hugetlb.c 76.7 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4 5 6 7 8
/*
 * Generic hugetlb support.
 * (C) William Irwin, April 2004
 */
#include <linux/list.h>
#include <linux/init.h>
#include <linux/module.h>
#include <linux/mm.h>
9
#include <linux/seq_file.h>
L
Linus Torvalds 已提交
10 11
#include <linux/sysctl.h>
#include <linux/highmem.h>
A
Andrea Arcangeli 已提交
12
#include <linux/mmu_notifier.h>
L
Linus Torvalds 已提交
13
#include <linux/nodemask.h>
D
David Gibson 已提交
14
#include <linux/pagemap.h>
15
#include <linux/mempolicy.h>
16
#include <linux/cpuset.h>
17
#include <linux/mutex.h>
18
#include <linux/bootmem.h>
19
#include <linux/sysfs.h>
20
#include <linux/slab.h>
21
#include <linux/rmap.h>
22 23
#include <linux/swap.h>
#include <linux/swapops.h>
24

D
David Gibson 已提交
25 26
#include <asm/page.h>
#include <asm/pgtable.h>
27
#include <asm/io.h>
D
David Gibson 已提交
28 29

#include <linux/hugetlb.h>
30
#include <linux/node.h>
31
#include "internal.h"
L
Linus Torvalds 已提交
32 33

const unsigned long hugetlb_zero = 0, hugetlb_infinity = ~0UL;
34 35
static gfp_t htlb_alloc_mask = GFP_HIGHUSER;
unsigned long hugepages_treat_as_movable;
36

37 38 39 40
static int max_hstate;
unsigned int default_hstate_idx;
struct hstate hstates[HUGE_MAX_HSTATE];

41 42
__initdata LIST_HEAD(huge_boot_pages);

43 44 45
/* for command line parsing */
static struct hstate * __initdata parsed_hstate;
static unsigned long __initdata default_hstate_max_huge_pages;
46
static unsigned long __initdata default_hstate_size;
47 48 49

#define for_each_hstate(h) \
	for ((h) = hstates; (h) < &hstates[max_hstate]; (h)++)
50

51 52 53 54
/*
 * Protects updates to hugepage_freelists, nr_huge_pages, and free_huge_pages
 */
static DEFINE_SPINLOCK(hugetlb_lock);
55

56 57 58
/*
 * Region tracking -- allows tracking of reservations and instantiated pages
 *                    across the pages in a mapping.
59 60 61 62 63 64 65 66 67 68
 *
 * The region data structures are protected by a combination of the mmap_sem
 * and the hugetlb_instantion_mutex.  To access or modify a region the caller
 * must either hold the mmap_sem for write, or the mmap_sem for read and
 * the hugetlb_instantiation mutex:
 *
 * 	down_write(&mm->mmap_sem);
 * or
 * 	down_read(&mm->mmap_sem);
 * 	mutex_lock(&hugetlb_instantiation_mutex);
69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190
 */
struct file_region {
	struct list_head link;
	long from;
	long to;
};

static long region_add(struct list_head *head, long f, long t)
{
	struct file_region *rg, *nrg, *trg;

	/* Locate the region we are either in or before. */
	list_for_each_entry(rg, head, link)
		if (f <= rg->to)
			break;

	/* Round our left edge to the current segment if it encloses us. */
	if (f > rg->from)
		f = rg->from;

	/* Check for and consume any regions we now overlap with. */
	nrg = rg;
	list_for_each_entry_safe(rg, trg, rg->link.prev, link) {
		if (&rg->link == head)
			break;
		if (rg->from > t)
			break;

		/* If this area reaches higher then extend our area to
		 * include it completely.  If this is not the first area
		 * which we intend to reuse, free it. */
		if (rg->to > t)
			t = rg->to;
		if (rg != nrg) {
			list_del(&rg->link);
			kfree(rg);
		}
	}
	nrg->from = f;
	nrg->to = t;
	return 0;
}

static long region_chg(struct list_head *head, long f, long t)
{
	struct file_region *rg, *nrg;
	long chg = 0;

	/* Locate the region we are before or in. */
	list_for_each_entry(rg, head, link)
		if (f <= rg->to)
			break;

	/* If we are below the current region then a new region is required.
	 * Subtle, allocate a new region at the position but make it zero
	 * size such that we can guarantee to record the reservation. */
	if (&rg->link == head || t < rg->from) {
		nrg = kmalloc(sizeof(*nrg), GFP_KERNEL);
		if (!nrg)
			return -ENOMEM;
		nrg->from = f;
		nrg->to   = f;
		INIT_LIST_HEAD(&nrg->link);
		list_add(&nrg->link, rg->link.prev);

		return t - f;
	}

	/* Round our left edge to the current segment if it encloses us. */
	if (f > rg->from)
		f = rg->from;
	chg = t - f;

	/* Check for and consume any regions we now overlap with. */
	list_for_each_entry(rg, rg->link.prev, link) {
		if (&rg->link == head)
			break;
		if (rg->from > t)
			return chg;

		/* We overlap with this area, if it extends futher than
		 * us then we must extend ourselves.  Account for its
		 * existing reservation. */
		if (rg->to > t) {
			chg += rg->to - t;
			t = rg->to;
		}
		chg -= rg->to - rg->from;
	}
	return chg;
}

static long region_truncate(struct list_head *head, long end)
{
	struct file_region *rg, *trg;
	long chg = 0;

	/* Locate the region we are either in or before. */
	list_for_each_entry(rg, head, link)
		if (end <= rg->to)
			break;
	if (&rg->link == head)
		return 0;

	/* If we are in the middle of a region then adjust it. */
	if (end > rg->from) {
		chg = rg->to - end;
		rg->to = end;
		rg = list_entry(rg->link.next, typeof(*rg), link);
	}

	/* Drop any remaining regions. */
	list_for_each_entry_safe(rg, trg, rg->link.prev, link) {
		if (&rg->link == head)
			break;
		chg += rg->to - rg->from;
		list_del(&rg->link);
		kfree(rg);
	}
	return chg;
}

191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214
static long region_count(struct list_head *head, long f, long t)
{
	struct file_region *rg;
	long chg = 0;

	/* Locate each segment we overlap with, and count that overlap. */
	list_for_each_entry(rg, head, link) {
		int seg_from;
		int seg_to;

		if (rg->to <= f)
			continue;
		if (rg->from >= t)
			break;

		seg_from = max(rg->from, f);
		seg_to = min(rg->to, t);

		chg += seg_to - seg_from;
	}

	return chg;
}

215 216 217 218
/*
 * Convert the address within this vma to the page offset within
 * the mapping, in pagecache page units; huge pages here.
 */
219 220
static pgoff_t vma_hugecache_offset(struct hstate *h,
			struct vm_area_struct *vma, unsigned long address)
221
{
222 223
	return ((address - vma->vm_start) >> huge_page_shift(h)) +
			(vma->vm_pgoff >> huge_page_order(h));
224 225
}

226 227 228 229 230 231
pgoff_t linear_hugepage_index(struct vm_area_struct *vma,
				     unsigned long address)
{
	return vma_hugecache_offset(hstate_vma(vma), vma, address);
}

232 233 234 235 236 237 238 239 240 241 242 243 244 245 246
/*
 * Return the size of the pages allocated when backing a VMA. In the majority
 * cases this will be same size as used by the page table entries.
 */
unsigned long vma_kernel_pagesize(struct vm_area_struct *vma)
{
	struct hstate *hstate;

	if (!is_vm_hugetlb_page(vma))
		return PAGE_SIZE;

	hstate = hstate_vma(vma);

	return 1UL << (hstate->order + PAGE_SHIFT);
}
247
EXPORT_SYMBOL_GPL(vma_kernel_pagesize);
248

249 250 251 252 253 254 255 256 257 258 259 260 261
/*
 * Return the page size being used by the MMU to back a VMA. In the majority
 * of cases, the page size used by the kernel matches the MMU size. On
 * architectures where it differs, an architecture-specific version of this
 * function is required.
 */
#ifndef vma_mmu_pagesize
unsigned long vma_mmu_pagesize(struct vm_area_struct *vma)
{
	return vma_kernel_pagesize(vma);
}
#endif

262 263 264 265 266 267 268
/*
 * Flags for MAP_PRIVATE reservations.  These are stored in the bottom
 * bits of the reservation map pointer, which are always clear due to
 * alignment.
 */
#define HPAGE_RESV_OWNER    (1UL << 0)
#define HPAGE_RESV_UNMAPPED (1UL << 1)
269
#define HPAGE_RESV_MASK (HPAGE_RESV_OWNER | HPAGE_RESV_UNMAPPED)
270

271 272 273 274 275 276 277 278 279
/*
 * These helpers are used to track how many pages are reserved for
 * faults in a MAP_PRIVATE mapping. Only the process that called mmap()
 * is guaranteed to have their future faults succeed.
 *
 * With the exception of reset_vma_resv_huge_pages() which is called at fork(),
 * the reserve counters are updated with the hugetlb_lock held. It is safe
 * to reset the VMA at fork() time as it is not in use yet and there is no
 * chance of the global counters getting corrupted as a result of the values.
280 281 282 283 284 285 286 287 288
 *
 * The private mapping reservation is represented in a subtly different
 * manner to a shared mapping.  A shared mapping has a region map associated
 * with the underlying file, this region map represents the backing file
 * pages which have ever had a reservation assigned which this persists even
 * after the page is instantiated.  A private mapping has a region map
 * associated with the original mmap which is attached to all VMAs which
 * reference it, this region map represents those offsets which have consumed
 * reservation ie. where pages have been instantiated.
289
 */
290 291 292 293 294 295 296 297 298 299 300
static unsigned long get_vma_private_data(struct vm_area_struct *vma)
{
	return (unsigned long)vma->vm_private_data;
}

static void set_vma_private_data(struct vm_area_struct *vma,
							unsigned long value)
{
	vma->vm_private_data = (void *)value;
}

301 302 303 304 305
struct resv_map {
	struct kref refs;
	struct list_head regions;
};

306
static struct resv_map *resv_map_alloc(void)
307 308 309 310 311 312 313 314 315 316 317
{
	struct resv_map *resv_map = kmalloc(sizeof(*resv_map), GFP_KERNEL);
	if (!resv_map)
		return NULL;

	kref_init(&resv_map->refs);
	INIT_LIST_HEAD(&resv_map->regions);

	return resv_map;
}

318
static void resv_map_release(struct kref *ref)
319 320 321 322 323 324 325 326 327
{
	struct resv_map *resv_map = container_of(ref, struct resv_map, refs);

	/* Clear out any active regions before we release the map. */
	region_truncate(&resv_map->regions, 0);
	kfree(resv_map);
}

static struct resv_map *vma_resv_map(struct vm_area_struct *vma)
328 329
{
	VM_BUG_ON(!is_vm_hugetlb_page(vma));
330
	if (!(vma->vm_flags & VM_MAYSHARE))
331 332
		return (struct resv_map *)(get_vma_private_data(vma) &
							~HPAGE_RESV_MASK);
333
	return NULL;
334 335
}

336
static void set_vma_resv_map(struct vm_area_struct *vma, struct resv_map *map)
337 338
{
	VM_BUG_ON(!is_vm_hugetlb_page(vma));
339
	VM_BUG_ON(vma->vm_flags & VM_MAYSHARE);
340

341 342
	set_vma_private_data(vma, (get_vma_private_data(vma) &
				HPAGE_RESV_MASK) | (unsigned long)map);
343 344 345 346 347
}

static void set_vma_resv_flags(struct vm_area_struct *vma, unsigned long flags)
{
	VM_BUG_ON(!is_vm_hugetlb_page(vma));
348
	VM_BUG_ON(vma->vm_flags & VM_MAYSHARE);
349 350

	set_vma_private_data(vma, get_vma_private_data(vma) | flags);
351 352 353 354 355
}

static int is_vma_resv_set(struct vm_area_struct *vma, unsigned long flag)
{
	VM_BUG_ON(!is_vm_hugetlb_page(vma));
356 357

	return (get_vma_private_data(vma) & flag) != 0;
358 359 360
}

/* Decrement the reserved pages in the hugepage pool by one */
361 362
static void decrement_hugepage_resv_vma(struct hstate *h,
			struct vm_area_struct *vma)
363
{
364 365 366
	if (vma->vm_flags & VM_NORESERVE)
		return;

367
	if (vma->vm_flags & VM_MAYSHARE) {
368
		/* Shared mappings always use reserves */
369
		h->resv_huge_pages--;
370
	} else if (is_vma_resv_set(vma, HPAGE_RESV_OWNER)) {
371 372 373 374
		/*
		 * Only the process that called mmap() has reserves for
		 * private mappings.
		 */
375
		h->resv_huge_pages--;
376 377 378
	}
}

379
/* Reset counters to 0 and clear all HPAGE_RESV_* flags */
380 381 382
void reset_vma_resv_huge_pages(struct vm_area_struct *vma)
{
	VM_BUG_ON(!is_vm_hugetlb_page(vma));
383
	if (!(vma->vm_flags & VM_MAYSHARE))
384 385 386 387
		vma->vm_private_data = (void *)0;
}

/* Returns true if the VMA has associated reserve pages */
388
static int vma_has_reserves(struct vm_area_struct *vma)
389
{
390
	if (vma->vm_flags & VM_MAYSHARE)
391 392 393 394
		return 1;
	if (is_vma_resv_set(vma, HPAGE_RESV_OWNER))
		return 1;
	return 0;
395 396
}

397 398 399 400 401 402 403 404 405 406 407 408
static void clear_gigantic_page(struct page *page,
			unsigned long addr, unsigned long sz)
{
	int i;
	struct page *p = page;

	might_sleep();
	for (i = 0; i < sz/PAGE_SIZE; i++, p = mem_map_next(p, page, i)) {
		cond_resched();
		clear_user_highpage(p, addr + i * PAGE_SIZE);
	}
}
409 410
static void clear_huge_page(struct page *page,
			unsigned long addr, unsigned long sz)
411 412 413
{
	int i;

414
	if (unlikely(sz/PAGE_SIZE > MAX_ORDER_NR_PAGES)) {
H
Hannes Eder 已提交
415 416 417
		clear_gigantic_page(page, addr, sz);
		return;
	}
418

419
	might_sleep();
420
	for (i = 0; i < sz/PAGE_SIZE; i++) {
421
		cond_resched();
422
		clear_user_highpage(page + i, addr + i * PAGE_SIZE);
423 424 425
	}
}

426
static void copy_user_gigantic_page(struct page *dst, struct page *src,
427 428 429 430 431 432
			   unsigned long addr, struct vm_area_struct *vma)
{
	int i;
	struct hstate *h = hstate_vma(vma);
	struct page *dst_base = dst;
	struct page *src_base = src;
433

434 435 436 437 438 439 440 441 442
	for (i = 0; i < pages_per_huge_page(h); ) {
		cond_resched();
		copy_user_highpage(dst, src, addr + i*PAGE_SIZE, vma);

		i++;
		dst = mem_map_next(dst, dst_base, i);
		src = mem_map_next(src, src_base, i);
	}
}
443 444

static void copy_user_huge_page(struct page *dst, struct page *src,
445
			   unsigned long addr, struct vm_area_struct *vma)
446 447
{
	int i;
448
	struct hstate *h = hstate_vma(vma);
449

H
Hannes Eder 已提交
450
	if (unlikely(pages_per_huge_page(h) > MAX_ORDER_NR_PAGES)) {
451
		copy_user_gigantic_page(dst, src, addr, vma);
H
Hannes Eder 已提交
452 453
		return;
	}
454

455
	might_sleep();
456
	for (i = 0; i < pages_per_huge_page(h); i++) {
457
		cond_resched();
458
		copy_user_highpage(dst + i, src + i, addr + i*PAGE_SIZE, vma);
459 460 461
	}
}

462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495
static void copy_gigantic_page(struct page *dst, struct page *src)
{
	int i;
	struct hstate *h = page_hstate(src);
	struct page *dst_base = dst;
	struct page *src_base = src;

	for (i = 0; i < pages_per_huge_page(h); ) {
		cond_resched();
		copy_highpage(dst, src);

		i++;
		dst = mem_map_next(dst, dst_base, i);
		src = mem_map_next(src, src_base, i);
	}
}

void copy_huge_page(struct page *dst, struct page *src)
{
	int i;
	struct hstate *h = page_hstate(src);

	if (unlikely(pages_per_huge_page(h) > MAX_ORDER_NR_PAGES)) {
		copy_gigantic_page(dst, src);
		return;
	}

	might_sleep();
	for (i = 0; i < pages_per_huge_page(h); i++) {
		cond_resched();
		copy_highpage(dst + i, src + i);
	}
}

496
static void enqueue_huge_page(struct hstate *h, struct page *page)
L
Linus Torvalds 已提交
497 498
{
	int nid = page_to_nid(page);
499 500 501
	list_add(&page->lru, &h->hugepage_freelists[nid]);
	h->free_huge_pages++;
	h->free_huge_pages_node[nid]++;
L
Linus Torvalds 已提交
502 503
}

504 505 506 507 508 509 510 511
static struct page *dequeue_huge_page_node(struct hstate *h, int nid)
{
	struct page *page;

	if (list_empty(&h->hugepage_freelists[nid]))
		return NULL;
	page = list_entry(h->hugepage_freelists[nid].next, struct page, lru);
	list_del(&page->lru);
512
	set_page_refcounted(page);
513 514 515 516 517
	h->free_huge_pages--;
	h->free_huge_pages_node[nid]--;
	return page;
}

518 519
static struct page *dequeue_huge_page_vma(struct hstate *h,
				struct vm_area_struct *vma,
520
				unsigned long address, int avoid_reserve)
L
Linus Torvalds 已提交
521 522
{
	struct page *page = NULL;
523
	struct mempolicy *mpol;
524
	nodemask_t *nodemask;
525
	struct zonelist *zonelist;
526 527
	struct zone *zone;
	struct zoneref *z;
L
Linus Torvalds 已提交
528

529 530 531
	get_mems_allowed();
	zonelist = huge_zonelist(vma, address,
					htlb_alloc_mask, &mpol, &nodemask);
532 533 534 535 536
	/*
	 * A child process with MAP_PRIVATE mappings created by their parent
	 * have no page reserves. This check ensures that reservations are
	 * not "stolen". The child may still get SIGKILLed
	 */
537
	if (!vma_has_reserves(vma) &&
538
			h->free_huge_pages - h->resv_huge_pages == 0)
539
		goto err;
540

541
	/* If reserves cannot be used, ensure enough pages are in the pool */
542
	if (avoid_reserve && h->free_huge_pages - h->resv_huge_pages == 0)
543
		goto err;;
544

545 546
	for_each_zone_zonelist_nodemask(zone, z, zonelist,
						MAX_NR_ZONES - 1, nodemask) {
547 548 549 550 551 552 553
		if (cpuset_zone_allowed_softwall(zone, htlb_alloc_mask)) {
			page = dequeue_huge_page_node(h, zone_to_nid(zone));
			if (page) {
				if (!avoid_reserve)
					decrement_hugepage_resv_vma(h, vma);
				break;
			}
A
Andrew Morton 已提交
554
		}
L
Linus Torvalds 已提交
555
	}
556
err:
557
	mpol_cond_put(mpol);
558
	put_mems_allowed();
L
Linus Torvalds 已提交
559 560 561
	return page;
}

562
static void update_and_free_page(struct hstate *h, struct page *page)
A
Adam Litke 已提交
563 564
{
	int i;
565

566 567
	VM_BUG_ON(h->order >= MAX_ORDER);

568 569 570
	h->nr_huge_pages--;
	h->nr_huge_pages_node[page_to_nid(page)]--;
	for (i = 0; i < pages_per_huge_page(h); i++) {
A
Adam Litke 已提交
571 572 573 574 575 576
		page[i].flags &= ~(1 << PG_locked | 1 << PG_error | 1 << PG_referenced |
				1 << PG_dirty | 1 << PG_active | 1 << PG_reserved |
				1 << PG_private | 1<< PG_writeback);
	}
	set_compound_page_dtor(page, NULL);
	set_page_refcounted(page);
577
	arch_release_hugepage(page);
578
	__free_pages(page, huge_page_order(h));
A
Adam Litke 已提交
579 580
}

581 582 583 584 585 586 587 588 589 590 591
struct hstate *size_to_hstate(unsigned long size)
{
	struct hstate *h;

	for_each_hstate(h) {
		if (huge_page_size(h) == size)
			return h;
	}
	return NULL;
}

592 593
static void free_huge_page(struct page *page)
{
594 595 596 597
	/*
	 * Can't pass hstate in here because it is called from the
	 * compound page destructor.
	 */
598
	struct hstate *h = page_hstate(page);
599
	int nid = page_to_nid(page);
600
	struct address_space *mapping;
601

602
	mapping = (struct address_space *) page_private(page);
603
	set_page_private(page, 0);
604
	page->mapping = NULL;
605
	BUG_ON(page_count(page));
606
	BUG_ON(page_mapcount(page));
607 608 609
	INIT_LIST_HEAD(&page->lru);

	spin_lock(&hugetlb_lock);
610
	if (h->surplus_huge_pages_node[nid] && huge_page_order(h) < MAX_ORDER) {
611 612 613
		update_and_free_page(h, page);
		h->surplus_huge_pages--;
		h->surplus_huge_pages_node[nid]--;
614
	} else {
615
		enqueue_huge_page(h, page);
616
	}
617
	spin_unlock(&hugetlb_lock);
618
	if (mapping)
619
		hugetlb_put_quota(mapping, 1);
620 621
}

622
static void prep_new_huge_page(struct hstate *h, struct page *page, int nid)
623 624 625
{
	set_compound_page_dtor(page, free_huge_page);
	spin_lock(&hugetlb_lock);
626 627
	h->nr_huge_pages++;
	h->nr_huge_pages_node[nid]++;
628 629 630 631
	spin_unlock(&hugetlb_lock);
	put_page(page); /* free it into the hugepage allocator */
}

632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659
static void prep_compound_gigantic_page(struct page *page, unsigned long order)
{
	int i;
	int nr_pages = 1 << order;
	struct page *p = page + 1;

	/* we rely on prep_new_huge_page to set the destructor */
	set_compound_order(page, order);
	__SetPageHead(page);
	for (i = 1; i < nr_pages; i++, p = mem_map_next(p, page, i)) {
		__SetPageTail(p);
		p->first_page = page;
	}
}

int PageHuge(struct page *page)
{
	compound_page_dtor *dtor;

	if (!PageCompound(page))
		return 0;

	page = compound_head(page);
	dtor = get_compound_page_dtor(page);

	return dtor == free_huge_page;
}

660 661
EXPORT_SYMBOL_GPL(PageHuge);

662
static struct page *alloc_fresh_huge_page_node(struct hstate *h, int nid)
L
Linus Torvalds 已提交
663 664
{
	struct page *page;
665

666 667 668
	if (h->order >= MAX_ORDER)
		return NULL;

669
	page = alloc_pages_exact_node(nid,
670 671
		htlb_alloc_mask|__GFP_COMP|__GFP_THISNODE|
						__GFP_REPEAT|__GFP_NOWARN,
672
		huge_page_order(h));
L
Linus Torvalds 已提交
673
	if (page) {
674
		if (arch_prepare_hugepage(page)) {
675
			__free_pages(page, huge_page_order(h));
676
			return NULL;
677
		}
678
		prep_new_huge_page(h, page, nid);
L
Linus Torvalds 已提交
679
	}
680 681 682 683

	return page;
}

684
/*
685 686 687 688 689
 * common helper functions for hstate_next_node_to_{alloc|free}.
 * We may have allocated or freed a huge page based on a different
 * nodes_allowed previously, so h->next_node_to_{alloc|free} might
 * be outside of *nodes_allowed.  Ensure that we use an allowed
 * node for alloc or free.
690
 */
691
static int next_node_allowed(int nid, nodemask_t *nodes_allowed)
692
{
693
	nid = next_node(nid, *nodes_allowed);
694
	if (nid == MAX_NUMNODES)
695
		nid = first_node(*nodes_allowed);
696 697 698 699 700
	VM_BUG_ON(nid >= MAX_NUMNODES);

	return nid;
}

701 702 703 704 705 706 707
static int get_valid_node_allowed(int nid, nodemask_t *nodes_allowed)
{
	if (!node_isset(nid, *nodes_allowed))
		nid = next_node_allowed(nid, nodes_allowed);
	return nid;
}

708
/*
709 710 711 712
 * returns the previously saved node ["this node"] from which to
 * allocate a persistent huge page for the pool and advance the
 * next node from which to allocate, handling wrap at end of node
 * mask.
713
 */
714 715
static int hstate_next_node_to_alloc(struct hstate *h,
					nodemask_t *nodes_allowed)
716
{
717 718 719 720 721 722
	int nid;

	VM_BUG_ON(!nodes_allowed);

	nid = get_valid_node_allowed(h->next_nid_to_alloc, nodes_allowed);
	h->next_nid_to_alloc = next_node_allowed(nid, nodes_allowed);
723 724

	return nid;
725 726
}

727
static int alloc_fresh_huge_page(struct hstate *h, nodemask_t *nodes_allowed)
728 729 730 731 732 733
{
	struct page *page;
	int start_nid;
	int next_nid;
	int ret = 0;

734
	start_nid = hstate_next_node_to_alloc(h, nodes_allowed);
735
	next_nid = start_nid;
736 737

	do {
738
		page = alloc_fresh_huge_page_node(h, next_nid);
739
		if (page) {
740
			ret = 1;
741 742
			break;
		}
743
		next_nid = hstate_next_node_to_alloc(h, nodes_allowed);
744
	} while (next_nid != start_nid);
745

746 747 748 749 750
	if (ret)
		count_vm_event(HTLB_BUDDY_PGALLOC);
	else
		count_vm_event(HTLB_BUDDY_PGALLOC_FAIL);

751
	return ret;
L
Linus Torvalds 已提交
752 753
}

754
/*
755 756 757 758
 * helper for free_pool_huge_page() - return the previously saved
 * node ["this node"] from which to free a huge page.  Advance the
 * next node id whether or not we find a free huge page to free so
 * that the next attempt to free addresses the next node.
759
 */
760
static int hstate_next_node_to_free(struct hstate *h, nodemask_t *nodes_allowed)
761
{
762 763 764 765 766 767
	int nid;

	VM_BUG_ON(!nodes_allowed);

	nid = get_valid_node_allowed(h->next_nid_to_free, nodes_allowed);
	h->next_nid_to_free = next_node_allowed(nid, nodes_allowed);
768 769

	return nid;
770 771 772 773 774 775 776 777
}

/*
 * Free huge page from pool from next node to free.
 * Attempt to keep persistent huge pages more or less
 * balanced over allowed nodes.
 * Called with hugetlb_lock locked.
 */
778 779
static int free_pool_huge_page(struct hstate *h, nodemask_t *nodes_allowed,
							 bool acct_surplus)
780 781 782 783 784
{
	int start_nid;
	int next_nid;
	int ret = 0;

785
	start_nid = hstate_next_node_to_free(h, nodes_allowed);
786 787 788
	next_nid = start_nid;

	do {
789 790 791 792 793 794
		/*
		 * If we're returning unused surplus pages, only examine
		 * nodes with surplus pages.
		 */
		if ((!acct_surplus || h->surplus_huge_pages_node[next_nid]) &&
		    !list_empty(&h->hugepage_freelists[next_nid])) {
795 796 797 798 799 800
			struct page *page =
				list_entry(h->hugepage_freelists[next_nid].next,
					  struct page, lru);
			list_del(&page->lru);
			h->free_huge_pages--;
			h->free_huge_pages_node[next_nid]--;
801 802 803 804
			if (acct_surplus) {
				h->surplus_huge_pages--;
				h->surplus_huge_pages_node[next_nid]--;
			}
805 806
			update_and_free_page(h, page);
			ret = 1;
807
			break;
808
		}
809
		next_nid = hstate_next_node_to_free(h, nodes_allowed);
810
	} while (next_nid != start_nid);
811 812 813 814

	return ret;
}

815
static struct page *alloc_buddy_huge_page(struct hstate *h, int nid)
816 817
{
	struct page *page;
818
	unsigned int r_nid;
819

820 821 822
	if (h->order >= MAX_ORDER)
		return NULL;

823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846
	/*
	 * Assume we will successfully allocate the surplus page to
	 * prevent racing processes from causing the surplus to exceed
	 * overcommit
	 *
	 * This however introduces a different race, where a process B
	 * tries to grow the static hugepage pool while alloc_pages() is
	 * called by process A. B will only examine the per-node
	 * counters in determining if surplus huge pages can be
	 * converted to normal huge pages in adjust_pool_surplus(). A
	 * won't be able to increment the per-node counter, until the
	 * lock is dropped by B, but B doesn't drop hugetlb_lock until
	 * no more huge pages can be converted from surplus to normal
	 * state (and doesn't try to convert again). Thus, we have a
	 * case where a surplus huge page exists, the pool is grown, and
	 * the surplus huge page still exists after, even though it
	 * should just have been converted to a normal huge page. This
	 * does not leak memory, though, as the hugepage will be freed
	 * once it is out of use. It also does not allow the counters to
	 * go out of whack in adjust_pool_surplus() as we don't modify
	 * the node values until we've gotten the hugepage and only the
	 * per-node value is checked there.
	 */
	spin_lock(&hugetlb_lock);
847
	if (h->surplus_huge_pages >= h->nr_overcommit_huge_pages) {
848 849 850
		spin_unlock(&hugetlb_lock);
		return NULL;
	} else {
851 852
		h->nr_huge_pages++;
		h->surplus_huge_pages++;
853 854 855
	}
	spin_unlock(&hugetlb_lock);

856 857 858 859 860 861 862 863
	if (nid == NUMA_NO_NODE)
		page = alloc_pages(htlb_alloc_mask|__GFP_COMP|
				   __GFP_REPEAT|__GFP_NOWARN,
				   huge_page_order(h));
	else
		page = alloc_pages_exact_node(nid,
			htlb_alloc_mask|__GFP_COMP|__GFP_THISNODE|
			__GFP_REPEAT|__GFP_NOWARN, huge_page_order(h));
864

865 866 867 868 869
	if (page && arch_prepare_hugepage(page)) {
		__free_pages(page, huge_page_order(h));
		return NULL;
	}

870
	spin_lock(&hugetlb_lock);
871
	if (page) {
872
		r_nid = page_to_nid(page);
873
		set_compound_page_dtor(page, free_huge_page);
874 875 876
		/*
		 * We incremented the global counters already
		 */
877 878
		h->nr_huge_pages_node[r_nid]++;
		h->surplus_huge_pages_node[r_nid]++;
879
		__count_vm_event(HTLB_BUDDY_PGALLOC);
880
	} else {
881 882
		h->nr_huge_pages--;
		h->surplus_huge_pages--;
883
		__count_vm_event(HTLB_BUDDY_PGALLOC_FAIL);
884
	}
885
	spin_unlock(&hugetlb_lock);
886 887 888 889

	return page;
}

890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908
/*
 * This allocation function is useful in the context where vma is irrelevant.
 * E.g. soft-offlining uses this function because it only cares physical
 * address of error page.
 */
struct page *alloc_huge_page_node(struct hstate *h, int nid)
{
	struct page *page;

	spin_lock(&hugetlb_lock);
	page = dequeue_huge_page_node(h, nid);
	spin_unlock(&hugetlb_lock);

	if (!page)
		page = alloc_buddy_huge_page(h, nid);

	return page;
}

909 910 911 912
/*
 * Increase the hugetlb pool such that it can accomodate a reservation
 * of size 'delta'.
 */
913
static int gather_surplus_pages(struct hstate *h, int delta)
914 915 916 917 918 919
{
	struct list_head surplus_list;
	struct page *page, *tmp;
	int ret, i;
	int needed, allocated;

920
	needed = (h->resv_huge_pages + delta) - h->free_huge_pages;
921
	if (needed <= 0) {
922
		h->resv_huge_pages += delta;
923
		return 0;
924
	}
925 926 927 928 929 930 931 932

	allocated = 0;
	INIT_LIST_HEAD(&surplus_list);

	ret = -ENOMEM;
retry:
	spin_unlock(&hugetlb_lock);
	for (i = 0; i < needed; i++) {
933
		page = alloc_buddy_huge_page(h, NUMA_NO_NODE);
934
		if (!page)
935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950
			/*
			 * We were not able to allocate enough pages to
			 * satisfy the entire reservation so we free what
			 * we've allocated so far.
			 */
			goto free;

		list_add(&page->lru, &surplus_list);
	}
	allocated += needed;

	/*
	 * After retaking hugetlb_lock, we need to recalculate 'needed'
	 * because either resv_huge_pages or free_huge_pages may have changed.
	 */
	spin_lock(&hugetlb_lock);
951 952
	needed = (h->resv_huge_pages + delta) -
			(h->free_huge_pages + allocated);
953 954 955 956 957 958 959
	if (needed > 0)
		goto retry;

	/*
	 * The surplus_list now contains _at_least_ the number of extra pages
	 * needed to accomodate the reservation.  Add the appropriate number
	 * of pages to the hugetlb pool and free the extras back to the buddy
960 961 962
	 * allocator.  Commit the entire reservation here to prevent another
	 * process from stealing the pages as they are added to the pool but
	 * before they are reserved.
963 964
	 */
	needed += allocated;
965
	h->resv_huge_pages += delta;
966
	ret = 0;
967 968

	spin_unlock(&hugetlb_lock);
969
	/* Free the needed pages to the hugetlb pool */
970
	list_for_each_entry_safe(page, tmp, &surplus_list, lru) {
971 972
		if ((--needed) < 0)
			break;
973
		list_del(&page->lru);
974 975 976 977 978 979
		/*
		 * This page is now managed by the hugetlb allocator and has
		 * no users -- drop the buddy allocator's reference.
		 */
		put_page_testzero(page);
		VM_BUG_ON(page_count(page));
980
		enqueue_huge_page(h, page);
981 982 983
	}

	/* Free unnecessary surplus pages to the buddy allocator */
984
free:
985 986 987
	if (!list_empty(&surplus_list)) {
		list_for_each_entry_safe(page, tmp, &surplus_list, lru) {
			list_del(&page->lru);
988
			put_page(page);
989
		}
990
	}
991
	spin_lock(&hugetlb_lock);
992 993 994 995 996 997 998 999

	return ret;
}

/*
 * When releasing a hugetlb pool reservation, any surplus pages that were
 * allocated to satisfy the reservation must be explicitly freed if they were
 * never used.
1000
 * Called with hugetlb_lock held.
1001
 */
1002 1003
static void return_unused_surplus_pages(struct hstate *h,
					unsigned long unused_resv_pages)
1004 1005 1006
{
	unsigned long nr_pages;

1007
	/* Uncommit the reservation */
1008
	h->resv_huge_pages -= unused_resv_pages;
1009

1010 1011 1012 1013
	/* Cannot return gigantic pages currently */
	if (h->order >= MAX_ORDER)
		return;

1014
	nr_pages = min(unused_resv_pages, h->surplus_huge_pages);
1015

1016 1017
	/*
	 * We want to release as many surplus pages as possible, spread
1018 1019 1020 1021 1022
	 * evenly across all nodes with memory. Iterate across these nodes
	 * until we can no longer free unreserved surplus pages. This occurs
	 * when the nodes with surplus pages have no free pages.
	 * free_pool_huge_page() will balance the the freed pages across the
	 * on-line nodes with memory and will handle the hstate accounting.
1023 1024
	 */
	while (nr_pages--) {
1025
		if (!free_pool_huge_page(h, &node_states[N_HIGH_MEMORY], 1))
1026
			break;
1027 1028 1029
	}
}

1030 1031 1032 1033 1034 1035 1036 1037 1038
/*
 * Determine if the huge page at addr within the vma has an associated
 * reservation.  Where it does not we will need to logically increase
 * reservation and actually increase quota before an allocation can occur.
 * Where any new reservation would be required the reservation change is
 * prepared, but not committed.  Once the page has been quota'd allocated
 * an instantiated the change should be committed via vma_commit_reservation.
 * No action is required on failure.
 */
1039
static long vma_needs_reservation(struct hstate *h,
1040
			struct vm_area_struct *vma, unsigned long addr)
1041 1042 1043 1044
{
	struct address_space *mapping = vma->vm_file->f_mapping;
	struct inode *inode = mapping->host;

1045
	if (vma->vm_flags & VM_MAYSHARE) {
1046
		pgoff_t idx = vma_hugecache_offset(h, vma, addr);
1047 1048 1049
		return region_chg(&inode->i_mapping->private_list,
							idx, idx + 1);

1050 1051
	} else if (!is_vma_resv_set(vma, HPAGE_RESV_OWNER)) {
		return 1;
1052

1053
	} else  {
1054
		long err;
1055
		pgoff_t idx = vma_hugecache_offset(h, vma, addr);
1056 1057 1058 1059 1060 1061 1062
		struct resv_map *reservations = vma_resv_map(vma);

		err = region_chg(&reservations->regions, idx, idx + 1);
		if (err < 0)
			return err;
		return 0;
	}
1063
}
1064 1065
static void vma_commit_reservation(struct hstate *h,
			struct vm_area_struct *vma, unsigned long addr)
1066 1067 1068 1069
{
	struct address_space *mapping = vma->vm_file->f_mapping;
	struct inode *inode = mapping->host;

1070
	if (vma->vm_flags & VM_MAYSHARE) {
1071
		pgoff_t idx = vma_hugecache_offset(h, vma, addr);
1072
		region_add(&inode->i_mapping->private_list, idx, idx + 1);
1073 1074

	} else if (is_vma_resv_set(vma, HPAGE_RESV_OWNER)) {
1075
		pgoff_t idx = vma_hugecache_offset(h, vma, addr);
1076 1077 1078 1079
		struct resv_map *reservations = vma_resv_map(vma);

		/* Mark this page used in the map. */
		region_add(&reservations->regions, idx, idx + 1);
1080 1081 1082
	}
}

1083
static struct page *alloc_huge_page(struct vm_area_struct *vma,
1084
				    unsigned long addr, int avoid_reserve)
L
Linus Torvalds 已提交
1085
{
1086
	struct hstate *h = hstate_vma(vma);
1087
	struct page *page;
1088 1089
	struct address_space *mapping = vma->vm_file->f_mapping;
	struct inode *inode = mapping->host;
1090
	long chg;
1091 1092 1093 1094 1095

	/*
	 * Processes that did not create the mapping will have no reserves and
	 * will not have accounted against quota. Check that the quota can be
	 * made before satisfying the allocation
1096 1097
	 * MAP_NORESERVE mappings may also need pages and quota allocated
	 * if no reserve mapping overlaps.
1098
	 */
1099
	chg = vma_needs_reservation(h, vma, addr);
1100 1101 1102
	if (chg < 0)
		return ERR_PTR(chg);
	if (chg)
1103 1104
		if (hugetlb_get_quota(inode->i_mapping, chg))
			return ERR_PTR(-ENOSPC);
L
Linus Torvalds 已提交
1105 1106

	spin_lock(&hugetlb_lock);
1107
	page = dequeue_huge_page_vma(h, vma, addr, avoid_reserve);
L
Linus Torvalds 已提交
1108
	spin_unlock(&hugetlb_lock);
1109

K
Ken Chen 已提交
1110
	if (!page) {
1111
		page = alloc_buddy_huge_page(h, NUMA_NO_NODE);
K
Ken Chen 已提交
1112
		if (!page) {
1113
			hugetlb_put_quota(inode->i_mapping, chg);
1114
			return ERR_PTR(-VM_FAULT_SIGBUS);
K
Ken Chen 已提交
1115 1116
		}
	}
1117

1118
	set_page_private(page, (unsigned long) mapping);
1119

1120
	vma_commit_reservation(h, vma, addr);
1121

1122
	return page;
1123 1124
}

1125
int __weak alloc_bootmem_huge_page(struct hstate *h)
1126 1127
{
	struct huge_bootmem_page *m;
1128
	int nr_nodes = nodes_weight(node_states[N_HIGH_MEMORY]);
1129 1130 1131 1132 1133

	while (nr_nodes) {
		void *addr;

		addr = __alloc_bootmem_node_nopanic(
1134
				NODE_DATA(hstate_next_node_to_alloc(h,
1135
						&node_states[N_HIGH_MEMORY])),
1136 1137 1138 1139 1140 1141 1142 1143 1144
				huge_page_size(h), huge_page_size(h), 0);

		if (addr) {
			/*
			 * Use the beginning of the huge page to store the
			 * huge_bootmem_page struct (until gather_bootmem
			 * puts them into the mem_map).
			 */
			m = addr;
1145
			goto found;
1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158
		}
		nr_nodes--;
	}
	return 0;

found:
	BUG_ON((unsigned long)virt_to_phys(m) & (huge_page_size(h) - 1));
	/* Put them into a private list first because mem_map is not up yet */
	list_add(&m->list, &huge_boot_pages);
	m->hstate = h;
	return 1;
}

1159 1160 1161 1162 1163 1164 1165 1166
static void prep_compound_huge_page(struct page *page, int order)
{
	if (unlikely(order > (MAX_ORDER - 1)))
		prep_compound_gigantic_page(page, order);
	else
		prep_compound_page(page, order);
}

1167 1168 1169 1170 1171 1172 1173 1174 1175 1176
/* Put bootmem huge pages into the standard lists after mem_map is up */
static void __init gather_bootmem_prealloc(void)
{
	struct huge_bootmem_page *m;

	list_for_each_entry(m, &huge_boot_pages, list) {
		struct page *page = virt_to_page(m);
		struct hstate *h = m->hstate;
		__ClearPageReserved(page);
		WARN_ON(page_count(page) != 1);
1177
		prep_compound_huge_page(page, h->order);
1178 1179 1180 1181
		prep_new_huge_page(h, page, page_to_nid(page));
	}
}

1182
static void __init hugetlb_hstate_alloc_pages(struct hstate *h)
L
Linus Torvalds 已提交
1183 1184
{
	unsigned long i;
1185

1186
	for (i = 0; i < h->max_huge_pages; ++i) {
1187 1188 1189
		if (h->order >= MAX_ORDER) {
			if (!alloc_bootmem_huge_page(h))
				break;
1190 1191
		} else if (!alloc_fresh_huge_page(h,
					 &node_states[N_HIGH_MEMORY]))
L
Linus Torvalds 已提交
1192 1193
			break;
	}
1194
	h->max_huge_pages = i;
1195 1196 1197 1198 1199 1200 1201
}

static void __init hugetlb_init_hstates(void)
{
	struct hstate *h;

	for_each_hstate(h) {
1202 1203 1204
		/* oversize hugepages were init'ed in early boot */
		if (h->order < MAX_ORDER)
			hugetlb_hstate_alloc_pages(h);
1205 1206 1207
	}
}

A
Andi Kleen 已提交
1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218
static char * __init memfmt(char *buf, unsigned long n)
{
	if (n >= (1UL << 30))
		sprintf(buf, "%lu GB", n >> 30);
	else if (n >= (1UL << 20))
		sprintf(buf, "%lu MB", n >> 20);
	else
		sprintf(buf, "%lu KB", n >> 10);
	return buf;
}

1219 1220 1221 1222 1223
static void __init report_hugepages(void)
{
	struct hstate *h;

	for_each_hstate(h) {
A
Andi Kleen 已提交
1224 1225 1226 1227 1228
		char buf[32];
		printk(KERN_INFO "HugeTLB registered %s page size, "
				 "pre-allocated %ld pages\n",
			memfmt(buf, huge_page_size(h)),
			h->free_huge_pages);
1229 1230 1231
	}
}

L
Linus Torvalds 已提交
1232
#ifdef CONFIG_HIGHMEM
1233 1234
static void try_to_free_low(struct hstate *h, unsigned long count,
						nodemask_t *nodes_allowed)
L
Linus Torvalds 已提交
1235
{
1236 1237
	int i;

1238 1239 1240
	if (h->order >= MAX_ORDER)
		return;

1241
	for_each_node_mask(i, *nodes_allowed) {
L
Linus Torvalds 已提交
1242
		struct page *page, *next;
1243 1244 1245
		struct list_head *freel = &h->hugepage_freelists[i];
		list_for_each_entry_safe(page, next, freel, lru) {
			if (count >= h->nr_huge_pages)
1246
				return;
L
Linus Torvalds 已提交
1247 1248 1249
			if (PageHighMem(page))
				continue;
			list_del(&page->lru);
1250
			update_and_free_page(h, page);
1251 1252
			h->free_huge_pages--;
			h->free_huge_pages_node[page_to_nid(page)]--;
L
Linus Torvalds 已提交
1253 1254 1255 1256
		}
	}
}
#else
1257 1258
static inline void try_to_free_low(struct hstate *h, unsigned long count,
						nodemask_t *nodes_allowed)
L
Linus Torvalds 已提交
1259 1260 1261 1262
{
}
#endif

1263 1264 1265 1266 1267
/*
 * Increment or decrement surplus_huge_pages.  Keep node-specific counters
 * balanced by operating on them in a round-robin fashion.
 * Returns 1 if an adjustment was made.
 */
1268 1269
static int adjust_pool_surplus(struct hstate *h, nodemask_t *nodes_allowed,
				int delta)
1270
{
1271
	int start_nid, next_nid;
1272 1273 1274 1275
	int ret = 0;

	VM_BUG_ON(delta != -1 && delta != 1);

1276
	if (delta < 0)
1277
		start_nid = hstate_next_node_to_alloc(h, nodes_allowed);
1278
	else
1279
		start_nid = hstate_next_node_to_free(h, nodes_allowed);
1280 1281 1282 1283 1284 1285 1286 1287
	next_nid = start_nid;

	do {
		int nid = next_nid;
		if (delta < 0)  {
			/*
			 * To shrink on this node, there must be a surplus page
			 */
1288
			if (!h->surplus_huge_pages_node[nid]) {
1289 1290
				next_nid = hstate_next_node_to_alloc(h,
								nodes_allowed);
1291
				continue;
1292
			}
1293 1294 1295 1296 1297 1298
		}
		if (delta > 0) {
			/*
			 * Surplus cannot exceed the total number of pages
			 */
			if (h->surplus_huge_pages_node[nid] >=
1299
						h->nr_huge_pages_node[nid]) {
1300 1301
				next_nid = hstate_next_node_to_free(h,
								nodes_allowed);
1302
				continue;
1303
			}
1304
		}
1305 1306 1307 1308 1309

		h->surplus_huge_pages += delta;
		h->surplus_huge_pages_node[nid] += delta;
		ret = 1;
		break;
1310
	} while (next_nid != start_nid);
1311 1312 1313 1314

	return ret;
}

1315
#define persistent_huge_pages(h) (h->nr_huge_pages - h->surplus_huge_pages)
1316 1317
static unsigned long set_max_huge_pages(struct hstate *h, unsigned long count,
						nodemask_t *nodes_allowed)
L
Linus Torvalds 已提交
1318
{
1319
	unsigned long min_count, ret;
L
Linus Torvalds 已提交
1320

1321 1322 1323
	if (h->order >= MAX_ORDER)
		return h->max_huge_pages;

1324 1325 1326 1327
	/*
	 * Increase the pool size
	 * First take pages out of surplus state.  Then make up the
	 * remaining difference by allocating fresh huge pages.
1328 1329 1330 1331 1332 1333
	 *
	 * We might race with alloc_buddy_huge_page() here and be unable
	 * to convert a surplus huge page to a normal huge page. That is
	 * not critical, though, it just means the overall size of the
	 * pool might be one hugepage larger than it needs to be, but
	 * within all the constraints specified by the sysctls.
1334
	 */
L
Linus Torvalds 已提交
1335
	spin_lock(&hugetlb_lock);
1336
	while (h->surplus_huge_pages && count > persistent_huge_pages(h)) {
1337
		if (!adjust_pool_surplus(h, nodes_allowed, -1))
1338 1339 1340
			break;
	}

1341
	while (count > persistent_huge_pages(h)) {
1342 1343 1344 1345 1346 1347
		/*
		 * If this allocation races such that we no longer need the
		 * page, free_huge_page will handle it by freeing the page
		 * and reducing the surplus.
		 */
		spin_unlock(&hugetlb_lock);
1348
		ret = alloc_fresh_huge_page(h, nodes_allowed);
1349 1350 1351 1352
		spin_lock(&hugetlb_lock);
		if (!ret)
			goto out;

1353 1354 1355
		/* Bail for signals. Probably ctrl-c from user */
		if (signal_pending(current))
			goto out;
1356 1357 1358 1359 1360 1361 1362 1363
	}

	/*
	 * Decrease the pool size
	 * First return free pages to the buddy allocator (being careful
	 * to keep enough around to satisfy reservations).  Then place
	 * pages into surplus state as needed so the pool will shrink
	 * to the desired size as pages become free.
1364 1365 1366 1367 1368 1369 1370 1371
	 *
	 * By placing pages into the surplus state independent of the
	 * overcommit value, we are allowing the surplus pool size to
	 * exceed overcommit. There are few sane options here. Since
	 * alloc_buddy_huge_page() is checking the global counter,
	 * though, we'll note that we're not allowed to exceed surplus
	 * and won't grow the pool anywhere else. Not until one of the
	 * sysctls are changed, or the surplus pages go out of use.
1372
	 */
1373
	min_count = h->resv_huge_pages + h->nr_huge_pages - h->free_huge_pages;
1374
	min_count = max(count, min_count);
1375
	try_to_free_low(h, min_count, nodes_allowed);
1376
	while (min_count < persistent_huge_pages(h)) {
1377
		if (!free_pool_huge_page(h, nodes_allowed, 0))
L
Linus Torvalds 已提交
1378 1379
			break;
	}
1380
	while (count < persistent_huge_pages(h)) {
1381
		if (!adjust_pool_surplus(h, nodes_allowed, 1))
1382 1383 1384
			break;
	}
out:
1385
	ret = persistent_huge_pages(h);
L
Linus Torvalds 已提交
1386
	spin_unlock(&hugetlb_lock);
1387
	return ret;
L
Linus Torvalds 已提交
1388 1389
}

1390 1391 1392 1393 1394 1395 1396 1397 1398 1399
#define HSTATE_ATTR_RO(_name) \
	static struct kobj_attribute _name##_attr = __ATTR_RO(_name)

#define HSTATE_ATTR(_name) \
	static struct kobj_attribute _name##_attr = \
		__ATTR(_name, 0644, _name##_show, _name##_store)

static struct kobject *hugepages_kobj;
static struct kobject *hstate_kobjs[HUGE_MAX_HSTATE];

1400 1401 1402
static struct hstate *kobj_to_node_hstate(struct kobject *kobj, int *nidp);

static struct hstate *kobj_to_hstate(struct kobject *kobj, int *nidp)
1403 1404
{
	int i;
1405

1406
	for (i = 0; i < HUGE_MAX_HSTATE; i++)
1407 1408 1409
		if (hstate_kobjs[i] == kobj) {
			if (nidp)
				*nidp = NUMA_NO_NODE;
1410
			return &hstates[i];
1411 1412 1413
		}

	return kobj_to_node_hstate(kobj, nidp);
1414 1415
}

1416
static ssize_t nr_hugepages_show_common(struct kobject *kobj,
1417 1418
					struct kobj_attribute *attr, char *buf)
{
1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429
	struct hstate *h;
	unsigned long nr_huge_pages;
	int nid;

	h = kobj_to_hstate(kobj, &nid);
	if (nid == NUMA_NO_NODE)
		nr_huge_pages = h->nr_huge_pages;
	else
		nr_huge_pages = h->nr_huge_pages_node[nid];

	return sprintf(buf, "%lu\n", nr_huge_pages);
1430
}
1431 1432 1433
static ssize_t nr_hugepages_store_common(bool obey_mempolicy,
			struct kobject *kobj, struct kobj_attribute *attr,
			const char *buf, size_t len)
1434 1435
{
	int err;
1436
	int nid;
1437
	unsigned long count;
1438
	struct hstate *h;
1439
	NODEMASK_ALLOC(nodemask_t, nodes_allowed, GFP_KERNEL | __GFP_NORETRY);
1440

1441
	err = strict_strtoul(buf, 10, &count);
1442 1443 1444
	if (err)
		return 0;

1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464
	h = kobj_to_hstate(kobj, &nid);
	if (nid == NUMA_NO_NODE) {
		/*
		 * global hstate attribute
		 */
		if (!(obey_mempolicy &&
				init_nodemask_of_mempolicy(nodes_allowed))) {
			NODEMASK_FREE(nodes_allowed);
			nodes_allowed = &node_states[N_HIGH_MEMORY];
		}
	} else if (nodes_allowed) {
		/*
		 * per node hstate attribute: adjust count to global,
		 * but restrict alloc/free to the specified node.
		 */
		count += h->nr_huge_pages - h->nr_huge_pages_node[nid];
		init_nodemask_of_node(nodes_allowed, nid);
	} else
		nodes_allowed = &node_states[N_HIGH_MEMORY];

1465
	h->max_huge_pages = set_max_huge_pages(h, count, nodes_allowed);
1466

1467
	if (nodes_allowed != &node_states[N_HIGH_MEMORY])
1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482
		NODEMASK_FREE(nodes_allowed);

	return len;
}

static ssize_t nr_hugepages_show(struct kobject *kobj,
				       struct kobj_attribute *attr, char *buf)
{
	return nr_hugepages_show_common(kobj, attr, buf);
}

static ssize_t nr_hugepages_store(struct kobject *kobj,
	       struct kobj_attribute *attr, const char *buf, size_t len)
{
	return nr_hugepages_store_common(false, kobj, attr, buf, len);
1483 1484 1485
}
HSTATE_ATTR(nr_hugepages);

1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506
#ifdef CONFIG_NUMA

/*
 * hstate attribute for optionally mempolicy-based constraint on persistent
 * huge page alloc/free.
 */
static ssize_t nr_hugepages_mempolicy_show(struct kobject *kobj,
				       struct kobj_attribute *attr, char *buf)
{
	return nr_hugepages_show_common(kobj, attr, buf);
}

static ssize_t nr_hugepages_mempolicy_store(struct kobject *kobj,
	       struct kobj_attribute *attr, const char *buf, size_t len)
{
	return nr_hugepages_store_common(true, kobj, attr, buf, len);
}
HSTATE_ATTR(nr_hugepages_mempolicy);
#endif


1507 1508 1509
static ssize_t nr_overcommit_hugepages_show(struct kobject *kobj,
					struct kobj_attribute *attr, char *buf)
{
1510
	struct hstate *h = kobj_to_hstate(kobj, NULL);
1511 1512 1513 1514 1515 1516 1517
	return sprintf(buf, "%lu\n", h->nr_overcommit_huge_pages);
}
static ssize_t nr_overcommit_hugepages_store(struct kobject *kobj,
		struct kobj_attribute *attr, const char *buf, size_t count)
{
	int err;
	unsigned long input;
1518
	struct hstate *h = kobj_to_hstate(kobj, NULL);
1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534

	err = strict_strtoul(buf, 10, &input);
	if (err)
		return 0;

	spin_lock(&hugetlb_lock);
	h->nr_overcommit_huge_pages = input;
	spin_unlock(&hugetlb_lock);

	return count;
}
HSTATE_ATTR(nr_overcommit_hugepages);

static ssize_t free_hugepages_show(struct kobject *kobj,
					struct kobj_attribute *attr, char *buf)
{
1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545
	struct hstate *h;
	unsigned long free_huge_pages;
	int nid;

	h = kobj_to_hstate(kobj, &nid);
	if (nid == NUMA_NO_NODE)
		free_huge_pages = h->free_huge_pages;
	else
		free_huge_pages = h->free_huge_pages_node[nid];

	return sprintf(buf, "%lu\n", free_huge_pages);
1546 1547 1548 1549 1550 1551
}
HSTATE_ATTR_RO(free_hugepages);

static ssize_t resv_hugepages_show(struct kobject *kobj,
					struct kobj_attribute *attr, char *buf)
{
1552
	struct hstate *h = kobj_to_hstate(kobj, NULL);
1553 1554 1555 1556 1557 1558 1559
	return sprintf(buf, "%lu\n", h->resv_huge_pages);
}
HSTATE_ATTR_RO(resv_hugepages);

static ssize_t surplus_hugepages_show(struct kobject *kobj,
					struct kobj_attribute *attr, char *buf)
{
1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570
	struct hstate *h;
	unsigned long surplus_huge_pages;
	int nid;

	h = kobj_to_hstate(kobj, &nid);
	if (nid == NUMA_NO_NODE)
		surplus_huge_pages = h->surplus_huge_pages;
	else
		surplus_huge_pages = h->surplus_huge_pages_node[nid];

	return sprintf(buf, "%lu\n", surplus_huge_pages);
1571 1572 1573 1574 1575 1576 1577 1578 1579
}
HSTATE_ATTR_RO(surplus_hugepages);

static struct attribute *hstate_attrs[] = {
	&nr_hugepages_attr.attr,
	&nr_overcommit_hugepages_attr.attr,
	&free_hugepages_attr.attr,
	&resv_hugepages_attr.attr,
	&surplus_hugepages_attr.attr,
1580 1581 1582
#ifdef CONFIG_NUMA
	&nr_hugepages_mempolicy_attr.attr,
#endif
1583 1584 1585 1586 1587 1588 1589
	NULL,
};

static struct attribute_group hstate_attr_group = {
	.attrs = hstate_attrs,
};

J
Jeff Mahoney 已提交
1590 1591 1592
static int hugetlb_sysfs_add_hstate(struct hstate *h, struct kobject *parent,
				    struct kobject **hstate_kobjs,
				    struct attribute_group *hstate_attr_group)
1593 1594
{
	int retval;
1595
	int hi = h - hstates;
1596

1597 1598
	hstate_kobjs[hi] = kobject_create_and_add(h->name, parent);
	if (!hstate_kobjs[hi])
1599 1600
		return -ENOMEM;

1601
	retval = sysfs_create_group(hstate_kobjs[hi], hstate_attr_group);
1602
	if (retval)
1603
		kobject_put(hstate_kobjs[hi]);
1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617

	return retval;
}

static void __init hugetlb_sysfs_init(void)
{
	struct hstate *h;
	int err;

	hugepages_kobj = kobject_create_and_add("hugepages", mm_kobj);
	if (!hugepages_kobj)
		return;

	for_each_hstate(h) {
1618 1619
		err = hugetlb_sysfs_add_hstate(h, hugepages_kobj,
					 hstate_kobjs, &hstate_attr_group);
1620 1621 1622 1623 1624 1625
		if (err)
			printk(KERN_ERR "Hugetlb: Unable to add hstate %s",
								h->name);
	}
}

1626 1627 1628 1629 1630 1631 1632 1633 1634 1635 1636 1637 1638 1639 1640 1641 1642 1643 1644 1645 1646 1647 1648 1649 1650 1651 1652 1653 1654 1655 1656 1657 1658 1659 1660 1661 1662 1663 1664 1665 1666 1667 1668 1669 1670 1671 1672 1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684 1685 1686 1687
#ifdef CONFIG_NUMA

/*
 * node_hstate/s - associate per node hstate attributes, via their kobjects,
 * with node sysdevs in node_devices[] using a parallel array.  The array
 * index of a node sysdev or _hstate == node id.
 * This is here to avoid any static dependency of the node sysdev driver, in
 * the base kernel, on the hugetlb module.
 */
struct node_hstate {
	struct kobject		*hugepages_kobj;
	struct kobject		*hstate_kobjs[HUGE_MAX_HSTATE];
};
struct node_hstate node_hstates[MAX_NUMNODES];

/*
 * A subset of global hstate attributes for node sysdevs
 */
static struct attribute *per_node_hstate_attrs[] = {
	&nr_hugepages_attr.attr,
	&free_hugepages_attr.attr,
	&surplus_hugepages_attr.attr,
	NULL,
};

static struct attribute_group per_node_hstate_attr_group = {
	.attrs = per_node_hstate_attrs,
};

/*
 * kobj_to_node_hstate - lookup global hstate for node sysdev hstate attr kobj.
 * Returns node id via non-NULL nidp.
 */
static struct hstate *kobj_to_node_hstate(struct kobject *kobj, int *nidp)
{
	int nid;

	for (nid = 0; nid < nr_node_ids; nid++) {
		struct node_hstate *nhs = &node_hstates[nid];
		int i;
		for (i = 0; i < HUGE_MAX_HSTATE; i++)
			if (nhs->hstate_kobjs[i] == kobj) {
				if (nidp)
					*nidp = nid;
				return &hstates[i];
			}
	}

	BUG();
	return NULL;
}

/*
 * Unregister hstate attributes from a single node sysdev.
 * No-op if no hstate attributes attached.
 */
void hugetlb_unregister_node(struct node *node)
{
	struct hstate *h;
	struct node_hstate *nhs = &node_hstates[node->sysdev.id];

	if (!nhs->hugepages_kobj)
1688
		return;		/* no hstate attributes */
1689 1690 1691 1692 1693 1694 1695 1696 1697 1698 1699 1700 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724 1725 1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746 1747 1748 1749 1750 1751 1752

	for_each_hstate(h)
		if (nhs->hstate_kobjs[h - hstates]) {
			kobject_put(nhs->hstate_kobjs[h - hstates]);
			nhs->hstate_kobjs[h - hstates] = NULL;
		}

	kobject_put(nhs->hugepages_kobj);
	nhs->hugepages_kobj = NULL;
}

/*
 * hugetlb module exit:  unregister hstate attributes from node sysdevs
 * that have them.
 */
static void hugetlb_unregister_all_nodes(void)
{
	int nid;

	/*
	 * disable node sysdev registrations.
	 */
	register_hugetlbfs_with_node(NULL, NULL);

	/*
	 * remove hstate attributes from any nodes that have them.
	 */
	for (nid = 0; nid < nr_node_ids; nid++)
		hugetlb_unregister_node(&node_devices[nid]);
}

/*
 * Register hstate attributes for a single node sysdev.
 * No-op if attributes already registered.
 */
void hugetlb_register_node(struct node *node)
{
	struct hstate *h;
	struct node_hstate *nhs = &node_hstates[node->sysdev.id];
	int err;

	if (nhs->hugepages_kobj)
		return;		/* already allocated */

	nhs->hugepages_kobj = kobject_create_and_add("hugepages",
							&node->sysdev.kobj);
	if (!nhs->hugepages_kobj)
		return;

	for_each_hstate(h) {
		err = hugetlb_sysfs_add_hstate(h, nhs->hugepages_kobj,
						nhs->hstate_kobjs,
						&per_node_hstate_attr_group);
		if (err) {
			printk(KERN_ERR "Hugetlb: Unable to add hstate %s"
					" for node %d\n",
						h->name, node->sysdev.id);
			hugetlb_unregister_node(node);
			break;
		}
	}
}

/*
1753 1754 1755
 * hugetlb init time:  register hstate attributes for all registered node
 * sysdevs of nodes that have memory.  All on-line nodes should have
 * registered their associated sysdev by this time.
1756 1757 1758 1759 1760
 */
static void hugetlb_register_all_nodes(void)
{
	int nid;

1761
	for_each_node_state(nid, N_HIGH_MEMORY) {
1762 1763 1764 1765 1766 1767 1768 1769 1770 1771 1772 1773 1774 1775 1776 1777 1778 1779 1780 1781 1782 1783 1784 1785 1786 1787 1788 1789
		struct node *node = &node_devices[nid];
		if (node->sysdev.id == nid)
			hugetlb_register_node(node);
	}

	/*
	 * Let the node sysdev driver know we're here so it can
	 * [un]register hstate attributes on node hotplug.
	 */
	register_hugetlbfs_with_node(hugetlb_register_node,
				     hugetlb_unregister_node);
}
#else	/* !CONFIG_NUMA */

static struct hstate *kobj_to_node_hstate(struct kobject *kobj, int *nidp)
{
	BUG();
	if (nidp)
		*nidp = -1;
	return NULL;
}

static void hugetlb_unregister_all_nodes(void) { }

static void hugetlb_register_all_nodes(void) { }

#endif

1790 1791 1792 1793
static void __exit hugetlb_exit(void)
{
	struct hstate *h;

1794 1795
	hugetlb_unregister_all_nodes();

1796 1797 1798 1799 1800 1801 1802 1803 1804 1805
	for_each_hstate(h) {
		kobject_put(hstate_kobjs[h - hstates]);
	}

	kobject_put(hugepages_kobj);
}
module_exit(hugetlb_exit);

static int __init hugetlb_init(void)
{
1806 1807 1808 1809 1810 1811
	/* Some platform decide whether they support huge pages at boot
	 * time. On these, such as powerpc, HPAGE_SHIFT is set to 0 when
	 * there is no such support
	 */
	if (HPAGE_SHIFT == 0)
		return 0;
1812

1813 1814 1815 1816
	if (!size_to_hstate(default_hstate_size)) {
		default_hstate_size = HPAGE_SIZE;
		if (!size_to_hstate(default_hstate_size))
			hugetlb_add_hstate(HUGETLB_PAGE_ORDER);
1817
	}
1818 1819 1820
	default_hstate_idx = size_to_hstate(default_hstate_size) - hstates;
	if (default_hstate_max_huge_pages)
		default_hstate.max_huge_pages = default_hstate_max_huge_pages;
1821 1822 1823

	hugetlb_init_hstates();

1824 1825
	gather_bootmem_prealloc();

1826 1827 1828 1829
	report_hugepages();

	hugetlb_sysfs_init();

1830 1831
	hugetlb_register_all_nodes();

1832 1833 1834 1835 1836 1837 1838 1839
	return 0;
}
module_init(hugetlb_init);

/* Should be called on processing a hugepagesz=... option */
void __init hugetlb_add_hstate(unsigned order)
{
	struct hstate *h;
1840 1841
	unsigned long i;

1842 1843 1844 1845 1846 1847 1848 1849 1850
	if (size_to_hstate(PAGE_SIZE << order)) {
		printk(KERN_WARNING "hugepagesz= specified twice, ignoring\n");
		return;
	}
	BUG_ON(max_hstate >= HUGE_MAX_HSTATE);
	BUG_ON(order == 0);
	h = &hstates[max_hstate++];
	h->order = order;
	h->mask = ~((1ULL << (order + PAGE_SHIFT)) - 1);
1851 1852 1853 1854
	h->nr_huge_pages = 0;
	h->free_huge_pages = 0;
	for (i = 0; i < MAX_NUMNODES; ++i)
		INIT_LIST_HEAD(&h->hugepage_freelists[i]);
1855 1856
	h->next_nid_to_alloc = first_node(node_states[N_HIGH_MEMORY]);
	h->next_nid_to_free = first_node(node_states[N_HIGH_MEMORY]);
1857 1858
	snprintf(h->name, HSTATE_NAME_LEN, "hugepages-%lukB",
					huge_page_size(h)/1024);
1859

1860 1861 1862
	parsed_hstate = h;
}

1863
static int __init hugetlb_nrpages_setup(char *s)
1864 1865
{
	unsigned long *mhp;
1866
	static unsigned long *last_mhp;
1867 1868 1869 1870 1871 1872 1873 1874 1875 1876

	/*
	 * !max_hstate means we haven't parsed a hugepagesz= parameter yet,
	 * so this hugepages= parameter goes to the "default hstate".
	 */
	if (!max_hstate)
		mhp = &default_hstate_max_huge_pages;
	else
		mhp = &parsed_hstate->max_huge_pages;

1877 1878 1879 1880 1881 1882
	if (mhp == last_mhp) {
		printk(KERN_WARNING "hugepages= specified twice without "
			"interleaving hugepagesz=, ignoring\n");
		return 1;
	}

1883 1884 1885
	if (sscanf(s, "%lu", mhp) <= 0)
		*mhp = 0;

1886 1887 1888 1889 1890 1891 1892 1893 1894 1895
	/*
	 * Global state is always initialized later in hugetlb_init.
	 * But we need to allocate >= MAX_ORDER hstates here early to still
	 * use the bootmem allocator.
	 */
	if (max_hstate && parsed_hstate->order >= MAX_ORDER)
		hugetlb_hstate_alloc_pages(parsed_hstate);

	last_mhp = mhp;

1896 1897
	return 1;
}
1898 1899 1900 1901 1902 1903 1904 1905
__setup("hugepages=", hugetlb_nrpages_setup);

static int __init hugetlb_default_setup(char *s)
{
	default_hstate_size = memparse(s, &s);
	return 1;
}
__setup("default_hugepagesz=", hugetlb_default_setup);
1906

1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918
static unsigned int cpuset_mems_nr(unsigned int *array)
{
	int node;
	unsigned int nr = 0;

	for_each_node_mask(node, cpuset_current_mems_allowed)
		nr += array[node];

	return nr;
}

#ifdef CONFIG_SYSCTL
1919 1920 1921
static int hugetlb_sysctl_handler_common(bool obey_mempolicy,
			 struct ctl_table *table, int write,
			 void __user *buffer, size_t *length, loff_t *ppos)
L
Linus Torvalds 已提交
1922
{
1923 1924 1925 1926 1927 1928 1929 1930
	struct hstate *h = &default_hstate;
	unsigned long tmp;

	if (!write)
		tmp = h->max_huge_pages;

	table->data = &tmp;
	table->maxlen = sizeof(unsigned long);
1931
	proc_doulongvec_minmax(table, write, buffer, length, ppos);
1932

1933
	if (write) {
1934 1935
		NODEMASK_ALLOC(nodemask_t, nodes_allowed,
						GFP_KERNEL | __GFP_NORETRY);
1936 1937 1938 1939 1940 1941 1942 1943 1944 1945
		if (!(obey_mempolicy &&
			       init_nodemask_of_mempolicy(nodes_allowed))) {
			NODEMASK_FREE(nodes_allowed);
			nodes_allowed = &node_states[N_HIGH_MEMORY];
		}
		h->max_huge_pages = set_max_huge_pages(h, tmp, nodes_allowed);

		if (nodes_allowed != &node_states[N_HIGH_MEMORY])
			NODEMASK_FREE(nodes_allowed);
	}
1946

L
Linus Torvalds 已提交
1947 1948
	return 0;
}
1949

1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 1966
int hugetlb_sysctl_handler(struct ctl_table *table, int write,
			  void __user *buffer, size_t *length, loff_t *ppos)
{

	return hugetlb_sysctl_handler_common(false, table, write,
							buffer, length, ppos);
}

#ifdef CONFIG_NUMA
int hugetlb_mempolicy_sysctl_handler(struct ctl_table *table, int write,
			  void __user *buffer, size_t *length, loff_t *ppos)
{
	return hugetlb_sysctl_handler_common(true, table, write,
							buffer, length, ppos);
}
#endif /* CONFIG_NUMA */

1967
int hugetlb_treat_movable_handler(struct ctl_table *table, int write,
1968
			void __user *buffer,
1969 1970
			size_t *length, loff_t *ppos)
{
1971
	proc_dointvec(table, write, buffer, length, ppos);
1972 1973 1974 1975 1976 1977 1978
	if (hugepages_treat_as_movable)
		htlb_alloc_mask = GFP_HIGHUSER_MOVABLE;
	else
		htlb_alloc_mask = GFP_HIGHUSER;
	return 0;
}

1979
int hugetlb_overcommit_handler(struct ctl_table *table, int write,
1980
			void __user *buffer,
1981 1982
			size_t *length, loff_t *ppos)
{
1983
	struct hstate *h = &default_hstate;
1984 1985 1986 1987 1988 1989 1990
	unsigned long tmp;

	if (!write)
		tmp = h->nr_overcommit_huge_pages;

	table->data = &tmp;
	table->maxlen = sizeof(unsigned long);
1991
	proc_doulongvec_minmax(table, write, buffer, length, ppos);
1992 1993 1994 1995 1996 1997 1998

	if (write) {
		spin_lock(&hugetlb_lock);
		h->nr_overcommit_huge_pages = tmp;
		spin_unlock(&hugetlb_lock);
	}

1999 2000 2001
	return 0;
}

L
Linus Torvalds 已提交
2002 2003
#endif /* CONFIG_SYSCTL */

2004
void hugetlb_report_meminfo(struct seq_file *m)
L
Linus Torvalds 已提交
2005
{
2006
	struct hstate *h = &default_hstate;
2007
	seq_printf(m,
2008 2009 2010 2011 2012
			"HugePages_Total:   %5lu\n"
			"HugePages_Free:    %5lu\n"
			"HugePages_Rsvd:    %5lu\n"
			"HugePages_Surp:    %5lu\n"
			"Hugepagesize:   %8lu kB\n",
2013 2014 2015 2016 2017
			h->nr_huge_pages,
			h->free_huge_pages,
			h->resv_huge_pages,
			h->surplus_huge_pages,
			1UL << (huge_page_order(h) + PAGE_SHIFT - 10));
L
Linus Torvalds 已提交
2018 2019 2020 2021
}

int hugetlb_report_node_meminfo(int nid, char *buf)
{
2022
	struct hstate *h = &default_hstate;
L
Linus Torvalds 已提交
2023 2024
	return sprintf(buf,
		"Node %d HugePages_Total: %5u\n"
2025 2026
		"Node %d HugePages_Free:  %5u\n"
		"Node %d HugePages_Surp:  %5u\n",
2027 2028 2029
		nid, h->nr_huge_pages_node[nid],
		nid, h->free_huge_pages_node[nid],
		nid, h->surplus_huge_pages_node[nid]);
L
Linus Torvalds 已提交
2030 2031 2032 2033 2034
}

/* Return the number pages of memory we physically have, in PAGE_SIZE units. */
unsigned long hugetlb_total_pages(void)
{
2035 2036
	struct hstate *h = &default_hstate;
	return h->nr_huge_pages * pages_per_huge_page(h);
L
Linus Torvalds 已提交
2037 2038
}

2039
static int hugetlb_acct_memory(struct hstate *h, long delta)
M
Mel Gorman 已提交
2040 2041 2042 2043 2044 2045 2046 2047 2048 2049 2050 2051 2052 2053 2054 2055 2056 2057 2058 2059 2060 2061
{
	int ret = -ENOMEM;

	spin_lock(&hugetlb_lock);
	/*
	 * When cpuset is configured, it breaks the strict hugetlb page
	 * reservation as the accounting is done on a global variable. Such
	 * reservation is completely rubbish in the presence of cpuset because
	 * the reservation is not checked against page availability for the
	 * current cpuset. Application can still potentially OOM'ed by kernel
	 * with lack of free htlb page in cpuset that the task is in.
	 * Attempt to enforce strict accounting with cpuset is almost
	 * impossible (or too ugly) because cpuset is too fluid that
	 * task or memory node can be dynamically moved between cpusets.
	 *
	 * The change of semantics for shared hugetlb mapping with cpuset is
	 * undesirable. However, in order to preserve some of the semantics,
	 * we fall back to check against current free page availability as
	 * a best attempt and hopefully to minimize the impact of changing
	 * semantics that cpuset has.
	 */
	if (delta > 0) {
2062
		if (gather_surplus_pages(h, delta) < 0)
M
Mel Gorman 已提交
2063 2064
			goto out;

2065 2066
		if (delta > cpuset_mems_nr(h->free_huge_pages_node)) {
			return_unused_surplus_pages(h, delta);
M
Mel Gorman 已提交
2067 2068 2069 2070 2071 2072
			goto out;
		}
	}

	ret = 0;
	if (delta < 0)
2073
		return_unused_surplus_pages(h, (unsigned long) -delta);
M
Mel Gorman 已提交
2074 2075 2076 2077 2078 2079

out:
	spin_unlock(&hugetlb_lock);
	return ret;
}

2080 2081 2082 2083 2084 2085 2086 2087 2088 2089 2090 2091 2092 2093 2094 2095
static void hugetlb_vm_op_open(struct vm_area_struct *vma)
{
	struct resv_map *reservations = vma_resv_map(vma);

	/*
	 * This new VMA should share its siblings reservation map if present.
	 * The VMA will only ever have a valid reservation map pointer where
	 * it is being copied for another still existing VMA.  As that VMA
	 * has a reference to the reservation map it cannot dissappear until
	 * after this open call completes.  It is therefore safe to take a
	 * new reference here without additional locking.
	 */
	if (reservations)
		kref_get(&reservations->refs);
}

2096 2097
static void hugetlb_vm_op_close(struct vm_area_struct *vma)
{
2098
	struct hstate *h = hstate_vma(vma);
2099 2100 2101 2102 2103 2104
	struct resv_map *reservations = vma_resv_map(vma);
	unsigned long reserve;
	unsigned long start;
	unsigned long end;

	if (reservations) {
2105 2106
		start = vma_hugecache_offset(h, vma, vma->vm_start);
		end = vma_hugecache_offset(h, vma, vma->vm_end);
2107 2108 2109 2110 2111 2112

		reserve = (end - start) -
			region_count(&reservations->regions, start, end);

		kref_put(&reservations->refs, resv_map_release);

2113
		if (reserve) {
2114
			hugetlb_acct_memory(h, -reserve);
2115 2116
			hugetlb_put_quota(vma->vm_file->f_mapping, reserve);
		}
2117
	}
2118 2119
}

L
Linus Torvalds 已提交
2120 2121 2122 2123 2124 2125
/*
 * We cannot handle pagefaults against hugetlb pages at all.  They cause
 * handle_mm_fault() to try to instantiate regular-sized pages in the
 * hugegpage VMA.  do_page_fault() is supposed to trap this, so BUG is we get
 * this far.
 */
N
Nick Piggin 已提交
2126
static int hugetlb_vm_op_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
L
Linus Torvalds 已提交
2127 2128
{
	BUG();
N
Nick Piggin 已提交
2129
	return 0;
L
Linus Torvalds 已提交
2130 2131
}

2132
const struct vm_operations_struct hugetlb_vm_ops = {
N
Nick Piggin 已提交
2133
	.fault = hugetlb_vm_op_fault,
2134
	.open = hugetlb_vm_op_open,
2135
	.close = hugetlb_vm_op_close,
L
Linus Torvalds 已提交
2136 2137
};

2138 2139
static pte_t make_huge_pte(struct vm_area_struct *vma, struct page *page,
				int writable)
D
David Gibson 已提交
2140 2141 2142
{
	pte_t entry;

2143
	if (writable) {
D
David Gibson 已提交
2144 2145 2146
		entry =
		    pte_mkwrite(pte_mkdirty(mk_pte(page, vma->vm_page_prot)));
	} else {
2147
		entry = huge_pte_wrprotect(mk_pte(page, vma->vm_page_prot));
D
David Gibson 已提交
2148 2149 2150 2151 2152 2153 2154
	}
	entry = pte_mkyoung(entry);
	entry = pte_mkhuge(entry);

	return entry;
}

2155 2156 2157 2158 2159
static void set_huge_ptep_writable(struct vm_area_struct *vma,
				   unsigned long address, pte_t *ptep)
{
	pte_t entry;

2160 2161
	entry = pte_mkwrite(pte_mkdirty(huge_ptep_get(ptep)));
	if (huge_ptep_set_access_flags(vma, address, ptep, entry, 1)) {
2162
		update_mmu_cache(vma, address, ptep);
2163
	}
2164 2165 2166
}


D
David Gibson 已提交
2167 2168 2169 2170 2171
int copy_hugetlb_page_range(struct mm_struct *dst, struct mm_struct *src,
			    struct vm_area_struct *vma)
{
	pte_t *src_pte, *dst_pte, entry;
	struct page *ptepage;
2172
	unsigned long addr;
2173
	int cow;
2174 2175
	struct hstate *h = hstate_vma(vma);
	unsigned long sz = huge_page_size(h);
2176 2177

	cow = (vma->vm_flags & (VM_SHARED | VM_MAYWRITE)) == VM_MAYWRITE;
D
David Gibson 已提交
2178

2179
	for (addr = vma->vm_start; addr < vma->vm_end; addr += sz) {
H
Hugh Dickins 已提交
2180 2181 2182
		src_pte = huge_pte_offset(src, addr);
		if (!src_pte)
			continue;
2183
		dst_pte = huge_pte_alloc(dst, addr, sz);
D
David Gibson 已提交
2184 2185
		if (!dst_pte)
			goto nomem;
2186 2187 2188 2189 2190

		/* If the pagetables are shared don't copy or take references */
		if (dst_pte == src_pte)
			continue;

H
Hugh Dickins 已提交
2191
		spin_lock(&dst->page_table_lock);
N
Nick Piggin 已提交
2192
		spin_lock_nested(&src->page_table_lock, SINGLE_DEPTH_NESTING);
2193
		if (!huge_pte_none(huge_ptep_get(src_pte))) {
2194
			if (cow)
2195 2196
				huge_ptep_set_wrprotect(src, addr, src_pte);
			entry = huge_ptep_get(src_pte);
2197 2198
			ptepage = pte_page(entry);
			get_page(ptepage);
2199
			page_dup_rmap(ptepage);
2200 2201 2202
			set_huge_pte_at(dst, addr, dst_pte, entry);
		}
		spin_unlock(&src->page_table_lock);
H
Hugh Dickins 已提交
2203
		spin_unlock(&dst->page_table_lock);
D
David Gibson 已提交
2204 2205 2206 2207 2208 2209 2210
	}
	return 0;

nomem:
	return -ENOMEM;
}

N
Naoya Horiguchi 已提交
2211 2212 2213 2214 2215 2216 2217 2218 2219 2220 2221 2222 2223
static int is_hugetlb_entry_migration(pte_t pte)
{
	swp_entry_t swp;

	if (huge_pte_none(pte) || pte_present(pte))
		return 0;
	swp = pte_to_swp_entry(pte);
	if (non_swap_entry(swp) && is_migration_entry(swp)) {
		return 1;
	} else
		return 0;
}

2224 2225 2226 2227 2228 2229 2230 2231 2232 2233 2234 2235 2236
static int is_hugetlb_entry_hwpoisoned(pte_t pte)
{
	swp_entry_t swp;

	if (huge_pte_none(pte) || pte_present(pte))
		return 0;
	swp = pte_to_swp_entry(pte);
	if (non_swap_entry(swp) && is_hwpoison_entry(swp)) {
		return 1;
	} else
		return 0;
}

2237
void __unmap_hugepage_range(struct vm_area_struct *vma, unsigned long start,
2238
			    unsigned long end, struct page *ref_page)
D
David Gibson 已提交
2239 2240 2241
{
	struct mm_struct *mm = vma->vm_mm;
	unsigned long address;
2242
	pte_t *ptep;
D
David Gibson 已提交
2243 2244
	pte_t pte;
	struct page *page;
2245
	struct page *tmp;
2246 2247 2248
	struct hstate *h = hstate_vma(vma);
	unsigned long sz = huge_page_size(h);

2249 2250 2251 2252 2253
	/*
	 * A page gathering list, protected by per file i_mmap_lock. The
	 * lock is used to avoid list corruption from multiple unmapping
	 * of the same page since we are using page->lru.
	 */
2254
	LIST_HEAD(page_list);
D
David Gibson 已提交
2255 2256

	WARN_ON(!is_vm_hugetlb_page(vma));
2257 2258
	BUG_ON(start & ~huge_page_mask(h));
	BUG_ON(end & ~huge_page_mask(h));
D
David Gibson 已提交
2259

A
Andrea Arcangeli 已提交
2260
	mmu_notifier_invalidate_range_start(mm, start, end);
2261
	spin_lock(&mm->page_table_lock);
2262
	for (address = start; address < end; address += sz) {
2263
		ptep = huge_pte_offset(mm, address);
A
Adam Litke 已提交
2264
		if (!ptep)
2265 2266
			continue;

2267 2268 2269
		if (huge_pmd_unshare(mm, &address, ptep))
			continue;

2270 2271 2272 2273 2274 2275 2276 2277 2278 2279 2280 2281 2282 2283 2284 2285 2286 2287 2288 2289 2290
		/*
		 * If a reference page is supplied, it is because a specific
		 * page is being unmapped, not a range. Ensure the page we
		 * are about to unmap is the actual page of interest.
		 */
		if (ref_page) {
			pte = huge_ptep_get(ptep);
			if (huge_pte_none(pte))
				continue;
			page = pte_page(pte);
			if (page != ref_page)
				continue;

			/*
			 * Mark the VMA as having unmapped its page so that
			 * future faults in this VMA will fail rather than
			 * looking like data was lost
			 */
			set_vma_resv_flags(vma, HPAGE_RESV_UNMAPPED);
		}

2291
		pte = huge_ptep_get_and_clear(mm, address, ptep);
2292
		if (huge_pte_none(pte))
D
David Gibson 已提交
2293
			continue;
2294

2295 2296 2297 2298 2299 2300
		/*
		 * HWPoisoned hugepage is already unmapped and dropped reference
		 */
		if (unlikely(is_hugetlb_entry_hwpoisoned(pte)))
			continue;

D
David Gibson 已提交
2301
		page = pte_page(pte);
2302 2303
		if (pte_dirty(pte))
			set_page_dirty(page);
2304
		list_add(&page->lru, &page_list);
D
David Gibson 已提交
2305
	}
L
Linus Torvalds 已提交
2306
	spin_unlock(&mm->page_table_lock);
2307
	flush_tlb_range(vma, start, end);
A
Andrea Arcangeli 已提交
2308
	mmu_notifier_invalidate_range_end(mm, start, end);
2309
	list_for_each_entry_safe(page, tmp, &page_list, lru) {
2310
		page_remove_rmap(page);
2311 2312 2313
		list_del(&page->lru);
		put_page(page);
	}
L
Linus Torvalds 已提交
2314
}
D
David Gibson 已提交
2315

2316
void unmap_hugepage_range(struct vm_area_struct *vma, unsigned long start,
2317
			  unsigned long end, struct page *ref_page)
2318
{
2319 2320 2321
	spin_lock(&vma->vm_file->f_mapping->i_mmap_lock);
	__unmap_hugepage_range(vma, start, end, ref_page);
	spin_unlock(&vma->vm_file->f_mapping->i_mmap_lock);
2322 2323
}

2324 2325 2326 2327 2328 2329
/*
 * This is called when the original mapper is failing to COW a MAP_PRIVATE
 * mappping it owns the reserve page for. The intention is to unmap the page
 * from other VMAs and let the children be SIGKILLed if they are faulting the
 * same region.
 */
2330 2331
static int unmap_ref_private(struct mm_struct *mm, struct vm_area_struct *vma,
				struct page *page, unsigned long address)
2332
{
2333
	struct hstate *h = hstate_vma(vma);
2334 2335 2336 2337 2338 2339 2340 2341 2342
	struct vm_area_struct *iter_vma;
	struct address_space *mapping;
	struct prio_tree_iter iter;
	pgoff_t pgoff;

	/*
	 * vm_pgoff is in PAGE_SIZE units, hence the different calculation
	 * from page cache lookup which is in HPAGE_SIZE units.
	 */
2343
	address = address & huge_page_mask(h);
2344 2345 2346 2347
	pgoff = ((address - vma->vm_start) >> PAGE_SHIFT)
		+ (vma->vm_pgoff >> PAGE_SHIFT);
	mapping = (struct address_space *)page_private(page);

2348 2349 2350 2351 2352 2353
	/*
	 * Take the mapping lock for the duration of the table walk. As
	 * this mapping should be shared between all the VMAs,
	 * __unmap_hugepage_range() is called as the lock is already held
	 */
	spin_lock(&mapping->i_mmap_lock);
2354 2355 2356 2357 2358 2359 2360 2361 2362 2363 2364 2365 2366
	vma_prio_tree_foreach(iter_vma, &iter, &mapping->i_mmap, pgoff, pgoff) {
		/* Do not unmap the current VMA */
		if (iter_vma == vma)
			continue;

		/*
		 * Unmap the page from other VMAs without their own reserves.
		 * They get marked to be SIGKILLed if they fault in these
		 * areas. This is because a future no-page fault on this VMA
		 * could insert a zeroed page instead of the data existing
		 * from the time of fork. This would look like data corruption
		 */
		if (!is_vma_resv_set(iter_vma, HPAGE_RESV_OWNER))
2367
			__unmap_hugepage_range(iter_vma,
2368
				address, address + huge_page_size(h),
2369 2370
				page);
	}
2371
	spin_unlock(&mapping->i_mmap_lock);
2372 2373 2374 2375

	return 1;
}

2376 2377 2378
/*
 * Hugetlb_cow() should be called with page lock of the original hugepage held.
 */
2379
static int hugetlb_cow(struct mm_struct *mm, struct vm_area_struct *vma,
2380 2381
			unsigned long address, pte_t *ptep, pte_t pte,
			struct page *pagecache_page)
2382
{
2383
	struct hstate *h = hstate_vma(vma);
2384
	struct page *old_page, *new_page;
2385
	int avoidcopy;
2386
	int outside_reserve = 0;
2387 2388 2389

	old_page = pte_page(pte);

2390
retry_avoidcopy:
2391 2392
	/* If no-one else is actually using this page, avoid the copy
	 * and just make the page writable */
2393
	avoidcopy = (page_mapcount(old_page) == 1);
2394
	if (avoidcopy) {
2395 2396
		if (PageAnon(old_page))
			page_move_anon_rmap(old_page, vma, address);
2397
		set_huge_ptep_writable(vma, address, ptep);
N
Nick Piggin 已提交
2398
		return 0;
2399 2400
	}

2401 2402 2403 2404 2405 2406 2407 2408 2409
	/*
	 * If the process that created a MAP_PRIVATE mapping is about to
	 * perform a COW due to a shared page count, attempt to satisfy
	 * the allocation without using the existing reserves. The pagecache
	 * page is used to determine if the reserve at this address was
	 * consumed or not. If reserves were used, a partial faulted mapping
	 * at the time of fork() could consume its reserves on COW instead
	 * of the full address range.
	 */
2410
	if (!(vma->vm_flags & VM_MAYSHARE) &&
2411 2412 2413 2414
			is_vma_resv_set(vma, HPAGE_RESV_OWNER) &&
			old_page != pagecache_page)
		outside_reserve = 1;

2415
	page_cache_get(old_page);
2416 2417 2418

	/* Drop page_table_lock as buddy allocator may be called */
	spin_unlock(&mm->page_table_lock);
2419
	new_page = alloc_huge_page(vma, address, outside_reserve);
2420

2421
	if (IS_ERR(new_page)) {
2422
		page_cache_release(old_page);
2423 2424 2425 2426 2427 2428 2429 2430 2431 2432 2433 2434 2435

		/*
		 * If a process owning a MAP_PRIVATE mapping fails to COW,
		 * it is due to references held by a child and an insufficient
		 * huge page pool. To guarantee the original mappers
		 * reliability, unmap the page from child processes. The child
		 * may get SIGKILLed if it later faults.
		 */
		if (outside_reserve) {
			BUG_ON(huge_pte_none(pte));
			if (unmap_ref_private(mm, vma, old_page, address)) {
				BUG_ON(page_count(old_page) != 1);
				BUG_ON(huge_pte_none(pte));
2436
				spin_lock(&mm->page_table_lock);
2437 2438 2439 2440 2441
				goto retry_avoidcopy;
			}
			WARN_ON_ONCE(1);
		}

2442 2443
		/* Caller expects lock to be held */
		spin_lock(&mm->page_table_lock);
2444
		return -PTR_ERR(new_page);
2445 2446
	}

2447 2448 2449 2450 2451 2452 2453
	/*
	 * When the original hugepage is shared one, it does not have
	 * anon_vma prepared.
	 */
	if (unlikely(anon_vma_prepare(vma)))
		return VM_FAULT_OOM;

2454
	copy_user_huge_page(new_page, old_page, address, vma);
N
Nick Piggin 已提交
2455
	__SetPageUptodate(new_page);
2456

2457 2458 2459 2460 2461
	/*
	 * Retake the page_table_lock to check for racing updates
	 * before the page tables are altered
	 */
	spin_lock(&mm->page_table_lock);
2462
	ptep = huge_pte_offset(mm, address & huge_page_mask(h));
2463
	if (likely(pte_same(huge_ptep_get(ptep), pte))) {
2464
		/* Break COW */
2465 2466 2467
		mmu_notifier_invalidate_range_start(mm,
			address & huge_page_mask(h),
			(address & huge_page_mask(h)) + huge_page_size(h));
2468
		huge_ptep_clear_flush(vma, address, ptep);
2469 2470
		set_huge_pte_at(mm, address, ptep,
				make_huge_pte(vma, new_page, 1));
2471
		page_remove_rmap(old_page);
2472
		hugepage_add_new_anon_rmap(new_page, vma, address);
2473 2474
		/* Make the old page be freed below */
		new_page = old_page;
2475 2476 2477
		mmu_notifier_invalidate_range_end(mm,
			address & huge_page_mask(h),
			(address & huge_page_mask(h)) + huge_page_size(h));
2478 2479 2480
	}
	page_cache_release(new_page);
	page_cache_release(old_page);
N
Nick Piggin 已提交
2481
	return 0;
2482 2483
}

2484
/* Return the pagecache page at a given address within a VMA */
2485 2486
static struct page *hugetlbfs_pagecache_page(struct hstate *h,
			struct vm_area_struct *vma, unsigned long address)
2487 2488
{
	struct address_space *mapping;
2489
	pgoff_t idx;
2490 2491

	mapping = vma->vm_file->f_mapping;
2492
	idx = vma_hugecache_offset(h, vma, address);
2493 2494 2495 2496

	return find_lock_page(mapping, idx);
}

H
Hugh Dickins 已提交
2497 2498 2499 2500 2501
/*
 * Return whether there is a pagecache page to back given address within VMA.
 * Caller follow_hugetlb_page() holds page_table_lock so we cannot lock_page.
 */
static bool hugetlbfs_pagecache_present(struct hstate *h,
H
Hugh Dickins 已提交
2502 2503 2504 2505 2506 2507 2508 2509 2510 2511 2512 2513 2514 2515 2516
			struct vm_area_struct *vma, unsigned long address)
{
	struct address_space *mapping;
	pgoff_t idx;
	struct page *page;

	mapping = vma->vm_file->f_mapping;
	idx = vma_hugecache_offset(h, vma, address);

	page = find_get_page(mapping, idx);
	if (page)
		put_page(page);
	return page != NULL;
}

2517
static int hugetlb_no_page(struct mm_struct *mm, struct vm_area_struct *vma,
2518
			unsigned long address, pte_t *ptep, unsigned int flags)
2519
{
2520
	struct hstate *h = hstate_vma(vma);
2521
	int ret = VM_FAULT_SIGBUS;
2522
	pgoff_t idx;
A
Adam Litke 已提交
2523 2524 2525
	unsigned long size;
	struct page *page;
	struct address_space *mapping;
2526
	pte_t new_pte;
A
Adam Litke 已提交
2527

2528 2529 2530 2531 2532 2533 2534 2535 2536 2537 2538 2539
	/*
	 * Currently, we are forced to kill the process in the event the
	 * original mapper has unmapped pages from the child due to a failed
	 * COW. Warn that such a situation has occured as it may not be obvious
	 */
	if (is_vma_resv_set(vma, HPAGE_RESV_UNMAPPED)) {
		printk(KERN_WARNING
			"PID %d killed due to inadequate hugepage pool\n",
			current->pid);
		return ret;
	}

A
Adam Litke 已提交
2540
	mapping = vma->vm_file->f_mapping;
2541
	idx = vma_hugecache_offset(h, vma, address);
A
Adam Litke 已提交
2542 2543 2544 2545 2546

	/*
	 * Use page lock to guard against racing truncation
	 * before we get page_table_lock.
	 */
2547 2548 2549
retry:
	page = find_lock_page(mapping, idx);
	if (!page) {
2550
		size = i_size_read(mapping->host) >> huge_page_shift(h);
2551 2552
		if (idx >= size)
			goto out;
2553
		page = alloc_huge_page(vma, address, 0);
2554 2555
		if (IS_ERR(page)) {
			ret = -PTR_ERR(page);
2556 2557
			goto out;
		}
2558
		clear_huge_page(page, address, huge_page_size(h));
N
Nick Piggin 已提交
2559
		__SetPageUptodate(page);
2560

2561
		if (vma->vm_flags & VM_MAYSHARE) {
2562
			int err;
K
Ken Chen 已提交
2563
			struct inode *inode = mapping->host;
2564 2565 2566 2567 2568 2569 2570 2571

			err = add_to_page_cache(page, mapping, idx, GFP_KERNEL);
			if (err) {
				put_page(page);
				if (err == -EEXIST)
					goto retry;
				goto out;
			}
K
Ken Chen 已提交
2572 2573

			spin_lock(&inode->i_lock);
2574
			inode->i_blocks += blocks_per_huge_page(h);
K
Ken Chen 已提交
2575
			spin_unlock(&inode->i_lock);
2576
			page_dup_rmap(page);
2577
		} else {
2578
			lock_page(page);
2579 2580 2581 2582 2583
			if (unlikely(anon_vma_prepare(vma))) {
				ret = VM_FAULT_OOM;
				goto backout_unlocked;
			}
			hugepage_add_new_anon_rmap(page, vma, address);
2584
		}
2585
	} else {
2586 2587 2588 2589 2590 2591 2592 2593 2594
		/*
		 * If memory error occurs between mmap() and fault, some process
		 * don't have hwpoisoned swap entry for errored virtual address.
		 * So we need to block hugepage fault by PG_hwpoison bit check.
		 */
		if (unlikely(PageHWPoison(page))) {
			ret = VM_FAULT_HWPOISON;
			goto backout_unlocked;
		}
2595
		page_dup_rmap(page);
2596
	}
2597

2598 2599 2600 2601 2602 2603
	/*
	 * If we are going to COW a private mapping later, we examine the
	 * pending reservations for this page now. This will ensure that
	 * any allocations necessary to record that reservation occur outside
	 * the spinlock.
	 */
2604
	if ((flags & FAULT_FLAG_WRITE) && !(vma->vm_flags & VM_SHARED))
2605 2606 2607 2608
		if (vma_needs_reservation(h, vma, address) < 0) {
			ret = VM_FAULT_OOM;
			goto backout_unlocked;
		}
2609

2610
	spin_lock(&mm->page_table_lock);
2611
	size = i_size_read(mapping->host) >> huge_page_shift(h);
A
Adam Litke 已提交
2612 2613 2614
	if (idx >= size)
		goto backout;

N
Nick Piggin 已提交
2615
	ret = 0;
2616
	if (!huge_pte_none(huge_ptep_get(ptep)))
A
Adam Litke 已提交
2617 2618
		goto backout;

2619 2620 2621 2622
	new_pte = make_huge_pte(vma, page, ((vma->vm_flags & VM_WRITE)
				&& (vma->vm_flags & VM_SHARED)));
	set_huge_pte_at(mm, address, ptep, new_pte);

2623
	if ((flags & FAULT_FLAG_WRITE) && !(vma->vm_flags & VM_SHARED)) {
2624
		/* Optimization, do the COW without a second fault */
2625
		ret = hugetlb_cow(mm, vma, address, ptep, new_pte, page);
2626 2627
	}

2628
	spin_unlock(&mm->page_table_lock);
A
Adam Litke 已提交
2629 2630
	unlock_page(page);
out:
2631
	return ret;
A
Adam Litke 已提交
2632 2633 2634

backout:
	spin_unlock(&mm->page_table_lock);
2635
backout_unlocked:
A
Adam Litke 已提交
2636 2637 2638
	unlock_page(page);
	put_page(page);
	goto out;
2639 2640
}

2641
int hugetlb_fault(struct mm_struct *mm, struct vm_area_struct *vma,
2642
			unsigned long address, unsigned int flags)
2643 2644 2645
{
	pte_t *ptep;
	pte_t entry;
2646
	int ret;
2647
	struct page *page = NULL;
2648
	struct page *pagecache_page = NULL;
2649
	static DEFINE_MUTEX(hugetlb_instantiation_mutex);
2650
	struct hstate *h = hstate_vma(vma);
2651

2652 2653 2654
	ptep = huge_pte_offset(mm, address);
	if (ptep) {
		entry = huge_ptep_get(ptep);
N
Naoya Horiguchi 已提交
2655 2656 2657 2658
		if (unlikely(is_hugetlb_entry_migration(entry))) {
			migration_entry_wait(mm, (pmd_t *)ptep, address);
			return 0;
		} else if (unlikely(is_hugetlb_entry_hwpoisoned(entry)))
2659 2660 2661
			return VM_FAULT_HWPOISON;
	}

2662
	ptep = huge_pte_alloc(mm, address, huge_page_size(h));
2663 2664 2665
	if (!ptep)
		return VM_FAULT_OOM;

2666 2667 2668 2669 2670 2671
	/*
	 * Serialize hugepage allocation and instantiation, so that we don't
	 * get spurious allocation failures if two CPUs race to instantiate
	 * the same page in the page cache.
	 */
	mutex_lock(&hugetlb_instantiation_mutex);
2672 2673
	entry = huge_ptep_get(ptep);
	if (huge_pte_none(entry)) {
2674
		ret = hugetlb_no_page(mm, vma, address, ptep, flags);
2675
		goto out_mutex;
2676
	}
2677

N
Nick Piggin 已提交
2678
	ret = 0;
2679

2680 2681 2682 2683 2684 2685 2686 2687
	/*
	 * If we are going to COW the mapping later, we examine the pending
	 * reservations for this page now. This will ensure that any
	 * allocations necessary to record that reservation occur outside the
	 * spinlock. For private mappings, we also lookup the pagecache
	 * page now as it is used to determine if a reservation has been
	 * consumed.
	 */
2688
	if ((flags & FAULT_FLAG_WRITE) && !pte_write(entry)) {
2689 2690
		if (vma_needs_reservation(h, vma, address) < 0) {
			ret = VM_FAULT_OOM;
2691
			goto out_mutex;
2692
		}
2693

2694
		if (!(vma->vm_flags & VM_MAYSHARE))
2695 2696 2697 2698
			pagecache_page = hugetlbfs_pagecache_page(h,
								vma, address);
	}

2699 2700 2701 2702 2703 2704 2705 2706 2707
	/*
	 * hugetlb_cow() requires page locks of pte_page(entry) and
	 * pagecache_page, so here we need take the former one
	 * when page != pagecache_page or !pagecache_page.
	 * Note that locking order is always pagecache_page -> page,
	 * so no worry about deadlock.
	 */
	page = pte_page(entry);
	if (page != pagecache_page)
2708 2709
		lock_page(page);

2710 2711
	spin_lock(&mm->page_table_lock);
	/* Check for a racing update before calling hugetlb_cow */
2712 2713 2714 2715
	if (unlikely(!pte_same(entry, huge_ptep_get(ptep))))
		goto out_page_table_lock;


2716
	if (flags & FAULT_FLAG_WRITE) {
2717
		if (!pte_write(entry)) {
2718 2719
			ret = hugetlb_cow(mm, vma, address, ptep, entry,
							pagecache_page);
2720 2721 2722 2723 2724
			goto out_page_table_lock;
		}
		entry = pte_mkdirty(entry);
	}
	entry = pte_mkyoung(entry);
2725 2726
	if (huge_ptep_set_access_flags(vma, address, ptep, entry,
						flags & FAULT_FLAG_WRITE))
2727
		update_mmu_cache(vma, address, ptep);
2728 2729

out_page_table_lock:
2730
	spin_unlock(&mm->page_table_lock);
2731 2732 2733 2734 2735

	if (pagecache_page) {
		unlock_page(pagecache_page);
		put_page(pagecache_page);
	}
2736
	unlock_page(page);
2737

2738
out_mutex:
2739
	mutex_unlock(&hugetlb_instantiation_mutex);
2740 2741

	return ret;
2742 2743
}

A
Andi Kleen 已提交
2744 2745 2746 2747 2748 2749 2750 2751 2752
/* Can be overriden by architectures */
__attribute__((weak)) struct page *
follow_huge_pud(struct mm_struct *mm, unsigned long address,
	       pud_t *pud, int write)
{
	BUG();
	return NULL;
}

D
David Gibson 已提交
2753 2754
int follow_hugetlb_page(struct mm_struct *mm, struct vm_area_struct *vma,
			struct page **pages, struct vm_area_struct **vmas,
2755
			unsigned long *position, int *length, int i,
H
Hugh Dickins 已提交
2756
			unsigned int flags)
D
David Gibson 已提交
2757
{
2758 2759
	unsigned long pfn_offset;
	unsigned long vaddr = *position;
D
David Gibson 已提交
2760
	int remainder = *length;
2761
	struct hstate *h = hstate_vma(vma);
D
David Gibson 已提交
2762

2763
	spin_lock(&mm->page_table_lock);
D
David Gibson 已提交
2764
	while (vaddr < vma->vm_end && remainder) {
A
Adam Litke 已提交
2765
		pte_t *pte;
H
Hugh Dickins 已提交
2766
		int absent;
A
Adam Litke 已提交
2767
		struct page *page;
D
David Gibson 已提交
2768

A
Adam Litke 已提交
2769 2770
		/*
		 * Some archs (sparc64, sh*) have multiple pte_ts to
H
Hugh Dickins 已提交
2771
		 * each hugepage.  We have to make sure we get the
A
Adam Litke 已提交
2772 2773
		 * first, for the page indexing below to work.
		 */
2774
		pte = huge_pte_offset(mm, vaddr & huge_page_mask(h));
H
Hugh Dickins 已提交
2775 2776 2777 2778
		absent = !pte || huge_pte_none(huge_ptep_get(pte));

		/*
		 * When coredumping, it suits get_dump_page if we just return
H
Hugh Dickins 已提交
2779 2780 2781 2782
		 * an error where there's an empty slot with no huge pagecache
		 * to back it.  This way, we avoid allocating a hugepage, and
		 * the sparse dumpfile avoids allocating disk blocks, but its
		 * huge holes still show up with zeroes where they need to be.
H
Hugh Dickins 已提交
2783
		 */
H
Hugh Dickins 已提交
2784 2785
		if (absent && (flags & FOLL_DUMP) &&
		    !hugetlbfs_pagecache_present(h, vma, vaddr)) {
H
Hugh Dickins 已提交
2786 2787 2788
			remainder = 0;
			break;
		}
D
David Gibson 已提交
2789

H
Hugh Dickins 已提交
2790 2791
		if (absent ||
		    ((flags & FOLL_WRITE) && !pte_write(huge_ptep_get(pte)))) {
A
Adam Litke 已提交
2792
			int ret;
D
David Gibson 已提交
2793

A
Adam Litke 已提交
2794
			spin_unlock(&mm->page_table_lock);
H
Hugh Dickins 已提交
2795 2796
			ret = hugetlb_fault(mm, vma, vaddr,
				(flags & FOLL_WRITE) ? FAULT_FLAG_WRITE : 0);
A
Adam Litke 已提交
2797
			spin_lock(&mm->page_table_lock);
2798
			if (!(ret & VM_FAULT_ERROR))
A
Adam Litke 已提交
2799
				continue;
D
David Gibson 已提交
2800

A
Adam Litke 已提交
2801 2802 2803 2804
			remainder = 0;
			break;
		}

2805
		pfn_offset = (vaddr & ~huge_page_mask(h)) >> PAGE_SHIFT;
2806
		page = pte_page(huge_ptep_get(pte));
2807
same_page:
2808
		if (pages) {
H
Hugh Dickins 已提交
2809
			pages[i] = mem_map_offset(page, pfn_offset);
K
KOSAKI Motohiro 已提交
2810
			get_page(pages[i]);
2811
		}
D
David Gibson 已提交
2812 2813 2814 2815 2816

		if (vmas)
			vmas[i] = vma;

		vaddr += PAGE_SIZE;
2817
		++pfn_offset;
D
David Gibson 已提交
2818 2819
		--remainder;
		++i;
2820
		if (vaddr < vma->vm_end && remainder &&
2821
				pfn_offset < pages_per_huge_page(h)) {
2822 2823 2824 2825 2826 2827
			/*
			 * We use pfn_offset to avoid touching the pageframes
			 * of this compound page.
			 */
			goto same_page;
		}
D
David Gibson 已提交
2828
	}
2829
	spin_unlock(&mm->page_table_lock);
D
David Gibson 已提交
2830 2831 2832
	*length = remainder;
	*position = vaddr;

H
Hugh Dickins 已提交
2833
	return i ? i : -EFAULT;
D
David Gibson 已提交
2834
}
2835 2836 2837 2838 2839 2840 2841 2842

void hugetlb_change_protection(struct vm_area_struct *vma,
		unsigned long address, unsigned long end, pgprot_t newprot)
{
	struct mm_struct *mm = vma->vm_mm;
	unsigned long start = address;
	pte_t *ptep;
	pte_t pte;
2843
	struct hstate *h = hstate_vma(vma);
2844 2845 2846 2847

	BUG_ON(address >= end);
	flush_cache_range(vma, address, end);

2848
	spin_lock(&vma->vm_file->f_mapping->i_mmap_lock);
2849
	spin_lock(&mm->page_table_lock);
2850
	for (; address < end; address += huge_page_size(h)) {
2851 2852 2853
		ptep = huge_pte_offset(mm, address);
		if (!ptep)
			continue;
2854 2855
		if (huge_pmd_unshare(mm, &address, ptep))
			continue;
2856
		if (!huge_pte_none(huge_ptep_get(ptep))) {
2857 2858 2859 2860 2861 2862
			pte = huge_ptep_get_and_clear(mm, address, ptep);
			pte = pte_mkhuge(pte_modify(pte, newprot));
			set_huge_pte_at(mm, address, ptep, pte);
		}
	}
	spin_unlock(&mm->page_table_lock);
2863
	spin_unlock(&vma->vm_file->f_mapping->i_mmap_lock);
2864 2865 2866 2867

	flush_tlb_range(vma, start, end);
}

2868 2869
int hugetlb_reserve_pages(struct inode *inode,
					long from, long to,
2870 2871
					struct vm_area_struct *vma,
					int acctflag)
2872
{
2873
	long ret, chg;
2874
	struct hstate *h = hstate_inode(inode);
2875

2876 2877 2878 2879 2880 2881 2882 2883
	/*
	 * Only apply hugepage reservation if asked. At fault time, an
	 * attempt will be made for VM_NORESERVE to allocate a page
	 * and filesystem quota without using reserves
	 */
	if (acctflag & VM_NORESERVE)
		return 0;

2884 2885 2886 2887 2888 2889
	/*
	 * Shared mappings base their reservation on the number of pages that
	 * are already allocated on behalf of the file. Private mappings need
	 * to reserve the full area even if read-only as mprotect() may be
	 * called to make the mapping read-write. Assume !vma is a shm mapping
	 */
2890
	if (!vma || vma->vm_flags & VM_MAYSHARE)
2891
		chg = region_chg(&inode->i_mapping->private_list, from, to);
2892 2893 2894 2895 2896
	else {
		struct resv_map *resv_map = resv_map_alloc();
		if (!resv_map)
			return -ENOMEM;

2897
		chg = to - from;
2898

2899 2900 2901 2902
		set_vma_resv_map(vma, resv_map);
		set_vma_resv_flags(vma, HPAGE_RESV_OWNER);
	}

2903 2904
	if (chg < 0)
		return chg;
2905

2906
	/* There must be enough filesystem quota for the mapping */
2907 2908
	if (hugetlb_get_quota(inode->i_mapping, chg))
		return -ENOSPC;
2909 2910

	/*
2911 2912
	 * Check enough hugepages are available for the reservation.
	 * Hand back the quota if there are not
2913
	 */
2914
	ret = hugetlb_acct_memory(h, chg);
K
Ken Chen 已提交
2915 2916
	if (ret < 0) {
		hugetlb_put_quota(inode->i_mapping, chg);
2917
		return ret;
K
Ken Chen 已提交
2918
	}
2919 2920 2921 2922 2923 2924 2925 2926 2927 2928 2929 2930

	/*
	 * Account for the reservations made. Shared mappings record regions
	 * that have reservations as they are shared by multiple VMAs.
	 * When the last VMA disappears, the region map says how much
	 * the reservation was and the page cache tells how much of
	 * the reservation was consumed. Private mappings are per-VMA and
	 * only the consumed reservations are tracked. When the VMA
	 * disappears, the original reservation is the VMA size and the
	 * consumed reservations are stored in the map. Hence, nothing
	 * else has to be done for private mappings here
	 */
2931
	if (!vma || vma->vm_flags & VM_MAYSHARE)
2932
		region_add(&inode->i_mapping->private_list, from, to);
2933 2934 2935 2936 2937
	return 0;
}

void hugetlb_unreserve_pages(struct inode *inode, long offset, long freed)
{
2938
	struct hstate *h = hstate_inode(inode);
2939
	long chg = region_truncate(&inode->i_mapping->private_list, offset);
K
Ken Chen 已提交
2940 2941

	spin_lock(&inode->i_lock);
2942
	inode->i_blocks -= (blocks_per_huge_page(h) * freed);
K
Ken Chen 已提交
2943 2944
	spin_unlock(&inode->i_lock);

2945
	hugetlb_put_quota(inode->i_mapping, (chg - freed));
2946
	hugetlb_acct_memory(h, -(chg - freed));
2947
}
2948

2949 2950 2951 2952 2953 2954 2955 2956 2957 2958 2959 2960 2961 2962 2963
/* Should be called in hugetlb_lock */
static int is_hugepage_on_freelist(struct page *hpage)
{
	struct page *page;
	struct page *tmp;
	struct hstate *h = page_hstate(hpage);
	int nid = page_to_nid(hpage);

	list_for_each_entry_safe(page, tmp, &h->hugepage_freelists[nid], lru)
		if (page == hpage)
			return 1;
	return 0;
}

#ifdef CONFIG_MEMORY_FAILURE
2964 2965 2966 2967
/*
 * This function is called from memory failure code.
 * Assume the caller holds page lock of the head page.
 */
2968
int dequeue_hwpoisoned_huge_page(struct page *hpage)
2969 2970 2971
{
	struct hstate *h = page_hstate(hpage);
	int nid = page_to_nid(hpage);
2972
	int ret = -EBUSY;
2973 2974

	spin_lock(&hugetlb_lock);
2975 2976
	if (is_hugepage_on_freelist(hpage)) {
		list_del(&hpage->lru);
2977
		set_page_refcounted(hpage);
2978 2979 2980 2981
		h->free_huge_pages--;
		h->free_huge_pages_node[nid]--;
		ret = 0;
	}
2982
	spin_unlock(&hugetlb_lock);
2983
	return ret;
2984
}
2985
#endif