morton 已提交
6541
	BUILD_BUG_ON(ARRAY_SIZE(pageflag_names) != __NR_PAGEFLAGS);
6542

6543 6544 6545 6546 6547
	printk(KERN_ALERT "page flags: %#lx(", flags);

	/* remove zone id */
	flags &= (1UL << NR_PAGEFLAGS) - 1;

A
Andrew Morton 已提交
6548
	for (i = 0; i < ARRAY_SIZE(pageflag_names) && flags; i++) {
6549 6550 6551 6552 6553 6554 6555 6556 6557 6558 6559 6560 6561 6562 6563 6564 6565

		mask = pageflag_names[i].mask;
		if ((flags & mask) != mask)
			continue;

		flags &= ~mask;
		printk("%s%s", delim, pageflag_names[i].name);
		delim = "|";
	}

	/* check for left over flags */
	if (flags)
		printk("%s%#lx", delim, flags);

	printk(")\n");
}

6566 6567
void dump_page_badflags(struct page *page, const char *reason,
		unsigned long badflags)
6568 6569 6570
{
	printk(KERN_ALERT
	       "page:%p count:%d mapcount:%d mapping:%p index:%#lx\n",
6571
		page, atomic_read(&page->_count), page_mapcount(page),
6572 6573
		page->mapping, page->index);
	dump_page_flags(page->flags);
6574 6575 6576 6577 6578 6579
	if (reason)
		pr_alert("page dumped because: %s\n", reason);
	if (page->flags & badflags) {
		pr_alert("bad because of flags:\n");
		dump_page_flags(page->flags & badflags);
	}
6580
	mem_cgroup_print_bad_page(page);
6581
}
6582

6583
void dump_page(struct page *page, const char *reason)
6584 6585 6586
{
	dump_page_badflags(page, reason, 0);
}
6587
EXPORT_SYMBOL(dump_page);