hugetlb.c 132.9 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
 */
#include <linux/list.h>
#include <linux/init.h>
#include <linux/mm.h>
8
#include <linux/seq_file.h>
L
Linus Torvalds 已提交
9 10
#include <linux/sysctl.h>
#include <linux/highmem.h>
A
Andrea Arcangeli 已提交
11
#include <linux/mmu_notifier.h>
L
Linus Torvalds 已提交
12
#include <linux/nodemask.h>
D
David Gibson 已提交
13
#include <linux/pagemap.h>
14
#include <linux/mempolicy.h>
15
#include <linux/compiler.h>
16
#include <linux/cpuset.h>
17
#include <linux/mutex.h>
18
#include <linux/memblock.h>
19
#include <linux/sysfs.h>
20
#include <linux/slab.h>
21
#include <linux/mmdebug.h>
22
#include <linux/sched/signal.h>
23
#include <linux/rmap.h>
24
#include <linux/string_helpers.h>
25 26
#include <linux/swap.h>
#include <linux/swapops.h>
27
#include <linux/jhash.h>
28

D
David Gibson 已提交
29 30
#include <asm/page.h>
#include <asm/pgtable.h>
31
#include <asm/tlb.h>
D
David Gibson 已提交
32

33
#include <linux/io.h>
D
David Gibson 已提交
34
#include <linux/hugetlb.h>
35
#include <linux/hugetlb_cgroup.h>
36
#include <linux/node.h>
37
#include <linux/userfaultfd_k.h>
38
#include <linux/page_owner.h>
39
#include "internal.h"
L
Linus Torvalds 已提交
40

41
int hugetlb_max_hstate __read_mostly;
42 43
unsigned int default_hstate_idx;
struct hstate hstates[HUGE_MAX_HSTATE];
44 45 46 47 48
/*
 * Minimum page order among possible hugepage sizes, set to a proper value
 * at boot time.
 */
static unsigned int minimum_order __read_mostly = UINT_MAX;
49

50 51
__initdata LIST_HEAD(huge_boot_pages);

52 53 54
/* for command line parsing */
static struct hstate * __initdata parsed_hstate;
static unsigned long __initdata default_hstate_max_huge_pages;
55
static unsigned long __initdata default_hstate_size;
56
static bool __initdata parsed_valid_hugepagesz = true;
57

58
/*
59 60
 * Protects updates to hugepage_freelists, hugepage_activelist, nr_huge_pages,
 * free_huge_pages, and surplus_huge_pages.
61
 */
62
DEFINE_SPINLOCK(hugetlb_lock);
63

64 65 66 67 68
/*
 * Serializes faults on the same logical page.  This is used to
 * prevent spurious OOMs when the hugepage pool is fully utilized.
 */
static int num_fault_mutexes;
69
struct mutex *hugetlb_fault_mutex_table ____cacheline_aligned_in_smp;
70

71 72 73
/* Forward declaration */
static int hugetlb_acct_memory(struct hstate *h, long delta);

74 75 76 77 78 79 80
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
81 82 83 84 85 86
	 * 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);
87
		kfree(spool);
88
	}
89 90
}

91 92
struct hugepage_subpool *hugepage_new_subpool(struct hstate *h, long max_hpages,
						long min_hpages)
93 94 95
{
	struct hugepage_subpool *spool;

96
	spool = kzalloc(sizeof(*spool), GFP_KERNEL);
97 98 99 100 101
	if (!spool)
		return NULL;

	spin_lock_init(&spool->lock);
	spool->count = 1;
102 103 104 105 106 107 108 109 110
	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;
111 112 113 114 115 116 117 118 119 120 121 122

	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);
}

123 124 125 126 127 128 129 130 131
/*
 * 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,
132 133
				      long delta)
{
134
	long ret = delta;
135 136

	if (!spool)
137
		return ret;
138 139

	spin_lock(&spool->lock);
140 141 142 143 144 145 146 147

	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;
		}
148 149
	}

150 151
	/* minimum size accounting */
	if (spool->min_hpages != -1 && spool->rsv_hpages) {
152 153 154 155 156 157 158 159 160 161 162 163 164 165 166
		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);
167 168 169
	return ret;
}

170 171 172 173 174 175 176
/*
 * 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,
177 178
				       long delta)
{
179 180
	long ret = delta;

181
	if (!spool)
182
		return delta;
183 184

	spin_lock(&spool->lock);
185 186 187 188

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

189 190
	 /* minimum size accounting */
	if (spool->min_hpages != -1 && spool->used_hpages < spool->min_hpages) {
191 192 193 194 195 196 197 198 199 200 201 202 203 204
		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.
	 */
205
	unlock_or_release_subpool(spool);
206 207

	return ret;
208 209 210 211 212 213 214 215 216
}

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
217
	return subpool_inode(file_inode(vma->vm_file));
218 219
}

220 221 222
/*
 * Region tracking -- allows tracking of reservations and instantiated pages
 *                    across the pages in a mapping.
223
 *
224 225 226 227 228 229 230 231 232 233 234 235 236 237
 * 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.
238 239 240 241 242 243 244
 */
struct file_region {
	struct list_head link;
	long from;
	long to;
};

245 246
/*
 * Add the huge page range represented by [f, t) to the reserve
247 248 249 250 251 252 253 254
 * 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.
255 256 257
 *
 * Return the number of new huge pages added to the map.  This
 * number is greater than or equal to zero.
258
 */
259
static long region_add(struct resv_map *resv, long f, long t)
260
{
261
	struct list_head *head = &resv->regions;
262
	struct file_region *rg, *nrg, *trg;
263
	long add = 0;
264

265
	spin_lock(&resv->lock);
266 267 268 269 270
	/* Locate the region we are either in or before. */
	list_for_each_entry(rg, head, link)
		if (f <= rg->to)
			break;

271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292
	/*
	 * 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;
	}

293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310
	/* 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) {
311 312 313 314 315
			/* 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);
316 317 318 319
			list_del(&rg->link);
			kfree(rg);
		}
	}
320 321

	add += (nrg->from - f);		/* Added to beginning of region */
322
	nrg->from = f;
323
	add += t - nrg->to;		/* Added to end of region */
324
	nrg->to = t;
325

326 327
out_locked:
	resv->adds_in_progress--;
328
	spin_unlock(&resv->lock);
329 330
	VM_BUG_ON(add < 0);
	return add;
331 332
}

333 334 335 336 337 338 339 340 341 342 343 344 345
/*
 * 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.
 *
346 347 348 349 350 351 352 353
 * 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.
354
 */
355
static long region_chg(struct resv_map *resv, long f, long t)
356
{
357
	struct list_head *head = &resv->regions;
358
	struct file_region *rg, *nrg = NULL;
359 360
	long chg = 0;

361 362
retry:
	spin_lock(&resv->lock);
363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378
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);
379 380
		if (!trg) {
			kfree(nrg);
381
			return -ENOMEM;
382
		}
383 384 385 386 387 388 389

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

390 391 392 393 394 395 396 397 398
	/* 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) {
399
		if (!nrg) {
400
			resv->adds_in_progress--;
401 402 403 404 405 406 407 408 409 410
			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;
		}
411

412 413 414
		list_add(&nrg->link, rg->link.prev);
		chg = t - f;
		goto out_nrg;
415 416 417 418 419 420 421 422 423 424 425 426
	}

	/* 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)
427
			goto out;
428

L
Lucas De Marchi 已提交
429
		/* We overlap with this area, if it extends further than
430 431 432 433 434 435 436 437
		 * 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;
	}
438 439 440 441 442 443 444 445

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);
446 447 448
	return chg;
}

449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467
/*
 * 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);
}

468
/*
469 470 471 472 473 474 475 476 477 478 479 480
 * 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.
481
 */
482
static long region_del(struct resv_map *resv, long f, long t)
483
{
484
	struct list_head *head = &resv->regions;
485
	struct file_region *rg, *trg;
486 487
	struct file_region *nrg = NULL;
	long del = 0;
488

489
retry:
490
	spin_lock(&resv->lock);
491
	list_for_each_entry_safe(rg, trg, head, link) {
492 493 494 495 496 497 498 499
		/*
		 * Skip regions before the range to be deleted.  file_region
		 * ranges are normally of the form [from, to).  However, there
		 * may be a "placeholder" entry in the map which is of the form
		 * (from, to) with from == to.  Check for placeholder entries
		 * at the beginning of the range to be deleted.
		 */
		if (rg->to <= f && (rg->to != rg->from || rg->to != f))
500
			continue;
501

502
		if (rg->from >= t)
503 504
			break;

505 506 507 508 509 510 511 512 513 514 515 516 517
		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--;
			}
518

519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538
			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;
539
			break;
540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555
		}

		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;
		}
556
	}
557 558

	spin_unlock(&resv->lock);
559 560
	kfree(nrg);
	return del;
561 562
}

563 564 565 566 567 568 569 570 571
/*
 * 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.
 */
572
void hugetlb_fix_reserve_counts(struct inode *inode)
573 574 575 576 577
{
	struct hugepage_subpool *spool = subpool_inode(inode);
	long rsv_adjust;

	rsv_adjust = hugepage_subpool_get_pages(spool, 1);
578
	if (rsv_adjust) {
579 580 581 582 583 584
		struct hstate *h = hstate_inode(inode);

		hugetlb_acct_memory(h, 1);
	}
}

585 586 587 588
/*
 * Count and return the number of huge pages in the reserve map
 * that intersect with the range [f, t).
 */
589
static long region_count(struct resv_map *resv, long f, long t)
590
{
591
	struct list_head *head = &resv->regions;
592 593 594
	struct file_region *rg;
	long chg = 0;

595
	spin_lock(&resv->lock);
596 597
	/* Locate each segment we overlap with, and count that overlap. */
	list_for_each_entry(rg, head, link) {
598 599
		long seg_from;
		long seg_to;
600 601 602 603 604 605 606 607 608 609 610

		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;
	}
611
	spin_unlock(&resv->lock);
612 613 614 615

	return chg;
}

616 617 618 619
/*
 * Convert the address within this vma to the page offset within
 * the mapping, in pagecache page units; huge pages here.
 */
620 621
static pgoff_t vma_hugecache_offset(struct hstate *h,
			struct vm_area_struct *vma, unsigned long address)
622
{
623 624
	return ((address - vma->vm_start) >> huge_page_shift(h)) +
			(vma->vm_pgoff >> huge_page_order(h));
625 626
}

627 628 629 630 631
pgoff_t linear_hugepage_index(struct vm_area_struct *vma,
				     unsigned long address)
{
	return vma_hugecache_offset(hstate_vma(vma), vma, address);
}
632
EXPORT_SYMBOL_GPL(linear_hugepage_index);
633

634 635 636 637 638 639
/*
 * 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)
{
640 641 642
	if (vma->vm_ops && vma->vm_ops->pagesize)
		return vma->vm_ops->pagesize(vma);
	return PAGE_SIZE;
643
}
644
EXPORT_SYMBOL_GPL(vma_kernel_pagesize);
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
649 650
 * architectures where it differs, an architecture-specific 'strong'
 * version of this symbol is required.
651
 */
652
__weak unsigned long vma_mmu_pagesize(struct vm_area_struct *vma)
653 654 655 656
{
	return vma_kernel_pagesize(vma);
}

657 658 659 660 661 662 663
/*
 * 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)
664
#define HPAGE_RESV_MASK (HPAGE_RESV_OWNER | HPAGE_RESV_UNMAPPED)
665

666 667 668 669 670 671 672 673 674
/*
 * 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.
675 676 677 678 679 680 681 682 683
 *
 * 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.
684
 */
685 686 687 688 689 690 691 692 693 694 695
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;
}

696
struct resv_map *resv_map_alloc(void)
697 698
{
	struct resv_map *resv_map = kmalloc(sizeof(*resv_map), GFP_KERNEL);
699 700 701 702 703
	struct file_region *rg = kmalloc(sizeof(*rg), GFP_KERNEL);

	if (!resv_map || !rg) {
		kfree(resv_map);
		kfree(rg);
704
		return NULL;
705
	}
706 707

	kref_init(&resv_map->refs);
708
	spin_lock_init(&resv_map->lock);
709 710
	INIT_LIST_HEAD(&resv_map->regions);

711 712 713 714 715 716
	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;

717 718 719
	return resv_map;
}

720
void resv_map_release(struct kref *ref)
721 722
{
	struct resv_map *resv_map = container_of(ref, struct resv_map, refs);
723 724
	struct list_head *head = &resv_map->region_cache;
	struct file_region *rg, *trg;
725 726

	/* Clear out any active regions before we release the map. */
727
	region_del(resv_map, 0, LONG_MAX);
728 729 730 731 732 733 734 735 736

	/* ... 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);

737 738 739
	kfree(resv_map);
}

740 741 742 743 744
static inline struct resv_map *inode_resv_map(struct inode *inode)
{
	return inode->i_mapping->private_data;
}

745
static struct resv_map *vma_resv_map(struct vm_area_struct *vma)
746
{
747
	VM_BUG_ON_VMA(!is_vm_hugetlb_page(vma), vma);
748 749 750 751 752 753 754
	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 {
755 756
		return (struct resv_map *)(get_vma_private_data(vma) &
							~HPAGE_RESV_MASK);
757
	}
758 759
}

760
static void set_vma_resv_map(struct vm_area_struct *vma, struct resv_map *map)
761
{
762 763
	VM_BUG_ON_VMA(!is_vm_hugetlb_page(vma), vma);
	VM_BUG_ON_VMA(vma->vm_flags & VM_MAYSHARE, vma);
764

765 766
	set_vma_private_data(vma, (get_vma_private_data(vma) &
				HPAGE_RESV_MASK) | (unsigned long)map);
767 768 769 770
}

static void set_vma_resv_flags(struct vm_area_struct *vma, unsigned long flags)
{
771 772
	VM_BUG_ON_VMA(!is_vm_hugetlb_page(vma), vma);
	VM_BUG_ON_VMA(vma->vm_flags & VM_MAYSHARE, vma);
773 774

	set_vma_private_data(vma, get_vma_private_data(vma) | flags);
775 776 777 778
}

static int is_vma_resv_set(struct vm_area_struct *vma, unsigned long flag)
{
779
	VM_BUG_ON_VMA(!is_vm_hugetlb_page(vma), vma);
780 781

	return (get_vma_private_data(vma) & flag) != 0;
782 783
}

784
/* Reset counters to 0 and clear all HPAGE_RESV_* flags */
785 786
void reset_vma_resv_huge_pages(struct vm_area_struct *vma)
{
787
	VM_BUG_ON_VMA(!is_vm_hugetlb_page(vma), vma);
788
	if (!(vma->vm_flags & VM_MAYSHARE))
789 790 791 792
		vma->vm_private_data = (void *)0;
}

/* Returns true if the VMA has associated reserve pages */
793
static bool vma_has_reserves(struct vm_area_struct *vma, long chg)
794
{
795 796 797 798 799 800 801 802 803 804 805
	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)
806
			return true;
807
		else
808
			return false;
809
	}
810 811

	/* Shared mappings always use reserves */
812 813 814 815 816 817 818 819 820 821 822 823 824
	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;
	}
825 826 827 828 829

	/*
	 * Only the process that called mmap() has reserves for
	 * private mappings.
	 */
830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850
	if (is_vma_resv_set(vma, HPAGE_RESV_OWNER)) {
		/*
		 * Like the shared case above, a hole punch or truncate
		 * could have been performed on the private mapping.
		 * Examine the value of chg to determine if reserves
		 * actually exist or were previously consumed.
		 * Very Subtle - The value of chg comes from a previous
		 * call to vma_needs_reserves().  The reserve map for
		 * private mappings has different (opposite) semantics
		 * than that of shared mappings.  vma_needs_reserves()
		 * has already taken this difference in semantics into
		 * account.  Therefore, the meaning of chg is the same
		 * as in the shared case above.  Code could easily be
		 * combined, but keeping it separate draws attention to
		 * subtle differences.
		 */
		if (chg)
			return false;
		else
			return true;
	}
851

852
	return false;
853 854
}

855
static void enqueue_huge_page(struct hstate *h, struct page *page)
L
Linus Torvalds 已提交
856 857
{
	int nid = page_to_nid(page);
858
	list_move(&page->lru, &h->hugepage_freelists[nid]);
859 860
	h->free_huge_pages++;
	h->free_huge_pages_node[nid]++;
L
Linus Torvalds 已提交
861 862
}

863
static struct page *dequeue_huge_page_node_exact(struct hstate *h, int nid)
864 865 866
{
	struct page *page;

867
	list_for_each_entry(page, &h->hugepage_freelists[nid], lru)
868
		if (!PageHWPoison(page))
869 870 871 872 873 874
			break;
	/*
	 * if 'non-isolated free hugepage' not found on the list,
	 * the allocation fails.
	 */
	if (&h->hugepage_freelists[nid] == &page->lru)
875
		return NULL;
876
	list_move(&page->lru, &h->hugepage_activelist);
877
	set_page_refcounted(page);
878 879 880 881 882
	h->free_huge_pages--;
	h->free_huge_pages_node[nid]--;
	return page;
}

883 884
static struct page *dequeue_huge_page_nodemask(struct hstate *h, gfp_t gfp_mask, int nid,
		nodemask_t *nmask)
885
{
886 887 888 889 890
	unsigned int cpuset_mems_cookie;
	struct zonelist *zonelist;
	struct zone *zone;
	struct zoneref *z;
	int node = -1;
891

892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907
	zonelist = node_zonelist(nid, gfp_mask);

retry_cpuset:
	cpuset_mems_cookie = read_mems_allowed_begin();
	for_each_zone_zonelist_nodemask(zone, z, zonelist, gfp_zone(gfp_mask), nmask) {
		struct page *page;

		if (!cpuset_zone_allowed(zone, gfp_mask))
			continue;
		/*
		 * no need to ask again on the same node. Pool is node rather than
		 * zone aware
		 */
		if (zone_to_nid(zone) == node)
			continue;
		node = zone_to_nid(zone);
908 909 910 911 912

		page = dequeue_huge_page_node_exact(h, node);
		if (page)
			return page;
	}
913 914 915
	if (unlikely(read_mems_allowed_retry(cpuset_mems_cookie)))
		goto retry_cpuset;

916 917 918
	return NULL;
}

919 920 921
/* Movability of hugepages depends on migration support. */
static inline gfp_t htlb_alloc_mask(struct hstate *h)
{
922
	if (hugepage_migration_supported(h))
923 924 925 926 927
		return GFP_HIGHUSER_MOVABLE;
	else
		return GFP_HIGHUSER;
}

928 929
static struct page *dequeue_huge_page_vma(struct hstate *h,
				struct vm_area_struct *vma,
930 931
				unsigned long address, int avoid_reserve,
				long chg)
L
Linus Torvalds 已提交
932
{
933
	struct page *page;
934
	struct mempolicy *mpol;
935
	gfp_t gfp_mask;
936
	nodemask_t *nodemask;
937
	int nid;
L
Linus Torvalds 已提交
938

939 940 941 942 943
	/*
	 * 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
	 */
944
	if (!vma_has_reserves(vma, chg) &&
945
			h->free_huge_pages - h->resv_huge_pages == 0)
946
		goto err;
947

948
	/* If reserves cannot be used, ensure enough pages are in the pool */
949
	if (avoid_reserve && h->free_huge_pages - h->resv_huge_pages == 0)
950
		goto err;
951

952 953
	gfp_mask = htlb_alloc_mask(h);
	nid = huge_node(vma, address, gfp_mask, &mpol, &nodemask);
954 955 956 957
	page = dequeue_huge_page_nodemask(h, gfp_mask, nid, nodemask);
	if (page && !avoid_reserve && vma_has_reserves(vma, chg)) {
		SetPagePrivate(page);
		h->resv_huge_pages--;
L
Linus Torvalds 已提交
958
	}
959

960
	mpol_cond_put(mpol);
L
Linus Torvalds 已提交
961
	return page;
962 963 964

err:
	return NULL;
L
Linus Torvalds 已提交
965 966
}

967 968 969 970 971 972 973 974 975
/*
 * 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)
{
976
	nid = next_node_in(nid, *nodes_allowed);
977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037
	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--)

1038
#ifdef CONFIG_ARCH_HAS_GIGANTIC_PAGE
1039
static void destroy_compound_gigantic_page(struct page *page,
1040
					unsigned int order)
1041 1042 1043 1044 1045
{
	int i;
	int nr_pages = 1 << order;
	struct page *p = page + 1;

1046
	atomic_set(compound_mapcount_ptr(page), 0);
1047
	for (i = 1; i < nr_pages; i++, p = mem_map_next(p, page, i)) {
1048
		clear_compound_head(p);
1049 1050 1051 1052 1053 1054 1055
		set_page_refcounted(p);
	}

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

1056
static void free_gigantic_page(struct page *page, unsigned int order)
1057 1058 1059 1060 1061
{
	free_contig_range(page_to_pfn(page), 1 << order);
}

static int __alloc_gigantic_page(unsigned long start_pfn,
1062
				unsigned long nr_pages, gfp_t gfp_mask)
1063 1064
{
	unsigned long end_pfn = start_pfn + nr_pages;
1065
	return alloc_contig_range(start_pfn, end_pfn, MIGRATE_MOVABLE,
1066
				  gfp_mask);
1067 1068
}

1069 1070
static bool pfn_range_valid_gigantic(struct zone *z,
			unsigned long start_pfn, unsigned long nr_pages)
1071 1072 1073 1074 1075 1076 1077 1078 1079 1080
{
	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);

1081 1082 1083
		if (page_zone(page) != z)
			return false;

1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103
		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);
}

1104 1105
static struct page *alloc_gigantic_page(struct hstate *h, gfp_t gfp_mask,
		int nid, nodemask_t *nodemask)
1106
{
1107
	unsigned int order = huge_page_order(h);
1108 1109
	unsigned long nr_pages = 1 << order;
	unsigned long ret, pfn, flags;
1110 1111 1112
	struct zonelist *zonelist;
	struct zone *zone;
	struct zoneref *z;
1113

1114
	zonelist = node_zonelist(nid, gfp_mask);
1115
	for_each_zone_zonelist_nodemask(zone, z, zonelist, gfp_zone(gfp_mask), nodemask) {
1116
		spin_lock_irqsave(&zone->lock, flags);
1117

1118 1119 1120
		pfn = ALIGN(zone->zone_start_pfn, nr_pages);
		while (zone_spans_last_pfn(zone, pfn, nr_pages)) {
			if (pfn_range_valid_gigantic(zone, pfn, nr_pages)) {
1121 1122 1123 1124 1125 1126 1127
				/*
				 * 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...
				 */
1128 1129
				spin_unlock_irqrestore(&zone->lock, flags);
				ret = __alloc_gigantic_page(pfn, nr_pages, gfp_mask);
1130 1131
				if (!ret)
					return pfn_to_page(pfn);
1132
				spin_lock_irqsave(&zone->lock, flags);
1133 1134 1135 1136
			}
			pfn += nr_pages;
		}

1137
		spin_unlock_irqrestore(&zone->lock, flags);
1138 1139 1140 1141 1142 1143
	}

	return NULL;
}

static void prep_new_huge_page(struct hstate *h, struct page *page, int nid);
1144
static void prep_compound_gigantic_page(struct page *page, unsigned int order);
1145

1146
#else /* !CONFIG_ARCH_HAS_GIGANTIC_PAGE */
1147
static inline bool gigantic_page_supported(void) { return false; }
1148 1149
static struct page *alloc_gigantic_page(struct hstate *h, gfp_t gfp_mask,
		int nid, nodemask_t *nodemask) { return NULL; }
1150
static inline void free_gigantic_page(struct page *page, unsigned int order) { }
1151
static inline void destroy_compound_gigantic_page(struct page *page,
1152
						unsigned int order) { }
1153 1154
#endif

1155
static void update_and_free_page(struct hstate *h, struct page *page)
A
Adam Litke 已提交
1156 1157
{
	int i;
1158

1159 1160
	if (hstate_is_gigantic(h) && !gigantic_page_supported())
		return;
1161

1162 1163 1164
	h->nr_huge_pages--;
	h->nr_huge_pages_node[page_to_nid(page)]--;
	for (i = 0; i < pages_per_huge_page(h); i++) {
1165 1166
		page[i].flags &= ~(1 << PG_locked | 1 << PG_error |
				1 << PG_referenced | 1 << PG_dirty |
1167 1168
				1 << PG_active | 1 << PG_private |
				1 << PG_writeback);
A
Adam Litke 已提交
1169
	}
1170
	VM_BUG_ON_PAGE(hugetlb_cgroup_from_page(page), page);
1171
	set_compound_page_dtor(page, NULL_COMPOUND_DTOR);
A
Adam Litke 已提交
1172
	set_page_refcounted(page);
1173 1174 1175 1176 1177 1178
	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
1179 1180
}

1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191
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;
}

1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216
/*
 * 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]);
}

1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238
/*
 * Internal hugetlb specific page flag. Do not use outside of the hugetlb
 * code
 */
static inline bool PageHugeTemporary(struct page *page)
{
	if (!PageHuge(page))
		return false;

	return (unsigned long)page[2].mapping == -1U;
}

static inline void SetPageHugeTemporary(struct page *page)
{
	page[2].mapping = (void *)-1U;
}

static inline void ClearPageHugeTemporary(struct page *page)
{
	page[2].mapping = NULL;
}

1239
void free_huge_page(struct page *page)
1240
{
1241 1242 1243 1244
	/*
	 * Can't pass hstate in here because it is called from the
	 * compound page destructor.
	 */
1245
	struct hstate *h = page_hstate(page);
1246
	int nid = page_to_nid(page);
1247 1248
	struct hugepage_subpool *spool =
		(struct hugepage_subpool *)page_private(page);
1249
	bool restore_reserve;
1250

1251 1252
	VM_BUG_ON_PAGE(page_count(page), page);
	VM_BUG_ON_PAGE(page_mapcount(page), page);
1253 1254 1255

	set_page_private(page, 0);
	page->mapping = NULL;
1256
	restore_reserve = PagePrivate(page);
1257
	ClearPagePrivate(page);
1258

1259 1260 1261 1262 1263 1264 1265 1266
	/*
	 * 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;

1267
	spin_lock(&hugetlb_lock);
1268
	clear_page_huge_active(page);
1269 1270
	hugetlb_cgroup_uncharge_page(hstate_index(h),
				     pages_per_huge_page(h), page);
1271 1272 1273
	if (restore_reserve)
		h->resv_huge_pages++;

1274 1275 1276 1277 1278
	if (PageHugeTemporary(page)) {
		list_del(&page->lru);
		ClearPageHugeTemporary(page);
		update_and_free_page(h, page);
	} else if (h->surplus_huge_pages_node[nid]) {
1279 1280
		/* remove the page from active list */
		list_del(&page->lru);
1281 1282 1283
		update_and_free_page(h, page);
		h->surplus_huge_pages--;
		h->surplus_huge_pages_node[nid]--;
1284
	} else {
1285
		arch_clear_hugepage_flags(page);
1286
		enqueue_huge_page(h, page);
1287
	}
1288 1289 1290
	spin_unlock(&hugetlb_lock);
}

1291
static void prep_new_huge_page(struct hstate *h, struct page *page, int nid)
1292
{
1293
	INIT_LIST_HEAD(&page->lru);
1294
	set_compound_page_dtor(page, HUGETLB_PAGE_DTOR);
1295
	spin_lock(&hugetlb_lock);
1296
	set_hugetlb_cgroup(page, NULL);
1297 1298
	h->nr_huge_pages++;
	h->nr_huge_pages_node[nid]++;
1299 1300 1301
	spin_unlock(&hugetlb_lock);
}

1302
static void prep_compound_gigantic_page(struct page *page, unsigned int order)
1303 1304 1305 1306 1307 1308 1309
{
	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);
1310
	__ClearPageReserved(page);
1311
	__SetPageHead(page);
1312
	for (i = 1; i < nr_pages; i++, p = mem_map_next(p, page, i)) {
1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325
		/*
		 * 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);
1326
		set_page_count(p, 0);
1327
		set_compound_head(p, page);
1328
	}
1329
	atomic_set(compound_mapcount_ptr(page), -1);
1330 1331
}

A
Andrew Morton 已提交
1332 1333 1334 1335 1336
/*
 * PageHuge() only returns true for hugetlbfs pages, but not for normal or
 * transparent huge pages.  See the PageTransHuge() documentation for more
 * details.
 */
1337 1338 1339 1340 1341 1342
int PageHuge(struct page *page)
{
	if (!PageCompound(page))
		return 0;

	page = compound_head(page);
1343
	return page[1].compound_dtor == HUGETLB_PAGE_DTOR;
1344
}
1345 1346
EXPORT_SYMBOL_GPL(PageHuge);

1347 1348 1349 1350 1351 1352 1353 1354 1355
/*
 * 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;

1356
	return get_compound_page_dtor(page_head) == free_huge_page;
1357 1358
}

1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375
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;
}

1376
static struct page *alloc_buddy_huge_page(struct hstate *h,
1377
		gfp_t gfp_mask, int nid, nodemask_t *nmask)
L
Linus Torvalds 已提交
1378
{
1379
	int order = huge_page_order(h);
L
Linus Torvalds 已提交
1380
	struct page *page;
1381

1382 1383 1384 1385 1386 1387 1388 1389
	gfp_mask |= __GFP_COMP|__GFP_RETRY_MAYFAIL|__GFP_NOWARN;
	if (nid == NUMA_NO_NODE)
		nid = numa_mem_id();
	page = __alloc_pages_nodemask(gfp_mask, order, nid, nmask);
	if (page)
		__count_vm_event(HTLB_BUDDY_PGALLOC);
	else
		__count_vm_event(HTLB_BUDDY_PGALLOC_FAIL);
1390 1391 1392 1393

	return page;
}

1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417
/*
 * Common helper to allocate a fresh hugetlb page. All specific allocators
 * should use this function to get new hugetlb pages
 */
static struct page *alloc_fresh_huge_page(struct hstate *h,
		gfp_t gfp_mask, int nid, nodemask_t *nmask)
{
	struct page *page;

	if (hstate_is_gigantic(h))
		page = alloc_gigantic_page(h, gfp_mask, nid, nmask);
	else
		page = alloc_buddy_huge_page(h, gfp_mask,
				nid, nmask);
	if (!page)
		return NULL;

	if (hstate_is_gigantic(h))
		prep_compound_gigantic_page(page, huge_page_order(h));
	prep_new_huge_page(h, page, page_to_nid(page));

	return page;
}

1418 1419 1420 1421
/*
 * Allocates a fresh page to the hugetlb allocator pool in the node interleaved
 * manner.
 */
1422
static int alloc_pool_huge_page(struct hstate *h, nodemask_t *nodes_allowed)
1423 1424 1425
{
	struct page *page;
	int nr_nodes, node;
1426
	gfp_t gfp_mask = htlb_alloc_mask(h) | __GFP_THISNODE;
1427 1428

	for_each_node_mask_to_alloc(h, nr_nodes, node, nodes_allowed) {
1429
		page = alloc_fresh_huge_page(h, gfp_mask, node, nodes_allowed);
1430
		if (page)
1431 1432 1433
			break;
	}

1434 1435
	if (!page)
		return 0;
1436

1437 1438 1439
	put_page(page); /* free it into the hugepage allocator */

	return 1;
1440 1441
}

1442 1443 1444 1445 1446 1447
/*
 * 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.
 */
1448 1449
static int free_pool_huge_page(struct hstate *h, nodemask_t *nodes_allowed,
							 bool acct_surplus)
1450
{
1451
	int nr_nodes, node;
1452 1453
	int ret = 0;

1454
	for_each_node_mask_to_free(h, nr_nodes, node, nodes_allowed) {
1455 1456 1457 1458
		/*
		 * If we're returning unused surplus pages, only examine
		 * nodes with surplus pages.
		 */
1459 1460
		if ((!acct_surplus || h->surplus_huge_pages_node[node]) &&
		    !list_empty(&h->hugepage_freelists[node])) {
1461
			struct page *page =
1462
				list_entry(h->hugepage_freelists[node].next,
1463 1464 1465
					  struct page, lru);
			list_del(&page->lru);
			h->free_huge_pages--;
1466
			h->free_huge_pages_node[node]--;
1467 1468
			if (acct_surplus) {
				h->surplus_huge_pages--;
1469
				h->surplus_huge_pages_node[node]--;
1470
			}
1471 1472
			update_and_free_page(h, page);
			ret = 1;
1473
			break;
1474
		}
1475
	}
1476 1477 1478 1479

	return ret;
}

1480 1481
/*
 * Dissolve a given free hugepage into free buddy pages. This function does
1482
 * nothing for in-use (including surplus) hugepages. Returns -EBUSY if the
1483 1484
 * dissolution fails because a give page is not a free hugepage, or because
 * free hugepages are fully reserved.
1485
 */
1486
int dissolve_free_huge_page(struct page *page)
1487
{
1488
	int rc = -EBUSY;
1489

1490 1491
	spin_lock(&hugetlb_lock);
	if (PageHuge(page) && !page_count(page)) {
1492 1493 1494
		struct page *head = compound_head(page);
		struct hstate *h = page_hstate(head);
		int nid = page_to_nid(head);
1495
		if (h->free_huge_pages - h->resv_huge_pages == 0)
1496
			goto out;
1497 1498 1499 1500 1501 1502 1503 1504
		/*
		 * Move PageHWPoison flag from head page to the raw error page,
		 * which makes any subpages rather than the error page reusable.
		 */
		if (PageHWPoison(head) && page != head) {
			SetPageHWPoison(page);
			ClearPageHWPoison(head);
		}
1505
		list_del(&head->lru);
1506 1507
		h->free_huge_pages--;
		h->free_huge_pages_node[nid]--;
1508
		h->max_huge_pages--;
1509
		update_and_free_page(h, head);
1510
		rc = 0;
1511
	}
1512
out:
1513
	spin_unlock(&hugetlb_lock);
1514
	return rc;
1515 1516 1517 1518 1519
}

/*
 * Dissolve free hugepages in a given pfn range. Used by memory hotplug to
 * make specified memory blocks removable from the system.
1520 1521
 * Note that this will dissolve a free gigantic hugepage completely, if any
 * part of it lies within the given range.
1522 1523
 * Also note that if dissolve_free_huge_page() returns with an error, all
 * free hugepages that were dissolved before that error are lost.
1524
 */
1525
int dissolve_free_huge_pages(unsigned long start_pfn, unsigned long end_pfn)
1526 1527
{
	unsigned long pfn;
1528
	struct page *page;
1529
	int rc = 0;
1530

1531
	if (!hugepages_supported())
1532
		return rc;
1533

1534 1535 1536 1537 1538 1539 1540 1541
	for (pfn = start_pfn; pfn < end_pfn; pfn += 1 << minimum_order) {
		page = pfn_to_page(pfn);
		if (PageHuge(page) && !page_count(page)) {
			rc = dissolve_free_huge_page(page);
			if (rc)
				break;
		}
	}
1542 1543

	return rc;
1544 1545
}

1546 1547 1548
/*
 * Allocates a fresh surplus page from the page allocator.
 */
1549
static struct page *alloc_surplus_huge_page(struct hstate *h, gfp_t gfp_mask,
1550
		int nid, nodemask_t *nmask)
1551
{
1552
	struct page *page = NULL;
1553

1554
	if (hstate_is_gigantic(h))
1555 1556
		return NULL;

1557
	spin_lock(&hugetlb_lock);
1558 1559
	if (h->surplus_huge_pages >= h->nr_overcommit_huge_pages)
		goto out_unlock;
1560 1561
	spin_unlock(&hugetlb_lock);

1562
	page = alloc_fresh_huge_page(h, gfp_mask, nid, nmask);
1563
	if (!page)
1564
		return NULL;
1565 1566

	spin_lock(&hugetlb_lock);
1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579
	/*
	 * We could have raced with the pool size change.
	 * Double check that and simply deallocate the new page
	 * if we would end up overcommiting the surpluses. Abuse
	 * temporary page to workaround the nasty free_huge_page
	 * codeflow
	 */
	if (h->surplus_huge_pages >= h->nr_overcommit_huge_pages) {
		SetPageHugeTemporary(page);
		put_page(page);
		page = NULL;
	} else {
		h->surplus_huge_pages++;
1580
		h->surplus_huge_pages_node[page_to_nid(page)]++;
1581
	}
1582 1583

out_unlock:
1584
	spin_unlock(&hugetlb_lock);
1585 1586 1587 1588

	return page;
}

1589
static struct page *alloc_migrate_huge_page(struct hstate *h, gfp_t gfp_mask,
1590 1591 1592 1593 1594 1595 1596
		int nid, nodemask_t *nmask)
{
	struct page *page;

	if (hstate_is_gigantic(h))
		return NULL;

1597
	page = alloc_fresh_huge_page(h, gfp_mask, nid, nmask);
1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609
	if (!page)
		return NULL;

	/*
	 * We do not account these pages as surplus because they are only
	 * temporary and will be released properly on the last reference
	 */
	SetPageHugeTemporary(page);

	return page;
}

1610 1611 1612
/*
 * Use the VMA's mpolicy to allocate a huge page from the buddy.
 */
D
Dave Hansen 已提交
1613
static
1614
struct page *alloc_buddy_huge_page_with_mpol(struct hstate *h,
1615 1616
		struct vm_area_struct *vma, unsigned long addr)
{
1617 1618 1619 1620 1621 1622 1623
	struct page *page;
	struct mempolicy *mpol;
	gfp_t gfp_mask = htlb_alloc_mask(h);
	int nid;
	nodemask_t *nodemask;

	nid = huge_node(vma, addr, gfp_mask, &mpol, &nodemask);
1624
	page = alloc_surplus_huge_page(h, gfp_mask, nid, nodemask);
1625 1626 1627
	mpol_cond_put(mpol);

	return page;
1628 1629
}

1630
/* page migration callback function */
1631 1632
struct page *alloc_huge_page_node(struct hstate *h, int nid)
{
1633
	gfp_t gfp_mask = htlb_alloc_mask(h);
1634
	struct page *page = NULL;
1635

1636 1637 1638
	if (nid != NUMA_NO_NODE)
		gfp_mask |= __GFP_THISNODE;

1639
	spin_lock(&hugetlb_lock);
1640
	if (h->free_huge_pages - h->resv_huge_pages > 0)
1641
		page = dequeue_huge_page_nodemask(h, gfp_mask, nid, NULL);
1642 1643
	spin_unlock(&hugetlb_lock);

1644
	if (!page)
1645
		page = alloc_migrate_huge_page(h, gfp_mask, nid, NULL);
1646 1647 1648 1649

	return page;
}

1650
/* page migration callback function */
1651 1652
struct page *alloc_huge_page_nodemask(struct hstate *h, int preferred_nid,
		nodemask_t *nmask)
1653
{
1654
	gfp_t gfp_mask = htlb_alloc_mask(h);
1655 1656 1657

	spin_lock(&hugetlb_lock);
	if (h->free_huge_pages - h->resv_huge_pages > 0) {
1658 1659 1660 1661 1662 1663
		struct page *page;

		page = dequeue_huge_page_nodemask(h, gfp_mask, preferred_nid, nmask);
		if (page) {
			spin_unlock(&hugetlb_lock);
			return page;
1664 1665 1666 1667
		}
	}
	spin_unlock(&hugetlb_lock);

1668
	return alloc_migrate_huge_page(h, gfp_mask, preferred_nid, nmask);
1669 1670
}

1671
/* mempolicy aware migration callback */
1672 1673
struct page *alloc_huge_page_vma(struct hstate *h, struct vm_area_struct *vma,
		unsigned long address)
1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684 1685 1686 1687 1688
{
	struct mempolicy *mpol;
	nodemask_t *nodemask;
	struct page *page;
	gfp_t gfp_mask;
	int node;

	gfp_mask = htlb_alloc_mask(h);
	node = huge_node(vma, address, gfp_mask, &mpol, &nodemask);
	page = alloc_huge_page_nodemask(h, node, nodemask);
	mpol_cond_put(mpol);

	return page;
}

1689
/*
L
Lucas De Marchi 已提交
1690
 * Increase the hugetlb pool such that it can accommodate a reservation
1691 1692
 * of size 'delta'.
 */
1693
static int gather_surplus_pages(struct hstate *h, int delta)
1694 1695 1696 1697 1698
{
	struct list_head surplus_list;
	struct page *page, *tmp;
	int ret, i;
	int needed, allocated;
1699
	bool alloc_ok = true;
1700

1701
	needed = (h->resv_huge_pages + delta) - h->free_huge_pages;
1702
	if (needed <= 0) {
1703
		h->resv_huge_pages += delta;
1704
		return 0;
1705
	}
1706 1707 1708 1709 1710 1711 1712 1713

	allocated = 0;
	INIT_LIST_HEAD(&surplus_list);

	ret = -ENOMEM;
retry:
	spin_unlock(&hugetlb_lock);
	for (i = 0; i < needed; i++) {
1714
		page = alloc_surplus_huge_page(h, htlb_alloc_mask(h),
1715
				NUMA_NO_NODE, NULL);
1716 1717 1718 1719
		if (!page) {
			alloc_ok = false;
			break;
		}
1720
		list_add(&page->lru, &surplus_list);
1721
		cond_resched();
1722
	}
1723
	allocated += i;
1724 1725 1726 1727 1728 1729

	/*
	 * 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);
1730 1731
	needed = (h->resv_huge_pages + delta) -
			(h->free_huge_pages + allocated);
1732 1733 1734 1735 1736 1737 1738 1739 1740 1741
	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;
	}
1742 1743
	/*
	 * The surplus_list now contains _at_least_ the number of extra pages
L
Lucas De Marchi 已提交
1744
	 * needed to accommodate the reservation.  Add the appropriate number
1745
	 * of pages to the hugetlb pool and free the extras back to the buddy
1746 1747 1748
	 * 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.
1749 1750
	 */
	needed += allocated;
1751
	h->resv_huge_pages += delta;
1752
	ret = 0;
1753

1754
	/* Free the needed pages to the hugetlb pool */
1755
	list_for_each_entry_safe(page, tmp, &surplus_list, lru) {
1756 1757
		if ((--needed) < 0)
			break;
1758 1759 1760 1761 1762
		/*
		 * This page is now managed by the hugetlb allocator and has
		 * no users -- drop the buddy allocator's reference.
		 */
		put_page_testzero(page);
1763
		VM_BUG_ON_PAGE(page_count(page), page);
1764
		enqueue_huge_page(h, page);
1765
	}
1766
free:
1767
	spin_unlock(&hugetlb_lock);
1768 1769

	/* Free unnecessary surplus pages to the buddy allocator */
1770 1771
	list_for_each_entry_safe(page, tmp, &surplus_list, lru)
		put_page(page);
1772
	spin_lock(&hugetlb_lock);
1773 1774 1775 1776 1777

	return ret;
}

/*
1778 1779 1780 1781 1782 1783 1784 1785 1786 1787 1788 1789
 * This routine has two main purposes:
 * 1) Decrement the reservation count (resv_huge_pages) by the value passed
 *    in unused_resv_pages.  This corresponds to the prior adjustments made
 *    to the associated reservation map.
 * 2) Free any unused surplus pages that may have been allocated to satisfy
 *    the reservation.  As many as unused_resv_pages may be freed.
 *
 * Called with hugetlb_lock held.  However, the lock could be dropped (and
 * reacquired) during calls to cond_resched_lock.  Whenever dropping the lock,
 * we must make sure nobody else can claim pages we are in the process of
 * freeing.  Do this by ensuring resv_huge_page always is greater than the
 * number of huge pages we plan to free when dropping the lock.
1790
 */
1791 1792
static void return_unused_surplus_pages(struct hstate *h,
					unsigned long unused_resv_pages)
1793 1794 1795
{
	unsigned long nr_pages;

1796
	/* Cannot return gigantic pages currently */
1797
	if (hstate_is_gigantic(h))
1798
		goto out;
1799

1800 1801 1802 1803
	/*
	 * Part (or even all) of the reservation could have been backed
	 * by pre-allocated pages. Only free surplus pages.
	 */
1804
	nr_pages = min(unused_resv_pages, h->surplus_huge_pages);
1805

1806 1807
	/*
	 * We want to release as many surplus pages as possible, spread
1808 1809 1810 1811 1812
	 * 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.
1813 1814 1815 1816
	 *
	 * Note that we decrement resv_huge_pages as we free the pages.  If
	 * we drop the lock, resv_huge_pages will still be sufficiently large
	 * to cover subsequent pages we may free.
1817 1818
	 */
	while (nr_pages--) {
1819 1820
		h->resv_huge_pages--;
		unused_resv_pages--;
1821
		if (!free_pool_huge_page(h, &node_states[N_MEMORY], 1))
1822
			goto out;
1823
		cond_resched_lock(&hugetlb_lock);
1824
	}
1825 1826 1827 1828

out:
	/* Fully uncommit the reservation */
	h->resv_huge_pages -= unused_resv_pages;
1829 1830
}

1831

1832
/*
1833
 * vma_needs_reservation, vma_commit_reservation and vma_end_reservation
1834
 * are used by the huge page allocation routines to manage reservations.
1835 1836 1837 1838 1839 1840
 *
 * 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
1841 1842 1843
 * 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.
1844 1845 1846 1847 1848 1849
 *
 * 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.
1850 1851 1852 1853 1854
 *
 * vma_add_reservation is used in error paths where a reservation must
 * be restored when a newly allocated huge page must be freed.  It is
 * to be called after calling vma_needs_reservation to determine if a
 * reservation exists.
1855
 */
1856 1857 1858
enum vma_resv_mode {
	VMA_NEEDS_RESV,
	VMA_COMMIT_RESV,
1859
	VMA_END_RESV,
1860
	VMA_ADD_RESV,
1861
};
1862 1863
static long __vma_reservation_common(struct hstate *h,
				struct vm_area_struct *vma, unsigned long addr,
1864
				enum vma_resv_mode mode)
1865
{
1866 1867
	struct resv_map *resv;
	pgoff_t idx;
1868
	long ret;
1869

1870 1871
	resv = vma_resv_map(vma);
	if (!resv)
1872
		return 1;
1873

1874
	idx = vma_hugecache_offset(h, vma, addr);
1875 1876
	switch (mode) {
	case VMA_NEEDS_RESV:
1877
		ret = region_chg(resv, idx, idx + 1);
1878 1879 1880 1881
		break;
	case VMA_COMMIT_RESV:
		ret = region_add(resv, idx, idx + 1);
		break;
1882
	case VMA_END_RESV:
1883 1884 1885
		region_abort(resv, idx, idx + 1);
		ret = 0;
		break;
1886 1887 1888 1889 1890 1891 1892 1893
	case VMA_ADD_RESV:
		if (vma->vm_flags & VM_MAYSHARE)
			ret = region_add(resv, idx, idx + 1);
		else {
			region_abort(resv, idx, idx + 1);
			ret = region_del(resv, idx, idx + 1);
		}
		break;
1894 1895 1896
	default:
		BUG();
	}
1897

1898
	if (vma->vm_flags & VM_MAYSHARE)
1899
		return ret;
1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918
	else if (is_vma_resv_set(vma, HPAGE_RESV_OWNER) && ret >= 0) {
		/*
		 * In most cases, reserves always exist for private mappings.
		 * However, a file associated with mapping could have been
		 * hole punched or truncated after reserves were consumed.
		 * As subsequent fault on such a range will not use reserves.
		 * Subtle - The reserve map for private mappings has the
		 * opposite meaning than that of shared mappings.  If NO
		 * entry is in the reserve map, it means a reservation exists.
		 * If an entry exists in the reserve map, it means the
		 * reservation has already been consumed.  As a result, the
		 * return value of this routine is the opposite of the
		 * value returned from reserve map manipulation routines above.
		 */
		if (ret)
			return 0;
		else
			return 1;
	}
1919
	else
1920
		return ret < 0 ? ret : 0;
1921
}
1922 1923

static long vma_needs_reservation(struct hstate *h,
1924
			struct vm_area_struct *vma, unsigned long addr)
1925
{
1926
	return __vma_reservation_common(h, vma, addr, VMA_NEEDS_RESV);
1927
}
1928

1929 1930 1931
static long vma_commit_reservation(struct hstate *h,
			struct vm_area_struct *vma, unsigned long addr)
{
1932 1933 1934
	return __vma_reservation_common(h, vma, addr, VMA_COMMIT_RESV);
}

1935
static void vma_end_reservation(struct hstate *h,
1936 1937
			struct vm_area_struct *vma, unsigned long addr)
{
1938
	(void)__vma_reservation_common(h, vma, addr, VMA_END_RESV);
1939 1940
}

1941 1942 1943 1944 1945 1946 1947 1948 1949 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990
static long vma_add_reservation(struct hstate *h,
			struct vm_area_struct *vma, unsigned long addr)
{
	return __vma_reservation_common(h, vma, addr, VMA_ADD_RESV);
}

/*
 * This routine is called to restore a reservation on error paths.  In the
 * specific error paths, a huge page was allocated (via alloc_huge_page)
 * and is about to be freed.  If a reservation for the page existed,
 * alloc_huge_page would have consumed the reservation and set PagePrivate
 * in the newly allocated page.  When the page is freed via free_huge_page,
 * the global reservation count will be incremented if PagePrivate is set.
 * However, free_huge_page can not adjust the reserve map.  Adjust the
 * reserve map here to be consistent with global reserve count adjustments
 * to be made by free_huge_page.
 */
static void restore_reserve_on_error(struct hstate *h,
			struct vm_area_struct *vma, unsigned long address,
			struct page *page)
{
	if (unlikely(PagePrivate(page))) {
		long rc = vma_needs_reservation(h, vma, address);

		if (unlikely(rc < 0)) {
			/*
			 * Rare out of memory condition in reserve map
			 * manipulation.  Clear PagePrivate so that
			 * global reserve count will not be incremented
			 * by free_huge_page.  This will make it appear
			 * as though the reservation for this page was
			 * consumed.  This may prevent the task from
			 * faulting in the page at a later time.  This
			 * is better than inconsistent global huge page
			 * accounting of reserve counts.
			 */
			ClearPagePrivate(page);
		} else if (rc) {
			rc = vma_add_reservation(h, vma, address);
			if (unlikely(rc < 0))
				/*
				 * See above comment about rare out of
				 * memory condition.
				 */
				ClearPagePrivate(page);
		} else
			vma_end_reservation(h, vma, address);
	}
}

1991
struct page *alloc_huge_page(struct vm_area_struct *vma,
1992
				    unsigned long addr, int avoid_reserve)
L
Linus Torvalds 已提交
1993
{
1994
	struct hugepage_subpool *spool = subpool_vma(vma);
1995
	struct hstate *h = hstate_vma(vma);
1996
	struct page *page;
1997 1998
	long map_chg, map_commit;
	long gbl_chg;
1999 2000
	int ret, idx;
	struct hugetlb_cgroup *h_cg;
2001

2002
	idx = hstate_index(h);
2003
	/*
2004 2005 2006
	 * 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).
2007
	 */
2008 2009
	map_chg = gbl_chg = vma_needs_reservation(h, vma, addr);
	if (map_chg < 0)
2010
		return ERR_PTR(-ENOMEM);
2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021

	/*
	 * 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) {
2022
			vma_end_reservation(h, vma, addr);
2023
			return ERR_PTR(-ENOSPC);
2024
		}
L
Linus Torvalds 已提交
2025

2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037
		/*
		 * 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;
	}

2038
	ret = hugetlb_cgroup_charge_cgroup(idx, pages_per_huge_page(h), &h_cg);
2039 2040 2041
	if (ret)
		goto out_subpool_put;

L
Linus Torvalds 已提交
2042
	spin_lock(&hugetlb_lock);
2043 2044 2045 2046 2047 2048
	/*
	 * 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);
2049
	if (!page) {
2050
		spin_unlock(&hugetlb_lock);
2051
		page = alloc_buddy_huge_page_with_mpol(h, vma, addr);
2052 2053
		if (!page)
			goto out_uncharge_cgroup;
2054 2055 2056 2057
		if (!avoid_reserve && vma_has_reserves(vma, gbl_chg)) {
			SetPagePrivate(page);
			h->resv_huge_pages--;
		}
2058 2059
		spin_lock(&hugetlb_lock);
		list_move(&page->lru, &h->hugepage_activelist);
2060
		/* Fall through */
K
Ken Chen 已提交
2061
	}
2062 2063
	hugetlb_cgroup_commit_charge(idx, pages_per_huge_page(h), h_cg, page);
	spin_unlock(&hugetlb_lock);
2064

2065
	set_page_private(page, (unsigned long)spool);
2066

2067 2068
	map_commit = vma_commit_reservation(h, vma, addr);
	if (unlikely(map_chg > map_commit)) {
2069 2070 2071 2072 2073 2074 2075 2076 2077 2078 2079 2080 2081 2082
		/*
		 * 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);
	}
2083
	return page;
2084 2085 2086 2087

out_uncharge_cgroup:
	hugetlb_cgroup_uncharge_cgroup(idx, pages_per_huge_page(h), h_cg);
out_subpool_put:
2088
	if (map_chg || avoid_reserve)
2089
		hugepage_subpool_put_pages(spool, 1);
2090
	vma_end_reservation(h, vma, addr);
2091
	return ERR_PTR(-ENOSPC);
2092 2093
}

2094 2095 2096
int alloc_bootmem_huge_page(struct hstate *h)
	__attribute__ ((weak, alias("__alloc_bootmem_huge_page")));
int __alloc_bootmem_huge_page(struct hstate *h)
2097 2098
{
	struct huge_bootmem_page *m;
2099
	int nr_nodes, node;
2100

2101
	for_each_node_mask_to_alloc(h, nr_nodes, node, &node_states[N_MEMORY]) {
2102 2103
		void *addr;

2104
		addr = memblock_alloc_try_nid_raw(
2105
				huge_page_size(h), huge_page_size(h),
2106
				0, MEMBLOCK_ALLOC_ACCESSIBLE, node);
2107 2108 2109 2110 2111 2112 2113
		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;
2114
			goto found;
2115 2116 2117 2118 2119
		}
	}
	return 0;

found:
2120
	BUG_ON(!IS_ALIGNED(virt_to_phys(m), huge_page_size(h)));
2121
	/* Put them into a private list first because mem_map is not up yet */
2122
	INIT_LIST_HEAD(&m->list);
2123 2124 2125 2126 2127
	list_add(&m->list, &huge_boot_pages);
	m->hstate = h;
	return 1;
}

2128 2129
static void __init prep_compound_huge_page(struct page *page,
		unsigned int order)
2130 2131 2132 2133 2134 2135 2136
{
	if (unlikely(order > (MAX_ORDER - 1)))
		prep_compound_gigantic_page(page, order);
	else
		prep_compound_page(page, order);
}

2137 2138 2139 2140 2141 2142
/* 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) {
2143
		struct page *page = virt_to_page(m);
2144
		struct hstate *h = m->hstate;
2145

2146
		WARN_ON(page_count(page) != 1);
2147
		prep_compound_huge_page(page, h->order);
2148
		WARN_ON(PageReserved(page));
2149
		prep_new_huge_page(h, page, page_to_nid(page));
2150 2151
		put_page(page); /* free it into the hugepage allocator */

2152 2153 2154 2155 2156 2157
		/*
		 * 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.
		 */
2158
		if (hstate_is_gigantic(h))
2159
			adjust_managed_page_count(page, 1 << h->order);
2160
		cond_resched();
2161 2162 2163
	}
}

2164
static void __init hugetlb_hstate_alloc_pages(struct hstate *h)
L
Linus Torvalds 已提交
2165 2166
{
	unsigned long i;
2167

2168
	for (i = 0; i < h->max_huge_pages; ++i) {
2169
		if (hstate_is_gigantic(h)) {
2170 2171
			if (!alloc_bootmem_huge_page(h))
				break;
2172
		} else if (!alloc_pool_huge_page(h,
2173
					 &node_states[N_MEMORY]))
L
Linus Torvalds 已提交
2174
			break;
2175
		cond_resched();
L
Linus Torvalds 已提交
2176
	}
2177 2178 2179
	if (i < h->max_huge_pages) {
		char buf[32];

2180
		string_get_size(huge_page_size(h), 1, STRING_UNITS_2, buf, 32);
2181 2182 2183 2184
		pr_warn("HugeTLB: allocating %lu of page size %s failed.  Only allocated %lu hugepages.\n",
			h->max_huge_pages, buf, i);
		h->max_huge_pages = i;
	}
2185 2186 2187 2188 2189 2190 2191
}

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

	for_each_hstate(h) {
2192 2193 2194
		if (minimum_order > huge_page_order(h))
			minimum_order = huge_page_order(h);

2195
		/* oversize hugepages were init'ed in early boot */
2196
		if (!hstate_is_gigantic(h))
2197
			hugetlb_hstate_alloc_pages(h);
2198
	}
2199
	VM_BUG_ON(minimum_order == UINT_MAX);
2200 2201 2202 2203 2204 2205 2206
}

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

	for_each_hstate(h) {
A
Andi Kleen 已提交
2207
		char buf[32];
2208 2209

		string_get_size(huge_page_size(h), 1, STRING_UNITS_2, buf, 32);
2210
		pr_info("HugeTLB registered %s page size, pre-allocated %ld pages\n",
2211
			buf, h->free_huge_pages);
2212 2213 2214
	}
}

L
Linus Torvalds 已提交
2215
#ifdef CONFIG_HIGHMEM
2216 2217
static void try_to_free_low(struct hstate *h, unsigned long count,
						nodemask_t *nodes_allowed)
L
Linus Torvalds 已提交
2218
{
2219 2220
	int i;

2221
	if (hstate_is_gigantic(h))
2222 2223
		return;

2224
	for_each_node_mask(i, *nodes_allowed) {
L
Linus Torvalds 已提交
2225
		struct page *page, *next;
2226 2227 2228
		struct list_head *freel = &h->hugepage_freelists[i];
		list_for_each_entry_safe(page, next, freel, lru) {
			if (count >= h->nr_huge_pages)
2229
				return;
L
Linus Torvalds 已提交
2230 2231 2232
			if (PageHighMem(page))
				continue;
			list_del(&page->lru);
2233
			update_and_free_page(h, page);
2234 2235
			h->free_huge_pages--;
			h->free_huge_pages_node[page_to_nid(page)]--;
L
Linus Torvalds 已提交
2236 2237 2238 2239
		}
	}
}
#else
2240 2241
static inline void try_to_free_low(struct hstate *h, unsigned long count,
						nodemask_t *nodes_allowed)
L
Linus Torvalds 已提交
2242 2243 2244 2245
{
}
#endif

2246 2247 2248 2249 2250
/*
 * 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.
 */
2251 2252
static int adjust_pool_surplus(struct hstate *h, nodemask_t *nodes_allowed,
				int delta)
2253
{
2254
	int nr_nodes, node;
2255 2256 2257

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

2258 2259 2260 2261
	if (delta < 0) {
		for_each_node_mask_to_alloc(h, nr_nodes, node, nodes_allowed) {
			if (h->surplus_huge_pages_node[node])
				goto found;
2262
		}
2263 2264 2265 2266 2267
	} 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;
2268
		}
2269 2270
	}
	return 0;
2271

2272 2273 2274 2275
found:
	h->surplus_huge_pages += delta;
	h->surplus_huge_pages_node[node] += delta;
	return 1;
2276 2277
}

2278
#define persistent_huge_pages(h) (h->nr_huge_pages - h->surplus_huge_pages)
2279 2280
static unsigned long set_max_huge_pages(struct hstate *h, unsigned long count,
						nodemask_t *nodes_allowed)
L
Linus Torvalds 已提交
2281
{
2282
	unsigned long min_count, ret;
L
Linus Torvalds 已提交
2283

2284
	if (hstate_is_gigantic(h) && !gigantic_page_supported())
2285 2286
		return h->max_huge_pages;

2287 2288 2289 2290
	/*
	 * Increase the pool size
	 * First take pages out of surplus state.  Then make up the
	 * remaining difference by allocating fresh huge pages.
2291
	 *
2292
	 * We might race with alloc_surplus_huge_page() here and be unable
2293 2294 2295 2296
	 * 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.
2297
	 */
L
Linus Torvalds 已提交
2298
	spin_lock(&hugetlb_lock);
2299
	while (h->surplus_huge_pages && count > persistent_huge_pages(h)) {
2300
		if (!adjust_pool_surplus(h, nodes_allowed, -1))
2301 2302 2303
			break;
	}

2304
	while (count > persistent_huge_pages(h)) {
2305 2306 2307 2308 2309 2310
		/*
		 * 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);
2311 2312 2313 2314

		/* yield cpu to avoid soft lockup */
		cond_resched();

2315
		ret = alloc_pool_huge_page(h, nodes_allowed);
2316 2317 2318 2319
		spin_lock(&hugetlb_lock);
		if (!ret)
			goto out;

2320 2321 2322
		/* Bail for signals. Probably ctrl-c from user */
		if (signal_pending(current))
			goto out;
2323 2324 2325 2326 2327 2328 2329 2330
	}

	/*
	 * 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.
2331 2332 2333 2334
	 *
	 * 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
2335
	 * alloc_surplus_huge_page() is checking the global counter,
2336 2337 2338
	 * 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.
2339
	 */
2340
	min_count = h->resv_huge_pages + h->nr_huge_pages - h->free_huge_pages;
2341
	min_count = max(count, min_count);
2342
	try_to_free_low(h, min_count, nodes_allowed);
2343
	while (min_count < persistent_huge_pages(h)) {
2344
		if (!free_pool_huge_page(h, nodes_allowed, 0))
L
Linus Torvalds 已提交
2345
			break;
2346
		cond_resched_lock(&hugetlb_lock);
L
Linus Torvalds 已提交
2347
	}
2348
	while (count < persistent_huge_pages(h)) {
2349
		if (!adjust_pool_surplus(h, nodes_allowed, 1))
2350 2351 2352
			break;
	}
out:
2353
	ret = persistent_huge_pages(h);
L
Linus Torvalds 已提交
2354
	spin_unlock(&hugetlb_lock);
2355
	return ret;
L
Linus Torvalds 已提交
2356 2357
}

2358 2359 2360 2361 2362 2363 2364 2365 2366 2367
#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];

2368 2369 2370
static struct hstate *kobj_to_node_hstate(struct kobject *kobj, int *nidp);

static struct hstate *kobj_to_hstate(struct kobject *kobj, int *nidp)
2371 2372
{
	int i;
2373

2374
	for (i = 0; i < HUGE_MAX_HSTATE; i++)
2375 2376 2377
		if (hstate_kobjs[i] == kobj) {
			if (nidp)
				*nidp = NUMA_NO_NODE;
2378
			return &hstates[i];
2379 2380 2381
		}

	return kobj_to_node_hstate(kobj, nidp);
2382 2383
}

2384
static ssize_t nr_hugepages_show_common(struct kobject *kobj,
2385 2386
					struct kobj_attribute *attr, char *buf)
{
2387 2388 2389 2390 2391 2392 2393 2394 2395 2396 2397
	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);
2398
}
2399

2400 2401 2402
static ssize_t __nr_hugepages_store_common(bool obey_mempolicy,
					   struct hstate *h, int nid,
					   unsigned long count, size_t len)
2403 2404
{
	int err;
2405
	NODEMASK_ALLOC(nodemask_t, nodes_allowed, GFP_KERNEL | __GFP_NORETRY);
2406

2407
	if (hstate_is_gigantic(h) && !gigantic_page_supported()) {
2408 2409 2410 2411
		err = -EINVAL;
		goto out;
	}

2412 2413 2414 2415 2416 2417 2418
	if (nid == NUMA_NO_NODE) {
		/*
		 * global hstate attribute
		 */
		if (!(obey_mempolicy &&
				init_nodemask_of_mempolicy(nodes_allowed))) {
			NODEMASK_FREE(nodes_allowed);
2419
			nodes_allowed = &node_states[N_MEMORY];
2420 2421 2422 2423 2424 2425 2426 2427 2428
		}
	} 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
2429
		nodes_allowed = &node_states[N_MEMORY];
2430

2431
	h->max_huge_pages = set_max_huge_pages(h, count, nodes_allowed);
2432

2433
	if (nodes_allowed != &node_states[N_MEMORY])
2434 2435 2436
		NODEMASK_FREE(nodes_allowed);

	return len;
2437 2438 2439
out:
	NODEMASK_FREE(nodes_allowed);
	return err;
2440 2441
}

2442 2443 2444 2445 2446 2447 2448 2449 2450 2451 2452 2453 2454 2455 2456 2457 2458
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);
}

2459 2460 2461 2462 2463 2464 2465 2466 2467
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)
{
2468
	return nr_hugepages_store_common(false, kobj, buf, len);
2469 2470 2471
}
HSTATE_ATTR(nr_hugepages);

2472 2473 2474 2475 2476 2477 2478 2479 2480 2481 2482 2483 2484 2485 2486
#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)
{
2487
	return nr_hugepages_store_common(true, kobj, buf, len);
2488 2489 2490 2491 2492
}
HSTATE_ATTR(nr_hugepages_mempolicy);
#endif


2493 2494 2495
static ssize_t nr_overcommit_hugepages_show(struct kobject *kobj,
					struct kobj_attribute *attr, char *buf)
{
2496
	struct hstate *h = kobj_to_hstate(kobj, NULL);
2497 2498
	return sprintf(buf, "%lu\n", h->nr_overcommit_huge_pages);
}
2499

2500 2501 2502 2503 2504
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;
2505
	struct hstate *h = kobj_to_hstate(kobj, NULL);
2506

2507
	if (hstate_is_gigantic(h))
2508 2509
		return -EINVAL;

2510
	err = kstrtoul(buf, 10, &input);
2511
	if (err)
2512
		return err;
2513 2514 2515 2516 2517 2518 2519 2520 2521 2522 2523 2524

	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)
{
2525 2526 2527 2528 2529 2530 2531 2532 2533 2534 2535
	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);
2536 2537 2538 2539 2540 2541
}
HSTATE_ATTR_RO(free_hugepages);

static ssize_t resv_hugepages_show(struct kobject *kobj,
					struct kobj_attribute *attr, char *buf)
{
2542
	struct hstate *h = kobj_to_hstate(kobj, NULL);
2543 2544 2545 2546 2547 2548 2549
	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)
{
2550 2551 2552 2553 2554 2555 2556 2557 2558 2559 2560
	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);
2561 2562 2563 2564 2565 2566 2567 2568 2569
}
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,
2570 2571 2572
#ifdef CONFIG_NUMA
	&nr_hugepages_mempolicy_attr.attr,
#endif
2573 2574 2575
	NULL,
};

2576
static const struct attribute_group hstate_attr_group = {
2577 2578 2579
	.attrs = hstate_attrs,
};

J
Jeff Mahoney 已提交
2580 2581
static int hugetlb_sysfs_add_hstate(struct hstate *h, struct kobject *parent,
				    struct kobject **hstate_kobjs,
2582
				    const struct attribute_group *hstate_attr_group)
2583 2584
{
	int retval;
2585
	int hi = hstate_index(h);
2586

2587 2588
	hstate_kobjs[hi] = kobject_create_and_add(h->name, parent);
	if (!hstate_kobjs[hi])
2589 2590
		return -ENOMEM;

2591
	retval = sysfs_create_group(hstate_kobjs[hi], hstate_attr_group);
2592
	if (retval)
2593
		kobject_put(hstate_kobjs[hi]);
2594 2595 2596 2597 2598 2599 2600 2601 2602 2603 2604 2605 2606 2607

	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) {
2608 2609
		err = hugetlb_sysfs_add_hstate(h, hugepages_kobj,
					 hstate_kobjs, &hstate_attr_group);
2610
		if (err)
2611
			pr_err("Hugetlb: Unable to add hstate %s", h->name);
2612 2613 2614
	}
}

2615 2616 2617 2618
#ifdef CONFIG_NUMA

/*
 * node_hstate/s - associate per node hstate attributes, via their kobjects,
2619 2620 2621
 * 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
2622 2623 2624 2625 2626 2627
 * the base kernel, on the hugetlb module.
 */
struct node_hstate {
	struct kobject		*hugepages_kobj;
	struct kobject		*hstate_kobjs[HUGE_MAX_HSTATE];
};
2628
static struct node_hstate node_hstates[MAX_NUMNODES];
2629 2630

/*
2631
 * A subset of global hstate attributes for node devices
2632 2633 2634 2635 2636 2637 2638 2639
 */
static struct attribute *per_node_hstate_attrs[] = {
	&nr_hugepages_attr.attr,
	&free_hugepages_attr.attr,
	&surplus_hugepages_attr.attr,
	NULL,
};

2640
static const struct attribute_group per_node_hstate_attr_group = {
2641 2642 2643 2644
	.attrs = per_node_hstate_attrs,
};

/*
2645
 * kobj_to_node_hstate - lookup global hstate for node device hstate attr kobj.
2646 2647 2648 2649 2650 2651 2652 2653 2654 2655 2656 2657 2658 2659 2660 2661 2662 2663 2664 2665 2666 2667
 * 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;
}

/*
2668
 * Unregister hstate attributes from a single node device.
2669 2670
 * No-op if no hstate attributes attached.
 */
2671
static void hugetlb_unregister_node(struct node *node)
2672 2673
{
	struct hstate *h;
2674
	struct node_hstate *nhs = &node_hstates[node->dev.id];
2675 2676

	if (!nhs->hugepages_kobj)
2677
		return;		/* no hstate attributes */
2678

2679 2680 2681 2682 2683
	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;
2684
		}
2685
	}
2686 2687 2688 2689 2690 2691 2692

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


/*
2693
 * Register hstate attributes for a single node device.
2694 2695
 * No-op if attributes already registered.
 */
2696
static void hugetlb_register_node(struct node *node)
2697 2698
{
	struct hstate *h;
2699
	struct node_hstate *nhs = &node_hstates[node->dev.id];
2700 2701 2702 2703 2704 2705
	int err;

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

	nhs->hugepages_kobj = kobject_create_and_add("hugepages",
2706
							&node->dev.kobj);
2707 2708 2709 2710 2711 2712 2713 2714
	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) {
2715 2716
			pr_err("Hugetlb: Unable to add hstate %s for node %d\n",
				h->name, node->dev.id);
2717 2718 2719 2720 2721 2722 2723
			hugetlb_unregister_node(node);
			break;
		}
	}
}

/*
2724
 * hugetlb init time:  register hstate attributes for all registered node
2725 2726
 * devices of nodes that have memory.  All on-line nodes should have
 * registered their associated device by this time.
2727
 */
2728
static void __init hugetlb_register_all_nodes(void)
2729 2730 2731
{
	int nid;

2732
	for_each_node_state(nid, N_MEMORY) {
2733
		struct node *node = node_devices[nid];
2734
		if (node->dev.id == nid)
2735 2736 2737 2738
			hugetlb_register_node(node);
	}

	/*
2739
	 * Let the node device driver know we're here so it can
2740 2741 2742 2743 2744 2745 2746 2747 2748 2749 2750 2751 2752 2753 2754 2755 2756 2757 2758
	 * [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_register_all_nodes(void) { }

#endif

2759 2760
static int __init hugetlb_init(void)
{
2761 2762
	int i;

2763
	if (!hugepages_supported())
2764
		return 0;
2765

2766
	if (!size_to_hstate(default_hstate_size)) {
2767 2768 2769 2770 2771
		if (default_hstate_size != 0) {
			pr_err("HugeTLB: unsupported default_hugepagesz %lu. Reverting to %lu\n",
			       default_hstate_size, HPAGE_SIZE);
		}

2772 2773 2774
		default_hstate_size = HPAGE_SIZE;
		if (!size_to_hstate(default_hstate_size))
			hugetlb_add_hstate(HUGETLB_PAGE_ORDER);
2775
	}
2776
	default_hstate_idx = hstate_index(size_to_hstate(default_hstate_size));
2777 2778 2779 2780
	if (default_hstate_max_huge_pages) {
		if (!default_hstate.max_huge_pages)
			default_hstate.max_huge_pages = default_hstate_max_huge_pages;
	}
2781 2782

	hugetlb_init_hstates();
2783
	gather_bootmem_prealloc();
2784 2785 2786
	report_hugepages();

	hugetlb_sysfs_init();
2787
	hugetlb_register_all_nodes();
2788
	hugetlb_cgroup_file_init();
2789

2790 2791 2792 2793 2794
#ifdef CONFIG_SMP
	num_fault_mutexes = roundup_pow_of_two(8 * num_possible_cpus());
#else
	num_fault_mutexes = 1;
#endif
2795
	hugetlb_fault_mutex_table =
2796 2797
		kmalloc_array(num_fault_mutexes, sizeof(struct mutex),
			      GFP_KERNEL);
2798
	BUG_ON(!hugetlb_fault_mutex_table);
2799 2800

	for (i = 0; i < num_fault_mutexes; i++)
2801
		mutex_init(&hugetlb_fault_mutex_table[i]);
2802 2803
	return 0;
}
2804
subsys_initcall(hugetlb_init);
2805 2806

/* Should be called on processing a hugepagesz=... option */
2807 2808 2809 2810 2811
void __init hugetlb_bad_size(void)
{
	parsed_valid_hugepagesz = false;
}

2812
void __init hugetlb_add_hstate(unsigned int order)
2813 2814
{
	struct hstate *h;
2815 2816
	unsigned long i;

2817
	if (size_to_hstate(PAGE_SIZE << order)) {
J
Joe Perches 已提交
2818
		pr_warn("hugepagesz= specified twice, ignoring\n");
2819 2820
		return;
	}
2821
	BUG_ON(hugetlb_max_hstate >= HUGE_MAX_HSTATE);
2822
	BUG_ON(order == 0);
2823
	h = &hstates[hugetlb_max_hstate++];
2824 2825
	h->order = order;
	h->mask = ~((1ULL << (order + PAGE_SHIFT)) - 1);
2826 2827 2828 2829
	h->nr_huge_pages = 0;
	h->free_huge_pages = 0;
	for (i = 0; i < MAX_NUMNODES; ++i)
		INIT_LIST_HEAD(&h->hugepage_freelists[i]);
2830
	INIT_LIST_HEAD(&h->hugepage_activelist);
2831 2832
	h->next_nid_to_alloc = first_memory_node;
	h->next_nid_to_free = first_memory_node;
2833 2834
	snprintf(h->name, HSTATE_NAME_LEN, "hugepages-%lukB",
					huge_page_size(h)/1024);
2835

2836 2837 2838
	parsed_hstate = h;
}

2839
static int __init hugetlb_nrpages_setup(char *s)
2840 2841
{
	unsigned long *mhp;
2842
	static unsigned long *last_mhp;
2843

2844 2845 2846 2847 2848 2849
	if (!parsed_valid_hugepagesz) {
		pr_warn("hugepages = %s preceded by "
			"an unsupported hugepagesz, ignoring\n", s);
		parsed_valid_hugepagesz = true;
		return 1;
	}
2850
	/*
2851
	 * !hugetlb_max_hstate means we haven't parsed a hugepagesz= parameter yet,
2852 2853
	 * so this hugepages= parameter goes to the "default hstate".
	 */
2854
	else if (!hugetlb_max_hstate)
2855 2856 2857 2858
		mhp = &default_hstate_max_huge_pages;
	else
		mhp = &parsed_hstate->max_huge_pages;

2859
	if (mhp == last_mhp) {
J
Joe Perches 已提交
2860
		pr_warn("hugepages= specified twice without interleaving hugepagesz=, ignoring\n");
2861 2862 2863
		return 1;
	}

2864 2865 2866
	if (sscanf(s, "%lu", mhp) <= 0)
		*mhp = 0;

2867 2868 2869 2870 2871
	/*
	 * 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.
	 */
2872
	if (hugetlb_max_hstate && parsed_hstate->order >= MAX_ORDER)
2873 2874 2875 2876
		hugetlb_hstate_alloc_pages(parsed_hstate);

	last_mhp = mhp;

2877 2878
	return 1;
}
2879 2880 2881 2882 2883 2884 2885 2886
__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);
2887

2888 2889 2890 2891 2892 2893 2894 2895 2896 2897 2898 2899
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
2900 2901 2902
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
2903
{
2904
	struct hstate *h = &default_hstate;
2905
	unsigned long tmp = h->max_huge_pages;
2906
	int ret;
2907

2908
	if (!hugepages_supported())
2909
		return -EOPNOTSUPP;
2910

2911 2912
	table->data = &tmp;
	table->maxlen = sizeof(unsigned long);
2913 2914 2915
	ret = proc_doulongvec_minmax(table, write, buffer, length, ppos);
	if (ret)
		goto out;
2916

2917 2918 2919
	if (write)
		ret = __nr_hugepages_store_common(obey_mempolicy, h,
						  NUMA_NO_NODE, tmp, *length);
2920 2921
out:
	return ret;
L
Linus Torvalds 已提交
2922
}
2923

2924 2925 2926 2927 2928 2929 2930 2931 2932 2933 2934 2935 2936 2937 2938 2939 2940
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 */

2941
int hugetlb_overcommit_handler(struct ctl_table *table, int write,
2942
			void __user *buffer,
2943 2944
			size_t *length, loff_t *ppos)
{
2945
	struct hstate *h = &default_hstate;
2946
	unsigned long tmp;
2947
	int ret;
2948

2949
	if (!hugepages_supported())
2950
		return -EOPNOTSUPP;
2951

2952
	tmp = h->nr_overcommit_huge_pages;
2953

2954
	if (write && hstate_is_gigantic(h))
2955 2956
		return -EINVAL;

2957 2958
	table->data = &tmp;
	table->maxlen = sizeof(unsigned long);
2959 2960 2961
	ret = proc_doulongvec_minmax(table, write, buffer, length, ppos);
	if (ret)
		goto out;
2962 2963 2964 2965 2966 2967

	if (write) {
		spin_lock(&hugetlb_lock);
		h->nr_overcommit_huge_pages = tmp;
		spin_unlock(&hugetlb_lock);
	}
2968 2969
out:
	return ret;
2970 2971
}

L
Linus Torvalds 已提交
2972 2973
#endif /* CONFIG_SYSCTL */

2974
void hugetlb_report_meminfo(struct seq_file *m)
L
Linus Torvalds 已提交
2975
{
2976 2977 2978
	struct hstate *h;
	unsigned long total = 0;

2979 2980
	if (!hugepages_supported())
		return;
2981 2982 2983 2984 2985 2986 2987 2988 2989 2990 2991 2992 2993 2994 2995 2996 2997 2998 2999 3000 3001

	for_each_hstate(h) {
		unsigned long count = h->nr_huge_pages;

		total += (PAGE_SIZE << huge_page_order(h)) * count;

		if (h == &default_hstate)
			seq_printf(m,
				   "HugePages_Total:   %5lu\n"
				   "HugePages_Free:    %5lu\n"
				   "HugePages_Rsvd:    %5lu\n"
				   "HugePages_Surp:    %5lu\n"
				   "Hugepagesize:   %8lu kB\n",
				   count,
				   h->free_huge_pages,
				   h->resv_huge_pages,
				   h->surplus_huge_pages,
				   (PAGE_SIZE << huge_page_order(h)) / 1024);
	}

	seq_printf(m, "Hugetlb:        %8lu kB\n", total / 1024);
L
Linus Torvalds 已提交
3002 3003 3004 3005
}

int hugetlb_report_node_meminfo(int nid, char *buf)
{
3006
	struct hstate *h = &default_hstate;
3007 3008
	if (!hugepages_supported())
		return 0;
L
Linus Torvalds 已提交
3009 3010
	return sprintf(buf,
		"Node %d HugePages_Total: %5u\n"
3011 3012
		"Node %d HugePages_Free:  %5u\n"
		"Node %d HugePages_Surp:  %5u\n",
3013 3014 3015
		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
3016 3017
}

3018 3019 3020 3021 3022
void hugetlb_show_meminfo(void)
{
	struct hstate *h;
	int nid;

3023 3024 3025
	if (!hugepages_supported())
		return;

3026 3027 3028 3029 3030 3031 3032 3033 3034 3035
	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));
}

3036 3037 3038 3039 3040 3041
void hugetlb_report_usage(struct seq_file *m, struct mm_struct *mm)
{
	seq_printf(m, "HugetlbPages:\t%8lu kB\n",
		   atomic_long_read(&mm->hugetlb_usage) << (PAGE_SHIFT - 10));
}

L
Linus Torvalds 已提交
3042 3043 3044
/* Return the number pages of memory we physically have, in PAGE_SIZE units. */
unsigned long hugetlb_total_pages(void)
{
3045 3046 3047 3048 3049 3050
	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
3051 3052
}

3053
static int hugetlb_acct_memory(struct hstate *h, long delta)
M
Mel Gorman 已提交
3054 3055 3056 3057 3058 3059 3060 3061 3062 3063 3064 3065 3066 3067 3068 3069 3070 3071 3072 3073 3074 3075
{
	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) {
3076
		if (gather_surplus_pages(h, delta) < 0)
M
Mel Gorman 已提交
3077 3078
			goto out;

3079 3080
		if (delta > cpuset_mems_nr(h->free_huge_pages_node)) {
			return_unused_surplus_pages(h, delta);
M
Mel Gorman 已提交
3081 3082 3083 3084 3085 3086
			goto out;
		}
	}

	ret = 0;
	if (delta < 0)
3087
		return_unused_surplus_pages(h, (unsigned long) -delta);
M
Mel Gorman 已提交
3088 3089 3090 3091 3092 3093

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

3094 3095
static void hugetlb_vm_op_open(struct vm_area_struct *vma)
{
3096
	struct resv_map *resv = vma_resv_map(vma);
3097 3098 3099 3100 3101

	/*
	 * 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
3102
	 * has a reference to the reservation map it cannot disappear until
3103 3104 3105
	 * after this open call completes.  It is therefore safe to take a
	 * new reference here without additional locking.
	 */
3106
	if (resv && is_vma_resv_set(vma, HPAGE_RESV_OWNER))
3107
		kref_get(&resv->refs);
3108 3109
}

3110 3111
static void hugetlb_vm_op_close(struct vm_area_struct *vma)
{
3112
	struct hstate *h = hstate_vma(vma);
3113
	struct resv_map *resv = vma_resv_map(vma);
3114
	struct hugepage_subpool *spool = subpool_vma(vma);
3115
	unsigned long reserve, start, end;
3116
	long gbl_reserve;
3117

3118 3119
	if (!resv || !is_vma_resv_set(vma, HPAGE_RESV_OWNER))
		return;
3120

3121 3122
	start = vma_hugecache_offset(h, vma, vma->vm_start);
	end = vma_hugecache_offset(h, vma, vma->vm_end);
3123

3124
	reserve = (end - start) - region_count(resv, start, end);
3125

3126 3127 3128
	kref_put(&resv->refs, resv_map_release);

	if (reserve) {
3129 3130 3131 3132 3133 3134
		/*
		 * 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);
3135
	}
3136 3137
}

3138 3139 3140 3141 3142 3143 3144
static int hugetlb_vm_op_split(struct vm_area_struct *vma, unsigned long addr)
{
	if (addr & ~(huge_page_mask(hstate_vma(vma))))
		return -EINVAL;
	return 0;
}

3145 3146 3147 3148 3149 3150 3151
static unsigned long hugetlb_vm_op_pagesize(struct vm_area_struct *vma)
{
	struct hstate *hstate = hstate_vma(vma);

	return 1UL << huge_page_shift(hstate);
}

L
Linus Torvalds 已提交
3152 3153 3154 3155 3156 3157
/*
 * 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.
 */
3158
static vm_fault_t hugetlb_vm_op_fault(struct vm_fault *vmf)
L
Linus Torvalds 已提交
3159 3160
{
	BUG();
N
Nick Piggin 已提交
3161
	return 0;
L
Linus Torvalds 已提交
3162 3163
}

3164 3165 3166 3167 3168 3169 3170
/*
 * When a new function is introduced to vm_operations_struct and added
 * to hugetlb_vm_ops, please consider adding the function to shm_vm_ops.
 * This is because under System V memory model, mappings created via
 * shmget/shmat with "huge page" specified are backed by hugetlbfs files,
 * their original vm_ops are overwritten with shm_vm_ops.
 */
3171
const struct vm_operations_struct hugetlb_vm_ops = {
N
Nick Piggin 已提交
3172
	.fault = hugetlb_vm_op_fault,
3173
	.open = hugetlb_vm_op_open,
3174
	.close = hugetlb_vm_op_close,
3175
	.split = hugetlb_vm_op_split,
3176
	.pagesize = hugetlb_vm_op_pagesize,
L
Linus Torvalds 已提交
3177 3178
};

3179 3180
static pte_t make_huge_pte(struct vm_area_struct *vma, struct page *page,
				int writable)
D
David Gibson 已提交
3181 3182 3183
{
	pte_t entry;

3184
	if (writable) {
3185 3186
		entry = huge_pte_mkwrite(huge_pte_mkdirty(mk_huge_pte(page,
					 vma->vm_page_prot)));
D
David Gibson 已提交
3187
	} else {
3188 3189
		entry = huge_pte_wrprotect(mk_huge_pte(page,
					   vma->vm_page_prot));
D
David Gibson 已提交
3190 3191 3192
	}
	entry = pte_mkyoung(entry);
	entry = pte_mkhuge(entry);
3193
	entry = arch_make_huge_pte(entry, vma, page, writable);
D
David Gibson 已提交
3194 3195 3196 3197

	return entry;
}

3198 3199 3200 3201 3202
static void set_huge_ptep_writable(struct vm_area_struct *vma,
				   unsigned long address, pte_t *ptep)
{
	pte_t entry;

3203
	entry = huge_pte_mkwrite(huge_pte_mkdirty(huge_ptep_get(ptep)));
3204
	if (huge_ptep_set_access_flags(vma, address, ptep, entry, 1))
3205
		update_mmu_cache(vma, address, ptep);
3206 3207
}

3208
bool is_hugetlb_entry_migration(pte_t pte)
3209 3210 3211 3212
{
	swp_entry_t swp;

	if (huge_pte_none(pte) || pte_present(pte))
3213
		return false;
3214 3215
	swp = pte_to_swp_entry(pte);
	if (non_swap_entry(swp) && is_migration_entry(swp))
3216
		return true;
3217
	else
3218
		return false;
3219 3220 3221 3222 3223 3224 3225 3226 3227 3228 3229 3230 3231 3232
}

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;
}
3233

D
David Gibson 已提交
3234 3235 3236
int copy_hugetlb_page_range(struct mm_struct *dst, struct mm_struct *src,
			    struct vm_area_struct *vma)
{
3237
	pte_t *src_pte, *dst_pte, entry, dst_entry;
D
David Gibson 已提交
3238
	struct page *ptepage;
3239
	unsigned long addr;
3240
	int cow;
3241 3242
	struct hstate *h = hstate_vma(vma);
	unsigned long sz = huge_page_size(h);
3243
	struct mmu_notifier_range range;
3244
	int ret = 0;
3245 3246

	cow = (vma->vm_flags & (VM_SHARED | VM_MAYWRITE)) == VM_MAYWRITE;
D
David Gibson 已提交
3247

3248 3249 3250 3251 3252
	if (cow) {
		mmu_notifier_range_init(&range, src, vma->vm_start,
					vma->vm_end);
		mmu_notifier_invalidate_range_start(&range);
	}
3253

3254
	for (addr = vma->vm_start; addr < vma->vm_end; addr += sz) {
3255
		spinlock_t *src_ptl, *dst_ptl;
3256
		src_pte = huge_pte_offset(src, addr, sz);
H
Hugh Dickins 已提交
3257 3258
		if (!src_pte)
			continue;
3259
		dst_pte = huge_pte_alloc(dst, addr, sz);
3260 3261 3262 3263
		if (!dst_pte) {
			ret = -ENOMEM;
			break;
		}
3264

3265 3266 3267 3268 3269 3270 3271 3272 3273 3274 3275
		/*
		 * If the pagetables are shared don't copy or take references.
		 * dst_pte == src_pte is the common case of src/dest sharing.
		 *
		 * However, src could have 'unshared' and dst shares with
		 * another vma.  If dst_pte !none, this implies sharing.
		 * Check here before taking page table lock, and once again
		 * after taking the lock below.
		 */
		dst_entry = huge_ptep_get(dst_pte);
		if ((dst_pte == src_pte) || !huge_pte_none(dst_entry))
3276 3277
			continue;

3278 3279 3280
		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);
3281
		entry = huge_ptep_get(src_pte);
3282 3283 3284 3285 3286 3287 3288
		dst_entry = huge_ptep_get(dst_pte);
		if (huge_pte_none(entry) || !huge_pte_none(dst_entry)) {
			/*
			 * Skip if src entry none.  Also, skip in the
			 * unlikely case dst entry !none as this implies
			 * sharing with another vma.
			 */
3289 3290 3291 3292 3293 3294 3295 3296 3297 3298 3299 3300
			;
		} 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);
3301 3302
				set_huge_swap_pte_at(src, addr, src_pte,
						     entry, sz);
3303
			}
3304
			set_huge_swap_pte_at(dst, addr, dst_pte, entry, sz);
3305
		} else {
3306
			if (cow) {
3307 3308 3309 3310 3311
				/*
				 * No need to notify as we are downgrading page
				 * table protection not changing it to point
				 * to a new page.
				 *
3312
				 * See Documentation/vm/mmu_notifier.rst
3313
				 */
3314
				huge_ptep_set_wrprotect(src, addr, src_pte);
3315
			}
3316
			entry = huge_ptep_get(src_pte);
3317 3318
			ptepage = pte_page(entry);
			get_page(ptepage);
3319
			page_dup_rmap(ptepage, true);
3320
			set_huge_pte_at(dst, addr, dst_pte, entry);
3321
			hugetlb_count_add(pages_per_huge_page(h), dst);
3322
		}
3323 3324
		spin_unlock(src_ptl);
		spin_unlock(dst_ptl);
D
David Gibson 已提交
3325 3326
	}

3327
	if (cow)
3328
		mmu_notifier_invalidate_range_end(&range);
3329 3330

	return ret;
D
David Gibson 已提交
3331 3332
}

3333 3334 3335
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
3336 3337 3338
{
	struct mm_struct *mm = vma->vm_mm;
	unsigned long address;
3339
	pte_t *ptep;
D
David Gibson 已提交
3340
	pte_t pte;
3341
	spinlock_t *ptl;
D
David Gibson 已提交
3342
	struct page *page;
3343 3344
	struct hstate *h = hstate_vma(vma);
	unsigned long sz = huge_page_size(h);
3345
	struct mmu_notifier_range range;
3346

D
David Gibson 已提交
3347
	WARN_ON(!is_vm_hugetlb_page(vma));
3348 3349
	BUG_ON(start & ~huge_page_mask(h));
	BUG_ON(end & ~huge_page_mask(h));
D
David Gibson 已提交
3350

3351 3352 3353 3354 3355
	/*
	 * This is a hugetlb vma, all the pte entries should point
	 * to huge page.
	 */
	tlb_remove_check_page_size_change(tlb, sz);
3356
	tlb_start_vma(tlb, vma);
3357 3358 3359 3360

	/*
	 * If sharing possible, alert mmu notifiers of worst case.
	 */
3361 3362 3363
	mmu_notifier_range_init(&range, mm, start, end);
	adjust_range_if_pmd_sharing_possible(vma, &range.start, &range.end);
	mmu_notifier_invalidate_range_start(&range);
3364 3365
	address = start;
	for (; address < end; address += sz) {
3366
		ptep = huge_pte_offset(mm, address, sz);
A
Adam Litke 已提交
3367
		if (!ptep)
3368 3369
			continue;

3370
		ptl = huge_pte_lock(h, mm, ptep);
3371 3372
		if (huge_pmd_unshare(mm, &address, ptep)) {
			spin_unlock(ptl);
3373 3374 3375 3376
			/*
			 * We just unmapped a page of PMDs by clearing a PUD.
			 * The caller's TLB flush range should cover this area.
			 */
3377 3378
			continue;
		}
3379

3380
		pte = huge_ptep_get(ptep);
3381 3382 3383 3384
		if (huge_pte_none(pte)) {
			spin_unlock(ptl);
			continue;
		}
3385 3386

		/*
3387 3388
		 * Migrating hugepage or HWPoisoned hugepage is already
		 * unmapped and its refcount is dropped, so just clear pte here.
3389
		 */
3390
		if (unlikely(!pte_present(pte))) {
3391
			huge_pte_clear(mm, address, ptep, sz);
3392 3393
			spin_unlock(ptl);
			continue;
3394
		}
3395 3396

		page = pte_page(pte);
3397 3398 3399 3400 3401 3402
		/*
		 * 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) {
3403 3404 3405 3406
			if (page != ref_page) {
				spin_unlock(ptl);
				continue;
			}
3407 3408 3409 3410 3411 3412 3413 3414
			/*
			 * 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);
		}

3415
		pte = huge_ptep_get_and_clear(mm, address, ptep);
3416
		tlb_remove_huge_tlb_entry(h, tlb, ptep, address);
3417
		if (huge_pte_dirty(pte))
3418
			set_page_dirty(page);
3419

3420
		hugetlb_count_sub(pages_per_huge_page(h), mm);
3421
		page_remove_rmap(page, true);
3422

3423
		spin_unlock(ptl);
3424
		tlb_remove_page_size(tlb, page, huge_page_size(h));
3425 3426 3427 3428 3429
		/*
		 * Bail out after unmapping reference page if supplied
		 */
		if (ref_page)
			break;
3430
	}
3431
	mmu_notifier_invalidate_range_end(&range);
3432
	tlb_end_vma(tlb, vma);
L
Linus Torvalds 已提交
3433
}
D
David Gibson 已提交
3434

3435 3436 3437 3438 3439 3440 3441 3442 3443 3444 3445 3446
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
3447
	 * is to clear it before releasing the i_mmap_rwsem. This works
3448
	 * because in the context this is called, the VMA is about to be
3449
	 * destroyed and the i_mmap_rwsem is held.
3450 3451 3452 3453
	 */
	vma->vm_flags &= ~VM_MAYSHARE;
}

3454
void unmap_hugepage_range(struct vm_area_struct *vma, unsigned long start,
3455
			  unsigned long end, struct page *ref_page)
3456
{
3457 3458
	struct mm_struct *mm;
	struct mmu_gather tlb;
3459 3460 3461 3462 3463 3464 3465 3466 3467 3468 3469
	unsigned long tlb_start = start;
	unsigned long tlb_end = end;

	/*
	 * If shared PMDs were possibly used within this vma range, adjust
	 * start/end for worst case tlb flushing.
	 * Note that we can not be sure if PMDs are shared until we try to
	 * unmap pages.  However, we want to make sure TLB flushing covers
	 * the largest possible range.
	 */
	adjust_range_if_pmd_sharing_possible(vma, &tlb_start, &tlb_end);
3470 3471 3472

	mm = vma->vm_mm;

3473
	tlb_gather_mmu(&tlb, mm, tlb_start, tlb_end);
3474
	__unmap_hugepage_range(&tlb, vma, start, end, ref_page);
3475
	tlb_finish_mmu(&tlb, tlb_start, tlb_end);
3476 3477
}

3478 3479 3480 3481 3482 3483
/*
 * 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.
 */
3484 3485
static void unmap_ref_private(struct mm_struct *mm, struct vm_area_struct *vma,
			      struct page *page, unsigned long address)
3486
{
3487
	struct hstate *h = hstate_vma(vma);
3488 3489 3490 3491 3492 3493 3494 3495
	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.
	 */
3496
	address = address & huge_page_mask(h);
3497 3498
	pgoff = ((address - vma->vm_start) >> PAGE_SHIFT) +
			vma->vm_pgoff;
3499
	mapping = vma->vm_file->f_mapping;
3500

3501 3502 3503 3504 3505
	/*
	 * 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
	 */
3506
	i_mmap_lock_write(mapping);
3507
	vma_interval_tree_foreach(iter_vma, &mapping->i_mmap, pgoff, pgoff) {
3508 3509 3510 3511
		/* Do not unmap the current VMA */
		if (iter_vma == vma)
			continue;

3512 3513 3514 3515 3516 3517 3518 3519
		/*
		 * 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;

3520 3521 3522 3523 3524 3525 3526 3527
		/*
		 * 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))
3528 3529
			unmap_hugepage_range(iter_vma, address,
					     address + huge_page_size(h), page);
3530
	}
3531
	i_mmap_unlock_write(mapping);
3532 3533
}

3534 3535
/*
 * Hugetlb_cow() should be called with page lock of the original hugepage held.
3536 3537 3538
 * 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.
3539
 */
3540
static vm_fault_t hugetlb_cow(struct mm_struct *mm, struct vm_area_struct *vma,
3541
		       unsigned long address, pte_t *ptep,
3542
		       struct page *pagecache_page, spinlock_t *ptl)
3543
{
3544
	pte_t pte;
3545
	struct hstate *h = hstate_vma(vma);
3546
	struct page *old_page, *new_page;
3547 3548
	int outside_reserve = 0;
	vm_fault_t ret = 0;
3549
	unsigned long haddr = address & huge_page_mask(h);
3550
	struct mmu_notifier_range range;
3551

3552
	pte = huge_ptep_get(ptep);
3553 3554
	old_page = pte_page(pte);

3555
retry_avoidcopy:
3556 3557
	/* If no-one else is actually using this page, avoid the copy
	 * and just make the page writable */
3558
	if (page_mapcount(old_page) == 1 && PageAnon(old_page)) {
3559
		page_move_anon_rmap(old_page, vma);
3560
		set_huge_ptep_writable(vma, haddr, ptep);
N
Nick Piggin 已提交
3561
		return 0;
3562 3563
	}

3564 3565 3566 3567 3568 3569 3570 3571 3572
	/*
	 * 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.
	 */
3573
	if (is_vma_resv_set(vma, HPAGE_RESV_OWNER) &&
3574 3575 3576
			old_page != pagecache_page)
		outside_reserve = 1;

3577
	get_page(old_page);
3578

3579 3580 3581 3582
	/*
	 * Drop page table lock as buddy allocator may be called. It will
	 * be acquired again before returning to the caller, as expected.
	 */
3583
	spin_unlock(ptl);
3584
	new_page = alloc_huge_page(vma, haddr, outside_reserve);
3585

3586
	if (IS_ERR(new_page)) {
3587 3588 3589 3590 3591 3592 3593 3594
		/*
		 * 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) {
3595
			put_page(old_page);
3596
			BUG_ON(huge_pte_none(pte));
3597
			unmap_ref_private(mm, vma, old_page, haddr);
3598 3599
			BUG_ON(huge_pte_none(pte));
			spin_lock(ptl);
3600
			ptep = huge_pte_offset(mm, haddr, huge_page_size(h));
3601 3602 3603 3604 3605 3606 3607 3608
			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;
3609 3610
		}

3611
		ret = vmf_error(PTR_ERR(new_page));
3612
		goto out_release_old;
3613 3614
	}

3615 3616 3617 3618
	/*
	 * When the original hugepage is shared one, it does not have
	 * anon_vma prepared.
	 */
3619
	if (unlikely(anon_vma_prepare(vma))) {
3620 3621
		ret = VM_FAULT_OOM;
		goto out_release_all;
3622
	}
3623

3624
	copy_user_huge_page(new_page, old_page, address, vma,
A
Andrea Arcangeli 已提交
3625
			    pages_per_huge_page(h));
N
Nick Piggin 已提交
3626
	__SetPageUptodate(new_page);
3627

3628 3629
	mmu_notifier_range_init(&range, mm, haddr, haddr + huge_page_size(h));
	mmu_notifier_invalidate_range_start(&range);
3630

3631
	/*
3632
	 * Retake the page table lock to check for racing updates
3633 3634
	 * before the page tables are altered
	 */
3635
	spin_lock(ptl);
3636
	ptep = huge_pte_offset(mm, haddr, huge_page_size(h));
3637
	if (likely(ptep && pte_same(huge_ptep_get(ptep), pte))) {
3638 3639
		ClearPagePrivate(new_page);

3640
		/* Break COW */
3641
		huge_ptep_clear_flush(vma, haddr, ptep);
3642
		mmu_notifier_invalidate_range(mm, range.start, range.end);
3643
		set_huge_pte_at(mm, haddr, ptep,
3644
				make_huge_pte(vma, new_page, 1));
3645
		page_remove_rmap(old_page, true);
3646
		hugepage_add_new_anon_rmap(new_page, vma, haddr);
3647
		set_page_huge_active(new_page);
3648 3649 3650
		/* Make the old page be freed below */
		new_page = old_page;
	}
3651
	spin_unlock(ptl);
3652
	mmu_notifier_invalidate_range_end(&range);
3653
out_release_all:
3654
	restore_reserve_on_error(h, vma, haddr, new_page);
3655
	put_page(new_page);
3656
out_release_old:
3657
	put_page(old_page);
3658

3659 3660
	spin_lock(ptl); /* Caller expects lock to be held */
	return ret;
3661 3662
}

3663
/* Return the pagecache page at a given address within a VMA */
3664 3665
static struct page *hugetlbfs_pagecache_page(struct hstate *h,
			struct vm_area_struct *vma, unsigned long address)
3666 3667
{
	struct address_space *mapping;
3668
	pgoff_t idx;
3669 3670

	mapping = vma->vm_file->f_mapping;
3671
	idx = vma_hugecache_offset(h, vma, address);
3672 3673 3674 3675

	return find_lock_page(mapping, idx);
}

H
Hugh Dickins 已提交
3676 3677 3678 3679 3680
/*
 * 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
3681 3682 3683 3684 3685 3686 3687 3688 3689 3690 3691 3692 3693 3694 3695
			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;
}

3696 3697 3698 3699 3700 3701 3702 3703 3704 3705 3706
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);

3707 3708 3709 3710 3711 3712
	/*
	 * set page dirty so that it will not be removed from cache/file
	 * by non-hugetlbfs specific code paths.
	 */
	set_page_dirty(page);

3713 3714 3715 3716 3717 3718
	spin_lock(&inode->i_lock);
	inode->i_blocks += blocks_per_huge_page(h);
	spin_unlock(&inode->i_lock);
	return 0;
}

3719 3720 3721 3722
static vm_fault_t hugetlb_no_page(struct mm_struct *mm,
			struct vm_area_struct *vma,
			struct address_space *mapping, pgoff_t idx,
			unsigned long address, pte_t *ptep, unsigned int flags)
3723
{
3724
	struct hstate *h = hstate_vma(vma);
3725
	vm_fault_t ret = VM_FAULT_SIGBUS;
3726
	int anon_rmap = 0;
A
Adam Litke 已提交
3727 3728
	unsigned long size;
	struct page *page;
3729
	pte_t new_pte;
3730
	spinlock_t *ptl;
3731
	unsigned long haddr = address & huge_page_mask(h);
3732
	bool new_page = false;
A
Adam Litke 已提交
3733

3734 3735 3736
	/*
	 * 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
3737
	 * COW. Warn that such a situation has occurred as it may not be obvious
3738 3739
	 */
	if (is_vma_resv_set(vma, HPAGE_RESV_UNMAPPED)) {
3740
		pr_warn_ratelimited("PID %d killed due to inadequate hugepage pool\n",
3741
			   current->pid);
3742 3743 3744
		return ret;
	}

A
Adam Litke 已提交
3745
	/*
3746 3747
	 * Use page lock to guard against racing truncation
	 * before we get page_table_lock.
A
Adam Litke 已提交
3748
	 */
3749 3750 3751
retry:
	page = find_lock_page(mapping, idx);
	if (!page) {
3752 3753 3754 3755
		size = i_size_read(mapping->host) >> huge_page_shift(h);
		if (idx >= size)
			goto out;

3756 3757 3758 3759 3760 3761 3762
		/*
		 * Check for page in userfault range
		 */
		if (userfaultfd_missing(vma)) {
			u32 hash;
			struct vm_fault vmf = {
				.vma = vma,
3763
				.address = haddr,
3764 3765 3766 3767 3768 3769 3770 3771 3772 3773 3774
				.flags = flags,
				/*
				 * Hard to debug if it ends up being
				 * used by a callee that assumes
				 * something about the other
				 * uninitialized fields... same as in
				 * memory.c
				 */
			};

			/*
3775 3776 3777
			 * hugetlb_fault_mutex must be dropped before
			 * handling userfault.  Reacquire after handling
			 * fault to make calling code simpler.
3778 3779
			 */
			hash = hugetlb_fault_mutex_hash(h, mm, vma, mapping,
3780
							idx, haddr);
3781 3782 3783 3784 3785 3786
			mutex_unlock(&hugetlb_fault_mutex_table[hash]);
			ret = handle_userfault(&vmf, VM_UFFD_MISSING);
			mutex_lock(&hugetlb_fault_mutex_table[hash]);
			goto out;
		}

3787
		page = alloc_huge_page(vma, haddr, 0);
3788
		if (IS_ERR(page)) {
3789
			ret = vmf_error(PTR_ERR(page));
3790 3791
			goto out;
		}
A
Andrea Arcangeli 已提交
3792
		clear_huge_page(page, address, pages_per_huge_page(h));
N
Nick Piggin 已提交
3793
		__SetPageUptodate(page);
3794
		new_page = true;
3795

3796
		if (vma->vm_flags & VM_MAYSHARE) {
3797
			int err = huge_add_to_page_cache(page, mapping, idx);
3798 3799 3800 3801 3802 3803
			if (err) {
				put_page(page);
				if (err == -EEXIST)
					goto retry;
				goto out;
			}
3804
		} else {
3805
			lock_page(page);
3806 3807 3808 3809
			if (unlikely(anon_vma_prepare(vma))) {
				ret = VM_FAULT_OOM;
				goto backout_unlocked;
			}
3810
			anon_rmap = 1;
3811
		}
3812
	} else {
3813 3814 3815 3816 3817 3818
		/*
		 * 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))) {
3819
			ret = VM_FAULT_HWPOISON |
3820
				VM_FAULT_SET_HINDEX(hstate_index(h));
3821 3822
			goto backout_unlocked;
		}
3823
	}
3824

3825 3826 3827 3828 3829 3830
	/*
	 * 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.
	 */
3831
	if ((flags & FAULT_FLAG_WRITE) && !(vma->vm_flags & VM_SHARED)) {
3832
		if (vma_needs_reservation(h, vma, haddr) < 0) {
3833 3834 3835
			ret = VM_FAULT_OOM;
			goto backout_unlocked;
		}
3836
		/* Just decrements count, does not deallocate */
3837
		vma_end_reservation(h, vma, haddr);
3838
	}
3839

3840
	ptl = huge_pte_lock(h, mm, ptep);
3841 3842 3843
	size = i_size_read(mapping->host) >> huge_page_shift(h);
	if (idx >= size)
		goto backout;
A
Adam Litke 已提交
3844

N
Nick Piggin 已提交
3845
	ret = 0;
3846
	if (!huge_pte_none(huge_ptep_get(ptep)))
A
Adam Litke 已提交
3847 3848
		goto backout;

3849 3850
	if (anon_rmap) {
		ClearPagePrivate(page);
3851
		hugepage_add_new_anon_rmap(page, vma, haddr);
3852
	} else
3853
		page_dup_rmap(page, true);
3854 3855
	new_pte = make_huge_pte(vma, page, ((vma->vm_flags & VM_WRITE)
				&& (vma->vm_flags & VM_SHARED)));
3856
	set_huge_pte_at(mm, haddr, ptep, new_pte);
3857

3858
	hugetlb_count_add(pages_per_huge_page(h), mm);
3859
	if ((flags & FAULT_FLAG_WRITE) && !(vma->vm_flags & VM_SHARED)) {
3860
		/* Optimization, do the COW without a second fault */
3861
		ret = hugetlb_cow(mm, vma, address, ptep, page, ptl);
3862 3863
	}

3864
	spin_unlock(ptl);
3865 3866 3867 3868 3869 3870 3871 3872 3873

	/*
	 * Only make newly allocated pages active.  Existing pages found
	 * in the pagecache could be !page_huge_active() if they have been
	 * isolated for migration.
	 */
	if (new_page)
		set_page_huge_active(page);

A
Adam Litke 已提交
3874 3875
	unlock_page(page);
out:
3876
	return ret;
A
Adam Litke 已提交
3877 3878

backout:
3879
	spin_unlock(ptl);
3880
backout_unlocked:
A
Adam Litke 已提交
3881
	unlock_page(page);
3882
	restore_reserve_on_error(h, vma, haddr, page);
A
Adam Litke 已提交
3883 3884
	put_page(page);
	goto out;
3885 3886
}

3887
#ifdef CONFIG_SMP
3888
u32 hugetlb_fault_mutex_hash(struct hstate *h, struct mm_struct *mm,
3889 3890 3891 3892 3893 3894 3895 3896 3897 3898 3899 3900 3901 3902 3903 3904 3905 3906 3907 3908 3909 3910 3911 3912
			    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.
 */
3913
u32 hugetlb_fault_mutex_hash(struct hstate *h, struct mm_struct *mm,
3914 3915 3916 3917 3918 3919 3920 3921
			    struct vm_area_struct *vma,
			    struct address_space *mapping,
			    pgoff_t idx, unsigned long address)
{
	return 0;
}
#endif

3922
vm_fault_t hugetlb_fault(struct mm_struct *mm, struct vm_area_struct *vma,
3923
			unsigned long address, unsigned int flags)
3924
{
3925
	pte_t *ptep, entry;
3926
	spinlock_t *ptl;
3927
	vm_fault_t ret;
3928 3929
	u32 hash;
	pgoff_t idx;
3930
	struct page *page = NULL;
3931
	struct page *pagecache_page = NULL;
3932
	struct hstate *h = hstate_vma(vma);
3933
	struct address_space *mapping;
3934
	int need_wait_lock = 0;
3935
	unsigned long haddr = address & huge_page_mask(h);
3936

3937
	ptep = huge_pte_offset(mm, haddr, huge_page_size(h));
3938 3939
	if (ptep) {
		entry = huge_ptep_get(ptep);
N
Naoya Horiguchi 已提交
3940
		if (unlikely(is_hugetlb_entry_migration(entry))) {
3941
			migration_entry_wait_huge(vma, mm, ptep);
N
Naoya Horiguchi 已提交
3942 3943
			return 0;
		} else if (unlikely(is_hugetlb_entry_hwpoisoned(entry)))
3944
			return VM_FAULT_HWPOISON_LARGE |
3945
				VM_FAULT_SET_HINDEX(hstate_index(h));
3946 3947 3948 3949
	} else {
		ptep = huge_pte_alloc(mm, haddr, huge_page_size(h));
		if (!ptep)
			return VM_FAULT_OOM;
3950 3951
	}

3952
	mapping = vma->vm_file->f_mapping;
3953
	idx = vma_hugecache_offset(h, vma, haddr);
3954

3955 3956 3957 3958 3959
	/*
	 * 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.
	 */
3960
	hash = hugetlb_fault_mutex_hash(h, mm, vma, mapping, idx, haddr);
3961
	mutex_lock(&hugetlb_fault_mutex_table[hash]);
3962

3963 3964
	entry = huge_ptep_get(ptep);
	if (huge_pte_none(entry)) {
3965
		ret = hugetlb_no_page(mm, vma, mapping, idx, address, ptep, flags);
3966
		goto out_mutex;
3967
	}
3968

N
Nick Piggin 已提交
3969
	ret = 0;
3970

3971 3972 3973 3974 3975 3976 3977 3978 3979 3980
	/*
	 * 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;

3981 3982 3983 3984 3985 3986 3987 3988
	/*
	 * 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.
	 */
3989
	if ((flags & FAULT_FLAG_WRITE) && !huge_pte_write(entry)) {
3990
		if (vma_needs_reservation(h, vma, haddr) < 0) {
3991
			ret = VM_FAULT_OOM;
3992
			goto out_mutex;
3993
		}
3994
		/* Just decrements count, does not deallocate */
3995
		vma_end_reservation(h, vma, haddr);
3996

3997
		if (!(vma->vm_flags & VM_MAYSHARE))
3998
			pagecache_page = hugetlbfs_pagecache_page(h,
3999
								vma, haddr);
4000 4001
	}

4002 4003 4004 4005 4006 4007
	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;

4008 4009 4010 4011 4012 4013 4014
	/*
	 * 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)
4015 4016 4017 4018
		if (!trylock_page(page)) {
			need_wait_lock = 1;
			goto out_ptl;
		}
4019

4020
	get_page(page);
4021

4022
	if (flags & FAULT_FLAG_WRITE) {
4023
		if (!huge_pte_write(entry)) {
4024
			ret = hugetlb_cow(mm, vma, address, ptep,
4025
					  pagecache_page, ptl);
4026
			goto out_put_page;
4027
		}
4028
		entry = huge_pte_mkdirty(entry);
4029 4030
	}
	entry = pte_mkyoung(entry);
4031
	if (huge_ptep_set_access_flags(vma, haddr, ptep, entry,
4032
						flags & FAULT_FLAG_WRITE))
4033
		update_mmu_cache(vma, haddr, ptep);
4034 4035 4036 4037
out_put_page:
	if (page != pagecache_page)
		unlock_page(page);
	put_page(page);
4038 4039
out_ptl:
	spin_unlock(ptl);
4040 4041 4042 4043 4044

	if (pagecache_page) {
		unlock_page(pagecache_page);
		put_page(pagecache_page);
	}
4045
out_mutex:
4046
	mutex_unlock(&hugetlb_fault_mutex_table[hash]);
4047 4048 4049 4050 4051 4052 4053 4054 4055
	/*
	 * 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);
4056
	return ret;
4057 4058
}

4059 4060 4061 4062 4063 4064 4065 4066 4067 4068 4069
/*
 * Used by userfaultfd UFFDIO_COPY.  Based on mcopy_atomic_pte with
 * modifications for huge pages.
 */
int hugetlb_mcopy_atomic_pte(struct mm_struct *dst_mm,
			    pte_t *dst_pte,
			    struct vm_area_struct *dst_vma,
			    unsigned long dst_addr,
			    unsigned long src_addr,
			    struct page **pagep)
{
4070 4071 4072
	struct address_space *mapping;
	pgoff_t idx;
	unsigned long size;
4073
	int vm_shared = dst_vma->vm_flags & VM_SHARED;
4074 4075 4076 4077 4078 4079 4080 4081 4082 4083 4084 4085 4086 4087
	struct hstate *h = hstate_vma(dst_vma);
	pte_t _dst_pte;
	spinlock_t *ptl;
	int ret;
	struct page *page;

	if (!*pagep) {
		ret = -ENOMEM;
		page = alloc_huge_page(dst_vma, dst_addr, 0);
		if (IS_ERR(page))
			goto out;

		ret = copy_huge_page_from_user(page,
						(const void __user *) src_addr,
4088
						pages_per_huge_page(h), false);
4089 4090 4091

		/* fallback to copy_from_user outside mmap_sem */
		if (unlikely(ret)) {
4092
			ret = -ENOENT;
4093 4094 4095 4096 4097 4098 4099 4100 4101 4102 4103 4104 4105 4106 4107 4108
			*pagep = page;
			/* don't free the page */
			goto out;
		}
	} else {
		page = *pagep;
		*pagep = NULL;
	}

	/*
	 * The memory barrier inside __SetPageUptodate makes sure that
	 * preceding stores to the page contents become visible before
	 * the set_pte_at() write.
	 */
	__SetPageUptodate(page);

4109 4110 4111
	mapping = dst_vma->vm_file->f_mapping;
	idx = vma_hugecache_offset(h, dst_vma, dst_addr);

4112 4113 4114 4115
	/*
	 * If shared, add to page cache
	 */
	if (vm_shared) {
4116 4117 4118 4119
		size = i_size_read(mapping->host) >> huge_page_shift(h);
		ret = -EFAULT;
		if (idx >= size)
			goto out_release_nounlock;
4120

4121 4122 4123 4124 4125 4126
		/*
		 * Serialization between remove_inode_hugepages() and
		 * huge_add_to_page_cache() below happens through the
		 * hugetlb_fault_mutex_table that here must be hold by
		 * the caller.
		 */
4127 4128 4129 4130 4131
		ret = huge_add_to_page_cache(page, mapping, idx);
		if (ret)
			goto out_release_nounlock;
	}

4132 4133 4134
	ptl = huge_pte_lockptr(h, dst_mm, dst_pte);
	spin_lock(ptl);

4135 4136 4137 4138 4139 4140 4141 4142 4143 4144 4145 4146 4147 4148
	/*
	 * Recheck the i_size after holding PT lock to make sure not
	 * to leave any page mapped (as page_mapped()) beyond the end
	 * of the i_size (remove_inode_hugepages() is strict about
	 * enforcing that). If we bail out here, we'll also leave a
	 * page in the radix tree in the vm_shared case beyond the end
	 * of the i_size, but remove_inode_hugepages() will take care
	 * of it as soon as we drop the hugetlb_fault_mutex_table.
	 */
	size = i_size_read(mapping->host) >> huge_page_shift(h);
	ret = -EFAULT;
	if (idx >= size)
		goto out_release_unlock;

4149 4150 4151 4152
	ret = -EEXIST;
	if (!huge_pte_none(huge_ptep_get(dst_pte)))
		goto out_release_unlock;

4153 4154 4155 4156 4157 4158
	if (vm_shared) {
		page_dup_rmap(page, true);
	} else {
		ClearPagePrivate(page);
		hugepage_add_new_anon_rmap(page, dst_vma, dst_addr);
	}
4159 4160 4161 4162 4163 4164 4165 4166 4167 4168 4169 4170 4171 4172 4173 4174

	_dst_pte = make_huge_pte(dst_vma, page, dst_vma->vm_flags & VM_WRITE);
	if (dst_vma->vm_flags & VM_WRITE)
		_dst_pte = huge_pte_mkdirty(_dst_pte);
	_dst_pte = pte_mkyoung(_dst_pte);

	set_huge_pte_at(dst_mm, dst_addr, dst_pte, _dst_pte);

	(void)huge_ptep_set_access_flags(dst_vma, dst_addr, dst_pte, _dst_pte,
					dst_vma->vm_flags & VM_WRITE);
	hugetlb_count_add(pages_per_huge_page(h), dst_mm);

	/* No need to invalidate - it was non-present before */
	update_mmu_cache(dst_vma, dst_addr, dst_pte);

	spin_unlock(ptl);
4175
	set_page_huge_active(page);
4176 4177
	if (vm_shared)
		unlock_page(page);
4178 4179 4180 4181 4182
	ret = 0;
out:
	return ret;
out_release_unlock:
	spin_unlock(ptl);
4183 4184
	if (vm_shared)
		unlock_page(page);
4185
out_release_nounlock:
4186 4187 4188 4189
	put_page(page);
	goto out;
}

4190 4191 4192
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,
4193
			 long i, unsigned int flags, int *nonblocking)
D
David Gibson 已提交
4194
{
4195 4196
	unsigned long pfn_offset;
	unsigned long vaddr = *position;
4197
	unsigned long remainder = *nr_pages;
4198
	struct hstate *h = hstate_vma(vma);
4199
	int err = -EFAULT;
D
David Gibson 已提交
4200 4201

	while (vaddr < vma->vm_end && remainder) {
A
Adam Litke 已提交
4202
		pte_t *pte;
4203
		spinlock_t *ptl = NULL;
H
Hugh Dickins 已提交
4204
		int absent;
A
Adam Litke 已提交
4205
		struct page *page;
D
David Gibson 已提交
4206

4207 4208 4209 4210
		/*
		 * If we have a pending SIGKILL, don't keep faulting pages and
		 * potentially allocating memory.
		 */
4211
		if (fatal_signal_pending(current)) {
4212 4213 4214 4215
			remainder = 0;
			break;
		}

A
Adam Litke 已提交
4216 4217
		/*
		 * Some archs (sparc64, sh*) have multiple pte_ts to
H
Hugh Dickins 已提交
4218
		 * each hugepage.  We have to make sure we get the
A
Adam Litke 已提交
4219
		 * first, for the page indexing below to work.
4220 4221
		 *
		 * Note that page table lock is not held when pte is null.
A
Adam Litke 已提交
4222
		 */
4223 4224
		pte = huge_pte_offset(mm, vaddr & huge_page_mask(h),
				      huge_page_size(h));
4225 4226
		if (pte)
			ptl = huge_pte_lock(h, mm, pte);
H
Hugh Dickins 已提交
4227 4228 4229 4230
		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
4231 4232 4233 4234
		 * 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
4235
		 */
H
Hugh Dickins 已提交
4236 4237
		if (absent && (flags & FOLL_DUMP) &&
		    !hugetlbfs_pagecache_present(h, vma, vaddr)) {
4238 4239
			if (pte)
				spin_unlock(ptl);
H
Hugh Dickins 已提交
4240 4241 4242
			remainder = 0;
			break;
		}
D
David Gibson 已提交
4243

4244 4245 4246 4247 4248 4249 4250 4251 4252 4253 4254
		/*
		 * 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)) ||
4255 4256
		    ((flags & FOLL_WRITE) &&
		      !huge_pte_write(huge_ptep_get(pte)))) {
4257
			vm_fault_t ret;
4258
			unsigned int fault_flags = 0;
D
David Gibson 已提交
4259

4260 4261
			if (pte)
				spin_unlock(ptl);
4262 4263 4264 4265 4266 4267 4268 4269 4270 4271 4272 4273 4274 4275
			if (flags & FOLL_WRITE)
				fault_flags |= FAULT_FLAG_WRITE;
			if (nonblocking)
				fault_flags |= FAULT_FLAG_ALLOW_RETRY;
			if (flags & FOLL_NOWAIT)
				fault_flags |= FAULT_FLAG_ALLOW_RETRY |
					FAULT_FLAG_RETRY_NOWAIT;
			if (flags & FOLL_TRIED) {
				VM_WARN_ON_ONCE(fault_flags &
						FAULT_FLAG_ALLOW_RETRY);
				fault_flags |= FAULT_FLAG_TRIED;
			}
			ret = hugetlb_fault(mm, vma, vaddr, fault_flags);
			if (ret & VM_FAULT_ERROR) {
4276
				err = vm_fault_to_errno(ret, flags);
4277 4278 4279 4280
				remainder = 0;
				break;
			}
			if (ret & VM_FAULT_RETRY) {
4281 4282
				if (nonblocking &&
				    !(fault_flags & FAULT_FLAG_RETRY_NOWAIT))
4283 4284 4285 4286 4287 4288 4289 4290 4291 4292 4293 4294 4295 4296
					*nonblocking = 0;
				*nr_pages = 0;
				/*
				 * VM_FAULT_RETRY must not return an
				 * error, it will return zero
				 * instead.
				 *
				 * No need to update "position" as the
				 * caller will not check it after
				 * *nr_pages is set to 0.
				 */
				return i;
			}
			continue;
A
Adam Litke 已提交
4297 4298
		}

4299
		pfn_offset = (vaddr & ~huge_page_mask(h)) >> PAGE_SHIFT;
4300
		page = pte_page(huge_ptep_get(pte));
4301
same_page:
4302
		if (pages) {
H
Hugh Dickins 已提交
4303
			pages[i] = mem_map_offset(page, pfn_offset);
4304
			get_page(pages[i]);
4305
		}
D
David Gibson 已提交
4306 4307 4308 4309 4310

		if (vmas)
			vmas[i] = vma;

		vaddr += PAGE_SIZE;
4311
		++pfn_offset;
D
David Gibson 已提交
4312 4313
		--remainder;
		++i;
4314
		if (vaddr < vma->vm_end && remainder &&
4315
				pfn_offset < pages_per_huge_page(h)) {
4316 4317 4318 4319 4320 4321
			/*
			 * We use pfn_offset to avoid touching the pageframes
			 * of this compound page.
			 */
			goto same_page;
		}
4322
		spin_unlock(ptl);
D
David Gibson 已提交
4323
	}
4324
	*nr_pages = remainder;
4325 4326 4327 4328 4329
	/*
	 * setting position is actually required only if remainder is
	 * not zero but it's faster not to add a "if (remainder)"
	 * branch.
	 */
D
David Gibson 已提交
4330 4331
	*position = vaddr;

4332
	return i ? i : err;
D
David Gibson 已提交
4333
}
4334

4335 4336 4337 4338 4339 4340 4341 4342
#ifndef __HAVE_ARCH_FLUSH_HUGETLB_TLB_RANGE
/*
 * ARCHes with special requirements for evicting HUGETLB backing TLB entries can
 * implement this.
 */
#define flush_hugetlb_tlb_range(vma, addr, end)	flush_tlb_range(vma, addr, end)
#endif

4343
unsigned long hugetlb_change_protection(struct vm_area_struct *vma,
4344 4345 4346 4347 4348 4349
		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;
4350
	struct hstate *h = hstate_vma(vma);
4351
	unsigned long pages = 0;
4352
	bool shared_pmd = false;
4353
	struct mmu_notifier_range range;
4354 4355 4356

	/*
	 * In the case of shared PMDs, the area to flush could be beyond
4357
	 * start/end.  Set range.start/range.end to cover the maximum possible
4358 4359
	 * range if PMD sharing is possible.
	 */
4360 4361
	mmu_notifier_range_init(&range, mm, start, end);
	adjust_range_if_pmd_sharing_possible(vma, &range.start, &range.end);
4362 4363

	BUG_ON(address >= end);
4364
	flush_cache_range(vma, range.start, range.end);
4365

4366
	mmu_notifier_invalidate_range_start(&range);
4367
	i_mmap_lock_write(vma->vm_file->f_mapping);
4368
	for (; address < end; address += huge_page_size(h)) {
4369
		spinlock_t *ptl;
4370
		ptep = huge_pte_offset(mm, address, huge_page_size(h));
4371 4372
		if (!ptep)
			continue;
4373
		ptl = huge_pte_lock(h, mm, ptep);
4374 4375
		if (huge_pmd_unshare(mm, &address, ptep)) {
			pages++;
4376
			spin_unlock(ptl);
4377
			shared_pmd = true;
4378
			continue;
4379
		}
4380 4381 4382 4383 4384 4385 4386 4387 4388 4389 4390 4391 4392
		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);
4393 4394
				set_huge_swap_pte_at(mm, address, ptep,
						     newpte, huge_page_size(h));
4395 4396 4397 4398 4399 4400
				pages++;
			}
			spin_unlock(ptl);
			continue;
		}
		if (!huge_pte_none(pte)) {
4401
			pte = huge_ptep_get_and_clear(mm, address, ptep);
4402
			pte = pte_mkhuge(huge_pte_modify(pte, newprot));
4403
			pte = arch_make_huge_pte(pte, vma, NULL, 0);
4404
			set_huge_pte_at(mm, address, ptep, pte);
4405
			pages++;
4406
		}
4407
		spin_unlock(ptl);
4408
	}
4409
	/*
4410
	 * Must flush TLB before releasing i_mmap_rwsem: x86's huge_pmd_unshare
4411
	 * may have cleared our pud entry and done put_page on the page table:
4412
	 * once we release i_mmap_rwsem, another task can do the final put_page
4413 4414
	 * and that page table be reused and filled with junk.  If we actually
	 * did unshare a page of pmds, flush the range corresponding to the pud.
4415
	 */
4416
	if (shared_pmd)
4417
		flush_hugetlb_tlb_range(vma, range.start, range.end);
4418 4419
	else
		flush_hugetlb_tlb_range(vma, start, end);
4420 4421 4422 4423
	/*
	 * No need to call mmu_notifier_invalidate_range() we are downgrading
	 * page table protection not changing it to point to a new page.
	 *
4424
	 * See Documentation/vm/mmu_notifier.rst
4425
	 */
4426
	i_mmap_unlock_write(vma->vm_file->f_mapping);
4427
	mmu_notifier_invalidate_range_end(&range);
4428 4429

	return pages << h->order;
4430 4431
}

4432 4433
int hugetlb_reserve_pages(struct inode *inode,
					long from, long to,
4434
					struct vm_area_struct *vma,
4435
					vm_flags_t vm_flags)
4436
{
4437
	long ret, chg;
4438
	struct hstate *h = hstate_inode(inode);
4439
	struct hugepage_subpool *spool = subpool_inode(inode);
4440
	struct resv_map *resv_map;
4441
	long gbl_reserve;
4442

4443 4444 4445 4446 4447 4448
	/* This should never happen */
	if (from > to) {
		VM_WARN(1, "%s called with a negative range\n", __func__);
		return -EINVAL;
	}

4449 4450 4451
	/*
	 * Only apply hugepage reservation if asked. At fault time, an
	 * attempt will be made for VM_NORESERVE to allocate a page
4452
	 * without using reserves
4453
	 */
4454
	if (vm_flags & VM_NORESERVE)
4455 4456
		return 0;

4457 4458 4459 4460 4461 4462
	/*
	 * 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
	 */
4463
	if (!vma || vma->vm_flags & VM_MAYSHARE) {
4464
		resv_map = inode_resv_map(inode);
4465

4466
		chg = region_chg(resv_map, from, to);
4467 4468 4469

	} else {
		resv_map = resv_map_alloc();
4470 4471 4472
		if (!resv_map)
			return -ENOMEM;

4473
		chg = to - from;
4474

4475 4476 4477 4478
		set_vma_resv_map(vma, resv_map);
		set_vma_resv_flags(vma, HPAGE_RESV_OWNER);
	}

4479 4480 4481 4482
	if (chg < 0) {
		ret = chg;
		goto out_err;
	}
4483

4484 4485 4486 4487 4488 4489 4490
	/*
	 * 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) {
4491 4492 4493
		ret = -ENOSPC;
		goto out_err;
	}
4494 4495

	/*
4496
	 * Check enough hugepages are available for the reservation.
4497
	 * Hand the pages back to the subpool if there are not
4498
	 */
4499
	ret = hugetlb_acct_memory(h, gbl_reserve);
K
Ken Chen 已提交
4500
	if (ret < 0) {
4501 4502
		/* put back original number of pages, chg */
		(void)hugepage_subpool_put_pages(spool, chg);
4503
		goto out_err;
K
Ken Chen 已提交
4504
	}
4505 4506 4507 4508 4509 4510 4511 4512 4513 4514 4515 4516

	/*
	 * 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
	 */
4517 4518 4519 4520 4521 4522 4523 4524 4525 4526 4527 4528 4529 4530 4531 4532 4533 4534
	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);
		}
	}
4535
	return 0;
4536
out_err:
4537
	if (!vma || vma->vm_flags & VM_MAYSHARE)
4538 4539 4540
		/* Don't call region_abort if region_chg failed */
		if (chg >= 0)
			region_abort(resv_map, from, to);
J
Joonsoo Kim 已提交
4541 4542
	if (vma && is_vma_resv_set(vma, HPAGE_RESV_OWNER))
		kref_put(&resv_map->refs, resv_map_release);
4543
	return ret;
4544 4545
}

4546 4547
long hugetlb_unreserve_pages(struct inode *inode, long start, long end,
								long freed)
4548
{
4549
	struct hstate *h = hstate_inode(inode);
4550
	struct resv_map *resv_map = inode_resv_map(inode);
4551
	long chg = 0;
4552
	struct hugepage_subpool *spool = subpool_inode(inode);
4553
	long gbl_reserve;
K
Ken Chen 已提交
4554

4555 4556 4557 4558 4559 4560 4561 4562 4563 4564 4565
	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
4566
	spin_lock(&inode->i_lock);
4567
	inode->i_blocks -= (blocks_per_huge_page(h) * freed);
K
Ken Chen 已提交
4568 4569
	spin_unlock(&inode->i_lock);

4570 4571 4572 4573 4574 4575
	/*
	 * 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);
4576 4577

	return 0;
4578
}
4579

4580 4581 4582 4583 4584 4585 4586 4587 4588 4589 4590
#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 */
E
Eric B Munson 已提交
4591 4592
	unsigned long vm_flags = vma->vm_flags & VM_LOCKED_CLEAR_MASK;
	unsigned long svm_flags = svma->vm_flags & VM_LOCKED_CLEAR_MASK;
4593 4594 4595 4596 4597 4598 4599 4600 4601 4602 4603 4604 4605

	/*
	 * 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;
}

4606
static bool vma_shareable(struct vm_area_struct *vma, unsigned long addr)
4607 4608 4609 4610 4611 4612 4613
{
	unsigned long base = addr & PUD_MASK;
	unsigned long end = base + PUD_SIZE;

	/*
	 * check on proper vm_flags and page table alignment
	 */
4614
	if (vma->vm_flags & VM_MAYSHARE && range_in_vma(vma, base, end))
4615 4616
		return true;
	return false;
4617 4618
}

4619 4620 4621 4622 4623 4624 4625 4626 4627 4628 4629 4630 4631 4632 4633 4634 4635 4636 4637 4638 4639 4640 4641 4642 4643 4644 4645 4646 4647
/*
 * Determine if start,end range within vma could be mapped by shared pmd.
 * If yes, adjust start and end to cover range associated with possible
 * shared pmd mappings.
 */
void adjust_range_if_pmd_sharing_possible(struct vm_area_struct *vma,
				unsigned long *start, unsigned long *end)
{
	unsigned long check_addr = *start;

	if (!(vma->vm_flags & VM_MAYSHARE))
		return;

	for (check_addr = *start; check_addr < *end; check_addr += PUD_SIZE) {
		unsigned long a_start = check_addr & PUD_MASK;
		unsigned long a_end = a_start + PUD_SIZE;

		/*
		 * If sharing is possible, adjust start/end if necessary.
		 */
		if (range_in_vma(vma, a_start, a_end)) {
			if (a_start < *start)
				*start = a_start;
			if (a_end > *end)
				*end = a_end;
		}
	}
}

4648 4649 4650 4651
/*
 * 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
4652 4653 4654 4655
 * code much cleaner. pmd allocation is essential for the shared case because
 * pud has to be populated inside the same i_mmap_rwsem section - otherwise
 * racing tasks could either miss the sharing (see huge_pte_offset) or select a
 * bad pmd for sharing.
4656 4657 4658 4659 4660 4661 4662 4663 4664 4665 4666
 */
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;
4667
	spinlock_t *ptl;
4668 4669 4670 4671

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

4672
	i_mmap_lock_write(mapping);
4673 4674 4675 4676 4677 4678
	vma_interval_tree_foreach(svma, &mapping->i_mmap, idx, idx) {
		if (svma == vma)
			continue;

		saddr = page_table_shareable(svma, vma, addr, idx);
		if (saddr) {
4679 4680
			spte = huge_pte_offset(svma->vm_mm, saddr,
					       vma_mmu_pagesize(svma));
4681 4682 4683 4684 4685 4686 4687 4688 4689 4690
			if (spte) {
				get_page(virt_to_page(spte));
				break;
			}
		}
	}

	if (!spte)
		goto out;

4691
	ptl = huge_pte_lock(hstate_vma(vma), mm, spte);
4692
	if (pud_none(*pud)) {
4693 4694
		pud_populate(mm, pud,
				(pmd_t *)((unsigned long)spte & PAGE_MASK));
4695
		mm_inc_nr_pmds(mm);
4696
	} else {
4697
		put_page(virt_to_page(spte));
4698
	}
4699
	spin_unlock(ptl);
4700 4701
out:
	pte = (pte_t *)pmd_alloc(mm, pud, addr);
4702
	i_mmap_unlock_write(mapping);
4703 4704 4705 4706 4707 4708 4709 4710 4711 4712
	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.
 *
4713
 * called with page table lock held.
4714 4715 4716 4717 4718 4719 4720
 *
 * 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);
4721 4722
	p4d_t *p4d = p4d_offset(pgd, *addr);
	pud_t *pud = pud_offset(p4d, *addr);
4723 4724 4725 4726 4727 4728 4729

	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));
4730
	mm_dec_nr_pmds(mm);
4731 4732 4733
	*addr = ALIGN(*addr, HPAGE_SIZE * PTRS_PER_PTE) - HPAGE_SIZE;
	return 1;
}
4734 4735 4736 4737 4738 4739
#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;
}
4740 4741 4742 4743 4744

int huge_pmd_unshare(struct mm_struct *mm, unsigned long *addr, pte_t *ptep)
{
	return 0;
}
4745 4746 4747 4748 4749

void adjust_range_if_pmd_sharing_possible(struct vm_area_struct *vma,
				unsigned long *start, unsigned long *end)
{
}
4750
#define want_pmd_share()	(0)
4751 4752
#endif /* CONFIG_ARCH_WANT_HUGE_PMD_SHARE */

4753 4754 4755 4756 4757
#ifdef CONFIG_ARCH_WANT_GENERAL_HUGETLB
pte_t *huge_pte_alloc(struct mm_struct *mm,
			unsigned long addr, unsigned long sz)
{
	pgd_t *pgd;
4758
	p4d_t *p4d;
4759 4760 4761 4762
	pud_t *pud;
	pte_t *pte = NULL;

	pgd = pgd_offset(mm, addr);
4763 4764 4765
	p4d = p4d_alloc(mm, pgd, addr);
	if (!p4d)
		return NULL;
4766
	pud = pud_alloc(mm, p4d, addr);
4767 4768 4769 4770 4771 4772 4773 4774 4775 4776 4777
	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);
		}
	}
4778
	BUG_ON(pte && pte_present(*pte) && !pte_huge(*pte));
4779 4780 4781 4782

	return pte;
}

4783 4784 4785 4786 4787 4788 4789 4790 4791
/*
 * huge_pte_offset() - Walk the page table to resolve the hugepage
 * entry at address @addr
 *
 * Return: Pointer to page table or swap entry (PUD or PMD) for
 * address @addr, or NULL if a p*d_none() entry is encountered and the
 * size @sz doesn't match the hugepage size at this level of the page
 * table.
 */
4792 4793
pte_t *huge_pte_offset(struct mm_struct *mm,
		       unsigned long addr, unsigned long sz)
4794 4795
{
	pgd_t *pgd;
4796
	p4d_t *p4d;
4797
	pud_t *pud;
4798
	pmd_t *pmd;
4799 4800

	pgd = pgd_offset(mm, addr);
4801 4802 4803 4804 4805
	if (!pgd_present(*pgd))
		return NULL;
	p4d = p4d_offset(pgd, addr);
	if (!p4d_present(*p4d))
		return NULL;
4806

4807
	pud = pud_offset(p4d, addr);
4808
	if (sz != PUD_SIZE && pud_none(*pud))
4809
		return NULL;
4810 4811
	/* hugepage or swap? */
	if (pud_huge(*pud) || !pud_present(*pud))
4812
		return (pte_t *)pud;
4813

4814
	pmd = pmd_offset(pud, addr);
4815 4816 4817 4818 4819 4820 4821
	if (sz != PMD_SIZE && pmd_none(*pmd))
		return NULL;
	/* hugepage or swap? */
	if (pmd_huge(*pmd) || !pmd_present(*pmd))
		return (pte_t *)pmd;

	return NULL;
4822 4823
}

4824 4825 4826 4827 4828 4829 4830 4831 4832 4833 4834 4835 4836
#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);
}

4837 4838 4839 4840 4841 4842 4843 4844
struct page * __weak
follow_huge_pd(struct vm_area_struct *vma,
	       unsigned long address, hugepd_t hpd, int flags, int pdshift)
{
	WARN(1, "hugepd follow called with no support for hugepage directory format\n");
	return NULL;
}

4845
struct page * __weak
4846
follow_huge_pmd(struct mm_struct *mm, unsigned long address,
4847
		pmd_t *pmd, int flags)
4848
{
4849 4850
	struct page *page = NULL;
	spinlock_t *ptl;
4851
	pte_t pte;
4852 4853 4854 4855 4856 4857 4858 4859 4860
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;
4861 4862
	pte = huge_ptep_get((pte_t *)pmd);
	if (pte_present(pte)) {
4863
		page = pmd_page(*pmd) + ((address & ~PMD_MASK) >> PAGE_SHIFT);
4864 4865 4866
		if (flags & FOLL_GET)
			get_page(page);
	} else {
4867
		if (is_hugetlb_entry_migration(pte)) {
4868 4869 4870 4871 4872 4873 4874 4875 4876 4877 4878
			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);
4879 4880 4881
	return page;
}

4882
struct page * __weak
4883
follow_huge_pud(struct mm_struct *mm, unsigned long address,
4884
		pud_t *pud, int flags)
4885
{
4886 4887
	if (flags & FOLL_GET)
		return NULL;
4888

4889
	return pte_page(*(pte_t *)pud) + ((address & ~PUD_MASK) >> PAGE_SHIFT);
4890 4891
}

4892 4893 4894 4895 4896 4897 4898 4899 4900
struct page * __weak
follow_huge_pgd(struct mm_struct *mm, unsigned long address, pgd_t *pgd, int flags)
{
	if (flags & FOLL_GET)
		return NULL;

	return pte_page(*(pte_t *)pgd) + ((address & ~PGDIR_MASK) >> PAGE_SHIFT);
}

4901 4902
bool isolate_huge_page(struct page *page, struct list_head *list)
{
4903 4904
	bool ret = true;

4905
	VM_BUG_ON_PAGE(!PageHead(page), page);
4906
	spin_lock(&hugetlb_lock);
4907 4908 4909 4910 4911
	if (!page_huge_active(page) || !get_page_unless_zero(page)) {
		ret = false;
		goto unlock;
	}
	clear_page_huge_active(page);
4912
	list_move_tail(&page->lru, list);
4913
unlock:
4914
	spin_unlock(&hugetlb_lock);
4915
	return ret;
4916 4917 4918 4919
}

void putback_active_hugepage(struct page *page)
{
4920
	VM_BUG_ON_PAGE(!PageHead(page), page);
4921
	spin_lock(&hugetlb_lock);
4922
	set_page_huge_active(page);
4923 4924 4925 4926
	list_move_tail(&page->lru, &(page_hstate(page))->hugepage_activelist);
	spin_unlock(&hugetlb_lock);
	put_page(page);
}
4927 4928 4929 4930 4931 4932 4933 4934 4935 4936 4937 4938 4939 4940 4941 4942 4943 4944 4945 4946 4947 4948 4949 4950 4951 4952 4953 4954 4955 4956 4957 4958 4959

void move_hugetlb_state(struct page *oldpage, struct page *newpage, int reason)
{
	struct hstate *h = page_hstate(oldpage);

	hugetlb_cgroup_migrate(oldpage, newpage);
	set_page_owner_migrate_reason(newpage, reason);

	/*
	 * transfer temporary state of the new huge page. This is
	 * reverse to other transitions because the newpage is going to
	 * be final while the old one will be freed so it takes over
	 * the temporary status.
	 *
	 * Also note that we have to transfer the per-node surplus state
	 * here as well otherwise the global surplus count will not match
	 * the per-node's.
	 */
	if (PageHugeTemporary(newpage)) {
		int old_nid = page_to_nid(oldpage);
		int new_nid = page_to_nid(newpage);

		SetPageHugeTemporary(oldpage);
		ClearPageHugeTemporary(newpage);

		spin_lock(&hugetlb_lock);
		if (h->surplus_huge_pages_node[old_nid]) {
			h->surplus_huge_pages_node[old_nid]--;
			h->surplus_huge_pages_node[new_nid]++;
		}
		spin_unlock(&hugetlb_lock);
	}
}