hugetlb.c 76.9 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 <linux/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
 *
 * 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:
 *
65
 *	down_write(&mm->mmap_sem);
66
 * or
67 68
 *	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
 */
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;

L
Lucas De Marchi 已提交
149
		/* We overlap with this area, if it extends further than
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
		 * 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
}

397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430
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);
	}
}

431
static void enqueue_huge_page(struct hstate *h, struct page *page)
L
Linus Torvalds 已提交
432 433
{
	int nid = page_to_nid(page);
434 435 436
	list_add(&page->lru, &h->hugepage_freelists[nid]);
	h->free_huge_pages++;
	h->free_huge_pages_node[nid]++;
L
Linus Torvalds 已提交
437 438
}

439 440 441 442 443 444 445 446
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);
447
	set_page_refcounted(page);
448 449 450 451 452
	h->free_huge_pages--;
	h->free_huge_pages_node[nid]--;
	return page;
}

453 454
static struct page *dequeue_huge_page_vma(struct hstate *h,
				struct vm_area_struct *vma,
455
				unsigned long address, int avoid_reserve)
L
Linus Torvalds 已提交
456 457
{
	struct page *page = NULL;
458
	struct mempolicy *mpol;
459
	nodemask_t *nodemask;
460
	struct zonelist *zonelist;
461 462
	struct zone *zone;
	struct zoneref *z;
L
Linus Torvalds 已提交
463

464 465 466
	get_mems_allowed();
	zonelist = huge_zonelist(vma, address,
					htlb_alloc_mask, &mpol, &nodemask);
467 468 469 470 471
	/*
	 * 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
	 */
472
	if (!vma_has_reserves(vma) &&
473
			h->free_huge_pages - h->resv_huge_pages == 0)
474
		goto err;
475

476
	/* If reserves cannot be used, ensure enough pages are in the pool */
477
	if (avoid_reserve && h->free_huge_pages - h->resv_huge_pages == 0)
478
		goto err;
479

480 481
	for_each_zone_zonelist_nodemask(zone, z, zonelist,
						MAX_NR_ZONES - 1, nodemask) {
482 483 484 485 486 487 488
		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
489
		}
L
Linus Torvalds 已提交
490
	}
491
err:
492
	mpol_cond_put(mpol);
493
	put_mems_allowed();
L
Linus Torvalds 已提交
494 495 496
	return page;
}

497
static void update_and_free_page(struct hstate *h, struct page *page)
A
Adam Litke 已提交
498 499
{
	int i;
500

501 502
	VM_BUG_ON(h->order >= MAX_ORDER);

503 504 505
	h->nr_huge_pages--;
	h->nr_huge_pages_node[page_to_nid(page)]--;
	for (i = 0; i < pages_per_huge_page(h); i++) {
506 507 508 509
		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);
A
Adam Litke 已提交
510 511 512
	}
	set_compound_page_dtor(page, NULL);
	set_page_refcounted(page);
513
	arch_release_hugepage(page);
514
	__free_pages(page, huge_page_order(h));
A
Adam Litke 已提交
515 516
}

517 518 519 520 521 522 523 524 525 526 527
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;
}

528 529
static void free_huge_page(struct page *page)
{
530 531 532 533
	/*
	 * Can't pass hstate in here because it is called from the
	 * compound page destructor.
	 */
534
	struct hstate *h = page_hstate(page);
535
	int nid = page_to_nid(page);
536
	struct address_space *mapping;
537

538
	mapping = (struct address_space *) page_private(page);
539
	set_page_private(page, 0);
540
	page->mapping = NULL;
541
	BUG_ON(page_count(page));
542
	BUG_ON(page_mapcount(page));
543 544 545
	INIT_LIST_HEAD(&page->lru);

	spin_lock(&hugetlb_lock);
546
	if (h->surplus_huge_pages_node[nid] && huge_page_order(h) < MAX_ORDER) {
547 548 549
		update_and_free_page(h, page);
		h->surplus_huge_pages--;
		h->surplus_huge_pages_node[nid]--;
550
	} else {
551
		enqueue_huge_page(h, page);
552
	}
553
	spin_unlock(&hugetlb_lock);
554
	if (mapping)
555
		hugetlb_put_quota(mapping, 1);
556 557
}

558
static void prep_new_huge_page(struct hstate *h, struct page *page, int nid)
559 560 561
{
	set_compound_page_dtor(page, free_huge_page);
	spin_lock(&hugetlb_lock);
562 563
	h->nr_huge_pages++;
	h->nr_huge_pages_node[nid]++;
564 565 566 567
	spin_unlock(&hugetlb_lock);
	put_page(page); /* free it into the hugepage allocator */
}

568 569 570 571 572 573 574 575 576 577 578
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);
579
		set_page_count(p, 0);
580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595
		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;
}
596 597
EXPORT_SYMBOL_GPL(PageHuge);

598
static struct page *alloc_fresh_huge_page_node(struct hstate *h, int nid)
L
Linus Torvalds 已提交
599 600
{
	struct page *page;
601

602 603 604
	if (h->order >= MAX_ORDER)
		return NULL;

605
	page = alloc_pages_exact_node(nid,
606 607
		htlb_alloc_mask|__GFP_COMP|__GFP_THISNODE|
						__GFP_REPEAT|__GFP_NOWARN,
608
		huge_page_order(h));
L
Linus Torvalds 已提交
609
	if (page) {
610
		if (arch_prepare_hugepage(page)) {
611
			__free_pages(page, huge_page_order(h));
612
			return NULL;
613
		}
614
		prep_new_huge_page(h, page, nid);
L
Linus Torvalds 已提交
615
	}
616 617 618 619

	return page;
}

620
/*
621 622 623 624 625
 * 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.
626
 */
627
static int next_node_allowed(int nid, nodemask_t *nodes_allowed)
628
{
629
	nid = next_node(nid, *nodes_allowed);
630
	if (nid == MAX_NUMNODES)
631
		nid = first_node(*nodes_allowed);
632 633 634 635 636
	VM_BUG_ON(nid >= MAX_NUMNODES);

	return nid;
}

637 638 639 640 641 642 643
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;
}

644
/*
645 646 647 648
 * 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.
649
 */
650 651
static int hstate_next_node_to_alloc(struct hstate *h,
					nodemask_t *nodes_allowed)
652
{
653 654 655 656 657 658
	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);
659 660

	return nid;
661 662
}

663
static int alloc_fresh_huge_page(struct hstate *h, nodemask_t *nodes_allowed)
664 665 666 667 668 669
{
	struct page *page;
	int start_nid;
	int next_nid;
	int ret = 0;

670
	start_nid = hstate_next_node_to_alloc(h, nodes_allowed);
671
	next_nid = start_nid;
672 673

	do {
674
		page = alloc_fresh_huge_page_node(h, next_nid);
675
		if (page) {
676
			ret = 1;
677 678
			break;
		}
679
		next_nid = hstate_next_node_to_alloc(h, nodes_allowed);
680
	} while (next_nid != start_nid);
681

682 683 684 685 686
	if (ret)
		count_vm_event(HTLB_BUDDY_PGALLOC);
	else
		count_vm_event(HTLB_BUDDY_PGALLOC_FAIL);

687
	return ret;
L
Linus Torvalds 已提交
688 689
}

690
/*
691 692 693 694
 * 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.
695
 */
696
static int hstate_next_node_to_free(struct hstate *h, nodemask_t *nodes_allowed)
697
{
698 699 700 701 702 703
	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);
704 705

	return nid;
706 707 708 709 710 711 712 713
}

/*
 * 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.
 */
714 715
static int free_pool_huge_page(struct hstate *h, nodemask_t *nodes_allowed,
							 bool acct_surplus)
716 717 718 719 720
{
	int start_nid;
	int next_nid;
	int ret = 0;

721
	start_nid = hstate_next_node_to_free(h, nodes_allowed);
722 723 724
	next_nid = start_nid;

	do {
725 726 727 728 729 730
		/*
		 * 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])) {
731 732 733 734 735 736
			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]--;
737 738 739 740
			if (acct_surplus) {
				h->surplus_huge_pages--;
				h->surplus_huge_pages_node[next_nid]--;
			}
741 742
			update_and_free_page(h, page);
			ret = 1;
743
			break;
744
		}
745
		next_nid = hstate_next_node_to_free(h, nodes_allowed);
746
	} while (next_nid != start_nid);
747 748 749 750

	return ret;
}

751
static struct page *alloc_buddy_huge_page(struct hstate *h, int nid)
752 753
{
	struct page *page;
754
	unsigned int r_nid;
755

756 757 758
	if (h->order >= MAX_ORDER)
		return NULL;

759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782
	/*
	 * 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);
783
	if (h->surplus_huge_pages >= h->nr_overcommit_huge_pages) {
784 785 786
		spin_unlock(&hugetlb_lock);
		return NULL;
	} else {
787 788
		h->nr_huge_pages++;
		h->surplus_huge_pages++;
789 790 791
	}
	spin_unlock(&hugetlb_lock);

792 793 794 795 796 797 798 799
	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));
800

801 802
	if (page && arch_prepare_hugepage(page)) {
		__free_pages(page, huge_page_order(h));
803
		page = NULL;
804 805
	}

806
	spin_lock(&hugetlb_lock);
807
	if (page) {
808
		r_nid = page_to_nid(page);
809
		set_compound_page_dtor(page, free_huge_page);
810 811 812
		/*
		 * We incremented the global counters already
		 */
813 814
		h->nr_huge_pages_node[r_nid]++;
		h->surplus_huge_pages_node[r_nid]++;
815
		__count_vm_event(HTLB_BUDDY_PGALLOC);
816
	} else {
817 818
		h->nr_huge_pages--;
		h->surplus_huge_pages--;
819
		__count_vm_event(HTLB_BUDDY_PGALLOC_FAIL);
820
	}
821
	spin_unlock(&hugetlb_lock);
822 823 824 825

	return page;
}

826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844
/*
 * 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;
}

845
/*
L
Lucas De Marchi 已提交
846
 * Increase the hugetlb pool such that it can accommodate a reservation
847 848
 * of size 'delta'.
 */
849
static int gather_surplus_pages(struct hstate *h, int delta)
850 851 852 853 854 855
{
	struct list_head surplus_list;
	struct page *page, *tmp;
	int ret, i;
	int needed, allocated;

856
	needed = (h->resv_huge_pages + delta) - h->free_huge_pages;
857
	if (needed <= 0) {
858
		h->resv_huge_pages += delta;
859
		return 0;
860
	}
861 862 863 864 865 866 867 868

	allocated = 0;
	INIT_LIST_HEAD(&surplus_list);

	ret = -ENOMEM;
retry:
	spin_unlock(&hugetlb_lock);
	for (i = 0; i < needed; i++) {
869
		page = alloc_buddy_huge_page(h, NUMA_NO_NODE);
870
		if (!page)
871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886
			/*
			 * 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);
887 888
	needed = (h->resv_huge_pages + delta) -
			(h->free_huge_pages + allocated);
889 890 891 892 893
	if (needed > 0)
		goto retry;

	/*
	 * The surplus_list now contains _at_least_ the number of extra pages
L
Lucas De Marchi 已提交
894
	 * needed to accommodate the reservation.  Add the appropriate number
895
	 * of pages to the hugetlb pool and free the extras back to the buddy
896 897 898
	 * 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.
899 900
	 */
	needed += allocated;
901
	h->resv_huge_pages += delta;
902
	ret = 0;
903

904
	/* Free the needed pages to the hugetlb pool */
905
	list_for_each_entry_safe(page, tmp, &surplus_list, lru) {
906 907
		if ((--needed) < 0)
			break;
908
		list_del(&page->lru);
909 910 911 912 913 914
		/*
		 * 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));
915
		enqueue_huge_page(h, page);
916
	}
917
	spin_unlock(&hugetlb_lock);
918 919

	/* Free unnecessary surplus pages to the buddy allocator */
920
free:
921 922 923
	if (!list_empty(&surplus_list)) {
		list_for_each_entry_safe(page, tmp, &surplus_list, lru) {
			list_del(&page->lru);
924
			put_page(page);
925
		}
926
	}
927
	spin_lock(&hugetlb_lock);
928 929 930 931 932 933 934 935

	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.
936
 * Called with hugetlb_lock held.
937
 */
938 939
static void return_unused_surplus_pages(struct hstate *h,
					unsigned long unused_resv_pages)
940 941 942
{
	unsigned long nr_pages;

943
	/* Uncommit the reservation */
944
	h->resv_huge_pages -= unused_resv_pages;
945

946 947 948 949
	/* Cannot return gigantic pages currently */
	if (h->order >= MAX_ORDER)
		return;

950
	nr_pages = min(unused_resv_pages, h->surplus_huge_pages);
951

952 953
	/*
	 * We want to release as many surplus pages as possible, spread
954 955 956 957 958
	 * 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.
959 960
	 */
	while (nr_pages--) {
961
		if (!free_pool_huge_page(h, &node_states[N_HIGH_MEMORY], 1))
962
			break;
963 964 965
	}
}

966 967 968 969 970 971 972 973 974
/*
 * 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.
 */
975
static long vma_needs_reservation(struct hstate *h,
976
			struct vm_area_struct *vma, unsigned long addr)
977 978 979 980
{
	struct address_space *mapping = vma->vm_file->f_mapping;
	struct inode *inode = mapping->host;

981
	if (vma->vm_flags & VM_MAYSHARE) {
982
		pgoff_t idx = vma_hugecache_offset(h, vma, addr);
983 984 985
		return region_chg(&inode->i_mapping->private_list,
							idx, idx + 1);

986 987
	} else if (!is_vma_resv_set(vma, HPAGE_RESV_OWNER)) {
		return 1;
988

989
	} else  {
990
		long err;
991
		pgoff_t idx = vma_hugecache_offset(h, vma, addr);
992 993 994 995 996 997 998
		struct resv_map *reservations = vma_resv_map(vma);

		err = region_chg(&reservations->regions, idx, idx + 1);
		if (err < 0)
			return err;
		return 0;
	}
999
}
1000 1001
static void vma_commit_reservation(struct hstate *h,
			struct vm_area_struct *vma, unsigned long addr)
1002 1003 1004 1005
{
	struct address_space *mapping = vma->vm_file->f_mapping;
	struct inode *inode = mapping->host;

1006
	if (vma->vm_flags & VM_MAYSHARE) {
1007
		pgoff_t idx = vma_hugecache_offset(h, vma, addr);
1008
		region_add(&inode->i_mapping->private_list, idx, idx + 1);
1009 1010

	} else if (is_vma_resv_set(vma, HPAGE_RESV_OWNER)) {
1011
		pgoff_t idx = vma_hugecache_offset(h, vma, addr);
1012 1013 1014 1015
		struct resv_map *reservations = vma_resv_map(vma);

		/* Mark this page used in the map. */
		region_add(&reservations->regions, idx, idx + 1);
1016 1017 1018
	}
}

1019
static struct page *alloc_huge_page(struct vm_area_struct *vma,
1020
				    unsigned long addr, int avoid_reserve)
L
Linus Torvalds 已提交
1021
{
1022
	struct hstate *h = hstate_vma(vma);
1023
	struct page *page;
1024 1025
	struct address_space *mapping = vma->vm_file->f_mapping;
	struct inode *inode = mapping->host;
1026
	long chg;
1027 1028 1029 1030 1031

	/*
	 * 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
1032 1033
	 * MAP_NORESERVE mappings may also need pages and quota allocated
	 * if no reserve mapping overlaps.
1034
	 */
1035
	chg = vma_needs_reservation(h, vma, addr);
1036
	if (chg < 0)
1037
		return ERR_PTR(-VM_FAULT_OOM);
1038
	if (chg)
1039
		if (hugetlb_get_quota(inode->i_mapping, chg))
1040
			return ERR_PTR(-VM_FAULT_SIGBUS);
L
Linus Torvalds 已提交
1041 1042

	spin_lock(&hugetlb_lock);
1043
	page = dequeue_huge_page_vma(h, vma, addr, avoid_reserve);
L
Linus Torvalds 已提交
1044
	spin_unlock(&hugetlb_lock);
1045

K
Ken Chen 已提交
1046
	if (!page) {
1047
		page = alloc_buddy_huge_page(h, NUMA_NO_NODE);
K
Ken Chen 已提交
1048
		if (!page) {
1049
			hugetlb_put_quota(inode->i_mapping, chg);
1050
			return ERR_PTR(-VM_FAULT_SIGBUS);
K
Ken Chen 已提交
1051 1052
		}
	}
1053

1054
	set_page_private(page, (unsigned long) mapping);
1055

1056
	vma_commit_reservation(h, vma, addr);
1057

1058
	return page;
1059 1060
}

1061
int __weak alloc_bootmem_huge_page(struct hstate *h)
1062 1063
{
	struct huge_bootmem_page *m;
1064
	int nr_nodes = nodes_weight(node_states[N_HIGH_MEMORY]);
1065 1066 1067 1068 1069

	while (nr_nodes) {
		void *addr;

		addr = __alloc_bootmem_node_nopanic(
1070
				NODE_DATA(hstate_next_node_to_alloc(h,
1071
						&node_states[N_HIGH_MEMORY])),
1072 1073 1074 1075 1076 1077 1078 1079 1080
				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;
1081
			goto found;
1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094
		}
		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;
}

1095 1096 1097 1098 1099 1100 1101 1102
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);
}

1103 1104 1105 1106 1107 1108 1109
/* 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 hstate *h = m->hstate;
1110 1111 1112 1113 1114 1115 1116 1117 1118
		struct page *page;

#ifdef CONFIG_HIGHMEM
		page = pfn_to_page(m->phys >> PAGE_SHIFT);
		free_bootmem_late((unsigned long)m,
				  sizeof(struct huge_bootmem_page));
#else
		page = virt_to_page(m);
#endif
1119 1120
		__ClearPageReserved(page);
		WARN_ON(page_count(page) != 1);
1121
		prep_compound_huge_page(page, h->order);
1122
		prep_new_huge_page(h, page, page_to_nid(page));
1123 1124 1125 1126 1127 1128 1129 1130
		/*
		 * If we had gigantic hugepages allocated at boot time, we need
		 * to restore the 'stolen' pages to totalram_pages in order to
		 * fix confusing memory reports from free(1) and another
		 * side-effects, like CommitLimit going negative.
		 */
		if (h->order > (MAX_ORDER - 1))
			totalram_pages += 1 << h->order;
1131 1132 1133
	}
}

1134
static void __init hugetlb_hstate_alloc_pages(struct hstate *h)
L
Linus Torvalds 已提交
1135 1136
{
	unsigned long i;
1137

1138
	for (i = 0; i < h->max_huge_pages; ++i) {
1139 1140 1141
		if (h->order >= MAX_ORDER) {
			if (!alloc_bootmem_huge_page(h))
				break;
1142 1143
		} else if (!alloc_fresh_huge_page(h,
					 &node_states[N_HIGH_MEMORY]))
L
Linus Torvalds 已提交
1144 1145
			break;
	}
1146
	h->max_huge_pages = i;
1147 1148 1149 1150 1151 1152 1153
}

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

	for_each_hstate(h) {
1154 1155 1156
		/* oversize hugepages were init'ed in early boot */
		if (h->order < MAX_ORDER)
			hugetlb_hstate_alloc_pages(h);
1157 1158 1159
	}
}

A
Andi Kleen 已提交
1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170
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;
}

1171 1172 1173 1174 1175
static void __init report_hugepages(void)
{
	struct hstate *h;

	for_each_hstate(h) {
A
Andi Kleen 已提交
1176 1177 1178 1179 1180
		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);
1181 1182 1183
	}
}

L
Linus Torvalds 已提交
1184
#ifdef CONFIG_HIGHMEM
1185 1186
static void try_to_free_low(struct hstate *h, unsigned long count,
						nodemask_t *nodes_allowed)
L
Linus Torvalds 已提交
1187
{
1188 1189
	int i;

1190 1191 1192
	if (h->order >= MAX_ORDER)
		return;

1193
	for_each_node_mask(i, *nodes_allowed) {
L
Linus Torvalds 已提交
1194
		struct page *page, *next;
1195 1196 1197
		struct list_head *freel = &h->hugepage_freelists[i];
		list_for_each_entry_safe(page, next, freel, lru) {
			if (count >= h->nr_huge_pages)
1198
				return;
L
Linus Torvalds 已提交
1199 1200 1201
			if (PageHighMem(page))
				continue;
			list_del(&page->lru);
1202
			update_and_free_page(h, page);
1203 1204
			h->free_huge_pages--;
			h->free_huge_pages_node[page_to_nid(page)]--;
L
Linus Torvalds 已提交
1205 1206 1207 1208
		}
	}
}
#else
1209 1210
static inline void try_to_free_low(struct hstate *h, unsigned long count,
						nodemask_t *nodes_allowed)
L
Linus Torvalds 已提交
1211 1212 1213 1214
{
}
#endif

1215 1216 1217 1218 1219
/*
 * 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.
 */
1220 1221
static int adjust_pool_surplus(struct hstate *h, nodemask_t *nodes_allowed,
				int delta)
1222
{
1223
	int start_nid, next_nid;
1224 1225 1226 1227
	int ret = 0;

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

1228
	if (delta < 0)
1229
		start_nid = hstate_next_node_to_alloc(h, nodes_allowed);
1230
	else
1231
		start_nid = hstate_next_node_to_free(h, nodes_allowed);
1232 1233 1234 1235 1236 1237 1238 1239
	next_nid = start_nid;

	do {
		int nid = next_nid;
		if (delta < 0)  {
			/*
			 * To shrink on this node, there must be a surplus page
			 */
1240
			if (!h->surplus_huge_pages_node[nid]) {
1241 1242
				next_nid = hstate_next_node_to_alloc(h,
								nodes_allowed);
1243
				continue;
1244
			}
1245 1246 1247 1248 1249 1250
		}
		if (delta > 0) {
			/*
			 * Surplus cannot exceed the total number of pages
			 */
			if (h->surplus_huge_pages_node[nid] >=
1251
						h->nr_huge_pages_node[nid]) {
1252 1253
				next_nid = hstate_next_node_to_free(h,
								nodes_allowed);
1254
				continue;
1255
			}
1256
		}
1257 1258 1259 1260 1261

		h->surplus_huge_pages += delta;
		h->surplus_huge_pages_node[nid] += delta;
		ret = 1;
		break;
1262
	} while (next_nid != start_nid);
1263 1264 1265 1266

	return ret;
}

1267
#define persistent_huge_pages(h) (h->nr_huge_pages - h->surplus_huge_pages)
1268 1269
static unsigned long set_max_huge_pages(struct hstate *h, unsigned long count,
						nodemask_t *nodes_allowed)
L
Linus Torvalds 已提交
1270
{
1271
	unsigned long min_count, ret;
L
Linus Torvalds 已提交
1272

1273 1274 1275
	if (h->order >= MAX_ORDER)
		return h->max_huge_pages;

1276 1277 1278 1279
	/*
	 * Increase the pool size
	 * First take pages out of surplus state.  Then make up the
	 * remaining difference by allocating fresh huge pages.
1280 1281 1282 1283 1284 1285
	 *
	 * 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.
1286
	 */
L
Linus Torvalds 已提交
1287
	spin_lock(&hugetlb_lock);
1288
	while (h->surplus_huge_pages && count > persistent_huge_pages(h)) {
1289
		if (!adjust_pool_surplus(h, nodes_allowed, -1))
1290 1291 1292
			break;
	}

1293
	while (count > persistent_huge_pages(h)) {
1294 1295 1296 1297 1298 1299
		/*
		 * 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);
1300
		ret = alloc_fresh_huge_page(h, nodes_allowed);
1301 1302 1303 1304
		spin_lock(&hugetlb_lock);
		if (!ret)
			goto out;

1305 1306 1307
		/* Bail for signals. Probably ctrl-c from user */
		if (signal_pending(current))
			goto out;
1308 1309 1310 1311 1312 1313 1314 1315
	}

	/*
	 * 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.
1316 1317 1318 1319 1320 1321 1322 1323
	 *
	 * 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.
1324
	 */
1325
	min_count = h->resv_huge_pages + h->nr_huge_pages - h->free_huge_pages;
1326
	min_count = max(count, min_count);
1327
	try_to_free_low(h, min_count, nodes_allowed);
1328
	while (min_count < persistent_huge_pages(h)) {
1329
		if (!free_pool_huge_page(h, nodes_allowed, 0))
L
Linus Torvalds 已提交
1330 1331
			break;
	}
1332
	while (count < persistent_huge_pages(h)) {
1333
		if (!adjust_pool_surplus(h, nodes_allowed, 1))
1334 1335 1336
			break;
	}
out:
1337
	ret = persistent_huge_pages(h);
L
Linus Torvalds 已提交
1338
	spin_unlock(&hugetlb_lock);
1339
	return ret;
L
Linus Torvalds 已提交
1340 1341
}

1342 1343 1344 1345 1346 1347 1348 1349 1350 1351
#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];

1352 1353 1354
static struct hstate *kobj_to_node_hstate(struct kobject *kobj, int *nidp);

static struct hstate *kobj_to_hstate(struct kobject *kobj, int *nidp)
1355 1356
{
	int i;
1357

1358
	for (i = 0; i < HUGE_MAX_HSTATE; i++)
1359 1360 1361
		if (hstate_kobjs[i] == kobj) {
			if (nidp)
				*nidp = NUMA_NO_NODE;
1362
			return &hstates[i];
1363 1364 1365
		}

	return kobj_to_node_hstate(kobj, nidp);
1366 1367
}

1368
static ssize_t nr_hugepages_show_common(struct kobject *kobj,
1369 1370
					struct kobj_attribute *attr, char *buf)
{
1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381
	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);
1382
}
1383

1384 1385 1386
static ssize_t nr_hugepages_store_common(bool obey_mempolicy,
			struct kobject *kobj, struct kobj_attribute *attr,
			const char *buf, size_t len)
1387 1388
{
	int err;
1389
	int nid;
1390
	unsigned long count;
1391
	struct hstate *h;
1392
	NODEMASK_ALLOC(nodemask_t, nodes_allowed, GFP_KERNEL | __GFP_NORETRY);
1393

1394
	err = strict_strtoul(buf, 10, &count);
1395
	if (err)
1396
		goto out;
1397

1398
	h = kobj_to_hstate(kobj, &nid);
1399 1400 1401 1402 1403
	if (h->order >= MAX_ORDER) {
		err = -EINVAL;
		goto out;
	}

1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422
	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];

1423
	h->max_huge_pages = set_max_huge_pages(h, count, nodes_allowed);
1424

1425
	if (nodes_allowed != &node_states[N_HIGH_MEMORY])
1426 1427 1428
		NODEMASK_FREE(nodes_allowed);

	return len;
1429 1430 1431
out:
	NODEMASK_FREE(nodes_allowed);
	return err;
1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443
}

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);
1444 1445 1446
}
HSTATE_ATTR(nr_hugepages);

1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467
#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


1468 1469 1470
static ssize_t nr_overcommit_hugepages_show(struct kobject *kobj,
					struct kobj_attribute *attr, char *buf)
{
1471
	struct hstate *h = kobj_to_hstate(kobj, NULL);
1472 1473
	return sprintf(buf, "%lu\n", h->nr_overcommit_huge_pages);
}
1474

1475 1476 1477 1478 1479
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;
1480
	struct hstate *h = kobj_to_hstate(kobj, NULL);
1481

1482 1483 1484
	if (h->order >= MAX_ORDER)
		return -EINVAL;

1485 1486
	err = strict_strtoul(buf, 10, &input);
	if (err)
1487
		return err;
1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499

	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)
{
1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510
	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);
1511 1512 1513 1514 1515 1516
}
HSTATE_ATTR_RO(free_hugepages);

static ssize_t resv_hugepages_show(struct kobject *kobj,
					struct kobj_attribute *attr, char *buf)
{
1517
	struct hstate *h = kobj_to_hstate(kobj, NULL);
1518 1519 1520 1521 1522 1523 1524
	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)
{
1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535
	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);
1536 1537 1538 1539 1540 1541 1542 1543 1544
}
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,
1545 1546 1547
#ifdef CONFIG_NUMA
	&nr_hugepages_mempolicy_attr.attr,
#endif
1548 1549 1550 1551 1552 1553 1554
	NULL,
};

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

J
Jeff Mahoney 已提交
1555 1556 1557
static int hugetlb_sysfs_add_hstate(struct hstate *h, struct kobject *parent,
				    struct kobject **hstate_kobjs,
				    struct attribute_group *hstate_attr_group)
1558 1559
{
	int retval;
1560
	int hi = h - hstates;
1561

1562 1563
	hstate_kobjs[hi] = kobject_create_and_add(h->name, parent);
	if (!hstate_kobjs[hi])
1564 1565
		return -ENOMEM;

1566
	retval = sysfs_create_group(hstate_kobjs[hi], hstate_attr_group);
1567
	if (retval)
1568
		kobject_put(hstate_kobjs[hi]);
1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582

	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) {
1583 1584
		err = hugetlb_sysfs_add_hstate(h, hugepages_kobj,
					 hstate_kobjs, &hstate_attr_group);
1585 1586 1587 1588 1589 1590
		if (err)
			printk(KERN_ERR "Hugetlb: Unable to add hstate %s",
								h->name);
	}
}

1591 1592 1593 1594
#ifdef CONFIG_NUMA

/*
 * node_hstate/s - associate per node hstate attributes, via their kobjects,
1595 1596 1597
 * with node devices in node_devices[] using a parallel array.  The array
 * index of a node device or _hstate == node id.
 * This is here to avoid any static dependency of the node device driver, in
1598 1599 1600 1601 1602 1603 1604 1605 1606
 * 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];

/*
1607
 * A subset of global hstate attributes for node devices
1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620
 */
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,
};

/*
1621
 * kobj_to_node_hstate - lookup global hstate for node device hstate attr kobj.
1622 1623 1624 1625 1626 1627 1628 1629 1630 1631 1632 1633 1634 1635 1636 1637 1638 1639 1640 1641 1642 1643
 * 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;
}

/*
1644
 * Unregister hstate attributes from a single node device.
1645 1646 1647 1648 1649
 * No-op if no hstate attributes attached.
 */
void hugetlb_unregister_node(struct node *node)
{
	struct hstate *h;
1650
	struct node_hstate *nhs = &node_hstates[node->dev.id];
1651 1652

	if (!nhs->hugepages_kobj)
1653
		return;		/* no hstate attributes */
1654 1655 1656 1657 1658 1659 1660 1661 1662 1663 1664 1665

	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;
}

/*
1666
 * hugetlb module exit:  unregister hstate attributes from node devices
1667 1668 1669 1670 1671 1672 1673
 * that have them.
 */
static void hugetlb_unregister_all_nodes(void)
{
	int nid;

	/*
1674
	 * disable node device registrations.
1675 1676 1677 1678 1679 1680 1681 1682 1683 1684 1685
	 */
	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]);
}

/*
1686
 * Register hstate attributes for a single node device.
1687 1688 1689 1690 1691
 * No-op if attributes already registered.
 */
void hugetlb_register_node(struct node *node)
{
	struct hstate *h;
1692
	struct node_hstate *nhs = &node_hstates[node->dev.id];
1693 1694 1695 1696 1697 1698
	int err;

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

	nhs->hugepages_kobj = kobject_create_and_add("hugepages",
1699
							&node->dev.kobj);
1700 1701 1702 1703 1704 1705 1706 1707 1708 1709
	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",
1710
						h->name, node->dev.id);
1711 1712 1713 1714 1715 1716 1717
			hugetlb_unregister_node(node);
			break;
		}
	}
}

/*
1718
 * hugetlb init time:  register hstate attributes for all registered node
1719 1720
 * devices of nodes that have memory.  All on-line nodes should have
 * registered their associated device by this time.
1721 1722 1723 1724 1725
 */
static void hugetlb_register_all_nodes(void)
{
	int nid;

1726
	for_each_node_state(nid, N_HIGH_MEMORY) {
1727
		struct node *node = &node_devices[nid];
1728
		if (node->dev.id == nid)
1729 1730 1731 1732
			hugetlb_register_node(node);
	}

	/*
1733
	 * Let the node device driver know we're here so it can
1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746 1747 1748 1749 1750 1751 1752 1753 1754
	 * [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

1755 1756 1757 1758
static void __exit hugetlb_exit(void)
{
	struct hstate *h;

1759 1760
	hugetlb_unregister_all_nodes();

1761 1762 1763 1764 1765 1766 1767 1768 1769 1770
	for_each_hstate(h) {
		kobject_put(hstate_kobjs[h - hstates]);
	}

	kobject_put(hugepages_kobj);
}
module_exit(hugetlb_exit);

static int __init hugetlb_init(void)
{
1771 1772 1773 1774 1775 1776
	/* 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;
1777

1778 1779 1780 1781
	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);
1782
	}
1783 1784 1785
	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;
1786 1787 1788

	hugetlb_init_hstates();

1789 1790
	gather_bootmem_prealloc();

1791 1792 1793 1794
	report_hugepages();

	hugetlb_sysfs_init();

1795 1796
	hugetlb_register_all_nodes();

1797 1798 1799 1800 1801 1802 1803 1804
	return 0;
}
module_init(hugetlb_init);

/* Should be called on processing a hugepagesz=... option */
void __init hugetlb_add_hstate(unsigned order)
{
	struct hstate *h;
1805 1806
	unsigned long i;

1807 1808 1809 1810 1811 1812 1813 1814 1815
	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);
1816 1817 1818 1819
	h->nr_huge_pages = 0;
	h->free_huge_pages = 0;
	for (i = 0; i < MAX_NUMNODES; ++i)
		INIT_LIST_HEAD(&h->hugepage_freelists[i]);
1820 1821
	h->next_nid_to_alloc = first_node(node_states[N_HIGH_MEMORY]);
	h->next_nid_to_free = first_node(node_states[N_HIGH_MEMORY]);
1822 1823
	snprintf(h->name, HSTATE_NAME_LEN, "hugepages-%lukB",
					huge_page_size(h)/1024);
1824

1825 1826 1827
	parsed_hstate = h;
}

1828
static int __init hugetlb_nrpages_setup(char *s)
1829 1830
{
	unsigned long *mhp;
1831
	static unsigned long *last_mhp;
1832 1833 1834 1835 1836 1837 1838 1839 1840 1841

	/*
	 * !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;

1842 1843 1844 1845 1846 1847
	if (mhp == last_mhp) {
		printk(KERN_WARNING "hugepages= specified twice without "
			"interleaving hugepagesz=, ignoring\n");
		return 1;
	}

1848 1849 1850
	if (sscanf(s, "%lu", mhp) <= 0)
		*mhp = 0;

1851 1852 1853 1854 1855 1856 1857 1858 1859 1860
	/*
	 * 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;

1861 1862
	return 1;
}
1863 1864 1865 1866 1867 1868 1869 1870
__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);
1871

1872 1873 1874 1875 1876 1877 1878 1879 1880 1881 1882 1883
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
1884 1885 1886
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
1887
{
1888 1889
	struct hstate *h = &default_hstate;
	unsigned long tmp;
1890
	int ret;
1891

1892
	tmp = h->max_huge_pages;
1893

1894 1895 1896
	if (write && h->order >= MAX_ORDER)
		return -EINVAL;

1897 1898
	table->data = &tmp;
	table->maxlen = sizeof(unsigned long);
1899 1900 1901
	ret = proc_doulongvec_minmax(table, write, buffer, length, ppos);
	if (ret)
		goto out;
1902

1903
	if (write) {
1904 1905
		NODEMASK_ALLOC(nodemask_t, nodes_allowed,
						GFP_KERNEL | __GFP_NORETRY);
1906 1907 1908 1909 1910 1911 1912 1913 1914 1915
		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);
	}
1916 1917
out:
	return ret;
L
Linus Torvalds 已提交
1918
}
1919

1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931 1932 1933 1934 1935 1936
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 */

1937
int hugetlb_treat_movable_handler(struct ctl_table *table, int write,
1938
			void __user *buffer,
1939 1940
			size_t *length, loff_t *ppos)
{
1941
	proc_dointvec(table, write, buffer, length, ppos);
1942 1943 1944 1945 1946 1947 1948
	if (hugepages_treat_as_movable)
		htlb_alloc_mask = GFP_HIGHUSER_MOVABLE;
	else
		htlb_alloc_mask = GFP_HIGHUSER;
	return 0;
}

1949
int hugetlb_overcommit_handler(struct ctl_table *table, int write,
1950
			void __user *buffer,
1951 1952
			size_t *length, loff_t *ppos)
{
1953
	struct hstate *h = &default_hstate;
1954
	unsigned long tmp;
1955
	int ret;
1956

1957
	tmp = h->nr_overcommit_huge_pages;
1958

1959 1960 1961
	if (write && h->order >= MAX_ORDER)
		return -EINVAL;

1962 1963
	table->data = &tmp;
	table->maxlen = sizeof(unsigned long);
1964 1965 1966
	ret = proc_doulongvec_minmax(table, write, buffer, length, ppos);
	if (ret)
		goto out;
1967 1968 1969 1970 1971 1972

	if (write) {
		spin_lock(&hugetlb_lock);
		h->nr_overcommit_huge_pages = tmp;
		spin_unlock(&hugetlb_lock);
	}
1973 1974
out:
	return ret;
1975 1976
}

L
Linus Torvalds 已提交
1977 1978
#endif /* CONFIG_SYSCTL */

1979
void hugetlb_report_meminfo(struct seq_file *m)
L
Linus Torvalds 已提交
1980
{
1981
	struct hstate *h = &default_hstate;
1982
	seq_printf(m,
1983 1984 1985 1986 1987
			"HugePages_Total:   %5lu\n"
			"HugePages_Free:    %5lu\n"
			"HugePages_Rsvd:    %5lu\n"
			"HugePages_Surp:    %5lu\n"
			"Hugepagesize:   %8lu kB\n",
1988 1989 1990 1991 1992
			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
1993 1994 1995 1996
}

int hugetlb_report_node_meminfo(int nid, char *buf)
{
1997
	struct hstate *h = &default_hstate;
L
Linus Torvalds 已提交
1998 1999
	return sprintf(buf,
		"Node %d HugePages_Total: %5u\n"
2000 2001
		"Node %d HugePages_Free:  %5u\n"
		"Node %d HugePages_Surp:  %5u\n",
2002 2003 2004
		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
2005 2006 2007 2008 2009
}

/* Return the number pages of memory we physically have, in PAGE_SIZE units. */
unsigned long hugetlb_total_pages(void)
{
2010 2011
	struct hstate *h = &default_hstate;
	return h->nr_huge_pages * pages_per_huge_page(h);
L
Linus Torvalds 已提交
2012 2013
}

2014
static int hugetlb_acct_memory(struct hstate *h, long delta)
M
Mel Gorman 已提交
2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036
{
	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) {
2037
		if (gather_surplus_pages(h, delta) < 0)
M
Mel Gorman 已提交
2038 2039
			goto out;

2040 2041
		if (delta > cpuset_mems_nr(h->free_huge_pages_node)) {
			return_unused_surplus_pages(h, delta);
M
Mel Gorman 已提交
2042 2043 2044 2045 2046 2047
			goto out;
		}
	}

	ret = 0;
	if (delta < 0)
2048
		return_unused_surplus_pages(h, (unsigned long) -delta);
M
Mel Gorman 已提交
2049 2050 2051 2052 2053 2054

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

2055 2056 2057 2058 2059 2060 2061 2062
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
L
Lucas De Marchi 已提交
2063
	 * has a reference to the reservation map it cannot disappear until
2064 2065 2066 2067 2068 2069 2070
	 * after this open call completes.  It is therefore safe to take a
	 * new reference here without additional locking.
	 */
	if (reservations)
		kref_get(&reservations->refs);
}

2071 2072
static void hugetlb_vm_op_close(struct vm_area_struct *vma)
{
2073
	struct hstate *h = hstate_vma(vma);
2074 2075 2076 2077 2078 2079
	struct resv_map *reservations = vma_resv_map(vma);
	unsigned long reserve;
	unsigned long start;
	unsigned long end;

	if (reservations) {
2080 2081
		start = vma_hugecache_offset(h, vma, vma->vm_start);
		end = vma_hugecache_offset(h, vma, vma->vm_end);
2082 2083 2084 2085 2086 2087

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

		kref_put(&reservations->refs, resv_map_release);

2088
		if (reserve) {
2089
			hugetlb_acct_memory(h, -reserve);
2090 2091
			hugetlb_put_quota(vma->vm_file->f_mapping, reserve);
		}
2092
	}
2093 2094
}

L
Linus Torvalds 已提交
2095 2096 2097 2098 2099 2100
/*
 * 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
2101
static int hugetlb_vm_op_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
L
Linus Torvalds 已提交
2102 2103
{
	BUG();
N
Nick Piggin 已提交
2104
	return 0;
L
Linus Torvalds 已提交
2105 2106
}

2107
const struct vm_operations_struct hugetlb_vm_ops = {
N
Nick Piggin 已提交
2108
	.fault = hugetlb_vm_op_fault,
2109
	.open = hugetlb_vm_op_open,
2110
	.close = hugetlb_vm_op_close,
L
Linus Torvalds 已提交
2111 2112
};

2113 2114
static pte_t make_huge_pte(struct vm_area_struct *vma, struct page *page,
				int writable)
D
David Gibson 已提交
2115 2116 2117
{
	pte_t entry;

2118
	if (writable) {
D
David Gibson 已提交
2119 2120 2121
		entry =
		    pte_mkwrite(pte_mkdirty(mk_pte(page, vma->vm_page_prot)));
	} else {
2122
		entry = huge_pte_wrprotect(mk_pte(page, vma->vm_page_prot));
D
David Gibson 已提交
2123 2124 2125 2126 2127 2128 2129
	}
	entry = pte_mkyoung(entry);
	entry = pte_mkhuge(entry);

	return entry;
}

2130 2131 2132 2133 2134
static void set_huge_ptep_writable(struct vm_area_struct *vma,
				   unsigned long address, pte_t *ptep)
{
	pte_t entry;

2135
	entry = pte_mkwrite(pte_mkdirty(huge_ptep_get(ptep)));
2136
	if (huge_ptep_set_access_flags(vma, address, ptep, entry, 1))
2137
		update_mmu_cache(vma, address, ptep);
2138 2139 2140
}


D
David Gibson 已提交
2141 2142 2143 2144 2145
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;
2146
	unsigned long addr;
2147
	int cow;
2148 2149
	struct hstate *h = hstate_vma(vma);
	unsigned long sz = huge_page_size(h);
2150 2151

	cow = (vma->vm_flags & (VM_SHARED | VM_MAYWRITE)) == VM_MAYWRITE;
D
David Gibson 已提交
2152

2153
	for (addr = vma->vm_start; addr < vma->vm_end; addr += sz) {
H
Hugh Dickins 已提交
2154 2155 2156
		src_pte = huge_pte_offset(src, addr);
		if (!src_pte)
			continue;
2157
		dst_pte = huge_pte_alloc(dst, addr, sz);
D
David Gibson 已提交
2158 2159
		if (!dst_pte)
			goto nomem;
2160 2161 2162 2163 2164

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

H
Hugh Dickins 已提交
2165
		spin_lock(&dst->page_table_lock);
N
Nick Piggin 已提交
2166
		spin_lock_nested(&src->page_table_lock, SINGLE_DEPTH_NESTING);
2167
		if (!huge_pte_none(huge_ptep_get(src_pte))) {
2168
			if (cow)
2169 2170
				huge_ptep_set_wrprotect(src, addr, src_pte);
			entry = huge_ptep_get(src_pte);
2171 2172
			ptepage = pte_page(entry);
			get_page(ptepage);
2173
			page_dup_rmap(ptepage);
2174 2175 2176
			set_huge_pte_at(dst, addr, dst_pte, entry);
		}
		spin_unlock(&src->page_table_lock);
H
Hugh Dickins 已提交
2177
		spin_unlock(&dst->page_table_lock);
D
David Gibson 已提交
2178 2179 2180 2181 2182 2183 2184
	}
	return 0;

nomem:
	return -ENOMEM;
}

N
Naoya Horiguchi 已提交
2185 2186 2187 2188 2189 2190 2191
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);
2192
	if (non_swap_entry(swp) && is_migration_entry(swp))
N
Naoya Horiguchi 已提交
2193
		return 1;
2194
	else
N
Naoya Horiguchi 已提交
2195 2196 2197
		return 0;
}

2198 2199 2200 2201 2202 2203 2204
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);
2205
	if (non_swap_entry(swp) && is_hwpoison_entry(swp))
2206
		return 1;
2207
	else
2208 2209 2210
		return 0;
}

2211
void __unmap_hugepage_range(struct vm_area_struct *vma, unsigned long start,
2212
			    unsigned long end, struct page *ref_page)
D
David Gibson 已提交
2213 2214 2215
{
	struct mm_struct *mm = vma->vm_mm;
	unsigned long address;
2216
	pte_t *ptep;
D
David Gibson 已提交
2217 2218
	pte_t pte;
	struct page *page;
2219
	struct page *tmp;
2220 2221 2222
	struct hstate *h = hstate_vma(vma);
	unsigned long sz = huge_page_size(h);

2223
	/*
2224
	 * A page gathering list, protected by per file i_mmap_mutex. The
2225 2226 2227
	 * lock is used to avoid list corruption from multiple unmapping
	 * of the same page since we are using page->lru.
	 */
2228
	LIST_HEAD(page_list);
D
David Gibson 已提交
2229 2230

	WARN_ON(!is_vm_hugetlb_page(vma));
2231 2232
	BUG_ON(start & ~huge_page_mask(h));
	BUG_ON(end & ~huge_page_mask(h));
D
David Gibson 已提交
2233

A
Andrea Arcangeli 已提交
2234
	mmu_notifier_invalidate_range_start(mm, start, end);
2235
	spin_lock(&mm->page_table_lock);
2236
	for (address = start; address < end; address += sz) {
2237
		ptep = huge_pte_offset(mm, address);
A
Adam Litke 已提交
2238
		if (!ptep)
2239 2240
			continue;

2241 2242 2243
		if (huge_pmd_unshare(mm, &address, ptep))
			continue;

2244 2245 2246 2247 2248 2249 2250 2251 2252 2253 2254 2255 2256 2257 2258 2259 2260 2261 2262 2263 2264
		/*
		 * 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);
		}

2265
		pte = huge_ptep_get_and_clear(mm, address, ptep);
2266
		if (huge_pte_none(pte))
D
David Gibson 已提交
2267
			continue;
2268

2269 2270 2271 2272 2273 2274
		/*
		 * HWPoisoned hugepage is already unmapped and dropped reference
		 */
		if (unlikely(is_hugetlb_entry_hwpoisoned(pte)))
			continue;

D
David Gibson 已提交
2275
		page = pte_page(pte);
2276 2277
		if (pte_dirty(pte))
			set_page_dirty(page);
2278
		list_add(&page->lru, &page_list);
D
David Gibson 已提交
2279
	}
L
Linus Torvalds 已提交
2280
	spin_unlock(&mm->page_table_lock);
2281
	flush_tlb_range(vma, start, end);
A
Andrea Arcangeli 已提交
2282
	mmu_notifier_invalidate_range_end(mm, start, end);
2283
	list_for_each_entry_safe(page, tmp, &page_list, lru) {
2284
		page_remove_rmap(page);
2285 2286 2287
		list_del(&page->lru);
		put_page(page);
	}
L
Linus Torvalds 已提交
2288
}
D
David Gibson 已提交
2289

2290
void unmap_hugepage_range(struct vm_area_struct *vma, unsigned long start,
2291
			  unsigned long end, struct page *ref_page)
2292
{
2293
	mutex_lock(&vma->vm_file->f_mapping->i_mmap_mutex);
2294
	__unmap_hugepage_range(vma, start, end, ref_page);
2295
	mutex_unlock(&vma->vm_file->f_mapping->i_mmap_mutex);
2296 2297
}

2298 2299 2300 2301 2302 2303
/*
 * 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.
 */
2304 2305
static int unmap_ref_private(struct mm_struct *mm, struct vm_area_struct *vma,
				struct page *page, unsigned long address)
2306
{
2307
	struct hstate *h = hstate_vma(vma);
2308 2309 2310 2311 2312 2313 2314 2315 2316
	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.
	 */
2317
	address = address & huge_page_mask(h);
2318
	pgoff = vma_hugecache_offset(h, vma, address);
2319 2320
	mapping = (struct address_space *)page_private(page);

2321 2322 2323 2324 2325
	/*
	 * 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
	 */
2326
	mutex_lock(&mapping->i_mmap_mutex);
2327 2328 2329 2330 2331 2332 2333 2334 2335 2336 2337 2338 2339
	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))
2340
			__unmap_hugepage_range(iter_vma,
2341
				address, address + huge_page_size(h),
2342 2343
				page);
	}
2344
	mutex_unlock(&mapping->i_mmap_mutex);
2345 2346 2347 2348

	return 1;
}

2349 2350
/*
 * Hugetlb_cow() should be called with page lock of the original hugepage held.
2351 2352 2353
 * Called with hugetlb_instantiation_mutex held and pte_page locked so we
 * cannot race with other handlers or page migration.
 * Keep the pte_same checks anyway to make transition from the mutex easier.
2354
 */
2355
static int hugetlb_cow(struct mm_struct *mm, struct vm_area_struct *vma,
2356 2357
			unsigned long address, pte_t *ptep, pte_t pte,
			struct page *pagecache_page)
2358
{
2359
	struct hstate *h = hstate_vma(vma);
2360
	struct page *old_page, *new_page;
2361
	int avoidcopy;
2362
	int outside_reserve = 0;
2363 2364 2365

	old_page = pte_page(pte);

2366
retry_avoidcopy:
2367 2368
	/* If no-one else is actually using this page, avoid the copy
	 * and just make the page writable */
2369
	avoidcopy = (page_mapcount(old_page) == 1);
2370
	if (avoidcopy) {
2371 2372
		if (PageAnon(old_page))
			page_move_anon_rmap(old_page, vma, address);
2373
		set_huge_ptep_writable(vma, address, ptep);
N
Nick Piggin 已提交
2374
		return 0;
2375 2376
	}

2377 2378 2379 2380 2381 2382 2383 2384 2385
	/*
	 * 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.
	 */
2386
	if (!(vma->vm_flags & VM_MAYSHARE) &&
2387 2388 2389 2390
			is_vma_resv_set(vma, HPAGE_RESV_OWNER) &&
			old_page != pagecache_page)
		outside_reserve = 1;

2391
	page_cache_get(old_page);
2392 2393 2394

	/* Drop page_table_lock as buddy allocator may be called */
	spin_unlock(&mm->page_table_lock);
2395
	new_page = alloc_huge_page(vma, address, outside_reserve);
2396

2397
	if (IS_ERR(new_page)) {
2398
		page_cache_release(old_page);
2399 2400 2401 2402 2403 2404 2405 2406 2407 2408 2409 2410 2411

		/*
		 * 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));
2412
				spin_lock(&mm->page_table_lock);
2413 2414 2415 2416 2417 2418 2419 2420
				ptep = huge_pte_offset(mm, address & huge_page_mask(h));
				if (likely(pte_same(huge_ptep_get(ptep), pte)))
					goto retry_avoidcopy;
				/*
				 * race occurs while re-acquiring page_table_lock, and
				 * our job is done.
				 */
				return 0;
2421 2422 2423 2424
			}
			WARN_ON_ONCE(1);
		}

2425 2426
		/* Caller expects lock to be held */
		spin_lock(&mm->page_table_lock);
2427
		return -PTR_ERR(new_page);
2428 2429
	}

2430 2431 2432 2433
	/*
	 * When the original hugepage is shared one, it does not have
	 * anon_vma prepared.
	 */
2434
	if (unlikely(anon_vma_prepare(vma))) {
2435 2436
		page_cache_release(new_page);
		page_cache_release(old_page);
2437 2438
		/* Caller expects lock to be held */
		spin_lock(&mm->page_table_lock);
2439
		return VM_FAULT_OOM;
2440
	}
2441

A
Andrea Arcangeli 已提交
2442 2443
	copy_user_huge_page(new_page, old_page, address, vma,
			    pages_per_huge_page(h));
N
Nick Piggin 已提交
2444
	__SetPageUptodate(new_page);
2445

2446 2447 2448 2449 2450
	/*
	 * Retake the page_table_lock to check for racing updates
	 * before the page tables are altered
	 */
	spin_lock(&mm->page_table_lock);
2451
	ptep = huge_pte_offset(mm, address & huge_page_mask(h));
2452
	if (likely(pte_same(huge_ptep_get(ptep), pte))) {
2453
		/* Break COW */
2454 2455 2456
		mmu_notifier_invalidate_range_start(mm,
			address & huge_page_mask(h),
			(address & huge_page_mask(h)) + huge_page_size(h));
2457
		huge_ptep_clear_flush(vma, address, ptep);
2458 2459
		set_huge_pte_at(mm, address, ptep,
				make_huge_pte(vma, new_page, 1));
2460
		page_remove_rmap(old_page);
2461
		hugepage_add_new_anon_rmap(new_page, vma, address);
2462 2463
		/* Make the old page be freed below */
		new_page = old_page;
2464 2465 2466
		mmu_notifier_invalidate_range_end(mm,
			address & huge_page_mask(h),
			(address & huge_page_mask(h)) + huge_page_size(h));
2467 2468 2469
	}
	page_cache_release(new_page);
	page_cache_release(old_page);
N
Nick Piggin 已提交
2470
	return 0;
2471 2472
}

2473
/* Return the pagecache page at a given address within a VMA */
2474 2475
static struct page *hugetlbfs_pagecache_page(struct hstate *h,
			struct vm_area_struct *vma, unsigned long address)
2476 2477
{
	struct address_space *mapping;
2478
	pgoff_t idx;
2479 2480

	mapping = vma->vm_file->f_mapping;
2481
	idx = vma_hugecache_offset(h, vma, address);
2482 2483 2484 2485

	return find_lock_page(mapping, idx);
}

H
Hugh Dickins 已提交
2486 2487 2488 2489 2490
/*
 * 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
2491 2492 2493 2494 2495 2496 2497 2498 2499 2500 2501 2502 2503 2504 2505
			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;
}

2506
static int hugetlb_no_page(struct mm_struct *mm, struct vm_area_struct *vma,
2507
			unsigned long address, pte_t *ptep, unsigned int flags)
2508
{
2509
	struct hstate *h = hstate_vma(vma);
2510
	int ret = VM_FAULT_SIGBUS;
2511
	int anon_rmap = 0;
2512
	pgoff_t idx;
A
Adam Litke 已提交
2513 2514 2515
	unsigned long size;
	struct page *page;
	struct address_space *mapping;
2516
	pte_t new_pte;
A
Adam Litke 已提交
2517

2518 2519 2520
	/*
	 * Currently, we are forced to kill the process in the event the
	 * original mapper has unmapped pages from the child due to a failed
L
Lucas De Marchi 已提交
2521
	 * COW. Warn that such a situation has occurred as it may not be obvious
2522 2523 2524 2525 2526 2527 2528 2529
	 */
	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
2530
	mapping = vma->vm_file->f_mapping;
2531
	idx = vma_hugecache_offset(h, vma, address);
A
Adam Litke 已提交
2532 2533 2534 2535 2536

	/*
	 * Use page lock to guard against racing truncation
	 * before we get page_table_lock.
	 */
2537 2538 2539
retry:
	page = find_lock_page(mapping, idx);
	if (!page) {
2540
		size = i_size_read(mapping->host) >> huge_page_shift(h);
2541 2542
		if (idx >= size)
			goto out;
2543
		page = alloc_huge_page(vma, address, 0);
2544 2545
		if (IS_ERR(page)) {
			ret = -PTR_ERR(page);
2546 2547
			goto out;
		}
A
Andrea Arcangeli 已提交
2548
		clear_huge_page(page, address, pages_per_huge_page(h));
N
Nick Piggin 已提交
2549
		__SetPageUptodate(page);
2550

2551
		if (vma->vm_flags & VM_MAYSHARE) {
2552
			int err;
K
Ken Chen 已提交
2553
			struct inode *inode = mapping->host;
2554 2555 2556 2557 2558 2559 2560 2561

			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
2562 2563

			spin_lock(&inode->i_lock);
2564
			inode->i_blocks += blocks_per_huge_page(h);
K
Ken Chen 已提交
2565
			spin_unlock(&inode->i_lock);
2566
		} else {
2567
			lock_page(page);
2568 2569 2570 2571
			if (unlikely(anon_vma_prepare(vma))) {
				ret = VM_FAULT_OOM;
				goto backout_unlocked;
			}
2572
			anon_rmap = 1;
2573
		}
2574
	} else {
2575 2576 2577 2578 2579 2580
		/*
		 * 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))) {
2581
			ret = VM_FAULT_HWPOISON |
2582
			      VM_FAULT_SET_HINDEX(h - hstates);
2583 2584
			goto backout_unlocked;
		}
2585
	}
2586

2587 2588 2589 2590 2591 2592
	/*
	 * 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.
	 */
2593
	if ((flags & FAULT_FLAG_WRITE) && !(vma->vm_flags & VM_SHARED))
2594 2595 2596 2597
		if (vma_needs_reservation(h, vma, address) < 0) {
			ret = VM_FAULT_OOM;
			goto backout_unlocked;
		}
2598

2599
	spin_lock(&mm->page_table_lock);
2600
	size = i_size_read(mapping->host) >> huge_page_shift(h);
A
Adam Litke 已提交
2601 2602 2603
	if (idx >= size)
		goto backout;

N
Nick Piggin 已提交
2604
	ret = 0;
2605
	if (!huge_pte_none(huge_ptep_get(ptep)))
A
Adam Litke 已提交
2606 2607
		goto backout;

2608 2609 2610 2611
	if (anon_rmap)
		hugepage_add_new_anon_rmap(page, vma, address);
	else
		page_dup_rmap(page);
2612 2613 2614 2615
	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);

2616
	if ((flags & FAULT_FLAG_WRITE) && !(vma->vm_flags & VM_SHARED)) {
2617
		/* Optimization, do the COW without a second fault */
2618
		ret = hugetlb_cow(mm, vma, address, ptep, new_pte, page);
2619 2620
	}

2621
	spin_unlock(&mm->page_table_lock);
A
Adam Litke 已提交
2622 2623
	unlock_page(page);
out:
2624
	return ret;
A
Adam Litke 已提交
2625 2626 2627

backout:
	spin_unlock(&mm->page_table_lock);
2628
backout_unlocked:
A
Adam Litke 已提交
2629 2630 2631
	unlock_page(page);
	put_page(page);
	goto out;
2632 2633
}

2634
int hugetlb_fault(struct mm_struct *mm, struct vm_area_struct *vma,
2635
			unsigned long address, unsigned int flags)
2636 2637 2638
{
	pte_t *ptep;
	pte_t entry;
2639
	int ret;
2640
	struct page *page = NULL;
2641
	struct page *pagecache_page = NULL;
2642
	static DEFINE_MUTEX(hugetlb_instantiation_mutex);
2643
	struct hstate *h = hstate_vma(vma);
2644

2645 2646
	address &= huge_page_mask(h);

2647 2648 2649
	ptep = huge_pte_offset(mm, address);
	if (ptep) {
		entry = huge_ptep_get(ptep);
N
Naoya Horiguchi 已提交
2650 2651 2652 2653
		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)))
2654
			return VM_FAULT_HWPOISON_LARGE |
2655
			       VM_FAULT_SET_HINDEX(h - hstates);
2656 2657
	}

2658
	ptep = huge_pte_alloc(mm, address, huge_page_size(h));
2659 2660 2661
	if (!ptep)
		return VM_FAULT_OOM;

2662 2663 2664 2665 2666 2667
	/*
	 * 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);
2668 2669
	entry = huge_ptep_get(ptep);
	if (huge_pte_none(entry)) {
2670
		ret = hugetlb_no_page(mm, vma, address, ptep, flags);
2671
		goto out_mutex;
2672
	}
2673

N
Nick Piggin 已提交
2674
	ret = 0;
2675

2676 2677 2678 2679 2680 2681 2682 2683
	/*
	 * 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.
	 */
2684
	if ((flags & FAULT_FLAG_WRITE) && !pte_write(entry)) {
2685 2686
		if (vma_needs_reservation(h, vma, address) < 0) {
			ret = VM_FAULT_OOM;
2687
			goto out_mutex;
2688
		}
2689

2690
		if (!(vma->vm_flags & VM_MAYSHARE))
2691 2692 2693 2694
			pagecache_page = hugetlbfs_pagecache_page(h,
								vma, address);
	}

2695 2696 2697 2698 2699 2700 2701 2702 2703
	/*
	 * 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)
2704 2705
		lock_page(page);

2706 2707
	spin_lock(&mm->page_table_lock);
	/* Check for a racing update before calling hugetlb_cow */
2708 2709 2710 2711
	if (unlikely(!pte_same(entry, huge_ptep_get(ptep))))
		goto out_page_table_lock;


2712
	if (flags & FAULT_FLAG_WRITE) {
2713
		if (!pte_write(entry)) {
2714 2715
			ret = hugetlb_cow(mm, vma, address, ptep, entry,
							pagecache_page);
2716 2717 2718 2719 2720
			goto out_page_table_lock;
		}
		entry = pte_mkdirty(entry);
	}
	entry = pte_mkyoung(entry);
2721 2722
	if (huge_ptep_set_access_flags(vma, address, ptep, entry,
						flags & FAULT_FLAG_WRITE))
2723
		update_mmu_cache(vma, address, ptep);
2724 2725

out_page_table_lock:
2726
	spin_unlock(&mm->page_table_lock);
2727 2728 2729 2730 2731

	if (pagecache_page) {
		unlock_page(pagecache_page);
		put_page(pagecache_page);
	}
2732 2733
	if (page != pagecache_page)
		unlock_page(page);
2734

2735
out_mutex:
2736
	mutex_unlock(&hugetlb_instantiation_mutex);
2737 2738

	return ret;
2739 2740
}

A
Andi Kleen 已提交
2741 2742 2743 2744 2745 2746 2747 2748 2749
/* 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
2750 2751
int follow_hugetlb_page(struct mm_struct *mm, struct vm_area_struct *vma,
			struct page **pages, struct vm_area_struct **vmas,
2752
			unsigned long *position, int *length, int i,
H
Hugh Dickins 已提交
2753
			unsigned int flags)
D
David Gibson 已提交
2754
{
2755 2756
	unsigned long pfn_offset;
	unsigned long vaddr = *position;
D
David Gibson 已提交
2757
	int remainder = *length;
2758
	struct hstate *h = hstate_vma(vma);
D
David Gibson 已提交
2759

2760
	spin_lock(&mm->page_table_lock);
D
David Gibson 已提交
2761
	while (vaddr < vma->vm_end && remainder) {
A
Adam Litke 已提交
2762
		pte_t *pte;
H
Hugh Dickins 已提交
2763
		int absent;
A
Adam Litke 已提交
2764
		struct page *page;
D
David Gibson 已提交
2765

A
Adam Litke 已提交
2766 2767
		/*
		 * Some archs (sparc64, sh*) have multiple pte_ts to
H
Hugh Dickins 已提交
2768
		 * each hugepage.  We have to make sure we get the
A
Adam Litke 已提交
2769 2770
		 * first, for the page indexing below to work.
		 */
2771
		pte = huge_pte_offset(mm, vaddr & huge_page_mask(h));
H
Hugh Dickins 已提交
2772 2773 2774 2775
		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
2776 2777 2778 2779
		 * 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
2780
		 */
H
Hugh Dickins 已提交
2781 2782
		if (absent && (flags & FOLL_DUMP) &&
		    !hugetlbfs_pagecache_present(h, vma, vaddr)) {
H
Hugh Dickins 已提交
2783 2784 2785
			remainder = 0;
			break;
		}
D
David Gibson 已提交
2786

H
Hugh Dickins 已提交
2787 2788
		if (absent ||
		    ((flags & FOLL_WRITE) && !pte_write(huge_ptep_get(pte)))) {
A
Adam Litke 已提交
2789
			int ret;
D
David Gibson 已提交
2790

A
Adam Litke 已提交
2791
			spin_unlock(&mm->page_table_lock);
H
Hugh Dickins 已提交
2792 2793
			ret = hugetlb_fault(mm, vma, vaddr,
				(flags & FOLL_WRITE) ? FAULT_FLAG_WRITE : 0);
A
Adam Litke 已提交
2794
			spin_lock(&mm->page_table_lock);
2795
			if (!(ret & VM_FAULT_ERROR))
A
Adam Litke 已提交
2796
				continue;
D
David Gibson 已提交
2797

A
Adam Litke 已提交
2798 2799 2800 2801
			remainder = 0;
			break;
		}

2802
		pfn_offset = (vaddr & ~huge_page_mask(h)) >> PAGE_SHIFT;
2803
		page = pte_page(huge_ptep_get(pte));
2804
same_page:
2805
		if (pages) {
H
Hugh Dickins 已提交
2806
			pages[i] = mem_map_offset(page, pfn_offset);
K
KOSAKI Motohiro 已提交
2807
			get_page(pages[i]);
2808
		}
D
David Gibson 已提交
2809 2810 2811 2812 2813

		if (vmas)
			vmas[i] = vma;

		vaddr += PAGE_SIZE;
2814
		++pfn_offset;
D
David Gibson 已提交
2815 2816
		--remainder;
		++i;
2817
		if (vaddr < vma->vm_end && remainder &&
2818
				pfn_offset < pages_per_huge_page(h)) {
2819 2820 2821 2822 2823 2824
			/*
			 * We use pfn_offset to avoid touching the pageframes
			 * of this compound page.
			 */
			goto same_page;
		}
D
David Gibson 已提交
2825
	}
2826
	spin_unlock(&mm->page_table_lock);
D
David Gibson 已提交
2827 2828 2829
	*length = remainder;
	*position = vaddr;

H
Hugh Dickins 已提交
2830
	return i ? i : -EFAULT;
D
David Gibson 已提交
2831
}
2832 2833 2834 2835 2836 2837 2838 2839

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;
2840
	struct hstate *h = hstate_vma(vma);
2841 2842 2843 2844

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

2845
	mutex_lock(&vma->vm_file->f_mapping->i_mmap_mutex);
2846
	spin_lock(&mm->page_table_lock);
2847
	for (; address < end; address += huge_page_size(h)) {
2848 2849 2850
		ptep = huge_pte_offset(mm, address);
		if (!ptep)
			continue;
2851 2852
		if (huge_pmd_unshare(mm, &address, ptep))
			continue;
2853
		if (!huge_pte_none(huge_ptep_get(ptep))) {
2854 2855 2856 2857 2858 2859
			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);
2860
	mutex_unlock(&vma->vm_file->f_mapping->i_mmap_mutex);
2861 2862 2863 2864

	flush_tlb_range(vma, start, end);
}

2865 2866
int hugetlb_reserve_pages(struct inode *inode,
					long from, long to,
2867
					struct vm_area_struct *vma,
2868
					vm_flags_t vm_flags)
2869
{
2870
	long ret, chg;
2871
	struct hstate *h = hstate_inode(inode);
2872

2873 2874 2875 2876 2877
	/*
	 * 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
	 */
2878
	if (vm_flags & VM_NORESERVE)
2879 2880
		return 0;

2881 2882 2883 2884 2885 2886
	/*
	 * 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
	 */
2887
	if (!vma || vma->vm_flags & VM_MAYSHARE)
2888
		chg = region_chg(&inode->i_mapping->private_list, from, to);
2889 2890 2891 2892 2893
	else {
		struct resv_map *resv_map = resv_map_alloc();
		if (!resv_map)
			return -ENOMEM;

2894
		chg = to - from;
2895

2896 2897 2898 2899
		set_vma_resv_map(vma, resv_map);
		set_vma_resv_flags(vma, HPAGE_RESV_OWNER);
	}

2900 2901
	if (chg < 0)
		return chg;
2902

2903
	/* There must be enough filesystem quota for the mapping */
2904 2905
	if (hugetlb_get_quota(inode->i_mapping, chg))
		return -ENOSPC;
2906 2907

	/*
2908 2909
	 * Check enough hugepages are available for the reservation.
	 * Hand back the quota if there are not
2910
	 */
2911
	ret = hugetlb_acct_memory(h, chg);
K
Ken Chen 已提交
2912 2913
	if (ret < 0) {
		hugetlb_put_quota(inode->i_mapping, chg);
2914
		return ret;
K
Ken Chen 已提交
2915
	}
2916 2917 2918 2919 2920 2921 2922 2923 2924 2925 2926 2927

	/*
	 * 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
	 */
2928
	if (!vma || vma->vm_flags & VM_MAYSHARE)
2929
		region_add(&inode->i_mapping->private_list, from, to);
2930 2931 2932 2933 2934
	return 0;
}

void hugetlb_unreserve_pages(struct inode *inode, long offset, long freed)
{
2935
	struct hstate *h = hstate_inode(inode);
2936
	long chg = region_truncate(&inode->i_mapping->private_list, offset);
K
Ken Chen 已提交
2937 2938

	spin_lock(&inode->i_lock);
2939
	inode->i_blocks -= (blocks_per_huge_page(h) * freed);
K
Ken Chen 已提交
2940 2941
	spin_unlock(&inode->i_lock);

2942
	hugetlb_put_quota(inode->i_mapping, (chg - freed));
2943
	hugetlb_acct_memory(h, -(chg - freed));
2944
}
2945

2946 2947
#ifdef CONFIG_MEMORY_FAILURE

2948 2949 2950 2951 2952 2953 2954 2955 2956 2957 2958 2959 2960 2961
/* 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;
}

2962 2963 2964 2965
/*
 * This function is called from memory failure code.
 * Assume the caller holds page lock of the head page.
 */
2966
int dequeue_hwpoisoned_huge_page(struct page *hpage)
2967 2968 2969
{
	struct hstate *h = page_hstate(hpage);
	int nid = page_to_nid(hpage);
2970
	int ret = -EBUSY;
2971 2972

	spin_lock(&hugetlb_lock);
2973 2974
	if (is_hugepage_on_freelist(hpage)) {
		list_del(&hpage->lru);
2975
		set_page_refcounted(hpage);
2976 2977 2978 2979
		h->free_huge_pages--;
		h->free_huge_pages_node[nid]--;
		ret = 0;
	}
2980
	spin_unlock(&hugetlb_lock);
2981
	return ret;
2982
}
2983
#endif