hugetlb.c 155.0 KB
Newer Older
1
// SPDX-License-Identifier: GPL-2.0-only
L
Linus Torvalds 已提交
2 3
/*
 * Generic hugetlb support.
4
 * (C) Nadia Yvette Chambers, April 2004
L
Linus Torvalds 已提交
5 6 7 8
 */
#include <linux/list.h>
#include <linux/init.h>
#include <linux/mm.h>
9
#include <linux/seq_file.h>
L
Linus Torvalds 已提交
10 11
#include <linux/sysctl.h>
#include <linux/highmem.h>
A
Andrea Arcangeli 已提交
12
#include <linux/mmu_notifier.h>
L
Linus Torvalds 已提交
13
#include <linux/nodemask.h>
D
David Gibson 已提交
14
#include <linux/pagemap.h>
15
#include <linux/mempolicy.h>
16
#include <linux/compiler.h>
17
#include <linux/cpuset.h>
18
#include <linux/mutex.h>
19
#include <linux/memblock.h>
20
#include <linux/sysfs.h>
21
#include <linux/slab.h>
22
#include <linux/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
/* Forward declaration */
static int hugetlb_acct_memory(struct hstate *h, long delta);

85
static inline bool subpool_is_free(struct hugepage_subpool *spool)
86
{
87 88 89 90 91 92 93 94 95
	if (spool->count)
		return false;
	if (spool->max_hpages != -1)
		return spool->used_hpages == 0;
	if (spool->min_hpages != -1)
		return spool->rsv_hpages == spool->min_hpages;

	return true;
}
96

97 98
static inline void unlock_or_release_subpool(struct hugepage_subpool *spool)
{
99 100 101
	spin_unlock(&spool->lock);

	/* If no pages are used, and no other handles to the subpool
E
Ethon Paul 已提交
102
	 * remain, give up any reservations based on minimum size and
103
	 * free the subpool */
104
	if (subpool_is_free(spool)) {
105 106 107
		if (spool->min_hpages != -1)
			hugetlb_acct_memory(spool->hstate,
						-spool->min_hpages);
108
		kfree(spool);
109
	}
110 111
}

112 113
struct hugepage_subpool *hugepage_new_subpool(struct hstate *h, long max_hpages,
						long min_hpages)
114 115 116
{
	struct hugepage_subpool *spool;

117
	spool = kzalloc(sizeof(*spool), GFP_KERNEL);
118 119 120 121 122
	if (!spool)
		return NULL;

	spin_lock_init(&spool->lock);
	spool->count = 1;
123 124 125 126 127 128 129 130 131
	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;
132 133 134 135 136 137 138 139 140 141 142 143

	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);
}

144 145 146
/*
 * Subpool accounting for allocating and reserving pages.
 * Return -ENOMEM if there are not enough resources to satisfy the
147
 * request.  Otherwise, return the number of pages by which the
148 149
 * 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
150
 * a subpool minimum size must be maintained.
151 152
 */
static long hugepage_subpool_get_pages(struct hugepage_subpool *spool,
153 154
				      long delta)
{
155
	long ret = delta;
156 157

	if (!spool)
158
		return ret;
159 160

	spin_lock(&spool->lock);
161 162 163 164 165 166 167 168

	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;
		}
169 170
	}

171 172
	/* minimum size accounting */
	if (spool->min_hpages != -1 && spool->rsv_hpages) {
173 174 175 176 177 178 179 180 181 182 183 184 185 186 187
		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);
188 189 190
	return ret;
}

191 192 193 194 195 196 197
/*
 * 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,
198 199
				       long delta)
{
200 201
	long ret = delta;

202
	if (!spool)
203
		return delta;
204 205

	spin_lock(&spool->lock);
206 207 208 209

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

210 211
	 /* minimum size accounting */
	if (spool->min_hpages != -1 && spool->used_hpages < spool->min_hpages) {
212 213 214 215 216 217 218 219 220 221 222 223 224 225
		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.
	 */
226
	unlock_or_release_subpool(spool);
227 228

	return ret;
229 230 231 232 233 234 235 236 237
}

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
238
	return subpool_inode(file_inode(vma->vm_file));
239 240
}

241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260
/* 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;
}

261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295
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
}

296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320
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);

321
		rg = prg;
322 323 324 325 326 327 328 329 330 331 332 333
	}

	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);
	}
}

334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351
static inline long
hugetlb_resv_map_add(struct resv_map *map, struct file_region *rg, long from,
		     long to, struct hstate *h, struct hugetlb_cgroup *cg,
		     long *regions_needed)
{
	struct file_region *nrg;

	if (!regions_needed) {
		nrg = get_file_region_entry_from_cache(map, from, to);
		record_hugetlb_cgroup_uncharge_info(cg, h, map, nrg);
		list_add(&nrg->link, rg->link.prev);
		coalesce_file_region(map, nrg);
	} else
		*regions_needed += 1;

	return to - from;
}

352 353 354 355 356 357 358
/*
 * 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
359 360
 */
static long add_reservation_in_range(struct resv_map *resv, long f, long t,
361
				     struct hugetlb_cgroup *h_cg,
362
				     struct hstate *h, long *regions_needed)
M
Mina Almasry 已提交
363
{
364
	long add = 0;
M
Mina Almasry 已提交
365
	struct list_head *head = &resv->regions;
366
	long last_accounted_offset = f;
367
	struct file_region *rg = NULL, *trg = NULL;
M
Mina Almasry 已提交
368

369 370
	if (regions_needed)
		*regions_needed = 0;
M
Mina Almasry 已提交
371

372 373 374 375 376 377 378 379 380 381 382 383 384 385
	/* 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
386

387 388 389
		/* When we find a region that starts beyond our range, we've
		 * finished.
		 */
390
		if (rg->from >= t)
M
Mina Almasry 已提交
391 392
			break;

393 394 395
		/* Add an entry for last_accounted_offset -> rg->from, and
		 * update last_accounted_offset.
		 */
396 397 398 399 400
		if (rg->from > last_accounted_offset)
			add += hugetlb_resv_map_add(resv, rg,
						    last_accounted_offset,
						    rg->from, h, h_cg,
						    regions_needed);
401 402 403 404 405 406 407

		last_accounted_offset = rg->to;
	}

	/* Handle the case where our range extends beyond
	 * last_accounted_offset.
	 */
408 409 410
	if (last_accounted_offset < t)
		add += hugetlb_resv_map_add(resv, rg, last_accounted_offset,
					    t, h, h_cg, regions_needed);
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

	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
447
		 */
448 449 450 451 452 453 454 455
		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
456 457
		}

458 459
		spin_lock(&resv->lock);

460 461
		list_splice(&allocated_regions, &resv->region_cache);
		resv->region_cache_count += to_allocate;
M
Mina Almasry 已提交
462 463
	}

464
	return 0;
M
Mina Almasry 已提交
465

466 467 468 469 470 471
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
472 473
}

474 475
/*
 * Add the huge page range represented by [f, t) to the reserve
476 477 478 479 480
 * 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.
481
 *
482 483 484 485
 * 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
486
 * this operation and we were not able to allocate, it returns -ENOMEM.
487 488 489
 * 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.
490
 */
491
static long region_add(struct resv_map *resv, long f, long t,
492 493
		       long in_regions_needed, struct hstate *h,
		       struct hugetlb_cgroup *h_cg)
494
{
495
	long add = 0, actual_regions_needed = 0;
496

497
	spin_lock(&resv->lock);
498 499 500
retry:

	/* Count how many regions are actually needed to execute this add. */
501 502
	add_reservation_in_range(resv, f, t, NULL, NULL,
				 &actual_regions_needed);
503

504
	/*
505 506 507 508 509 510 511
	 * 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.
512
	 */
513 514 515 516 517 518 519 520
	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);
521

522 523 524 525
		if (allocate_file_region_entries(
			    resv, actual_regions_needed - in_regions_needed)) {
			return -ENOMEM;
		}
526

527
		goto retry;
528 529
	}

530
	add = add_reservation_in_range(resv, f, t, h_cg, h, NULL);
531 532

	resv->adds_in_progress -= in_regions_needed;
533

534
	spin_unlock(&resv->lock);
535 536
	VM_BUG_ON(add < 0);
	return add;
537 538
}

539 540 541 542 543 544 545
/*
 * 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
546 547 548 549 550 551 552
 * 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.
553 554 555 556 557
 *
 * 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.
558
 */
559 560
static long region_chg(struct resv_map *resv, long f, long t,
		       long *out_regions_needed)
561 562 563
{
	long chg = 0;

564
	spin_lock(&resv->lock);
565

566
	/* Count how many hugepages in this range are NOT represented. */
567
	chg = add_reservation_in_range(resv, f, t, NULL, NULL,
568
				       out_regions_needed);
569

570 571
	if (*out_regions_needed == 0)
		*out_regions_needed = 1;
572

573 574
	if (allocate_file_region_entries(resv, *out_regions_needed))
		return -ENOMEM;
575

576
	resv->adds_in_progress += *out_regions_needed;
577 578

	spin_unlock(&resv->lock);
579 580 581
	return chg;
}

582 583 584 585 586
/*
 * 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
587 588 589
 * 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.
590 591 592 593 594
 *
 * 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.
 */
595 596
static void region_abort(struct resv_map *resv, long f, long t,
			 long regions_needed)
597 598 599
{
	spin_lock(&resv->lock);
	VM_BUG_ON(!resv->region_cache_count);
600
	resv->adds_in_progress -= regions_needed;
601 602 603
	spin_unlock(&resv->lock);
}

604
/*
605 606 607 608 609 610 611 612 613 614 615 616
 * 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.
617
 */
618
static long region_del(struct resv_map *resv, long f, long t)
619
{
620
	struct list_head *head = &resv->regions;
621
	struct file_region *rg, *trg;
622 623
	struct file_region *nrg = NULL;
	long del = 0;
624

625
retry:
626
	spin_lock(&resv->lock);
627
	list_for_each_entry_safe(rg, trg, head, link) {
628 629 630 631 632 633 634 635
		/*
		 * 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))
636
			continue;
637

638
		if (rg->from >= t)
639 640
			break;

641 642 643 644 645 646 647 648 649 650 651 652 653
		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--;
			}
654

655 656 657 658 659 660 661 662 663
			if (!nrg) {
				spin_unlock(&resv->lock);
				nrg = kmalloc(sizeof(*nrg), GFP_KERNEL);
				if (!nrg)
					return -ENOMEM;
				goto retry;
			}

			del += t - f;
664 665
			hugetlb_cgroup_uncharge_file_region(
				resv, rg, t - f);
666 667 668 669

			/* New entry for end of split region */
			nrg->from = t;
			nrg->to = rg->to;
670 671 672

			copy_hugetlb_cgroup_uncharge_info(nrg, rg);

673 674 675 676 677 678 679
			INIT_LIST_HEAD(&nrg->link);

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

			list_add(&nrg->link, &rg->link);
			nrg = NULL;
680
			break;
681 682 683 684
		}

		if (f <= rg->from && t >= rg->to) { /* Remove entire region */
			del += rg->to - rg->from;
685 686
			hugetlb_cgroup_uncharge_file_region(resv, rg,
							    rg->to - rg->from);
687 688 689 690 691 692
			list_del(&rg->link);
			kfree(rg);
			continue;
		}

		if (f <= rg->from) {	/* Trim beginning of region */
693 694 695
			hugetlb_cgroup_uncharge_file_region(resv, rg,
							    t - rg->from);

696 697 698
			del += t - rg->from;
			rg->from = t;
		} else {		/* Trim end of region */
699 700
			hugetlb_cgroup_uncharge_file_region(resv, rg,
							    rg->to - f);
701 702 703

			del += rg->to - f;
			rg->to = f;
704
		}
705
	}
706 707

	spin_unlock(&resv->lock);
708 709
	kfree(nrg);
	return del;
710 711
}

712 713 714 715 716 717 718 719 720
/*
 * 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.
 */
721
void hugetlb_fix_reserve_counts(struct inode *inode)
722 723 724 725 726
{
	struct hugepage_subpool *spool = subpool_inode(inode);
	long rsv_adjust;

	rsv_adjust = hugepage_subpool_get_pages(spool, 1);
727
	if (rsv_adjust) {
728 729 730 731 732 733
		struct hstate *h = hstate_inode(inode);

		hugetlb_acct_memory(h, 1);
	}
}

734 735 736 737
/*
 * Count and return the number of huge pages in the reserve map
 * that intersect with the range [f, t).
 */
738
static long region_count(struct resv_map *resv, long f, long t)
739
{
740
	struct list_head *head = &resv->regions;
741 742 743
	struct file_region *rg;
	long chg = 0;

744
	spin_lock(&resv->lock);
745 746
	/* Locate each segment we overlap with, and count that overlap. */
	list_for_each_entry(rg, head, link) {
747 748
		long seg_from;
		long seg_to;
749 750 751 752 753 754 755 756 757 758 759

		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;
	}
760
	spin_unlock(&resv->lock);
761 762 763 764

	return chg;
}

765 766 767 768
/*
 * Convert the address within this vma to the page offset within
 * the mapping, in pagecache page units; huge pages here.
 */
769 770
static pgoff_t vma_hugecache_offset(struct hstate *h,
			struct vm_area_struct *vma, unsigned long address)
771
{
772 773
	return ((address - vma->vm_start) >> huge_page_shift(h)) +
			(vma->vm_pgoff >> huge_page_order(h));
774 775
}

776 777 778 779 780
pgoff_t linear_hugepage_index(struct vm_area_struct *vma,
				     unsigned long address)
{
	return vma_hugecache_offset(hstate_vma(vma), vma, address);
}
781
EXPORT_SYMBOL_GPL(linear_hugepage_index);
782

783 784 785 786 787 788
/*
 * 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)
{
789 790 791
	if (vma->vm_ops && vma->vm_ops->pagesize)
		return vma->vm_ops->pagesize(vma);
	return PAGE_SIZE;
792
}
793
EXPORT_SYMBOL_GPL(vma_kernel_pagesize);
794

795 796 797
/*
 * 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
798 799
 * architectures where it differs, an architecture-specific 'strong'
 * version of this symbol is required.
800
 */
801
__weak unsigned long vma_mmu_pagesize(struct vm_area_struct *vma)
802 803 804 805
{
	return vma_kernel_pagesize(vma);
}

806 807 808 809 810 811 812
/*
 * 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)
813
#define HPAGE_RESV_MASK (HPAGE_RESV_OWNER | HPAGE_RESV_UNMAPPED)
814

815 816 817 818 819 820 821 822 823
/*
 * 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.
824 825 826 827 828 829 830 831 832
 *
 * 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.
833
 */
834 835 836 837 838 839 840 841 842 843 844
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;
}

845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863
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
}

864
struct resv_map *resv_map_alloc(void)
865 866
{
	struct resv_map *resv_map = kmalloc(sizeof(*resv_map), GFP_KERNEL);
867 868 869 870 871
	struct file_region *rg = kmalloc(sizeof(*rg), GFP_KERNEL);

	if (!resv_map || !rg) {
		kfree(resv_map);
		kfree(rg);
872
		return NULL;
873
	}
874 875

	kref_init(&resv_map->refs);
876
	spin_lock_init(&resv_map->lock);
877 878
	INIT_LIST_HEAD(&resv_map->regions);

879
	resv_map->adds_in_progress = 0;
880 881 882 883 884 885 886
	/*
	 * 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);
887 888 889 890 891

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

892 893 894
	return resv_map;
}

895
void resv_map_release(struct kref *ref)
896 897
{
	struct resv_map *resv_map = container_of(ref, struct resv_map, refs);
898 899
	struct list_head *head = &resv_map->region_cache;
	struct file_region *rg, *trg;
900 901

	/* Clear out any active regions before we release the map. */
902
	region_del(resv_map, 0, LONG_MAX);
903 904 905 906 907 908 909 910 911

	/* ... 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);

912 913 914
	kfree(resv_map);
}

915 916
static inline struct resv_map *inode_resv_map(struct inode *inode)
{
917 918 919 920 921 922 923 924 925
	/*
	 * 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;
926 927
}

928
static struct resv_map *vma_resv_map(struct vm_area_struct *vma)
929
{
930
	VM_BUG_ON_VMA(!is_vm_hugetlb_page(vma), vma);
931 932 933 934 935 936 937
	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 {
938 939
		return (struct resv_map *)(get_vma_private_data(vma) &
							~HPAGE_RESV_MASK);
940
	}
941 942
}

943
static void set_vma_resv_map(struct vm_area_struct *vma, struct resv_map *map)
944
{
945 946
	VM_BUG_ON_VMA(!is_vm_hugetlb_page(vma), vma);
	VM_BUG_ON_VMA(vma->vm_flags & VM_MAYSHARE, vma);
947

948 949
	set_vma_private_data(vma, (get_vma_private_data(vma) &
				HPAGE_RESV_MASK) | (unsigned long)map);
950 951 952 953
}

static void set_vma_resv_flags(struct vm_area_struct *vma, unsigned long flags)
{
954 955
	VM_BUG_ON_VMA(!is_vm_hugetlb_page(vma), vma);
	VM_BUG_ON_VMA(vma->vm_flags & VM_MAYSHARE, vma);
956 957

	set_vma_private_data(vma, get_vma_private_data(vma) | flags);
958 959 960 961
}

static int is_vma_resv_set(struct vm_area_struct *vma, unsigned long flag)
{
962
	VM_BUG_ON_VMA(!is_vm_hugetlb_page(vma), vma);
963 964

	return (get_vma_private_data(vma) & flag) != 0;
965 966
}

967
/* Reset counters to 0 and clear all HPAGE_RESV_* flags */
968 969
void reset_vma_resv_huge_pages(struct vm_area_struct *vma)
{
970
	VM_BUG_ON_VMA(!is_vm_hugetlb_page(vma), vma);
971
	if (!(vma->vm_flags & VM_MAYSHARE))
972 973 974 975
		vma->vm_private_data = (void *)0;
}

/* Returns true if the VMA has associated reserve pages */
976
static bool vma_has_reserves(struct vm_area_struct *vma, long chg)
977
{
978 979 980 981 982 983 984 985 986 987 988
	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)
989
			return true;
990
		else
991
			return false;
992
	}
993 994

	/* Shared mappings always use reserves */
995 996 997 998 999
	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
1000
		 * fallocate.  In this case, there really are no reserves to
1001 1002 1003 1004 1005 1006 1007
		 * use.  This situation is indicated if chg != 0.
		 */
		if (chg)
			return false;
		else
			return true;
	}
1008 1009 1010 1011 1012

	/*
	 * Only the process that called mmap() has reserves for
	 * private mappings.
	 */
1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033
	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;
	}
1034

1035
	return false;
1036 1037
}

1038
static void enqueue_huge_page(struct hstate *h, struct page *page)
L
Linus Torvalds 已提交
1039 1040
{
	int nid = page_to_nid(page);
1041
	list_move(&page->lru, &h->hugepage_freelists[nid]);
1042 1043
	h->free_huge_pages++;
	h->free_huge_pages_node[nid]++;
1044
	SetHPageFreed(page);
L
Linus Torvalds 已提交
1045 1046
}

1047
static struct page *dequeue_huge_page_node_exact(struct hstate *h, int nid)
1048 1049
{
	struct page *page;
1050 1051 1052 1053 1054
	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;
1055

1056 1057 1058 1059 1060
		if (PageHWPoison(page))
			continue;

		list_move(&page->lru, &h->hugepage_activelist);
		set_page_refcounted(page);
1061
		ClearHPageFreed(page);
1062 1063 1064
		h->free_huge_pages--;
		h->free_huge_pages_node[nid]--;
		return page;
1065 1066
	}

1067
	return NULL;
1068 1069
}

1070 1071
static struct page *dequeue_huge_page_nodemask(struct hstate *h, gfp_t gfp_mask, int nid,
		nodemask_t *nmask)
1072
{
1073 1074 1075 1076
	unsigned int cpuset_mems_cookie;
	struct zonelist *zonelist;
	struct zone *zone;
	struct zoneref *z;
1077
	int node = NUMA_NO_NODE;
1078

1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094
	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);
1095 1096 1097 1098 1099

		page = dequeue_huge_page_node_exact(h, node);
		if (page)
			return page;
	}
1100 1101 1102
	if (unlikely(read_mems_allowed_retry(cpuset_mems_cookie)))
		goto retry_cpuset;

1103 1104 1105
	return NULL;
}

1106 1107
static struct page *dequeue_huge_page_vma(struct hstate *h,
				struct vm_area_struct *vma,
1108 1109
				unsigned long address, int avoid_reserve,
				long chg)
L
Linus Torvalds 已提交
1110
{
1111
	struct page *page;
1112
	struct mempolicy *mpol;
1113
	gfp_t gfp_mask;
1114
	nodemask_t *nodemask;
1115
	int nid;
L
Linus Torvalds 已提交
1116

1117 1118 1119 1120 1121
	/*
	 * 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
	 */
1122
	if (!vma_has_reserves(vma, chg) &&
1123
			h->free_huge_pages - h->resv_huge_pages == 0)
1124
		goto err;
1125

1126
	/* If reserves cannot be used, ensure enough pages are in the pool */
1127
	if (avoid_reserve && h->free_huge_pages - h->resv_huge_pages == 0)
1128
		goto err;
1129

1130 1131
	gfp_mask = htlb_alloc_mask(h);
	nid = huge_node(vma, address, gfp_mask, &mpol, &nodemask);
1132 1133
	page = dequeue_huge_page_nodemask(h, gfp_mask, nid, nodemask);
	if (page && !avoid_reserve && vma_has_reserves(vma, chg)) {
1134
		SetHPageRestoreReserve(page);
1135
		h->resv_huge_pages--;
L
Linus Torvalds 已提交
1136
	}
1137

1138
	mpol_cond_put(mpol);
L
Linus Torvalds 已提交
1139
	return page;
1140 1141 1142

err:
	return NULL;
L
Linus Torvalds 已提交
1143 1144
}

1145 1146 1147 1148 1149 1150 1151 1152 1153
/*
 * 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)
{
1154
	nid = next_node_in(nid, *nodes_allowed);
1155 1156 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
	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--)

1216
#ifdef CONFIG_ARCH_HAS_GIGANTIC_PAGE
1217
static void destroy_compound_gigantic_page(struct page *page,
1218
					unsigned int order)
1219 1220 1221 1222 1223
{
	int i;
	int nr_pages = 1 << order;
	struct page *p = page + 1;

1224
	atomic_set(compound_mapcount_ptr(page), 0);
1225
	atomic_set(compound_pincount_ptr(page), 0);
1226

1227
	for (i = 1; i < nr_pages; i++, p = mem_map_next(p, page, i)) {
1228
		clear_compound_head(p);
1229 1230 1231 1232
		set_page_refcounted(p);
	}

	set_compound_order(page, 0);
1233
	page[1].compound_nr = 0;
1234 1235 1236
	__ClearPageHead(page);
}

1237
static void free_gigantic_page(struct page *page, unsigned int order)
1238
{
1239 1240 1241 1242
	/*
	 * If the page isn't allocated using the cma allocator,
	 * cma_release() returns false.
	 */
1243 1244
#ifdef CONFIG_CMA
	if (cma_release(hugetlb_cma[page_to_nid(page)], page, 1 << order))
1245
		return;
1246
#endif
1247

1248 1249 1250
	free_contig_range(page_to_pfn(page), 1 << order);
}

1251
#ifdef CONFIG_CONTIG_ALLOC
1252 1253
static struct page *alloc_gigantic_page(struct hstate *h, gfp_t gfp_mask,
		int nid, nodemask_t *nodemask)
1254
{
1255
	unsigned long nr_pages = 1UL << huge_page_order(h);
1256 1257
	if (nid == NUMA_NO_NODE)
		nid = numa_mem_id();
1258

1259 1260
#ifdef CONFIG_CMA
	{
1261 1262 1263
		struct page *page;
		int node;

1264 1265 1266
		if (hugetlb_cma[nid]) {
			page = cma_alloc(hugetlb_cma[nid], nr_pages,
					huge_page_order(h), true);
1267 1268 1269
			if (page)
				return page;
		}
1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281

		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;
			}
		}
1282
	}
1283
#endif
1284

1285
	return alloc_contig_pages(nr_pages, gfp_mask, nid, nodemask);
1286 1287 1288
}

static void prep_new_huge_page(struct hstate *h, struct page *page, int nid);
1289
static void prep_compound_gigantic_page(struct page *page, unsigned int order);
1290 1291 1292 1293 1294 1295 1296
#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 */
1297

1298
#else /* !CONFIG_ARCH_HAS_GIGANTIC_PAGE */
1299
static struct page *alloc_gigantic_page(struct hstate *h, gfp_t gfp_mask,
1300 1301 1302 1303
					int nid, nodemask_t *nodemask)
{
	return NULL;
}
1304
static inline void free_gigantic_page(struct page *page, unsigned int order) { }
1305
static inline void destroy_compound_gigantic_page(struct page *page,
1306
						unsigned int order) { }
1307 1308
#endif

1309
static void update_and_free_page(struct hstate *h, struct page *page)
A
Adam Litke 已提交
1310 1311
{
	int i;
1312
	struct page *subpage = page;
1313

1314
	if (hstate_is_gigantic(h) && !gigantic_page_runtime_supported())
1315
		return;
1316

1317 1318
	h->nr_huge_pages--;
	h->nr_huge_pages_node[page_to_nid(page)]--;
1319 1320 1321
	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 |
1322
				1 << PG_referenced | 1 << PG_dirty |
1323 1324
				1 << PG_active | 1 << PG_private |
				1 << PG_writeback);
A
Adam Litke 已提交
1325
	}
1326
	VM_BUG_ON_PAGE(hugetlb_cgroup_from_page(page), page);
1327
	VM_BUG_ON_PAGE(hugetlb_cgroup_from_page_rsvd(page), page);
1328
	set_compound_page_dtor(page, NULL_COMPOUND_DTOR);
A
Adam Litke 已提交
1329
	set_page_refcounted(page);
1330
	if (hstate_is_gigantic(h)) {
1331 1332 1333 1334 1335
		/*
		 * Temporarily drop the hugetlb_lock, because
		 * we might block in free_gigantic_page().
		 */
		spin_unlock(&hugetlb_lock);
1336 1337
		destroy_compound_gigantic_page(page, huge_page_order(h));
		free_gigantic_page(page, huge_page_order(h));
1338
		spin_lock(&hugetlb_lock);
1339 1340 1341
	} else {
		__free_pages(page, huge_page_order(h));
	}
A
Adam Litke 已提交
1342 1343
}

1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354
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;
}

1355
static void __free_huge_page(struct page *page)
1356
{
1357 1358 1359 1360
	/*
	 * Can't pass hstate in here because it is called from the
	 * compound page destructor.
	 */
1361
	struct hstate *h = page_hstate(page);
1362
	int nid = page_to_nid(page);
1363
	struct hugepage_subpool *spool = hugetlb_page_subpool(page);
1364
	bool restore_reserve;
1365

1366 1367
	VM_BUG_ON_PAGE(page_count(page), page);
	VM_BUG_ON_PAGE(page_mapcount(page), page);
1368

1369
	hugetlb_set_page_subpool(page, NULL);
1370
	page->mapping = NULL;
1371 1372
	restore_reserve = HPageRestoreReserve(page);
	ClearHPageRestoreReserve(page);
1373

1374
	/*
1375
	 * If HPageRestoreReserve was set on page, page allocation consumed a
1376 1377 1378
	 * 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
M
Miaohe Lin 已提交
1379
	 * reservation, do not call hugepage_subpool_put_pages() as this will
1380
	 * remove the reserved page from the subpool.
1381
	 */
1382 1383 1384 1385 1386 1387 1388 1389 1390 1391
	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;
	}
1392

1393
	spin_lock(&hugetlb_lock);
1394
	ClearHPageMigratable(page);
1395 1396
	hugetlb_cgroup_uncharge_page(hstate_index(h),
				     pages_per_huge_page(h), page);
1397 1398
	hugetlb_cgroup_uncharge_page_rsvd(hstate_index(h),
					  pages_per_huge_page(h), page);
1399 1400 1401
	if (restore_reserve)
		h->resv_huge_pages++;

1402
	if (HPageTemporary(page)) {
1403
		list_del(&page->lru);
1404
		ClearHPageTemporary(page);
1405 1406
		update_and_free_page(h, page);
	} else if (h->surplus_huge_pages_node[nid]) {
1407 1408
		/* remove the page from active list */
		list_del(&page->lru);
1409 1410 1411
		update_and_free_page(h, page);
		h->surplus_huge_pages--;
		h->surplus_huge_pages_node[nid]--;
1412
	} else {
1413
		arch_clear_hugepage_flags(page);
1414
		enqueue_huge_page(h, page);
1415
	}
1416 1417 1418
	spin_unlock(&hugetlb_lock);
}

1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466
/*
 * 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);
}

1467
static void prep_new_huge_page(struct hstate *h, struct page *page, int nid)
1468
{
1469
	INIT_LIST_HEAD(&page->lru);
1470
	set_compound_page_dtor(page, HUGETLB_PAGE_DTOR);
1471
	hugetlb_set_page_subpool(page, NULL);
1472
	set_hugetlb_cgroup(page, NULL);
1473
	set_hugetlb_cgroup_rsvd(page, NULL);
1474
	spin_lock(&hugetlb_lock);
1475 1476
	h->nr_huge_pages++;
	h->nr_huge_pages_node[nid]++;
1477
	ClearHPageFreed(page);
1478 1479 1480
	spin_unlock(&hugetlb_lock);
}

1481
static void prep_compound_gigantic_page(struct page *page, unsigned int order)
1482 1483 1484 1485 1486 1487 1488
{
	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);
1489
	__ClearPageReserved(page);
1490
	__SetPageHead(page);
1491
	for (i = 1; i < nr_pages; i++, p = mem_map_next(p, page, i)) {
1492 1493 1494 1495
		/*
		 * 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
1496
		 * too.  Otherwise drivers using get_user_pages() to access tail
1497 1498 1499 1500 1501 1502 1503 1504
		 * 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);
1505
		set_page_count(p, 0);
1506
		set_compound_head(p, page);
1507
	}
1508
	atomic_set(compound_mapcount_ptr(page), -1);
1509
	atomic_set(compound_pincount_ptr(page), 0);
1510 1511
}

A
Andrew Morton 已提交
1512 1513 1514 1515 1516
/*
 * PageHuge() only returns true for hugetlbfs pages, but not for normal or
 * transparent huge pages.  See the PageTransHuge() documentation for more
 * details.
 */
1517 1518 1519 1520 1521 1522
int PageHuge(struct page *page)
{
	if (!PageCompound(page))
		return 0;

	page = compound_head(page);
1523
	return page[1].compound_dtor == HUGETLB_PAGE_DTOR;
1524
}
1525 1526
EXPORT_SYMBOL_GPL(PageHuge);

1527 1528 1529 1530 1531 1532 1533 1534 1535
/*
 * 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;

1536
	return page_head[1].compound_dtor == HUGETLB_PAGE_DTOR;
1537 1538
}

1539 1540 1541
/*
 * Find and lock address space (mapping) in write mode.
 *
1542 1543 1544
 * 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.
1545 1546 1547
 */
struct address_space *hugetlb_page_mapping_lock_write(struct page *hpage)
{
1548
	struct address_space *mapping = page_mapping(hpage);
1549 1550 1551 1552 1553 1554 1555

	if (!mapping)
		return mapping;

	if (i_mmap_trylock_write(mapping))
		return mapping;

1556
	return NULL;
1557 1558
}

1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575
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;
}

1576
static struct page *alloc_buddy_huge_page(struct hstate *h,
1577 1578
		gfp_t gfp_mask, int nid, nodemask_t *nmask,
		nodemask_t *node_alloc_noretry)
L
Linus Torvalds 已提交
1579
{
1580
	int order = huge_page_order(h);
L
Linus Torvalds 已提交
1581
	struct page *page;
1582
	bool alloc_try_hard = true;
1583

1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595
	/*
	 * 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;
1596 1597 1598 1599 1600 1601 1602
	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);
1603

1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619
	/*
	 * 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);

1620 1621 1622
	return page;
}

1623 1624 1625 1626 1627
/*
 * 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,
1628 1629
		gfp_t gfp_mask, int nid, nodemask_t *nmask,
		nodemask_t *node_alloc_noretry)
1630 1631 1632 1633 1634 1635 1636
{
	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,
1637
				nid, nmask, node_alloc_noretry);
1638 1639 1640 1641 1642 1643 1644 1645 1646 1647
	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;
}

1648 1649 1650 1651
/*
 * Allocates a fresh page to the hugetlb allocator pool in the node interleaved
 * manner.
 */
1652 1653
static int alloc_pool_huge_page(struct hstate *h, nodemask_t *nodes_allowed,
				nodemask_t *node_alloc_noretry)
1654 1655 1656
{
	struct page *page;
	int nr_nodes, node;
1657
	gfp_t gfp_mask = htlb_alloc_mask(h) | __GFP_THISNODE;
1658 1659

	for_each_node_mask_to_alloc(h, nr_nodes, node, nodes_allowed) {
1660 1661
		page = alloc_fresh_huge_page(h, gfp_mask, node, nodes_allowed,
						node_alloc_noretry);
1662
		if (page)
1663 1664 1665
			break;
	}

1666 1667
	if (!page)
		return 0;
1668

1669 1670 1671
	put_page(page); /* free it into the hugepage allocator */

	return 1;
1672 1673
}

1674 1675 1676 1677 1678 1679
/*
 * 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.
 */
1680 1681
static int free_pool_huge_page(struct hstate *h, nodemask_t *nodes_allowed,
							 bool acct_surplus)
1682
{
1683
	int nr_nodes, node;
1684 1685
	int ret = 0;

1686
	for_each_node_mask_to_free(h, nr_nodes, node, nodes_allowed) {
1687 1688 1689 1690
		/*
		 * If we're returning unused surplus pages, only examine
		 * nodes with surplus pages.
		 */
1691 1692
		if ((!acct_surplus || h->surplus_huge_pages_node[node]) &&
		    !list_empty(&h->hugepage_freelists[node])) {
1693
			struct page *page =
1694
				list_entry(h->hugepage_freelists[node].next,
1695 1696 1697
					  struct page, lru);
			list_del(&page->lru);
			h->free_huge_pages--;
1698
			h->free_huge_pages_node[node]--;
1699 1700
			if (acct_surplus) {
				h->surplus_huge_pages--;
1701
				h->surplus_huge_pages_node[node]--;
1702
			}
1703 1704
			update_and_free_page(h, page);
			ret = 1;
1705
			break;
1706
		}
1707
	}
1708 1709 1710 1711

	return ret;
}

1712 1713
/*
 * Dissolve a given free hugepage into free buddy pages. This function does
1714 1715 1716 1717 1718 1719 1720
 * 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)
1721
 */
1722
int dissolve_free_huge_page(struct page *page)
1723
{
1724
	int rc = -EBUSY;
1725

1726
retry:
1727 1728 1729 1730
	/* Not to disrupt normal path by vainly holding hugetlb_lock */
	if (!PageHuge(page))
		return 0;

1731
	spin_lock(&hugetlb_lock);
1732 1733 1734 1735 1736 1737
	if (!PageHuge(page)) {
		rc = 0;
		goto out;
	}

	if (!page_count(page)) {
1738 1739 1740
		struct page *head = compound_head(page);
		struct hstate *h = page_hstate(head);
		int nid = page_to_nid(head);
1741
		if (h->free_huge_pages - h->resv_huge_pages == 0)
1742
			goto out;
1743 1744 1745 1746 1747

		/*
		 * We should make sure that the page is already on the free list
		 * when it is dissolved.
		 */
1748
		if (unlikely(!HPageFreed(head))) {
1749 1750 1751 1752 1753 1754 1755 1756 1757 1758 1759 1760 1761 1762
			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;
		}

1763 1764 1765 1766 1767 1768 1769 1770
		/*
		 * 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);
		}
1771
		list_del(&head->lru);
1772 1773
		h->free_huge_pages--;
		h->free_huge_pages_node[nid]--;
1774
		h->max_huge_pages--;
1775
		update_and_free_page(h, head);
1776
		rc = 0;
1777
	}
1778
out:
1779
	spin_unlock(&hugetlb_lock);
1780
	return rc;
1781 1782 1783 1784 1785
}

/*
 * Dissolve free hugepages in a given pfn range. Used by memory hotplug to
 * make specified memory blocks removable from the system.
1786 1787
 * Note that this will dissolve a free gigantic hugepage completely, if any
 * part of it lies within the given range.
1788 1789
 * Also note that if dissolve_free_huge_page() returns with an error, all
 * free hugepages that were dissolved before that error are lost.
1790
 */
1791
int dissolve_free_huge_pages(unsigned long start_pfn, unsigned long end_pfn)
1792 1793
{
	unsigned long pfn;
1794
	struct page *page;
1795
	int rc = 0;
1796

1797
	if (!hugepages_supported())
1798
		return rc;
1799

1800 1801
	for (pfn = start_pfn; pfn < end_pfn; pfn += 1 << minimum_order) {
		page = pfn_to_page(pfn);
1802 1803 1804
		rc = dissolve_free_huge_page(page);
		if (rc)
			break;
1805
	}
1806 1807

	return rc;
1808 1809
}

1810 1811 1812
/*
 * Allocates a fresh surplus page from the page allocator.
 */
1813
static struct page *alloc_surplus_huge_page(struct hstate *h, gfp_t gfp_mask,
1814
		int nid, nodemask_t *nmask)
1815
{
1816
	struct page *page = NULL;
1817

1818
	if (hstate_is_gigantic(h))
1819 1820
		return NULL;

1821
	spin_lock(&hugetlb_lock);
1822 1823
	if (h->surplus_huge_pages >= h->nr_overcommit_huge_pages)
		goto out_unlock;
1824 1825
	spin_unlock(&hugetlb_lock);

1826
	page = alloc_fresh_huge_page(h, gfp_mask, nid, nmask, NULL);
1827
	if (!page)
1828
		return NULL;
1829 1830

	spin_lock(&hugetlb_lock);
1831 1832 1833 1834 1835 1836 1837 1838
	/*
	 * 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) {
1839
		SetHPageTemporary(page);
1840
		spin_unlock(&hugetlb_lock);
1841
		put_page(page);
1842
		return NULL;
1843 1844
	} else {
		h->surplus_huge_pages++;
1845
		h->surplus_huge_pages_node[page_to_nid(page)]++;
1846
	}
1847 1848

out_unlock:
1849
	spin_unlock(&hugetlb_lock);
1850 1851 1852 1853

	return page;
}

1854
static struct page *alloc_migrate_huge_page(struct hstate *h, gfp_t gfp_mask,
1855
				     int nid, nodemask_t *nmask)
1856 1857 1858 1859 1860 1861
{
	struct page *page;

	if (hstate_is_gigantic(h))
		return NULL;

1862
	page = alloc_fresh_huge_page(h, gfp_mask, nid, nmask, NULL);
1863 1864 1865 1866 1867 1868 1869
	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
	 */
1870
	SetHPageTemporary(page);
1871 1872 1873 1874

	return page;
}

1875 1876 1877
/*
 * Use the VMA's mpolicy to allocate a huge page from the buddy.
 */
D
Dave Hansen 已提交
1878
static
1879
struct page *alloc_buddy_huge_page_with_mpol(struct hstate *h,
1880 1881
		struct vm_area_struct *vma, unsigned long addr)
{
1882 1883 1884 1885 1886 1887 1888
	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);
1889
	page = alloc_surplus_huge_page(h, gfp_mask, nid, nodemask);
1890 1891 1892
	mpol_cond_put(mpol);

	return page;
1893 1894
}

1895
/* page migration callback function */
1896
struct page *alloc_huge_page_nodemask(struct hstate *h, int preferred_nid,
1897
		nodemask_t *nmask, gfp_t gfp_mask)
1898 1899 1900
{
	spin_lock(&hugetlb_lock);
	if (h->free_huge_pages - h->resv_huge_pages > 0) {
1901 1902 1903 1904 1905 1906
		struct page *page;

		page = dequeue_huge_page_nodemask(h, gfp_mask, preferred_nid, nmask);
		if (page) {
			spin_unlock(&hugetlb_lock);
			return page;
1907 1908 1909 1910
		}
	}
	spin_unlock(&hugetlb_lock);

1911
	return alloc_migrate_huge_page(h, gfp_mask, preferred_nid, nmask);
1912 1913
}

1914
/* mempolicy aware migration callback */
1915 1916
struct page *alloc_huge_page_vma(struct hstate *h, struct vm_area_struct *vma,
		unsigned long address)
1917 1918 1919 1920 1921 1922 1923 1924 1925
{
	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);
1926
	page = alloc_huge_page_nodemask(h, node, nodemask, gfp_mask);
1927 1928 1929 1930 1931
	mpol_cond_put(mpol);

	return page;
}

1932
/*
L
Lucas De Marchi 已提交
1933
 * Increase the hugetlb pool such that it can accommodate a reservation
1934 1935
 * of size 'delta'.
 */
1936
static int gather_surplus_pages(struct hstate *h, long delta)
1937
	__must_hold(&hugetlb_lock)
1938 1939 1940
{
	struct list_head surplus_list;
	struct page *page, *tmp;
1941 1942 1943
	int ret;
	long i;
	long needed, allocated;
1944
	bool alloc_ok = true;
1945

1946
	needed = (h->resv_huge_pages + delta) - h->free_huge_pages;
1947
	if (needed <= 0) {
1948
		h->resv_huge_pages += delta;
1949
		return 0;
1950
	}
1951 1952 1953 1954 1955 1956 1957 1958

	allocated = 0;
	INIT_LIST_HEAD(&surplus_list);

	ret = -ENOMEM;
retry:
	spin_unlock(&hugetlb_lock);
	for (i = 0; i < needed; i++) {
1959
		page = alloc_surplus_huge_page(h, htlb_alloc_mask(h),
1960
				NUMA_NO_NODE, NULL);
1961 1962 1963 1964
		if (!page) {
			alloc_ok = false;
			break;
		}
1965
		list_add(&page->lru, &surplus_list);
1966
		cond_resched();
1967
	}
1968
	allocated += i;
1969 1970 1971 1972 1973 1974

	/*
	 * 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);
1975 1976
	needed = (h->resv_huge_pages + delta) -
			(h->free_huge_pages + allocated);
1977 1978 1979 1980 1981 1982 1983 1984 1985 1986
	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;
	}
1987 1988
	/*
	 * The surplus_list now contains _at_least_ the number of extra pages
L
Lucas De Marchi 已提交
1989
	 * needed to accommodate the reservation.  Add the appropriate number
1990
	 * of pages to the hugetlb pool and free the extras back to the buddy
1991 1992 1993
	 * 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.
1994 1995
	 */
	needed += allocated;
1996
	h->resv_huge_pages += delta;
1997
	ret = 0;
1998

1999
	/* Free the needed pages to the hugetlb pool */
2000
	list_for_each_entry_safe(page, tmp, &surplus_list, lru) {
2001 2002
		int zeroed;

2003 2004
		if ((--needed) < 0)
			break;
2005 2006 2007 2008
		/*
		 * This page is now managed by the hugetlb allocator and has
		 * no users -- drop the buddy allocator's reference.
		 */
2009 2010
		zeroed = put_page_testzero(page);
		VM_BUG_ON_PAGE(!zeroed, page);
2011
		enqueue_huge_page(h, page);
2012
	}
2013
free:
2014
	spin_unlock(&hugetlb_lock);
2015 2016

	/* Free unnecessary surplus pages to the buddy allocator */
2017 2018
	list_for_each_entry_safe(page, tmp, &surplus_list, lru)
		put_page(page);
2019
	spin_lock(&hugetlb_lock);
2020 2021 2022 2023 2024

	return ret;
}

/*
2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036
 * 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.
2037
 */
2038 2039
static void return_unused_surplus_pages(struct hstate *h,
					unsigned long unused_resv_pages)
2040 2041 2042
{
	unsigned long nr_pages;

2043
	/* Cannot return gigantic pages currently */
2044
	if (hstate_is_gigantic(h))
2045
		goto out;
2046

2047 2048 2049 2050
	/*
	 * Part (or even all) of the reservation could have been backed
	 * by pre-allocated pages. Only free surplus pages.
	 */
2051
	nr_pages = min(unused_resv_pages, h->surplus_huge_pages);
2052

2053 2054
	/*
	 * We want to release as many surplus pages as possible, spread
2055 2056 2057
	 * 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.
2058
	 * free_pool_huge_page() will balance the freed pages across the
2059
	 * on-line nodes with memory and will handle the hstate accounting.
2060 2061 2062 2063
	 *
	 * 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.
2064 2065
	 */
	while (nr_pages--) {
2066 2067
		h->resv_huge_pages--;
		unused_resv_pages--;
2068
		if (!free_pool_huge_page(h, &node_states[N_MEMORY], 1))
2069
			goto out;
2070
		cond_resched_lock(&hugetlb_lock);
2071
	}
2072 2073 2074 2075

out:
	/* Fully uncommit the reservation */
	h->resv_huge_pages -= unused_resv_pages;
2076 2077
}

2078

2079
/*
2080
 * vma_needs_reservation, vma_commit_reservation and vma_end_reservation
2081
 * are used by the huge page allocation routines to manage reservations.
2082 2083 2084 2085 2086 2087
 *
 * 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
2088 2089 2090
 * 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.
2091 2092 2093 2094 2095 2096
 *
 * 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.
2097 2098 2099 2100 2101
 *
 * 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.
2102
 */
2103 2104 2105
enum vma_resv_mode {
	VMA_NEEDS_RESV,
	VMA_COMMIT_RESV,
2106
	VMA_END_RESV,
2107
	VMA_ADD_RESV,
2108
};
2109 2110
static long __vma_reservation_common(struct hstate *h,
				struct vm_area_struct *vma, unsigned long addr,
2111
				enum vma_resv_mode mode)
2112
{
2113 2114
	struct resv_map *resv;
	pgoff_t idx;
2115
	long ret;
2116
	long dummy_out_regions_needed;
2117

2118 2119
	resv = vma_resv_map(vma);
	if (!resv)
2120
		return 1;
2121

2122
	idx = vma_hugecache_offset(h, vma, addr);
2123 2124
	switch (mode) {
	case VMA_NEEDS_RESV:
2125 2126 2127 2128 2129 2130
		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);
2131 2132
		break;
	case VMA_COMMIT_RESV:
2133
		ret = region_add(resv, idx, idx + 1, 1, NULL, NULL);
2134 2135
		/* region_add calls of range 1 should never fail. */
		VM_BUG_ON(ret < 0);
2136
		break;
2137
	case VMA_END_RESV:
2138
		region_abort(resv, idx, idx + 1, 1);
2139 2140
		ret = 0;
		break;
2141
	case VMA_ADD_RESV:
2142
		if (vma->vm_flags & VM_MAYSHARE) {
2143
			ret = region_add(resv, idx, idx + 1, 1, NULL, NULL);
2144 2145 2146 2147
			/* region_add calls of range 1 should never fail. */
			VM_BUG_ON(ret < 0);
		} else {
			region_abort(resv, idx, idx + 1, 1);
2148 2149 2150
			ret = region_del(resv, idx, idx + 1);
		}
		break;
2151 2152 2153
	default:
		BUG();
	}
2154

2155
	if (vma->vm_flags & VM_MAYSHARE)
2156
		return ret;
2157 2158 2159 2160 2161 2162 2163 2164 2165 2166 2167 2168 2169 2170 2171 2172 2173 2174 2175
	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;
	}
2176
	else
2177
		return ret < 0 ? ret : 0;
2178
}
2179 2180

static long vma_needs_reservation(struct hstate *h,
2181
			struct vm_area_struct *vma, unsigned long addr)
2182
{
2183
	return __vma_reservation_common(h, vma, addr, VMA_NEEDS_RESV);
2184
}
2185

2186 2187 2188
static long vma_commit_reservation(struct hstate *h,
			struct vm_area_struct *vma, unsigned long addr)
{
2189 2190 2191
	return __vma_reservation_common(h, vma, addr, VMA_COMMIT_RESV);
}

2192
static void vma_end_reservation(struct hstate *h,
2193 2194
			struct vm_area_struct *vma, unsigned long addr)
{
2195
	(void)__vma_reservation_common(h, vma, addr, VMA_END_RESV);
2196 2197
}

2198 2199 2200 2201 2202 2203 2204 2205 2206 2207
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,
2208 2209 2210 2211 2212 2213
 * alloc_huge_page would have consumed the reservation and set
 * HPageRestoreReserve in the newly allocated page.  When the page is freed
 * via free_huge_page, the global reservation count will be incremented if
 * HPageRestoreReserve 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.
2214 2215 2216 2217 2218
 */
static void restore_reserve_on_error(struct hstate *h,
			struct vm_area_struct *vma, unsigned long address,
			struct page *page)
{
2219
	if (unlikely(HPageRestoreReserve(page))) {
2220 2221 2222 2223 2224
		long rc = vma_needs_reservation(h, vma, address);

		if (unlikely(rc < 0)) {
			/*
			 * Rare out of memory condition in reserve map
2225
			 * manipulation.  Clear HPageRestoreReserve so that
2226 2227 2228 2229 2230 2231 2232 2233
			 * 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.
			 */
2234
			ClearHPageRestoreReserve(page);
2235 2236 2237 2238 2239 2240 2241
		} else if (rc) {
			rc = vma_add_reservation(h, vma, address);
			if (unlikely(rc < 0))
				/*
				 * See above comment about rare out of
				 * memory condition.
				 */
2242
				ClearHPageRestoreReserve(page);
2243 2244 2245 2246 2247
		} else
			vma_end_reservation(h, vma, address);
	}
}

2248
struct page *alloc_huge_page(struct vm_area_struct *vma,
2249
				    unsigned long addr, int avoid_reserve)
L
Linus Torvalds 已提交
2250
{
2251
	struct hugepage_subpool *spool = subpool_vma(vma);
2252
	struct hstate *h = hstate_vma(vma);
2253
	struct page *page;
2254 2255
	long map_chg, map_commit;
	long gbl_chg;
2256 2257
	int ret, idx;
	struct hugetlb_cgroup *h_cg;
2258
	bool deferred_reserve;
2259

2260
	idx = hstate_index(h);
2261
	/*
2262 2263 2264
	 * 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).
2265
	 */
2266 2267
	map_chg = gbl_chg = vma_needs_reservation(h, vma, addr);
	if (map_chg < 0)
2268
		return ERR_PTR(-ENOMEM);
2269 2270 2271 2272 2273 2274 2275 2276 2277 2278 2279

	/*
	 * 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) {
2280
			vma_end_reservation(h, vma, addr);
2281
			return ERR_PTR(-ENOSPC);
2282
		}
L
Linus Torvalds 已提交
2283

2284 2285 2286 2287 2288 2289 2290 2291 2292 2293 2294 2295
		/*
		 * 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;
	}

2296 2297 2298 2299 2300 2301 2302 2303 2304 2305
	/* 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;
	}

2306
	ret = hugetlb_cgroup_charge_cgroup(idx, pages_per_huge_page(h), &h_cg);
2307
	if (ret)
2308
		goto out_uncharge_cgroup_reservation;
2309

L
Linus Torvalds 已提交
2310
	spin_lock(&hugetlb_lock);
2311 2312 2313 2314 2315 2316
	/*
	 * 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);
2317
	if (!page) {
2318
		spin_unlock(&hugetlb_lock);
2319
		page = alloc_buddy_huge_page_with_mpol(h, vma, addr);
2320 2321
		if (!page)
			goto out_uncharge_cgroup;
2322
		if (!avoid_reserve && vma_has_reserves(vma, gbl_chg)) {
2323
			SetHPageRestoreReserve(page);
2324 2325
			h->resv_huge_pages--;
		}
2326
		spin_lock(&hugetlb_lock);
2327
		list_add(&page->lru, &h->hugepage_activelist);
2328
		/* Fall through */
K
Ken Chen 已提交
2329
	}
2330
	hugetlb_cgroup_commit_charge(idx, pages_per_huge_page(h), h_cg, page);
2331 2332 2333 2334 2335 2336 2337 2338
	/* 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);
	}

2339
	spin_unlock(&hugetlb_lock);
2340

2341
	hugetlb_set_page_subpool(page, spool);
2342

2343 2344
	map_commit = vma_commit_reservation(h, vma, addr);
	if (unlikely(map_chg > map_commit)) {
2345 2346 2347 2348 2349 2350 2351 2352 2353 2354 2355 2356 2357
		/*
		 * 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);
2358 2359 2360
		if (deferred_reserve)
			hugetlb_cgroup_uncharge_page_rsvd(hstate_index(h),
					pages_per_huge_page(h), page);
2361
	}
2362
	return page;
2363 2364 2365

out_uncharge_cgroup:
	hugetlb_cgroup_uncharge_cgroup(idx, pages_per_huge_page(h), h_cg);
2366 2367 2368 2369
out_uncharge_cgroup_reservation:
	if (deferred_reserve)
		hugetlb_cgroup_uncharge_cgroup_rsvd(idx, pages_per_huge_page(h),
						    h_cg);
2370
out_subpool_put:
2371
	if (map_chg || avoid_reserve)
2372
		hugepage_subpool_put_pages(spool, 1);
2373
	vma_end_reservation(h, vma, addr);
2374
	return ERR_PTR(-ENOSPC);
2375 2376
}

2377 2378 2379
int alloc_bootmem_huge_page(struct hstate *h)
	__attribute__ ((weak, alias("__alloc_bootmem_huge_page")));
int __alloc_bootmem_huge_page(struct hstate *h)
2380 2381
{
	struct huge_bootmem_page *m;
2382
	int nr_nodes, node;
2383

2384
	for_each_node_mask_to_alloc(h, nr_nodes, node, &node_states[N_MEMORY]) {
2385 2386
		void *addr;

2387
		addr = memblock_alloc_try_nid_raw(
2388
				huge_page_size(h), huge_page_size(h),
2389
				0, MEMBLOCK_ALLOC_ACCESSIBLE, node);
2390 2391 2392 2393 2394 2395 2396
		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;
2397
			goto found;
2398 2399 2400 2401 2402
		}
	}
	return 0;

found:
2403
	BUG_ON(!IS_ALIGNED(virt_to_phys(m), huge_page_size(h)));
2404
	/* Put them into a private list first because mem_map is not up yet */
2405
	INIT_LIST_HEAD(&m->list);
2406 2407 2408 2409 2410
	list_add(&m->list, &huge_boot_pages);
	m->hstate = h;
	return 1;
}

2411 2412
static void __init prep_compound_huge_page(struct page *page,
		unsigned int order)
2413 2414 2415 2416 2417 2418 2419
{
	if (unlikely(order > (MAX_ORDER - 1)))
		prep_compound_gigantic_page(page, order);
	else
		prep_compound_page(page, order);
}

2420 2421 2422 2423 2424 2425
/* 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) {
2426
		struct page *page = virt_to_page(m);
2427
		struct hstate *h = m->hstate;
2428

2429
		WARN_ON(page_count(page) != 1);
2430
		prep_compound_huge_page(page, huge_page_order(h));
2431
		WARN_ON(PageReserved(page));
2432
		prep_new_huge_page(h, page, page_to_nid(page));
2433 2434
		put_page(page); /* free it into the hugepage allocator */

2435 2436 2437 2438 2439 2440
		/*
		 * 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.
		 */
2441
		if (hstate_is_gigantic(h))
2442
			adjust_managed_page_count(page, pages_per_huge_page(h));
2443
		cond_resched();
2444 2445 2446
	}
}

2447
static void __init hugetlb_hstate_alloc_pages(struct hstate *h)
L
Linus Torvalds 已提交
2448 2449
{
	unsigned long i;
2450 2451 2452 2453 2454 2455 2456 2457 2458 2459 2460 2461 2462 2463 2464 2465 2466 2467 2468
	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);
2469

2470
	for (i = 0; i < h->max_huge_pages; ++i) {
2471
		if (hstate_is_gigantic(h)) {
2472
			if (hugetlb_cma_size) {
2473
				pr_warn_once("HugeTLB: hugetlb_cma is enabled, skip boot time allocation\n");
2474
				goto free;
2475
			}
2476 2477
			if (!alloc_bootmem_huge_page(h))
				break;
2478
		} else if (!alloc_pool_huge_page(h,
2479 2480
					 &node_states[N_MEMORY],
					 node_alloc_noretry))
L
Linus Torvalds 已提交
2481
			break;
2482
		cond_resched();
L
Linus Torvalds 已提交
2483
	}
2484 2485 2486
	if (i < h->max_huge_pages) {
		char buf[32];

2487
		string_get_size(huge_page_size(h), 1, STRING_UNITS_2, buf, 32);
2488 2489 2490 2491
		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;
	}
2492
free:
2493
	kfree(node_alloc_noretry);
2494 2495 2496 2497 2498 2499 2500
}

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

	for_each_hstate(h) {
2501 2502 2503
		if (minimum_order > huge_page_order(h))
			minimum_order = huge_page_order(h);

2504
		/* oversize hugepages were init'ed in early boot */
2505
		if (!hstate_is_gigantic(h))
2506
			hugetlb_hstate_alloc_pages(h);
2507
	}
2508
	VM_BUG_ON(minimum_order == UINT_MAX);
2509 2510 2511 2512 2513 2514 2515
}

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

	for_each_hstate(h) {
A
Andi Kleen 已提交
2516
		char buf[32];
2517 2518

		string_get_size(huge_page_size(h), 1, STRING_UNITS_2, buf, 32);
2519
		pr_info("HugeTLB registered %s page size, pre-allocated %ld pages\n",
2520
			buf, h->free_huge_pages);
2521 2522 2523
	}
}

L
Linus Torvalds 已提交
2524
#ifdef CONFIG_HIGHMEM
2525 2526
static void try_to_free_low(struct hstate *h, unsigned long count,
						nodemask_t *nodes_allowed)
L
Linus Torvalds 已提交
2527
{
2528 2529
	int i;

2530
	if (hstate_is_gigantic(h))
2531 2532
		return;

2533
	for_each_node_mask(i, *nodes_allowed) {
L
Linus Torvalds 已提交
2534
		struct page *page, *next;
2535 2536 2537
		struct list_head *freel = &h->hugepage_freelists[i];
		list_for_each_entry_safe(page, next, freel, lru) {
			if (count >= h->nr_huge_pages)
2538
				return;
L
Linus Torvalds 已提交
2539 2540 2541
			if (PageHighMem(page))
				continue;
			list_del(&page->lru);
2542
			update_and_free_page(h, page);
2543 2544
			h->free_huge_pages--;
			h->free_huge_pages_node[page_to_nid(page)]--;
L
Linus Torvalds 已提交
2545 2546 2547 2548
		}
	}
}
#else
2549 2550
static inline void try_to_free_low(struct hstate *h, unsigned long count,
						nodemask_t *nodes_allowed)
L
Linus Torvalds 已提交
2551 2552 2553 2554
{
}
#endif

2555 2556 2557 2558 2559
/*
 * 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.
 */
2560 2561
static int adjust_pool_surplus(struct hstate *h, nodemask_t *nodes_allowed,
				int delta)
2562
{
2563
	int nr_nodes, node;
2564 2565 2566

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

2567 2568 2569 2570
	if (delta < 0) {
		for_each_node_mask_to_alloc(h, nr_nodes, node, nodes_allowed) {
			if (h->surplus_huge_pages_node[node])
				goto found;
2571
		}
2572 2573 2574 2575 2576
	} 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;
2577
		}
2578 2579
	}
	return 0;
2580

2581 2582 2583 2584
found:
	h->surplus_huge_pages += delta;
	h->surplus_huge_pages_node[node] += delta;
	return 1;
2585 2586
}

2587
#define persistent_huge_pages(h) (h->nr_huge_pages - h->surplus_huge_pages)
2588
static int set_max_huge_pages(struct hstate *h, unsigned long count, int nid,
2589
			      nodemask_t *nodes_allowed)
L
Linus Torvalds 已提交
2590
{
2591
	unsigned long min_count, ret;
2592 2593 2594 2595 2596 2597 2598 2599 2600 2601 2602
	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
2603

2604 2605
	spin_lock(&hugetlb_lock);

2606 2607 2608 2609 2610 2611 2612 2613 2614 2615 2616 2617 2618 2619 2620 2621 2622 2623 2624 2625
	/*
	 * 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;
	}

2626 2627 2628 2629 2630 2631 2632 2633 2634 2635
	/*
	 * 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);
2636
			NODEMASK_FREE(node_alloc_noretry);
2637 2638 2639 2640
			return -EINVAL;
		}
		/* Fall through to decrease pool */
	}
2641

2642 2643 2644 2645
	/*
	 * Increase the pool size
	 * First take pages out of surplus state.  Then make up the
	 * remaining difference by allocating fresh huge pages.
2646
	 *
2647
	 * We might race with alloc_surplus_huge_page() here and be unable
2648 2649 2650 2651
	 * 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.
2652
	 */
2653
	while (h->surplus_huge_pages && count > persistent_huge_pages(h)) {
2654
		if (!adjust_pool_surplus(h, nodes_allowed, -1))
2655 2656 2657
			break;
	}

2658
	while (count > persistent_huge_pages(h)) {
2659 2660 2661 2662 2663 2664
		/*
		 * 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);
2665 2666 2667 2668

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

2669 2670
		ret = alloc_pool_huge_page(h, nodes_allowed,
						node_alloc_noretry);
2671 2672 2673 2674
		spin_lock(&hugetlb_lock);
		if (!ret)
			goto out;

2675 2676 2677
		/* Bail for signals. Probably ctrl-c from user */
		if (signal_pending(current))
			goto out;
2678 2679 2680 2681 2682 2683 2684 2685
	}

	/*
	 * 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.
2686 2687 2688 2689
	 *
	 * 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
2690
	 * alloc_surplus_huge_page() is checking the global counter,
2691 2692 2693
	 * 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.
2694
	 */
2695
	min_count = h->resv_huge_pages + h->nr_huge_pages - h->free_huge_pages;
2696
	min_count = max(count, min_count);
2697
	try_to_free_low(h, min_count, nodes_allowed);
2698
	while (min_count < persistent_huge_pages(h)) {
2699
		if (!free_pool_huge_page(h, nodes_allowed, 0))
L
Linus Torvalds 已提交
2700
			break;
2701
		cond_resched_lock(&hugetlb_lock);
L
Linus Torvalds 已提交
2702
	}
2703
	while (count < persistent_huge_pages(h)) {
2704
		if (!adjust_pool_surplus(h, nodes_allowed, 1))
2705 2706 2707
			break;
	}
out:
2708
	h->max_huge_pages = persistent_huge_pages(h);
L
Linus Torvalds 已提交
2709
	spin_unlock(&hugetlb_lock);
2710

2711 2712
	NODEMASK_FREE(node_alloc_noretry);

2713
	return 0;
L
Linus Torvalds 已提交
2714 2715
}

2716 2717 2718 2719 2720 2721 2722 2723 2724 2725
#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];

2726 2727 2728
static struct hstate *kobj_to_node_hstate(struct kobject *kobj, int *nidp);

static struct hstate *kobj_to_hstate(struct kobject *kobj, int *nidp)
2729 2730
{
	int i;
2731

2732
	for (i = 0; i < HUGE_MAX_HSTATE; i++)
2733 2734 2735
		if (hstate_kobjs[i] == kobj) {
			if (nidp)
				*nidp = NUMA_NO_NODE;
2736
			return &hstates[i];
2737 2738 2739
		}

	return kobj_to_node_hstate(kobj, nidp);
2740 2741
}

2742
static ssize_t nr_hugepages_show_common(struct kobject *kobj,
2743 2744
					struct kobj_attribute *attr, char *buf)
{
2745 2746 2747 2748 2749 2750 2751 2752 2753 2754
	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];

2755
	return sysfs_emit(buf, "%lu\n", nr_huge_pages);
2756
}
2757

2758 2759 2760
static ssize_t __nr_hugepages_store_common(bool obey_mempolicy,
					   struct hstate *h, int nid,
					   unsigned long count, size_t len)
2761 2762
{
	int err;
2763
	nodemask_t nodes_allowed, *n_mask;
2764

2765 2766
	if (hstate_is_gigantic(h) && !gigantic_page_runtime_supported())
		return -EINVAL;
2767

2768 2769 2770 2771 2772
	if (nid == NUMA_NO_NODE) {
		/*
		 * global hstate attribute
		 */
		if (!(obey_mempolicy &&
2773 2774 2775 2776 2777
				init_nodemask_of_mempolicy(&nodes_allowed)))
			n_mask = &node_states[N_MEMORY];
		else
			n_mask = &nodes_allowed;
	} else {
2778
		/*
2779 2780
		 * Node specific request.  count adjustment happens in
		 * set_max_huge_pages() after acquiring hugetlb_lock.
2781
		 */
2782 2783
		init_nodemask_of_node(&nodes_allowed, nid);
		n_mask = &nodes_allowed;
2784
	}
2785

2786
	err = set_max_huge_pages(h, count, nid, n_mask);
2787

2788
	return err ? err : len;
2789 2790
}

2791 2792 2793 2794 2795 2796 2797 2798 2799 2800 2801 2802 2803 2804 2805 2806 2807
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);
}

2808 2809 2810 2811 2812 2813 2814 2815 2816
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)
{
2817
	return nr_hugepages_store_common(false, kobj, buf, len);
2818 2819 2820
}
HSTATE_ATTR(nr_hugepages);

2821 2822 2823 2824 2825 2826 2827
#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,
2828 2829
					   struct kobj_attribute *attr,
					   char *buf)
2830 2831 2832 2833 2834 2835 2836
{
	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)
{
2837
	return nr_hugepages_store_common(true, kobj, buf, len);
2838 2839 2840 2841 2842
}
HSTATE_ATTR(nr_hugepages_mempolicy);
#endif


2843 2844 2845
static ssize_t nr_overcommit_hugepages_show(struct kobject *kobj,
					struct kobj_attribute *attr, char *buf)
{
2846
	struct hstate *h = kobj_to_hstate(kobj, NULL);
2847
	return sysfs_emit(buf, "%lu\n", h->nr_overcommit_huge_pages);
2848
}
2849

2850 2851 2852 2853 2854
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;
2855
	struct hstate *h = kobj_to_hstate(kobj, NULL);
2856

2857
	if (hstate_is_gigantic(h))
2858 2859
		return -EINVAL;

2860
	err = kstrtoul(buf, 10, &input);
2861
	if (err)
2862
		return err;
2863 2864 2865 2866 2867 2868 2869 2870 2871 2872 2873 2874

	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)
{
2875 2876 2877 2878 2879 2880 2881 2882 2883 2884
	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];

2885
	return sysfs_emit(buf, "%lu\n", free_huge_pages);
2886 2887 2888 2889 2890 2891
}
HSTATE_ATTR_RO(free_hugepages);

static ssize_t resv_hugepages_show(struct kobject *kobj,
					struct kobj_attribute *attr, char *buf)
{
2892
	struct hstate *h = kobj_to_hstate(kobj, NULL);
2893
	return sysfs_emit(buf, "%lu\n", h->resv_huge_pages);
2894 2895 2896 2897 2898 2899
}
HSTATE_ATTR_RO(resv_hugepages);

static ssize_t surplus_hugepages_show(struct kobject *kobj,
					struct kobj_attribute *attr, char *buf)
{
2900 2901 2902 2903 2904 2905 2906 2907 2908 2909
	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];

2910
	return sysfs_emit(buf, "%lu\n", surplus_huge_pages);
2911 2912 2913 2914 2915 2916 2917 2918 2919
}
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,
2920 2921 2922
#ifdef CONFIG_NUMA
	&nr_hugepages_mempolicy_attr.attr,
#endif
2923 2924 2925
	NULL,
};

2926
static const struct attribute_group hstate_attr_group = {
2927 2928 2929
	.attrs = hstate_attrs,
};

J
Jeff Mahoney 已提交
2930 2931
static int hugetlb_sysfs_add_hstate(struct hstate *h, struct kobject *parent,
				    struct kobject **hstate_kobjs,
2932
				    const struct attribute_group *hstate_attr_group)
2933 2934
{
	int retval;
2935
	int hi = hstate_index(h);
2936

2937 2938
	hstate_kobjs[hi] = kobject_create_and_add(h->name, parent);
	if (!hstate_kobjs[hi])
2939 2940
		return -ENOMEM;

2941
	retval = sysfs_create_group(hstate_kobjs[hi], hstate_attr_group);
2942
	if (retval) {
2943
		kobject_put(hstate_kobjs[hi]);
2944 2945
		hstate_kobjs[hi] = NULL;
	}
2946 2947 2948 2949 2950 2951 2952 2953 2954 2955 2956 2957 2958 2959

	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) {
2960 2961
		err = hugetlb_sysfs_add_hstate(h, hugepages_kobj,
					 hstate_kobjs, &hstate_attr_group);
2962
		if (err)
2963
			pr_err("HugeTLB: Unable to add hstate %s", h->name);
2964 2965 2966
	}
}

2967 2968 2969 2970
#ifdef CONFIG_NUMA

/*
 * node_hstate/s - associate per node hstate attributes, via their kobjects,
2971 2972 2973
 * 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
2974 2975 2976 2977 2978 2979
 * the base kernel, on the hugetlb module.
 */
struct node_hstate {
	struct kobject		*hugepages_kobj;
	struct kobject		*hstate_kobjs[HUGE_MAX_HSTATE];
};
2980
static struct node_hstate node_hstates[MAX_NUMNODES];
2981 2982

/*
2983
 * A subset of global hstate attributes for node devices
2984 2985 2986 2987 2988 2989 2990 2991
 */
static struct attribute *per_node_hstate_attrs[] = {
	&nr_hugepages_attr.attr,
	&free_hugepages_attr.attr,
	&surplus_hugepages_attr.attr,
	NULL,
};

2992
static const struct attribute_group per_node_hstate_attr_group = {
2993 2994 2995 2996
	.attrs = per_node_hstate_attrs,
};

/*
2997
 * kobj_to_node_hstate - lookup global hstate for node device hstate attr kobj.
2998 2999 3000 3001 3002 3003 3004 3005 3006 3007 3008 3009 3010 3011 3012 3013 3014 3015 3016 3017 3018 3019
 * 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;
}

/*
3020
 * Unregister hstate attributes from a single node device.
3021 3022
 * No-op if no hstate attributes attached.
 */
3023
static void hugetlb_unregister_node(struct node *node)
3024 3025
{
	struct hstate *h;
3026
	struct node_hstate *nhs = &node_hstates[node->dev.id];
3027 3028

	if (!nhs->hugepages_kobj)
3029
		return;		/* no hstate attributes */
3030

3031 3032 3033 3034 3035
	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;
3036
		}
3037
	}
3038 3039 3040 3041 3042 3043 3044

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


/*
3045
 * Register hstate attributes for a single node device.
3046 3047
 * No-op if attributes already registered.
 */
3048
static void hugetlb_register_node(struct node *node)
3049 3050
{
	struct hstate *h;
3051
	struct node_hstate *nhs = &node_hstates[node->dev.id];
3052 3053 3054 3055 3056 3057
	int err;

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

	nhs->hugepages_kobj = kobject_create_and_add("hugepages",
3058
							&node->dev.kobj);
3059 3060 3061 3062 3063 3064 3065 3066
	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) {
3067
			pr_err("HugeTLB: Unable to add hstate %s for node %d\n",
3068
				h->name, node->dev.id);
3069 3070 3071 3072 3073 3074 3075
			hugetlb_unregister_node(node);
			break;
		}
	}
}

/*
3076
 * hugetlb init time:  register hstate attributes for all registered node
3077 3078
 * devices of nodes that have memory.  All on-line nodes should have
 * registered their associated device by this time.
3079
 */
3080
static void __init hugetlb_register_all_nodes(void)
3081 3082 3083
{
	int nid;

3084
	for_each_node_state(nid, N_MEMORY) {
3085
		struct node *node = node_devices[nid];
3086
		if (node->dev.id == nid)
3087 3088 3089 3090
			hugetlb_register_node(node);
	}

	/*
3091
	 * Let the node device driver know we're here so it can
3092 3093 3094 3095 3096 3097 3098 3099 3100 3101 3102 3103 3104 3105 3106 3107 3108 3109 3110
	 * [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

3111 3112
static int __init hugetlb_init(void)
{
3113 3114
	int i;

3115 3116 3117
	BUILD_BUG_ON(sizeof_field(struct page, private) * BITS_PER_BYTE <
			__NR_HPAGEFLAGS);

3118 3119 3120
	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");
3121
		return 0;
3122
	}
3123

3124 3125 3126 3127 3128 3129 3130 3131 3132 3133 3134 3135 3136 3137 3138 3139 3140 3141 3142 3143 3144 3145 3146 3147 3148 3149 3150 3151
	/*
	 * 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;
3152
		}
3153
	}
3154

3155
	hugetlb_cma_check();
3156
	hugetlb_init_hstates();
3157
	gather_bootmem_prealloc();
3158 3159 3160
	report_hugepages();

	hugetlb_sysfs_init();
3161
	hugetlb_register_all_nodes();
3162
	hugetlb_cgroup_file_init();
3163

3164 3165 3166 3167 3168
#ifdef CONFIG_SMP
	num_fault_mutexes = roundup_pow_of_two(8 * num_possible_cpus());
#else
	num_fault_mutexes = 1;
#endif
3169
	hugetlb_fault_mutex_table =
3170 3171
		kmalloc_array(num_fault_mutexes, sizeof(struct mutex),
			      GFP_KERNEL);
3172
	BUG_ON(!hugetlb_fault_mutex_table);
3173 3174

	for (i = 0; i < num_fault_mutexes; i++)
3175
		mutex_init(&hugetlb_fault_mutex_table[i]);
3176 3177
	return 0;
}
3178
subsys_initcall(hugetlb_init);
3179

3180 3181
/* Overwritten by architectures with more huge page sizes */
bool __init __attribute((weak)) arch_hugetlb_valid_size(unsigned long size)
3182
{
3183
	return size == HPAGE_SIZE;
3184 3185
}

3186
void __init hugetlb_add_hstate(unsigned int order)
3187 3188
{
	struct hstate *h;
3189 3190
	unsigned long i;

3191 3192 3193
	if (size_to_hstate(PAGE_SIZE << order)) {
		return;
	}
3194
	BUG_ON(hugetlb_max_hstate >= HUGE_MAX_HSTATE);
3195
	BUG_ON(order == 0);
3196
	h = &hstates[hugetlb_max_hstate++];
3197
	h->order = order;
3198
	h->mask = ~(huge_page_size(h) - 1);
3199 3200
	for (i = 0; i < MAX_NUMNODES; ++i)
		INIT_LIST_HEAD(&h->hugepage_freelists[i]);
3201
	INIT_LIST_HEAD(&h->hugepage_activelist);
3202 3203
	h->next_nid_to_alloc = first_memory_node;
	h->next_nid_to_free = first_memory_node;
3204 3205
	snprintf(h->name, HSTATE_NAME_LEN, "hugepages-%lukB",
					huge_page_size(h)/1024);
3206

3207 3208 3209
	parsed_hstate = h;
}

3210 3211 3212 3213 3214 3215 3216 3217
/*
 * 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)
3218 3219
{
	unsigned long *mhp;
3220
	static unsigned long *last_mhp;
3221

3222
	if (!parsed_valid_hugepagesz) {
3223
		pr_warn("HugeTLB: hugepages=%s does not follow a valid hugepagesz, ignoring\n", s);
3224
		parsed_valid_hugepagesz = true;
3225
		return 0;
3226
	}
3227

3228
	/*
3229 3230 3231 3232
	 * !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.
3233
	 */
3234
	else if (!hugetlb_max_hstate)
3235 3236 3237 3238
		mhp = &default_hstate_max_huge_pages;
	else
		mhp = &parsed_hstate->max_huge_pages;

3239
	if (mhp == last_mhp) {
3240 3241
		pr_warn("HugeTLB: hugepages= specified twice without interleaving hugepagesz=, ignoring hugepages=%s\n", s);
		return 0;
3242 3243
	}

3244 3245 3246
	if (sscanf(s, "%lu", mhp) <= 0)
		*mhp = 0;

3247 3248 3249 3250 3251
	/*
	 * 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.
	 */
3252
	if (hugetlb_max_hstate && parsed_hstate->order >= MAX_ORDER)
3253 3254 3255 3256
		hugetlb_hstate_alloc_pages(parsed_hstate);

	last_mhp = mhp;

3257 3258
	return 1;
}
3259
__setup("hugepages=", hugepages_setup);
3260

3261 3262 3263 3264 3265 3266 3267
/*
 * 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.
 */
3268
static int __init hugepagesz_setup(char *s)
3269
{
3270
	unsigned long size;
3271 3272 3273
	struct hstate *h;

	parsed_valid_hugepagesz = false;
3274 3275 3276
	size = (unsigned long)memparse(s, NULL);

	if (!arch_hugetlb_valid_size(size)) {
3277
		pr_err("HugeTLB: unsupported hugepagesz=%s\n", s);
3278 3279 3280
		return 0;
	}

3281 3282 3283 3284 3285 3286 3287 3288 3289 3290 3291 3292 3293 3294 3295 3296 3297 3298 3299 3300 3301 3302 3303
	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;
3304 3305
	}

3306
	hugetlb_add_hstate(ilog2(size) - PAGE_SHIFT);
3307
	parsed_valid_hugepagesz = true;
3308 3309
	return 1;
}
3310 3311
__setup("hugepagesz=", hugepagesz_setup);

3312 3313 3314 3315
/*
 * default_hugepagesz command line input
 * Only one instance of default_hugepagesz allowed on command line.
 */
3316
static int __init default_hugepagesz_setup(char *s)
3317
{
3318 3319
	unsigned long size;

3320 3321 3322 3323 3324 3325
	parsed_valid_hugepagesz = false;
	if (parsed_default_hugepagesz) {
		pr_err("HugeTLB: default_hugepagesz previously specified, ignoring %s\n", s);
		return 0;
	}

3326 3327 3328
	size = (unsigned long)memparse(s, NULL);

	if (!arch_hugetlb_valid_size(size)) {
3329
		pr_err("HugeTLB: unsupported default_hugepagesz=%s\n", s);
3330 3331 3332
		return 0;
	}

3333 3334 3335 3336 3337 3338 3339 3340 3341 3342 3343 3344 3345 3346 3347 3348 3349 3350 3351
	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;
	}

3352 3353
	return 1;
}
3354
__setup("default_hugepagesz=", default_hugepagesz_setup);
3355

3356
static unsigned int allowed_mems_nr(struct hstate *h)
3357 3358 3359
{
	int node;
	unsigned int nr = 0;
3360 3361 3362 3363 3364
	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);
3365

3366
	for_each_node_mask(node, cpuset_current_mems_allowed) {
3367
		if (!mpol_allowed || node_isset(node, *mpol_allowed))
3368 3369
			nr += array[node];
	}
3370 3371 3372 3373 3374

	return nr;
}

#ifdef CONFIG_SYSCTL
3375 3376 3377 3378 3379 3380 3381 3382 3383 3384 3385 3386 3387 3388 3389 3390
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);
}

3391 3392
static int hugetlb_sysctl_handler_common(bool obey_mempolicy,
			 struct ctl_table *table, int write,
3393
			 void *buffer, size_t *length, loff_t *ppos)
L
Linus Torvalds 已提交
3394
{
3395
	struct hstate *h = &default_hstate;
3396
	unsigned long tmp = h->max_huge_pages;
3397
	int ret;
3398

3399
	if (!hugepages_supported())
3400
		return -EOPNOTSUPP;
3401

3402 3403
	ret = proc_hugetlb_doulongvec_minmax(table, write, buffer, length, ppos,
					     &tmp);
3404 3405
	if (ret)
		goto out;
3406

3407 3408 3409
	if (write)
		ret = __nr_hugepages_store_common(obey_mempolicy, h,
						  NUMA_NO_NODE, tmp, *length);
3410 3411
out:
	return ret;
L
Linus Torvalds 已提交
3412
}
3413

3414
int hugetlb_sysctl_handler(struct ctl_table *table, int write,
3415
			  void *buffer, size_t *length, loff_t *ppos)
3416 3417 3418 3419 3420 3421 3422 3423
{

	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,
3424
			  void *buffer, size_t *length, loff_t *ppos)
3425 3426 3427 3428 3429 3430
{
	return hugetlb_sysctl_handler_common(true, table, write,
							buffer, length, ppos);
}
#endif /* CONFIG_NUMA */

3431
int hugetlb_overcommit_handler(struct ctl_table *table, int write,
3432
		void *buffer, size_t *length, loff_t *ppos)
3433
{
3434
	struct hstate *h = &default_hstate;
3435
	unsigned long tmp;
3436
	int ret;
3437

3438
	if (!hugepages_supported())
3439
		return -EOPNOTSUPP;
3440

3441
	tmp = h->nr_overcommit_huge_pages;
3442

3443
	if (write && hstate_is_gigantic(h))
3444 3445
		return -EINVAL;

3446 3447
	ret = proc_hugetlb_doulongvec_minmax(table, write, buffer, length, ppos,
					     &tmp);
3448 3449
	if (ret)
		goto out;
3450 3451 3452 3453 3454 3455

	if (write) {
		spin_lock(&hugetlb_lock);
		h->nr_overcommit_huge_pages = tmp;
		spin_unlock(&hugetlb_lock);
	}
3456 3457
out:
	return ret;
3458 3459
}

L
Linus Torvalds 已提交
3460 3461
#endif /* CONFIG_SYSCTL */

3462
void hugetlb_report_meminfo(struct seq_file *m)
L
Linus Torvalds 已提交
3463
{
3464 3465 3466
	struct hstate *h;
	unsigned long total = 0;

3467 3468
	if (!hugepages_supported())
		return;
3469 3470 3471 3472

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

3473
		total += huge_page_size(h) * count;
3474 3475 3476 3477 3478 3479 3480 3481 3482 3483 3484 3485

		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,
3486
				   huge_page_size(h) / SZ_1K);
3487 3488
	}

3489
	seq_printf(m, "Hugetlb:        %8lu kB\n", total / SZ_1K);
L
Linus Torvalds 已提交
3490 3491
}

3492
int hugetlb_report_node_meminfo(char *buf, int len, int nid)
L
Linus Torvalds 已提交
3493
{
3494
	struct hstate *h = &default_hstate;
3495

3496 3497
	if (!hugepages_supported())
		return 0;
3498 3499 3500 3501 3502 3503 3504 3505

	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
3506 3507
}

3508 3509 3510 3511 3512
void hugetlb_show_meminfo(void)
{
	struct hstate *h;
	int nid;

3513 3514 3515
	if (!hugepages_supported())
		return;

3516 3517 3518 3519 3520 3521 3522
	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],
3523
				huge_page_size(h) / SZ_1K);
3524 3525
}

3526 3527 3528 3529 3530 3531
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
3532 3533 3534
/* Return the number pages of memory we physically have, in PAGE_SIZE units. */
unsigned long hugetlb_total_pages(void)
{
3535 3536 3537 3538 3539 3540
	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
3541 3542
}

3543
static int hugetlb_acct_memory(struct hstate *h, long delta)
M
Mel Gorman 已提交
3544 3545 3546
{
	int ret = -ENOMEM;

3547 3548 3549
	if (!delta)
		return 0;

M
Mel Gorman 已提交
3550 3551 3552 3553 3554 3555 3556 3557 3558 3559 3560 3561 3562 3563 3564 3565 3566
	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.
3567 3568 3569 3570 3571 3572
	 *
	 * 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
3573 3574
	 */
	if (delta > 0) {
3575
		if (gather_surplus_pages(h, delta) < 0)
M
Mel Gorman 已提交
3576 3577
			goto out;

3578
		if (delta > allowed_mems_nr(h)) {
3579
			return_unused_surplus_pages(h, delta);
M
Mel Gorman 已提交
3580 3581 3582 3583 3584 3585
			goto out;
		}
	}

	ret = 0;
	if (delta < 0)
3586
		return_unused_surplus_pages(h, (unsigned long) -delta);
M
Mel Gorman 已提交
3587 3588 3589 3590 3591 3592

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

3593 3594
static void hugetlb_vm_op_open(struct vm_area_struct *vma)
{
3595
	struct resv_map *resv = vma_resv_map(vma);
3596 3597 3598 3599 3600

	/*
	 * 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
3601
	 * has a reference to the reservation map it cannot disappear until
3602 3603 3604
	 * after this open call completes.  It is therefore safe to take a
	 * new reference here without additional locking.
	 */
3605
	if (resv && is_vma_resv_set(vma, HPAGE_RESV_OWNER))
3606
		kref_get(&resv->refs);
3607 3608
}

3609 3610
static void hugetlb_vm_op_close(struct vm_area_struct *vma)
{
3611
	struct hstate *h = hstate_vma(vma);
3612
	struct resv_map *resv = vma_resv_map(vma);
3613
	struct hugepage_subpool *spool = subpool_vma(vma);
3614
	unsigned long reserve, start, end;
3615
	long gbl_reserve;
3616

3617 3618
	if (!resv || !is_vma_resv_set(vma, HPAGE_RESV_OWNER))
		return;
3619

3620 3621
	start = vma_hugecache_offset(h, vma, vma->vm_start);
	end = vma_hugecache_offset(h, vma, vma->vm_end);
3622

3623
	reserve = (end - start) - region_count(resv, start, end);
3624
	hugetlb_cgroup_uncharge_counter(resv, start, end);
3625
	if (reserve) {
3626 3627 3628 3629 3630 3631
		/*
		 * 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);
3632
	}
3633 3634

	kref_put(&resv->refs, resv_map_release);
3635 3636
}

3637 3638 3639 3640 3641 3642 3643
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;
}

3644 3645
static unsigned long hugetlb_vm_op_pagesize(struct vm_area_struct *vma)
{
3646
	return huge_page_size(hstate_vma(vma));
3647 3648
}

L
Linus Torvalds 已提交
3649 3650 3651
/*
 * We cannot handle pagefaults against hugetlb pages at all.  They cause
 * handle_mm_fault() to try to instantiate regular-sized pages in the
M
Miaohe Lin 已提交
3652
 * hugepage VMA.  do_page_fault() is supposed to trap this, so BUG is we get
L
Linus Torvalds 已提交
3653 3654
 * this far.
 */
3655
static vm_fault_t hugetlb_vm_op_fault(struct vm_fault *vmf)
L
Linus Torvalds 已提交
3656 3657
{
	BUG();
N
Nick Piggin 已提交
3658
	return 0;
L
Linus Torvalds 已提交
3659 3660
}

3661 3662 3663 3664 3665 3666 3667
/*
 * 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.
 */
3668
const struct vm_operations_struct hugetlb_vm_ops = {
N
Nick Piggin 已提交
3669
	.fault = hugetlb_vm_op_fault,
3670
	.open = hugetlb_vm_op_open,
3671
	.close = hugetlb_vm_op_close,
3672
	.may_split = hugetlb_vm_op_split,
3673
	.pagesize = hugetlb_vm_op_pagesize,
L
Linus Torvalds 已提交
3674 3675
};

3676 3677
static pte_t make_huge_pte(struct vm_area_struct *vma, struct page *page,
				int writable)
D
David Gibson 已提交
3678 3679 3680
{
	pte_t entry;

3681
	if (writable) {
3682 3683
		entry = huge_pte_mkwrite(huge_pte_mkdirty(mk_huge_pte(page,
					 vma->vm_page_prot)));
D
David Gibson 已提交
3684
	} else {
3685 3686
		entry = huge_pte_wrprotect(mk_huge_pte(page,
					   vma->vm_page_prot));
D
David Gibson 已提交
3687 3688 3689
	}
	entry = pte_mkyoung(entry);
	entry = pte_mkhuge(entry);
3690
	entry = arch_make_huge_pte(entry, vma, page, writable);
D
David Gibson 已提交
3691 3692 3693 3694

	return entry;
}

3695 3696 3697 3698 3699
static void set_huge_ptep_writable(struct vm_area_struct *vma,
				   unsigned long address, pte_t *ptep)
{
	pte_t entry;

3700
	entry = huge_pte_mkwrite(huge_pte_mkdirty(huge_ptep_get(ptep)));
3701
	if (huge_ptep_set_access_flags(vma, address, ptep, entry, 1))
3702
		update_mmu_cache(vma, address, ptep);
3703 3704
}

3705
bool is_hugetlb_entry_migration(pte_t pte)
3706 3707 3708 3709
{
	swp_entry_t swp;

	if (huge_pte_none(pte) || pte_present(pte))
3710
		return false;
3711
	swp = pte_to_swp_entry(pte);
3712
	if (is_migration_entry(swp))
3713
		return true;
3714
	else
3715
		return false;
3716 3717
}

3718
static bool is_hugetlb_entry_hwpoisoned(pte_t pte)
3719 3720 3721 3722
{
	swp_entry_t swp;

	if (huge_pte_none(pte) || pte_present(pte))
3723
		return false;
3724
	swp = pte_to_swp_entry(pte);
3725
	if (is_hwpoison_entry(swp))
3726
		return true;
3727
	else
3728
		return false;
3729
}
3730

D
David Gibson 已提交
3731 3732 3733
int copy_hugetlb_page_range(struct mm_struct *dst, struct mm_struct *src,
			    struct vm_area_struct *vma)
{
3734
	pte_t *src_pte, *dst_pte, entry, dst_entry;
D
David Gibson 已提交
3735
	struct page *ptepage;
3736
	unsigned long addr;
3737
	bool cow = is_cow_mapping(vma->vm_flags);
3738 3739
	struct hstate *h = hstate_vma(vma);
	unsigned long sz = huge_page_size(h);
3740
	struct address_space *mapping = vma->vm_file->f_mapping;
3741
	struct mmu_notifier_range range;
3742
	int ret = 0;
3743

3744
	if (cow) {
3745
		mmu_notifier_range_init(&range, MMU_NOTIFY_CLEAR, 0, vma, src,
3746
					vma->vm_start,
3747 3748
					vma->vm_end);
		mmu_notifier_invalidate_range_start(&range);
3749 3750 3751 3752 3753 3754 3755 3756
	} 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);
3757
	}
3758

3759
	for (addr = vma->vm_start; addr < vma->vm_end; addr += sz) {
3760
		spinlock_t *src_ptl, *dst_ptl;
3761
		src_pte = huge_pte_offset(src, addr, sz);
H
Hugh Dickins 已提交
3762 3763
		if (!src_pte)
			continue;
3764
		dst_pte = huge_pte_alloc(dst, addr, sz);
3765 3766 3767 3768
		if (!dst_pte) {
			ret = -ENOMEM;
			break;
		}
3769

3770 3771 3772 3773 3774 3775 3776 3777 3778 3779 3780
		/*
		 * 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))
3781 3782
			continue;

3783 3784 3785
		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);
3786
		entry = huge_ptep_get(src_pte);
3787 3788 3789 3790 3791 3792 3793
		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.
			 */
3794 3795 3796 3797 3798 3799 3800 3801 3802 3803 3804 3805
			;
		} 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);
3806 3807
				set_huge_swap_pte_at(src, addr, src_pte,
						     entry, sz);
3808
			}
3809
			set_huge_swap_pte_at(dst, addr, dst_pte, entry, sz);
3810
		} else {
3811
			if (cow) {
3812 3813 3814 3815 3816
				/*
				 * No need to notify as we are downgrading page
				 * table protection not changing it to point
				 * to a new page.
				 *
3817
				 * See Documentation/vm/mmu_notifier.rst
3818
				 */
3819
				huge_ptep_set_wrprotect(src, addr, src_pte);
3820
			}
3821
			entry = huge_ptep_get(src_pte);
3822 3823
			ptepage = pte_page(entry);
			get_page(ptepage);
3824
			page_dup_rmap(ptepage, true);
3825
			set_huge_pte_at(dst, addr, dst_pte, entry);
3826
			hugetlb_count_add(pages_per_huge_page(h), dst);
3827
		}
3828 3829
		spin_unlock(src_ptl);
		spin_unlock(dst_ptl);
D
David Gibson 已提交
3830 3831
	}

3832
	if (cow)
3833
		mmu_notifier_invalidate_range_end(&range);
3834 3835
	else
		i_mmap_unlock_read(mapping);
3836 3837

	return ret;
D
David Gibson 已提交
3838 3839
}

3840 3841 3842
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
3843 3844 3845
{
	struct mm_struct *mm = vma->vm_mm;
	unsigned long address;
3846
	pte_t *ptep;
D
David Gibson 已提交
3847
	pte_t pte;
3848
	spinlock_t *ptl;
D
David Gibson 已提交
3849
	struct page *page;
3850 3851
	struct hstate *h = hstate_vma(vma);
	unsigned long sz = huge_page_size(h);
3852
	struct mmu_notifier_range range;
3853

D
David Gibson 已提交
3854
	WARN_ON(!is_vm_hugetlb_page(vma));
3855 3856
	BUG_ON(start & ~huge_page_mask(h));
	BUG_ON(end & ~huge_page_mask(h));
D
David Gibson 已提交
3857

3858 3859 3860 3861
	/*
	 * This is a hugetlb vma, all the pte entries should point
	 * to huge page.
	 */
3862
	tlb_change_page_size(tlb, sz);
3863
	tlb_start_vma(tlb, vma);
3864 3865 3866 3867

	/*
	 * If sharing possible, alert mmu notifiers of worst case.
	 */
3868 3869
	mmu_notifier_range_init(&range, MMU_NOTIFY_UNMAP, 0, vma, mm, start,
				end);
3870 3871
	adjust_range_if_pmd_sharing_possible(vma, &range.start, &range.end);
	mmu_notifier_invalidate_range_start(&range);
3872 3873
	address = start;
	for (; address < end; address += sz) {
3874
		ptep = huge_pte_offset(mm, address, sz);
A
Adam Litke 已提交
3875
		if (!ptep)
3876 3877
			continue;

3878
		ptl = huge_pte_lock(h, mm, ptep);
3879
		if (huge_pmd_unshare(mm, vma, &address, ptep)) {
3880
			spin_unlock(ptl);
3881 3882 3883 3884
			/*
			 * We just unmapped a page of PMDs by clearing a PUD.
			 * The caller's TLB flush range should cover this area.
			 */
3885 3886
			continue;
		}
3887

3888
		pte = huge_ptep_get(ptep);
3889 3890 3891 3892
		if (huge_pte_none(pte)) {
			spin_unlock(ptl);
			continue;
		}
3893 3894

		/*
3895 3896
		 * Migrating hugepage or HWPoisoned hugepage is already
		 * unmapped and its refcount is dropped, so just clear pte here.
3897
		 */
3898
		if (unlikely(!pte_present(pte))) {
3899
			huge_pte_clear(mm, address, ptep, sz);
3900 3901
			spin_unlock(ptl);
			continue;
3902
		}
3903 3904

		page = pte_page(pte);
3905 3906 3907 3908 3909 3910
		/*
		 * 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) {
3911 3912 3913 3914
			if (page != ref_page) {
				spin_unlock(ptl);
				continue;
			}
3915 3916 3917 3918 3919 3920 3921 3922
			/*
			 * 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);
		}

3923
		pte = huge_ptep_get_and_clear(mm, address, ptep);
3924
		tlb_remove_huge_tlb_entry(h, tlb, ptep, address);
3925
		if (huge_pte_dirty(pte))
3926
			set_page_dirty(page);
3927

3928
		hugetlb_count_sub(pages_per_huge_page(h), mm);
3929
		page_remove_rmap(page, true);
3930

3931
		spin_unlock(ptl);
3932
		tlb_remove_page_size(tlb, page, huge_page_size(h));
3933 3934 3935 3936 3937
		/*
		 * Bail out after unmapping reference page if supplied
		 */
		if (ref_page)
			break;
3938
	}
3939
	mmu_notifier_invalidate_range_end(&range);
3940
	tlb_end_vma(tlb, vma);
L
Linus Torvalds 已提交
3941
}
D
David Gibson 已提交
3942

3943 3944 3945 3946 3947 3948 3949 3950 3951 3952 3953 3954
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
3955
	 * is to clear it before releasing the i_mmap_rwsem. This works
3956
	 * because in the context this is called, the VMA is about to be
3957
	 * destroyed and the i_mmap_rwsem is held.
3958 3959 3960 3961
	 */
	vma->vm_flags &= ~VM_MAYSHARE;
}

3962
void unmap_hugepage_range(struct vm_area_struct *vma, unsigned long start,
3963
			  unsigned long end, struct page *ref_page)
3964
{
3965
	struct mmu_gather tlb;
3966

3967
	tlb_gather_mmu(&tlb, vma->vm_mm);
3968
	__unmap_hugepage_range(&tlb, vma, start, end, ref_page);
3969
	tlb_finish_mmu(&tlb);
3970 3971
}

3972 3973
/*
 * This is called when the original mapper is failing to COW a MAP_PRIVATE
3974
 * mapping it owns the reserve page for. The intention is to unmap the page
3975 3976 3977
 * from other VMAs and let the children be SIGKILLed if they are faulting the
 * same region.
 */
3978 3979
static void unmap_ref_private(struct mm_struct *mm, struct vm_area_struct *vma,
			      struct page *page, unsigned long address)
3980
{
3981
	struct hstate *h = hstate_vma(vma);
3982 3983 3984 3985 3986 3987 3988 3989
	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.
	 */
3990
	address = address & huge_page_mask(h);
3991 3992
	pgoff = ((address - vma->vm_start) >> PAGE_SHIFT) +
			vma->vm_pgoff;
3993
	mapping = vma->vm_file->f_mapping;
3994

3995 3996 3997 3998 3999
	/*
	 * 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
	 */
4000
	i_mmap_lock_write(mapping);
4001
	vma_interval_tree_foreach(iter_vma, &mapping->i_mmap, pgoff, pgoff) {
4002 4003 4004 4005
		/* Do not unmap the current VMA */
		if (iter_vma == vma)
			continue;

4006 4007 4008 4009 4010 4011 4012 4013
		/*
		 * 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;

4014 4015 4016 4017 4018 4019 4020 4021
		/*
		 * 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))
4022 4023
			unmap_hugepage_range(iter_vma, address,
					     address + huge_page_size(h), page);
4024
	}
4025
	i_mmap_unlock_write(mapping);
4026 4027
}

4028 4029
/*
 * Hugetlb_cow() should be called with page lock of the original hugepage held.
4030 4031 4032
 * 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.
4033
 */
4034
static vm_fault_t hugetlb_cow(struct mm_struct *mm, struct vm_area_struct *vma,
4035
		       unsigned long address, pte_t *ptep,
4036
		       struct page *pagecache_page, spinlock_t *ptl)
4037
{
4038
	pte_t pte;
4039
	struct hstate *h = hstate_vma(vma);
4040
	struct page *old_page, *new_page;
4041 4042
	int outside_reserve = 0;
	vm_fault_t ret = 0;
4043
	unsigned long haddr = address & huge_page_mask(h);
4044
	struct mmu_notifier_range range;
4045

4046
	pte = huge_ptep_get(ptep);
4047 4048
	old_page = pte_page(pte);

4049
retry_avoidcopy:
4050 4051
	/* If no-one else is actually using this page, avoid the copy
	 * and just make the page writable */
4052
	if (page_mapcount(old_page) == 1 && PageAnon(old_page)) {
4053
		page_move_anon_rmap(old_page, vma);
4054
		set_huge_ptep_writable(vma, haddr, ptep);
N
Nick Piggin 已提交
4055
		return 0;
4056 4057
	}

4058 4059 4060 4061 4062 4063 4064 4065 4066
	/*
	 * 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.
	 */
4067
	if (is_vma_resv_set(vma, HPAGE_RESV_OWNER) &&
4068 4069 4070
			old_page != pagecache_page)
		outside_reserve = 1;

4071
	get_page(old_page);
4072

4073 4074 4075 4076
	/*
	 * Drop page table lock as buddy allocator may be called. It will
	 * be acquired again before returning to the caller, as expected.
	 */
4077
	spin_unlock(ptl);
4078
	new_page = alloc_huge_page(vma, haddr, outside_reserve);
4079

4080
	if (IS_ERR(new_page)) {
4081 4082 4083 4084 4085 4086 4087 4088
		/*
		 * 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) {
4089 4090 4091 4092
			struct address_space *mapping = vma->vm_file->f_mapping;
			pgoff_t idx;
			u32 hash;

4093
			put_page(old_page);
4094
			BUG_ON(huge_pte_none(pte));
4095 4096 4097 4098 4099 4100 4101 4102 4103 4104 4105 4106 4107 4108
			/*
			 * 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);

4109
			unmap_ref_private(mm, vma, old_page, haddr);
4110 4111 4112

			i_mmap_lock_read(mapping);
			mutex_lock(&hugetlb_fault_mutex_table[hash]);
4113
			spin_lock(ptl);
4114
			ptep = huge_pte_offset(mm, haddr, huge_page_size(h));
4115 4116 4117 4118 4119 4120 4121 4122
			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;
4123 4124
		}

4125
		ret = vmf_error(PTR_ERR(new_page));
4126
		goto out_release_old;
4127 4128
	}

4129 4130 4131 4132
	/*
	 * When the original hugepage is shared one, it does not have
	 * anon_vma prepared.
	 */
4133
	if (unlikely(anon_vma_prepare(vma))) {
4134 4135
		ret = VM_FAULT_OOM;
		goto out_release_all;
4136
	}
4137

4138
	copy_user_huge_page(new_page, old_page, address, vma,
A
Andrea Arcangeli 已提交
4139
			    pages_per_huge_page(h));
N
Nick Piggin 已提交
4140
	__SetPageUptodate(new_page);
4141

4142
	mmu_notifier_range_init(&range, MMU_NOTIFY_CLEAR, 0, vma, mm, haddr,
4143
				haddr + huge_page_size(h));
4144
	mmu_notifier_invalidate_range_start(&range);
4145

4146
	/*
4147
	 * Retake the page table lock to check for racing updates
4148 4149
	 * before the page tables are altered
	 */
4150
	spin_lock(ptl);
4151
	ptep = huge_pte_offset(mm, haddr, huge_page_size(h));
4152
	if (likely(ptep && pte_same(huge_ptep_get(ptep), pte))) {
4153
		ClearHPageRestoreReserve(new_page);
4154

4155
		/* Break COW */
4156
		huge_ptep_clear_flush(vma, haddr, ptep);
4157
		mmu_notifier_invalidate_range(mm, range.start, range.end);
4158
		set_huge_pte_at(mm, haddr, ptep,
4159
				make_huge_pte(vma, new_page, 1));
4160
		page_remove_rmap(old_page, true);
4161
		hugepage_add_new_anon_rmap(new_page, vma, haddr);
4162
		SetHPageMigratable(new_page);
4163 4164 4165
		/* Make the old page be freed below */
		new_page = old_page;
	}
4166
	spin_unlock(ptl);
4167
	mmu_notifier_invalidate_range_end(&range);
4168
out_release_all:
4169
	restore_reserve_on_error(h, vma, haddr, new_page);
4170
	put_page(new_page);
4171
out_release_old:
4172
	put_page(old_page);
4173

4174 4175
	spin_lock(ptl); /* Caller expects lock to be held */
	return ret;
4176 4177
}

4178
/* Return the pagecache page at a given address within a VMA */
4179 4180
static struct page *hugetlbfs_pagecache_page(struct hstate *h,
			struct vm_area_struct *vma, unsigned long address)
4181 4182
{
	struct address_space *mapping;
4183
	pgoff_t idx;
4184 4185

	mapping = vma->vm_file->f_mapping;
4186
	idx = vma_hugecache_offset(h, vma, address);
4187 4188 4189 4190

	return find_lock_page(mapping, idx);
}

H
Hugh Dickins 已提交
4191 4192 4193 4194 4195
/*
 * 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
4196 4197 4198 4199 4200 4201 4202 4203 4204 4205 4206 4207 4208 4209 4210
			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;
}

4211 4212 4213 4214 4215 4216 4217 4218 4219
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;
4220
	ClearHPageRestoreReserve(page);
4221

4222 4223 4224 4225 4226 4227
	/*
	 * set page dirty so that it will not be removed from cache/file
	 * by non-hugetlbfs specific code paths.
	 */
	set_page_dirty(page);

4228 4229 4230 4231 4232 4233
	spin_lock(&inode->i_lock);
	inode->i_blocks += blocks_per_huge_page(h);
	spin_unlock(&inode->i_lock);
	return 0;
}

4234 4235 4236 4237
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)
4238
{
4239
	struct hstate *h = hstate_vma(vma);
4240
	vm_fault_t ret = VM_FAULT_SIGBUS;
4241
	int anon_rmap = 0;
A
Adam Litke 已提交
4242 4243
	unsigned long size;
	struct page *page;
4244
	pte_t new_pte;
4245
	spinlock_t *ptl;
4246
	unsigned long haddr = address & huge_page_mask(h);
4247
	bool new_page = false;
A
Adam Litke 已提交
4248

4249 4250 4251
	/*
	 * 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
4252
	 * COW. Warn that such a situation has occurred as it may not be obvious
4253 4254
	 */
	if (is_vma_resv_set(vma, HPAGE_RESV_UNMAPPED)) {
4255
		pr_warn_ratelimited("PID %d killed due to inadequate hugepage pool\n",
4256
			   current->pid);
4257 4258 4259
		return ret;
	}

A
Adam Litke 已提交
4260
	/*
4261 4262 4263
	 * 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
4264
	 */
4265 4266 4267 4268
	size = i_size_read(mapping->host) >> huge_page_shift(h);
	if (idx >= size)
		goto out;

4269 4270 4271
retry:
	page = find_lock_page(mapping, idx);
	if (!page) {
4272 4273 4274 4275 4276 4277 4278
		/*
		 * Check for page in userfault range
		 */
		if (userfaultfd_missing(vma)) {
			u32 hash;
			struct vm_fault vmf = {
				.vma = vma,
4279
				.address = haddr,
4280 4281 4282 4283 4284 4285 4286 4287 4288 4289 4290
				.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
				 */
			};

			/*
4291 4292 4293
			 * hugetlb_fault_mutex and i_mmap_rwsem must be
			 * dropped before handling userfault.  Reacquire
			 * after handling fault to make calling code simpler.
4294
			 */
4295
			hash = hugetlb_fault_mutex_hash(mapping, idx);
4296
			mutex_unlock(&hugetlb_fault_mutex_table[hash]);
4297
			i_mmap_unlock_read(mapping);
4298
			ret = handle_userfault(&vmf, VM_UFFD_MISSING);
4299
			i_mmap_lock_read(mapping);
4300 4301 4302 4303
			mutex_lock(&hugetlb_fault_mutex_table[hash]);
			goto out;
		}

4304
		page = alloc_huge_page(vma, haddr, 0);
4305
		if (IS_ERR(page)) {
4306 4307 4308 4309 4310 4311 4312 4313 4314 4315 4316 4317 4318 4319 4320 4321 4322 4323 4324
			/*
			 * 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);
4325
			ret = vmf_error(PTR_ERR(page));
4326 4327
			goto out;
		}
A
Andrea Arcangeli 已提交
4328
		clear_huge_page(page, address, pages_per_huge_page(h));
N
Nick Piggin 已提交
4329
		__SetPageUptodate(page);
4330
		new_page = true;
4331

4332
		if (vma->vm_flags & VM_MAYSHARE) {
4333
			int err = huge_add_to_page_cache(page, mapping, idx);
4334 4335 4336 4337 4338 4339
			if (err) {
				put_page(page);
				if (err == -EEXIST)
					goto retry;
				goto out;
			}
4340
		} else {
4341
			lock_page(page);
4342 4343 4344 4345
			if (unlikely(anon_vma_prepare(vma))) {
				ret = VM_FAULT_OOM;
				goto backout_unlocked;
			}
4346
			anon_rmap = 1;
4347
		}
4348
	} else {
4349 4350 4351 4352 4353 4354
		/*
		 * 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))) {
4355
			ret = VM_FAULT_HWPOISON_LARGE |
4356
				VM_FAULT_SET_HINDEX(hstate_index(h));
4357 4358
			goto backout_unlocked;
		}
4359
	}
4360

4361 4362 4363 4364 4365 4366
	/*
	 * 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.
	 */
4367
	if ((flags & FAULT_FLAG_WRITE) && !(vma->vm_flags & VM_SHARED)) {
4368
		if (vma_needs_reservation(h, vma, haddr) < 0) {
4369 4370 4371
			ret = VM_FAULT_OOM;
			goto backout_unlocked;
		}
4372
		/* Just decrements count, does not deallocate */
4373
		vma_end_reservation(h, vma, haddr);
4374
	}
4375

4376
	ptl = huge_pte_lock(h, mm, ptep);
N
Nick Piggin 已提交
4377
	ret = 0;
4378
	if (!huge_pte_none(huge_ptep_get(ptep)))
A
Adam Litke 已提交
4379 4380
		goto backout;

4381
	if (anon_rmap) {
4382
		ClearHPageRestoreReserve(page);
4383
		hugepage_add_new_anon_rmap(page, vma, haddr);
4384
	} else
4385
		page_dup_rmap(page, true);
4386 4387
	new_pte = make_huge_pte(vma, page, ((vma->vm_flags & VM_WRITE)
				&& (vma->vm_flags & VM_SHARED)));
4388
	set_huge_pte_at(mm, haddr, ptep, new_pte);
4389

4390
	hugetlb_count_add(pages_per_huge_page(h), mm);
4391
	if ((flags & FAULT_FLAG_WRITE) && !(vma->vm_flags & VM_SHARED)) {
4392
		/* Optimization, do the COW without a second fault */
4393
		ret = hugetlb_cow(mm, vma, address, ptep, page, ptl);
4394 4395
	}

4396
	spin_unlock(ptl);
4397 4398

	/*
4399 4400 4401
	 * Only set HPageMigratable in newly allocated pages.  Existing pages
	 * found in the pagecache may not have HPageMigratableset if they have
	 * been isolated for migration.
4402 4403
	 */
	if (new_page)
4404
		SetHPageMigratable(page);
4405

A
Adam Litke 已提交
4406 4407
	unlock_page(page);
out:
4408
	return ret;
A
Adam Litke 已提交
4409 4410

backout:
4411
	spin_unlock(ptl);
4412
backout_unlocked:
A
Adam Litke 已提交
4413
	unlock_page(page);
4414
	restore_reserve_on_error(h, vma, haddr, page);
A
Adam Litke 已提交
4415 4416
	put_page(page);
	goto out;
4417 4418
}

4419
#ifdef CONFIG_SMP
4420
u32 hugetlb_fault_mutex_hash(struct address_space *mapping, pgoff_t idx)
4421 4422 4423 4424
{
	unsigned long key[2];
	u32 hash;

4425 4426
	key[0] = (unsigned long) mapping;
	key[1] = idx;
4427

4428
	hash = jhash2((u32 *)&key, sizeof(key)/(sizeof(u32)), 0);
4429 4430 4431 4432 4433

	return hash & (num_fault_mutexes - 1);
}
#else
/*
M
Miaohe Lin 已提交
4434
 * For uniprocessor systems we always use a single mutex, so just
4435 4436
 * return 0 and avoid the hashing overhead.
 */
4437
u32 hugetlb_fault_mutex_hash(struct address_space *mapping, pgoff_t idx)
4438 4439 4440 4441 4442
{
	return 0;
}
#endif

4443
vm_fault_t hugetlb_fault(struct mm_struct *mm, struct vm_area_struct *vma,
4444
			unsigned long address, unsigned int flags)
4445
{
4446
	pte_t *ptep, entry;
4447
	spinlock_t *ptl;
4448
	vm_fault_t ret;
4449 4450
	u32 hash;
	pgoff_t idx;
4451
	struct page *page = NULL;
4452
	struct page *pagecache_page = NULL;
4453
	struct hstate *h = hstate_vma(vma);
4454
	struct address_space *mapping;
4455
	int need_wait_lock = 0;
4456
	unsigned long haddr = address & huge_page_mask(h);
4457

4458
	ptep = huge_pte_offset(mm, haddr, huge_page_size(h));
4459
	if (ptep) {
4460 4461 4462 4463 4464
		/*
		 * 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.
		 */
4465
		entry = huge_ptep_get(ptep);
N
Naoya Horiguchi 已提交
4466
		if (unlikely(is_hugetlb_entry_migration(entry))) {
4467
			migration_entry_wait_huge(vma, mm, ptep);
N
Naoya Horiguchi 已提交
4468 4469
			return 0;
		} else if (unlikely(is_hugetlb_entry_hwpoisoned(entry)))
4470
			return VM_FAULT_HWPOISON_LARGE |
4471
				VM_FAULT_SET_HINDEX(hstate_index(h));
4472 4473
	}

4474 4475
	/*
	 * Acquire i_mmap_rwsem before calling huge_pte_alloc and hold
4476 4477 4478 4479
	 * 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.
4480 4481 4482 4483 4484
	 *
	 * 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.
	 */
4485
	mapping = vma->vm_file->f_mapping;
4486 4487 4488 4489 4490 4491
	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;
	}
4492

4493 4494 4495 4496 4497
	/*
	 * 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.
	 */
4498
	idx = vma_hugecache_offset(h, vma, haddr);
4499
	hash = hugetlb_fault_mutex_hash(mapping, idx);
4500
	mutex_lock(&hugetlb_fault_mutex_table[hash]);
4501

4502 4503
	entry = huge_ptep_get(ptep);
	if (huge_pte_none(entry)) {
4504
		ret = hugetlb_no_page(mm, vma, mapping, idx, address, ptep, flags);
4505
		goto out_mutex;
4506
	}
4507

N
Nick Piggin 已提交
4508
	ret = 0;
4509

4510 4511 4512
	/*
	 * 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
4513 4514 4515
	 * an active hugepage in pagecache. This goto expects the 2nd page
	 * fault, and is_hugetlb_entry_(migration|hwpoisoned) check will
	 * properly handle it.
4516 4517 4518 4519
	 */
	if (!pte_present(entry))
		goto out_mutex;

4520 4521 4522 4523 4524 4525 4526 4527
	/*
	 * 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.
	 */
4528
	if ((flags & FAULT_FLAG_WRITE) && !huge_pte_write(entry)) {
4529
		if (vma_needs_reservation(h, vma, haddr) < 0) {
4530
			ret = VM_FAULT_OOM;
4531
			goto out_mutex;
4532
		}
4533
		/* Just decrements count, does not deallocate */
4534
		vma_end_reservation(h, vma, haddr);
4535

4536
		if (!(vma->vm_flags & VM_MAYSHARE))
4537
			pagecache_page = hugetlbfs_pagecache_page(h,
4538
								vma, haddr);
4539 4540
	}

4541 4542 4543 4544 4545 4546
	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;

4547 4548 4549 4550 4551 4552 4553
	/*
	 * 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)
4554 4555 4556 4557
		if (!trylock_page(page)) {
			need_wait_lock = 1;
			goto out_ptl;
		}
4558

4559
	get_page(page);
4560

4561
	if (flags & FAULT_FLAG_WRITE) {
4562
		if (!huge_pte_write(entry)) {
4563
			ret = hugetlb_cow(mm, vma, address, ptep,
4564
					  pagecache_page, ptl);
4565
			goto out_put_page;
4566
		}
4567
		entry = huge_pte_mkdirty(entry);
4568 4569
	}
	entry = pte_mkyoung(entry);
4570
	if (huge_ptep_set_access_flags(vma, haddr, ptep, entry,
4571
						flags & FAULT_FLAG_WRITE))
4572
		update_mmu_cache(vma, haddr, ptep);
4573 4574 4575 4576
out_put_page:
	if (page != pagecache_page)
		unlock_page(page);
	put_page(page);
4577 4578
out_ptl:
	spin_unlock(ptl);
4579 4580 4581 4582 4583

	if (pagecache_page) {
		unlock_page(pagecache_page);
		put_page(pagecache_page);
	}
4584
out_mutex:
4585
	mutex_unlock(&hugetlb_fault_mutex_table[hash]);
4586
	i_mmap_unlock_read(mapping);
4587 4588 4589 4590 4591 4592 4593 4594 4595
	/*
	 * 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);
4596
	return ret;
4597 4598
}

4599 4600 4601 4602 4603 4604 4605 4606 4607 4608 4609
/*
 * 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)
{
4610 4611 4612
	struct address_space *mapping;
	pgoff_t idx;
	unsigned long size;
4613
	int vm_shared = dst_vma->vm_flags & VM_SHARED;
4614 4615 4616 4617 4618 4619 4620 4621 4622 4623 4624 4625 4626 4627
	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,
4628
						pages_per_huge_page(h), false);
4629

4630
		/* fallback to copy_from_user outside mmap_lock */
4631
		if (unlikely(ret)) {
4632
			ret = -ENOENT;
4633 4634 4635 4636 4637 4638 4639 4640 4641 4642 4643 4644 4645 4646 4647 4648
			*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);

4649 4650 4651
	mapping = dst_vma->vm_file->f_mapping;
	idx = vma_hugecache_offset(h, dst_vma, dst_addr);

4652 4653 4654 4655
	/*
	 * If shared, add to page cache
	 */
	if (vm_shared) {
4656 4657 4658 4659
		size = i_size_read(mapping->host) >> huge_page_shift(h);
		ret = -EFAULT;
		if (idx >= size)
			goto out_release_nounlock;
4660

4661 4662 4663 4664 4665 4666
		/*
		 * 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.
		 */
4667 4668 4669 4670 4671
		ret = huge_add_to_page_cache(page, mapping, idx);
		if (ret)
			goto out_release_nounlock;
	}

4672 4673 4674
	ptl = huge_pte_lockptr(h, dst_mm, dst_pte);
	spin_lock(ptl);

4675 4676 4677 4678 4679 4680 4681 4682 4683 4684 4685 4686 4687 4688
	/*
	 * 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;

4689 4690 4691 4692
	ret = -EEXIST;
	if (!huge_pte_none(huge_ptep_get(dst_pte)))
		goto out_release_unlock;

4693 4694 4695
	if (vm_shared) {
		page_dup_rmap(page, true);
	} else {
4696
		ClearHPageRestoreReserve(page);
4697 4698
		hugepage_add_new_anon_rmap(page, dst_vma, dst_addr);
	}
4699 4700 4701 4702 4703 4704 4705 4706 4707 4708 4709 4710 4711 4712 4713 4714

	_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);
4715
	SetHPageMigratable(page);
4716 4717
	if (vm_shared)
		unlock_page(page);
4718 4719 4720 4721 4722
	ret = 0;
out:
	return ret;
out_release_unlock:
	spin_unlock(ptl);
4723 4724
	if (vm_shared)
		unlock_page(page);
4725
out_release_nounlock:
4726 4727 4728 4729
	put_page(page);
	goto out;
}

4730 4731 4732 4733 4734 4735 4736 4737 4738 4739 4740 4741 4742 4743
static void record_subpages_vmas(struct page *page, struct vm_area_struct *vma,
				 int refs, struct page **pages,
				 struct vm_area_struct **vmas)
{
	int nr;

	for (nr = 0; nr < refs; nr++) {
		if (likely(pages))
			pages[nr] = mem_map_offset(page, nr);
		if (vmas)
			vmas[nr] = vma;
	}
}

4744 4745 4746
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,
4747
			 long i, unsigned int flags, int *locked)
D
David Gibson 已提交
4748
{
4749 4750
	unsigned long pfn_offset;
	unsigned long vaddr = *position;
4751
	unsigned long remainder = *nr_pages;
4752
	struct hstate *h = hstate_vma(vma);
4753
	int err = -EFAULT, refs;
D
David Gibson 已提交
4754 4755

	while (vaddr < vma->vm_end && remainder) {
A
Adam Litke 已提交
4756
		pte_t *pte;
4757
		spinlock_t *ptl = NULL;
H
Hugh Dickins 已提交
4758
		int absent;
A
Adam Litke 已提交
4759
		struct page *page;
D
David Gibson 已提交
4760

4761 4762 4763 4764
		/*
		 * If we have a pending SIGKILL, don't keep faulting pages and
		 * potentially allocating memory.
		 */
4765
		if (fatal_signal_pending(current)) {
4766 4767 4768 4769
			remainder = 0;
			break;
		}

A
Adam Litke 已提交
4770 4771
		/*
		 * Some archs (sparc64, sh*) have multiple pte_ts to
H
Hugh Dickins 已提交
4772
		 * each hugepage.  We have to make sure we get the
A
Adam Litke 已提交
4773
		 * first, for the page indexing below to work.
4774 4775
		 *
		 * Note that page table lock is not held when pte is null.
A
Adam Litke 已提交
4776
		 */
4777 4778
		pte = huge_pte_offset(mm, vaddr & huge_page_mask(h),
				      huge_page_size(h));
4779 4780
		if (pte)
			ptl = huge_pte_lock(h, mm, pte);
H
Hugh Dickins 已提交
4781 4782 4783 4784
		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
4785 4786 4787 4788
		 * 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
4789
		 */
H
Hugh Dickins 已提交
4790 4791
		if (absent && (flags & FOLL_DUMP) &&
		    !hugetlbfs_pagecache_present(h, vma, vaddr)) {
4792 4793
			if (pte)
				spin_unlock(ptl);
H
Hugh Dickins 已提交
4794 4795 4796
			remainder = 0;
			break;
		}
D
David Gibson 已提交
4797

4798 4799 4800 4801 4802 4803 4804 4805 4806 4807 4808
		/*
		 * 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)) ||
4809 4810
		    ((flags & FOLL_WRITE) &&
		      !huge_pte_write(huge_ptep_get(pte)))) {
4811
			vm_fault_t ret;
4812
			unsigned int fault_flags = 0;
D
David Gibson 已提交
4813

4814 4815
			if (pte)
				spin_unlock(ptl);
4816 4817
			if (flags & FOLL_WRITE)
				fault_flags |= FAULT_FLAG_WRITE;
4818
			if (locked)
4819 4820
				fault_flags |= FAULT_FLAG_ALLOW_RETRY |
					FAULT_FLAG_KILLABLE;
4821 4822 4823 4824
			if (flags & FOLL_NOWAIT)
				fault_flags |= FAULT_FLAG_ALLOW_RETRY |
					FAULT_FLAG_RETRY_NOWAIT;
			if (flags & FOLL_TRIED) {
4825 4826 4827 4828
				/*
				 * Note: FAULT_FLAG_ALLOW_RETRY and
				 * FAULT_FLAG_TRIED can co-exist
				 */
4829 4830 4831 4832
				fault_flags |= FAULT_FLAG_TRIED;
			}
			ret = hugetlb_fault(mm, vma, vaddr, fault_flags);
			if (ret & VM_FAULT_ERROR) {
4833
				err = vm_fault_to_errno(ret, flags);
4834 4835 4836 4837
				remainder = 0;
				break;
			}
			if (ret & VM_FAULT_RETRY) {
4838
				if (locked &&
4839
				    !(fault_flags & FAULT_FLAG_RETRY_NOWAIT))
4840
					*locked = 0;
4841 4842 4843 4844 4845 4846 4847 4848 4849 4850 4851 4852 4853
				*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
4854 4855
		}

4856
		pfn_offset = (vaddr & ~huge_page_mask(h)) >> PAGE_SHIFT;
4857
		page = pte_page(huge_ptep_get(pte));
4858

4859 4860 4861 4862 4863 4864 4865 4866 4867 4868 4869 4870 4871 4872
		/*
		 * 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;
		}

4873 4874
		refs = min3(pages_per_huge_page(h) - pfn_offset,
			    (vma->vm_end - vaddr) >> PAGE_SHIFT, remainder);
4875

4876 4877 4878 4879 4880
		if (pages || vmas)
			record_subpages_vmas(mem_map_offset(page, pfn_offset),
					     vma, refs,
					     likely(pages) ? pages + i : NULL,
					     vmas ? vmas + i : NULL);
D
David Gibson 已提交
4881

4882
		if (pages) {
4883 4884 4885 4886 4887 4888 4889 4890 4891 4892
			/*
			 * try_grab_compound_head() 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:
			 */
4893
			if (WARN_ON_ONCE(!try_grab_compound_head(pages[i],
4894 4895 4896 4897 4898 4899 4900
								 refs,
								 flags))) {
				spin_unlock(ptl);
				remainder = 0;
				err = -ENOMEM;
				break;
			}
4901
		}
4902 4903 4904 4905 4906

		vaddr += (refs << PAGE_SHIFT);
		remainder -= refs;
		i += refs;

4907
		spin_unlock(ptl);
D
David Gibson 已提交
4908
	}
4909
	*nr_pages = remainder;
4910 4911 4912 4913 4914
	/*
	 * 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
4915 4916
	*position = vaddr;

4917
	return i ? i : err;
D
David Gibson 已提交
4918
}
4919

4920 4921 4922 4923 4924 4925 4926 4927
#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

4928
unsigned long hugetlb_change_protection(struct vm_area_struct *vma,
4929 4930 4931 4932 4933 4934
		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;
4935
	struct hstate *h = hstate_vma(vma);
4936
	unsigned long pages = 0;
4937
	bool shared_pmd = false;
4938
	struct mmu_notifier_range range;
4939 4940 4941

	/*
	 * In the case of shared PMDs, the area to flush could be beyond
4942
	 * start/end.  Set range.start/range.end to cover the maximum possible
4943 4944
	 * range if PMD sharing is possible.
	 */
4945 4946
	mmu_notifier_range_init(&range, MMU_NOTIFY_PROTECTION_VMA,
				0, vma, mm, start, end);
4947
	adjust_range_if_pmd_sharing_possible(vma, &range.start, &range.end);
4948 4949

	BUG_ON(address >= end);
4950
	flush_cache_range(vma, range.start, range.end);
4951

4952
	mmu_notifier_invalidate_range_start(&range);
4953
	i_mmap_lock_write(vma->vm_file->f_mapping);
4954
	for (; address < end; address += huge_page_size(h)) {
4955
		spinlock_t *ptl;
4956
		ptep = huge_pte_offset(mm, address, huge_page_size(h));
4957 4958
		if (!ptep)
			continue;
4959
		ptl = huge_pte_lock(h, mm, ptep);
4960
		if (huge_pmd_unshare(mm, vma, &address, ptep)) {
4961
			pages++;
4962
			spin_unlock(ptl);
4963
			shared_pmd = true;
4964
			continue;
4965
		}
4966 4967 4968 4969 4970 4971 4972 4973 4974 4975 4976 4977 4978
		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);
4979 4980
				set_huge_swap_pte_at(mm, address, ptep,
						     newpte, huge_page_size(h));
4981 4982 4983 4984 4985 4986
				pages++;
			}
			spin_unlock(ptl);
			continue;
		}
		if (!huge_pte_none(pte)) {
4987 4988 4989 4990
			pte_t old_pte;

			old_pte = huge_ptep_modify_prot_start(vma, address, ptep);
			pte = pte_mkhuge(huge_pte_modify(old_pte, newprot));
4991
			pte = arch_make_huge_pte(pte, vma, NULL, 0);
4992
			huge_ptep_modify_prot_commit(vma, address, ptep, old_pte, pte);
4993
			pages++;
4994
		}
4995
		spin_unlock(ptl);
4996
	}
4997
	/*
4998
	 * Must flush TLB before releasing i_mmap_rwsem: x86's huge_pmd_unshare
4999
	 * may have cleared our pud entry and done put_page on the page table:
5000
	 * once we release i_mmap_rwsem, another task can do the final put_page
5001 5002
	 * 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.
5003
	 */
5004
	if (shared_pmd)
5005
		flush_hugetlb_tlb_range(vma, range.start, range.end);
5006 5007
	else
		flush_hugetlb_tlb_range(vma, start, end);
5008 5009 5010 5011
	/*
	 * No need to call mmu_notifier_invalidate_range() we are downgrading
	 * page table protection not changing it to point to a new page.
	 *
5012
	 * See Documentation/vm/mmu_notifier.rst
5013
	 */
5014
	i_mmap_unlock_write(vma->vm_file->f_mapping);
5015
	mmu_notifier_invalidate_range_end(&range);
5016 5017

	return pages << h->order;
5018 5019
}

5020 5021
/* Return true if reservation was successful, false otherwise.  */
bool hugetlb_reserve_pages(struct inode *inode,
5022
					long from, long to,
5023
					struct vm_area_struct *vma,
5024
					vm_flags_t vm_flags)
5025
{
5026
	long chg, add = -1;
5027
	struct hstate *h = hstate_inode(inode);
5028
	struct hugepage_subpool *spool = subpool_inode(inode);
5029
	struct resv_map *resv_map;
5030
	struct hugetlb_cgroup *h_cg = NULL;
5031
	long gbl_reserve, regions_needed = 0;
5032

5033 5034 5035
	/* This should never happen */
	if (from > to) {
		VM_WARN(1, "%s called with a negative range\n", __func__);
5036
		return false;
5037 5038
	}

5039 5040 5041
	/*
	 * Only apply hugepage reservation if asked. At fault time, an
	 * attempt will be made for VM_NORESERVE to allocate a page
5042
	 * without using reserves
5043
	 */
5044
	if (vm_flags & VM_NORESERVE)
5045
		return true;
5046

5047 5048 5049 5050 5051 5052
	/*
	 * 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
	 */
5053
	if (!vma || vma->vm_flags & VM_MAYSHARE) {
5054 5055 5056 5057 5058
		/*
		 * 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).
		 */
5059
		resv_map = inode_resv_map(inode);
5060

5061
		chg = region_chg(resv_map, from, to, &regions_needed);
5062 5063

	} else {
5064
		/* Private mapping. */
5065
		resv_map = resv_map_alloc();
5066
		if (!resv_map)
5067
			return false;
5068

5069
		chg = to - from;
5070

5071 5072 5073 5074
		set_vma_resv_map(vma, resv_map);
		set_vma_resv_flags(vma, HPAGE_RESV_OWNER);
	}

5075
	if (chg < 0)
5076
		goto out_err;
5077

5078 5079
	if (hugetlb_cgroup_charge_cgroup_rsvd(hstate_index(h),
				chg * pages_per_huge_page(h), &h_cg) < 0)
5080 5081 5082 5083 5084 5085 5086 5087 5088
		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);
	}

5089 5090 5091 5092 5093 5094
	/*
	 * 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);
5095
	if (gbl_reserve < 0)
5096
		goto out_uncharge_cgroup;
5097 5098

	/*
5099
	 * Check enough hugepages are available for the reservation.
5100
	 * Hand the pages back to the subpool if there are not
5101
	 */
5102
	if (hugetlb_acct_memory(h, gbl_reserve) < 0)
5103
		goto out_put_pages;
5104 5105 5106 5107 5108 5109 5110 5111 5112 5113 5114 5115

	/*
	 * 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
	 */
5116
	if (!vma || vma->vm_flags & VM_MAYSHARE) {
5117
		add = region_add(resv_map, from, to, regions_needed, h, h_cg);
5118 5119 5120

		if (unlikely(add < 0)) {
			hugetlb_acct_memory(h, -gbl_reserve);
5121
			goto out_put_pages;
5122
		} else if (unlikely(chg > add)) {
5123 5124 5125 5126 5127 5128 5129 5130 5131
			/*
			 * 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;

5132 5133 5134 5135
			hugetlb_cgroup_uncharge_cgroup_rsvd(
				hstate_index(h),
				(chg - add) * pages_per_huge_page(h), h_cg);

5136 5137 5138 5139 5140
			rsv_adjust = hugepage_subpool_put_pages(spool,
								chg - add);
			hugetlb_acct_memory(h, -rsv_adjust);
		}
	}
5141 5142
	return true;

5143 5144 5145 5146 5147 5148
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);
5149
out_err:
5150
	if (!vma || vma->vm_flags & VM_MAYSHARE)
5151 5152 5153 5154 5155
		/* 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
5156 5157
	if (vma && is_vma_resv_set(vma, HPAGE_RESV_OWNER))
		kref_put(&resv_map->refs, resv_map_release);
5158
	return false;
5159 5160
}

5161 5162
long hugetlb_unreserve_pages(struct inode *inode, long start, long end,
								long freed)
5163
{
5164
	struct hstate *h = hstate_inode(inode);
5165
	struct resv_map *resv_map = inode_resv_map(inode);
5166
	long chg = 0;
5167
	struct hugepage_subpool *spool = subpool_inode(inode);
5168
	long gbl_reserve;
K
Ken Chen 已提交
5169

5170 5171 5172 5173
	/*
	 * Since this routine can be called in the evict inode path for all
	 * hugetlbfs inodes, resv_map could be NULL.
	 */
5174 5175 5176 5177 5178 5179 5180 5181 5182 5183 5184
	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
5185
	spin_lock(&inode->i_lock);
5186
	inode->i_blocks -= (blocks_per_huge_page(h) * freed);
K
Ken Chen 已提交
5187 5188
	spin_unlock(&inode->i_lock);

5189 5190 5191 5192 5193 5194
	/*
	 * 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);
5195 5196

	return 0;
5197
}
5198

5199 5200 5201 5202 5203 5204 5205 5206 5207 5208 5209
#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
5210 5211
	unsigned long vm_flags = vma->vm_flags & VM_LOCKED_CLEAR_MASK;
	unsigned long svm_flags = svma->vm_flags & VM_LOCKED_CLEAR_MASK;
5212 5213 5214 5215 5216 5217 5218

	/*
	 * match the virtual addresses, permission and the alignment of the
	 * page table page.
	 */
	if (pmd_index(addr) != pmd_index(saddr) ||
	    vm_flags != svm_flags ||
5219
	    !range_in_vma(svma, sbase, s_end))
5220 5221 5222 5223 5224
		return 0;

	return saddr;
}

5225
static bool vma_shareable(struct vm_area_struct *vma, unsigned long addr)
5226 5227 5228 5229 5230 5231 5232
{
	unsigned long base = addr & PUD_MASK;
	unsigned long end = base + PUD_SIZE;

	/*
	 * check on proper vm_flags and page table alignment
	 */
5233
	if (vma->vm_flags & VM_MAYSHARE && range_in_vma(vma, base, end))
5234 5235
		return true;
	return false;
5236 5237
}

5238 5239 5240 5241 5242 5243 5244 5245
/*
 * 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)
{
5246 5247
	unsigned long v_start = ALIGN(vma->vm_start, PUD_SIZE),
		v_end = ALIGN_DOWN(vma->vm_end, PUD_SIZE);
5248

5249 5250 5251 5252 5253 5254
	/*
	 * 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))
5255 5256
		return;

5257
	/* Extend the range to be PUD aligned for a worst case scenario */
5258 5259
	if (*start > v_start)
		*start = ALIGN_DOWN(*start, PUD_SIZE);
5260

5261 5262
	if (*end < v_end)
		*end = ALIGN(*end, PUD_SIZE);
5263 5264
}

5265 5266 5267 5268
/*
 * 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
5269 5270
 * code much cleaner.
 *
5271 5272 5273 5274 5275 5276 5277 5278 5279 5280
 * 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.
5281 5282 5283 5284 5285 5286 5287 5288 5289 5290 5291
 */
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;
5292
	spinlock_t *ptl;
5293 5294 5295 5296

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

5297
	i_mmap_assert_locked(mapping);
5298 5299 5300 5301 5302 5303
	vma_interval_tree_foreach(svma, &mapping->i_mmap, idx, idx) {
		if (svma == vma)
			continue;

		saddr = page_table_shareable(svma, vma, addr, idx);
		if (saddr) {
5304 5305
			spte = huge_pte_offset(svma->vm_mm, saddr,
					       vma_mmu_pagesize(svma));
5306 5307 5308 5309 5310 5311 5312 5313 5314 5315
			if (spte) {
				get_page(virt_to_page(spte));
				break;
			}
		}
	}

	if (!spte)
		goto out;

5316
	ptl = huge_pte_lock(hstate_vma(vma), mm, spte);
5317
	if (pud_none(*pud)) {
5318 5319
		pud_populate(mm, pud,
				(pmd_t *)((unsigned long)spte & PAGE_MASK));
5320
		mm_inc_nr_pmds(mm);
5321
	} else {
5322
		put_page(virt_to_page(spte));
5323
	}
5324
	spin_unlock(ptl);
5325 5326 5327 5328 5329 5330 5331 5332 5333 5334 5335 5336
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.
 *
5337
 * Called with page table lock held and i_mmap_rwsem held in write mode.
5338 5339 5340 5341
 *
 * returns: 1 successfully unmapped a shared pte page
 *	    0 the underlying pte page is not shared, or it is the last user
 */
5342 5343
int huge_pmd_unshare(struct mm_struct *mm, struct vm_area_struct *vma,
					unsigned long *addr, pte_t *ptep)
5344 5345
{
	pgd_t *pgd = pgd_offset(mm, *addr);
5346 5347
	p4d_t *p4d = p4d_offset(pgd, *addr);
	pud_t *pud = pud_offset(p4d, *addr);
5348

5349
	i_mmap_assert_write_locked(vma->vm_file->f_mapping);
5350 5351 5352 5353 5354 5355
	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));
5356
	mm_dec_nr_pmds(mm);
5357 5358 5359
	*addr = ALIGN(*addr, HPAGE_SIZE * PTRS_PER_PTE) - HPAGE_SIZE;
	return 1;
}
5360 5361 5362 5363 5364 5365
#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;
}
5366

5367 5368
int huge_pmd_unshare(struct mm_struct *mm, struct vm_area_struct *vma,
				unsigned long *addr, pte_t *ptep)
5369 5370 5371
{
	return 0;
}
5372 5373 5374 5375 5376

void adjust_range_if_pmd_sharing_possible(struct vm_area_struct *vma,
				unsigned long *start, unsigned long *end)
{
}
5377
#define want_pmd_share()	(0)
5378 5379
#endif /* CONFIG_ARCH_WANT_HUGE_PMD_SHARE */

5380 5381 5382 5383 5384
#ifdef CONFIG_ARCH_WANT_GENERAL_HUGETLB
pte_t *huge_pte_alloc(struct mm_struct *mm,
			unsigned long addr, unsigned long sz)
{
	pgd_t *pgd;
5385
	p4d_t *p4d;
5386 5387 5388 5389
	pud_t *pud;
	pte_t *pte = NULL;

	pgd = pgd_offset(mm, addr);
5390 5391 5392
	p4d = p4d_alloc(mm, pgd, addr);
	if (!p4d)
		return NULL;
5393
	pud = pud_alloc(mm, p4d, addr);
5394 5395 5396 5397 5398 5399 5400 5401 5402 5403 5404
	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);
		}
	}
5405
	BUG_ON(pte && pte_present(*pte) && !pte_huge(*pte));
5406 5407 5408 5409

	return pte;
}

5410 5411 5412 5413
/*
 * huge_pte_offset() - Walk the page table to resolve the hugepage
 * entry at address @addr
 *
5414 5415
 * Return: Pointer to page table entry (PUD or PMD) for
 * address @addr, or NULL if a !p*d_present() entry is encountered and the
5416 5417 5418
 * size @sz doesn't match the hugepage size at this level of the page
 * table.
 */
5419 5420
pte_t *huge_pte_offset(struct mm_struct *mm,
		       unsigned long addr, unsigned long sz)
5421 5422
{
	pgd_t *pgd;
5423
	p4d_t *p4d;
5424 5425
	pud_t *pud;
	pmd_t *pmd;
5426 5427

	pgd = pgd_offset(mm, addr);
5428 5429 5430 5431 5432
	if (!pgd_present(*pgd))
		return NULL;
	p4d = p4d_offset(pgd, addr);
	if (!p4d_present(*p4d))
		return NULL;
5433

5434
	pud = pud_offset(p4d, addr);
5435 5436
	if (sz == PUD_SIZE)
		/* must be pud huge, non-present or none */
5437
		return (pte_t *)pud;
5438
	if (!pud_present(*pud))
5439
		return NULL;
5440
	/* must have a valid entry and size to go further */
5441

5442 5443 5444
	pmd = pmd_offset(pud, addr);
	/* must be pmd huge, non-present or none */
	return (pte_t *)pmd;
5445 5446
}

5447 5448 5449 5450 5451 5452 5453 5454 5455 5456 5457 5458 5459
#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);
}

5460 5461 5462 5463 5464 5465 5466 5467
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;
}

5468
struct page * __weak
5469
follow_huge_pmd(struct mm_struct *mm, unsigned long address,
5470
		pmd_t *pmd, int flags)
5471
{
5472 5473
	struct page *page = NULL;
	spinlock_t *ptl;
5474
	pte_t pte;
J
John Hubbard 已提交
5475 5476 5477 5478 5479 5480

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

5481 5482 5483 5484 5485 5486 5487 5488 5489
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;
5490 5491
	pte = huge_ptep_get((pte_t *)pmd);
	if (pte_present(pte)) {
5492
		page = pmd_page(*pmd) + ((address & ~PMD_MASK) >> PAGE_SHIFT);
J
John Hubbard 已提交
5493 5494 5495 5496 5497 5498 5499 5500 5501 5502 5503 5504
		/*
		 * 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;
		}
5505
	} else {
5506
		if (is_hugetlb_entry_migration(pte)) {
5507 5508 5509 5510 5511 5512 5513 5514 5515 5516 5517
			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);
5518 5519 5520
	return page;
}

5521
struct page * __weak
5522
follow_huge_pud(struct mm_struct *mm, unsigned long address,
5523
		pud_t *pud, int flags)
5524
{
J
John Hubbard 已提交
5525
	if (flags & (FOLL_GET | FOLL_PIN))
5526
		return NULL;
5527

5528
	return pte_page(*(pte_t *)pud) + ((address & ~PUD_MASK) >> PAGE_SHIFT);
5529 5530
}

5531 5532 5533
struct page * __weak
follow_huge_pgd(struct mm_struct *mm, unsigned long address, pgd_t *pgd, int flags)
{
J
John Hubbard 已提交
5534
	if (flags & (FOLL_GET | FOLL_PIN))
5535 5536 5537 5538 5539
		return NULL;

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

5540 5541
bool isolate_huge_page(struct page *page, struct list_head *list)
{
5542 5543
	bool ret = true;

5544
	spin_lock(&hugetlb_lock);
5545 5546
	if (!PageHeadHuge(page) ||
	    !HPageMigratable(page) ||
5547
	    !get_page_unless_zero(page)) {
5548 5549 5550
		ret = false;
		goto unlock;
	}
5551
	ClearHPageMigratable(page);
5552
	list_move_tail(&page->lru, list);
5553
unlock:
5554
	spin_unlock(&hugetlb_lock);
5555
	return ret;
5556 5557 5558 5559 5560
}

void putback_active_hugepage(struct page *page)
{
	spin_lock(&hugetlb_lock);
5561
	SetHPageMigratable(page);
5562 5563 5564 5565
	list_move_tail(&page->lru, &(page_hstate(page))->hugepage_activelist);
	spin_unlock(&hugetlb_lock);
	put_page(page);
}
5566 5567 5568 5569 5570 5571 5572 5573 5574 5575 5576 5577 5578 5579 5580 5581 5582 5583

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.
	 */
5584
	if (HPageTemporary(newpage)) {
5585 5586 5587
		int old_nid = page_to_nid(oldpage);
		int new_nid = page_to_nid(newpage);

5588 5589
		SetHPageTemporary(oldpage);
		ClearHPageTemporary(newpage);
5590 5591 5592 5593 5594 5595 5596 5597 5598

		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);
	}
}
5599 5600 5601 5602 5603 5604 5605 5606 5607 5608 5609 5610 5611 5612 5613 5614 5615 5616 5617 5618 5619 5620 5621 5622 5623 5624 5625 5626 5627 5628 5629 5630 5631 5632 5633 5634 5635 5636 5637

#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;
5638
		char name[CMA_MAX_NAME];
5639 5640 5641 5642

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

5643
		snprintf(name, sizeof(name), "hugetlb%d", nid);
5644
		res = cma_declare_contiguous_nid(0, size, 0, PAGE_SIZE << order,
5645
						 0, false, name,
5646 5647 5648 5649 5650 5651 5652 5653 5654 5655 5656 5657 5658 5659 5660 5661 5662 5663 5664 5665 5666 5667 5668 5669 5670
						 &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 */