hugetlb.c 97.8 KB
Newer Older
L
Linus Torvalds 已提交
1 2
/*
 * Generic hugetlb support.
3
 * (C) Nadia Yvette Chambers, April 2004
L
Linus Torvalds 已提交
4 5 6 7 8
 */
#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/compiler.h>
17
#include <linux/cpuset.h>
18
#include <linux/mutex.h>
19
#include <linux/bootmem.h>
20
#include <linux/sysfs.h>
21
#include <linux/slab.h>
22
#include <linux/rmap.h>
23 24
#include <linux/swap.h>
#include <linux/swapops.h>
25
#include <linux/page-isolation.h>
26
#include <linux/jhash.h>
27

D
David Gibson 已提交
28 29
#include <asm/page.h>
#include <asm/pgtable.h>
30
#include <asm/tlb.h>
D
David Gibson 已提交
31

32
#include <linux/io.h>
D
David Gibson 已提交
33
#include <linux/hugetlb.h>
34
#include <linux/hugetlb_cgroup.h>
35
#include <linux/node.h>
36
#include "internal.h"
L
Linus Torvalds 已提交
37 38

const unsigned long hugetlb_zero = 0, hugetlb_infinity = ~0UL;
39
unsigned long hugepages_treat_as_movable;
40

41
int hugetlb_max_hstate __read_mostly;
42 43 44
unsigned int default_hstate_idx;
struct hstate hstates[HUGE_MAX_HSTATE];

45 46
__initdata LIST_HEAD(huge_boot_pages);

47 48 49
/* for command line parsing */
static struct hstate * __initdata parsed_hstate;
static unsigned long __initdata default_hstate_max_huge_pages;
50
static unsigned long __initdata default_hstate_size;
51

52
/*
53 54
 * Protects updates to hugepage_freelists, hugepage_activelist, nr_huge_pages,
 * free_huge_pages, and surplus_huge_pages.
55
 */
56
DEFINE_SPINLOCK(hugetlb_lock);
57

58 59 60 61 62 63 64
/*
 * Serializes faults on the same logical page.  This is used to
 * prevent spurious OOMs when the hugepage pool is fully utilized.
 */
static int num_fault_mutexes;
static struct mutex *htlb_fault_mutex_table ____cacheline_aligned_in_smp;

65 66 67 68 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
static inline void unlock_or_release_subpool(struct hugepage_subpool *spool)
{
	bool free = (spool->count == 0) && (spool->used_hpages == 0);

	spin_unlock(&spool->lock);

	/* If no pages are used, and no other handles to the subpool
	 * remain, free the subpool the subpool remain */
	if (free)
		kfree(spool);
}

struct hugepage_subpool *hugepage_new_subpool(long nr_blocks)
{
	struct hugepage_subpool *spool;

	spool = kmalloc(sizeof(*spool), GFP_KERNEL);
	if (!spool)
		return NULL;

	spin_lock_init(&spool->lock);
	spool->count = 1;
	spool->max_hpages = nr_blocks;
	spool->used_hpages = 0;

	return spool;
}

void hugepage_put_subpool(struct hugepage_subpool *spool)
{
	spin_lock(&spool->lock);
	BUG_ON(!spool->count);
	spool->count--;
	unlock_or_release_subpool(spool);
}

static int hugepage_subpool_get_pages(struct hugepage_subpool *spool,
				      long delta)
{
	int ret = 0;

	if (!spool)
		return 0;

	spin_lock(&spool->lock);
	if ((spool->used_hpages + delta) <= spool->max_hpages) {
		spool->used_hpages += delta;
	} else {
		ret = -ENOMEM;
	}
	spin_unlock(&spool->lock);

	return ret;
}

static void hugepage_subpool_put_pages(struct hugepage_subpool *spool,
				       long delta)
{
	if (!spool)
		return;

	spin_lock(&spool->lock);
	spool->used_hpages -= delta;
	/* If hugetlbfs_put_super couldn't free spool due to
	* an outstanding quota reference, free it now. */
	unlock_or_release_subpool(spool);
}

static inline struct hugepage_subpool *subpool_inode(struct inode *inode)
{
	return HUGETLBFS_SB(inode->i_sb)->spool;
}

static inline struct hugepage_subpool *subpool_vma(struct vm_area_struct *vma)
{
A
Al Viro 已提交
140
	return subpool_inode(file_inode(vma->vm_file));
141 142
}

143 144 145
/*
 * Region tracking -- allows tracking of reservations and instantiated pages
 *                    across the pages in a mapping.
146
 *
147 148
 * The region data structures are embedded into a resv_map and
 * protected by a resv_map's lock
149 150 151 152 153 154 155
 */
struct file_region {
	struct list_head link;
	long from;
	long to;
};

156
static long region_add(struct resv_map *resv, long f, long t)
157
{
158
	struct list_head *head = &resv->regions;
159 160
	struct file_region *rg, *nrg, *trg;

161
	spin_lock(&resv->lock);
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
	/* 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;
191
	spin_unlock(&resv->lock);
192 193 194
	return 0;
}

195
static long region_chg(struct resv_map *resv, long f, long t)
196
{
197
	struct list_head *head = &resv->regions;
198
	struct file_region *rg, *nrg = NULL;
199 200
	long chg = 0;

201 202
retry:
	spin_lock(&resv->lock);
203 204 205 206 207 208 209 210 211
	/* 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) {
212 213 214 215 216 217 218 219 220 221 222
		if (!nrg) {
			spin_unlock(&resv->lock);
			nrg = kmalloc(sizeof(*nrg), GFP_KERNEL);
			if (!nrg)
				return -ENOMEM;

			nrg->from = f;
			nrg->to   = f;
			INIT_LIST_HEAD(&nrg->link);
			goto retry;
		}
223

224 225 226
		list_add(&nrg->link, rg->link.prev);
		chg = t - f;
		goto out_nrg;
227 228 229 230 231 232 233 234 235 236 237 238
	}

	/* 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)
239
			goto out;
240

L
Lucas De Marchi 已提交
241
		/* We overlap with this area, if it extends further than
242 243 244 245 246 247 248 249
		 * 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;
	}
250 251 252 253 254 255 256 257

out:
	spin_unlock(&resv->lock);
	/*  We already know we raced and no longer need the new region */
	kfree(nrg);
	return chg;
out_nrg:
	spin_unlock(&resv->lock);
258 259 260
	return chg;
}

261
static long region_truncate(struct resv_map *resv, long end)
262
{
263
	struct list_head *head = &resv->regions;
264 265 266
	struct file_region *rg, *trg;
	long chg = 0;

267
	spin_lock(&resv->lock);
268 269 270 271 272
	/* 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)
273
		goto out;
274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289

	/* 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);
	}
290 291 292

out:
	spin_unlock(&resv->lock);
293 294 295
	return chg;
}

296
static long region_count(struct resv_map *resv, long f, long t)
297
{
298
	struct list_head *head = &resv->regions;
299 300 301
	struct file_region *rg;
	long chg = 0;

302
	spin_lock(&resv->lock);
303 304
	/* Locate each segment we overlap with, and count that overlap. */
	list_for_each_entry(rg, head, link) {
305 306
		long seg_from;
		long seg_to;
307 308 309 310 311 312 313 314 315 316 317

		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;
	}
318
	spin_unlock(&resv->lock);
319 320 321 322

	return chg;
}

323 324 325 326
/*
 * Convert the address within this vma to the page offset within
 * the mapping, in pagecache page units; huge pages here.
 */
327 328
static pgoff_t vma_hugecache_offset(struct hstate *h,
			struct vm_area_struct *vma, unsigned long address)
329
{
330 331
	return ((address - vma->vm_start) >> huge_page_shift(h)) +
			(vma->vm_pgoff >> huge_page_order(h));
332 333
}

334 335 336 337 338 339
pgoff_t linear_hugepage_index(struct vm_area_struct *vma,
				     unsigned long address)
{
	return vma_hugecache_offset(hstate_vma(vma), vma, address);
}

340 341 342 343 344 345 346 347 348 349 350 351 352
/*
 * 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);

353
	return 1UL << huge_page_shift(hstate);
354
}
355
EXPORT_SYMBOL_GPL(vma_kernel_pagesize);
356

357 358 359 360 361 362 363 364 365 366 367 368 369
/*
 * 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

370 371 372 373 374 375 376
/*
 * 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)
377
#define HPAGE_RESV_MASK (HPAGE_RESV_OWNER | HPAGE_RESV_UNMAPPED)
378

379 380 381 382 383 384 385 386 387
/*
 * 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.
388 389 390 391 392 393 394 395 396
 *
 * 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.
397
 */
398 399 400 401 402 403 404 405 406 407 408
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;
}

409
struct resv_map *resv_map_alloc(void)
410 411 412 413 414 415
{
	struct resv_map *resv_map = kmalloc(sizeof(*resv_map), GFP_KERNEL);
	if (!resv_map)
		return NULL;

	kref_init(&resv_map->refs);
416
	spin_lock_init(&resv_map->lock);
417 418 419 420 421
	INIT_LIST_HEAD(&resv_map->regions);

	return resv_map;
}

422
void resv_map_release(struct kref *ref)
423 424 425 426
{
	struct resv_map *resv_map = container_of(ref, struct resv_map, refs);

	/* Clear out any active regions before we release the map. */
427
	region_truncate(resv_map, 0);
428 429 430
	kfree(resv_map);
}

431 432 433 434 435
static inline struct resv_map *inode_resv_map(struct inode *inode)
{
	return inode->i_mapping->private_data;
}

436
static struct resv_map *vma_resv_map(struct vm_area_struct *vma)
437 438
{
	VM_BUG_ON(!is_vm_hugetlb_page(vma));
439 440 441 442 443 444 445
	if (vma->vm_flags & VM_MAYSHARE) {
		struct address_space *mapping = vma->vm_file->f_mapping;
		struct inode *inode = mapping->host;

		return inode_resv_map(inode);

	} else {
446 447
		return (struct resv_map *)(get_vma_private_data(vma) &
							~HPAGE_RESV_MASK);
448
	}
449 450
}

451
static void set_vma_resv_map(struct vm_area_struct *vma, struct resv_map *map)
452 453
{
	VM_BUG_ON(!is_vm_hugetlb_page(vma));
454
	VM_BUG_ON(vma->vm_flags & VM_MAYSHARE);
455

456 457
	set_vma_private_data(vma, (get_vma_private_data(vma) &
				HPAGE_RESV_MASK) | (unsigned long)map);
458 459 460 461 462
}

static void set_vma_resv_flags(struct vm_area_struct *vma, unsigned long flags)
{
	VM_BUG_ON(!is_vm_hugetlb_page(vma));
463
	VM_BUG_ON(vma->vm_flags & VM_MAYSHARE);
464 465

	set_vma_private_data(vma, get_vma_private_data(vma) | flags);
466 467 468 469 470
}

static int is_vma_resv_set(struct vm_area_struct *vma, unsigned long flag)
{
	VM_BUG_ON(!is_vm_hugetlb_page(vma));
471 472

	return (get_vma_private_data(vma) & flag) != 0;
473 474
}

475
/* Reset counters to 0 and clear all HPAGE_RESV_* flags */
476 477 478
void reset_vma_resv_huge_pages(struct vm_area_struct *vma)
{
	VM_BUG_ON(!is_vm_hugetlb_page(vma));
479
	if (!(vma->vm_flags & VM_MAYSHARE))
480 481 482 483
		vma->vm_private_data = (void *)0;
}

/* Returns true if the VMA has associated reserve pages */
484
static int vma_has_reserves(struct vm_area_struct *vma, long chg)
485
{
486 487 488 489 490 491 492 493 494 495 496 497 498 499 500
	if (vma->vm_flags & VM_NORESERVE) {
		/*
		 * This address is already reserved by other process(chg == 0),
		 * so, we should decrement reserved count. Without decrementing,
		 * reserve count remains after releasing inode, because this
		 * allocated page will go into page cache and is regarded as
		 * coming from reserved pool in releasing step.  Currently, we
		 * don't have any other solution to deal with this situation
		 * properly, so add work-around here.
		 */
		if (vma->vm_flags & VM_MAYSHARE && chg == 0)
			return 1;
		else
			return 0;
	}
501 502

	/* Shared mappings always use reserves */
503
	if (vma->vm_flags & VM_MAYSHARE)
504
		return 1;
505 506 507 508 509

	/*
	 * Only the process that called mmap() has reserves for
	 * private mappings.
	 */
510 511
	if (is_vma_resv_set(vma, HPAGE_RESV_OWNER))
		return 1;
512

513
	return 0;
514 515
}

516
static void enqueue_huge_page(struct hstate *h, struct page *page)
L
Linus Torvalds 已提交
517 518
{
	int nid = page_to_nid(page);
519
	list_move(&page->lru, &h->hugepage_freelists[nid]);
520 521
	h->free_huge_pages++;
	h->free_huge_pages_node[nid]++;
L
Linus Torvalds 已提交
522 523
}

524 525 526 527
static struct page *dequeue_huge_page_node(struct hstate *h, int nid)
{
	struct page *page;

528 529 530 531 532 533 534 535
	list_for_each_entry(page, &h->hugepage_freelists[nid], lru)
		if (!is_migrate_isolate_page(page))
			break;
	/*
	 * if 'non-isolated free hugepage' not found on the list,
	 * the allocation fails.
	 */
	if (&h->hugepage_freelists[nid] == &page->lru)
536
		return NULL;
537
	list_move(&page->lru, &h->hugepage_activelist);
538
	set_page_refcounted(page);
539 540 541 542 543
	h->free_huge_pages--;
	h->free_huge_pages_node[nid]--;
	return page;
}

544 545 546 547 548 549 550 551 552
/* Movability of hugepages depends on migration support. */
static inline gfp_t htlb_alloc_mask(struct hstate *h)
{
	if (hugepages_treat_as_movable || hugepage_migration_support(h))
		return GFP_HIGHUSER_MOVABLE;
	else
		return GFP_HIGHUSER;
}

553 554
static struct page *dequeue_huge_page_vma(struct hstate *h,
				struct vm_area_struct *vma,
555 556
				unsigned long address, int avoid_reserve,
				long chg)
L
Linus Torvalds 已提交
557
{
558
	struct page *page = NULL;
559
	struct mempolicy *mpol;
560
	nodemask_t *nodemask;
561
	struct zonelist *zonelist;
562 563
	struct zone *zone;
	struct zoneref *z;
564
	unsigned int cpuset_mems_cookie;
L
Linus Torvalds 已提交
565

566 567 568 569 570
	/*
	 * 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
	 */
571
	if (!vma_has_reserves(vma, chg) &&
572
			h->free_huge_pages - h->resv_huge_pages == 0)
573
		goto err;
574

575
	/* If reserves cannot be used, ensure enough pages are in the pool */
576
	if (avoid_reserve && h->free_huge_pages - h->resv_huge_pages == 0)
577
		goto err;
578

579
retry_cpuset:
580
	cpuset_mems_cookie = read_mems_allowed_begin();
581
	zonelist = huge_zonelist(vma, address,
582
					htlb_alloc_mask(h), &mpol, &nodemask);
583

584 585
	for_each_zone_zonelist_nodemask(zone, z, zonelist,
						MAX_NR_ZONES - 1, nodemask) {
586
		if (cpuset_zone_allowed_softwall(zone, htlb_alloc_mask(h))) {
587 588
			page = dequeue_huge_page_node(h, zone_to_nid(zone));
			if (page) {
589 590 591 592 593
				if (avoid_reserve)
					break;
				if (!vma_has_reserves(vma, chg))
					break;

594
				SetPagePrivate(page);
595
				h->resv_huge_pages--;
596 597
				break;
			}
A
Andrew Morton 已提交
598
		}
L
Linus Torvalds 已提交
599
	}
600

601
	mpol_cond_put(mpol);
602
	if (unlikely(!page && read_mems_allowed_retry(cpuset_mems_cookie)))
603
		goto retry_cpuset;
L
Linus Torvalds 已提交
604
	return page;
605 606 607

err:
	return NULL;
L
Linus Torvalds 已提交
608 609
}

610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682
/*
 * 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.
 */
static int next_node_allowed(int nid, nodemask_t *nodes_allowed)
{
	nid = next_node(nid, *nodes_allowed);
	if (nid == MAX_NUMNODES)
		nid = first_node(*nodes_allowed);
	VM_BUG_ON(nid >= MAX_NUMNODES);

	return nid;
}

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;
}

/*
 * 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.
 */
static int hstate_next_node_to_alloc(struct hstate *h,
					nodemask_t *nodes_allowed)
{
	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);

	return nid;
}

/*
 * 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.
 */
static int hstate_next_node_to_free(struct hstate *h, nodemask_t *nodes_allowed)
{
	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);

	return nid;
}

#define for_each_node_mask_to_alloc(hs, nr_nodes, node, mask)		\
	for (nr_nodes = nodes_weight(*mask);				\
		nr_nodes > 0 &&						\
		((node = hstate_next_node_to_alloc(hs, mask)) || 1);	\
		nr_nodes--)

#define for_each_node_mask_to_free(hs, nr_nodes, node, mask)		\
	for (nr_nodes = nodes_weight(*mask);				\
		nr_nodes > 0 &&						\
		((node = hstate_next_node_to_free(hs, mask)) || 1);	\
		nr_nodes--)

683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820
#if defined(CONFIG_CMA) && defined(CONFIG_X86_64)
static void destroy_compound_gigantic_page(struct page *page,
					unsigned long order)
{
	int i;
	int nr_pages = 1 << order;
	struct page *p = page + 1;

	for (i = 1; i < nr_pages; i++, p = mem_map_next(p, page, i)) {
		__ClearPageTail(p);
		set_page_refcounted(p);
		p->first_page = NULL;
	}

	set_compound_order(page, 0);
	__ClearPageHead(page);
}

static void free_gigantic_page(struct page *page, unsigned order)
{
	free_contig_range(page_to_pfn(page), 1 << order);
}

static int __alloc_gigantic_page(unsigned long start_pfn,
				unsigned long nr_pages)
{
	unsigned long end_pfn = start_pfn + nr_pages;
	return alloc_contig_range(start_pfn, end_pfn, MIGRATE_MOVABLE);
}

static bool pfn_range_valid_gigantic(unsigned long start_pfn,
				unsigned long nr_pages)
{
	unsigned long i, end_pfn = start_pfn + nr_pages;
	struct page *page;

	for (i = start_pfn; i < end_pfn; i++) {
		if (!pfn_valid(i))
			return false;

		page = pfn_to_page(i);

		if (PageReserved(page))
			return false;

		if (page_count(page) > 0)
			return false;

		if (PageHuge(page))
			return false;
	}

	return true;
}

static bool zone_spans_last_pfn(const struct zone *zone,
			unsigned long start_pfn, unsigned long nr_pages)
{
	unsigned long last_pfn = start_pfn + nr_pages - 1;
	return zone_spans_pfn(zone, last_pfn);
}

static struct page *alloc_gigantic_page(int nid, unsigned order)
{
	unsigned long nr_pages = 1 << order;
	unsigned long ret, pfn, flags;
	struct zone *z;

	z = NODE_DATA(nid)->node_zones;
	for (; z - NODE_DATA(nid)->node_zones < MAX_NR_ZONES; z++) {
		spin_lock_irqsave(&z->lock, flags);

		pfn = ALIGN(z->zone_start_pfn, nr_pages);
		while (zone_spans_last_pfn(z, pfn, nr_pages)) {
			if (pfn_range_valid_gigantic(pfn, nr_pages)) {
				/*
				 * We release the zone lock here because
				 * alloc_contig_range() will also lock the zone
				 * at some point. If there's an allocation
				 * spinning on this lock, it may win the race
				 * and cause alloc_contig_range() to fail...
				 */
				spin_unlock_irqrestore(&z->lock, flags);
				ret = __alloc_gigantic_page(pfn, nr_pages);
				if (!ret)
					return pfn_to_page(pfn);
				spin_lock_irqsave(&z->lock, flags);
			}
			pfn += nr_pages;
		}

		spin_unlock_irqrestore(&z->lock, flags);
	}

	return NULL;
}

static void prep_new_huge_page(struct hstate *h, struct page *page, int nid);
static void prep_compound_gigantic_page(struct page *page, unsigned long order);

static struct page *alloc_fresh_gigantic_page_node(struct hstate *h, int nid)
{
	struct page *page;

	page = alloc_gigantic_page(nid, huge_page_order(h));
	if (page) {
		prep_compound_gigantic_page(page, huge_page_order(h));
		prep_new_huge_page(h, page, nid);
	}

	return page;
}

static int alloc_fresh_gigantic_page(struct hstate *h,
				nodemask_t *nodes_allowed)
{
	struct page *page = NULL;
	int nr_nodes, node;

	for_each_node_mask_to_alloc(h, nr_nodes, node, nodes_allowed) {
		page = alloc_fresh_gigantic_page_node(h, node);
		if (page)
			return 1;
	}

	return 0;
}

static inline bool gigantic_page_supported(void) { return true; }
#else
static inline bool gigantic_page_supported(void) { return false; }
static inline void free_gigantic_page(struct page *page, unsigned order) { }
static inline void destroy_compound_gigantic_page(struct page *page,
						unsigned long order) { }
static inline int alloc_fresh_gigantic_page(struct hstate *h,
					nodemask_t *nodes_allowed) { return 0; }
#endif

821
static void update_and_free_page(struct hstate *h, struct page *page)
A
Adam Litke 已提交
822 823
{
	int i;
824

825 826
	if (hstate_is_gigantic(h) && !gigantic_page_supported())
		return;
827

828 829 830
	h->nr_huge_pages--;
	h->nr_huge_pages_node[page_to_nid(page)]--;
	for (i = 0; i < pages_per_huge_page(h); i++) {
831 832
		page[i].flags &= ~(1 << PG_locked | 1 << PG_error |
				1 << PG_referenced | 1 << PG_dirty |
833 834
				1 << PG_active | 1 << PG_private |
				1 << PG_writeback);
A
Adam Litke 已提交
835
	}
836
	VM_BUG_ON_PAGE(hugetlb_cgroup_from_page(page), page);
A
Adam Litke 已提交
837 838
	set_compound_page_dtor(page, NULL);
	set_page_refcounted(page);
839 840 841 842 843 844 845
	if (hstate_is_gigantic(h)) {
		destroy_compound_gigantic_page(page, huge_page_order(h));
		free_gigantic_page(page, huge_page_order(h));
	} else {
		arch_release_hugepage(page);
		__free_pages(page, huge_page_order(h));
	}
A
Adam Litke 已提交
846 847
}

848 849 850 851 852 853 854 855 856 857 858
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;
}

859 860
static void free_huge_page(struct page *page)
{
861 862 863 864
	/*
	 * Can't pass hstate in here because it is called from the
	 * compound page destructor.
	 */
865
	struct hstate *h = page_hstate(page);
866
	int nid = page_to_nid(page);
867 868
	struct hugepage_subpool *spool =
		(struct hugepage_subpool *)page_private(page);
869
	bool restore_reserve;
870

871
	set_page_private(page, 0);
872
	page->mapping = NULL;
873
	BUG_ON(page_count(page));
874
	BUG_ON(page_mapcount(page));
875
	restore_reserve = PagePrivate(page);
876
	ClearPagePrivate(page);
877 878

	spin_lock(&hugetlb_lock);
879 880
	hugetlb_cgroup_uncharge_page(hstate_index(h),
				     pages_per_huge_page(h), page);
881 882 883
	if (restore_reserve)
		h->resv_huge_pages++;

884
	if (h->surplus_huge_pages_node[nid]) {
885 886
		/* remove the page from active list */
		list_del(&page->lru);
887 888 889
		update_and_free_page(h, page);
		h->surplus_huge_pages--;
		h->surplus_huge_pages_node[nid]--;
890
	} else {
891
		arch_clear_hugepage_flags(page);
892
		enqueue_huge_page(h, page);
893
	}
894
	spin_unlock(&hugetlb_lock);
895
	hugepage_subpool_put_pages(spool, 1);
896 897
}

898
static void prep_new_huge_page(struct hstate *h, struct page *page, int nid)
899
{
900
	INIT_LIST_HEAD(&page->lru);
901 902
	set_compound_page_dtor(page, free_huge_page);
	spin_lock(&hugetlb_lock);
903
	set_hugetlb_cgroup(page, NULL);
904 905
	h->nr_huge_pages++;
	h->nr_huge_pages_node[nid]++;
906 907 908 909
	spin_unlock(&hugetlb_lock);
	put_page(page); /* free it into the hugepage allocator */
}

910
static void prep_compound_gigantic_page(struct page *page, unsigned long order)
911 912 913 914 915 916 917 918
{
	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);
919
	__ClearPageReserved(page);
920 921
	for (i = 1; i < nr_pages; i++, p = mem_map_next(p, page, i)) {
		__SetPageTail(p);
922 923 924 925 926 927 928 929 930 931 932 933 934
		/*
		 * For gigantic hugepages allocated through bootmem at
		 * boot, it's safer to be consistent with the not-gigantic
		 * hugepages and clear the PG_reserved bit from all tail pages
		 * too.  Otherwse drivers using get_user_pages() to access tail
		 * pages may get the reference counting wrong if they see
		 * PG_reserved set on a tail page (despite the head page not
		 * having PG_reserved set).  Enforcing this consistency between
		 * head and tail pages allows drivers to optimize away a check
		 * on the head page when they need know if put_page() is needed
		 * after get_user_pages().
		 */
		__ClearPageReserved(p);
935
		set_page_count(p, 0);
936 937 938 939
		p->first_page = page;
	}
}

A
Andrew Morton 已提交
940 941 942 943 944
/*
 * PageHuge() only returns true for hugetlbfs pages, but not for normal or
 * transparent huge pages.  See the PageTransHuge() documentation for more
 * details.
 */
945 946 947 948 949 950
int PageHuge(struct page *page)
{
	if (!PageCompound(page))
		return 0;

	page = compound_head(page);
951
	return get_compound_page_dtor(page) == free_huge_page;
952
}
953 954
EXPORT_SYMBOL_GPL(PageHuge);

955 956 957 958 959 960 961 962 963
/*
 * PageHeadHuge() only returns true for hugetlbfs head page, but not for
 * normal or transparent huge pages.
 */
int PageHeadHuge(struct page *page_head)
{
	if (!PageHead(page_head))
		return 0;

964
	return get_compound_page_dtor(page_head) == free_huge_page;
965 966
}

967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983
pgoff_t __basepage_index(struct page *page)
{
	struct page *page_head = compound_head(page);
	pgoff_t index = page_index(page_head);
	unsigned long compound_idx;

	if (!PageHuge(page_head))
		return page_index(page);

	if (compound_order(page_head) >= MAX_ORDER)
		compound_idx = page_to_pfn(page) - page_to_pfn(page_head);
	else
		compound_idx = page - page_head;

	return (index << compound_order(page_head)) + compound_idx;
}

984
static struct page *alloc_fresh_huge_page_node(struct hstate *h, int nid)
L
Linus Torvalds 已提交
985 986
{
	struct page *page;
987

988
	page = alloc_pages_exact_node(nid,
989
		htlb_alloc_mask(h)|__GFP_COMP|__GFP_THISNODE|
990
						__GFP_REPEAT|__GFP_NOWARN,
991
		huge_page_order(h));
L
Linus Torvalds 已提交
992
	if (page) {
993
		if (arch_prepare_hugepage(page)) {
994
			__free_pages(page, huge_page_order(h));
995
			return NULL;
996
		}
997
		prep_new_huge_page(h, page, nid);
L
Linus Torvalds 已提交
998
	}
999 1000 1001 1002

	return page;
}

1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024
static int alloc_fresh_huge_page(struct hstate *h, nodemask_t *nodes_allowed)
{
	struct page *page;
	int nr_nodes, node;
	int ret = 0;

	for_each_node_mask_to_alloc(h, nr_nodes, node, nodes_allowed) {
		page = alloc_fresh_huge_page_node(h, node);
		if (page) {
			ret = 1;
			break;
		}
	}

	if (ret)
		count_vm_event(HTLB_BUDDY_PGALLOC);
	else
		count_vm_event(HTLB_BUDDY_PGALLOC_FAIL);

	return ret;
}

1025 1026 1027 1028 1029 1030
/*
 * 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.
 */
1031 1032
static int free_pool_huge_page(struct hstate *h, nodemask_t *nodes_allowed,
							 bool acct_surplus)
1033
{
1034
	int nr_nodes, node;
1035 1036
	int ret = 0;

1037
	for_each_node_mask_to_free(h, nr_nodes, node, nodes_allowed) {
1038 1039 1040 1041
		/*
		 * If we're returning unused surplus pages, only examine
		 * nodes with surplus pages.
		 */
1042 1043
		if ((!acct_surplus || h->surplus_huge_pages_node[node]) &&
		    !list_empty(&h->hugepage_freelists[node])) {
1044
			struct page *page =
1045
				list_entry(h->hugepage_freelists[node].next,
1046 1047 1048
					  struct page, lru);
			list_del(&page->lru);
			h->free_huge_pages--;
1049
			h->free_huge_pages_node[node]--;
1050 1051
			if (acct_surplus) {
				h->surplus_huge_pages--;
1052
				h->surplus_huge_pages_node[node]--;
1053
			}
1054 1055
			update_and_free_page(h, page);
			ret = 1;
1056
			break;
1057
		}
1058
	}
1059 1060 1061 1062

	return ret;
}

1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100
/*
 * Dissolve a given free hugepage into free buddy pages. This function does
 * nothing for in-use (including surplus) hugepages.
 */
static void dissolve_free_huge_page(struct page *page)
{
	spin_lock(&hugetlb_lock);
	if (PageHuge(page) && !page_count(page)) {
		struct hstate *h = page_hstate(page);
		int nid = page_to_nid(page);
		list_del(&page->lru);
		h->free_huge_pages--;
		h->free_huge_pages_node[nid]--;
		update_and_free_page(h, page);
	}
	spin_unlock(&hugetlb_lock);
}

/*
 * Dissolve free hugepages in a given pfn range. Used by memory hotplug to
 * make specified memory blocks removable from the system.
 * Note that start_pfn should aligned with (minimum) hugepage size.
 */
void dissolve_free_huge_pages(unsigned long start_pfn, unsigned long end_pfn)
{
	unsigned int order = 8 * sizeof(void *);
	unsigned long pfn;
	struct hstate *h;

	/* Set scan step to minimum hugepage size */
	for_each_hstate(h)
		if (order > huge_page_order(h))
			order = huge_page_order(h);
	VM_BUG_ON(!IS_ALIGNED(start_pfn, 1 << order));
	for (pfn = start_pfn; pfn < end_pfn; pfn += 1 << order)
		dissolve_free_huge_page(pfn_to_page(pfn));
}

1101
static struct page *alloc_buddy_huge_page(struct hstate *h, int nid)
1102 1103
{
	struct page *page;
1104
	unsigned int r_nid;
1105

1106
	if (hstate_is_gigantic(h))
1107 1108
		return NULL;

1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132
	/*
	 * 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);
1133
	if (h->surplus_huge_pages >= h->nr_overcommit_huge_pages) {
1134 1135 1136
		spin_unlock(&hugetlb_lock);
		return NULL;
	} else {
1137 1138
		h->nr_huge_pages++;
		h->surplus_huge_pages++;
1139 1140 1141
	}
	spin_unlock(&hugetlb_lock);

1142
	if (nid == NUMA_NO_NODE)
1143
		page = alloc_pages(htlb_alloc_mask(h)|__GFP_COMP|
1144 1145 1146 1147
				   __GFP_REPEAT|__GFP_NOWARN,
				   huge_page_order(h));
	else
		page = alloc_pages_exact_node(nid,
1148
			htlb_alloc_mask(h)|__GFP_COMP|__GFP_THISNODE|
1149
			__GFP_REPEAT|__GFP_NOWARN, huge_page_order(h));
1150

1151 1152
	if (page && arch_prepare_hugepage(page)) {
		__free_pages(page, huge_page_order(h));
1153
		page = NULL;
1154 1155
	}

1156
	spin_lock(&hugetlb_lock);
1157
	if (page) {
1158
		INIT_LIST_HEAD(&page->lru);
1159
		r_nid = page_to_nid(page);
1160
		set_compound_page_dtor(page, free_huge_page);
1161
		set_hugetlb_cgroup(page, NULL);
1162 1163 1164
		/*
		 * We incremented the global counters already
		 */
1165 1166
		h->nr_huge_pages_node[r_nid]++;
		h->surplus_huge_pages_node[r_nid]++;
1167
		__count_vm_event(HTLB_BUDDY_PGALLOC);
1168
	} else {
1169 1170
		h->nr_huge_pages--;
		h->surplus_huge_pages--;
1171
		__count_vm_event(HTLB_BUDDY_PGALLOC_FAIL);
1172
	}
1173
	spin_unlock(&hugetlb_lock);
1174 1175 1176 1177

	return page;
}

1178 1179 1180 1181 1182 1183 1184
/*
 * 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)
{
1185
	struct page *page = NULL;
1186 1187

	spin_lock(&hugetlb_lock);
1188 1189
	if (h->free_huge_pages - h->resv_huge_pages > 0)
		page = dequeue_huge_page_node(h, nid);
1190 1191
	spin_unlock(&hugetlb_lock);

1192
	if (!page)
1193 1194 1195 1196 1197
		page = alloc_buddy_huge_page(h, nid);

	return page;
}

1198
/*
L
Lucas De Marchi 已提交
1199
 * Increase the hugetlb pool such that it can accommodate a reservation
1200 1201
 * of size 'delta'.
 */
1202
static int gather_surplus_pages(struct hstate *h, int delta)
1203 1204 1205 1206 1207
{
	struct list_head surplus_list;
	struct page *page, *tmp;
	int ret, i;
	int needed, allocated;
1208
	bool alloc_ok = true;
1209

1210
	needed = (h->resv_huge_pages + delta) - h->free_huge_pages;
1211
	if (needed <= 0) {
1212
		h->resv_huge_pages += delta;
1213
		return 0;
1214
	}
1215 1216 1217 1218 1219 1220 1221 1222

	allocated = 0;
	INIT_LIST_HEAD(&surplus_list);

	ret = -ENOMEM;
retry:
	spin_unlock(&hugetlb_lock);
	for (i = 0; i < needed; i++) {
1223
		page = alloc_buddy_huge_page(h, NUMA_NO_NODE);
1224 1225 1226 1227
		if (!page) {
			alloc_ok = false;
			break;
		}
1228 1229
		list_add(&page->lru, &surplus_list);
	}
1230
	allocated += i;
1231 1232 1233 1234 1235 1236

	/*
	 * 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);
1237 1238
	needed = (h->resv_huge_pages + delta) -
			(h->free_huge_pages + allocated);
1239 1240 1241 1242 1243 1244 1245 1246 1247 1248
	if (needed > 0) {
		if (alloc_ok)
			goto retry;
		/*
		 * We were not able to allocate enough pages to
		 * satisfy the entire reservation so we free what
		 * we've allocated so far.
		 */
		goto free;
	}
1249 1250
	/*
	 * The surplus_list now contains _at_least_ the number of extra pages
L
Lucas De Marchi 已提交
1251
	 * needed to accommodate the reservation.  Add the appropriate number
1252
	 * of pages to the hugetlb pool and free the extras back to the buddy
1253 1254 1255
	 * 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.
1256 1257
	 */
	needed += allocated;
1258
	h->resv_huge_pages += delta;
1259
	ret = 0;
1260

1261
	/* Free the needed pages to the hugetlb pool */
1262
	list_for_each_entry_safe(page, tmp, &surplus_list, lru) {
1263 1264
		if ((--needed) < 0)
			break;
1265 1266 1267 1268 1269
		/*
		 * This page is now managed by the hugetlb allocator and has
		 * no users -- drop the buddy allocator's reference.
		 */
		put_page_testzero(page);
1270
		VM_BUG_ON_PAGE(page_count(page), page);
1271
		enqueue_huge_page(h, page);
1272
	}
1273
free:
1274
	spin_unlock(&hugetlb_lock);
1275 1276

	/* Free unnecessary surplus pages to the buddy allocator */
1277 1278
	list_for_each_entry_safe(page, tmp, &surplus_list, lru)
		put_page(page);
1279
	spin_lock(&hugetlb_lock);
1280 1281 1282 1283 1284 1285 1286 1287

	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.
1288
 * Called with hugetlb_lock held.
1289
 */
1290 1291
static void return_unused_surplus_pages(struct hstate *h,
					unsigned long unused_resv_pages)
1292 1293 1294
{
	unsigned long nr_pages;

1295
	/* Uncommit the reservation */
1296
	h->resv_huge_pages -= unused_resv_pages;
1297

1298
	/* Cannot return gigantic pages currently */
1299
	if (hstate_is_gigantic(h))
1300 1301
		return;

1302
	nr_pages = min(unused_resv_pages, h->surplus_huge_pages);
1303

1304 1305
	/*
	 * We want to release as many surplus pages as possible, spread
1306 1307 1308 1309 1310
	 * 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.
1311 1312
	 */
	while (nr_pages--) {
1313
		if (!free_pool_huge_page(h, &node_states[N_MEMORY], 1))
1314
			break;
1315
		cond_resched_lock(&hugetlb_lock);
1316 1317 1318
	}
}

1319 1320 1321
/*
 * Determine if the huge page at addr within the vma has an associated
 * reservation.  Where it does not we will need to logically increase
1322 1323 1324 1325 1326 1327
 * reservation and actually increase subpool usage 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 allocated from the subpool and instantiated the change should
 * be committed via vma_commit_reservation.  No action is required on
 * failure.
1328
 */
1329
static long vma_needs_reservation(struct hstate *h,
1330
			struct vm_area_struct *vma, unsigned long addr)
1331
{
1332 1333 1334
	struct resv_map *resv;
	pgoff_t idx;
	long chg;
1335

1336 1337
	resv = vma_resv_map(vma);
	if (!resv)
1338
		return 1;
1339

1340 1341
	idx = vma_hugecache_offset(h, vma, addr);
	chg = region_chg(resv, idx, idx + 1);
1342

1343 1344 1345 1346
	if (vma->vm_flags & VM_MAYSHARE)
		return chg;
	else
		return chg < 0 ? chg : 0;
1347
}
1348 1349
static void vma_commit_reservation(struct hstate *h,
			struct vm_area_struct *vma, unsigned long addr)
1350
{
1351 1352
	struct resv_map *resv;
	pgoff_t idx;
1353

1354 1355 1356
	resv = vma_resv_map(vma);
	if (!resv)
		return;
1357

1358 1359
	idx = vma_hugecache_offset(h, vma, addr);
	region_add(resv, idx, idx + 1);
1360 1361
}

1362
static struct page *alloc_huge_page(struct vm_area_struct *vma,
1363
				    unsigned long addr, int avoid_reserve)
L
Linus Torvalds 已提交
1364
{
1365
	struct hugepage_subpool *spool = subpool_vma(vma);
1366
	struct hstate *h = hstate_vma(vma);
1367
	struct page *page;
1368
	long chg;
1369 1370
	int ret, idx;
	struct hugetlb_cgroup *h_cg;
1371

1372
	idx = hstate_index(h);
1373
	/*
1374 1375 1376 1377 1378 1379
	 * Processes that did not create the mapping will have no
	 * reserves and will not have accounted against subpool
	 * limit. Check that the subpool limit can be made before
	 * satisfying the allocation MAP_NORESERVE mappings may also
	 * need pages and subpool limit allocated allocated if no reserve
	 * mapping overlaps.
1380
	 */
1381
	chg = vma_needs_reservation(h, vma, addr);
1382
	if (chg < 0)
1383
		return ERR_PTR(-ENOMEM);
1384 1385
	if (chg || avoid_reserve)
		if (hugepage_subpool_get_pages(spool, 1))
1386
			return ERR_PTR(-ENOSPC);
L
Linus Torvalds 已提交
1387

1388
	ret = hugetlb_cgroup_charge_cgroup(idx, pages_per_huge_page(h), &h_cg);
1389 1390 1391
	if (ret)
		goto out_subpool_put;

L
Linus Torvalds 已提交
1392
	spin_lock(&hugetlb_lock);
1393
	page = dequeue_huge_page_vma(h, vma, addr, avoid_reserve, chg);
1394
	if (!page) {
1395
		spin_unlock(&hugetlb_lock);
1396
		page = alloc_buddy_huge_page(h, NUMA_NO_NODE);
1397 1398 1399
		if (!page)
			goto out_uncharge_cgroup;

1400 1401
		spin_lock(&hugetlb_lock);
		list_move(&page->lru, &h->hugepage_activelist);
1402
		/* Fall through */
K
Ken Chen 已提交
1403
	}
1404 1405
	hugetlb_cgroup_commit_charge(idx, pages_per_huge_page(h), h_cg, page);
	spin_unlock(&hugetlb_lock);
1406

1407
	set_page_private(page, (unsigned long)spool);
1408

1409
	vma_commit_reservation(h, vma, addr);
1410
	return page;
1411 1412 1413 1414 1415 1416 1417

out_uncharge_cgroup:
	hugetlb_cgroup_uncharge_cgroup(idx, pages_per_huge_page(h), h_cg);
out_subpool_put:
	if (chg || avoid_reserve)
		hugepage_subpool_put_pages(spool, 1);
	return ERR_PTR(-ENOSPC);
1418 1419
}

1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433
/*
 * alloc_huge_page()'s wrapper which simply returns the page if allocation
 * succeeds, otherwise NULL. This function is called from new_vma_page(),
 * where no ERR_VALUE is expected to be returned.
 */
struct page *alloc_huge_page_noerr(struct vm_area_struct *vma,
				unsigned long addr, int avoid_reserve)
{
	struct page *page = alloc_huge_page(vma, addr, avoid_reserve);
	if (IS_ERR(page))
		page = NULL;
	return page;
}

1434
int __weak alloc_bootmem_huge_page(struct hstate *h)
1435 1436
{
	struct huge_bootmem_page *m;
1437
	int nr_nodes, node;
1438

1439
	for_each_node_mask_to_alloc(h, nr_nodes, node, &node_states[N_MEMORY]) {
1440 1441
		void *addr;

1442 1443 1444
		addr = memblock_virt_alloc_try_nid_nopanic(
				huge_page_size(h), huge_page_size(h),
				0, BOOTMEM_ALLOC_ACCESSIBLE, node);
1445 1446 1447 1448 1449 1450 1451
		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;
1452
			goto found;
1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464
		}
	}
	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;
}

1465
static void __init prep_compound_huge_page(struct page *page, int order)
1466 1467 1468 1469 1470 1471 1472
{
	if (unlikely(order > (MAX_ORDER - 1)))
		prep_compound_gigantic_page(page, order);
	else
		prep_compound_page(page, order);
}

1473 1474 1475 1476 1477 1478 1479
/* 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;
1480 1481 1482 1483
		struct page *page;

#ifdef CONFIG_HIGHMEM
		page = pfn_to_page(m->phys >> PAGE_SHIFT);
1484 1485
		memblock_free_late(__pa(m),
				   sizeof(struct huge_bootmem_page));
1486 1487 1488
#else
		page = virt_to_page(m);
#endif
1489
		WARN_ON(page_count(page) != 1);
1490
		prep_compound_huge_page(page, h->order);
1491
		WARN_ON(PageReserved(page));
1492
		prep_new_huge_page(h, page, page_to_nid(page));
1493 1494 1495 1496 1497 1498
		/*
		 * 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.
		 */
1499
		if (hstate_is_gigantic(h))
1500
			adjust_managed_page_count(page, 1 << h->order);
1501 1502 1503
	}
}

1504
static void __init hugetlb_hstate_alloc_pages(struct hstate *h)
L
Linus Torvalds 已提交
1505 1506
{
	unsigned long i;
1507

1508
	for (i = 0; i < h->max_huge_pages; ++i) {
1509
		if (hstate_is_gigantic(h)) {
1510 1511
			if (!alloc_bootmem_huge_page(h))
				break;
1512
		} else if (!alloc_fresh_huge_page(h,
1513
					 &node_states[N_MEMORY]))
L
Linus Torvalds 已提交
1514 1515
			break;
	}
1516
	h->max_huge_pages = i;
1517 1518 1519 1520 1521 1522 1523
}

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

	for_each_hstate(h) {
1524
		/* oversize hugepages were init'ed in early boot */
1525
		if (!hstate_is_gigantic(h))
1526
			hugetlb_hstate_alloc_pages(h);
1527 1528 1529
	}
}

A
Andi Kleen 已提交
1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540
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;
}

1541 1542 1543 1544 1545
static void __init report_hugepages(void)
{
	struct hstate *h;

	for_each_hstate(h) {
A
Andi Kleen 已提交
1546
		char buf[32];
1547
		pr_info("HugeTLB registered %s page size, pre-allocated %ld pages\n",
A
Andi Kleen 已提交
1548 1549
			memfmt(buf, huge_page_size(h)),
			h->free_huge_pages);
1550 1551 1552
	}
}

L
Linus Torvalds 已提交
1553
#ifdef CONFIG_HIGHMEM
1554 1555
static void try_to_free_low(struct hstate *h, unsigned long count,
						nodemask_t *nodes_allowed)
L
Linus Torvalds 已提交
1556
{
1557 1558
	int i;

1559
	if (hstate_is_gigantic(h))
1560 1561
		return;

1562
	for_each_node_mask(i, *nodes_allowed) {
L
Linus Torvalds 已提交
1563
		struct page *page, *next;
1564 1565 1566
		struct list_head *freel = &h->hugepage_freelists[i];
		list_for_each_entry_safe(page, next, freel, lru) {
			if (count >= h->nr_huge_pages)
1567
				return;
L
Linus Torvalds 已提交
1568 1569 1570
			if (PageHighMem(page))
				continue;
			list_del(&page->lru);
1571
			update_and_free_page(h, page);
1572 1573
			h->free_huge_pages--;
			h->free_huge_pages_node[page_to_nid(page)]--;
L
Linus Torvalds 已提交
1574 1575 1576 1577
		}
	}
}
#else
1578 1579
static inline void try_to_free_low(struct hstate *h, unsigned long count,
						nodemask_t *nodes_allowed)
L
Linus Torvalds 已提交
1580 1581 1582 1583
{
}
#endif

1584 1585 1586 1587 1588
/*
 * 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.
 */
1589 1590
static int adjust_pool_surplus(struct hstate *h, nodemask_t *nodes_allowed,
				int delta)
1591
{
1592
	int nr_nodes, node;
1593 1594 1595

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

1596 1597 1598 1599
	if (delta < 0) {
		for_each_node_mask_to_alloc(h, nr_nodes, node, nodes_allowed) {
			if (h->surplus_huge_pages_node[node])
				goto found;
1600
		}
1601 1602 1603 1604 1605
	} else {
		for_each_node_mask_to_free(h, nr_nodes, node, nodes_allowed) {
			if (h->surplus_huge_pages_node[node] <
					h->nr_huge_pages_node[node])
				goto found;
1606
		}
1607 1608
	}
	return 0;
1609

1610 1611 1612 1613
found:
	h->surplus_huge_pages += delta;
	h->surplus_huge_pages_node[node] += delta;
	return 1;
1614 1615
}

1616
#define persistent_huge_pages(h) (h->nr_huge_pages - h->surplus_huge_pages)
1617 1618
static unsigned long set_max_huge_pages(struct hstate *h, unsigned long count,
						nodemask_t *nodes_allowed)
L
Linus Torvalds 已提交
1619
{
1620
	unsigned long min_count, ret;
L
Linus Torvalds 已提交
1621

1622
	if (hstate_is_gigantic(h) && !gigantic_page_supported())
1623 1624
		return h->max_huge_pages;

1625 1626 1627 1628
	/*
	 * Increase the pool size
	 * First take pages out of surplus state.  Then make up the
	 * remaining difference by allocating fresh huge pages.
1629 1630 1631 1632 1633 1634
	 *
	 * 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.
1635
	 */
L
Linus Torvalds 已提交
1636
	spin_lock(&hugetlb_lock);
1637
	while (h->surplus_huge_pages && count > persistent_huge_pages(h)) {
1638
		if (!adjust_pool_surplus(h, nodes_allowed, -1))
1639 1640 1641
			break;
	}

1642
	while (count > persistent_huge_pages(h)) {
1643 1644 1645 1646 1647 1648
		/*
		 * 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);
1649 1650 1651 1652
		if (hstate_is_gigantic(h))
			ret = alloc_fresh_gigantic_page(h, nodes_allowed);
		else
			ret = alloc_fresh_huge_page(h, nodes_allowed);
1653 1654 1655 1656
		spin_lock(&hugetlb_lock);
		if (!ret)
			goto out;

1657 1658 1659
		/* Bail for signals. Probably ctrl-c from user */
		if (signal_pending(current))
			goto out;
1660 1661 1662 1663 1664 1665 1666 1667
	}

	/*
	 * 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.
1668 1669 1670 1671 1672 1673 1674 1675
	 *
	 * 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.
1676
	 */
1677
	min_count = h->resv_huge_pages + h->nr_huge_pages - h->free_huge_pages;
1678
	min_count = max(count, min_count);
1679
	try_to_free_low(h, min_count, nodes_allowed);
1680
	while (min_count < persistent_huge_pages(h)) {
1681
		if (!free_pool_huge_page(h, nodes_allowed, 0))
L
Linus Torvalds 已提交
1682
			break;
1683
		cond_resched_lock(&hugetlb_lock);
L
Linus Torvalds 已提交
1684
	}
1685
	while (count < persistent_huge_pages(h)) {
1686
		if (!adjust_pool_surplus(h, nodes_allowed, 1))
1687 1688 1689
			break;
	}
out:
1690
	ret = persistent_huge_pages(h);
L
Linus Torvalds 已提交
1691
	spin_unlock(&hugetlb_lock);
1692
	return ret;
L
Linus Torvalds 已提交
1693 1694
}

1695 1696 1697 1698 1699 1700 1701 1702 1703 1704
#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];

1705 1706 1707
static struct hstate *kobj_to_node_hstate(struct kobject *kobj, int *nidp);

static struct hstate *kobj_to_hstate(struct kobject *kobj, int *nidp)
1708 1709
{
	int i;
1710

1711
	for (i = 0; i < HUGE_MAX_HSTATE; i++)
1712 1713 1714
		if (hstate_kobjs[i] == kobj) {
			if (nidp)
				*nidp = NUMA_NO_NODE;
1715
			return &hstates[i];
1716 1717 1718
		}

	return kobj_to_node_hstate(kobj, nidp);
1719 1720
}

1721
static ssize_t nr_hugepages_show_common(struct kobject *kobj,
1722 1723
					struct kobj_attribute *attr, char *buf)
{
1724 1725 1726 1727 1728 1729 1730 1731 1732 1733 1734
	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);
1735
}
1736

1737 1738 1739
static ssize_t nr_hugepages_store_common(bool obey_mempolicy,
			struct kobject *kobj, struct kobj_attribute *attr,
			const char *buf, size_t len)
1740 1741
{
	int err;
1742
	int nid;
1743
	unsigned long count;
1744
	struct hstate *h;
1745
	NODEMASK_ALLOC(nodemask_t, nodes_allowed, GFP_KERNEL | __GFP_NORETRY);
1746

1747
	err = kstrtoul(buf, 10, &count);
1748
	if (err)
1749
		goto out;
1750

1751
	h = kobj_to_hstate(kobj, &nid);
1752
	if (hstate_is_gigantic(h) && !gigantic_page_supported()) {
1753 1754 1755 1756
		err = -EINVAL;
		goto out;
	}

1757 1758 1759 1760 1761 1762 1763
	if (nid == NUMA_NO_NODE) {
		/*
		 * global hstate attribute
		 */
		if (!(obey_mempolicy &&
				init_nodemask_of_mempolicy(nodes_allowed))) {
			NODEMASK_FREE(nodes_allowed);
1764
			nodes_allowed = &node_states[N_MEMORY];
1765 1766 1767 1768 1769 1770 1771 1772 1773
		}
	} 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
1774
		nodes_allowed = &node_states[N_MEMORY];
1775

1776
	h->max_huge_pages = set_max_huge_pages(h, count, nodes_allowed);
1777

1778
	if (nodes_allowed != &node_states[N_MEMORY])
1779 1780 1781
		NODEMASK_FREE(nodes_allowed);

	return len;
1782 1783 1784
out:
	NODEMASK_FREE(nodes_allowed);
	return err;
1785 1786 1787 1788 1789 1790 1791 1792 1793 1794 1795 1796
}

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);
1797 1798 1799
}
HSTATE_ATTR(nr_hugepages);

1800 1801 1802 1803 1804 1805 1806 1807 1808 1809 1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820
#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


1821 1822 1823
static ssize_t nr_overcommit_hugepages_show(struct kobject *kobj,
					struct kobj_attribute *attr, char *buf)
{
1824
	struct hstate *h = kobj_to_hstate(kobj, NULL);
1825 1826
	return sprintf(buf, "%lu\n", h->nr_overcommit_huge_pages);
}
1827

1828 1829 1830 1831 1832
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;
1833
	struct hstate *h = kobj_to_hstate(kobj, NULL);
1834

1835
	if (hstate_is_gigantic(h))
1836 1837
		return -EINVAL;

1838
	err = kstrtoul(buf, 10, &input);
1839
	if (err)
1840
		return err;
1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852

	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)
{
1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863
	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);
1864 1865 1866 1867 1868 1869
}
HSTATE_ATTR_RO(free_hugepages);

static ssize_t resv_hugepages_show(struct kobject *kobj,
					struct kobj_attribute *attr, char *buf)
{
1870
	struct hstate *h = kobj_to_hstate(kobj, NULL);
1871 1872 1873 1874 1875 1876 1877
	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)
{
1878 1879 1880 1881 1882 1883 1884 1885 1886 1887 1888
	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);
1889 1890 1891 1892 1893 1894 1895 1896 1897
}
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,
1898 1899 1900
#ifdef CONFIG_NUMA
	&nr_hugepages_mempolicy_attr.attr,
#endif
1901 1902 1903 1904 1905 1906 1907
	NULL,
};

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

J
Jeff Mahoney 已提交
1908 1909 1910
static int hugetlb_sysfs_add_hstate(struct hstate *h, struct kobject *parent,
				    struct kobject **hstate_kobjs,
				    struct attribute_group *hstate_attr_group)
1911 1912
{
	int retval;
1913
	int hi = hstate_index(h);
1914

1915 1916
	hstate_kobjs[hi] = kobject_create_and_add(h->name, parent);
	if (!hstate_kobjs[hi])
1917 1918
		return -ENOMEM;

1919
	retval = sysfs_create_group(hstate_kobjs[hi], hstate_attr_group);
1920
	if (retval)
1921
		kobject_put(hstate_kobjs[hi]);
1922 1923 1924 1925 1926 1927 1928 1929 1930 1931 1932 1933 1934 1935

	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) {
1936 1937
		err = hugetlb_sysfs_add_hstate(h, hugepages_kobj,
					 hstate_kobjs, &hstate_attr_group);
1938
		if (err)
1939
			pr_err("Hugetlb: Unable to add hstate %s", h->name);
1940 1941 1942
	}
}

1943 1944 1945 1946
#ifdef CONFIG_NUMA

/*
 * node_hstate/s - associate per node hstate attributes, via their kobjects,
1947 1948 1949
 * 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
1950 1951 1952 1953 1954 1955 1956 1957 1958
 * 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];

/*
1959
 * A subset of global hstate attributes for node devices
1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972
 */
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,
};

/*
1973
 * kobj_to_node_hstate - lookup global hstate for node device hstate attr kobj.
1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995
 * 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;
}

/*
1996
 * Unregister hstate attributes from a single node device.
1997 1998
 * No-op if no hstate attributes attached.
 */
1999
static void hugetlb_unregister_node(struct node *node)
2000 2001
{
	struct hstate *h;
2002
	struct node_hstate *nhs = &node_hstates[node->dev.id];
2003 2004

	if (!nhs->hugepages_kobj)
2005
		return;		/* no hstate attributes */
2006

2007 2008 2009 2010 2011
	for_each_hstate(h) {
		int idx = hstate_index(h);
		if (nhs->hstate_kobjs[idx]) {
			kobject_put(nhs->hstate_kobjs[idx]);
			nhs->hstate_kobjs[idx] = NULL;
2012
		}
2013
	}
2014 2015 2016 2017 2018 2019

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

/*
2020
 * hugetlb module exit:  unregister hstate attributes from node devices
2021 2022 2023 2024 2025 2026 2027
 * that have them.
 */
static void hugetlb_unregister_all_nodes(void)
{
	int nid;

	/*
2028
	 * disable node device registrations.
2029 2030 2031 2032 2033 2034 2035
	 */
	register_hugetlbfs_with_node(NULL, NULL);

	/*
	 * remove hstate attributes from any nodes that have them.
	 */
	for (nid = 0; nid < nr_node_ids; nid++)
2036
		hugetlb_unregister_node(node_devices[nid]);
2037 2038 2039
}

/*
2040
 * Register hstate attributes for a single node device.
2041 2042
 * No-op if attributes already registered.
 */
2043
static void hugetlb_register_node(struct node *node)
2044 2045
{
	struct hstate *h;
2046
	struct node_hstate *nhs = &node_hstates[node->dev.id];
2047 2048 2049 2050 2051 2052
	int err;

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

	nhs->hugepages_kobj = kobject_create_and_add("hugepages",
2053
							&node->dev.kobj);
2054 2055 2056 2057 2058 2059 2060 2061
	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) {
2062 2063
			pr_err("Hugetlb: Unable to add hstate %s for node %d\n",
				h->name, node->dev.id);
2064 2065 2066 2067 2068 2069 2070
			hugetlb_unregister_node(node);
			break;
		}
	}
}

/*
2071
 * hugetlb init time:  register hstate attributes for all registered node
2072 2073
 * devices of nodes that have memory.  All on-line nodes should have
 * registered their associated device by this time.
2074 2075 2076 2077 2078
 */
static void hugetlb_register_all_nodes(void)
{
	int nid;

2079
	for_each_node_state(nid, N_MEMORY) {
2080
		struct node *node = node_devices[nid];
2081
		if (node->dev.id == nid)
2082 2083 2084 2085
			hugetlb_register_node(node);
	}

	/*
2086
	 * Let the node device driver know we're here so it can
2087 2088 2089 2090 2091 2092 2093 2094 2095 2096 2097 2098 2099 2100 2101 2102 2103 2104 2105 2106 2107
	 * [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

2108 2109 2110 2111
static void __exit hugetlb_exit(void)
{
	struct hstate *h;

2112 2113
	hugetlb_unregister_all_nodes();

2114
	for_each_hstate(h) {
2115
		kobject_put(hstate_kobjs[hstate_index(h)]);
2116 2117 2118
	}

	kobject_put(hugepages_kobj);
2119
	kfree(htlb_fault_mutex_table);
2120 2121 2122 2123 2124
}
module_exit(hugetlb_exit);

static int __init hugetlb_init(void)
{
2125 2126
	int i;

2127
	if (!hugepages_supported())
2128
		return 0;
2129

2130 2131 2132 2133
	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);
2134
	}
2135
	default_hstate_idx = hstate_index(size_to_hstate(default_hstate_size));
2136 2137
	if (default_hstate_max_huge_pages)
		default_hstate.max_huge_pages = default_hstate_max_huge_pages;
2138 2139

	hugetlb_init_hstates();
2140
	gather_bootmem_prealloc();
2141 2142 2143
	report_hugepages();

	hugetlb_sysfs_init();
2144
	hugetlb_register_all_nodes();
2145
	hugetlb_cgroup_file_init();
2146

2147 2148 2149 2150 2151 2152 2153 2154 2155 2156 2157
#ifdef CONFIG_SMP
	num_fault_mutexes = roundup_pow_of_two(8 * num_possible_cpus());
#else
	num_fault_mutexes = 1;
#endif
	htlb_fault_mutex_table =
		kmalloc(sizeof(struct mutex) * num_fault_mutexes, GFP_KERNEL);
	BUG_ON(!htlb_fault_mutex_table);

	for (i = 0; i < num_fault_mutexes; i++)
		mutex_init(&htlb_fault_mutex_table[i]);
2158 2159 2160 2161 2162 2163 2164 2165
	return 0;
}
module_init(hugetlb_init);

/* Should be called on processing a hugepagesz=... option */
void __init hugetlb_add_hstate(unsigned order)
{
	struct hstate *h;
2166 2167
	unsigned long i;

2168
	if (size_to_hstate(PAGE_SIZE << order)) {
2169
		pr_warning("hugepagesz= specified twice, ignoring\n");
2170 2171
		return;
	}
2172
	BUG_ON(hugetlb_max_hstate >= HUGE_MAX_HSTATE);
2173
	BUG_ON(order == 0);
2174
	h = &hstates[hugetlb_max_hstate++];
2175 2176
	h->order = order;
	h->mask = ~((1ULL << (order + PAGE_SHIFT)) - 1);
2177 2178 2179 2180
	h->nr_huge_pages = 0;
	h->free_huge_pages = 0;
	for (i = 0; i < MAX_NUMNODES; ++i)
		INIT_LIST_HEAD(&h->hugepage_freelists[i]);
2181
	INIT_LIST_HEAD(&h->hugepage_activelist);
2182 2183
	h->next_nid_to_alloc = first_node(node_states[N_MEMORY]);
	h->next_nid_to_free = first_node(node_states[N_MEMORY]);
2184 2185
	snprintf(h->name, HSTATE_NAME_LEN, "hugepages-%lukB",
					huge_page_size(h)/1024);
2186

2187 2188 2189
	parsed_hstate = h;
}

2190
static int __init hugetlb_nrpages_setup(char *s)
2191 2192
{
	unsigned long *mhp;
2193
	static unsigned long *last_mhp;
2194 2195

	/*
2196
	 * !hugetlb_max_hstate means we haven't parsed a hugepagesz= parameter yet,
2197 2198
	 * so this hugepages= parameter goes to the "default hstate".
	 */
2199
	if (!hugetlb_max_hstate)
2200 2201 2202 2203
		mhp = &default_hstate_max_huge_pages;
	else
		mhp = &parsed_hstate->max_huge_pages;

2204
	if (mhp == last_mhp) {
2205 2206
		pr_warning("hugepages= specified twice without "
			   "interleaving hugepagesz=, ignoring\n");
2207 2208 2209
		return 1;
	}

2210 2211 2212
	if (sscanf(s, "%lu", mhp) <= 0)
		*mhp = 0;

2213 2214 2215 2216 2217
	/*
	 * 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.
	 */
2218
	if (hugetlb_max_hstate && parsed_hstate->order >= MAX_ORDER)
2219 2220 2221 2222
		hugetlb_hstate_alloc_pages(parsed_hstate);

	last_mhp = mhp;

2223 2224
	return 1;
}
2225 2226 2227 2228 2229 2230 2231 2232
__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);
2233

2234 2235 2236 2237 2238 2239 2240 2241 2242 2243 2244 2245
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
2246 2247 2248
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
2249
{
2250 2251
	struct hstate *h = &default_hstate;
	unsigned long tmp;
2252
	int ret;
2253

2254 2255 2256
	if (!hugepages_supported())
		return -ENOTSUPP;

2257
	tmp = h->max_huge_pages;
2258

2259
	if (write && hstate_is_gigantic(h) && !gigantic_page_supported())
2260 2261
		return -EINVAL;

2262 2263
	table->data = &tmp;
	table->maxlen = sizeof(unsigned long);
2264 2265 2266
	ret = proc_doulongvec_minmax(table, write, buffer, length, ppos);
	if (ret)
		goto out;
2267

2268
	if (write) {
2269 2270
		NODEMASK_ALLOC(nodemask_t, nodes_allowed,
						GFP_KERNEL | __GFP_NORETRY);
2271 2272 2273
		if (!(obey_mempolicy &&
			       init_nodemask_of_mempolicy(nodes_allowed))) {
			NODEMASK_FREE(nodes_allowed);
2274
			nodes_allowed = &node_states[N_MEMORY];
2275 2276 2277
		}
		h->max_huge_pages = set_max_huge_pages(h, tmp, nodes_allowed);

2278
		if (nodes_allowed != &node_states[N_MEMORY])
2279 2280
			NODEMASK_FREE(nodes_allowed);
	}
2281 2282
out:
	return ret;
L
Linus Torvalds 已提交
2283
}
2284

2285 2286 2287 2288 2289 2290 2291 2292 2293 2294 2295 2296 2297 2298 2299 2300 2301
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 */

2302
int hugetlb_overcommit_handler(struct ctl_table *table, int write,
2303
			void __user *buffer,
2304 2305
			size_t *length, loff_t *ppos)
{
2306
	struct hstate *h = &default_hstate;
2307
	unsigned long tmp;
2308
	int ret;
2309

2310 2311 2312
	if (!hugepages_supported())
		return -ENOTSUPP;

2313
	tmp = h->nr_overcommit_huge_pages;
2314

2315
	if (write && hstate_is_gigantic(h))
2316 2317
		return -EINVAL;

2318 2319
	table->data = &tmp;
	table->maxlen = sizeof(unsigned long);
2320 2321 2322
	ret = proc_doulongvec_minmax(table, write, buffer, length, ppos);
	if (ret)
		goto out;
2323 2324 2325 2326 2327 2328

	if (write) {
		spin_lock(&hugetlb_lock);
		h->nr_overcommit_huge_pages = tmp;
		spin_unlock(&hugetlb_lock);
	}
2329 2330
out:
	return ret;
2331 2332
}

L
Linus Torvalds 已提交
2333 2334
#endif /* CONFIG_SYSCTL */

2335
void hugetlb_report_meminfo(struct seq_file *m)
L
Linus Torvalds 已提交
2336
{
2337
	struct hstate *h = &default_hstate;
2338 2339
	if (!hugepages_supported())
		return;
2340
	seq_printf(m,
2341 2342 2343 2344 2345
			"HugePages_Total:   %5lu\n"
			"HugePages_Free:    %5lu\n"
			"HugePages_Rsvd:    %5lu\n"
			"HugePages_Surp:    %5lu\n"
			"Hugepagesize:   %8lu kB\n",
2346 2347 2348 2349 2350
			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
2351 2352 2353 2354
}

int hugetlb_report_node_meminfo(int nid, char *buf)
{
2355
	struct hstate *h = &default_hstate;
2356 2357
	if (!hugepages_supported())
		return 0;
L
Linus Torvalds 已提交
2358 2359
	return sprintf(buf,
		"Node %d HugePages_Total: %5u\n"
2360 2361
		"Node %d HugePages_Free:  %5u\n"
		"Node %d HugePages_Surp:  %5u\n",
2362 2363 2364
		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
2365 2366
}

2367 2368 2369 2370 2371
void hugetlb_show_meminfo(void)
{
	struct hstate *h;
	int nid;

2372 2373 2374
	if (!hugepages_supported())
		return;

2375 2376 2377 2378 2379 2380 2381 2382 2383 2384
	for_each_node_state(nid, N_MEMORY)
		for_each_hstate(h)
			pr_info("Node %d hugepages_total=%u hugepages_free=%u hugepages_surp=%u hugepages_size=%lukB\n",
				nid,
				h->nr_huge_pages_node[nid],
				h->free_huge_pages_node[nid],
				h->surplus_huge_pages_node[nid],
				1UL << (huge_page_order(h) + PAGE_SHIFT - 10));
}

L
Linus Torvalds 已提交
2385 2386 2387
/* Return the number pages of memory we physically have, in PAGE_SIZE units. */
unsigned long hugetlb_total_pages(void)
{
2388 2389 2390 2391 2392 2393
	struct hstate *h;
	unsigned long nr_total_pages = 0;

	for_each_hstate(h)
		nr_total_pages += h->nr_huge_pages * pages_per_huge_page(h);
	return nr_total_pages;
L
Linus Torvalds 已提交
2394 2395
}

2396
static int hugetlb_acct_memory(struct hstate *h, long delta)
M
Mel Gorman 已提交
2397 2398 2399 2400 2401 2402 2403 2404 2405 2406 2407 2408 2409 2410 2411 2412 2413 2414 2415 2416 2417 2418
{
	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) {
2419
		if (gather_surplus_pages(h, delta) < 0)
M
Mel Gorman 已提交
2420 2421
			goto out;

2422 2423
		if (delta > cpuset_mems_nr(h->free_huge_pages_node)) {
			return_unused_surplus_pages(h, delta);
M
Mel Gorman 已提交
2424 2425 2426 2427 2428 2429
			goto out;
		}
	}

	ret = 0;
	if (delta < 0)
2430
		return_unused_surplus_pages(h, (unsigned long) -delta);
M
Mel Gorman 已提交
2431 2432 2433 2434 2435 2436

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

2437 2438
static void hugetlb_vm_op_open(struct vm_area_struct *vma)
{
2439
	struct resv_map *resv = vma_resv_map(vma);
2440 2441 2442 2443 2444

	/*
	 * 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
2445
	 * has a reference to the reservation map it cannot disappear until
2446 2447 2448
	 * after this open call completes.  It is therefore safe to take a
	 * new reference here without additional locking.
	 */
2449
	if (resv && is_vma_resv_set(vma, HPAGE_RESV_OWNER))
2450
		kref_get(&resv->refs);
2451 2452
}

2453 2454
static void hugetlb_vm_op_close(struct vm_area_struct *vma)
{
2455
	struct hstate *h = hstate_vma(vma);
2456
	struct resv_map *resv = vma_resv_map(vma);
2457
	struct hugepage_subpool *spool = subpool_vma(vma);
2458
	unsigned long reserve, start, end;
2459

2460 2461
	if (!resv || !is_vma_resv_set(vma, HPAGE_RESV_OWNER))
		return;
2462

2463 2464
	start = vma_hugecache_offset(h, vma, vma->vm_start);
	end = vma_hugecache_offset(h, vma, vma->vm_end);
2465

2466
	reserve = (end - start) - region_count(resv, start, end);
2467

2468 2469 2470 2471 2472
	kref_put(&resv->refs, resv_map_release);

	if (reserve) {
		hugetlb_acct_memory(h, -reserve);
		hugepage_subpool_put_pages(spool, reserve);
2473
	}
2474 2475
}

L
Linus Torvalds 已提交
2476 2477 2478 2479 2480 2481
/*
 * 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
2482
static int hugetlb_vm_op_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
L
Linus Torvalds 已提交
2483 2484
{
	BUG();
N
Nick Piggin 已提交
2485
	return 0;
L
Linus Torvalds 已提交
2486 2487
}

2488
const struct vm_operations_struct hugetlb_vm_ops = {
N
Nick Piggin 已提交
2489
	.fault = hugetlb_vm_op_fault,
2490
	.open = hugetlb_vm_op_open,
2491
	.close = hugetlb_vm_op_close,
L
Linus Torvalds 已提交
2492 2493
};

2494 2495
static pte_t make_huge_pte(struct vm_area_struct *vma, struct page *page,
				int writable)
D
David Gibson 已提交
2496 2497 2498
{
	pte_t entry;

2499
	if (writable) {
2500 2501
		entry = huge_pte_mkwrite(huge_pte_mkdirty(mk_huge_pte(page,
					 vma->vm_page_prot)));
D
David Gibson 已提交
2502
	} else {
2503 2504
		entry = huge_pte_wrprotect(mk_huge_pte(page,
					   vma->vm_page_prot));
D
David Gibson 已提交
2505 2506 2507
	}
	entry = pte_mkyoung(entry);
	entry = pte_mkhuge(entry);
2508
	entry = arch_make_huge_pte(entry, vma, page, writable);
D
David Gibson 已提交
2509 2510 2511 2512

	return entry;
}

2513 2514 2515 2516 2517
static void set_huge_ptep_writable(struct vm_area_struct *vma,
				   unsigned long address, pte_t *ptep)
{
	pte_t entry;

2518
	entry = huge_pte_mkwrite(huge_pte_mkdirty(huge_ptep_get(ptep)));
2519
	if (huge_ptep_set_access_flags(vma, address, ptep, entry, 1))
2520
		update_mmu_cache(vma, address, ptep);
2521 2522 2523
}


D
David Gibson 已提交
2524 2525 2526 2527 2528
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;
2529
	unsigned long addr;
2530
	int cow;
2531 2532
	struct hstate *h = hstate_vma(vma);
	unsigned long sz = huge_page_size(h);
2533 2534 2535
	unsigned long mmun_start;	/* For mmu_notifiers */
	unsigned long mmun_end;		/* For mmu_notifiers */
	int ret = 0;
2536 2537

	cow = (vma->vm_flags & (VM_SHARED | VM_MAYWRITE)) == VM_MAYWRITE;
D
David Gibson 已提交
2538

2539 2540 2541 2542 2543
	mmun_start = vma->vm_start;
	mmun_end = vma->vm_end;
	if (cow)
		mmu_notifier_invalidate_range_start(src, mmun_start, mmun_end);

2544
	for (addr = vma->vm_start; addr < vma->vm_end; addr += sz) {
2545
		spinlock_t *src_ptl, *dst_ptl;
H
Hugh Dickins 已提交
2546 2547 2548
		src_pte = huge_pte_offset(src, addr);
		if (!src_pte)
			continue;
2549
		dst_pte = huge_pte_alloc(dst, addr, sz);
2550 2551 2552 2553
		if (!dst_pte) {
			ret = -ENOMEM;
			break;
		}
2554 2555 2556 2557 2558

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

2559 2560 2561
		dst_ptl = huge_pte_lock(h, dst, dst_pte);
		src_ptl = huge_pte_lockptr(h, src, src_pte);
		spin_lock_nested(src_ptl, SINGLE_DEPTH_NESTING);
2562
		if (!huge_pte_none(huge_ptep_get(src_pte))) {
2563
			if (cow)
2564 2565
				huge_ptep_set_wrprotect(src, addr, src_pte);
			entry = huge_ptep_get(src_pte);
2566 2567
			ptepage = pte_page(entry);
			get_page(ptepage);
2568
			page_dup_rmap(ptepage);
2569 2570
			set_huge_pte_at(dst, addr, dst_pte, entry);
		}
2571 2572
		spin_unlock(src_ptl);
		spin_unlock(dst_ptl);
D
David Gibson 已提交
2573 2574
	}

2575 2576 2577 2578
	if (cow)
		mmu_notifier_invalidate_range_end(src, mmun_start, mmun_end);

	return ret;
D
David Gibson 已提交
2579 2580
}

N
Naoya Horiguchi 已提交
2581 2582 2583 2584 2585 2586 2587
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);
2588
	if (non_swap_entry(swp) && is_migration_entry(swp))
N
Naoya Horiguchi 已提交
2589
		return 1;
2590
	else
N
Naoya Horiguchi 已提交
2591 2592 2593
		return 0;
}

2594 2595 2596 2597 2598 2599 2600
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);
2601
	if (non_swap_entry(swp) && is_hwpoison_entry(swp))
2602
		return 1;
2603
	else
2604 2605 2606
		return 0;
}

2607 2608 2609
void __unmap_hugepage_range(struct mmu_gather *tlb, struct vm_area_struct *vma,
			    unsigned long start, unsigned long end,
			    struct page *ref_page)
D
David Gibson 已提交
2610
{
2611
	int force_flush = 0;
D
David Gibson 已提交
2612 2613
	struct mm_struct *mm = vma->vm_mm;
	unsigned long address;
2614
	pte_t *ptep;
D
David Gibson 已提交
2615
	pte_t pte;
2616
	spinlock_t *ptl;
D
David Gibson 已提交
2617
	struct page *page;
2618 2619
	struct hstate *h = hstate_vma(vma);
	unsigned long sz = huge_page_size(h);
2620 2621
	const unsigned long mmun_start = start;	/* For mmu_notifiers */
	const unsigned long mmun_end   = end;	/* For mmu_notifiers */
2622

D
David Gibson 已提交
2623
	WARN_ON(!is_vm_hugetlb_page(vma));
2624 2625
	BUG_ON(start & ~huge_page_mask(h));
	BUG_ON(end & ~huge_page_mask(h));
D
David Gibson 已提交
2626

2627
	tlb_start_vma(tlb, vma);
2628
	mmu_notifier_invalidate_range_start(mm, mmun_start, mmun_end);
2629
again:
2630
	for (address = start; address < end; address += sz) {
2631
		ptep = huge_pte_offset(mm, address);
A
Adam Litke 已提交
2632
		if (!ptep)
2633 2634
			continue;

2635
		ptl = huge_pte_lock(h, mm, ptep);
2636
		if (huge_pmd_unshare(mm, &address, ptep))
2637
			goto unlock;
2638

2639 2640
		pte = huge_ptep_get(ptep);
		if (huge_pte_none(pte))
2641
			goto unlock;
2642 2643 2644 2645

		/*
		 * HWPoisoned hugepage is already unmapped and dropped reference
		 */
2646
		if (unlikely(is_hugetlb_entry_hwpoisoned(pte))) {
2647
			huge_pte_clear(mm, address, ptep);
2648
			goto unlock;
2649
		}
2650 2651

		page = pte_page(pte);
2652 2653 2654 2655 2656 2657 2658
		/*
		 * 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) {
			if (page != ref_page)
2659
				goto unlock;
2660 2661 2662 2663 2664 2665 2666 2667 2668

			/*
			 * 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);
		}

2669
		pte = huge_ptep_get_and_clear(mm, address, ptep);
2670
		tlb_remove_tlb_entry(tlb, ptep, address);
2671
		if (huge_pte_dirty(pte))
2672
			set_page_dirty(page);
2673

2674 2675
		page_remove_rmap(page);
		force_flush = !__tlb_remove_page(tlb, page);
2676 2677
		if (force_flush) {
			spin_unlock(ptl);
2678
			break;
2679
		}
2680
		/* Bail out after unmapping reference page if supplied */
2681 2682
		if (ref_page) {
			spin_unlock(ptl);
2683
			break;
2684 2685 2686
		}
unlock:
		spin_unlock(ptl);
D
David Gibson 已提交
2687
	}
2688 2689 2690 2691 2692 2693 2694 2695 2696 2697
	/*
	 * mmu_gather ran out of room to batch pages, we break out of
	 * the PTE lock to avoid doing the potential expensive TLB invalidate
	 * and page-free while holding it.
	 */
	if (force_flush) {
		force_flush = 0;
		tlb_flush_mmu(tlb);
		if (address < end && !ref_page)
			goto again;
2698
	}
2699
	mmu_notifier_invalidate_range_end(mm, mmun_start, mmun_end);
2700
	tlb_end_vma(tlb, vma);
L
Linus Torvalds 已提交
2701
}
D
David Gibson 已提交
2702

2703 2704 2705 2706 2707 2708 2709 2710 2711 2712 2713 2714 2715 2716 2717 2718 2719 2720 2721
void __unmap_hugepage_range_final(struct mmu_gather *tlb,
			  struct vm_area_struct *vma, unsigned long start,
			  unsigned long end, struct page *ref_page)
{
	__unmap_hugepage_range(tlb, vma, start, end, ref_page);

	/*
	 * Clear this flag so that x86's huge_pmd_share page_table_shareable
	 * test will fail on a vma being torn down, and not grab a page table
	 * on its way out.  We're lucky that the flag has such an appropriate
	 * name, and can in fact be safely cleared here. We could clear it
	 * before the __unmap_hugepage_range above, but all that's necessary
	 * is to clear it before releasing the i_mmap_mutex. This works
	 * because in the context this is called, the VMA is about to be
	 * destroyed and the i_mmap_mutex is held.
	 */
	vma->vm_flags &= ~VM_MAYSHARE;
}

2722
void unmap_hugepage_range(struct vm_area_struct *vma, unsigned long start,
2723
			  unsigned long end, struct page *ref_page)
2724
{
2725 2726 2727 2728 2729
	struct mm_struct *mm;
	struct mmu_gather tlb;

	mm = vma->vm_mm;

2730
	tlb_gather_mmu(&tlb, mm, start, end);
2731 2732
	__unmap_hugepage_range(&tlb, vma, start, end, ref_page);
	tlb_finish_mmu(&tlb, start, end);
2733 2734
}

2735 2736 2737 2738 2739 2740
/*
 * 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.
 */
2741 2742
static int unmap_ref_private(struct mm_struct *mm, struct vm_area_struct *vma,
				struct page *page, unsigned long address)
2743
{
2744
	struct hstate *h = hstate_vma(vma);
2745 2746 2747 2748 2749 2750 2751 2752
	struct vm_area_struct *iter_vma;
	struct address_space *mapping;
	pgoff_t pgoff;

	/*
	 * vm_pgoff is in PAGE_SIZE units, hence the different calculation
	 * from page cache lookup which is in HPAGE_SIZE units.
	 */
2753
	address = address & huge_page_mask(h);
2754 2755
	pgoff = ((address - vma->vm_start) >> PAGE_SHIFT) +
			vma->vm_pgoff;
A
Al Viro 已提交
2756
	mapping = file_inode(vma->vm_file)->i_mapping;
2757

2758 2759 2760 2761 2762
	/*
	 * 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
	 */
2763
	mutex_lock(&mapping->i_mmap_mutex);
2764
	vma_interval_tree_foreach(iter_vma, &mapping->i_mmap, pgoff, pgoff) {
2765 2766 2767 2768 2769 2770 2771 2772 2773 2774 2775 2776
		/* 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))
2777 2778
			unmap_hugepage_range(iter_vma, address,
					     address + huge_page_size(h), page);
2779
	}
2780
	mutex_unlock(&mapping->i_mmap_mutex);
2781 2782 2783 2784

	return 1;
}

2785 2786
/*
 * Hugetlb_cow() should be called with page lock of the original hugepage held.
2787 2788 2789
 * 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.
2790
 */
2791
static int hugetlb_cow(struct mm_struct *mm, struct vm_area_struct *vma,
2792
			unsigned long address, pte_t *ptep, pte_t pte,
2793
			struct page *pagecache_page, spinlock_t *ptl)
2794
{
2795
	struct hstate *h = hstate_vma(vma);
2796
	struct page *old_page, *new_page;
2797
	int outside_reserve = 0;
2798 2799
	unsigned long mmun_start;	/* For mmu_notifiers */
	unsigned long mmun_end;		/* For mmu_notifiers */
2800 2801 2802

	old_page = pte_page(pte);

2803
retry_avoidcopy:
2804 2805
	/* If no-one else is actually using this page, avoid the copy
	 * and just make the page writable */
2806 2807
	if (page_mapcount(old_page) == 1 && PageAnon(old_page)) {
		page_move_anon_rmap(old_page, vma, address);
2808
		set_huge_ptep_writable(vma, address, ptep);
N
Nick Piggin 已提交
2809
		return 0;
2810 2811
	}

2812 2813 2814 2815 2816 2817 2818 2819 2820
	/*
	 * 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.
	 */
2821
	if (is_vma_resv_set(vma, HPAGE_RESV_OWNER) &&
2822 2823 2824
			old_page != pagecache_page)
		outside_reserve = 1;

2825
	page_cache_get(old_page);
2826

2827 2828
	/* Drop page table lock as buddy allocator may be called */
	spin_unlock(ptl);
2829
	new_page = alloc_huge_page(vma, address, outside_reserve);
2830

2831
	if (IS_ERR(new_page)) {
2832
		long err = PTR_ERR(new_page);
2833
		page_cache_release(old_page);
2834 2835 2836 2837 2838 2839 2840 2841 2842 2843 2844 2845

		/*
		 * 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(huge_pte_none(pte));
2846
				spin_lock(ptl);
2847
				ptep = huge_pte_offset(mm, address & huge_page_mask(h));
2848 2849
				if (likely(ptep &&
					   pte_same(huge_ptep_get(ptep), pte)))
2850 2851
					goto retry_avoidcopy;
				/*
2852 2853
				 * race occurs while re-acquiring page table
				 * lock, and our job is done.
2854 2855
				 */
				return 0;
2856 2857 2858 2859
			}
			WARN_ON_ONCE(1);
		}

2860
		/* Caller expects lock to be held */
2861
		spin_lock(ptl);
2862 2863 2864 2865
		if (err == -ENOMEM)
			return VM_FAULT_OOM;
		else
			return VM_FAULT_SIGBUS;
2866 2867
	}

2868 2869 2870 2871
	/*
	 * When the original hugepage is shared one, it does not have
	 * anon_vma prepared.
	 */
2872
	if (unlikely(anon_vma_prepare(vma))) {
2873 2874
		page_cache_release(new_page);
		page_cache_release(old_page);
2875
		/* Caller expects lock to be held */
2876
		spin_lock(ptl);
2877
		return VM_FAULT_OOM;
2878
	}
2879

A
Andrea Arcangeli 已提交
2880 2881
	copy_user_huge_page(new_page, old_page, address, vma,
			    pages_per_huge_page(h));
N
Nick Piggin 已提交
2882
	__SetPageUptodate(new_page);
2883

2884 2885 2886
	mmun_start = address & huge_page_mask(h);
	mmun_end = mmun_start + huge_page_size(h);
	mmu_notifier_invalidate_range_start(mm, mmun_start, mmun_end);
2887
	/*
2888
	 * Retake the page table lock to check for racing updates
2889 2890
	 * before the page tables are altered
	 */
2891
	spin_lock(ptl);
2892
	ptep = huge_pte_offset(mm, address & huge_page_mask(h));
2893
	if (likely(ptep && pte_same(huge_ptep_get(ptep), pte))) {
2894 2895
		ClearPagePrivate(new_page);

2896
		/* Break COW */
2897
		huge_ptep_clear_flush(vma, address, ptep);
2898 2899
		set_huge_pte_at(mm, address, ptep,
				make_huge_pte(vma, new_page, 1));
2900
		page_remove_rmap(old_page);
2901
		hugepage_add_new_anon_rmap(new_page, vma, address);
2902 2903 2904
		/* Make the old page be freed below */
		new_page = old_page;
	}
2905
	spin_unlock(ptl);
2906
	mmu_notifier_invalidate_range_end(mm, mmun_start, mmun_end);
2907 2908
	page_cache_release(new_page);
	page_cache_release(old_page);
2909 2910

	/* Caller expects lock to be held */
2911
	spin_lock(ptl);
N
Nick Piggin 已提交
2912
	return 0;
2913 2914
}

2915
/* Return the pagecache page at a given address within a VMA */
2916 2917
static struct page *hugetlbfs_pagecache_page(struct hstate *h,
			struct vm_area_struct *vma, unsigned long address)
2918 2919
{
	struct address_space *mapping;
2920
	pgoff_t idx;
2921 2922

	mapping = vma->vm_file->f_mapping;
2923
	idx = vma_hugecache_offset(h, vma, address);
2924 2925 2926 2927

	return find_lock_page(mapping, idx);
}

H
Hugh Dickins 已提交
2928 2929 2930 2931 2932
/*
 * 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
2933 2934 2935 2936 2937 2938 2939 2940 2941 2942 2943 2944 2945 2946 2947
			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;
}

2948
static int hugetlb_no_page(struct mm_struct *mm, struct vm_area_struct *vma,
2949 2950
			   struct address_space *mapping, pgoff_t idx,
			   unsigned long address, pte_t *ptep, unsigned int flags)
2951
{
2952
	struct hstate *h = hstate_vma(vma);
2953
	int ret = VM_FAULT_SIGBUS;
2954
	int anon_rmap = 0;
A
Adam Litke 已提交
2955 2956
	unsigned long size;
	struct page *page;
2957
	pte_t new_pte;
2958
	spinlock_t *ptl;
A
Adam Litke 已提交
2959

2960 2961 2962
	/*
	 * 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
2963
	 * COW. Warn that such a situation has occurred as it may not be obvious
2964 2965
	 */
	if (is_vma_resv_set(vma, HPAGE_RESV_UNMAPPED)) {
2966 2967
		pr_warning("PID %d killed due to inadequate hugepage pool\n",
			   current->pid);
2968 2969 2970
		return ret;
	}

A
Adam Litke 已提交
2971 2972 2973 2974
	/*
	 * Use page lock to guard against racing truncation
	 * before we get page_table_lock.
	 */
2975 2976 2977
retry:
	page = find_lock_page(mapping, idx);
	if (!page) {
2978
		size = i_size_read(mapping->host) >> huge_page_shift(h);
2979 2980
		if (idx >= size)
			goto out;
2981
		page = alloc_huge_page(vma, address, 0);
2982
		if (IS_ERR(page)) {
2983 2984 2985 2986 2987
			ret = PTR_ERR(page);
			if (ret == -ENOMEM)
				ret = VM_FAULT_OOM;
			else
				ret = VM_FAULT_SIGBUS;
2988 2989
			goto out;
		}
A
Andrea Arcangeli 已提交
2990
		clear_huge_page(page, address, pages_per_huge_page(h));
N
Nick Piggin 已提交
2991
		__SetPageUptodate(page);
2992

2993
		if (vma->vm_flags & VM_MAYSHARE) {
2994
			int err;
K
Ken Chen 已提交
2995
			struct inode *inode = mapping->host;
2996 2997 2998 2999 3000 3001 3002 3003

			err = add_to_page_cache(page, mapping, idx, GFP_KERNEL);
			if (err) {
				put_page(page);
				if (err == -EEXIST)
					goto retry;
				goto out;
			}
3004
			ClearPagePrivate(page);
K
Ken Chen 已提交
3005 3006

			spin_lock(&inode->i_lock);
3007
			inode->i_blocks += blocks_per_huge_page(h);
K
Ken Chen 已提交
3008
			spin_unlock(&inode->i_lock);
3009
		} else {
3010
			lock_page(page);
3011 3012 3013 3014
			if (unlikely(anon_vma_prepare(vma))) {
				ret = VM_FAULT_OOM;
				goto backout_unlocked;
			}
3015
			anon_rmap = 1;
3016
		}
3017
	} else {
3018 3019 3020 3021 3022 3023
		/*
		 * 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))) {
3024
			ret = VM_FAULT_HWPOISON |
3025
				VM_FAULT_SET_HINDEX(hstate_index(h));
3026 3027
			goto backout_unlocked;
		}
3028
	}
3029

3030 3031 3032 3033 3034 3035
	/*
	 * 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.
	 */
3036
	if ((flags & FAULT_FLAG_WRITE) && !(vma->vm_flags & VM_SHARED))
3037 3038 3039 3040
		if (vma_needs_reservation(h, vma, address) < 0) {
			ret = VM_FAULT_OOM;
			goto backout_unlocked;
		}
3041

3042 3043
	ptl = huge_pte_lockptr(h, mm, ptep);
	spin_lock(ptl);
3044
	size = i_size_read(mapping->host) >> huge_page_shift(h);
A
Adam Litke 已提交
3045 3046 3047
	if (idx >= size)
		goto backout;

N
Nick Piggin 已提交
3048
	ret = 0;
3049
	if (!huge_pte_none(huge_ptep_get(ptep)))
A
Adam Litke 已提交
3050 3051
		goto backout;

3052 3053
	if (anon_rmap) {
		ClearPagePrivate(page);
3054
		hugepage_add_new_anon_rmap(page, vma, address);
3055
	} else
3056
		page_dup_rmap(page);
3057 3058 3059 3060
	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);

3061
	if ((flags & FAULT_FLAG_WRITE) && !(vma->vm_flags & VM_SHARED)) {
3062
		/* Optimization, do the COW without a second fault */
3063
		ret = hugetlb_cow(mm, vma, address, ptep, new_pte, page, ptl);
3064 3065
	}

3066
	spin_unlock(ptl);
A
Adam Litke 已提交
3067 3068
	unlock_page(page);
out:
3069
	return ret;
A
Adam Litke 已提交
3070 3071

backout:
3072
	spin_unlock(ptl);
3073
backout_unlocked:
A
Adam Litke 已提交
3074 3075 3076
	unlock_page(page);
	put_page(page);
	goto out;
3077 3078
}

3079 3080 3081 3082 3083 3084 3085 3086 3087 3088 3089 3090 3091 3092 3093 3094 3095 3096 3097 3098 3099 3100 3101 3102 3103 3104 3105 3106 3107 3108 3109 3110 3111 3112 3113
#ifdef CONFIG_SMP
static u32 fault_mutex_hash(struct hstate *h, struct mm_struct *mm,
			    struct vm_area_struct *vma,
			    struct address_space *mapping,
			    pgoff_t idx, unsigned long address)
{
	unsigned long key[2];
	u32 hash;

	if (vma->vm_flags & VM_SHARED) {
		key[0] = (unsigned long) mapping;
		key[1] = idx;
	} else {
		key[0] = (unsigned long) mm;
		key[1] = address >> huge_page_shift(h);
	}

	hash = jhash2((u32 *)&key, sizeof(key)/sizeof(u32), 0);

	return hash & (num_fault_mutexes - 1);
}
#else
/*
 * For uniprocesor systems we always use a single mutex, so just
 * return 0 and avoid the hashing overhead.
 */
static u32 fault_mutex_hash(struct hstate *h, struct mm_struct *mm,
			    struct vm_area_struct *vma,
			    struct address_space *mapping,
			    pgoff_t idx, unsigned long address)
{
	return 0;
}
#endif

3114
int hugetlb_fault(struct mm_struct *mm, struct vm_area_struct *vma,
3115
			unsigned long address, unsigned int flags)
3116
{
3117
	pte_t *ptep, entry;
3118
	spinlock_t *ptl;
3119
	int ret;
3120 3121
	u32 hash;
	pgoff_t idx;
3122
	struct page *page = NULL;
3123
	struct page *pagecache_page = NULL;
3124
	struct hstate *h = hstate_vma(vma);
3125
	struct address_space *mapping;
3126

3127 3128
	address &= huge_page_mask(h);

3129 3130 3131
	ptep = huge_pte_offset(mm, address);
	if (ptep) {
		entry = huge_ptep_get(ptep);
N
Naoya Horiguchi 已提交
3132
		if (unlikely(is_hugetlb_entry_migration(entry))) {
3133
			migration_entry_wait_huge(vma, mm, ptep);
N
Naoya Horiguchi 已提交
3134 3135
			return 0;
		} else if (unlikely(is_hugetlb_entry_hwpoisoned(entry)))
3136
			return VM_FAULT_HWPOISON_LARGE |
3137
				VM_FAULT_SET_HINDEX(hstate_index(h));
3138 3139
	}

3140
	ptep = huge_pte_alloc(mm, address, huge_page_size(h));
3141 3142 3143
	if (!ptep)
		return VM_FAULT_OOM;

3144 3145 3146
	mapping = vma->vm_file->f_mapping;
	idx = vma_hugecache_offset(h, vma, address);

3147 3148 3149 3150 3151
	/*
	 * 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.
	 */
3152 3153 3154
	hash = fault_mutex_hash(h, mm, vma, mapping, idx, address);
	mutex_lock(&htlb_fault_mutex_table[hash]);

3155 3156
	entry = huge_ptep_get(ptep);
	if (huge_pte_none(entry)) {
3157
		ret = hugetlb_no_page(mm, vma, mapping, idx, address, ptep, flags);
3158
		goto out_mutex;
3159
	}
3160

N
Nick Piggin 已提交
3161
	ret = 0;
3162

3163 3164 3165 3166 3167 3168 3169 3170
	/*
	 * 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.
	 */
3171
	if ((flags & FAULT_FLAG_WRITE) && !huge_pte_write(entry)) {
3172 3173
		if (vma_needs_reservation(h, vma, address) < 0) {
			ret = VM_FAULT_OOM;
3174
			goto out_mutex;
3175
		}
3176

3177
		if (!(vma->vm_flags & VM_MAYSHARE))
3178 3179 3180 3181
			pagecache_page = hugetlbfs_pagecache_page(h,
								vma, address);
	}

3182 3183 3184 3185 3186 3187 3188 3189
	/*
	 * 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);
3190
	get_page(page);
3191
	if (page != pagecache_page)
3192 3193
		lock_page(page);

3194 3195
	ptl = huge_pte_lockptr(h, mm, ptep);
	spin_lock(ptl);
3196
	/* Check for a racing update before calling hugetlb_cow */
3197
	if (unlikely(!pte_same(entry, huge_ptep_get(ptep))))
3198
		goto out_ptl;
3199 3200


3201
	if (flags & FAULT_FLAG_WRITE) {
3202
		if (!huge_pte_write(entry)) {
3203
			ret = hugetlb_cow(mm, vma, address, ptep, entry,
3204 3205
					pagecache_page, ptl);
			goto out_ptl;
3206
		}
3207
		entry = huge_pte_mkdirty(entry);
3208 3209
	}
	entry = pte_mkyoung(entry);
3210 3211
	if (huge_ptep_set_access_flags(vma, address, ptep, entry,
						flags & FAULT_FLAG_WRITE))
3212
		update_mmu_cache(vma, address, ptep);
3213

3214 3215
out_ptl:
	spin_unlock(ptl);
3216 3217 3218 3219 3220

	if (pagecache_page) {
		unlock_page(pagecache_page);
		put_page(pagecache_page);
	}
3221 3222
	if (page != pagecache_page)
		unlock_page(page);
3223
	put_page(page);
3224

3225
out_mutex:
3226
	mutex_unlock(&htlb_fault_mutex_table[hash]);
3227
	return ret;
3228 3229
}

3230 3231 3232 3233
long follow_hugetlb_page(struct mm_struct *mm, struct vm_area_struct *vma,
			 struct page **pages, struct vm_area_struct **vmas,
			 unsigned long *position, unsigned long *nr_pages,
			 long i, unsigned int flags)
D
David Gibson 已提交
3234
{
3235 3236
	unsigned long pfn_offset;
	unsigned long vaddr = *position;
3237
	unsigned long remainder = *nr_pages;
3238
	struct hstate *h = hstate_vma(vma);
D
David Gibson 已提交
3239 3240

	while (vaddr < vma->vm_end && remainder) {
A
Adam Litke 已提交
3241
		pte_t *pte;
3242
		spinlock_t *ptl = NULL;
H
Hugh Dickins 已提交
3243
		int absent;
A
Adam Litke 已提交
3244
		struct page *page;
D
David Gibson 已提交
3245

A
Adam Litke 已提交
3246 3247
		/*
		 * Some archs (sparc64, sh*) have multiple pte_ts to
H
Hugh Dickins 已提交
3248
		 * each hugepage.  We have to make sure we get the
A
Adam Litke 已提交
3249
		 * first, for the page indexing below to work.
3250 3251
		 *
		 * Note that page table lock is not held when pte is null.
A
Adam Litke 已提交
3252
		 */
3253
		pte = huge_pte_offset(mm, vaddr & huge_page_mask(h));
3254 3255
		if (pte)
			ptl = huge_pte_lock(h, mm, pte);
H
Hugh Dickins 已提交
3256 3257 3258 3259
		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
3260 3261 3262 3263
		 * 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
3264
		 */
H
Hugh Dickins 已提交
3265 3266
		if (absent && (flags & FOLL_DUMP) &&
		    !hugetlbfs_pagecache_present(h, vma, vaddr)) {
3267 3268
			if (pte)
				spin_unlock(ptl);
H
Hugh Dickins 已提交
3269 3270 3271
			remainder = 0;
			break;
		}
D
David Gibson 已提交
3272

3273 3274 3275 3276 3277 3278 3279 3280 3281 3282 3283
		/*
		 * We need call hugetlb_fault for both hugepages under migration
		 * (in which case hugetlb_fault waits for the migration,) and
		 * hwpoisoned hugepages (in which case we need to prevent the
		 * caller from accessing to them.) In order to do this, we use
		 * here is_swap_pte instead of is_hugetlb_entry_migration and
		 * is_hugetlb_entry_hwpoisoned. This is because it simply covers
		 * both cases, and because we can't follow correct pages
		 * directly from any kind of swap entries.
		 */
		if (absent || is_swap_pte(huge_ptep_get(pte)) ||
3284 3285
		    ((flags & FOLL_WRITE) &&
		      !huge_pte_write(huge_ptep_get(pte)))) {
A
Adam Litke 已提交
3286
			int ret;
D
David Gibson 已提交
3287

3288 3289
			if (pte)
				spin_unlock(ptl);
H
Hugh Dickins 已提交
3290 3291
			ret = hugetlb_fault(mm, vma, vaddr,
				(flags & FOLL_WRITE) ? FAULT_FLAG_WRITE : 0);
3292
			if (!(ret & VM_FAULT_ERROR))
A
Adam Litke 已提交
3293
				continue;
D
David Gibson 已提交
3294

A
Adam Litke 已提交
3295 3296 3297 3298
			remainder = 0;
			break;
		}

3299
		pfn_offset = (vaddr & ~huge_page_mask(h)) >> PAGE_SHIFT;
3300
		page = pte_page(huge_ptep_get(pte));
3301
same_page:
3302
		if (pages) {
H
Hugh Dickins 已提交
3303
			pages[i] = mem_map_offset(page, pfn_offset);
3304
			get_page_foll(pages[i]);
3305
		}
D
David Gibson 已提交
3306 3307 3308 3309 3310

		if (vmas)
			vmas[i] = vma;

		vaddr += PAGE_SIZE;
3311
		++pfn_offset;
D
David Gibson 已提交
3312 3313
		--remainder;
		++i;
3314
		if (vaddr < vma->vm_end && remainder &&
3315
				pfn_offset < pages_per_huge_page(h)) {
3316 3317 3318 3319 3320 3321
			/*
			 * We use pfn_offset to avoid touching the pageframes
			 * of this compound page.
			 */
			goto same_page;
		}
3322
		spin_unlock(ptl);
D
David Gibson 已提交
3323
	}
3324
	*nr_pages = remainder;
D
David Gibson 已提交
3325 3326
	*position = vaddr;

H
Hugh Dickins 已提交
3327
	return i ? i : -EFAULT;
D
David Gibson 已提交
3328
}
3329

3330
unsigned long hugetlb_change_protection(struct vm_area_struct *vma,
3331 3332 3333 3334 3335 3336
		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;
3337
	struct hstate *h = hstate_vma(vma);
3338
	unsigned long pages = 0;
3339 3340 3341 3342

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

3343
	mmu_notifier_invalidate_range_start(mm, start, end);
3344
	mutex_lock(&vma->vm_file->f_mapping->i_mmap_mutex);
3345
	for (; address < end; address += huge_page_size(h)) {
3346
		spinlock_t *ptl;
3347 3348 3349
		ptep = huge_pte_offset(mm, address);
		if (!ptep)
			continue;
3350
		ptl = huge_pte_lock(h, mm, ptep);
3351 3352
		if (huge_pmd_unshare(mm, &address, ptep)) {
			pages++;
3353
			spin_unlock(ptl);
3354
			continue;
3355
		}
3356
		if (!huge_pte_none(huge_ptep_get(ptep))) {
3357
			pte = huge_ptep_get_and_clear(mm, address, ptep);
3358
			pte = pte_mkhuge(huge_pte_modify(pte, newprot));
3359
			pte = arch_make_huge_pte(pte, vma, NULL, 0);
3360
			set_huge_pte_at(mm, address, ptep, pte);
3361
			pages++;
3362
		}
3363
		spin_unlock(ptl);
3364
	}
3365 3366 3367 3368 3369 3370
	/*
	 * Must flush TLB before releasing i_mmap_mutex: x86's huge_pmd_unshare
	 * may have cleared our pud entry and done put_page on the page table:
	 * once we release i_mmap_mutex, another task can do the final put_page
	 * and that page table be reused and filled with junk.
	 */
3371
	flush_tlb_range(vma, start, end);
3372
	mutex_unlock(&vma->vm_file->f_mapping->i_mmap_mutex);
3373
	mmu_notifier_invalidate_range_end(mm, start, end);
3374 3375

	return pages << h->order;
3376 3377
}

3378 3379
int hugetlb_reserve_pages(struct inode *inode,
					long from, long to,
3380
					struct vm_area_struct *vma,
3381
					vm_flags_t vm_flags)
3382
{
3383
	long ret, chg;
3384
	struct hstate *h = hstate_inode(inode);
3385
	struct hugepage_subpool *spool = subpool_inode(inode);
3386
	struct resv_map *resv_map;
3387

3388 3389 3390
	/*
	 * Only apply hugepage reservation if asked. At fault time, an
	 * attempt will be made for VM_NORESERVE to allocate a page
3391
	 * without using reserves
3392
	 */
3393
	if (vm_flags & VM_NORESERVE)
3394 3395
		return 0;

3396 3397 3398 3399 3400 3401
	/*
	 * 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
	 */
3402
	if (!vma || vma->vm_flags & VM_MAYSHARE) {
3403
		resv_map = inode_resv_map(inode);
3404

3405
		chg = region_chg(resv_map, from, to);
3406 3407 3408

	} else {
		resv_map = resv_map_alloc();
3409 3410 3411
		if (!resv_map)
			return -ENOMEM;

3412
		chg = to - from;
3413

3414 3415 3416 3417
		set_vma_resv_map(vma, resv_map);
		set_vma_resv_flags(vma, HPAGE_RESV_OWNER);
	}

3418 3419 3420 3421
	if (chg < 0) {
		ret = chg;
		goto out_err;
	}
3422

3423
	/* There must be enough pages in the subpool for the mapping */
3424 3425 3426 3427
	if (hugepage_subpool_get_pages(spool, chg)) {
		ret = -ENOSPC;
		goto out_err;
	}
3428 3429

	/*
3430
	 * Check enough hugepages are available for the reservation.
3431
	 * Hand the pages back to the subpool if there are not
3432
	 */
3433
	ret = hugetlb_acct_memory(h, chg);
K
Ken Chen 已提交
3434
	if (ret < 0) {
3435
		hugepage_subpool_put_pages(spool, chg);
3436
		goto out_err;
K
Ken Chen 已提交
3437
	}
3438 3439 3440 3441 3442 3443 3444 3445 3446 3447 3448 3449

	/*
	 * 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
	 */
3450
	if (!vma || vma->vm_flags & VM_MAYSHARE)
3451
		region_add(resv_map, from, to);
3452
	return 0;
3453
out_err:
J
Joonsoo Kim 已提交
3454 3455
	if (vma && is_vma_resv_set(vma, HPAGE_RESV_OWNER))
		kref_put(&resv_map->refs, resv_map_release);
3456
	return ret;
3457 3458 3459 3460
}

void hugetlb_unreserve_pages(struct inode *inode, long offset, long freed)
{
3461
	struct hstate *h = hstate_inode(inode);
3462
	struct resv_map *resv_map = inode_resv_map(inode);
3463
	long chg = 0;
3464
	struct hugepage_subpool *spool = subpool_inode(inode);
K
Ken Chen 已提交
3465

3466
	if (resv_map)
3467
		chg = region_truncate(resv_map, offset);
K
Ken Chen 已提交
3468
	spin_lock(&inode->i_lock);
3469
	inode->i_blocks -= (blocks_per_huge_page(h) * freed);
K
Ken Chen 已提交
3470 3471
	spin_unlock(&inode->i_lock);

3472
	hugepage_subpool_put_pages(spool, (chg - freed));
3473
	hugetlb_acct_memory(h, -(chg - freed));
3474
}
3475

3476 3477 3478 3479 3480 3481 3482 3483 3484 3485 3486 3487 3488 3489 3490 3491 3492 3493 3494 3495 3496 3497 3498 3499 3500 3501 3502 3503 3504 3505 3506 3507 3508 3509 3510 3511 3512 3513 3514 3515 3516 3517 3518 3519 3520 3521 3522 3523 3524 3525 3526 3527 3528 3529 3530 3531 3532 3533 3534
#ifdef CONFIG_ARCH_WANT_HUGE_PMD_SHARE
static unsigned long page_table_shareable(struct vm_area_struct *svma,
				struct vm_area_struct *vma,
				unsigned long addr, pgoff_t idx)
{
	unsigned long saddr = ((idx - svma->vm_pgoff) << PAGE_SHIFT) +
				svma->vm_start;
	unsigned long sbase = saddr & PUD_MASK;
	unsigned long s_end = sbase + PUD_SIZE;

	/* Allow segments to share if only one is marked locked */
	unsigned long vm_flags = vma->vm_flags & ~VM_LOCKED;
	unsigned long svm_flags = svma->vm_flags & ~VM_LOCKED;

	/*
	 * match the virtual addresses, permission and the alignment of the
	 * page table page.
	 */
	if (pmd_index(addr) != pmd_index(saddr) ||
	    vm_flags != svm_flags ||
	    sbase < svma->vm_start || svma->vm_end < s_end)
		return 0;

	return saddr;
}

static int vma_shareable(struct vm_area_struct *vma, unsigned long addr)
{
	unsigned long base = addr & PUD_MASK;
	unsigned long end = base + PUD_SIZE;

	/*
	 * check on proper vm_flags and page table alignment
	 */
	if (vma->vm_flags & VM_MAYSHARE &&
	    vma->vm_start <= base && end <= vma->vm_end)
		return 1;
	return 0;
}

/*
 * Search for a shareable pmd page for hugetlb. In any case calls pmd_alloc()
 * and returns the corresponding pte. While this is not necessary for the
 * !shared pmd case because we can allocate the pmd later as well, it makes the
 * code much cleaner. pmd allocation is essential for the shared case because
 * pud has to be populated inside the same i_mmap_mutex section - otherwise
 * racing tasks could either miss the sharing (see huge_pte_offset) or select a
 * bad pmd for sharing.
 */
pte_t *huge_pmd_share(struct mm_struct *mm, unsigned long addr, pud_t *pud)
{
	struct vm_area_struct *vma = find_vma(mm, addr);
	struct address_space *mapping = vma->vm_file->f_mapping;
	pgoff_t idx = ((addr - vma->vm_start) >> PAGE_SHIFT) +
			vma->vm_pgoff;
	struct vm_area_struct *svma;
	unsigned long saddr;
	pte_t *spte = NULL;
	pte_t *pte;
3535
	spinlock_t *ptl;
3536 3537 3538 3539 3540 3541 3542 3543 3544 3545 3546 3547 3548 3549 3550 3551 3552 3553 3554 3555 3556 3557

	if (!vma_shareable(vma, addr))
		return (pte_t *)pmd_alloc(mm, pud, addr);

	mutex_lock(&mapping->i_mmap_mutex);
	vma_interval_tree_foreach(svma, &mapping->i_mmap, idx, idx) {
		if (svma == vma)
			continue;

		saddr = page_table_shareable(svma, vma, addr, idx);
		if (saddr) {
			spte = huge_pte_offset(svma->vm_mm, saddr);
			if (spte) {
				get_page(virt_to_page(spte));
				break;
			}
		}
	}

	if (!spte)
		goto out;

3558 3559
	ptl = huge_pte_lockptr(hstate_vma(vma), mm, spte);
	spin_lock(ptl);
3560 3561 3562 3563 3564
	if (pud_none(*pud))
		pud_populate(mm, pud,
				(pmd_t *)((unsigned long)spte & PAGE_MASK));
	else
		put_page(virt_to_page(spte));
3565
	spin_unlock(ptl);
3566 3567 3568 3569 3570 3571 3572 3573 3574 3575 3576 3577 3578
out:
	pte = (pte_t *)pmd_alloc(mm, pud, addr);
	mutex_unlock(&mapping->i_mmap_mutex);
	return pte;
}

/*
 * unmap huge page backed by shared pte.
 *
 * Hugetlb pte page is ref counted at the time of mapping.  If pte is shared
 * indicated by page_count > 1, unmap is achieved by clearing pud and
 * decrementing the ref count. If count == 1, the pte page is not shared.
 *
3579
 * called with page table lock held.
3580 3581 3582 3583 3584 3585 3586 3587 3588 3589 3590 3591 3592 3593 3594 3595 3596 3597
 *
 * returns: 1 successfully unmapped a shared pte page
 *	    0 the underlying pte page is not shared, or it is the last user
 */
int huge_pmd_unshare(struct mm_struct *mm, unsigned long *addr, pte_t *ptep)
{
	pgd_t *pgd = pgd_offset(mm, *addr);
	pud_t *pud = pud_offset(pgd, *addr);

	BUG_ON(page_count(virt_to_page(ptep)) == 0);
	if (page_count(virt_to_page(ptep)) == 1)
		return 0;

	pud_clear(pud);
	put_page(virt_to_page(ptep));
	*addr = ALIGN(*addr, HPAGE_SIZE * PTRS_PER_PTE) - HPAGE_SIZE;
	return 1;
}
3598 3599 3600 3601 3602 3603 3604
#define want_pmd_share()	(1)
#else /* !CONFIG_ARCH_WANT_HUGE_PMD_SHARE */
pte_t *huge_pmd_share(struct mm_struct *mm, unsigned long addr, pud_t *pud)
{
	return NULL;
}
#define want_pmd_share()	(0)
3605 3606
#endif /* CONFIG_ARCH_WANT_HUGE_PMD_SHARE */

3607 3608 3609 3610 3611 3612 3613 3614 3615 3616 3617 3618 3619 3620 3621 3622 3623 3624 3625 3626 3627 3628 3629 3630 3631 3632 3633 3634 3635 3636 3637 3638 3639 3640 3641 3642 3643 3644 3645 3646 3647 3648 3649 3650 3651 3652 3653 3654 3655 3656 3657 3658 3659 3660 3661 3662 3663 3664 3665 3666 3667 3668 3669 3670 3671 3672 3673 3674 3675 3676 3677
#ifdef CONFIG_ARCH_WANT_GENERAL_HUGETLB
pte_t *huge_pte_alloc(struct mm_struct *mm,
			unsigned long addr, unsigned long sz)
{
	pgd_t *pgd;
	pud_t *pud;
	pte_t *pte = NULL;

	pgd = pgd_offset(mm, addr);
	pud = pud_alloc(mm, pgd, addr);
	if (pud) {
		if (sz == PUD_SIZE) {
			pte = (pte_t *)pud;
		} else {
			BUG_ON(sz != PMD_SIZE);
			if (want_pmd_share() && pud_none(*pud))
				pte = huge_pmd_share(mm, addr, pud);
			else
				pte = (pte_t *)pmd_alloc(mm, pud, addr);
		}
	}
	BUG_ON(pte && !pte_none(*pte) && !pte_huge(*pte));

	return pte;
}

pte_t *huge_pte_offset(struct mm_struct *mm, unsigned long addr)
{
	pgd_t *pgd;
	pud_t *pud;
	pmd_t *pmd = NULL;

	pgd = pgd_offset(mm, addr);
	if (pgd_present(*pgd)) {
		pud = pud_offset(pgd, addr);
		if (pud_present(*pud)) {
			if (pud_huge(*pud))
				return (pte_t *)pud;
			pmd = pmd_offset(pud, addr);
		}
	}
	return (pte_t *) pmd;
}

struct page *
follow_huge_pmd(struct mm_struct *mm, unsigned long address,
		pmd_t *pmd, int write)
{
	struct page *page;

	page = pte_page(*(pte_t *)pmd);
	if (page)
		page += ((address & ~PMD_MASK) >> PAGE_SHIFT);
	return page;
}

struct page *
follow_huge_pud(struct mm_struct *mm, unsigned long address,
		pud_t *pud, int write)
{
	struct page *page;

	page = pte_page(*(pte_t *)pud);
	if (page)
		page += ((address & ~PUD_MASK) >> PAGE_SHIFT);
	return page;
}

#else /* !CONFIG_ARCH_WANT_GENERAL_HUGETLB */

/* Can be overriden by architectures */
3678
struct page * __weak
3679 3680 3681 3682 3683 3684 3685 3686 3687
follow_huge_pud(struct mm_struct *mm, unsigned long address,
	       pud_t *pud, int write)
{
	BUG();
	return NULL;
}

#endif /* CONFIG_ARCH_WANT_GENERAL_HUGETLB */

3688 3689
#ifdef CONFIG_MEMORY_FAILURE

3690 3691 3692 3693 3694 3695 3696 3697 3698 3699 3700 3701 3702 3703
/* 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;
}

3704 3705 3706 3707
/*
 * This function is called from memory failure code.
 * Assume the caller holds page lock of the head page.
 */
3708
int dequeue_hwpoisoned_huge_page(struct page *hpage)
3709 3710 3711
{
	struct hstate *h = page_hstate(hpage);
	int nid = page_to_nid(hpage);
3712
	int ret = -EBUSY;
3713 3714

	spin_lock(&hugetlb_lock);
3715
	if (is_hugepage_on_freelist(hpage)) {
3716 3717 3718 3719 3720 3721 3722
		/*
		 * Hwpoisoned hugepage isn't linked to activelist or freelist,
		 * but dangling hpage->lru can trigger list-debug warnings
		 * (this happens when we call unpoison_memory() on it),
		 * so let it point to itself with list_del_init().
		 */
		list_del_init(&hpage->lru);
3723
		set_page_refcounted(hpage);
3724 3725 3726 3727
		h->free_huge_pages--;
		h->free_huge_pages_node[nid]--;
		ret = 0;
	}
3728
	spin_unlock(&hugetlb_lock);
3729
	return ret;
3730
}
3731
#endif
3732 3733 3734

bool isolate_huge_page(struct page *page, struct list_head *list)
{
3735
	VM_BUG_ON_PAGE(!PageHead(page), page);
3736 3737 3738 3739 3740 3741 3742 3743 3744 3745
	if (!get_page_unless_zero(page))
		return false;
	spin_lock(&hugetlb_lock);
	list_move_tail(&page->lru, list);
	spin_unlock(&hugetlb_lock);
	return true;
}

void putback_active_hugepage(struct page *page)
{
3746
	VM_BUG_ON_PAGE(!PageHead(page), page);
3747 3748 3749 3750 3751
	spin_lock(&hugetlb_lock);
	list_move_tail(&page->lru, &(page_hstate(page))->hugepage_activelist);
	spin_unlock(&hugetlb_lock);
	put_page(page);
}
3752 3753 3754

bool is_hugepage_active(struct page *page)
{
3755
	VM_BUG_ON_PAGE(!PageHuge(page), page);
3756 3757 3758 3759 3760 3761 3762 3763 3764 3765 3766 3767 3768 3769 3770 3771 3772 3773
	/*
	 * This function can be called for a tail page because the caller,
	 * scan_movable_pages, scans through a given pfn-range which typically
	 * covers one memory block. In systems using gigantic hugepage (1GB
	 * for x86_64,) a hugepage is larger than a memory block, and we don't
	 * support migrating such large hugepages for now, so return false
	 * when called for tail pages.
	 */
	if (PageTail(page))
		return false;
	/*
	 * Refcount of a hwpoisoned hugepages is 1, but they are not active,
	 * so we should return false for them.
	 */
	if (unlikely(PageHWPoison(page)))
		return false;
	return page_count(page) > 0;
}