hugetlb.c 129.0 KB
Newer Older
L
Linus Torvalds 已提交
1 2
/*
 * Generic hugetlb support.
3
 * (C) Nadia Yvette Chambers, April 2004
L
Linus Torvalds 已提交
4 5 6 7
 */
#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 80 81 82 83
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;
}

84 85 86 87 88 89 90
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
91 92 93 94 95 96
	 * 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);
97
		kfree(spool);
98
	}
99 100
}

101 102
struct hugepage_subpool *hugepage_new_subpool(struct hstate *h, long max_hpages,
						long min_hpages)
103 104 105
{
	struct hugepage_subpool *spool;

106
	spool = kzalloc(sizeof(*spool), GFP_KERNEL);
107 108 109 110 111
	if (!spool)
		return NULL;

	spin_lock_init(&spool->lock);
	spool->count = 1;
112 113 114 115 116 117 118 119 120
	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;
121 122 123 124 125 126 127 128 129 130 131 132

	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);
}

133 134 135 136 137 138 139 140 141
/*
 * 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,
142 143
				      long delta)
{
144
	long ret = delta;
145 146

	if (!spool)
147
		return ret;
148 149

	spin_lock(&spool->lock);
150 151 152 153 154 155 156 157

	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;
		}
158 159
	}

160 161
	/* minimum size accounting */
	if (spool->min_hpages != -1 && spool->rsv_hpages) {
162 163 164 165 166 167 168 169 170 171 172 173 174 175 176
		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);
177 178 179
	return ret;
}

180 181 182 183 184 185 186
/*
 * 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,
187 188
				       long delta)
{
189 190
	long ret = delta;

191
	if (!spool)
192
		return delta;
193 194

	spin_lock(&spool->lock);
195 196 197 198

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

199 200
	 /* minimum size accounting */
	if (spool->min_hpages != -1 && spool->used_hpages < spool->min_hpages) {
201 202 203 204 205 206 207 208 209 210 211 212 213 214
		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.
	 */
215
	unlock_or_release_subpool(spool);
216 217

	return ret;
218 219 220 221 222 223 224 225 226
}

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
227
	return subpool_inode(file_inode(vma->vm_file));
228 229
}

230 231 232
/*
 * Region tracking -- allows tracking of reservations and instantiated pages
 *                    across the pages in a mapping.
233
 *
234 235 236 237 238 239 240 241 242 243 244 245 246 247
 * 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.
248 249 250 251 252 253 254
 */
struct file_region {
	struct list_head link;
	long from;
	long to;
};

255 256
/*
 * Add the huge page range represented by [f, t) to the reserve
257 258 259 260 261 262 263 264
 * 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.
265 266 267
 *
 * Return the number of new huge pages added to the map.  This
 * number is greater than or equal to zero.
268
 */
269
static long region_add(struct resv_map *resv, long f, long t)
270
{
271
	struct list_head *head = &resv->regions;
272
	struct file_region *rg, *nrg, *trg;
273
	long add = 0;
274

275
	spin_lock(&resv->lock);
276 277 278 279 280
	/* Locate the region we are either in or before. */
	list_for_each_entry(rg, head, link)
		if (f <= rg->to)
			break;

281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302
	/*
	 * 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;
	}

303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320
	/* 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) {
321 322 323 324 325
			/* 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);
326 327 328 329
			list_del(&rg->link);
			kfree(rg);
		}
	}
330 331

	add += (nrg->from - f);		/* Added to beginning of region */
332
	nrg->from = f;
333
	add += t - nrg->to;		/* Added to end of region */
334
	nrg->to = t;
335

336 337
out_locked:
	resv->adds_in_progress--;
338
	spin_unlock(&resv->lock);
339 340
	VM_BUG_ON(add < 0);
	return add;
341 342
}

343 344 345 346 347 348 349 350 351 352 353 354 355
/*
 * 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.
 *
356 357 358 359 360 361 362 363
 * 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.
364
 */
365
static long region_chg(struct resv_map *resv, long f, long t)
366
{
367
	struct list_head *head = &resv->regions;
368
	struct file_region *rg, *nrg = NULL;
369 370
	long chg = 0;

371 372
retry:
	spin_lock(&resv->lock);
373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388
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);
389 390
		if (!trg) {
			kfree(nrg);
391
			return -ENOMEM;
392
		}
393 394 395 396 397 398 399

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

400 401 402 403 404 405 406 407 408
	/* 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) {
409
		if (!nrg) {
410
			resv->adds_in_progress--;
411 412 413 414 415 416 417 418 419 420
			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;
		}
421

422 423 424
		list_add(&nrg->link, rg->link.prev);
		chg = t - f;
		goto out_nrg;
425 426 427 428 429 430 431 432 433 434 435 436
	}

	/* 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)
437
			goto out;
438

L
Lucas De Marchi 已提交
439
		/* We overlap with this area, if it extends further than
440 441 442 443 444 445 446 447
		 * 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;
	}
448 449 450 451 452 453 454 455

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);
456 457 458
	return chg;
}

459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477
/*
 * 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);
}

478
/*
479 480 481 482 483 484 485 486 487 488 489 490
 * 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.
491
 */
492
static long region_del(struct resv_map *resv, long f, long t)
493
{
494
	struct list_head *head = &resv->regions;
495
	struct file_region *rg, *trg;
496 497
	struct file_region *nrg = NULL;
	long del = 0;
498

499
retry:
500
	spin_lock(&resv->lock);
501
	list_for_each_entry_safe(rg, trg, head, link) {
502 503 504 505 506 507 508 509
		/*
		 * 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))
510
			continue;
511

512
		if (rg->from >= t)
513 514
			break;

515 516 517 518 519 520 521 522 523 524 525 526 527
		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--;
			}
528

529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548
			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;
549
			break;
550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565
		}

		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;
		}
566
	}
567 568

	spin_unlock(&resv->lock);
569 570
	kfree(nrg);
	return del;
571 572
}

573 574 575 576 577 578 579 580 581
/*
 * 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.
 */
582
void hugetlb_fix_reserve_counts(struct inode *inode)
583 584 585 586 587
{
	struct hugepage_subpool *spool = subpool_inode(inode);
	long rsv_adjust;

	rsv_adjust = hugepage_subpool_get_pages(spool, 1);
588
	if (rsv_adjust) {
589 590 591 592 593 594
		struct hstate *h = hstate_inode(inode);

		hugetlb_acct_memory(h, 1);
	}
}

595 596 597 598
/*
 * Count and return the number of huge pages in the reserve map
 * that intersect with the range [f, t).
 */
599
static long region_count(struct resv_map *resv, long f, long t)
600
{
601
	struct list_head *head = &resv->regions;
602 603 604
	struct file_region *rg;
	long chg = 0;

605
	spin_lock(&resv->lock);
606 607
	/* Locate each segment we overlap with, and count that overlap. */
	list_for_each_entry(rg, head, link) {
608 609
		long seg_from;
		long seg_to;
610 611 612 613 614 615 616 617 618 619 620

		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;
	}
621
	spin_unlock(&resv->lock);
622 623 624 625

	return chg;
}

626 627 628 629
/*
 * Convert the address within this vma to the page offset within
 * the mapping, in pagecache page units; huge pages here.
 */
630 631
static pgoff_t vma_hugecache_offset(struct hstate *h,
			struct vm_area_struct *vma, unsigned long address)
632
{
633 634
	return ((address - vma->vm_start) >> huge_page_shift(h)) +
			(vma->vm_pgoff >> huge_page_order(h));
635 636
}

637 638 639 640 641
pgoff_t linear_hugepage_index(struct vm_area_struct *vma,
				     unsigned long address)
{
	return vma_hugecache_offset(hstate_vma(vma), vma, address);
}
642
EXPORT_SYMBOL_GPL(linear_hugepage_index);
643

644 645 646 647 648 649 650 651 652 653 654 655 656
/*
 * 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);

657
	return 1UL << huge_page_shift(hstate);
658
}
659
EXPORT_SYMBOL_GPL(vma_kernel_pagesize);
660

661 662 663 664 665 666 667 668 669 670 671 672 673
/*
 * 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

674 675 676 677 678 679 680
/*
 * 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)
681
#define HPAGE_RESV_MASK (HPAGE_RESV_OWNER | HPAGE_RESV_UNMAPPED)
682

683 684 685 686 687 688 689 690 691
/*
 * 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.
692 693 694 695 696 697 698 699 700
 *
 * 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.
701
 */
702 703 704 705 706 707 708 709 710 711 712
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;
}

713
struct resv_map *resv_map_alloc(void)
714 715
{
	struct resv_map *resv_map = kmalloc(sizeof(*resv_map), GFP_KERNEL);
716 717 718 719 720
	struct file_region *rg = kmalloc(sizeof(*rg), GFP_KERNEL);

	if (!resv_map || !rg) {
		kfree(resv_map);
		kfree(rg);
721
		return NULL;
722
	}
723 724

	kref_init(&resv_map->refs);
725
	spin_lock_init(&resv_map->lock);
726 727
	INIT_LIST_HEAD(&resv_map->regions);

728 729 730 731 732 733
	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;

734 735 736
	return resv_map;
}

737
void resv_map_release(struct kref *ref)
738 739
{
	struct resv_map *resv_map = container_of(ref, struct resv_map, refs);
740 741
	struct list_head *head = &resv_map->region_cache;
	struct file_region *rg, *trg;
742 743

	/* Clear out any active regions before we release the map. */
744
	region_del(resv_map, 0, LONG_MAX);
745 746 747 748 749 750 751 752 753

	/* ... 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);

754 755 756
	kfree(resv_map);
}

757 758 759 760 761
static inline struct resv_map *inode_resv_map(struct inode *inode)
{
	return inode->i_mapping->private_data;
}

762
static struct resv_map *vma_resv_map(struct vm_area_struct *vma)
763
{
764
	VM_BUG_ON_VMA(!is_vm_hugetlb_page(vma), vma);
765 766 767 768 769 770 771
	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 {
772 773
		return (struct resv_map *)(get_vma_private_data(vma) &
							~HPAGE_RESV_MASK);
774
	}
775 776
}

777
static void set_vma_resv_map(struct vm_area_struct *vma, struct resv_map *map)
778
{
779 780
	VM_BUG_ON_VMA(!is_vm_hugetlb_page(vma), vma);
	VM_BUG_ON_VMA(vma->vm_flags & VM_MAYSHARE, vma);
781

782 783
	set_vma_private_data(vma, (get_vma_private_data(vma) &
				HPAGE_RESV_MASK) | (unsigned long)map);
784 785 786 787
}

static void set_vma_resv_flags(struct vm_area_struct *vma, unsigned long flags)
{
788 789
	VM_BUG_ON_VMA(!is_vm_hugetlb_page(vma), vma);
	VM_BUG_ON_VMA(vma->vm_flags & VM_MAYSHARE, vma);
790 791

	set_vma_private_data(vma, get_vma_private_data(vma) | flags);
792 793 794 795
}

static int is_vma_resv_set(struct vm_area_struct *vma, unsigned long flag)
{
796
	VM_BUG_ON_VMA(!is_vm_hugetlb_page(vma), vma);
797 798

	return (get_vma_private_data(vma) & flag) != 0;
799 800
}

801
/* Reset counters to 0 and clear all HPAGE_RESV_* flags */
802 803
void reset_vma_resv_huge_pages(struct vm_area_struct *vma)
{
804
	VM_BUG_ON_VMA(!is_vm_hugetlb_page(vma), vma);
805
	if (!(vma->vm_flags & VM_MAYSHARE))
806 807 808 809
		vma->vm_private_data = (void *)0;
}

/* Returns true if the VMA has associated reserve pages */
810
static bool vma_has_reserves(struct vm_area_struct *vma, long chg)
811
{
812 813 814 815 816 817 818 819 820 821 822
	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)
823
			return true;
824
		else
825
			return false;
826
	}
827 828

	/* Shared mappings always use reserves */
829 830 831 832 833 834 835 836 837 838 839 840 841
	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;
	}
842 843 844 845 846

	/*
	 * Only the process that called mmap() has reserves for
	 * private mappings.
	 */
847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867
	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;
	}
868

869
	return false;
870 871
}

872
static void enqueue_huge_page(struct hstate *h, struct page *page)
L
Linus Torvalds 已提交
873 874
{
	int nid = page_to_nid(page);
875
	list_move(&page->lru, &h->hugepage_freelists[nid]);
876 877
	h->free_huge_pages++;
	h->free_huge_pages_node[nid]++;
L
Linus Torvalds 已提交
878 879
}

880
static struct page *dequeue_huge_page_node_exact(struct hstate *h, int nid)
881 882 883
{
	struct page *page;

884
	list_for_each_entry(page, &h->hugepage_freelists[nid], lru)
885
		if (!PageHWPoison(page))
886 887 888 889 890 891
			break;
	/*
	 * if 'non-isolated free hugepage' not found on the list,
	 * the allocation fails.
	 */
	if (&h->hugepage_freelists[nid] == &page->lru)
892
		return NULL;
893
	list_move(&page->lru, &h->hugepage_activelist);
894
	set_page_refcounted(page);
895 896 897 898 899
	h->free_huge_pages--;
	h->free_huge_pages_node[nid]--;
	return page;
}

900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915
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;
}

916 917 918
/* Movability of hugepages depends on migration support. */
static inline gfp_t htlb_alloc_mask(struct hstate *h)
{
919
	if (hugepages_treat_as_movable || hugepage_migration_supported(h))
920 921 922 923 924
		return GFP_HIGHUSER_MOVABLE;
	else
		return GFP_HIGHUSER;
}

925 926
static struct page *dequeue_huge_page_vma(struct hstate *h,
				struct vm_area_struct *vma,
927 928
				unsigned long address, int avoid_reserve,
				long chg)
L
Linus Torvalds 已提交
929
{
930
	struct page *page = NULL;
931
	struct mempolicy *mpol;
932
	nodemask_t *nodemask;
933 934
	gfp_t gfp_mask;
	int nid;
935
	struct zonelist *zonelist;
936 937
	struct zone *zone;
	struct zoneref *z;
938
	unsigned int cpuset_mems_cookie;
L
Linus Torvalds 已提交
939

940 941 942 943 944
	/*
	 * 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
	 */
945
	if (!vma_has_reserves(vma, chg) &&
946
			h->free_huge_pages - h->resv_huge_pages == 0)
947
		goto err;
948

949
	/* If reserves cannot be used, ensure enough pages are in the pool */
950
	if (avoid_reserve && h->free_huge_pages - h->resv_huge_pages == 0)
951
		goto err;
952

953
retry_cpuset:
954
	cpuset_mems_cookie = read_mems_allowed_begin();
955 956 957
	gfp_mask = htlb_alloc_mask(h);
	nid = huge_node(vma, address, gfp_mask, &mpol, &nodemask);
	zonelist = node_zonelist(nid, gfp_mask);
958

959 960
	for_each_zone_zonelist_nodemask(zone, z, zonelist,
						MAX_NR_ZONES - 1, nodemask) {
961
		if (cpuset_zone_allowed(zone, gfp_mask)) {
962 963
			page = dequeue_huge_page_node(h, zone_to_nid(zone));
			if (page) {
964 965 966 967 968
				if (avoid_reserve)
					break;
				if (!vma_has_reserves(vma, chg))
					break;

969
				SetPagePrivate(page);
970
				h->resv_huge_pages--;
971 972
				break;
			}
A
Andrew Morton 已提交
973
		}
L
Linus Torvalds 已提交
974
	}
975

976
	mpol_cond_put(mpol);
977
	if (unlikely(!page && read_mems_allowed_retry(cpuset_mems_cookie)))
978
		goto retry_cpuset;
L
Linus Torvalds 已提交
979
	return page;
980 981 982

err:
	return NULL;
L
Linus Torvalds 已提交
983 984
}

985 986 987 988 989 990 991 992 993
/*
 * 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)
{
994
	nid = next_node_in(nid, *nodes_allowed);
995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055
	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--)

1056
#ifdef CONFIG_ARCH_HAS_GIGANTIC_PAGE
1057
static void destroy_compound_gigantic_page(struct page *page,
1058
					unsigned int order)
1059 1060 1061 1062 1063
{
	int i;
	int nr_pages = 1 << order;
	struct page *p = page + 1;

1064
	atomic_set(compound_mapcount_ptr(page), 0);
1065
	for (i = 1; i < nr_pages; i++, p = mem_map_next(p, page, i)) {
1066
		clear_compound_head(p);
1067 1068 1069 1070 1071 1072 1073
		set_page_refcounted(p);
	}

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

1074
static void free_gigantic_page(struct page *page, unsigned int order)
1075 1076 1077 1078 1079 1080 1081 1082
{
	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;
1083 1084
	return alloc_contig_range(start_pfn, end_pfn, MIGRATE_MOVABLE,
				  GFP_KERNEL);
1085 1086
}

1087 1088
static bool pfn_range_valid_gigantic(struct zone *z,
			unsigned long start_pfn, unsigned long nr_pages)
1089 1090 1091 1092 1093 1094 1095 1096 1097 1098
{
	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);

1099 1100 1101
		if (page_zone(page) != z)
			return false;

1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121
		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);
}

1122
static struct page *alloc_gigantic_page(int nid, unsigned int order)
1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133
{
	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)) {
1134
			if (pfn_range_valid_gigantic(z, pfn, nr_pages)) {
1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157
				/*
				 * 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);
1158
static void prep_compound_gigantic_page(struct page *page, unsigned int order);
1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187

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;
}

1188
#else /* !CONFIG_ARCH_HAS_GIGANTIC_PAGE */
1189
static inline bool gigantic_page_supported(void) { return false; }
1190
static inline void free_gigantic_page(struct page *page, unsigned int order) { }
1191
static inline void destroy_compound_gigantic_page(struct page *page,
1192
						unsigned int order) { }
1193 1194 1195 1196
static inline int alloc_fresh_gigantic_page(struct hstate *h,
					nodemask_t *nodes_allowed) { return 0; }
#endif

1197
static void update_and_free_page(struct hstate *h, struct page *page)
A
Adam Litke 已提交
1198 1199
{
	int i;
1200

1201 1202
	if (hstate_is_gigantic(h) && !gigantic_page_supported())
		return;
1203

1204 1205 1206
	h->nr_huge_pages--;
	h->nr_huge_pages_node[page_to_nid(page)]--;
	for (i = 0; i < pages_per_huge_page(h); i++) {
1207 1208
		page[i].flags &= ~(1 << PG_locked | 1 << PG_error |
				1 << PG_referenced | 1 << PG_dirty |
1209 1210
				1 << PG_active | 1 << PG_private |
				1 << PG_writeback);
A
Adam Litke 已提交
1211
	}
1212
	VM_BUG_ON_PAGE(hugetlb_cgroup_from_page(page), page);
1213
	set_compound_page_dtor(page, NULL_COMPOUND_DTOR);
A
Adam Litke 已提交
1214
	set_page_refcounted(page);
1215 1216 1217 1218 1219 1220
	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
1221 1222
}

1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233
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;
}

1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258
/*
 * 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]);
}

1259
void free_huge_page(struct page *page)
1260
{
1261 1262 1263 1264
	/*
	 * Can't pass hstate in here because it is called from the
	 * compound page destructor.
	 */
1265
	struct hstate *h = page_hstate(page);
1266
	int nid = page_to_nid(page);
1267 1268
	struct hugepage_subpool *spool =
		(struct hugepage_subpool *)page_private(page);
1269
	bool restore_reserve;
1270

1271
	set_page_private(page, 0);
1272
	page->mapping = NULL;
1273 1274
	VM_BUG_ON_PAGE(page_count(page), page);
	VM_BUG_ON_PAGE(page_mapcount(page), page);
1275
	restore_reserve = PagePrivate(page);
1276
	ClearPagePrivate(page);
1277

1278 1279 1280 1281 1282 1283 1284 1285
	/*
	 * 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;

1286
	spin_lock(&hugetlb_lock);
1287
	clear_page_huge_active(page);
1288 1289
	hugetlb_cgroup_uncharge_page(hstate_index(h),
				     pages_per_huge_page(h), page);
1290 1291 1292
	if (restore_reserve)
		h->resv_huge_pages++;

1293
	if (h->surplus_huge_pages_node[nid]) {
1294 1295
		/* remove the page from active list */
		list_del(&page->lru);
1296 1297 1298
		update_and_free_page(h, page);
		h->surplus_huge_pages--;
		h->surplus_huge_pages_node[nid]--;
1299
	} else {
1300
		arch_clear_hugepage_flags(page);
1301
		enqueue_huge_page(h, page);
1302
	}
1303 1304 1305
	spin_unlock(&hugetlb_lock);
}

1306
static void prep_new_huge_page(struct hstate *h, struct page *page, int nid)
1307
{
1308
	INIT_LIST_HEAD(&page->lru);
1309
	set_compound_page_dtor(page, HUGETLB_PAGE_DTOR);
1310
	spin_lock(&hugetlb_lock);
1311
	set_hugetlb_cgroup(page, NULL);
1312 1313
	h->nr_huge_pages++;
	h->nr_huge_pages_node[nid]++;
1314 1315 1316 1317
	spin_unlock(&hugetlb_lock);
	put_page(page); /* free it into the hugepage allocator */
}

1318
static void prep_compound_gigantic_page(struct page *page, unsigned int order)
1319 1320 1321 1322 1323 1324 1325
{
	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);
1326
	__ClearPageReserved(page);
1327
	__SetPageHead(page);
1328
	for (i = 1; i < nr_pages; i++, p = mem_map_next(p, page, i)) {
1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341
		/*
		 * 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);
1342
		set_page_count(p, 0);
1343
		set_compound_head(p, page);
1344
	}
1345
	atomic_set(compound_mapcount_ptr(page), -1);
1346 1347
}

A
Andrew Morton 已提交
1348 1349 1350 1351 1352
/*
 * PageHuge() only returns true for hugetlbfs pages, but not for normal or
 * transparent huge pages.  See the PageTransHuge() documentation for more
 * details.
 */
1353 1354 1355 1356 1357 1358
int PageHuge(struct page *page)
{
	if (!PageCompound(page))
		return 0;

	page = compound_head(page);
1359
	return page[1].compound_dtor == HUGETLB_PAGE_DTOR;
1360
}
1361 1362
EXPORT_SYMBOL_GPL(PageHuge);

1363 1364 1365 1366 1367 1368 1369 1370 1371
/*
 * 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;

1372
	return get_compound_page_dtor(page_head) == free_huge_page;
1373 1374
}

1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391
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;
}

1392
static struct page *alloc_fresh_huge_page_node(struct hstate *h, int nid)
L
Linus Torvalds 已提交
1393 1394
{
	struct page *page;
1395

1396
	page = __alloc_pages_node(nid,
1397
		htlb_alloc_mask(h)|__GFP_COMP|__GFP_THISNODE|
1398
						__GFP_REPEAT|__GFP_NOWARN,
1399
		huge_page_order(h));
L
Linus Torvalds 已提交
1400
	if (page) {
1401
		prep_new_huge_page(h, page, nid);
L
Linus Torvalds 已提交
1402
	}
1403 1404 1405 1406

	return page;
}

1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428
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;
}

1429 1430 1431 1432 1433 1434
/*
 * 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.
 */
1435 1436
static int free_pool_huge_page(struct hstate *h, nodemask_t *nodes_allowed,
							 bool acct_surplus)
1437
{
1438
	int nr_nodes, node;
1439 1440
	int ret = 0;

1441
	for_each_node_mask_to_free(h, nr_nodes, node, nodes_allowed) {
1442 1443 1444 1445
		/*
		 * If we're returning unused surplus pages, only examine
		 * nodes with surplus pages.
		 */
1446 1447
		if ((!acct_surplus || h->surplus_huge_pages_node[node]) &&
		    !list_empty(&h->hugepage_freelists[node])) {
1448
			struct page *page =
1449
				list_entry(h->hugepage_freelists[node].next,
1450 1451 1452
					  struct page, lru);
			list_del(&page->lru);
			h->free_huge_pages--;
1453
			h->free_huge_pages_node[node]--;
1454 1455
			if (acct_surplus) {
				h->surplus_huge_pages--;
1456
				h->surplus_huge_pages_node[node]--;
1457
			}
1458 1459
			update_and_free_page(h, page);
			ret = 1;
1460
			break;
1461
		}
1462
	}
1463 1464 1465 1466

	return ret;
}

1467 1468
/*
 * Dissolve a given free hugepage into free buddy pages. This function does
1469 1470 1471
 * nothing for in-use (including surplus) hugepages. Returns -EBUSY if the
 * number of free hugepages would be reduced below the number of reserved
 * hugepages.
1472
 */
1473
int dissolve_free_huge_page(struct page *page)
1474
{
1475 1476
	int rc = 0;

1477 1478
	spin_lock(&hugetlb_lock);
	if (PageHuge(page) && !page_count(page)) {
1479 1480 1481
		struct page *head = compound_head(page);
		struct hstate *h = page_hstate(head);
		int nid = page_to_nid(head);
1482 1483 1484 1485
		if (h->free_huge_pages - h->resv_huge_pages == 0) {
			rc = -EBUSY;
			goto out;
		}
1486 1487 1488 1489 1490 1491 1492 1493
		/*
		 * Move PageHWPoison flag from head page to the raw error page,
		 * which makes any subpages rather than the error page reusable.
		 */
		if (PageHWPoison(head) && page != head) {
			SetPageHWPoison(page);
			ClearPageHWPoison(head);
		}
1494
		list_del(&head->lru);
1495 1496
		h->free_huge_pages--;
		h->free_huge_pages_node[nid]--;
1497
		h->max_huge_pages--;
1498
		update_and_free_page(h, head);
1499
	}
1500
out:
1501
	spin_unlock(&hugetlb_lock);
1502
	return rc;
1503 1504 1505 1506 1507
}

/*
 * Dissolve free hugepages in a given pfn range. Used by memory hotplug to
 * make specified memory blocks removable from the system.
1508 1509
 * Note that this will dissolve a free gigantic hugepage completely, if any
 * part of it lies within the given range.
1510 1511
 * Also note that if dissolve_free_huge_page() returns with an error, all
 * free hugepages that were dissolved before that error are lost.
1512
 */
1513
int dissolve_free_huge_pages(unsigned long start_pfn, unsigned long end_pfn)
1514 1515
{
	unsigned long pfn;
1516
	struct page *page;
1517
	int rc = 0;
1518

1519
	if (!hugepages_supported())
1520
		return rc;
1521

1522 1523 1524 1525 1526 1527 1528 1529
	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;
		}
	}
1530 1531

	return rc;
1532 1533
}

1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550 1551
/*
 * 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
1552 1553 1554 1555 1556 1557
	 * 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.
1558
	 */
D
Dave Hansen 已提交
1559
	if (!IS_ENABLED(CONFIG_NUMA) || !vma) {
1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575
		/*
		 * 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
1576 1577
	 * allocate a huge page with it.  We will only reach this
	 * when CONFIG_NUMA=y.
1578 1579 1580 1581
	 */
	do {
		struct page *page;
		struct mempolicy *mpol;
1582
		int nid;
1583 1584 1585
		nodemask_t *nodemask;

		cpuset_mems_cookie = read_mems_allowed_begin();
1586
		nid = huge_node(vma, addr, gfp, &mpol, &nodemask);
1587
		mpol_cond_put(mpol);
1588
		page = __alloc_pages_nodemask(gfp, order, nid, nodemask);
1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609
		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)
1610 1611
{
	struct page *page;
1612
	unsigned int r_nid;
1613

1614
	if (hstate_is_gigantic(h))
1615 1616
		return NULL;

1617 1618 1619 1620 1621 1622
	/*
	 * 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
1623 1624
		VM_WARN_ON_ONCE(addr == -1);
		VM_WARN_ON_ONCE(nid != NUMA_NO_NODE);
1625
	}
1626 1627 1628 1629 1630 1631 1632 1633 1634 1635 1636 1637 1638 1639 1640 1641 1642 1643 1644 1645 1646 1647 1648 1649
	/*
	 * 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);
1650
	if (h->surplus_huge_pages >= h->nr_overcommit_huge_pages) {
1651 1652 1653
		spin_unlock(&hugetlb_lock);
		return NULL;
	} else {
1654 1655
		h->nr_huge_pages++;
		h->surplus_huge_pages++;
1656 1657 1658
	}
	spin_unlock(&hugetlb_lock);

1659
	page = __hugetlb_alloc_buddy_huge_page(h, vma, addr, nid);
1660 1661

	spin_lock(&hugetlb_lock);
1662
	if (page) {
1663
		INIT_LIST_HEAD(&page->lru);
1664
		r_nid = page_to_nid(page);
1665
		set_compound_page_dtor(page, HUGETLB_PAGE_DTOR);
1666
		set_hugetlb_cgroup(page, NULL);
1667 1668 1669
		/*
		 * We incremented the global counters already
		 */
1670 1671
		h->nr_huge_pages_node[r_nid]++;
		h->surplus_huge_pages_node[r_nid]++;
1672
		__count_vm_event(HTLB_BUDDY_PGALLOC);
1673
	} else {
1674 1675
		h->nr_huge_pages--;
		h->surplus_huge_pages--;
1676
		__count_vm_event(HTLB_BUDDY_PGALLOC_FAIL);
1677
	}
1678
	spin_unlock(&hugetlb_lock);
1679 1680 1681 1682

	return page;
}

1683 1684 1685 1686 1687
/*
 * 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
1688
static
1689 1690 1691 1692 1693 1694 1695 1696 1697 1698
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
1699
static
1700 1701 1702 1703 1704 1705
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);
}

1706 1707 1708 1709 1710 1711 1712
/*
 * 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)
{
1713
	struct page *page = NULL;
1714 1715

	spin_lock(&hugetlb_lock);
1716 1717
	if (h->free_huge_pages - h->resv_huge_pages > 0)
		page = dequeue_huge_page_node(h, nid);
1718 1719
	spin_unlock(&hugetlb_lock);

1720
	if (!page)
1721
		page = __alloc_buddy_huge_page_no_mpol(h, nid);
1722 1723 1724 1725

	return page;
}

1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746 1747 1748 1749 1750 1751 1752
struct page *alloc_huge_page_nodemask(struct hstate *h, const nodemask_t *nmask)
{
	struct page *page = NULL;
	int node;

	spin_lock(&hugetlb_lock);
	if (h->free_huge_pages - h->resv_huge_pages > 0) {
		for_each_node_mask(node, *nmask) {
			page = dequeue_huge_page_node_exact(h, node);
			if (page)
				break;
		}
	}
	spin_unlock(&hugetlb_lock);
	if (page)
		return page;

	/* No reservations, try to overcommit */
	for_each_node_mask(node, *nmask) {
		page = __alloc_buddy_huge_page_no_mpol(h, node);
		if (page)
			return page;
	}

	return NULL;
}

1753
/*
L
Lucas De Marchi 已提交
1754
 * Increase the hugetlb pool such that it can accommodate a reservation
1755 1756
 * of size 'delta'.
 */
1757
static int gather_surplus_pages(struct hstate *h, int delta)
1758 1759 1760 1761 1762
{
	struct list_head surplus_list;
	struct page *page, *tmp;
	int ret, i;
	int needed, allocated;
1763
	bool alloc_ok = true;
1764

1765
	needed = (h->resv_huge_pages + delta) - h->free_huge_pages;
1766
	if (needed <= 0) {
1767
		h->resv_huge_pages += delta;
1768
		return 0;
1769
	}
1770 1771 1772 1773 1774 1775 1776 1777

	allocated = 0;
	INIT_LIST_HEAD(&surplus_list);

	ret = -ENOMEM;
retry:
	spin_unlock(&hugetlb_lock);
	for (i = 0; i < needed; i++) {
1778
		page = __alloc_buddy_huge_page_no_mpol(h, NUMA_NO_NODE);
1779 1780 1781 1782
		if (!page) {
			alloc_ok = false;
			break;
		}
1783
		list_add(&page->lru, &surplus_list);
1784
		cond_resched();
1785
	}
1786
	allocated += i;
1787 1788 1789 1790 1791 1792

	/*
	 * 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);
1793 1794
	needed = (h->resv_huge_pages + delta) -
			(h->free_huge_pages + allocated);
1795 1796 1797 1798 1799 1800 1801 1802 1803 1804
	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;
	}
1805 1806
	/*
	 * The surplus_list now contains _at_least_ the number of extra pages
L
Lucas De Marchi 已提交
1807
	 * needed to accommodate the reservation.  Add the appropriate number
1808
	 * of pages to the hugetlb pool and free the extras back to the buddy
1809 1810 1811
	 * 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.
1812 1813
	 */
	needed += allocated;
1814
	h->resv_huge_pages += delta;
1815
	ret = 0;
1816

1817
	/* Free the needed pages to the hugetlb pool */
1818
	list_for_each_entry_safe(page, tmp, &surplus_list, lru) {
1819 1820
		if ((--needed) < 0)
			break;
1821 1822 1823 1824 1825
		/*
		 * This page is now managed by the hugetlb allocator and has
		 * no users -- drop the buddy allocator's reference.
		 */
		put_page_testzero(page);
1826
		VM_BUG_ON_PAGE(page_count(page), page);
1827
		enqueue_huge_page(h, page);
1828
	}
1829
free:
1830
	spin_unlock(&hugetlb_lock);
1831 1832

	/* Free unnecessary surplus pages to the buddy allocator */
1833 1834
	list_for_each_entry_safe(page, tmp, &surplus_list, lru)
		put_page(page);
1835
	spin_lock(&hugetlb_lock);
1836 1837 1838 1839 1840

	return ret;
}

/*
1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852
 * 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.
1853
 */
1854 1855
static void return_unused_surplus_pages(struct hstate *h,
					unsigned long unused_resv_pages)
1856 1857 1858
{
	unsigned long nr_pages;

1859
	/* Cannot return gigantic pages currently */
1860
	if (hstate_is_gigantic(h))
1861
		goto out;
1862

1863 1864 1865 1866
	/*
	 * Part (or even all) of the reservation could have been backed
	 * by pre-allocated pages. Only free surplus pages.
	 */
1867
	nr_pages = min(unused_resv_pages, h->surplus_huge_pages);
1868

1869 1870
	/*
	 * We want to release as many surplus pages as possible, spread
1871 1872 1873 1874 1875
	 * 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.
1876 1877 1878 1879
	 *
	 * 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.
1880 1881
	 */
	while (nr_pages--) {
1882 1883
		h->resv_huge_pages--;
		unused_resv_pages--;
1884
		if (!free_pool_huge_page(h, &node_states[N_MEMORY], 1))
1885
			goto out;
1886
		cond_resched_lock(&hugetlb_lock);
1887
	}
1888 1889 1890 1891

out:
	/* Fully uncommit the reservation */
	h->resv_huge_pages -= unused_resv_pages;
1892 1893
}

1894

1895
/*
1896
 * vma_needs_reservation, vma_commit_reservation and vma_end_reservation
1897
 * are used by the huge page allocation routines to manage reservations.
1898 1899 1900 1901 1902 1903
 *
 * 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
1904 1905 1906
 * 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.
1907 1908 1909 1910 1911 1912
 *
 * 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.
1913 1914 1915 1916 1917
 *
 * 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.
1918
 */
1919 1920 1921
enum vma_resv_mode {
	VMA_NEEDS_RESV,
	VMA_COMMIT_RESV,
1922
	VMA_END_RESV,
1923
	VMA_ADD_RESV,
1924
};
1925 1926
static long __vma_reservation_common(struct hstate *h,
				struct vm_area_struct *vma, unsigned long addr,
1927
				enum vma_resv_mode mode)
1928
{
1929 1930
	struct resv_map *resv;
	pgoff_t idx;
1931
	long ret;
1932

1933 1934
	resv = vma_resv_map(vma);
	if (!resv)
1935
		return 1;
1936

1937
	idx = vma_hugecache_offset(h, vma, addr);
1938 1939
	switch (mode) {
	case VMA_NEEDS_RESV:
1940
		ret = region_chg(resv, idx, idx + 1);
1941 1942 1943 1944
		break;
	case VMA_COMMIT_RESV:
		ret = region_add(resv, idx, idx + 1);
		break;
1945
	case VMA_END_RESV:
1946 1947 1948
		region_abort(resv, idx, idx + 1);
		ret = 0;
		break;
1949 1950 1951 1952 1953 1954 1955 1956
	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;
1957 1958 1959
	default:
		BUG();
	}
1960

1961
	if (vma->vm_flags & VM_MAYSHARE)
1962
		return ret;
1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981
	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;
	}
1982
	else
1983
		return ret < 0 ? ret : 0;
1984
}
1985 1986

static long vma_needs_reservation(struct hstate *h,
1987
			struct vm_area_struct *vma, unsigned long addr)
1988
{
1989
	return __vma_reservation_common(h, vma, addr, VMA_NEEDS_RESV);
1990
}
1991

1992 1993 1994
static long vma_commit_reservation(struct hstate *h,
			struct vm_area_struct *vma, unsigned long addr)
{
1995 1996 1997
	return __vma_reservation_common(h, vma, addr, VMA_COMMIT_RESV);
}

1998
static void vma_end_reservation(struct hstate *h,
1999 2000
			struct vm_area_struct *vma, unsigned long addr)
{
2001
	(void)__vma_reservation_common(h, vma, addr, VMA_END_RESV);
2002 2003
}

2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046 2047 2048 2049 2050 2051 2052 2053
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);
	}
}

2054
struct page *alloc_huge_page(struct vm_area_struct *vma,
2055
				    unsigned long addr, int avoid_reserve)
L
Linus Torvalds 已提交
2056
{
2057
	struct hugepage_subpool *spool = subpool_vma(vma);
2058
	struct hstate *h = hstate_vma(vma);
2059
	struct page *page;
2060 2061
	long map_chg, map_commit;
	long gbl_chg;
2062 2063
	int ret, idx;
	struct hugetlb_cgroup *h_cg;
2064

2065
	idx = hstate_index(h);
2066
	/*
2067 2068 2069
	 * 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).
2070
	 */
2071 2072
	map_chg = gbl_chg = vma_needs_reservation(h, vma, addr);
	if (map_chg < 0)
2073
		return ERR_PTR(-ENOMEM);
2074 2075 2076 2077 2078 2079 2080 2081 2082 2083 2084

	/*
	 * 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) {
2085
			vma_end_reservation(h, vma, addr);
2086
			return ERR_PTR(-ENOSPC);
2087
		}
L
Linus Torvalds 已提交
2088

2089 2090 2091 2092 2093 2094 2095 2096 2097 2098 2099 2100
		/*
		 * 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;
	}

2101
	ret = hugetlb_cgroup_charge_cgroup(idx, pages_per_huge_page(h), &h_cg);
2102 2103 2104
	if (ret)
		goto out_subpool_put;

L
Linus Torvalds 已提交
2105
	spin_lock(&hugetlb_lock);
2106 2107 2108 2109 2110 2111
	/*
	 * 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);
2112
	if (!page) {
2113
		spin_unlock(&hugetlb_lock);
2114
		page = __alloc_buddy_huge_page_with_mpol(h, vma, addr);
2115 2116
		if (!page)
			goto out_uncharge_cgroup;
2117 2118 2119 2120
		if (!avoid_reserve && vma_has_reserves(vma, gbl_chg)) {
			SetPagePrivate(page);
			h->resv_huge_pages--;
		}
2121 2122
		spin_lock(&hugetlb_lock);
		list_move(&page->lru, &h->hugepage_activelist);
2123
		/* Fall through */
K
Ken Chen 已提交
2124
	}
2125 2126
	hugetlb_cgroup_commit_charge(idx, pages_per_huge_page(h), h_cg, page);
	spin_unlock(&hugetlb_lock);
2127

2128
	set_page_private(page, (unsigned long)spool);
2129

2130 2131
	map_commit = vma_commit_reservation(h, vma, addr);
	if (unlikely(map_chg > map_commit)) {
2132 2133 2134 2135 2136 2137 2138 2139 2140 2141 2142 2143 2144 2145
		/*
		 * 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);
	}
2146
	return page;
2147 2148 2149 2150

out_uncharge_cgroup:
	hugetlb_cgroup_uncharge_cgroup(idx, pages_per_huge_page(h), h_cg);
out_subpool_put:
2151
	if (map_chg || avoid_reserve)
2152
		hugepage_subpool_put_pages(spool, 1);
2153
	vma_end_reservation(h, vma, addr);
2154
	return ERR_PTR(-ENOSPC);
2155 2156
}

2157 2158 2159 2160 2161 2162 2163 2164 2165 2166 2167 2168 2169 2170
/*
 * 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;
}

2171
int __weak alloc_bootmem_huge_page(struct hstate *h)
2172 2173
{
	struct huge_bootmem_page *m;
2174
	int nr_nodes, node;
2175

2176
	for_each_node_mask_to_alloc(h, nr_nodes, node, &node_states[N_MEMORY]) {
2177 2178
		void *addr;

2179 2180 2181
		addr = memblock_virt_alloc_try_nid_nopanic(
				huge_page_size(h), huge_page_size(h),
				0, BOOTMEM_ALLOC_ACCESSIBLE, node);
2182 2183 2184 2185 2186 2187 2188
		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;
2189
			goto found;
2190 2191 2192 2193 2194
		}
	}
	return 0;

found:
2195
	BUG_ON(!IS_ALIGNED(virt_to_phys(m), huge_page_size(h)));
2196 2197 2198 2199 2200 2201
	/* 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;
}

2202 2203
static void __init prep_compound_huge_page(struct page *page,
		unsigned int order)
2204 2205 2206 2207 2208 2209 2210
{
	if (unlikely(order > (MAX_ORDER - 1)))
		prep_compound_gigantic_page(page, order);
	else
		prep_compound_page(page, order);
}

2211 2212 2213 2214 2215 2216 2217
/* 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;
2218 2219 2220 2221
		struct page *page;

#ifdef CONFIG_HIGHMEM
		page = pfn_to_page(m->phys >> PAGE_SHIFT);
2222 2223
		memblock_free_late(__pa(m),
				   sizeof(struct huge_bootmem_page));
2224 2225 2226
#else
		page = virt_to_page(m);
#endif
2227
		WARN_ON(page_count(page) != 1);
2228
		prep_compound_huge_page(page, h->order);
2229
		WARN_ON(PageReserved(page));
2230
		prep_new_huge_page(h, page, page_to_nid(page));
2231 2232 2233 2234 2235 2236
		/*
		 * 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.
		 */
2237
		if (hstate_is_gigantic(h))
2238
			adjust_managed_page_count(page, 1 << h->order);
2239 2240 2241
	}
}

2242
static void __init hugetlb_hstate_alloc_pages(struct hstate *h)
L
Linus Torvalds 已提交
2243 2244
{
	unsigned long i;
2245

2246
	for (i = 0; i < h->max_huge_pages; ++i) {
2247
		if (hstate_is_gigantic(h)) {
2248 2249
			if (!alloc_bootmem_huge_page(h))
				break;
2250
		} else if (!alloc_fresh_huge_page(h,
2251
					 &node_states[N_MEMORY]))
L
Linus Torvalds 已提交
2252
			break;
2253
		cond_resched();
L
Linus Torvalds 已提交
2254
	}
2255 2256 2257 2258 2259 2260 2261 2262
	if (i < h->max_huge_pages) {
		char buf[32];

		memfmt(buf, huge_page_size(h)),
		pr_warn("HugeTLB: allocating %lu of page size %s failed.  Only allocated %lu hugepages.\n",
			h->max_huge_pages, buf, i);
		h->max_huge_pages = i;
	}
2263 2264 2265 2266 2267 2268 2269
}

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

	for_each_hstate(h) {
2270 2271 2272
		if (minimum_order > huge_page_order(h))
			minimum_order = huge_page_order(h);

2273
		/* oversize hugepages were init'ed in early boot */
2274
		if (!hstate_is_gigantic(h))
2275
			hugetlb_hstate_alloc_pages(h);
2276
	}
2277
	VM_BUG_ON(minimum_order == UINT_MAX);
2278 2279 2280 2281 2282 2283 2284
}

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

	for_each_hstate(h) {
A
Andi Kleen 已提交
2285
		char buf[32];
2286
		pr_info("HugeTLB registered %s page size, pre-allocated %ld pages\n",
A
Andi Kleen 已提交
2287 2288
			memfmt(buf, huge_page_size(h)),
			h->free_huge_pages);
2289 2290 2291
	}
}

L
Linus Torvalds 已提交
2292
#ifdef CONFIG_HIGHMEM
2293 2294
static void try_to_free_low(struct hstate *h, unsigned long count,
						nodemask_t *nodes_allowed)
L
Linus Torvalds 已提交
2295
{
2296 2297
	int i;

2298
	if (hstate_is_gigantic(h))
2299 2300
		return;

2301
	for_each_node_mask(i, *nodes_allowed) {
L
Linus Torvalds 已提交
2302
		struct page *page, *next;
2303 2304 2305
		struct list_head *freel = &h->hugepage_freelists[i];
		list_for_each_entry_safe(page, next, freel, lru) {
			if (count >= h->nr_huge_pages)
2306
				return;
L
Linus Torvalds 已提交
2307 2308 2309
			if (PageHighMem(page))
				continue;
			list_del(&page->lru);
2310
			update_and_free_page(h, page);
2311 2312
			h->free_huge_pages--;
			h->free_huge_pages_node[page_to_nid(page)]--;
L
Linus Torvalds 已提交
2313 2314 2315 2316
		}
	}
}
#else
2317 2318
static inline void try_to_free_low(struct hstate *h, unsigned long count,
						nodemask_t *nodes_allowed)
L
Linus Torvalds 已提交
2319 2320 2321 2322
{
}
#endif

2323 2324 2325 2326 2327
/*
 * 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.
 */
2328 2329
static int adjust_pool_surplus(struct hstate *h, nodemask_t *nodes_allowed,
				int delta)
2330
{
2331
	int nr_nodes, node;
2332 2333 2334

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

2335 2336 2337 2338
	if (delta < 0) {
		for_each_node_mask_to_alloc(h, nr_nodes, node, nodes_allowed) {
			if (h->surplus_huge_pages_node[node])
				goto found;
2339
		}
2340 2341 2342 2343 2344
	} 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;
2345
		}
2346 2347
	}
	return 0;
2348

2349 2350 2351 2352
found:
	h->surplus_huge_pages += delta;
	h->surplus_huge_pages_node[node] += delta;
	return 1;
2353 2354
}

2355
#define persistent_huge_pages(h) (h->nr_huge_pages - h->surplus_huge_pages)
2356 2357
static unsigned long set_max_huge_pages(struct hstate *h, unsigned long count,
						nodemask_t *nodes_allowed)
L
Linus Torvalds 已提交
2358
{
2359
	unsigned long min_count, ret;
L
Linus Torvalds 已提交
2360

2361
	if (hstate_is_gigantic(h) && !gigantic_page_supported())
2362 2363
		return h->max_huge_pages;

2364 2365 2366 2367
	/*
	 * Increase the pool size
	 * First take pages out of surplus state.  Then make up the
	 * remaining difference by allocating fresh huge pages.
2368
	 *
N
Naoya Horiguchi 已提交
2369
	 * We might race with __alloc_buddy_huge_page() here and be unable
2370 2371 2372 2373
	 * 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.
2374
	 */
L
Linus Torvalds 已提交
2375
	spin_lock(&hugetlb_lock);
2376
	while (h->surplus_huge_pages && count > persistent_huge_pages(h)) {
2377
		if (!adjust_pool_surplus(h, nodes_allowed, -1))
2378 2379 2380
			break;
	}

2381
	while (count > persistent_huge_pages(h)) {
2382 2383 2384 2385 2386 2387
		/*
		 * 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);
2388 2389 2390 2391

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

2392 2393 2394 2395
		if (hstate_is_gigantic(h))
			ret = alloc_fresh_gigantic_page(h, nodes_allowed);
		else
			ret = alloc_fresh_huge_page(h, nodes_allowed);
2396 2397 2398 2399
		spin_lock(&hugetlb_lock);
		if (!ret)
			goto out;

2400 2401 2402
		/* Bail for signals. Probably ctrl-c from user */
		if (signal_pending(current))
			goto out;
2403 2404 2405 2406 2407 2408 2409 2410
	}

	/*
	 * 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.
2411 2412 2413 2414
	 *
	 * 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
2415
	 * __alloc_buddy_huge_page() is checking the global counter,
2416 2417 2418
	 * 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.
2419
	 */
2420
	min_count = h->resv_huge_pages + h->nr_huge_pages - h->free_huge_pages;
2421
	min_count = max(count, min_count);
2422
	try_to_free_low(h, min_count, nodes_allowed);
2423
	while (min_count < persistent_huge_pages(h)) {
2424
		if (!free_pool_huge_page(h, nodes_allowed, 0))
L
Linus Torvalds 已提交
2425
			break;
2426
		cond_resched_lock(&hugetlb_lock);
L
Linus Torvalds 已提交
2427
	}
2428
	while (count < persistent_huge_pages(h)) {
2429
		if (!adjust_pool_surplus(h, nodes_allowed, 1))
2430 2431 2432
			break;
	}
out:
2433
	ret = persistent_huge_pages(h);
L
Linus Torvalds 已提交
2434
	spin_unlock(&hugetlb_lock);
2435
	return ret;
L
Linus Torvalds 已提交
2436 2437
}

2438 2439 2440 2441 2442 2443 2444 2445 2446 2447
#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];

2448 2449 2450
static struct hstate *kobj_to_node_hstate(struct kobject *kobj, int *nidp);

static struct hstate *kobj_to_hstate(struct kobject *kobj, int *nidp)
2451 2452
{
	int i;
2453

2454
	for (i = 0; i < HUGE_MAX_HSTATE; i++)
2455 2456 2457
		if (hstate_kobjs[i] == kobj) {
			if (nidp)
				*nidp = NUMA_NO_NODE;
2458
			return &hstates[i];
2459 2460 2461
		}

	return kobj_to_node_hstate(kobj, nidp);
2462 2463
}

2464
static ssize_t nr_hugepages_show_common(struct kobject *kobj,
2465 2466
					struct kobj_attribute *attr, char *buf)
{
2467 2468 2469 2470 2471 2472 2473 2474 2475 2476 2477
	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);
2478
}
2479

2480 2481 2482
static ssize_t __nr_hugepages_store_common(bool obey_mempolicy,
					   struct hstate *h, int nid,
					   unsigned long count, size_t len)
2483 2484
{
	int err;
2485
	NODEMASK_ALLOC(nodemask_t, nodes_allowed, GFP_KERNEL | __GFP_NORETRY);
2486

2487
	if (hstate_is_gigantic(h) && !gigantic_page_supported()) {
2488 2489 2490 2491
		err = -EINVAL;
		goto out;
	}

2492 2493 2494 2495 2496 2497 2498
	if (nid == NUMA_NO_NODE) {
		/*
		 * global hstate attribute
		 */
		if (!(obey_mempolicy &&
				init_nodemask_of_mempolicy(nodes_allowed))) {
			NODEMASK_FREE(nodes_allowed);
2499
			nodes_allowed = &node_states[N_MEMORY];
2500 2501 2502 2503 2504 2505 2506 2507 2508
		}
	} 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
2509
		nodes_allowed = &node_states[N_MEMORY];
2510

2511
	h->max_huge_pages = set_max_huge_pages(h, count, nodes_allowed);
2512

2513
	if (nodes_allowed != &node_states[N_MEMORY])
2514 2515 2516
		NODEMASK_FREE(nodes_allowed);

	return len;
2517 2518 2519
out:
	NODEMASK_FREE(nodes_allowed);
	return err;
2520 2521
}

2522 2523 2524 2525 2526 2527 2528 2529 2530 2531 2532 2533 2534 2535 2536 2537 2538
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);
}

2539 2540 2541 2542 2543 2544 2545 2546 2547
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)
{
2548
	return nr_hugepages_store_common(false, kobj, buf, len);
2549 2550 2551
}
HSTATE_ATTR(nr_hugepages);

2552 2553 2554 2555 2556 2557 2558 2559 2560 2561 2562 2563 2564 2565 2566
#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)
{
2567
	return nr_hugepages_store_common(true, kobj, buf, len);
2568 2569 2570 2571 2572
}
HSTATE_ATTR(nr_hugepages_mempolicy);
#endif


2573 2574 2575
static ssize_t nr_overcommit_hugepages_show(struct kobject *kobj,
					struct kobj_attribute *attr, char *buf)
{
2576
	struct hstate *h = kobj_to_hstate(kobj, NULL);
2577 2578
	return sprintf(buf, "%lu\n", h->nr_overcommit_huge_pages);
}
2579

2580 2581 2582 2583 2584
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;
2585
	struct hstate *h = kobj_to_hstate(kobj, NULL);
2586

2587
	if (hstate_is_gigantic(h))
2588 2589
		return -EINVAL;

2590
	err = kstrtoul(buf, 10, &input);
2591
	if (err)
2592
		return err;
2593 2594 2595 2596 2597 2598 2599 2600 2601 2602 2603 2604

	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)
{
2605 2606 2607 2608 2609 2610 2611 2612 2613 2614 2615
	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);
2616 2617 2618 2619 2620 2621
}
HSTATE_ATTR_RO(free_hugepages);

static ssize_t resv_hugepages_show(struct kobject *kobj,
					struct kobj_attribute *attr, char *buf)
{
2622
	struct hstate *h = kobj_to_hstate(kobj, NULL);
2623 2624 2625 2626 2627 2628 2629
	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)
{
2630 2631 2632 2633 2634 2635 2636 2637 2638 2639 2640
	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);
2641 2642 2643 2644 2645 2646 2647 2648 2649
}
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,
2650 2651 2652
#ifdef CONFIG_NUMA
	&nr_hugepages_mempolicy_attr.attr,
#endif
2653 2654 2655 2656 2657 2658 2659
	NULL,
};

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

J
Jeff Mahoney 已提交
2660 2661 2662
static int hugetlb_sysfs_add_hstate(struct hstate *h, struct kobject *parent,
				    struct kobject **hstate_kobjs,
				    struct attribute_group *hstate_attr_group)
2663 2664
{
	int retval;
2665
	int hi = hstate_index(h);
2666

2667 2668
	hstate_kobjs[hi] = kobject_create_and_add(h->name, parent);
	if (!hstate_kobjs[hi])
2669 2670
		return -ENOMEM;

2671
	retval = sysfs_create_group(hstate_kobjs[hi], hstate_attr_group);
2672
	if (retval)
2673
		kobject_put(hstate_kobjs[hi]);
2674 2675 2676 2677 2678 2679 2680 2681 2682 2683 2684 2685 2686 2687

	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) {
2688 2689
		err = hugetlb_sysfs_add_hstate(h, hugepages_kobj,
					 hstate_kobjs, &hstate_attr_group);
2690
		if (err)
2691
			pr_err("Hugetlb: Unable to add hstate %s", h->name);
2692 2693 2694
	}
}

2695 2696 2697 2698
#ifdef CONFIG_NUMA

/*
 * node_hstate/s - associate per node hstate attributes, via their kobjects,
2699 2700 2701
 * 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
2702 2703 2704 2705 2706 2707
 * the base kernel, on the hugetlb module.
 */
struct node_hstate {
	struct kobject		*hugepages_kobj;
	struct kobject		*hstate_kobjs[HUGE_MAX_HSTATE];
};
2708
static struct node_hstate node_hstates[MAX_NUMNODES];
2709 2710

/*
2711
 * A subset of global hstate attributes for node devices
2712 2713 2714 2715 2716 2717 2718 2719 2720 2721 2722 2723 2724
 */
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,
};

/*
2725
 * kobj_to_node_hstate - lookup global hstate for node device hstate attr kobj.
2726 2727 2728 2729 2730 2731 2732 2733 2734 2735 2736 2737 2738 2739 2740 2741 2742 2743 2744 2745 2746 2747
 * 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;
}

/*
2748
 * Unregister hstate attributes from a single node device.
2749 2750
 * No-op if no hstate attributes attached.
 */
2751
static void hugetlb_unregister_node(struct node *node)
2752 2753
{
	struct hstate *h;
2754
	struct node_hstate *nhs = &node_hstates[node->dev.id];
2755 2756

	if (!nhs->hugepages_kobj)
2757
		return;		/* no hstate attributes */
2758

2759 2760 2761 2762 2763
	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;
2764
		}
2765
	}
2766 2767 2768 2769 2770 2771 2772

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


/*
2773
 * Register hstate attributes for a single node device.
2774 2775
 * No-op if attributes already registered.
 */
2776
static void hugetlb_register_node(struct node *node)
2777 2778
{
	struct hstate *h;
2779
	struct node_hstate *nhs = &node_hstates[node->dev.id];
2780 2781 2782 2783 2784 2785
	int err;

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

	nhs->hugepages_kobj = kobject_create_and_add("hugepages",
2786
							&node->dev.kobj);
2787 2788 2789 2790 2791 2792 2793 2794
	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) {
2795 2796
			pr_err("Hugetlb: Unable to add hstate %s for node %d\n",
				h->name, node->dev.id);
2797 2798 2799 2800 2801 2802 2803
			hugetlb_unregister_node(node);
			break;
		}
	}
}

/*
2804
 * hugetlb init time:  register hstate attributes for all registered node
2805 2806
 * devices of nodes that have memory.  All on-line nodes should have
 * registered their associated device by this time.
2807
 */
2808
static void __init hugetlb_register_all_nodes(void)
2809 2810 2811
{
	int nid;

2812
	for_each_node_state(nid, N_MEMORY) {
2813
		struct node *node = node_devices[nid];
2814
		if (node->dev.id == nid)
2815 2816 2817 2818
			hugetlb_register_node(node);
	}

	/*
2819
	 * Let the node device driver know we're here so it can
2820 2821 2822 2823 2824 2825 2826 2827 2828 2829 2830 2831 2832 2833 2834 2835 2836 2837 2838
	 * [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

2839 2840
static int __init hugetlb_init(void)
{
2841 2842
	int i;

2843
	if (!hugepages_supported())
2844
		return 0;
2845

2846
	if (!size_to_hstate(default_hstate_size)) {
2847 2848 2849 2850 2851
		if (default_hstate_size != 0) {
			pr_err("HugeTLB: unsupported default_hugepagesz %lu. Reverting to %lu\n",
			       default_hstate_size, HPAGE_SIZE);
		}

2852 2853 2854
		default_hstate_size = HPAGE_SIZE;
		if (!size_to_hstate(default_hstate_size))
			hugetlb_add_hstate(HUGETLB_PAGE_ORDER);
2855
	}
2856
	default_hstate_idx = hstate_index(size_to_hstate(default_hstate_size));
2857 2858 2859 2860
	if (default_hstate_max_huge_pages) {
		if (!default_hstate.max_huge_pages)
			default_hstate.max_huge_pages = default_hstate_max_huge_pages;
	}
2861 2862

	hugetlb_init_hstates();
2863
	gather_bootmem_prealloc();
2864 2865 2866
	report_hugepages();

	hugetlb_sysfs_init();
2867
	hugetlb_register_all_nodes();
2868
	hugetlb_cgroup_file_init();
2869

2870 2871 2872 2873 2874
#ifdef CONFIG_SMP
	num_fault_mutexes = roundup_pow_of_two(8 * num_possible_cpus());
#else
	num_fault_mutexes = 1;
#endif
2875
	hugetlb_fault_mutex_table =
2876
		kmalloc(sizeof(struct mutex) * num_fault_mutexes, GFP_KERNEL);
2877
	BUG_ON(!hugetlb_fault_mutex_table);
2878 2879

	for (i = 0; i < num_fault_mutexes; i++)
2880
		mutex_init(&hugetlb_fault_mutex_table[i]);
2881 2882
	return 0;
}
2883
subsys_initcall(hugetlb_init);
2884 2885

/* Should be called on processing a hugepagesz=... option */
2886 2887 2888 2889 2890
void __init hugetlb_bad_size(void)
{
	parsed_valid_hugepagesz = false;
}

2891
void __init hugetlb_add_hstate(unsigned int order)
2892 2893
{
	struct hstate *h;
2894 2895
	unsigned long i;

2896
	if (size_to_hstate(PAGE_SIZE << order)) {
J
Joe Perches 已提交
2897
		pr_warn("hugepagesz= specified twice, ignoring\n");
2898 2899
		return;
	}
2900
	BUG_ON(hugetlb_max_hstate >= HUGE_MAX_HSTATE);
2901
	BUG_ON(order == 0);
2902
	h = &hstates[hugetlb_max_hstate++];
2903 2904
	h->order = order;
	h->mask = ~((1ULL << (order + PAGE_SHIFT)) - 1);
2905 2906 2907 2908
	h->nr_huge_pages = 0;
	h->free_huge_pages = 0;
	for (i = 0; i < MAX_NUMNODES; ++i)
		INIT_LIST_HEAD(&h->hugepage_freelists[i]);
2909
	INIT_LIST_HEAD(&h->hugepage_activelist);
2910 2911
	h->next_nid_to_alloc = first_memory_node;
	h->next_nid_to_free = first_memory_node;
2912 2913
	snprintf(h->name, HSTATE_NAME_LEN, "hugepages-%lukB",
					huge_page_size(h)/1024);
2914

2915 2916 2917
	parsed_hstate = h;
}

2918
static int __init hugetlb_nrpages_setup(char *s)
2919 2920
{
	unsigned long *mhp;
2921
	static unsigned long *last_mhp;
2922

2923 2924 2925 2926 2927 2928
	if (!parsed_valid_hugepagesz) {
		pr_warn("hugepages = %s preceded by "
			"an unsupported hugepagesz, ignoring\n", s);
		parsed_valid_hugepagesz = true;
		return 1;
	}
2929
	/*
2930
	 * !hugetlb_max_hstate means we haven't parsed a hugepagesz= parameter yet,
2931 2932
	 * so this hugepages= parameter goes to the "default hstate".
	 */
2933
	else if (!hugetlb_max_hstate)
2934 2935 2936 2937
		mhp = &default_hstate_max_huge_pages;
	else
		mhp = &parsed_hstate->max_huge_pages;

2938
	if (mhp == last_mhp) {
J
Joe Perches 已提交
2939
		pr_warn("hugepages= specified twice without interleaving hugepagesz=, ignoring\n");
2940 2941 2942
		return 1;
	}

2943 2944 2945
	if (sscanf(s, "%lu", mhp) <= 0)
		*mhp = 0;

2946 2947 2948 2949 2950
	/*
	 * 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.
	 */
2951
	if (hugetlb_max_hstate && parsed_hstate->order >= MAX_ORDER)
2952 2953 2954 2955
		hugetlb_hstate_alloc_pages(parsed_hstate);

	last_mhp = mhp;

2956 2957
	return 1;
}
2958 2959 2960 2961 2962 2963 2964 2965
__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);
2966

2967 2968 2969 2970 2971 2972 2973 2974 2975 2976 2977 2978
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
2979 2980 2981
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
2982
{
2983
	struct hstate *h = &default_hstate;
2984
	unsigned long tmp = h->max_huge_pages;
2985
	int ret;
2986

2987
	if (!hugepages_supported())
2988
		return -EOPNOTSUPP;
2989

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

2996 2997 2998
	if (write)
		ret = __nr_hugepages_store_common(obey_mempolicy, h,
						  NUMA_NO_NODE, tmp, *length);
2999 3000
out:
	return ret;
L
Linus Torvalds 已提交
3001
}
3002

3003 3004 3005 3006 3007 3008 3009 3010 3011 3012 3013 3014 3015 3016 3017 3018 3019
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 */

3020
int hugetlb_overcommit_handler(struct ctl_table *table, int write,
3021
			void __user *buffer,
3022 3023
			size_t *length, loff_t *ppos)
{
3024
	struct hstate *h = &default_hstate;
3025
	unsigned long tmp;
3026
	int ret;
3027

3028
	if (!hugepages_supported())
3029
		return -EOPNOTSUPP;
3030

3031
	tmp = h->nr_overcommit_huge_pages;
3032

3033
	if (write && hstate_is_gigantic(h))
3034 3035
		return -EINVAL;

3036 3037
	table->data = &tmp;
	table->maxlen = sizeof(unsigned long);
3038 3039 3040
	ret = proc_doulongvec_minmax(table, write, buffer, length, ppos);
	if (ret)
		goto out;
3041 3042 3043 3044 3045 3046

	if (write) {
		spin_lock(&hugetlb_lock);
		h->nr_overcommit_huge_pages = tmp;
		spin_unlock(&hugetlb_lock);
	}
3047 3048
out:
	return ret;
3049 3050
}

L
Linus Torvalds 已提交
3051 3052
#endif /* CONFIG_SYSCTL */

3053
void hugetlb_report_meminfo(struct seq_file *m)
L
Linus Torvalds 已提交
3054
{
3055
	struct hstate *h = &default_hstate;
3056 3057
	if (!hugepages_supported())
		return;
3058
	seq_printf(m,
3059 3060 3061 3062 3063
			"HugePages_Total:   %5lu\n"
			"HugePages_Free:    %5lu\n"
			"HugePages_Rsvd:    %5lu\n"
			"HugePages_Surp:    %5lu\n"
			"Hugepagesize:   %8lu kB\n",
3064 3065 3066 3067 3068
			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
3069 3070 3071 3072
}

int hugetlb_report_node_meminfo(int nid, char *buf)
{
3073
	struct hstate *h = &default_hstate;
3074 3075
	if (!hugepages_supported())
		return 0;
L
Linus Torvalds 已提交
3076 3077
	return sprintf(buf,
		"Node %d HugePages_Total: %5u\n"
3078 3079
		"Node %d HugePages_Free:  %5u\n"
		"Node %d HugePages_Surp:  %5u\n",
3080 3081 3082
		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
3083 3084
}

3085 3086 3087 3088 3089
void hugetlb_show_meminfo(void)
{
	struct hstate *h;
	int nid;

3090 3091 3092
	if (!hugepages_supported())
		return;

3093 3094 3095 3096 3097 3098 3099 3100 3101 3102
	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));
}

3103 3104 3105 3106 3107 3108
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
3109 3110 3111
/* Return the number pages of memory we physically have, in PAGE_SIZE units. */
unsigned long hugetlb_total_pages(void)
{
3112 3113 3114 3115 3116 3117
	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
3118 3119
}

3120
static int hugetlb_acct_memory(struct hstate *h, long delta)
M
Mel Gorman 已提交
3121 3122 3123 3124 3125 3126 3127 3128 3129 3130 3131 3132 3133 3134 3135 3136 3137 3138 3139 3140 3141 3142
{
	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) {
3143
		if (gather_surplus_pages(h, delta) < 0)
M
Mel Gorman 已提交
3144 3145
			goto out;

3146 3147
		if (delta > cpuset_mems_nr(h->free_huge_pages_node)) {
			return_unused_surplus_pages(h, delta);
M
Mel Gorman 已提交
3148 3149 3150 3151 3152 3153
			goto out;
		}
	}

	ret = 0;
	if (delta < 0)
3154
		return_unused_surplus_pages(h, (unsigned long) -delta);
M
Mel Gorman 已提交
3155 3156 3157 3158 3159 3160

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

3161 3162
static void hugetlb_vm_op_open(struct vm_area_struct *vma)
{
3163
	struct resv_map *resv = vma_resv_map(vma);
3164 3165 3166 3167 3168

	/*
	 * 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
3169
	 * has a reference to the reservation map it cannot disappear until
3170 3171 3172
	 * after this open call completes.  It is therefore safe to take a
	 * new reference here without additional locking.
	 */
3173
	if (resv && is_vma_resv_set(vma, HPAGE_RESV_OWNER))
3174
		kref_get(&resv->refs);
3175 3176
}

3177 3178
static void hugetlb_vm_op_close(struct vm_area_struct *vma)
{
3179
	struct hstate *h = hstate_vma(vma);
3180
	struct resv_map *resv = vma_resv_map(vma);
3181
	struct hugepage_subpool *spool = subpool_vma(vma);
3182
	unsigned long reserve, start, end;
3183
	long gbl_reserve;
3184

3185 3186
	if (!resv || !is_vma_resv_set(vma, HPAGE_RESV_OWNER))
		return;
3187

3188 3189
	start = vma_hugecache_offset(h, vma, vma->vm_start);
	end = vma_hugecache_offset(h, vma, vma->vm_end);
3190

3191
	reserve = (end - start) - region_count(resv, start, end);
3192

3193 3194 3195
	kref_put(&resv->refs, resv_map_release);

	if (reserve) {
3196 3197 3198 3199 3200 3201
		/*
		 * 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);
3202
	}
3203 3204
}

L
Linus Torvalds 已提交
3205 3206 3207 3208 3209 3210
/*
 * 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.
 */
3211
static int hugetlb_vm_op_fault(struct vm_fault *vmf)
L
Linus Torvalds 已提交
3212 3213
{
	BUG();
N
Nick Piggin 已提交
3214
	return 0;
L
Linus Torvalds 已提交
3215 3216
}

3217
const struct vm_operations_struct hugetlb_vm_ops = {
N
Nick Piggin 已提交
3218
	.fault = hugetlb_vm_op_fault,
3219
	.open = hugetlb_vm_op_open,
3220
	.close = hugetlb_vm_op_close,
L
Linus Torvalds 已提交
3221 3222
};

3223 3224
static pte_t make_huge_pte(struct vm_area_struct *vma, struct page *page,
				int writable)
D
David Gibson 已提交
3225 3226 3227
{
	pte_t entry;

3228
	if (writable) {
3229 3230
		entry = huge_pte_mkwrite(huge_pte_mkdirty(mk_huge_pte(page,
					 vma->vm_page_prot)));
D
David Gibson 已提交
3231
	} else {
3232 3233
		entry = huge_pte_wrprotect(mk_huge_pte(page,
					   vma->vm_page_prot));
D
David Gibson 已提交
3234 3235 3236
	}
	entry = pte_mkyoung(entry);
	entry = pte_mkhuge(entry);
3237
	entry = arch_make_huge_pte(entry, vma, page, writable);
D
David Gibson 已提交
3238 3239 3240 3241

	return entry;
}

3242 3243 3244 3245 3246
static void set_huge_ptep_writable(struct vm_area_struct *vma,
				   unsigned long address, pte_t *ptep)
{
	pte_t entry;

3247
	entry = huge_pte_mkwrite(huge_pte_mkdirty(huge_ptep_get(ptep)));
3248
	if (huge_ptep_set_access_flags(vma, address, ptep, entry, 1))
3249
		update_mmu_cache(vma, address, ptep);
3250 3251
}

3252
bool is_hugetlb_entry_migration(pte_t pte)
3253 3254 3255 3256
{
	swp_entry_t swp;

	if (huge_pte_none(pte) || pte_present(pte))
3257
		return false;
3258 3259
	swp = pte_to_swp_entry(pte);
	if (non_swap_entry(swp) && is_migration_entry(swp))
3260
		return true;
3261
	else
3262
		return false;
3263 3264 3265 3266 3267 3268 3269 3270 3271 3272 3273 3274 3275 3276
}

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;
}
3277

D
David Gibson 已提交
3278 3279 3280 3281 3282
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;
3283
	unsigned long addr;
3284
	int cow;
3285 3286
	struct hstate *h = hstate_vma(vma);
	unsigned long sz = huge_page_size(h);
3287 3288 3289
	unsigned long mmun_start;	/* For mmu_notifiers */
	unsigned long mmun_end;		/* For mmu_notifiers */
	int ret = 0;
3290 3291

	cow = (vma->vm_flags & (VM_SHARED | VM_MAYWRITE)) == VM_MAYWRITE;
D
David Gibson 已提交
3292

3293 3294 3295 3296 3297
	mmun_start = vma->vm_start;
	mmun_end = vma->vm_end;
	if (cow)
		mmu_notifier_invalidate_range_start(src, mmun_start, mmun_end);

3298
	for (addr = vma->vm_start; addr < vma->vm_end; addr += sz) {
3299
		spinlock_t *src_ptl, *dst_ptl;
3300
		src_pte = huge_pte_offset(src, addr, sz);
H
Hugh Dickins 已提交
3301 3302
		if (!src_pte)
			continue;
3303
		dst_pte = huge_pte_alloc(dst, addr, sz);
3304 3305 3306 3307
		if (!dst_pte) {
			ret = -ENOMEM;
			break;
		}
3308 3309 3310 3311 3312

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

3313 3314 3315
		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);
3316 3317 3318 3319 3320 3321 3322 3323 3324 3325 3326 3327 3328 3329
		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);
3330 3331
				set_huge_swap_pte_at(src, addr, src_pte,
						     entry, sz);
3332
			}
3333
			set_huge_swap_pte_at(dst, addr, dst_pte, entry, sz);
3334
		} else {
3335
			if (cow) {
3336
				huge_ptep_set_wrprotect(src, addr, src_pte);
3337 3338 3339
				mmu_notifier_invalidate_range(src, mmun_start,
								   mmun_end);
			}
3340
			entry = huge_ptep_get(src_pte);
3341 3342
			ptepage = pte_page(entry);
			get_page(ptepage);
3343
			page_dup_rmap(ptepage, true);
3344
			set_huge_pte_at(dst, addr, dst_pte, entry);
3345
			hugetlb_count_add(pages_per_huge_page(h), dst);
3346
		}
3347 3348
		spin_unlock(src_ptl);
		spin_unlock(dst_ptl);
D
David Gibson 已提交
3349 3350
	}

3351 3352 3353 3354
	if (cow)
		mmu_notifier_invalidate_range_end(src, mmun_start, mmun_end);

	return ret;
D
David Gibson 已提交
3355 3356
}

3357 3358 3359
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
3360 3361 3362
{
	struct mm_struct *mm = vma->vm_mm;
	unsigned long address;
3363
	pte_t *ptep;
D
David Gibson 已提交
3364
	pte_t pte;
3365
	spinlock_t *ptl;
D
David Gibson 已提交
3366
	struct page *page;
3367 3368
	struct hstate *h = hstate_vma(vma);
	unsigned long sz = huge_page_size(h);
3369 3370
	const unsigned long mmun_start = start;	/* For mmu_notifiers */
	const unsigned long mmun_end   = end;	/* For mmu_notifiers */
3371

D
David Gibson 已提交
3372
	WARN_ON(!is_vm_hugetlb_page(vma));
3373 3374
	BUG_ON(start & ~huge_page_mask(h));
	BUG_ON(end & ~huge_page_mask(h));
D
David Gibson 已提交
3375

3376 3377 3378 3379 3380
	/*
	 * This is a hugetlb vma, all the pte entries should point
	 * to huge page.
	 */
	tlb_remove_check_page_size_change(tlb, sz);
3381
	tlb_start_vma(tlb, vma);
3382
	mmu_notifier_invalidate_range_start(mm, mmun_start, mmun_end);
3383 3384
	address = start;
	for (; address < end; address += sz) {
3385
		ptep = huge_pte_offset(mm, address, sz);
A
Adam Litke 已提交
3386
		if (!ptep)
3387 3388
			continue;

3389
		ptl = huge_pte_lock(h, mm, ptep);
3390 3391 3392 3393
		if (huge_pmd_unshare(mm, &address, ptep)) {
			spin_unlock(ptl);
			continue;
		}
3394

3395
		pte = huge_ptep_get(ptep);
3396 3397 3398 3399
		if (huge_pte_none(pte)) {
			spin_unlock(ptl);
			continue;
		}
3400 3401

		/*
3402 3403
		 * Migrating hugepage or HWPoisoned hugepage is already
		 * unmapped and its refcount is dropped, so just clear pte here.
3404
		 */
3405
		if (unlikely(!pte_present(pte))) {
3406
			huge_pte_clear(mm, address, ptep, sz);
3407 3408
			spin_unlock(ptl);
			continue;
3409
		}
3410 3411

		page = pte_page(pte);
3412 3413 3414 3415 3416 3417
		/*
		 * 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) {
3418 3419 3420 3421
			if (page != ref_page) {
				spin_unlock(ptl);
				continue;
			}
3422 3423 3424 3425 3426 3427 3428 3429
			/*
			 * 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);
		}

3430
		pte = huge_ptep_get_and_clear(mm, address, ptep);
3431
		tlb_remove_huge_tlb_entry(h, tlb, ptep, address);
3432
		if (huge_pte_dirty(pte))
3433
			set_page_dirty(page);
3434

3435
		hugetlb_count_sub(pages_per_huge_page(h), mm);
3436
		page_remove_rmap(page, true);
3437

3438
		spin_unlock(ptl);
3439
		tlb_remove_page_size(tlb, page, huge_page_size(h));
3440 3441 3442 3443 3444
		/*
		 * Bail out after unmapping reference page if supplied
		 */
		if (ref_page)
			break;
3445
	}
3446
	mmu_notifier_invalidate_range_end(mm, mmun_start, mmun_end);
3447
	tlb_end_vma(tlb, vma);
L
Linus Torvalds 已提交
3448
}
D
David Gibson 已提交
3449

3450 3451 3452 3453 3454 3455 3456 3457 3458 3459 3460 3461
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
3462
	 * is to clear it before releasing the i_mmap_rwsem. This works
3463
	 * because in the context this is called, the VMA is about to be
3464
	 * destroyed and the i_mmap_rwsem is held.
3465 3466 3467 3468
	 */
	vma->vm_flags &= ~VM_MAYSHARE;
}

3469
void unmap_hugepage_range(struct vm_area_struct *vma, unsigned long start,
3470
			  unsigned long end, struct page *ref_page)
3471
{
3472 3473 3474 3475 3476
	struct mm_struct *mm;
	struct mmu_gather tlb;

	mm = vma->vm_mm;

3477
	tlb_gather_mmu(&tlb, mm, start, end);
3478 3479
	__unmap_hugepage_range(&tlb, vma, start, end, ref_page);
	tlb_finish_mmu(&tlb, start, end);
3480 3481
}

3482 3483 3484 3485 3486 3487
/*
 * 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.
 */
3488 3489
static void unmap_ref_private(struct mm_struct *mm, struct vm_area_struct *vma,
			      struct page *page, unsigned long address)
3490
{
3491
	struct hstate *h = hstate_vma(vma);
3492 3493 3494 3495 3496 3497 3498 3499
	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.
	 */
3500
	address = address & huge_page_mask(h);
3501 3502
	pgoff = ((address - vma->vm_start) >> PAGE_SHIFT) +
			vma->vm_pgoff;
3503
	mapping = vma->vm_file->f_mapping;
3504

3505 3506 3507 3508 3509
	/*
	 * 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
	 */
3510
	i_mmap_lock_write(mapping);
3511
	vma_interval_tree_foreach(iter_vma, &mapping->i_mmap, pgoff, pgoff) {
3512 3513 3514 3515
		/* Do not unmap the current VMA */
		if (iter_vma == vma)
			continue;

3516 3517 3518 3519 3520 3521 3522 3523
		/*
		 * 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;

3524 3525 3526 3527 3528 3529 3530 3531
		/*
		 * 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))
3532 3533
			unmap_hugepage_range(iter_vma, address,
					     address + huge_page_size(h), page);
3534
	}
3535
	i_mmap_unlock_write(mapping);
3536 3537
}

3538 3539
/*
 * Hugetlb_cow() should be called with page lock of the original hugepage held.
3540 3541 3542
 * 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.
3543
 */
3544
static int hugetlb_cow(struct mm_struct *mm, struct vm_area_struct *vma,
3545 3546
		       unsigned long address, pte_t *ptep,
		       struct page *pagecache_page, spinlock_t *ptl)
3547
{
3548
	pte_t pte;
3549
	struct hstate *h = hstate_vma(vma);
3550
	struct page *old_page, *new_page;
3551
	int ret = 0, outside_reserve = 0;
3552 3553
	unsigned long mmun_start;	/* For mmu_notifiers */
	unsigned long mmun_end;		/* For mmu_notifiers */
3554

3555
	pte = huge_ptep_get(ptep);
3556 3557
	old_page = pte_page(pte);

3558
retry_avoidcopy:
3559 3560
	/* If no-one else is actually using this page, avoid the copy
	 * and just make the page writable */
3561
	if (page_mapcount(old_page) == 1 && PageAnon(old_page)) {
3562
		page_move_anon_rmap(old_page, vma);
3563
		set_huge_ptep_writable(vma, address, ptep);
N
Nick Piggin 已提交
3564
		return 0;
3565 3566
	}

3567 3568 3569 3570 3571 3572 3573 3574 3575
	/*
	 * 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.
	 */
3576
	if (is_vma_resv_set(vma, HPAGE_RESV_OWNER) &&
3577 3578 3579
			old_page != pagecache_page)
		outside_reserve = 1;

3580
	get_page(old_page);
3581

3582 3583 3584 3585
	/*
	 * Drop page table lock as buddy allocator may be called. It will
	 * be acquired again before returning to the caller, as expected.
	 */
3586
	spin_unlock(ptl);
3587
	new_page = alloc_huge_page(vma, address, outside_reserve);
3588

3589
	if (IS_ERR(new_page)) {
3590 3591 3592 3593 3594 3595 3596 3597
		/*
		 * 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) {
3598
			put_page(old_page);
3599
			BUG_ON(huge_pte_none(pte));
3600 3601 3602
			unmap_ref_private(mm, vma, old_page, address);
			BUG_ON(huge_pte_none(pte));
			spin_lock(ptl);
3603 3604
			ptep = huge_pte_offset(mm, address & huge_page_mask(h),
					       huge_page_size(h));
3605 3606 3607 3608 3609 3610 3611 3612
			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;
3613 3614
		}

3615 3616 3617
		ret = (PTR_ERR(new_page) == -ENOMEM) ?
			VM_FAULT_OOM : VM_FAULT_SIGBUS;
		goto out_release_old;
3618 3619
	}

3620 3621 3622 3623
	/*
	 * When the original hugepage is shared one, it does not have
	 * anon_vma prepared.
	 */
3624
	if (unlikely(anon_vma_prepare(vma))) {
3625 3626
		ret = VM_FAULT_OOM;
		goto out_release_all;
3627
	}
3628

A
Andrea Arcangeli 已提交
3629 3630
	copy_user_huge_page(new_page, old_page, address, vma,
			    pages_per_huge_page(h));
N
Nick Piggin 已提交
3631
	__SetPageUptodate(new_page);
3632
	set_page_huge_active(new_page);
3633

3634 3635 3636
	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);
3637

3638
	/*
3639
	 * Retake the page table lock to check for racing updates
3640 3641
	 * before the page tables are altered
	 */
3642
	spin_lock(ptl);
3643 3644
	ptep = huge_pte_offset(mm, address & huge_page_mask(h),
			       huge_page_size(h));
3645
	if (likely(ptep && pte_same(huge_ptep_get(ptep), pte))) {
3646 3647
		ClearPagePrivate(new_page);

3648
		/* Break COW */
3649
		huge_ptep_clear_flush(vma, address, ptep);
3650
		mmu_notifier_invalidate_range(mm, mmun_start, mmun_end);
3651 3652
		set_huge_pte_at(mm, address, ptep,
				make_huge_pte(vma, new_page, 1));
3653
		page_remove_rmap(old_page, true);
3654
		hugepage_add_new_anon_rmap(new_page, vma, address);
3655 3656 3657
		/* Make the old page be freed below */
		new_page = old_page;
	}
3658
	spin_unlock(ptl);
3659
	mmu_notifier_invalidate_range_end(mm, mmun_start, mmun_end);
3660
out_release_all:
3661
	restore_reserve_on_error(h, vma, address, new_page);
3662
	put_page(new_page);
3663
out_release_old:
3664
	put_page(old_page);
3665

3666 3667
	spin_lock(ptl); /* Caller expects lock to be held */
	return ret;
3668 3669
}

3670
/* Return the pagecache page at a given address within a VMA */
3671 3672
static struct page *hugetlbfs_pagecache_page(struct hstate *h,
			struct vm_area_struct *vma, unsigned long address)
3673 3674
{
	struct address_space *mapping;
3675
	pgoff_t idx;
3676 3677

	mapping = vma->vm_file->f_mapping;
3678
	idx = vma_hugecache_offset(h, vma, address);
3679 3680 3681 3682

	return find_lock_page(mapping, idx);
}

H
Hugh Dickins 已提交
3683 3684 3685 3686 3687
/*
 * 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
3688 3689 3690 3691 3692 3693 3694 3695 3696 3697 3698 3699 3700 3701 3702
			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;
}

3703 3704 3705 3706 3707 3708 3709 3710 3711 3712 3713 3714 3715 3716 3717 3718 3719
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;
}

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

3732 3733 3734
	/*
	 * 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
3735
	 * COW. Warn that such a situation has occurred as it may not be obvious
3736 3737
	 */
	if (is_vma_resv_set(vma, HPAGE_RESV_UNMAPPED)) {
3738
		pr_warn_ratelimited("PID %d killed due to inadequate hugepage pool\n",
3739
			   current->pid);
3740 3741 3742
		return ret;
	}

A
Adam Litke 已提交
3743 3744 3745 3746
	/*
	 * Use page lock to guard against racing truncation
	 * before we get page_table_lock.
	 */
3747 3748 3749
retry:
	page = find_lock_page(mapping, idx);
	if (!page) {
3750
		size = i_size_read(mapping->host) >> huge_page_shift(h);
3751 3752
		if (idx >= size)
			goto out;
3753 3754 3755 3756 3757 3758 3759 3760 3761 3762 3763 3764 3765 3766 3767 3768 3769 3770 3771 3772 3773 3774 3775 3776 3777 3778 3779 3780 3781 3782 3783 3784

		/*
		 * 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;
		}

3785
		page = alloc_huge_page(vma, address, 0);
3786
		if (IS_ERR(page)) {
3787 3788 3789 3790 3791
			ret = PTR_ERR(page);
			if (ret == -ENOMEM)
				ret = VM_FAULT_OOM;
			else
				ret = VM_FAULT_SIGBUS;
3792 3793
			goto out;
		}
A
Andrea Arcangeli 已提交
3794
		clear_huge_page(page, address, pages_per_huge_page(h));
N
Nick Piggin 已提交
3795
		__SetPageUptodate(page);
3796
		set_page_huge_active(page);
3797

3798
		if (vma->vm_flags & VM_MAYSHARE) {
3799
			int err = huge_add_to_page_cache(page, mapping, idx);
3800 3801 3802 3803 3804 3805
			if (err) {
				put_page(page);
				if (err == -EEXIST)
					goto retry;
				goto out;
			}
3806
		} else {
3807
			lock_page(page);
3808 3809 3810 3811
			if (unlikely(anon_vma_prepare(vma))) {
				ret = VM_FAULT_OOM;
				goto backout_unlocked;
			}
3812
			anon_rmap = 1;
3813
		}
3814
	} else {
3815 3816 3817 3818 3819 3820
		/*
		 * 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))) {
3821
			ret = VM_FAULT_HWPOISON |
3822
				VM_FAULT_SET_HINDEX(hstate_index(h));
3823 3824
			goto backout_unlocked;
		}
3825
	}
3826

3827 3828 3829 3830 3831 3832
	/*
	 * 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.
	 */
3833
	if ((flags & FAULT_FLAG_WRITE) && !(vma->vm_flags & VM_SHARED)) {
3834 3835 3836 3837
		if (vma_needs_reservation(h, vma, address) < 0) {
			ret = VM_FAULT_OOM;
			goto backout_unlocked;
		}
3838
		/* Just decrements count, does not deallocate */
3839
		vma_end_reservation(h, vma, address);
3840
	}
3841

3842
	ptl = huge_pte_lock(h, mm, ptep);
3843
	size = i_size_read(mapping->host) >> huge_page_shift(h);
A
Adam Litke 已提交
3844 3845 3846
	if (idx >= size)
		goto backout;

N
Nick Piggin 已提交
3847
	ret = 0;
3848
	if (!huge_pte_none(huge_ptep_get(ptep)))
A
Adam Litke 已提交
3849 3850
		goto backout;

3851 3852
	if (anon_rmap) {
		ClearPagePrivate(page);
3853
		hugepage_add_new_anon_rmap(page, vma, address);
3854
	} else
3855
		page_dup_rmap(page, true);
3856 3857 3858 3859
	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);

3860
	hugetlb_count_add(pages_per_huge_page(h), mm);
3861
	if ((flags & FAULT_FLAG_WRITE) && !(vma->vm_flags & VM_SHARED)) {
3862
		/* Optimization, do the COW without a second fault */
3863
		ret = hugetlb_cow(mm, vma, address, ptep, page, ptl);
3864 3865
	}

3866
	spin_unlock(ptl);
A
Adam Litke 已提交
3867 3868
	unlock_page(page);
out:
3869
	return ret;
A
Adam Litke 已提交
3870 3871

backout:
3872
	spin_unlock(ptl);
3873
backout_unlocked:
A
Adam Litke 已提交
3874
	unlock_page(page);
3875
	restore_reserve_on_error(h, vma, address, page);
A
Adam Litke 已提交
3876 3877
	put_page(page);
	goto out;
3878 3879
}

3880
#ifdef CONFIG_SMP
3881
u32 hugetlb_fault_mutex_hash(struct hstate *h, struct mm_struct *mm,
3882 3883 3884 3885 3886 3887 3888 3889 3890 3891 3892 3893 3894 3895 3896 3897 3898 3899 3900 3901 3902 3903 3904 3905
			    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.
 */
3906
u32 hugetlb_fault_mutex_hash(struct hstate *h, struct mm_struct *mm,
3907 3908 3909 3910 3911 3912 3913 3914
			    struct vm_area_struct *vma,
			    struct address_space *mapping,
			    pgoff_t idx, unsigned long address)
{
	return 0;
}
#endif

3915
int hugetlb_fault(struct mm_struct *mm, struct vm_area_struct *vma,
3916
			unsigned long address, unsigned int flags)
3917
{
3918
	pte_t *ptep, entry;
3919
	spinlock_t *ptl;
3920
	int ret;
3921 3922
	u32 hash;
	pgoff_t idx;
3923
	struct page *page = NULL;
3924
	struct page *pagecache_page = NULL;
3925
	struct hstate *h = hstate_vma(vma);
3926
	struct address_space *mapping;
3927
	int need_wait_lock = 0;
3928

3929 3930
	address &= huge_page_mask(h);

3931
	ptep = huge_pte_offset(mm, address, huge_page_size(h));
3932 3933
	if (ptep) {
		entry = huge_ptep_get(ptep);
N
Naoya Horiguchi 已提交
3934
		if (unlikely(is_hugetlb_entry_migration(entry))) {
3935
			migration_entry_wait_huge(vma, mm, ptep);
N
Naoya Horiguchi 已提交
3936 3937
			return 0;
		} else if (unlikely(is_hugetlb_entry_hwpoisoned(entry)))
3938
			return VM_FAULT_HWPOISON_LARGE |
3939
				VM_FAULT_SET_HINDEX(hstate_index(h));
3940 3941 3942 3943
	} else {
		ptep = huge_pte_alloc(mm, address, huge_page_size(h));
		if (!ptep)
			return VM_FAULT_OOM;
3944 3945
	}

3946 3947 3948
	mapping = vma->vm_file->f_mapping;
	idx = vma_hugecache_offset(h, vma, address);

3949 3950 3951 3952 3953
	/*
	 * 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.
	 */
3954 3955
	hash = hugetlb_fault_mutex_hash(h, mm, vma, mapping, idx, address);
	mutex_lock(&hugetlb_fault_mutex_table[hash]);
3956

3957 3958
	entry = huge_ptep_get(ptep);
	if (huge_pte_none(entry)) {
3959
		ret = hugetlb_no_page(mm, vma, mapping, idx, address, ptep, flags);
3960
		goto out_mutex;
3961
	}
3962

N
Nick Piggin 已提交
3963
	ret = 0;
3964

3965 3966 3967 3968 3969 3970 3971 3972 3973 3974
	/*
	 * 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;

3975 3976 3977 3978 3979 3980 3981 3982
	/*
	 * 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.
	 */
3983
	if ((flags & FAULT_FLAG_WRITE) && !huge_pte_write(entry)) {
3984 3985
		if (vma_needs_reservation(h, vma, address) < 0) {
			ret = VM_FAULT_OOM;
3986
			goto out_mutex;
3987
		}
3988
		/* Just decrements count, does not deallocate */
3989
		vma_end_reservation(h, vma, address);
3990

3991
		if (!(vma->vm_flags & VM_MAYSHARE))
3992 3993 3994 3995
			pagecache_page = hugetlbfs_pagecache_page(h,
								vma, address);
	}

3996 3997 3998 3999 4000 4001
	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;

4002 4003 4004 4005 4006 4007 4008
	/*
	 * 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)
4009 4010 4011 4012
		if (!trylock_page(page)) {
			need_wait_lock = 1;
			goto out_ptl;
		}
4013

4014
	get_page(page);
4015

4016
	if (flags & FAULT_FLAG_WRITE) {
4017
		if (!huge_pte_write(entry)) {
4018 4019
			ret = hugetlb_cow(mm, vma, address, ptep,
					  pagecache_page, ptl);
4020
			goto out_put_page;
4021
		}
4022
		entry = huge_pte_mkdirty(entry);
4023 4024
	}
	entry = pte_mkyoung(entry);
4025 4026
	if (huge_ptep_set_access_flags(vma, address, ptep, entry,
						flags & FAULT_FLAG_WRITE))
4027
		update_mmu_cache(vma, address, ptep);
4028 4029 4030 4031
out_put_page:
	if (page != pagecache_page)
		unlock_page(page);
	put_page(page);
4032 4033
out_ptl:
	spin_unlock(ptl);
4034 4035 4036 4037 4038

	if (pagecache_page) {
		unlock_page(pagecache_page);
		put_page(pagecache_page);
	}
4039
out_mutex:
4040
	mutex_unlock(&hugetlb_fault_mutex_table[hash]);
4041 4042 4043 4044 4045 4046 4047 4048 4049
	/*
	 * 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);
4050
	return ret;
4051 4052
}

4053 4054 4055 4056 4057 4058 4059 4060 4061 4062 4063
/*
 * 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)
{
4064
	int vm_shared = dst_vma->vm_flags & VM_SHARED;
4065 4066 4067 4068 4069 4070 4071 4072 4073 4074 4075 4076 4077 4078
	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,
4079
						pages_per_huge_page(h), false);
4080 4081 4082 4083 4084 4085 4086 4087 4088 4089 4090 4091 4092 4093 4094 4095 4096 4097 4098 4099 4100

		/* 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);

4101 4102 4103 4104 4105 4106 4107 4108 4109 4110 4111 4112
	/*
	 * 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;
	}

4113 4114 4115 4116 4117 4118 4119
	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;

4120 4121 4122 4123 4124 4125
	if (vm_shared) {
		page_dup_rmap(page, true);
	} else {
		ClearPagePrivate(page);
		hugepage_add_new_anon_rmap(page, dst_vma, dst_addr);
	}
4126 4127 4128 4129 4130 4131 4132 4133 4134 4135 4136 4137 4138 4139 4140 4141

	_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);
4142 4143
	if (vm_shared)
		unlock_page(page);
4144 4145 4146 4147 4148
	ret = 0;
out:
	return ret;
out_release_unlock:
	spin_unlock(ptl);
4149 4150 4151
out_release_nounlock:
	if (vm_shared)
		unlock_page(page);
4152 4153 4154 4155
	put_page(page);
	goto out;
}

4156 4157 4158
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,
4159
			 long i, unsigned int flags, int *nonblocking)
D
David Gibson 已提交
4160
{
4161 4162
	unsigned long pfn_offset;
	unsigned long vaddr = *position;
4163
	unsigned long remainder = *nr_pages;
4164
	struct hstate *h = hstate_vma(vma);
D
David Gibson 已提交
4165 4166

	while (vaddr < vma->vm_end && remainder) {
A
Adam Litke 已提交
4167
		pte_t *pte;
4168
		spinlock_t *ptl = NULL;
H
Hugh Dickins 已提交
4169
		int absent;
A
Adam Litke 已提交
4170
		struct page *page;
D
David Gibson 已提交
4171

4172 4173 4174 4175 4176 4177 4178 4179 4180
		/*
		 * 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
4181 4182
		/*
		 * Some archs (sparc64, sh*) have multiple pte_ts to
H
Hugh Dickins 已提交
4183
		 * each hugepage.  We have to make sure we get the
A
Adam Litke 已提交
4184
		 * first, for the page indexing below to work.
4185 4186
		 *
		 * Note that page table lock is not held when pte is null.
A
Adam Litke 已提交
4187
		 */
4188 4189
		pte = huge_pte_offset(mm, vaddr & huge_page_mask(h),
				      huge_page_size(h));
4190 4191
		if (pte)
			ptl = huge_pte_lock(h, mm, pte);
H
Hugh Dickins 已提交
4192 4193 4194 4195
		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
4196 4197 4198 4199
		 * 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
4200
		 */
H
Hugh Dickins 已提交
4201 4202
		if (absent && (flags & FOLL_DUMP) &&
		    !hugetlbfs_pagecache_present(h, vma, vaddr)) {
4203 4204
			if (pte)
				spin_unlock(ptl);
H
Hugh Dickins 已提交
4205 4206 4207
			remainder = 0;
			break;
		}
D
David Gibson 已提交
4208

4209 4210 4211 4212 4213 4214 4215 4216 4217 4218 4219
		/*
		 * 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)) ||
4220 4221
		    ((flags & FOLL_WRITE) &&
		      !huge_pte_write(huge_ptep_get(pte)))) {
A
Adam Litke 已提交
4222
			int ret;
4223
			unsigned int fault_flags = 0;
D
David Gibson 已提交
4224

4225 4226
			if (pte)
				spin_unlock(ptl);
4227 4228 4229 4230 4231 4232 4233 4234 4235 4236 4237 4238 4239 4240
			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) {
4241 4242 4243 4244 4245
				int err = vm_fault_to_errno(ret, flags);

				if (err)
					return err;

4246 4247 4248 4249 4250 4251 4252 4253 4254 4255 4256 4257 4258 4259 4260 4261 4262 4263 4264
				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
4265 4266
		}

4267
		pfn_offset = (vaddr & ~huge_page_mask(h)) >> PAGE_SHIFT;
4268
		page = pte_page(huge_ptep_get(pte));
4269
same_page:
4270
		if (pages) {
H
Hugh Dickins 已提交
4271
			pages[i] = mem_map_offset(page, pfn_offset);
4272
			get_page(pages[i]);
4273
		}
D
David Gibson 已提交
4274 4275 4276 4277 4278

		if (vmas)
			vmas[i] = vma;

		vaddr += PAGE_SIZE;
4279
		++pfn_offset;
D
David Gibson 已提交
4280 4281
		--remainder;
		++i;
4282
		if (vaddr < vma->vm_end && remainder &&
4283
				pfn_offset < pages_per_huge_page(h)) {
4284 4285 4286 4287 4288 4289
			/*
			 * We use pfn_offset to avoid touching the pageframes
			 * of this compound page.
			 */
			goto same_page;
		}
4290
		spin_unlock(ptl);
D
David Gibson 已提交
4291
	}
4292
	*nr_pages = remainder;
4293 4294 4295 4296 4297
	/*
	 * 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
4298 4299
	*position = vaddr;

H
Hugh Dickins 已提交
4300
	return i ? i : -EFAULT;
D
David Gibson 已提交
4301
}
4302

4303 4304 4305 4306 4307 4308 4309 4310
#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

4311
unsigned long hugetlb_change_protection(struct vm_area_struct *vma,
4312 4313 4314 4315 4316 4317
		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;
4318
	struct hstate *h = hstate_vma(vma);
4319
	unsigned long pages = 0;
4320 4321 4322 4323

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

4324
	mmu_notifier_invalidate_range_start(mm, start, end);
4325
	i_mmap_lock_write(vma->vm_file->f_mapping);
4326
	for (; address < end; address += huge_page_size(h)) {
4327
		spinlock_t *ptl;
4328
		ptep = huge_pte_offset(mm, address, huge_page_size(h));
4329 4330
		if (!ptep)
			continue;
4331
		ptl = huge_pte_lock(h, mm, ptep);
4332 4333
		if (huge_pmd_unshare(mm, &address, ptep)) {
			pages++;
4334
			spin_unlock(ptl);
4335
			continue;
4336
		}
4337 4338 4339 4340 4341 4342 4343 4344 4345 4346 4347 4348 4349
		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);
4350 4351
				set_huge_swap_pte_at(mm, address, ptep,
						     newpte, huge_page_size(h));
4352 4353 4354 4355 4356 4357
				pages++;
			}
			spin_unlock(ptl);
			continue;
		}
		if (!huge_pte_none(pte)) {
4358
			pte = huge_ptep_get_and_clear(mm, address, ptep);
4359
			pte = pte_mkhuge(huge_pte_modify(pte, newprot));
4360
			pte = arch_make_huge_pte(pte, vma, NULL, 0);
4361
			set_huge_pte_at(mm, address, ptep, pte);
4362
			pages++;
4363
		}
4364
		spin_unlock(ptl);
4365
	}
4366
	/*
4367
	 * Must flush TLB before releasing i_mmap_rwsem: x86's huge_pmd_unshare
4368
	 * may have cleared our pud entry and done put_page on the page table:
4369
	 * once we release i_mmap_rwsem, another task can do the final put_page
4370 4371
	 * and that page table be reused and filled with junk.
	 */
4372
	flush_hugetlb_tlb_range(vma, start, end);
4373
	mmu_notifier_invalidate_range(mm, start, end);
4374
	i_mmap_unlock_write(vma->vm_file->f_mapping);
4375
	mmu_notifier_invalidate_range_end(mm, start, end);
4376 4377

	return pages << h->order;
4378 4379
}

4380 4381
int hugetlb_reserve_pages(struct inode *inode,
					long from, long to,
4382
					struct vm_area_struct *vma,
4383
					vm_flags_t vm_flags)
4384
{
4385
	long ret, chg;
4386
	struct hstate *h = hstate_inode(inode);
4387
	struct hugepage_subpool *spool = subpool_inode(inode);
4388
	struct resv_map *resv_map;
4389
	long gbl_reserve;
4390

4391 4392 4393
	/*
	 * Only apply hugepage reservation if asked. At fault time, an
	 * attempt will be made for VM_NORESERVE to allocate a page
4394
	 * without using reserves
4395
	 */
4396
	if (vm_flags & VM_NORESERVE)
4397 4398
		return 0;

4399 4400 4401 4402 4403 4404
	/*
	 * 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
	 */
4405
	if (!vma || vma->vm_flags & VM_MAYSHARE) {
4406
		resv_map = inode_resv_map(inode);
4407

4408
		chg = region_chg(resv_map, from, to);
4409 4410 4411

	} else {
		resv_map = resv_map_alloc();
4412 4413 4414
		if (!resv_map)
			return -ENOMEM;

4415
		chg = to - from;
4416

4417 4418 4419 4420
		set_vma_resv_map(vma, resv_map);
		set_vma_resv_flags(vma, HPAGE_RESV_OWNER);
	}

4421 4422 4423 4424
	if (chg < 0) {
		ret = chg;
		goto out_err;
	}
4425

4426 4427 4428 4429 4430 4431 4432
	/*
	 * 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) {
4433 4434 4435
		ret = -ENOSPC;
		goto out_err;
	}
4436 4437

	/*
4438
	 * Check enough hugepages are available for the reservation.
4439
	 * Hand the pages back to the subpool if there are not
4440
	 */
4441
	ret = hugetlb_acct_memory(h, gbl_reserve);
K
Ken Chen 已提交
4442
	if (ret < 0) {
4443 4444
		/* put back original number of pages, chg */
		(void)hugepage_subpool_put_pages(spool, chg);
4445
		goto out_err;
K
Ken Chen 已提交
4446
	}
4447 4448 4449 4450 4451 4452 4453 4454 4455 4456 4457 4458

	/*
	 * 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
	 */
4459 4460 4461 4462 4463 4464 4465 4466 4467 4468 4469 4470 4471 4472 4473 4474 4475 4476
	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);
		}
	}
4477
	return 0;
4478
out_err:
4479
	if (!vma || vma->vm_flags & VM_MAYSHARE)
4480 4481 4482
		/* Don't call region_abort if region_chg failed */
		if (chg >= 0)
			region_abort(resv_map, from, to);
J
Joonsoo Kim 已提交
4483 4484
	if (vma && is_vma_resv_set(vma, HPAGE_RESV_OWNER))
		kref_put(&resv_map->refs, resv_map_release);
4485
	return ret;
4486 4487
}

4488 4489
long hugetlb_unreserve_pages(struct inode *inode, long start, long end,
								long freed)
4490
{
4491
	struct hstate *h = hstate_inode(inode);
4492
	struct resv_map *resv_map = inode_resv_map(inode);
4493
	long chg = 0;
4494
	struct hugepage_subpool *spool = subpool_inode(inode);
4495
	long gbl_reserve;
K
Ken Chen 已提交
4496

4497 4498 4499 4500 4501 4502 4503 4504 4505 4506 4507
	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
4508
	spin_lock(&inode->i_lock);
4509
	inode->i_blocks -= (blocks_per_huge_page(h) * freed);
K
Ken Chen 已提交
4510 4511
	spin_unlock(&inode->i_lock);

4512 4513 4514 4515 4516 4517
	/*
	 * 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);
4518 4519

	return 0;
4520
}
4521

4522 4523 4524 4525 4526 4527 4528 4529 4530 4531 4532
#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
4533 4534
	unsigned long vm_flags = vma->vm_flags & VM_LOCKED_CLEAR_MASK;
	unsigned long svm_flags = svma->vm_flags & VM_LOCKED_CLEAR_MASK;
4535 4536 4537 4538 4539 4540 4541 4542 4543 4544 4545 4546 4547

	/*
	 * 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;
}

4548
static bool vma_shareable(struct vm_area_struct *vma, unsigned long addr)
4549 4550 4551 4552 4553 4554 4555 4556 4557
{
	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)
4558 4559
		return true;
	return false;
4560 4561 4562 4563 4564 4565 4566
}

/*
 * 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
4567
 * pud has to be populated inside the same i_mmap_rwsem section - otherwise
4568 4569 4570 4571 4572 4573 4574 4575 4576 4577 4578 4579 4580
 * 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;
4581
	spinlock_t *ptl;
4582 4583 4584 4585

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

4586
	i_mmap_lock_write(mapping);
4587 4588 4589 4590 4591 4592
	vma_interval_tree_foreach(svma, &mapping->i_mmap, idx, idx) {
		if (svma == vma)
			continue;

		saddr = page_table_shareable(svma, vma, addr, idx);
		if (saddr) {
4593 4594
			spte = huge_pte_offset(svma->vm_mm, saddr,
					       vma_mmu_pagesize(svma));
4595 4596 4597 4598 4599 4600 4601 4602 4603 4604
			if (spte) {
				get_page(virt_to_page(spte));
				break;
			}
		}
	}

	if (!spte)
		goto out;

4605
	ptl = huge_pte_lock(hstate_vma(vma), mm, spte);
4606
	if (pud_none(*pud)) {
4607 4608
		pud_populate(mm, pud,
				(pmd_t *)((unsigned long)spte & PAGE_MASK));
4609
		mm_inc_nr_pmds(mm);
4610
	} else {
4611
		put_page(virt_to_page(spte));
4612
	}
4613
	spin_unlock(ptl);
4614 4615
out:
	pte = (pte_t *)pmd_alloc(mm, pud, addr);
4616
	i_mmap_unlock_write(mapping);
4617 4618 4619 4620 4621 4622 4623 4624 4625 4626
	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.
 *
4627
 * called with page table lock held.
4628 4629 4630 4631 4632 4633 4634
 *
 * 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);
4635 4636
	p4d_t *p4d = p4d_offset(pgd, *addr);
	pud_t *pud = pud_offset(p4d, *addr);
4637 4638 4639 4640 4641 4642 4643

	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));
4644
	mm_dec_nr_pmds(mm);
4645 4646 4647
	*addr = ALIGN(*addr, HPAGE_SIZE * PTRS_PER_PTE) - HPAGE_SIZE;
	return 1;
}
4648 4649 4650 4651 4652 4653
#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;
}
4654 4655 4656 4657 4658

int huge_pmd_unshare(struct mm_struct *mm, unsigned long *addr, pte_t *ptep)
{
	return 0;
}
4659
#define want_pmd_share()	(0)
4660 4661
#endif /* CONFIG_ARCH_WANT_HUGE_PMD_SHARE */

4662 4663 4664 4665 4666
#ifdef CONFIG_ARCH_WANT_GENERAL_HUGETLB
pte_t *huge_pte_alloc(struct mm_struct *mm,
			unsigned long addr, unsigned long sz)
{
	pgd_t *pgd;
4667
	p4d_t *p4d;
4668 4669 4670 4671
	pud_t *pud;
	pte_t *pte = NULL;

	pgd = pgd_offset(mm, addr);
4672 4673
	p4d = p4d_offset(pgd, addr);
	pud = pud_alloc(mm, p4d, addr);
4674 4675 4676 4677 4678 4679 4680 4681 4682 4683 4684
	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);
		}
	}
4685
	BUG_ON(pte && pte_present(*pte) && !pte_huge(*pte));
4686 4687 4688 4689

	return pte;
}

4690 4691
pte_t *huge_pte_offset(struct mm_struct *mm,
		       unsigned long addr, unsigned long sz)
4692 4693
{
	pgd_t *pgd;
4694
	p4d_t *p4d;
4695
	pud_t *pud;
4696
	pmd_t *pmd;
4697 4698

	pgd = pgd_offset(mm, addr);
4699 4700 4701 4702 4703 4704 4705 4706 4707 4708 4709
	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);
4710 4711 4712
	return (pte_t *) pmd;
}

4713 4714 4715 4716 4717 4718 4719 4720 4721 4722 4723 4724 4725
#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);
}

4726 4727 4728 4729 4730 4731 4732 4733
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;
}

4734
struct page * __weak
4735
follow_huge_pmd(struct mm_struct *mm, unsigned long address,
4736
		pmd_t *pmd, int flags)
4737
{
4738 4739
	struct page *page = NULL;
	spinlock_t *ptl;
4740
	pte_t pte;
4741 4742 4743 4744 4745 4746 4747 4748 4749
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;
4750 4751
	pte = huge_ptep_get((pte_t *)pmd);
	if (pte_present(pte)) {
4752
		page = pmd_page(*pmd) + ((address & ~PMD_MASK) >> PAGE_SHIFT);
4753 4754 4755
		if (flags & FOLL_GET)
			get_page(page);
	} else {
4756
		if (is_hugetlb_entry_migration(pte)) {
4757 4758 4759 4760 4761 4762 4763 4764 4765 4766 4767
			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);
4768 4769 4770
	return page;
}

4771
struct page * __weak
4772
follow_huge_pud(struct mm_struct *mm, unsigned long address,
4773
		pud_t *pud, int flags)
4774
{
4775 4776
	if (flags & FOLL_GET)
		return NULL;
4777

4778
	return pte_page(*(pte_t *)pud) + ((address & ~PUD_MASK) >> PAGE_SHIFT);
4779 4780
}

4781 4782 4783 4784 4785 4786 4787 4788 4789
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);
}

4790 4791
bool isolate_huge_page(struct page *page, struct list_head *list)
{
4792 4793
	bool ret = true;

4794
	VM_BUG_ON_PAGE(!PageHead(page), page);
4795
	spin_lock(&hugetlb_lock);
4796 4797 4798 4799 4800
	if (!page_huge_active(page) || !get_page_unless_zero(page)) {
		ret = false;
		goto unlock;
	}
	clear_page_huge_active(page);
4801
	list_move_tail(&page->lru, list);
4802
unlock:
4803
	spin_unlock(&hugetlb_lock);
4804
	return ret;
4805 4806 4807 4808
}

void putback_active_hugepage(struct page *page)
{
4809
	VM_BUG_ON_PAGE(!PageHead(page), page);
4810
	spin_lock(&hugetlb_lock);
4811
	set_page_huge_active(page);
4812 4813 4814 4815
	list_move_tail(&page->lru, &(page_hstate(page))->hugepage_activelist);
	spin_unlock(&hugetlb_lock);
	put_page(page);
}