hugetlb.c 107.0 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
int hugepages_treat_as_movable;
39

40
int hugetlb_max_hstate __read_mostly;
41 42
unsigned int default_hstate_idx;
struct hstate hstates[HUGE_MAX_HSTATE];
43 44 45 46 47
/*
 * Minimum page order among possible hugepage sizes, set to a proper value
 * at boot time.
 */
static unsigned int minimum_order __read_mostly = UINT_MAX;
48

49 50
__initdata LIST_HEAD(huge_boot_pages);

51 52 53
/* for command line parsing */
static struct hstate * __initdata parsed_hstate;
static unsigned long __initdata default_hstate_max_huge_pages;
54
static unsigned long __initdata default_hstate_size;
55

56
/*
57 58
 * Protects updates to hugepage_freelists, hugepage_activelist, nr_huge_pages,
 * free_huge_pages, and surplus_huge_pages.
59
 */
60
DEFINE_SPINLOCK(hugetlb_lock);
61

62 63 64 65 66 67 68
/*
 * 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;

69 70 71
/* Forward declaration */
static int hugetlb_acct_memory(struct hstate *h, long delta);

72 73 74 75 76 77 78
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
79 80 81 82 83 84
	 * remain, give up any reservations mased on minimum size and
	 * free the subpool */
	if (free) {
		if (spool->min_hpages != -1)
			hugetlb_acct_memory(spool->hstate,
						-spool->min_hpages);
85
		kfree(spool);
86
	}
87 88
}

89 90
struct hugepage_subpool *hugepage_new_subpool(struct hstate *h, long max_hpages,
						long min_hpages)
91 92 93
{
	struct hugepage_subpool *spool;

94
	spool = kzalloc(sizeof(*spool), GFP_KERNEL);
95 96 97 98 99
	if (!spool)
		return NULL;

	spin_lock_init(&spool->lock);
	spool->count = 1;
100 101 102 103 104 105 106 107 108
	spool->max_hpages = max_hpages;
	spool->hstate = h;
	spool->min_hpages = min_hpages;

	if (min_hpages != -1 && hugetlb_acct_memory(h, min_hpages)) {
		kfree(spool);
		return NULL;
	}
	spool->rsv_hpages = min_hpages;
109 110 111 112 113 114 115 116 117 118 119 120

	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);
}

121 122 123 124 125 126 127 128 129
/*
 * Subpool accounting for allocating and reserving pages.
 * Return -ENOMEM if there are not enough resources to satisfy the
 * the request.  Otherwise, return the number of pages by which the
 * global pools must be adjusted (upward).  The returned value may
 * only be different than the passed value (delta) in the case where
 * a subpool minimum size must be manitained.
 */
static long hugepage_subpool_get_pages(struct hugepage_subpool *spool,
130 131
				      long delta)
{
132
	long ret = delta;
133 134

	if (!spool)
135
		return ret;
136 137

	spin_lock(&spool->lock);
138 139 140 141 142 143 144 145

	if (spool->max_hpages != -1) {		/* maximum size accounting */
		if ((spool->used_hpages + delta) <= spool->max_hpages)
			spool->used_hpages += delta;
		else {
			ret = -ENOMEM;
			goto unlock_ret;
		}
146 147
	}

148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163
	if (spool->min_hpages != -1) {		/* minimum size accounting */
		if (delta > spool->rsv_hpages) {
			/*
			 * Asking for more reserves than those already taken on
			 * behalf of subpool.  Return difference.
			 */
			ret = delta - spool->rsv_hpages;
			spool->rsv_hpages = 0;
		} else {
			ret = 0;	/* reserves already accounted for */
			spool->rsv_hpages -= delta;
		}
	}

unlock_ret:
	spin_unlock(&spool->lock);
164 165 166
	return ret;
}

167 168 169 170 171 172 173
/*
 * Subpool accounting for freeing and unreserving pages.
 * Return the number of global page reservations that must be dropped.
 * The return value may only be different than the passed value (delta)
 * in the case where a subpool minimum size must be maintained.
 */
static long hugepage_subpool_put_pages(struct hugepage_subpool *spool,
174 175
				       long delta)
{
176 177
	long ret = delta;

178
	if (!spool)
179
		return delta;
180 181

	spin_lock(&spool->lock);
182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200

	if (spool->max_hpages != -1)		/* maximum size accounting */
		spool->used_hpages -= delta;

	if (spool->min_hpages != -1) {		/* minimum size accounting */
		if (spool->rsv_hpages + delta <= spool->min_hpages)
			ret = 0;
		else
			ret = spool->rsv_hpages + delta - spool->min_hpages;

		spool->rsv_hpages += delta;
		if (spool->rsv_hpages > spool->min_hpages)
			spool->rsv_hpages = spool->min_hpages;
	}

	/*
	 * If hugetlbfs_put_super couldn't free spool due to an outstanding
	 * quota reference, free it now.
	 */
201
	unlock_or_release_subpool(spool);
202 203

	return ret;
204 205 206 207 208 209 210 211 212
}

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
213
	return subpool_inode(file_inode(vma->vm_file));
214 215
}

216 217 218
/*
 * Region tracking -- allows tracking of reservations and instantiated pages
 *                    across the pages in a mapping.
219
 *
220 221 222 223 224 225 226 227 228 229 230 231 232 233
 * The region data structures are embedded into a resv_map and protected
 * by a resv_map's lock.  The set of regions within the resv_map represent
 * reservations for huge pages, or huge pages that have already been
 * instantiated within the map.  The from and to elements are huge page
 * indicies into the associated mapping.  from indicates the starting index
 * of the region.  to represents the first index past the end of  the region.
 *
 * For example, a file region structure with from == 0 and to == 4 represents
 * four huge pages in a mapping.  It is important to note that the to element
 * represents the first element past the end of the region. This is used in
 * arithmetic as 4(to) - 0(from) = 4 huge pages in the region.
 *
 * Interval notation of the form [from, to) will be used to indicate that
 * the endpoint from is inclusive and to is exclusive.
234 235 236 237 238 239 240
 */
struct file_region {
	struct list_head link;
	long from;
	long to;
};

241 242 243 244 245 246 247
/*
 * Add the huge page range represented by [f, t) to the reserve
 * map.  Existing regions will be expanded to accommodate the
 * specified range.  We know only existing regions need to be
 * expanded, because region_add is only called after region_chg
 * with the same range.  If a new file_region structure must
 * be allocated, it is done in region_chg.
248 249 250
 *
 * Return the number of new huge pages added to the map.  This
 * number is greater than or equal to zero.
251
 */
252
static long region_add(struct resv_map *resv, long f, long t)
253
{
254
	struct list_head *head = &resv->regions;
255
	struct file_region *rg, *nrg, *trg;
256
	long add = 0;
257

258
	spin_lock(&resv->lock);
259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281
	/* 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) {
282 283 284 285 286
			/* Decrement return value by the deleted range.
			 * Another range will span this area so that by
			 * end of routine add will be >= zero
			 */
			add -= (rg->to - rg->from);
287 288 289 290
			list_del(&rg->link);
			kfree(rg);
		}
	}
291 292

	add += (nrg->from - f);		/* Added to beginning of region */
293
	nrg->from = f;
294
	add += t - nrg->to;		/* Added to end of region */
295
	nrg->to = t;
296

297
	spin_unlock(&resv->lock);
298 299
	VM_BUG_ON(add < 0);
	return add;
300 301
}

302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320
/*
 * Examine the existing reserve map and determine how many
 * huge pages in the specified range [f, t) are NOT currently
 * represented.  This routine is called before a subsequent
 * call to region_add that will actually modify the reserve
 * map to add the specified range [f, t).  region_chg does
 * not change the number of huge pages represented by the
 * map.  However, if the existing regions in the map can not
 * be expanded to represent the new range, a new file_region
 * structure is added to the map as a placeholder.  This is
 * so that the subsequent region_add call will have all the
 * regions it needs and will not fail.
 *
 * Returns the number of huge pages that need to be added
 * to the existing reservation map for the range [f, t).
 * This number is greater or equal to zero.  -ENOMEM is
 * returned if a new file_region structure is needed and can
 * not be allocated.
 */
321
static long region_chg(struct resv_map *resv, long f, long t)
322
{
323
	struct list_head *head = &resv->regions;
324
	struct file_region *rg, *nrg = NULL;
325 326
	long chg = 0;

327 328
retry:
	spin_lock(&resv->lock);
329 330 331 332 333 334 335 336 337
	/* 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) {
338 339 340 341 342 343 344 345 346 347 348
		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;
		}
349

350 351 352
		list_add(&nrg->link, rg->link.prev);
		chg = t - f;
		goto out_nrg;
353 354 355 356 357 358 359 360 361 362 363 364
	}

	/* 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)
365
			goto out;
366

L
Lucas De Marchi 已提交
367
		/* We overlap with this area, if it extends further than
368 369 370 371 372 373 374 375
		 * 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;
	}
376 377 378 379 380 381 382 383

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);
384 385 386
	return chg;
}

387 388 389 390 391
/*
 * Truncate the reserve map at index 'end'.  Modify/truncate any
 * region which contains end.  Delete any regions past end.
 * Return the number of huge pages removed from the map.
 */
392
static long region_truncate(struct resv_map *resv, long end)
393
{
394
	struct list_head *head = &resv->regions;
395 396 397
	struct file_region *rg, *trg;
	long chg = 0;

398
	spin_lock(&resv->lock);
399 400 401 402 403
	/* 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)
404
		goto out;
405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420

	/* 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);
	}
421 422 423

out:
	spin_unlock(&resv->lock);
424 425 426
	return chg;
}

427 428 429 430
/*
 * Count and return the number of huge pages in the reserve map
 * that intersect with the range [f, t).
 */
431
static long region_count(struct resv_map *resv, long f, long t)
432
{
433
	struct list_head *head = &resv->regions;
434 435 436
	struct file_region *rg;
	long chg = 0;

437
	spin_lock(&resv->lock);
438 439
	/* Locate each segment we overlap with, and count that overlap. */
	list_for_each_entry(rg, head, link) {
440 441
		long seg_from;
		long seg_to;
442 443 444 445 446 447 448 449 450 451 452

		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;
	}
453
	spin_unlock(&resv->lock);
454 455 456 457

	return chg;
}

458 459 460 461
/*
 * Convert the address within this vma to the page offset within
 * the mapping, in pagecache page units; huge pages here.
 */
462 463
static pgoff_t vma_hugecache_offset(struct hstate *h,
			struct vm_area_struct *vma, unsigned long address)
464
{
465 466
	return ((address - vma->vm_start) >> huge_page_shift(h)) +
			(vma->vm_pgoff >> huge_page_order(h));
467 468
}

469 470 471 472 473 474
pgoff_t linear_hugepage_index(struct vm_area_struct *vma,
				     unsigned long address)
{
	return vma_hugecache_offset(hstate_vma(vma), vma, address);
}

475 476 477 478 479 480 481 482 483 484 485 486 487
/*
 * 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);

488
	return 1UL << huge_page_shift(hstate);
489
}
490
EXPORT_SYMBOL_GPL(vma_kernel_pagesize);
491

492 493 494 495 496 497 498 499 500 501 502 503 504
/*
 * 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

505 506 507 508 509 510 511
/*
 * 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)
512
#define HPAGE_RESV_MASK (HPAGE_RESV_OWNER | HPAGE_RESV_UNMAPPED)
513

514 515 516 517 518 519 520 521 522
/*
 * 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.
523 524 525 526 527 528 529 530 531
 *
 * 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.
532
 */
533 534 535 536 537 538 539 540 541 542 543
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;
}

544
struct resv_map *resv_map_alloc(void)
545 546 547 548 549 550
{
	struct resv_map *resv_map = kmalloc(sizeof(*resv_map), GFP_KERNEL);
	if (!resv_map)
		return NULL;

	kref_init(&resv_map->refs);
551
	spin_lock_init(&resv_map->lock);
552 553 554 555 556
	INIT_LIST_HEAD(&resv_map->regions);

	return resv_map;
}

557
void resv_map_release(struct kref *ref)
558 559 560 561
{
	struct resv_map *resv_map = container_of(ref, struct resv_map, refs);

	/* Clear out any active regions before we release the map. */
562
	region_truncate(resv_map, 0);
563 564 565
	kfree(resv_map);
}

566 567 568 569 570
static inline struct resv_map *inode_resv_map(struct inode *inode)
{
	return inode->i_mapping->private_data;
}

571
static struct resv_map *vma_resv_map(struct vm_area_struct *vma)
572
{
573
	VM_BUG_ON_VMA(!is_vm_hugetlb_page(vma), vma);
574 575 576 577 578 579 580
	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 {
581 582
		return (struct resv_map *)(get_vma_private_data(vma) &
							~HPAGE_RESV_MASK);
583
	}
584 585
}

586
static void set_vma_resv_map(struct vm_area_struct *vma, struct resv_map *map)
587
{
588 589
	VM_BUG_ON_VMA(!is_vm_hugetlb_page(vma), vma);
	VM_BUG_ON_VMA(vma->vm_flags & VM_MAYSHARE, vma);
590

591 592
	set_vma_private_data(vma, (get_vma_private_data(vma) &
				HPAGE_RESV_MASK) | (unsigned long)map);
593 594 595 596
}

static void set_vma_resv_flags(struct vm_area_struct *vma, unsigned long flags)
{
597 598
	VM_BUG_ON_VMA(!is_vm_hugetlb_page(vma), vma);
	VM_BUG_ON_VMA(vma->vm_flags & VM_MAYSHARE, vma);
599 600

	set_vma_private_data(vma, get_vma_private_data(vma) | flags);
601 602 603 604
}

static int is_vma_resv_set(struct vm_area_struct *vma, unsigned long flag)
{
605
	VM_BUG_ON_VMA(!is_vm_hugetlb_page(vma), vma);
606 607

	return (get_vma_private_data(vma) & flag) != 0;
608 609
}

610
/* Reset counters to 0 and clear all HPAGE_RESV_* flags */
611 612
void reset_vma_resv_huge_pages(struct vm_area_struct *vma)
{
613
	VM_BUG_ON_VMA(!is_vm_hugetlb_page(vma), vma);
614
	if (!(vma->vm_flags & VM_MAYSHARE))
615 616 617 618
		vma->vm_private_data = (void *)0;
}

/* Returns true if the VMA has associated reserve pages */
619
static int vma_has_reserves(struct vm_area_struct *vma, long chg)
620
{
621 622 623 624 625 626 627 628 629 630 631 632 633 634 635
	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;
	}
636 637

	/* Shared mappings always use reserves */
638
	if (vma->vm_flags & VM_MAYSHARE)
639
		return 1;
640 641 642 643 644

	/*
	 * Only the process that called mmap() has reserves for
	 * private mappings.
	 */
645 646
	if (is_vma_resv_set(vma, HPAGE_RESV_OWNER))
		return 1;
647

648
	return 0;
649 650
}

651
static void enqueue_huge_page(struct hstate *h, struct page *page)
L
Linus Torvalds 已提交
652 653
{
	int nid = page_to_nid(page);
654
	list_move(&page->lru, &h->hugepage_freelists[nid]);
655 656
	h->free_huge_pages++;
	h->free_huge_pages_node[nid]++;
L
Linus Torvalds 已提交
657 658
}

659 660 661 662
static struct page *dequeue_huge_page_node(struct hstate *h, int nid)
{
	struct page *page;

663 664 665 666 667 668 669 670
	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)
671
		return NULL;
672
	list_move(&page->lru, &h->hugepage_activelist);
673
	set_page_refcounted(page);
674 675 676 677 678
	h->free_huge_pages--;
	h->free_huge_pages_node[nid]--;
	return page;
}

679 680 681
/* Movability of hugepages depends on migration support. */
static inline gfp_t htlb_alloc_mask(struct hstate *h)
{
682
	if (hugepages_treat_as_movable || hugepage_migration_supported(h))
683 684 685 686 687
		return GFP_HIGHUSER_MOVABLE;
	else
		return GFP_HIGHUSER;
}

688 689
static struct page *dequeue_huge_page_vma(struct hstate *h,
				struct vm_area_struct *vma,
690 691
				unsigned long address, int avoid_reserve,
				long chg)
L
Linus Torvalds 已提交
692
{
693
	struct page *page = NULL;
694
	struct mempolicy *mpol;
695
	nodemask_t *nodemask;
696
	struct zonelist *zonelist;
697 698
	struct zone *zone;
	struct zoneref *z;
699
	unsigned int cpuset_mems_cookie;
L
Linus Torvalds 已提交
700

701 702 703 704 705
	/*
	 * 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
	 */
706
	if (!vma_has_reserves(vma, chg) &&
707
			h->free_huge_pages - h->resv_huge_pages == 0)
708
		goto err;
709

710
	/* If reserves cannot be used, ensure enough pages are in the pool */
711
	if (avoid_reserve && h->free_huge_pages - h->resv_huge_pages == 0)
712
		goto err;
713

714
retry_cpuset:
715
	cpuset_mems_cookie = read_mems_allowed_begin();
716
	zonelist = huge_zonelist(vma, address,
717
					htlb_alloc_mask(h), &mpol, &nodemask);
718

719 720
	for_each_zone_zonelist_nodemask(zone, z, zonelist,
						MAX_NR_ZONES - 1, nodemask) {
721
		if (cpuset_zone_allowed(zone, htlb_alloc_mask(h))) {
722 723
			page = dequeue_huge_page_node(h, zone_to_nid(zone));
			if (page) {
724 725 726 727 728
				if (avoid_reserve)
					break;
				if (!vma_has_reserves(vma, chg))
					break;

729
				SetPagePrivate(page);
730
				h->resv_huge_pages--;
731 732
				break;
			}
A
Andrew Morton 已提交
733
		}
L
Linus Torvalds 已提交
734
	}
735

736
	mpol_cond_put(mpol);
737
	if (unlikely(!page && read_mems_allowed_retry(cpuset_mems_cookie)))
738
		goto retry_cpuset;
L
Linus Torvalds 已提交
739
	return page;
740 741 742

err:
	return NULL;
L
Linus Torvalds 已提交
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
/*
 * 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--)

818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955
#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

956
static void update_and_free_page(struct hstate *h, struct page *page)
A
Adam Litke 已提交
957 958
{
	int i;
959

960 961
	if (hstate_is_gigantic(h) && !gigantic_page_supported())
		return;
962

963 964 965
	h->nr_huge_pages--;
	h->nr_huge_pages_node[page_to_nid(page)]--;
	for (i = 0; i < pages_per_huge_page(h); i++) {
966 967
		page[i].flags &= ~(1 << PG_locked | 1 << PG_error |
				1 << PG_referenced | 1 << PG_dirty |
968 969
				1 << PG_active | 1 << PG_private |
				1 << PG_writeback);
A
Adam Litke 已提交
970
	}
971
	VM_BUG_ON_PAGE(hugetlb_cgroup_from_page(page), page);
A
Adam Litke 已提交
972 973
	set_compound_page_dtor(page, NULL);
	set_page_refcounted(page);
974 975 976 977 978 979
	if (hstate_is_gigantic(h)) {
		destroy_compound_gigantic_page(page, huge_page_order(h));
		free_gigantic_page(page, huge_page_order(h));
	} else {
		__free_pages(page, huge_page_order(h));
	}
A
Adam Litke 已提交
980 981
}

982 983 984 985 986 987 988 989 990 991 992
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;
}

993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017
/*
 * Test to determine whether the hugepage is "active/in-use" (i.e. being linked
 * to hstate->hugepage_activelist.)
 *
 * This function can be called for tail pages, but never returns true for them.
 */
bool page_huge_active(struct page *page)
{
	VM_BUG_ON_PAGE(!PageHuge(page), page);
	return PageHead(page) && PagePrivate(&page[1]);
}

/* never called for tail page */
static void set_page_huge_active(struct page *page)
{
	VM_BUG_ON_PAGE(!PageHeadHuge(page), page);
	SetPagePrivate(&page[1]);
}

static void clear_page_huge_active(struct page *page)
{
	VM_BUG_ON_PAGE(!PageHeadHuge(page), page);
	ClearPagePrivate(&page[1]);
}

1018
void free_huge_page(struct page *page)
1019
{
1020 1021 1022 1023
	/*
	 * Can't pass hstate in here because it is called from the
	 * compound page destructor.
	 */
1024
	struct hstate *h = page_hstate(page);
1025
	int nid = page_to_nid(page);
1026 1027
	struct hugepage_subpool *spool =
		(struct hugepage_subpool *)page_private(page);
1028
	bool restore_reserve;
1029

1030
	set_page_private(page, 0);
1031
	page->mapping = NULL;
1032
	BUG_ON(page_count(page));
1033
	BUG_ON(page_mapcount(page));
1034
	restore_reserve = PagePrivate(page);
1035
	ClearPagePrivate(page);
1036

1037 1038 1039 1040 1041 1042 1043 1044
	/*
	 * A return code of zero implies that the subpool will be under its
	 * minimum size if the reservation is not restored after page is free.
	 * Therefore, force restore_reserve operation.
	 */
	if (hugepage_subpool_put_pages(spool, 1) == 0)
		restore_reserve = true;

1045
	spin_lock(&hugetlb_lock);
1046
	clear_page_huge_active(page);
1047 1048
	hugetlb_cgroup_uncharge_page(hstate_index(h),
				     pages_per_huge_page(h), page);
1049 1050 1051
	if (restore_reserve)
		h->resv_huge_pages++;

1052
	if (h->surplus_huge_pages_node[nid]) {
1053 1054
		/* remove the page from active list */
		list_del(&page->lru);
1055 1056 1057
		update_and_free_page(h, page);
		h->surplus_huge_pages--;
		h->surplus_huge_pages_node[nid]--;
1058
	} else {
1059
		arch_clear_hugepage_flags(page);
1060
		enqueue_huge_page(h, page);
1061
	}
1062 1063 1064
	spin_unlock(&hugetlb_lock);
}

1065
static void prep_new_huge_page(struct hstate *h, struct page *page, int nid)
1066
{
1067
	INIT_LIST_HEAD(&page->lru);
1068 1069
	set_compound_page_dtor(page, free_huge_page);
	spin_lock(&hugetlb_lock);
1070
	set_hugetlb_cgroup(page, NULL);
1071 1072
	h->nr_huge_pages++;
	h->nr_huge_pages_node[nid]++;
1073 1074 1075 1076
	spin_unlock(&hugetlb_lock);
	put_page(page); /* free it into the hugepage allocator */
}

1077
static void prep_compound_gigantic_page(struct page *page, unsigned long order)
1078 1079 1080 1081 1082 1083 1084 1085
{
	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);
1086
	__ClearPageReserved(page);
1087
	for (i = 1; i < nr_pages; i++, p = mem_map_next(p, page, i)) {
1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100
		/*
		 * 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);
1101
		set_page_count(p, 0);
1102
		p->first_page = page;
1103 1104 1105
		/* Make sure p->first_page is always valid for PageTail() */
		smp_wmb();
		__SetPageTail(p);
1106 1107 1108
	}
}

A
Andrew Morton 已提交
1109 1110 1111 1112 1113
/*
 * PageHuge() only returns true for hugetlbfs pages, but not for normal or
 * transparent huge pages.  See the PageTransHuge() documentation for more
 * details.
 */
1114 1115 1116 1117 1118 1119
int PageHuge(struct page *page)
{
	if (!PageCompound(page))
		return 0;

	page = compound_head(page);
1120
	return get_compound_page_dtor(page) == free_huge_page;
1121
}
1122 1123
EXPORT_SYMBOL_GPL(PageHuge);

1124 1125 1126 1127 1128 1129 1130 1131 1132
/*
 * 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;

1133
	return get_compound_page_dtor(page_head) == free_huge_page;
1134 1135
}

1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152
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;
}

1153
static struct page *alloc_fresh_huge_page_node(struct hstate *h, int nid)
L
Linus Torvalds 已提交
1154 1155
{
	struct page *page;
1156

1157
	page = alloc_pages_exact_node(nid,
1158
		htlb_alloc_mask(h)|__GFP_COMP|__GFP_THISNODE|
1159
						__GFP_REPEAT|__GFP_NOWARN,
1160
		huge_page_order(h));
L
Linus Torvalds 已提交
1161
	if (page) {
1162
		prep_new_huge_page(h, page, nid);
L
Linus Torvalds 已提交
1163
	}
1164 1165 1166 1167

	return page;
}

1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189
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;
}

1190 1191 1192 1193 1194 1195
/*
 * 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.
 */
1196 1197
static int free_pool_huge_page(struct hstate *h, nodemask_t *nodes_allowed,
							 bool acct_surplus)
1198
{
1199
	int nr_nodes, node;
1200 1201
	int ret = 0;

1202
	for_each_node_mask_to_free(h, nr_nodes, node, nodes_allowed) {
1203 1204 1205 1206
		/*
		 * If we're returning unused surplus pages, only examine
		 * nodes with surplus pages.
		 */
1207 1208
		if ((!acct_surplus || h->surplus_huge_pages_node[node]) &&
		    !list_empty(&h->hugepage_freelists[node])) {
1209
			struct page *page =
1210
				list_entry(h->hugepage_freelists[node].next,
1211 1212 1213
					  struct page, lru);
			list_del(&page->lru);
			h->free_huge_pages--;
1214
			h->free_huge_pages_node[node]--;
1215 1216
			if (acct_surplus) {
				h->surplus_huge_pages--;
1217
				h->surplus_huge_pages_node[node]--;
1218
			}
1219 1220
			update_and_free_page(h, page);
			ret = 1;
1221
			break;
1222
		}
1223
	}
1224 1225 1226 1227

	return ret;
}

1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254
/*
 * 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 long pfn;

1255 1256 1257
	if (!hugepages_supported())
		return;

1258 1259
	VM_BUG_ON(!IS_ALIGNED(start_pfn, 1 << minimum_order));
	for (pfn = start_pfn; pfn < end_pfn; pfn += 1 << minimum_order)
1260 1261 1262
		dissolve_free_huge_page(pfn_to_page(pfn));
}

1263
static struct page *alloc_buddy_huge_page(struct hstate *h, int nid)
1264 1265
{
	struct page *page;
1266
	unsigned int r_nid;
1267

1268
	if (hstate_is_gigantic(h))
1269 1270
		return NULL;

1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294
	/*
	 * 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);
1295
	if (h->surplus_huge_pages >= h->nr_overcommit_huge_pages) {
1296 1297 1298
		spin_unlock(&hugetlb_lock);
		return NULL;
	} else {
1299 1300
		h->nr_huge_pages++;
		h->surplus_huge_pages++;
1301 1302 1303
	}
	spin_unlock(&hugetlb_lock);

1304
	if (nid == NUMA_NO_NODE)
1305
		page = alloc_pages(htlb_alloc_mask(h)|__GFP_COMP|
1306 1307 1308 1309
				   __GFP_REPEAT|__GFP_NOWARN,
				   huge_page_order(h));
	else
		page = alloc_pages_exact_node(nid,
1310
			htlb_alloc_mask(h)|__GFP_COMP|__GFP_THISNODE|
1311
			__GFP_REPEAT|__GFP_NOWARN, huge_page_order(h));
1312 1313

	spin_lock(&hugetlb_lock);
1314
	if (page) {
1315
		INIT_LIST_HEAD(&page->lru);
1316
		r_nid = page_to_nid(page);
1317
		set_compound_page_dtor(page, free_huge_page);
1318
		set_hugetlb_cgroup(page, NULL);
1319 1320 1321
		/*
		 * We incremented the global counters already
		 */
1322 1323
		h->nr_huge_pages_node[r_nid]++;
		h->surplus_huge_pages_node[r_nid]++;
1324
		__count_vm_event(HTLB_BUDDY_PGALLOC);
1325
	} else {
1326 1327
		h->nr_huge_pages--;
		h->surplus_huge_pages--;
1328
		__count_vm_event(HTLB_BUDDY_PGALLOC_FAIL);
1329
	}
1330
	spin_unlock(&hugetlb_lock);
1331 1332 1333 1334

	return page;
}

1335 1336 1337 1338 1339 1340 1341
/*
 * 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)
{
1342
	struct page *page = NULL;
1343 1344

	spin_lock(&hugetlb_lock);
1345 1346
	if (h->free_huge_pages - h->resv_huge_pages > 0)
		page = dequeue_huge_page_node(h, nid);
1347 1348
	spin_unlock(&hugetlb_lock);

1349
	if (!page)
1350 1351 1352 1353 1354
		page = alloc_buddy_huge_page(h, nid);

	return page;
}

1355
/*
L
Lucas De Marchi 已提交
1356
 * Increase the hugetlb pool such that it can accommodate a reservation
1357 1358
 * of size 'delta'.
 */
1359
static int gather_surplus_pages(struct hstate *h, int delta)
1360 1361 1362 1363 1364
{
	struct list_head surplus_list;
	struct page *page, *tmp;
	int ret, i;
	int needed, allocated;
1365
	bool alloc_ok = true;
1366

1367
	needed = (h->resv_huge_pages + delta) - h->free_huge_pages;
1368
	if (needed <= 0) {
1369
		h->resv_huge_pages += delta;
1370
		return 0;
1371
	}
1372 1373 1374 1375 1376 1377 1378 1379

	allocated = 0;
	INIT_LIST_HEAD(&surplus_list);

	ret = -ENOMEM;
retry:
	spin_unlock(&hugetlb_lock);
	for (i = 0; i < needed; i++) {
1380
		page = alloc_buddy_huge_page(h, NUMA_NO_NODE);
1381 1382 1383 1384
		if (!page) {
			alloc_ok = false;
			break;
		}
1385 1386
		list_add(&page->lru, &surplus_list);
	}
1387
	allocated += i;
1388 1389 1390 1391 1392 1393

	/*
	 * 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);
1394 1395
	needed = (h->resv_huge_pages + delta) -
			(h->free_huge_pages + allocated);
1396 1397 1398 1399 1400 1401 1402 1403 1404 1405
	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;
	}
1406 1407
	/*
	 * The surplus_list now contains _at_least_ the number of extra pages
L
Lucas De Marchi 已提交
1408
	 * needed to accommodate the reservation.  Add the appropriate number
1409
	 * of pages to the hugetlb pool and free the extras back to the buddy
1410 1411 1412
	 * 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.
1413 1414
	 */
	needed += allocated;
1415
	h->resv_huge_pages += delta;
1416
	ret = 0;
1417

1418
	/* Free the needed pages to the hugetlb pool */
1419
	list_for_each_entry_safe(page, tmp, &surplus_list, lru) {
1420 1421
		if ((--needed) < 0)
			break;
1422 1423 1424 1425 1426
		/*
		 * This page is now managed by the hugetlb allocator and has
		 * no users -- drop the buddy allocator's reference.
		 */
		put_page_testzero(page);
1427
		VM_BUG_ON_PAGE(page_count(page), page);
1428
		enqueue_huge_page(h, page);
1429
	}
1430
free:
1431
	spin_unlock(&hugetlb_lock);
1432 1433

	/* Free unnecessary surplus pages to the buddy allocator */
1434 1435
	list_for_each_entry_safe(page, tmp, &surplus_list, lru)
		put_page(page);
1436
	spin_lock(&hugetlb_lock);
1437 1438 1439 1440 1441 1442 1443 1444

	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.
1445
 * Called with hugetlb_lock held.
1446
 */
1447 1448
static void return_unused_surplus_pages(struct hstate *h,
					unsigned long unused_resv_pages)
1449 1450 1451
{
	unsigned long nr_pages;

1452
	/* Uncommit the reservation */
1453
	h->resv_huge_pages -= unused_resv_pages;
1454

1455
	/* Cannot return gigantic pages currently */
1456
	if (hstate_is_gigantic(h))
1457 1458
		return;

1459
	nr_pages = min(unused_resv_pages, h->surplus_huge_pages);
1460

1461 1462
	/*
	 * We want to release as many surplus pages as possible, spread
1463 1464 1465 1466 1467
	 * 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.
1468 1469
	 */
	while (nr_pages--) {
1470
		if (!free_pool_huge_page(h, &node_states[N_MEMORY], 1))
1471
			break;
1472
		cond_resched_lock(&hugetlb_lock);
1473 1474 1475
	}
}

1476
/*
1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491
 * vma_needs_reservation and vma_commit_reservation are used by the huge
 * page allocation routines to manage reservations.
 *
 * vma_needs_reservation is called to determine if the huge page at addr
 * within the vma has an associated reservation.  If a reservation is
 * needed, the value 1 is returned.  The caller is then responsible for
 * managing the global reservation and subpool usage counts.  After
 * the huge page has been allocated, vma_commit_reservation is called
 * to add the page to the reservation map.
 *
 * In the normal case, vma_commit_reservation returns the same value
 * as the preceding vma_needs_reservation call.  The only time this
 * is not the case is if a reserve map was changed between calls.  It
 * is the responsibility of the caller to notice the difference and
 * take appropriate action.
1492
 */
1493 1494 1495
static long __vma_reservation_common(struct hstate *h,
				struct vm_area_struct *vma, unsigned long addr,
				bool commit)
1496
{
1497 1498
	struct resv_map *resv;
	pgoff_t idx;
1499
	long ret;
1500

1501 1502
	resv = vma_resv_map(vma);
	if (!resv)
1503
		return 1;
1504

1505
	idx = vma_hugecache_offset(h, vma, addr);
1506 1507 1508 1509
	if (commit)
		ret = region_add(resv, idx, idx + 1);
	else
		ret = region_chg(resv, idx, idx + 1);
1510

1511
	if (vma->vm_flags & VM_MAYSHARE)
1512
		return ret;
1513
	else
1514
		return ret < 0 ? ret : 0;
1515
}
1516 1517

static long vma_needs_reservation(struct hstate *h,
1518
			struct vm_area_struct *vma, unsigned long addr)
1519
{
1520 1521
	return __vma_reservation_common(h, vma, addr, false);
}
1522

1523 1524 1525 1526
static long vma_commit_reservation(struct hstate *h,
			struct vm_area_struct *vma, unsigned long addr)
{
	return __vma_reservation_common(h, vma, addr, true);
1527 1528
}

1529
static struct page *alloc_huge_page(struct vm_area_struct *vma,
1530
				    unsigned long addr, int avoid_reserve)
L
Linus Torvalds 已提交
1531
{
1532
	struct hugepage_subpool *spool = subpool_vma(vma);
1533
	struct hstate *h = hstate_vma(vma);
1534
	struct page *page;
1535
	long chg, commit;
1536 1537
	int ret, idx;
	struct hugetlb_cgroup *h_cg;
1538

1539
	idx = hstate_index(h);
1540
	/*
1541 1542 1543 1544 1545 1546
	 * 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.
1547
	 */
1548
	chg = vma_needs_reservation(h, vma, addr);
1549
	if (chg < 0)
1550
		return ERR_PTR(-ENOMEM);
1551
	if (chg || avoid_reserve)
1552
		if (hugepage_subpool_get_pages(spool, 1) < 0)
1553
			return ERR_PTR(-ENOSPC);
L
Linus Torvalds 已提交
1554

1555
	ret = hugetlb_cgroup_charge_cgroup(idx, pages_per_huge_page(h), &h_cg);
1556 1557 1558
	if (ret)
		goto out_subpool_put;

L
Linus Torvalds 已提交
1559
	spin_lock(&hugetlb_lock);
1560
	page = dequeue_huge_page_vma(h, vma, addr, avoid_reserve, chg);
1561
	if (!page) {
1562
		spin_unlock(&hugetlb_lock);
1563
		page = alloc_buddy_huge_page(h, NUMA_NO_NODE);
1564 1565 1566
		if (!page)
			goto out_uncharge_cgroup;

1567 1568
		spin_lock(&hugetlb_lock);
		list_move(&page->lru, &h->hugepage_activelist);
1569
		/* Fall through */
K
Ken Chen 已提交
1570
	}
1571 1572
	hugetlb_cgroup_commit_charge(idx, pages_per_huge_page(h), h_cg, page);
	spin_unlock(&hugetlb_lock);
1573

1574
	set_page_private(page, (unsigned long)spool);
1575

1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591
	commit = vma_commit_reservation(h, vma, addr);
	if (unlikely(chg > commit)) {
		/*
		 * The page was added to the reservation map between
		 * vma_needs_reservation and vma_commit_reservation.
		 * This indicates a race with hugetlb_reserve_pages.
		 * Adjust for the subpool count incremented above AND
		 * in hugetlb_reserve_pages for the same page.  Also,
		 * the reservation count added in hugetlb_reserve_pages
		 * no longer applies.
		 */
		long rsv_adjust;

		rsv_adjust = hugepage_subpool_put_pages(spool, 1);
		hugetlb_acct_memory(h, -rsv_adjust);
	}
1592
	return page;
1593 1594 1595 1596 1597 1598 1599

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);
1600 1601
}

1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615
/*
 * 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;
}

1616
int __weak alloc_bootmem_huge_page(struct hstate *h)
1617 1618
{
	struct huge_bootmem_page *m;
1619
	int nr_nodes, node;
1620

1621
	for_each_node_mask_to_alloc(h, nr_nodes, node, &node_states[N_MEMORY]) {
1622 1623
		void *addr;

1624 1625 1626
		addr = memblock_virt_alloc_try_nid_nopanic(
				huge_page_size(h), huge_page_size(h),
				0, BOOTMEM_ALLOC_ACCESSIBLE, node);
1627 1628 1629 1630 1631 1632 1633
		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;
1634
			goto found;
1635 1636 1637 1638 1639
		}
	}
	return 0;

found:
1640
	BUG_ON(!IS_ALIGNED(virt_to_phys(m), huge_page_size(h)));
1641 1642 1643 1644 1645 1646
	/* 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;
}

1647
static void __init prep_compound_huge_page(struct page *page, int order)
1648 1649 1650 1651 1652 1653 1654
{
	if (unlikely(order > (MAX_ORDER - 1)))
		prep_compound_gigantic_page(page, order);
	else
		prep_compound_page(page, order);
}

1655 1656 1657 1658 1659 1660 1661
/* 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;
1662 1663 1664 1665
		struct page *page;

#ifdef CONFIG_HIGHMEM
		page = pfn_to_page(m->phys >> PAGE_SHIFT);
1666 1667
		memblock_free_late(__pa(m),
				   sizeof(struct huge_bootmem_page));
1668 1669 1670
#else
		page = virt_to_page(m);
#endif
1671
		WARN_ON(page_count(page) != 1);
1672
		prep_compound_huge_page(page, h->order);
1673
		WARN_ON(PageReserved(page));
1674
		prep_new_huge_page(h, page, page_to_nid(page));
1675 1676 1677 1678 1679 1680
		/*
		 * 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.
		 */
1681
		if (hstate_is_gigantic(h))
1682
			adjust_managed_page_count(page, 1 << h->order);
1683 1684 1685
	}
}

1686
static void __init hugetlb_hstate_alloc_pages(struct hstate *h)
L
Linus Torvalds 已提交
1687 1688
{
	unsigned long i;
1689

1690
	for (i = 0; i < h->max_huge_pages; ++i) {
1691
		if (hstate_is_gigantic(h)) {
1692 1693
			if (!alloc_bootmem_huge_page(h))
				break;
1694
		} else if (!alloc_fresh_huge_page(h,
1695
					 &node_states[N_MEMORY]))
L
Linus Torvalds 已提交
1696 1697
			break;
	}
1698
	h->max_huge_pages = i;
1699 1700 1701 1702 1703 1704 1705
}

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

	for_each_hstate(h) {
1706 1707 1708
		if (minimum_order > huge_page_order(h))
			minimum_order = huge_page_order(h);

1709
		/* oversize hugepages were init'ed in early boot */
1710
		if (!hstate_is_gigantic(h))
1711
			hugetlb_hstate_alloc_pages(h);
1712
	}
1713
	VM_BUG_ON(minimum_order == UINT_MAX);
1714 1715
}

A
Andi Kleen 已提交
1716 1717 1718 1719 1720 1721 1722 1723 1724 1725 1726
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;
}

1727 1728 1729 1730 1731
static void __init report_hugepages(void)
{
	struct hstate *h;

	for_each_hstate(h) {
A
Andi Kleen 已提交
1732
		char buf[32];
1733
		pr_info("HugeTLB registered %s page size, pre-allocated %ld pages\n",
A
Andi Kleen 已提交
1734 1735
			memfmt(buf, huge_page_size(h)),
			h->free_huge_pages);
1736 1737 1738
	}
}

L
Linus Torvalds 已提交
1739
#ifdef CONFIG_HIGHMEM
1740 1741
static void try_to_free_low(struct hstate *h, unsigned long count,
						nodemask_t *nodes_allowed)
L
Linus Torvalds 已提交
1742
{
1743 1744
	int i;

1745
	if (hstate_is_gigantic(h))
1746 1747
		return;

1748
	for_each_node_mask(i, *nodes_allowed) {
L
Linus Torvalds 已提交
1749
		struct page *page, *next;
1750 1751 1752
		struct list_head *freel = &h->hugepage_freelists[i];
		list_for_each_entry_safe(page, next, freel, lru) {
			if (count >= h->nr_huge_pages)
1753
				return;
L
Linus Torvalds 已提交
1754 1755 1756
			if (PageHighMem(page))
				continue;
			list_del(&page->lru);
1757
			update_and_free_page(h, page);
1758 1759
			h->free_huge_pages--;
			h->free_huge_pages_node[page_to_nid(page)]--;
L
Linus Torvalds 已提交
1760 1761 1762 1763
		}
	}
}
#else
1764 1765
static inline void try_to_free_low(struct hstate *h, unsigned long count,
						nodemask_t *nodes_allowed)
L
Linus Torvalds 已提交
1766 1767 1768 1769
{
}
#endif

1770 1771 1772 1773 1774
/*
 * 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.
 */
1775 1776
static int adjust_pool_surplus(struct hstate *h, nodemask_t *nodes_allowed,
				int delta)
1777
{
1778
	int nr_nodes, node;
1779 1780 1781

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

1782 1783 1784 1785
	if (delta < 0) {
		for_each_node_mask_to_alloc(h, nr_nodes, node, nodes_allowed) {
			if (h->surplus_huge_pages_node[node])
				goto found;
1786
		}
1787 1788 1789 1790 1791
	} 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;
1792
		}
1793 1794
	}
	return 0;
1795

1796 1797 1798 1799
found:
	h->surplus_huge_pages += delta;
	h->surplus_huge_pages_node[node] += delta;
	return 1;
1800 1801
}

1802
#define persistent_huge_pages(h) (h->nr_huge_pages - h->surplus_huge_pages)
1803 1804
static unsigned long set_max_huge_pages(struct hstate *h, unsigned long count,
						nodemask_t *nodes_allowed)
L
Linus Torvalds 已提交
1805
{
1806
	unsigned long min_count, ret;
L
Linus Torvalds 已提交
1807

1808
	if (hstate_is_gigantic(h) && !gigantic_page_supported())
1809 1810
		return h->max_huge_pages;

1811 1812 1813 1814
	/*
	 * Increase the pool size
	 * First take pages out of surplus state.  Then make up the
	 * remaining difference by allocating fresh huge pages.
1815 1816 1817 1818 1819 1820
	 *
	 * 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.
1821
	 */
L
Linus Torvalds 已提交
1822
	spin_lock(&hugetlb_lock);
1823
	while (h->surplus_huge_pages && count > persistent_huge_pages(h)) {
1824
		if (!adjust_pool_surplus(h, nodes_allowed, -1))
1825 1826 1827
			break;
	}

1828
	while (count > persistent_huge_pages(h)) {
1829 1830 1831 1832 1833 1834
		/*
		 * 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);
1835 1836 1837 1838
		if (hstate_is_gigantic(h))
			ret = alloc_fresh_gigantic_page(h, nodes_allowed);
		else
			ret = alloc_fresh_huge_page(h, nodes_allowed);
1839 1840 1841 1842
		spin_lock(&hugetlb_lock);
		if (!ret)
			goto out;

1843 1844 1845
		/* Bail for signals. Probably ctrl-c from user */
		if (signal_pending(current))
			goto out;
1846 1847 1848 1849 1850 1851 1852 1853
	}

	/*
	 * 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.
1854 1855 1856 1857 1858 1859 1860 1861
	 *
	 * 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.
1862
	 */
1863
	min_count = h->resv_huge_pages + h->nr_huge_pages - h->free_huge_pages;
1864
	min_count = max(count, min_count);
1865
	try_to_free_low(h, min_count, nodes_allowed);
1866
	while (min_count < persistent_huge_pages(h)) {
1867
		if (!free_pool_huge_page(h, nodes_allowed, 0))
L
Linus Torvalds 已提交
1868
			break;
1869
		cond_resched_lock(&hugetlb_lock);
L
Linus Torvalds 已提交
1870
	}
1871
	while (count < persistent_huge_pages(h)) {
1872
		if (!adjust_pool_surplus(h, nodes_allowed, 1))
1873 1874 1875
			break;
	}
out:
1876
	ret = persistent_huge_pages(h);
L
Linus Torvalds 已提交
1877
	spin_unlock(&hugetlb_lock);
1878
	return ret;
L
Linus Torvalds 已提交
1879 1880
}

1881 1882 1883 1884 1885 1886 1887 1888 1889 1890
#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];

1891 1892 1893
static struct hstate *kobj_to_node_hstate(struct kobject *kobj, int *nidp);

static struct hstate *kobj_to_hstate(struct kobject *kobj, int *nidp)
1894 1895
{
	int i;
1896

1897
	for (i = 0; i < HUGE_MAX_HSTATE; i++)
1898 1899 1900
		if (hstate_kobjs[i] == kobj) {
			if (nidp)
				*nidp = NUMA_NO_NODE;
1901
			return &hstates[i];
1902 1903 1904
		}

	return kobj_to_node_hstate(kobj, nidp);
1905 1906
}

1907
static ssize_t nr_hugepages_show_common(struct kobject *kobj,
1908 1909
					struct kobj_attribute *attr, char *buf)
{
1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920
	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);
1921
}
1922

1923 1924 1925
static ssize_t __nr_hugepages_store_common(bool obey_mempolicy,
					   struct hstate *h, int nid,
					   unsigned long count, size_t len)
1926 1927
{
	int err;
1928
	NODEMASK_ALLOC(nodemask_t, nodes_allowed, GFP_KERNEL | __GFP_NORETRY);
1929

1930
	if (hstate_is_gigantic(h) && !gigantic_page_supported()) {
1931 1932 1933 1934
		err = -EINVAL;
		goto out;
	}

1935 1936 1937 1938 1939 1940 1941
	if (nid == NUMA_NO_NODE) {
		/*
		 * global hstate attribute
		 */
		if (!(obey_mempolicy &&
				init_nodemask_of_mempolicy(nodes_allowed))) {
			NODEMASK_FREE(nodes_allowed);
1942
			nodes_allowed = &node_states[N_MEMORY];
1943 1944 1945 1946 1947 1948 1949 1950 1951
		}
	} 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
1952
		nodes_allowed = &node_states[N_MEMORY];
1953

1954
	h->max_huge_pages = set_max_huge_pages(h, count, nodes_allowed);
1955

1956
	if (nodes_allowed != &node_states[N_MEMORY])
1957 1958 1959
		NODEMASK_FREE(nodes_allowed);

	return len;
1960 1961 1962
out:
	NODEMASK_FREE(nodes_allowed);
	return err;
1963 1964
}

1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981
static ssize_t nr_hugepages_store_common(bool obey_mempolicy,
					 struct kobject *kobj, const char *buf,
					 size_t len)
{
	struct hstate *h;
	unsigned long count;
	int nid;
	int err;

	err = kstrtoul(buf, 10, &count);
	if (err)
		return err;

	h = kobj_to_hstate(kobj, &nid);
	return __nr_hugepages_store_common(obey_mempolicy, h, nid, count, len);
}

1982 1983 1984 1985 1986 1987 1988 1989 1990
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)
{
1991
	return nr_hugepages_store_common(false, kobj, buf, len);
1992 1993 1994
}
HSTATE_ATTR(nr_hugepages);

1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009
#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)
{
2010
	return nr_hugepages_store_common(true, kobj, buf, len);
2011 2012 2013 2014 2015
}
HSTATE_ATTR(nr_hugepages_mempolicy);
#endif


2016 2017 2018
static ssize_t nr_overcommit_hugepages_show(struct kobject *kobj,
					struct kobj_attribute *attr, char *buf)
{
2019
	struct hstate *h = kobj_to_hstate(kobj, NULL);
2020 2021
	return sprintf(buf, "%lu\n", h->nr_overcommit_huge_pages);
}
2022

2023 2024 2025 2026 2027
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;
2028
	struct hstate *h = kobj_to_hstate(kobj, NULL);
2029

2030
	if (hstate_is_gigantic(h))
2031 2032
		return -EINVAL;

2033
	err = kstrtoul(buf, 10, &input);
2034
	if (err)
2035
		return err;
2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046 2047

	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)
{
2048 2049 2050 2051 2052 2053 2054 2055 2056 2057 2058
	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);
2059 2060 2061 2062 2063 2064
}
HSTATE_ATTR_RO(free_hugepages);

static ssize_t resv_hugepages_show(struct kobject *kobj,
					struct kobj_attribute *attr, char *buf)
{
2065
	struct hstate *h = kobj_to_hstate(kobj, NULL);
2066 2067 2068 2069 2070 2071 2072
	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)
{
2073 2074 2075 2076 2077 2078 2079 2080 2081 2082 2083
	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);
2084 2085 2086 2087 2088 2089 2090 2091 2092
}
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,
2093 2094 2095
#ifdef CONFIG_NUMA
	&nr_hugepages_mempolicy_attr.attr,
#endif
2096 2097 2098 2099 2100 2101 2102
	NULL,
};

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

J
Jeff Mahoney 已提交
2103 2104 2105
static int hugetlb_sysfs_add_hstate(struct hstate *h, struct kobject *parent,
				    struct kobject **hstate_kobjs,
				    struct attribute_group *hstate_attr_group)
2106 2107
{
	int retval;
2108
	int hi = hstate_index(h);
2109

2110 2111
	hstate_kobjs[hi] = kobject_create_and_add(h->name, parent);
	if (!hstate_kobjs[hi])
2112 2113
		return -ENOMEM;

2114
	retval = sysfs_create_group(hstate_kobjs[hi], hstate_attr_group);
2115
	if (retval)
2116
		kobject_put(hstate_kobjs[hi]);
2117 2118 2119 2120 2121 2122 2123 2124 2125 2126 2127 2128 2129 2130

	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) {
2131 2132
		err = hugetlb_sysfs_add_hstate(h, hugepages_kobj,
					 hstate_kobjs, &hstate_attr_group);
2133
		if (err)
2134
			pr_err("Hugetlb: Unable to add hstate %s", h->name);
2135 2136 2137
	}
}

2138 2139 2140 2141
#ifdef CONFIG_NUMA

/*
 * node_hstate/s - associate per node hstate attributes, via their kobjects,
2142 2143 2144
 * 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
2145 2146 2147 2148 2149 2150 2151 2152 2153
 * 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];

/*
2154
 * A subset of global hstate attributes for node devices
2155 2156 2157 2158 2159 2160 2161 2162 2163 2164 2165 2166 2167
 */
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,
};

/*
2168
 * kobj_to_node_hstate - lookup global hstate for node device hstate attr kobj.
2169 2170 2171 2172 2173 2174 2175 2176 2177 2178 2179 2180 2181 2182 2183 2184 2185 2186 2187 2188 2189 2190
 * 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;
}

/*
2191
 * Unregister hstate attributes from a single node device.
2192 2193
 * No-op if no hstate attributes attached.
 */
2194
static void hugetlb_unregister_node(struct node *node)
2195 2196
{
	struct hstate *h;
2197
	struct node_hstate *nhs = &node_hstates[node->dev.id];
2198 2199

	if (!nhs->hugepages_kobj)
2200
		return;		/* no hstate attributes */
2201

2202 2203 2204 2205 2206
	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;
2207
		}
2208
	}
2209 2210 2211 2212 2213 2214

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

/*
2215
 * hugetlb module exit:  unregister hstate attributes from node devices
2216 2217 2218 2219 2220 2221 2222
 * that have them.
 */
static void hugetlb_unregister_all_nodes(void)
{
	int nid;

	/*
2223
	 * disable node device registrations.
2224 2225 2226 2227 2228 2229 2230
	 */
	register_hugetlbfs_with_node(NULL, NULL);

	/*
	 * remove hstate attributes from any nodes that have them.
	 */
	for (nid = 0; nid < nr_node_ids; nid++)
2231
		hugetlb_unregister_node(node_devices[nid]);
2232 2233 2234
}

/*
2235
 * Register hstate attributes for a single node device.
2236 2237
 * No-op if attributes already registered.
 */
2238
static void hugetlb_register_node(struct node *node)
2239 2240
{
	struct hstate *h;
2241
	struct node_hstate *nhs = &node_hstates[node->dev.id];
2242 2243 2244 2245 2246 2247
	int err;

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

	nhs->hugepages_kobj = kobject_create_and_add("hugepages",
2248
							&node->dev.kobj);
2249 2250 2251 2252 2253 2254 2255 2256
	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) {
2257 2258
			pr_err("Hugetlb: Unable to add hstate %s for node %d\n",
				h->name, node->dev.id);
2259 2260 2261 2262 2263 2264 2265
			hugetlb_unregister_node(node);
			break;
		}
	}
}

/*
2266
 * hugetlb init time:  register hstate attributes for all registered node
2267 2268
 * devices of nodes that have memory.  All on-line nodes should have
 * registered their associated device by this time.
2269
 */
2270
static void __init hugetlb_register_all_nodes(void)
2271 2272 2273
{
	int nid;

2274
	for_each_node_state(nid, N_MEMORY) {
2275
		struct node *node = node_devices[nid];
2276
		if (node->dev.id == nid)
2277 2278 2279 2280
			hugetlb_register_node(node);
	}

	/*
2281
	 * Let the node device driver know we're here so it can
2282 2283 2284 2285 2286 2287 2288 2289 2290 2291 2292 2293 2294 2295 2296 2297 2298 2299 2300 2301 2302
	 * [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

2303 2304 2305 2306
static void __exit hugetlb_exit(void)
{
	struct hstate *h;

2307 2308
	hugetlb_unregister_all_nodes();

2309
	for_each_hstate(h) {
2310
		kobject_put(hstate_kobjs[hstate_index(h)]);
2311 2312 2313
	}

	kobject_put(hugepages_kobj);
2314
	kfree(htlb_fault_mutex_table);
2315 2316 2317 2318 2319
}
module_exit(hugetlb_exit);

static int __init hugetlb_init(void)
{
2320 2321
	int i;

2322
	if (!hugepages_supported())
2323
		return 0;
2324

2325 2326 2327 2328
	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);
2329
	}
2330
	default_hstate_idx = hstate_index(size_to_hstate(default_hstate_size));
2331 2332
	if (default_hstate_max_huge_pages)
		default_hstate.max_huge_pages = default_hstate_max_huge_pages;
2333 2334

	hugetlb_init_hstates();
2335
	gather_bootmem_prealloc();
2336 2337 2338
	report_hugepages();

	hugetlb_sysfs_init();
2339
	hugetlb_register_all_nodes();
2340
	hugetlb_cgroup_file_init();
2341

2342 2343 2344 2345 2346 2347 2348 2349 2350 2351 2352
#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]);
2353 2354 2355 2356 2357 2358 2359 2360
	return 0;
}
module_init(hugetlb_init);

/* Should be called on processing a hugepagesz=... option */
void __init hugetlb_add_hstate(unsigned order)
{
	struct hstate *h;
2361 2362
	unsigned long i;

2363
	if (size_to_hstate(PAGE_SIZE << order)) {
2364
		pr_warning("hugepagesz= specified twice, ignoring\n");
2365 2366
		return;
	}
2367
	BUG_ON(hugetlb_max_hstate >= HUGE_MAX_HSTATE);
2368
	BUG_ON(order == 0);
2369
	h = &hstates[hugetlb_max_hstate++];
2370 2371
	h->order = order;
	h->mask = ~((1ULL << (order + PAGE_SHIFT)) - 1);
2372 2373 2374 2375
	h->nr_huge_pages = 0;
	h->free_huge_pages = 0;
	for (i = 0; i < MAX_NUMNODES; ++i)
		INIT_LIST_HEAD(&h->hugepage_freelists[i]);
2376
	INIT_LIST_HEAD(&h->hugepage_activelist);
2377 2378
	h->next_nid_to_alloc = first_node(node_states[N_MEMORY]);
	h->next_nid_to_free = first_node(node_states[N_MEMORY]);
2379 2380
	snprintf(h->name, HSTATE_NAME_LEN, "hugepages-%lukB",
					huge_page_size(h)/1024);
2381

2382 2383 2384
	parsed_hstate = h;
}

2385
static int __init hugetlb_nrpages_setup(char *s)
2386 2387
{
	unsigned long *mhp;
2388
	static unsigned long *last_mhp;
2389 2390

	/*
2391
	 * !hugetlb_max_hstate means we haven't parsed a hugepagesz= parameter yet,
2392 2393
	 * so this hugepages= parameter goes to the "default hstate".
	 */
2394
	if (!hugetlb_max_hstate)
2395 2396 2397 2398
		mhp = &default_hstate_max_huge_pages;
	else
		mhp = &parsed_hstate->max_huge_pages;

2399
	if (mhp == last_mhp) {
2400 2401
		pr_warning("hugepages= specified twice without "
			   "interleaving hugepagesz=, ignoring\n");
2402 2403 2404
		return 1;
	}

2405 2406 2407
	if (sscanf(s, "%lu", mhp) <= 0)
		*mhp = 0;

2408 2409 2410 2411 2412
	/*
	 * 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.
	 */
2413
	if (hugetlb_max_hstate && parsed_hstate->order >= MAX_ORDER)
2414 2415 2416 2417
		hugetlb_hstate_alloc_pages(parsed_hstate);

	last_mhp = mhp;

2418 2419
	return 1;
}
2420 2421 2422 2423 2424 2425 2426 2427
__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);
2428

2429 2430 2431 2432 2433 2434 2435 2436 2437 2438 2439 2440
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
2441 2442 2443
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
2444
{
2445
	struct hstate *h = &default_hstate;
2446
	unsigned long tmp = h->max_huge_pages;
2447
	int ret;
2448

2449 2450 2451
	if (!hugepages_supported())
		return -ENOTSUPP;

2452 2453
	table->data = &tmp;
	table->maxlen = sizeof(unsigned long);
2454 2455 2456
	ret = proc_doulongvec_minmax(table, write, buffer, length, ppos);
	if (ret)
		goto out;
2457

2458 2459 2460
	if (write)
		ret = __nr_hugepages_store_common(obey_mempolicy, h,
						  NUMA_NO_NODE, tmp, *length);
2461 2462
out:
	return ret;
L
Linus Torvalds 已提交
2463
}
2464

2465 2466 2467 2468 2469 2470 2471 2472 2473 2474 2475 2476 2477 2478 2479 2480 2481
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 */

2482
int hugetlb_overcommit_handler(struct ctl_table *table, int write,
2483
			void __user *buffer,
2484 2485
			size_t *length, loff_t *ppos)
{
2486
	struct hstate *h = &default_hstate;
2487
	unsigned long tmp;
2488
	int ret;
2489

2490 2491 2492
	if (!hugepages_supported())
		return -ENOTSUPP;

2493
	tmp = h->nr_overcommit_huge_pages;
2494

2495
	if (write && hstate_is_gigantic(h))
2496 2497
		return -EINVAL;

2498 2499
	table->data = &tmp;
	table->maxlen = sizeof(unsigned long);
2500 2501 2502
	ret = proc_doulongvec_minmax(table, write, buffer, length, ppos);
	if (ret)
		goto out;
2503 2504 2505 2506 2507 2508

	if (write) {
		spin_lock(&hugetlb_lock);
		h->nr_overcommit_huge_pages = tmp;
		spin_unlock(&hugetlb_lock);
	}
2509 2510
out:
	return ret;
2511 2512
}

L
Linus Torvalds 已提交
2513 2514
#endif /* CONFIG_SYSCTL */

2515
void hugetlb_report_meminfo(struct seq_file *m)
L
Linus Torvalds 已提交
2516
{
2517
	struct hstate *h = &default_hstate;
2518 2519
	if (!hugepages_supported())
		return;
2520
	seq_printf(m,
2521 2522 2523 2524 2525
			"HugePages_Total:   %5lu\n"
			"HugePages_Free:    %5lu\n"
			"HugePages_Rsvd:    %5lu\n"
			"HugePages_Surp:    %5lu\n"
			"Hugepagesize:   %8lu kB\n",
2526 2527 2528 2529 2530
			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
2531 2532 2533 2534
}

int hugetlb_report_node_meminfo(int nid, char *buf)
{
2535
	struct hstate *h = &default_hstate;
2536 2537
	if (!hugepages_supported())
		return 0;
L
Linus Torvalds 已提交
2538 2539
	return sprintf(buf,
		"Node %d HugePages_Total: %5u\n"
2540 2541
		"Node %d HugePages_Free:  %5u\n"
		"Node %d HugePages_Surp:  %5u\n",
2542 2543 2544
		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
2545 2546
}

2547 2548 2549 2550 2551
void hugetlb_show_meminfo(void)
{
	struct hstate *h;
	int nid;

2552 2553 2554
	if (!hugepages_supported())
		return;

2555 2556 2557 2558 2559 2560 2561 2562 2563 2564
	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
2565 2566 2567
/* Return the number pages of memory we physically have, in PAGE_SIZE units. */
unsigned long hugetlb_total_pages(void)
{
2568 2569 2570 2571 2572 2573
	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
2574 2575
}

2576
static int hugetlb_acct_memory(struct hstate *h, long delta)
M
Mel Gorman 已提交
2577 2578 2579 2580 2581 2582 2583 2584 2585 2586 2587 2588 2589 2590 2591 2592 2593 2594 2595 2596 2597 2598
{
	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) {
2599
		if (gather_surplus_pages(h, delta) < 0)
M
Mel Gorman 已提交
2600 2601
			goto out;

2602 2603
		if (delta > cpuset_mems_nr(h->free_huge_pages_node)) {
			return_unused_surplus_pages(h, delta);
M
Mel Gorman 已提交
2604 2605 2606 2607 2608 2609
			goto out;
		}
	}

	ret = 0;
	if (delta < 0)
2610
		return_unused_surplus_pages(h, (unsigned long) -delta);
M
Mel Gorman 已提交
2611 2612 2613 2614 2615 2616

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

2617 2618
static void hugetlb_vm_op_open(struct vm_area_struct *vma)
{
2619
	struct resv_map *resv = vma_resv_map(vma);
2620 2621 2622 2623 2624

	/*
	 * 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
2625
	 * has a reference to the reservation map it cannot disappear until
2626 2627 2628
	 * after this open call completes.  It is therefore safe to take a
	 * new reference here without additional locking.
	 */
2629
	if (resv && is_vma_resv_set(vma, HPAGE_RESV_OWNER))
2630
		kref_get(&resv->refs);
2631 2632
}

2633 2634
static void hugetlb_vm_op_close(struct vm_area_struct *vma)
{
2635
	struct hstate *h = hstate_vma(vma);
2636
	struct resv_map *resv = vma_resv_map(vma);
2637
	struct hugepage_subpool *spool = subpool_vma(vma);
2638
	unsigned long reserve, start, end;
2639
	long gbl_reserve;
2640

2641 2642
	if (!resv || !is_vma_resv_set(vma, HPAGE_RESV_OWNER))
		return;
2643

2644 2645
	start = vma_hugecache_offset(h, vma, vma->vm_start);
	end = vma_hugecache_offset(h, vma, vma->vm_end);
2646

2647
	reserve = (end - start) - region_count(resv, start, end);
2648

2649 2650 2651
	kref_put(&resv->refs, resv_map_release);

	if (reserve) {
2652 2653 2654 2655 2656 2657
		/*
		 * Decrement reserve counts.  The global reserve count may be
		 * adjusted if the subpool has a minimum size.
		 */
		gbl_reserve = hugepage_subpool_put_pages(spool, reserve);
		hugetlb_acct_memory(h, -gbl_reserve);
2658
	}
2659 2660
}

L
Linus Torvalds 已提交
2661 2662 2663 2664 2665 2666
/*
 * 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
2667
static int hugetlb_vm_op_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
L
Linus Torvalds 已提交
2668 2669
{
	BUG();
N
Nick Piggin 已提交
2670
	return 0;
L
Linus Torvalds 已提交
2671 2672
}

2673
const struct vm_operations_struct hugetlb_vm_ops = {
N
Nick Piggin 已提交
2674
	.fault = hugetlb_vm_op_fault,
2675
	.open = hugetlb_vm_op_open,
2676
	.close = hugetlb_vm_op_close,
L
Linus Torvalds 已提交
2677 2678
};

2679 2680
static pte_t make_huge_pte(struct vm_area_struct *vma, struct page *page,
				int writable)
D
David Gibson 已提交
2681 2682 2683
{
	pte_t entry;

2684
	if (writable) {
2685 2686
		entry = huge_pte_mkwrite(huge_pte_mkdirty(mk_huge_pte(page,
					 vma->vm_page_prot)));
D
David Gibson 已提交
2687
	} else {
2688 2689
		entry = huge_pte_wrprotect(mk_huge_pte(page,
					   vma->vm_page_prot));
D
David Gibson 已提交
2690 2691 2692
	}
	entry = pte_mkyoung(entry);
	entry = pte_mkhuge(entry);
2693
	entry = arch_make_huge_pte(entry, vma, page, writable);
D
David Gibson 已提交
2694 2695 2696 2697

	return entry;
}

2698 2699 2700 2701 2702
static void set_huge_ptep_writable(struct vm_area_struct *vma,
				   unsigned long address, pte_t *ptep)
{
	pte_t entry;

2703
	entry = huge_pte_mkwrite(huge_pte_mkdirty(huge_ptep_get(ptep)));
2704
	if (huge_ptep_set_access_flags(vma, address, ptep, entry, 1))
2705
		update_mmu_cache(vma, address, ptep);
2706 2707
}

2708 2709 2710 2711 2712 2713 2714 2715 2716 2717 2718 2719 2720 2721 2722 2723 2724 2725 2726 2727 2728 2729 2730 2731 2732
static int is_hugetlb_entry_migration(pte_t pte)
{
	swp_entry_t swp;

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

static int is_hugetlb_entry_hwpoisoned(pte_t pte)
{
	swp_entry_t swp;

	if (huge_pte_none(pte) || pte_present(pte))
		return 0;
	swp = pte_to_swp_entry(pte);
	if (non_swap_entry(swp) && is_hwpoison_entry(swp))
		return 1;
	else
		return 0;
}
2733

D
David Gibson 已提交
2734 2735 2736 2737 2738
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;
2739
	unsigned long addr;
2740
	int cow;
2741 2742
	struct hstate *h = hstate_vma(vma);
	unsigned long sz = huge_page_size(h);
2743 2744 2745
	unsigned long mmun_start;	/* For mmu_notifiers */
	unsigned long mmun_end;		/* For mmu_notifiers */
	int ret = 0;
2746 2747

	cow = (vma->vm_flags & (VM_SHARED | VM_MAYWRITE)) == VM_MAYWRITE;
D
David Gibson 已提交
2748

2749 2750 2751 2752 2753
	mmun_start = vma->vm_start;
	mmun_end = vma->vm_end;
	if (cow)
		mmu_notifier_invalidate_range_start(src, mmun_start, mmun_end);

2754
	for (addr = vma->vm_start; addr < vma->vm_end; addr += sz) {
2755
		spinlock_t *src_ptl, *dst_ptl;
H
Hugh Dickins 已提交
2756 2757 2758
		src_pte = huge_pte_offset(src, addr);
		if (!src_pte)
			continue;
2759
		dst_pte = huge_pte_alloc(dst, addr, sz);
2760 2761 2762 2763
		if (!dst_pte) {
			ret = -ENOMEM;
			break;
		}
2764 2765 2766 2767 2768

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

2769 2770 2771
		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);
2772 2773 2774 2775 2776 2777 2778 2779 2780 2781 2782 2783 2784 2785 2786 2787 2788 2789
		entry = huge_ptep_get(src_pte);
		if (huge_pte_none(entry)) { /* skip none entry */
			;
		} else if (unlikely(is_hugetlb_entry_migration(entry) ||
				    is_hugetlb_entry_hwpoisoned(entry))) {
			swp_entry_t swp_entry = pte_to_swp_entry(entry);

			if (is_write_migration_entry(swp_entry) && cow) {
				/*
				 * COW mappings require pages in both
				 * parent and child to be set to read.
				 */
				make_migration_entry_read(&swp_entry);
				entry = swp_entry_to_pte(swp_entry);
				set_huge_pte_at(src, addr, src_pte, entry);
			}
			set_huge_pte_at(dst, addr, dst_pte, entry);
		} else {
2790
			if (cow) {
2791
				huge_ptep_set_wrprotect(src, addr, src_pte);
2792 2793 2794
				mmu_notifier_invalidate_range(src, mmun_start,
								   mmun_end);
			}
2795
			entry = huge_ptep_get(src_pte);
2796 2797
			ptepage = pte_page(entry);
			get_page(ptepage);
2798
			page_dup_rmap(ptepage);
2799 2800
			set_huge_pte_at(dst, addr, dst_pte, entry);
		}
2801 2802
		spin_unlock(src_ptl);
		spin_unlock(dst_ptl);
D
David Gibson 已提交
2803 2804
	}

2805 2806 2807 2808
	if (cow)
		mmu_notifier_invalidate_range_end(src, mmun_start, mmun_end);

	return ret;
D
David Gibson 已提交
2809 2810
}

2811 2812 2813
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
2814
{
2815
	int force_flush = 0;
D
David Gibson 已提交
2816 2817
	struct mm_struct *mm = vma->vm_mm;
	unsigned long address;
2818
	pte_t *ptep;
D
David Gibson 已提交
2819
	pte_t pte;
2820
	spinlock_t *ptl;
D
David Gibson 已提交
2821
	struct page *page;
2822 2823
	struct hstate *h = hstate_vma(vma);
	unsigned long sz = huge_page_size(h);
2824 2825
	const unsigned long mmun_start = start;	/* For mmu_notifiers */
	const unsigned long mmun_end   = end;	/* For mmu_notifiers */
2826

D
David Gibson 已提交
2827
	WARN_ON(!is_vm_hugetlb_page(vma));
2828 2829
	BUG_ON(start & ~huge_page_mask(h));
	BUG_ON(end & ~huge_page_mask(h));
D
David Gibson 已提交
2830

2831
	tlb_start_vma(tlb, vma);
2832
	mmu_notifier_invalidate_range_start(mm, mmun_start, mmun_end);
2833
	address = start;
2834
again:
2835
	for (; address < end; address += sz) {
2836
		ptep = huge_pte_offset(mm, address);
A
Adam Litke 已提交
2837
		if (!ptep)
2838 2839
			continue;

2840
		ptl = huge_pte_lock(h, mm, ptep);
2841
		if (huge_pmd_unshare(mm, &address, ptep))
2842
			goto unlock;
2843

2844 2845
		pte = huge_ptep_get(ptep);
		if (huge_pte_none(pte))
2846
			goto unlock;
2847 2848

		/*
2849 2850
		 * Migrating hugepage or HWPoisoned hugepage is already
		 * unmapped and its refcount is dropped, so just clear pte here.
2851
		 */
2852
		if (unlikely(!pte_present(pte))) {
2853
			huge_pte_clear(mm, address, ptep);
2854
			goto unlock;
2855
		}
2856 2857

		page = pte_page(pte);
2858 2859 2860 2861 2862 2863 2864
		/*
		 * 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)
2865
				goto unlock;
2866 2867 2868 2869 2870 2871 2872 2873 2874

			/*
			 * 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);
		}

2875
		pte = huge_ptep_get_and_clear(mm, address, ptep);
2876
		tlb_remove_tlb_entry(tlb, ptep, address);
2877
		if (huge_pte_dirty(pte))
2878
			set_page_dirty(page);
2879

2880 2881
		page_remove_rmap(page);
		force_flush = !__tlb_remove_page(tlb, page);
2882
		if (force_flush) {
2883
			address += sz;
2884
			spin_unlock(ptl);
2885
			break;
2886
		}
2887
		/* Bail out after unmapping reference page if supplied */
2888 2889
		if (ref_page) {
			spin_unlock(ptl);
2890
			break;
2891 2892 2893
		}
unlock:
		spin_unlock(ptl);
D
David Gibson 已提交
2894
	}
2895 2896 2897 2898 2899 2900 2901 2902 2903 2904
	/*
	 * 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;
2905
	}
2906
	mmu_notifier_invalidate_range_end(mm, mmun_start, mmun_end);
2907
	tlb_end_vma(tlb, vma);
L
Linus Torvalds 已提交
2908
}
D
David Gibson 已提交
2909

2910 2911 2912 2913 2914 2915 2916 2917 2918 2919 2920 2921
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
2922
	 * is to clear it before releasing the i_mmap_rwsem. This works
2923
	 * because in the context this is called, the VMA is about to be
2924
	 * destroyed and the i_mmap_rwsem is held.
2925 2926 2927 2928
	 */
	vma->vm_flags &= ~VM_MAYSHARE;
}

2929
void unmap_hugepage_range(struct vm_area_struct *vma, unsigned long start,
2930
			  unsigned long end, struct page *ref_page)
2931
{
2932 2933 2934 2935 2936
	struct mm_struct *mm;
	struct mmu_gather tlb;

	mm = vma->vm_mm;

2937
	tlb_gather_mmu(&tlb, mm, start, end);
2938 2939
	__unmap_hugepage_range(&tlb, vma, start, end, ref_page);
	tlb_finish_mmu(&tlb, start, end);
2940 2941
}

2942 2943 2944 2945 2946 2947
/*
 * 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.
 */
2948 2949
static void unmap_ref_private(struct mm_struct *mm, struct vm_area_struct *vma,
			      struct page *page, unsigned long address)
2950
{
2951
	struct hstate *h = hstate_vma(vma);
2952 2953 2954 2955 2956 2957 2958 2959
	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.
	 */
2960
	address = address & huge_page_mask(h);
2961 2962
	pgoff = ((address - vma->vm_start) >> PAGE_SHIFT) +
			vma->vm_pgoff;
A
Al Viro 已提交
2963
	mapping = file_inode(vma->vm_file)->i_mapping;
2964

2965 2966 2967 2968 2969
	/*
	 * 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
	 */
2970
	i_mmap_lock_write(mapping);
2971
	vma_interval_tree_foreach(iter_vma, &mapping->i_mmap, pgoff, pgoff) {
2972 2973 2974 2975 2976 2977 2978 2979 2980 2981 2982 2983
		/* 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))
2984 2985
			unmap_hugepage_range(iter_vma, address,
					     address + huge_page_size(h), page);
2986
	}
2987
	i_mmap_unlock_write(mapping);
2988 2989
}

2990 2991
/*
 * Hugetlb_cow() should be called with page lock of the original hugepage held.
2992 2993 2994
 * 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.
2995
 */
2996
static int hugetlb_cow(struct mm_struct *mm, struct vm_area_struct *vma,
2997
			unsigned long address, pte_t *ptep, pte_t pte,
2998
			struct page *pagecache_page, spinlock_t *ptl)
2999
{
3000
	struct hstate *h = hstate_vma(vma);
3001
	struct page *old_page, *new_page;
3002
	int ret = 0, outside_reserve = 0;
3003 3004
	unsigned long mmun_start;	/* For mmu_notifiers */
	unsigned long mmun_end;		/* For mmu_notifiers */
3005 3006 3007

	old_page = pte_page(pte);

3008
retry_avoidcopy:
3009 3010
	/* If no-one else is actually using this page, avoid the copy
	 * and just make the page writable */
3011 3012
	if (page_mapcount(old_page) == 1 && PageAnon(old_page)) {
		page_move_anon_rmap(old_page, vma, address);
3013
		set_huge_ptep_writable(vma, address, ptep);
N
Nick Piggin 已提交
3014
		return 0;
3015 3016
	}

3017 3018 3019 3020 3021 3022 3023 3024 3025
	/*
	 * 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.
	 */
3026
	if (is_vma_resv_set(vma, HPAGE_RESV_OWNER) &&
3027 3028 3029
			old_page != pagecache_page)
		outside_reserve = 1;

3030
	page_cache_get(old_page);
3031

3032 3033 3034 3035
	/*
	 * Drop page table lock as buddy allocator may be called. It will
	 * be acquired again before returning to the caller, as expected.
	 */
3036
	spin_unlock(ptl);
3037
	new_page = alloc_huge_page(vma, address, outside_reserve);
3038

3039
	if (IS_ERR(new_page)) {
3040 3041 3042 3043 3044 3045 3046 3047
		/*
		 * 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) {
3048
			page_cache_release(old_page);
3049
			BUG_ON(huge_pte_none(pte));
3050 3051 3052 3053 3054 3055 3056 3057 3058 3059 3060 3061
			unmap_ref_private(mm, vma, old_page, address);
			BUG_ON(huge_pte_none(pte));
			spin_lock(ptl);
			ptep = huge_pte_offset(mm, address & huge_page_mask(h));
			if (likely(ptep &&
				   pte_same(huge_ptep_get(ptep), pte)))
				goto retry_avoidcopy;
			/*
			 * race occurs while re-acquiring page table
			 * lock, and our job is done.
			 */
			return 0;
3062 3063
		}

3064 3065 3066
		ret = (PTR_ERR(new_page) == -ENOMEM) ?
			VM_FAULT_OOM : VM_FAULT_SIGBUS;
		goto out_release_old;
3067 3068
	}

3069 3070 3071 3072
	/*
	 * When the original hugepage is shared one, it does not have
	 * anon_vma prepared.
	 */
3073
	if (unlikely(anon_vma_prepare(vma))) {
3074 3075
		ret = VM_FAULT_OOM;
		goto out_release_all;
3076
	}
3077

A
Andrea Arcangeli 已提交
3078 3079
	copy_user_huge_page(new_page, old_page, address, vma,
			    pages_per_huge_page(h));
N
Nick Piggin 已提交
3080
	__SetPageUptodate(new_page);
3081
	set_page_huge_active(new_page);
3082

3083 3084 3085
	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);
3086

3087
	/*
3088
	 * Retake the page table lock to check for racing updates
3089 3090
	 * before the page tables are altered
	 */
3091
	spin_lock(ptl);
3092
	ptep = huge_pte_offset(mm, address & huge_page_mask(h));
3093
	if (likely(ptep && pte_same(huge_ptep_get(ptep), pte))) {
3094 3095
		ClearPagePrivate(new_page);

3096
		/* Break COW */
3097
		huge_ptep_clear_flush(vma, address, ptep);
3098
		mmu_notifier_invalidate_range(mm, mmun_start, mmun_end);
3099 3100
		set_huge_pte_at(mm, address, ptep,
				make_huge_pte(vma, new_page, 1));
3101
		page_remove_rmap(old_page);
3102
		hugepage_add_new_anon_rmap(new_page, vma, address);
3103 3104 3105
		/* Make the old page be freed below */
		new_page = old_page;
	}
3106
	spin_unlock(ptl);
3107
	mmu_notifier_invalidate_range_end(mm, mmun_start, mmun_end);
3108
out_release_all:
3109
	page_cache_release(new_page);
3110
out_release_old:
3111
	page_cache_release(old_page);
3112

3113 3114
	spin_lock(ptl); /* Caller expects lock to be held */
	return ret;
3115 3116
}

3117
/* Return the pagecache page at a given address within a VMA */
3118 3119
static struct page *hugetlbfs_pagecache_page(struct hstate *h,
			struct vm_area_struct *vma, unsigned long address)
3120 3121
{
	struct address_space *mapping;
3122
	pgoff_t idx;
3123 3124

	mapping = vma->vm_file->f_mapping;
3125
	idx = vma_hugecache_offset(h, vma, address);
3126 3127 3128 3129

	return find_lock_page(mapping, idx);
}

H
Hugh Dickins 已提交
3130 3131 3132 3133 3134
/*
 * 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
3135 3136 3137 3138 3139 3140 3141 3142 3143 3144 3145 3146 3147 3148 3149
			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;
}

3150
static int hugetlb_no_page(struct mm_struct *mm, struct vm_area_struct *vma,
3151 3152
			   struct address_space *mapping, pgoff_t idx,
			   unsigned long address, pte_t *ptep, unsigned int flags)
3153
{
3154
	struct hstate *h = hstate_vma(vma);
3155
	int ret = VM_FAULT_SIGBUS;
3156
	int anon_rmap = 0;
A
Adam Litke 已提交
3157 3158
	unsigned long size;
	struct page *page;
3159
	pte_t new_pte;
3160
	spinlock_t *ptl;
A
Adam Litke 已提交
3161

3162 3163 3164
	/*
	 * 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
3165
	 * COW. Warn that such a situation has occurred as it may not be obvious
3166 3167
	 */
	if (is_vma_resv_set(vma, HPAGE_RESV_UNMAPPED)) {
3168 3169
		pr_warning("PID %d killed due to inadequate hugepage pool\n",
			   current->pid);
3170 3171 3172
		return ret;
	}

A
Adam Litke 已提交
3173 3174 3175 3176
	/*
	 * Use page lock to guard against racing truncation
	 * before we get page_table_lock.
	 */
3177 3178 3179
retry:
	page = find_lock_page(mapping, idx);
	if (!page) {
3180
		size = i_size_read(mapping->host) >> huge_page_shift(h);
3181 3182
		if (idx >= size)
			goto out;
3183
		page = alloc_huge_page(vma, address, 0);
3184
		if (IS_ERR(page)) {
3185 3186 3187 3188 3189
			ret = PTR_ERR(page);
			if (ret == -ENOMEM)
				ret = VM_FAULT_OOM;
			else
				ret = VM_FAULT_SIGBUS;
3190 3191
			goto out;
		}
A
Andrea Arcangeli 已提交
3192
		clear_huge_page(page, address, pages_per_huge_page(h));
N
Nick Piggin 已提交
3193
		__SetPageUptodate(page);
3194
		set_page_huge_active(page);
3195

3196
		if (vma->vm_flags & VM_MAYSHARE) {
3197
			int err;
K
Ken Chen 已提交
3198
			struct inode *inode = mapping->host;
3199 3200 3201 3202 3203 3204 3205 3206

			err = add_to_page_cache(page, mapping, idx, GFP_KERNEL);
			if (err) {
				put_page(page);
				if (err == -EEXIST)
					goto retry;
				goto out;
			}
3207
			ClearPagePrivate(page);
K
Ken Chen 已提交
3208 3209

			spin_lock(&inode->i_lock);
3210
			inode->i_blocks += blocks_per_huge_page(h);
K
Ken Chen 已提交
3211
			spin_unlock(&inode->i_lock);
3212
		} else {
3213
			lock_page(page);
3214 3215 3216 3217
			if (unlikely(anon_vma_prepare(vma))) {
				ret = VM_FAULT_OOM;
				goto backout_unlocked;
			}
3218
			anon_rmap = 1;
3219
		}
3220
	} else {
3221 3222 3223 3224 3225 3226
		/*
		 * 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))) {
3227
			ret = VM_FAULT_HWPOISON |
3228
				VM_FAULT_SET_HINDEX(hstate_index(h));
3229 3230
			goto backout_unlocked;
		}
3231
	}
3232

3233 3234 3235 3236 3237 3238
	/*
	 * 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.
	 */
3239
	if ((flags & FAULT_FLAG_WRITE) && !(vma->vm_flags & VM_SHARED))
3240 3241 3242 3243
		if (vma_needs_reservation(h, vma, address) < 0) {
			ret = VM_FAULT_OOM;
			goto backout_unlocked;
		}
3244

3245 3246
	ptl = huge_pte_lockptr(h, mm, ptep);
	spin_lock(ptl);
3247
	size = i_size_read(mapping->host) >> huge_page_shift(h);
A
Adam Litke 已提交
3248 3249 3250
	if (idx >= size)
		goto backout;

N
Nick Piggin 已提交
3251
	ret = 0;
3252
	if (!huge_pte_none(huge_ptep_get(ptep)))
A
Adam Litke 已提交
3253 3254
		goto backout;

3255 3256
	if (anon_rmap) {
		ClearPagePrivate(page);
3257
		hugepage_add_new_anon_rmap(page, vma, address);
3258
	} else
3259
		page_dup_rmap(page);
3260 3261 3262 3263
	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);

3264
	if ((flags & FAULT_FLAG_WRITE) && !(vma->vm_flags & VM_SHARED)) {
3265
		/* Optimization, do the COW without a second fault */
3266
		ret = hugetlb_cow(mm, vma, address, ptep, new_pte, page, ptl);
3267 3268
	}

3269
	spin_unlock(ptl);
A
Adam Litke 已提交
3270 3271
	unlock_page(page);
out:
3272
	return ret;
A
Adam Litke 已提交
3273 3274

backout:
3275
	spin_unlock(ptl);
3276
backout_unlocked:
A
Adam Litke 已提交
3277 3278 3279
	unlock_page(page);
	put_page(page);
	goto out;
3280 3281
}

3282 3283 3284 3285 3286 3287 3288 3289 3290 3291 3292 3293 3294 3295 3296 3297 3298 3299 3300 3301 3302 3303 3304 3305 3306 3307 3308 3309 3310 3311 3312 3313 3314 3315 3316
#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

3317
int hugetlb_fault(struct mm_struct *mm, struct vm_area_struct *vma,
3318
			unsigned long address, unsigned int flags)
3319
{
3320
	pte_t *ptep, entry;
3321
	spinlock_t *ptl;
3322
	int ret;
3323 3324
	u32 hash;
	pgoff_t idx;
3325
	struct page *page = NULL;
3326
	struct page *pagecache_page = NULL;
3327
	struct hstate *h = hstate_vma(vma);
3328
	struct address_space *mapping;
3329
	int need_wait_lock = 0;
3330

3331 3332
	address &= huge_page_mask(h);

3333 3334 3335
	ptep = huge_pte_offset(mm, address);
	if (ptep) {
		entry = huge_ptep_get(ptep);
N
Naoya Horiguchi 已提交
3336
		if (unlikely(is_hugetlb_entry_migration(entry))) {
3337
			migration_entry_wait_huge(vma, mm, ptep);
N
Naoya Horiguchi 已提交
3338 3339
			return 0;
		} else if (unlikely(is_hugetlb_entry_hwpoisoned(entry)))
3340
			return VM_FAULT_HWPOISON_LARGE |
3341
				VM_FAULT_SET_HINDEX(hstate_index(h));
3342 3343
	}

3344
	ptep = huge_pte_alloc(mm, address, huge_page_size(h));
3345 3346 3347
	if (!ptep)
		return VM_FAULT_OOM;

3348 3349 3350
	mapping = vma->vm_file->f_mapping;
	idx = vma_hugecache_offset(h, vma, address);

3351 3352 3353 3354 3355
	/*
	 * 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.
	 */
3356 3357 3358
	hash = fault_mutex_hash(h, mm, vma, mapping, idx, address);
	mutex_lock(&htlb_fault_mutex_table[hash]);

3359 3360
	entry = huge_ptep_get(ptep);
	if (huge_pte_none(entry)) {
3361
		ret = hugetlb_no_page(mm, vma, mapping, idx, address, ptep, flags);
3362
		goto out_mutex;
3363
	}
3364

N
Nick Piggin 已提交
3365
	ret = 0;
3366

3367 3368 3369 3370 3371 3372 3373 3374 3375 3376
	/*
	 * entry could be a migration/hwpoison entry at this point, so this
	 * check prevents the kernel from going below assuming that we have
	 * a active hugepage in pagecache. This goto expects the 2nd page fault,
	 * and is_hugetlb_entry_(migration|hwpoisoned) check will properly
	 * handle it.
	 */
	if (!pte_present(entry))
		goto out_mutex;

3377 3378 3379 3380 3381 3382 3383 3384
	/*
	 * 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.
	 */
3385
	if ((flags & FAULT_FLAG_WRITE) && !huge_pte_write(entry)) {
3386 3387
		if (vma_needs_reservation(h, vma, address) < 0) {
			ret = VM_FAULT_OOM;
3388
			goto out_mutex;
3389
		}
3390

3391
		if (!(vma->vm_flags & VM_MAYSHARE))
3392 3393 3394 3395
			pagecache_page = hugetlbfs_pagecache_page(h,
								vma, address);
	}

3396 3397 3398 3399 3400 3401
	ptl = huge_pte_lock(h, mm, ptep);

	/* Check for a racing update before calling hugetlb_cow */
	if (unlikely(!pte_same(entry, huge_ptep_get(ptep))))
		goto out_ptl;

3402 3403 3404 3405 3406 3407 3408
	/*
	 * 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.
	 */
	page = pte_page(entry);
	if (page != pagecache_page)
3409 3410 3411 3412
		if (!trylock_page(page)) {
			need_wait_lock = 1;
			goto out_ptl;
		}
3413

3414
	get_page(page);
3415

3416
	if (flags & FAULT_FLAG_WRITE) {
3417
		if (!huge_pte_write(entry)) {
3418
			ret = hugetlb_cow(mm, vma, address, ptep, entry,
3419
					pagecache_page, ptl);
3420
			goto out_put_page;
3421
		}
3422
		entry = huge_pte_mkdirty(entry);
3423 3424
	}
	entry = pte_mkyoung(entry);
3425 3426
	if (huge_ptep_set_access_flags(vma, address, ptep, entry,
						flags & FAULT_FLAG_WRITE))
3427
		update_mmu_cache(vma, address, ptep);
3428 3429 3430 3431
out_put_page:
	if (page != pagecache_page)
		unlock_page(page);
	put_page(page);
3432 3433
out_ptl:
	spin_unlock(ptl);
3434 3435 3436 3437 3438

	if (pagecache_page) {
		unlock_page(pagecache_page);
		put_page(pagecache_page);
	}
3439
out_mutex:
3440
	mutex_unlock(&htlb_fault_mutex_table[hash]);
3441 3442 3443 3444 3445 3446 3447 3448 3449
	/*
	 * Generally it's safe to hold refcount during waiting page lock. But
	 * here we just wait to defer the next page fault to avoid busy loop and
	 * the page is not used after unlocked before returning from the current
	 * page fault. So we are safe from accessing freed page, even if we wait
	 * here without taking refcount.
	 */
	if (need_wait_lock)
		wait_on_page_locked(page);
3450
	return ret;
3451 3452
}

3453 3454 3455 3456
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
3457
{
3458 3459
	unsigned long pfn_offset;
	unsigned long vaddr = *position;
3460
	unsigned long remainder = *nr_pages;
3461
	struct hstate *h = hstate_vma(vma);
D
David Gibson 已提交
3462 3463

	while (vaddr < vma->vm_end && remainder) {
A
Adam Litke 已提交
3464
		pte_t *pte;
3465
		spinlock_t *ptl = NULL;
H
Hugh Dickins 已提交
3466
		int absent;
A
Adam Litke 已提交
3467
		struct page *page;
D
David Gibson 已提交
3468

3469 3470 3471 3472 3473 3474 3475 3476 3477
		/*
		 * If we have a pending SIGKILL, don't keep faulting pages and
		 * potentially allocating memory.
		 */
		if (unlikely(fatal_signal_pending(current))) {
			remainder = 0;
			break;
		}

A
Adam Litke 已提交
3478 3479
		/*
		 * Some archs (sparc64, sh*) have multiple pte_ts to
H
Hugh Dickins 已提交
3480
		 * each hugepage.  We have to make sure we get the
A
Adam Litke 已提交
3481
		 * first, for the page indexing below to work.
3482 3483
		 *
		 * Note that page table lock is not held when pte is null.
A
Adam Litke 已提交
3484
		 */
3485
		pte = huge_pte_offset(mm, vaddr & huge_page_mask(h));
3486 3487
		if (pte)
			ptl = huge_pte_lock(h, mm, pte);
H
Hugh Dickins 已提交
3488 3489 3490 3491
		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
3492 3493 3494 3495
		 * 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
3496
		 */
H
Hugh Dickins 已提交
3497 3498
		if (absent && (flags & FOLL_DUMP) &&
		    !hugetlbfs_pagecache_present(h, vma, vaddr)) {
3499 3500
			if (pte)
				spin_unlock(ptl);
H
Hugh Dickins 已提交
3501 3502 3503
			remainder = 0;
			break;
		}
D
David Gibson 已提交
3504

3505 3506 3507 3508 3509 3510 3511 3512 3513 3514 3515
		/*
		 * 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)) ||
3516 3517
		    ((flags & FOLL_WRITE) &&
		      !huge_pte_write(huge_ptep_get(pte)))) {
A
Adam Litke 已提交
3518
			int ret;
D
David Gibson 已提交
3519

3520 3521
			if (pte)
				spin_unlock(ptl);
H
Hugh Dickins 已提交
3522 3523
			ret = hugetlb_fault(mm, vma, vaddr,
				(flags & FOLL_WRITE) ? FAULT_FLAG_WRITE : 0);
3524
			if (!(ret & VM_FAULT_ERROR))
A
Adam Litke 已提交
3525
				continue;
D
David Gibson 已提交
3526

A
Adam Litke 已提交
3527 3528 3529 3530
			remainder = 0;
			break;
		}

3531
		pfn_offset = (vaddr & ~huge_page_mask(h)) >> PAGE_SHIFT;
3532
		page = pte_page(huge_ptep_get(pte));
3533
same_page:
3534
		if (pages) {
H
Hugh Dickins 已提交
3535
			pages[i] = mem_map_offset(page, pfn_offset);
3536
			get_page_foll(pages[i]);
3537
		}
D
David Gibson 已提交
3538 3539 3540 3541 3542

		if (vmas)
			vmas[i] = vma;

		vaddr += PAGE_SIZE;
3543
		++pfn_offset;
D
David Gibson 已提交
3544 3545
		--remainder;
		++i;
3546
		if (vaddr < vma->vm_end && remainder &&
3547
				pfn_offset < pages_per_huge_page(h)) {
3548 3549 3550 3551 3552 3553
			/*
			 * We use pfn_offset to avoid touching the pageframes
			 * of this compound page.
			 */
			goto same_page;
		}
3554
		spin_unlock(ptl);
D
David Gibson 已提交
3555
	}
3556
	*nr_pages = remainder;
D
David Gibson 已提交
3557 3558
	*position = vaddr;

H
Hugh Dickins 已提交
3559
	return i ? i : -EFAULT;
D
David Gibson 已提交
3560
}
3561

3562
unsigned long hugetlb_change_protection(struct vm_area_struct *vma,
3563 3564 3565 3566 3567 3568
		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;
3569
	struct hstate *h = hstate_vma(vma);
3570
	unsigned long pages = 0;
3571 3572 3573 3574

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

3575
	mmu_notifier_invalidate_range_start(mm, start, end);
3576
	i_mmap_lock_write(vma->vm_file->f_mapping);
3577
	for (; address < end; address += huge_page_size(h)) {
3578
		spinlock_t *ptl;
3579 3580 3581
		ptep = huge_pte_offset(mm, address);
		if (!ptep)
			continue;
3582
		ptl = huge_pte_lock(h, mm, ptep);
3583 3584
		if (huge_pmd_unshare(mm, &address, ptep)) {
			pages++;
3585
			spin_unlock(ptl);
3586
			continue;
3587
		}
3588 3589 3590 3591 3592 3593 3594 3595 3596 3597 3598 3599 3600 3601 3602 3603 3604 3605 3606 3607
		pte = huge_ptep_get(ptep);
		if (unlikely(is_hugetlb_entry_hwpoisoned(pte))) {
			spin_unlock(ptl);
			continue;
		}
		if (unlikely(is_hugetlb_entry_migration(pte))) {
			swp_entry_t entry = pte_to_swp_entry(pte);

			if (is_write_migration_entry(entry)) {
				pte_t newpte;

				make_migration_entry_read(&entry);
				newpte = swp_entry_to_pte(entry);
				set_huge_pte_at(mm, address, ptep, newpte);
				pages++;
			}
			spin_unlock(ptl);
			continue;
		}
		if (!huge_pte_none(pte)) {
3608
			pte = huge_ptep_get_and_clear(mm, address, ptep);
3609
			pte = pte_mkhuge(huge_pte_modify(pte, newprot));
3610
			pte = arch_make_huge_pte(pte, vma, NULL, 0);
3611
			set_huge_pte_at(mm, address, ptep, pte);
3612
			pages++;
3613
		}
3614
		spin_unlock(ptl);
3615
	}
3616
	/*
3617
	 * Must flush TLB before releasing i_mmap_rwsem: x86's huge_pmd_unshare
3618
	 * may have cleared our pud entry and done put_page on the page table:
3619
	 * once we release i_mmap_rwsem, another task can do the final put_page
3620 3621
	 * and that page table be reused and filled with junk.
	 */
3622
	flush_tlb_range(vma, start, end);
3623
	mmu_notifier_invalidate_range(mm, start, end);
3624
	i_mmap_unlock_write(vma->vm_file->f_mapping);
3625
	mmu_notifier_invalidate_range_end(mm, start, end);
3626 3627

	return pages << h->order;
3628 3629
}

3630 3631
int hugetlb_reserve_pages(struct inode *inode,
					long from, long to,
3632
					struct vm_area_struct *vma,
3633
					vm_flags_t vm_flags)
3634
{
3635
	long ret, chg;
3636
	struct hstate *h = hstate_inode(inode);
3637
	struct hugepage_subpool *spool = subpool_inode(inode);
3638
	struct resv_map *resv_map;
3639
	long gbl_reserve;
3640

3641 3642 3643
	/*
	 * Only apply hugepage reservation if asked. At fault time, an
	 * attempt will be made for VM_NORESERVE to allocate a page
3644
	 * without using reserves
3645
	 */
3646
	if (vm_flags & VM_NORESERVE)
3647 3648
		return 0;

3649 3650 3651 3652 3653 3654
	/*
	 * 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
	 */
3655
	if (!vma || vma->vm_flags & VM_MAYSHARE) {
3656
		resv_map = inode_resv_map(inode);
3657

3658
		chg = region_chg(resv_map, from, to);
3659 3660 3661

	} else {
		resv_map = resv_map_alloc();
3662 3663 3664
		if (!resv_map)
			return -ENOMEM;

3665
		chg = to - from;
3666

3667 3668 3669 3670
		set_vma_resv_map(vma, resv_map);
		set_vma_resv_flags(vma, HPAGE_RESV_OWNER);
	}

3671 3672 3673 3674
	if (chg < 0) {
		ret = chg;
		goto out_err;
	}
3675

3676 3677 3678 3679 3680 3681 3682
	/*
	 * There must be enough pages in the subpool for the mapping. If
	 * the subpool has a minimum size, there may be some global
	 * reservations already in place (gbl_reserve).
	 */
	gbl_reserve = hugepage_subpool_get_pages(spool, chg);
	if (gbl_reserve < 0) {
3683 3684 3685
		ret = -ENOSPC;
		goto out_err;
	}
3686 3687

	/*
3688
	 * Check enough hugepages are available for the reservation.
3689
	 * Hand the pages back to the subpool if there are not
3690
	 */
3691
	ret = hugetlb_acct_memory(h, gbl_reserve);
K
Ken Chen 已提交
3692
	if (ret < 0) {
3693 3694
		/* put back original number of pages, chg */
		(void)hugepage_subpool_put_pages(spool, chg);
3695
		goto out_err;
K
Ken Chen 已提交
3696
	}
3697 3698 3699 3700 3701 3702 3703 3704 3705 3706 3707 3708

	/*
	 * 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
	 */
3709 3710 3711 3712 3713 3714 3715 3716 3717 3718 3719 3720 3721 3722 3723 3724 3725 3726
	if (!vma || vma->vm_flags & VM_MAYSHARE) {
		long add = region_add(resv_map, from, to);

		if (unlikely(chg > add)) {
			/*
			 * pages in this range were added to the reserve
			 * map between region_chg and region_add.  This
			 * indicates a race with alloc_huge_page.  Adjust
			 * the subpool and reserve counts modified above
			 * based on the difference.
			 */
			long rsv_adjust;

			rsv_adjust = hugepage_subpool_put_pages(spool,
								chg - add);
			hugetlb_acct_memory(h, -rsv_adjust);
		}
	}
3727
	return 0;
3728
out_err:
J
Joonsoo Kim 已提交
3729 3730
	if (vma && is_vma_resv_set(vma, HPAGE_RESV_OWNER))
		kref_put(&resv_map->refs, resv_map_release);
3731
	return ret;
3732 3733 3734 3735
}

void hugetlb_unreserve_pages(struct inode *inode, long offset, long freed)
{
3736
	struct hstate *h = hstate_inode(inode);
3737
	struct resv_map *resv_map = inode_resv_map(inode);
3738
	long chg = 0;
3739
	struct hugepage_subpool *spool = subpool_inode(inode);
3740
	long gbl_reserve;
K
Ken Chen 已提交
3741

3742
	if (resv_map)
3743
		chg = region_truncate(resv_map, offset);
K
Ken Chen 已提交
3744
	spin_lock(&inode->i_lock);
3745
	inode->i_blocks -= (blocks_per_huge_page(h) * freed);
K
Ken Chen 已提交
3746 3747
	spin_unlock(&inode->i_lock);

3748 3749 3750 3751 3752 3753
	/*
	 * If the subpool has a minimum size, the number of global
	 * reservations to be released may be adjusted.
	 */
	gbl_reserve = hugepage_subpool_put_pages(spool, (chg - freed));
	hugetlb_acct_memory(h, -gbl_reserve);
3754
}
3755

3756 3757 3758 3759 3760 3761 3762 3763 3764 3765 3766 3767 3768 3769 3770 3771 3772 3773 3774 3775 3776 3777 3778 3779 3780 3781
#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;
}

3782
static bool vma_shareable(struct vm_area_struct *vma, unsigned long addr)
3783 3784 3785 3786 3787 3788 3789 3790 3791
{
	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)
3792 3793
		return true;
	return false;
3794 3795 3796 3797 3798 3799 3800
}

/*
 * 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
3801
 * pud has to be populated inside the same i_mmap_rwsem section - otherwise
3802 3803 3804 3805 3806 3807 3808 3809 3810 3811 3812 3813 3814
 * 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;
3815
	spinlock_t *ptl;
3816 3817 3818 3819

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

3820
	i_mmap_lock_write(mapping);
3821 3822 3823 3824 3825 3826 3827 3828
	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) {
3829
				mm_inc_nr_pmds(mm);
3830 3831 3832 3833 3834 3835 3836 3837 3838
				get_page(virt_to_page(spte));
				break;
			}
		}
	}

	if (!spte)
		goto out;

3839 3840
	ptl = huge_pte_lockptr(hstate_vma(vma), mm, spte);
	spin_lock(ptl);
3841
	if (pud_none(*pud)) {
3842 3843
		pud_populate(mm, pud,
				(pmd_t *)((unsigned long)spte & PAGE_MASK));
3844
	} else {
3845
		put_page(virt_to_page(spte));
3846 3847
		mm_inc_nr_pmds(mm);
	}
3848
	spin_unlock(ptl);
3849 3850
out:
	pte = (pte_t *)pmd_alloc(mm, pud, addr);
3851
	i_mmap_unlock_write(mapping);
3852 3853 3854 3855 3856 3857 3858 3859 3860 3861
	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.
 *
3862
 * called with page table lock held.
3863 3864 3865 3866 3867 3868 3869 3870 3871 3872 3873 3874 3875 3876 3877
 *
 * 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));
3878
	mm_dec_nr_pmds(mm);
3879 3880 3881
	*addr = ALIGN(*addr, HPAGE_SIZE * PTRS_PER_PTE) - HPAGE_SIZE;
	return 1;
}
3882 3883 3884 3885 3886 3887
#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;
}
3888 3889 3890 3891 3892

int huge_pmd_unshare(struct mm_struct *mm, unsigned long *addr, pte_t *ptep)
{
	return 0;
}
3893
#define want_pmd_share()	(0)
3894 3895
#endif /* CONFIG_ARCH_WANT_HUGE_PMD_SHARE */

3896 3897 3898 3899 3900 3901 3902 3903 3904 3905 3906 3907 3908 3909 3910 3911 3912 3913 3914 3915 3916 3917 3918 3919 3920 3921 3922 3923 3924 3925 3926 3927 3928 3929 3930 3931 3932 3933 3934 3935 3936 3937 3938 3939
#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;
}

3940 3941 3942 3943 3944 3945 3946 3947 3948 3949 3950 3951 3952 3953
#endif /* CONFIG_ARCH_WANT_GENERAL_HUGETLB */

/*
 * These functions are overwritable if your architecture needs its own
 * behavior.
 */
struct page * __weak
follow_huge_addr(struct mm_struct *mm, unsigned long address,
			      int write)
{
	return ERR_PTR(-EINVAL);
}

struct page * __weak
3954
follow_huge_pmd(struct mm_struct *mm, unsigned long address,
3955
		pmd_t *pmd, int flags)
3956
{
3957 3958 3959 3960 3961 3962 3963 3964 3965 3966 3967 3968
	struct page *page = NULL;
	spinlock_t *ptl;
retry:
	ptl = pmd_lockptr(mm, pmd);
	spin_lock(ptl);
	/*
	 * make sure that the address range covered by this pmd is not
	 * unmapped from other threads.
	 */
	if (!pmd_huge(*pmd))
		goto out;
	if (pmd_present(*pmd)) {
3969
		page = pmd_page(*pmd) + ((address & ~PMD_MASK) >> PAGE_SHIFT);
3970 3971 3972 3973 3974 3975 3976 3977 3978 3979 3980 3981 3982 3983 3984
		if (flags & FOLL_GET)
			get_page(page);
	} else {
		if (is_hugetlb_entry_migration(huge_ptep_get((pte_t *)pmd))) {
			spin_unlock(ptl);
			__migration_entry_wait(mm, (pte_t *)pmd, ptl);
			goto retry;
		}
		/*
		 * hwpoisoned entry is treated as no_page_table in
		 * follow_page_mask().
		 */
	}
out:
	spin_unlock(ptl);
3985 3986 3987
	return page;
}

3988
struct page * __weak
3989
follow_huge_pud(struct mm_struct *mm, unsigned long address,
3990
		pud_t *pud, int flags)
3991
{
3992 3993
	if (flags & FOLL_GET)
		return NULL;
3994

3995
	return pte_page(*(pte_t *)pud) + ((address & ~PUD_MASK) >> PAGE_SHIFT);
3996 3997
}

3998 3999
#ifdef CONFIG_MEMORY_FAILURE

4000 4001 4002 4003
/*
 * This function is called from memory failure code.
 * Assume the caller holds page lock of the head page.
 */
4004
int dequeue_hwpoisoned_huge_page(struct page *hpage)
4005 4006 4007
{
	struct hstate *h = page_hstate(hpage);
	int nid = page_to_nid(hpage);
4008
	int ret = -EBUSY;
4009 4010

	spin_lock(&hugetlb_lock);
4011 4012 4013 4014 4015
	/*
	 * Just checking !page_huge_active is not enough, because that could be
	 * an isolated/hwpoisoned hugepage (which have >0 refcount).
	 */
	if (!page_huge_active(hpage) && !page_count(hpage)) {
4016 4017 4018 4019 4020 4021 4022
		/*
		 * 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);
4023
		set_page_refcounted(hpage);
4024 4025 4026 4027
		h->free_huge_pages--;
		h->free_huge_pages_node[nid]--;
		ret = 0;
	}
4028
	spin_unlock(&hugetlb_lock);
4029
	return ret;
4030
}
4031
#endif
4032 4033 4034

bool isolate_huge_page(struct page *page, struct list_head *list)
{
4035 4036
	bool ret = true;

4037
	VM_BUG_ON_PAGE(!PageHead(page), page);
4038
	spin_lock(&hugetlb_lock);
4039 4040 4041 4042 4043
	if (!page_huge_active(page) || !get_page_unless_zero(page)) {
		ret = false;
		goto unlock;
	}
	clear_page_huge_active(page);
4044
	list_move_tail(&page->lru, list);
4045
unlock:
4046
	spin_unlock(&hugetlb_lock);
4047
	return ret;
4048 4049 4050 4051
}

void putback_active_hugepage(struct page *page)
{
4052
	VM_BUG_ON_PAGE(!PageHead(page), page);
4053
	spin_lock(&hugetlb_lock);
4054
	set_page_huge_active(page);
4055 4056 4057 4058
	list_move_tail(&page->lru, &(page_hstate(page))->hugepage_activelist);
	spin_unlock(&hugetlb_lock);
	put_page(page);
}