hugetlb.c 143.0 KB
Newer Older
1
// SPDX-License-Identifier: GPL-2.0-only
L
Linus Torvalds 已提交
2 3
/*
 * Generic hugetlb support.
4
 * (C) Nadia Yvette Chambers, April 2004
L
Linus Torvalds 已提交
5 6 7 8
 */
#include <linux/list.h>
#include <linux/init.h>
#include <linux/mm.h>
9
#include <linux/seq_file.h>
L
Linus Torvalds 已提交
10 11
#include <linux/sysctl.h>
#include <linux/highmem.h>
A
Andrea Arcangeli 已提交
12
#include <linux/mmu_notifier.h>
L
Linus Torvalds 已提交
13
#include <linux/nodemask.h>
D
David Gibson 已提交
14
#include <linux/pagemap.h>
15
#include <linux/mempolicy.h>
16
#include <linux/compiler.h>
17
#include <linux/cpuset.h>
18
#include <linux/mutex.h>
19
#include <linux/memblock.h>
20
#include <linux/sysfs.h>
21
#include <linux/slab.h>
22
#include <linux/mmdebug.h>
23
#include <linux/sched/signal.h>
24
#include <linux/rmap.h>
25
#include <linux/string_helpers.h>
26 27
#include <linux/swap.h>
#include <linux/swapops.h>
28
#include <linux/jhash.h>
29
#include <linux/numa.h>
30
#include <linux/llist.h>
31

D
David Gibson 已提交
32 33
#include <asm/page.h>
#include <asm/pgtable.h>
34
#include <asm/tlb.h>
D
David Gibson 已提交
35

36
#include <linux/io.h>
D
David Gibson 已提交
37
#include <linux/hugetlb.h>
38
#include <linux/hugetlb_cgroup.h>
39
#include <linux/node.h>
40
#include <linux/userfaultfd_k.h>
41
#include <linux/page_owner.h>
42
#include "internal.h"
L
Linus Torvalds 已提交
43

44
int hugetlb_max_hstate __read_mostly;
45 46
unsigned int default_hstate_idx;
struct hstate hstates[HUGE_MAX_HSTATE];
47 48 49 50 51
/*
 * Minimum page order among possible hugepage sizes, set to a proper value
 * at boot time.
 */
static unsigned int minimum_order __read_mostly = UINT_MAX;
52

53 54
__initdata LIST_HEAD(huge_boot_pages);

55 56 57
/* for command line parsing */
static struct hstate * __initdata parsed_hstate;
static unsigned long __initdata default_hstate_max_huge_pages;
58
static unsigned long __initdata default_hstate_size;
59
static bool __initdata parsed_valid_hugepagesz = true;
60

61
/*
62 63
 * Protects updates to hugepage_freelists, hugepage_activelist, nr_huge_pages,
 * free_huge_pages, and surplus_huge_pages.
64
 */
65
DEFINE_SPINLOCK(hugetlb_lock);
66

67 68 69 70 71
/*
 * 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;
72
struct mutex *hugetlb_fault_mutex_table ____cacheline_aligned_in_smp;
73

74 75 76
/* Forward declaration */
static int hugetlb_acct_memory(struct hstate *h, long delta);

77 78 79 80 81 82 83
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
84 85 86 87 88 89
	 * 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);
90
		kfree(spool);
91
	}
92 93
}

94 95
struct hugepage_subpool *hugepage_new_subpool(struct hstate *h, long max_hpages,
						long min_hpages)
96 97 98
{
	struct hugepage_subpool *spool;

99
	spool = kzalloc(sizeof(*spool), GFP_KERNEL);
100 101 102 103 104
	if (!spool)
		return NULL;

	spin_lock_init(&spool->lock);
	spool->count = 1;
105 106 107 108 109 110 111 112 113
	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;
114 115 116 117 118 119 120 121 122 123 124 125

	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);
}

126 127 128 129 130 131 132 133 134
/*
 * 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,
135 136
				      long delta)
{
137
	long ret = delta;
138 139

	if (!spool)
140
		return ret;
141 142

	spin_lock(&spool->lock);
143 144 145 146 147 148 149 150

	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;
		}
151 152
	}

153 154
	/* minimum size accounting */
	if (spool->min_hpages != -1 && spool->rsv_hpages) {
155 156 157 158 159 160 161 162 163 164 165 166 167 168 169
		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);
170 171 172
	return ret;
}

173 174 175 176 177 178 179
/*
 * 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,
180 181
				       long delta)
{
182 183
	long ret = delta;

184
	if (!spool)
185
		return delta;
186 187

	spin_lock(&spool->lock);
188 189 190 191

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

192 193
	 /* minimum size accounting */
	if (spool->min_hpages != -1 && spool->used_hpages < spool->min_hpages) {
194 195 196 197 198 199 200 201 202 203 204 205 206 207
		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.
	 */
208
	unlock_or_release_subpool(spool);
209 210

	return ret;
211 212 213 214 215 216 217 218 219
}

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
220
	return subpool_inode(file_inode(vma->vm_file));
221 222
}

223 224 225
/*
 * Region tracking -- allows tracking of reservations and instantiated pages
 *                    across the pages in a mapping.
226
 *
227 228 229 230 231 232 233 234 235 236 237 238 239 240
 * 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.
241 242 243 244 245 246 247
 */
struct file_region {
	struct list_head link;
	long from;
	long to;
};

M
Mina Almasry 已提交
248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301
/* Must be called with resv->lock held. Calling this with count_only == true
 * will count the number of pages to be added but will not modify the linked
 * list.
 */
static long add_reservation_in_range(struct resv_map *resv, long f, long t,
				     bool count_only)
{
	long chg = 0;
	struct list_head *head = &resv->regions;
	struct file_region *rg = NULL, *trg = NULL, *nrg = NULL;

	/* Locate the region we are before or in. */
	list_for_each_entry(rg, head, link)
		if (f <= rg->to)
			break;

	/* Round our left edge to the current segment if it encloses us. */
	if (f > rg->from)
		f = rg->from;

	chg = t - f;

	/* 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;

		/* We overlap with this area, if it extends further than
		 * 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;

		if (!count_only && rg != nrg) {
			list_del(&rg->link);
			kfree(rg);
		}
	}

	if (!count_only) {
		nrg->from = f;
		nrg->to = t;
	}

	return chg;
}

302 303
/*
 * Add the huge page range represented by [f, t) to the reserve
304 305 306 307
 * map.  Existing regions will be expanded to accommodate the specified
 * range, or a region will be taken from the cache.  Sufficient regions
 * must exist in the cache due to the previous call to region_chg with
 * the same range.
308 309 310
 *
 * Return the number of new huge pages added to the map.  This
 * number is greater than or equal to zero.
311
 */
312
static long region_add(struct resv_map *resv, long f, long t)
313
{
314
	struct list_head *head = &resv->regions;
M
Mina Almasry 已提交
315
	struct file_region *rg, *nrg;
316
	long add = 0;
317

318
	spin_lock(&resv->lock);
319 320 321 322 323
	/* Locate the region we are either in or before. */
	list_for_each_entry(rg, head, link)
		if (f <= rg->to)
			break;

324 325
	/*
	 * If no region exists which can be expanded to include the
326 327
	 * specified range, pull a region descriptor from the cache
	 * and use it for this range.
328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344
	 */
	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;
	}

M
Mina Almasry 已提交
345
	add = add_reservation_in_range(resv, f, t, false);
346

347 348
out_locked:
	resv->adds_in_progress--;
349
	spin_unlock(&resv->lock);
350 351
	VM_BUG_ON(add < 0);
	return add;
352 353
}

354 355 356 357 358 359 360
/*
 * 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
361 362 363
 * map.  A new file_region structure is added to the cache
 * as a placeholder, so that the subsequent region_add
 * call will have all the regions it needs and will not fail.
364 365 366 367 368
 *
 * 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.
369
 */
370
static long region_chg(struct resv_map *resv, long f, long t)
371 372 373
{
	long chg = 0;

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

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

M
Mina Almasry 已提交
400
	chg = add_reservation_in_range(resv, f, t, true);
401 402

	spin_unlock(&resv->lock);
403 404 405
	return chg;
}

406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424
/*
 * 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);
}

425
/*
426 427 428 429 430 431 432 433 434 435 436 437
 * 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.
438
 */
439
static long region_del(struct resv_map *resv, long f, long t)
440
{
441
	struct list_head *head = &resv->regions;
442
	struct file_region *rg, *trg;
443 444
	struct file_region *nrg = NULL;
	long del = 0;
445

446
retry:
447
	spin_lock(&resv->lock);
448
	list_for_each_entry_safe(rg, trg, head, link) {
449 450 451 452 453 454 455 456
		/*
		 * 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))
457
			continue;
458

459
		if (rg->from >= t)
460 461
			break;

462 463 464 465 466 467 468 469 470 471 472 473 474
		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--;
			}
475

476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495
			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;
496
			break;
497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512
		}

		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;
		}
513
	}
514 515

	spin_unlock(&resv->lock);
516 517
	kfree(nrg);
	return del;
518 519
}

520 521 522 523 524 525 526 527 528
/*
 * 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.
 */
529
void hugetlb_fix_reserve_counts(struct inode *inode)
530 531 532 533 534
{
	struct hugepage_subpool *spool = subpool_inode(inode);
	long rsv_adjust;

	rsv_adjust = hugepage_subpool_get_pages(spool, 1);
535
	if (rsv_adjust) {
536 537 538 539 540 541
		struct hstate *h = hstate_inode(inode);

		hugetlb_acct_memory(h, 1);
	}
}

542 543 544 545
/*
 * Count and return the number of huge pages in the reserve map
 * that intersect with the range [f, t).
 */
546
static long region_count(struct resv_map *resv, long f, long t)
547
{
548
	struct list_head *head = &resv->regions;
549 550 551
	struct file_region *rg;
	long chg = 0;

552
	spin_lock(&resv->lock);
553 554
	/* Locate each segment we overlap with, and count that overlap. */
	list_for_each_entry(rg, head, link) {
555 556
		long seg_from;
		long seg_to;
557 558 559 560 561 562 563 564 565 566 567

		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;
	}
568
	spin_unlock(&resv->lock);
569 570 571 572

	return chg;
}

573 574 575 576
/*
 * Convert the address within this vma to the page offset within
 * the mapping, in pagecache page units; huge pages here.
 */
577 578
static pgoff_t vma_hugecache_offset(struct hstate *h,
			struct vm_area_struct *vma, unsigned long address)
579
{
580 581
	return ((address - vma->vm_start) >> huge_page_shift(h)) +
			(vma->vm_pgoff >> huge_page_order(h));
582 583
}

584 585 586 587 588
pgoff_t linear_hugepage_index(struct vm_area_struct *vma,
				     unsigned long address)
{
	return vma_hugecache_offset(hstate_vma(vma), vma, address);
}
589
EXPORT_SYMBOL_GPL(linear_hugepage_index);
590

591 592 593 594 595 596
/*
 * 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)
{
597 598 599
	if (vma->vm_ops && vma->vm_ops->pagesize)
		return vma->vm_ops->pagesize(vma);
	return PAGE_SIZE;
600
}
601
EXPORT_SYMBOL_GPL(vma_kernel_pagesize);
602

603 604 605
/*
 * 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
606 607
 * architectures where it differs, an architecture-specific 'strong'
 * version of this symbol is required.
608
 */
609
__weak unsigned long vma_mmu_pagesize(struct vm_area_struct *vma)
610 611 612 613
{
	return vma_kernel_pagesize(vma);
}

614 615 616 617 618 619 620
/*
 * 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)
621
#define HPAGE_RESV_MASK (HPAGE_RESV_OWNER | HPAGE_RESV_UNMAPPED)
622

623 624 625 626 627 628 629 630 631
/*
 * 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.
632 633 634 635 636 637 638 639 640
 *
 * 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.
641
 */
642 643 644 645 646 647 648 649 650 651 652
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;
}

653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671
static void
resv_map_set_hugetlb_cgroup_uncharge_info(struct resv_map *resv_map,
					  struct hugetlb_cgroup *h_cg,
					  struct hstate *h)
{
#ifdef CONFIG_CGROUP_HUGETLB
	if (!h_cg || !h) {
		resv_map->reservation_counter = NULL;
		resv_map->pages_per_hpage = 0;
		resv_map->css = NULL;
	} else {
		resv_map->reservation_counter =
			&h_cg->rsvd_hugepage[hstate_index(h)];
		resv_map->pages_per_hpage = pages_per_huge_page(h);
		resv_map->css = &h_cg->css;
	}
#endif
}

672
struct resv_map *resv_map_alloc(void)
673 674
{
	struct resv_map *resv_map = kmalloc(sizeof(*resv_map), GFP_KERNEL);
675 676 677 678 679
	struct file_region *rg = kmalloc(sizeof(*rg), GFP_KERNEL);

	if (!resv_map || !rg) {
		kfree(resv_map);
		kfree(rg);
680
		return NULL;
681
	}
682 683

	kref_init(&resv_map->refs);
684
	spin_lock_init(&resv_map->lock);
685 686
	INIT_LIST_HEAD(&resv_map->regions);

687
	resv_map->adds_in_progress = 0;
688 689 690 691 692 693 694
	/*
	 * Initialize these to 0. On shared mappings, 0's here indicate these
	 * fields don't do cgroup accounting. On private mappings, these will be
	 * re-initialized to the proper values, to indicate that hugetlb cgroup
	 * reservations are to be un-charged from here.
	 */
	resv_map_set_hugetlb_cgroup_uncharge_info(resv_map, NULL, NULL);
695 696 697 698 699

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

700 701 702
	return resv_map;
}

703
void resv_map_release(struct kref *ref)
704 705
{
	struct resv_map *resv_map = container_of(ref, struct resv_map, refs);
706 707
	struct list_head *head = &resv_map->region_cache;
	struct file_region *rg, *trg;
708 709

	/* Clear out any active regions before we release the map. */
710
	region_del(resv_map, 0, LONG_MAX);
711 712 713 714 715 716 717 718 719

	/* ... 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);

720 721 722
	kfree(resv_map);
}

723 724
static inline struct resv_map *inode_resv_map(struct inode *inode)
{
725 726 727 728 729 730 731 732 733
	/*
	 * At inode evict time, i_mapping may not point to the original
	 * address space within the inode.  This original address space
	 * contains the pointer to the resv_map.  So, always use the
	 * address space embedded within the inode.
	 * The VERY common case is inode->mapping == &inode->i_data but,
	 * this may not be true for device special inodes.
	 */
	return (struct resv_map *)(&inode->i_data)->private_data;
734 735
}

736
static struct resv_map *vma_resv_map(struct vm_area_struct *vma)
737
{
738
	VM_BUG_ON_VMA(!is_vm_hugetlb_page(vma), vma);
739 740 741 742 743 744 745
	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 {
746 747
		return (struct resv_map *)(get_vma_private_data(vma) &
							~HPAGE_RESV_MASK);
748
	}
749 750
}

751
static void set_vma_resv_map(struct vm_area_struct *vma, struct resv_map *map)
752
{
753 754
	VM_BUG_ON_VMA(!is_vm_hugetlb_page(vma), vma);
	VM_BUG_ON_VMA(vma->vm_flags & VM_MAYSHARE, vma);
755

756 757
	set_vma_private_data(vma, (get_vma_private_data(vma) &
				HPAGE_RESV_MASK) | (unsigned long)map);
758 759 760 761
}

static void set_vma_resv_flags(struct vm_area_struct *vma, unsigned long flags)
{
762 763
	VM_BUG_ON_VMA(!is_vm_hugetlb_page(vma), vma);
	VM_BUG_ON_VMA(vma->vm_flags & VM_MAYSHARE, vma);
764 765

	set_vma_private_data(vma, get_vma_private_data(vma) | flags);
766 767 768 769
}

static int is_vma_resv_set(struct vm_area_struct *vma, unsigned long flag)
{
770
	VM_BUG_ON_VMA(!is_vm_hugetlb_page(vma), vma);
771 772

	return (get_vma_private_data(vma) & flag) != 0;
773 774
}

775
/* Reset counters to 0 and clear all HPAGE_RESV_* flags */
776 777
void reset_vma_resv_huge_pages(struct vm_area_struct *vma)
{
778
	VM_BUG_ON_VMA(!is_vm_hugetlb_page(vma), vma);
779
	if (!(vma->vm_flags & VM_MAYSHARE))
780 781 782 783
		vma->vm_private_data = (void *)0;
}

/* Returns true if the VMA has associated reserve pages */
784
static bool vma_has_reserves(struct vm_area_struct *vma, long chg)
785
{
786 787 788 789 790 791 792 793 794 795 796
	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)
797
			return true;
798
		else
799
			return false;
800
	}
801 802

	/* Shared mappings always use reserves */
803 804 805 806 807 808 809 810 811 812 813 814 815
	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;
	}
816 817 818 819 820

	/*
	 * Only the process that called mmap() has reserves for
	 * private mappings.
	 */
821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841
	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;
	}
842

843
	return false;
844 845
}

846
static void enqueue_huge_page(struct hstate *h, struct page *page)
L
Linus Torvalds 已提交
847 848
{
	int nid = page_to_nid(page);
849
	list_move(&page->lru, &h->hugepage_freelists[nid]);
850 851
	h->free_huge_pages++;
	h->free_huge_pages_node[nid]++;
L
Linus Torvalds 已提交
852 853
}

854
static struct page *dequeue_huge_page_node_exact(struct hstate *h, int nid)
855 856 857
{
	struct page *page;

858
	list_for_each_entry(page, &h->hugepage_freelists[nid], lru)
859
		if (!PageHWPoison(page))
860 861 862 863 864 865
			break;
	/*
	 * if 'non-isolated free hugepage' not found on the list,
	 * the allocation fails.
	 */
	if (&h->hugepage_freelists[nid] == &page->lru)
866
		return NULL;
867
	list_move(&page->lru, &h->hugepage_activelist);
868
	set_page_refcounted(page);
869 870 871 872 873
	h->free_huge_pages--;
	h->free_huge_pages_node[nid]--;
	return page;
}

874 875
static struct page *dequeue_huge_page_nodemask(struct hstate *h, gfp_t gfp_mask, int nid,
		nodemask_t *nmask)
876
{
877 878 879 880
	unsigned int cpuset_mems_cookie;
	struct zonelist *zonelist;
	struct zone *zone;
	struct zoneref *z;
881
	int node = NUMA_NO_NODE;
882

883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898
	zonelist = node_zonelist(nid, gfp_mask);

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

		if (!cpuset_zone_allowed(zone, gfp_mask))
			continue;
		/*
		 * no need to ask again on the same node. Pool is node rather than
		 * zone aware
		 */
		if (zone_to_nid(zone) == node)
			continue;
		node = zone_to_nid(zone);
899 900 901 902 903

		page = dequeue_huge_page_node_exact(h, node);
		if (page)
			return page;
	}
904 905 906
	if (unlikely(read_mems_allowed_retry(cpuset_mems_cookie)))
		goto retry_cpuset;

907 908 909
	return NULL;
}

910 911 912
/* Movability of hugepages depends on migration support. */
static inline gfp_t htlb_alloc_mask(struct hstate *h)
{
913
	if (hugepage_movable_supported(h))
914 915 916 917 918
		return GFP_HIGHUSER_MOVABLE;
	else
		return GFP_HIGHUSER;
}

919 920
static struct page *dequeue_huge_page_vma(struct hstate *h,
				struct vm_area_struct *vma,
921 922
				unsigned long address, int avoid_reserve,
				long chg)
L
Linus Torvalds 已提交
923
{
924
	struct page *page;
925
	struct mempolicy *mpol;
926
	gfp_t gfp_mask;
927
	nodemask_t *nodemask;
928
	int nid;
L
Linus Torvalds 已提交
929

930 931 932 933 934
	/*
	 * 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
	 */
935
	if (!vma_has_reserves(vma, chg) &&
936
			h->free_huge_pages - h->resv_huge_pages == 0)
937
		goto err;
938

939
	/* If reserves cannot be used, ensure enough pages are in the pool */
940
	if (avoid_reserve && h->free_huge_pages - h->resv_huge_pages == 0)
941
		goto err;
942

943 944
	gfp_mask = htlb_alloc_mask(h);
	nid = huge_node(vma, address, gfp_mask, &mpol, &nodemask);
945 946 947 948
	page = dequeue_huge_page_nodemask(h, gfp_mask, nid, nodemask);
	if (page && !avoid_reserve && vma_has_reserves(vma, chg)) {
		SetPagePrivate(page);
		h->resv_huge_pages--;
L
Linus Torvalds 已提交
949
	}
950

951
	mpol_cond_put(mpol);
L
Linus Torvalds 已提交
952
	return page;
953 954 955

err:
	return NULL;
L
Linus Torvalds 已提交
956 957
}

958 959 960 961 962 963 964 965 966
/*
 * 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)
{
967
	nid = next_node_in(nid, *nodes_allowed);
968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 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
	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--)

1029
#ifdef CONFIG_ARCH_HAS_GIGANTIC_PAGE
1030
static void destroy_compound_gigantic_page(struct page *page,
1031
					unsigned int order)
1032 1033 1034 1035 1036
{
	int i;
	int nr_pages = 1 << order;
	struct page *p = page + 1;

1037
	atomic_set(compound_mapcount_ptr(page), 0);
1038 1039 1040
	if (hpage_pincount_available(page))
		atomic_set(compound_pincount_ptr(page), 0);

1041
	for (i = 1; i < nr_pages; i++, p = mem_map_next(p, page, i)) {
1042
		clear_compound_head(p);
1043 1044 1045 1046 1047 1048 1049
		set_page_refcounted(p);
	}

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

1050
static void free_gigantic_page(struct page *page, unsigned int order)
1051 1052 1053 1054
{
	free_contig_range(page_to_pfn(page), 1 << order);
}

1055
#ifdef CONFIG_CONTIG_ALLOC
1056 1057
static struct page *alloc_gigantic_page(struct hstate *h, gfp_t gfp_mask,
		int nid, nodemask_t *nodemask)
1058
{
1059
	unsigned long nr_pages = 1UL << huge_page_order(h);
1060

1061
	return alloc_contig_pages(nr_pages, gfp_mask, nid, nodemask);
1062 1063 1064
}

static void prep_new_huge_page(struct hstate *h, struct page *page, int nid);
1065
static void prep_compound_gigantic_page(struct page *page, unsigned int order);
1066 1067 1068 1069 1070 1071 1072
#else /* !CONFIG_CONTIG_ALLOC */
static struct page *alloc_gigantic_page(struct hstate *h, gfp_t gfp_mask,
					int nid, nodemask_t *nodemask)
{
	return NULL;
}
#endif /* CONFIG_CONTIG_ALLOC */
1073

1074
#else /* !CONFIG_ARCH_HAS_GIGANTIC_PAGE */
1075
static struct page *alloc_gigantic_page(struct hstate *h, gfp_t gfp_mask,
1076 1077 1078 1079
					int nid, nodemask_t *nodemask)
{
	return NULL;
}
1080
static inline void free_gigantic_page(struct page *page, unsigned int order) { }
1081
static inline void destroy_compound_gigantic_page(struct page *page,
1082
						unsigned int order) { }
1083 1084
#endif

1085
static void update_and_free_page(struct hstate *h, struct page *page)
A
Adam Litke 已提交
1086 1087
{
	int i;
1088

1089
	if (hstate_is_gigantic(h) && !gigantic_page_runtime_supported())
1090
		return;
1091

1092 1093 1094
	h->nr_huge_pages--;
	h->nr_huge_pages_node[page_to_nid(page)]--;
	for (i = 0; i < pages_per_huge_page(h); i++) {
1095 1096
		page[i].flags &= ~(1 << PG_locked | 1 << PG_error |
				1 << PG_referenced | 1 << PG_dirty |
1097 1098
				1 << PG_active | 1 << PG_private |
				1 << PG_writeback);
A
Adam Litke 已提交
1099
	}
1100
	VM_BUG_ON_PAGE(hugetlb_cgroup_from_page(page), page);
1101
	VM_BUG_ON_PAGE(hugetlb_cgroup_from_page_rsvd(page), page);
1102
	set_compound_page_dtor(page, NULL_COMPOUND_DTOR);
A
Adam Litke 已提交
1103
	set_page_refcounted(page);
1104 1105 1106 1107 1108 1109
	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
1110 1111
}

1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122
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;
}

1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147
/*
 * 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]);
}

1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169
/*
 * Internal hugetlb specific page flag. Do not use outside of the hugetlb
 * code
 */
static inline bool PageHugeTemporary(struct page *page)
{
	if (!PageHuge(page))
		return false;

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

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

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

1170
static void __free_huge_page(struct page *page)
1171
{
1172 1173 1174 1175
	/*
	 * Can't pass hstate in here because it is called from the
	 * compound page destructor.
	 */
1176
	struct hstate *h = page_hstate(page);
1177
	int nid = page_to_nid(page);
1178 1179
	struct hugepage_subpool *spool =
		(struct hugepage_subpool *)page_private(page);
1180
	bool restore_reserve;
1181

1182 1183
	VM_BUG_ON_PAGE(page_count(page), page);
	VM_BUG_ON_PAGE(page_mapcount(page), page);
1184 1185 1186

	set_page_private(page, 0);
	page->mapping = NULL;
1187
	restore_reserve = PagePrivate(page);
1188
	ClearPagePrivate(page);
1189

1190
	/*
1191 1192 1193 1194 1195 1196
	 * If PagePrivate() was set on page, page allocation consumed a
	 * reservation.  If the page was associated with a subpool, there
	 * would have been a page reserved in the subpool before allocation
	 * via hugepage_subpool_get_pages().  Since we are 'restoring' the
	 * reservtion, do not call hugepage_subpool_put_pages() as this will
	 * remove the reserved page from the subpool.
1197
	 */
1198 1199 1200 1201 1202 1203 1204 1205 1206 1207
	if (!restore_reserve) {
		/*
		 * 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;
	}
1208

1209
	spin_lock(&hugetlb_lock);
1210
	clear_page_huge_active(page);
1211 1212
	hugetlb_cgroup_uncharge_page(hstate_index(h),
				     pages_per_huge_page(h), page);
1213 1214 1215
	if (restore_reserve)
		h->resv_huge_pages++;

1216 1217 1218 1219 1220
	if (PageHugeTemporary(page)) {
		list_del(&page->lru);
		ClearPageHugeTemporary(page);
		update_and_free_page(h, page);
	} else if (h->surplus_huge_pages_node[nid]) {
1221 1222
		/* remove the page from active list */
		list_del(&page->lru);
1223 1224 1225
		update_and_free_page(h, page);
		h->surplus_huge_pages--;
		h->surplus_huge_pages_node[nid]--;
1226
	} else {
1227
		arch_clear_hugepage_flags(page);
1228
		enqueue_huge_page(h, page);
1229
	}
1230 1231 1232
	spin_unlock(&hugetlb_lock);
}

1233 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 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280
/*
 * As free_huge_page() can be called from a non-task context, we have
 * to defer the actual freeing in a workqueue to prevent potential
 * hugetlb_lock deadlock.
 *
 * free_hpage_workfn() locklessly retrieves the linked list of pages to
 * be freed and frees them one-by-one. As the page->mapping pointer is
 * going to be cleared in __free_huge_page() anyway, it is reused as the
 * llist_node structure of a lockless linked list of huge pages to be freed.
 */
static LLIST_HEAD(hpage_freelist);

static void free_hpage_workfn(struct work_struct *work)
{
	struct llist_node *node;
	struct page *page;

	node = llist_del_all(&hpage_freelist);

	while (node) {
		page = container_of((struct address_space **)node,
				     struct page, mapping);
		node = node->next;
		__free_huge_page(page);
	}
}
static DECLARE_WORK(free_hpage_work, free_hpage_workfn);

void free_huge_page(struct page *page)
{
	/*
	 * Defer freeing if in non-task context to avoid hugetlb_lock deadlock.
	 */
	if (!in_task()) {
		/*
		 * Only call schedule_work() if hpage_freelist is previously
		 * empty. Otherwise, schedule_work() had been called but the
		 * workfn hasn't retrieved the list yet.
		 */
		if (llist_add((struct llist_node *)&page->mapping,
			      &hpage_freelist))
			schedule_work(&free_hpage_work);
		return;
	}

	__free_huge_page(page);
}

1281
static void prep_new_huge_page(struct hstate *h, struct page *page, int nid)
1282
{
1283
	INIT_LIST_HEAD(&page->lru);
1284
	set_compound_page_dtor(page, HUGETLB_PAGE_DTOR);
1285
	spin_lock(&hugetlb_lock);
1286
	set_hugetlb_cgroup(page, NULL);
1287
	set_hugetlb_cgroup_rsvd(page, NULL);
1288 1289
	h->nr_huge_pages++;
	h->nr_huge_pages_node[nid]++;
1290 1291 1292
	spin_unlock(&hugetlb_lock);
}

1293
static void prep_compound_gigantic_page(struct page *page, unsigned int order)
1294 1295 1296 1297 1298 1299 1300
{
	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);
1301
	__ClearPageReserved(page);
1302
	__SetPageHead(page);
1303
	for (i = 1; i < nr_pages; i++, p = mem_map_next(p, page, i)) {
1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316
		/*
		 * 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);
1317
		set_page_count(p, 0);
1318
		set_compound_head(p, page);
1319
	}
1320
	atomic_set(compound_mapcount_ptr(page), -1);
1321 1322 1323

	if (hpage_pincount_available(page))
		atomic_set(compound_pincount_ptr(page), 0);
1324 1325
}

A
Andrew Morton 已提交
1326 1327 1328 1329 1330
/*
 * PageHuge() only returns true for hugetlbfs pages, but not for normal or
 * transparent huge pages.  See the PageTransHuge() documentation for more
 * details.
 */
1331 1332 1333 1334 1335 1336
int PageHuge(struct page *page)
{
	if (!PageCompound(page))
		return 0;

	page = compound_head(page);
1337
	return page[1].compound_dtor == HUGETLB_PAGE_DTOR;
1338
}
1339 1340
EXPORT_SYMBOL_GPL(PageHuge);

1341 1342 1343 1344 1345 1346 1347 1348 1349
/*
 * 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;

1350
	return get_compound_page_dtor(page_head) == free_huge_page;
1351 1352
}

1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452
/*
 * Find address_space associated with hugetlbfs page.
 * Upon entry page is locked and page 'was' mapped although mapped state
 * could change.  If necessary, use anon_vma to find vma and associated
 * address space.  The returned mapping may be stale, but it can not be
 * invalid as page lock (which is held) is required to destroy mapping.
 */
static struct address_space *_get_hugetlb_page_mapping(struct page *hpage)
{
	struct anon_vma *anon_vma;
	pgoff_t pgoff_start, pgoff_end;
	struct anon_vma_chain *avc;
	struct address_space *mapping = page_mapping(hpage);

	/* Simple file based mapping */
	if (mapping)
		return mapping;

	/*
	 * Even anonymous hugetlbfs mappings are associated with an
	 * underlying hugetlbfs file (see hugetlb_file_setup in mmap
	 * code).  Find a vma associated with the anonymous vma, and
	 * use the file pointer to get address_space.
	 */
	anon_vma = page_lock_anon_vma_read(hpage);
	if (!anon_vma)
		return mapping;  /* NULL */

	/* Use first found vma */
	pgoff_start = page_to_pgoff(hpage);
	pgoff_end = pgoff_start + hpage_nr_pages(hpage) - 1;
	anon_vma_interval_tree_foreach(avc, &anon_vma->rb_root,
					pgoff_start, pgoff_end) {
		struct vm_area_struct *vma = avc->vma;

		mapping = vma->vm_file->f_mapping;
		break;
	}

	anon_vma_unlock_read(anon_vma);
	return mapping;
}

/*
 * Find and lock address space (mapping) in write mode.
 *
 * Upon entry, the page is locked which allows us to find the mapping
 * even in the case of an anon page.  However, locking order dictates
 * the i_mmap_rwsem be acquired BEFORE the page lock.  This is hugetlbfs
 * specific.  So, we first try to lock the sema while still holding the
 * page lock.  If this works, great!  If not, then we need to drop the
 * page lock and then acquire i_mmap_rwsem and reacquire page lock.  Of
 * course, need to revalidate state along the way.
 */
struct address_space *hugetlb_page_mapping_lock_write(struct page *hpage)
{
	struct address_space *mapping, *mapping2;

	mapping = _get_hugetlb_page_mapping(hpage);
retry:
	if (!mapping)
		return mapping;

	/*
	 * If no contention, take lock and return
	 */
	if (i_mmap_trylock_write(mapping))
		return mapping;

	/*
	 * Must drop page lock and wait on mapping sema.
	 * Note:  Once page lock is dropped, mapping could become invalid.
	 * As a hack, increase map count until we lock page again.
	 */
	atomic_inc(&hpage->_mapcount);
	unlock_page(hpage);
	i_mmap_lock_write(mapping);
	lock_page(hpage);
	atomic_add_negative(-1, &hpage->_mapcount);

	/* verify page is still mapped */
	if (!page_mapped(hpage)) {
		i_mmap_unlock_write(mapping);
		return NULL;
	}

	/*
	 * Get address space again and verify it is the same one
	 * we locked.  If not, drop lock and retry.
	 */
	mapping2 = _get_hugetlb_page_mapping(hpage);
	if (mapping2 != mapping) {
		i_mmap_unlock_write(mapping);
		mapping = mapping2;
		goto retry;
	}

	return mapping;
}

1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469
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;
}

1470
static struct page *alloc_buddy_huge_page(struct hstate *h,
1471 1472
		gfp_t gfp_mask, int nid, nodemask_t *nmask,
		nodemask_t *node_alloc_noretry)
L
Linus Torvalds 已提交
1473
{
1474
	int order = huge_page_order(h);
L
Linus Torvalds 已提交
1475
	struct page *page;
1476
	bool alloc_try_hard = true;
1477

1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489
	/*
	 * By default we always try hard to allocate the page with
	 * __GFP_RETRY_MAYFAIL flag.  However, if we are allocating pages in
	 * a loop (to adjust global huge page counts) and previous allocation
	 * failed, do not continue to try hard on the same node.  Use the
	 * node_alloc_noretry bitmap to manage this state information.
	 */
	if (node_alloc_noretry && node_isset(nid, *node_alloc_noretry))
		alloc_try_hard = false;
	gfp_mask |= __GFP_COMP|__GFP_NOWARN;
	if (alloc_try_hard)
		gfp_mask |= __GFP_RETRY_MAYFAIL;
1490 1491 1492 1493 1494 1495 1496
	if (nid == NUMA_NO_NODE)
		nid = numa_mem_id();
	page = __alloc_pages_nodemask(gfp_mask, order, nid, nmask);
	if (page)
		__count_vm_event(HTLB_BUDDY_PGALLOC);
	else
		__count_vm_event(HTLB_BUDDY_PGALLOC_FAIL);
1497

1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512 1513
	/*
	 * If we did not specify __GFP_RETRY_MAYFAIL, but still got a page this
	 * indicates an overall state change.  Clear bit so that we resume
	 * normal 'try hard' allocations.
	 */
	if (node_alloc_noretry && page && !alloc_try_hard)
		node_clear(nid, *node_alloc_noretry);

	/*
	 * If we tried hard to get a page but failed, set bit so that
	 * subsequent attempts will not try as hard until there is an
	 * overall state change.
	 */
	if (node_alloc_noretry && !page && alloc_try_hard)
		node_set(nid, *node_alloc_noretry);

1514 1515 1516
	return page;
}

1517 1518 1519 1520 1521
/*
 * Common helper to allocate a fresh hugetlb page. All specific allocators
 * should use this function to get new hugetlb pages
 */
static struct page *alloc_fresh_huge_page(struct hstate *h,
1522 1523
		gfp_t gfp_mask, int nid, nodemask_t *nmask,
		nodemask_t *node_alloc_noretry)
1524 1525 1526 1527 1528 1529 1530
{
	struct page *page;

	if (hstate_is_gigantic(h))
		page = alloc_gigantic_page(h, gfp_mask, nid, nmask);
	else
		page = alloc_buddy_huge_page(h, gfp_mask,
1531
				nid, nmask, node_alloc_noretry);
1532 1533 1534 1535 1536 1537 1538 1539 1540 1541
	if (!page)
		return NULL;

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

	return page;
}

1542 1543 1544 1545
/*
 * Allocates a fresh page to the hugetlb allocator pool in the node interleaved
 * manner.
 */
1546 1547
static int alloc_pool_huge_page(struct hstate *h, nodemask_t *nodes_allowed,
				nodemask_t *node_alloc_noretry)
1548 1549 1550
{
	struct page *page;
	int nr_nodes, node;
1551
	gfp_t gfp_mask = htlb_alloc_mask(h) | __GFP_THISNODE;
1552 1553

	for_each_node_mask_to_alloc(h, nr_nodes, node, nodes_allowed) {
1554 1555
		page = alloc_fresh_huge_page(h, gfp_mask, node, nodes_allowed,
						node_alloc_noretry);
1556
		if (page)
1557 1558 1559
			break;
	}

1560 1561
	if (!page)
		return 0;
1562

1563 1564 1565
	put_page(page); /* free it into the hugepage allocator */

	return 1;
1566 1567
}

1568 1569 1570 1571 1572 1573
/*
 * 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.
 */
1574 1575
static int free_pool_huge_page(struct hstate *h, nodemask_t *nodes_allowed,
							 bool acct_surplus)
1576
{
1577
	int nr_nodes, node;
1578 1579
	int ret = 0;

1580
	for_each_node_mask_to_free(h, nr_nodes, node, nodes_allowed) {
1581 1582 1583 1584
		/*
		 * If we're returning unused surplus pages, only examine
		 * nodes with surplus pages.
		 */
1585 1586
		if ((!acct_surplus || h->surplus_huge_pages_node[node]) &&
		    !list_empty(&h->hugepage_freelists[node])) {
1587
			struct page *page =
1588
				list_entry(h->hugepage_freelists[node].next,
1589 1590 1591
					  struct page, lru);
			list_del(&page->lru);
			h->free_huge_pages--;
1592
			h->free_huge_pages_node[node]--;
1593 1594
			if (acct_surplus) {
				h->surplus_huge_pages--;
1595
				h->surplus_huge_pages_node[node]--;
1596
			}
1597 1598
			update_and_free_page(h, page);
			ret = 1;
1599
			break;
1600
		}
1601
	}
1602 1603 1604 1605

	return ret;
}

1606 1607
/*
 * Dissolve a given free hugepage into free buddy pages. This function does
1608 1609 1610 1611 1612 1613 1614
 * nothing for in-use hugepages and non-hugepages.
 * This function returns values like below:
 *
 *  -EBUSY: failed to dissolved free hugepages or the hugepage is in-use
 *          (allocated or reserved.)
 *       0: successfully dissolved free hugepages or the page is not a
 *          hugepage (considered as already dissolved)
1615
 */
1616
int dissolve_free_huge_page(struct page *page)
1617
{
1618
	int rc = -EBUSY;
1619

1620 1621 1622 1623
	/* Not to disrupt normal path by vainly holding hugetlb_lock */
	if (!PageHuge(page))
		return 0;

1624
	spin_lock(&hugetlb_lock);
1625 1626 1627 1628 1629 1630
	if (!PageHuge(page)) {
		rc = 0;
		goto out;
	}

	if (!page_count(page)) {
1631 1632 1633
		struct page *head = compound_head(page);
		struct hstate *h = page_hstate(head);
		int nid = page_to_nid(head);
1634
		if (h->free_huge_pages - h->resv_huge_pages == 0)
1635
			goto out;
1636 1637 1638 1639 1640 1641 1642 1643
		/*
		 * 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);
		}
1644
		list_del(&head->lru);
1645 1646
		h->free_huge_pages--;
		h->free_huge_pages_node[nid]--;
1647
		h->max_huge_pages--;
1648
		update_and_free_page(h, head);
1649
		rc = 0;
1650
	}
1651
out:
1652
	spin_unlock(&hugetlb_lock);
1653
	return rc;
1654 1655 1656 1657 1658
}

/*
 * Dissolve free hugepages in a given pfn range. Used by memory hotplug to
 * make specified memory blocks removable from the system.
1659 1660
 * Note that this will dissolve a free gigantic hugepage completely, if any
 * part of it lies within the given range.
1661 1662
 * Also note that if dissolve_free_huge_page() returns with an error, all
 * free hugepages that were dissolved before that error are lost.
1663
 */
1664
int dissolve_free_huge_pages(unsigned long start_pfn, unsigned long end_pfn)
1665 1666
{
	unsigned long pfn;
1667
	struct page *page;
1668
	int rc = 0;
1669

1670
	if (!hugepages_supported())
1671
		return rc;
1672

1673 1674
	for (pfn = start_pfn; pfn < end_pfn; pfn += 1 << minimum_order) {
		page = pfn_to_page(pfn);
1675 1676 1677
		rc = dissolve_free_huge_page(page);
		if (rc)
			break;
1678
	}
1679 1680

	return rc;
1681 1682
}

1683 1684 1685
/*
 * Allocates a fresh surplus page from the page allocator.
 */
1686
static struct page *alloc_surplus_huge_page(struct hstate *h, gfp_t gfp_mask,
1687
		int nid, nodemask_t *nmask)
1688
{
1689
	struct page *page = NULL;
1690

1691
	if (hstate_is_gigantic(h))
1692 1693
		return NULL;

1694
	spin_lock(&hugetlb_lock);
1695 1696
	if (h->surplus_huge_pages >= h->nr_overcommit_huge_pages)
		goto out_unlock;
1697 1698
	spin_unlock(&hugetlb_lock);

1699
	page = alloc_fresh_huge_page(h, gfp_mask, nid, nmask, NULL);
1700
	if (!page)
1701
		return NULL;
1702 1703

	spin_lock(&hugetlb_lock);
1704 1705 1706 1707 1708 1709 1710 1711 1712
	/*
	 * We could have raced with the pool size change.
	 * Double check that and simply deallocate the new page
	 * if we would end up overcommiting the surpluses. Abuse
	 * temporary page to workaround the nasty free_huge_page
	 * codeflow
	 */
	if (h->surplus_huge_pages >= h->nr_overcommit_huge_pages) {
		SetPageHugeTemporary(page);
1713
		spin_unlock(&hugetlb_lock);
1714
		put_page(page);
1715
		return NULL;
1716 1717
	} else {
		h->surplus_huge_pages++;
1718
		h->surplus_huge_pages_node[page_to_nid(page)]++;
1719
	}
1720 1721

out_unlock:
1722
	spin_unlock(&hugetlb_lock);
1723 1724 1725 1726

	return page;
}

1727 1728
struct page *alloc_migrate_huge_page(struct hstate *h, gfp_t gfp_mask,
				     int nid, nodemask_t *nmask)
1729 1730 1731 1732 1733 1734
{
	struct page *page;

	if (hstate_is_gigantic(h))
		return NULL;

1735
	page = alloc_fresh_huge_page(h, gfp_mask, nid, nmask, NULL);
1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746 1747
	if (!page)
		return NULL;

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

	return page;
}

1748 1749 1750
/*
 * Use the VMA's mpolicy to allocate a huge page from the buddy.
 */
D
Dave Hansen 已提交
1751
static
1752
struct page *alloc_buddy_huge_page_with_mpol(struct hstate *h,
1753 1754
		struct vm_area_struct *vma, unsigned long addr)
{
1755 1756 1757 1758 1759 1760 1761
	struct page *page;
	struct mempolicy *mpol;
	gfp_t gfp_mask = htlb_alloc_mask(h);
	int nid;
	nodemask_t *nodemask;

	nid = huge_node(vma, addr, gfp_mask, &mpol, &nodemask);
1762
	page = alloc_surplus_huge_page(h, gfp_mask, nid, nodemask);
1763 1764 1765
	mpol_cond_put(mpol);

	return page;
1766 1767
}

1768
/* page migration callback function */
1769 1770
struct page *alloc_huge_page_node(struct hstate *h, int nid)
{
1771
	gfp_t gfp_mask = htlb_alloc_mask(h);
1772
	struct page *page = NULL;
1773

1774 1775 1776
	if (nid != NUMA_NO_NODE)
		gfp_mask |= __GFP_THISNODE;

1777
	spin_lock(&hugetlb_lock);
1778
	if (h->free_huge_pages - h->resv_huge_pages > 0)
1779
		page = dequeue_huge_page_nodemask(h, gfp_mask, nid, NULL);
1780 1781
	spin_unlock(&hugetlb_lock);

1782
	if (!page)
1783
		page = alloc_migrate_huge_page(h, gfp_mask, nid, NULL);
1784 1785 1786 1787

	return page;
}

1788
/* page migration callback function */
1789 1790
struct page *alloc_huge_page_nodemask(struct hstate *h, int preferred_nid,
		nodemask_t *nmask)
1791
{
1792
	gfp_t gfp_mask = htlb_alloc_mask(h);
1793 1794 1795

	spin_lock(&hugetlb_lock);
	if (h->free_huge_pages - h->resv_huge_pages > 0) {
1796 1797 1798 1799 1800 1801
		struct page *page;

		page = dequeue_huge_page_nodemask(h, gfp_mask, preferred_nid, nmask);
		if (page) {
			spin_unlock(&hugetlb_lock);
			return page;
1802 1803 1804 1805
		}
	}
	spin_unlock(&hugetlb_lock);

1806
	return alloc_migrate_huge_page(h, gfp_mask, preferred_nid, nmask);
1807 1808
}

1809
/* mempolicy aware migration callback */
1810 1811
struct page *alloc_huge_page_vma(struct hstate *h, struct vm_area_struct *vma,
		unsigned long address)
1812 1813 1814 1815 1816 1817 1818 1819 1820 1821 1822 1823 1824 1825 1826
{
	struct mempolicy *mpol;
	nodemask_t *nodemask;
	struct page *page;
	gfp_t gfp_mask;
	int node;

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

	return page;
}

1827
/*
L
Lucas De Marchi 已提交
1828
 * Increase the hugetlb pool such that it can accommodate a reservation
1829 1830
 * of size 'delta'.
 */
1831
static int gather_surplus_pages(struct hstate *h, int delta)
1832 1833 1834 1835 1836
{
	struct list_head surplus_list;
	struct page *page, *tmp;
	int ret, i;
	int needed, allocated;
1837
	bool alloc_ok = true;
1838

1839
	needed = (h->resv_huge_pages + delta) - h->free_huge_pages;
1840
	if (needed <= 0) {
1841
		h->resv_huge_pages += delta;
1842
		return 0;
1843
	}
1844 1845 1846 1847 1848 1849 1850 1851

	allocated = 0;
	INIT_LIST_HEAD(&surplus_list);

	ret = -ENOMEM;
retry:
	spin_unlock(&hugetlb_lock);
	for (i = 0; i < needed; i++) {
1852
		page = alloc_surplus_huge_page(h, htlb_alloc_mask(h),
1853
				NUMA_NO_NODE, NULL);
1854 1855 1856 1857
		if (!page) {
			alloc_ok = false;
			break;
		}
1858
		list_add(&page->lru, &surplus_list);
1859
		cond_resched();
1860
	}
1861
	allocated += i;
1862 1863 1864 1865 1866 1867

	/*
	 * 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);
1868 1869
	needed = (h->resv_huge_pages + delta) -
			(h->free_huge_pages + allocated);
1870 1871 1872 1873 1874 1875 1876 1877 1878 1879
	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;
	}
1880 1881
	/*
	 * The surplus_list now contains _at_least_ the number of extra pages
L
Lucas De Marchi 已提交
1882
	 * needed to accommodate the reservation.  Add the appropriate number
1883
	 * of pages to the hugetlb pool and free the extras back to the buddy
1884 1885 1886
	 * 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.
1887 1888
	 */
	needed += allocated;
1889
	h->resv_huge_pages += delta;
1890
	ret = 0;
1891

1892
	/* Free the needed pages to the hugetlb pool */
1893
	list_for_each_entry_safe(page, tmp, &surplus_list, lru) {
1894 1895
		if ((--needed) < 0)
			break;
1896 1897 1898 1899 1900
		/*
		 * This page is now managed by the hugetlb allocator and has
		 * no users -- drop the buddy allocator's reference.
		 */
		put_page_testzero(page);
1901
		VM_BUG_ON_PAGE(page_count(page), page);
1902
		enqueue_huge_page(h, page);
1903
	}
1904
free:
1905
	spin_unlock(&hugetlb_lock);
1906 1907

	/* Free unnecessary surplus pages to the buddy allocator */
1908 1909
	list_for_each_entry_safe(page, tmp, &surplus_list, lru)
		put_page(page);
1910
	spin_lock(&hugetlb_lock);
1911 1912 1913 1914 1915

	return ret;
}

/*
1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927
 * 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.
1928
 */
1929 1930
static void return_unused_surplus_pages(struct hstate *h,
					unsigned long unused_resv_pages)
1931 1932 1933
{
	unsigned long nr_pages;

1934
	/* Cannot return gigantic pages currently */
1935
	if (hstate_is_gigantic(h))
1936
		goto out;
1937

1938 1939 1940 1941
	/*
	 * Part (or even all) of the reservation could have been backed
	 * by pre-allocated pages. Only free surplus pages.
	 */
1942
	nr_pages = min(unused_resv_pages, h->surplus_huge_pages);
1943

1944 1945
	/*
	 * We want to release as many surplus pages as possible, spread
1946 1947 1948 1949 1950
	 * 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.
1951 1952 1953 1954
	 *
	 * 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.
1955 1956
	 */
	while (nr_pages--) {
1957 1958
		h->resv_huge_pages--;
		unused_resv_pages--;
1959
		if (!free_pool_huge_page(h, &node_states[N_MEMORY], 1))
1960
			goto out;
1961
		cond_resched_lock(&hugetlb_lock);
1962
	}
1963 1964 1965 1966

out:
	/* Fully uncommit the reservation */
	h->resv_huge_pages -= unused_resv_pages;
1967 1968
}

1969

1970
/*
1971
 * vma_needs_reservation, vma_commit_reservation and vma_end_reservation
1972
 * are used by the huge page allocation routines to manage reservations.
1973 1974 1975 1976 1977 1978
 *
 * 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
1979 1980 1981
 * 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.
1982 1983 1984 1985 1986 1987
 *
 * 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.
1988 1989 1990 1991 1992
 *
 * 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.
1993
 */
1994 1995 1996
enum vma_resv_mode {
	VMA_NEEDS_RESV,
	VMA_COMMIT_RESV,
1997
	VMA_END_RESV,
1998
	VMA_ADD_RESV,
1999
};
2000 2001
static long __vma_reservation_common(struct hstate *h,
				struct vm_area_struct *vma, unsigned long addr,
2002
				enum vma_resv_mode mode)
2003
{
2004 2005
	struct resv_map *resv;
	pgoff_t idx;
2006
	long ret;
2007

2008 2009
	resv = vma_resv_map(vma);
	if (!resv)
2010
		return 1;
2011

2012
	idx = vma_hugecache_offset(h, vma, addr);
2013 2014
	switch (mode) {
	case VMA_NEEDS_RESV:
2015
		ret = region_chg(resv, idx, idx + 1);
2016 2017 2018 2019
		break;
	case VMA_COMMIT_RESV:
		ret = region_add(resv, idx, idx + 1);
		break;
2020
	case VMA_END_RESV:
2021 2022 2023
		region_abort(resv, idx, idx + 1);
		ret = 0;
		break;
2024 2025 2026 2027 2028 2029 2030 2031
	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;
2032 2033 2034
	default:
		BUG();
	}
2035

2036
	if (vma->vm_flags & VM_MAYSHARE)
2037
		return ret;
2038 2039 2040 2041 2042 2043 2044 2045 2046 2047 2048 2049 2050 2051 2052 2053 2054 2055 2056
	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;
	}
2057
	else
2058
		return ret < 0 ? ret : 0;
2059
}
2060 2061

static long vma_needs_reservation(struct hstate *h,
2062
			struct vm_area_struct *vma, unsigned long addr)
2063
{
2064
	return __vma_reservation_common(h, vma, addr, VMA_NEEDS_RESV);
2065
}
2066

2067 2068 2069
static long vma_commit_reservation(struct hstate *h,
			struct vm_area_struct *vma, unsigned long addr)
{
2070 2071 2072
	return __vma_reservation_common(h, vma, addr, VMA_COMMIT_RESV);
}

2073
static void vma_end_reservation(struct hstate *h,
2074 2075
			struct vm_area_struct *vma, unsigned long addr)
{
2076
	(void)__vma_reservation_common(h, vma, addr, VMA_END_RESV);
2077 2078
}

2079 2080 2081 2082 2083 2084 2085 2086 2087 2088 2089 2090 2091 2092 2093 2094 2095 2096 2097 2098 2099 2100 2101 2102 2103 2104 2105 2106 2107 2108 2109 2110 2111 2112 2113 2114 2115 2116 2117 2118 2119 2120 2121 2122 2123 2124 2125 2126 2127 2128
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);
	}
}

2129
struct page *alloc_huge_page(struct vm_area_struct *vma,
2130
				    unsigned long addr, int avoid_reserve)
L
Linus Torvalds 已提交
2131
{
2132
	struct hugepage_subpool *spool = subpool_vma(vma);
2133
	struct hstate *h = hstate_vma(vma);
2134
	struct page *page;
2135 2136
	long map_chg, map_commit;
	long gbl_chg;
2137 2138
	int ret, idx;
	struct hugetlb_cgroup *h_cg;
2139

2140
	idx = hstate_index(h);
2141
	/*
2142 2143 2144
	 * 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).
2145
	 */
2146 2147
	map_chg = gbl_chg = vma_needs_reservation(h, vma, addr);
	if (map_chg < 0)
2148
		return ERR_PTR(-ENOMEM);
2149 2150 2151 2152 2153 2154 2155 2156 2157 2158 2159

	/*
	 * 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) {
2160
			vma_end_reservation(h, vma, addr);
2161
			return ERR_PTR(-ENOSPC);
2162
		}
L
Linus Torvalds 已提交
2163

2164 2165 2166 2167 2168 2169 2170 2171 2172 2173 2174 2175
		/*
		 * 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;
	}

2176
	ret = hugetlb_cgroup_charge_cgroup(idx, pages_per_huge_page(h), &h_cg);
2177 2178 2179
	if (ret)
		goto out_subpool_put;

L
Linus Torvalds 已提交
2180
	spin_lock(&hugetlb_lock);
2181 2182 2183 2184 2185 2186
	/*
	 * 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);
2187
	if (!page) {
2188
		spin_unlock(&hugetlb_lock);
2189
		page = alloc_buddy_huge_page_with_mpol(h, vma, addr);
2190 2191
		if (!page)
			goto out_uncharge_cgroup;
2192 2193 2194 2195
		if (!avoid_reserve && vma_has_reserves(vma, gbl_chg)) {
			SetPagePrivate(page);
			h->resv_huge_pages--;
		}
2196 2197
		spin_lock(&hugetlb_lock);
		list_move(&page->lru, &h->hugepage_activelist);
2198
		/* Fall through */
K
Ken Chen 已提交
2199
	}
2200 2201
	hugetlb_cgroup_commit_charge(idx, pages_per_huge_page(h), h_cg, page);
	spin_unlock(&hugetlb_lock);
2202

2203
	set_page_private(page, (unsigned long)spool);
2204

2205 2206
	map_commit = vma_commit_reservation(h, vma, addr);
	if (unlikely(map_chg > map_commit)) {
2207 2208 2209 2210 2211 2212 2213 2214 2215 2216 2217 2218 2219 2220
		/*
		 * 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);
	}
2221
	return page;
2222 2223 2224 2225

out_uncharge_cgroup:
	hugetlb_cgroup_uncharge_cgroup(idx, pages_per_huge_page(h), h_cg);
out_subpool_put:
2226
	if (map_chg || avoid_reserve)
2227
		hugepage_subpool_put_pages(spool, 1);
2228
	vma_end_reservation(h, vma, addr);
2229
	return ERR_PTR(-ENOSPC);
2230 2231
}

2232 2233 2234
int alloc_bootmem_huge_page(struct hstate *h)
	__attribute__ ((weak, alias("__alloc_bootmem_huge_page")));
int __alloc_bootmem_huge_page(struct hstate *h)
2235 2236
{
	struct huge_bootmem_page *m;
2237
	int nr_nodes, node;
2238

2239
	for_each_node_mask_to_alloc(h, nr_nodes, node, &node_states[N_MEMORY]) {
2240 2241
		void *addr;

2242
		addr = memblock_alloc_try_nid_raw(
2243
				huge_page_size(h), huge_page_size(h),
2244
				0, MEMBLOCK_ALLOC_ACCESSIBLE, node);
2245 2246 2247 2248 2249 2250 2251
		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;
2252
			goto found;
2253 2254 2255 2256 2257
		}
	}
	return 0;

found:
2258
	BUG_ON(!IS_ALIGNED(virt_to_phys(m), huge_page_size(h)));
2259
	/* Put them into a private list first because mem_map is not up yet */
2260
	INIT_LIST_HEAD(&m->list);
2261 2262 2263 2264 2265
	list_add(&m->list, &huge_boot_pages);
	m->hstate = h;
	return 1;
}

2266 2267
static void __init prep_compound_huge_page(struct page *page,
		unsigned int order)
2268 2269 2270 2271 2272 2273 2274
{
	if (unlikely(order > (MAX_ORDER - 1)))
		prep_compound_gigantic_page(page, order);
	else
		prep_compound_page(page, order);
}

2275 2276 2277 2278 2279 2280
/* 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) {
2281
		struct page *page = virt_to_page(m);
2282
		struct hstate *h = m->hstate;
2283

2284
		WARN_ON(page_count(page) != 1);
2285
		prep_compound_huge_page(page, h->order);
2286
		WARN_ON(PageReserved(page));
2287
		prep_new_huge_page(h, page, page_to_nid(page));
2288 2289
		put_page(page); /* free it into the hugepage allocator */

2290 2291 2292 2293 2294 2295
		/*
		 * 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.
		 */
2296
		if (hstate_is_gigantic(h))
2297
			adjust_managed_page_count(page, 1 << h->order);
2298
		cond_resched();
2299 2300 2301
	}
}

2302
static void __init hugetlb_hstate_alloc_pages(struct hstate *h)
L
Linus Torvalds 已提交
2303 2304
{
	unsigned long i;
2305 2306 2307 2308 2309 2310 2311 2312 2313 2314 2315 2316 2317 2318 2319 2320 2321 2322 2323
	nodemask_t *node_alloc_noretry;

	if (!hstate_is_gigantic(h)) {
		/*
		 * Bit mask controlling how hard we retry per-node allocations.
		 * Ignore errors as lower level routines can deal with
		 * node_alloc_noretry == NULL.  If this kmalloc fails at boot
		 * time, we are likely in bigger trouble.
		 */
		node_alloc_noretry = kmalloc(sizeof(*node_alloc_noretry),
						GFP_KERNEL);
	} else {
		/* allocations done at boot time */
		node_alloc_noretry = NULL;
	}

	/* bit mask controlling how hard we retry per-node allocations */
	if (node_alloc_noretry)
		nodes_clear(*node_alloc_noretry);
2324

2325
	for (i = 0; i < h->max_huge_pages; ++i) {
2326
		if (hstate_is_gigantic(h)) {
2327 2328
			if (!alloc_bootmem_huge_page(h))
				break;
2329
		} else if (!alloc_pool_huge_page(h,
2330 2331
					 &node_states[N_MEMORY],
					 node_alloc_noretry))
L
Linus Torvalds 已提交
2332
			break;
2333
		cond_resched();
L
Linus Torvalds 已提交
2334
	}
2335 2336 2337
	if (i < h->max_huge_pages) {
		char buf[32];

2338
		string_get_size(huge_page_size(h), 1, STRING_UNITS_2, buf, 32);
2339 2340 2341 2342
		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;
	}
2343 2344

	kfree(node_alloc_noretry);
2345 2346 2347 2348 2349 2350 2351
}

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

	for_each_hstate(h) {
2352 2353 2354
		if (minimum_order > huge_page_order(h))
			minimum_order = huge_page_order(h);

2355
		/* oversize hugepages were init'ed in early boot */
2356
		if (!hstate_is_gigantic(h))
2357
			hugetlb_hstate_alloc_pages(h);
2358
	}
2359
	VM_BUG_ON(minimum_order == UINT_MAX);
2360 2361 2362 2363 2364 2365 2366
}

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

	for_each_hstate(h) {
A
Andi Kleen 已提交
2367
		char buf[32];
2368 2369

		string_get_size(huge_page_size(h), 1, STRING_UNITS_2, buf, 32);
2370
		pr_info("HugeTLB registered %s page size, pre-allocated %ld pages\n",
2371
			buf, h->free_huge_pages);
2372 2373 2374
	}
}

L
Linus Torvalds 已提交
2375
#ifdef CONFIG_HIGHMEM
2376 2377
static void try_to_free_low(struct hstate *h, unsigned long count,
						nodemask_t *nodes_allowed)
L
Linus Torvalds 已提交
2378
{
2379 2380
	int i;

2381
	if (hstate_is_gigantic(h))
2382 2383
		return;

2384
	for_each_node_mask(i, *nodes_allowed) {
L
Linus Torvalds 已提交
2385
		struct page *page, *next;
2386 2387 2388
		struct list_head *freel = &h->hugepage_freelists[i];
		list_for_each_entry_safe(page, next, freel, lru) {
			if (count >= h->nr_huge_pages)
2389
				return;
L
Linus Torvalds 已提交
2390 2391 2392
			if (PageHighMem(page))
				continue;
			list_del(&page->lru);
2393
			update_and_free_page(h, page);
2394 2395
			h->free_huge_pages--;
			h->free_huge_pages_node[page_to_nid(page)]--;
L
Linus Torvalds 已提交
2396 2397 2398 2399
		}
	}
}
#else
2400 2401
static inline void try_to_free_low(struct hstate *h, unsigned long count,
						nodemask_t *nodes_allowed)
L
Linus Torvalds 已提交
2402 2403 2404 2405
{
}
#endif

2406 2407 2408 2409 2410
/*
 * 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.
 */
2411 2412
static int adjust_pool_surplus(struct hstate *h, nodemask_t *nodes_allowed,
				int delta)
2413
{
2414
	int nr_nodes, node;
2415 2416 2417

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

2418 2419 2420 2421
	if (delta < 0) {
		for_each_node_mask_to_alloc(h, nr_nodes, node, nodes_allowed) {
			if (h->surplus_huge_pages_node[node])
				goto found;
2422
		}
2423 2424 2425 2426 2427
	} 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;
2428
		}
2429 2430
	}
	return 0;
2431

2432 2433 2434 2435
found:
	h->surplus_huge_pages += delta;
	h->surplus_huge_pages_node[node] += delta;
	return 1;
2436 2437
}

2438
#define persistent_huge_pages(h) (h->nr_huge_pages - h->surplus_huge_pages)
2439
static int set_max_huge_pages(struct hstate *h, unsigned long count, int nid,
2440
			      nodemask_t *nodes_allowed)
L
Linus Torvalds 已提交
2441
{
2442
	unsigned long min_count, ret;
2443 2444 2445 2446 2447 2448 2449 2450 2451 2452 2453
	NODEMASK_ALLOC(nodemask_t, node_alloc_noretry, GFP_KERNEL);

	/*
	 * Bit mask controlling how hard we retry per-node allocations.
	 * If we can not allocate the bit mask, do not attempt to allocate
	 * the requested huge pages.
	 */
	if (node_alloc_noretry)
		nodes_clear(*node_alloc_noretry);
	else
		return -ENOMEM;
L
Linus Torvalds 已提交
2454

2455 2456
	spin_lock(&hugetlb_lock);

2457 2458 2459 2460 2461 2462 2463 2464 2465 2466 2467 2468 2469 2470 2471 2472 2473 2474 2475 2476
	/*
	 * Check for a node specific request.
	 * Changing node specific huge page count may require a corresponding
	 * change to the global count.  In any case, the passed node mask
	 * (nodes_allowed) will restrict alloc/free to the specified node.
	 */
	if (nid != NUMA_NO_NODE) {
		unsigned long old_count = count;

		count += h->nr_huge_pages - h->nr_huge_pages_node[nid];
		/*
		 * User may have specified a large count value which caused the
		 * above calculation to overflow.  In this case, they wanted
		 * to allocate as many huge pages as possible.  Set count to
		 * largest possible value to align with their intention.
		 */
		if (count < old_count)
			count = ULONG_MAX;
	}

2477 2478 2479 2480 2481 2482 2483 2484 2485 2486
	/*
	 * Gigantic pages runtime allocation depend on the capability for large
	 * page range allocation.
	 * If the system does not provide this feature, return an error when
	 * the user tries to allocate gigantic pages but let the user free the
	 * boottime allocated gigantic pages.
	 */
	if (hstate_is_gigantic(h) && !IS_ENABLED(CONFIG_CONTIG_ALLOC)) {
		if (count > persistent_huge_pages(h)) {
			spin_unlock(&hugetlb_lock);
2487
			NODEMASK_FREE(node_alloc_noretry);
2488 2489 2490 2491
			return -EINVAL;
		}
		/* Fall through to decrease pool */
	}
2492

2493 2494 2495 2496
	/*
	 * Increase the pool size
	 * First take pages out of surplus state.  Then make up the
	 * remaining difference by allocating fresh huge pages.
2497
	 *
2498
	 * We might race with alloc_surplus_huge_page() here and be unable
2499 2500 2501 2502
	 * 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.
2503
	 */
2504
	while (h->surplus_huge_pages && count > persistent_huge_pages(h)) {
2505
		if (!adjust_pool_surplus(h, nodes_allowed, -1))
2506 2507 2508
			break;
	}

2509
	while (count > persistent_huge_pages(h)) {
2510 2511 2512 2513 2514 2515
		/*
		 * 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);
2516 2517 2518 2519

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

2520 2521
		ret = alloc_pool_huge_page(h, nodes_allowed,
						node_alloc_noretry);
2522 2523 2524 2525
		spin_lock(&hugetlb_lock);
		if (!ret)
			goto out;

2526 2527 2528
		/* Bail for signals. Probably ctrl-c from user */
		if (signal_pending(current))
			goto out;
2529 2530 2531 2532 2533 2534 2535 2536
	}

	/*
	 * 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.
2537 2538 2539 2540
	 *
	 * 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
2541
	 * alloc_surplus_huge_page() is checking the global counter,
2542 2543 2544
	 * 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.
2545
	 */
2546
	min_count = h->resv_huge_pages + h->nr_huge_pages - h->free_huge_pages;
2547
	min_count = max(count, min_count);
2548
	try_to_free_low(h, min_count, nodes_allowed);
2549
	while (min_count < persistent_huge_pages(h)) {
2550
		if (!free_pool_huge_page(h, nodes_allowed, 0))
L
Linus Torvalds 已提交
2551
			break;
2552
		cond_resched_lock(&hugetlb_lock);
L
Linus Torvalds 已提交
2553
	}
2554
	while (count < persistent_huge_pages(h)) {
2555
		if (!adjust_pool_surplus(h, nodes_allowed, 1))
2556 2557 2558
			break;
	}
out:
2559
	h->max_huge_pages = persistent_huge_pages(h);
L
Linus Torvalds 已提交
2560
	spin_unlock(&hugetlb_lock);
2561

2562 2563
	NODEMASK_FREE(node_alloc_noretry);

2564
	return 0;
L
Linus Torvalds 已提交
2565 2566
}

2567 2568 2569 2570 2571 2572 2573 2574 2575 2576
#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];

2577 2578 2579
static struct hstate *kobj_to_node_hstate(struct kobject *kobj, int *nidp);

static struct hstate *kobj_to_hstate(struct kobject *kobj, int *nidp)
2580 2581
{
	int i;
2582

2583
	for (i = 0; i < HUGE_MAX_HSTATE; i++)
2584 2585 2586
		if (hstate_kobjs[i] == kobj) {
			if (nidp)
				*nidp = NUMA_NO_NODE;
2587
			return &hstates[i];
2588 2589 2590
		}

	return kobj_to_node_hstate(kobj, nidp);
2591 2592
}

2593
static ssize_t nr_hugepages_show_common(struct kobject *kobj,
2594 2595
					struct kobj_attribute *attr, char *buf)
{
2596 2597 2598 2599 2600 2601 2602 2603 2604 2605 2606
	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);
2607
}
2608

2609 2610 2611
static ssize_t __nr_hugepages_store_common(bool obey_mempolicy,
					   struct hstate *h, int nid,
					   unsigned long count, size_t len)
2612 2613
{
	int err;
2614
	nodemask_t nodes_allowed, *n_mask;
2615

2616 2617
	if (hstate_is_gigantic(h) && !gigantic_page_runtime_supported())
		return -EINVAL;
2618

2619 2620 2621 2622 2623
	if (nid == NUMA_NO_NODE) {
		/*
		 * global hstate attribute
		 */
		if (!(obey_mempolicy &&
2624 2625 2626 2627 2628
				init_nodemask_of_mempolicy(&nodes_allowed)))
			n_mask = &node_states[N_MEMORY];
		else
			n_mask = &nodes_allowed;
	} else {
2629
		/*
2630 2631
		 * Node specific request.  count adjustment happens in
		 * set_max_huge_pages() after acquiring hugetlb_lock.
2632
		 */
2633 2634
		init_nodemask_of_node(&nodes_allowed, nid);
		n_mask = &nodes_allowed;
2635
	}
2636

2637
	err = set_max_huge_pages(h, count, nid, n_mask);
2638

2639
	return err ? err : len;
2640 2641
}

2642 2643 2644 2645 2646 2647 2648 2649 2650 2651 2652 2653 2654 2655 2656 2657 2658
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);
}

2659 2660 2661 2662 2663 2664 2665 2666 2667
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)
{
2668
	return nr_hugepages_store_common(false, kobj, buf, len);
2669 2670 2671
}
HSTATE_ATTR(nr_hugepages);

2672 2673 2674 2675 2676 2677 2678 2679 2680 2681 2682 2683 2684 2685 2686
#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)
{
2687
	return nr_hugepages_store_common(true, kobj, buf, len);
2688 2689 2690 2691 2692
}
HSTATE_ATTR(nr_hugepages_mempolicy);
#endif


2693 2694 2695
static ssize_t nr_overcommit_hugepages_show(struct kobject *kobj,
					struct kobj_attribute *attr, char *buf)
{
2696
	struct hstate *h = kobj_to_hstate(kobj, NULL);
2697 2698
	return sprintf(buf, "%lu\n", h->nr_overcommit_huge_pages);
}
2699

2700 2701 2702 2703 2704
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;
2705
	struct hstate *h = kobj_to_hstate(kobj, NULL);
2706

2707
	if (hstate_is_gigantic(h))
2708 2709
		return -EINVAL;

2710
	err = kstrtoul(buf, 10, &input);
2711
	if (err)
2712
		return err;
2713 2714 2715 2716 2717 2718 2719 2720 2721 2722 2723 2724

	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)
{
2725 2726 2727 2728 2729 2730 2731 2732 2733 2734 2735
	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);
2736 2737 2738 2739 2740 2741
}
HSTATE_ATTR_RO(free_hugepages);

static ssize_t resv_hugepages_show(struct kobject *kobj,
					struct kobj_attribute *attr, char *buf)
{
2742
	struct hstate *h = kobj_to_hstate(kobj, NULL);
2743 2744 2745 2746 2747 2748 2749
	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)
{
2750 2751 2752 2753 2754 2755 2756 2757 2758 2759 2760
	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);
2761 2762 2763 2764 2765 2766 2767 2768 2769
}
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,
2770 2771 2772
#ifdef CONFIG_NUMA
	&nr_hugepages_mempolicy_attr.attr,
#endif
2773 2774 2775
	NULL,
};

2776
static const struct attribute_group hstate_attr_group = {
2777 2778 2779
	.attrs = hstate_attrs,
};

J
Jeff Mahoney 已提交
2780 2781
static int hugetlb_sysfs_add_hstate(struct hstate *h, struct kobject *parent,
				    struct kobject **hstate_kobjs,
2782
				    const struct attribute_group *hstate_attr_group)
2783 2784
{
	int retval;
2785
	int hi = hstate_index(h);
2786

2787 2788
	hstate_kobjs[hi] = kobject_create_and_add(h->name, parent);
	if (!hstate_kobjs[hi])
2789 2790
		return -ENOMEM;

2791
	retval = sysfs_create_group(hstate_kobjs[hi], hstate_attr_group);
2792
	if (retval)
2793
		kobject_put(hstate_kobjs[hi]);
2794 2795 2796 2797 2798 2799 2800 2801 2802 2803 2804 2805 2806 2807

	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) {
2808 2809
		err = hugetlb_sysfs_add_hstate(h, hugepages_kobj,
					 hstate_kobjs, &hstate_attr_group);
2810
		if (err)
2811
			pr_err("Hugetlb: Unable to add hstate %s", h->name);
2812 2813 2814
	}
}

2815 2816 2817 2818
#ifdef CONFIG_NUMA

/*
 * node_hstate/s - associate per node hstate attributes, via their kobjects,
2819 2820 2821
 * 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
2822 2823 2824 2825 2826 2827
 * the base kernel, on the hugetlb module.
 */
struct node_hstate {
	struct kobject		*hugepages_kobj;
	struct kobject		*hstate_kobjs[HUGE_MAX_HSTATE];
};
2828
static struct node_hstate node_hstates[MAX_NUMNODES];
2829 2830

/*
2831
 * A subset of global hstate attributes for node devices
2832 2833 2834 2835 2836 2837 2838 2839
 */
static struct attribute *per_node_hstate_attrs[] = {
	&nr_hugepages_attr.attr,
	&free_hugepages_attr.attr,
	&surplus_hugepages_attr.attr,
	NULL,
};

2840
static const struct attribute_group per_node_hstate_attr_group = {
2841 2842 2843 2844
	.attrs = per_node_hstate_attrs,
};

/*
2845
 * kobj_to_node_hstate - lookup global hstate for node device hstate attr kobj.
2846 2847 2848 2849 2850 2851 2852 2853 2854 2855 2856 2857 2858 2859 2860 2861 2862 2863 2864 2865 2866 2867
 * 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;
}

/*
2868
 * Unregister hstate attributes from a single node device.
2869 2870
 * No-op if no hstate attributes attached.
 */
2871
static void hugetlb_unregister_node(struct node *node)
2872 2873
{
	struct hstate *h;
2874
	struct node_hstate *nhs = &node_hstates[node->dev.id];
2875 2876

	if (!nhs->hugepages_kobj)
2877
		return;		/* no hstate attributes */
2878

2879 2880 2881 2882 2883
	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;
2884
		}
2885
	}
2886 2887 2888 2889 2890 2891 2892

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


/*
2893
 * Register hstate attributes for a single node device.
2894 2895
 * No-op if attributes already registered.
 */
2896
static void hugetlb_register_node(struct node *node)
2897 2898
{
	struct hstate *h;
2899
	struct node_hstate *nhs = &node_hstates[node->dev.id];
2900 2901 2902 2903 2904 2905
	int err;

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

	nhs->hugepages_kobj = kobject_create_and_add("hugepages",
2906
							&node->dev.kobj);
2907 2908 2909 2910 2911 2912 2913 2914
	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) {
2915 2916
			pr_err("Hugetlb: Unable to add hstate %s for node %d\n",
				h->name, node->dev.id);
2917 2918 2919 2920 2921 2922 2923
			hugetlb_unregister_node(node);
			break;
		}
	}
}

/*
2924
 * hugetlb init time:  register hstate attributes for all registered node
2925 2926
 * devices of nodes that have memory.  All on-line nodes should have
 * registered their associated device by this time.
2927
 */
2928
static void __init hugetlb_register_all_nodes(void)
2929 2930 2931
{
	int nid;

2932
	for_each_node_state(nid, N_MEMORY) {
2933
		struct node *node = node_devices[nid];
2934
		if (node->dev.id == nid)
2935 2936 2937 2938
			hugetlb_register_node(node);
	}

	/*
2939
	 * Let the node device driver know we're here so it can
2940 2941 2942 2943 2944 2945 2946 2947 2948 2949 2950 2951 2952 2953 2954 2955 2956 2957 2958
	 * [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

2959 2960
static int __init hugetlb_init(void)
{
2961 2962
	int i;

2963
	if (!hugepages_supported())
2964
		return 0;
2965

2966
	if (!size_to_hstate(default_hstate_size)) {
2967 2968 2969 2970 2971
		if (default_hstate_size != 0) {
			pr_err("HugeTLB: unsupported default_hugepagesz %lu. Reverting to %lu\n",
			       default_hstate_size, HPAGE_SIZE);
		}

2972 2973 2974
		default_hstate_size = HPAGE_SIZE;
		if (!size_to_hstate(default_hstate_size))
			hugetlb_add_hstate(HUGETLB_PAGE_ORDER);
2975
	}
2976
	default_hstate_idx = hstate_index(size_to_hstate(default_hstate_size));
2977 2978 2979 2980
	if (default_hstate_max_huge_pages) {
		if (!default_hstate.max_huge_pages)
			default_hstate.max_huge_pages = default_hstate_max_huge_pages;
	}
2981 2982

	hugetlb_init_hstates();
2983
	gather_bootmem_prealloc();
2984 2985 2986
	report_hugepages();

	hugetlb_sysfs_init();
2987
	hugetlb_register_all_nodes();
2988
	hugetlb_cgroup_file_init();
2989

2990 2991 2992 2993 2994
#ifdef CONFIG_SMP
	num_fault_mutexes = roundup_pow_of_two(8 * num_possible_cpus());
#else
	num_fault_mutexes = 1;
#endif
2995
	hugetlb_fault_mutex_table =
2996 2997
		kmalloc_array(num_fault_mutexes, sizeof(struct mutex),
			      GFP_KERNEL);
2998
	BUG_ON(!hugetlb_fault_mutex_table);
2999 3000

	for (i = 0; i < num_fault_mutexes; i++)
3001
		mutex_init(&hugetlb_fault_mutex_table[i]);
3002 3003
	return 0;
}
3004
subsys_initcall(hugetlb_init);
3005 3006

/* Should be called on processing a hugepagesz=... option */
3007 3008 3009 3010 3011
void __init hugetlb_bad_size(void)
{
	parsed_valid_hugepagesz = false;
}

3012
void __init hugetlb_add_hstate(unsigned int order)
3013 3014
{
	struct hstate *h;
3015 3016
	unsigned long i;

3017
	if (size_to_hstate(PAGE_SIZE << order)) {
J
Joe Perches 已提交
3018
		pr_warn("hugepagesz= specified twice, ignoring\n");
3019 3020
		return;
	}
3021
	BUG_ON(hugetlb_max_hstate >= HUGE_MAX_HSTATE);
3022
	BUG_ON(order == 0);
3023
	h = &hstates[hugetlb_max_hstate++];
3024 3025
	h->order = order;
	h->mask = ~((1ULL << (order + PAGE_SHIFT)) - 1);
3026 3027 3028 3029
	h->nr_huge_pages = 0;
	h->free_huge_pages = 0;
	for (i = 0; i < MAX_NUMNODES; ++i)
		INIT_LIST_HEAD(&h->hugepage_freelists[i]);
3030
	INIT_LIST_HEAD(&h->hugepage_activelist);
3031 3032
	h->next_nid_to_alloc = first_memory_node;
	h->next_nid_to_free = first_memory_node;
3033 3034
	snprintf(h->name, HSTATE_NAME_LEN, "hugepages-%lukB",
					huge_page_size(h)/1024);
3035

3036 3037 3038
	parsed_hstate = h;
}

3039
static int __init hugetlb_nrpages_setup(char *s)
3040 3041
{
	unsigned long *mhp;
3042
	static unsigned long *last_mhp;
3043

3044 3045 3046 3047 3048 3049
	if (!parsed_valid_hugepagesz) {
		pr_warn("hugepages = %s preceded by "
			"an unsupported hugepagesz, ignoring\n", s);
		parsed_valid_hugepagesz = true;
		return 1;
	}
3050
	/*
3051
	 * !hugetlb_max_hstate means we haven't parsed a hugepagesz= parameter yet,
3052 3053
	 * so this hugepages= parameter goes to the "default hstate".
	 */
3054
	else if (!hugetlb_max_hstate)
3055 3056 3057 3058
		mhp = &default_hstate_max_huge_pages;
	else
		mhp = &parsed_hstate->max_huge_pages;

3059
	if (mhp == last_mhp) {
J
Joe Perches 已提交
3060
		pr_warn("hugepages= specified twice without interleaving hugepagesz=, ignoring\n");
3061 3062 3063
		return 1;
	}

3064 3065 3066
	if (sscanf(s, "%lu", mhp) <= 0)
		*mhp = 0;

3067 3068 3069 3070 3071
	/*
	 * 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.
	 */
3072
	if (hugetlb_max_hstate && parsed_hstate->order >= MAX_ORDER)
3073 3074 3075 3076
		hugetlb_hstate_alloc_pages(parsed_hstate);

	last_mhp = mhp;

3077 3078
	return 1;
}
3079 3080 3081 3082 3083 3084 3085 3086
__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);
3087

3088 3089 3090 3091 3092 3093 3094 3095 3096 3097 3098 3099
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
3100 3101 3102
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
3103
{
3104
	struct hstate *h = &default_hstate;
3105
	unsigned long tmp = h->max_huge_pages;
3106
	int ret;
3107

3108
	if (!hugepages_supported())
3109
		return -EOPNOTSUPP;
3110

3111 3112
	table->data = &tmp;
	table->maxlen = sizeof(unsigned long);
3113 3114 3115
	ret = proc_doulongvec_minmax(table, write, buffer, length, ppos);
	if (ret)
		goto out;
3116

3117 3118 3119
	if (write)
		ret = __nr_hugepages_store_common(obey_mempolicy, h,
						  NUMA_NO_NODE, tmp, *length);
3120 3121
out:
	return ret;
L
Linus Torvalds 已提交
3122
}
3123

3124 3125 3126 3127 3128 3129 3130 3131 3132 3133 3134 3135 3136 3137 3138 3139 3140
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 */

3141
int hugetlb_overcommit_handler(struct ctl_table *table, int write,
3142
			void __user *buffer,
3143 3144
			size_t *length, loff_t *ppos)
{
3145
	struct hstate *h = &default_hstate;
3146
	unsigned long tmp;
3147
	int ret;
3148

3149
	if (!hugepages_supported())
3150
		return -EOPNOTSUPP;
3151

3152
	tmp = h->nr_overcommit_huge_pages;
3153

3154
	if (write && hstate_is_gigantic(h))
3155 3156
		return -EINVAL;

3157 3158
	table->data = &tmp;
	table->maxlen = sizeof(unsigned long);
3159 3160 3161
	ret = proc_doulongvec_minmax(table, write, buffer, length, ppos);
	if (ret)
		goto out;
3162 3163 3164 3165 3166 3167

	if (write) {
		spin_lock(&hugetlb_lock);
		h->nr_overcommit_huge_pages = tmp;
		spin_unlock(&hugetlb_lock);
	}
3168 3169
out:
	return ret;
3170 3171
}

L
Linus Torvalds 已提交
3172 3173
#endif /* CONFIG_SYSCTL */

3174
void hugetlb_report_meminfo(struct seq_file *m)
L
Linus Torvalds 已提交
3175
{
3176 3177 3178
	struct hstate *h;
	unsigned long total = 0;

3179 3180
	if (!hugepages_supported())
		return;
3181 3182 3183 3184 3185 3186 3187 3188 3189 3190 3191 3192 3193 3194 3195 3196 3197 3198 3199 3200 3201

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

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

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

	seq_printf(m, "Hugetlb:        %8lu kB\n", total / 1024);
L
Linus Torvalds 已提交
3202 3203 3204 3205
}

int hugetlb_report_node_meminfo(int nid, char *buf)
{
3206
	struct hstate *h = &default_hstate;
3207 3208
	if (!hugepages_supported())
		return 0;
L
Linus Torvalds 已提交
3209 3210
	return sprintf(buf,
		"Node %d HugePages_Total: %5u\n"
3211 3212
		"Node %d HugePages_Free:  %5u\n"
		"Node %d HugePages_Surp:  %5u\n",
3213 3214 3215
		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
3216 3217
}

3218 3219 3220 3221 3222
void hugetlb_show_meminfo(void)
{
	struct hstate *h;
	int nid;

3223 3224 3225
	if (!hugepages_supported())
		return;

3226 3227 3228 3229 3230 3231 3232 3233 3234 3235
	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));
}

3236 3237 3238 3239 3240 3241
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
3242 3243 3244
/* Return the number pages of memory we physically have, in PAGE_SIZE units. */
unsigned long hugetlb_total_pages(void)
{
3245 3246 3247 3248 3249 3250
	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
3251 3252
}

3253
static int hugetlb_acct_memory(struct hstate *h, long delta)
M
Mel Gorman 已提交
3254 3255 3256 3257 3258 3259 3260 3261 3262 3263 3264 3265 3266 3267 3268 3269 3270 3271 3272 3273 3274 3275
{
	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) {
3276
		if (gather_surplus_pages(h, delta) < 0)
M
Mel Gorman 已提交
3277 3278
			goto out;

3279 3280
		if (delta > cpuset_mems_nr(h->free_huge_pages_node)) {
			return_unused_surplus_pages(h, delta);
M
Mel Gorman 已提交
3281 3282 3283 3284 3285 3286
			goto out;
		}
	}

	ret = 0;
	if (delta < 0)
3287
		return_unused_surplus_pages(h, (unsigned long) -delta);
M
Mel Gorman 已提交
3288 3289 3290 3291 3292 3293

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

3294 3295
static void hugetlb_vm_op_open(struct vm_area_struct *vma)
{
3296
	struct resv_map *resv = vma_resv_map(vma);
3297 3298 3299 3300 3301

	/*
	 * 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
3302
	 * has a reference to the reservation map it cannot disappear until
3303 3304 3305
	 * after this open call completes.  It is therefore safe to take a
	 * new reference here without additional locking.
	 */
3306
	if (resv && is_vma_resv_set(vma, HPAGE_RESV_OWNER))
3307
		kref_get(&resv->refs);
3308 3309
}

3310 3311
static void hugetlb_vm_op_close(struct vm_area_struct *vma)
{
3312
	struct hstate *h = hstate_vma(vma);
3313
	struct resv_map *resv = vma_resv_map(vma);
3314
	struct hugepage_subpool *spool = subpool_vma(vma);
3315
	unsigned long reserve, start, end;
3316
	long gbl_reserve;
3317

3318 3319
	if (!resv || !is_vma_resv_set(vma, HPAGE_RESV_OWNER))
		return;
3320

3321 3322
	start = vma_hugecache_offset(h, vma, vma->vm_start);
	end = vma_hugecache_offset(h, vma, vma->vm_end);
3323

3324
	reserve = (end - start) - region_count(resv, start, end);
3325
	hugetlb_cgroup_uncharge_counter(resv, start, end);
3326
	if (reserve) {
3327 3328 3329 3330 3331 3332
		/*
		 * 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);
3333
	}
3334 3335

	kref_put(&resv->refs, resv_map_release);
3336 3337
}

3338 3339 3340 3341 3342 3343 3344
static int hugetlb_vm_op_split(struct vm_area_struct *vma, unsigned long addr)
{
	if (addr & ~(huge_page_mask(hstate_vma(vma))))
		return -EINVAL;
	return 0;
}

3345 3346 3347 3348 3349 3350 3351
static unsigned long hugetlb_vm_op_pagesize(struct vm_area_struct *vma)
{
	struct hstate *hstate = hstate_vma(vma);

	return 1UL << huge_page_shift(hstate);
}

L
Linus Torvalds 已提交
3352 3353 3354 3355 3356 3357
/*
 * 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.
 */
3358
static vm_fault_t hugetlb_vm_op_fault(struct vm_fault *vmf)
L
Linus Torvalds 已提交
3359 3360
{
	BUG();
N
Nick Piggin 已提交
3361
	return 0;
L
Linus Torvalds 已提交
3362 3363
}

3364 3365 3366 3367 3368 3369 3370
/*
 * When a new function is introduced to vm_operations_struct and added
 * to hugetlb_vm_ops, please consider adding the function to shm_vm_ops.
 * This is because under System V memory model, mappings created via
 * shmget/shmat with "huge page" specified are backed by hugetlbfs files,
 * their original vm_ops are overwritten with shm_vm_ops.
 */
3371
const struct vm_operations_struct hugetlb_vm_ops = {
N
Nick Piggin 已提交
3372
	.fault = hugetlb_vm_op_fault,
3373
	.open = hugetlb_vm_op_open,
3374
	.close = hugetlb_vm_op_close,
3375
	.split = hugetlb_vm_op_split,
3376
	.pagesize = hugetlb_vm_op_pagesize,
L
Linus Torvalds 已提交
3377 3378
};

3379 3380
static pte_t make_huge_pte(struct vm_area_struct *vma, struct page *page,
				int writable)
D
David Gibson 已提交
3381 3382 3383
{
	pte_t entry;

3384
	if (writable) {
3385 3386
		entry = huge_pte_mkwrite(huge_pte_mkdirty(mk_huge_pte(page,
					 vma->vm_page_prot)));
D
David Gibson 已提交
3387
	} else {
3388 3389
		entry = huge_pte_wrprotect(mk_huge_pte(page,
					   vma->vm_page_prot));
D
David Gibson 已提交
3390 3391 3392
	}
	entry = pte_mkyoung(entry);
	entry = pte_mkhuge(entry);
3393
	entry = arch_make_huge_pte(entry, vma, page, writable);
D
David Gibson 已提交
3394 3395 3396 3397

	return entry;
}

3398 3399 3400 3401 3402
static void set_huge_ptep_writable(struct vm_area_struct *vma,
				   unsigned long address, pte_t *ptep)
{
	pte_t entry;

3403
	entry = huge_pte_mkwrite(huge_pte_mkdirty(huge_ptep_get(ptep)));
3404
	if (huge_ptep_set_access_flags(vma, address, ptep, entry, 1))
3405
		update_mmu_cache(vma, address, ptep);
3406 3407
}

3408
bool is_hugetlb_entry_migration(pte_t pte)
3409 3410 3411 3412
{
	swp_entry_t swp;

	if (huge_pte_none(pte) || pte_present(pte))
3413
		return false;
3414 3415
	swp = pte_to_swp_entry(pte);
	if (non_swap_entry(swp) && is_migration_entry(swp))
3416
		return true;
3417
	else
3418
		return false;
3419 3420 3421 3422 3423 3424 3425 3426 3427 3428 3429 3430 3431 3432
}

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;
}
3433

D
David Gibson 已提交
3434 3435 3436
int copy_hugetlb_page_range(struct mm_struct *dst, struct mm_struct *src,
			    struct vm_area_struct *vma)
{
3437
	pte_t *src_pte, *dst_pte, entry, dst_entry;
D
David Gibson 已提交
3438
	struct page *ptepage;
3439
	unsigned long addr;
3440
	int cow;
3441 3442
	struct hstate *h = hstate_vma(vma);
	unsigned long sz = huge_page_size(h);
3443
	struct address_space *mapping = vma->vm_file->f_mapping;
3444
	struct mmu_notifier_range range;
3445
	int ret = 0;
3446 3447

	cow = (vma->vm_flags & (VM_SHARED | VM_MAYWRITE)) == VM_MAYWRITE;
D
David Gibson 已提交
3448

3449
	if (cow) {
3450
		mmu_notifier_range_init(&range, MMU_NOTIFY_CLEAR, 0, vma, src,
3451
					vma->vm_start,
3452 3453
					vma->vm_end);
		mmu_notifier_invalidate_range_start(&range);
3454 3455 3456 3457 3458 3459 3460 3461
	} else {
		/*
		 * For shared mappings i_mmap_rwsem must be held to call
		 * huge_pte_alloc, otherwise the returned ptep could go
		 * away if part of a shared pmd and another thread calls
		 * huge_pmd_unshare.
		 */
		i_mmap_lock_read(mapping);
3462
	}
3463

3464
	for (addr = vma->vm_start; addr < vma->vm_end; addr += sz) {
3465
		spinlock_t *src_ptl, *dst_ptl;
3466
		src_pte = huge_pte_offset(src, addr, sz);
H
Hugh Dickins 已提交
3467 3468
		if (!src_pte)
			continue;
3469
		dst_pte = huge_pte_alloc(dst, addr, sz);
3470 3471 3472 3473
		if (!dst_pte) {
			ret = -ENOMEM;
			break;
		}
3474

3475 3476 3477 3478 3479 3480 3481 3482 3483 3484 3485
		/*
		 * If the pagetables are shared don't copy or take references.
		 * dst_pte == src_pte is the common case of src/dest sharing.
		 *
		 * However, src could have 'unshared' and dst shares with
		 * another vma.  If dst_pte !none, this implies sharing.
		 * Check here before taking page table lock, and once again
		 * after taking the lock below.
		 */
		dst_entry = huge_ptep_get(dst_pte);
		if ((dst_pte == src_pte) || !huge_pte_none(dst_entry))
3486 3487
			continue;

3488 3489 3490
		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);
3491
		entry = huge_ptep_get(src_pte);
3492 3493 3494 3495 3496 3497 3498
		dst_entry = huge_ptep_get(dst_pte);
		if (huge_pte_none(entry) || !huge_pte_none(dst_entry)) {
			/*
			 * Skip if src entry none.  Also, skip in the
			 * unlikely case dst entry !none as this implies
			 * sharing with another vma.
			 */
3499 3500 3501 3502 3503 3504 3505 3506 3507 3508 3509 3510
			;
		} 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);
3511 3512
				set_huge_swap_pte_at(src, addr, src_pte,
						     entry, sz);
3513
			}
3514
			set_huge_swap_pte_at(dst, addr, dst_pte, entry, sz);
3515
		} else {
3516
			if (cow) {
3517 3518 3519 3520 3521
				/*
				 * No need to notify as we are downgrading page
				 * table protection not changing it to point
				 * to a new page.
				 *
3522
				 * See Documentation/vm/mmu_notifier.rst
3523
				 */
3524
				huge_ptep_set_wrprotect(src, addr, src_pte);
3525
			}
3526
			entry = huge_ptep_get(src_pte);
3527 3528
			ptepage = pte_page(entry);
			get_page(ptepage);
3529
			page_dup_rmap(ptepage, true);
3530
			set_huge_pte_at(dst, addr, dst_pte, entry);
3531
			hugetlb_count_add(pages_per_huge_page(h), dst);
3532
		}
3533 3534
		spin_unlock(src_ptl);
		spin_unlock(dst_ptl);
D
David Gibson 已提交
3535 3536
	}

3537
	if (cow)
3538
		mmu_notifier_invalidate_range_end(&range);
3539 3540
	else
		i_mmap_unlock_read(mapping);
3541 3542

	return ret;
D
David Gibson 已提交
3543 3544
}

3545 3546 3547
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
3548 3549 3550
{
	struct mm_struct *mm = vma->vm_mm;
	unsigned long address;
3551
	pte_t *ptep;
D
David Gibson 已提交
3552
	pte_t pte;
3553
	spinlock_t *ptl;
D
David Gibson 已提交
3554
	struct page *page;
3555 3556
	struct hstate *h = hstate_vma(vma);
	unsigned long sz = huge_page_size(h);
3557
	struct mmu_notifier_range range;
3558

D
David Gibson 已提交
3559
	WARN_ON(!is_vm_hugetlb_page(vma));
3560 3561
	BUG_ON(start & ~huge_page_mask(h));
	BUG_ON(end & ~huge_page_mask(h));
D
David Gibson 已提交
3562

3563 3564 3565 3566
	/*
	 * This is a hugetlb vma, all the pte entries should point
	 * to huge page.
	 */
3567
	tlb_change_page_size(tlb, sz);
3568
	tlb_start_vma(tlb, vma);
3569 3570 3571 3572

	/*
	 * If sharing possible, alert mmu notifiers of worst case.
	 */
3573 3574
	mmu_notifier_range_init(&range, MMU_NOTIFY_UNMAP, 0, vma, mm, start,
				end);
3575 3576
	adjust_range_if_pmd_sharing_possible(vma, &range.start, &range.end);
	mmu_notifier_invalidate_range_start(&range);
3577 3578
	address = start;
	for (; address < end; address += sz) {
3579
		ptep = huge_pte_offset(mm, address, sz);
A
Adam Litke 已提交
3580
		if (!ptep)
3581 3582
			continue;

3583
		ptl = huge_pte_lock(h, mm, ptep);
3584 3585
		if (huge_pmd_unshare(mm, &address, ptep)) {
			spin_unlock(ptl);
3586 3587 3588 3589
			/*
			 * We just unmapped a page of PMDs by clearing a PUD.
			 * The caller's TLB flush range should cover this area.
			 */
3590 3591
			continue;
		}
3592

3593
		pte = huge_ptep_get(ptep);
3594 3595 3596 3597
		if (huge_pte_none(pte)) {
			spin_unlock(ptl);
			continue;
		}
3598 3599

		/*
3600 3601
		 * Migrating hugepage or HWPoisoned hugepage is already
		 * unmapped and its refcount is dropped, so just clear pte here.
3602
		 */
3603
		if (unlikely(!pte_present(pte))) {
3604
			huge_pte_clear(mm, address, ptep, sz);
3605 3606
			spin_unlock(ptl);
			continue;
3607
		}
3608 3609

		page = pte_page(pte);
3610 3611 3612 3613 3614 3615
		/*
		 * 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) {
3616 3617 3618 3619
			if (page != ref_page) {
				spin_unlock(ptl);
				continue;
			}
3620 3621 3622 3623 3624 3625 3626 3627
			/*
			 * 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);
		}

3628
		pte = huge_ptep_get_and_clear(mm, address, ptep);
3629
		tlb_remove_huge_tlb_entry(h, tlb, ptep, address);
3630
		if (huge_pte_dirty(pte))
3631
			set_page_dirty(page);
3632

3633
		hugetlb_count_sub(pages_per_huge_page(h), mm);
3634
		page_remove_rmap(page, true);
3635

3636
		spin_unlock(ptl);
3637
		tlb_remove_page_size(tlb, page, huge_page_size(h));
3638 3639 3640 3641 3642
		/*
		 * Bail out after unmapping reference page if supplied
		 */
		if (ref_page)
			break;
3643
	}
3644
	mmu_notifier_invalidate_range_end(&range);
3645
	tlb_end_vma(tlb, vma);
L
Linus Torvalds 已提交
3646
}
D
David Gibson 已提交
3647

3648 3649 3650 3651 3652 3653 3654 3655 3656 3657 3658 3659
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
3660
	 * is to clear it before releasing the i_mmap_rwsem. This works
3661
	 * because in the context this is called, the VMA is about to be
3662
	 * destroyed and the i_mmap_rwsem is held.
3663 3664 3665 3666
	 */
	vma->vm_flags &= ~VM_MAYSHARE;
}

3667
void unmap_hugepage_range(struct vm_area_struct *vma, unsigned long start,
3668
			  unsigned long end, struct page *ref_page)
3669
{
3670 3671
	struct mm_struct *mm;
	struct mmu_gather tlb;
3672 3673 3674 3675 3676 3677 3678 3679 3680 3681 3682
	unsigned long tlb_start = start;
	unsigned long tlb_end = end;

	/*
	 * If shared PMDs were possibly used within this vma range, adjust
	 * start/end for worst case tlb flushing.
	 * Note that we can not be sure if PMDs are shared until we try to
	 * unmap pages.  However, we want to make sure TLB flushing covers
	 * the largest possible range.
	 */
	adjust_range_if_pmd_sharing_possible(vma, &tlb_start, &tlb_end);
3683 3684 3685

	mm = vma->vm_mm;

3686
	tlb_gather_mmu(&tlb, mm, tlb_start, tlb_end);
3687
	__unmap_hugepage_range(&tlb, vma, start, end, ref_page);
3688
	tlb_finish_mmu(&tlb, tlb_start, tlb_end);
3689 3690
}

3691 3692 3693 3694 3695 3696
/*
 * 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.
 */
3697 3698
static void unmap_ref_private(struct mm_struct *mm, struct vm_area_struct *vma,
			      struct page *page, unsigned long address)
3699
{
3700
	struct hstate *h = hstate_vma(vma);
3701 3702 3703 3704 3705 3706 3707 3708
	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.
	 */
3709
	address = address & huge_page_mask(h);
3710 3711
	pgoff = ((address - vma->vm_start) >> PAGE_SHIFT) +
			vma->vm_pgoff;
3712
	mapping = vma->vm_file->f_mapping;
3713

3714 3715 3716 3717 3718
	/*
	 * 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
	 */
3719
	i_mmap_lock_write(mapping);
3720
	vma_interval_tree_foreach(iter_vma, &mapping->i_mmap, pgoff, pgoff) {
3721 3722 3723 3724
		/* Do not unmap the current VMA */
		if (iter_vma == vma)
			continue;

3725 3726 3727 3728 3729 3730 3731 3732
		/*
		 * 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;

3733 3734 3735 3736 3737 3738 3739 3740
		/*
		 * 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))
3741 3742
			unmap_hugepage_range(iter_vma, address,
					     address + huge_page_size(h), page);
3743
	}
3744
	i_mmap_unlock_write(mapping);
3745 3746
}

3747 3748
/*
 * Hugetlb_cow() should be called with page lock of the original hugepage held.
3749 3750 3751
 * 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.
3752
 */
3753
static vm_fault_t hugetlb_cow(struct mm_struct *mm, struct vm_area_struct *vma,
3754
		       unsigned long address, pte_t *ptep,
3755
		       struct page *pagecache_page, spinlock_t *ptl)
3756
{
3757
	pte_t pte;
3758
	struct hstate *h = hstate_vma(vma);
3759
	struct page *old_page, *new_page;
3760 3761
	int outside_reserve = 0;
	vm_fault_t ret = 0;
3762
	unsigned long haddr = address & huge_page_mask(h);
3763
	struct mmu_notifier_range range;
3764

3765
	pte = huge_ptep_get(ptep);
3766 3767
	old_page = pte_page(pte);

3768
retry_avoidcopy:
3769 3770
	/* If no-one else is actually using this page, avoid the copy
	 * and just make the page writable */
3771
	if (page_mapcount(old_page) == 1 && PageAnon(old_page)) {
3772
		page_move_anon_rmap(old_page, vma);
3773
		set_huge_ptep_writable(vma, haddr, ptep);
N
Nick Piggin 已提交
3774
		return 0;
3775 3776
	}

3777 3778 3779 3780 3781 3782 3783 3784 3785
	/*
	 * 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.
	 */
3786
	if (is_vma_resv_set(vma, HPAGE_RESV_OWNER) &&
3787 3788 3789
			old_page != pagecache_page)
		outside_reserve = 1;

3790
	get_page(old_page);
3791

3792 3793 3794 3795
	/*
	 * Drop page table lock as buddy allocator may be called. It will
	 * be acquired again before returning to the caller, as expected.
	 */
3796
	spin_unlock(ptl);
3797
	new_page = alloc_huge_page(vma, haddr, outside_reserve);
3798

3799
	if (IS_ERR(new_page)) {
3800 3801 3802 3803 3804 3805 3806 3807
		/*
		 * 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) {
3808
			put_page(old_page);
3809
			BUG_ON(huge_pte_none(pte));
3810
			unmap_ref_private(mm, vma, old_page, haddr);
3811 3812
			BUG_ON(huge_pte_none(pte));
			spin_lock(ptl);
3813
			ptep = huge_pte_offset(mm, haddr, huge_page_size(h));
3814 3815 3816 3817 3818 3819 3820 3821
			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;
3822 3823
		}

3824
		ret = vmf_error(PTR_ERR(new_page));
3825
		goto out_release_old;
3826 3827
	}

3828 3829 3830 3831
	/*
	 * When the original hugepage is shared one, it does not have
	 * anon_vma prepared.
	 */
3832
	if (unlikely(anon_vma_prepare(vma))) {
3833 3834
		ret = VM_FAULT_OOM;
		goto out_release_all;
3835
	}
3836

3837
	copy_user_huge_page(new_page, old_page, address, vma,
A
Andrea Arcangeli 已提交
3838
			    pages_per_huge_page(h));
N
Nick Piggin 已提交
3839
	__SetPageUptodate(new_page);
3840

3841
	mmu_notifier_range_init(&range, MMU_NOTIFY_CLEAR, 0, vma, mm, haddr,
3842
				haddr + huge_page_size(h));
3843
	mmu_notifier_invalidate_range_start(&range);
3844

3845
	/*
3846
	 * Retake the page table lock to check for racing updates
3847 3848
	 * before the page tables are altered
	 */
3849
	spin_lock(ptl);
3850
	ptep = huge_pte_offset(mm, haddr, huge_page_size(h));
3851
	if (likely(ptep && pte_same(huge_ptep_get(ptep), pte))) {
3852 3853
		ClearPagePrivate(new_page);

3854
		/* Break COW */
3855
		huge_ptep_clear_flush(vma, haddr, ptep);
3856
		mmu_notifier_invalidate_range(mm, range.start, range.end);
3857
		set_huge_pte_at(mm, haddr, ptep,
3858
				make_huge_pte(vma, new_page, 1));
3859
		page_remove_rmap(old_page, true);
3860
		hugepage_add_new_anon_rmap(new_page, vma, haddr);
3861
		set_page_huge_active(new_page);
3862 3863 3864
		/* Make the old page be freed below */
		new_page = old_page;
	}
3865
	spin_unlock(ptl);
3866
	mmu_notifier_invalidate_range_end(&range);
3867
out_release_all:
3868
	restore_reserve_on_error(h, vma, haddr, new_page);
3869
	put_page(new_page);
3870
out_release_old:
3871
	put_page(old_page);
3872

3873 3874
	spin_lock(ptl); /* Caller expects lock to be held */
	return ret;
3875 3876
}

3877
/* Return the pagecache page at a given address within a VMA */
3878 3879
static struct page *hugetlbfs_pagecache_page(struct hstate *h,
			struct vm_area_struct *vma, unsigned long address)
3880 3881
{
	struct address_space *mapping;
3882
	pgoff_t idx;
3883 3884

	mapping = vma->vm_file->f_mapping;
3885
	idx = vma_hugecache_offset(h, vma, address);
3886 3887 3888 3889

	return find_lock_page(mapping, idx);
}

H
Hugh Dickins 已提交
3890 3891 3892 3893 3894
/*
 * 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
3895 3896 3897 3898 3899 3900 3901 3902 3903 3904 3905 3906 3907 3908 3909
			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;
}

3910 3911 3912 3913 3914 3915 3916 3917 3918 3919 3920
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);

3921 3922 3923 3924 3925 3926
	/*
	 * set page dirty so that it will not be removed from cache/file
	 * by non-hugetlbfs specific code paths.
	 */
	set_page_dirty(page);

3927 3928 3929 3930 3931 3932
	spin_lock(&inode->i_lock);
	inode->i_blocks += blocks_per_huge_page(h);
	spin_unlock(&inode->i_lock);
	return 0;
}

3933 3934 3935 3936
static vm_fault_t hugetlb_no_page(struct mm_struct *mm,
			struct vm_area_struct *vma,
			struct address_space *mapping, pgoff_t idx,
			unsigned long address, pte_t *ptep, unsigned int flags)
3937
{
3938
	struct hstate *h = hstate_vma(vma);
3939
	vm_fault_t ret = VM_FAULT_SIGBUS;
3940
	int anon_rmap = 0;
A
Adam Litke 已提交
3941 3942
	unsigned long size;
	struct page *page;
3943
	pte_t new_pte;
3944
	spinlock_t *ptl;
3945
	unsigned long haddr = address & huge_page_mask(h);
3946
	bool new_page = false;
A
Adam Litke 已提交
3947

3948 3949 3950
	/*
	 * 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
3951
	 * COW. Warn that such a situation has occurred as it may not be obvious
3952 3953
	 */
	if (is_vma_resv_set(vma, HPAGE_RESV_UNMAPPED)) {
3954
		pr_warn_ratelimited("PID %d killed due to inadequate hugepage pool\n",
3955
			   current->pid);
3956 3957 3958
		return ret;
	}

A
Adam Litke 已提交
3959
	/*
3960 3961 3962
	 * We can not race with truncation due to holding i_mmap_rwsem.
	 * i_size is modified when holding i_mmap_rwsem, so check here
	 * once for faults beyond end of file.
A
Adam Litke 已提交
3963
	 */
3964 3965 3966 3967
	size = i_size_read(mapping->host) >> huge_page_shift(h);
	if (idx >= size)
		goto out;

3968 3969 3970
retry:
	page = find_lock_page(mapping, idx);
	if (!page) {
3971 3972 3973 3974 3975 3976 3977
		/*
		 * Check for page in userfault range
		 */
		if (userfaultfd_missing(vma)) {
			u32 hash;
			struct vm_fault vmf = {
				.vma = vma,
3978
				.address = haddr,
3979 3980 3981 3982 3983 3984 3985 3986 3987 3988 3989
				.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
				 */
			};

			/*
3990 3991 3992
			 * hugetlb_fault_mutex and i_mmap_rwsem must be
			 * dropped before handling userfault.  Reacquire
			 * after handling fault to make calling code simpler.
3993
			 */
3994
			hash = hugetlb_fault_mutex_hash(mapping, idx);
3995
			mutex_unlock(&hugetlb_fault_mutex_table[hash]);
3996
			i_mmap_unlock_read(mapping);
3997
			ret = handle_userfault(&vmf, VM_UFFD_MISSING);
3998
			i_mmap_lock_read(mapping);
3999 4000 4001 4002
			mutex_lock(&hugetlb_fault_mutex_table[hash]);
			goto out;
		}

4003
		page = alloc_huge_page(vma, haddr, 0);
4004
		if (IS_ERR(page)) {
4005 4006 4007 4008 4009 4010 4011 4012 4013 4014 4015 4016 4017 4018 4019 4020 4021 4022 4023
			/*
			 * Returning error will result in faulting task being
			 * sent SIGBUS.  The hugetlb fault mutex prevents two
			 * tasks from racing to fault in the same page which
			 * could result in false unable to allocate errors.
			 * Page migration does not take the fault mutex, but
			 * does a clear then write of pte's under page table
			 * lock.  Page fault code could race with migration,
			 * notice the clear pte and try to allocate a page
			 * here.  Before returning error, get ptl and make
			 * sure there really is no pte entry.
			 */
			ptl = huge_pte_lock(h, mm, ptep);
			if (!huge_pte_none(huge_ptep_get(ptep))) {
				ret = 0;
				spin_unlock(ptl);
				goto out;
			}
			spin_unlock(ptl);
4024
			ret = vmf_error(PTR_ERR(page));
4025 4026
			goto out;
		}
A
Andrea Arcangeli 已提交
4027
		clear_huge_page(page, address, pages_per_huge_page(h));
N
Nick Piggin 已提交
4028
		__SetPageUptodate(page);
4029
		new_page = true;
4030

4031
		if (vma->vm_flags & VM_MAYSHARE) {
4032
			int err = huge_add_to_page_cache(page, mapping, idx);
4033 4034 4035 4036 4037 4038
			if (err) {
				put_page(page);
				if (err == -EEXIST)
					goto retry;
				goto out;
			}
4039
		} else {
4040
			lock_page(page);
4041 4042 4043 4044
			if (unlikely(anon_vma_prepare(vma))) {
				ret = VM_FAULT_OOM;
				goto backout_unlocked;
			}
4045
			anon_rmap = 1;
4046
		}
4047
	} else {
4048 4049 4050 4051 4052 4053
		/*
		 * 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))) {
4054
			ret = VM_FAULT_HWPOISON |
4055
				VM_FAULT_SET_HINDEX(hstate_index(h));
4056 4057
			goto backout_unlocked;
		}
4058
	}
4059

4060 4061 4062 4063 4064 4065
	/*
	 * 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.
	 */
4066
	if ((flags & FAULT_FLAG_WRITE) && !(vma->vm_flags & VM_SHARED)) {
4067
		if (vma_needs_reservation(h, vma, haddr) < 0) {
4068 4069 4070
			ret = VM_FAULT_OOM;
			goto backout_unlocked;
		}
4071
		/* Just decrements count, does not deallocate */
4072
		vma_end_reservation(h, vma, haddr);
4073
	}
4074

4075
	ptl = huge_pte_lock(h, mm, ptep);
N
Nick Piggin 已提交
4076
	ret = 0;
4077
	if (!huge_pte_none(huge_ptep_get(ptep)))
A
Adam Litke 已提交
4078 4079
		goto backout;

4080 4081
	if (anon_rmap) {
		ClearPagePrivate(page);
4082
		hugepage_add_new_anon_rmap(page, vma, haddr);
4083
	} else
4084
		page_dup_rmap(page, true);
4085 4086
	new_pte = make_huge_pte(vma, page, ((vma->vm_flags & VM_WRITE)
				&& (vma->vm_flags & VM_SHARED)));
4087
	set_huge_pte_at(mm, haddr, ptep, new_pte);
4088

4089
	hugetlb_count_add(pages_per_huge_page(h), mm);
4090
	if ((flags & FAULT_FLAG_WRITE) && !(vma->vm_flags & VM_SHARED)) {
4091
		/* Optimization, do the COW without a second fault */
4092
		ret = hugetlb_cow(mm, vma, address, ptep, page, ptl);
4093 4094
	}

4095
	spin_unlock(ptl);
4096 4097 4098 4099 4100 4101 4102 4103 4104

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

A
Adam Litke 已提交
4105 4106
	unlock_page(page);
out:
4107
	return ret;
A
Adam Litke 已提交
4108 4109

backout:
4110
	spin_unlock(ptl);
4111
backout_unlocked:
A
Adam Litke 已提交
4112
	unlock_page(page);
4113
	restore_reserve_on_error(h, vma, haddr, page);
A
Adam Litke 已提交
4114 4115
	put_page(page);
	goto out;
4116 4117
}

4118
#ifdef CONFIG_SMP
4119
u32 hugetlb_fault_mutex_hash(struct address_space *mapping, pgoff_t idx)
4120 4121 4122 4123
{
	unsigned long key[2];
	u32 hash;

4124 4125
	key[0] = (unsigned long) mapping;
	key[1] = idx;
4126

4127
	hash = jhash2((u32 *)&key, sizeof(key)/(sizeof(u32)), 0);
4128 4129 4130 4131 4132 4133 4134 4135

	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.
 */
4136
u32 hugetlb_fault_mutex_hash(struct address_space *mapping, pgoff_t idx)
4137 4138 4139 4140 4141
{
	return 0;
}
#endif

4142
vm_fault_t hugetlb_fault(struct mm_struct *mm, struct vm_area_struct *vma,
4143
			unsigned long address, unsigned int flags)
4144
{
4145
	pte_t *ptep, entry;
4146
	spinlock_t *ptl;
4147
	vm_fault_t ret;
4148 4149
	u32 hash;
	pgoff_t idx;
4150
	struct page *page = NULL;
4151
	struct page *pagecache_page = NULL;
4152
	struct hstate *h = hstate_vma(vma);
4153
	struct address_space *mapping;
4154
	int need_wait_lock = 0;
4155
	unsigned long haddr = address & huge_page_mask(h);
4156

4157
	ptep = huge_pte_offset(mm, haddr, huge_page_size(h));
4158
	if (ptep) {
4159 4160 4161 4162 4163
		/*
		 * Since we hold no locks, ptep could be stale.  That is
		 * OK as we are only making decisions based on content and
		 * not actually modifying content here.
		 */
4164
		entry = huge_ptep_get(ptep);
N
Naoya Horiguchi 已提交
4165
		if (unlikely(is_hugetlb_entry_migration(entry))) {
4166
			migration_entry_wait_huge(vma, mm, ptep);
N
Naoya Horiguchi 已提交
4167 4168
			return 0;
		} else if (unlikely(is_hugetlb_entry_hwpoisoned(entry)))
4169
			return VM_FAULT_HWPOISON_LARGE |
4170
				VM_FAULT_SET_HINDEX(hstate_index(h));
4171 4172 4173 4174
	} else {
		ptep = huge_pte_alloc(mm, haddr, huge_page_size(h));
		if (!ptep)
			return VM_FAULT_OOM;
4175 4176
	}

4177 4178
	/*
	 * Acquire i_mmap_rwsem before calling huge_pte_alloc and hold
4179 4180 4181 4182
	 * until finished with ptep.  This serves two purposes:
	 * 1) It prevents huge_pmd_unshare from being called elsewhere
	 *    and making the ptep no longer valid.
	 * 2) It synchronizes us with i_size modifications during truncation.
4183 4184 4185 4186 4187
	 *
	 * ptep could have already be assigned via huge_pte_offset.  That
	 * is OK, as huge_pte_alloc will return the same value unless
	 * something has changed.
	 */
4188
	mapping = vma->vm_file->f_mapping;
4189 4190 4191 4192 4193 4194
	i_mmap_lock_read(mapping);
	ptep = huge_pte_alloc(mm, haddr, huge_page_size(h));
	if (!ptep) {
		i_mmap_unlock_read(mapping);
		return VM_FAULT_OOM;
	}
4195

4196 4197 4198 4199 4200
	/*
	 * 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.
	 */
4201
	idx = vma_hugecache_offset(h, vma, haddr);
4202
	hash = hugetlb_fault_mutex_hash(mapping, idx);
4203
	mutex_lock(&hugetlb_fault_mutex_table[hash]);
4204

4205 4206
	entry = huge_ptep_get(ptep);
	if (huge_pte_none(entry)) {
4207
		ret = hugetlb_no_page(mm, vma, mapping, idx, address, ptep, flags);
4208
		goto out_mutex;
4209
	}
4210

N
Nick Piggin 已提交
4211
	ret = 0;
4212

4213 4214 4215 4216 4217 4218 4219 4220 4221 4222
	/*
	 * 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;

4223 4224 4225 4226 4227 4228 4229 4230
	/*
	 * 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.
	 */
4231
	if ((flags & FAULT_FLAG_WRITE) && !huge_pte_write(entry)) {
4232
		if (vma_needs_reservation(h, vma, haddr) < 0) {
4233
			ret = VM_FAULT_OOM;
4234
			goto out_mutex;
4235
		}
4236
		/* Just decrements count, does not deallocate */
4237
		vma_end_reservation(h, vma, haddr);
4238

4239
		if (!(vma->vm_flags & VM_MAYSHARE))
4240
			pagecache_page = hugetlbfs_pagecache_page(h,
4241
								vma, haddr);
4242 4243
	}

4244 4245 4246 4247 4248 4249
	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;

4250 4251 4252 4253 4254 4255 4256
	/*
	 * 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)
4257 4258 4259 4260
		if (!trylock_page(page)) {
			need_wait_lock = 1;
			goto out_ptl;
		}
4261

4262
	get_page(page);
4263

4264
	if (flags & FAULT_FLAG_WRITE) {
4265
		if (!huge_pte_write(entry)) {
4266
			ret = hugetlb_cow(mm, vma, address, ptep,
4267
					  pagecache_page, ptl);
4268
			goto out_put_page;
4269
		}
4270
		entry = huge_pte_mkdirty(entry);
4271 4272
	}
	entry = pte_mkyoung(entry);
4273
	if (huge_ptep_set_access_flags(vma, haddr, ptep, entry,
4274
						flags & FAULT_FLAG_WRITE))
4275
		update_mmu_cache(vma, haddr, ptep);
4276 4277 4278 4279
out_put_page:
	if (page != pagecache_page)
		unlock_page(page);
	put_page(page);
4280 4281
out_ptl:
	spin_unlock(ptl);
4282 4283 4284 4285 4286

	if (pagecache_page) {
		unlock_page(pagecache_page);
		put_page(pagecache_page);
	}
4287
out_mutex:
4288
	mutex_unlock(&hugetlb_fault_mutex_table[hash]);
4289
	i_mmap_unlock_read(mapping);
4290 4291 4292 4293 4294 4295 4296 4297 4298
	/*
	 * 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);
4299
	return ret;
4300 4301
}

4302 4303 4304 4305 4306 4307 4308 4309 4310 4311 4312
/*
 * 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)
{
4313 4314 4315
	struct address_space *mapping;
	pgoff_t idx;
	unsigned long size;
4316
	int vm_shared = dst_vma->vm_flags & VM_SHARED;
4317 4318 4319 4320 4321 4322 4323 4324 4325 4326 4327 4328 4329 4330
	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,
4331
						pages_per_huge_page(h), false);
4332 4333 4334

		/* fallback to copy_from_user outside mmap_sem */
		if (unlikely(ret)) {
4335
			ret = -ENOENT;
4336 4337 4338 4339 4340 4341 4342 4343 4344 4345 4346 4347 4348 4349 4350 4351
			*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);

4352 4353 4354
	mapping = dst_vma->vm_file->f_mapping;
	idx = vma_hugecache_offset(h, dst_vma, dst_addr);

4355 4356 4357 4358
	/*
	 * If shared, add to page cache
	 */
	if (vm_shared) {
4359 4360 4361 4362
		size = i_size_read(mapping->host) >> huge_page_shift(h);
		ret = -EFAULT;
		if (idx >= size)
			goto out_release_nounlock;
4363

4364 4365 4366 4367 4368 4369
		/*
		 * Serialization between remove_inode_hugepages() and
		 * huge_add_to_page_cache() below happens through the
		 * hugetlb_fault_mutex_table that here must be hold by
		 * the caller.
		 */
4370 4371 4372 4373 4374
		ret = huge_add_to_page_cache(page, mapping, idx);
		if (ret)
			goto out_release_nounlock;
	}

4375 4376 4377
	ptl = huge_pte_lockptr(h, dst_mm, dst_pte);
	spin_lock(ptl);

4378 4379 4380 4381 4382 4383 4384 4385 4386 4387 4388 4389 4390 4391
	/*
	 * Recheck the i_size after holding PT lock to make sure not
	 * to leave any page mapped (as page_mapped()) beyond the end
	 * of the i_size (remove_inode_hugepages() is strict about
	 * enforcing that). If we bail out here, we'll also leave a
	 * page in the radix tree in the vm_shared case beyond the end
	 * of the i_size, but remove_inode_hugepages() will take care
	 * of it as soon as we drop the hugetlb_fault_mutex_table.
	 */
	size = i_size_read(mapping->host) >> huge_page_shift(h);
	ret = -EFAULT;
	if (idx >= size)
		goto out_release_unlock;

4392 4393 4394 4395
	ret = -EEXIST;
	if (!huge_pte_none(huge_ptep_get(dst_pte)))
		goto out_release_unlock;

4396 4397 4398 4399 4400 4401
	if (vm_shared) {
		page_dup_rmap(page, true);
	} else {
		ClearPagePrivate(page);
		hugepage_add_new_anon_rmap(page, dst_vma, dst_addr);
	}
4402 4403 4404 4405 4406 4407 4408 4409 4410 4411 4412 4413 4414 4415 4416 4417

	_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);
4418
	set_page_huge_active(page);
4419 4420
	if (vm_shared)
		unlock_page(page);
4421 4422 4423 4424 4425
	ret = 0;
out:
	return ret;
out_release_unlock:
	spin_unlock(ptl);
4426 4427
	if (vm_shared)
		unlock_page(page);
4428
out_release_nounlock:
4429 4430 4431 4432
	put_page(page);
	goto out;
}

4433 4434 4435
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,
4436
			 long i, unsigned int flags, int *locked)
D
David Gibson 已提交
4437
{
4438 4439
	unsigned long pfn_offset;
	unsigned long vaddr = *position;
4440
	unsigned long remainder = *nr_pages;
4441
	struct hstate *h = hstate_vma(vma);
4442
	int err = -EFAULT;
D
David Gibson 已提交
4443 4444

	while (vaddr < vma->vm_end && remainder) {
A
Adam Litke 已提交
4445
		pte_t *pte;
4446
		spinlock_t *ptl = NULL;
H
Hugh Dickins 已提交
4447
		int absent;
A
Adam Litke 已提交
4448
		struct page *page;
D
David Gibson 已提交
4449

4450 4451 4452 4453
		/*
		 * If we have a pending SIGKILL, don't keep faulting pages and
		 * potentially allocating memory.
		 */
4454
		if (fatal_signal_pending(current)) {
4455 4456 4457 4458
			remainder = 0;
			break;
		}

A
Adam Litke 已提交
4459 4460
		/*
		 * Some archs (sparc64, sh*) have multiple pte_ts to
H
Hugh Dickins 已提交
4461
		 * each hugepage.  We have to make sure we get the
A
Adam Litke 已提交
4462
		 * first, for the page indexing below to work.
4463 4464
		 *
		 * Note that page table lock is not held when pte is null.
A
Adam Litke 已提交
4465
		 */
4466 4467
		pte = huge_pte_offset(mm, vaddr & huge_page_mask(h),
				      huge_page_size(h));
4468 4469
		if (pte)
			ptl = huge_pte_lock(h, mm, pte);
H
Hugh Dickins 已提交
4470 4471 4472 4473
		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
4474 4475 4476 4477
		 * 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
4478
		 */
H
Hugh Dickins 已提交
4479 4480
		if (absent && (flags & FOLL_DUMP) &&
		    !hugetlbfs_pagecache_present(h, vma, vaddr)) {
4481 4482
			if (pte)
				spin_unlock(ptl);
H
Hugh Dickins 已提交
4483 4484 4485
			remainder = 0;
			break;
		}
D
David Gibson 已提交
4486

4487 4488 4489 4490 4491 4492 4493 4494 4495 4496 4497
		/*
		 * 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)) ||
4498 4499
		    ((flags & FOLL_WRITE) &&
		      !huge_pte_write(huge_ptep_get(pte)))) {
4500
			vm_fault_t ret;
4501
			unsigned int fault_flags = 0;
D
David Gibson 已提交
4502

4503 4504
			if (pte)
				spin_unlock(ptl);
4505 4506
			if (flags & FOLL_WRITE)
				fault_flags |= FAULT_FLAG_WRITE;
4507
			if (locked)
4508 4509
				fault_flags |= FAULT_FLAG_ALLOW_RETRY |
					FAULT_FLAG_KILLABLE;
4510 4511 4512 4513
			if (flags & FOLL_NOWAIT)
				fault_flags |= FAULT_FLAG_ALLOW_RETRY |
					FAULT_FLAG_RETRY_NOWAIT;
			if (flags & FOLL_TRIED) {
4514 4515 4516 4517
				/*
				 * Note: FAULT_FLAG_ALLOW_RETRY and
				 * FAULT_FLAG_TRIED can co-exist
				 */
4518 4519 4520 4521
				fault_flags |= FAULT_FLAG_TRIED;
			}
			ret = hugetlb_fault(mm, vma, vaddr, fault_flags);
			if (ret & VM_FAULT_ERROR) {
4522
				err = vm_fault_to_errno(ret, flags);
4523 4524 4525 4526
				remainder = 0;
				break;
			}
			if (ret & VM_FAULT_RETRY) {
4527
				if (locked &&
4528
				    !(fault_flags & FAULT_FLAG_RETRY_NOWAIT))
4529
					*locked = 0;
4530 4531 4532 4533 4534 4535 4536 4537 4538 4539 4540 4541 4542
				*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
4543 4544
		}

4545
		pfn_offset = (vaddr & ~huge_page_mask(h)) >> PAGE_SHIFT;
4546
		page = pte_page(huge_ptep_get(pte));
4547

4548 4549 4550 4551 4552 4553 4554 4555 4556 4557 4558 4559 4560 4561
		/*
		 * If subpage information not requested, update counters
		 * and skip the same_page loop below.
		 */
		if (!pages && !vmas && !pfn_offset &&
		    (vaddr + huge_page_size(h) < vma->vm_end) &&
		    (remainder >= pages_per_huge_page(h))) {
			vaddr += huge_page_size(h);
			remainder -= pages_per_huge_page(h);
			i += pages_per_huge_page(h);
			spin_unlock(ptl);
			continue;
		}

4562
same_page:
4563
		if (pages) {
H
Hugh Dickins 已提交
4564
			pages[i] = mem_map_offset(page, pfn_offset);
J
John Hubbard 已提交
4565 4566 4567 4568 4569 4570 4571 4572 4573 4574 4575 4576 4577 4578 4579 4580
			/*
			 * try_grab_page() should always succeed here, because:
			 * a) we hold the ptl lock, and b) we've just checked
			 * that the huge page is present in the page tables. If
			 * the huge page is present, then the tail pages must
			 * also be present. The ptl prevents the head page and
			 * tail pages from being rearranged in any way. So this
			 * page must be available at this point, unless the page
			 * refcount overflowed:
			 */
			if (WARN_ON_ONCE(!try_grab_page(pages[i], flags))) {
				spin_unlock(ptl);
				remainder = 0;
				err = -ENOMEM;
				break;
			}
4581
		}
D
David Gibson 已提交
4582 4583 4584 4585 4586

		if (vmas)
			vmas[i] = vma;

		vaddr += PAGE_SIZE;
4587
		++pfn_offset;
D
David Gibson 已提交
4588 4589
		--remainder;
		++i;
4590
		if (vaddr < vma->vm_end && remainder &&
4591
				pfn_offset < pages_per_huge_page(h)) {
4592 4593 4594 4595 4596 4597
			/*
			 * We use pfn_offset to avoid touching the pageframes
			 * of this compound page.
			 */
			goto same_page;
		}
4598
		spin_unlock(ptl);
D
David Gibson 已提交
4599
	}
4600
	*nr_pages = remainder;
4601 4602 4603 4604 4605
	/*
	 * 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
4606 4607
	*position = vaddr;

4608
	return i ? i : err;
D
David Gibson 已提交
4609
}
4610

4611 4612 4613 4614 4615 4616 4617 4618
#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

4619
unsigned long hugetlb_change_protection(struct vm_area_struct *vma,
4620 4621 4622 4623 4624 4625
		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;
4626
	struct hstate *h = hstate_vma(vma);
4627
	unsigned long pages = 0;
4628
	bool shared_pmd = false;
4629
	struct mmu_notifier_range range;
4630 4631 4632

	/*
	 * In the case of shared PMDs, the area to flush could be beyond
4633
	 * start/end.  Set range.start/range.end to cover the maximum possible
4634 4635
	 * range if PMD sharing is possible.
	 */
4636 4637
	mmu_notifier_range_init(&range, MMU_NOTIFY_PROTECTION_VMA,
				0, vma, mm, start, end);
4638
	adjust_range_if_pmd_sharing_possible(vma, &range.start, &range.end);
4639 4640

	BUG_ON(address >= end);
4641
	flush_cache_range(vma, range.start, range.end);
4642

4643
	mmu_notifier_invalidate_range_start(&range);
4644
	i_mmap_lock_write(vma->vm_file->f_mapping);
4645
	for (; address < end; address += huge_page_size(h)) {
4646
		spinlock_t *ptl;
4647
		ptep = huge_pte_offset(mm, address, huge_page_size(h));
4648 4649
		if (!ptep)
			continue;
4650
		ptl = huge_pte_lock(h, mm, ptep);
4651 4652
		if (huge_pmd_unshare(mm, &address, ptep)) {
			pages++;
4653
			spin_unlock(ptl);
4654
			shared_pmd = true;
4655
			continue;
4656
		}
4657 4658 4659 4660 4661 4662 4663 4664 4665 4666 4667 4668 4669
		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);
4670 4671
				set_huge_swap_pte_at(mm, address, ptep,
						     newpte, huge_page_size(h));
4672 4673 4674 4675 4676 4677
				pages++;
			}
			spin_unlock(ptl);
			continue;
		}
		if (!huge_pte_none(pte)) {
4678 4679 4680 4681
			pte_t old_pte;

			old_pte = huge_ptep_modify_prot_start(vma, address, ptep);
			pte = pte_mkhuge(huge_pte_modify(old_pte, newprot));
4682
			pte = arch_make_huge_pte(pte, vma, NULL, 0);
4683
			huge_ptep_modify_prot_commit(vma, address, ptep, old_pte, pte);
4684
			pages++;
4685
		}
4686
		spin_unlock(ptl);
4687
	}
4688
	/*
4689
	 * Must flush TLB before releasing i_mmap_rwsem: x86's huge_pmd_unshare
4690
	 * may have cleared our pud entry and done put_page on the page table:
4691
	 * once we release i_mmap_rwsem, another task can do the final put_page
4692 4693
	 * and that page table be reused and filled with junk.  If we actually
	 * did unshare a page of pmds, flush the range corresponding to the pud.
4694
	 */
4695
	if (shared_pmd)
4696
		flush_hugetlb_tlb_range(vma, range.start, range.end);
4697 4698
	else
		flush_hugetlb_tlb_range(vma, start, end);
4699 4700 4701 4702
	/*
	 * No need to call mmu_notifier_invalidate_range() we are downgrading
	 * page table protection not changing it to point to a new page.
	 *
4703
	 * See Documentation/vm/mmu_notifier.rst
4704
	 */
4705
	i_mmap_unlock_write(vma->vm_file->f_mapping);
4706
	mmu_notifier_invalidate_range_end(&range);
4707 4708

	return pages << h->order;
4709 4710
}

4711 4712
int hugetlb_reserve_pages(struct inode *inode,
					long from, long to,
4713
					struct vm_area_struct *vma,
4714
					vm_flags_t vm_flags)
4715
{
4716
	long ret, chg;
4717
	struct hstate *h = hstate_inode(inode);
4718
	struct hugepage_subpool *spool = subpool_inode(inode);
4719
	struct resv_map *resv_map;
4720
	struct hugetlb_cgroup *h_cg;
4721
	long gbl_reserve;
4722

4723 4724 4725 4726 4727 4728
	/* This should never happen */
	if (from > to) {
		VM_WARN(1, "%s called with a negative range\n", __func__);
		return -EINVAL;
	}

4729 4730 4731
	/*
	 * Only apply hugepage reservation if asked. At fault time, an
	 * attempt will be made for VM_NORESERVE to allocate a page
4732
	 * without using reserves
4733
	 */
4734
	if (vm_flags & VM_NORESERVE)
4735 4736
		return 0;

4737 4738 4739 4740 4741 4742
	/*
	 * 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
	 */
4743
	if (!vma || vma->vm_flags & VM_MAYSHARE) {
4744 4745 4746 4747 4748
		/*
		 * resv_map can not be NULL as hugetlb_reserve_pages is only
		 * called for inodes for which resv_maps were created (see
		 * hugetlbfs_get_inode).
		 */
4749
		resv_map = inode_resv_map(inode);
4750

4751
		chg = region_chg(resv_map, from, to);
4752 4753

	} else {
4754
		/* Private mapping. */
4755
		resv_map = resv_map_alloc();
4756 4757 4758
		if (!resv_map)
			return -ENOMEM;

4759
		chg = to - from;
4760

4761 4762 4763 4764 4765 4766 4767 4768 4769 4770 4771 4772 4773
		if (hugetlb_cgroup_charge_cgroup_rsvd(
			    hstate_index(h), chg * pages_per_huge_page(h),
			    &h_cg)) {
			kref_put(&resv_map->refs, resv_map_release);
			return -ENOMEM;
		}

		/*
		 * Since this branch handles private mappings, we attach the
		 * counter to uncharge for this reservation off resv_map.
		 */
		resv_map_set_hugetlb_cgroup_uncharge_info(resv_map, h_cg, h);

4774 4775 4776 4777
		set_vma_resv_map(vma, resv_map);
		set_vma_resv_flags(vma, HPAGE_RESV_OWNER);
	}

4778 4779 4780 4781
	if (chg < 0) {
		ret = chg;
		goto out_err;
	}
4782

4783 4784 4785 4786 4787 4788 4789
	/*
	 * 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) {
4790 4791 4792
		ret = -ENOSPC;
		goto out_err;
	}
4793 4794

	/*
4795
	 * Check enough hugepages are available for the reservation.
4796
	 * Hand the pages back to the subpool if there are not
4797
	 */
4798
	ret = hugetlb_acct_memory(h, gbl_reserve);
K
Ken Chen 已提交
4799
	if (ret < 0) {
4800 4801
		/* put back original number of pages, chg */
		(void)hugepage_subpool_put_pages(spool, chg);
4802
		goto out_err;
K
Ken Chen 已提交
4803
	}
4804 4805 4806 4807 4808 4809 4810 4811 4812 4813 4814 4815

	/*
	 * 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
	 */
4816 4817 4818 4819 4820 4821 4822 4823 4824 4825 4826 4827 4828 4829 4830 4831 4832 4833
	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);
		}
	}
4834
	return 0;
4835
out_err:
4836
	if (!vma || vma->vm_flags & VM_MAYSHARE)
4837 4838 4839
		/* Don't call region_abort if region_chg failed */
		if (chg >= 0)
			region_abort(resv_map, from, to);
J
Joonsoo Kim 已提交
4840 4841
	if (vma && is_vma_resv_set(vma, HPAGE_RESV_OWNER))
		kref_put(&resv_map->refs, resv_map_release);
4842
	return ret;
4843 4844
}

4845 4846
long hugetlb_unreserve_pages(struct inode *inode, long start, long end,
								long freed)
4847
{
4848
	struct hstate *h = hstate_inode(inode);
4849
	struct resv_map *resv_map = inode_resv_map(inode);
4850
	long chg = 0;
4851
	struct hugepage_subpool *spool = subpool_inode(inode);
4852
	long gbl_reserve;
K
Ken Chen 已提交
4853

4854 4855 4856 4857
	/*
	 * Since this routine can be called in the evict inode path for all
	 * hugetlbfs inodes, resv_map could be NULL.
	 */
4858 4859 4860 4861 4862 4863 4864 4865 4866 4867 4868
	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
4869
	spin_lock(&inode->i_lock);
4870
	inode->i_blocks -= (blocks_per_huge_page(h) * freed);
K
Ken Chen 已提交
4871 4872
	spin_unlock(&inode->i_lock);

4873 4874 4875 4876 4877 4878
	/*
	 * 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);
4879 4880

	return 0;
4881
}
4882

4883 4884 4885 4886 4887 4888 4889 4890 4891 4892 4893
#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
4894 4895
	unsigned long vm_flags = vma->vm_flags & VM_LOCKED_CLEAR_MASK;
	unsigned long svm_flags = svma->vm_flags & VM_LOCKED_CLEAR_MASK;
4896 4897 4898 4899 4900 4901 4902 4903 4904 4905 4906 4907 4908

	/*
	 * 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;
}

4909
static bool vma_shareable(struct vm_area_struct *vma, unsigned long addr)
4910 4911 4912 4913 4914 4915 4916
{
	unsigned long base = addr & PUD_MASK;
	unsigned long end = base + PUD_SIZE;

	/*
	 * check on proper vm_flags and page table alignment
	 */
4917
	if (vma->vm_flags & VM_MAYSHARE && range_in_vma(vma, base, end))
4918 4919
		return true;
	return false;
4920 4921
}

4922 4923 4924 4925 4926 4927 4928 4929 4930 4931 4932 4933 4934 4935 4936 4937 4938 4939 4940 4941 4942 4943 4944 4945 4946 4947 4948 4949 4950
/*
 * Determine if start,end range within vma could be mapped by shared pmd.
 * If yes, adjust start and end to cover range associated with possible
 * shared pmd mappings.
 */
void adjust_range_if_pmd_sharing_possible(struct vm_area_struct *vma,
				unsigned long *start, unsigned long *end)
{
	unsigned long check_addr = *start;

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

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

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

4951 4952 4953 4954
/*
 * 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
4955 4956 4957 4958 4959 4960
 * code much cleaner.
 *
 * This routine must be called with i_mmap_rwsem held in at least read mode.
 * For hugetlbfs, this prevents removal of any page table entries associated
 * with the address space.  This is important as we are setting up sharing
 * based on existing page table entries (mappings).
4961 4962 4963 4964 4965 4966 4967 4968 4969 4970 4971
 */
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;
4972
	spinlock_t *ptl;
4973 4974 4975 4976 4977 4978 4979 4980 4981 4982

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

	vma_interval_tree_foreach(svma, &mapping->i_mmap, idx, idx) {
		if (svma == vma)
			continue;

		saddr = page_table_shareable(svma, vma, addr, idx);
		if (saddr) {
4983 4984
			spte = huge_pte_offset(svma->vm_mm, saddr,
					       vma_mmu_pagesize(svma));
4985 4986 4987 4988 4989 4990 4991 4992 4993 4994
			if (spte) {
				get_page(virt_to_page(spte));
				break;
			}
		}
	}

	if (!spte)
		goto out;

4995
	ptl = huge_pte_lock(hstate_vma(vma), mm, spte);
4996
	if (pud_none(*pud)) {
4997 4998
		pud_populate(mm, pud,
				(pmd_t *)((unsigned long)spte & PAGE_MASK));
4999
		mm_inc_nr_pmds(mm);
5000
	} else {
5001
		put_page(virt_to_page(spte));
5002
	}
5003
	spin_unlock(ptl);
5004 5005 5006 5007 5008 5009 5010 5011 5012 5013 5014 5015
out:
	pte = (pte_t *)pmd_alloc(mm, pud, addr);
	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.
 *
5016
 * Called with page table lock held and i_mmap_rwsem held in write mode.
5017 5018 5019 5020 5021 5022 5023
 *
 * 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);
5024 5025
	p4d_t *p4d = p4d_offset(pgd, *addr);
	pud_t *pud = pud_offset(p4d, *addr);
5026 5027 5028 5029 5030 5031 5032

	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));
5033
	mm_dec_nr_pmds(mm);
5034 5035 5036
	*addr = ALIGN(*addr, HPAGE_SIZE * PTRS_PER_PTE) - HPAGE_SIZE;
	return 1;
}
5037 5038 5039 5040 5041 5042
#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;
}
5043 5044 5045 5046 5047

int huge_pmd_unshare(struct mm_struct *mm, unsigned long *addr, pte_t *ptep)
{
	return 0;
}
5048 5049 5050 5051 5052

void adjust_range_if_pmd_sharing_possible(struct vm_area_struct *vma,
				unsigned long *start, unsigned long *end)
{
}
5053
#define want_pmd_share()	(0)
5054 5055
#endif /* CONFIG_ARCH_WANT_HUGE_PMD_SHARE */

5056 5057 5058 5059 5060
#ifdef CONFIG_ARCH_WANT_GENERAL_HUGETLB
pte_t *huge_pte_alloc(struct mm_struct *mm,
			unsigned long addr, unsigned long sz)
{
	pgd_t *pgd;
5061
	p4d_t *p4d;
5062 5063 5064 5065
	pud_t *pud;
	pte_t *pte = NULL;

	pgd = pgd_offset(mm, addr);
5066 5067 5068
	p4d = p4d_alloc(mm, pgd, addr);
	if (!p4d)
		return NULL;
5069
	pud = pud_alloc(mm, p4d, addr);
5070 5071 5072 5073 5074 5075 5076 5077 5078 5079 5080
	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);
		}
	}
5081
	BUG_ON(pte && pte_present(*pte) && !pte_huge(*pte));
5082 5083 5084 5085

	return pte;
}

5086 5087 5088 5089 5090 5091 5092 5093 5094
/*
 * huge_pte_offset() - Walk the page table to resolve the hugepage
 * entry at address @addr
 *
 * Return: Pointer to page table or swap entry (PUD or PMD) for
 * address @addr, or NULL if a p*d_none() entry is encountered and the
 * size @sz doesn't match the hugepage size at this level of the page
 * table.
 */
5095 5096
pte_t *huge_pte_offset(struct mm_struct *mm,
		       unsigned long addr, unsigned long sz)
5097 5098
{
	pgd_t *pgd;
5099
	p4d_t *p4d;
5100
	pud_t *pud;
5101
	pmd_t *pmd;
5102 5103

	pgd = pgd_offset(mm, addr);
5104 5105 5106 5107 5108
	if (!pgd_present(*pgd))
		return NULL;
	p4d = p4d_offset(pgd, addr);
	if (!p4d_present(*p4d))
		return NULL;
5109

5110
	pud = pud_offset(p4d, addr);
5111
	if (sz != PUD_SIZE && pud_none(*pud))
5112
		return NULL;
5113 5114
	/* hugepage or swap? */
	if (pud_huge(*pud) || !pud_present(*pud))
5115
		return (pte_t *)pud;
5116

5117
	pmd = pmd_offset(pud, addr);
5118 5119 5120 5121 5122 5123 5124
	if (sz != PMD_SIZE && pmd_none(*pmd))
		return NULL;
	/* hugepage or swap? */
	if (pmd_huge(*pmd) || !pmd_present(*pmd))
		return (pte_t *)pmd;

	return NULL;
5125 5126
}

5127 5128 5129 5130 5131 5132 5133 5134 5135 5136 5137 5138 5139
#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);
}

5140 5141 5142 5143 5144 5145 5146 5147
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;
}

5148
struct page * __weak
5149
follow_huge_pmd(struct mm_struct *mm, unsigned long address,
5150
		pmd_t *pmd, int flags)
5151
{
5152 5153
	struct page *page = NULL;
	spinlock_t *ptl;
5154
	pte_t pte;
J
John Hubbard 已提交
5155 5156 5157 5158 5159 5160

	/* FOLL_GET and FOLL_PIN are mutually exclusive. */
	if (WARN_ON_ONCE((flags & (FOLL_PIN | FOLL_GET)) ==
			 (FOLL_PIN | FOLL_GET)))
		return NULL;

5161 5162 5163 5164 5165 5166 5167 5168 5169
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;
5170 5171
	pte = huge_ptep_get((pte_t *)pmd);
	if (pte_present(pte)) {
5172
		page = pmd_page(*pmd) + ((address & ~PMD_MASK) >> PAGE_SHIFT);
J
John Hubbard 已提交
5173 5174 5175 5176 5177 5178 5179 5180 5181 5182 5183 5184
		/*
		 * try_grab_page() should always succeed here, because: a) we
		 * hold the pmd (ptl) lock, and b) we've just checked that the
		 * huge pmd (head) page is present in the page tables. The ptl
		 * prevents the head page and tail pages from being rearranged
		 * in any way. So this page must be available at this point,
		 * unless the page refcount overflowed:
		 */
		if (WARN_ON_ONCE(!try_grab_page(page, flags))) {
			page = NULL;
			goto out;
		}
5185
	} else {
5186
		if (is_hugetlb_entry_migration(pte)) {
5187 5188 5189 5190 5191 5192 5193 5194 5195 5196 5197
			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);
5198 5199 5200
	return page;
}

5201
struct page * __weak
5202
follow_huge_pud(struct mm_struct *mm, unsigned long address,
5203
		pud_t *pud, int flags)
5204
{
J
John Hubbard 已提交
5205
	if (flags & (FOLL_GET | FOLL_PIN))
5206
		return NULL;
5207

5208
	return pte_page(*(pte_t *)pud) + ((address & ~PUD_MASK) >> PAGE_SHIFT);
5209 5210
}

5211 5212 5213
struct page * __weak
follow_huge_pgd(struct mm_struct *mm, unsigned long address, pgd_t *pgd, int flags)
{
J
John Hubbard 已提交
5214
	if (flags & (FOLL_GET | FOLL_PIN))
5215 5216 5217 5218 5219
		return NULL;

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

5220 5221
bool isolate_huge_page(struct page *page, struct list_head *list)
{
5222 5223
	bool ret = true;

5224
	VM_BUG_ON_PAGE(!PageHead(page), page);
5225
	spin_lock(&hugetlb_lock);
5226 5227 5228 5229 5230
	if (!page_huge_active(page) || !get_page_unless_zero(page)) {
		ret = false;
		goto unlock;
	}
	clear_page_huge_active(page);
5231
	list_move_tail(&page->lru, list);
5232
unlock:
5233
	spin_unlock(&hugetlb_lock);
5234
	return ret;
5235 5236 5237 5238
}

void putback_active_hugepage(struct page *page)
{
5239
	VM_BUG_ON_PAGE(!PageHead(page), page);
5240
	spin_lock(&hugetlb_lock);
5241
	set_page_huge_active(page);
5242 5243 5244 5245
	list_move_tail(&page->lru, &(page_hstate(page))->hugepage_activelist);
	spin_unlock(&hugetlb_lock);
	put_page(page);
}
5246 5247 5248 5249 5250 5251 5252 5253 5254 5255 5256 5257 5258 5259 5260 5261 5262 5263 5264 5265 5266 5267 5268 5269 5270 5271 5272 5273 5274 5275 5276 5277 5278

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

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

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

		SetPageHugeTemporary(oldpage);
		ClearPageHugeTemporary(newpage);

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