hugetlb.c 156.3 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/sched/mm.h>
23
#include <linux/mmdebug.h>
24
#include <linux/sched/signal.h>
25
#include <linux/rmap.h>
26
#include <linux/string_helpers.h>
27 28
#include <linux/swap.h>
#include <linux/swapops.h>
29
#include <linux/jhash.h>
30
#include <linux/numa.h>
31
#include <linux/llist.h>
32
#include <linux/cma.h>
33

D
David Gibson 已提交
34
#include <asm/page.h>
35
#include <asm/pgalloc.h>
36
#include <asm/tlb.h>
D
David Gibson 已提交
37

38
#include <linux/io.h>
D
David Gibson 已提交
39
#include <linux/hugetlb.h>
40
#include <linux/hugetlb_cgroup.h>
41
#include <linux/node.h>
42
#include <linux/userfaultfd_k.h>
43
#include <linux/page_owner.h>
44
#include "internal.h"
L
Linus Torvalds 已提交
45

46
int hugetlb_max_hstate __read_mostly;
47 48
unsigned int default_hstate_idx;
struct hstate hstates[HUGE_MAX_HSTATE];
49

50
#ifdef CONFIG_CMA
51
static struct cma *hugetlb_cma[MAX_NUMNODES];
52 53
#endif
static unsigned long hugetlb_cma_size __initdata;
54

55 56 57 58 59
/*
 * Minimum page order among possible hugepage sizes, set to a proper value
 * at boot time.
 */
static unsigned int minimum_order __read_mostly = UINT_MAX;
60

61 62
__initdata LIST_HEAD(huge_boot_pages);

63 64 65
/* for command line parsing */
static struct hstate * __initdata parsed_hstate;
static unsigned long __initdata default_hstate_max_huge_pages;
66
static bool __initdata parsed_valid_hugepagesz = true;
67
static bool __initdata parsed_default_hugepagesz;
68

69
/*
70 71
 * Protects updates to hugepage_freelists, hugepage_activelist, nr_huge_pages,
 * free_huge_pages, and surplus_huge_pages.
72
 */
73
DEFINE_SPINLOCK(hugetlb_lock);
74

75 76 77 78 79
/*
 * 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;
80
struct mutex *hugetlb_fault_mutex_table ____cacheline_aligned_in_smp;
81

82 83 84 85 86 87 88 89 90 91 92 93 94 95 96
static inline bool PageHugeFreed(struct page *head)
{
	return page_private(head + 4) == -1UL;
}

static inline void SetPageHugeFreed(struct page *head)
{
	set_page_private(head + 4, -1UL);
}

static inline void ClearPageHugeFreed(struct page *head)
{
	set_page_private(head + 4, 0);
}

97 98 99
/* Forward declaration */
static int hugetlb_acct_memory(struct hstate *h, long delta);

100 101 102 103 104 105 106
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
E
Ethon Paul 已提交
107
	 * remain, give up any reservations based on minimum size and
108 109 110 111 112
	 * free the subpool */
	if (free) {
		if (spool->min_hpages != -1)
			hugetlb_acct_memory(spool->hstate,
						-spool->min_hpages);
113
		kfree(spool);
114
	}
115 116
}

117 118
struct hugepage_subpool *hugepage_new_subpool(struct hstate *h, long max_hpages,
						long min_hpages)
119 120 121
{
	struct hugepage_subpool *spool;

122
	spool = kzalloc(sizeof(*spool), GFP_KERNEL);
123 124 125 126 127
	if (!spool)
		return NULL;

	spin_lock_init(&spool->lock);
	spool->count = 1;
128 129 130 131 132 133 134 135 136
	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;
137 138 139 140 141 142 143 144 145 146 147 148

	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);
}

149 150 151
/*
 * Subpool accounting for allocating and reserving pages.
 * Return -ENOMEM if there are not enough resources to satisfy the
152
 * request.  Otherwise, return the number of pages by which the
153 154
 * global pools must be adjusted (upward).  The returned value may
 * only be different than the passed value (delta) in the case where
E
Ethon Paul 已提交
155
 * a subpool minimum size must be maintained.
156 157
 */
static long hugepage_subpool_get_pages(struct hugepage_subpool *spool,
158 159
				      long delta)
{
160
	long ret = delta;
161 162

	if (!spool)
163
		return ret;
164 165

	spin_lock(&spool->lock);
166 167 168 169 170 171 172 173

	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;
		}
174 175
	}

176 177
	/* minimum size accounting */
	if (spool->min_hpages != -1 && spool->rsv_hpages) {
178 179 180 181 182 183 184 185 186 187 188 189 190 191 192
		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);
193 194 195
	return ret;
}

196 197 198 199 200 201 202
/*
 * 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,
203 204
				       long delta)
{
205 206
	long ret = delta;

207
	if (!spool)
208
		return delta;
209 210

	spin_lock(&spool->lock);
211 212 213 214

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

215 216
	 /* minimum size accounting */
	if (spool->min_hpages != -1 && spool->used_hpages < spool->min_hpages) {
217 218 219 220 221 222 223 224 225 226 227 228 229 230
		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.
	 */
231
	unlock_or_release_subpool(spool);
232 233

	return ret;
234 235 236 237 238 239 240 241 242
}

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
243
	return subpool_inode(file_inode(vma->vm_file));
244 245
}

246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265
/* Helper that removes a struct file_region from the resv_map cache and returns
 * it for use.
 */
static struct file_region *
get_file_region_entry_from_cache(struct resv_map *resv, long from, long to)
{
	struct file_region *nrg = NULL;

	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 = from;
	nrg->to = to;

	return nrg;
}

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
static void copy_hugetlb_cgroup_uncharge_info(struct file_region *nrg,
					      struct file_region *rg)
{
#ifdef CONFIG_CGROUP_HUGETLB
	nrg->reservation_counter = rg->reservation_counter;
	nrg->css = rg->css;
	if (rg->css)
		css_get(rg->css);
#endif
}

/* Helper that records hugetlb_cgroup uncharge info. */
static void record_hugetlb_cgroup_uncharge_info(struct hugetlb_cgroup *h_cg,
						struct hstate *h,
						struct resv_map *resv,
						struct file_region *nrg)
{
#ifdef CONFIG_CGROUP_HUGETLB
	if (h_cg) {
		nrg->reservation_counter =
			&h_cg->rsvd_hugepage[hstate_index(h)];
		nrg->css = &h_cg->css;
		if (!resv->pages_per_hpage)
			resv->pages_per_hpage = pages_per_huge_page(h);
		/* pages_per_hpage should be the same for all entries in
		 * a resv_map.
		 */
		VM_BUG_ON(resv->pages_per_hpage != pages_per_huge_page(h));
	} else {
		nrg->reservation_counter = NULL;
		nrg->css = NULL;
	}
#endif
}

301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325
static bool has_same_uncharge_info(struct file_region *rg,
				   struct file_region *org)
{
#ifdef CONFIG_CGROUP_HUGETLB
	return rg && org &&
	       rg->reservation_counter == org->reservation_counter &&
	       rg->css == org->css;

#else
	return true;
#endif
}

static void coalesce_file_region(struct resv_map *resv, struct file_region *rg)
{
	struct file_region *nrg = NULL, *prg = NULL;

	prg = list_prev_entry(rg, link);
	if (&prg->link != &resv->regions && prg->to == rg->from &&
	    has_same_uncharge_info(prg, rg)) {
		prg->to = rg->to;

		list_del(&rg->link);
		kfree(rg);

326
		rg = prg;
327 328 329 330 331 332 333 334 335 336 337 338
	}

	nrg = list_next_entry(rg, link);
	if (&nrg->link != &resv->regions && nrg->from == rg->to &&
	    has_same_uncharge_info(nrg, rg)) {
		nrg->from = rg->from;

		list_del(&rg->link);
		kfree(rg);
	}
}

339 340 341 342 343 344 345
/*
 * Must be called with resv->lock held.
 *
 * Calling this with regions_needed != NULL will count the number of pages
 * to be added but will not modify the linked list. And regions_needed will
 * indicate the number of file_regions needed in the cache to carry out to add
 * the regions for this range.
M
Mina Almasry 已提交
346 347
 */
static long add_reservation_in_range(struct resv_map *resv, long f, long t,
348
				     struct hugetlb_cgroup *h_cg,
349
				     struct hstate *h, long *regions_needed)
M
Mina Almasry 已提交
350
{
351
	long add = 0;
M
Mina Almasry 已提交
352
	struct list_head *head = &resv->regions;
353
	long last_accounted_offset = f;
M
Mina Almasry 已提交
354 355
	struct file_region *rg = NULL, *trg = NULL, *nrg = NULL;

356 357
	if (regions_needed)
		*regions_needed = 0;
M
Mina Almasry 已提交
358

359 360 361 362 363 364 365 366 367 368 369 370 371 372
	/* In this loop, we essentially handle an entry for the range
	 * [last_accounted_offset, rg->from), at every iteration, with some
	 * bounds checking.
	 */
	list_for_each_entry_safe(rg, trg, head, link) {
		/* Skip irrelevant regions that start before our range. */
		if (rg->from < f) {
			/* If this region ends after the last accounted offset,
			 * then we need to update last_accounted_offset.
			 */
			if (rg->to > last_accounted_offset)
				last_accounted_offset = rg->to;
			continue;
		}
M
Mina Almasry 已提交
373

374 375 376
		/* When we find a region that starts beyond our range, we've
		 * finished.
		 */
M
Mina Almasry 已提交
377 378 379
		if (rg->from > t)
			break;

380 381 382 383 384
		/* Add an entry for last_accounted_offset -> rg->from, and
		 * update last_accounted_offset.
		 */
		if (rg->from > last_accounted_offset) {
			add += rg->from - last_accounted_offset;
385
			if (!regions_needed) {
386 387
				nrg = get_file_region_entry_from_cache(
					resv, last_accounted_offset, rg->from);
388 389
				record_hugetlb_cgroup_uncharge_info(h_cg, h,
								    resv, nrg);
390
				list_add(&nrg->link, rg->link.prev);
391
				coalesce_file_region(resv, nrg);
392
			} else
393 394 395 396 397 398 399 400 401 402 403
				*regions_needed += 1;
		}

		last_accounted_offset = rg->to;
	}

	/* Handle the case where our range extends beyond
	 * last_accounted_offset.
	 */
	if (last_accounted_offset < t) {
		add += t - last_accounted_offset;
404
		if (!regions_needed) {
405 406
			nrg = get_file_region_entry_from_cache(
				resv, last_accounted_offset, t);
407
			record_hugetlb_cgroup_uncharge_info(h_cg, h, resv, nrg);
408
			list_add(&nrg->link, rg->link.prev);
409
			coalesce_file_region(resv, nrg);
410
		} else
411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448
			*regions_needed += 1;
	}

	VM_BUG_ON(add < 0);
	return add;
}

/* Must be called with resv->lock acquired. Will drop lock to allocate entries.
 */
static int allocate_file_region_entries(struct resv_map *resv,
					int regions_needed)
	__must_hold(&resv->lock)
{
	struct list_head allocated_regions;
	int to_allocate = 0, i = 0;
	struct file_region *trg = NULL, *rg = NULL;

	VM_BUG_ON(regions_needed < 0);

	INIT_LIST_HEAD(&allocated_regions);

	/*
	 * Check for sufficient descriptors in the cache to accommodate
	 * the number of in progress add operations plus regions_needed.
	 *
	 * This is a while loop because when we drop the lock, some other call
	 * to region_add or region_del may have consumed some region_entries,
	 * so we keep looping here until we finally have enough entries for
	 * (adds_in_progress + regions_needed).
	 */
	while (resv->region_cache_count <
	       (resv->adds_in_progress + regions_needed)) {
		to_allocate = resv->adds_in_progress + regions_needed -
			      resv->region_cache_count;

		/* At this point, we should have enough entries in the cache
		 * for all the existings adds_in_progress. We should only be
		 * needing to allocate for regions_needed.
M
Mina Almasry 已提交
449
		 */
450 451 452 453 454 455 456 457
		VM_BUG_ON(resv->region_cache_count < resv->adds_in_progress);

		spin_unlock(&resv->lock);
		for (i = 0; i < to_allocate; i++) {
			trg = kmalloc(sizeof(*trg), GFP_KERNEL);
			if (!trg)
				goto out_of_memory;
			list_add(&trg->link, &allocated_regions);
M
Mina Almasry 已提交
458 459
		}

460 461
		spin_lock(&resv->lock);

462 463
		list_splice(&allocated_regions, &resv->region_cache);
		resv->region_cache_count += to_allocate;
M
Mina Almasry 已提交
464 465
	}

466
	return 0;
M
Mina Almasry 已提交
467

468 469 470 471 472 473
out_of_memory:
	list_for_each_entry_safe(rg, trg, &allocated_regions, link) {
		list_del(&rg->link);
		kfree(rg);
	}
	return -ENOMEM;
M
Mina Almasry 已提交
474 475
}

476 477
/*
 * Add the huge page range represented by [f, t) to the reserve
478 479 480 481 482
 * map.  Regions will be taken from the cache to fill in this range.
 * Sufficient regions should exist in the cache due to the previous
 * call to region_chg with the same range, but in some cases the cache will not
 * have sufficient entries due to races with other code doing region_add or
 * region_del.  The extra needed entries will be allocated.
483
 *
484 485 486 487
 * regions_needed is the out value provided by a previous call to region_chg.
 *
 * Return the number of new huge pages added to the map.  This number is greater
 * than or equal to zero.  If file_region entries needed to be allocated for
E
Ethon Paul 已提交
488
 * this operation and we were not able to allocate, it returns -ENOMEM.
489 490 491
 * region_add of regions of length 1 never allocate file_regions and cannot
 * fail; region_chg will always allocate at least 1 entry and a region_add for
 * 1 page will only require at most 1 entry.
492
 */
493
static long region_add(struct resv_map *resv, long f, long t,
494 495
		       long in_regions_needed, struct hstate *h,
		       struct hugetlb_cgroup *h_cg)
496
{
497
	long add = 0, actual_regions_needed = 0;
498

499
	spin_lock(&resv->lock);
500 501 502
retry:

	/* Count how many regions are actually needed to execute this add. */
503 504
	add_reservation_in_range(resv, f, t, NULL, NULL,
				 &actual_regions_needed);
505

506
	/*
507 508 509 510 511 512 513
	 * Check for sufficient descriptors in the cache to accommodate
	 * this add operation. Note that actual_regions_needed may be greater
	 * than in_regions_needed, as the resv_map may have been modified since
	 * the region_chg call. In this case, we need to make sure that we
	 * allocate extra entries, such that we have enough for all the
	 * existing adds_in_progress, plus the excess needed for this
	 * operation.
514
	 */
515 516 517 518 519 520 521 522
	if (actual_regions_needed > in_regions_needed &&
	    resv->region_cache_count <
		    resv->adds_in_progress +
			    (actual_regions_needed - in_regions_needed)) {
		/* region_add operation of range 1 should never need to
		 * allocate file_region entries.
		 */
		VM_BUG_ON(t - f <= 1);
523

524 525 526 527
		if (allocate_file_region_entries(
			    resv, actual_regions_needed - in_regions_needed)) {
			return -ENOMEM;
		}
528

529
		goto retry;
530 531
	}

532
	add = add_reservation_in_range(resv, f, t, h_cg, h, NULL);
533 534

	resv->adds_in_progress -= in_regions_needed;
535

536
	spin_unlock(&resv->lock);
537 538
	VM_BUG_ON(add < 0);
	return add;
539 540
}

541 542 543 544 545 546 547
/*
 * 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
548 549 550 551 552 553 554
 * map.  A number of new file_region structures is added to the cache as a
 * placeholder, for the subsequent region_add call to use. At least 1
 * file_region structure is added.
 *
 * out_regions_needed is the number of regions added to the
 * resv->adds_in_progress.  This value needs to be provided to a follow up call
 * to region_add or region_abort for proper accounting.
555 556 557 558 559
 *
 * 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.
560
 */
561 562
static long region_chg(struct resv_map *resv, long f, long t,
		       long *out_regions_needed)
563 564 565
{
	long chg = 0;

566
	spin_lock(&resv->lock);
567

568
	/* Count how many hugepages in this range are NOT represented. */
569
	chg = add_reservation_in_range(resv, f, t, NULL, NULL,
570
				       out_regions_needed);
571

572 573
	if (*out_regions_needed == 0)
		*out_regions_needed = 1;
574

575 576
	if (allocate_file_region_entries(resv, *out_regions_needed))
		return -ENOMEM;
577

578
	resv->adds_in_progress += *out_regions_needed;
579 580

	spin_unlock(&resv->lock);
581 582 583
	return chg;
}

584 585 586 587 588
/*
 * 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
589 590 591
 * is called to decrement the adds_in_progress counter. regions_needed
 * is the value returned by the region_chg call, it is used to decrement
 * the adds_in_progress counter.
592 593 594 595 596
 *
 * 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.
 */
597 598
static void region_abort(struct resv_map *resv, long f, long t,
			 long regions_needed)
599 600 601
{
	spin_lock(&resv->lock);
	VM_BUG_ON(!resv->region_cache_count);
602
	resv->adds_in_progress -= regions_needed;
603 604 605
	spin_unlock(&resv->lock);
}

606
/*
607 608 609 610 611 612 613 614 615 616 617 618
 * 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.
619
 */
620
static long region_del(struct resv_map *resv, long f, long t)
621
{
622
	struct list_head *head = &resv->regions;
623
	struct file_region *rg, *trg;
624 625
	struct file_region *nrg = NULL;
	long del = 0;
626

627
retry:
628
	spin_lock(&resv->lock);
629
	list_for_each_entry_safe(rg, trg, head, link) {
630 631 632 633 634 635 636 637
		/*
		 * 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))
638
			continue;
639

640
		if (rg->from >= t)
641 642
			break;

643 644 645 646 647 648 649 650 651 652 653 654 655
		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--;
			}
656

657 658 659 660 661 662 663 664 665
			if (!nrg) {
				spin_unlock(&resv->lock);
				nrg = kmalloc(sizeof(*nrg), GFP_KERNEL);
				if (!nrg)
					return -ENOMEM;
				goto retry;
			}

			del += t - f;
666 667
			hugetlb_cgroup_uncharge_file_region(
				resv, rg, t - f);
668 669 670 671

			/* New entry for end of split region */
			nrg->from = t;
			nrg->to = rg->to;
672 673 674

			copy_hugetlb_cgroup_uncharge_info(nrg, rg);

675 676 677 678 679 680 681
			INIT_LIST_HEAD(&nrg->link);

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

			list_add(&nrg->link, &rg->link);
			nrg = NULL;
682
			break;
683 684 685 686
		}

		if (f <= rg->from && t >= rg->to) { /* Remove entire region */
			del += rg->to - rg->from;
687 688
			hugetlb_cgroup_uncharge_file_region(resv, rg,
							    rg->to - rg->from);
689 690 691 692 693 694
			list_del(&rg->link);
			kfree(rg);
			continue;
		}

		if (f <= rg->from) {	/* Trim beginning of region */
695 696 697
			hugetlb_cgroup_uncharge_file_region(resv, rg,
							    t - rg->from);

698 699 700
			del += t - rg->from;
			rg->from = t;
		} else {		/* Trim end of region */
701 702
			hugetlb_cgroup_uncharge_file_region(resv, rg,
							    rg->to - f);
703 704 705

			del += rg->to - f;
			rg->to = f;
706
		}
707
	}
708 709

	spin_unlock(&resv->lock);
710 711
	kfree(nrg);
	return del;
712 713
}

714 715 716 717 718 719 720 721 722
/*
 * 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.
 */
723
void hugetlb_fix_reserve_counts(struct inode *inode)
724 725 726 727 728
{
	struct hugepage_subpool *spool = subpool_inode(inode);
	long rsv_adjust;

	rsv_adjust = hugepage_subpool_get_pages(spool, 1);
729
	if (rsv_adjust) {
730 731 732 733 734 735
		struct hstate *h = hstate_inode(inode);

		hugetlb_acct_memory(h, 1);
	}
}

736 737 738 739
/*
 * Count and return the number of huge pages in the reserve map
 * that intersect with the range [f, t).
 */
740
static long region_count(struct resv_map *resv, long f, long t)
741
{
742
	struct list_head *head = &resv->regions;
743 744 745
	struct file_region *rg;
	long chg = 0;

746
	spin_lock(&resv->lock);
747 748
	/* Locate each segment we overlap with, and count that overlap. */
	list_for_each_entry(rg, head, link) {
749 750
		long seg_from;
		long seg_to;
751 752 753 754 755 756 757 758 759 760 761

		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;
	}
762
	spin_unlock(&resv->lock);
763 764 765 766

	return chg;
}

767 768 769 770
/*
 * Convert the address within this vma to the page offset within
 * the mapping, in pagecache page units; huge pages here.
 */
771 772
static pgoff_t vma_hugecache_offset(struct hstate *h,
			struct vm_area_struct *vma, unsigned long address)
773
{
774 775
	return ((address - vma->vm_start) >> huge_page_shift(h)) +
			(vma->vm_pgoff >> huge_page_order(h));
776 777
}

778 779 780 781 782
pgoff_t linear_hugepage_index(struct vm_area_struct *vma,
				     unsigned long address)
{
	return vma_hugecache_offset(hstate_vma(vma), vma, address);
}
783
EXPORT_SYMBOL_GPL(linear_hugepage_index);
784

785 786 787 788 789 790
/*
 * 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)
{
791 792 793
	if (vma->vm_ops && vma->vm_ops->pagesize)
		return vma->vm_ops->pagesize(vma);
	return PAGE_SIZE;
794
}
795
EXPORT_SYMBOL_GPL(vma_kernel_pagesize);
796

797 798 799
/*
 * 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
800 801
 * architectures where it differs, an architecture-specific 'strong'
 * version of this symbol is required.
802
 */
803
__weak unsigned long vma_mmu_pagesize(struct vm_area_struct *vma)
804 805 806 807
{
	return vma_kernel_pagesize(vma);
}

808 809 810 811 812 813 814
/*
 * 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)
815
#define HPAGE_RESV_MASK (HPAGE_RESV_OWNER | HPAGE_RESV_UNMAPPED)
816

817 818 819 820 821 822 823 824 825
/*
 * 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.
826 827 828 829 830 831 832 833 834
 *
 * 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.
835
 */
836 837 838 839 840 841 842 843 844 845 846
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;
}

847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865
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
}

866
struct resv_map *resv_map_alloc(void)
867 868
{
	struct resv_map *resv_map = kmalloc(sizeof(*resv_map), GFP_KERNEL);
869 870 871 872 873
	struct file_region *rg = kmalloc(sizeof(*rg), GFP_KERNEL);

	if (!resv_map || !rg) {
		kfree(resv_map);
		kfree(rg);
874
		return NULL;
875
	}
876 877

	kref_init(&resv_map->refs);
878
	spin_lock_init(&resv_map->lock);
879 880
	INIT_LIST_HEAD(&resv_map->regions);

881
	resv_map->adds_in_progress = 0;
882 883 884 885 886 887 888
	/*
	 * 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);
889 890 891 892 893

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

894 895 896
	return resv_map;
}

897
void resv_map_release(struct kref *ref)
898 899
{
	struct resv_map *resv_map = container_of(ref, struct resv_map, refs);
900 901
	struct list_head *head = &resv_map->region_cache;
	struct file_region *rg, *trg;
902 903

	/* Clear out any active regions before we release the map. */
904
	region_del(resv_map, 0, LONG_MAX);
905 906 907 908 909 910 911 912 913

	/* ... 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);

914 915 916
	kfree(resv_map);
}

917 918
static inline struct resv_map *inode_resv_map(struct inode *inode)
{
919 920 921 922 923 924 925 926 927
	/*
	 * 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;
928 929
}

930
static struct resv_map *vma_resv_map(struct vm_area_struct *vma)
931
{
932
	VM_BUG_ON_VMA(!is_vm_hugetlb_page(vma), vma);
933 934 935 936 937 938 939
	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 {
940 941
		return (struct resv_map *)(get_vma_private_data(vma) &
							~HPAGE_RESV_MASK);
942
	}
943 944
}

945
static void set_vma_resv_map(struct vm_area_struct *vma, struct resv_map *map)
946
{
947 948
	VM_BUG_ON_VMA(!is_vm_hugetlb_page(vma), vma);
	VM_BUG_ON_VMA(vma->vm_flags & VM_MAYSHARE, vma);
949

950 951
	set_vma_private_data(vma, (get_vma_private_data(vma) &
				HPAGE_RESV_MASK) | (unsigned long)map);
952 953 954 955
}

static void set_vma_resv_flags(struct vm_area_struct *vma, unsigned long flags)
{
956 957
	VM_BUG_ON_VMA(!is_vm_hugetlb_page(vma), vma);
	VM_BUG_ON_VMA(vma->vm_flags & VM_MAYSHARE, vma);
958 959

	set_vma_private_data(vma, get_vma_private_data(vma) | flags);
960 961 962 963
}

static int is_vma_resv_set(struct vm_area_struct *vma, unsigned long flag)
{
964
	VM_BUG_ON_VMA(!is_vm_hugetlb_page(vma), vma);
965 966

	return (get_vma_private_data(vma) & flag) != 0;
967 968
}

969
/* Reset counters to 0 and clear all HPAGE_RESV_* flags */
970 971
void reset_vma_resv_huge_pages(struct vm_area_struct *vma)
{
972
	VM_BUG_ON_VMA(!is_vm_hugetlb_page(vma), vma);
973
	if (!(vma->vm_flags & VM_MAYSHARE))
974 975 976 977
		vma->vm_private_data = (void *)0;
}

/* Returns true if the VMA has associated reserve pages */
978
static bool vma_has_reserves(struct vm_area_struct *vma, long chg)
979
{
980 981 982 983 984 985 986 987 988 989 990
	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)
991
			return true;
992
		else
993
			return false;
994
	}
995 996

	/* Shared mappings always use reserves */
997 998 999 1000 1001
	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
E
Ethon Paul 已提交
1002
		 * fallocate.  In this case, there really are no reserves to
1003 1004 1005 1006 1007 1008 1009
		 * use.  This situation is indicated if chg != 0.
		 */
		if (chg)
			return false;
		else
			return true;
	}
1010 1011 1012 1013 1014

	/*
	 * Only the process that called mmap() has reserves for
	 * private mappings.
	 */
1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035
	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;
	}
1036

1037
	return false;
1038 1039
}

1040
static void enqueue_huge_page(struct hstate *h, struct page *page)
L
Linus Torvalds 已提交
1041 1042
{
	int nid = page_to_nid(page);
1043
	list_move(&page->lru, &h->hugepage_freelists[nid]);
1044 1045
	h->free_huge_pages++;
	h->free_huge_pages_node[nid]++;
1046
	SetPageHugeFreed(page);
L
Linus Torvalds 已提交
1047 1048
}

1049
static struct page *dequeue_huge_page_node_exact(struct hstate *h, int nid)
1050 1051
{
	struct page *page;
1052 1053 1054 1055 1056
	bool nocma = !!(current->flags & PF_MEMALLOC_NOCMA);

	list_for_each_entry(page, &h->hugepage_freelists[nid], lru) {
		if (nocma && is_migrate_cma_page(page))
			continue;
1057

1058 1059 1060 1061 1062
		if (PageHWPoison(page))
			continue;

		list_move(&page->lru, &h->hugepage_activelist);
		set_page_refcounted(page);
1063
		ClearPageHugeFreed(page);
1064 1065 1066
		h->free_huge_pages--;
		h->free_huge_pages_node[nid]--;
		return page;
1067 1068
	}

1069
	return NULL;
1070 1071
}

1072 1073
static struct page *dequeue_huge_page_nodemask(struct hstate *h, gfp_t gfp_mask, int nid,
		nodemask_t *nmask)
1074
{
1075 1076 1077 1078
	unsigned int cpuset_mems_cookie;
	struct zonelist *zonelist;
	struct zone *zone;
	struct zoneref *z;
1079
	int node = NUMA_NO_NODE;
1080

1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096
	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);
1097 1098 1099 1100 1101

		page = dequeue_huge_page_node_exact(h, node);
		if (page)
			return page;
	}
1102 1103 1104
	if (unlikely(read_mems_allowed_retry(cpuset_mems_cookie)))
		goto retry_cpuset;

1105 1106 1107
	return NULL;
}

1108 1109
static struct page *dequeue_huge_page_vma(struct hstate *h,
				struct vm_area_struct *vma,
1110 1111
				unsigned long address, int avoid_reserve,
				long chg)
L
Linus Torvalds 已提交
1112
{
1113
	struct page *page;
1114
	struct mempolicy *mpol;
1115
	gfp_t gfp_mask;
1116
	nodemask_t *nodemask;
1117
	int nid;
L
Linus Torvalds 已提交
1118

1119 1120 1121 1122 1123
	/*
	 * 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
	 */
1124
	if (!vma_has_reserves(vma, chg) &&
1125
			h->free_huge_pages - h->resv_huge_pages == 0)
1126
		goto err;
1127

1128
	/* If reserves cannot be used, ensure enough pages are in the pool */
1129
	if (avoid_reserve && h->free_huge_pages - h->resv_huge_pages == 0)
1130
		goto err;
1131

1132 1133
	gfp_mask = htlb_alloc_mask(h);
	nid = huge_node(vma, address, gfp_mask, &mpol, &nodemask);
1134 1135 1136 1137
	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
1138
	}
1139

1140
	mpol_cond_put(mpol);
L
Linus Torvalds 已提交
1141
	return page;
1142 1143 1144

err:
	return NULL;
L
Linus Torvalds 已提交
1145 1146
}

1147 1148 1149 1150 1151 1152 1153 1154 1155
/*
 * 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)
{
1156
	nid = next_node_in(nid, *nodes_allowed);
1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217
	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--)

1218
#ifdef CONFIG_ARCH_HAS_GIGANTIC_PAGE
1219
static void destroy_compound_gigantic_page(struct page *page,
1220
					unsigned int order)
1221 1222 1223 1224 1225
{
	int i;
	int nr_pages = 1 << order;
	struct page *p = page + 1;

1226
	atomic_set(compound_mapcount_ptr(page), 0);
1227 1228 1229
	if (hpage_pincount_available(page))
		atomic_set(compound_pincount_ptr(page), 0);

1230
	for (i = 1; i < nr_pages; i++, p = mem_map_next(p, page, i)) {
1231
		clear_compound_head(p);
1232 1233 1234 1235
		set_page_refcounted(p);
	}

	set_compound_order(page, 0);
1236
	page[1].compound_nr = 0;
1237 1238 1239
	__ClearPageHead(page);
}

1240
static void free_gigantic_page(struct page *page, unsigned int order)
1241
{
1242 1243 1244 1245
	/*
	 * If the page isn't allocated using the cma allocator,
	 * cma_release() returns false.
	 */
1246 1247
#ifdef CONFIG_CMA
	if (cma_release(hugetlb_cma[page_to_nid(page)], page, 1 << order))
1248
		return;
1249
#endif
1250

1251 1252 1253
	free_contig_range(page_to_pfn(page), 1 << order);
}

1254
#ifdef CONFIG_CONTIG_ALLOC
1255 1256
static struct page *alloc_gigantic_page(struct hstate *h, gfp_t gfp_mask,
		int nid, nodemask_t *nodemask)
1257
{
1258
	unsigned long nr_pages = 1UL << huge_page_order(h);
1259 1260
	if (nid == NUMA_NO_NODE)
		nid = numa_mem_id();
1261

1262 1263
#ifdef CONFIG_CMA
	{
1264 1265 1266
		struct page *page;
		int node;

1267 1268 1269
		if (hugetlb_cma[nid]) {
			page = cma_alloc(hugetlb_cma[nid], nr_pages,
					huge_page_order(h), true);
1270 1271 1272
			if (page)
				return page;
		}
1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284

		if (!(gfp_mask & __GFP_THISNODE)) {
			for_each_node_mask(node, *nodemask) {
				if (node == nid || !hugetlb_cma[node])
					continue;

				page = cma_alloc(hugetlb_cma[node], nr_pages,
						huge_page_order(h), true);
				if (page)
					return page;
			}
		}
1285
	}
1286
#endif
1287

1288
	return alloc_contig_pages(nr_pages, gfp_mask, nid, nodemask);
1289 1290 1291
}

static void prep_new_huge_page(struct hstate *h, struct page *page, int nid);
1292
static void prep_compound_gigantic_page(struct page *page, unsigned int order);
1293 1294 1295 1296 1297 1298 1299
#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 */
1300

1301
#else /* !CONFIG_ARCH_HAS_GIGANTIC_PAGE */
1302
static struct page *alloc_gigantic_page(struct hstate *h, gfp_t gfp_mask,
1303 1304 1305 1306
					int nid, nodemask_t *nodemask)
{
	return NULL;
}
1307
static inline void free_gigantic_page(struct page *page, unsigned int order) { }
1308
static inline void destroy_compound_gigantic_page(struct page *page,
1309
						unsigned int order) { }
1310 1311
#endif

1312
static void update_and_free_page(struct hstate *h, struct page *page)
A
Adam Litke 已提交
1313 1314
{
	int i;
1315
	struct page *subpage = page;
1316

1317
	if (hstate_is_gigantic(h) && !gigantic_page_runtime_supported())
1318
		return;
1319

1320 1321
	h->nr_huge_pages--;
	h->nr_huge_pages_node[page_to_nid(page)]--;
1322 1323 1324
	for (i = 0; i < pages_per_huge_page(h);
	     i++, subpage = mem_map_next(subpage, page, i)) {
		subpage->flags &= ~(1 << PG_locked | 1 << PG_error |
1325
				1 << PG_referenced | 1 << PG_dirty |
1326 1327
				1 << PG_active | 1 << PG_private |
				1 << PG_writeback);
A
Adam Litke 已提交
1328
	}
1329
	VM_BUG_ON_PAGE(hugetlb_cgroup_from_page(page), page);
1330
	VM_BUG_ON_PAGE(hugetlb_cgroup_from_page_rsvd(page), page);
1331
	set_compound_page_dtor(page, NULL_COMPOUND_DTOR);
A
Adam Litke 已提交
1332
	set_page_refcounted(page);
1333
	if (hstate_is_gigantic(h)) {
1334 1335 1336 1337 1338
		/*
		 * Temporarily drop the hugetlb_lock, because
		 * we might block in free_gigantic_page().
		 */
		spin_unlock(&hugetlb_lock);
1339 1340
		destroy_compound_gigantic_page(page, huge_page_order(h));
		free_gigantic_page(page, huge_page_order(h));
1341
		spin_lock(&hugetlb_lock);
1342 1343 1344
	} else {
		__free_pages(page, huge_page_order(h));
	}
A
Adam Litke 已提交
1345 1346
}

1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357
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;
}

1358 1359 1360 1361 1362 1363 1364 1365
/*
 * 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)
{
1366
	return PageHeadHuge(page) && PagePrivate(&page[1]);
1367 1368 1369
}

/* never called for tail page */
1370
void set_page_huge_active(struct page *page)
1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381
{
	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]);
}

1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403
/*
 * 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;
}

1404
static void __free_huge_page(struct page *page)
1405
{
1406 1407 1408 1409
	/*
	 * Can't pass hstate in here because it is called from the
	 * compound page destructor.
	 */
1410
	struct hstate *h = page_hstate(page);
1411
	int nid = page_to_nid(page);
1412 1413
	struct hugepage_subpool *spool =
		(struct hugepage_subpool *)page_private(page);
1414
	bool restore_reserve;
1415

1416 1417
	VM_BUG_ON_PAGE(page_count(page), page);
	VM_BUG_ON_PAGE(page_mapcount(page), page);
1418 1419 1420

	set_page_private(page, 0);
	page->mapping = NULL;
1421
	restore_reserve = PagePrivate(page);
1422
	ClearPagePrivate(page);
1423

1424
	/*
1425 1426 1427 1428 1429 1430
	 * 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.
1431
	 */
1432 1433 1434 1435 1436 1437 1438 1439 1440 1441
	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;
	}
1442

1443
	spin_lock(&hugetlb_lock);
1444
	clear_page_huge_active(page);
1445 1446
	hugetlb_cgroup_uncharge_page(hstate_index(h),
				     pages_per_huge_page(h), page);
1447 1448
	hugetlb_cgroup_uncharge_page_rsvd(hstate_index(h),
					  pages_per_huge_page(h), page);
1449 1450 1451
	if (restore_reserve)
		h->resv_huge_pages++;

1452 1453 1454 1455 1456
	if (PageHugeTemporary(page)) {
		list_del(&page->lru);
		ClearPageHugeTemporary(page);
		update_and_free_page(h, page);
	} else if (h->surplus_huge_pages_node[nid]) {
1457 1458
		/* remove the page from active list */
		list_del(&page->lru);
1459 1460 1461
		update_and_free_page(h, page);
		h->surplus_huge_pages--;
		h->surplus_huge_pages_node[nid]--;
1462
	} else {
1463
		arch_clear_hugepage_flags(page);
1464
		enqueue_huge_page(h, page);
1465
	}
1466 1467 1468
	spin_unlock(&hugetlb_lock);
}

1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516
/*
 * 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);
}

1517
static void prep_new_huge_page(struct hstate *h, struct page *page, int nid)
1518
{
1519
	INIT_LIST_HEAD(&page->lru);
1520
	set_compound_page_dtor(page, HUGETLB_PAGE_DTOR);
1521
	set_hugetlb_cgroup(page, NULL);
1522
	set_hugetlb_cgroup_rsvd(page, NULL);
1523
	spin_lock(&hugetlb_lock);
1524 1525
	h->nr_huge_pages++;
	h->nr_huge_pages_node[nid]++;
1526
	ClearPageHugeFreed(page);
1527 1528 1529
	spin_unlock(&hugetlb_lock);
}

1530
static void prep_compound_gigantic_page(struct page *page, unsigned int order)
1531 1532 1533 1534 1535 1536 1537
{
	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);
1538
	__ClearPageReserved(page);
1539
	__SetPageHead(page);
1540
	for (i = 1; i < nr_pages; i++, p = mem_map_next(p, page, i)) {
1541 1542 1543 1544
		/*
		 * 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
E
Ethon Paul 已提交
1545
		 * too.  Otherwise drivers using get_user_pages() to access tail
1546 1547 1548 1549 1550 1551 1552 1553
		 * 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);
1554
		set_page_count(p, 0);
1555
		set_compound_head(p, page);
1556
	}
1557
	atomic_set(compound_mapcount_ptr(page), -1);
1558 1559 1560

	if (hpage_pincount_available(page))
		atomic_set(compound_pincount_ptr(page), 0);
1561 1562
}

A
Andrew Morton 已提交
1563 1564 1565 1566 1567
/*
 * PageHuge() only returns true for hugetlbfs pages, but not for normal or
 * transparent huge pages.  See the PageTransHuge() documentation for more
 * details.
 */
1568 1569 1570 1571 1572 1573
int PageHuge(struct page *page)
{
	if (!PageCompound(page))
		return 0;

	page = compound_head(page);
1574
	return page[1].compound_dtor == HUGETLB_PAGE_DTOR;
1575
}
1576 1577
EXPORT_SYMBOL_GPL(PageHuge);

1578 1579 1580 1581 1582 1583 1584 1585 1586
/*
 * 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;

1587
	return page_head[1].compound_dtor == HUGETLB_PAGE_DTOR;
1588 1589
}

1590 1591 1592
/*
 * Find and lock address space (mapping) in write mode.
 *
1593 1594 1595
 * Upon entry, the page is locked which means that page_mapping() is
 * stable.  Due to locking order, we can only trylock_write.  If we can
 * not get the lock, simply return NULL to caller.
1596 1597 1598
 */
struct address_space *hugetlb_page_mapping_lock_write(struct page *hpage)
{
1599
	struct address_space *mapping = page_mapping(hpage);
1600 1601 1602 1603 1604 1605 1606

	if (!mapping)
		return mapping;

	if (i_mmap_trylock_write(mapping))
		return mapping;

1607
	return NULL;
1608 1609
}

1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622 1623 1624 1625 1626
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;
}

1627
static struct page *alloc_buddy_huge_page(struct hstate *h,
1628 1629
		gfp_t gfp_mask, int nid, nodemask_t *nmask,
		nodemask_t *node_alloc_noretry)
L
Linus Torvalds 已提交
1630
{
1631
	int order = huge_page_order(h);
L
Linus Torvalds 已提交
1632
	struct page *page;
1633
	bool alloc_try_hard = true;
1634

1635 1636 1637 1638 1639 1640 1641 1642 1643 1644 1645 1646
	/*
	 * 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;
1647 1648 1649 1650 1651 1652 1653
	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);
1654

1655 1656 1657 1658 1659 1660 1661 1662 1663 1664 1665 1666 1667 1668 1669 1670
	/*
	 * 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);

1671 1672 1673
	return page;
}

1674 1675 1676 1677 1678
/*
 * 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,
1679 1680
		gfp_t gfp_mask, int nid, nodemask_t *nmask,
		nodemask_t *node_alloc_noretry)
1681 1682 1683 1684 1685 1686 1687
{
	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,
1688
				nid, nmask, node_alloc_noretry);
1689 1690 1691 1692 1693 1694 1695 1696 1697 1698
	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;
}

1699 1700 1701 1702
/*
 * Allocates a fresh page to the hugetlb allocator pool in the node interleaved
 * manner.
 */
1703 1704
static int alloc_pool_huge_page(struct hstate *h, nodemask_t *nodes_allowed,
				nodemask_t *node_alloc_noretry)
1705 1706 1707
{
	struct page *page;
	int nr_nodes, node;
1708
	gfp_t gfp_mask = htlb_alloc_mask(h) | __GFP_THISNODE;
1709 1710

	for_each_node_mask_to_alloc(h, nr_nodes, node, nodes_allowed) {
1711 1712
		page = alloc_fresh_huge_page(h, gfp_mask, node, nodes_allowed,
						node_alloc_noretry);
1713
		if (page)
1714 1715 1716
			break;
	}

1717 1718
	if (!page)
		return 0;
1719

1720 1721 1722
	put_page(page); /* free it into the hugepage allocator */

	return 1;
1723 1724
}

1725 1726 1727 1728 1729 1730
/*
 * 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.
 */
1731 1732
static int free_pool_huge_page(struct hstate *h, nodemask_t *nodes_allowed,
							 bool acct_surplus)
1733
{
1734
	int nr_nodes, node;
1735 1736
	int ret = 0;

1737
	for_each_node_mask_to_free(h, nr_nodes, node, nodes_allowed) {
1738 1739 1740 1741
		/*
		 * If we're returning unused surplus pages, only examine
		 * nodes with surplus pages.
		 */
1742 1743
		if ((!acct_surplus || h->surplus_huge_pages_node[node]) &&
		    !list_empty(&h->hugepage_freelists[node])) {
1744
			struct page *page =
1745
				list_entry(h->hugepage_freelists[node].next,
1746 1747 1748
					  struct page, lru);
			list_del(&page->lru);
			h->free_huge_pages--;
1749
			h->free_huge_pages_node[node]--;
1750 1751
			if (acct_surplus) {
				h->surplus_huge_pages--;
1752
				h->surplus_huge_pages_node[node]--;
1753
			}
1754 1755
			update_and_free_page(h, page);
			ret = 1;
1756
			break;
1757
		}
1758
	}
1759 1760 1761 1762

	return ret;
}

1763 1764
/*
 * Dissolve a given free hugepage into free buddy pages. This function does
1765 1766 1767 1768 1769 1770 1771
 * 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)
1772
 */
1773
int dissolve_free_huge_page(struct page *page)
1774
{
1775
	int rc = -EBUSY;
1776

1777
retry:
1778 1779 1780 1781
	/* Not to disrupt normal path by vainly holding hugetlb_lock */
	if (!PageHuge(page))
		return 0;

1782
	spin_lock(&hugetlb_lock);
1783 1784 1785 1786 1787 1788
	if (!PageHuge(page)) {
		rc = 0;
		goto out;
	}

	if (!page_count(page)) {
1789 1790 1791
		struct page *head = compound_head(page);
		struct hstate *h = page_hstate(head);
		int nid = page_to_nid(head);
1792
		if (h->free_huge_pages - h->resv_huge_pages == 0)
1793
			goto out;
1794 1795 1796 1797 1798 1799 1800 1801 1802 1803 1804 1805 1806 1807 1808 1809 1810 1811 1812 1813

		/*
		 * We should make sure that the page is already on the free list
		 * when it is dissolved.
		 */
		if (unlikely(!PageHugeFreed(head))) {
			spin_unlock(&hugetlb_lock);
			cond_resched();

			/*
			 * Theoretically, we should return -EBUSY when we
			 * encounter this race. In fact, we have a chance
			 * to successfully dissolve the page if we do a
			 * retry. Because the race window is quite small.
			 * If we seize this opportunity, it is an optimization
			 * for increasing the success rate of dissolving page.
			 */
			goto retry;
		}

1814 1815 1816 1817 1818 1819 1820 1821
		/*
		 * 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);
		}
1822
		list_del(&head->lru);
1823 1824
		h->free_huge_pages--;
		h->free_huge_pages_node[nid]--;
1825
		h->max_huge_pages--;
1826
		update_and_free_page(h, head);
1827
		rc = 0;
1828
	}
1829
out:
1830
	spin_unlock(&hugetlb_lock);
1831
	return rc;
1832 1833 1834 1835 1836
}

/*
 * Dissolve free hugepages in a given pfn range. Used by memory hotplug to
 * make specified memory blocks removable from the system.
1837 1838
 * Note that this will dissolve a free gigantic hugepage completely, if any
 * part of it lies within the given range.
1839 1840
 * Also note that if dissolve_free_huge_page() returns with an error, all
 * free hugepages that were dissolved before that error are lost.
1841
 */
1842
int dissolve_free_huge_pages(unsigned long start_pfn, unsigned long end_pfn)
1843 1844
{
	unsigned long pfn;
1845
	struct page *page;
1846
	int rc = 0;
1847

1848
	if (!hugepages_supported())
1849
		return rc;
1850

1851 1852
	for (pfn = start_pfn; pfn < end_pfn; pfn += 1 << minimum_order) {
		page = pfn_to_page(pfn);
1853 1854 1855
		rc = dissolve_free_huge_page(page);
		if (rc)
			break;
1856
	}
1857 1858

	return rc;
1859 1860
}

1861 1862 1863
/*
 * Allocates a fresh surplus page from the page allocator.
 */
1864
static struct page *alloc_surplus_huge_page(struct hstate *h, gfp_t gfp_mask,
1865
		int nid, nodemask_t *nmask)
1866
{
1867
	struct page *page = NULL;
1868

1869
	if (hstate_is_gigantic(h))
1870 1871
		return NULL;

1872
	spin_lock(&hugetlb_lock);
1873 1874
	if (h->surplus_huge_pages >= h->nr_overcommit_huge_pages)
		goto out_unlock;
1875 1876
	spin_unlock(&hugetlb_lock);

1877
	page = alloc_fresh_huge_page(h, gfp_mask, nid, nmask, NULL);
1878
	if (!page)
1879
		return NULL;
1880 1881

	spin_lock(&hugetlb_lock);
1882 1883 1884 1885 1886 1887 1888 1889 1890
	/*
	 * 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);
1891
		spin_unlock(&hugetlb_lock);
1892
		put_page(page);
1893
		return NULL;
1894 1895
	} else {
		h->surplus_huge_pages++;
1896
		h->surplus_huge_pages_node[page_to_nid(page)]++;
1897
	}
1898 1899

out_unlock:
1900
	spin_unlock(&hugetlb_lock);
1901 1902 1903 1904

	return page;
}

1905
static struct page *alloc_migrate_huge_page(struct hstate *h, gfp_t gfp_mask,
1906
				     int nid, nodemask_t *nmask)
1907 1908 1909 1910 1911 1912
{
	struct page *page;

	if (hstate_is_gigantic(h))
		return NULL;

1913
	page = alloc_fresh_huge_page(h, gfp_mask, nid, nmask, NULL);
1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925
	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;
}

1926 1927 1928
/*
 * Use the VMA's mpolicy to allocate a huge page from the buddy.
 */
D
Dave Hansen 已提交
1929
static
1930
struct page *alloc_buddy_huge_page_with_mpol(struct hstate *h,
1931 1932
		struct vm_area_struct *vma, unsigned long addr)
{
1933 1934 1935 1936 1937 1938 1939
	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);
1940
	page = alloc_surplus_huge_page(h, gfp_mask, nid, nodemask);
1941 1942 1943
	mpol_cond_put(mpol);

	return page;
1944 1945
}

1946
/* page migration callback function */
1947
struct page *alloc_huge_page_nodemask(struct hstate *h, int preferred_nid,
1948
		nodemask_t *nmask, gfp_t gfp_mask)
1949 1950 1951
{
	spin_lock(&hugetlb_lock);
	if (h->free_huge_pages - h->resv_huge_pages > 0) {
1952 1953 1954 1955 1956 1957
		struct page *page;

		page = dequeue_huge_page_nodemask(h, gfp_mask, preferred_nid, nmask);
		if (page) {
			spin_unlock(&hugetlb_lock);
			return page;
1958 1959 1960 1961
		}
	}
	spin_unlock(&hugetlb_lock);

1962
	return alloc_migrate_huge_page(h, gfp_mask, preferred_nid, nmask);
1963 1964
}

1965
/* mempolicy aware migration callback */
1966 1967
struct page *alloc_huge_page_vma(struct hstate *h, struct vm_area_struct *vma,
		unsigned long address)
1968 1969 1970 1971 1972 1973 1974 1975 1976
{
	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);
1977
	page = alloc_huge_page_nodemask(h, node, nodemask, gfp_mask);
1978 1979 1980 1981 1982
	mpol_cond_put(mpol);

	return page;
}

1983
/*
L
Lucas De Marchi 已提交
1984
 * Increase the hugetlb pool such that it can accommodate a reservation
1985 1986
 * of size 'delta'.
 */
1987
static int gather_surplus_pages(struct hstate *h, int delta)
1988
	__must_hold(&hugetlb_lock)
1989 1990 1991 1992 1993
{
	struct list_head surplus_list;
	struct page *page, *tmp;
	int ret, i;
	int needed, allocated;
1994
	bool alloc_ok = true;
1995

1996
	needed = (h->resv_huge_pages + delta) - h->free_huge_pages;
1997
	if (needed <= 0) {
1998
		h->resv_huge_pages += delta;
1999
		return 0;
2000
	}
2001 2002 2003 2004 2005 2006 2007 2008

	allocated = 0;
	INIT_LIST_HEAD(&surplus_list);

	ret = -ENOMEM;
retry:
	spin_unlock(&hugetlb_lock);
	for (i = 0; i < needed; i++) {
2009
		page = alloc_surplus_huge_page(h, htlb_alloc_mask(h),
2010
				NUMA_NO_NODE, NULL);
2011 2012 2013 2014
		if (!page) {
			alloc_ok = false;
			break;
		}
2015
		list_add(&page->lru, &surplus_list);
2016
		cond_resched();
2017
	}
2018
	allocated += i;
2019 2020 2021 2022 2023 2024

	/*
	 * 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);
2025 2026
	needed = (h->resv_huge_pages + delta) -
			(h->free_huge_pages + allocated);
2027 2028 2029 2030 2031 2032 2033 2034 2035 2036
	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;
	}
2037 2038
	/*
	 * The surplus_list now contains _at_least_ the number of extra pages
L
Lucas De Marchi 已提交
2039
	 * needed to accommodate the reservation.  Add the appropriate number
2040
	 * of pages to the hugetlb pool and free the extras back to the buddy
2041 2042 2043
	 * 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.
2044 2045
	 */
	needed += allocated;
2046
	h->resv_huge_pages += delta;
2047
	ret = 0;
2048

2049
	/* Free the needed pages to the hugetlb pool */
2050
	list_for_each_entry_safe(page, tmp, &surplus_list, lru) {
2051 2052
		if ((--needed) < 0)
			break;
2053 2054 2055 2056 2057
		/*
		 * This page is now managed by the hugetlb allocator and has
		 * no users -- drop the buddy allocator's reference.
		 */
		put_page_testzero(page);
2058
		VM_BUG_ON_PAGE(page_count(page), page);
2059
		enqueue_huge_page(h, page);
2060
	}
2061
free:
2062
	spin_unlock(&hugetlb_lock);
2063 2064

	/* Free unnecessary surplus pages to the buddy allocator */
2065 2066
	list_for_each_entry_safe(page, tmp, &surplus_list, lru)
		put_page(page);
2067
	spin_lock(&hugetlb_lock);
2068 2069 2070 2071 2072

	return ret;
}

/*
2073 2074 2075 2076 2077 2078 2079 2080 2081 2082 2083 2084
 * 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.
2085
 */
2086 2087
static void return_unused_surplus_pages(struct hstate *h,
					unsigned long unused_resv_pages)
2088 2089 2090
{
	unsigned long nr_pages;

2091
	/* Cannot return gigantic pages currently */
2092
	if (hstate_is_gigantic(h))
2093
		goto out;
2094

2095 2096 2097 2098
	/*
	 * Part (or even all) of the reservation could have been backed
	 * by pre-allocated pages. Only free surplus pages.
	 */
2099
	nr_pages = min(unused_resv_pages, h->surplus_huge_pages);
2100

2101 2102
	/*
	 * We want to release as many surplus pages as possible, spread
2103 2104 2105
	 * 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.
2106
	 * free_pool_huge_page() will balance the freed pages across the
2107
	 * on-line nodes with memory and will handle the hstate accounting.
2108 2109 2110 2111
	 *
	 * 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.
2112 2113
	 */
	while (nr_pages--) {
2114 2115
		h->resv_huge_pages--;
		unused_resv_pages--;
2116
		if (!free_pool_huge_page(h, &node_states[N_MEMORY], 1))
2117
			goto out;
2118
		cond_resched_lock(&hugetlb_lock);
2119
	}
2120 2121 2122 2123

out:
	/* Fully uncommit the reservation */
	h->resv_huge_pages -= unused_resv_pages;
2124 2125
}

2126

2127
/*
2128
 * vma_needs_reservation, vma_commit_reservation and vma_end_reservation
2129
 * are used by the huge page allocation routines to manage reservations.
2130 2131 2132 2133 2134 2135
 *
 * 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
2136 2137 2138
 * 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.
2139 2140 2141 2142 2143 2144
 *
 * 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.
2145 2146 2147 2148 2149
 *
 * 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.
2150
 */
2151 2152 2153
enum vma_resv_mode {
	VMA_NEEDS_RESV,
	VMA_COMMIT_RESV,
2154
	VMA_END_RESV,
2155
	VMA_ADD_RESV,
2156
};
2157 2158
static long __vma_reservation_common(struct hstate *h,
				struct vm_area_struct *vma, unsigned long addr,
2159
				enum vma_resv_mode mode)
2160
{
2161 2162
	struct resv_map *resv;
	pgoff_t idx;
2163
	long ret;
2164
	long dummy_out_regions_needed;
2165

2166 2167
	resv = vma_resv_map(vma);
	if (!resv)
2168
		return 1;
2169

2170
	idx = vma_hugecache_offset(h, vma, addr);
2171 2172
	switch (mode) {
	case VMA_NEEDS_RESV:
2173 2174 2175 2176 2177 2178
		ret = region_chg(resv, idx, idx + 1, &dummy_out_regions_needed);
		/* We assume that vma_reservation_* routines always operate on
		 * 1 page, and that adding to resv map a 1 page entry can only
		 * ever require 1 region.
		 */
		VM_BUG_ON(dummy_out_regions_needed != 1);
2179 2180
		break;
	case VMA_COMMIT_RESV:
2181
		ret = region_add(resv, idx, idx + 1, 1, NULL, NULL);
2182 2183
		/* region_add calls of range 1 should never fail. */
		VM_BUG_ON(ret < 0);
2184
		break;
2185
	case VMA_END_RESV:
2186
		region_abort(resv, idx, idx + 1, 1);
2187 2188
		ret = 0;
		break;
2189
	case VMA_ADD_RESV:
2190
		if (vma->vm_flags & VM_MAYSHARE) {
2191
			ret = region_add(resv, idx, idx + 1, 1, NULL, NULL);
2192 2193 2194 2195
			/* region_add calls of range 1 should never fail. */
			VM_BUG_ON(ret < 0);
		} else {
			region_abort(resv, idx, idx + 1, 1);
2196 2197 2198
			ret = region_del(resv, idx, idx + 1);
		}
		break;
2199 2200 2201
	default:
		BUG();
	}
2202

2203
	if (vma->vm_flags & VM_MAYSHARE)
2204
		return ret;
2205 2206 2207 2208 2209 2210 2211 2212 2213 2214 2215 2216 2217 2218 2219 2220 2221 2222 2223
	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;
	}
2224
	else
2225
		return ret < 0 ? ret : 0;
2226
}
2227 2228

static long vma_needs_reservation(struct hstate *h,
2229
			struct vm_area_struct *vma, unsigned long addr)
2230
{
2231
	return __vma_reservation_common(h, vma, addr, VMA_NEEDS_RESV);
2232
}
2233

2234 2235 2236
static long vma_commit_reservation(struct hstate *h,
			struct vm_area_struct *vma, unsigned long addr)
{
2237 2238 2239
	return __vma_reservation_common(h, vma, addr, VMA_COMMIT_RESV);
}

2240
static void vma_end_reservation(struct hstate *h,
2241 2242
			struct vm_area_struct *vma, unsigned long addr)
{
2243
	(void)__vma_reservation_common(h, vma, addr, VMA_END_RESV);
2244 2245
}

2246 2247 2248 2249 2250 2251 2252 2253 2254 2255 2256 2257 2258 2259 2260 2261 2262 2263 2264 2265 2266 2267 2268 2269 2270 2271 2272 2273 2274 2275 2276 2277 2278 2279 2280 2281 2282 2283 2284 2285 2286 2287 2288 2289 2290 2291 2292 2293 2294 2295
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);
	}
}

2296
struct page *alloc_huge_page(struct vm_area_struct *vma,
2297
				    unsigned long addr, int avoid_reserve)
L
Linus Torvalds 已提交
2298
{
2299
	struct hugepage_subpool *spool = subpool_vma(vma);
2300
	struct hstate *h = hstate_vma(vma);
2301
	struct page *page;
2302 2303
	long map_chg, map_commit;
	long gbl_chg;
2304 2305
	int ret, idx;
	struct hugetlb_cgroup *h_cg;
2306
	bool deferred_reserve;
2307

2308
	idx = hstate_index(h);
2309
	/*
2310 2311 2312
	 * 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).
2313
	 */
2314 2315
	map_chg = gbl_chg = vma_needs_reservation(h, vma, addr);
	if (map_chg < 0)
2316
		return ERR_PTR(-ENOMEM);
2317 2318 2319 2320 2321 2322 2323 2324 2325 2326 2327

	/*
	 * 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) {
2328
			vma_end_reservation(h, vma, addr);
2329
			return ERR_PTR(-ENOSPC);
2330
		}
L
Linus Torvalds 已提交
2331

2332 2333 2334 2335 2336 2337 2338 2339 2340 2341 2342 2343
		/*
		 * 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;
	}

2344 2345 2346 2347 2348 2349 2350 2351 2352 2353
	/* If this allocation is not consuming a reservation, charge it now.
	 */
	deferred_reserve = map_chg || avoid_reserve || !vma_resv_map(vma);
	if (deferred_reserve) {
		ret = hugetlb_cgroup_charge_cgroup_rsvd(
			idx, pages_per_huge_page(h), &h_cg);
		if (ret)
			goto out_subpool_put;
	}

2354
	ret = hugetlb_cgroup_charge_cgroup(idx, pages_per_huge_page(h), &h_cg);
2355
	if (ret)
2356
		goto out_uncharge_cgroup_reservation;
2357

L
Linus Torvalds 已提交
2358
	spin_lock(&hugetlb_lock);
2359 2360 2361 2362 2363 2364
	/*
	 * 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);
2365
	if (!page) {
2366
		spin_unlock(&hugetlb_lock);
2367
		page = alloc_buddy_huge_page_with_mpol(h, vma, addr);
2368 2369
		if (!page)
			goto out_uncharge_cgroup;
2370 2371 2372 2373
		if (!avoid_reserve && vma_has_reserves(vma, gbl_chg)) {
			SetPagePrivate(page);
			h->resv_huge_pages--;
		}
2374
		spin_lock(&hugetlb_lock);
2375
		list_add(&page->lru, &h->hugepage_activelist);
2376
		/* Fall through */
K
Ken Chen 已提交
2377
	}
2378
	hugetlb_cgroup_commit_charge(idx, pages_per_huge_page(h), h_cg, page);
2379 2380 2381 2382 2383 2384 2385 2386
	/* If allocation is not consuming a reservation, also store the
	 * hugetlb_cgroup pointer on the page.
	 */
	if (deferred_reserve) {
		hugetlb_cgroup_commit_charge_rsvd(idx, pages_per_huge_page(h),
						  h_cg, page);
	}

2387
	spin_unlock(&hugetlb_lock);
2388

2389
	set_page_private(page, (unsigned long)spool);
2390

2391 2392
	map_commit = vma_commit_reservation(h, vma, addr);
	if (unlikely(map_chg > map_commit)) {
2393 2394 2395 2396 2397 2398 2399 2400 2401 2402 2403 2404 2405
		/*
		 * 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);
2406 2407 2408
		if (deferred_reserve)
			hugetlb_cgroup_uncharge_page_rsvd(hstate_index(h),
					pages_per_huge_page(h), page);
2409
	}
2410
	return page;
2411 2412 2413

out_uncharge_cgroup:
	hugetlb_cgroup_uncharge_cgroup(idx, pages_per_huge_page(h), h_cg);
2414 2415 2416 2417
out_uncharge_cgroup_reservation:
	if (deferred_reserve)
		hugetlb_cgroup_uncharge_cgroup_rsvd(idx, pages_per_huge_page(h),
						    h_cg);
2418
out_subpool_put:
2419
	if (map_chg || avoid_reserve)
2420
		hugepage_subpool_put_pages(spool, 1);
2421
	vma_end_reservation(h, vma, addr);
2422
	return ERR_PTR(-ENOSPC);
2423 2424
}

2425 2426 2427
int alloc_bootmem_huge_page(struct hstate *h)
	__attribute__ ((weak, alias("__alloc_bootmem_huge_page")));
int __alloc_bootmem_huge_page(struct hstate *h)
2428 2429
{
	struct huge_bootmem_page *m;
2430
	int nr_nodes, node;
2431

2432
	for_each_node_mask_to_alloc(h, nr_nodes, node, &node_states[N_MEMORY]) {
2433 2434
		void *addr;

2435
		addr = memblock_alloc_try_nid_raw(
2436
				huge_page_size(h), huge_page_size(h),
2437
				0, MEMBLOCK_ALLOC_ACCESSIBLE, node);
2438 2439 2440 2441 2442 2443 2444
		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;
2445
			goto found;
2446 2447 2448 2449 2450
		}
	}
	return 0;

found:
2451
	BUG_ON(!IS_ALIGNED(virt_to_phys(m), huge_page_size(h)));
2452
	/* Put them into a private list first because mem_map is not up yet */
2453
	INIT_LIST_HEAD(&m->list);
2454 2455 2456 2457 2458
	list_add(&m->list, &huge_boot_pages);
	m->hstate = h;
	return 1;
}

2459 2460
static void __init prep_compound_huge_page(struct page *page,
		unsigned int order)
2461 2462 2463 2464 2465 2466 2467
{
	if (unlikely(order > (MAX_ORDER - 1)))
		prep_compound_gigantic_page(page, order);
	else
		prep_compound_page(page, order);
}

2468 2469 2470 2471 2472 2473
/* 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) {
2474
		struct page *page = virt_to_page(m);
2475
		struct hstate *h = m->hstate;
2476

2477
		WARN_ON(page_count(page) != 1);
2478
		prep_compound_huge_page(page, h->order);
2479
		WARN_ON(PageReserved(page));
2480
		prep_new_huge_page(h, page, page_to_nid(page));
2481 2482
		put_page(page); /* free it into the hugepage allocator */

2483 2484 2485 2486 2487 2488
		/*
		 * 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.
		 */
2489
		if (hstate_is_gigantic(h))
2490
			adjust_managed_page_count(page, 1 << h->order);
2491
		cond_resched();
2492 2493 2494
	}
}

2495
static void __init hugetlb_hstate_alloc_pages(struct hstate *h)
L
Linus Torvalds 已提交
2496 2497
{
	unsigned long i;
2498 2499 2500 2501 2502 2503 2504 2505 2506 2507 2508 2509 2510 2511 2512 2513 2514 2515 2516
	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);
2517

2518
	for (i = 0; i < h->max_huge_pages; ++i) {
2519
		if (hstate_is_gigantic(h)) {
2520
			if (hugetlb_cma_size) {
2521
				pr_warn_once("HugeTLB: hugetlb_cma is enabled, skip boot time allocation\n");
2522
				goto free;
2523
			}
2524 2525
			if (!alloc_bootmem_huge_page(h))
				break;
2526
		} else if (!alloc_pool_huge_page(h,
2527 2528
					 &node_states[N_MEMORY],
					 node_alloc_noretry))
L
Linus Torvalds 已提交
2529
			break;
2530
		cond_resched();
L
Linus Torvalds 已提交
2531
	}
2532 2533 2534
	if (i < h->max_huge_pages) {
		char buf[32];

2535
		string_get_size(huge_page_size(h), 1, STRING_UNITS_2, buf, 32);
2536 2537 2538 2539
		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;
	}
2540
free:
2541
	kfree(node_alloc_noretry);
2542 2543 2544 2545 2546 2547 2548
}

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

	for_each_hstate(h) {
2549 2550 2551
		if (minimum_order > huge_page_order(h))
			minimum_order = huge_page_order(h);

2552
		/* oversize hugepages were init'ed in early boot */
2553
		if (!hstate_is_gigantic(h))
2554
			hugetlb_hstate_alloc_pages(h);
2555
	}
2556
	VM_BUG_ON(minimum_order == UINT_MAX);
2557 2558 2559 2560 2561 2562 2563
}

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

	for_each_hstate(h) {
A
Andi Kleen 已提交
2564
		char buf[32];
2565 2566

		string_get_size(huge_page_size(h), 1, STRING_UNITS_2, buf, 32);
2567
		pr_info("HugeTLB registered %s page size, pre-allocated %ld pages\n",
2568
			buf, h->free_huge_pages);
2569 2570 2571
	}
}

L
Linus Torvalds 已提交
2572
#ifdef CONFIG_HIGHMEM
2573 2574
static void try_to_free_low(struct hstate *h, unsigned long count,
						nodemask_t *nodes_allowed)
L
Linus Torvalds 已提交
2575
{
2576 2577
	int i;

2578
	if (hstate_is_gigantic(h))
2579 2580
		return;

2581
	for_each_node_mask(i, *nodes_allowed) {
L
Linus Torvalds 已提交
2582
		struct page *page, *next;
2583 2584 2585
		struct list_head *freel = &h->hugepage_freelists[i];
		list_for_each_entry_safe(page, next, freel, lru) {
			if (count >= h->nr_huge_pages)
2586
				return;
L
Linus Torvalds 已提交
2587 2588 2589
			if (PageHighMem(page))
				continue;
			list_del(&page->lru);
2590
			update_and_free_page(h, page);
2591 2592
			h->free_huge_pages--;
			h->free_huge_pages_node[page_to_nid(page)]--;
L
Linus Torvalds 已提交
2593 2594 2595 2596
		}
	}
}
#else
2597 2598
static inline void try_to_free_low(struct hstate *h, unsigned long count,
						nodemask_t *nodes_allowed)
L
Linus Torvalds 已提交
2599 2600 2601 2602
{
}
#endif

2603 2604 2605 2606 2607
/*
 * 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.
 */
2608 2609
static int adjust_pool_surplus(struct hstate *h, nodemask_t *nodes_allowed,
				int delta)
2610
{
2611
	int nr_nodes, node;
2612 2613 2614

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

2615 2616 2617 2618
	if (delta < 0) {
		for_each_node_mask_to_alloc(h, nr_nodes, node, nodes_allowed) {
			if (h->surplus_huge_pages_node[node])
				goto found;
2619
		}
2620 2621 2622 2623 2624
	} 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;
2625
		}
2626 2627
	}
	return 0;
2628

2629 2630 2631 2632
found:
	h->surplus_huge_pages += delta;
	h->surplus_huge_pages_node[node] += delta;
	return 1;
2633 2634
}

2635
#define persistent_huge_pages(h) (h->nr_huge_pages - h->surplus_huge_pages)
2636
static int set_max_huge_pages(struct hstate *h, unsigned long count, int nid,
2637
			      nodemask_t *nodes_allowed)
L
Linus Torvalds 已提交
2638
{
2639
	unsigned long min_count, ret;
2640 2641 2642 2643 2644 2645 2646 2647 2648 2649 2650
	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
2651

2652 2653
	spin_lock(&hugetlb_lock);

2654 2655 2656 2657 2658 2659 2660 2661 2662 2663 2664 2665 2666 2667 2668 2669 2670 2671 2672 2673
	/*
	 * 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;
	}

2674 2675 2676 2677 2678 2679 2680 2681 2682 2683
	/*
	 * 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);
2684
			NODEMASK_FREE(node_alloc_noretry);
2685 2686 2687 2688
			return -EINVAL;
		}
		/* Fall through to decrease pool */
	}
2689

2690 2691 2692 2693
	/*
	 * Increase the pool size
	 * First take pages out of surplus state.  Then make up the
	 * remaining difference by allocating fresh huge pages.
2694
	 *
2695
	 * We might race with alloc_surplus_huge_page() here and be unable
2696 2697 2698 2699
	 * 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.
2700
	 */
2701
	while (h->surplus_huge_pages && count > persistent_huge_pages(h)) {
2702
		if (!adjust_pool_surplus(h, nodes_allowed, -1))
2703 2704 2705
			break;
	}

2706
	while (count > persistent_huge_pages(h)) {
2707 2708 2709 2710 2711 2712
		/*
		 * 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);
2713 2714 2715 2716

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

2717 2718
		ret = alloc_pool_huge_page(h, nodes_allowed,
						node_alloc_noretry);
2719 2720 2721 2722
		spin_lock(&hugetlb_lock);
		if (!ret)
			goto out;

2723 2724 2725
		/* Bail for signals. Probably ctrl-c from user */
		if (signal_pending(current))
			goto out;
2726 2727 2728 2729 2730 2731 2732 2733
	}

	/*
	 * 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.
2734 2735 2736 2737
	 *
	 * 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
2738
	 * alloc_surplus_huge_page() is checking the global counter,
2739 2740 2741
	 * 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.
2742
	 */
2743
	min_count = h->resv_huge_pages + h->nr_huge_pages - h->free_huge_pages;
2744
	min_count = max(count, min_count);
2745
	try_to_free_low(h, min_count, nodes_allowed);
2746
	while (min_count < persistent_huge_pages(h)) {
2747
		if (!free_pool_huge_page(h, nodes_allowed, 0))
L
Linus Torvalds 已提交
2748
			break;
2749
		cond_resched_lock(&hugetlb_lock);
L
Linus Torvalds 已提交
2750
	}
2751
	while (count < persistent_huge_pages(h)) {
2752
		if (!adjust_pool_surplus(h, nodes_allowed, 1))
2753 2754 2755
			break;
	}
out:
2756
	h->max_huge_pages = persistent_huge_pages(h);
L
Linus Torvalds 已提交
2757
	spin_unlock(&hugetlb_lock);
2758

2759 2760
	NODEMASK_FREE(node_alloc_noretry);

2761
	return 0;
L
Linus Torvalds 已提交
2762 2763
}

2764 2765 2766 2767 2768 2769 2770 2771 2772 2773
#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];

2774 2775 2776
static struct hstate *kobj_to_node_hstate(struct kobject *kobj, int *nidp);

static struct hstate *kobj_to_hstate(struct kobject *kobj, int *nidp)
2777 2778
{
	int i;
2779

2780
	for (i = 0; i < HUGE_MAX_HSTATE; i++)
2781 2782 2783
		if (hstate_kobjs[i] == kobj) {
			if (nidp)
				*nidp = NUMA_NO_NODE;
2784
			return &hstates[i];
2785 2786 2787
		}

	return kobj_to_node_hstate(kobj, nidp);
2788 2789
}

2790
static ssize_t nr_hugepages_show_common(struct kobject *kobj,
2791 2792
					struct kobj_attribute *attr, char *buf)
{
2793 2794 2795 2796 2797 2798 2799 2800 2801 2802 2803
	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);
2804
}
2805

2806 2807 2808
static ssize_t __nr_hugepages_store_common(bool obey_mempolicy,
					   struct hstate *h, int nid,
					   unsigned long count, size_t len)
2809 2810
{
	int err;
2811
	nodemask_t nodes_allowed, *n_mask;
2812

2813 2814
	if (hstate_is_gigantic(h) && !gigantic_page_runtime_supported())
		return -EINVAL;
2815

2816 2817 2818 2819 2820
	if (nid == NUMA_NO_NODE) {
		/*
		 * global hstate attribute
		 */
		if (!(obey_mempolicy &&
2821 2822 2823 2824 2825
				init_nodemask_of_mempolicy(&nodes_allowed)))
			n_mask = &node_states[N_MEMORY];
		else
			n_mask = &nodes_allowed;
	} else {
2826
		/*
2827 2828
		 * Node specific request.  count adjustment happens in
		 * set_max_huge_pages() after acquiring hugetlb_lock.
2829
		 */
2830 2831
		init_nodemask_of_node(&nodes_allowed, nid);
		n_mask = &nodes_allowed;
2832
	}
2833

2834
	err = set_max_huge_pages(h, count, nid, n_mask);
2835

2836
	return err ? err : len;
2837 2838
}

2839 2840 2841 2842 2843 2844 2845 2846 2847 2848 2849 2850 2851 2852 2853 2854 2855
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);
}

2856 2857 2858 2859 2860 2861 2862 2863 2864
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)
{
2865
	return nr_hugepages_store_common(false, kobj, buf, len);
2866 2867 2868
}
HSTATE_ATTR(nr_hugepages);

2869 2870 2871 2872 2873 2874 2875 2876 2877 2878 2879 2880 2881 2882 2883
#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)
{
2884
	return nr_hugepages_store_common(true, kobj, buf, len);
2885 2886 2887 2888 2889
}
HSTATE_ATTR(nr_hugepages_mempolicy);
#endif


2890 2891 2892
static ssize_t nr_overcommit_hugepages_show(struct kobject *kobj,
					struct kobj_attribute *attr, char *buf)
{
2893
	struct hstate *h = kobj_to_hstate(kobj, NULL);
2894 2895
	return sprintf(buf, "%lu\n", h->nr_overcommit_huge_pages);
}
2896

2897 2898 2899 2900 2901
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;
2902
	struct hstate *h = kobj_to_hstate(kobj, NULL);
2903

2904
	if (hstate_is_gigantic(h))
2905 2906
		return -EINVAL;

2907
	err = kstrtoul(buf, 10, &input);
2908
	if (err)
2909
		return err;
2910 2911 2912 2913 2914 2915 2916 2917 2918 2919 2920 2921

	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)
{
2922 2923 2924 2925 2926 2927 2928 2929 2930 2931 2932
	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);
2933 2934 2935 2936 2937 2938
}
HSTATE_ATTR_RO(free_hugepages);

static ssize_t resv_hugepages_show(struct kobject *kobj,
					struct kobj_attribute *attr, char *buf)
{
2939
	struct hstate *h = kobj_to_hstate(kobj, NULL);
2940 2941 2942 2943 2944 2945 2946
	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)
{
2947 2948 2949 2950 2951 2952 2953 2954 2955 2956 2957
	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);
2958 2959 2960 2961 2962 2963 2964 2965 2966
}
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,
2967 2968 2969
#ifdef CONFIG_NUMA
	&nr_hugepages_mempolicy_attr.attr,
#endif
2970 2971 2972
	NULL,
};

2973
static const struct attribute_group hstate_attr_group = {
2974 2975 2976
	.attrs = hstate_attrs,
};

J
Jeff Mahoney 已提交
2977 2978
static int hugetlb_sysfs_add_hstate(struct hstate *h, struct kobject *parent,
				    struct kobject **hstate_kobjs,
2979
				    const struct attribute_group *hstate_attr_group)
2980 2981
{
	int retval;
2982
	int hi = hstate_index(h);
2983

2984 2985
	hstate_kobjs[hi] = kobject_create_and_add(h->name, parent);
	if (!hstate_kobjs[hi])
2986 2987
		return -ENOMEM;

2988
	retval = sysfs_create_group(hstate_kobjs[hi], hstate_attr_group);
2989
	if (retval) {
2990
		kobject_put(hstate_kobjs[hi]);
2991 2992
		hstate_kobjs[hi] = NULL;
	}
2993 2994 2995 2996 2997 2998 2999 3000 3001 3002 3003 3004 3005 3006

	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) {
3007 3008
		err = hugetlb_sysfs_add_hstate(h, hugepages_kobj,
					 hstate_kobjs, &hstate_attr_group);
3009
		if (err)
3010
			pr_err("HugeTLB: Unable to add hstate %s", h->name);
3011 3012 3013
	}
}

3014 3015 3016 3017
#ifdef CONFIG_NUMA

/*
 * node_hstate/s - associate per node hstate attributes, via their kobjects,
3018 3019 3020
 * 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
3021 3022 3023 3024 3025 3026
 * the base kernel, on the hugetlb module.
 */
struct node_hstate {
	struct kobject		*hugepages_kobj;
	struct kobject		*hstate_kobjs[HUGE_MAX_HSTATE];
};
3027
static struct node_hstate node_hstates[MAX_NUMNODES];
3028 3029

/*
3030
 * A subset of global hstate attributes for node devices
3031 3032 3033 3034 3035 3036 3037 3038
 */
static struct attribute *per_node_hstate_attrs[] = {
	&nr_hugepages_attr.attr,
	&free_hugepages_attr.attr,
	&surplus_hugepages_attr.attr,
	NULL,
};

3039
static const struct attribute_group per_node_hstate_attr_group = {
3040 3041 3042 3043
	.attrs = per_node_hstate_attrs,
};

/*
3044
 * kobj_to_node_hstate - lookup global hstate for node device hstate attr kobj.
3045 3046 3047 3048 3049 3050 3051 3052 3053 3054 3055 3056 3057 3058 3059 3060 3061 3062 3063 3064 3065 3066
 * 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;
}

/*
3067
 * Unregister hstate attributes from a single node device.
3068 3069
 * No-op if no hstate attributes attached.
 */
3070
static void hugetlb_unregister_node(struct node *node)
3071 3072
{
	struct hstate *h;
3073
	struct node_hstate *nhs = &node_hstates[node->dev.id];
3074 3075

	if (!nhs->hugepages_kobj)
3076
		return;		/* no hstate attributes */
3077

3078 3079 3080 3081 3082
	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;
3083
		}
3084
	}
3085 3086 3087 3088 3089 3090 3091

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


/*
3092
 * Register hstate attributes for a single node device.
3093 3094
 * No-op if attributes already registered.
 */
3095
static void hugetlb_register_node(struct node *node)
3096 3097
{
	struct hstate *h;
3098
	struct node_hstate *nhs = &node_hstates[node->dev.id];
3099 3100 3101 3102 3103 3104
	int err;

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

	nhs->hugepages_kobj = kobject_create_and_add("hugepages",
3105
							&node->dev.kobj);
3106 3107 3108 3109 3110 3111 3112 3113
	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) {
3114
			pr_err("HugeTLB: Unable to add hstate %s for node %d\n",
3115
				h->name, node->dev.id);
3116 3117 3118 3119 3120 3121 3122
			hugetlb_unregister_node(node);
			break;
		}
	}
}

/*
3123
 * hugetlb init time:  register hstate attributes for all registered node
3124 3125
 * devices of nodes that have memory.  All on-line nodes should have
 * registered their associated device by this time.
3126
 */
3127
static void __init hugetlb_register_all_nodes(void)
3128 3129 3130
{
	int nid;

3131
	for_each_node_state(nid, N_MEMORY) {
3132
		struct node *node = node_devices[nid];
3133
		if (node->dev.id == nid)
3134 3135 3136 3137
			hugetlb_register_node(node);
	}

	/*
3138
	 * Let the node device driver know we're here so it can
3139 3140 3141 3142 3143 3144 3145 3146 3147 3148 3149 3150 3151 3152 3153 3154 3155 3156 3157
	 * [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

3158 3159
static int __init hugetlb_init(void)
{
3160 3161
	int i;

3162 3163 3164
	if (!hugepages_supported()) {
		if (hugetlb_max_hstate || default_hstate_max_huge_pages)
			pr_warn("HugeTLB: huge pages not supported, ignoring associated command-line parameters\n");
3165
		return 0;
3166
	}
3167

3168 3169 3170 3171 3172 3173 3174 3175 3176 3177 3178 3179 3180 3181 3182 3183 3184 3185 3186 3187 3188 3189 3190 3191 3192 3193 3194 3195
	/*
	 * Make sure HPAGE_SIZE (HUGETLB_PAGE_ORDER) hstate exists.  Some
	 * architectures depend on setup being done here.
	 */
	hugetlb_add_hstate(HUGETLB_PAGE_ORDER);
	if (!parsed_default_hugepagesz) {
		/*
		 * If we did not parse a default huge page size, set
		 * default_hstate_idx to HPAGE_SIZE hstate. And, if the
		 * number of huge pages for this default size was implicitly
		 * specified, set that here as well.
		 * Note that the implicit setting will overwrite an explicit
		 * setting.  A warning will be printed in this case.
		 */
		default_hstate_idx = hstate_index(size_to_hstate(HPAGE_SIZE));
		if (default_hstate_max_huge_pages) {
			if (default_hstate.max_huge_pages) {
				char buf[32];

				string_get_size(huge_page_size(&default_hstate),
					1, STRING_UNITS_2, buf, 32);
				pr_warn("HugeTLB: Ignoring hugepages=%lu associated with %s page size\n",
					default_hstate.max_huge_pages, buf);
				pr_warn("HugeTLB: Using hugepages=%lu for number of default huge pages\n",
					default_hstate_max_huge_pages);
			}
			default_hstate.max_huge_pages =
				default_hstate_max_huge_pages;
3196
		}
3197
	}
3198

3199
	hugetlb_cma_check();
3200
	hugetlb_init_hstates();
3201
	gather_bootmem_prealloc();
3202 3203 3204
	report_hugepages();

	hugetlb_sysfs_init();
3205
	hugetlb_register_all_nodes();
3206
	hugetlb_cgroup_file_init();
3207

3208 3209 3210 3211 3212
#ifdef CONFIG_SMP
	num_fault_mutexes = roundup_pow_of_two(8 * num_possible_cpus());
#else
	num_fault_mutexes = 1;
#endif
3213
	hugetlb_fault_mutex_table =
3214 3215
		kmalloc_array(num_fault_mutexes, sizeof(struct mutex),
			      GFP_KERNEL);
3216
	BUG_ON(!hugetlb_fault_mutex_table);
3217 3218

	for (i = 0; i < num_fault_mutexes; i++)
3219
		mutex_init(&hugetlb_fault_mutex_table[i]);
3220 3221
	return 0;
}
3222
subsys_initcall(hugetlb_init);
3223

3224 3225
/* Overwritten by architectures with more huge page sizes */
bool __init __attribute((weak)) arch_hugetlb_valid_size(unsigned long size)
3226
{
3227
	return size == HPAGE_SIZE;
3228 3229
}

3230
void __init hugetlb_add_hstate(unsigned int order)
3231 3232
{
	struct hstate *h;
3233 3234
	unsigned long i;

3235 3236 3237
	if (size_to_hstate(PAGE_SIZE << order)) {
		return;
	}
3238
	BUG_ON(hugetlb_max_hstate >= HUGE_MAX_HSTATE);
3239
	BUG_ON(order == 0);
3240
	h = &hstates[hugetlb_max_hstate++];
3241 3242
	h->order = order;
	h->mask = ~((1ULL << (order + PAGE_SHIFT)) - 1);
3243 3244 3245 3246
	h->nr_huge_pages = 0;
	h->free_huge_pages = 0;
	for (i = 0; i < MAX_NUMNODES; ++i)
		INIT_LIST_HEAD(&h->hugepage_freelists[i]);
3247
	INIT_LIST_HEAD(&h->hugepage_activelist);
3248 3249
	h->next_nid_to_alloc = first_memory_node;
	h->next_nid_to_free = first_memory_node;
3250 3251
	snprintf(h->name, HSTATE_NAME_LEN, "hugepages-%lukB",
					huge_page_size(h)/1024);
3252

3253 3254 3255
	parsed_hstate = h;
}

3256 3257 3258 3259 3260 3261 3262 3263
/*
 * hugepages command line processing
 * hugepages normally follows a valid hugepagsz or default_hugepagsz
 * specification.  If not, ignore the hugepages value.  hugepages can also
 * be the first huge page command line  option in which case it implicitly
 * specifies the number of huge pages for the default size.
 */
static int __init hugepages_setup(char *s)
3264 3265
{
	unsigned long *mhp;
3266
	static unsigned long *last_mhp;
3267

3268
	if (!parsed_valid_hugepagesz) {
3269
		pr_warn("HugeTLB: hugepages=%s does not follow a valid hugepagesz, ignoring\n", s);
3270
		parsed_valid_hugepagesz = true;
3271
		return 0;
3272
	}
3273

3274
	/*
3275 3276 3277 3278
	 * !hugetlb_max_hstate means we haven't parsed a hugepagesz= parameter
	 * yet, so this hugepages= parameter goes to the "default hstate".
	 * Otherwise, it goes with the previously parsed hugepagesz or
	 * default_hugepagesz.
3279
	 */
3280
	else if (!hugetlb_max_hstate)
3281 3282 3283 3284
		mhp = &default_hstate_max_huge_pages;
	else
		mhp = &parsed_hstate->max_huge_pages;

3285
	if (mhp == last_mhp) {
3286 3287
		pr_warn("HugeTLB: hugepages= specified twice without interleaving hugepagesz=, ignoring hugepages=%s\n", s);
		return 0;
3288 3289
	}

3290 3291 3292
	if (sscanf(s, "%lu", mhp) <= 0)
		*mhp = 0;

3293 3294 3295 3296 3297
	/*
	 * 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.
	 */
3298
	if (hugetlb_max_hstate && parsed_hstate->order >= MAX_ORDER)
3299 3300 3301 3302
		hugetlb_hstate_alloc_pages(parsed_hstate);

	last_mhp = mhp;

3303 3304
	return 1;
}
3305
__setup("hugepages=", hugepages_setup);
3306

3307 3308 3309 3310 3311 3312 3313
/*
 * hugepagesz command line processing
 * A specific huge page size can only be specified once with hugepagesz.
 * hugepagesz is followed by hugepages on the command line.  The global
 * variable 'parsed_valid_hugepagesz' is used to determine if prior
 * hugepagesz argument was valid.
 */
3314
static int __init hugepagesz_setup(char *s)
3315
{
3316
	unsigned long size;
3317 3318 3319
	struct hstate *h;

	parsed_valid_hugepagesz = false;
3320 3321 3322
	size = (unsigned long)memparse(s, NULL);

	if (!arch_hugetlb_valid_size(size)) {
3323
		pr_err("HugeTLB: unsupported hugepagesz=%s\n", s);
3324 3325 3326
		return 0;
	}

3327 3328 3329 3330 3331 3332 3333 3334 3335 3336 3337 3338 3339 3340 3341 3342 3343 3344 3345 3346 3347 3348 3349
	h = size_to_hstate(size);
	if (h) {
		/*
		 * hstate for this size already exists.  This is normally
		 * an error, but is allowed if the existing hstate is the
		 * default hstate.  More specifically, it is only allowed if
		 * the number of huge pages for the default hstate was not
		 * previously specified.
		 */
		if (!parsed_default_hugepagesz ||  h != &default_hstate ||
		    default_hstate.max_huge_pages) {
			pr_warn("HugeTLB: hugepagesz=%s specified twice, ignoring\n", s);
			return 0;
		}

		/*
		 * No need to call hugetlb_add_hstate() as hstate already
		 * exists.  But, do set parsed_hstate so that a following
		 * hugepages= parameter will be applied to this hstate.
		 */
		parsed_hstate = h;
		parsed_valid_hugepagesz = true;
		return 1;
3350 3351
	}

3352
	hugetlb_add_hstate(ilog2(size) - PAGE_SHIFT);
3353
	parsed_valid_hugepagesz = true;
3354 3355
	return 1;
}
3356 3357
__setup("hugepagesz=", hugepagesz_setup);

3358 3359 3360 3361
/*
 * default_hugepagesz command line input
 * Only one instance of default_hugepagesz allowed on command line.
 */
3362
static int __init default_hugepagesz_setup(char *s)
3363
{
3364 3365
	unsigned long size;

3366 3367 3368 3369 3370 3371
	parsed_valid_hugepagesz = false;
	if (parsed_default_hugepagesz) {
		pr_err("HugeTLB: default_hugepagesz previously specified, ignoring %s\n", s);
		return 0;
	}

3372 3373 3374
	size = (unsigned long)memparse(s, NULL);

	if (!arch_hugetlb_valid_size(size)) {
3375
		pr_err("HugeTLB: unsupported default_hugepagesz=%s\n", s);
3376 3377 3378
		return 0;
	}

3379 3380 3381 3382 3383 3384 3385 3386 3387 3388 3389 3390 3391 3392 3393 3394 3395 3396 3397
	hugetlb_add_hstate(ilog2(size) - PAGE_SHIFT);
	parsed_valid_hugepagesz = true;
	parsed_default_hugepagesz = true;
	default_hstate_idx = hstate_index(size_to_hstate(size));

	/*
	 * The number of default huge pages (for this size) could have been
	 * specified as the first hugetlb parameter: hugepages=X.  If so,
	 * then default_hstate_max_huge_pages is set.  If the default huge
	 * page size is gigantic (>= MAX_ORDER), then the pages must be
	 * allocated here from bootmem allocator.
	 */
	if (default_hstate_max_huge_pages) {
		default_hstate.max_huge_pages = default_hstate_max_huge_pages;
		if (hstate_is_gigantic(&default_hstate))
			hugetlb_hstate_alloc_pages(&default_hstate);
		default_hstate_max_huge_pages = 0;
	}

3398 3399
	return 1;
}
3400
__setup("default_hugepagesz=", default_hugepagesz_setup);
3401

3402
static unsigned int allowed_mems_nr(struct hstate *h)
3403 3404 3405
{
	int node;
	unsigned int nr = 0;
3406 3407 3408 3409 3410
	nodemask_t *mpol_allowed;
	unsigned int *array = h->free_huge_pages_node;
	gfp_t gfp_mask = htlb_alloc_mask(h);

	mpol_allowed = policy_nodemask_current(gfp_mask);
3411

3412 3413 3414 3415 3416
	for_each_node_mask(node, cpuset_current_mems_allowed) {
		if (!mpol_allowed ||
		    (mpol_allowed && node_isset(node, *mpol_allowed)))
			nr += array[node];
	}
3417 3418 3419 3420 3421

	return nr;
}

#ifdef CONFIG_SYSCTL
3422 3423 3424 3425 3426 3427 3428 3429 3430 3431 3432 3433 3434 3435 3436 3437
static int proc_hugetlb_doulongvec_minmax(struct ctl_table *table, int write,
					  void *buffer, size_t *length,
					  loff_t *ppos, unsigned long *out)
{
	struct ctl_table dup_table;

	/*
	 * In order to avoid races with __do_proc_doulongvec_minmax(), we
	 * can duplicate the @table and alter the duplicate of it.
	 */
	dup_table = *table;
	dup_table.data = out;

	return proc_doulongvec_minmax(&dup_table, write, buffer, length, ppos);
}

3438 3439
static int hugetlb_sysctl_handler_common(bool obey_mempolicy,
			 struct ctl_table *table, int write,
3440
			 void *buffer, size_t *length, loff_t *ppos)
L
Linus Torvalds 已提交
3441
{
3442
	struct hstate *h = &default_hstate;
3443
	unsigned long tmp = h->max_huge_pages;
3444
	int ret;
3445

3446
	if (!hugepages_supported())
3447
		return -EOPNOTSUPP;
3448

3449 3450
	ret = proc_hugetlb_doulongvec_minmax(table, write, buffer, length, ppos,
					     &tmp);
3451 3452
	if (ret)
		goto out;
3453

3454 3455 3456
	if (write)
		ret = __nr_hugepages_store_common(obey_mempolicy, h,
						  NUMA_NO_NODE, tmp, *length);
3457 3458
out:
	return ret;
L
Linus Torvalds 已提交
3459
}
3460

3461
int hugetlb_sysctl_handler(struct ctl_table *table, int write,
3462
			  void *buffer, size_t *length, loff_t *ppos)
3463 3464 3465 3466 3467 3468 3469 3470
{

	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,
3471
			  void *buffer, size_t *length, loff_t *ppos)
3472 3473 3474 3475 3476 3477
{
	return hugetlb_sysctl_handler_common(true, table, write,
							buffer, length, ppos);
}
#endif /* CONFIG_NUMA */

3478
int hugetlb_overcommit_handler(struct ctl_table *table, int write,
3479
		void *buffer, size_t *length, loff_t *ppos)
3480
{
3481
	struct hstate *h = &default_hstate;
3482
	unsigned long tmp;
3483
	int ret;
3484

3485
	if (!hugepages_supported())
3486
		return -EOPNOTSUPP;
3487

3488
	tmp = h->nr_overcommit_huge_pages;
3489

3490
	if (write && hstate_is_gigantic(h))
3491 3492
		return -EINVAL;

3493 3494
	ret = proc_hugetlb_doulongvec_minmax(table, write, buffer, length, ppos,
					     &tmp);
3495 3496
	if (ret)
		goto out;
3497 3498 3499 3500 3501 3502

	if (write) {
		spin_lock(&hugetlb_lock);
		h->nr_overcommit_huge_pages = tmp;
		spin_unlock(&hugetlb_lock);
	}
3503 3504
out:
	return ret;
3505 3506
}

L
Linus Torvalds 已提交
3507 3508
#endif /* CONFIG_SYSCTL */

3509
void hugetlb_report_meminfo(struct seq_file *m)
L
Linus Torvalds 已提交
3510
{
3511 3512 3513
	struct hstate *h;
	unsigned long total = 0;

3514 3515
	if (!hugepages_supported())
		return;
3516 3517 3518 3519 3520 3521 3522 3523 3524 3525 3526 3527 3528 3529 3530 3531 3532 3533 3534 3535 3536

	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
3537 3538
}

3539
int hugetlb_report_node_meminfo(char *buf, int len, int nid)
L
Linus Torvalds 已提交
3540
{
3541
	struct hstate *h = &default_hstate;
3542

3543 3544
	if (!hugepages_supported())
		return 0;
3545 3546 3547 3548 3549 3550 3551 3552

	return sysfs_emit_at(buf, len,
			     "Node %d HugePages_Total: %5u\n"
			     "Node %d HugePages_Free:  %5u\n"
			     "Node %d HugePages_Surp:  %5u\n",
			     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
3553 3554
}

3555 3556 3557 3558 3559
void hugetlb_show_meminfo(void)
{
	struct hstate *h;
	int nid;

3560 3561 3562
	if (!hugepages_supported())
		return;

3563 3564 3565 3566 3567 3568 3569 3570 3571 3572
	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));
}

3573 3574 3575 3576 3577 3578
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
3579 3580 3581
/* Return the number pages of memory we physically have, in PAGE_SIZE units. */
unsigned long hugetlb_total_pages(void)
{
3582 3583 3584 3585 3586 3587
	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
3588 3589
}

3590
static int hugetlb_acct_memory(struct hstate *h, long delta)
M
Mel Gorman 已提交
3591 3592 3593 3594 3595 3596 3597 3598 3599 3600 3601 3602 3603 3604 3605 3606 3607 3608 3609 3610
{
	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.
3611 3612 3613 3614 3615 3616
	 *
	 * Apart from cpuset, we also have memory policy mechanism that
	 * also determines from which node the kernel will allocate memory
	 * in a NUMA system. So similar to cpuset, we also should consider
	 * the memory policy of the current task. Similar to the description
	 * above.
M
Mel Gorman 已提交
3617 3618
	 */
	if (delta > 0) {
3619
		if (gather_surplus_pages(h, delta) < 0)
M
Mel Gorman 已提交
3620 3621
			goto out;

3622
		if (delta > allowed_mems_nr(h)) {
3623
			return_unused_surplus_pages(h, delta);
M
Mel Gorman 已提交
3624 3625 3626 3627 3628 3629
			goto out;
		}
	}

	ret = 0;
	if (delta < 0)
3630
		return_unused_surplus_pages(h, (unsigned long) -delta);
M
Mel Gorman 已提交
3631 3632 3633 3634 3635 3636

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

3637 3638
static void hugetlb_vm_op_open(struct vm_area_struct *vma)
{
3639
	struct resv_map *resv = vma_resv_map(vma);
3640 3641 3642 3643 3644

	/*
	 * 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
3645
	 * has a reference to the reservation map it cannot disappear until
3646 3647 3648
	 * after this open call completes.  It is therefore safe to take a
	 * new reference here without additional locking.
	 */
3649
	if (resv && is_vma_resv_set(vma, HPAGE_RESV_OWNER))
3650
		kref_get(&resv->refs);
3651 3652
}

3653 3654
static void hugetlb_vm_op_close(struct vm_area_struct *vma)
{
3655
	struct hstate *h = hstate_vma(vma);
3656
	struct resv_map *resv = vma_resv_map(vma);
3657
	struct hugepage_subpool *spool = subpool_vma(vma);
3658
	unsigned long reserve, start, end;
3659
	long gbl_reserve;
3660

3661 3662
	if (!resv || !is_vma_resv_set(vma, HPAGE_RESV_OWNER))
		return;
3663

3664 3665
	start = vma_hugecache_offset(h, vma, vma->vm_start);
	end = vma_hugecache_offset(h, vma, vma->vm_end);
3666

3667
	reserve = (end - start) - region_count(resv, start, end);
3668
	hugetlb_cgroup_uncharge_counter(resv, start, end);
3669
	if (reserve) {
3670 3671 3672 3673 3674 3675
		/*
		 * 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);
3676
	}
3677 3678

	kref_put(&resv->refs, resv_map_release);
3679 3680
}

3681 3682 3683 3684 3685 3686 3687
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;
}

3688 3689 3690 3691 3692 3693 3694
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
3695 3696 3697 3698 3699 3700
/*
 * 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.
 */
3701
static vm_fault_t hugetlb_vm_op_fault(struct vm_fault *vmf)
L
Linus Torvalds 已提交
3702 3703
{
	BUG();
N
Nick Piggin 已提交
3704
	return 0;
L
Linus Torvalds 已提交
3705 3706
}

3707 3708 3709 3710 3711 3712 3713
/*
 * 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.
 */
3714
const struct vm_operations_struct hugetlb_vm_ops = {
N
Nick Piggin 已提交
3715
	.fault = hugetlb_vm_op_fault,
3716
	.open = hugetlb_vm_op_open,
3717
	.close = hugetlb_vm_op_close,
3718
	.split = hugetlb_vm_op_split,
3719
	.pagesize = hugetlb_vm_op_pagesize,
L
Linus Torvalds 已提交
3720 3721
};

3722 3723
static pte_t make_huge_pte(struct vm_area_struct *vma, struct page *page,
				int writable)
D
David Gibson 已提交
3724 3725 3726
{
	pte_t entry;

3727
	if (writable) {
3728 3729
		entry = huge_pte_mkwrite(huge_pte_mkdirty(mk_huge_pte(page,
					 vma->vm_page_prot)));
D
David Gibson 已提交
3730
	} else {
3731 3732
		entry = huge_pte_wrprotect(mk_huge_pte(page,
					   vma->vm_page_prot));
D
David Gibson 已提交
3733 3734 3735
	}
	entry = pte_mkyoung(entry);
	entry = pte_mkhuge(entry);
3736
	entry = arch_make_huge_pte(entry, vma, page, writable);
D
David Gibson 已提交
3737 3738 3739 3740

	return entry;
}

3741 3742 3743 3744 3745
static void set_huge_ptep_writable(struct vm_area_struct *vma,
				   unsigned long address, pte_t *ptep)
{
	pte_t entry;

3746
	entry = huge_pte_mkwrite(huge_pte_mkdirty(huge_ptep_get(ptep)));
3747
	if (huge_ptep_set_access_flags(vma, address, ptep, entry, 1))
3748
		update_mmu_cache(vma, address, ptep);
3749 3750
}

3751
bool is_hugetlb_entry_migration(pte_t pte)
3752 3753 3754 3755
{
	swp_entry_t swp;

	if (huge_pte_none(pte) || pte_present(pte))
3756
		return false;
3757
	swp = pte_to_swp_entry(pte);
3758
	if (is_migration_entry(swp))
3759
		return true;
3760
	else
3761
		return false;
3762 3763
}

3764
static bool is_hugetlb_entry_hwpoisoned(pte_t pte)
3765 3766 3767 3768
{
	swp_entry_t swp;

	if (huge_pte_none(pte) || pte_present(pte))
3769
		return false;
3770
	swp = pte_to_swp_entry(pte);
3771
	if (is_hwpoison_entry(swp))
3772
		return true;
3773
	else
3774
		return false;
3775
}
3776

D
David Gibson 已提交
3777 3778 3779
int copy_hugetlb_page_range(struct mm_struct *dst, struct mm_struct *src,
			    struct vm_area_struct *vma)
{
3780
	pte_t *src_pte, *dst_pte, entry, dst_entry;
D
David Gibson 已提交
3781
	struct page *ptepage;
3782
	unsigned long addr;
3783
	int cow;
3784 3785
	struct hstate *h = hstate_vma(vma);
	unsigned long sz = huge_page_size(h);
3786
	struct address_space *mapping = vma->vm_file->f_mapping;
3787
	struct mmu_notifier_range range;
3788
	int ret = 0;
3789 3790

	cow = (vma->vm_flags & (VM_SHARED | VM_MAYWRITE)) == VM_MAYWRITE;
D
David Gibson 已提交
3791

3792
	if (cow) {
3793
		mmu_notifier_range_init(&range, MMU_NOTIFY_CLEAR, 0, vma, src,
3794
					vma->vm_start,
3795 3796
					vma->vm_end);
		mmu_notifier_invalidate_range_start(&range);
3797 3798 3799 3800 3801 3802 3803 3804
	} 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);
3805
	}
3806

3807
	for (addr = vma->vm_start; addr < vma->vm_end; addr += sz) {
3808
		spinlock_t *src_ptl, *dst_ptl;
3809
		src_pte = huge_pte_offset(src, addr, sz);
H
Hugh Dickins 已提交
3810 3811
		if (!src_pte)
			continue;
3812
		dst_pte = huge_pte_alloc(dst, addr, sz);
3813 3814 3815 3816
		if (!dst_pte) {
			ret = -ENOMEM;
			break;
		}
3817

3818 3819 3820 3821 3822 3823 3824 3825 3826 3827 3828
		/*
		 * 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))
3829 3830
			continue;

3831 3832 3833
		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);
3834
		entry = huge_ptep_get(src_pte);
3835 3836 3837 3838 3839 3840 3841
		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.
			 */
3842 3843 3844 3845 3846 3847 3848 3849 3850 3851 3852 3853
			;
		} 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);
3854 3855
				set_huge_swap_pte_at(src, addr, src_pte,
						     entry, sz);
3856
			}
3857
			set_huge_swap_pte_at(dst, addr, dst_pte, entry, sz);
3858
		} else {
3859
			if (cow) {
3860 3861 3862 3863 3864
				/*
				 * No need to notify as we are downgrading page
				 * table protection not changing it to point
				 * to a new page.
				 *
3865
				 * See Documentation/vm/mmu_notifier.rst
3866
				 */
3867
				huge_ptep_set_wrprotect(src, addr, src_pte);
3868
			}
3869
			entry = huge_ptep_get(src_pte);
3870 3871
			ptepage = pte_page(entry);
			get_page(ptepage);
3872
			page_dup_rmap(ptepage, true);
3873
			set_huge_pte_at(dst, addr, dst_pte, entry);
3874
			hugetlb_count_add(pages_per_huge_page(h), dst);
3875
		}
3876 3877
		spin_unlock(src_ptl);
		spin_unlock(dst_ptl);
D
David Gibson 已提交
3878 3879
	}

3880
	if (cow)
3881
		mmu_notifier_invalidate_range_end(&range);
3882 3883
	else
		i_mmap_unlock_read(mapping);
3884 3885

	return ret;
D
David Gibson 已提交
3886 3887
}

3888 3889 3890
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
3891 3892 3893
{
	struct mm_struct *mm = vma->vm_mm;
	unsigned long address;
3894
	pte_t *ptep;
D
David Gibson 已提交
3895
	pte_t pte;
3896
	spinlock_t *ptl;
D
David Gibson 已提交
3897
	struct page *page;
3898 3899
	struct hstate *h = hstate_vma(vma);
	unsigned long sz = huge_page_size(h);
3900
	struct mmu_notifier_range range;
3901

D
David Gibson 已提交
3902
	WARN_ON(!is_vm_hugetlb_page(vma));
3903 3904
	BUG_ON(start & ~huge_page_mask(h));
	BUG_ON(end & ~huge_page_mask(h));
D
David Gibson 已提交
3905

3906 3907 3908 3909
	/*
	 * This is a hugetlb vma, all the pte entries should point
	 * to huge page.
	 */
3910
	tlb_change_page_size(tlb, sz);
3911
	tlb_start_vma(tlb, vma);
3912 3913 3914 3915

	/*
	 * If sharing possible, alert mmu notifiers of worst case.
	 */
3916 3917
	mmu_notifier_range_init(&range, MMU_NOTIFY_UNMAP, 0, vma, mm, start,
				end);
3918 3919
	adjust_range_if_pmd_sharing_possible(vma, &range.start, &range.end);
	mmu_notifier_invalidate_range_start(&range);
3920 3921
	address = start;
	for (; address < end; address += sz) {
3922
		ptep = huge_pte_offset(mm, address, sz);
A
Adam Litke 已提交
3923
		if (!ptep)
3924 3925
			continue;

3926
		ptl = huge_pte_lock(h, mm, ptep);
3927
		if (huge_pmd_unshare(mm, vma, &address, ptep)) {
3928
			spin_unlock(ptl);
3929 3930 3931 3932
			/*
			 * We just unmapped a page of PMDs by clearing a PUD.
			 * The caller's TLB flush range should cover this area.
			 */
3933 3934
			continue;
		}
3935

3936
		pte = huge_ptep_get(ptep);
3937 3938 3939 3940
		if (huge_pte_none(pte)) {
			spin_unlock(ptl);
			continue;
		}
3941 3942

		/*
3943 3944
		 * Migrating hugepage or HWPoisoned hugepage is already
		 * unmapped and its refcount is dropped, so just clear pte here.
3945
		 */
3946
		if (unlikely(!pte_present(pte))) {
3947
			huge_pte_clear(mm, address, ptep, sz);
3948 3949
			spin_unlock(ptl);
			continue;
3950
		}
3951 3952

		page = pte_page(pte);
3953 3954 3955 3956 3957 3958
		/*
		 * 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) {
3959 3960 3961 3962
			if (page != ref_page) {
				spin_unlock(ptl);
				continue;
			}
3963 3964 3965 3966 3967 3968 3969 3970
			/*
			 * 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);
		}

3971
		pte = huge_ptep_get_and_clear(mm, address, ptep);
3972
		tlb_remove_huge_tlb_entry(h, tlb, ptep, address);
3973
		if (huge_pte_dirty(pte))
3974
			set_page_dirty(page);
3975

3976
		hugetlb_count_sub(pages_per_huge_page(h), mm);
3977
		page_remove_rmap(page, true);
3978

3979
		spin_unlock(ptl);
3980
		tlb_remove_page_size(tlb, page, huge_page_size(h));
3981 3982 3983 3984 3985
		/*
		 * Bail out after unmapping reference page if supplied
		 */
		if (ref_page)
			break;
3986
	}
3987
	mmu_notifier_invalidate_range_end(&range);
3988
	tlb_end_vma(tlb, vma);
L
Linus Torvalds 已提交
3989
}
D
David Gibson 已提交
3990

3991 3992 3993 3994 3995 3996 3997 3998 3999 4000 4001 4002
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
4003
	 * is to clear it before releasing the i_mmap_rwsem. This works
4004
	 * because in the context this is called, the VMA is about to be
4005
	 * destroyed and the i_mmap_rwsem is held.
4006 4007 4008 4009
	 */
	vma->vm_flags &= ~VM_MAYSHARE;
}

4010
void unmap_hugepage_range(struct vm_area_struct *vma, unsigned long start,
4011
			  unsigned long end, struct page *ref_page)
4012
{
4013 4014
	struct mm_struct *mm;
	struct mmu_gather tlb;
4015 4016 4017 4018 4019 4020 4021 4022 4023 4024 4025
	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);
4026 4027 4028

	mm = vma->vm_mm;

4029
	tlb_gather_mmu(&tlb, mm, tlb_start, tlb_end);
4030
	__unmap_hugepage_range(&tlb, vma, start, end, ref_page);
4031
	tlb_finish_mmu(&tlb, tlb_start, tlb_end);
4032 4033
}

4034 4035 4036 4037 4038 4039
/*
 * 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.
 */
4040 4041
static void unmap_ref_private(struct mm_struct *mm, struct vm_area_struct *vma,
			      struct page *page, unsigned long address)
4042
{
4043
	struct hstate *h = hstate_vma(vma);
4044 4045 4046 4047 4048 4049 4050 4051
	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.
	 */
4052
	address = address & huge_page_mask(h);
4053 4054
	pgoff = ((address - vma->vm_start) >> PAGE_SHIFT) +
			vma->vm_pgoff;
4055
	mapping = vma->vm_file->f_mapping;
4056

4057 4058 4059 4060 4061
	/*
	 * 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
	 */
4062
	i_mmap_lock_write(mapping);
4063
	vma_interval_tree_foreach(iter_vma, &mapping->i_mmap, pgoff, pgoff) {
4064 4065 4066 4067
		/* Do not unmap the current VMA */
		if (iter_vma == vma)
			continue;

4068 4069 4070 4071 4072 4073 4074 4075
		/*
		 * 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;

4076 4077 4078 4079 4080 4081 4082 4083
		/*
		 * 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))
4084 4085
			unmap_hugepage_range(iter_vma, address,
					     address + huge_page_size(h), page);
4086
	}
4087
	i_mmap_unlock_write(mapping);
4088 4089
}

4090 4091
/*
 * Hugetlb_cow() should be called with page lock of the original hugepage held.
4092 4093 4094
 * 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.
4095
 */
4096
static vm_fault_t hugetlb_cow(struct mm_struct *mm, struct vm_area_struct *vma,
4097
		       unsigned long address, pte_t *ptep,
4098
		       struct page *pagecache_page, spinlock_t *ptl)
4099
{
4100
	pte_t pte;
4101
	struct hstate *h = hstate_vma(vma);
4102
	struct page *old_page, *new_page;
4103 4104
	int outside_reserve = 0;
	vm_fault_t ret = 0;
4105
	unsigned long haddr = address & huge_page_mask(h);
4106
	struct mmu_notifier_range range;
4107

4108
	pte = huge_ptep_get(ptep);
4109 4110
	old_page = pte_page(pte);

4111
retry_avoidcopy:
4112 4113
	/* If no-one else is actually using this page, avoid the copy
	 * and just make the page writable */
4114
	if (page_mapcount(old_page) == 1 && PageAnon(old_page)) {
4115
		page_move_anon_rmap(old_page, vma);
4116
		set_huge_ptep_writable(vma, haddr, ptep);
N
Nick Piggin 已提交
4117
		return 0;
4118 4119
	}

4120 4121 4122 4123 4124 4125 4126 4127 4128
	/*
	 * 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.
	 */
4129
	if (is_vma_resv_set(vma, HPAGE_RESV_OWNER) &&
4130 4131 4132
			old_page != pagecache_page)
		outside_reserve = 1;

4133
	get_page(old_page);
4134

4135 4136 4137 4138
	/*
	 * Drop page table lock as buddy allocator may be called. It will
	 * be acquired again before returning to the caller, as expected.
	 */
4139
	spin_unlock(ptl);
4140
	new_page = alloc_huge_page(vma, haddr, outside_reserve);
4141

4142
	if (IS_ERR(new_page)) {
4143 4144 4145 4146 4147 4148 4149 4150
		/*
		 * 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) {
4151 4152 4153 4154
			struct address_space *mapping = vma->vm_file->f_mapping;
			pgoff_t idx;
			u32 hash;

4155
			put_page(old_page);
4156
			BUG_ON(huge_pte_none(pte));
4157 4158 4159 4160 4161 4162 4163 4164 4165 4166 4167 4168 4169 4170
			/*
			 * Drop hugetlb_fault_mutex and i_mmap_rwsem before
			 * unmapping.  unmapping needs to hold i_mmap_rwsem
			 * in write mode.  Dropping i_mmap_rwsem in read mode
			 * here is OK as COW mappings do not interact with
			 * PMD sharing.
			 *
			 * Reacquire both after unmap operation.
			 */
			idx = vma_hugecache_offset(h, vma, haddr);
			hash = hugetlb_fault_mutex_hash(mapping, idx);
			mutex_unlock(&hugetlb_fault_mutex_table[hash]);
			i_mmap_unlock_read(mapping);

4171
			unmap_ref_private(mm, vma, old_page, haddr);
4172 4173 4174

			i_mmap_lock_read(mapping);
			mutex_lock(&hugetlb_fault_mutex_table[hash]);
4175
			spin_lock(ptl);
4176
			ptep = huge_pte_offset(mm, haddr, huge_page_size(h));
4177 4178 4179 4180 4181 4182 4183 4184
			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;
4185 4186
		}

4187
		ret = vmf_error(PTR_ERR(new_page));
4188
		goto out_release_old;
4189 4190
	}

4191 4192 4193 4194
	/*
	 * When the original hugepage is shared one, it does not have
	 * anon_vma prepared.
	 */
4195
	if (unlikely(anon_vma_prepare(vma))) {
4196 4197
		ret = VM_FAULT_OOM;
		goto out_release_all;
4198
	}
4199

4200
	copy_user_huge_page(new_page, old_page, address, vma,
A
Andrea Arcangeli 已提交
4201
			    pages_per_huge_page(h));
N
Nick Piggin 已提交
4202
	__SetPageUptodate(new_page);
4203

4204
	mmu_notifier_range_init(&range, MMU_NOTIFY_CLEAR, 0, vma, mm, haddr,
4205
				haddr + huge_page_size(h));
4206
	mmu_notifier_invalidate_range_start(&range);
4207

4208
	/*
4209
	 * Retake the page table lock to check for racing updates
4210 4211
	 * before the page tables are altered
	 */
4212
	spin_lock(ptl);
4213
	ptep = huge_pte_offset(mm, haddr, huge_page_size(h));
4214
	if (likely(ptep && pte_same(huge_ptep_get(ptep), pte))) {
4215 4216
		ClearPagePrivate(new_page);

4217
		/* Break COW */
4218
		huge_ptep_clear_flush(vma, haddr, ptep);
4219
		mmu_notifier_invalidate_range(mm, range.start, range.end);
4220
		set_huge_pte_at(mm, haddr, ptep,
4221
				make_huge_pte(vma, new_page, 1));
4222
		page_remove_rmap(old_page, true);
4223
		hugepage_add_new_anon_rmap(new_page, vma, haddr);
4224
		set_page_huge_active(new_page);
4225 4226 4227
		/* Make the old page be freed below */
		new_page = old_page;
	}
4228
	spin_unlock(ptl);
4229
	mmu_notifier_invalidate_range_end(&range);
4230
out_release_all:
4231
	restore_reserve_on_error(h, vma, haddr, new_page);
4232
	put_page(new_page);
4233
out_release_old:
4234
	put_page(old_page);
4235

4236 4237
	spin_lock(ptl); /* Caller expects lock to be held */
	return ret;
4238 4239
}

4240
/* Return the pagecache page at a given address within a VMA */
4241 4242
static struct page *hugetlbfs_pagecache_page(struct hstate *h,
			struct vm_area_struct *vma, unsigned long address)
4243 4244
{
	struct address_space *mapping;
4245
	pgoff_t idx;
4246 4247

	mapping = vma->vm_file->f_mapping;
4248
	idx = vma_hugecache_offset(h, vma, address);
4249 4250 4251 4252

	return find_lock_page(mapping, idx);
}

H
Hugh Dickins 已提交
4253 4254 4255 4256 4257
/*
 * 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
4258 4259 4260 4261 4262 4263 4264 4265 4266 4267 4268 4269 4270 4271 4272
			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;
}

4273 4274 4275 4276 4277 4278 4279 4280 4281 4282 4283
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);

4284 4285 4286 4287 4288 4289
	/*
	 * set page dirty so that it will not be removed from cache/file
	 * by non-hugetlbfs specific code paths.
	 */
	set_page_dirty(page);

4290 4291 4292 4293 4294 4295
	spin_lock(&inode->i_lock);
	inode->i_blocks += blocks_per_huge_page(h);
	spin_unlock(&inode->i_lock);
	return 0;
}

4296 4297 4298 4299
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)
4300
{
4301
	struct hstate *h = hstate_vma(vma);
4302
	vm_fault_t ret = VM_FAULT_SIGBUS;
4303
	int anon_rmap = 0;
A
Adam Litke 已提交
4304 4305
	unsigned long size;
	struct page *page;
4306
	pte_t new_pte;
4307
	spinlock_t *ptl;
4308
	unsigned long haddr = address & huge_page_mask(h);
4309
	bool new_page = false;
A
Adam Litke 已提交
4310

4311 4312 4313
	/*
	 * 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
4314
	 * COW. Warn that such a situation has occurred as it may not be obvious
4315 4316
	 */
	if (is_vma_resv_set(vma, HPAGE_RESV_UNMAPPED)) {
4317
		pr_warn_ratelimited("PID %d killed due to inadequate hugepage pool\n",
4318
			   current->pid);
4319 4320 4321
		return ret;
	}

A
Adam Litke 已提交
4322
	/*
4323 4324 4325
	 * 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
4326
	 */
4327 4328 4329 4330
	size = i_size_read(mapping->host) >> huge_page_shift(h);
	if (idx >= size)
		goto out;

4331 4332 4333
retry:
	page = find_lock_page(mapping, idx);
	if (!page) {
4334 4335 4336 4337 4338 4339 4340
		/*
		 * Check for page in userfault range
		 */
		if (userfaultfd_missing(vma)) {
			u32 hash;
			struct vm_fault vmf = {
				.vma = vma,
4341
				.address = haddr,
4342 4343 4344 4345 4346 4347 4348 4349 4350 4351 4352
				.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
				 */
			};

			/*
4353 4354 4355
			 * hugetlb_fault_mutex and i_mmap_rwsem must be
			 * dropped before handling userfault.  Reacquire
			 * after handling fault to make calling code simpler.
4356
			 */
4357
			hash = hugetlb_fault_mutex_hash(mapping, idx);
4358
			mutex_unlock(&hugetlb_fault_mutex_table[hash]);
4359
			i_mmap_unlock_read(mapping);
4360
			ret = handle_userfault(&vmf, VM_UFFD_MISSING);
4361
			i_mmap_lock_read(mapping);
4362 4363 4364 4365
			mutex_lock(&hugetlb_fault_mutex_table[hash]);
			goto out;
		}

4366
		page = alloc_huge_page(vma, haddr, 0);
4367
		if (IS_ERR(page)) {
4368 4369 4370 4371 4372 4373 4374 4375 4376 4377 4378 4379 4380 4381 4382 4383 4384 4385 4386
			/*
			 * 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);
4387
			ret = vmf_error(PTR_ERR(page));
4388 4389
			goto out;
		}
A
Andrea Arcangeli 已提交
4390
		clear_huge_page(page, address, pages_per_huge_page(h));
N
Nick Piggin 已提交
4391
		__SetPageUptodate(page);
4392
		new_page = true;
4393

4394
		if (vma->vm_flags & VM_MAYSHARE) {
4395
			int err = huge_add_to_page_cache(page, mapping, idx);
4396 4397 4398 4399 4400 4401
			if (err) {
				put_page(page);
				if (err == -EEXIST)
					goto retry;
				goto out;
			}
4402
		} else {
4403
			lock_page(page);
4404 4405 4406 4407
			if (unlikely(anon_vma_prepare(vma))) {
				ret = VM_FAULT_OOM;
				goto backout_unlocked;
			}
4408
			anon_rmap = 1;
4409
		}
4410
	} else {
4411 4412 4413 4414 4415 4416
		/*
		 * 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))) {
4417
			ret = VM_FAULT_HWPOISON_LARGE |
4418
				VM_FAULT_SET_HINDEX(hstate_index(h));
4419 4420
			goto backout_unlocked;
		}
4421
	}
4422

4423 4424 4425 4426 4427 4428
	/*
	 * 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.
	 */
4429
	if ((flags & FAULT_FLAG_WRITE) && !(vma->vm_flags & VM_SHARED)) {
4430
		if (vma_needs_reservation(h, vma, haddr) < 0) {
4431 4432 4433
			ret = VM_FAULT_OOM;
			goto backout_unlocked;
		}
4434
		/* Just decrements count, does not deallocate */
4435
		vma_end_reservation(h, vma, haddr);
4436
	}
4437

4438
	ptl = huge_pte_lock(h, mm, ptep);
N
Nick Piggin 已提交
4439
	ret = 0;
4440
	if (!huge_pte_none(huge_ptep_get(ptep)))
A
Adam Litke 已提交
4441 4442
		goto backout;

4443 4444
	if (anon_rmap) {
		ClearPagePrivate(page);
4445
		hugepage_add_new_anon_rmap(page, vma, haddr);
4446
	} else
4447
		page_dup_rmap(page, true);
4448 4449
	new_pte = make_huge_pte(vma, page, ((vma->vm_flags & VM_WRITE)
				&& (vma->vm_flags & VM_SHARED)));
4450
	set_huge_pte_at(mm, haddr, ptep, new_pte);
4451

4452
	hugetlb_count_add(pages_per_huge_page(h), mm);
4453
	if ((flags & FAULT_FLAG_WRITE) && !(vma->vm_flags & VM_SHARED)) {
4454
		/* Optimization, do the COW without a second fault */
4455
		ret = hugetlb_cow(mm, vma, address, ptep, page, ptl);
4456 4457
	}

4458
	spin_unlock(ptl);
4459 4460 4461 4462 4463 4464 4465 4466 4467

	/*
	 * 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
4468 4469
	unlock_page(page);
out:
4470
	return ret;
A
Adam Litke 已提交
4471 4472

backout:
4473
	spin_unlock(ptl);
4474
backout_unlocked:
A
Adam Litke 已提交
4475
	unlock_page(page);
4476
	restore_reserve_on_error(h, vma, haddr, page);
A
Adam Litke 已提交
4477 4478
	put_page(page);
	goto out;
4479 4480
}

4481
#ifdef CONFIG_SMP
4482
u32 hugetlb_fault_mutex_hash(struct address_space *mapping, pgoff_t idx)
4483 4484 4485 4486
{
	unsigned long key[2];
	u32 hash;

4487 4488
	key[0] = (unsigned long) mapping;
	key[1] = idx;
4489

4490
	hash = jhash2((u32 *)&key, sizeof(key)/(sizeof(u32)), 0);
4491 4492 4493 4494 4495 4496 4497 4498

	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.
 */
4499
u32 hugetlb_fault_mutex_hash(struct address_space *mapping, pgoff_t idx)
4500 4501 4502 4503 4504
{
	return 0;
}
#endif

4505
vm_fault_t hugetlb_fault(struct mm_struct *mm, struct vm_area_struct *vma,
4506
			unsigned long address, unsigned int flags)
4507
{
4508
	pte_t *ptep, entry;
4509
	spinlock_t *ptl;
4510
	vm_fault_t ret;
4511 4512
	u32 hash;
	pgoff_t idx;
4513
	struct page *page = NULL;
4514
	struct page *pagecache_page = NULL;
4515
	struct hstate *h = hstate_vma(vma);
4516
	struct address_space *mapping;
4517
	int need_wait_lock = 0;
4518
	unsigned long haddr = address & huge_page_mask(h);
4519

4520
	ptep = huge_pte_offset(mm, haddr, huge_page_size(h));
4521
	if (ptep) {
4522 4523 4524 4525 4526
		/*
		 * 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.
		 */
4527
		entry = huge_ptep_get(ptep);
N
Naoya Horiguchi 已提交
4528
		if (unlikely(is_hugetlb_entry_migration(entry))) {
4529
			migration_entry_wait_huge(vma, mm, ptep);
N
Naoya Horiguchi 已提交
4530 4531
			return 0;
		} else if (unlikely(is_hugetlb_entry_hwpoisoned(entry)))
4532
			return VM_FAULT_HWPOISON_LARGE |
4533
				VM_FAULT_SET_HINDEX(hstate_index(h));
4534 4535
	}

4536 4537
	/*
	 * Acquire i_mmap_rwsem before calling huge_pte_alloc and hold
4538 4539 4540 4541
	 * 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.
4542 4543 4544 4545 4546
	 *
	 * 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.
	 */
4547
	mapping = vma->vm_file->f_mapping;
4548 4549 4550 4551 4552 4553
	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;
	}
4554

4555 4556 4557 4558 4559
	/*
	 * 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.
	 */
4560
	idx = vma_hugecache_offset(h, vma, haddr);
4561
	hash = hugetlb_fault_mutex_hash(mapping, idx);
4562
	mutex_lock(&hugetlb_fault_mutex_table[hash]);
4563

4564 4565
	entry = huge_ptep_get(ptep);
	if (huge_pte_none(entry)) {
4566
		ret = hugetlb_no_page(mm, vma, mapping, idx, address, ptep, flags);
4567
		goto out_mutex;
4568
	}
4569

N
Nick Piggin 已提交
4570
	ret = 0;
4571

4572 4573 4574
	/*
	 * entry could be a migration/hwpoison entry at this point, so this
	 * check prevents the kernel from going below assuming that we have
E
Ethon Paul 已提交
4575 4576 4577
	 * an active hugepage in pagecache. This goto expects the 2nd page
	 * fault, and is_hugetlb_entry_(migration|hwpoisoned) check will
	 * properly handle it.
4578 4579 4580 4581
	 */
	if (!pte_present(entry))
		goto out_mutex;

4582 4583 4584 4585 4586 4587 4588 4589
	/*
	 * 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.
	 */
4590
	if ((flags & FAULT_FLAG_WRITE) && !huge_pte_write(entry)) {
4591
		if (vma_needs_reservation(h, vma, haddr) < 0) {
4592
			ret = VM_FAULT_OOM;
4593
			goto out_mutex;
4594
		}
4595
		/* Just decrements count, does not deallocate */
4596
		vma_end_reservation(h, vma, haddr);
4597

4598
		if (!(vma->vm_flags & VM_MAYSHARE))
4599
			pagecache_page = hugetlbfs_pagecache_page(h,
4600
								vma, haddr);
4601 4602
	}

4603 4604 4605 4606 4607 4608
	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;

4609 4610 4611 4612 4613 4614 4615
	/*
	 * 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)
4616 4617 4618 4619
		if (!trylock_page(page)) {
			need_wait_lock = 1;
			goto out_ptl;
		}
4620

4621
	get_page(page);
4622

4623
	if (flags & FAULT_FLAG_WRITE) {
4624
		if (!huge_pte_write(entry)) {
4625
			ret = hugetlb_cow(mm, vma, address, ptep,
4626
					  pagecache_page, ptl);
4627
			goto out_put_page;
4628
		}
4629
		entry = huge_pte_mkdirty(entry);
4630 4631
	}
	entry = pte_mkyoung(entry);
4632
	if (huge_ptep_set_access_flags(vma, haddr, ptep, entry,
4633
						flags & FAULT_FLAG_WRITE))
4634
		update_mmu_cache(vma, haddr, ptep);
4635 4636 4637 4638
out_put_page:
	if (page != pagecache_page)
		unlock_page(page);
	put_page(page);
4639 4640
out_ptl:
	spin_unlock(ptl);
4641 4642 4643 4644 4645

	if (pagecache_page) {
		unlock_page(pagecache_page);
		put_page(pagecache_page);
	}
4646
out_mutex:
4647
	mutex_unlock(&hugetlb_fault_mutex_table[hash]);
4648
	i_mmap_unlock_read(mapping);
4649 4650 4651 4652 4653 4654 4655 4656 4657
	/*
	 * 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);
4658
	return ret;
4659 4660
}

4661 4662 4663 4664 4665 4666 4667 4668 4669 4670 4671
/*
 * 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)
{
4672 4673 4674
	struct address_space *mapping;
	pgoff_t idx;
	unsigned long size;
4675
	int vm_shared = dst_vma->vm_flags & VM_SHARED;
4676 4677 4678 4679 4680 4681 4682 4683 4684 4685 4686 4687 4688 4689
	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,
4690
						pages_per_huge_page(h), false);
4691

4692
		/* fallback to copy_from_user outside mmap_lock */
4693
		if (unlikely(ret)) {
4694
			ret = -ENOENT;
4695 4696 4697 4698 4699 4700 4701 4702 4703 4704 4705 4706 4707 4708 4709 4710
			*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);

4711 4712 4713
	mapping = dst_vma->vm_file->f_mapping;
	idx = vma_hugecache_offset(h, dst_vma, dst_addr);

4714 4715 4716 4717
	/*
	 * If shared, add to page cache
	 */
	if (vm_shared) {
4718 4719 4720 4721
		size = i_size_read(mapping->host) >> huge_page_shift(h);
		ret = -EFAULT;
		if (idx >= size)
			goto out_release_nounlock;
4722

4723 4724 4725 4726 4727 4728
		/*
		 * 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.
		 */
4729 4730 4731 4732 4733
		ret = huge_add_to_page_cache(page, mapping, idx);
		if (ret)
			goto out_release_nounlock;
	}

4734 4735 4736
	ptl = huge_pte_lockptr(h, dst_mm, dst_pte);
	spin_lock(ptl);

4737 4738 4739 4740 4741 4742 4743 4744 4745 4746 4747 4748 4749 4750
	/*
	 * 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;

4751 4752 4753 4754
	ret = -EEXIST;
	if (!huge_pte_none(huge_ptep_get(dst_pte)))
		goto out_release_unlock;

4755 4756 4757 4758 4759 4760
	if (vm_shared) {
		page_dup_rmap(page, true);
	} else {
		ClearPagePrivate(page);
		hugepage_add_new_anon_rmap(page, dst_vma, dst_addr);
	}
4761 4762 4763 4764 4765 4766 4767 4768 4769 4770 4771 4772 4773 4774 4775 4776

	_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);
4777
	set_page_huge_active(page);
4778 4779
	if (vm_shared)
		unlock_page(page);
4780 4781 4782 4783 4784
	ret = 0;
out:
	return ret;
out_release_unlock:
	spin_unlock(ptl);
4785 4786
	if (vm_shared)
		unlock_page(page);
4787
out_release_nounlock:
4788 4789 4790 4791
	put_page(page);
	goto out;
}

4792 4793 4794
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,
4795
			 long i, unsigned int flags, int *locked)
D
David Gibson 已提交
4796
{
4797 4798
	unsigned long pfn_offset;
	unsigned long vaddr = *position;
4799
	unsigned long remainder = *nr_pages;
4800
	struct hstate *h = hstate_vma(vma);
4801
	int err = -EFAULT;
D
David Gibson 已提交
4802 4803

	while (vaddr < vma->vm_end && remainder) {
A
Adam Litke 已提交
4804
		pte_t *pte;
4805
		spinlock_t *ptl = NULL;
H
Hugh Dickins 已提交
4806
		int absent;
A
Adam Litke 已提交
4807
		struct page *page;
D
David Gibson 已提交
4808

4809 4810 4811 4812
		/*
		 * If we have a pending SIGKILL, don't keep faulting pages and
		 * potentially allocating memory.
		 */
4813
		if (fatal_signal_pending(current)) {
4814 4815 4816 4817
			remainder = 0;
			break;
		}

A
Adam Litke 已提交
4818 4819
		/*
		 * Some archs (sparc64, sh*) have multiple pte_ts to
H
Hugh Dickins 已提交
4820
		 * each hugepage.  We have to make sure we get the
A
Adam Litke 已提交
4821
		 * first, for the page indexing below to work.
4822 4823
		 *
		 * Note that page table lock is not held when pte is null.
A
Adam Litke 已提交
4824
		 */
4825 4826
		pte = huge_pte_offset(mm, vaddr & huge_page_mask(h),
				      huge_page_size(h));
4827 4828
		if (pte)
			ptl = huge_pte_lock(h, mm, pte);
H
Hugh Dickins 已提交
4829 4830 4831 4832
		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
4833 4834 4835 4836
		 * 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
4837
		 */
H
Hugh Dickins 已提交
4838 4839
		if (absent && (flags & FOLL_DUMP) &&
		    !hugetlbfs_pagecache_present(h, vma, vaddr)) {
4840 4841
			if (pte)
				spin_unlock(ptl);
H
Hugh Dickins 已提交
4842 4843 4844
			remainder = 0;
			break;
		}
D
David Gibson 已提交
4845

4846 4847 4848 4849 4850 4851 4852 4853 4854 4855 4856
		/*
		 * 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)) ||
4857 4858
		    ((flags & FOLL_WRITE) &&
		      !huge_pte_write(huge_ptep_get(pte)))) {
4859
			vm_fault_t ret;
4860
			unsigned int fault_flags = 0;
D
David Gibson 已提交
4861

4862 4863
			if (pte)
				spin_unlock(ptl);
4864 4865
			if (flags & FOLL_WRITE)
				fault_flags |= FAULT_FLAG_WRITE;
4866
			if (locked)
4867 4868
				fault_flags |= FAULT_FLAG_ALLOW_RETRY |
					FAULT_FLAG_KILLABLE;
4869 4870 4871 4872
			if (flags & FOLL_NOWAIT)
				fault_flags |= FAULT_FLAG_ALLOW_RETRY |
					FAULT_FLAG_RETRY_NOWAIT;
			if (flags & FOLL_TRIED) {
4873 4874 4875 4876
				/*
				 * Note: FAULT_FLAG_ALLOW_RETRY and
				 * FAULT_FLAG_TRIED can co-exist
				 */
4877 4878 4879 4880
				fault_flags |= FAULT_FLAG_TRIED;
			}
			ret = hugetlb_fault(mm, vma, vaddr, fault_flags);
			if (ret & VM_FAULT_ERROR) {
4881
				err = vm_fault_to_errno(ret, flags);
4882 4883 4884 4885
				remainder = 0;
				break;
			}
			if (ret & VM_FAULT_RETRY) {
4886
				if (locked &&
4887
				    !(fault_flags & FAULT_FLAG_RETRY_NOWAIT))
4888
					*locked = 0;
4889 4890 4891 4892 4893 4894 4895 4896 4897 4898 4899 4900 4901
				*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
4902 4903
		}

4904
		pfn_offset = (vaddr & ~huge_page_mask(h)) >> PAGE_SHIFT;
4905
		page = pte_page(huge_ptep_get(pte));
4906

4907 4908 4909 4910 4911 4912 4913 4914 4915 4916 4917 4918 4919 4920
		/*
		 * 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;
		}

4921
same_page:
4922
		if (pages) {
H
Hugh Dickins 已提交
4923
			pages[i] = mem_map_offset(page, pfn_offset);
J
John Hubbard 已提交
4924 4925 4926 4927 4928 4929 4930 4931 4932 4933 4934 4935 4936 4937 4938 4939
			/*
			 * 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;
			}
4940
		}
D
David Gibson 已提交
4941 4942 4943 4944 4945

		if (vmas)
			vmas[i] = vma;

		vaddr += PAGE_SIZE;
4946
		++pfn_offset;
D
David Gibson 已提交
4947 4948
		--remainder;
		++i;
4949
		if (vaddr < vma->vm_end && remainder &&
4950
				pfn_offset < pages_per_huge_page(h)) {
4951 4952 4953 4954 4955 4956
			/*
			 * We use pfn_offset to avoid touching the pageframes
			 * of this compound page.
			 */
			goto same_page;
		}
4957
		spin_unlock(ptl);
D
David Gibson 已提交
4958
	}
4959
	*nr_pages = remainder;
4960 4961 4962 4963 4964
	/*
	 * 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
4965 4966
	*position = vaddr;

4967
	return i ? i : err;
D
David Gibson 已提交
4968
}
4969

4970 4971 4972 4973 4974 4975 4976 4977
#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

4978
unsigned long hugetlb_change_protection(struct vm_area_struct *vma,
4979 4980 4981 4982 4983 4984
		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;
4985
	struct hstate *h = hstate_vma(vma);
4986
	unsigned long pages = 0;
4987
	bool shared_pmd = false;
4988
	struct mmu_notifier_range range;
4989 4990 4991

	/*
	 * In the case of shared PMDs, the area to flush could be beyond
4992
	 * start/end.  Set range.start/range.end to cover the maximum possible
4993 4994
	 * range if PMD sharing is possible.
	 */
4995 4996
	mmu_notifier_range_init(&range, MMU_NOTIFY_PROTECTION_VMA,
				0, vma, mm, start, end);
4997
	adjust_range_if_pmd_sharing_possible(vma, &range.start, &range.end);
4998 4999

	BUG_ON(address >= end);
5000
	flush_cache_range(vma, range.start, range.end);
5001

5002
	mmu_notifier_invalidate_range_start(&range);
5003
	i_mmap_lock_write(vma->vm_file->f_mapping);
5004
	for (; address < end; address += huge_page_size(h)) {
5005
		spinlock_t *ptl;
5006
		ptep = huge_pte_offset(mm, address, huge_page_size(h));
5007 5008
		if (!ptep)
			continue;
5009
		ptl = huge_pte_lock(h, mm, ptep);
5010
		if (huge_pmd_unshare(mm, vma, &address, ptep)) {
5011
			pages++;
5012
			spin_unlock(ptl);
5013
			shared_pmd = true;
5014
			continue;
5015
		}
5016 5017 5018 5019 5020 5021 5022 5023 5024 5025 5026 5027 5028
		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);
5029 5030
				set_huge_swap_pte_at(mm, address, ptep,
						     newpte, huge_page_size(h));
5031 5032 5033 5034 5035 5036
				pages++;
			}
			spin_unlock(ptl);
			continue;
		}
		if (!huge_pte_none(pte)) {
5037 5038 5039 5040
			pte_t old_pte;

			old_pte = huge_ptep_modify_prot_start(vma, address, ptep);
			pte = pte_mkhuge(huge_pte_modify(old_pte, newprot));
5041
			pte = arch_make_huge_pte(pte, vma, NULL, 0);
5042
			huge_ptep_modify_prot_commit(vma, address, ptep, old_pte, pte);
5043
			pages++;
5044
		}
5045
		spin_unlock(ptl);
5046
	}
5047
	/*
5048
	 * Must flush TLB before releasing i_mmap_rwsem: x86's huge_pmd_unshare
5049
	 * may have cleared our pud entry and done put_page on the page table:
5050
	 * once we release i_mmap_rwsem, another task can do the final put_page
5051 5052
	 * 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.
5053
	 */
5054
	if (shared_pmd)
5055
		flush_hugetlb_tlb_range(vma, range.start, range.end);
5056 5057
	else
		flush_hugetlb_tlb_range(vma, start, end);
5058 5059 5060 5061
	/*
	 * No need to call mmu_notifier_invalidate_range() we are downgrading
	 * page table protection not changing it to point to a new page.
	 *
5062
	 * See Documentation/vm/mmu_notifier.rst
5063
	 */
5064
	i_mmap_unlock_write(vma->vm_file->f_mapping);
5065
	mmu_notifier_invalidate_range_end(&range);
5066 5067

	return pages << h->order;
5068 5069
}

5070 5071
int hugetlb_reserve_pages(struct inode *inode,
					long from, long to,
5072
					struct vm_area_struct *vma,
5073
					vm_flags_t vm_flags)
5074
{
5075
	long ret, chg, add = -1;
5076
	struct hstate *h = hstate_inode(inode);
5077
	struct hugepage_subpool *spool = subpool_inode(inode);
5078
	struct resv_map *resv_map;
5079
	struct hugetlb_cgroup *h_cg = NULL;
5080
	long gbl_reserve, regions_needed = 0;
5081

5082 5083 5084 5085 5086 5087
	/* This should never happen */
	if (from > to) {
		VM_WARN(1, "%s called with a negative range\n", __func__);
		return -EINVAL;
	}

5088 5089 5090
	/*
	 * Only apply hugepage reservation if asked. At fault time, an
	 * attempt will be made for VM_NORESERVE to allocate a page
5091
	 * without using reserves
5092
	 */
5093
	if (vm_flags & VM_NORESERVE)
5094 5095
		return 0;

5096 5097 5098 5099 5100 5101
	/*
	 * 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
	 */
5102
	if (!vma || vma->vm_flags & VM_MAYSHARE) {
5103 5104 5105 5106 5107
		/*
		 * 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).
		 */
5108
		resv_map = inode_resv_map(inode);
5109

5110
		chg = region_chg(resv_map, from, to, &regions_needed);
5111 5112

	} else {
5113
		/* Private mapping. */
5114
		resv_map = resv_map_alloc();
5115 5116 5117
		if (!resv_map)
			return -ENOMEM;

5118
		chg = to - from;
5119

5120 5121 5122 5123
		set_vma_resv_map(vma, resv_map);
		set_vma_resv_flags(vma, HPAGE_RESV_OWNER);
	}

5124 5125 5126 5127
	if (chg < 0) {
		ret = chg;
		goto out_err;
	}
5128

5129 5130 5131 5132 5133 5134 5135 5136 5137 5138 5139 5140 5141 5142 5143
	ret = hugetlb_cgroup_charge_cgroup_rsvd(
		hstate_index(h), chg * pages_per_huge_page(h), &h_cg);

	if (ret < 0) {
		ret = -ENOMEM;
		goto out_err;
	}

	if (vma && !(vma->vm_flags & VM_MAYSHARE) && h_cg) {
		/* For private mappings, the hugetlb_cgroup uncharge info hangs
		 * of the resv_map.
		 */
		resv_map_set_hugetlb_cgroup_uncharge_info(resv_map, h_cg, h);
	}

5144 5145 5146 5147 5148 5149 5150
	/*
	 * 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) {
5151
		ret = -ENOSPC;
5152
		goto out_uncharge_cgroup;
5153
	}
5154 5155

	/*
5156
	 * Check enough hugepages are available for the reservation.
5157
	 * Hand the pages back to the subpool if there are not
5158
	 */
5159
	ret = hugetlb_acct_memory(h, gbl_reserve);
K
Ken Chen 已提交
5160
	if (ret < 0) {
5161
		goto out_put_pages;
K
Ken Chen 已提交
5162
	}
5163 5164 5165 5166 5167 5168 5169 5170 5171 5172 5173 5174

	/*
	 * 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
	 */
5175
	if (!vma || vma->vm_flags & VM_MAYSHARE) {
5176
		add = region_add(resv_map, from, to, regions_needed, h, h_cg);
5177 5178 5179

		if (unlikely(add < 0)) {
			hugetlb_acct_memory(h, -gbl_reserve);
5180
			ret = add;
5181
			goto out_put_pages;
5182
		} else if (unlikely(chg > add)) {
5183 5184 5185 5186 5187 5188 5189 5190 5191
			/*
			 * 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;

5192 5193 5194 5195
			hugetlb_cgroup_uncharge_cgroup_rsvd(
				hstate_index(h),
				(chg - add) * pages_per_huge_page(h), h_cg);

5196 5197 5198 5199 5200
			rsv_adjust = hugepage_subpool_put_pages(spool,
								chg - add);
			hugetlb_acct_memory(h, -rsv_adjust);
		}
	}
5201
	return 0;
5202 5203 5204 5205 5206 5207
out_put_pages:
	/* put back original number of pages, chg */
	(void)hugepage_subpool_put_pages(spool, chg);
out_uncharge_cgroup:
	hugetlb_cgroup_uncharge_cgroup_rsvd(hstate_index(h),
					    chg * pages_per_huge_page(h), h_cg);
5208
out_err:
5209
	if (!vma || vma->vm_flags & VM_MAYSHARE)
5210 5211 5212 5213 5214
		/* Only call region_abort if the region_chg succeeded but the
		 * region_add failed or didn't run.
		 */
		if (chg >= 0 && add < 0)
			region_abort(resv_map, from, to, regions_needed);
J
Joonsoo Kim 已提交
5215 5216
	if (vma && is_vma_resv_set(vma, HPAGE_RESV_OWNER))
		kref_put(&resv_map->refs, resv_map_release);
5217
	return ret;
5218 5219
}

5220 5221
long hugetlb_unreserve_pages(struct inode *inode, long start, long end,
								long freed)
5222
{
5223
	struct hstate *h = hstate_inode(inode);
5224
	struct resv_map *resv_map = inode_resv_map(inode);
5225
	long chg = 0;
5226
	struct hugepage_subpool *spool = subpool_inode(inode);
5227
	long gbl_reserve;
K
Ken Chen 已提交
5228

5229 5230 5231 5232
	/*
	 * Since this routine can be called in the evict inode path for all
	 * hugetlbfs inodes, resv_map could be NULL.
	 */
5233 5234 5235 5236 5237 5238 5239 5240 5241 5242 5243
	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
5244
	spin_lock(&inode->i_lock);
5245
	inode->i_blocks -= (blocks_per_huge_page(h) * freed);
K
Ken Chen 已提交
5246 5247
	spin_unlock(&inode->i_lock);

5248 5249 5250 5251 5252 5253
	/*
	 * 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);
5254 5255

	return 0;
5256
}
5257

5258 5259 5260 5261 5262 5263 5264 5265 5266 5267 5268
#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
5269 5270
	unsigned long vm_flags = vma->vm_flags & VM_LOCKED_CLEAR_MASK;
	unsigned long svm_flags = svma->vm_flags & VM_LOCKED_CLEAR_MASK;
5271 5272 5273 5274 5275 5276 5277 5278 5279 5280 5281 5282 5283

	/*
	 * 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;
}

5284
static bool vma_shareable(struct vm_area_struct *vma, unsigned long addr)
5285 5286 5287 5288 5289 5290 5291
{
	unsigned long base = addr & PUD_MASK;
	unsigned long end = base + PUD_SIZE;

	/*
	 * check on proper vm_flags and page table alignment
	 */
5292
	if (vma->vm_flags & VM_MAYSHARE && range_in_vma(vma, base, end))
5293 5294
		return true;
	return false;
5295 5296
}

5297 5298 5299 5300 5301 5302 5303 5304
/*
 * 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)
{
5305 5306
	unsigned long v_start = ALIGN(vma->vm_start, PUD_SIZE),
		v_end = ALIGN_DOWN(vma->vm_end, PUD_SIZE);
5307

5308 5309 5310 5311 5312 5313
	/*
	 * vma need span at least one aligned PUD size and the start,end range
	 * must at least partialy within it.
	 */
	if (!(vma->vm_flags & VM_MAYSHARE) || !(v_end > v_start) ||
		(*end <= v_start) || (*start >= v_end))
5314 5315
		return;

5316
	/* Extend the range to be PUD aligned for a worst case scenario */
5317 5318
	if (*start > v_start)
		*start = ALIGN_DOWN(*start, PUD_SIZE);
5319

5320 5321
	if (*end < v_end)
		*end = ALIGN(*end, PUD_SIZE);
5322 5323
}

5324 5325 5326 5327
/*
 * 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
5328 5329
 * code much cleaner.
 *
5330 5331 5332 5333 5334 5335 5336 5337 5338 5339
 * This routine must be called with i_mmap_rwsem held in at least read mode if
 * sharing is possible.  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).
 *
 * NOTE: This routine is only called from huge_pte_alloc.  Some callers of
 * huge_pte_alloc know that sharing is not possible and do not take
 * i_mmap_rwsem as a performance optimization.  This is handled by the
 * if !vma_shareable check at the beginning of the routine. i_mmap_rwsem is
 * only required for subsequent processing.
5340 5341 5342 5343 5344 5345 5346 5347 5348 5349 5350
 */
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;
5351
	spinlock_t *ptl;
5352 5353 5354 5355

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

5356
	i_mmap_assert_locked(mapping);
5357 5358 5359 5360 5361 5362
	vma_interval_tree_foreach(svma, &mapping->i_mmap, idx, idx) {
		if (svma == vma)
			continue;

		saddr = page_table_shareable(svma, vma, addr, idx);
		if (saddr) {
5363 5364
			spte = huge_pte_offset(svma->vm_mm, saddr,
					       vma_mmu_pagesize(svma));
5365 5366 5367 5368 5369 5370 5371 5372 5373 5374
			if (spte) {
				get_page(virt_to_page(spte));
				break;
			}
		}
	}

	if (!spte)
		goto out;

5375
	ptl = huge_pte_lock(hstate_vma(vma), mm, spte);
5376
	if (pud_none(*pud)) {
5377 5378
		pud_populate(mm, pud,
				(pmd_t *)((unsigned long)spte & PAGE_MASK));
5379
		mm_inc_nr_pmds(mm);
5380
	} else {
5381
		put_page(virt_to_page(spte));
5382
	}
5383
	spin_unlock(ptl);
5384 5385 5386 5387 5388 5389 5390 5391 5392 5393 5394 5395
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.
 *
5396
 * Called with page table lock held and i_mmap_rwsem held in write mode.
5397 5398 5399 5400
 *
 * returns: 1 successfully unmapped a shared pte page
 *	    0 the underlying pte page is not shared, or it is the last user
 */
5401 5402
int huge_pmd_unshare(struct mm_struct *mm, struct vm_area_struct *vma,
					unsigned long *addr, pte_t *ptep)
5403 5404
{
	pgd_t *pgd = pgd_offset(mm, *addr);
5405 5406
	p4d_t *p4d = p4d_offset(pgd, *addr);
	pud_t *pud = pud_offset(p4d, *addr);
5407

5408
	i_mmap_assert_write_locked(vma->vm_file->f_mapping);
5409 5410 5411 5412 5413 5414
	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));
5415
	mm_dec_nr_pmds(mm);
5416 5417 5418
	*addr = ALIGN(*addr, HPAGE_SIZE * PTRS_PER_PTE) - HPAGE_SIZE;
	return 1;
}
5419 5420 5421 5422 5423 5424
#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;
}
5425

5426 5427
int huge_pmd_unshare(struct mm_struct *mm, struct vm_area_struct *vma,
				unsigned long *addr, pte_t *ptep)
5428 5429 5430
{
	return 0;
}
5431 5432 5433 5434 5435

void adjust_range_if_pmd_sharing_possible(struct vm_area_struct *vma,
				unsigned long *start, unsigned long *end)
{
}
5436
#define want_pmd_share()	(0)
5437 5438
#endif /* CONFIG_ARCH_WANT_HUGE_PMD_SHARE */

5439 5440 5441 5442 5443
#ifdef CONFIG_ARCH_WANT_GENERAL_HUGETLB
pte_t *huge_pte_alloc(struct mm_struct *mm,
			unsigned long addr, unsigned long sz)
{
	pgd_t *pgd;
5444
	p4d_t *p4d;
5445 5446 5447 5448
	pud_t *pud;
	pte_t *pte = NULL;

	pgd = pgd_offset(mm, addr);
5449 5450 5451
	p4d = p4d_alloc(mm, pgd, addr);
	if (!p4d)
		return NULL;
5452
	pud = pud_alloc(mm, p4d, addr);
5453 5454 5455 5456 5457 5458 5459 5460 5461 5462 5463
	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);
		}
	}
5464
	BUG_ON(pte && pte_present(*pte) && !pte_huge(*pte));
5465 5466 5467 5468

	return pte;
}

5469 5470 5471 5472
/*
 * huge_pte_offset() - Walk the page table to resolve the hugepage
 * entry at address @addr
 *
5473 5474
 * Return: Pointer to page table entry (PUD or PMD) for
 * address @addr, or NULL if a !p*d_present() entry is encountered and the
5475 5476 5477
 * size @sz doesn't match the hugepage size at this level of the page
 * table.
 */
5478 5479
pte_t *huge_pte_offset(struct mm_struct *mm,
		       unsigned long addr, unsigned long sz)
5480 5481
{
	pgd_t *pgd;
5482
	p4d_t *p4d;
5483 5484
	pud_t *pud;
	pmd_t *pmd;
5485 5486

	pgd = pgd_offset(mm, addr);
5487 5488 5489 5490 5491
	if (!pgd_present(*pgd))
		return NULL;
	p4d = p4d_offset(pgd, addr);
	if (!p4d_present(*p4d))
		return NULL;
5492

5493
	pud = pud_offset(p4d, addr);
5494 5495
	if (sz == PUD_SIZE)
		/* must be pud huge, non-present or none */
5496
		return (pte_t *)pud;
5497
	if (!pud_present(*pud))
5498
		return NULL;
5499
	/* must have a valid entry and size to go further */
5500

5501 5502 5503
	pmd = pmd_offset(pud, addr);
	/* must be pmd huge, non-present or none */
	return (pte_t *)pmd;
5504 5505
}

5506 5507 5508 5509 5510 5511 5512 5513 5514 5515 5516 5517 5518
#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);
}

5519 5520 5521 5522 5523 5524 5525 5526
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;
}

5527
struct page * __weak
5528
follow_huge_pmd(struct mm_struct *mm, unsigned long address,
5529
		pmd_t *pmd, int flags)
5530
{
5531 5532
	struct page *page = NULL;
	spinlock_t *ptl;
5533
	pte_t pte;
J
John Hubbard 已提交
5534 5535 5536 5537 5538 5539

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

5540 5541 5542 5543 5544 5545 5546 5547 5548
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;
5549 5550
	pte = huge_ptep_get((pte_t *)pmd);
	if (pte_present(pte)) {
5551
		page = pmd_page(*pmd) + ((address & ~PMD_MASK) >> PAGE_SHIFT);
J
John Hubbard 已提交
5552 5553 5554 5555 5556 5557 5558 5559 5560 5561 5562 5563
		/*
		 * 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;
		}
5564
	} else {
5565
		if (is_hugetlb_entry_migration(pte)) {
5566 5567 5568 5569 5570 5571 5572 5573 5574 5575 5576
			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);
5577 5578 5579
	return page;
}

5580
struct page * __weak
5581
follow_huge_pud(struct mm_struct *mm, unsigned long address,
5582
		pud_t *pud, int flags)
5583
{
J
John Hubbard 已提交
5584
	if (flags & (FOLL_GET | FOLL_PIN))
5585
		return NULL;
5586

5587
	return pte_page(*(pte_t *)pud) + ((address & ~PUD_MASK) >> PAGE_SHIFT);
5588 5589
}

5590 5591 5592
struct page * __weak
follow_huge_pgd(struct mm_struct *mm, unsigned long address, pgd_t *pgd, int flags)
{
J
John Hubbard 已提交
5593
	if (flags & (FOLL_GET | FOLL_PIN))
5594 5595 5596 5597 5598
		return NULL;

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

5599 5600
bool isolate_huge_page(struct page *page, struct list_head *list)
{
5601 5602
	bool ret = true;

5603
	spin_lock(&hugetlb_lock);
5604 5605
	if (!PageHeadHuge(page) || !page_huge_active(page) ||
	    !get_page_unless_zero(page)) {
5606 5607 5608 5609
		ret = false;
		goto unlock;
	}
	clear_page_huge_active(page);
5610
	list_move_tail(&page->lru, list);
5611
unlock:
5612
	spin_unlock(&hugetlb_lock);
5613
	return ret;
5614 5615 5616 5617
}

void putback_active_hugepage(struct page *page)
{
5618
	VM_BUG_ON_PAGE(!PageHead(page), page);
5619
	spin_lock(&hugetlb_lock);
5620
	set_page_huge_active(page);
5621 5622 5623 5624
	list_move_tail(&page->lru, &(page_hstate(page))->hugepage_activelist);
	spin_unlock(&hugetlb_lock);
	put_page(page);
}
5625 5626 5627 5628 5629 5630 5631 5632 5633 5634 5635 5636 5637 5638 5639 5640 5641 5642 5643 5644 5645 5646 5647 5648 5649 5650 5651 5652 5653 5654 5655 5656 5657

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);
	}
}
5658 5659 5660 5661 5662 5663 5664 5665 5666 5667 5668 5669 5670 5671 5672 5673 5674 5675 5676 5677 5678 5679 5680 5681 5682 5683 5684 5685 5686 5687 5688 5689 5690 5691 5692 5693 5694 5695 5696

#ifdef CONFIG_CMA
static bool cma_reserve_called __initdata;

static int __init cmdline_parse_hugetlb_cma(char *p)
{
	hugetlb_cma_size = memparse(p, &p);
	return 0;
}

early_param("hugetlb_cma", cmdline_parse_hugetlb_cma);

void __init hugetlb_cma_reserve(int order)
{
	unsigned long size, reserved, per_node;
	int nid;

	cma_reserve_called = true;

	if (!hugetlb_cma_size)
		return;

	if (hugetlb_cma_size < (PAGE_SIZE << order)) {
		pr_warn("hugetlb_cma: cma area should be at least %lu MiB\n",
			(PAGE_SIZE << order) / SZ_1M);
		return;
	}

	/*
	 * If 3 GB area is requested on a machine with 4 numa nodes,
	 * let's allocate 1 GB on first three nodes and ignore the last one.
	 */
	per_node = DIV_ROUND_UP(hugetlb_cma_size, nr_online_nodes);
	pr_info("hugetlb_cma: reserve %lu MiB, up to %lu MiB per node\n",
		hugetlb_cma_size / SZ_1M, per_node / SZ_1M);

	reserved = 0;
	for_each_node_state(nid, N_ONLINE) {
		int res;
5697
		char name[CMA_MAX_NAME];
5698 5699 5700 5701

		size = min(per_node, hugetlb_cma_size - reserved);
		size = round_up(size, PAGE_SIZE << order);

5702
		snprintf(name, sizeof(name), "hugetlb%d", nid);
5703
		res = cma_declare_contiguous_nid(0, size, 0, PAGE_SIZE << order,
5704
						 0, false, name,
5705 5706 5707 5708 5709 5710 5711 5712 5713 5714 5715 5716 5717 5718 5719 5720 5721 5722 5723 5724 5725 5726 5727 5728 5729
						 &hugetlb_cma[nid], nid);
		if (res) {
			pr_warn("hugetlb_cma: reservation failed: err %d, node %d",
				res, nid);
			continue;
		}

		reserved += size;
		pr_info("hugetlb_cma: reserved %lu MiB on node %d\n",
			size / SZ_1M, nid);

		if (reserved >= hugetlb_cma_size)
			break;
	}
}

void __init hugetlb_cma_check(void)
{
	if (!hugetlb_cma_size || cma_reserve_called)
		return;

	pr_warn("hugetlb_cma: the option isn't supported by current arch\n");
}

#endif /* CONFIG_CMA */