hugetlb.c 128.7 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/bootmem.h>
19
#include <linux/sysfs.h>
20
#include <linux/slab.h>
21
#include <linux/sched/signal.h>
22
#include <linux/rmap.h>
23 24
#include <linux/swap.h>
#include <linux/swapops.h>
25
#include <linux/jhash.h>
26

D
David Gibson 已提交
27 28
#include <asm/page.h>
#include <asm/pgtable.h>
29
#include <asm/tlb.h>
D
David Gibson 已提交
30

31
#include <linux/io.h>
D
David Gibson 已提交
32
#include <linux/hugetlb.h>
33
#include <linux/hugetlb_cgroup.h>
34
#include <linux/node.h>
35
#include <linux/userfaultfd_k.h>
36
#include "internal.h"
L
Linus Torvalds 已提交
37

38
int hugepages_treat_as_movable;
39

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

49 50
__initdata LIST_HEAD(huge_boot_pages);

51 52 53
/* for command line parsing */
static struct hstate * __initdata parsed_hstate;
static unsigned long __initdata default_hstate_max_huge_pages;
54
static unsigned long __initdata default_hstate_size;
55
static bool __initdata parsed_valid_hugepagesz = true;
56

57
/*
58 59
 * Protects updates to hugepage_freelists, hugepage_activelist, nr_huge_pages,
 * free_huge_pages, and surplus_huge_pages.
60
 */
61
DEFINE_SPINLOCK(hugetlb_lock);
62

63 64 65 66 67
/*
 * 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;
68
struct mutex *hugetlb_fault_mutex_table ____cacheline_aligned_in_smp;
69

70 71 72
/* Forward declaration */
static int hugetlb_acct_memory(struct hstate *h, long delta);

73 74 75 76 77 78 79
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
80 81 82 83 84 85
	 * 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);
86
		kfree(spool);
87
	}
88 89
}

90 91
struct hugepage_subpool *hugepage_new_subpool(struct hstate *h, long max_hpages,
						long min_hpages)
92 93 94
{
	struct hugepage_subpool *spool;

95
	spool = kzalloc(sizeof(*spool), GFP_KERNEL);
96 97 98 99 100
	if (!spool)
		return NULL;

	spin_lock_init(&spool->lock);
	spool->count = 1;
101 102 103 104 105 106 107 108 109
	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;
110 111 112 113 114 115 116 117 118 119 120 121

	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);
}

122 123 124 125 126 127 128 129 130
/*
 * 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,
131 132
				      long delta)
{
133
	long ret = delta;
134 135

	if (!spool)
136
		return ret;
137 138

	spin_lock(&spool->lock);
139 140 141 142 143 144 145 146

	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;
		}
147 148
	}

149 150
	/* minimum size accounting */
	if (spool->min_hpages != -1 && spool->rsv_hpages) {
151 152 153 154 155 156 157 158 159 160 161 162 163 164 165
		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);
166 167 168
	return ret;
}

169 170 171 172 173 174 175
/*
 * 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,
176 177
				       long delta)
{
178 179
	long ret = delta;

180
	if (!spool)
181
		return delta;
182 183

	spin_lock(&spool->lock);
184 185 186 187

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

188 189
	 /* minimum size accounting */
	if (spool->min_hpages != -1 && spool->used_hpages < spool->min_hpages) {
190 191 192 193 194 195 196 197 198 199 200 201 202 203
		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.
	 */
204
	unlock_or_release_subpool(spool);
205 206

	return ret;
207 208 209 210 211 212 213 214 215
}

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
216
	return subpool_inode(file_inode(vma->vm_file));
217 218
}

219 220 221
/*
 * Region tracking -- allows tracking of reservations and instantiated pages
 *                    across the pages in a mapping.
222
 *
223 224 225 226 227 228 229 230 231 232 233 234 235 236
 * 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.
237 238 239 240 241 242 243
 */
struct file_region {
	struct list_head link;
	long from;
	long to;
};

244 245
/*
 * Add the huge page range represented by [f, t) to the reserve
246 247 248 249 250 251 252 253
 * 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.
254 255 256
 *
 * Return the number of new huge pages added to the map.  This
 * number is greater than or equal to zero.
257
 */
258
static long region_add(struct resv_map *resv, long f, long t)
259
{
260
	struct list_head *head = &resv->regions;
261
	struct file_region *rg, *nrg, *trg;
262
	long add = 0;
263

264
	spin_lock(&resv->lock);
265 266 267 268 269
	/* Locate the region we are either in or before. */
	list_for_each_entry(rg, head, link)
		if (f <= rg->to)
			break;

270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291
	/*
	 * 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;
	}

292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309
	/* 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) {
310 311 312 313 314
			/* 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);
315 316 317 318
			list_del(&rg->link);
			kfree(rg);
		}
	}
319 320

	add += (nrg->from - f);		/* Added to beginning of region */
321
	nrg->from = f;
322
	add += t - nrg->to;		/* Added to end of region */
323
	nrg->to = t;
324

325 326
out_locked:
	resv->adds_in_progress--;
327
	spin_unlock(&resv->lock);
328 329
	VM_BUG_ON(add < 0);
	return add;
330 331
}

332 333 334 335 336 337 338 339 340 341 342 343 344
/*
 * 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.
 *
345 346 347 348 349 350 351 352
 * 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.
353
 */
354
static long region_chg(struct resv_map *resv, long f, long t)
355
{
356
	struct list_head *head = &resv->regions;
357
	struct file_region *rg, *nrg = NULL;
358 359
	long chg = 0;

360 361
retry:
	spin_lock(&resv->lock);
362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377
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);
378 379
		if (!trg) {
			kfree(nrg);
380
			return -ENOMEM;
381
		}
382 383 384 385 386 387 388

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

389 390 391 392 393 394 395 396 397
	/* 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) {
398
		if (!nrg) {
399
			resv->adds_in_progress--;
400 401 402 403 404 405 406 407 408 409
			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;
		}
410

411 412 413
		list_add(&nrg->link, rg->link.prev);
		chg = t - f;
		goto out_nrg;
414 415 416 417 418 419 420 421 422 423 424 425
	}

	/* 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)
426
			goto out;
427

L
Lucas De Marchi 已提交
428
		/* We overlap with this area, if it extends further than
429 430 431 432 433 434 435 436
		 * 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;
	}
437 438 439 440 441 442 443 444

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);
445 446 447
	return chg;
}

448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466
/*
 * 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);
}

467
/*
468 469 470 471 472 473 474 475 476 477 478 479
 * 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.
480
 */
481
static long region_del(struct resv_map *resv, long f, long t)
482
{
483
	struct list_head *head = &resv->regions;
484
	struct file_region *rg, *trg;
485 486
	struct file_region *nrg = NULL;
	long del = 0;
487

488
retry:
489
	spin_lock(&resv->lock);
490
	list_for_each_entry_safe(rg, trg, head, link) {
491 492 493 494 495 496 497 498
		/*
		 * 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))
499
			continue;
500

501
		if (rg->from >= t)
502 503
			break;

504 505 506 507 508 509 510 511 512 513 514 515 516
		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--;
			}
517

518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537
			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;
538
			break;
539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554
		}

		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;
		}
555
	}
556 557

	spin_unlock(&resv->lock);
558 559
	kfree(nrg);
	return del;
560 561
}

562 563 564 565 566 567 568 569 570
/*
 * A rare out of memory error was encountered which prevented removal of
 * the reserve map region for a page.  The huge page itself was free'ed
 * and removed from the page cache.  This routine will adjust the subpool
 * usage count, and the global reserve count if needed.  By incrementing
 * these counts, the reserve map entry which could not be deleted will
 * appear as a "reserved" entry instead of simply dangling with incorrect
 * counts.
 */
571
void hugetlb_fix_reserve_counts(struct inode *inode)
572 573 574 575 576
{
	struct hugepage_subpool *spool = subpool_inode(inode);
	long rsv_adjust;

	rsv_adjust = hugepage_subpool_get_pages(spool, 1);
577
	if (rsv_adjust) {
578 579 580 581 582 583
		struct hstate *h = hstate_inode(inode);

		hugetlb_acct_memory(h, 1);
	}
}

584 585 586 587
/*
 * Count and return the number of huge pages in the reserve map
 * that intersect with the range [f, t).
 */
588
static long region_count(struct resv_map *resv, long f, long t)
589
{
590
	struct list_head *head = &resv->regions;
591 592 593
	struct file_region *rg;
	long chg = 0;

594
	spin_lock(&resv->lock);
595 596
	/* Locate each segment we overlap with, and count that overlap. */
	list_for_each_entry(rg, head, link) {
597 598
		long seg_from;
		long seg_to;
599 600 601 602 603 604 605 606 607 608 609

		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;
	}
610
	spin_unlock(&resv->lock);
611 612 613 614

	return chg;
}

615 616 617 618
/*
 * Convert the address within this vma to the page offset within
 * the mapping, in pagecache page units; huge pages here.
 */
619 620
static pgoff_t vma_hugecache_offset(struct hstate *h,
			struct vm_area_struct *vma, unsigned long address)
621
{
622 623
	return ((address - vma->vm_start) >> huge_page_shift(h)) +
			(vma->vm_pgoff >> huge_page_order(h));
624 625
}

626 627 628 629 630
pgoff_t linear_hugepage_index(struct vm_area_struct *vma,
				     unsigned long address)
{
	return vma_hugecache_offset(hstate_vma(vma), vma, address);
}
631
EXPORT_SYMBOL_GPL(linear_hugepage_index);
632

633 634 635 636 637 638 639 640 641 642 643 644 645
/*
 * Return the size of the pages allocated when backing a VMA. In the majority
 * cases this will be same size as used by the page table entries.
 */
unsigned long vma_kernel_pagesize(struct vm_area_struct *vma)
{
	struct hstate *hstate;

	if (!is_vm_hugetlb_page(vma))
		return PAGE_SIZE;

	hstate = hstate_vma(vma);

646
	return 1UL << huge_page_shift(hstate);
647
}
648
EXPORT_SYMBOL_GPL(vma_kernel_pagesize);
649

650 651 652 653 654 655 656 657 658 659 660 661 662
/*
 * Return the page size being used by the MMU to back a VMA. In the majority
 * of cases, the page size used by the kernel matches the MMU size. On
 * architectures where it differs, an architecture-specific version of this
 * function is required.
 */
#ifndef vma_mmu_pagesize
unsigned long vma_mmu_pagesize(struct vm_area_struct *vma)
{
	return vma_kernel_pagesize(vma);
}
#endif

663 664 665 666 667 668 669
/*
 * 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)
670
#define HPAGE_RESV_MASK (HPAGE_RESV_OWNER | HPAGE_RESV_UNMAPPED)
671

672 673 674 675 676 677 678 679 680
/*
 * 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.
681 682 683 684 685 686 687 688 689
 *
 * 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.
690
 */
691 692 693 694 695 696 697 698 699 700 701
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;
}

702
struct resv_map *resv_map_alloc(void)
703 704
{
	struct resv_map *resv_map = kmalloc(sizeof(*resv_map), GFP_KERNEL);
705 706 707 708 709
	struct file_region *rg = kmalloc(sizeof(*rg), GFP_KERNEL);

	if (!resv_map || !rg) {
		kfree(resv_map);
		kfree(rg);
710
		return NULL;
711
	}
712 713

	kref_init(&resv_map->refs);
714
	spin_lock_init(&resv_map->lock);
715 716
	INIT_LIST_HEAD(&resv_map->regions);

717 718 719 720 721 722
	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;

723 724 725
	return resv_map;
}

726
void resv_map_release(struct kref *ref)
727 728
{
	struct resv_map *resv_map = container_of(ref, struct resv_map, refs);
729 730
	struct list_head *head = &resv_map->region_cache;
	struct file_region *rg, *trg;
731 732

	/* Clear out any active regions before we release the map. */
733
	region_del(resv_map, 0, LONG_MAX);
734 735 736 737 738 739 740 741 742

	/* ... 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);

743 744 745
	kfree(resv_map);
}

746 747 748 749 750
static inline struct resv_map *inode_resv_map(struct inode *inode)
{
	return inode->i_mapping->private_data;
}

751
static struct resv_map *vma_resv_map(struct vm_area_struct *vma)
752
{
753
	VM_BUG_ON_VMA(!is_vm_hugetlb_page(vma), vma);
754 755 756 757 758 759 760
	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 {
761 762
		return (struct resv_map *)(get_vma_private_data(vma) &
							~HPAGE_RESV_MASK);
763
	}
764 765
}

766
static void set_vma_resv_map(struct vm_area_struct *vma, struct resv_map *map)
767
{
768 769
	VM_BUG_ON_VMA(!is_vm_hugetlb_page(vma), vma);
	VM_BUG_ON_VMA(vma->vm_flags & VM_MAYSHARE, vma);
770

771 772
	set_vma_private_data(vma, (get_vma_private_data(vma) &
				HPAGE_RESV_MASK) | (unsigned long)map);
773 774 775 776
}

static void set_vma_resv_flags(struct vm_area_struct *vma, unsigned long flags)
{
777 778
	VM_BUG_ON_VMA(!is_vm_hugetlb_page(vma), vma);
	VM_BUG_ON_VMA(vma->vm_flags & VM_MAYSHARE, vma);
779 780

	set_vma_private_data(vma, get_vma_private_data(vma) | flags);
781 782 783 784
}

static int is_vma_resv_set(struct vm_area_struct *vma, unsigned long flag)
{
785
	VM_BUG_ON_VMA(!is_vm_hugetlb_page(vma), vma);
786 787

	return (get_vma_private_data(vma) & flag) != 0;
788 789
}

790
/* Reset counters to 0 and clear all HPAGE_RESV_* flags */
791 792
void reset_vma_resv_huge_pages(struct vm_area_struct *vma)
{
793
	VM_BUG_ON_VMA(!is_vm_hugetlb_page(vma), vma);
794
	if (!(vma->vm_flags & VM_MAYSHARE))
795 796 797 798
		vma->vm_private_data = (void *)0;
}

/* Returns true if the VMA has associated reserve pages */
799
static bool vma_has_reserves(struct vm_area_struct *vma, long chg)
800
{
801 802 803 804 805 806 807 808 809 810 811
	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)
812
			return true;
813
		else
814
			return false;
815
	}
816 817

	/* Shared mappings always use reserves */
818 819 820 821 822 823 824 825 826 827 828 829 830
	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;
	}
831 832 833 834 835

	/*
	 * Only the process that called mmap() has reserves for
	 * private mappings.
	 */
836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856
	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;
	}
857

858
	return false;
859 860
}

861
static void enqueue_huge_page(struct hstate *h, struct page *page)
L
Linus Torvalds 已提交
862 863
{
	int nid = page_to_nid(page);
864
	list_move(&page->lru, &h->hugepage_freelists[nid]);
865 866
	h->free_huge_pages++;
	h->free_huge_pages_node[nid]++;
L
Linus Torvalds 已提交
867 868
}

869
static struct page *dequeue_huge_page_node_exact(struct hstate *h, int nid)
870 871 872
{
	struct page *page;

873
	list_for_each_entry(page, &h->hugepage_freelists[nid], lru)
874
		if (!PageHWPoison(page))
875 876 877 878 879 880
			break;
	/*
	 * if 'non-isolated free hugepage' not found on the list,
	 * the allocation fails.
	 */
	if (&h->hugepage_freelists[nid] == &page->lru)
881
		return NULL;
882
	list_move(&page->lru, &h->hugepage_activelist);
883
	set_page_refcounted(page);
884 885 886 887 888
	h->free_huge_pages--;
	h->free_huge_pages_node[nid]--;
	return page;
}

889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904
static struct page *dequeue_huge_page_node(struct hstate *h, int nid)
{
	struct page *page;
	int node;

	if (nid != NUMA_NO_NODE)
		return dequeue_huge_page_node_exact(h, nid);

	for_each_online_node(node) {
		page = dequeue_huge_page_node_exact(h, node);
		if (page)
			return page;
	}
	return NULL;
}

905 906 907
/* Movability of hugepages depends on migration support. */
static inline gfp_t htlb_alloc_mask(struct hstate *h)
{
908
	if (hugepages_treat_as_movable || hugepage_migration_supported(h))
909 910 911 912 913
		return GFP_HIGHUSER_MOVABLE;
	else
		return GFP_HIGHUSER;
}

914 915
static struct page *dequeue_huge_page_vma(struct hstate *h,
				struct vm_area_struct *vma,
916 917
				unsigned long address, int avoid_reserve,
				long chg)
L
Linus Torvalds 已提交
918
{
919
	struct page *page = NULL;
920
	struct mempolicy *mpol;
921
	nodemask_t *nodemask;
922 923
	gfp_t gfp_mask;
	int nid;
924
	struct zonelist *zonelist;
925 926
	struct zone *zone;
	struct zoneref *z;
927
	unsigned int cpuset_mems_cookie;
L
Linus Torvalds 已提交
928

929 930 931 932 933
	/*
	 * 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
	 */
934
	if (!vma_has_reserves(vma, chg) &&
935
			h->free_huge_pages - h->resv_huge_pages == 0)
936
		goto err;
937

938
	/* If reserves cannot be used, ensure enough pages are in the pool */
939
	if (avoid_reserve && h->free_huge_pages - h->resv_huge_pages == 0)
940
		goto err;
941

942
retry_cpuset:
943
	cpuset_mems_cookie = read_mems_allowed_begin();
944 945 946
	gfp_mask = htlb_alloc_mask(h);
	nid = huge_node(vma, address, gfp_mask, &mpol, &nodemask);
	zonelist = node_zonelist(nid, gfp_mask);
947

948 949
	for_each_zone_zonelist_nodemask(zone, z, zonelist,
						MAX_NR_ZONES - 1, nodemask) {
950
		if (cpuset_zone_allowed(zone, gfp_mask)) {
951 952
			page = dequeue_huge_page_node(h, zone_to_nid(zone));
			if (page) {
953 954 955 956 957
				if (avoid_reserve)
					break;
				if (!vma_has_reserves(vma, chg))
					break;

958
				SetPagePrivate(page);
959
				h->resv_huge_pages--;
960 961
				break;
			}
A
Andrew Morton 已提交
962
		}
L
Linus Torvalds 已提交
963
	}
964

965
	mpol_cond_put(mpol);
966
	if (unlikely(!page && read_mems_allowed_retry(cpuset_mems_cookie)))
967
		goto retry_cpuset;
L
Linus Torvalds 已提交
968
	return page;
969 970 971

err:
	return NULL;
L
Linus Torvalds 已提交
972 973
}

974 975 976 977 978 979 980 981 982
/*
 * 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)
{
983
	nid = next_node_in(nid, *nodes_allowed);
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 1038 1039 1040 1041 1042 1043 1044
	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--)

1045
#ifdef CONFIG_ARCH_HAS_GIGANTIC_PAGE
1046
static void destroy_compound_gigantic_page(struct page *page,
1047
					unsigned int order)
1048 1049 1050 1051 1052
{
	int i;
	int nr_pages = 1 << order;
	struct page *p = page + 1;

1053
	atomic_set(compound_mapcount_ptr(page), 0);
1054
	for (i = 1; i < nr_pages; i++, p = mem_map_next(p, page, i)) {
1055
		clear_compound_head(p);
1056 1057 1058 1059 1060 1061 1062
		set_page_refcounted(p);
	}

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

1063
static void free_gigantic_page(struct page *page, unsigned int order)
1064 1065 1066 1067 1068 1069 1070 1071
{
	free_contig_range(page_to_pfn(page), 1 << order);
}

static int __alloc_gigantic_page(unsigned long start_pfn,
				unsigned long nr_pages)
{
	unsigned long end_pfn = start_pfn + nr_pages;
1072 1073
	return alloc_contig_range(start_pfn, end_pfn, MIGRATE_MOVABLE,
				  GFP_KERNEL);
1074 1075
}

1076 1077
static bool pfn_range_valid_gigantic(struct zone *z,
			unsigned long start_pfn, unsigned long nr_pages)
1078 1079 1080 1081 1082 1083 1084 1085 1086 1087
{
	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);

1088 1089 1090
		if (page_zone(page) != z)
			return false;

1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110
		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);
}

1111
static struct page *alloc_gigantic_page(int nid, unsigned int order)
1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122
{
	unsigned long nr_pages = 1 << order;
	unsigned long ret, pfn, flags;
	struct zone *z;

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

		pfn = ALIGN(z->zone_start_pfn, nr_pages);
		while (zone_spans_last_pfn(z, pfn, nr_pages)) {
1123
			if (pfn_range_valid_gigantic(z, pfn, nr_pages)) {
1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146
				/*
				 * We release the zone lock here because
				 * alloc_contig_range() will also lock the zone
				 * at some point. If there's an allocation
				 * spinning on this lock, it may win the race
				 * and cause alloc_contig_range() to fail...
				 */
				spin_unlock_irqrestore(&z->lock, flags);
				ret = __alloc_gigantic_page(pfn, nr_pages);
				if (!ret)
					return pfn_to_page(pfn);
				spin_lock_irqsave(&z->lock, flags);
			}
			pfn += nr_pages;
		}

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

	return NULL;
}

static void prep_new_huge_page(struct hstate *h, struct page *page, int nid);
1147
static void prep_compound_gigantic_page(struct page *page, unsigned int order);
1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176

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

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

	return page;
}

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

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

	return 0;
}

1177
#else /* !CONFIG_ARCH_HAS_GIGANTIC_PAGE */
1178
static inline bool gigantic_page_supported(void) { return false; }
1179
static inline void free_gigantic_page(struct page *page, unsigned int order) { }
1180
static inline void destroy_compound_gigantic_page(struct page *page,
1181
						unsigned int order) { }
1182 1183 1184 1185
static inline int alloc_fresh_gigantic_page(struct hstate *h,
					nodemask_t *nodes_allowed) { return 0; }
#endif

1186
static void update_and_free_page(struct hstate *h, struct page *page)
A
Adam Litke 已提交
1187 1188
{
	int i;
1189

1190 1191
	if (hstate_is_gigantic(h) && !gigantic_page_supported())
		return;
1192

1193 1194 1195
	h->nr_huge_pages--;
	h->nr_huge_pages_node[page_to_nid(page)]--;
	for (i = 0; i < pages_per_huge_page(h); i++) {
1196 1197
		page[i].flags &= ~(1 << PG_locked | 1 << PG_error |
				1 << PG_referenced | 1 << PG_dirty |
1198 1199
				1 << PG_active | 1 << PG_private |
				1 << PG_writeback);
A
Adam Litke 已提交
1200
	}
1201
	VM_BUG_ON_PAGE(hugetlb_cgroup_from_page(page), page);
1202
	set_compound_page_dtor(page, NULL_COMPOUND_DTOR);
A
Adam Litke 已提交
1203
	set_page_refcounted(page);
1204 1205 1206 1207 1208 1209
	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
1210 1211
}

1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222
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;
}

1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247
/*
 * 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]);
}

1248
void free_huge_page(struct page *page)
1249
{
1250 1251 1252 1253
	/*
	 * Can't pass hstate in here because it is called from the
	 * compound page destructor.
	 */
1254
	struct hstate *h = page_hstate(page);
1255
	int nid = page_to_nid(page);
1256 1257
	struct hugepage_subpool *spool =
		(struct hugepage_subpool *)page_private(page);
1258
	bool restore_reserve;
1259

1260
	set_page_private(page, 0);
1261
	page->mapping = NULL;
1262 1263
	VM_BUG_ON_PAGE(page_count(page), page);
	VM_BUG_ON_PAGE(page_mapcount(page), page);
1264
	restore_reserve = PagePrivate(page);
1265
	ClearPagePrivate(page);
1266

1267 1268 1269 1270 1271 1272 1273 1274
	/*
	 * 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;

1275
	spin_lock(&hugetlb_lock);
1276
	clear_page_huge_active(page);
1277 1278
	hugetlb_cgroup_uncharge_page(hstate_index(h),
				     pages_per_huge_page(h), page);
1279 1280 1281
	if (restore_reserve)
		h->resv_huge_pages++;

1282
	if (h->surplus_huge_pages_node[nid]) {
1283 1284
		/* remove the page from active list */
		list_del(&page->lru);
1285 1286 1287
		update_and_free_page(h, page);
		h->surplus_huge_pages--;
		h->surplus_huge_pages_node[nid]--;
1288
	} else {
1289
		arch_clear_hugepage_flags(page);
1290
		enqueue_huge_page(h, page);
1291
	}
1292 1293 1294
	spin_unlock(&hugetlb_lock);
}

1295
static void prep_new_huge_page(struct hstate *h, struct page *page, int nid)
1296
{
1297
	INIT_LIST_HEAD(&page->lru);
1298
	set_compound_page_dtor(page, HUGETLB_PAGE_DTOR);
1299
	spin_lock(&hugetlb_lock);
1300
	set_hugetlb_cgroup(page, NULL);
1301 1302
	h->nr_huge_pages++;
	h->nr_huge_pages_node[nid]++;
1303 1304 1305 1306
	spin_unlock(&hugetlb_lock);
	put_page(page); /* free it into the hugepage allocator */
}

1307
static void prep_compound_gigantic_page(struct page *page, unsigned int order)
1308 1309 1310 1311 1312 1313 1314
{
	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);
1315
	__ClearPageReserved(page);
1316
	__SetPageHead(page);
1317
	for (i = 1; i < nr_pages; i++, p = mem_map_next(p, page, i)) {
1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330
		/*
		 * 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);
1331
		set_page_count(p, 0);
1332
		set_compound_head(p, page);
1333
	}
1334
	atomic_set(compound_mapcount_ptr(page), -1);
1335 1336
}

A
Andrew Morton 已提交
1337 1338 1339 1340 1341
/*
 * PageHuge() only returns true for hugetlbfs pages, but not for normal or
 * transparent huge pages.  See the PageTransHuge() documentation for more
 * details.
 */
1342 1343 1344 1345 1346 1347
int PageHuge(struct page *page)
{
	if (!PageCompound(page))
		return 0;

	page = compound_head(page);
1348
	return page[1].compound_dtor == HUGETLB_PAGE_DTOR;
1349
}
1350 1351
EXPORT_SYMBOL_GPL(PageHuge);

1352 1353 1354 1355 1356 1357 1358 1359 1360
/*
 * 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;

1361
	return get_compound_page_dtor(page_head) == free_huge_page;
1362 1363
}

1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380
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;
}

1381
static struct page *alloc_fresh_huge_page_node(struct hstate *h, int nid)
L
Linus Torvalds 已提交
1382 1383
{
	struct page *page;
1384

1385
	page = __alloc_pages_node(nid,
1386
		htlb_alloc_mask(h)|__GFP_COMP|__GFP_THISNODE|
1387
						__GFP_REPEAT|__GFP_NOWARN,
1388
		huge_page_order(h));
L
Linus Torvalds 已提交
1389
	if (page) {
1390
		prep_new_huge_page(h, page, nid);
L
Linus Torvalds 已提交
1391
	}
1392 1393 1394 1395

	return page;
}

1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417
static int alloc_fresh_huge_page(struct hstate *h, nodemask_t *nodes_allowed)
{
	struct page *page;
	int nr_nodes, node;
	int ret = 0;

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

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

	return ret;
}

1418 1419 1420 1421 1422 1423
/*
 * 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.
 */
1424 1425
static int free_pool_huge_page(struct hstate *h, nodemask_t *nodes_allowed,
							 bool acct_surplus)
1426
{
1427
	int nr_nodes, node;
1428 1429
	int ret = 0;

1430
	for_each_node_mask_to_free(h, nr_nodes, node, nodes_allowed) {
1431 1432 1433 1434
		/*
		 * If we're returning unused surplus pages, only examine
		 * nodes with surplus pages.
		 */
1435 1436
		if ((!acct_surplus || h->surplus_huge_pages_node[node]) &&
		    !list_empty(&h->hugepage_freelists[node])) {
1437
			struct page *page =
1438
				list_entry(h->hugepage_freelists[node].next,
1439 1440 1441
					  struct page, lru);
			list_del(&page->lru);
			h->free_huge_pages--;
1442
			h->free_huge_pages_node[node]--;
1443 1444
			if (acct_surplus) {
				h->surplus_huge_pages--;
1445
				h->surplus_huge_pages_node[node]--;
1446
			}
1447 1448
			update_and_free_page(h, page);
			ret = 1;
1449
			break;
1450
		}
1451
	}
1452 1453 1454 1455

	return ret;
}

1456 1457
/*
 * Dissolve a given free hugepage into free buddy pages. This function does
1458 1459 1460
 * nothing for in-use (including surplus) hugepages. Returns -EBUSY if the
 * number of free hugepages would be reduced below the number of reserved
 * hugepages.
1461
 */
1462
static int dissolve_free_huge_page(struct page *page)
1463
{
1464 1465
	int rc = 0;

1466 1467
	spin_lock(&hugetlb_lock);
	if (PageHuge(page) && !page_count(page)) {
1468 1469 1470
		struct page *head = compound_head(page);
		struct hstate *h = page_hstate(head);
		int nid = page_to_nid(head);
1471 1472 1473 1474
		if (h->free_huge_pages - h->resv_huge_pages == 0) {
			rc = -EBUSY;
			goto out;
		}
1475
		list_del(&head->lru);
1476 1477
		h->free_huge_pages--;
		h->free_huge_pages_node[nid]--;
1478
		h->max_huge_pages--;
1479
		update_and_free_page(h, head);
1480
	}
1481
out:
1482
	spin_unlock(&hugetlb_lock);
1483
	return rc;
1484 1485 1486 1487 1488
}

/*
 * Dissolve free hugepages in a given pfn range. Used by memory hotplug to
 * make specified memory blocks removable from the system.
1489 1490
 * Note that this will dissolve a free gigantic hugepage completely, if any
 * part of it lies within the given range.
1491 1492
 * Also note that if dissolve_free_huge_page() returns with an error, all
 * free hugepages that were dissolved before that error are lost.
1493
 */
1494
int dissolve_free_huge_pages(unsigned long start_pfn, unsigned long end_pfn)
1495 1496
{
	unsigned long pfn;
1497
	struct page *page;
1498
	int rc = 0;
1499

1500
	if (!hugepages_supported())
1501
		return rc;
1502

1503 1504 1505 1506 1507 1508 1509 1510
	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;
		}
	}
1511 1512

	return rc;
1513 1514
}

1515 1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532
/*
 * There are 3 ways this can get called:
 * 1. With vma+addr: we use the VMA's memory policy
 * 2. With !vma, but nid=NUMA_NO_NODE:  We try to allocate a huge
 *    page from any node, and let the buddy allocator itself figure
 *    it out.
 * 3. With !vma, but nid!=NUMA_NO_NODE.  We allocate a huge page
 *    strictly from 'nid'
 */
static struct page *__hugetlb_alloc_buddy_huge_page(struct hstate *h,
		struct vm_area_struct *vma, unsigned long addr, int nid)
{
	int order = huge_page_order(h);
	gfp_t gfp = htlb_alloc_mask(h)|__GFP_COMP|__GFP_REPEAT|__GFP_NOWARN;
	unsigned int cpuset_mems_cookie;

	/*
	 * We need a VMA to get a memory policy.  If we do not
D
Dave Hansen 已提交
1533 1534 1535 1536 1537 1538
	 * have one, we use the 'nid' argument.
	 *
	 * The mempolicy stuff below has some non-inlined bits
	 * and calls ->vm_ops.  That makes it hard to optimize at
	 * compile-time, even when NUMA is off and it does
	 * nothing.  This helps the compiler optimize it out.
1539
	 */
D
Dave Hansen 已提交
1540
	if (!IS_ENABLED(CONFIG_NUMA) || !vma) {
1541 1542 1543 1544 1545 1546 1547 1548 1549 1550 1551 1552 1553 1554 1555 1556
		/*
		 * If a specific node is requested, make sure to
		 * get memory from there, but only when a node
		 * is explicitly specified.
		 */
		if (nid != NUMA_NO_NODE)
			gfp |= __GFP_THISNODE;
		/*
		 * Make sure to call something that can handle
		 * nid=NUMA_NO_NODE
		 */
		return alloc_pages_node(nid, gfp, order);
	}

	/*
	 * OK, so we have a VMA.  Fetch the mempolicy and try to
D
Dave Hansen 已提交
1557 1558
	 * allocate a huge page with it.  We will only reach this
	 * when CONFIG_NUMA=y.
1559 1560 1561 1562
	 */
	do {
		struct page *page;
		struct mempolicy *mpol;
1563
		int nid;
1564 1565 1566
		nodemask_t *nodemask;

		cpuset_mems_cookie = read_mems_allowed_begin();
1567
		nid = huge_node(vma, addr, gfp, &mpol, &nodemask);
1568
		mpol_cond_put(mpol);
1569
		page = __alloc_pages_nodemask(gfp, order, nid, nodemask);
1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590
		if (page)
			return page;
	} while (read_mems_allowed_retry(cpuset_mems_cookie));

	return NULL;
}

/*
 * There are two ways to allocate a huge page:
 * 1. When you have a VMA and an address (like a fault)
 * 2. When you have no VMA (like when setting /proc/.../nr_hugepages)
 *
 * 'vma' and 'addr' are only for (1).  'nid' is always NUMA_NO_NODE in
 * this case which signifies that the allocation should be done with
 * respect for the VMA's memory policy.
 *
 * For (2), we ignore 'vma' and 'addr' and use 'nid' exclusively. This
 * implies that memory policies will not be taken in to account.
 */
static struct page *__alloc_buddy_huge_page(struct hstate *h,
		struct vm_area_struct *vma, unsigned long addr, int nid)
1591 1592
{
	struct page *page;
1593
	unsigned int r_nid;
1594

1595
	if (hstate_is_gigantic(h))
1596 1597
		return NULL;

1598 1599 1600 1601 1602 1603
	/*
	 * Make sure that anyone specifying 'nid' is not also specifying a VMA.
	 * This makes sure the caller is picking _one_ of the modes with which
	 * we can call this function, not both.
	 */
	if (vma || (addr != -1)) {
D
Dave Hansen 已提交
1604 1605
		VM_WARN_ON_ONCE(addr == -1);
		VM_WARN_ON_ONCE(nid != NUMA_NO_NODE);
1606
	}
1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622 1623 1624 1625 1626 1627 1628 1629 1630
	/*
	 * Assume we will successfully allocate the surplus page to
	 * prevent racing processes from causing the surplus to exceed
	 * overcommit
	 *
	 * This however introduces a different race, where a process B
	 * tries to grow the static hugepage pool while alloc_pages() is
	 * called by process A. B will only examine the per-node
	 * counters in determining if surplus huge pages can be
	 * converted to normal huge pages in adjust_pool_surplus(). A
	 * won't be able to increment the per-node counter, until the
	 * lock is dropped by B, but B doesn't drop hugetlb_lock until
	 * no more huge pages can be converted from surplus to normal
	 * state (and doesn't try to convert again). Thus, we have a
	 * case where a surplus huge page exists, the pool is grown, and
	 * the surplus huge page still exists after, even though it
	 * should just have been converted to a normal huge page. This
	 * does not leak memory, though, as the hugepage will be freed
	 * once it is out of use. It also does not allow the counters to
	 * go out of whack in adjust_pool_surplus() as we don't modify
	 * the node values until we've gotten the hugepage and only the
	 * per-node value is checked there.
	 */
	spin_lock(&hugetlb_lock);
1631
	if (h->surplus_huge_pages >= h->nr_overcommit_huge_pages) {
1632 1633 1634
		spin_unlock(&hugetlb_lock);
		return NULL;
	} else {
1635 1636
		h->nr_huge_pages++;
		h->surplus_huge_pages++;
1637 1638 1639
	}
	spin_unlock(&hugetlb_lock);

1640
	page = __hugetlb_alloc_buddy_huge_page(h, vma, addr, nid);
1641 1642

	spin_lock(&hugetlb_lock);
1643
	if (page) {
1644
		INIT_LIST_HEAD(&page->lru);
1645
		r_nid = page_to_nid(page);
1646
		set_compound_page_dtor(page, HUGETLB_PAGE_DTOR);
1647
		set_hugetlb_cgroup(page, NULL);
1648 1649 1650
		/*
		 * We incremented the global counters already
		 */
1651 1652
		h->nr_huge_pages_node[r_nid]++;
		h->surplus_huge_pages_node[r_nid]++;
1653
		__count_vm_event(HTLB_BUDDY_PGALLOC);
1654
	} else {
1655 1656
		h->nr_huge_pages--;
		h->surplus_huge_pages--;
1657
		__count_vm_event(HTLB_BUDDY_PGALLOC_FAIL);
1658
	}
1659
	spin_unlock(&hugetlb_lock);
1660 1661 1662 1663

	return page;
}

1664 1665 1666 1667 1668
/*
 * Allocate a huge page from 'nid'.  Note, 'nid' may be
 * NUMA_NO_NODE, which means that it may be allocated
 * anywhere.
 */
D
Dave Hansen 已提交
1669
static
1670 1671 1672 1673 1674 1675 1676 1677 1678 1679
struct page *__alloc_buddy_huge_page_no_mpol(struct hstate *h, int nid)
{
	unsigned long addr = -1;

	return __alloc_buddy_huge_page(h, NULL, addr, nid);
}

/*
 * Use the VMA's mpolicy to allocate a huge page from the buddy.
 */
D
Dave Hansen 已提交
1680
static
1681 1682 1683 1684 1685 1686
struct page *__alloc_buddy_huge_page_with_mpol(struct hstate *h,
		struct vm_area_struct *vma, unsigned long addr)
{
	return __alloc_buddy_huge_page(h, vma, addr, NUMA_NO_NODE);
}

1687 1688 1689 1690 1691 1692 1693
/*
 * This allocation function is useful in the context where vma is irrelevant.
 * E.g. soft-offlining uses this function because it only cares physical
 * address of error page.
 */
struct page *alloc_huge_page_node(struct hstate *h, int nid)
{
1694
	struct page *page = NULL;
1695 1696

	spin_lock(&hugetlb_lock);
1697 1698
	if (h->free_huge_pages - h->resv_huge_pages > 0)
		page = dequeue_huge_page_node(h, nid);
1699 1700
	spin_unlock(&hugetlb_lock);

1701
	if (!page)
1702
		page = __alloc_buddy_huge_page_no_mpol(h, nid);
1703 1704 1705 1706

	return page;
}

1707
/*
L
Lucas De Marchi 已提交
1708
 * Increase the hugetlb pool such that it can accommodate a reservation
1709 1710
 * of size 'delta'.
 */
1711
static int gather_surplus_pages(struct hstate *h, int delta)
1712 1713 1714 1715 1716
{
	struct list_head surplus_list;
	struct page *page, *tmp;
	int ret, i;
	int needed, allocated;
1717
	bool alloc_ok = true;
1718

1719
	needed = (h->resv_huge_pages + delta) - h->free_huge_pages;
1720
	if (needed <= 0) {
1721
		h->resv_huge_pages += delta;
1722
		return 0;
1723
	}
1724 1725 1726 1727 1728 1729 1730 1731

	allocated = 0;
	INIT_LIST_HEAD(&surplus_list);

	ret = -ENOMEM;
retry:
	spin_unlock(&hugetlb_lock);
	for (i = 0; i < needed; i++) {
1732
		page = __alloc_buddy_huge_page_no_mpol(h, NUMA_NO_NODE);
1733 1734 1735 1736
		if (!page) {
			alloc_ok = false;
			break;
		}
1737 1738
		list_add(&page->lru, &surplus_list);
	}
1739
	allocated += i;
1740 1741 1742 1743 1744 1745

	/*
	 * 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);
1746 1747
	needed = (h->resv_huge_pages + delta) -
			(h->free_huge_pages + allocated);
1748 1749 1750 1751 1752 1753 1754 1755 1756 1757
	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;
	}
1758 1759
	/*
	 * The surplus_list now contains _at_least_ the number of extra pages
L
Lucas De Marchi 已提交
1760
	 * needed to accommodate the reservation.  Add the appropriate number
1761
	 * of pages to the hugetlb pool and free the extras back to the buddy
1762 1763 1764
	 * 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.
1765 1766
	 */
	needed += allocated;
1767
	h->resv_huge_pages += delta;
1768
	ret = 0;
1769

1770
	/* Free the needed pages to the hugetlb pool */
1771
	list_for_each_entry_safe(page, tmp, &surplus_list, lru) {
1772 1773
		if ((--needed) < 0)
			break;
1774 1775 1776 1777 1778
		/*
		 * This page is now managed by the hugetlb allocator and has
		 * no users -- drop the buddy allocator's reference.
		 */
		put_page_testzero(page);
1779
		VM_BUG_ON_PAGE(page_count(page), page);
1780
		enqueue_huge_page(h, page);
1781
	}
1782
free:
1783
	spin_unlock(&hugetlb_lock);
1784 1785

	/* Free unnecessary surplus pages to the buddy allocator */
1786 1787
	list_for_each_entry_safe(page, tmp, &surplus_list, lru)
		put_page(page);
1788
	spin_lock(&hugetlb_lock);
1789 1790 1791 1792 1793

	return ret;
}

/*
1794 1795 1796 1797 1798 1799 1800 1801 1802 1803 1804 1805
 * 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.
1806
 */
1807 1808
static void return_unused_surplus_pages(struct hstate *h,
					unsigned long unused_resv_pages)
1809 1810 1811
{
	unsigned long nr_pages;

1812
	/* Cannot return gigantic pages currently */
1813
	if (hstate_is_gigantic(h))
1814
		goto out;
1815

1816 1817 1818 1819
	/*
	 * Part (or even all) of the reservation could have been backed
	 * by pre-allocated pages. Only free surplus pages.
	 */
1820
	nr_pages = min(unused_resv_pages, h->surplus_huge_pages);
1821

1822 1823
	/*
	 * We want to release as many surplus pages as possible, spread
1824 1825 1826 1827 1828
	 * 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.
1829 1830 1831 1832
	 *
	 * 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.
1833 1834
	 */
	while (nr_pages--) {
1835 1836
		h->resv_huge_pages--;
		unused_resv_pages--;
1837
		if (!free_pool_huge_page(h, &node_states[N_MEMORY], 1))
1838
			goto out;
1839
		cond_resched_lock(&hugetlb_lock);
1840
	}
1841 1842 1843 1844

out:
	/* Fully uncommit the reservation */
	h->resv_huge_pages -= unused_resv_pages;
1845 1846
}

1847

1848
/*
1849
 * vma_needs_reservation, vma_commit_reservation and vma_end_reservation
1850
 * are used by the huge page allocation routines to manage reservations.
1851 1852 1853 1854 1855 1856
 *
 * 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
1857 1858 1859
 * 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.
1860 1861 1862 1863 1864 1865
 *
 * 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.
1866 1867 1868 1869 1870
 *
 * 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.
1871
 */
1872 1873 1874
enum vma_resv_mode {
	VMA_NEEDS_RESV,
	VMA_COMMIT_RESV,
1875
	VMA_END_RESV,
1876
	VMA_ADD_RESV,
1877
};
1878 1879
static long __vma_reservation_common(struct hstate *h,
				struct vm_area_struct *vma, unsigned long addr,
1880
				enum vma_resv_mode mode)
1881
{
1882 1883
	struct resv_map *resv;
	pgoff_t idx;
1884
	long ret;
1885

1886 1887
	resv = vma_resv_map(vma);
	if (!resv)
1888
		return 1;
1889

1890
	idx = vma_hugecache_offset(h, vma, addr);
1891 1892
	switch (mode) {
	case VMA_NEEDS_RESV:
1893
		ret = region_chg(resv, idx, idx + 1);
1894 1895 1896 1897
		break;
	case VMA_COMMIT_RESV:
		ret = region_add(resv, idx, idx + 1);
		break;
1898
	case VMA_END_RESV:
1899 1900 1901
		region_abort(resv, idx, idx + 1);
		ret = 0;
		break;
1902 1903 1904 1905 1906 1907 1908 1909
	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;
1910 1911 1912
	default:
		BUG();
	}
1913

1914
	if (vma->vm_flags & VM_MAYSHARE)
1915
		return ret;
1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931 1932 1933 1934
	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;
	}
1935
	else
1936
		return ret < 0 ? ret : 0;
1937
}
1938 1939

static long vma_needs_reservation(struct hstate *h,
1940
			struct vm_area_struct *vma, unsigned long addr)
1941
{
1942
	return __vma_reservation_common(h, vma, addr, VMA_NEEDS_RESV);
1943
}
1944

1945 1946 1947
static long vma_commit_reservation(struct hstate *h,
			struct vm_area_struct *vma, unsigned long addr)
{
1948 1949 1950
	return __vma_reservation_common(h, vma, addr, VMA_COMMIT_RESV);
}

1951
static void vma_end_reservation(struct hstate *h,
1952 1953
			struct vm_area_struct *vma, unsigned long addr)
{
1954
	(void)__vma_reservation_common(h, vma, addr, VMA_END_RESV);
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 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006
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);
	}
}

2007
struct page *alloc_huge_page(struct vm_area_struct *vma,
2008
				    unsigned long addr, int avoid_reserve)
L
Linus Torvalds 已提交
2009
{
2010
	struct hugepage_subpool *spool = subpool_vma(vma);
2011
	struct hstate *h = hstate_vma(vma);
2012
	struct page *page;
2013 2014
	long map_chg, map_commit;
	long gbl_chg;
2015 2016
	int ret, idx;
	struct hugetlb_cgroup *h_cg;
2017

2018
	idx = hstate_index(h);
2019
	/*
2020 2021 2022
	 * 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).
2023
	 */
2024 2025
	map_chg = gbl_chg = vma_needs_reservation(h, vma, addr);
	if (map_chg < 0)
2026
		return ERR_PTR(-ENOMEM);
2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037

	/*
	 * 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) {
2038
			vma_end_reservation(h, vma, addr);
2039
			return ERR_PTR(-ENOSPC);
2040
		}
L
Linus Torvalds 已提交
2041

2042 2043 2044 2045 2046 2047 2048 2049 2050 2051 2052 2053
		/*
		 * 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;
	}

2054
	ret = hugetlb_cgroup_charge_cgroup(idx, pages_per_huge_page(h), &h_cg);
2055 2056 2057
	if (ret)
		goto out_subpool_put;

L
Linus Torvalds 已提交
2058
	spin_lock(&hugetlb_lock);
2059 2060 2061 2062 2063 2064
	/*
	 * 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);
2065
	if (!page) {
2066
		spin_unlock(&hugetlb_lock);
2067
		page = __alloc_buddy_huge_page_with_mpol(h, vma, addr);
2068 2069
		if (!page)
			goto out_uncharge_cgroup;
2070 2071 2072 2073
		if (!avoid_reserve && vma_has_reserves(vma, gbl_chg)) {
			SetPagePrivate(page);
			h->resv_huge_pages--;
		}
2074 2075
		spin_lock(&hugetlb_lock);
		list_move(&page->lru, &h->hugepage_activelist);
2076
		/* Fall through */
K
Ken Chen 已提交
2077
	}
2078 2079
	hugetlb_cgroup_commit_charge(idx, pages_per_huge_page(h), h_cg, page);
	spin_unlock(&hugetlb_lock);
2080

2081
	set_page_private(page, (unsigned long)spool);
2082

2083 2084
	map_commit = vma_commit_reservation(h, vma, addr);
	if (unlikely(map_chg > map_commit)) {
2085 2086 2087 2088 2089 2090 2091 2092 2093 2094 2095 2096 2097 2098
		/*
		 * 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);
	}
2099
	return page;
2100 2101 2102 2103

out_uncharge_cgroup:
	hugetlb_cgroup_uncharge_cgroup(idx, pages_per_huge_page(h), h_cg);
out_subpool_put:
2104
	if (map_chg || avoid_reserve)
2105
		hugepage_subpool_put_pages(spool, 1);
2106
	vma_end_reservation(h, vma, addr);
2107
	return ERR_PTR(-ENOSPC);
2108 2109
}

2110 2111 2112 2113 2114 2115 2116 2117 2118 2119 2120 2121 2122 2123
/*
 * alloc_huge_page()'s wrapper which simply returns the page if allocation
 * succeeds, otherwise NULL. This function is called from new_vma_page(),
 * where no ERR_VALUE is expected to be returned.
 */
struct page *alloc_huge_page_noerr(struct vm_area_struct *vma,
				unsigned long addr, int avoid_reserve)
{
	struct page *page = alloc_huge_page(vma, addr, avoid_reserve);
	if (IS_ERR(page))
		page = NULL;
	return page;
}

2124
int __weak alloc_bootmem_huge_page(struct hstate *h)
2125 2126
{
	struct huge_bootmem_page *m;
2127
	int nr_nodes, node;
2128

2129
	for_each_node_mask_to_alloc(h, nr_nodes, node, &node_states[N_MEMORY]) {
2130 2131
		void *addr;

2132 2133 2134
		addr = memblock_virt_alloc_try_nid_nopanic(
				huge_page_size(h), huge_page_size(h),
				0, BOOTMEM_ALLOC_ACCESSIBLE, node);
2135 2136 2137 2138 2139 2140 2141
		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;
2142
			goto found;
2143 2144 2145 2146 2147
		}
	}
	return 0;

found:
2148
	BUG_ON(!IS_ALIGNED(virt_to_phys(m), huge_page_size(h)));
2149 2150 2151 2152 2153 2154
	/* Put them into a private list first because mem_map is not up yet */
	list_add(&m->list, &huge_boot_pages);
	m->hstate = h;
	return 1;
}

2155 2156
static void __init prep_compound_huge_page(struct page *page,
		unsigned int order)
2157 2158 2159 2160 2161 2162 2163
{
	if (unlikely(order > (MAX_ORDER - 1)))
		prep_compound_gigantic_page(page, order);
	else
		prep_compound_page(page, order);
}

2164 2165 2166 2167 2168 2169 2170
/* Put bootmem huge pages into the standard lists after mem_map is up */
static void __init gather_bootmem_prealloc(void)
{
	struct huge_bootmem_page *m;

	list_for_each_entry(m, &huge_boot_pages, list) {
		struct hstate *h = m->hstate;
2171 2172 2173 2174
		struct page *page;

#ifdef CONFIG_HIGHMEM
		page = pfn_to_page(m->phys >> PAGE_SHIFT);
2175 2176
		memblock_free_late(__pa(m),
				   sizeof(struct huge_bootmem_page));
2177 2178 2179
#else
		page = virt_to_page(m);
#endif
2180
		WARN_ON(page_count(page) != 1);
2181
		prep_compound_huge_page(page, h->order);
2182
		WARN_ON(PageReserved(page));
2183
		prep_new_huge_page(h, page, page_to_nid(page));
2184 2185 2186 2187 2188 2189
		/*
		 * 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.
		 */
2190
		if (hstate_is_gigantic(h))
2191
			adjust_managed_page_count(page, 1 << h->order);
2192 2193 2194
	}
}

2195
static void __init hugetlb_hstate_alloc_pages(struct hstate *h)
L
Linus Torvalds 已提交
2196 2197
{
	unsigned long i;
2198

2199
	for (i = 0; i < h->max_huge_pages; ++i) {
2200
		if (hstate_is_gigantic(h)) {
2201 2202
			if (!alloc_bootmem_huge_page(h))
				break;
2203
		} else if (!alloc_fresh_huge_page(h,
2204
					 &node_states[N_MEMORY]))
L
Linus Torvalds 已提交
2205 2206
			break;
	}
2207
	h->max_huge_pages = i;
2208 2209 2210 2211 2212 2213 2214
}

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

	for_each_hstate(h) {
2215 2216 2217
		if (minimum_order > huge_page_order(h))
			minimum_order = huge_page_order(h);

2218
		/* oversize hugepages were init'ed in early boot */
2219
		if (!hstate_is_gigantic(h))
2220
			hugetlb_hstate_alloc_pages(h);
2221
	}
2222
	VM_BUG_ON(minimum_order == UINT_MAX);
2223 2224
}

A
Andi Kleen 已提交
2225 2226 2227 2228 2229 2230 2231 2232 2233 2234 2235
static char * __init memfmt(char *buf, unsigned long n)
{
	if (n >= (1UL << 30))
		sprintf(buf, "%lu GB", n >> 30);
	else if (n >= (1UL << 20))
		sprintf(buf, "%lu MB", n >> 20);
	else
		sprintf(buf, "%lu KB", n >> 10);
	return buf;
}

2236 2237 2238 2239 2240
static void __init report_hugepages(void)
{
	struct hstate *h;

	for_each_hstate(h) {
A
Andi Kleen 已提交
2241
		char buf[32];
2242
		pr_info("HugeTLB registered %s page size, pre-allocated %ld pages\n",
A
Andi Kleen 已提交
2243 2244
			memfmt(buf, huge_page_size(h)),
			h->free_huge_pages);
2245 2246 2247
	}
}

L
Linus Torvalds 已提交
2248
#ifdef CONFIG_HIGHMEM
2249 2250
static void try_to_free_low(struct hstate *h, unsigned long count,
						nodemask_t *nodes_allowed)
L
Linus Torvalds 已提交
2251
{
2252 2253
	int i;

2254
	if (hstate_is_gigantic(h))
2255 2256
		return;

2257
	for_each_node_mask(i, *nodes_allowed) {
L
Linus Torvalds 已提交
2258
		struct page *page, *next;
2259 2260 2261
		struct list_head *freel = &h->hugepage_freelists[i];
		list_for_each_entry_safe(page, next, freel, lru) {
			if (count >= h->nr_huge_pages)
2262
				return;
L
Linus Torvalds 已提交
2263 2264 2265
			if (PageHighMem(page))
				continue;
			list_del(&page->lru);
2266
			update_and_free_page(h, page);
2267 2268
			h->free_huge_pages--;
			h->free_huge_pages_node[page_to_nid(page)]--;
L
Linus Torvalds 已提交
2269 2270 2271 2272
		}
	}
}
#else
2273 2274
static inline void try_to_free_low(struct hstate *h, unsigned long count,
						nodemask_t *nodes_allowed)
L
Linus Torvalds 已提交
2275 2276 2277 2278
{
}
#endif

2279 2280 2281 2282 2283
/*
 * 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.
 */
2284 2285
static int adjust_pool_surplus(struct hstate *h, nodemask_t *nodes_allowed,
				int delta)
2286
{
2287
	int nr_nodes, node;
2288 2289 2290

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

2291 2292 2293 2294
	if (delta < 0) {
		for_each_node_mask_to_alloc(h, nr_nodes, node, nodes_allowed) {
			if (h->surplus_huge_pages_node[node])
				goto found;
2295
		}
2296 2297 2298 2299 2300
	} 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;
2301
		}
2302 2303
	}
	return 0;
2304

2305 2306 2307 2308
found:
	h->surplus_huge_pages += delta;
	h->surplus_huge_pages_node[node] += delta;
	return 1;
2309 2310
}

2311
#define persistent_huge_pages(h) (h->nr_huge_pages - h->surplus_huge_pages)
2312 2313
static unsigned long set_max_huge_pages(struct hstate *h, unsigned long count,
						nodemask_t *nodes_allowed)
L
Linus Torvalds 已提交
2314
{
2315
	unsigned long min_count, ret;
L
Linus Torvalds 已提交
2316

2317
	if (hstate_is_gigantic(h) && !gigantic_page_supported())
2318 2319
		return h->max_huge_pages;

2320 2321 2322 2323
	/*
	 * Increase the pool size
	 * First take pages out of surplus state.  Then make up the
	 * remaining difference by allocating fresh huge pages.
2324
	 *
N
Naoya Horiguchi 已提交
2325
	 * We might race with __alloc_buddy_huge_page() here and be unable
2326 2327 2328 2329
	 * 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.
2330
	 */
L
Linus Torvalds 已提交
2331
	spin_lock(&hugetlb_lock);
2332
	while (h->surplus_huge_pages && count > persistent_huge_pages(h)) {
2333
		if (!adjust_pool_surplus(h, nodes_allowed, -1))
2334 2335 2336
			break;
	}

2337
	while (count > persistent_huge_pages(h)) {
2338 2339 2340 2341 2342 2343
		/*
		 * 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);
2344 2345 2346 2347

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

2348 2349 2350 2351
		if (hstate_is_gigantic(h))
			ret = alloc_fresh_gigantic_page(h, nodes_allowed);
		else
			ret = alloc_fresh_huge_page(h, nodes_allowed);
2352 2353 2354 2355
		spin_lock(&hugetlb_lock);
		if (!ret)
			goto out;

2356 2357 2358
		/* Bail for signals. Probably ctrl-c from user */
		if (signal_pending(current))
			goto out;
2359 2360 2361 2362 2363 2364 2365 2366
	}

	/*
	 * 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.
2367 2368 2369 2370
	 *
	 * 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
N
Naoya Horiguchi 已提交
2371
	 * __alloc_buddy_huge_page() is checking the global counter,
2372 2373 2374
	 * 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.
2375
	 */
2376
	min_count = h->resv_huge_pages + h->nr_huge_pages - h->free_huge_pages;
2377
	min_count = max(count, min_count);
2378
	try_to_free_low(h, min_count, nodes_allowed);
2379
	while (min_count < persistent_huge_pages(h)) {
2380
		if (!free_pool_huge_page(h, nodes_allowed, 0))
L
Linus Torvalds 已提交
2381
			break;
2382
		cond_resched_lock(&hugetlb_lock);
L
Linus Torvalds 已提交
2383
	}
2384
	while (count < persistent_huge_pages(h)) {
2385
		if (!adjust_pool_surplus(h, nodes_allowed, 1))
2386 2387 2388
			break;
	}
out:
2389
	ret = persistent_huge_pages(h);
L
Linus Torvalds 已提交
2390
	spin_unlock(&hugetlb_lock);
2391
	return ret;
L
Linus Torvalds 已提交
2392 2393
}

2394 2395 2396 2397 2398 2399 2400 2401 2402 2403
#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];

2404 2405 2406
static struct hstate *kobj_to_node_hstate(struct kobject *kobj, int *nidp);

static struct hstate *kobj_to_hstate(struct kobject *kobj, int *nidp)
2407 2408
{
	int i;
2409

2410
	for (i = 0; i < HUGE_MAX_HSTATE; i++)
2411 2412 2413
		if (hstate_kobjs[i] == kobj) {
			if (nidp)
				*nidp = NUMA_NO_NODE;
2414
			return &hstates[i];
2415 2416 2417
		}

	return kobj_to_node_hstate(kobj, nidp);
2418 2419
}

2420
static ssize_t nr_hugepages_show_common(struct kobject *kobj,
2421 2422
					struct kobj_attribute *attr, char *buf)
{
2423 2424 2425 2426 2427 2428 2429 2430 2431 2432 2433
	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);
2434
}
2435

2436 2437 2438
static ssize_t __nr_hugepages_store_common(bool obey_mempolicy,
					   struct hstate *h, int nid,
					   unsigned long count, size_t len)
2439 2440
{
	int err;
2441
	NODEMASK_ALLOC(nodemask_t, nodes_allowed, GFP_KERNEL | __GFP_NORETRY);
2442

2443
	if (hstate_is_gigantic(h) && !gigantic_page_supported()) {
2444 2445 2446 2447
		err = -EINVAL;
		goto out;
	}

2448 2449 2450 2451 2452 2453 2454
	if (nid == NUMA_NO_NODE) {
		/*
		 * global hstate attribute
		 */
		if (!(obey_mempolicy &&
				init_nodemask_of_mempolicy(nodes_allowed))) {
			NODEMASK_FREE(nodes_allowed);
2455
			nodes_allowed = &node_states[N_MEMORY];
2456 2457 2458 2459 2460 2461 2462 2463 2464
		}
	} 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
2465
		nodes_allowed = &node_states[N_MEMORY];
2466

2467
	h->max_huge_pages = set_max_huge_pages(h, count, nodes_allowed);
2468

2469
	if (nodes_allowed != &node_states[N_MEMORY])
2470 2471 2472
		NODEMASK_FREE(nodes_allowed);

	return len;
2473 2474 2475
out:
	NODEMASK_FREE(nodes_allowed);
	return err;
2476 2477
}

2478 2479 2480 2481 2482 2483 2484 2485 2486 2487 2488 2489 2490 2491 2492 2493 2494
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);
}

2495 2496 2497 2498 2499 2500 2501 2502 2503
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)
{
2504
	return nr_hugepages_store_common(false, kobj, buf, len);
2505 2506 2507
}
HSTATE_ATTR(nr_hugepages);

2508 2509 2510 2511 2512 2513 2514 2515 2516 2517 2518 2519 2520 2521 2522
#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)
{
2523
	return nr_hugepages_store_common(true, kobj, buf, len);
2524 2525 2526 2527 2528
}
HSTATE_ATTR(nr_hugepages_mempolicy);
#endif


2529 2530 2531
static ssize_t nr_overcommit_hugepages_show(struct kobject *kobj,
					struct kobj_attribute *attr, char *buf)
{
2532
	struct hstate *h = kobj_to_hstate(kobj, NULL);
2533 2534
	return sprintf(buf, "%lu\n", h->nr_overcommit_huge_pages);
}
2535

2536 2537 2538 2539 2540
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;
2541
	struct hstate *h = kobj_to_hstate(kobj, NULL);
2542

2543
	if (hstate_is_gigantic(h))
2544 2545
		return -EINVAL;

2546
	err = kstrtoul(buf, 10, &input);
2547
	if (err)
2548
		return err;
2549 2550 2551 2552 2553 2554 2555 2556 2557 2558 2559 2560

	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)
{
2561 2562 2563 2564 2565 2566 2567 2568 2569 2570 2571
	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);
2572 2573 2574 2575 2576 2577
}
HSTATE_ATTR_RO(free_hugepages);

static ssize_t resv_hugepages_show(struct kobject *kobj,
					struct kobj_attribute *attr, char *buf)
{
2578
	struct hstate *h = kobj_to_hstate(kobj, NULL);
2579 2580 2581 2582 2583 2584 2585
	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)
{
2586 2587 2588 2589 2590 2591 2592 2593 2594 2595 2596
	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);
2597 2598 2599 2600 2601 2602 2603 2604 2605
}
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,
2606 2607 2608
#ifdef CONFIG_NUMA
	&nr_hugepages_mempolicy_attr.attr,
#endif
2609 2610 2611 2612 2613 2614 2615
	NULL,
};

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

J
Jeff Mahoney 已提交
2616 2617 2618
static int hugetlb_sysfs_add_hstate(struct hstate *h, struct kobject *parent,
				    struct kobject **hstate_kobjs,
				    struct attribute_group *hstate_attr_group)
2619 2620
{
	int retval;
2621
	int hi = hstate_index(h);
2622

2623 2624
	hstate_kobjs[hi] = kobject_create_and_add(h->name, parent);
	if (!hstate_kobjs[hi])
2625 2626
		return -ENOMEM;

2627
	retval = sysfs_create_group(hstate_kobjs[hi], hstate_attr_group);
2628
	if (retval)
2629
		kobject_put(hstate_kobjs[hi]);
2630 2631 2632 2633 2634 2635 2636 2637 2638 2639 2640 2641 2642 2643

	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) {
2644 2645
		err = hugetlb_sysfs_add_hstate(h, hugepages_kobj,
					 hstate_kobjs, &hstate_attr_group);
2646
		if (err)
2647
			pr_err("Hugetlb: Unable to add hstate %s", h->name);
2648 2649 2650
	}
}

2651 2652 2653 2654
#ifdef CONFIG_NUMA

/*
 * node_hstate/s - associate per node hstate attributes, via their kobjects,
2655 2656 2657
 * 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
2658 2659 2660 2661 2662 2663
 * the base kernel, on the hugetlb module.
 */
struct node_hstate {
	struct kobject		*hugepages_kobj;
	struct kobject		*hstate_kobjs[HUGE_MAX_HSTATE];
};
2664
static struct node_hstate node_hstates[MAX_NUMNODES];
2665 2666

/*
2667
 * A subset of global hstate attributes for node devices
2668 2669 2670 2671 2672 2673 2674 2675 2676 2677 2678 2679 2680
 */
static struct attribute *per_node_hstate_attrs[] = {
	&nr_hugepages_attr.attr,
	&free_hugepages_attr.attr,
	&surplus_hugepages_attr.attr,
	NULL,
};

static struct attribute_group per_node_hstate_attr_group = {
	.attrs = per_node_hstate_attrs,
};

/*
2681
 * kobj_to_node_hstate - lookup global hstate for node device hstate attr kobj.
2682 2683 2684 2685 2686 2687 2688 2689 2690 2691 2692 2693 2694 2695 2696 2697 2698 2699 2700 2701 2702 2703
 * 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;
}

/*
2704
 * Unregister hstate attributes from a single node device.
2705 2706
 * No-op if no hstate attributes attached.
 */
2707
static void hugetlb_unregister_node(struct node *node)
2708 2709
{
	struct hstate *h;
2710
	struct node_hstate *nhs = &node_hstates[node->dev.id];
2711 2712

	if (!nhs->hugepages_kobj)
2713
		return;		/* no hstate attributes */
2714

2715 2716 2717 2718 2719
	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;
2720
		}
2721
	}
2722 2723 2724 2725 2726 2727 2728

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


/*
2729
 * Register hstate attributes for a single node device.
2730 2731
 * No-op if attributes already registered.
 */
2732
static void hugetlb_register_node(struct node *node)
2733 2734
{
	struct hstate *h;
2735
	struct node_hstate *nhs = &node_hstates[node->dev.id];
2736 2737 2738 2739 2740 2741
	int err;

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

	nhs->hugepages_kobj = kobject_create_and_add("hugepages",
2742
							&node->dev.kobj);
2743 2744 2745 2746 2747 2748 2749 2750
	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) {
2751 2752
			pr_err("Hugetlb: Unable to add hstate %s for node %d\n",
				h->name, node->dev.id);
2753 2754 2755 2756 2757 2758 2759
			hugetlb_unregister_node(node);
			break;
		}
	}
}

/*
2760
 * hugetlb init time:  register hstate attributes for all registered node
2761 2762
 * devices of nodes that have memory.  All on-line nodes should have
 * registered their associated device by this time.
2763
 */
2764
static void __init hugetlb_register_all_nodes(void)
2765 2766 2767
{
	int nid;

2768
	for_each_node_state(nid, N_MEMORY) {
2769
		struct node *node = node_devices[nid];
2770
		if (node->dev.id == nid)
2771 2772 2773 2774
			hugetlb_register_node(node);
	}

	/*
2775
	 * Let the node device driver know we're here so it can
2776 2777 2778 2779 2780 2781 2782 2783 2784 2785 2786 2787 2788 2789 2790 2791 2792 2793 2794
	 * [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

2795 2796
static int __init hugetlb_init(void)
{
2797 2798
	int i;

2799
	if (!hugepages_supported())
2800
		return 0;
2801

2802 2803 2804 2805
	if (!size_to_hstate(default_hstate_size)) {
		default_hstate_size = HPAGE_SIZE;
		if (!size_to_hstate(default_hstate_size))
			hugetlb_add_hstate(HUGETLB_PAGE_ORDER);
2806
	}
2807
	default_hstate_idx = hstate_index(size_to_hstate(default_hstate_size));
2808 2809 2810 2811
	if (default_hstate_max_huge_pages) {
		if (!default_hstate.max_huge_pages)
			default_hstate.max_huge_pages = default_hstate_max_huge_pages;
	}
2812 2813

	hugetlb_init_hstates();
2814
	gather_bootmem_prealloc();
2815 2816 2817
	report_hugepages();

	hugetlb_sysfs_init();
2818
	hugetlb_register_all_nodes();
2819
	hugetlb_cgroup_file_init();
2820

2821 2822 2823 2824 2825
#ifdef CONFIG_SMP
	num_fault_mutexes = roundup_pow_of_two(8 * num_possible_cpus());
#else
	num_fault_mutexes = 1;
#endif
2826
	hugetlb_fault_mutex_table =
2827
		kmalloc(sizeof(struct mutex) * num_fault_mutexes, GFP_KERNEL);
2828
	BUG_ON(!hugetlb_fault_mutex_table);
2829 2830

	for (i = 0; i < num_fault_mutexes; i++)
2831
		mutex_init(&hugetlb_fault_mutex_table[i]);
2832 2833
	return 0;
}
2834
subsys_initcall(hugetlb_init);
2835 2836

/* Should be called on processing a hugepagesz=... option */
2837 2838 2839 2840 2841
void __init hugetlb_bad_size(void)
{
	parsed_valid_hugepagesz = false;
}

2842
void __init hugetlb_add_hstate(unsigned int order)
2843 2844
{
	struct hstate *h;
2845 2846
	unsigned long i;

2847
	if (size_to_hstate(PAGE_SIZE << order)) {
J
Joe Perches 已提交
2848
		pr_warn("hugepagesz= specified twice, ignoring\n");
2849 2850
		return;
	}
2851
	BUG_ON(hugetlb_max_hstate >= HUGE_MAX_HSTATE);
2852
	BUG_ON(order == 0);
2853
	h = &hstates[hugetlb_max_hstate++];
2854 2855
	h->order = order;
	h->mask = ~((1ULL << (order + PAGE_SHIFT)) - 1);
2856 2857 2858 2859
	h->nr_huge_pages = 0;
	h->free_huge_pages = 0;
	for (i = 0; i < MAX_NUMNODES; ++i)
		INIT_LIST_HEAD(&h->hugepage_freelists[i]);
2860
	INIT_LIST_HEAD(&h->hugepage_activelist);
2861 2862
	h->next_nid_to_alloc = first_memory_node;
	h->next_nid_to_free = first_memory_node;
2863 2864
	snprintf(h->name, HSTATE_NAME_LEN, "hugepages-%lukB",
					huge_page_size(h)/1024);
2865

2866 2867 2868
	parsed_hstate = h;
}

2869
static int __init hugetlb_nrpages_setup(char *s)
2870 2871
{
	unsigned long *mhp;
2872
	static unsigned long *last_mhp;
2873

2874 2875 2876 2877 2878 2879
	if (!parsed_valid_hugepagesz) {
		pr_warn("hugepages = %s preceded by "
			"an unsupported hugepagesz, ignoring\n", s);
		parsed_valid_hugepagesz = true;
		return 1;
	}
2880
	/*
2881
	 * !hugetlb_max_hstate means we haven't parsed a hugepagesz= parameter yet,
2882 2883
	 * so this hugepages= parameter goes to the "default hstate".
	 */
2884
	else if (!hugetlb_max_hstate)
2885 2886 2887 2888
		mhp = &default_hstate_max_huge_pages;
	else
		mhp = &parsed_hstate->max_huge_pages;

2889
	if (mhp == last_mhp) {
J
Joe Perches 已提交
2890
		pr_warn("hugepages= specified twice without interleaving hugepagesz=, ignoring\n");
2891 2892 2893
		return 1;
	}

2894 2895 2896
	if (sscanf(s, "%lu", mhp) <= 0)
		*mhp = 0;

2897 2898 2899 2900 2901
	/*
	 * 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.
	 */
2902
	if (hugetlb_max_hstate && parsed_hstate->order >= MAX_ORDER)
2903 2904 2905 2906
		hugetlb_hstate_alloc_pages(parsed_hstate);

	last_mhp = mhp;

2907 2908
	return 1;
}
2909 2910 2911 2912 2913 2914 2915 2916
__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);
2917

2918 2919 2920 2921 2922 2923 2924 2925 2926 2927 2928 2929
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
2930 2931 2932
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
2933
{
2934
	struct hstate *h = &default_hstate;
2935
	unsigned long tmp = h->max_huge_pages;
2936
	int ret;
2937

2938
	if (!hugepages_supported())
2939
		return -EOPNOTSUPP;
2940

2941 2942
	table->data = &tmp;
	table->maxlen = sizeof(unsigned long);
2943 2944 2945
	ret = proc_doulongvec_minmax(table, write, buffer, length, ppos);
	if (ret)
		goto out;
2946

2947 2948 2949
	if (write)
		ret = __nr_hugepages_store_common(obey_mempolicy, h,
						  NUMA_NO_NODE, tmp, *length);
2950 2951
out:
	return ret;
L
Linus Torvalds 已提交
2952
}
2953

2954 2955 2956 2957 2958 2959 2960 2961 2962 2963 2964 2965 2966 2967 2968 2969 2970
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 */

2971
int hugetlb_overcommit_handler(struct ctl_table *table, int write,
2972
			void __user *buffer,
2973 2974
			size_t *length, loff_t *ppos)
{
2975
	struct hstate *h = &default_hstate;
2976
	unsigned long tmp;
2977
	int ret;
2978

2979
	if (!hugepages_supported())
2980
		return -EOPNOTSUPP;
2981

2982
	tmp = h->nr_overcommit_huge_pages;
2983

2984
	if (write && hstate_is_gigantic(h))
2985 2986
		return -EINVAL;

2987 2988
	table->data = &tmp;
	table->maxlen = sizeof(unsigned long);
2989 2990 2991
	ret = proc_doulongvec_minmax(table, write, buffer, length, ppos);
	if (ret)
		goto out;
2992 2993 2994 2995 2996 2997

	if (write) {
		spin_lock(&hugetlb_lock);
		h->nr_overcommit_huge_pages = tmp;
		spin_unlock(&hugetlb_lock);
	}
2998 2999
out:
	return ret;
3000 3001
}

L
Linus Torvalds 已提交
3002 3003
#endif /* CONFIG_SYSCTL */

3004
void hugetlb_report_meminfo(struct seq_file *m)
L
Linus Torvalds 已提交
3005
{
3006
	struct hstate *h = &default_hstate;
3007 3008
	if (!hugepages_supported())
		return;
3009
	seq_printf(m,
3010 3011 3012 3013 3014
			"HugePages_Total:   %5lu\n"
			"HugePages_Free:    %5lu\n"
			"HugePages_Rsvd:    %5lu\n"
			"HugePages_Surp:    %5lu\n"
			"Hugepagesize:   %8lu kB\n",
3015 3016 3017 3018 3019
			h->nr_huge_pages,
			h->free_huge_pages,
			h->resv_huge_pages,
			h->surplus_huge_pages,
			1UL << (huge_page_order(h) + PAGE_SHIFT - 10));
L
Linus Torvalds 已提交
3020 3021 3022 3023
}

int hugetlb_report_node_meminfo(int nid, char *buf)
{
3024
	struct hstate *h = &default_hstate;
3025 3026
	if (!hugepages_supported())
		return 0;
L
Linus Torvalds 已提交
3027 3028
	return sprintf(buf,
		"Node %d HugePages_Total: %5u\n"
3029 3030
		"Node %d HugePages_Free:  %5u\n"
		"Node %d HugePages_Surp:  %5u\n",
3031 3032 3033
		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
3034 3035
}

3036 3037 3038 3039 3040
void hugetlb_show_meminfo(void)
{
	struct hstate *h;
	int nid;

3041 3042 3043
	if (!hugepages_supported())
		return;

3044 3045 3046 3047 3048 3049 3050 3051 3052 3053
	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));
}

3054 3055 3056 3057 3058 3059
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
3060 3061 3062
/* Return the number pages of memory we physically have, in PAGE_SIZE units. */
unsigned long hugetlb_total_pages(void)
{
3063 3064 3065 3066 3067 3068
	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
3069 3070
}

3071
static int hugetlb_acct_memory(struct hstate *h, long delta)
M
Mel Gorman 已提交
3072 3073 3074 3075 3076 3077 3078 3079 3080 3081 3082 3083 3084 3085 3086 3087 3088 3089 3090 3091 3092 3093
{
	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) {
3094
		if (gather_surplus_pages(h, delta) < 0)
M
Mel Gorman 已提交
3095 3096
			goto out;

3097 3098
		if (delta > cpuset_mems_nr(h->free_huge_pages_node)) {
			return_unused_surplus_pages(h, delta);
M
Mel Gorman 已提交
3099 3100 3101 3102 3103 3104
			goto out;
		}
	}

	ret = 0;
	if (delta < 0)
3105
		return_unused_surplus_pages(h, (unsigned long) -delta);
M
Mel Gorman 已提交
3106 3107 3108 3109 3110 3111

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

3112 3113
static void hugetlb_vm_op_open(struct vm_area_struct *vma)
{
3114
	struct resv_map *resv = vma_resv_map(vma);
3115 3116 3117 3118 3119

	/*
	 * 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
3120
	 * has a reference to the reservation map it cannot disappear until
3121 3122 3123
	 * after this open call completes.  It is therefore safe to take a
	 * new reference here without additional locking.
	 */
3124
	if (resv && is_vma_resv_set(vma, HPAGE_RESV_OWNER))
3125
		kref_get(&resv->refs);
3126 3127
}

3128 3129
static void hugetlb_vm_op_close(struct vm_area_struct *vma)
{
3130
	struct hstate *h = hstate_vma(vma);
3131
	struct resv_map *resv = vma_resv_map(vma);
3132
	struct hugepage_subpool *spool = subpool_vma(vma);
3133
	unsigned long reserve, start, end;
3134
	long gbl_reserve;
3135

3136 3137
	if (!resv || !is_vma_resv_set(vma, HPAGE_RESV_OWNER))
		return;
3138

3139 3140
	start = vma_hugecache_offset(h, vma, vma->vm_start);
	end = vma_hugecache_offset(h, vma, vma->vm_end);
3141

3142
	reserve = (end - start) - region_count(resv, start, end);
3143

3144 3145 3146
	kref_put(&resv->refs, resv_map_release);

	if (reserve) {
3147 3148 3149 3150 3151 3152
		/*
		 * 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);
3153
	}
3154 3155
}

L
Linus Torvalds 已提交
3156 3157 3158 3159 3160 3161
/*
 * 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.
 */
3162
static int hugetlb_vm_op_fault(struct vm_fault *vmf)
L
Linus Torvalds 已提交
3163 3164
{
	BUG();
N
Nick Piggin 已提交
3165
	return 0;
L
Linus Torvalds 已提交
3166 3167
}

3168
const struct vm_operations_struct hugetlb_vm_ops = {
N
Nick Piggin 已提交
3169
	.fault = hugetlb_vm_op_fault,
3170
	.open = hugetlb_vm_op_open,
3171
	.close = hugetlb_vm_op_close,
L
Linus Torvalds 已提交
3172 3173
};

3174 3175
static pte_t make_huge_pte(struct vm_area_struct *vma, struct page *page,
				int writable)
D
David Gibson 已提交
3176 3177 3178
{
	pte_t entry;

3179
	if (writable) {
3180 3181
		entry = huge_pte_mkwrite(huge_pte_mkdirty(mk_huge_pte(page,
					 vma->vm_page_prot)));
D
David Gibson 已提交
3182
	} else {
3183 3184
		entry = huge_pte_wrprotect(mk_huge_pte(page,
					   vma->vm_page_prot));
D
David Gibson 已提交
3185 3186 3187
	}
	entry = pte_mkyoung(entry);
	entry = pte_mkhuge(entry);
3188
	entry = arch_make_huge_pte(entry, vma, page, writable);
D
David Gibson 已提交
3189 3190 3191 3192

	return entry;
}

3193 3194 3195 3196 3197
static void set_huge_ptep_writable(struct vm_area_struct *vma,
				   unsigned long address, pte_t *ptep)
{
	pte_t entry;

3198
	entry = huge_pte_mkwrite(huge_pte_mkdirty(huge_ptep_get(ptep)));
3199
	if (huge_ptep_set_access_flags(vma, address, ptep, entry, 1))
3200
		update_mmu_cache(vma, address, ptep);
3201 3202
}

3203
bool is_hugetlb_entry_migration(pte_t pte)
3204 3205 3206 3207
{
	swp_entry_t swp;

	if (huge_pte_none(pte) || pte_present(pte))
3208
		return false;
3209 3210
	swp = pte_to_swp_entry(pte);
	if (non_swap_entry(swp) && is_migration_entry(swp))
3211
		return true;
3212
	else
3213
		return false;
3214 3215 3216 3217 3218 3219 3220 3221 3222 3223 3224 3225 3226 3227
}

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;
}
3228

D
David Gibson 已提交
3229 3230 3231 3232 3233
int copy_hugetlb_page_range(struct mm_struct *dst, struct mm_struct *src,
			    struct vm_area_struct *vma)
{
	pte_t *src_pte, *dst_pte, entry;
	struct page *ptepage;
3234
	unsigned long addr;
3235
	int cow;
3236 3237
	struct hstate *h = hstate_vma(vma);
	unsigned long sz = huge_page_size(h);
3238 3239 3240
	unsigned long mmun_start;	/* For mmu_notifiers */
	unsigned long mmun_end;		/* For mmu_notifiers */
	int ret = 0;
3241 3242

	cow = (vma->vm_flags & (VM_SHARED | VM_MAYWRITE)) == VM_MAYWRITE;
D
David Gibson 已提交
3243

3244 3245 3246 3247 3248
	mmun_start = vma->vm_start;
	mmun_end = vma->vm_end;
	if (cow)
		mmu_notifier_invalidate_range_start(src, mmun_start, mmun_end);

3249
	for (addr = vma->vm_start; addr < vma->vm_end; addr += sz) {
3250
		spinlock_t *src_ptl, *dst_ptl;
3251
		src_pte = huge_pte_offset(src, addr, sz);
H
Hugh Dickins 已提交
3252 3253
		if (!src_pte)
			continue;
3254
		dst_pte = huge_pte_alloc(dst, addr, sz);
3255 3256 3257 3258
		if (!dst_pte) {
			ret = -ENOMEM;
			break;
		}
3259 3260 3261 3262 3263

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

3264 3265 3266
		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);
3267 3268 3269 3270 3271 3272 3273 3274 3275 3276 3277 3278 3279 3280
		entry = huge_ptep_get(src_pte);
		if (huge_pte_none(entry)) { /* skip none entry */
			;
		} else if (unlikely(is_hugetlb_entry_migration(entry) ||
				    is_hugetlb_entry_hwpoisoned(entry))) {
			swp_entry_t swp_entry = pte_to_swp_entry(entry);

			if (is_write_migration_entry(swp_entry) && cow) {
				/*
				 * COW mappings require pages in both
				 * parent and child to be set to read.
				 */
				make_migration_entry_read(&swp_entry);
				entry = swp_entry_to_pte(swp_entry);
3281 3282
				set_huge_swap_pte_at(src, addr, src_pte,
						     entry, sz);
3283
			}
3284
			set_huge_swap_pte_at(dst, addr, dst_pte, entry, sz);
3285
		} else {
3286
			if (cow) {
3287
				huge_ptep_set_wrprotect(src, addr, src_pte);
3288 3289 3290
				mmu_notifier_invalidate_range(src, mmun_start,
								   mmun_end);
			}
3291
			entry = huge_ptep_get(src_pte);
3292 3293
			ptepage = pte_page(entry);
			get_page(ptepage);
3294
			page_dup_rmap(ptepage, true);
3295
			set_huge_pte_at(dst, addr, dst_pte, entry);
3296
			hugetlb_count_add(pages_per_huge_page(h), dst);
3297
		}
3298 3299
		spin_unlock(src_ptl);
		spin_unlock(dst_ptl);
D
David Gibson 已提交
3300 3301
	}

3302 3303 3304 3305
	if (cow)
		mmu_notifier_invalidate_range_end(src, mmun_start, mmun_end);

	return ret;
D
David Gibson 已提交
3306 3307
}

3308 3309 3310
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
3311 3312 3313
{
	struct mm_struct *mm = vma->vm_mm;
	unsigned long address;
3314
	pte_t *ptep;
D
David Gibson 已提交
3315
	pte_t pte;
3316
	spinlock_t *ptl;
D
David Gibson 已提交
3317
	struct page *page;
3318 3319
	struct hstate *h = hstate_vma(vma);
	unsigned long sz = huge_page_size(h);
3320 3321
	const unsigned long mmun_start = start;	/* For mmu_notifiers */
	const unsigned long mmun_end   = end;	/* For mmu_notifiers */
3322

D
David Gibson 已提交
3323
	WARN_ON(!is_vm_hugetlb_page(vma));
3324 3325
	BUG_ON(start & ~huge_page_mask(h));
	BUG_ON(end & ~huge_page_mask(h));
D
David Gibson 已提交
3326

3327 3328 3329 3330 3331
	/*
	 * This is a hugetlb vma, all the pte entries should point
	 * to huge page.
	 */
	tlb_remove_check_page_size_change(tlb, sz);
3332
	tlb_start_vma(tlb, vma);
3333
	mmu_notifier_invalidate_range_start(mm, mmun_start, mmun_end);
3334 3335
	address = start;
	for (; address < end; address += sz) {
3336
		ptep = huge_pte_offset(mm, address, sz);
A
Adam Litke 已提交
3337
		if (!ptep)
3338 3339
			continue;

3340
		ptl = huge_pte_lock(h, mm, ptep);
3341 3342 3343 3344
		if (huge_pmd_unshare(mm, &address, ptep)) {
			spin_unlock(ptl);
			continue;
		}
3345

3346
		pte = huge_ptep_get(ptep);
3347 3348 3349 3350
		if (huge_pte_none(pte)) {
			spin_unlock(ptl);
			continue;
		}
3351 3352

		/*
3353 3354
		 * Migrating hugepage or HWPoisoned hugepage is already
		 * unmapped and its refcount is dropped, so just clear pte here.
3355
		 */
3356
		if (unlikely(!pte_present(pte))) {
3357
			huge_pte_clear(mm, address, ptep, sz);
3358 3359
			spin_unlock(ptl);
			continue;
3360
		}
3361 3362

		page = pte_page(pte);
3363 3364 3365 3366 3367 3368
		/*
		 * 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) {
3369 3370 3371 3372
			if (page != ref_page) {
				spin_unlock(ptl);
				continue;
			}
3373 3374 3375 3376 3377 3378 3379 3380
			/*
			 * 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);
		}

3381
		pte = huge_ptep_get_and_clear(mm, address, ptep);
3382
		tlb_remove_huge_tlb_entry(h, tlb, ptep, address);
3383
		if (huge_pte_dirty(pte))
3384
			set_page_dirty(page);
3385

3386
		hugetlb_count_sub(pages_per_huge_page(h), mm);
3387
		page_remove_rmap(page, true);
3388

3389
		spin_unlock(ptl);
3390
		tlb_remove_page_size(tlb, page, huge_page_size(h));
3391 3392 3393 3394 3395
		/*
		 * Bail out after unmapping reference page if supplied
		 */
		if (ref_page)
			break;
3396
	}
3397
	mmu_notifier_invalidate_range_end(mm, mmun_start, mmun_end);
3398
	tlb_end_vma(tlb, vma);
L
Linus Torvalds 已提交
3399
}
D
David Gibson 已提交
3400

3401 3402 3403 3404 3405 3406 3407 3408 3409 3410 3411 3412
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
3413
	 * is to clear it before releasing the i_mmap_rwsem. This works
3414
	 * because in the context this is called, the VMA is about to be
3415
	 * destroyed and the i_mmap_rwsem is held.
3416 3417 3418 3419
	 */
	vma->vm_flags &= ~VM_MAYSHARE;
}

3420
void unmap_hugepage_range(struct vm_area_struct *vma, unsigned long start,
3421
			  unsigned long end, struct page *ref_page)
3422
{
3423 3424 3425 3426 3427
	struct mm_struct *mm;
	struct mmu_gather tlb;

	mm = vma->vm_mm;

3428
	tlb_gather_mmu(&tlb, mm, start, end);
3429 3430
	__unmap_hugepage_range(&tlb, vma, start, end, ref_page);
	tlb_finish_mmu(&tlb, start, end);
3431 3432
}

3433 3434 3435 3436 3437 3438
/*
 * 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.
 */
3439 3440
static void unmap_ref_private(struct mm_struct *mm, struct vm_area_struct *vma,
			      struct page *page, unsigned long address)
3441
{
3442
	struct hstate *h = hstate_vma(vma);
3443 3444 3445 3446 3447 3448 3449 3450
	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.
	 */
3451
	address = address & huge_page_mask(h);
3452 3453
	pgoff = ((address - vma->vm_start) >> PAGE_SHIFT) +
			vma->vm_pgoff;
3454
	mapping = vma->vm_file->f_mapping;
3455

3456 3457 3458 3459 3460
	/*
	 * 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
	 */
3461
	i_mmap_lock_write(mapping);
3462
	vma_interval_tree_foreach(iter_vma, &mapping->i_mmap, pgoff, pgoff) {
3463 3464 3465 3466
		/* Do not unmap the current VMA */
		if (iter_vma == vma)
			continue;

3467 3468 3469 3470 3471 3472 3473 3474
		/*
		 * 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;

3475 3476 3477 3478 3479 3480 3481 3482
		/*
		 * 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))
3483 3484
			unmap_hugepage_range(iter_vma, address,
					     address + huge_page_size(h), page);
3485
	}
3486
	i_mmap_unlock_write(mapping);
3487 3488
}

3489 3490
/*
 * Hugetlb_cow() should be called with page lock of the original hugepage held.
3491 3492 3493
 * 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.
3494
 */
3495
static int hugetlb_cow(struct mm_struct *mm, struct vm_area_struct *vma,
3496 3497
		       unsigned long address, pte_t *ptep,
		       struct page *pagecache_page, spinlock_t *ptl)
3498
{
3499
	pte_t pte;
3500
	struct hstate *h = hstate_vma(vma);
3501
	struct page *old_page, *new_page;
3502
	int ret = 0, outside_reserve = 0;
3503 3504
	unsigned long mmun_start;	/* For mmu_notifiers */
	unsigned long mmun_end;		/* For mmu_notifiers */
3505

3506
	pte = huge_ptep_get(ptep);
3507 3508
	old_page = pte_page(pte);

3509
retry_avoidcopy:
3510 3511
	/* If no-one else is actually using this page, avoid the copy
	 * and just make the page writable */
3512
	if (page_mapcount(old_page) == 1 && PageAnon(old_page)) {
3513
		page_move_anon_rmap(old_page, vma);
3514
		set_huge_ptep_writable(vma, address, ptep);
N
Nick Piggin 已提交
3515
		return 0;
3516 3517
	}

3518 3519 3520 3521 3522 3523 3524 3525 3526
	/*
	 * 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.
	 */
3527
	if (is_vma_resv_set(vma, HPAGE_RESV_OWNER) &&
3528 3529 3530
			old_page != pagecache_page)
		outside_reserve = 1;

3531
	get_page(old_page);
3532

3533 3534 3535 3536
	/*
	 * Drop page table lock as buddy allocator may be called. It will
	 * be acquired again before returning to the caller, as expected.
	 */
3537
	spin_unlock(ptl);
3538
	new_page = alloc_huge_page(vma, address, outside_reserve);
3539

3540
	if (IS_ERR(new_page)) {
3541 3542 3543 3544 3545 3546 3547 3548
		/*
		 * 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) {
3549
			put_page(old_page);
3550
			BUG_ON(huge_pte_none(pte));
3551 3552 3553
			unmap_ref_private(mm, vma, old_page, address);
			BUG_ON(huge_pte_none(pte));
			spin_lock(ptl);
3554 3555
			ptep = huge_pte_offset(mm, address & huge_page_mask(h),
					       huge_page_size(h));
3556 3557 3558 3559 3560 3561 3562 3563
			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;
3564 3565
		}

3566 3567 3568
		ret = (PTR_ERR(new_page) == -ENOMEM) ?
			VM_FAULT_OOM : VM_FAULT_SIGBUS;
		goto out_release_old;
3569 3570
	}

3571 3572 3573 3574
	/*
	 * When the original hugepage is shared one, it does not have
	 * anon_vma prepared.
	 */
3575
	if (unlikely(anon_vma_prepare(vma))) {
3576 3577
		ret = VM_FAULT_OOM;
		goto out_release_all;
3578
	}
3579

A
Andrea Arcangeli 已提交
3580 3581
	copy_user_huge_page(new_page, old_page, address, vma,
			    pages_per_huge_page(h));
N
Nick Piggin 已提交
3582
	__SetPageUptodate(new_page);
3583
	set_page_huge_active(new_page);
3584

3585 3586 3587
	mmun_start = address & huge_page_mask(h);
	mmun_end = mmun_start + huge_page_size(h);
	mmu_notifier_invalidate_range_start(mm, mmun_start, mmun_end);
3588

3589
	/*
3590
	 * Retake the page table lock to check for racing updates
3591 3592
	 * before the page tables are altered
	 */
3593
	spin_lock(ptl);
3594 3595
	ptep = huge_pte_offset(mm, address & huge_page_mask(h),
			       huge_page_size(h));
3596
	if (likely(ptep && pte_same(huge_ptep_get(ptep), pte))) {
3597 3598
		ClearPagePrivate(new_page);

3599
		/* Break COW */
3600
		huge_ptep_clear_flush(vma, address, ptep);
3601
		mmu_notifier_invalidate_range(mm, mmun_start, mmun_end);
3602 3603
		set_huge_pte_at(mm, address, ptep,
				make_huge_pte(vma, new_page, 1));
3604
		page_remove_rmap(old_page, true);
3605
		hugepage_add_new_anon_rmap(new_page, vma, address);
3606 3607 3608
		/* Make the old page be freed below */
		new_page = old_page;
	}
3609
	spin_unlock(ptl);
3610
	mmu_notifier_invalidate_range_end(mm, mmun_start, mmun_end);
3611
out_release_all:
3612
	restore_reserve_on_error(h, vma, address, new_page);
3613
	put_page(new_page);
3614
out_release_old:
3615
	put_page(old_page);
3616

3617 3618
	spin_lock(ptl); /* Caller expects lock to be held */
	return ret;
3619 3620
}

3621
/* Return the pagecache page at a given address within a VMA */
3622 3623
static struct page *hugetlbfs_pagecache_page(struct hstate *h,
			struct vm_area_struct *vma, unsigned long address)
3624 3625
{
	struct address_space *mapping;
3626
	pgoff_t idx;
3627 3628

	mapping = vma->vm_file->f_mapping;
3629
	idx = vma_hugecache_offset(h, vma, address);
3630 3631 3632 3633

	return find_lock_page(mapping, idx);
}

H
Hugh Dickins 已提交
3634 3635 3636 3637 3638
/*
 * 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
3639 3640 3641 3642 3643 3644 3645 3646 3647 3648 3649 3650 3651 3652 3653
			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;
}

3654 3655 3656 3657 3658 3659 3660 3661 3662 3663 3664 3665 3666 3667 3668 3669 3670
int huge_add_to_page_cache(struct page *page, struct address_space *mapping,
			   pgoff_t idx)
{
	struct inode *inode = mapping->host;
	struct hstate *h = hstate_inode(inode);
	int err = add_to_page_cache(page, mapping, idx, GFP_KERNEL);

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

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

3671
static int hugetlb_no_page(struct mm_struct *mm, struct vm_area_struct *vma,
3672 3673
			   struct address_space *mapping, pgoff_t idx,
			   unsigned long address, pte_t *ptep, unsigned int flags)
3674
{
3675
	struct hstate *h = hstate_vma(vma);
3676
	int ret = VM_FAULT_SIGBUS;
3677
	int anon_rmap = 0;
A
Adam Litke 已提交
3678 3679
	unsigned long size;
	struct page *page;
3680
	pte_t new_pte;
3681
	spinlock_t *ptl;
A
Adam Litke 已提交
3682

3683 3684 3685
	/*
	 * 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
3686
	 * COW. Warn that such a situation has occurred as it may not be obvious
3687 3688
	 */
	if (is_vma_resv_set(vma, HPAGE_RESV_UNMAPPED)) {
3689
		pr_warn_ratelimited("PID %d killed due to inadequate hugepage pool\n",
3690
			   current->pid);
3691 3692 3693
		return ret;
	}

A
Adam Litke 已提交
3694 3695 3696 3697
	/*
	 * Use page lock to guard against racing truncation
	 * before we get page_table_lock.
	 */
3698 3699 3700
retry:
	page = find_lock_page(mapping, idx);
	if (!page) {
3701
		size = i_size_read(mapping->host) >> huge_page_shift(h);
3702 3703
		if (idx >= size)
			goto out;
3704 3705 3706 3707 3708 3709 3710 3711 3712 3713 3714 3715 3716 3717 3718 3719 3720 3721 3722 3723 3724 3725 3726 3727 3728 3729 3730 3731 3732 3733 3734 3735

		/*
		 * Check for page in userfault range
		 */
		if (userfaultfd_missing(vma)) {
			u32 hash;
			struct vm_fault vmf = {
				.vma = vma,
				.address = address,
				.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
				 */
			};

			/*
			 * hugetlb_fault_mutex must be dropped before
			 * handling userfault.  Reacquire after handling
			 * fault to make calling code simpler.
			 */
			hash = hugetlb_fault_mutex_hash(h, mm, vma, mapping,
							idx, address);
			mutex_unlock(&hugetlb_fault_mutex_table[hash]);
			ret = handle_userfault(&vmf, VM_UFFD_MISSING);
			mutex_lock(&hugetlb_fault_mutex_table[hash]);
			goto out;
		}

3736
		page = alloc_huge_page(vma, address, 0);
3737
		if (IS_ERR(page)) {
3738 3739 3740 3741 3742
			ret = PTR_ERR(page);
			if (ret == -ENOMEM)
				ret = VM_FAULT_OOM;
			else
				ret = VM_FAULT_SIGBUS;
3743 3744
			goto out;
		}
A
Andrea Arcangeli 已提交
3745
		clear_huge_page(page, address, pages_per_huge_page(h));
N
Nick Piggin 已提交
3746
		__SetPageUptodate(page);
3747
		set_page_huge_active(page);
3748

3749
		if (vma->vm_flags & VM_MAYSHARE) {
3750
			int err = huge_add_to_page_cache(page, mapping, idx);
3751 3752 3753 3754 3755 3756
			if (err) {
				put_page(page);
				if (err == -EEXIST)
					goto retry;
				goto out;
			}
3757
		} else {
3758
			lock_page(page);
3759 3760 3761 3762
			if (unlikely(anon_vma_prepare(vma))) {
				ret = VM_FAULT_OOM;
				goto backout_unlocked;
			}
3763
			anon_rmap = 1;
3764
		}
3765
	} else {
3766 3767 3768 3769 3770 3771
		/*
		 * 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))) {
3772
			ret = VM_FAULT_HWPOISON |
3773
				VM_FAULT_SET_HINDEX(hstate_index(h));
3774 3775
			goto backout_unlocked;
		}
3776
	}
3777

3778 3779 3780 3781 3782 3783
	/*
	 * 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.
	 */
3784
	if ((flags & FAULT_FLAG_WRITE) && !(vma->vm_flags & VM_SHARED)) {
3785 3786 3787 3788
		if (vma_needs_reservation(h, vma, address) < 0) {
			ret = VM_FAULT_OOM;
			goto backout_unlocked;
		}
3789
		/* Just decrements count, does not deallocate */
3790
		vma_end_reservation(h, vma, address);
3791
	}
3792

3793
	ptl = huge_pte_lock(h, mm, ptep);
3794
	size = i_size_read(mapping->host) >> huge_page_shift(h);
A
Adam Litke 已提交
3795 3796 3797
	if (idx >= size)
		goto backout;

N
Nick Piggin 已提交
3798
	ret = 0;
3799
	if (!huge_pte_none(huge_ptep_get(ptep)))
A
Adam Litke 已提交
3800 3801
		goto backout;

3802 3803
	if (anon_rmap) {
		ClearPagePrivate(page);
3804
		hugepage_add_new_anon_rmap(page, vma, address);
3805
	} else
3806
		page_dup_rmap(page, true);
3807 3808 3809 3810
	new_pte = make_huge_pte(vma, page, ((vma->vm_flags & VM_WRITE)
				&& (vma->vm_flags & VM_SHARED)));
	set_huge_pte_at(mm, address, ptep, new_pte);

3811
	hugetlb_count_add(pages_per_huge_page(h), mm);
3812
	if ((flags & FAULT_FLAG_WRITE) && !(vma->vm_flags & VM_SHARED)) {
3813
		/* Optimization, do the COW without a second fault */
3814
		ret = hugetlb_cow(mm, vma, address, ptep, page, ptl);
3815 3816
	}

3817
	spin_unlock(ptl);
A
Adam Litke 已提交
3818 3819
	unlock_page(page);
out:
3820
	return ret;
A
Adam Litke 已提交
3821 3822

backout:
3823
	spin_unlock(ptl);
3824
backout_unlocked:
A
Adam Litke 已提交
3825
	unlock_page(page);
3826
	restore_reserve_on_error(h, vma, address, page);
A
Adam Litke 已提交
3827 3828
	put_page(page);
	goto out;
3829 3830
}

3831
#ifdef CONFIG_SMP
3832
u32 hugetlb_fault_mutex_hash(struct hstate *h, struct mm_struct *mm,
3833 3834 3835 3836 3837 3838 3839 3840 3841 3842 3843 3844 3845 3846 3847 3848 3849 3850 3851 3852 3853 3854 3855 3856
			    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.
 */
3857
u32 hugetlb_fault_mutex_hash(struct hstate *h, struct mm_struct *mm,
3858 3859 3860 3861 3862 3863 3864 3865
			    struct vm_area_struct *vma,
			    struct address_space *mapping,
			    pgoff_t idx, unsigned long address)
{
	return 0;
}
#endif

3866
int hugetlb_fault(struct mm_struct *mm, struct vm_area_struct *vma,
3867
			unsigned long address, unsigned int flags)
3868
{
3869
	pte_t *ptep, entry;
3870
	spinlock_t *ptl;
3871
	int ret;
3872 3873
	u32 hash;
	pgoff_t idx;
3874
	struct page *page = NULL;
3875
	struct page *pagecache_page = NULL;
3876
	struct hstate *h = hstate_vma(vma);
3877
	struct address_space *mapping;
3878
	int need_wait_lock = 0;
3879

3880 3881
	address &= huge_page_mask(h);

3882
	ptep = huge_pte_offset(mm, address, huge_page_size(h));
3883 3884
	if (ptep) {
		entry = huge_ptep_get(ptep);
N
Naoya Horiguchi 已提交
3885
		if (unlikely(is_hugetlb_entry_migration(entry))) {
3886
			migration_entry_wait_huge(vma, mm, ptep);
N
Naoya Horiguchi 已提交
3887 3888
			return 0;
		} else if (unlikely(is_hugetlb_entry_hwpoisoned(entry)))
3889
			return VM_FAULT_HWPOISON_LARGE |
3890
				VM_FAULT_SET_HINDEX(hstate_index(h));
3891 3892 3893 3894
	} else {
		ptep = huge_pte_alloc(mm, address, huge_page_size(h));
		if (!ptep)
			return VM_FAULT_OOM;
3895 3896
	}

3897 3898 3899
	mapping = vma->vm_file->f_mapping;
	idx = vma_hugecache_offset(h, vma, address);

3900 3901 3902 3903 3904
	/*
	 * 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.
	 */
3905 3906
	hash = hugetlb_fault_mutex_hash(h, mm, vma, mapping, idx, address);
	mutex_lock(&hugetlb_fault_mutex_table[hash]);
3907

3908 3909
	entry = huge_ptep_get(ptep);
	if (huge_pte_none(entry)) {
3910
		ret = hugetlb_no_page(mm, vma, mapping, idx, address, ptep, flags);
3911
		goto out_mutex;
3912
	}
3913

N
Nick Piggin 已提交
3914
	ret = 0;
3915

3916 3917 3918 3919 3920 3921 3922 3923 3924 3925
	/*
	 * 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;

3926 3927 3928 3929 3930 3931 3932 3933
	/*
	 * 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.
	 */
3934
	if ((flags & FAULT_FLAG_WRITE) && !huge_pte_write(entry)) {
3935 3936
		if (vma_needs_reservation(h, vma, address) < 0) {
			ret = VM_FAULT_OOM;
3937
			goto out_mutex;
3938
		}
3939
		/* Just decrements count, does not deallocate */
3940
		vma_end_reservation(h, vma, address);
3941

3942
		if (!(vma->vm_flags & VM_MAYSHARE))
3943 3944 3945 3946
			pagecache_page = hugetlbfs_pagecache_page(h,
								vma, address);
	}

3947 3948 3949 3950 3951 3952
	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;

3953 3954 3955 3956 3957 3958 3959
	/*
	 * 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)
3960 3961 3962 3963
		if (!trylock_page(page)) {
			need_wait_lock = 1;
			goto out_ptl;
		}
3964

3965
	get_page(page);
3966

3967
	if (flags & FAULT_FLAG_WRITE) {
3968
		if (!huge_pte_write(entry)) {
3969 3970
			ret = hugetlb_cow(mm, vma, address, ptep,
					  pagecache_page, ptl);
3971
			goto out_put_page;
3972
		}
3973
		entry = huge_pte_mkdirty(entry);
3974 3975
	}
	entry = pte_mkyoung(entry);
3976 3977
	if (huge_ptep_set_access_flags(vma, address, ptep, entry,
						flags & FAULT_FLAG_WRITE))
3978
		update_mmu_cache(vma, address, ptep);
3979 3980 3981 3982
out_put_page:
	if (page != pagecache_page)
		unlock_page(page);
	put_page(page);
3983 3984
out_ptl:
	spin_unlock(ptl);
3985 3986 3987 3988 3989

	if (pagecache_page) {
		unlock_page(pagecache_page);
		put_page(pagecache_page);
	}
3990
out_mutex:
3991
	mutex_unlock(&hugetlb_fault_mutex_table[hash]);
3992 3993 3994 3995 3996 3997 3998 3999 4000
	/*
	 * 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);
4001
	return ret;
4002 4003
}

4004 4005 4006 4007 4008 4009 4010 4011 4012 4013 4014
/*
 * 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)
{
4015
	int vm_shared = dst_vma->vm_flags & VM_SHARED;
4016 4017 4018 4019 4020 4021 4022 4023 4024 4025 4026 4027 4028 4029
	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,
4030
						pages_per_huge_page(h), false);
4031 4032 4033 4034 4035 4036 4037 4038 4039 4040 4041 4042 4043 4044 4045 4046 4047 4048 4049 4050 4051

		/* fallback to copy_from_user outside mmap_sem */
		if (unlikely(ret)) {
			ret = -EFAULT;
			*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);
	set_page_huge_active(page);

4052 4053 4054 4055 4056 4057 4058 4059 4060 4061 4062 4063
	/*
	 * If shared, add to page cache
	 */
	if (vm_shared) {
		struct address_space *mapping = dst_vma->vm_file->f_mapping;
		pgoff_t idx = vma_hugecache_offset(h, dst_vma, dst_addr);

		ret = huge_add_to_page_cache(page, mapping, idx);
		if (ret)
			goto out_release_nounlock;
	}

4064 4065 4066 4067 4068 4069 4070
	ptl = huge_pte_lockptr(h, dst_mm, dst_pte);
	spin_lock(ptl);

	ret = -EEXIST;
	if (!huge_pte_none(huge_ptep_get(dst_pte)))
		goto out_release_unlock;

4071 4072 4073 4074 4075 4076
	if (vm_shared) {
		page_dup_rmap(page, true);
	} else {
		ClearPagePrivate(page);
		hugepage_add_new_anon_rmap(page, dst_vma, dst_addr);
	}
4077 4078 4079 4080 4081 4082 4083 4084 4085 4086 4087 4088 4089 4090 4091 4092

	_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);
4093 4094
	if (vm_shared)
		unlock_page(page);
4095 4096 4097 4098 4099
	ret = 0;
out:
	return ret;
out_release_unlock:
	spin_unlock(ptl);
4100 4101 4102
out_release_nounlock:
	if (vm_shared)
		unlock_page(page);
4103 4104 4105 4106
	put_page(page);
	goto out;
}

4107 4108 4109
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,
4110
			 long i, unsigned int flags, int *nonblocking)
D
David Gibson 已提交
4111
{
4112 4113
	unsigned long pfn_offset;
	unsigned long vaddr = *position;
4114
	unsigned long remainder = *nr_pages;
4115
	struct hstate *h = hstate_vma(vma);
D
David Gibson 已提交
4116 4117

	while (vaddr < vma->vm_end && remainder) {
A
Adam Litke 已提交
4118
		pte_t *pte;
4119
		spinlock_t *ptl = NULL;
H
Hugh Dickins 已提交
4120
		int absent;
A
Adam Litke 已提交
4121
		struct page *page;
D
David Gibson 已提交
4122

4123 4124 4125 4126 4127 4128 4129 4130 4131
		/*
		 * If we have a pending SIGKILL, don't keep faulting pages and
		 * potentially allocating memory.
		 */
		if (unlikely(fatal_signal_pending(current))) {
			remainder = 0;
			break;
		}

A
Adam Litke 已提交
4132 4133
		/*
		 * Some archs (sparc64, sh*) have multiple pte_ts to
H
Hugh Dickins 已提交
4134
		 * each hugepage.  We have to make sure we get the
A
Adam Litke 已提交
4135
		 * first, for the page indexing below to work.
4136 4137
		 *
		 * Note that page table lock is not held when pte is null.
A
Adam Litke 已提交
4138
		 */
4139 4140
		pte = huge_pte_offset(mm, vaddr & huge_page_mask(h),
				      huge_page_size(h));
4141 4142
		if (pte)
			ptl = huge_pte_lock(h, mm, pte);
H
Hugh Dickins 已提交
4143 4144 4145 4146
		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
4147 4148 4149 4150
		 * 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
4151
		 */
H
Hugh Dickins 已提交
4152 4153
		if (absent && (flags & FOLL_DUMP) &&
		    !hugetlbfs_pagecache_present(h, vma, vaddr)) {
4154 4155
			if (pte)
				spin_unlock(ptl);
H
Hugh Dickins 已提交
4156 4157 4158
			remainder = 0;
			break;
		}
D
David Gibson 已提交
4159

4160 4161 4162 4163 4164 4165 4166 4167 4168 4169 4170
		/*
		 * 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)) ||
4171 4172
		    ((flags & FOLL_WRITE) &&
		      !huge_pte_write(huge_ptep_get(pte)))) {
A
Adam Litke 已提交
4173
			int ret;
4174
			unsigned int fault_flags = 0;
D
David Gibson 已提交
4175

4176 4177
			if (pte)
				spin_unlock(ptl);
4178 4179 4180 4181 4182 4183 4184 4185 4186 4187 4188 4189 4190 4191
			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) {
4192 4193 4194 4195 4196
				int err = vm_fault_to_errno(ret, flags);

				if (err)
					return err;

4197 4198 4199 4200 4201 4202 4203 4204 4205 4206 4207 4208 4209 4210 4211 4212 4213 4214 4215
				remainder = 0;
				break;
			}
			if (ret & VM_FAULT_RETRY) {
				if (nonblocking)
					*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
4216 4217
		}

4218
		pfn_offset = (vaddr & ~huge_page_mask(h)) >> PAGE_SHIFT;
4219
		page = pte_page(huge_ptep_get(pte));
4220
same_page:
4221
		if (pages) {
H
Hugh Dickins 已提交
4222
			pages[i] = mem_map_offset(page, pfn_offset);
4223
			get_page(pages[i]);
4224
		}
D
David Gibson 已提交
4225 4226 4227 4228 4229

		if (vmas)
			vmas[i] = vma;

		vaddr += PAGE_SIZE;
4230
		++pfn_offset;
D
David Gibson 已提交
4231 4232
		--remainder;
		++i;
4233
		if (vaddr < vma->vm_end && remainder &&
4234
				pfn_offset < pages_per_huge_page(h)) {
4235 4236 4237 4238 4239 4240
			/*
			 * We use pfn_offset to avoid touching the pageframes
			 * of this compound page.
			 */
			goto same_page;
		}
4241
		spin_unlock(ptl);
D
David Gibson 已提交
4242
	}
4243
	*nr_pages = remainder;
4244 4245 4246 4247 4248
	/*
	 * 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
4249 4250
	*position = vaddr;

H
Hugh Dickins 已提交
4251
	return i ? i : -EFAULT;
D
David Gibson 已提交
4252
}
4253

4254 4255 4256 4257 4258 4259 4260 4261
#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

4262
unsigned long hugetlb_change_protection(struct vm_area_struct *vma,
4263 4264 4265 4266 4267 4268
		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;
4269
	struct hstate *h = hstate_vma(vma);
4270
	unsigned long pages = 0;
4271 4272 4273 4274

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

4275
	mmu_notifier_invalidate_range_start(mm, start, end);
4276
	i_mmap_lock_write(vma->vm_file->f_mapping);
4277
	for (; address < end; address += huge_page_size(h)) {
4278
		spinlock_t *ptl;
4279
		ptep = huge_pte_offset(mm, address, huge_page_size(h));
4280 4281
		if (!ptep)
			continue;
4282
		ptl = huge_pte_lock(h, mm, ptep);
4283 4284
		if (huge_pmd_unshare(mm, &address, ptep)) {
			pages++;
4285
			spin_unlock(ptl);
4286
			continue;
4287
		}
4288 4289 4290 4291 4292 4293 4294 4295 4296 4297 4298 4299 4300
		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);
4301 4302
				set_huge_swap_pte_at(mm, address, ptep,
						     newpte, huge_page_size(h));
4303 4304 4305 4306 4307 4308
				pages++;
			}
			spin_unlock(ptl);
			continue;
		}
		if (!huge_pte_none(pte)) {
4309
			pte = huge_ptep_get_and_clear(mm, address, ptep);
4310
			pte = pte_mkhuge(huge_pte_modify(pte, newprot));
4311
			pte = arch_make_huge_pte(pte, vma, NULL, 0);
4312
			set_huge_pte_at(mm, address, ptep, pte);
4313
			pages++;
4314
		}
4315
		spin_unlock(ptl);
4316
	}
4317
	/*
4318
	 * Must flush TLB before releasing i_mmap_rwsem: x86's huge_pmd_unshare
4319
	 * may have cleared our pud entry and done put_page on the page table:
4320
	 * once we release i_mmap_rwsem, another task can do the final put_page
4321 4322
	 * and that page table be reused and filled with junk.
	 */
4323
	flush_hugetlb_tlb_range(vma, start, end);
4324
	mmu_notifier_invalidate_range(mm, start, end);
4325
	i_mmap_unlock_write(vma->vm_file->f_mapping);
4326
	mmu_notifier_invalidate_range_end(mm, start, end);
4327 4328

	return pages << h->order;
4329 4330
}

4331 4332
int hugetlb_reserve_pages(struct inode *inode,
					long from, long to,
4333
					struct vm_area_struct *vma,
4334
					vm_flags_t vm_flags)
4335
{
4336
	long ret, chg;
4337
	struct hstate *h = hstate_inode(inode);
4338
	struct hugepage_subpool *spool = subpool_inode(inode);
4339
	struct resv_map *resv_map;
4340
	long gbl_reserve;
4341

4342 4343 4344
	/*
	 * Only apply hugepage reservation if asked. At fault time, an
	 * attempt will be made for VM_NORESERVE to allocate a page
4345
	 * without using reserves
4346
	 */
4347
	if (vm_flags & VM_NORESERVE)
4348 4349
		return 0;

4350 4351 4352 4353 4354 4355
	/*
	 * 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
	 */
4356
	if (!vma || vma->vm_flags & VM_MAYSHARE) {
4357
		resv_map = inode_resv_map(inode);
4358

4359
		chg = region_chg(resv_map, from, to);
4360 4361 4362

	} else {
		resv_map = resv_map_alloc();
4363 4364 4365
		if (!resv_map)
			return -ENOMEM;

4366
		chg = to - from;
4367

4368 4369 4370 4371
		set_vma_resv_map(vma, resv_map);
		set_vma_resv_flags(vma, HPAGE_RESV_OWNER);
	}

4372 4373 4374 4375
	if (chg < 0) {
		ret = chg;
		goto out_err;
	}
4376

4377 4378 4379 4380 4381 4382 4383
	/*
	 * 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) {
4384 4385 4386
		ret = -ENOSPC;
		goto out_err;
	}
4387 4388

	/*
4389
	 * Check enough hugepages are available for the reservation.
4390
	 * Hand the pages back to the subpool if there are not
4391
	 */
4392
	ret = hugetlb_acct_memory(h, gbl_reserve);
K
Ken Chen 已提交
4393
	if (ret < 0) {
4394 4395
		/* put back original number of pages, chg */
		(void)hugepage_subpool_put_pages(spool, chg);
4396
		goto out_err;
K
Ken Chen 已提交
4397
	}
4398 4399 4400 4401 4402 4403 4404 4405 4406 4407 4408 4409

	/*
	 * 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
	 */
4410 4411 4412 4413 4414 4415 4416 4417 4418 4419 4420 4421 4422 4423 4424 4425 4426 4427
	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);
		}
	}
4428
	return 0;
4429
out_err:
4430
	if (!vma || vma->vm_flags & VM_MAYSHARE)
4431 4432 4433
		/* Don't call region_abort if region_chg failed */
		if (chg >= 0)
			region_abort(resv_map, from, to);
J
Joonsoo Kim 已提交
4434 4435
	if (vma && is_vma_resv_set(vma, HPAGE_RESV_OWNER))
		kref_put(&resv_map->refs, resv_map_release);
4436
	return ret;
4437 4438
}

4439 4440
long hugetlb_unreserve_pages(struct inode *inode, long start, long end,
								long freed)
4441
{
4442
	struct hstate *h = hstate_inode(inode);
4443
	struct resv_map *resv_map = inode_resv_map(inode);
4444
	long chg = 0;
4445
	struct hugepage_subpool *spool = subpool_inode(inode);
4446
	long gbl_reserve;
K
Ken Chen 已提交
4447

4448 4449 4450 4451 4452 4453 4454 4455 4456 4457 4458
	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
4459
	spin_lock(&inode->i_lock);
4460
	inode->i_blocks -= (blocks_per_huge_page(h) * freed);
K
Ken Chen 已提交
4461 4462
	spin_unlock(&inode->i_lock);

4463 4464 4465 4466 4467 4468
	/*
	 * 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);
4469 4470

	return 0;
4471
}
4472

4473 4474 4475 4476 4477 4478 4479 4480 4481 4482 4483
#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
4484 4485
	unsigned long vm_flags = vma->vm_flags & VM_LOCKED_CLEAR_MASK;
	unsigned long svm_flags = svma->vm_flags & VM_LOCKED_CLEAR_MASK;
4486 4487 4488 4489 4490 4491 4492 4493 4494 4495 4496 4497 4498

	/*
	 * 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;
}

4499
static bool vma_shareable(struct vm_area_struct *vma, unsigned long addr)
4500 4501 4502 4503 4504 4505 4506 4507 4508
{
	unsigned long base = addr & PUD_MASK;
	unsigned long end = base + PUD_SIZE;

	/*
	 * check on proper vm_flags and page table alignment
	 */
	if (vma->vm_flags & VM_MAYSHARE &&
	    vma->vm_start <= base && end <= vma->vm_end)
4509 4510
		return true;
	return false;
4511 4512 4513 4514 4515 4516 4517
}

/*
 * Search for a shareable pmd page for hugetlb. In any case calls pmd_alloc()
 * and returns the corresponding pte. While this is not necessary for the
 * !shared pmd case because we can allocate the pmd later as well, it makes the
 * code much cleaner. pmd allocation is essential for the shared case because
4518
 * pud has to be populated inside the same i_mmap_rwsem section - otherwise
4519 4520 4521 4522 4523 4524 4525 4526 4527 4528 4529 4530 4531
 * racing tasks could either miss the sharing (see huge_pte_offset) or select a
 * bad pmd for sharing.
 */
pte_t *huge_pmd_share(struct mm_struct *mm, unsigned long addr, pud_t *pud)
{
	struct vm_area_struct *vma = find_vma(mm, addr);
	struct address_space *mapping = vma->vm_file->f_mapping;
	pgoff_t idx = ((addr - vma->vm_start) >> PAGE_SHIFT) +
			vma->vm_pgoff;
	struct vm_area_struct *svma;
	unsigned long saddr;
	pte_t *spte = NULL;
	pte_t *pte;
4532
	spinlock_t *ptl;
4533 4534 4535 4536

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

4537
	i_mmap_lock_write(mapping);
4538 4539 4540 4541 4542 4543
	vma_interval_tree_foreach(svma, &mapping->i_mmap, idx, idx) {
		if (svma == vma)
			continue;

		saddr = page_table_shareable(svma, vma, addr, idx);
		if (saddr) {
4544 4545
			spte = huge_pte_offset(svma->vm_mm, saddr,
					       vma_mmu_pagesize(svma));
4546 4547 4548 4549 4550 4551 4552 4553 4554 4555
			if (spte) {
				get_page(virt_to_page(spte));
				break;
			}
		}
	}

	if (!spte)
		goto out;

4556
	ptl = huge_pte_lock(hstate_vma(vma), mm, spte);
4557
	if (pud_none(*pud)) {
4558 4559
		pud_populate(mm, pud,
				(pmd_t *)((unsigned long)spte & PAGE_MASK));
4560
		mm_inc_nr_pmds(mm);
4561
	} else {
4562
		put_page(virt_to_page(spte));
4563
	}
4564
	spin_unlock(ptl);
4565 4566
out:
	pte = (pte_t *)pmd_alloc(mm, pud, addr);
4567
	i_mmap_unlock_write(mapping);
4568 4569 4570 4571 4572 4573 4574 4575 4576 4577
	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.
 *
4578
 * called with page table lock held.
4579 4580 4581 4582 4583 4584 4585
 *
 * 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);
4586 4587
	p4d_t *p4d = p4d_offset(pgd, *addr);
	pud_t *pud = pud_offset(p4d, *addr);
4588 4589 4590 4591 4592 4593 4594

	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));
4595
	mm_dec_nr_pmds(mm);
4596 4597 4598
	*addr = ALIGN(*addr, HPAGE_SIZE * PTRS_PER_PTE) - HPAGE_SIZE;
	return 1;
}
4599 4600 4601 4602 4603 4604
#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;
}
4605 4606 4607 4608 4609

int huge_pmd_unshare(struct mm_struct *mm, unsigned long *addr, pte_t *ptep)
{
	return 0;
}
4610
#define want_pmd_share()	(0)
4611 4612
#endif /* CONFIG_ARCH_WANT_HUGE_PMD_SHARE */

4613 4614 4615 4616 4617
#ifdef CONFIG_ARCH_WANT_GENERAL_HUGETLB
pte_t *huge_pte_alloc(struct mm_struct *mm,
			unsigned long addr, unsigned long sz)
{
	pgd_t *pgd;
4618
	p4d_t *p4d;
4619 4620 4621 4622
	pud_t *pud;
	pte_t *pte = NULL;

	pgd = pgd_offset(mm, addr);
4623 4624
	p4d = p4d_offset(pgd, addr);
	pud = pud_alloc(mm, p4d, addr);
4625 4626 4627 4628 4629 4630 4631 4632 4633 4634 4635
	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);
		}
	}
4636
	BUG_ON(pte && pte_present(*pte) && !pte_huge(*pte));
4637 4638 4639 4640

	return pte;
}

4641 4642
pte_t *huge_pte_offset(struct mm_struct *mm,
		       unsigned long addr, unsigned long sz)
4643 4644
{
	pgd_t *pgd;
4645
	p4d_t *p4d;
4646
	pud_t *pud;
4647
	pmd_t *pmd;
4648 4649

	pgd = pgd_offset(mm, addr);
4650 4651 4652 4653 4654 4655 4656 4657 4658 4659 4660
	if (!pgd_present(*pgd))
		return NULL;
	p4d = p4d_offset(pgd, addr);
	if (!p4d_present(*p4d))
		return NULL;
	pud = pud_offset(p4d, addr);
	if (!pud_present(*pud))
		return NULL;
	if (pud_huge(*pud))
		return (pte_t *)pud;
	pmd = pmd_offset(pud, addr);
4661 4662 4663
	return (pte_t *) pmd;
}

4664 4665 4666 4667 4668 4669 4670 4671 4672 4673 4674 4675 4676
#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);
}

4677 4678 4679 4680 4681 4682 4683 4684
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;
}

4685
struct page * __weak
4686
follow_huge_pmd(struct mm_struct *mm, unsigned long address,
4687
		pmd_t *pmd, int flags)
4688
{
4689 4690
	struct page *page = NULL;
	spinlock_t *ptl;
4691
	pte_t pte;
4692 4693 4694 4695 4696 4697 4698 4699 4700
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;
4701 4702
	pte = huge_ptep_get((pte_t *)pmd);
	if (pte_present(pte)) {
4703
		page = pmd_page(*pmd) + ((address & ~PMD_MASK) >> PAGE_SHIFT);
4704 4705 4706
		if (flags & FOLL_GET)
			get_page(page);
	} else {
4707
		if (is_hugetlb_entry_migration(pte)) {
4708 4709 4710 4711 4712 4713 4714 4715 4716 4717 4718
			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);
4719 4720 4721
	return page;
}

4722
struct page * __weak
4723
follow_huge_pud(struct mm_struct *mm, unsigned long address,
4724
		pud_t *pud, int flags)
4725
{
4726 4727
	if (flags & FOLL_GET)
		return NULL;
4728

4729
	return pte_page(*(pte_t *)pud) + ((address & ~PUD_MASK) >> PAGE_SHIFT);
4730 4731
}

4732 4733 4734 4735 4736 4737 4738 4739 4740
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);
}

4741 4742
#ifdef CONFIG_MEMORY_FAILURE

4743 4744 4745
/*
 * This function is called from memory failure code.
 */
4746
int dequeue_hwpoisoned_huge_page(struct page *hpage)
4747 4748 4749
{
	struct hstate *h = page_hstate(hpage);
	int nid = page_to_nid(hpage);
4750
	int ret = -EBUSY;
4751 4752

	spin_lock(&hugetlb_lock);
4753 4754 4755 4756 4757
	/*
	 * Just checking !page_huge_active is not enough, because that could be
	 * an isolated/hwpoisoned hugepage (which have >0 refcount).
	 */
	if (!page_huge_active(hpage) && !page_count(hpage)) {
4758 4759 4760 4761 4762 4763 4764
		/*
		 * Hwpoisoned hugepage isn't linked to activelist or freelist,
		 * but dangling hpage->lru can trigger list-debug warnings
		 * (this happens when we call unpoison_memory() on it),
		 * so let it point to itself with list_del_init().
		 */
		list_del_init(&hpage->lru);
4765
		set_page_refcounted(hpage);
4766 4767 4768 4769
		h->free_huge_pages--;
		h->free_huge_pages_node[nid]--;
		ret = 0;
	}
4770
	spin_unlock(&hugetlb_lock);
4771
	return ret;
4772
}
4773
#endif
4774 4775 4776

bool isolate_huge_page(struct page *page, struct list_head *list)
{
4777 4778
	bool ret = true;

4779
	VM_BUG_ON_PAGE(!PageHead(page), page);
4780
	spin_lock(&hugetlb_lock);
4781 4782 4783 4784 4785
	if (!page_huge_active(page) || !get_page_unless_zero(page)) {
		ret = false;
		goto unlock;
	}
	clear_page_huge_active(page);
4786
	list_move_tail(&page->lru, list);
4787
unlock:
4788
	spin_unlock(&hugetlb_lock);
4789
	return ret;
4790 4791 4792 4793
}

void putback_active_hugepage(struct page *page)
{
4794
	VM_BUG_ON_PAGE(!PageHead(page), page);
4795
	spin_lock(&hugetlb_lock);
4796
	set_page_huge_active(page);
4797 4798 4799 4800
	list_move_tail(&page->lru, &(page_hstate(page))->hugepage_activelist);
	spin_unlock(&hugetlb_lock);
	put_page(page);
}