hugetlb.c 114.6 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
/*
 * 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;
67
struct mutex *hugetlb_fault_mutex_table ____cacheline_aligned_in_smp;
68

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
/*
 * Add the huge page range represented by [f, t) to the reserve
243 244 245 246 247 248 249 250
 * map.  In the normal case, existing regions will be expanded
 * to accommodate the specified range.  Sufficient regions should
 * exist for expansion due to the previous call to region_chg
 * with the same range.  However, it is possible that region_del
 * could have been called after region_chg and modifed the map
 * in such a way that no region exists to be expanded.  In this
 * case, pull a region descriptor from the cache associated with
 * the map and use that for the new range.
251 252 253
 *
 * Return the number of new huge pages added to the map.  This
 * number is greater than or equal to zero.
254
 */
255
static long region_add(struct resv_map *resv, long f, long t)
256
{
257
	struct list_head *head = &resv->regions;
258
	struct file_region *rg, *nrg, *trg;
259
	long add = 0;
260

261
	spin_lock(&resv->lock);
262 263 264 265 266
	/* Locate the region we are either in or before. */
	list_for_each_entry(rg, head, link)
		if (f <= rg->to)
			break;

267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288
	/*
	 * If no region exists which can be expanded to include the
	 * specified range, the list must have been modified by an
	 * interleving call to region_del().  Pull a region descriptor
	 * from the cache and use it for this range.
	 */
	if (&rg->link == head || t < rg->from) {
		VM_BUG_ON(resv->region_cache_count <= 0);

		resv->region_cache_count--;
		nrg = list_first_entry(&resv->region_cache, struct file_region,
					link);
		list_del(&nrg->link);

		nrg->from = f;
		nrg->to = t;
		list_add(&nrg->link, rg->link.prev);

		add += t - f;
		goto out_locked;
	}

289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306
	/* 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) {
307 308 309 310 311
			/* 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);
312 313 314 315
			list_del(&rg->link);
			kfree(rg);
		}
	}
316 317

	add += (nrg->from - f);		/* Added to beginning of region */
318
	nrg->from = f;
319
	add += t - nrg->to;		/* Added to end of region */
320
	nrg->to = t;
321

322 323
out_locked:
	resv->adds_in_progress--;
324
	spin_unlock(&resv->lock);
325 326
	VM_BUG_ON(add < 0);
	return add;
327 328
}

329 330 331 332 333 334 335 336 337 338 339 340 341
/*
 * 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.
 *
342 343 344 345 346 347 348 349
 * Upon entry, region_chg will also examine the cache of region descriptors
 * associated with the map.  If there are not enough descriptors cached, one
 * will be allocated for the in progress add operation.
 *
 * 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 or cache entry
 * is needed and can not be allocated.
350
 */
351
static long region_chg(struct resv_map *resv, long f, long t)
352
{
353
	struct list_head *head = &resv->regions;
354
	struct file_region *rg, *nrg = NULL;
355 356
	long chg = 0;

357 358
retry:
	spin_lock(&resv->lock);
359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383
retry_locked:
	resv->adds_in_progress++;

	/*
	 * Check for sufficient descriptors in the cache to accommodate
	 * the number of in progress add operations.
	 */
	if (resv->adds_in_progress > resv->region_cache_count) {
		struct file_region *trg;

		VM_BUG_ON(resv->adds_in_progress - resv->region_cache_count > 1);
		/* Must drop lock to allocate a new descriptor. */
		resv->adds_in_progress--;
		spin_unlock(&resv->lock);

		trg = kmalloc(sizeof(*trg), GFP_KERNEL);
		if (!trg)
			return -ENOMEM;

		spin_lock(&resv->lock);
		list_add(&trg->link, &resv->region_cache);
		resv->region_cache_count++;
		goto retry_locked;
	}

384 385 386 387 388 389 390 391 392
	/* 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) {
393
		if (!nrg) {
394
			resv->adds_in_progress--;
395 396 397 398 399 400 401 402 403 404
			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;
		}
405

406 407 408
		list_add(&nrg->link, rg->link.prev);
		chg = t - f;
		goto out_nrg;
409 410 411 412 413 414 415 416 417 418 419 420
	}

	/* 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)
421
			goto out;
422

L
Lucas De Marchi 已提交
423
		/* We overlap with this area, if it extends further than
424 425 426 427 428 429 430 431
		 * 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;
	}
432 433 434 435 436 437 438 439

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);
440 441 442
	return chg;
}

443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461
/*
 * Abort the in progress add operation.  The adds_in_progress field
 * of the resv_map keeps track of the operations in progress between
 * calls to region_chg and region_add.  Operations are sometimes
 * aborted after the call to region_chg.  In such cases, region_abort
 * is called to decrement the adds_in_progress counter.
 *
 * NOTE: The range arguments [f, t) are not needed or used in this
 * routine.  They are kept to make reading the calling code easier as
 * arguments will match the associated region_chg call.
 */
static void region_abort(struct resv_map *resv, long f, long t)
{
	spin_lock(&resv->lock);
	VM_BUG_ON(!resv->region_cache_count);
	resv->adds_in_progress--;
	spin_unlock(&resv->lock);
}

462
/*
463 464 465 466 467 468 469 470 471 472 473 474
 * Delete the specified range [f, t) from the reserve map.  If the
 * t parameter is LONG_MAX, this indicates that ALL regions after f
 * should be deleted.  Locate the regions which intersect [f, t)
 * and either trim, delete or split the existing regions.
 *
 * Returns the number of huge pages deleted from the reserve map.
 * In the normal case, the return value is zero or more.  In the
 * case where a region must be split, a new region descriptor must
 * be allocated.  If the allocation fails, -ENOMEM will be returned.
 * NOTE: If the parameter t == LONG_MAX, then we will never split
 * a region and possibly return -ENOMEM.  Callers specifying
 * t == LONG_MAX do not need to check for -ENOMEM error.
475
 */
476
static long region_del(struct resv_map *resv, long f, long t)
477
{
478
	struct list_head *head = &resv->regions;
479
	struct file_region *rg, *trg;
480 481
	struct file_region *nrg = NULL;
	long del = 0;
482

483
retry:
484
	spin_lock(&resv->lock);
485 486 487 488
	list_for_each_entry_safe(rg, trg, head, link) {
		if (rg->to <= f)
			continue;
		if (rg->from >= t)
489 490
			break;

491 492 493 494 495 496 497 498 499 500 501 502 503
		if (f > rg->from && t < rg->to) { /* Must split region */
			/*
			 * Check for an entry in the cache before dropping
			 * lock and attempting allocation.
			 */
			if (!nrg &&
			    resv->region_cache_count > resv->adds_in_progress) {
				nrg = list_first_entry(&resv->region_cache,
							struct file_region,
							link);
				list_del(&nrg->link);
				resv->region_cache_count--;
			}
504

505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524
			if (!nrg) {
				spin_unlock(&resv->lock);
				nrg = kmalloc(sizeof(*nrg), GFP_KERNEL);
				if (!nrg)
					return -ENOMEM;
				goto retry;
			}

			del += t - f;

			/* New entry for end of split region */
			nrg->from = t;
			nrg->to = rg->to;
			INIT_LIST_HEAD(&nrg->link);

			/* Original entry is trimmed */
			rg->to = f;

			list_add(&nrg->link, &rg->link);
			nrg = NULL;
525
			break;
526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541
		}

		if (f <= rg->from && t >= rg->to) { /* Remove entire region */
			del += rg->to - rg->from;
			list_del(&rg->link);
			kfree(rg);
			continue;
		}

		if (f <= rg->from) {	/* Trim beginning of region */
			del += t - rg->from;
			rg->from = t;
		} else {		/* Trim end of region */
			del += rg->to - f;
			rg->to = f;
		}
542
	}
543 544

	spin_unlock(&resv->lock);
545 546
	kfree(nrg);
	return del;
547 548
}

549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570
/*
 * A rare out of memory error was encountered which prevented removal of
 * the reserve map region for a page.  The huge page itself was free'ed
 * and removed from the page cache.  This routine will adjust the subpool
 * usage count, and the global reserve count if needed.  By incrementing
 * these counts, the reserve map entry which could not be deleted will
 * appear as a "reserved" entry instead of simply dangling with incorrect
 * counts.
 */
void hugetlb_fix_reserve_counts(struct inode *inode, bool restore_reserve)
{
	struct hugepage_subpool *spool = subpool_inode(inode);
	long rsv_adjust;

	rsv_adjust = hugepage_subpool_get_pages(spool, 1);
	if (restore_reserve && rsv_adjust) {
		struct hstate *h = hstate_inode(inode);

		hugetlb_acct_memory(h, 1);
	}
}

571 572 573 574
/*
 * Count and return the number of huge pages in the reserve map
 * that intersect with the range [f, t).
 */
575
static long region_count(struct resv_map *resv, long f, long t)
576
{
577
	struct list_head *head = &resv->regions;
578 579 580
	struct file_region *rg;
	long chg = 0;

581
	spin_lock(&resv->lock);
582 583
	/* Locate each segment we overlap with, and count that overlap. */
	list_for_each_entry(rg, head, link) {
584 585
		long seg_from;
		long seg_to;
586 587 588 589 590 591 592 593 594 595 596

		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;
	}
597
	spin_unlock(&resv->lock);
598 599 600 601

	return chg;
}

602 603 604 605
/*
 * Convert the address within this vma to the page offset within
 * the mapping, in pagecache page units; huge pages here.
 */
606 607
static pgoff_t vma_hugecache_offset(struct hstate *h,
			struct vm_area_struct *vma, unsigned long address)
608
{
609 610
	return ((address - vma->vm_start) >> huge_page_shift(h)) +
			(vma->vm_pgoff >> huge_page_order(h));
611 612
}

613 614 615 616 617 618
pgoff_t linear_hugepage_index(struct vm_area_struct *vma,
				     unsigned long address)
{
	return vma_hugecache_offset(hstate_vma(vma), vma, address);
}

619 620 621 622 623 624 625 626 627 628 629 630 631
/*
 * 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);

632
	return 1UL << huge_page_shift(hstate);
633
}
634
EXPORT_SYMBOL_GPL(vma_kernel_pagesize);
635

636 637 638 639 640 641 642 643 644 645 646 647 648
/*
 * 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

649 650 651 652 653 654 655
/*
 * 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)
656
#define HPAGE_RESV_MASK (HPAGE_RESV_OWNER | HPAGE_RESV_UNMAPPED)
657

658 659 660 661 662 663 664 665 666
/*
 * 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.
667 668 669 670 671 672 673 674 675
 *
 * 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.
676
 */
677 678 679 680 681 682 683 684 685 686 687
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;
}

688
struct resv_map *resv_map_alloc(void)
689 690
{
	struct resv_map *resv_map = kmalloc(sizeof(*resv_map), GFP_KERNEL);
691 692 693 694 695
	struct file_region *rg = kmalloc(sizeof(*rg), GFP_KERNEL);

	if (!resv_map || !rg) {
		kfree(resv_map);
		kfree(rg);
696
		return NULL;
697
	}
698 699

	kref_init(&resv_map->refs);
700
	spin_lock_init(&resv_map->lock);
701 702
	INIT_LIST_HEAD(&resv_map->regions);

703 704 705 706 707 708
	resv_map->adds_in_progress = 0;

	INIT_LIST_HEAD(&resv_map->region_cache);
	list_add(&rg->link, &resv_map->region_cache);
	resv_map->region_cache_count = 1;

709 710 711
	return resv_map;
}

712
void resv_map_release(struct kref *ref)
713 714
{
	struct resv_map *resv_map = container_of(ref, struct resv_map, refs);
715 716
	struct list_head *head = &resv_map->region_cache;
	struct file_region *rg, *trg;
717 718

	/* Clear out any active regions before we release the map. */
719
	region_del(resv_map, 0, LONG_MAX);
720 721 722 723 724 725 726 727 728

	/* ... and any entries left in the cache */
	list_for_each_entry_safe(rg, trg, head, link) {
		list_del(&rg->link);
		kfree(rg);
	}

	VM_BUG_ON(resv_map->adds_in_progress);

729 730 731
	kfree(resv_map);
}

732 733 734 735 736
static inline struct resv_map *inode_resv_map(struct inode *inode)
{
	return inode->i_mapping->private_data;
}

737
static struct resv_map *vma_resv_map(struct vm_area_struct *vma)
738
{
739
	VM_BUG_ON_VMA(!is_vm_hugetlb_page(vma), vma);
740 741 742 743 744 745 746
	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 {
747 748
		return (struct resv_map *)(get_vma_private_data(vma) &
							~HPAGE_RESV_MASK);
749
	}
750 751
}

752
static void set_vma_resv_map(struct vm_area_struct *vma, struct resv_map *map)
753
{
754 755
	VM_BUG_ON_VMA(!is_vm_hugetlb_page(vma), vma);
	VM_BUG_ON_VMA(vma->vm_flags & VM_MAYSHARE, vma);
756

757 758
	set_vma_private_data(vma, (get_vma_private_data(vma) &
				HPAGE_RESV_MASK) | (unsigned long)map);
759 760 761 762
}

static void set_vma_resv_flags(struct vm_area_struct *vma, unsigned long flags)
{
763 764
	VM_BUG_ON_VMA(!is_vm_hugetlb_page(vma), vma);
	VM_BUG_ON_VMA(vma->vm_flags & VM_MAYSHARE, vma);
765 766

	set_vma_private_data(vma, get_vma_private_data(vma) | flags);
767 768 769 770
}

static int is_vma_resv_set(struct vm_area_struct *vma, unsigned long flag)
{
771
	VM_BUG_ON_VMA(!is_vm_hugetlb_page(vma), vma);
772 773

	return (get_vma_private_data(vma) & flag) != 0;
774 775
}

776
/* Reset counters to 0 and clear all HPAGE_RESV_* flags */
777 778
void reset_vma_resv_huge_pages(struct vm_area_struct *vma)
{
779
	VM_BUG_ON_VMA(!is_vm_hugetlb_page(vma), vma);
780
	if (!(vma->vm_flags & VM_MAYSHARE))
781 782 783 784
		vma->vm_private_data = (void *)0;
}

/* Returns true if the VMA has associated reserve pages */
785
static bool vma_has_reserves(struct vm_area_struct *vma, long chg)
786
{
787 788 789 790 791 792 793 794 795 796 797
	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)
798
			return true;
799
		else
800
			return false;
801
	}
802 803

	/* Shared mappings always use reserves */
804 805 806 807 808 809 810 811 812 813 814 815 816
	if (vma->vm_flags & VM_MAYSHARE) {
		/*
		 * We know VM_NORESERVE is not set.  Therefore, there SHOULD
		 * be a region map for all pages.  The only situation where
		 * there is no region map is if a hole was punched via
		 * fallocate.  In this case, there really are no reverves to
		 * use.  This situation is indicated if chg != 0.
		 */
		if (chg)
			return false;
		else
			return true;
	}
817 818 819 820 821

	/*
	 * Only the process that called mmap() has reserves for
	 * private mappings.
	 */
822
	if (is_vma_resv_set(vma, HPAGE_RESV_OWNER))
823
		return true;
824

825
	return false;
826 827
}

828
static void enqueue_huge_page(struct hstate *h, struct page *page)
L
Linus Torvalds 已提交
829 830
{
	int nid = page_to_nid(page);
831
	list_move(&page->lru, &h->hugepage_freelists[nid]);
832 833
	h->free_huge_pages++;
	h->free_huge_pages_node[nid]++;
L
Linus Torvalds 已提交
834 835
}

836 837 838 839
static struct page *dequeue_huge_page_node(struct hstate *h, int nid)
{
	struct page *page;

840 841 842 843 844 845 846 847
	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)
848
		return NULL;
849
	list_move(&page->lru, &h->hugepage_activelist);
850
	set_page_refcounted(page);
851 852 853 854 855
	h->free_huge_pages--;
	h->free_huge_pages_node[nid]--;
	return page;
}

856 857 858
/* Movability of hugepages depends on migration support. */
static inline gfp_t htlb_alloc_mask(struct hstate *h)
{
859
	if (hugepages_treat_as_movable || hugepage_migration_supported(h))
860 861 862 863 864
		return GFP_HIGHUSER_MOVABLE;
	else
		return GFP_HIGHUSER;
}

865 866
static struct page *dequeue_huge_page_vma(struct hstate *h,
				struct vm_area_struct *vma,
867 868
				unsigned long address, int avoid_reserve,
				long chg)
L
Linus Torvalds 已提交
869
{
870
	struct page *page = NULL;
871
	struct mempolicy *mpol;
872
	nodemask_t *nodemask;
873
	struct zonelist *zonelist;
874 875
	struct zone *zone;
	struct zoneref *z;
876
	unsigned int cpuset_mems_cookie;
L
Linus Torvalds 已提交
877

878 879 880 881 882
	/*
	 * 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
	 */
883
	if (!vma_has_reserves(vma, chg) &&
884
			h->free_huge_pages - h->resv_huge_pages == 0)
885
		goto err;
886

887
	/* If reserves cannot be used, ensure enough pages are in the pool */
888
	if (avoid_reserve && h->free_huge_pages - h->resv_huge_pages == 0)
889
		goto err;
890

891
retry_cpuset:
892
	cpuset_mems_cookie = read_mems_allowed_begin();
893
	zonelist = huge_zonelist(vma, address,
894
					htlb_alloc_mask(h), &mpol, &nodemask);
895

896 897
	for_each_zone_zonelist_nodemask(zone, z, zonelist,
						MAX_NR_ZONES - 1, nodemask) {
898
		if (cpuset_zone_allowed(zone, htlb_alloc_mask(h))) {
899 900
			page = dequeue_huge_page_node(h, zone_to_nid(zone));
			if (page) {
901 902 903 904 905
				if (avoid_reserve)
					break;
				if (!vma_has_reserves(vma, chg))
					break;

906
				SetPagePrivate(page);
907
				h->resv_huge_pages--;
908 909
				break;
			}
A
Andrew Morton 已提交
910
		}
L
Linus Torvalds 已提交
911
	}
912

913
	mpol_cond_put(mpol);
914
	if (unlikely(!page && read_mems_allowed_retry(cpuset_mems_cookie)))
915
		goto retry_cpuset;
L
Linus Torvalds 已提交
916
	return page;
917 918 919

err:
	return NULL;
L
Linus Torvalds 已提交
920 921
}

922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994
/*
 * 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--)

995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132
#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

1133
static void update_and_free_page(struct hstate *h, struct page *page)
A
Adam Litke 已提交
1134 1135
{
	int i;
1136

1137 1138
	if (hstate_is_gigantic(h) && !gigantic_page_supported())
		return;
1139

1140 1141 1142
	h->nr_huge_pages--;
	h->nr_huge_pages_node[page_to_nid(page)]--;
	for (i = 0; i < pages_per_huge_page(h); i++) {
1143 1144
		page[i].flags &= ~(1 << PG_locked | 1 << PG_error |
				1 << PG_referenced | 1 << PG_dirty |
1145 1146
				1 << PG_active | 1 << PG_private |
				1 << PG_writeback);
A
Adam Litke 已提交
1147
	}
1148
	VM_BUG_ON_PAGE(hugetlb_cgroup_from_page(page), page);
A
Adam Litke 已提交
1149 1150
	set_compound_page_dtor(page, NULL);
	set_page_refcounted(page);
1151 1152 1153 1154 1155 1156
	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
1157 1158
}

1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169
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;
}

1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194
/*
 * 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]);
}

1195
void free_huge_page(struct page *page)
1196
{
1197 1198 1199 1200
	/*
	 * Can't pass hstate in here because it is called from the
	 * compound page destructor.
	 */
1201
	struct hstate *h = page_hstate(page);
1202
	int nid = page_to_nid(page);
1203 1204
	struct hugepage_subpool *spool =
		(struct hugepage_subpool *)page_private(page);
1205
	bool restore_reserve;
1206

1207
	set_page_private(page, 0);
1208
	page->mapping = NULL;
1209
	BUG_ON(page_count(page));
1210
	BUG_ON(page_mapcount(page));
1211
	restore_reserve = PagePrivate(page);
1212
	ClearPagePrivate(page);
1213

1214 1215 1216 1217 1218 1219 1220 1221
	/*
	 * 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;

1222
	spin_lock(&hugetlb_lock);
1223
	clear_page_huge_active(page);
1224 1225
	hugetlb_cgroup_uncharge_page(hstate_index(h),
				     pages_per_huge_page(h), page);
1226 1227 1228
	if (restore_reserve)
		h->resv_huge_pages++;

1229
	if (h->surplus_huge_pages_node[nid]) {
1230 1231
		/* remove the page from active list */
		list_del(&page->lru);
1232 1233 1234
		update_and_free_page(h, page);
		h->surplus_huge_pages--;
		h->surplus_huge_pages_node[nid]--;
1235
	} else {
1236
		arch_clear_hugepage_flags(page);
1237
		enqueue_huge_page(h, page);
1238
	}
1239 1240 1241
	spin_unlock(&hugetlb_lock);
}

1242
static void prep_new_huge_page(struct hstate *h, struct page *page, int nid)
1243
{
1244
	INIT_LIST_HEAD(&page->lru);
1245 1246
	set_compound_page_dtor(page, free_huge_page);
	spin_lock(&hugetlb_lock);
1247
	set_hugetlb_cgroup(page, NULL);
1248 1249
	h->nr_huge_pages++;
	h->nr_huge_pages_node[nid]++;
1250 1251 1252 1253
	spin_unlock(&hugetlb_lock);
	put_page(page); /* free it into the hugepage allocator */
}

1254
static void prep_compound_gigantic_page(struct page *page, unsigned long order)
1255 1256 1257 1258 1259 1260 1261 1262
{
	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);
1263
	__ClearPageReserved(page);
1264
	for (i = 1; i < nr_pages; i++, p = mem_map_next(p, page, i)) {
1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277
		/*
		 * 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);
1278
		set_page_count(p, 0);
1279
		p->first_page = page;
1280 1281 1282
		/* Make sure p->first_page is always valid for PageTail() */
		smp_wmb();
		__SetPageTail(p);
1283 1284 1285
	}
}

A
Andrew Morton 已提交
1286 1287 1288 1289 1290
/*
 * PageHuge() only returns true for hugetlbfs pages, but not for normal or
 * transparent huge pages.  See the PageTransHuge() documentation for more
 * details.
 */
1291 1292 1293 1294 1295 1296
int PageHuge(struct page *page)
{
	if (!PageCompound(page))
		return 0;

	page = compound_head(page);
1297
	return get_compound_page_dtor(page) == free_huge_page;
1298
}
1299 1300
EXPORT_SYMBOL_GPL(PageHuge);

1301 1302 1303 1304 1305 1306 1307 1308 1309
/*
 * 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;

1310
	return get_compound_page_dtor(page_head) == free_huge_page;
1311 1312
}

1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329
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;
}

1330
static struct page *alloc_fresh_huge_page_node(struct hstate *h, int nid)
L
Linus Torvalds 已提交
1331 1332
{
	struct page *page;
1333

1334
	page = __alloc_pages_node(nid,
1335
		htlb_alloc_mask(h)|__GFP_COMP|__GFP_THISNODE|
1336
						__GFP_REPEAT|__GFP_NOWARN,
1337
		huge_page_order(h));
L
Linus Torvalds 已提交
1338
	if (page) {
1339
		prep_new_huge_page(h, page, nid);
L
Linus Torvalds 已提交
1340
	}
1341 1342 1343 1344

	return page;
}

1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366
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;
}

1367 1368 1369 1370 1371 1372
/*
 * 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.
 */
1373 1374
static int free_pool_huge_page(struct hstate *h, nodemask_t *nodes_allowed,
							 bool acct_surplus)
1375
{
1376
	int nr_nodes, node;
1377 1378
	int ret = 0;

1379
	for_each_node_mask_to_free(h, nr_nodes, node, nodes_allowed) {
1380 1381 1382 1383
		/*
		 * If we're returning unused surplus pages, only examine
		 * nodes with surplus pages.
		 */
1384 1385
		if ((!acct_surplus || h->surplus_huge_pages_node[node]) &&
		    !list_empty(&h->hugepage_freelists[node])) {
1386
			struct page *page =
1387
				list_entry(h->hugepage_freelists[node].next,
1388 1389 1390
					  struct page, lru);
			list_del(&page->lru);
			h->free_huge_pages--;
1391
			h->free_huge_pages_node[node]--;
1392 1393
			if (acct_surplus) {
				h->surplus_huge_pages--;
1394
				h->surplus_huge_pages_node[node]--;
1395
			}
1396 1397
			update_and_free_page(h, page);
			ret = 1;
1398
			break;
1399
		}
1400
	}
1401 1402 1403 1404

	return ret;
}

1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431
/*
 * 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;

1432 1433 1434
	if (!hugepages_supported())
		return;

1435 1436
	VM_BUG_ON(!IS_ALIGNED(start_pfn, 1 << minimum_order));
	for (pfn = start_pfn; pfn < end_pfn; pfn += 1 << minimum_order)
1437 1438 1439
		dissolve_free_huge_page(pfn_to_page(pfn));
}

1440
static struct page *alloc_buddy_huge_page(struct hstate *h, int nid)
1441 1442
{
	struct page *page;
1443
	unsigned int r_nid;
1444

1445
	if (hstate_is_gigantic(h))
1446 1447
		return NULL;

1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471
	/*
	 * 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);
1472
	if (h->surplus_huge_pages >= h->nr_overcommit_huge_pages) {
1473 1474 1475
		spin_unlock(&hugetlb_lock);
		return NULL;
	} else {
1476 1477
		h->nr_huge_pages++;
		h->surplus_huge_pages++;
1478 1479 1480
	}
	spin_unlock(&hugetlb_lock);

1481
	if (nid == NUMA_NO_NODE)
1482
		page = alloc_pages(htlb_alloc_mask(h)|__GFP_COMP|
1483 1484 1485
				   __GFP_REPEAT|__GFP_NOWARN,
				   huge_page_order(h));
	else
1486
		page = __alloc_pages_node(nid,
1487
			htlb_alloc_mask(h)|__GFP_COMP|__GFP_THISNODE|
1488
			__GFP_REPEAT|__GFP_NOWARN, huge_page_order(h));
1489 1490

	spin_lock(&hugetlb_lock);
1491
	if (page) {
1492
		INIT_LIST_HEAD(&page->lru);
1493
		r_nid = page_to_nid(page);
1494
		set_compound_page_dtor(page, free_huge_page);
1495
		set_hugetlb_cgroup(page, NULL);
1496 1497 1498
		/*
		 * We incremented the global counters already
		 */
1499 1500
		h->nr_huge_pages_node[r_nid]++;
		h->surplus_huge_pages_node[r_nid]++;
1501
		__count_vm_event(HTLB_BUDDY_PGALLOC);
1502
	} else {
1503 1504
		h->nr_huge_pages--;
		h->surplus_huge_pages--;
1505
		__count_vm_event(HTLB_BUDDY_PGALLOC_FAIL);
1506
	}
1507
	spin_unlock(&hugetlb_lock);
1508 1509 1510 1511

	return page;
}

1512 1513 1514 1515 1516 1517 1518
/*
 * 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)
{
1519
	struct page *page = NULL;
1520 1521

	spin_lock(&hugetlb_lock);
1522 1523
	if (h->free_huge_pages - h->resv_huge_pages > 0)
		page = dequeue_huge_page_node(h, nid);
1524 1525
	spin_unlock(&hugetlb_lock);

1526
	if (!page)
1527 1528 1529 1530 1531
		page = alloc_buddy_huge_page(h, nid);

	return page;
}

1532
/*
L
Lucas De Marchi 已提交
1533
 * Increase the hugetlb pool such that it can accommodate a reservation
1534 1535
 * of size 'delta'.
 */
1536
static int gather_surplus_pages(struct hstate *h, int delta)
1537 1538 1539 1540 1541
{
	struct list_head surplus_list;
	struct page *page, *tmp;
	int ret, i;
	int needed, allocated;
1542
	bool alloc_ok = true;
1543

1544
	needed = (h->resv_huge_pages + delta) - h->free_huge_pages;
1545
	if (needed <= 0) {
1546
		h->resv_huge_pages += delta;
1547
		return 0;
1548
	}
1549 1550 1551 1552 1553 1554 1555 1556

	allocated = 0;
	INIT_LIST_HEAD(&surplus_list);

	ret = -ENOMEM;
retry:
	spin_unlock(&hugetlb_lock);
	for (i = 0; i < needed; i++) {
1557
		page = alloc_buddy_huge_page(h, NUMA_NO_NODE);
1558 1559 1560 1561
		if (!page) {
			alloc_ok = false;
			break;
		}
1562 1563
		list_add(&page->lru, &surplus_list);
	}
1564
	allocated += i;
1565 1566 1567 1568 1569 1570

	/*
	 * 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);
1571 1572
	needed = (h->resv_huge_pages + delta) -
			(h->free_huge_pages + allocated);
1573 1574 1575 1576 1577 1578 1579 1580 1581 1582
	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;
	}
1583 1584
	/*
	 * The surplus_list now contains _at_least_ the number of extra pages
L
Lucas De Marchi 已提交
1585
	 * needed to accommodate the reservation.  Add the appropriate number
1586
	 * of pages to the hugetlb pool and free the extras back to the buddy
1587 1588 1589
	 * 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.
1590 1591
	 */
	needed += allocated;
1592
	h->resv_huge_pages += delta;
1593
	ret = 0;
1594

1595
	/* Free the needed pages to the hugetlb pool */
1596
	list_for_each_entry_safe(page, tmp, &surplus_list, lru) {
1597 1598
		if ((--needed) < 0)
			break;
1599 1600 1601 1602 1603
		/*
		 * This page is now managed by the hugetlb allocator and has
		 * no users -- drop the buddy allocator's reference.
		 */
		put_page_testzero(page);
1604
		VM_BUG_ON_PAGE(page_count(page), page);
1605
		enqueue_huge_page(h, page);
1606
	}
1607
free:
1608
	spin_unlock(&hugetlb_lock);
1609 1610

	/* Free unnecessary surplus pages to the buddy allocator */
1611 1612
	list_for_each_entry_safe(page, tmp, &surplus_list, lru)
		put_page(page);
1613
	spin_lock(&hugetlb_lock);
1614 1615 1616 1617 1618 1619 1620 1621

	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.
1622
 * Called with hugetlb_lock held.
1623
 */
1624 1625
static void return_unused_surplus_pages(struct hstate *h,
					unsigned long unused_resv_pages)
1626 1627 1628
{
	unsigned long nr_pages;

1629
	/* Uncommit the reservation */
1630
	h->resv_huge_pages -= unused_resv_pages;
1631

1632
	/* Cannot return gigantic pages currently */
1633
	if (hstate_is_gigantic(h))
1634 1635
		return;

1636
	nr_pages = min(unused_resv_pages, h->surplus_huge_pages);
1637

1638 1639
	/*
	 * We want to release as many surplus pages as possible, spread
1640 1641 1642 1643 1644
	 * 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.
1645 1646
	 */
	while (nr_pages--) {
1647
		if (!free_pool_huge_page(h, &node_states[N_MEMORY], 1))
1648
			break;
1649
		cond_resched_lock(&hugetlb_lock);
1650 1651 1652
	}
}

1653

1654
/*
1655
 * vma_needs_reservation, vma_commit_reservation and vma_end_reservation
1656
 * are used by the huge page allocation routines to manage reservations.
1657 1658 1659 1660 1661 1662
 *
 * 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
1663 1664 1665
 * to add the page to the reservation map.  If the page allocation fails,
 * the reservation must be ended instead of committed.  vma_end_reservation
 * is called in such cases.
1666 1667 1668 1669 1670 1671
 *
 * 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.
1672
 */
1673 1674 1675
enum vma_resv_mode {
	VMA_NEEDS_RESV,
	VMA_COMMIT_RESV,
1676
	VMA_END_RESV,
1677
};
1678 1679
static long __vma_reservation_common(struct hstate *h,
				struct vm_area_struct *vma, unsigned long addr,
1680
				enum vma_resv_mode mode)
1681
{
1682 1683
	struct resv_map *resv;
	pgoff_t idx;
1684
	long ret;
1685

1686 1687
	resv = vma_resv_map(vma);
	if (!resv)
1688
		return 1;
1689

1690
	idx = vma_hugecache_offset(h, vma, addr);
1691 1692
	switch (mode) {
	case VMA_NEEDS_RESV:
1693
		ret = region_chg(resv, idx, idx + 1);
1694 1695 1696 1697
		break;
	case VMA_COMMIT_RESV:
		ret = region_add(resv, idx, idx + 1);
		break;
1698
	case VMA_END_RESV:
1699 1700 1701 1702 1703 1704
		region_abort(resv, idx, idx + 1);
		ret = 0;
		break;
	default:
		BUG();
	}
1705

1706
	if (vma->vm_flags & VM_MAYSHARE)
1707
		return ret;
1708
	else
1709
		return ret < 0 ? ret : 0;
1710
}
1711 1712

static long vma_needs_reservation(struct hstate *h,
1713
			struct vm_area_struct *vma, unsigned long addr)
1714
{
1715
	return __vma_reservation_common(h, vma, addr, VMA_NEEDS_RESV);
1716
}
1717

1718 1719 1720
static long vma_commit_reservation(struct hstate *h,
			struct vm_area_struct *vma, unsigned long addr)
{
1721 1722 1723
	return __vma_reservation_common(h, vma, addr, VMA_COMMIT_RESV);
}

1724
static void vma_end_reservation(struct hstate *h,
1725 1726
			struct vm_area_struct *vma, unsigned long addr)
{
1727
	(void)__vma_reservation_common(h, vma, addr, VMA_END_RESV);
1728 1729
}

1730
struct page *alloc_huge_page(struct vm_area_struct *vma,
1731
				    unsigned long addr, int avoid_reserve)
L
Linus Torvalds 已提交
1732
{
1733
	struct hugepage_subpool *spool = subpool_vma(vma);
1734
	struct hstate *h = hstate_vma(vma);
1735
	struct page *page;
1736 1737
	long map_chg, map_commit;
	long gbl_chg;
1738 1739
	int ret, idx;
	struct hugetlb_cgroup *h_cg;
1740

1741
	idx = hstate_index(h);
1742
	/*
1743 1744 1745
	 * Examine the region/reserve map to determine if the process
	 * has a reservation for the page to be allocated.  A return
	 * code of zero indicates a reservation exists (no change).
1746
	 */
1747 1748
	map_chg = gbl_chg = vma_needs_reservation(h, vma, addr);
	if (map_chg < 0)
1749
		return ERR_PTR(-ENOMEM);
1750 1751 1752 1753 1754 1755 1756 1757 1758 1759 1760

	/*
	 * Processes that did not create the mapping will have no
	 * reserves as indicated by the region/reserve map. Check
	 * that the allocation will not exceed the subpool limit.
	 * Allocations for MAP_NORESERVE mappings also need to be
	 * checked against any subpool limit.
	 */
	if (map_chg || avoid_reserve) {
		gbl_chg = hugepage_subpool_get_pages(spool, 1);
		if (gbl_chg < 0) {
1761
			vma_end_reservation(h, vma, addr);
1762
			return ERR_PTR(-ENOSPC);
1763
		}
L
Linus Torvalds 已提交
1764

1765 1766 1767 1768 1769 1770 1771 1772 1773 1774 1775 1776
		/*
		 * Even though there was no reservation in the region/reserve
		 * map, there could be reservations associated with the
		 * subpool that can be used.  This would be indicated if the
		 * return value of hugepage_subpool_get_pages() is zero.
		 * However, if avoid_reserve is specified we still avoid even
		 * the subpool reservations.
		 */
		if (avoid_reserve)
			gbl_chg = 1;
	}

1777
	ret = hugetlb_cgroup_charge_cgroup(idx, pages_per_huge_page(h), &h_cg);
1778 1779 1780
	if (ret)
		goto out_subpool_put;

L
Linus Torvalds 已提交
1781
	spin_lock(&hugetlb_lock);
1782 1783 1784 1785 1786 1787
	/*
	 * glb_chg is passed to indicate whether or not a page must be taken
	 * from the global free pool (global change).  gbl_chg == 0 indicates
	 * a reservation exists for the allocation.
	 */
	page = dequeue_huge_page_vma(h, vma, addr, avoid_reserve, gbl_chg);
1788
	if (!page) {
1789
		spin_unlock(&hugetlb_lock);
1790
		page = alloc_buddy_huge_page(h, NUMA_NO_NODE);
1791 1792 1793
		if (!page)
			goto out_uncharge_cgroup;

1794 1795
		spin_lock(&hugetlb_lock);
		list_move(&page->lru, &h->hugepage_activelist);
1796
		/* Fall through */
K
Ken Chen 已提交
1797
	}
1798 1799
	hugetlb_cgroup_commit_charge(idx, pages_per_huge_page(h), h_cg, page);
	spin_unlock(&hugetlb_lock);
1800

1801
	set_page_private(page, (unsigned long)spool);
1802

1803 1804
	map_commit = vma_commit_reservation(h, vma, addr);
	if (unlikely(map_chg > map_commit)) {
1805 1806 1807 1808 1809 1810 1811 1812 1813 1814 1815 1816 1817 1818
		/*
		 * 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);
	}
1819
	return page;
1820 1821 1822 1823

out_uncharge_cgroup:
	hugetlb_cgroup_uncharge_cgroup(idx, pages_per_huge_page(h), h_cg);
out_subpool_put:
1824
	if (map_chg || avoid_reserve)
1825
		hugepage_subpool_put_pages(spool, 1);
1826
	vma_end_reservation(h, vma, addr);
1827
	return ERR_PTR(-ENOSPC);
1828 1829
}

1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843
/*
 * 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;
}

1844
int __weak alloc_bootmem_huge_page(struct hstate *h)
1845 1846
{
	struct huge_bootmem_page *m;
1847
	int nr_nodes, node;
1848

1849
	for_each_node_mask_to_alloc(h, nr_nodes, node, &node_states[N_MEMORY]) {
1850 1851
		void *addr;

1852 1853 1854
		addr = memblock_virt_alloc_try_nid_nopanic(
				huge_page_size(h), huge_page_size(h),
				0, BOOTMEM_ALLOC_ACCESSIBLE, node);
1855 1856 1857 1858 1859 1860 1861
		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;
1862
			goto found;
1863 1864 1865 1866 1867
		}
	}
	return 0;

found:
1868
	BUG_ON(!IS_ALIGNED(virt_to_phys(m), huge_page_size(h)));
1869 1870 1871 1872 1873 1874
	/* 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;
}

1875
static void __init prep_compound_huge_page(struct page *page, int order)
1876 1877 1878 1879 1880 1881 1882
{
	if (unlikely(order > (MAX_ORDER - 1)))
		prep_compound_gigantic_page(page, order);
	else
		prep_compound_page(page, order);
}

1883 1884 1885 1886 1887 1888 1889
/* 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;
1890 1891 1892 1893
		struct page *page;

#ifdef CONFIG_HIGHMEM
		page = pfn_to_page(m->phys >> PAGE_SHIFT);
1894 1895
		memblock_free_late(__pa(m),
				   sizeof(struct huge_bootmem_page));
1896 1897 1898
#else
		page = virt_to_page(m);
#endif
1899
		WARN_ON(page_count(page) != 1);
1900
		prep_compound_huge_page(page, h->order);
1901
		WARN_ON(PageReserved(page));
1902
		prep_new_huge_page(h, page, page_to_nid(page));
1903 1904 1905 1906 1907 1908
		/*
		 * 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.
		 */
1909
		if (hstate_is_gigantic(h))
1910
			adjust_managed_page_count(page, 1 << h->order);
1911 1912 1913
	}
}

1914
static void __init hugetlb_hstate_alloc_pages(struct hstate *h)
L
Linus Torvalds 已提交
1915 1916
{
	unsigned long i;
1917

1918
	for (i = 0; i < h->max_huge_pages; ++i) {
1919
		if (hstate_is_gigantic(h)) {
1920 1921
			if (!alloc_bootmem_huge_page(h))
				break;
1922
		} else if (!alloc_fresh_huge_page(h,
1923
					 &node_states[N_MEMORY]))
L
Linus Torvalds 已提交
1924 1925
			break;
	}
1926
	h->max_huge_pages = i;
1927 1928 1929 1930 1931 1932 1933
}

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

	for_each_hstate(h) {
1934 1935 1936
		if (minimum_order > huge_page_order(h))
			minimum_order = huge_page_order(h);

1937
		/* oversize hugepages were init'ed in early boot */
1938
		if (!hstate_is_gigantic(h))
1939
			hugetlb_hstate_alloc_pages(h);
1940
	}
1941
	VM_BUG_ON(minimum_order == UINT_MAX);
1942 1943
}

A
Andi Kleen 已提交
1944 1945 1946 1947 1948 1949 1950 1951 1952 1953 1954
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;
}

1955 1956 1957 1958 1959
static void __init report_hugepages(void)
{
	struct hstate *h;

	for_each_hstate(h) {
A
Andi Kleen 已提交
1960
		char buf[32];
1961
		pr_info("HugeTLB registered %s page size, pre-allocated %ld pages\n",
A
Andi Kleen 已提交
1962 1963
			memfmt(buf, huge_page_size(h)),
			h->free_huge_pages);
1964 1965 1966
	}
}

L
Linus Torvalds 已提交
1967
#ifdef CONFIG_HIGHMEM
1968 1969
static void try_to_free_low(struct hstate *h, unsigned long count,
						nodemask_t *nodes_allowed)
L
Linus Torvalds 已提交
1970
{
1971 1972
	int i;

1973
	if (hstate_is_gigantic(h))
1974 1975
		return;

1976
	for_each_node_mask(i, *nodes_allowed) {
L
Linus Torvalds 已提交
1977
		struct page *page, *next;
1978 1979 1980
		struct list_head *freel = &h->hugepage_freelists[i];
		list_for_each_entry_safe(page, next, freel, lru) {
			if (count >= h->nr_huge_pages)
1981
				return;
L
Linus Torvalds 已提交
1982 1983 1984
			if (PageHighMem(page))
				continue;
			list_del(&page->lru);
1985
			update_and_free_page(h, page);
1986 1987
			h->free_huge_pages--;
			h->free_huge_pages_node[page_to_nid(page)]--;
L
Linus Torvalds 已提交
1988 1989 1990 1991
		}
	}
}
#else
1992 1993
static inline void try_to_free_low(struct hstate *h, unsigned long count,
						nodemask_t *nodes_allowed)
L
Linus Torvalds 已提交
1994 1995 1996 1997
{
}
#endif

1998 1999 2000 2001 2002
/*
 * 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.
 */
2003 2004
static int adjust_pool_surplus(struct hstate *h, nodemask_t *nodes_allowed,
				int delta)
2005
{
2006
	int nr_nodes, node;
2007 2008 2009

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

2010 2011 2012 2013
	if (delta < 0) {
		for_each_node_mask_to_alloc(h, nr_nodes, node, nodes_allowed) {
			if (h->surplus_huge_pages_node[node])
				goto found;
2014
		}
2015 2016 2017 2018 2019
	} 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;
2020
		}
2021 2022
	}
	return 0;
2023

2024 2025 2026 2027
found:
	h->surplus_huge_pages += delta;
	h->surplus_huge_pages_node[node] += delta;
	return 1;
2028 2029
}

2030
#define persistent_huge_pages(h) (h->nr_huge_pages - h->surplus_huge_pages)
2031 2032
static unsigned long set_max_huge_pages(struct hstate *h, unsigned long count,
						nodemask_t *nodes_allowed)
L
Linus Torvalds 已提交
2033
{
2034
	unsigned long min_count, ret;
L
Linus Torvalds 已提交
2035

2036
	if (hstate_is_gigantic(h) && !gigantic_page_supported())
2037 2038
		return h->max_huge_pages;

2039 2040 2041 2042
	/*
	 * Increase the pool size
	 * First take pages out of surplus state.  Then make up the
	 * remaining difference by allocating fresh huge pages.
2043 2044 2045 2046 2047 2048
	 *
	 * 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.
2049
	 */
L
Linus Torvalds 已提交
2050
	spin_lock(&hugetlb_lock);
2051
	while (h->surplus_huge_pages && count > persistent_huge_pages(h)) {
2052
		if (!adjust_pool_surplus(h, nodes_allowed, -1))
2053 2054 2055
			break;
	}

2056
	while (count > persistent_huge_pages(h)) {
2057 2058 2059 2060 2061 2062
		/*
		 * 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);
2063 2064 2065 2066
		if (hstate_is_gigantic(h))
			ret = alloc_fresh_gigantic_page(h, nodes_allowed);
		else
			ret = alloc_fresh_huge_page(h, nodes_allowed);
2067 2068 2069 2070
		spin_lock(&hugetlb_lock);
		if (!ret)
			goto out;

2071 2072 2073
		/* Bail for signals. Probably ctrl-c from user */
		if (signal_pending(current))
			goto out;
2074 2075 2076 2077 2078 2079 2080 2081
	}

	/*
	 * 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.
2082 2083 2084 2085 2086 2087 2088 2089
	 *
	 * 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.
2090
	 */
2091
	min_count = h->resv_huge_pages + h->nr_huge_pages - h->free_huge_pages;
2092
	min_count = max(count, min_count);
2093
	try_to_free_low(h, min_count, nodes_allowed);
2094
	while (min_count < persistent_huge_pages(h)) {
2095
		if (!free_pool_huge_page(h, nodes_allowed, 0))
L
Linus Torvalds 已提交
2096
			break;
2097
		cond_resched_lock(&hugetlb_lock);
L
Linus Torvalds 已提交
2098
	}
2099
	while (count < persistent_huge_pages(h)) {
2100
		if (!adjust_pool_surplus(h, nodes_allowed, 1))
2101 2102 2103
			break;
	}
out:
2104
	ret = persistent_huge_pages(h);
L
Linus Torvalds 已提交
2105
	spin_unlock(&hugetlb_lock);
2106
	return ret;
L
Linus Torvalds 已提交
2107 2108
}

2109 2110 2111 2112 2113 2114 2115 2116 2117 2118
#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];

2119 2120 2121
static struct hstate *kobj_to_node_hstate(struct kobject *kobj, int *nidp);

static struct hstate *kobj_to_hstate(struct kobject *kobj, int *nidp)
2122 2123
{
	int i;
2124

2125
	for (i = 0; i < HUGE_MAX_HSTATE; i++)
2126 2127 2128
		if (hstate_kobjs[i] == kobj) {
			if (nidp)
				*nidp = NUMA_NO_NODE;
2129
			return &hstates[i];
2130 2131 2132
		}

	return kobj_to_node_hstate(kobj, nidp);
2133 2134
}

2135
static ssize_t nr_hugepages_show_common(struct kobject *kobj,
2136 2137
					struct kobj_attribute *attr, char *buf)
{
2138 2139 2140 2141 2142 2143 2144 2145 2146 2147 2148
	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);
2149
}
2150

2151 2152 2153
static ssize_t __nr_hugepages_store_common(bool obey_mempolicy,
					   struct hstate *h, int nid,
					   unsigned long count, size_t len)
2154 2155
{
	int err;
2156
	NODEMASK_ALLOC(nodemask_t, nodes_allowed, GFP_KERNEL | __GFP_NORETRY);
2157

2158
	if (hstate_is_gigantic(h) && !gigantic_page_supported()) {
2159 2160 2161 2162
		err = -EINVAL;
		goto out;
	}

2163 2164 2165 2166 2167 2168 2169
	if (nid == NUMA_NO_NODE) {
		/*
		 * global hstate attribute
		 */
		if (!(obey_mempolicy &&
				init_nodemask_of_mempolicy(nodes_allowed))) {
			NODEMASK_FREE(nodes_allowed);
2170
			nodes_allowed = &node_states[N_MEMORY];
2171 2172 2173 2174 2175 2176 2177 2178 2179
		}
	} 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
2180
		nodes_allowed = &node_states[N_MEMORY];
2181

2182
	h->max_huge_pages = set_max_huge_pages(h, count, nodes_allowed);
2183

2184
	if (nodes_allowed != &node_states[N_MEMORY])
2185 2186 2187
		NODEMASK_FREE(nodes_allowed);

	return len;
2188 2189 2190
out:
	NODEMASK_FREE(nodes_allowed);
	return err;
2191 2192
}

2193 2194 2195 2196 2197 2198 2199 2200 2201 2202 2203 2204 2205 2206 2207 2208 2209
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);
}

2210 2211 2212 2213 2214 2215 2216 2217 2218
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)
{
2219
	return nr_hugepages_store_common(false, kobj, buf, len);
2220 2221 2222
}
HSTATE_ATTR(nr_hugepages);

2223 2224 2225 2226 2227 2228 2229 2230 2231 2232 2233 2234 2235 2236 2237
#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)
{
2238
	return nr_hugepages_store_common(true, kobj, buf, len);
2239 2240 2241 2242 2243
}
HSTATE_ATTR(nr_hugepages_mempolicy);
#endif


2244 2245 2246
static ssize_t nr_overcommit_hugepages_show(struct kobject *kobj,
					struct kobj_attribute *attr, char *buf)
{
2247
	struct hstate *h = kobj_to_hstate(kobj, NULL);
2248 2249
	return sprintf(buf, "%lu\n", h->nr_overcommit_huge_pages);
}
2250

2251 2252 2253 2254 2255
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;
2256
	struct hstate *h = kobj_to_hstate(kobj, NULL);
2257

2258
	if (hstate_is_gigantic(h))
2259 2260
		return -EINVAL;

2261
	err = kstrtoul(buf, 10, &input);
2262
	if (err)
2263
		return err;
2264 2265 2266 2267 2268 2269 2270 2271 2272 2273 2274 2275

	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)
{
2276 2277 2278 2279 2280 2281 2282 2283 2284 2285 2286
	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);
2287 2288 2289 2290 2291 2292
}
HSTATE_ATTR_RO(free_hugepages);

static ssize_t resv_hugepages_show(struct kobject *kobj,
					struct kobj_attribute *attr, char *buf)
{
2293
	struct hstate *h = kobj_to_hstate(kobj, NULL);
2294 2295 2296 2297 2298 2299 2300
	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)
{
2301 2302 2303 2304 2305 2306 2307 2308 2309 2310 2311
	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);
2312 2313 2314 2315 2316 2317 2318 2319 2320
}
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,
2321 2322 2323
#ifdef CONFIG_NUMA
	&nr_hugepages_mempolicy_attr.attr,
#endif
2324 2325 2326 2327 2328 2329 2330
	NULL,
};

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

J
Jeff Mahoney 已提交
2331 2332 2333
static int hugetlb_sysfs_add_hstate(struct hstate *h, struct kobject *parent,
				    struct kobject **hstate_kobjs,
				    struct attribute_group *hstate_attr_group)
2334 2335
{
	int retval;
2336
	int hi = hstate_index(h);
2337

2338 2339
	hstate_kobjs[hi] = kobject_create_and_add(h->name, parent);
	if (!hstate_kobjs[hi])
2340 2341
		return -ENOMEM;

2342
	retval = sysfs_create_group(hstate_kobjs[hi], hstate_attr_group);
2343
	if (retval)
2344
		kobject_put(hstate_kobjs[hi]);
2345 2346 2347 2348 2349 2350 2351 2352 2353 2354 2355 2356 2357 2358

	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) {
2359 2360
		err = hugetlb_sysfs_add_hstate(h, hugepages_kobj,
					 hstate_kobjs, &hstate_attr_group);
2361
		if (err)
2362
			pr_err("Hugetlb: Unable to add hstate %s", h->name);
2363 2364 2365
	}
}

2366 2367 2368 2369
#ifdef CONFIG_NUMA

/*
 * node_hstate/s - associate per node hstate attributes, via their kobjects,
2370 2371 2372
 * 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
2373 2374 2375 2376 2377 2378 2379 2380 2381
 * 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];

/*
2382
 * A subset of global hstate attributes for node devices
2383 2384 2385 2386 2387 2388 2389 2390 2391 2392 2393 2394 2395
 */
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,
};

/*
2396
 * kobj_to_node_hstate - lookup global hstate for node device hstate attr kobj.
2397 2398 2399 2400 2401 2402 2403 2404 2405 2406 2407 2408 2409 2410 2411 2412 2413 2414 2415 2416 2417 2418
 * 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;
}

/*
2419
 * Unregister hstate attributes from a single node device.
2420 2421
 * No-op if no hstate attributes attached.
 */
2422
static void hugetlb_unregister_node(struct node *node)
2423 2424
{
	struct hstate *h;
2425
	struct node_hstate *nhs = &node_hstates[node->dev.id];
2426 2427

	if (!nhs->hugepages_kobj)
2428
		return;		/* no hstate attributes */
2429

2430 2431 2432 2433 2434
	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;
2435
		}
2436
	}
2437 2438 2439 2440 2441 2442

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

/*
2443
 * hugetlb module exit:  unregister hstate attributes from node devices
2444 2445 2446 2447 2448 2449 2450
 * that have them.
 */
static void hugetlb_unregister_all_nodes(void)
{
	int nid;

	/*
2451
	 * disable node device registrations.
2452 2453 2454 2455 2456 2457 2458
	 */
	register_hugetlbfs_with_node(NULL, NULL);

	/*
	 * remove hstate attributes from any nodes that have them.
	 */
	for (nid = 0; nid < nr_node_ids; nid++)
2459
		hugetlb_unregister_node(node_devices[nid]);
2460 2461 2462
}

/*
2463
 * Register hstate attributes for a single node device.
2464 2465
 * No-op if attributes already registered.
 */
2466
static void hugetlb_register_node(struct node *node)
2467 2468
{
	struct hstate *h;
2469
	struct node_hstate *nhs = &node_hstates[node->dev.id];
2470 2471 2472 2473 2474 2475
	int err;

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

	nhs->hugepages_kobj = kobject_create_and_add("hugepages",
2476
							&node->dev.kobj);
2477 2478 2479 2480 2481 2482 2483 2484
	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) {
2485 2486
			pr_err("Hugetlb: Unable to add hstate %s for node %d\n",
				h->name, node->dev.id);
2487 2488 2489 2490 2491 2492 2493
			hugetlb_unregister_node(node);
			break;
		}
	}
}

/*
2494
 * hugetlb init time:  register hstate attributes for all registered node
2495 2496
 * devices of nodes that have memory.  All on-line nodes should have
 * registered their associated device by this time.
2497
 */
2498
static void __init hugetlb_register_all_nodes(void)
2499 2500 2501
{
	int nid;

2502
	for_each_node_state(nid, N_MEMORY) {
2503
		struct node *node = node_devices[nid];
2504
		if (node->dev.id == nid)
2505 2506 2507 2508
			hugetlb_register_node(node);
	}

	/*
2509
	 * Let the node device driver know we're here so it can
2510 2511 2512 2513 2514 2515 2516 2517 2518 2519 2520 2521 2522 2523 2524 2525 2526 2527 2528 2529 2530
	 * [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

2531 2532 2533 2534
static void __exit hugetlb_exit(void)
{
	struct hstate *h;

2535 2536
	hugetlb_unregister_all_nodes();

2537
	for_each_hstate(h) {
2538
		kobject_put(hstate_kobjs[hstate_index(h)]);
2539 2540 2541
	}

	kobject_put(hugepages_kobj);
2542
	kfree(hugetlb_fault_mutex_table);
2543 2544 2545 2546 2547
}
module_exit(hugetlb_exit);

static int __init hugetlb_init(void)
{
2548 2549
	int i;

2550
	if (!hugepages_supported())
2551
		return 0;
2552

2553 2554 2555 2556
	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);
2557
	}
2558
	default_hstate_idx = hstate_index(size_to_hstate(default_hstate_size));
2559 2560
	if (default_hstate_max_huge_pages)
		default_hstate.max_huge_pages = default_hstate_max_huge_pages;
2561 2562

	hugetlb_init_hstates();
2563
	gather_bootmem_prealloc();
2564 2565 2566
	report_hugepages();

	hugetlb_sysfs_init();
2567
	hugetlb_register_all_nodes();
2568
	hugetlb_cgroup_file_init();
2569

2570 2571 2572 2573 2574
#ifdef CONFIG_SMP
	num_fault_mutexes = roundup_pow_of_two(8 * num_possible_cpus());
#else
	num_fault_mutexes = 1;
#endif
2575
	hugetlb_fault_mutex_table =
2576
		kmalloc(sizeof(struct mutex) * num_fault_mutexes, GFP_KERNEL);
2577
	BUG_ON(!hugetlb_fault_mutex_table);
2578 2579

	for (i = 0; i < num_fault_mutexes; i++)
2580
		mutex_init(&hugetlb_fault_mutex_table[i]);
2581 2582 2583 2584 2585 2586 2587 2588
	return 0;
}
module_init(hugetlb_init);

/* Should be called on processing a hugepagesz=... option */
void __init hugetlb_add_hstate(unsigned order)
{
	struct hstate *h;
2589 2590
	unsigned long i;

2591
	if (size_to_hstate(PAGE_SIZE << order)) {
2592
		pr_warning("hugepagesz= specified twice, ignoring\n");
2593 2594
		return;
	}
2595
	BUG_ON(hugetlb_max_hstate >= HUGE_MAX_HSTATE);
2596
	BUG_ON(order == 0);
2597
	h = &hstates[hugetlb_max_hstate++];
2598 2599
	h->order = order;
	h->mask = ~((1ULL << (order + PAGE_SHIFT)) - 1);
2600 2601 2602 2603
	h->nr_huge_pages = 0;
	h->free_huge_pages = 0;
	for (i = 0; i < MAX_NUMNODES; ++i)
		INIT_LIST_HEAD(&h->hugepage_freelists[i]);
2604
	INIT_LIST_HEAD(&h->hugepage_activelist);
2605 2606
	h->next_nid_to_alloc = first_node(node_states[N_MEMORY]);
	h->next_nid_to_free = first_node(node_states[N_MEMORY]);
2607 2608
	snprintf(h->name, HSTATE_NAME_LEN, "hugepages-%lukB",
					huge_page_size(h)/1024);
2609

2610 2611 2612
	parsed_hstate = h;
}

2613
static int __init hugetlb_nrpages_setup(char *s)
2614 2615
{
	unsigned long *mhp;
2616
	static unsigned long *last_mhp;
2617 2618

	/*
2619
	 * !hugetlb_max_hstate means we haven't parsed a hugepagesz= parameter yet,
2620 2621
	 * so this hugepages= parameter goes to the "default hstate".
	 */
2622
	if (!hugetlb_max_hstate)
2623 2624 2625 2626
		mhp = &default_hstate_max_huge_pages;
	else
		mhp = &parsed_hstate->max_huge_pages;

2627
	if (mhp == last_mhp) {
2628 2629
		pr_warning("hugepages= specified twice without "
			   "interleaving hugepagesz=, ignoring\n");
2630 2631 2632
		return 1;
	}

2633 2634 2635
	if (sscanf(s, "%lu", mhp) <= 0)
		*mhp = 0;

2636 2637 2638 2639 2640
	/*
	 * 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.
	 */
2641
	if (hugetlb_max_hstate && parsed_hstate->order >= MAX_ORDER)
2642 2643 2644 2645
		hugetlb_hstate_alloc_pages(parsed_hstate);

	last_mhp = mhp;

2646 2647
	return 1;
}
2648 2649 2650 2651 2652 2653 2654 2655
__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);
2656

2657 2658 2659 2660 2661 2662 2663 2664 2665 2666 2667 2668
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
2669 2670 2671
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
2672
{
2673
	struct hstate *h = &default_hstate;
2674
	unsigned long tmp = h->max_huge_pages;
2675
	int ret;
2676

2677 2678 2679
	if (!hugepages_supported())
		return -ENOTSUPP;

2680 2681
	table->data = &tmp;
	table->maxlen = sizeof(unsigned long);
2682 2683 2684
	ret = proc_doulongvec_minmax(table, write, buffer, length, ppos);
	if (ret)
		goto out;
2685

2686 2687 2688
	if (write)
		ret = __nr_hugepages_store_common(obey_mempolicy, h,
						  NUMA_NO_NODE, tmp, *length);
2689 2690
out:
	return ret;
L
Linus Torvalds 已提交
2691
}
2692

2693 2694 2695 2696 2697 2698 2699 2700 2701 2702 2703 2704 2705 2706 2707 2708 2709
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 */

2710
int hugetlb_overcommit_handler(struct ctl_table *table, int write,
2711
			void __user *buffer,
2712 2713
			size_t *length, loff_t *ppos)
{
2714
	struct hstate *h = &default_hstate;
2715
	unsigned long tmp;
2716
	int ret;
2717

2718 2719 2720
	if (!hugepages_supported())
		return -ENOTSUPP;

2721
	tmp = h->nr_overcommit_huge_pages;
2722

2723
	if (write && hstate_is_gigantic(h))
2724 2725
		return -EINVAL;

2726 2727
	table->data = &tmp;
	table->maxlen = sizeof(unsigned long);
2728 2729 2730
	ret = proc_doulongvec_minmax(table, write, buffer, length, ppos);
	if (ret)
		goto out;
2731 2732 2733 2734 2735 2736

	if (write) {
		spin_lock(&hugetlb_lock);
		h->nr_overcommit_huge_pages = tmp;
		spin_unlock(&hugetlb_lock);
	}
2737 2738
out:
	return ret;
2739 2740
}

L
Linus Torvalds 已提交
2741 2742
#endif /* CONFIG_SYSCTL */

2743
void hugetlb_report_meminfo(struct seq_file *m)
L
Linus Torvalds 已提交
2744
{
2745
	struct hstate *h = &default_hstate;
2746 2747
	if (!hugepages_supported())
		return;
2748
	seq_printf(m,
2749 2750 2751 2752 2753
			"HugePages_Total:   %5lu\n"
			"HugePages_Free:    %5lu\n"
			"HugePages_Rsvd:    %5lu\n"
			"HugePages_Surp:    %5lu\n"
			"Hugepagesize:   %8lu kB\n",
2754 2755 2756 2757 2758
			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
2759 2760 2761 2762
}

int hugetlb_report_node_meminfo(int nid, char *buf)
{
2763
	struct hstate *h = &default_hstate;
2764 2765
	if (!hugepages_supported())
		return 0;
L
Linus Torvalds 已提交
2766 2767
	return sprintf(buf,
		"Node %d HugePages_Total: %5u\n"
2768 2769
		"Node %d HugePages_Free:  %5u\n"
		"Node %d HugePages_Surp:  %5u\n",
2770 2771 2772
		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
2773 2774
}

2775 2776 2777 2778 2779
void hugetlb_show_meminfo(void)
{
	struct hstate *h;
	int nid;

2780 2781 2782
	if (!hugepages_supported())
		return;

2783 2784 2785 2786 2787 2788 2789 2790 2791 2792
	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
2793 2794 2795
/* Return the number pages of memory we physically have, in PAGE_SIZE units. */
unsigned long hugetlb_total_pages(void)
{
2796 2797 2798 2799 2800 2801
	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
2802 2803
}

2804
static int hugetlb_acct_memory(struct hstate *h, long delta)
M
Mel Gorman 已提交
2805 2806 2807 2808 2809 2810 2811 2812 2813 2814 2815 2816 2817 2818 2819 2820 2821 2822 2823 2824 2825 2826
{
	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) {
2827
		if (gather_surplus_pages(h, delta) < 0)
M
Mel Gorman 已提交
2828 2829
			goto out;

2830 2831
		if (delta > cpuset_mems_nr(h->free_huge_pages_node)) {
			return_unused_surplus_pages(h, delta);
M
Mel Gorman 已提交
2832 2833 2834 2835 2836 2837
			goto out;
		}
	}

	ret = 0;
	if (delta < 0)
2838
		return_unused_surplus_pages(h, (unsigned long) -delta);
M
Mel Gorman 已提交
2839 2840 2841 2842 2843 2844

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

2845 2846
static void hugetlb_vm_op_open(struct vm_area_struct *vma)
{
2847
	struct resv_map *resv = vma_resv_map(vma);
2848 2849 2850 2851 2852

	/*
	 * 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
2853
	 * has a reference to the reservation map it cannot disappear until
2854 2855 2856
	 * after this open call completes.  It is therefore safe to take a
	 * new reference here without additional locking.
	 */
2857
	if (resv && is_vma_resv_set(vma, HPAGE_RESV_OWNER))
2858
		kref_get(&resv->refs);
2859 2860
}

2861 2862
static void hugetlb_vm_op_close(struct vm_area_struct *vma)
{
2863
	struct hstate *h = hstate_vma(vma);
2864
	struct resv_map *resv = vma_resv_map(vma);
2865
	struct hugepage_subpool *spool = subpool_vma(vma);
2866
	unsigned long reserve, start, end;
2867
	long gbl_reserve;
2868

2869 2870
	if (!resv || !is_vma_resv_set(vma, HPAGE_RESV_OWNER))
		return;
2871

2872 2873
	start = vma_hugecache_offset(h, vma, vma->vm_start);
	end = vma_hugecache_offset(h, vma, vma->vm_end);
2874

2875
	reserve = (end - start) - region_count(resv, start, end);
2876

2877 2878 2879
	kref_put(&resv->refs, resv_map_release);

	if (reserve) {
2880 2881 2882 2883 2884 2885
		/*
		 * 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);
2886
	}
2887 2888
}

L
Linus Torvalds 已提交
2889 2890 2891 2892 2893 2894
/*
 * 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
2895
static int hugetlb_vm_op_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
L
Linus Torvalds 已提交
2896 2897
{
	BUG();
N
Nick Piggin 已提交
2898
	return 0;
L
Linus Torvalds 已提交
2899 2900
}

2901
const struct vm_operations_struct hugetlb_vm_ops = {
N
Nick Piggin 已提交
2902
	.fault = hugetlb_vm_op_fault,
2903
	.open = hugetlb_vm_op_open,
2904
	.close = hugetlb_vm_op_close,
L
Linus Torvalds 已提交
2905 2906
};

2907 2908
static pte_t make_huge_pte(struct vm_area_struct *vma, struct page *page,
				int writable)
D
David Gibson 已提交
2909 2910 2911
{
	pte_t entry;

2912
	if (writable) {
2913 2914
		entry = huge_pte_mkwrite(huge_pte_mkdirty(mk_huge_pte(page,
					 vma->vm_page_prot)));
D
David Gibson 已提交
2915
	} else {
2916 2917
		entry = huge_pte_wrprotect(mk_huge_pte(page,
					   vma->vm_page_prot));
D
David Gibson 已提交
2918 2919 2920
	}
	entry = pte_mkyoung(entry);
	entry = pte_mkhuge(entry);
2921
	entry = arch_make_huge_pte(entry, vma, page, writable);
D
David Gibson 已提交
2922 2923 2924 2925

	return entry;
}

2926 2927 2928 2929 2930
static void set_huge_ptep_writable(struct vm_area_struct *vma,
				   unsigned long address, pte_t *ptep)
{
	pte_t entry;

2931
	entry = huge_pte_mkwrite(huge_pte_mkdirty(huge_ptep_get(ptep)));
2932
	if (huge_ptep_set_access_flags(vma, address, ptep, entry, 1))
2933
		update_mmu_cache(vma, address, ptep);
2934 2935
}

2936 2937 2938 2939 2940 2941 2942 2943 2944 2945 2946 2947 2948 2949 2950 2951 2952 2953 2954 2955 2956 2957 2958 2959 2960
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;
}
2961

D
David Gibson 已提交
2962 2963 2964 2965 2966
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;
2967
	unsigned long addr;
2968
	int cow;
2969 2970
	struct hstate *h = hstate_vma(vma);
	unsigned long sz = huge_page_size(h);
2971 2972 2973
	unsigned long mmun_start;	/* For mmu_notifiers */
	unsigned long mmun_end;		/* For mmu_notifiers */
	int ret = 0;
2974 2975

	cow = (vma->vm_flags & (VM_SHARED | VM_MAYWRITE)) == VM_MAYWRITE;
D
David Gibson 已提交
2976

2977 2978 2979 2980 2981
	mmun_start = vma->vm_start;
	mmun_end = vma->vm_end;
	if (cow)
		mmu_notifier_invalidate_range_start(src, mmun_start, mmun_end);

2982
	for (addr = vma->vm_start; addr < vma->vm_end; addr += sz) {
2983
		spinlock_t *src_ptl, *dst_ptl;
H
Hugh Dickins 已提交
2984 2985 2986
		src_pte = huge_pte_offset(src, addr);
		if (!src_pte)
			continue;
2987
		dst_pte = huge_pte_alloc(dst, addr, sz);
2988 2989 2990 2991
		if (!dst_pte) {
			ret = -ENOMEM;
			break;
		}
2992 2993 2994 2995 2996

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

2997 2998 2999
		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);
3000 3001 3002 3003 3004 3005 3006 3007 3008 3009 3010 3011 3012 3013 3014 3015 3016 3017
		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 {
3018
			if (cow) {
3019
				huge_ptep_set_wrprotect(src, addr, src_pte);
3020 3021 3022
				mmu_notifier_invalidate_range(src, mmun_start,
								   mmun_end);
			}
3023
			entry = huge_ptep_get(src_pte);
3024 3025
			ptepage = pte_page(entry);
			get_page(ptepage);
3026
			page_dup_rmap(ptepage);
3027 3028
			set_huge_pte_at(dst, addr, dst_pte, entry);
		}
3029 3030
		spin_unlock(src_ptl);
		spin_unlock(dst_ptl);
D
David Gibson 已提交
3031 3032
	}

3033 3034 3035 3036
	if (cow)
		mmu_notifier_invalidate_range_end(src, mmun_start, mmun_end);

	return ret;
D
David Gibson 已提交
3037 3038
}

3039 3040 3041
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
3042
{
3043
	int force_flush = 0;
D
David Gibson 已提交
3044 3045
	struct mm_struct *mm = vma->vm_mm;
	unsigned long address;
3046
	pte_t *ptep;
D
David Gibson 已提交
3047
	pte_t pte;
3048
	spinlock_t *ptl;
D
David Gibson 已提交
3049
	struct page *page;
3050 3051
	struct hstate *h = hstate_vma(vma);
	unsigned long sz = huge_page_size(h);
3052 3053
	const unsigned long mmun_start = start;	/* For mmu_notifiers */
	const unsigned long mmun_end   = end;	/* For mmu_notifiers */
3054

D
David Gibson 已提交
3055
	WARN_ON(!is_vm_hugetlb_page(vma));
3056 3057
	BUG_ON(start & ~huge_page_mask(h));
	BUG_ON(end & ~huge_page_mask(h));
D
David Gibson 已提交
3058

3059
	tlb_start_vma(tlb, vma);
3060
	mmu_notifier_invalidate_range_start(mm, mmun_start, mmun_end);
3061
	address = start;
3062
again:
3063
	for (; address < end; address += sz) {
3064
		ptep = huge_pte_offset(mm, address);
A
Adam Litke 已提交
3065
		if (!ptep)
3066 3067
			continue;

3068
		ptl = huge_pte_lock(h, mm, ptep);
3069
		if (huge_pmd_unshare(mm, &address, ptep))
3070
			goto unlock;
3071

3072 3073
		pte = huge_ptep_get(ptep);
		if (huge_pte_none(pte))
3074
			goto unlock;
3075 3076

		/*
3077 3078
		 * Migrating hugepage or HWPoisoned hugepage is already
		 * unmapped and its refcount is dropped, so just clear pte here.
3079
		 */
3080
		if (unlikely(!pte_present(pte))) {
3081
			huge_pte_clear(mm, address, ptep);
3082
			goto unlock;
3083
		}
3084 3085

		page = pte_page(pte);
3086 3087 3088 3089 3090 3091 3092
		/*
		 * 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)
3093
				goto unlock;
3094 3095 3096 3097 3098 3099 3100 3101 3102

			/*
			 * 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);
		}

3103
		pte = huge_ptep_get_and_clear(mm, address, ptep);
3104
		tlb_remove_tlb_entry(tlb, ptep, address);
3105
		if (huge_pte_dirty(pte))
3106
			set_page_dirty(page);
3107

3108 3109
		page_remove_rmap(page);
		force_flush = !__tlb_remove_page(tlb, page);
3110
		if (force_flush) {
3111
			address += sz;
3112
			spin_unlock(ptl);
3113
			break;
3114
		}
3115
		/* Bail out after unmapping reference page if supplied */
3116 3117
		if (ref_page) {
			spin_unlock(ptl);
3118
			break;
3119 3120 3121
		}
unlock:
		spin_unlock(ptl);
D
David Gibson 已提交
3122
	}
3123 3124 3125 3126 3127 3128 3129 3130 3131 3132
	/*
	 * 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;
3133
	}
3134
	mmu_notifier_invalidate_range_end(mm, mmun_start, mmun_end);
3135
	tlb_end_vma(tlb, vma);
L
Linus Torvalds 已提交
3136
}
D
David Gibson 已提交
3137

3138 3139 3140 3141 3142 3143 3144 3145 3146 3147 3148 3149
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
3150
	 * is to clear it before releasing the i_mmap_rwsem. This works
3151
	 * because in the context this is called, the VMA is about to be
3152
	 * destroyed and the i_mmap_rwsem is held.
3153 3154 3155 3156
	 */
	vma->vm_flags &= ~VM_MAYSHARE;
}

3157
void unmap_hugepage_range(struct vm_area_struct *vma, unsigned long start,
3158
			  unsigned long end, struct page *ref_page)
3159
{
3160 3161 3162 3163 3164
	struct mm_struct *mm;
	struct mmu_gather tlb;

	mm = vma->vm_mm;

3165
	tlb_gather_mmu(&tlb, mm, start, end);
3166 3167
	__unmap_hugepage_range(&tlb, vma, start, end, ref_page);
	tlb_finish_mmu(&tlb, start, end);
3168 3169
}

3170 3171 3172 3173 3174 3175
/*
 * 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.
 */
3176 3177
static void unmap_ref_private(struct mm_struct *mm, struct vm_area_struct *vma,
			      struct page *page, unsigned long address)
3178
{
3179
	struct hstate *h = hstate_vma(vma);
3180 3181 3182 3183 3184 3185 3186 3187
	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.
	 */
3188
	address = address & huge_page_mask(h);
3189 3190
	pgoff = ((address - vma->vm_start) >> PAGE_SHIFT) +
			vma->vm_pgoff;
A
Al Viro 已提交
3191
	mapping = file_inode(vma->vm_file)->i_mapping;
3192

3193 3194 3195 3196 3197
	/*
	 * 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
	 */
3198
	i_mmap_lock_write(mapping);
3199
	vma_interval_tree_foreach(iter_vma, &mapping->i_mmap, pgoff, pgoff) {
3200 3201 3202 3203
		/* Do not unmap the current VMA */
		if (iter_vma == vma)
			continue;

3204 3205 3206 3207 3208 3209 3210 3211
		/*
		 * Shared VMAs have their own reserves and do not affect
		 * MAP_PRIVATE accounting but it is possible that a shared
		 * VMA is using the same page so check and skip such VMAs.
		 */
		if (iter_vma->vm_flags & VM_MAYSHARE)
			continue;

3212 3213 3214 3215 3216 3217 3218 3219
		/*
		 * 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))
3220 3221
			unmap_hugepage_range(iter_vma, address,
					     address + huge_page_size(h), page);
3222
	}
3223
	i_mmap_unlock_write(mapping);
3224 3225
}

3226 3227
/*
 * Hugetlb_cow() should be called with page lock of the original hugepage held.
3228 3229 3230
 * 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.
3231
 */
3232
static int hugetlb_cow(struct mm_struct *mm, struct vm_area_struct *vma,
3233
			unsigned long address, pte_t *ptep, pte_t pte,
3234
			struct page *pagecache_page, spinlock_t *ptl)
3235
{
3236
	struct hstate *h = hstate_vma(vma);
3237
	struct page *old_page, *new_page;
3238
	int ret = 0, outside_reserve = 0;
3239 3240
	unsigned long mmun_start;	/* For mmu_notifiers */
	unsigned long mmun_end;		/* For mmu_notifiers */
3241 3242 3243

	old_page = pte_page(pte);

3244
retry_avoidcopy:
3245 3246
	/* If no-one else is actually using this page, avoid the copy
	 * and just make the page writable */
3247 3248
	if (page_mapcount(old_page) == 1 && PageAnon(old_page)) {
		page_move_anon_rmap(old_page, vma, address);
3249
		set_huge_ptep_writable(vma, address, ptep);
N
Nick Piggin 已提交
3250
		return 0;
3251 3252
	}

3253 3254 3255 3256 3257 3258 3259 3260 3261
	/*
	 * 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.
	 */
3262
	if (is_vma_resv_set(vma, HPAGE_RESV_OWNER) &&
3263 3264 3265
			old_page != pagecache_page)
		outside_reserve = 1;

3266
	page_cache_get(old_page);
3267

3268 3269 3270 3271
	/*
	 * Drop page table lock as buddy allocator may be called. It will
	 * be acquired again before returning to the caller, as expected.
	 */
3272
	spin_unlock(ptl);
3273
	new_page = alloc_huge_page(vma, address, outside_reserve);
3274

3275
	if (IS_ERR(new_page)) {
3276 3277 3278 3279 3280 3281 3282 3283
		/*
		 * 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) {
3284
			page_cache_release(old_page);
3285
			BUG_ON(huge_pte_none(pte));
3286 3287 3288 3289 3290 3291 3292 3293 3294 3295 3296 3297
			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;
3298 3299
		}

3300 3301 3302
		ret = (PTR_ERR(new_page) == -ENOMEM) ?
			VM_FAULT_OOM : VM_FAULT_SIGBUS;
		goto out_release_old;
3303 3304
	}

3305 3306 3307 3308
	/*
	 * When the original hugepage is shared one, it does not have
	 * anon_vma prepared.
	 */
3309
	if (unlikely(anon_vma_prepare(vma))) {
3310 3311
		ret = VM_FAULT_OOM;
		goto out_release_all;
3312
	}
3313

A
Andrea Arcangeli 已提交
3314 3315
	copy_user_huge_page(new_page, old_page, address, vma,
			    pages_per_huge_page(h));
N
Nick Piggin 已提交
3316
	__SetPageUptodate(new_page);
3317
	set_page_huge_active(new_page);
3318

3319 3320 3321
	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);
3322

3323
	/*
3324
	 * Retake the page table lock to check for racing updates
3325 3326
	 * before the page tables are altered
	 */
3327
	spin_lock(ptl);
3328
	ptep = huge_pte_offset(mm, address & huge_page_mask(h));
3329
	if (likely(ptep && pte_same(huge_ptep_get(ptep), pte))) {
3330 3331
		ClearPagePrivate(new_page);

3332
		/* Break COW */
3333
		huge_ptep_clear_flush(vma, address, ptep);
3334
		mmu_notifier_invalidate_range(mm, mmun_start, mmun_end);
3335 3336
		set_huge_pte_at(mm, address, ptep,
				make_huge_pte(vma, new_page, 1));
3337
		page_remove_rmap(old_page);
3338
		hugepage_add_new_anon_rmap(new_page, vma, address);
3339 3340 3341
		/* Make the old page be freed below */
		new_page = old_page;
	}
3342
	spin_unlock(ptl);
3343
	mmu_notifier_invalidate_range_end(mm, mmun_start, mmun_end);
3344
out_release_all:
3345
	page_cache_release(new_page);
3346
out_release_old:
3347
	page_cache_release(old_page);
3348

3349 3350
	spin_lock(ptl); /* Caller expects lock to be held */
	return ret;
3351 3352
}

3353
/* Return the pagecache page at a given address within a VMA */
3354 3355
static struct page *hugetlbfs_pagecache_page(struct hstate *h,
			struct vm_area_struct *vma, unsigned long address)
3356 3357
{
	struct address_space *mapping;
3358
	pgoff_t idx;
3359 3360

	mapping = vma->vm_file->f_mapping;
3361
	idx = vma_hugecache_offset(h, vma, address);
3362 3363 3364 3365

	return find_lock_page(mapping, idx);
}

H
Hugh Dickins 已提交
3366 3367 3368 3369 3370
/*
 * 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
3371 3372 3373 3374 3375 3376 3377 3378 3379 3380 3381 3382 3383 3384 3385
			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;
}

3386 3387 3388 3389 3390 3391 3392 3393 3394 3395 3396 3397 3398 3399 3400 3401 3402
int huge_add_to_page_cache(struct page *page, struct address_space *mapping,
			   pgoff_t idx)
{
	struct inode *inode = mapping->host;
	struct hstate *h = hstate_inode(inode);
	int err = add_to_page_cache(page, mapping, idx, GFP_KERNEL);

	if (err)
		return err;
	ClearPagePrivate(page);

	spin_lock(&inode->i_lock);
	inode->i_blocks += blocks_per_huge_page(h);
	spin_unlock(&inode->i_lock);
	return 0;
}

3403
static int hugetlb_no_page(struct mm_struct *mm, struct vm_area_struct *vma,
3404 3405
			   struct address_space *mapping, pgoff_t idx,
			   unsigned long address, pte_t *ptep, unsigned int flags)
3406
{
3407
	struct hstate *h = hstate_vma(vma);
3408
	int ret = VM_FAULT_SIGBUS;
3409
	int anon_rmap = 0;
A
Adam Litke 已提交
3410 3411
	unsigned long size;
	struct page *page;
3412
	pte_t new_pte;
3413
	spinlock_t *ptl;
A
Adam Litke 已提交
3414

3415 3416 3417
	/*
	 * 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
3418
	 * COW. Warn that such a situation has occurred as it may not be obvious
3419 3420
	 */
	if (is_vma_resv_set(vma, HPAGE_RESV_UNMAPPED)) {
3421 3422
		pr_warning("PID %d killed due to inadequate hugepage pool\n",
			   current->pid);
3423 3424 3425
		return ret;
	}

A
Adam Litke 已提交
3426 3427 3428 3429
	/*
	 * Use page lock to guard against racing truncation
	 * before we get page_table_lock.
	 */
3430 3431 3432
retry:
	page = find_lock_page(mapping, idx);
	if (!page) {
3433
		size = i_size_read(mapping->host) >> huge_page_shift(h);
3434 3435
		if (idx >= size)
			goto out;
3436
		page = alloc_huge_page(vma, address, 0);
3437
		if (IS_ERR(page)) {
3438 3439 3440 3441 3442
			ret = PTR_ERR(page);
			if (ret == -ENOMEM)
				ret = VM_FAULT_OOM;
			else
				ret = VM_FAULT_SIGBUS;
3443 3444
			goto out;
		}
A
Andrea Arcangeli 已提交
3445
		clear_huge_page(page, address, pages_per_huge_page(h));
N
Nick Piggin 已提交
3446
		__SetPageUptodate(page);
3447
		set_page_huge_active(page);
3448

3449
		if (vma->vm_flags & VM_MAYSHARE) {
3450
			int err = huge_add_to_page_cache(page, mapping, idx);
3451 3452 3453 3454 3455 3456
			if (err) {
				put_page(page);
				if (err == -EEXIST)
					goto retry;
				goto out;
			}
3457
		} else {
3458
			lock_page(page);
3459 3460 3461 3462
			if (unlikely(anon_vma_prepare(vma))) {
				ret = VM_FAULT_OOM;
				goto backout_unlocked;
			}
3463
			anon_rmap = 1;
3464
		}
3465
	} else {
3466 3467 3468 3469 3470 3471
		/*
		 * 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))) {
3472
			ret = VM_FAULT_HWPOISON |
3473
				VM_FAULT_SET_HINDEX(hstate_index(h));
3474 3475
			goto backout_unlocked;
		}
3476
	}
3477

3478 3479 3480 3481 3482 3483
	/*
	 * 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.
	 */
3484
	if ((flags & FAULT_FLAG_WRITE) && !(vma->vm_flags & VM_SHARED)) {
3485 3486 3487 3488
		if (vma_needs_reservation(h, vma, address) < 0) {
			ret = VM_FAULT_OOM;
			goto backout_unlocked;
		}
3489
		/* Just decrements count, does not deallocate */
3490
		vma_end_reservation(h, vma, address);
3491
	}
3492

3493 3494
	ptl = huge_pte_lockptr(h, mm, ptep);
	spin_lock(ptl);
3495
	size = i_size_read(mapping->host) >> huge_page_shift(h);
A
Adam Litke 已提交
3496 3497 3498
	if (idx >= size)
		goto backout;

N
Nick Piggin 已提交
3499
	ret = 0;
3500
	if (!huge_pte_none(huge_ptep_get(ptep)))
A
Adam Litke 已提交
3501 3502
		goto backout;

3503 3504
	if (anon_rmap) {
		ClearPagePrivate(page);
3505
		hugepage_add_new_anon_rmap(page, vma, address);
3506
	} else
3507
		page_dup_rmap(page);
3508 3509 3510 3511
	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);

3512
	if ((flags & FAULT_FLAG_WRITE) && !(vma->vm_flags & VM_SHARED)) {
3513
		/* Optimization, do the COW without a second fault */
3514
		ret = hugetlb_cow(mm, vma, address, ptep, new_pte, page, ptl);
3515 3516
	}

3517
	spin_unlock(ptl);
A
Adam Litke 已提交
3518 3519
	unlock_page(page);
out:
3520
	return ret;
A
Adam Litke 已提交
3521 3522

backout:
3523
	spin_unlock(ptl);
3524
backout_unlocked:
A
Adam Litke 已提交
3525 3526 3527
	unlock_page(page);
	put_page(page);
	goto out;
3528 3529
}

3530
#ifdef CONFIG_SMP
3531
u32 hugetlb_fault_mutex_hash(struct hstate *h, struct mm_struct *mm,
3532 3533 3534 3535 3536 3537 3538 3539 3540 3541 3542 3543 3544 3545 3546 3547 3548 3549 3550 3551 3552 3553 3554 3555
			    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.
 */
3556
u32 hugetlb_fault_mutex_hash(struct hstate *h, struct mm_struct *mm,
3557 3558 3559 3560 3561 3562 3563 3564
			    struct vm_area_struct *vma,
			    struct address_space *mapping,
			    pgoff_t idx, unsigned long address)
{
	return 0;
}
#endif

3565
int hugetlb_fault(struct mm_struct *mm, struct vm_area_struct *vma,
3566
			unsigned long address, unsigned int flags)
3567
{
3568
	pte_t *ptep, entry;
3569
	spinlock_t *ptl;
3570
	int ret;
3571 3572
	u32 hash;
	pgoff_t idx;
3573
	struct page *page = NULL;
3574
	struct page *pagecache_page = NULL;
3575
	struct hstate *h = hstate_vma(vma);
3576
	struct address_space *mapping;
3577
	int need_wait_lock = 0;
3578

3579 3580
	address &= huge_page_mask(h);

3581 3582 3583
	ptep = huge_pte_offset(mm, address);
	if (ptep) {
		entry = huge_ptep_get(ptep);
N
Naoya Horiguchi 已提交
3584
		if (unlikely(is_hugetlb_entry_migration(entry))) {
3585
			migration_entry_wait_huge(vma, mm, ptep);
N
Naoya Horiguchi 已提交
3586 3587
			return 0;
		} else if (unlikely(is_hugetlb_entry_hwpoisoned(entry)))
3588
			return VM_FAULT_HWPOISON_LARGE |
3589
				VM_FAULT_SET_HINDEX(hstate_index(h));
3590 3591
	}

3592
	ptep = huge_pte_alloc(mm, address, huge_page_size(h));
3593 3594 3595
	if (!ptep)
		return VM_FAULT_OOM;

3596 3597 3598
	mapping = vma->vm_file->f_mapping;
	idx = vma_hugecache_offset(h, vma, address);

3599 3600 3601 3602 3603
	/*
	 * 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.
	 */
3604 3605
	hash = hugetlb_fault_mutex_hash(h, mm, vma, mapping, idx, address);
	mutex_lock(&hugetlb_fault_mutex_table[hash]);
3606

3607 3608
	entry = huge_ptep_get(ptep);
	if (huge_pte_none(entry)) {
3609
		ret = hugetlb_no_page(mm, vma, mapping, idx, address, ptep, flags);
3610
		goto out_mutex;
3611
	}
3612

N
Nick Piggin 已提交
3613
	ret = 0;
3614

3615 3616 3617 3618 3619 3620 3621 3622 3623 3624
	/*
	 * 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;

3625 3626 3627 3628 3629 3630 3631 3632
	/*
	 * 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.
	 */
3633
	if ((flags & FAULT_FLAG_WRITE) && !huge_pte_write(entry)) {
3634 3635
		if (vma_needs_reservation(h, vma, address) < 0) {
			ret = VM_FAULT_OOM;
3636
			goto out_mutex;
3637
		}
3638
		/* Just decrements count, does not deallocate */
3639
		vma_end_reservation(h, vma, address);
3640

3641
		if (!(vma->vm_flags & VM_MAYSHARE))
3642 3643 3644 3645
			pagecache_page = hugetlbfs_pagecache_page(h,
								vma, address);
	}

3646 3647 3648 3649 3650 3651
	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;

3652 3653 3654 3655 3656 3657 3658
	/*
	 * 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)
3659 3660 3661 3662
		if (!trylock_page(page)) {
			need_wait_lock = 1;
			goto out_ptl;
		}
3663

3664
	get_page(page);
3665

3666
	if (flags & FAULT_FLAG_WRITE) {
3667
		if (!huge_pte_write(entry)) {
3668
			ret = hugetlb_cow(mm, vma, address, ptep, entry,
3669
					pagecache_page, ptl);
3670
			goto out_put_page;
3671
		}
3672
		entry = huge_pte_mkdirty(entry);
3673 3674
	}
	entry = pte_mkyoung(entry);
3675 3676
	if (huge_ptep_set_access_flags(vma, address, ptep, entry,
						flags & FAULT_FLAG_WRITE))
3677
		update_mmu_cache(vma, address, ptep);
3678 3679 3680 3681
out_put_page:
	if (page != pagecache_page)
		unlock_page(page);
	put_page(page);
3682 3683
out_ptl:
	spin_unlock(ptl);
3684 3685 3686 3687 3688

	if (pagecache_page) {
		unlock_page(pagecache_page);
		put_page(pagecache_page);
	}
3689
out_mutex:
3690
	mutex_unlock(&hugetlb_fault_mutex_table[hash]);
3691 3692 3693 3694 3695 3696 3697 3698 3699
	/*
	 * 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);
3700
	return ret;
3701 3702
}

3703 3704 3705 3706
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
3707
{
3708 3709
	unsigned long pfn_offset;
	unsigned long vaddr = *position;
3710
	unsigned long remainder = *nr_pages;
3711
	struct hstate *h = hstate_vma(vma);
D
David Gibson 已提交
3712 3713

	while (vaddr < vma->vm_end && remainder) {
A
Adam Litke 已提交
3714
		pte_t *pte;
3715
		spinlock_t *ptl = NULL;
H
Hugh Dickins 已提交
3716
		int absent;
A
Adam Litke 已提交
3717
		struct page *page;
D
David Gibson 已提交
3718

3719 3720 3721 3722 3723 3724 3725 3726 3727
		/*
		 * 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
3728 3729
		/*
		 * Some archs (sparc64, sh*) have multiple pte_ts to
H
Hugh Dickins 已提交
3730
		 * each hugepage.  We have to make sure we get the
A
Adam Litke 已提交
3731
		 * first, for the page indexing below to work.
3732 3733
		 *
		 * Note that page table lock is not held when pte is null.
A
Adam Litke 已提交
3734
		 */
3735
		pte = huge_pte_offset(mm, vaddr & huge_page_mask(h));
3736 3737
		if (pte)
			ptl = huge_pte_lock(h, mm, pte);
H
Hugh Dickins 已提交
3738 3739 3740 3741
		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
3742 3743 3744 3745
		 * 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
3746
		 */
H
Hugh Dickins 已提交
3747 3748
		if (absent && (flags & FOLL_DUMP) &&
		    !hugetlbfs_pagecache_present(h, vma, vaddr)) {
3749 3750
			if (pte)
				spin_unlock(ptl);
H
Hugh Dickins 已提交
3751 3752 3753
			remainder = 0;
			break;
		}
D
David Gibson 已提交
3754

3755 3756 3757 3758 3759 3760 3761 3762 3763 3764 3765
		/*
		 * 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)) ||
3766 3767
		    ((flags & FOLL_WRITE) &&
		      !huge_pte_write(huge_ptep_get(pte)))) {
A
Adam Litke 已提交
3768
			int ret;
D
David Gibson 已提交
3769

3770 3771
			if (pte)
				spin_unlock(ptl);
H
Hugh Dickins 已提交
3772 3773
			ret = hugetlb_fault(mm, vma, vaddr,
				(flags & FOLL_WRITE) ? FAULT_FLAG_WRITE : 0);
3774
			if (!(ret & VM_FAULT_ERROR))
A
Adam Litke 已提交
3775
				continue;
D
David Gibson 已提交
3776

A
Adam Litke 已提交
3777 3778 3779 3780
			remainder = 0;
			break;
		}

3781
		pfn_offset = (vaddr & ~huge_page_mask(h)) >> PAGE_SHIFT;
3782
		page = pte_page(huge_ptep_get(pte));
3783
same_page:
3784
		if (pages) {
H
Hugh Dickins 已提交
3785
			pages[i] = mem_map_offset(page, pfn_offset);
3786
			get_page_foll(pages[i]);
3787
		}
D
David Gibson 已提交
3788 3789 3790 3791 3792

		if (vmas)
			vmas[i] = vma;

		vaddr += PAGE_SIZE;
3793
		++pfn_offset;
D
David Gibson 已提交
3794 3795
		--remainder;
		++i;
3796
		if (vaddr < vma->vm_end && remainder &&
3797
				pfn_offset < pages_per_huge_page(h)) {
3798 3799 3800 3801 3802 3803
			/*
			 * We use pfn_offset to avoid touching the pageframes
			 * of this compound page.
			 */
			goto same_page;
		}
3804
		spin_unlock(ptl);
D
David Gibson 已提交
3805
	}
3806
	*nr_pages = remainder;
D
David Gibson 已提交
3807 3808
	*position = vaddr;

H
Hugh Dickins 已提交
3809
	return i ? i : -EFAULT;
D
David Gibson 已提交
3810
}
3811

3812
unsigned long hugetlb_change_protection(struct vm_area_struct *vma,
3813 3814 3815 3816 3817 3818
		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;
3819
	struct hstate *h = hstate_vma(vma);
3820
	unsigned long pages = 0;
3821 3822 3823 3824

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

3825
	mmu_notifier_invalidate_range_start(mm, start, end);
3826
	i_mmap_lock_write(vma->vm_file->f_mapping);
3827
	for (; address < end; address += huge_page_size(h)) {
3828
		spinlock_t *ptl;
3829 3830 3831
		ptep = huge_pte_offset(mm, address);
		if (!ptep)
			continue;
3832
		ptl = huge_pte_lock(h, mm, ptep);
3833 3834
		if (huge_pmd_unshare(mm, &address, ptep)) {
			pages++;
3835
			spin_unlock(ptl);
3836
			continue;
3837
		}
3838 3839 3840 3841 3842 3843 3844 3845 3846 3847 3848 3849 3850 3851 3852 3853 3854 3855 3856 3857
		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)) {
3858
			pte = huge_ptep_get_and_clear(mm, address, ptep);
3859
			pte = pte_mkhuge(huge_pte_modify(pte, newprot));
3860
			pte = arch_make_huge_pte(pte, vma, NULL, 0);
3861
			set_huge_pte_at(mm, address, ptep, pte);
3862
			pages++;
3863
		}
3864
		spin_unlock(ptl);
3865
	}
3866
	/*
3867
	 * Must flush TLB before releasing i_mmap_rwsem: x86's huge_pmd_unshare
3868
	 * may have cleared our pud entry and done put_page on the page table:
3869
	 * once we release i_mmap_rwsem, another task can do the final put_page
3870 3871
	 * and that page table be reused and filled with junk.
	 */
3872
	flush_tlb_range(vma, start, end);
3873
	mmu_notifier_invalidate_range(mm, start, end);
3874
	i_mmap_unlock_write(vma->vm_file->f_mapping);
3875
	mmu_notifier_invalidate_range_end(mm, start, end);
3876 3877

	return pages << h->order;
3878 3879
}

3880 3881
int hugetlb_reserve_pages(struct inode *inode,
					long from, long to,
3882
					struct vm_area_struct *vma,
3883
					vm_flags_t vm_flags)
3884
{
3885
	long ret, chg;
3886
	struct hstate *h = hstate_inode(inode);
3887
	struct hugepage_subpool *spool = subpool_inode(inode);
3888
	struct resv_map *resv_map;
3889
	long gbl_reserve;
3890

3891 3892 3893
	/*
	 * Only apply hugepage reservation if asked. At fault time, an
	 * attempt will be made for VM_NORESERVE to allocate a page
3894
	 * without using reserves
3895
	 */
3896
	if (vm_flags & VM_NORESERVE)
3897 3898
		return 0;

3899 3900 3901 3902 3903 3904
	/*
	 * 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
	 */
3905
	if (!vma || vma->vm_flags & VM_MAYSHARE) {
3906
		resv_map = inode_resv_map(inode);
3907

3908
		chg = region_chg(resv_map, from, to);
3909 3910 3911

	} else {
		resv_map = resv_map_alloc();
3912 3913 3914
		if (!resv_map)
			return -ENOMEM;

3915
		chg = to - from;
3916

3917 3918 3919 3920
		set_vma_resv_map(vma, resv_map);
		set_vma_resv_flags(vma, HPAGE_RESV_OWNER);
	}

3921 3922 3923 3924
	if (chg < 0) {
		ret = chg;
		goto out_err;
	}
3925

3926 3927 3928 3929 3930 3931 3932
	/*
	 * 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) {
3933 3934 3935
		ret = -ENOSPC;
		goto out_err;
	}
3936 3937

	/*
3938
	 * Check enough hugepages are available for the reservation.
3939
	 * Hand the pages back to the subpool if there are not
3940
	 */
3941
	ret = hugetlb_acct_memory(h, gbl_reserve);
K
Ken Chen 已提交
3942
	if (ret < 0) {
3943 3944
		/* put back original number of pages, chg */
		(void)hugepage_subpool_put_pages(spool, chg);
3945
		goto out_err;
K
Ken Chen 已提交
3946
	}
3947 3948 3949 3950 3951 3952 3953 3954 3955 3956 3957 3958

	/*
	 * 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
	 */
3959 3960 3961 3962 3963 3964 3965 3966 3967 3968 3969 3970 3971 3972 3973 3974 3975 3976
	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);
		}
	}
3977
	return 0;
3978
out_err:
3979 3980
	if (!vma || vma->vm_flags & VM_MAYSHARE)
		region_abort(resv_map, from, to);
J
Joonsoo Kim 已提交
3981 3982
	if (vma && is_vma_resv_set(vma, HPAGE_RESV_OWNER))
		kref_put(&resv_map->refs, resv_map_release);
3983
	return ret;
3984 3985
}

3986 3987
long hugetlb_unreserve_pages(struct inode *inode, long start, long end,
								long freed)
3988
{
3989
	struct hstate *h = hstate_inode(inode);
3990
	struct resv_map *resv_map = inode_resv_map(inode);
3991
	long chg = 0;
3992
	struct hugepage_subpool *spool = subpool_inode(inode);
3993
	long gbl_reserve;
K
Ken Chen 已提交
3994

3995 3996 3997 3998 3999 4000 4001 4002 4003 4004 4005
	if (resv_map) {
		chg = region_del(resv_map, start, end);
		/*
		 * region_del() can fail in the rare case where a region
		 * must be split and another region descriptor can not be
		 * allocated.  If end == LONG_MAX, it will not fail.
		 */
		if (chg < 0)
			return chg;
	}

K
Ken Chen 已提交
4006
	spin_lock(&inode->i_lock);
4007
	inode->i_blocks -= (blocks_per_huge_page(h) * freed);
K
Ken Chen 已提交
4008 4009
	spin_unlock(&inode->i_lock);

4010 4011 4012 4013 4014 4015
	/*
	 * 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);
4016 4017

	return 0;
4018
}
4019

4020 4021 4022 4023 4024 4025 4026 4027 4028 4029 4030 4031 4032 4033 4034 4035 4036 4037 4038 4039 4040 4041 4042 4043 4044 4045
#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;
}

4046
static bool vma_shareable(struct vm_area_struct *vma, unsigned long addr)
4047 4048 4049 4050 4051 4052 4053 4054 4055
{
	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)
4056 4057
		return true;
	return false;
4058 4059 4060 4061 4062 4063 4064
}

/*
 * 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
4065
 * pud has to be populated inside the same i_mmap_rwsem section - otherwise
4066 4067 4068 4069 4070 4071 4072 4073 4074 4075 4076 4077 4078
 * 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;
4079
	spinlock_t *ptl;
4080 4081 4082 4083

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

4084
	i_mmap_lock_write(mapping);
4085 4086 4087 4088 4089 4090 4091 4092
	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) {
4093
				mm_inc_nr_pmds(mm);
4094 4095 4096 4097 4098 4099 4100 4101 4102
				get_page(virt_to_page(spte));
				break;
			}
		}
	}

	if (!spte)
		goto out;

4103 4104
	ptl = huge_pte_lockptr(hstate_vma(vma), mm, spte);
	spin_lock(ptl);
4105
	if (pud_none(*pud)) {
4106 4107
		pud_populate(mm, pud,
				(pmd_t *)((unsigned long)spte & PAGE_MASK));
4108
	} else {
4109
		put_page(virt_to_page(spte));
4110 4111
		mm_inc_nr_pmds(mm);
	}
4112
	spin_unlock(ptl);
4113 4114
out:
	pte = (pte_t *)pmd_alloc(mm, pud, addr);
4115
	i_mmap_unlock_write(mapping);
4116 4117 4118 4119 4120 4121 4122 4123 4124 4125
	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.
 *
4126
 * called with page table lock held.
4127 4128 4129 4130 4131 4132 4133 4134 4135 4136 4137 4138 4139 4140 4141
 *
 * 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));
4142
	mm_dec_nr_pmds(mm);
4143 4144 4145
	*addr = ALIGN(*addr, HPAGE_SIZE * PTRS_PER_PTE) - HPAGE_SIZE;
	return 1;
}
4146 4147 4148 4149 4150 4151
#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;
}
4152 4153 4154 4155 4156

int huge_pmd_unshare(struct mm_struct *mm, unsigned long *addr, pte_t *ptep)
{
	return 0;
}
4157
#define want_pmd_share()	(0)
4158 4159
#endif /* CONFIG_ARCH_WANT_HUGE_PMD_SHARE */

4160 4161 4162 4163 4164 4165 4166 4167 4168 4169 4170 4171 4172 4173 4174 4175 4176 4177 4178 4179 4180 4181 4182 4183 4184 4185 4186 4187 4188 4189 4190 4191 4192 4193 4194 4195 4196 4197 4198 4199 4200 4201 4202 4203
#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;
}

4204 4205 4206 4207 4208 4209 4210 4211 4212 4213 4214 4215 4216 4217
#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
4218
follow_huge_pmd(struct mm_struct *mm, unsigned long address,
4219
		pmd_t *pmd, int flags)
4220
{
4221 4222 4223 4224 4225 4226 4227 4228 4229 4230 4231 4232
	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)) {
4233
		page = pmd_page(*pmd) + ((address & ~PMD_MASK) >> PAGE_SHIFT);
4234 4235 4236 4237 4238 4239 4240 4241 4242 4243 4244 4245 4246 4247 4248
		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);
4249 4250 4251
	return page;
}

4252
struct page * __weak
4253
follow_huge_pud(struct mm_struct *mm, unsigned long address,
4254
		pud_t *pud, int flags)
4255
{
4256 4257
	if (flags & FOLL_GET)
		return NULL;
4258

4259
	return pte_page(*(pte_t *)pud) + ((address & ~PUD_MASK) >> PAGE_SHIFT);
4260 4261
}

4262 4263
#ifdef CONFIG_MEMORY_FAILURE

4264 4265 4266 4267
/*
 * This function is called from memory failure code.
 * Assume the caller holds page lock of the head page.
 */
4268
int dequeue_hwpoisoned_huge_page(struct page *hpage)
4269 4270 4271
{
	struct hstate *h = page_hstate(hpage);
	int nid = page_to_nid(hpage);
4272
	int ret = -EBUSY;
4273 4274

	spin_lock(&hugetlb_lock);
4275 4276 4277 4278 4279
	/*
	 * 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)) {
4280 4281 4282 4283 4284 4285 4286
		/*
		 * 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);
4287
		set_page_refcounted(hpage);
4288 4289 4290 4291
		h->free_huge_pages--;
		h->free_huge_pages_node[nid]--;
		ret = 0;
	}
4292
	spin_unlock(&hugetlb_lock);
4293
	return ret;
4294
}
4295
#endif
4296 4297 4298

bool isolate_huge_page(struct page *page, struct list_head *list)
{
4299 4300
	bool ret = true;

4301
	VM_BUG_ON_PAGE(!PageHead(page), page);
4302
	spin_lock(&hugetlb_lock);
4303 4304 4305 4306 4307
	if (!page_huge_active(page) || !get_page_unless_zero(page)) {
		ret = false;
		goto unlock;
	}
	clear_page_huge_active(page);
4308
	list_move_tail(&page->lru, list);
4309
unlock:
4310
	spin_unlock(&hugetlb_lock);
4311
	return ret;
4312 4313 4314 4315
}

void putback_active_hugepage(struct page *page)
{
4316
	VM_BUG_ON_PAGE(!PageHead(page), page);
4317
	spin_lock(&hugetlb_lock);
4318
	set_page_huge_active(page);
4319 4320 4321 4322
	list_move_tail(&page->lru, &(page_hstate(page))->hugepage_activelist);
	spin_unlock(&hugetlb_lock);
	put_page(page);
}