hugetlb.c 163.4 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,
						unsigned long irq_flags)
99
{
100
	spin_unlock_irqrestore(&spool->lock, irq_flags);
101 102

	/* If no pages are used, and no other handles to the subpool
E
Ethon Paul 已提交
103
	 * remain, give up any reservations based on minimum size and
104
	 * free the subpool */
105
	if (subpool_is_free(spool)) {
106 107 108
		if (spool->min_hpages != -1)
			hugetlb_acct_memory(spool->hstate,
						-spool->min_hpages);
109
		kfree(spool);
110
	}
111 112
}

113 114
struct hugepage_subpool *hugepage_new_subpool(struct hstate *h, long max_hpages,
						long min_hpages)
115 116 117
{
	struct hugepage_subpool *spool;

118
	spool = kzalloc(sizeof(*spool), GFP_KERNEL);
119 120 121 122 123
	if (!spool)
		return NULL;

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

	return spool;
}

void hugepage_put_subpool(struct hugepage_subpool *spool)
{
139 140 141
	unsigned long flags;

	spin_lock_irqsave(&spool->lock, flags);
142 143
	BUG_ON(!spool->count);
	spool->count--;
144
	unlock_or_release_subpool(spool, flags);
145 146
}

147 148 149
/*
 * Subpool accounting for allocating and reserving pages.
 * Return -ENOMEM if there are not enough resources to satisfy the
150
 * request.  Otherwise, return the number of pages by which the
151 152
 * 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
153
 * a subpool minimum size must be maintained.
154 155
 */
static long hugepage_subpool_get_pages(struct hugepage_subpool *spool,
156 157
				      long delta)
{
158
	long ret = delta;
159 160

	if (!spool)
161
		return ret;
162

163
	spin_lock_irq(&spool->lock);
164 165 166 167 168 169 170 171

	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;
		}
172 173
	}

174 175
	/* minimum size accounting */
	if (spool->min_hpages != -1 && spool->rsv_hpages) {
176 177 178 179 180 181 182 183 184 185 186 187 188 189
		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:
190
	spin_unlock_irq(&spool->lock);
191 192 193
	return ret;
}

194 195 196 197 198 199 200
/*
 * 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,
201 202
				       long delta)
{
203
	long ret = delta;
204
	unsigned long flags;
205

206
	if (!spool)
207
		return delta;
208

209
	spin_lock_irqsave(&spool->lock, flags);
210 211 212 213

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

214 215
	 /* minimum size accounting */
	if (spool->min_hpages != -1 && spool->used_hpages < spool->min_hpages) {
216 217 218 219 220 221 222 223 224 225 226 227 228 229
		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.
	 */
230
	unlock_or_release_subpool(spool, flags);
231 232

	return ret;
233 234 235 236 237 238 239 240 241
}

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
242
	return subpool_inode(file_inode(vma->vm_file));
243 244
}

245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264
/* 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;
}

265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286
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;
287 288 289 290 291 292 293 294 295 296 297
		/*
		 * The caller will hold exactly one h_cg->css reference for the
		 * whole contiguous reservation region. But this area might be
		 * scattered when there are already some file_regions reside in
		 * it. As a result, many file_regions may share only one css
		 * reference. In order to ensure that one file_region must hold
		 * exactly one h_cg->css reference, we should do css_get for
		 * each file_region and leave the reference held by caller
		 * untouched.
		 */
		css_get(&h_cg->css);
298 299 300 301 302 303 304 305 306 307 308 309 310
		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
}

311 312 313 314 315 316 317 318
static void put_uncharge_info(struct file_region *rg)
{
#ifdef CONFIG_CGROUP_HUGETLB
	if (rg->css)
		css_put(rg->css);
#endif
}

319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341
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);
342
		put_uncharge_info(rg);
343 344
		kfree(rg);

345
		rg = prg;
346 347 348 349 350 351 352 353
	}

	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);
354
		put_uncharge_info(rg);
355 356 357 358
		kfree(rg);
	}
}

359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376
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;
}

377 378 379 380 381 382 383
/*
 * 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
384 385
 */
static long add_reservation_in_range(struct resv_map *resv, long f, long t,
386
				     struct hugetlb_cgroup *h_cg,
387
				     struct hstate *h, long *regions_needed)
M
Mina Almasry 已提交
388
{
389
	long add = 0;
M
Mina Almasry 已提交
390
	struct list_head *head = &resv->regions;
391
	long last_accounted_offset = f;
392
	struct file_region *rg = NULL, *trg = NULL;
M
Mina Almasry 已提交
393

394 395
	if (regions_needed)
		*regions_needed = 0;
M
Mina Almasry 已提交
396

397 398 399 400 401 402 403 404 405 406 407 408 409 410
	/* 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
411

412 413 414
		/* When we find a region that starts beyond our range, we've
		 * finished.
		 */
415
		if (rg->from >= t)
M
Mina Almasry 已提交
416 417
			break;

418 419 420
		/* Add an entry for last_accounted_offset -> rg->from, and
		 * update last_accounted_offset.
		 */
421 422 423 424 425
		if (rg->from > last_accounted_offset)
			add += hugetlb_resv_map_add(resv, rg,
						    last_accounted_offset,
						    rg->from, h, h_cg,
						    regions_needed);
426 427 428 429 430 431 432

		last_accounted_offset = rg->to;
	}

	/* Handle the case where our range extends beyond
	 * last_accounted_offset.
	 */
433 434 435
	if (last_accounted_offset < t)
		add += hugetlb_resv_map_add(resv, rg, last_accounted_offset,
					    t, h, h_cg, regions_needed);
436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471

	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
472
		 */
473 474 475 476 477 478 479 480
		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
481 482
		}

483 484
		spin_lock(&resv->lock);

485 486
		list_splice(&allocated_regions, &resv->region_cache);
		resv->region_cache_count += to_allocate;
M
Mina Almasry 已提交
487 488
	}

489
	return 0;
M
Mina Almasry 已提交
490

491 492 493 494 495 496
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
497 498
}

499 500
/*
 * Add the huge page range represented by [f, t) to the reserve
501 502 503 504 505
 * 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.
506
 *
507 508 509 510
 * 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
511
 * this operation and we were not able to allocate, it returns -ENOMEM.
512 513 514
 * 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.
515
 */
516
static long region_add(struct resv_map *resv, long f, long t,
517 518
		       long in_regions_needed, struct hstate *h,
		       struct hugetlb_cgroup *h_cg)
519
{
520
	long add = 0, actual_regions_needed = 0;
521

522
	spin_lock(&resv->lock);
523 524 525
retry:

	/* Count how many regions are actually needed to execute this add. */
526 527
	add_reservation_in_range(resv, f, t, NULL, NULL,
				 &actual_regions_needed);
528

529
	/*
530 531 532 533 534 535 536
	 * 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.
537
	 */
538 539 540 541 542 543 544 545
	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);
546

547 548 549 550
		if (allocate_file_region_entries(
			    resv, actual_regions_needed - in_regions_needed)) {
			return -ENOMEM;
		}
551

552
		goto retry;
553 554
	}

555
	add = add_reservation_in_range(resv, f, t, h_cg, h, NULL);
556 557

	resv->adds_in_progress -= in_regions_needed;
558

559
	spin_unlock(&resv->lock);
560
	return add;
561 562
}

563 564 565 566 567 568 569
/*
 * 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
570 571 572 573 574 575 576
 * 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.
577 578 579 580 581
 *
 * 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.
582
 */
583 584
static long region_chg(struct resv_map *resv, long f, long t,
		       long *out_regions_needed)
585 586 587
{
	long chg = 0;

588
	spin_lock(&resv->lock);
589

590
	/* Count how many hugepages in this range are NOT represented. */
591
	chg = add_reservation_in_range(resv, f, t, NULL, NULL,
592
				       out_regions_needed);
593

594 595
	if (*out_regions_needed == 0)
		*out_regions_needed = 1;
596

597 598
	if (allocate_file_region_entries(resv, *out_regions_needed))
		return -ENOMEM;
599

600
	resv->adds_in_progress += *out_regions_needed;
601 602

	spin_unlock(&resv->lock);
603 604 605
	return chg;
}

606 607 608 609 610
/*
 * 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
611 612 613
 * 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.
614 615 616 617 618
 *
 * 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.
 */
619 620
static void region_abort(struct resv_map *resv, long f, long t,
			 long regions_needed)
621 622 623
{
	spin_lock(&resv->lock);
	VM_BUG_ON(!resv->region_cache_count);
624
	resv->adds_in_progress -= regions_needed;
625 626 627
	spin_unlock(&resv->lock);
}

628
/*
629 630 631 632 633 634 635 636 637 638 639 640
 * 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.
641
 */
642
static long region_del(struct resv_map *resv, long f, long t)
643
{
644
	struct list_head *head = &resv->regions;
645
	struct file_region *rg, *trg;
646 647
	struct file_region *nrg = NULL;
	long del = 0;
648

649
retry:
650
	spin_lock(&resv->lock);
651
	list_for_each_entry_safe(rg, trg, head, link) {
652 653 654 655 656 657 658 659
		/*
		 * 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))
660
			continue;
661

662
		if (rg->from >= t)
663 664
			break;

665 666 667 668 669 670 671 672 673 674 675 676 677
		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--;
			}
678

679 680 681 682 683 684 685 686 687
			if (!nrg) {
				spin_unlock(&resv->lock);
				nrg = kmalloc(sizeof(*nrg), GFP_KERNEL);
				if (!nrg)
					return -ENOMEM;
				goto retry;
			}

			del += t - f;
688
			hugetlb_cgroup_uncharge_file_region(
689
				resv, rg, t - f, false);
690 691 692 693

			/* New entry for end of split region */
			nrg->from = t;
			nrg->to = rg->to;
694 695 696

			copy_hugetlb_cgroup_uncharge_info(nrg, rg);

697 698 699 700 701 702 703
			INIT_LIST_HEAD(&nrg->link);

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

			list_add(&nrg->link, &rg->link);
			nrg = NULL;
704
			break;
705 706 707 708
		}

		if (f <= rg->from && t >= rg->to) { /* Remove entire region */
			del += rg->to - rg->from;
709
			hugetlb_cgroup_uncharge_file_region(resv, rg,
710
							    rg->to - rg->from, true);
711 712 713 714 715 716
			list_del(&rg->link);
			kfree(rg);
			continue;
		}

		if (f <= rg->from) {	/* Trim beginning of region */
717
			hugetlb_cgroup_uncharge_file_region(resv, rg,
718
							    t - rg->from, false);
719

720 721 722
			del += t - rg->from;
			rg->from = t;
		} else {		/* Trim end of region */
723
			hugetlb_cgroup_uncharge_file_region(resv, rg,
724
							    rg->to - f, false);
725 726 727

			del += rg->to - f;
			rg->to = f;
728
		}
729
	}
730 731

	spin_unlock(&resv->lock);
732 733
	kfree(nrg);
	return del;
734 735
}

736 737 738 739 740 741 742 743 744
/*
 * 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.
 */
745
void hugetlb_fix_reserve_counts(struct inode *inode)
746 747 748
{
	struct hugepage_subpool *spool = subpool_inode(inode);
	long rsv_adjust;
749
	bool reserved = false;
750 751

	rsv_adjust = hugepage_subpool_get_pages(spool, 1);
752
	if (rsv_adjust > 0) {
753 754
		struct hstate *h = hstate_inode(inode);

755 756 757 758
		if (!hugetlb_acct_memory(h, 1))
			reserved = true;
	} else if (!rsv_adjust) {
		reserved = true;
759
	}
760 761 762

	if (!reserved)
		pr_warn("hugetlb: Huge Page Reserved count may go negative.\n");
763 764
}

765 766 767 768
/*
 * Count and return the number of huge pages in the reserve map
 * that intersect with the range [f, t).
 */
769
static long region_count(struct resv_map *resv, long f, long t)
770
{
771
	struct list_head *head = &resv->regions;
772 773 774
	struct file_region *rg;
	long chg = 0;

775
	spin_lock(&resv->lock);
776 777
	/* Locate each segment we overlap with, and count that overlap. */
	list_for_each_entry(rg, head, link) {
778 779
		long seg_from;
		long seg_to;
780 781 782 783 784 785 786 787 788 789 790

		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;
	}
791
	spin_unlock(&resv->lock);
792 793 794 795

	return chg;
}

796 797 798 799
/*
 * Convert the address within this vma to the page offset within
 * the mapping, in pagecache page units; huge pages here.
 */
800 801
static pgoff_t vma_hugecache_offset(struct hstate *h,
			struct vm_area_struct *vma, unsigned long address)
802
{
803 804
	return ((address - vma->vm_start) >> huge_page_shift(h)) +
			(vma->vm_pgoff >> huge_page_order(h));
805 806
}

807 808 809 810 811
pgoff_t linear_hugepage_index(struct vm_area_struct *vma,
				     unsigned long address)
{
	return vma_hugecache_offset(hstate_vma(vma), vma, address);
}
812
EXPORT_SYMBOL_GPL(linear_hugepage_index);
813

814 815 816 817 818 819
/*
 * 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)
{
820 821 822
	if (vma->vm_ops && vma->vm_ops->pagesize)
		return vma->vm_ops->pagesize(vma);
	return PAGE_SIZE;
823
}
824
EXPORT_SYMBOL_GPL(vma_kernel_pagesize);
825

826 827 828
/*
 * 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
829 830
 * architectures where it differs, an architecture-specific 'strong'
 * version of this symbol is required.
831
 */
832
__weak unsigned long vma_mmu_pagesize(struct vm_area_struct *vma)
833 834 835 836
{
	return vma_kernel_pagesize(vma);
}

837 838 839 840 841 842 843
/*
 * 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)
844
#define HPAGE_RESV_MASK (HPAGE_RESV_OWNER | HPAGE_RESV_UNMAPPED)
845

846 847 848 849 850 851 852 853 854
/*
 * 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.
855 856 857 858 859 860 861 862 863
 *
 * 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.
864
 */
865 866 867 868 869 870 871 872 873 874 875
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;
}

876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894
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
}

895
struct resv_map *resv_map_alloc(void)
896 897
{
	struct resv_map *resv_map = kmalloc(sizeof(*resv_map), GFP_KERNEL);
898 899 900 901 902
	struct file_region *rg = kmalloc(sizeof(*rg), GFP_KERNEL);

	if (!resv_map || !rg) {
		kfree(resv_map);
		kfree(rg);
903
		return NULL;
904
	}
905 906

	kref_init(&resv_map->refs);
907
	spin_lock_init(&resv_map->lock);
908 909
	INIT_LIST_HEAD(&resv_map->regions);

910
	resv_map->adds_in_progress = 0;
911 912 913 914 915 916 917
	/*
	 * 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);
918 919 920 921 922

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

923 924 925
	return resv_map;
}

926
void resv_map_release(struct kref *ref)
927 928
{
	struct resv_map *resv_map = container_of(ref, struct resv_map, refs);
929 930
	struct list_head *head = &resv_map->region_cache;
	struct file_region *rg, *trg;
931 932

	/* Clear out any active regions before we release the map. */
933
	region_del(resv_map, 0, LONG_MAX);
934 935 936 937 938 939 940 941 942

	/* ... 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);

943 944 945
	kfree(resv_map);
}

946 947
static inline struct resv_map *inode_resv_map(struct inode *inode)
{
948 949 950 951 952 953 954 955 956
	/*
	 * 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;
957 958
}

959
static struct resv_map *vma_resv_map(struct vm_area_struct *vma)
960
{
961
	VM_BUG_ON_VMA(!is_vm_hugetlb_page(vma), vma);
962 963 964 965 966 967 968
	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 {
969 970
		return (struct resv_map *)(get_vma_private_data(vma) &
							~HPAGE_RESV_MASK);
971
	}
972 973
}

974
static void set_vma_resv_map(struct vm_area_struct *vma, struct resv_map *map)
975
{
976 977
	VM_BUG_ON_VMA(!is_vm_hugetlb_page(vma), vma);
	VM_BUG_ON_VMA(vma->vm_flags & VM_MAYSHARE, vma);
978

979 980
	set_vma_private_data(vma, (get_vma_private_data(vma) &
				HPAGE_RESV_MASK) | (unsigned long)map);
981 982 983 984
}

static void set_vma_resv_flags(struct vm_area_struct *vma, unsigned long flags)
{
985 986
	VM_BUG_ON_VMA(!is_vm_hugetlb_page(vma), vma);
	VM_BUG_ON_VMA(vma->vm_flags & VM_MAYSHARE, vma);
987 988

	set_vma_private_data(vma, get_vma_private_data(vma) | flags);
989 990 991 992
}

static int is_vma_resv_set(struct vm_area_struct *vma, unsigned long flag)
{
993
	VM_BUG_ON_VMA(!is_vm_hugetlb_page(vma), vma);
994 995

	return (get_vma_private_data(vma) & flag) != 0;
996 997
}

998
/* Reset counters to 0 and clear all HPAGE_RESV_* flags */
999 1000
void reset_vma_resv_huge_pages(struct vm_area_struct *vma)
{
1001
	VM_BUG_ON_VMA(!is_vm_hugetlb_page(vma), vma);
1002
	if (!(vma->vm_flags & VM_MAYSHARE))
1003 1004 1005 1006
		vma->vm_private_data = (void *)0;
}

/* Returns true if the VMA has associated reserve pages */
1007
static bool vma_has_reserves(struct vm_area_struct *vma, long chg)
1008
{
1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019
	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)
1020
			return true;
1021
		else
1022
			return false;
1023
	}
1024 1025

	/* Shared mappings always use reserves */
1026 1027 1028 1029 1030
	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
1031
		 * fallocate.  In this case, there really are no reserves to
1032 1033 1034 1035 1036 1037 1038
		 * use.  This situation is indicated if chg != 0.
		 */
		if (chg)
			return false;
		else
			return true;
	}
1039 1040 1041 1042 1043

	/*
	 * Only the process that called mmap() has reserves for
	 * private mappings.
	 */
1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064
	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;
	}
1065

1066
	return false;
1067 1068
}

1069
static void enqueue_huge_page(struct hstate *h, struct page *page)
L
Linus Torvalds 已提交
1070 1071
{
	int nid = page_to_nid(page);
1072 1073

	lockdep_assert_held(&hugetlb_lock);
1074
	list_move(&page->lru, &h->hugepage_freelists[nid]);
1075 1076
	h->free_huge_pages++;
	h->free_huge_pages_node[nid]++;
1077
	SetHPageFreed(page);
L
Linus Torvalds 已提交
1078 1079
}

1080
static struct page *dequeue_huge_page_node_exact(struct hstate *h, int nid)
1081 1082
{
	struct page *page;
1083 1084
	bool nocma = !!(current->flags & PF_MEMALLOC_NOCMA);

1085
	lockdep_assert_held(&hugetlb_lock);
1086 1087 1088
	list_for_each_entry(page, &h->hugepage_freelists[nid], lru) {
		if (nocma && is_migrate_cma_page(page))
			continue;
1089

1090 1091 1092 1093 1094
		if (PageHWPoison(page))
			continue;

		list_move(&page->lru, &h->hugepage_activelist);
		set_page_refcounted(page);
1095
		ClearHPageFreed(page);
1096 1097 1098
		h->free_huge_pages--;
		h->free_huge_pages_node[nid]--;
		return page;
1099 1100
	}

1101
	return NULL;
1102 1103
}

1104 1105
static struct page *dequeue_huge_page_nodemask(struct hstate *h, gfp_t gfp_mask, int nid,
		nodemask_t *nmask)
1106
{
1107 1108 1109 1110
	unsigned int cpuset_mems_cookie;
	struct zonelist *zonelist;
	struct zone *zone;
	struct zoneref *z;
1111
	int node = NUMA_NO_NODE;
1112

1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128
	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);
1129 1130 1131 1132 1133

		page = dequeue_huge_page_node_exact(h, node);
		if (page)
			return page;
	}
1134 1135 1136
	if (unlikely(read_mems_allowed_retry(cpuset_mems_cookie)))
		goto retry_cpuset;

1137 1138 1139
	return NULL;
}

1140 1141
static struct page *dequeue_huge_page_vma(struct hstate *h,
				struct vm_area_struct *vma,
1142 1143
				unsigned long address, int avoid_reserve,
				long chg)
L
Linus Torvalds 已提交
1144
{
1145
	struct page *page;
1146
	struct mempolicy *mpol;
1147
	gfp_t gfp_mask;
1148
	nodemask_t *nodemask;
1149
	int nid;
L
Linus Torvalds 已提交
1150

1151 1152 1153 1154 1155
	/*
	 * 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
	 */
1156
	if (!vma_has_reserves(vma, chg) &&
1157
			h->free_huge_pages - h->resv_huge_pages == 0)
1158
		goto err;
1159

1160
	/* If reserves cannot be used, ensure enough pages are in the pool */
1161
	if (avoid_reserve && h->free_huge_pages - h->resv_huge_pages == 0)
1162
		goto err;
1163

1164 1165
	gfp_mask = htlb_alloc_mask(h);
	nid = huge_node(vma, address, gfp_mask, &mpol, &nodemask);
1166 1167
	page = dequeue_huge_page_nodemask(h, gfp_mask, nid, nodemask);
	if (page && !avoid_reserve && vma_has_reserves(vma, chg)) {
1168
		SetHPageRestoreReserve(page);
1169
		h->resv_huge_pages--;
L
Linus Torvalds 已提交
1170
	}
1171

1172
	mpol_cond_put(mpol);
L
Linus Torvalds 已提交
1173
	return page;
1174 1175 1176

err:
	return NULL;
L
Linus Torvalds 已提交
1177 1178
}

1179 1180 1181 1182 1183 1184 1185 1186 1187
/*
 * 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)
{
1188
	nid = next_node_in(nid, *nodes_allowed);
1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220
	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;
}

/*
1221
 * helper for remove_pool_huge_page() - return the previously saved
1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249
 * 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--)

1250
#ifdef CONFIG_ARCH_HAS_GIGANTIC_PAGE
1251
static void destroy_compound_gigantic_page(struct page *page,
1252
					unsigned int order)
1253 1254 1255 1256 1257
{
	int i;
	int nr_pages = 1 << order;
	struct page *p = page + 1;

1258
	atomic_set(compound_mapcount_ptr(page), 0);
1259
	atomic_set(compound_pincount_ptr(page), 0);
1260

1261
	for (i = 1; i < nr_pages; i++, p = mem_map_next(p, page, i)) {
1262
		clear_compound_head(p);
1263 1264 1265 1266
		set_page_refcounted(p);
	}

	set_compound_order(page, 0);
1267
	page[1].compound_nr = 0;
1268 1269 1270
	__ClearPageHead(page);
}

1271
static void free_gigantic_page(struct page *page, unsigned int order)
1272
{
1273 1274 1275 1276
	/*
	 * If the page isn't allocated using the cma allocator,
	 * cma_release() returns false.
	 */
1277 1278
#ifdef CONFIG_CMA
	if (cma_release(hugetlb_cma[page_to_nid(page)], page, 1 << order))
1279
		return;
1280
#endif
1281

1282 1283 1284
	free_contig_range(page_to_pfn(page), 1 << order);
}

1285
#ifdef CONFIG_CONTIG_ALLOC
1286 1287
static struct page *alloc_gigantic_page(struct hstate *h, gfp_t gfp_mask,
		int nid, nodemask_t *nodemask)
1288
{
1289
	unsigned long nr_pages = pages_per_huge_page(h);
1290 1291
	if (nid == NUMA_NO_NODE)
		nid = numa_mem_id();
1292

1293 1294
#ifdef CONFIG_CMA
	{
1295 1296 1297
		struct page *page;
		int node;

1298 1299 1300
		if (hugetlb_cma[nid]) {
			page = cma_alloc(hugetlb_cma[nid], nr_pages,
					huge_page_order(h), true);
1301 1302 1303
			if (page)
				return page;
		}
1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315

		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;
			}
		}
1316
	}
1317
#endif
1318

1319
	return alloc_contig_pages(nr_pages, gfp_mask, nid, nodemask);
1320 1321 1322
}

static void prep_new_huge_page(struct hstate *h, struct page *page, int nid);
1323
static void prep_compound_gigantic_page(struct page *page, unsigned int order);
1324 1325 1326 1327 1328 1329 1330
#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 */
1331

1332
#else /* !CONFIG_ARCH_HAS_GIGANTIC_PAGE */
1333
static struct page *alloc_gigantic_page(struct hstate *h, gfp_t gfp_mask,
1334 1335 1336 1337
					int nid, nodemask_t *nodemask)
{
	return NULL;
}
1338
static inline void free_gigantic_page(struct page *page, unsigned int order) { }
1339
static inline void destroy_compound_gigantic_page(struct page *page,
1340
						unsigned int order) { }
1341 1342
#endif

1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356
/*
 * Remove hugetlb page from lists, and update dtor so that page appears
 * as just a compound page.  A reference is held on the page.
 *
 * Must be called with hugetlb lock held.
 */
static void remove_hugetlb_page(struct hstate *h, struct page *page,
							bool adjust_surplus)
{
	int nid = page_to_nid(page);

	VM_BUG_ON_PAGE(hugetlb_cgroup_from_page(page), page);
	VM_BUG_ON_PAGE(hugetlb_cgroup_from_page_rsvd(page), page);

1357
	lockdep_assert_held(&hugetlb_lock);
1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378
	if (hstate_is_gigantic(h) && !gigantic_page_runtime_supported())
		return;

	list_del(&page->lru);

	if (HPageFreed(page)) {
		h->free_huge_pages--;
		h->free_huge_pages_node[nid]--;
	}
	if (adjust_surplus) {
		h->surplus_huge_pages--;
		h->surplus_huge_pages_node[nid]--;
	}

	set_page_refcounted(page);
	set_compound_page_dtor(page, NULL_COMPOUND_DTOR);

	h->nr_huge_pages--;
	h->nr_huge_pages_node[nid]--;
}

1379
static void update_and_free_page(struct hstate *h, struct page *page)
A
Adam Litke 已提交
1380 1381
{
	int i;
1382
	struct page *subpage = page;
1383

1384
	if (hstate_is_gigantic(h) && !gigantic_page_runtime_supported())
1385
		return;
1386

1387 1388 1389
	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 |
1390
				1 << PG_referenced | 1 << PG_dirty |
1391 1392
				1 << PG_active | 1 << PG_private |
				1 << PG_writeback);
A
Adam Litke 已提交
1393
	}
1394 1395 1396 1397 1398 1399
	if (hstate_is_gigantic(h)) {
		destroy_compound_gigantic_page(page, huge_page_order(h));
		free_gigantic_page(page, huge_page_order(h));
	} else {
		__free_pages(page, huge_page_order(h));
	}
A
Adam Litke 已提交
1400 1401
}

1402 1403 1404 1405 1406 1407 1408 1409 1410 1411
static void update_and_free_pages_bulk(struct hstate *h, struct list_head *list)
{
	struct page *page, *t_page;

	list_for_each_entry_safe(page, t_page, list, lru) {
		update_and_free_page(h, page);
		cond_resched();
	}
}

1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422
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;
}

1423
void free_huge_page(struct page *page)
1424
{
1425 1426 1427 1428
	/*
	 * Can't pass hstate in here because it is called from the
	 * compound page destructor.
	 */
1429
	struct hstate *h = page_hstate(page);
1430
	int nid = page_to_nid(page);
1431
	struct hugepage_subpool *spool = hugetlb_page_subpool(page);
1432
	bool restore_reserve;
1433
	unsigned long flags;
1434

1435 1436
	VM_BUG_ON_PAGE(page_count(page), page);
	VM_BUG_ON_PAGE(page_mapcount(page), page);
1437

1438
	hugetlb_set_page_subpool(page, NULL);
1439
	page->mapping = NULL;
1440 1441
	restore_reserve = HPageRestoreReserve(page);
	ClearHPageRestoreReserve(page);
1442

1443
	/*
1444
	 * If HPageRestoreReserve was set on page, page allocation consumed a
1445 1446 1447
	 * 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
1448
	 * reservation, do not call hugepage_subpool_put_pages() as this will
1449
	 * remove the reserved page from the subpool.
1450
	 */
1451 1452 1453 1454 1455 1456 1457 1458 1459 1460
	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;
	}
1461

1462
	spin_lock_irqsave(&hugetlb_lock, flags);
1463
	ClearHPageMigratable(page);
1464 1465
	hugetlb_cgroup_uncharge_page(hstate_index(h),
				     pages_per_huge_page(h), page);
1466 1467
	hugetlb_cgroup_uncharge_page_rsvd(hstate_index(h),
					  pages_per_huge_page(h), page);
1468 1469 1470
	if (restore_reserve)
		h->resv_huge_pages++;

1471
	if (HPageTemporary(page)) {
1472
		remove_hugetlb_page(h, page, false);
1473
		spin_unlock_irqrestore(&hugetlb_lock, flags);
1474 1475
		update_and_free_page(h, page);
	} else if (h->surplus_huge_pages_node[nid]) {
1476
		/* remove the page from active list */
1477
		remove_hugetlb_page(h, page, true);
1478
		spin_unlock_irqrestore(&hugetlb_lock, flags);
1479
		update_and_free_page(h, page);
1480
	} else {
1481
		arch_clear_hugepage_flags(page);
1482
		enqueue_huge_page(h, page);
1483
		spin_unlock_irqrestore(&hugetlb_lock, flags);
1484 1485 1486
	}
}

1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497
/*
 * Must be called with the hugetlb lock held
 */
static void __prep_account_new_huge_page(struct hstate *h, int nid)
{
	lockdep_assert_held(&hugetlb_lock);
	h->nr_huge_pages++;
	h->nr_huge_pages_node[nid]++;
}

static void __prep_new_huge_page(struct page *page)
1498
{
1499
	INIT_LIST_HEAD(&page->lru);
1500
	set_compound_page_dtor(page, HUGETLB_PAGE_DTOR);
1501
	hugetlb_set_page_subpool(page, NULL);
1502
	set_hugetlb_cgroup(page, NULL);
1503
	set_hugetlb_cgroup_rsvd(page, NULL);
1504 1505 1506 1507 1508
}

static void prep_new_huge_page(struct hstate *h, struct page *page, int nid)
{
	__prep_new_huge_page(page);
1509
	spin_lock_irq(&hugetlb_lock);
1510
	__prep_account_new_huge_page(h, nid);
1511
	spin_unlock_irq(&hugetlb_lock);
1512 1513
}

1514
static void prep_compound_gigantic_page(struct page *page, unsigned int order)
1515 1516 1517 1518 1519 1520 1521
{
	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);
1522
	__ClearPageReserved(page);
1523
	__SetPageHead(page);
1524
	for (i = 1; i < nr_pages; i++, p = mem_map_next(p, page, i)) {
1525 1526 1527 1528
		/*
		 * 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
1529
		 * too.  Otherwise drivers using get_user_pages() to access tail
1530 1531 1532 1533 1534 1535 1536 1537
		 * 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);
1538
		set_page_count(p, 0);
1539
		set_compound_head(p, page);
1540
	}
1541
	atomic_set(compound_mapcount_ptr(page), -1);
1542
	atomic_set(compound_pincount_ptr(page), 0);
1543 1544
}

A
Andrew Morton 已提交
1545 1546 1547 1548 1549
/*
 * PageHuge() only returns true for hugetlbfs pages, but not for normal or
 * transparent huge pages.  See the PageTransHuge() documentation for more
 * details.
 */
1550 1551 1552 1553 1554 1555
int PageHuge(struct page *page)
{
	if (!PageCompound(page))
		return 0;

	page = compound_head(page);
1556
	return page[1].compound_dtor == HUGETLB_PAGE_DTOR;
1557
}
1558 1559
EXPORT_SYMBOL_GPL(PageHuge);

1560 1561 1562 1563 1564 1565 1566 1567 1568
/*
 * 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;

1569
	return page_head[1].compound_dtor == HUGETLB_PAGE_DTOR;
1570 1571
}

1572 1573 1574
/*
 * Find and lock address space (mapping) in write mode.
 *
1575 1576 1577
 * 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.
1578 1579 1580
 */
struct address_space *hugetlb_page_mapping_lock_write(struct page *hpage)
{
1581
	struct address_space *mapping = page_mapping(hpage);
1582 1583 1584 1585 1586 1587 1588

	if (!mapping)
		return mapping;

	if (i_mmap_trylock_write(mapping))
		return mapping;

1589
	return NULL;
1590 1591
}

1592 1593 1594 1595 1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608
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;
}

1609
static struct page *alloc_buddy_huge_page(struct hstate *h,
1610 1611
		gfp_t gfp_mask, int nid, nodemask_t *nmask,
		nodemask_t *node_alloc_noretry)
L
Linus Torvalds 已提交
1612
{
1613
	int order = huge_page_order(h);
L
Linus Torvalds 已提交
1614
	struct page *page;
1615
	bool alloc_try_hard = true;
1616

1617 1618 1619 1620 1621 1622 1623 1624 1625 1626 1627 1628
	/*
	 * 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;
1629 1630
	if (nid == NUMA_NO_NODE)
		nid = numa_mem_id();
1631
	page = __alloc_pages(gfp_mask, order, nid, nmask);
1632 1633 1634 1635
	if (page)
		__count_vm_event(HTLB_BUDDY_PGALLOC);
	else
		__count_vm_event(HTLB_BUDDY_PGALLOC_FAIL);
1636

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

1653 1654 1655
	return page;
}

1656 1657 1658 1659 1660
/*
 * 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,
1661 1662
		gfp_t gfp_mask, int nid, nodemask_t *nmask,
		nodemask_t *node_alloc_noretry)
1663 1664 1665 1666 1667 1668 1669
{
	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,
1670
				nid, nmask, node_alloc_noretry);
1671 1672 1673 1674 1675 1676 1677 1678 1679 1680
	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;
}

1681 1682 1683 1684
/*
 * Allocates a fresh page to the hugetlb allocator pool in the node interleaved
 * manner.
 */
1685 1686
static int alloc_pool_huge_page(struct hstate *h, nodemask_t *nodes_allowed,
				nodemask_t *node_alloc_noretry)
1687 1688 1689
{
	struct page *page;
	int nr_nodes, node;
1690
	gfp_t gfp_mask = htlb_alloc_mask(h) | __GFP_THISNODE;
1691 1692

	for_each_node_mask_to_alloc(h, nr_nodes, node, nodes_allowed) {
1693 1694
		page = alloc_fresh_huge_page(h, gfp_mask, node, nodes_allowed,
						node_alloc_noretry);
1695
		if (page)
1696 1697 1698
			break;
	}

1699 1700
	if (!page)
		return 0;
1701

1702 1703 1704
	put_page(page); /* free it into the hugepage allocator */

	return 1;
1705 1706
}

1707
/*
1708 1709 1710 1711
 * Remove huge page from pool from next node to free.  Attempt to keep
 * persistent huge pages more or less balanced over allowed nodes.
 * This routine only 'removes' the hugetlb page.  The caller must make
 * an additional call to free the page to low level allocators.
1712 1713
 * Called with hugetlb_lock locked.
 */
1714 1715 1716
static struct page *remove_pool_huge_page(struct hstate *h,
						nodemask_t *nodes_allowed,
						 bool acct_surplus)
1717
{
1718
	int nr_nodes, node;
1719
	struct page *page = NULL;
1720

1721
	lockdep_assert_held(&hugetlb_lock);
1722
	for_each_node_mask_to_free(h, nr_nodes, node, nodes_allowed) {
1723 1724 1725 1726
		/*
		 * If we're returning unused surplus pages, only examine
		 * nodes with surplus pages.
		 */
1727 1728
		if ((!acct_surplus || h->surplus_huge_pages_node[node]) &&
		    !list_empty(&h->hugepage_freelists[node])) {
1729
			page = list_entry(h->hugepage_freelists[node].next,
1730
					  struct page, lru);
1731
			remove_hugetlb_page(h, page, acct_surplus);
1732
			break;
1733
		}
1734
	}
1735

1736
	return page;
1737 1738
}

1739 1740
/*
 * Dissolve a given free hugepage into free buddy pages. This function does
1741 1742 1743 1744 1745 1746 1747
 * 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)
1748
 */
1749
int dissolve_free_huge_page(struct page *page)
1750
{
1751
	int rc = -EBUSY;
1752

1753
retry:
1754 1755 1756 1757
	/* Not to disrupt normal path by vainly holding hugetlb_lock */
	if (!PageHuge(page))
		return 0;

1758
	spin_lock_irq(&hugetlb_lock);
1759 1760 1761 1762 1763 1764
	if (!PageHuge(page)) {
		rc = 0;
		goto out;
	}

	if (!page_count(page)) {
1765 1766
		struct page *head = compound_head(page);
		struct hstate *h = page_hstate(head);
1767
		if (h->free_huge_pages - h->resv_huge_pages == 0)
1768
			goto out;
1769 1770 1771 1772 1773

		/*
		 * We should make sure that the page is already on the free list
		 * when it is dissolved.
		 */
1774
		if (unlikely(!HPageFreed(head))) {
1775
			spin_unlock_irq(&hugetlb_lock);
1776 1777 1778 1779 1780 1781 1782 1783 1784 1785 1786 1787 1788
			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;
		}

1789 1790 1791 1792 1793 1794 1795 1796
		/*
		 * 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);
		}
1797
		remove_hugetlb_page(h, page, false);
1798
		h->max_huge_pages--;
1799
		spin_unlock_irq(&hugetlb_lock);
1800
		update_and_free_page(h, head);
1801
		return 0;
1802
	}
1803
out:
1804
	spin_unlock_irq(&hugetlb_lock);
1805
	return rc;
1806 1807 1808 1809 1810
}

/*
 * Dissolve free hugepages in a given pfn range. Used by memory hotplug to
 * make specified memory blocks removable from the system.
1811 1812
 * Note that this will dissolve a free gigantic hugepage completely, if any
 * part of it lies within the given range.
1813 1814
 * Also note that if dissolve_free_huge_page() returns with an error, all
 * free hugepages that were dissolved before that error are lost.
1815
 */
1816
int dissolve_free_huge_pages(unsigned long start_pfn, unsigned long end_pfn)
1817 1818
{
	unsigned long pfn;
1819
	struct page *page;
1820
	int rc = 0;
1821

1822
	if (!hugepages_supported())
1823
		return rc;
1824

1825 1826
	for (pfn = start_pfn; pfn < end_pfn; pfn += 1 << minimum_order) {
		page = pfn_to_page(pfn);
1827 1828 1829
		rc = dissolve_free_huge_page(page);
		if (rc)
			break;
1830
	}
1831 1832

	return rc;
1833 1834
}

1835 1836 1837
/*
 * Allocates a fresh surplus page from the page allocator.
 */
1838
static struct page *alloc_surplus_huge_page(struct hstate *h, gfp_t gfp_mask,
1839
		int nid, nodemask_t *nmask)
1840
{
1841
	struct page *page = NULL;
1842

1843
	if (hstate_is_gigantic(h))
1844 1845
		return NULL;

1846
	spin_lock_irq(&hugetlb_lock);
1847 1848
	if (h->surplus_huge_pages >= h->nr_overcommit_huge_pages)
		goto out_unlock;
1849
	spin_unlock_irq(&hugetlb_lock);
1850

1851
	page = alloc_fresh_huge_page(h, gfp_mask, nid, nmask, NULL);
1852
	if (!page)
1853
		return NULL;
1854

1855
	spin_lock_irq(&hugetlb_lock);
1856 1857 1858 1859 1860 1861 1862 1863
	/*
	 * 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) {
1864
		SetHPageTemporary(page);
1865
		spin_unlock_irq(&hugetlb_lock);
1866
		put_page(page);
1867
		return NULL;
1868 1869
	} else {
		h->surplus_huge_pages++;
1870
		h->surplus_huge_pages_node[page_to_nid(page)]++;
1871
	}
1872 1873

out_unlock:
1874
	spin_unlock_irq(&hugetlb_lock);
1875 1876 1877 1878

	return page;
}

1879
static struct page *alloc_migrate_huge_page(struct hstate *h, gfp_t gfp_mask,
1880
				     int nid, nodemask_t *nmask)
1881 1882 1883 1884 1885 1886
{
	struct page *page;

	if (hstate_is_gigantic(h))
		return NULL;

1887
	page = alloc_fresh_huge_page(h, gfp_mask, nid, nmask, NULL);
1888 1889 1890 1891 1892 1893 1894
	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
	 */
1895
	SetHPageTemporary(page);
1896 1897 1898 1899

	return page;
}

1900 1901 1902
/*
 * Use the VMA's mpolicy to allocate a huge page from the buddy.
 */
D
Dave Hansen 已提交
1903
static
1904
struct page *alloc_buddy_huge_page_with_mpol(struct hstate *h,
1905 1906
		struct vm_area_struct *vma, unsigned long addr)
{
1907 1908 1909 1910 1911 1912 1913
	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);
1914
	page = alloc_surplus_huge_page(h, gfp_mask, nid, nodemask);
1915 1916 1917
	mpol_cond_put(mpol);

	return page;
1918 1919
}

1920
/* page migration callback function */
1921
struct page *alloc_huge_page_nodemask(struct hstate *h, int preferred_nid,
1922
		nodemask_t *nmask, gfp_t gfp_mask)
1923
{
1924
	spin_lock_irq(&hugetlb_lock);
1925
	if (h->free_huge_pages - h->resv_huge_pages > 0) {
1926 1927 1928 1929
		struct page *page;

		page = dequeue_huge_page_nodemask(h, gfp_mask, preferred_nid, nmask);
		if (page) {
1930
			spin_unlock_irq(&hugetlb_lock);
1931
			return page;
1932 1933
		}
	}
1934
	spin_unlock_irq(&hugetlb_lock);
1935

1936
	return alloc_migrate_huge_page(h, gfp_mask, preferred_nid, nmask);
1937 1938
}

1939
/* mempolicy aware migration callback */
1940 1941
struct page *alloc_huge_page_vma(struct hstate *h, struct vm_area_struct *vma,
		unsigned long address)
1942 1943 1944 1945 1946 1947 1948 1949 1950
{
	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);
1951
	page = alloc_huge_page_nodemask(h, node, nodemask, gfp_mask);
1952 1953 1954 1955 1956
	mpol_cond_put(mpol);

	return page;
}

1957
/*
L
Lucas De Marchi 已提交
1958
 * Increase the hugetlb pool such that it can accommodate a reservation
1959 1960
 * of size 'delta'.
 */
1961
static int gather_surplus_pages(struct hstate *h, long delta)
1962
	__must_hold(&hugetlb_lock)
1963 1964 1965
{
	struct list_head surplus_list;
	struct page *page, *tmp;
1966 1967 1968
	int ret;
	long i;
	long needed, allocated;
1969
	bool alloc_ok = true;
1970

1971
	lockdep_assert_held(&hugetlb_lock);
1972
	needed = (h->resv_huge_pages + delta) - h->free_huge_pages;
1973
	if (needed <= 0) {
1974
		h->resv_huge_pages += delta;
1975
		return 0;
1976
	}
1977 1978 1979 1980 1981 1982

	allocated = 0;
	INIT_LIST_HEAD(&surplus_list);

	ret = -ENOMEM;
retry:
1983
	spin_unlock_irq(&hugetlb_lock);
1984
	for (i = 0; i < needed; i++) {
1985
		page = alloc_surplus_huge_page(h, htlb_alloc_mask(h),
1986
				NUMA_NO_NODE, NULL);
1987 1988 1989 1990
		if (!page) {
			alloc_ok = false;
			break;
		}
1991
		list_add(&page->lru, &surplus_list);
1992
		cond_resched();
1993
	}
1994
	allocated += i;
1995 1996 1997 1998 1999

	/*
	 * After retaking hugetlb_lock, we need to recalculate 'needed'
	 * because either resv_huge_pages or free_huge_pages may have changed.
	 */
2000
	spin_lock_irq(&hugetlb_lock);
2001 2002
	needed = (h->resv_huge_pages + delta) -
			(h->free_huge_pages + allocated);
2003 2004 2005 2006 2007 2008 2009 2010 2011 2012
	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;
	}
2013 2014
	/*
	 * The surplus_list now contains _at_least_ the number of extra pages
L
Lucas De Marchi 已提交
2015
	 * needed to accommodate the reservation.  Add the appropriate number
2016
	 * of pages to the hugetlb pool and free the extras back to the buddy
2017 2018 2019
	 * 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.
2020 2021
	 */
	needed += allocated;
2022
	h->resv_huge_pages += delta;
2023
	ret = 0;
2024

2025
	/* Free the needed pages to the hugetlb pool */
2026
	list_for_each_entry_safe(page, tmp, &surplus_list, lru) {
2027 2028
		int zeroed;

2029 2030
		if ((--needed) < 0)
			break;
2031 2032 2033 2034
		/*
		 * This page is now managed by the hugetlb allocator and has
		 * no users -- drop the buddy allocator's reference.
		 */
2035 2036
		zeroed = put_page_testzero(page);
		VM_BUG_ON_PAGE(!zeroed, page);
2037
		enqueue_huge_page(h, page);
2038
	}
2039
free:
2040
	spin_unlock_irq(&hugetlb_lock);
2041 2042

	/* Free unnecessary surplus pages to the buddy allocator */
2043 2044
	list_for_each_entry_safe(page, tmp, &surplus_list, lru)
		put_page(page);
2045
	spin_lock_irq(&hugetlb_lock);
2046 2047 2048 2049 2050

	return ret;
}

/*
2051 2052 2053 2054 2055 2056
 * 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.
2057
 */
2058 2059
static void return_unused_surplus_pages(struct hstate *h,
					unsigned long unused_resv_pages)
2060 2061
{
	unsigned long nr_pages;
2062 2063 2064
	struct page *page;
	LIST_HEAD(page_list);

2065
	lockdep_assert_held(&hugetlb_lock);
2066 2067
	/* Uncommit the reservation */
	h->resv_huge_pages -= unused_resv_pages;
2068

2069
	/* Cannot return gigantic pages currently */
2070
	if (hstate_is_gigantic(h))
2071
		goto out;
2072

2073 2074 2075 2076
	/*
	 * Part (or even all) of the reservation could have been backed
	 * by pre-allocated pages. Only free surplus pages.
	 */
2077
	nr_pages = min(unused_resv_pages, h->surplus_huge_pages);
2078

2079 2080
	/*
	 * We want to release as many surplus pages as possible, spread
2081 2082 2083
	 * 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.
2084
	 * remove_pool_huge_page() will balance the freed pages across the
2085
	 * on-line nodes with memory and will handle the hstate accounting.
2086 2087
	 */
	while (nr_pages--) {
2088 2089
		page = remove_pool_huge_page(h, &node_states[N_MEMORY], 1);
		if (!page)
2090
			goto out;
2091 2092

		list_add(&page->lru, &page_list);
2093
	}
2094 2095

out:
2096
	spin_unlock_irq(&hugetlb_lock);
2097
	update_and_free_pages_bulk(h, &page_list);
2098
	spin_lock_irq(&hugetlb_lock);
2099 2100
}

2101

2102
/*
2103
 * vma_needs_reservation, vma_commit_reservation and vma_end_reservation
2104
 * are used by the huge page allocation routines to manage reservations.
2105 2106 2107 2108 2109 2110
 *
 * 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
2111 2112 2113
 * 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.
2114 2115 2116 2117 2118 2119
 *
 * 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.
2120 2121 2122 2123 2124
 *
 * 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.
2125
 */
2126 2127 2128
enum vma_resv_mode {
	VMA_NEEDS_RESV,
	VMA_COMMIT_RESV,
2129
	VMA_END_RESV,
2130
	VMA_ADD_RESV,
2131
};
2132 2133
static long __vma_reservation_common(struct hstate *h,
				struct vm_area_struct *vma, unsigned long addr,
2134
				enum vma_resv_mode mode)
2135
{
2136 2137
	struct resv_map *resv;
	pgoff_t idx;
2138
	long ret;
2139
	long dummy_out_regions_needed;
2140

2141 2142
	resv = vma_resv_map(vma);
	if (!resv)
2143
		return 1;
2144

2145
	idx = vma_hugecache_offset(h, vma, addr);
2146 2147
	switch (mode) {
	case VMA_NEEDS_RESV:
2148 2149 2150 2151 2152 2153
		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);
2154 2155
		break;
	case VMA_COMMIT_RESV:
2156
		ret = region_add(resv, idx, idx + 1, 1, NULL, NULL);
2157 2158
		/* region_add calls of range 1 should never fail. */
		VM_BUG_ON(ret < 0);
2159
		break;
2160
	case VMA_END_RESV:
2161
		region_abort(resv, idx, idx + 1, 1);
2162 2163
		ret = 0;
		break;
2164
	case VMA_ADD_RESV:
2165
		if (vma->vm_flags & VM_MAYSHARE) {
2166
			ret = region_add(resv, idx, idx + 1, 1, NULL, NULL);
2167 2168 2169 2170
			/* region_add calls of range 1 should never fail. */
			VM_BUG_ON(ret < 0);
		} else {
			region_abort(resv, idx, idx + 1, 1);
2171 2172 2173
			ret = region_del(resv, idx, idx + 1);
		}
		break;
2174 2175 2176
	default:
		BUG();
	}
2177

2178
	if (vma->vm_flags & VM_MAYSHARE)
2179
		return ret;
2180 2181 2182 2183 2184 2185 2186 2187 2188 2189 2190 2191 2192 2193 2194 2195 2196 2197 2198 2199
	/*
	 * We know private mapping must have HPAGE_RESV_OWNER set.
	 *
	 * 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 > 0)
		return 0;
	if (ret == 0)
		return 1;
	return ret;
2200
}
2201 2202

static long vma_needs_reservation(struct hstate *h,
2203
			struct vm_area_struct *vma, unsigned long addr)
2204
{
2205
	return __vma_reservation_common(h, vma, addr, VMA_NEEDS_RESV);
2206
}
2207

2208 2209 2210
static long vma_commit_reservation(struct hstate *h,
			struct vm_area_struct *vma, unsigned long addr)
{
2211 2212 2213
	return __vma_reservation_common(h, vma, addr, VMA_COMMIT_RESV);
}

2214
static void vma_end_reservation(struct hstate *h,
2215 2216
			struct vm_area_struct *vma, unsigned long addr)
{
2217
	(void)__vma_reservation_common(h, vma, addr, VMA_END_RESV);
2218 2219
}

2220 2221 2222 2223 2224 2225 2226 2227 2228 2229
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,
2230 2231 2232 2233 2234 2235
 * 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.
2236 2237 2238 2239 2240
 */
static void restore_reserve_on_error(struct hstate *h,
			struct vm_area_struct *vma, unsigned long address,
			struct page *page)
{
2241
	if (unlikely(HPageRestoreReserve(page))) {
2242 2243 2244 2245 2246
		long rc = vma_needs_reservation(h, vma, address);

		if (unlikely(rc < 0)) {
			/*
			 * Rare out of memory condition in reserve map
2247
			 * manipulation.  Clear HPageRestoreReserve so that
2248 2249 2250 2251 2252 2253 2254 2255
			 * 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.
			 */
2256
			ClearHPageRestoreReserve(page);
2257 2258 2259 2260 2261 2262 2263
		} else if (rc) {
			rc = vma_add_reservation(h, vma, address);
			if (unlikely(rc < 0))
				/*
				 * See above comment about rare out of
				 * memory condition.
				 */
2264
				ClearHPageRestoreReserve(page);
2265 2266 2267 2268 2269
		} else
			vma_end_reservation(h, vma, address);
	}
}

2270 2271 2272 2273
/*
 * alloc_and_dissolve_huge_page - Allocate a new page and dissolve the old one
 * @h: struct hstate old page belongs to
 * @old_page: Old page to dissolve
2274
 * @list: List to isolate the page in case we need to
2275 2276
 * Returns 0 on success, otherwise negated error.
 */
2277 2278
static int alloc_and_dissolve_huge_page(struct hstate *h, struct page *old_page,
					struct list_head *list)
2279 2280 2281 2282 2283 2284 2285 2286 2287 2288 2289 2290 2291 2292 2293 2294 2295 2296 2297 2298 2299 2300 2301 2302 2303 2304
{
	gfp_t gfp_mask = htlb_alloc_mask(h) | __GFP_THISNODE;
	int nid = page_to_nid(old_page);
	struct page *new_page;
	int ret = 0;

	/*
	 * Before dissolving the page, we need to allocate a new one for the
	 * pool to remain stable. Using alloc_buddy_huge_page() allows us to
	 * not having to deal with prep_new_huge_page() and avoids dealing of any
	 * counters. This simplifies and let us do the whole thing under the
	 * lock.
	 */
	new_page = alloc_buddy_huge_page(h, gfp_mask, nid, NULL, NULL);
	if (!new_page)
		return -ENOMEM;

retry:
	spin_lock_irq(&hugetlb_lock);
	if (!PageHuge(old_page)) {
		/*
		 * Freed from under us. Drop new_page too.
		 */
		goto free_new;
	} else if (page_count(old_page)) {
		/*
2305 2306
		 * Someone has grabbed the page, try to isolate it here.
		 * Fail with -EBUSY if not possible.
2307
		 */
2308 2309 2310 2311
		spin_unlock_irq(&hugetlb_lock);
		if (!isolate_huge_page(old_page, list))
			ret = -EBUSY;
		spin_lock_irq(&hugetlb_lock);
2312 2313 2314 2315 2316 2317 2318 2319 2320 2321 2322 2323 2324 2325 2326 2327 2328 2329 2330 2331 2332 2333 2334 2335 2336 2337 2338 2339 2340 2341 2342 2343 2344 2345 2346 2347 2348 2349 2350 2351 2352 2353 2354 2355 2356 2357 2358 2359 2360
		goto free_new;
	} else if (!HPageFreed(old_page)) {
		/*
		 * Page's refcount is 0 but it has not been enqueued in the
		 * freelist yet. Race window is small, so we can succeed here if
		 * we retry.
		 */
		spin_unlock_irq(&hugetlb_lock);
		cond_resched();
		goto retry;
	} else {
		/*
		 * Ok, old_page is still a genuine free hugepage. Remove it from
		 * the freelist and decrease the counters. These will be
		 * incremented again when calling __prep_account_new_huge_page()
		 * and enqueue_huge_page() for new_page. The counters will remain
		 * stable since this happens under the lock.
		 */
		remove_hugetlb_page(h, old_page, false);

		/*
		 * new_page needs to be initialized with the standard hugetlb
		 * state. This is normally done by prep_new_huge_page() but
		 * that takes hugetlb_lock which is already held so we need to
		 * open code it here.
		 * Reference count trick is needed because allocator gives us
		 * referenced page but the pool requires pages with 0 refcount.
		 */
		__prep_new_huge_page(new_page);
		__prep_account_new_huge_page(h, nid);
		page_ref_dec(new_page);
		enqueue_huge_page(h, new_page);

		/*
		 * Pages have been replaced, we can safely free the old one.
		 */
		spin_unlock_irq(&hugetlb_lock);
		update_and_free_page(h, old_page);
	}

	return ret;

free_new:
	spin_unlock_irq(&hugetlb_lock);
	__free_pages(new_page, huge_page_order(h));

	return ret;
}

2361
int isolate_or_dissolve_huge_page(struct page *page, struct list_head *list)
2362 2363 2364
{
	struct hstate *h;
	struct page *head;
2365
	int ret = -EBUSY;
2366 2367 2368 2369 2370 2371 2372 2373 2374 2375 2376 2377 2378 2379 2380 2381 2382 2383 2384 2385 2386 2387 2388 2389

	/*
	 * The page might have been dissolved from under our feet, so make sure
	 * to carefully check the state under the lock.
	 * Return success when racing as if we dissolved the page ourselves.
	 */
	spin_lock_irq(&hugetlb_lock);
	if (PageHuge(page)) {
		head = compound_head(page);
		h = page_hstate(head);
	} else {
		spin_unlock_irq(&hugetlb_lock);
		return 0;
	}
	spin_unlock_irq(&hugetlb_lock);

	/*
	 * Fence off gigantic pages as there is a cyclic dependency between
	 * alloc_contig_range and them. Return -ENOMEM as this has the effect
	 * of bailing out right away without further retrying.
	 */
	if (hstate_is_gigantic(h))
		return -ENOMEM;

2390 2391 2392 2393 2394 2395
	if (page_count(head) && isolate_huge_page(head, list))
		ret = 0;
	else if (!page_count(head))
		ret = alloc_and_dissolve_huge_page(h, head, list);

	return ret;
2396 2397
}

2398
struct page *alloc_huge_page(struct vm_area_struct *vma,
2399
				    unsigned long addr, int avoid_reserve)
L
Linus Torvalds 已提交
2400
{
2401
	struct hugepage_subpool *spool = subpool_vma(vma);
2402
	struct hstate *h = hstate_vma(vma);
2403
	struct page *page;
2404 2405
	long map_chg, map_commit;
	long gbl_chg;
2406 2407
	int ret, idx;
	struct hugetlb_cgroup *h_cg;
2408
	bool deferred_reserve;
2409

2410
	idx = hstate_index(h);
2411
	/*
2412 2413 2414
	 * 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).
2415
	 */
2416 2417
	map_chg = gbl_chg = vma_needs_reservation(h, vma, addr);
	if (map_chg < 0)
2418
		return ERR_PTR(-ENOMEM);
2419 2420 2421 2422 2423 2424 2425 2426 2427 2428 2429

	/*
	 * 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) {
2430
			vma_end_reservation(h, vma, addr);
2431
			return ERR_PTR(-ENOSPC);
2432
		}
L
Linus Torvalds 已提交
2433

2434 2435 2436 2437 2438 2439 2440 2441 2442 2443 2444 2445
		/*
		 * 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;
	}

2446 2447
	/* If this allocation is not consuming a reservation, charge it now.
	 */
2448
	deferred_reserve = map_chg || avoid_reserve;
2449 2450 2451 2452 2453 2454 2455
	if (deferred_reserve) {
		ret = hugetlb_cgroup_charge_cgroup_rsvd(
			idx, pages_per_huge_page(h), &h_cg);
		if (ret)
			goto out_subpool_put;
	}

2456
	ret = hugetlb_cgroup_charge_cgroup(idx, pages_per_huge_page(h), &h_cg);
2457
	if (ret)
2458
		goto out_uncharge_cgroup_reservation;
2459

2460
	spin_lock_irq(&hugetlb_lock);
2461 2462 2463 2464 2465 2466
	/*
	 * 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);
2467
	if (!page) {
2468
		spin_unlock_irq(&hugetlb_lock);
2469
		page = alloc_buddy_huge_page_with_mpol(h, vma, addr);
2470 2471
		if (!page)
			goto out_uncharge_cgroup;
2472
		if (!avoid_reserve && vma_has_reserves(vma, gbl_chg)) {
2473
			SetHPageRestoreReserve(page);
2474 2475
			h->resv_huge_pages--;
		}
2476
		spin_lock_irq(&hugetlb_lock);
2477
		list_add(&page->lru, &h->hugepage_activelist);
2478
		/* Fall through */
K
Ken Chen 已提交
2479
	}
2480
	hugetlb_cgroup_commit_charge(idx, pages_per_huge_page(h), h_cg, page);
2481 2482 2483 2484 2485 2486 2487 2488
	/* 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);
	}

2489
	spin_unlock_irq(&hugetlb_lock);
2490

2491
	hugetlb_set_page_subpool(page, spool);
2492

2493 2494
	map_commit = vma_commit_reservation(h, vma, addr);
	if (unlikely(map_chg > map_commit)) {
2495 2496 2497 2498 2499 2500 2501 2502 2503 2504 2505 2506 2507
		/*
		 * 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);
2508 2509 2510
		if (deferred_reserve)
			hugetlb_cgroup_uncharge_page_rsvd(hstate_index(h),
					pages_per_huge_page(h), page);
2511
	}
2512
	return page;
2513 2514 2515

out_uncharge_cgroup:
	hugetlb_cgroup_uncharge_cgroup(idx, pages_per_huge_page(h), h_cg);
2516 2517 2518 2519
out_uncharge_cgroup_reservation:
	if (deferred_reserve)
		hugetlb_cgroup_uncharge_cgroup_rsvd(idx, pages_per_huge_page(h),
						    h_cg);
2520
out_subpool_put:
2521
	if (map_chg || avoid_reserve)
2522
		hugepage_subpool_put_pages(spool, 1);
2523
	vma_end_reservation(h, vma, addr);
2524
	return ERR_PTR(-ENOSPC);
2525 2526
}

2527 2528 2529
int alloc_bootmem_huge_page(struct hstate *h)
	__attribute__ ((weak, alias("__alloc_bootmem_huge_page")));
int __alloc_bootmem_huge_page(struct hstate *h)
2530 2531
{
	struct huge_bootmem_page *m;
2532
	int nr_nodes, node;
2533

2534
	for_each_node_mask_to_alloc(h, nr_nodes, node, &node_states[N_MEMORY]) {
2535 2536
		void *addr;

2537
		addr = memblock_alloc_try_nid_raw(
2538
				huge_page_size(h), huge_page_size(h),
2539
				0, MEMBLOCK_ALLOC_ACCESSIBLE, node);
2540 2541 2542 2543 2544 2545 2546
		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;
2547
			goto found;
2548 2549 2550 2551 2552
		}
	}
	return 0;

found:
2553
	BUG_ON(!IS_ALIGNED(virt_to_phys(m), huge_page_size(h)));
2554
	/* Put them into a private list first because mem_map is not up yet */
2555
	INIT_LIST_HEAD(&m->list);
2556 2557 2558 2559 2560
	list_add(&m->list, &huge_boot_pages);
	m->hstate = h;
	return 1;
}

2561 2562
static void __init prep_compound_huge_page(struct page *page,
		unsigned int order)
2563 2564 2565 2566 2567 2568 2569
{
	if (unlikely(order > (MAX_ORDER - 1)))
		prep_compound_gigantic_page(page, order);
	else
		prep_compound_page(page, order);
}

2570 2571 2572 2573 2574 2575
/* 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) {
2576
		struct page *page = virt_to_page(m);
2577
		struct hstate *h = m->hstate;
2578

2579
		WARN_ON(page_count(page) != 1);
2580
		prep_compound_huge_page(page, huge_page_order(h));
2581
		WARN_ON(PageReserved(page));
2582
		prep_new_huge_page(h, page, page_to_nid(page));
2583 2584
		put_page(page); /* free it into the hugepage allocator */

2585 2586 2587 2588 2589 2590
		/*
		 * 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.
		 */
2591
		if (hstate_is_gigantic(h))
2592
			adjust_managed_page_count(page, pages_per_huge_page(h));
2593
		cond_resched();
2594 2595 2596
	}
}

2597
static void __init hugetlb_hstate_alloc_pages(struct hstate *h)
L
Linus Torvalds 已提交
2598 2599
{
	unsigned long i;
2600 2601 2602 2603 2604 2605 2606 2607 2608 2609 2610 2611 2612 2613 2614 2615 2616 2617 2618
	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);
2619

2620
	for (i = 0; i < h->max_huge_pages; ++i) {
2621
		if (hstate_is_gigantic(h)) {
2622
			if (hugetlb_cma_size) {
2623
				pr_warn_once("HugeTLB: hugetlb_cma is enabled, skip boot time allocation\n");
2624
				goto free;
2625
			}
2626 2627
			if (!alloc_bootmem_huge_page(h))
				break;
2628
		} else if (!alloc_pool_huge_page(h,
2629 2630
					 &node_states[N_MEMORY],
					 node_alloc_noretry))
L
Linus Torvalds 已提交
2631
			break;
2632
		cond_resched();
L
Linus Torvalds 已提交
2633
	}
2634 2635 2636
	if (i < h->max_huge_pages) {
		char buf[32];

2637
		string_get_size(huge_page_size(h), 1, STRING_UNITS_2, buf, 32);
2638 2639 2640 2641
		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;
	}
2642
free:
2643
	kfree(node_alloc_noretry);
2644 2645 2646 2647 2648 2649 2650
}

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

	for_each_hstate(h) {
2651 2652 2653
		if (minimum_order > huge_page_order(h))
			minimum_order = huge_page_order(h);

2654
		/* oversize hugepages were init'ed in early boot */
2655
		if (!hstate_is_gigantic(h))
2656
			hugetlb_hstate_alloc_pages(h);
2657
	}
2658
	VM_BUG_ON(minimum_order == UINT_MAX);
2659 2660 2661 2662 2663 2664 2665
}

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

	for_each_hstate(h) {
A
Andi Kleen 已提交
2666
		char buf[32];
2667 2668

		string_get_size(huge_page_size(h), 1, STRING_UNITS_2, buf, 32);
2669
		pr_info("HugeTLB registered %s page size, pre-allocated %ld pages\n",
2670
			buf, h->free_huge_pages);
2671 2672 2673
	}
}

L
Linus Torvalds 已提交
2674
#ifdef CONFIG_HIGHMEM
2675 2676
static void try_to_free_low(struct hstate *h, unsigned long count,
						nodemask_t *nodes_allowed)
L
Linus Torvalds 已提交
2677
{
2678
	int i;
2679
	LIST_HEAD(page_list);
2680

2681
	lockdep_assert_held(&hugetlb_lock);
2682
	if (hstate_is_gigantic(h))
2683 2684
		return;

2685 2686 2687
	/*
	 * Collect pages to be freed on a list, and free after dropping lock
	 */
2688
	for_each_node_mask(i, *nodes_allowed) {
2689
		struct page *page, *next;
2690 2691 2692
		struct list_head *freel = &h->hugepage_freelists[i];
		list_for_each_entry_safe(page, next, freel, lru) {
			if (count >= h->nr_huge_pages)
2693
				goto out;
L
Linus Torvalds 已提交
2694 2695
			if (PageHighMem(page))
				continue;
2696
			remove_hugetlb_page(h, page, false);
2697
			list_add(&page->lru, &page_list);
L
Linus Torvalds 已提交
2698 2699
		}
	}
2700 2701

out:
2702
	spin_unlock_irq(&hugetlb_lock);
2703
	update_and_free_pages_bulk(h, &page_list);
2704
	spin_lock_irq(&hugetlb_lock);
L
Linus Torvalds 已提交
2705 2706
}
#else
2707 2708
static inline void try_to_free_low(struct hstate *h, unsigned long count,
						nodemask_t *nodes_allowed)
L
Linus Torvalds 已提交
2709 2710 2711 2712
{
}
#endif

2713 2714 2715 2716 2717
/*
 * 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.
 */
2718 2719
static int adjust_pool_surplus(struct hstate *h, nodemask_t *nodes_allowed,
				int delta)
2720
{
2721
	int nr_nodes, node;
2722

2723
	lockdep_assert_held(&hugetlb_lock);
2724 2725
	VM_BUG_ON(delta != -1 && delta != 1);

2726 2727 2728 2729
	if (delta < 0) {
		for_each_node_mask_to_alloc(h, nr_nodes, node, nodes_allowed) {
			if (h->surplus_huge_pages_node[node])
				goto found;
2730
		}
2731 2732 2733 2734 2735
	} 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;
2736
		}
2737 2738
	}
	return 0;
2739

2740 2741 2742 2743
found:
	h->surplus_huge_pages += delta;
	h->surplus_huge_pages_node[node] += delta;
	return 1;
2744 2745
}

2746
#define persistent_huge_pages(h) (h->nr_huge_pages - h->surplus_huge_pages)
2747
static int set_max_huge_pages(struct hstate *h, unsigned long count, int nid,
2748
			      nodemask_t *nodes_allowed)
L
Linus Torvalds 已提交
2749
{
2750
	unsigned long min_count, ret;
2751 2752
	struct page *page;
	LIST_HEAD(page_list);
2753 2754 2755 2756 2757 2758 2759 2760 2761 2762 2763
	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
2764

2765 2766 2767 2768 2769
	/*
	 * resize_lock mutex prevents concurrent adjustments to number of
	 * pages in hstate via the proc/sysfs interfaces.
	 */
	mutex_lock(&h->resize_lock);
2770
	spin_lock_irq(&hugetlb_lock);
2771

2772 2773 2774 2775 2776 2777 2778 2779 2780 2781 2782 2783 2784 2785 2786 2787 2788 2789 2790 2791
	/*
	 * 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;
	}

2792 2793 2794 2795 2796 2797 2798 2799 2800
	/*
	 * 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)) {
2801
			spin_unlock_irq(&hugetlb_lock);
2802
			mutex_unlock(&h->resize_lock);
2803
			NODEMASK_FREE(node_alloc_noretry);
2804 2805 2806 2807
			return -EINVAL;
		}
		/* Fall through to decrease pool */
	}
2808

2809 2810 2811 2812
	/*
	 * Increase the pool size
	 * First take pages out of surplus state.  Then make up the
	 * remaining difference by allocating fresh huge pages.
2813
	 *
2814
	 * We might race with alloc_surplus_huge_page() here and be unable
2815 2816 2817 2818
	 * 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.
2819
	 */
2820
	while (h->surplus_huge_pages && count > persistent_huge_pages(h)) {
2821
		if (!adjust_pool_surplus(h, nodes_allowed, -1))
2822 2823 2824
			break;
	}

2825
	while (count > persistent_huge_pages(h)) {
2826 2827 2828 2829 2830
		/*
		 * 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.
		 */
2831
		spin_unlock_irq(&hugetlb_lock);
2832 2833 2834 2835

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

2836 2837
		ret = alloc_pool_huge_page(h, nodes_allowed,
						node_alloc_noretry);
2838
		spin_lock_irq(&hugetlb_lock);
2839 2840 2841
		if (!ret)
			goto out;

2842 2843 2844
		/* Bail for signals. Probably ctrl-c from user */
		if (signal_pending(current))
			goto out;
2845 2846 2847 2848 2849 2850 2851 2852
	}

	/*
	 * 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.
2853 2854 2855 2856
	 *
	 * 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
2857
	 * alloc_surplus_huge_page() is checking the global counter,
2858 2859 2860
	 * 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.
2861
	 */
2862
	min_count = h->resv_huge_pages + h->nr_huge_pages - h->free_huge_pages;
2863
	min_count = max(count, min_count);
2864
	try_to_free_low(h, min_count, nodes_allowed);
2865 2866 2867 2868

	/*
	 * Collect pages to be removed on list without dropping lock
	 */
2869
	while (min_count < persistent_huge_pages(h)) {
2870 2871
		page = remove_pool_huge_page(h, nodes_allowed, 0);
		if (!page)
L
Linus Torvalds 已提交
2872
			break;
2873 2874

		list_add(&page->lru, &page_list);
L
Linus Torvalds 已提交
2875
	}
2876
	/* free the pages after dropping lock */
2877
	spin_unlock_irq(&hugetlb_lock);
2878
	update_and_free_pages_bulk(h, &page_list);
2879
	spin_lock_irq(&hugetlb_lock);
2880

2881
	while (count < persistent_huge_pages(h)) {
2882
		if (!adjust_pool_surplus(h, nodes_allowed, 1))
2883 2884 2885
			break;
	}
out:
2886
	h->max_huge_pages = persistent_huge_pages(h);
2887
	spin_unlock_irq(&hugetlb_lock);
2888
	mutex_unlock(&h->resize_lock);
2889

2890 2891
	NODEMASK_FREE(node_alloc_noretry);

2892
	return 0;
L
Linus Torvalds 已提交
2893 2894
}

2895 2896 2897 2898 2899 2900 2901 2902 2903 2904
#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];

2905 2906 2907
static struct hstate *kobj_to_node_hstate(struct kobject *kobj, int *nidp);

static struct hstate *kobj_to_hstate(struct kobject *kobj, int *nidp)
2908 2909
{
	int i;
2910

2911
	for (i = 0; i < HUGE_MAX_HSTATE; i++)
2912 2913 2914
		if (hstate_kobjs[i] == kobj) {
			if (nidp)
				*nidp = NUMA_NO_NODE;
2915
			return &hstates[i];
2916 2917 2918
		}

	return kobj_to_node_hstate(kobj, nidp);
2919 2920
}

2921
static ssize_t nr_hugepages_show_common(struct kobject *kobj,
2922 2923
					struct kobj_attribute *attr, char *buf)
{
2924 2925 2926 2927 2928 2929 2930 2931 2932 2933
	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];

2934
	return sysfs_emit(buf, "%lu\n", nr_huge_pages);
2935
}
2936

2937 2938 2939
static ssize_t __nr_hugepages_store_common(bool obey_mempolicy,
					   struct hstate *h, int nid,
					   unsigned long count, size_t len)
2940 2941
{
	int err;
2942
	nodemask_t nodes_allowed, *n_mask;
2943

2944 2945
	if (hstate_is_gigantic(h) && !gigantic_page_runtime_supported())
		return -EINVAL;
2946

2947 2948 2949 2950 2951
	if (nid == NUMA_NO_NODE) {
		/*
		 * global hstate attribute
		 */
		if (!(obey_mempolicy &&
2952 2953 2954 2955 2956
				init_nodemask_of_mempolicy(&nodes_allowed)))
			n_mask = &node_states[N_MEMORY];
		else
			n_mask = &nodes_allowed;
	} else {
2957
		/*
2958 2959
		 * Node specific request.  count adjustment happens in
		 * set_max_huge_pages() after acquiring hugetlb_lock.
2960
		 */
2961 2962
		init_nodemask_of_node(&nodes_allowed, nid);
		n_mask = &nodes_allowed;
2963
	}
2964

2965
	err = set_max_huge_pages(h, count, nid, n_mask);
2966

2967
	return err ? err : len;
2968 2969
}

2970 2971 2972 2973 2974 2975 2976 2977 2978 2979 2980 2981 2982 2983 2984 2985 2986
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);
}

2987 2988 2989 2990 2991 2992 2993 2994 2995
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)
{
2996
	return nr_hugepages_store_common(false, kobj, buf, len);
2997 2998 2999
}
HSTATE_ATTR(nr_hugepages);

3000 3001 3002 3003 3004 3005 3006
#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,
3007 3008
					   struct kobj_attribute *attr,
					   char *buf)
3009 3010 3011 3012 3013 3014 3015
{
	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)
{
3016
	return nr_hugepages_store_common(true, kobj, buf, len);
3017 3018 3019 3020 3021
}
HSTATE_ATTR(nr_hugepages_mempolicy);
#endif


3022 3023 3024
static ssize_t nr_overcommit_hugepages_show(struct kobject *kobj,
					struct kobj_attribute *attr, char *buf)
{
3025
	struct hstate *h = kobj_to_hstate(kobj, NULL);
3026
	return sysfs_emit(buf, "%lu\n", h->nr_overcommit_huge_pages);
3027
}
3028

3029 3030 3031 3032 3033
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;
3034
	struct hstate *h = kobj_to_hstate(kobj, NULL);
3035

3036
	if (hstate_is_gigantic(h))
3037 3038
		return -EINVAL;

3039
	err = kstrtoul(buf, 10, &input);
3040
	if (err)
3041
		return err;
3042

3043
	spin_lock_irq(&hugetlb_lock);
3044
	h->nr_overcommit_huge_pages = input;
3045
	spin_unlock_irq(&hugetlb_lock);
3046 3047 3048 3049 3050 3051 3052 3053

	return count;
}
HSTATE_ATTR(nr_overcommit_hugepages);

static ssize_t free_hugepages_show(struct kobject *kobj,
					struct kobj_attribute *attr, char *buf)
{
3054 3055 3056 3057 3058 3059 3060 3061 3062 3063
	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];

3064
	return sysfs_emit(buf, "%lu\n", free_huge_pages);
3065 3066 3067 3068 3069 3070
}
HSTATE_ATTR_RO(free_hugepages);

static ssize_t resv_hugepages_show(struct kobject *kobj,
					struct kobj_attribute *attr, char *buf)
{
3071
	struct hstate *h = kobj_to_hstate(kobj, NULL);
3072
	return sysfs_emit(buf, "%lu\n", h->resv_huge_pages);
3073 3074 3075 3076 3077 3078
}
HSTATE_ATTR_RO(resv_hugepages);

static ssize_t surplus_hugepages_show(struct kobject *kobj,
					struct kobj_attribute *attr, char *buf)
{
3079 3080 3081 3082 3083 3084 3085 3086 3087 3088
	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];

3089
	return sysfs_emit(buf, "%lu\n", surplus_huge_pages);
3090 3091 3092 3093 3094 3095 3096 3097 3098
}
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,
3099 3100 3101
#ifdef CONFIG_NUMA
	&nr_hugepages_mempolicy_attr.attr,
#endif
3102 3103 3104
	NULL,
};

3105
static const struct attribute_group hstate_attr_group = {
3106 3107 3108
	.attrs = hstate_attrs,
};

J
Jeff Mahoney 已提交
3109 3110
static int hugetlb_sysfs_add_hstate(struct hstate *h, struct kobject *parent,
				    struct kobject **hstate_kobjs,
3111
				    const struct attribute_group *hstate_attr_group)
3112 3113
{
	int retval;
3114
	int hi = hstate_index(h);
3115

3116 3117
	hstate_kobjs[hi] = kobject_create_and_add(h->name, parent);
	if (!hstate_kobjs[hi])
3118 3119
		return -ENOMEM;

3120
	retval = sysfs_create_group(hstate_kobjs[hi], hstate_attr_group);
3121
	if (retval) {
3122
		kobject_put(hstate_kobjs[hi]);
3123 3124
		hstate_kobjs[hi] = NULL;
	}
3125 3126 3127 3128 3129 3130 3131 3132 3133 3134 3135 3136 3137 3138

	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) {
3139 3140
		err = hugetlb_sysfs_add_hstate(h, hugepages_kobj,
					 hstate_kobjs, &hstate_attr_group);
3141
		if (err)
3142
			pr_err("HugeTLB: Unable to add hstate %s", h->name);
3143 3144 3145
	}
}

3146 3147 3148 3149
#ifdef CONFIG_NUMA

/*
 * node_hstate/s - associate per node hstate attributes, via their kobjects,
3150 3151 3152
 * 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
3153 3154 3155 3156 3157 3158
 * the base kernel, on the hugetlb module.
 */
struct node_hstate {
	struct kobject		*hugepages_kobj;
	struct kobject		*hstate_kobjs[HUGE_MAX_HSTATE];
};
3159
static struct node_hstate node_hstates[MAX_NUMNODES];
3160 3161

/*
3162
 * A subset of global hstate attributes for node devices
3163 3164 3165 3166 3167 3168 3169 3170
 */
static struct attribute *per_node_hstate_attrs[] = {
	&nr_hugepages_attr.attr,
	&free_hugepages_attr.attr,
	&surplus_hugepages_attr.attr,
	NULL,
};

3171
static const struct attribute_group per_node_hstate_attr_group = {
3172 3173 3174 3175
	.attrs = per_node_hstate_attrs,
};

/*
3176
 * kobj_to_node_hstate - lookup global hstate for node device hstate attr kobj.
3177 3178 3179 3180 3181 3182 3183 3184 3185 3186 3187 3188 3189 3190 3191 3192 3193 3194 3195 3196 3197 3198
 * 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;
}

/*
3199
 * Unregister hstate attributes from a single node device.
3200 3201
 * No-op if no hstate attributes attached.
 */
3202
static void hugetlb_unregister_node(struct node *node)
3203 3204
{
	struct hstate *h;
3205
	struct node_hstate *nhs = &node_hstates[node->dev.id];
3206 3207

	if (!nhs->hugepages_kobj)
3208
		return;		/* no hstate attributes */
3209

3210 3211 3212 3213 3214
	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;
3215
		}
3216
	}
3217 3218 3219 3220 3221 3222 3223

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


/*
3224
 * Register hstate attributes for a single node device.
3225 3226
 * No-op if attributes already registered.
 */
3227
static void hugetlb_register_node(struct node *node)
3228 3229
{
	struct hstate *h;
3230
	struct node_hstate *nhs = &node_hstates[node->dev.id];
3231 3232 3233 3234 3235 3236
	int err;

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

	nhs->hugepages_kobj = kobject_create_and_add("hugepages",
3237
							&node->dev.kobj);
3238 3239 3240 3241 3242 3243 3244 3245
	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) {
3246
			pr_err("HugeTLB: Unable to add hstate %s for node %d\n",
3247
				h->name, node->dev.id);
3248 3249 3250 3251 3252 3253 3254
			hugetlb_unregister_node(node);
			break;
		}
	}
}

/*
3255
 * hugetlb init time:  register hstate attributes for all registered node
3256 3257
 * devices of nodes that have memory.  All on-line nodes should have
 * registered their associated device by this time.
3258
 */
3259
static void __init hugetlb_register_all_nodes(void)
3260 3261 3262
{
	int nid;

3263
	for_each_node_state(nid, N_MEMORY) {
3264
		struct node *node = node_devices[nid];
3265
		if (node->dev.id == nid)
3266 3267 3268 3269
			hugetlb_register_node(node);
	}

	/*
3270
	 * Let the node device driver know we're here so it can
3271 3272 3273 3274 3275 3276 3277 3278 3279 3280 3281 3282 3283 3284 3285 3286 3287 3288 3289
	 * [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

3290 3291
static int __init hugetlb_init(void)
{
3292 3293
	int i;

3294 3295 3296
	BUILD_BUG_ON(sizeof_field(struct page, private) * BITS_PER_BYTE <
			__NR_HPAGEFLAGS);

3297 3298 3299
	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");
3300
		return 0;
3301
	}
3302

3303 3304 3305 3306 3307 3308 3309 3310 3311 3312 3313 3314 3315 3316 3317 3318 3319 3320 3321 3322 3323 3324 3325 3326 3327 3328 3329 3330
	/*
	 * 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;
3331
		}
3332
	}
3333

3334
	hugetlb_cma_check();
3335
	hugetlb_init_hstates();
3336
	gather_bootmem_prealloc();
3337 3338 3339
	report_hugepages();

	hugetlb_sysfs_init();
3340
	hugetlb_register_all_nodes();
3341
	hugetlb_cgroup_file_init();
3342

3343 3344 3345 3346 3347
#ifdef CONFIG_SMP
	num_fault_mutexes = roundup_pow_of_two(8 * num_possible_cpus());
#else
	num_fault_mutexes = 1;
#endif
3348
	hugetlb_fault_mutex_table =
3349 3350
		kmalloc_array(num_fault_mutexes, sizeof(struct mutex),
			      GFP_KERNEL);
3351
	BUG_ON(!hugetlb_fault_mutex_table);
3352 3353

	for (i = 0; i < num_fault_mutexes; i++)
3354
		mutex_init(&hugetlb_fault_mutex_table[i]);
3355 3356
	return 0;
}
3357
subsys_initcall(hugetlb_init);
3358

3359 3360
/* Overwritten by architectures with more huge page sizes */
bool __init __attribute((weak)) arch_hugetlb_valid_size(unsigned long size)
3361
{
3362
	return size == HPAGE_SIZE;
3363 3364
}

3365
void __init hugetlb_add_hstate(unsigned int order)
3366 3367
{
	struct hstate *h;
3368 3369
	unsigned long i;

3370 3371 3372
	if (size_to_hstate(PAGE_SIZE << order)) {
		return;
	}
3373
	BUG_ON(hugetlb_max_hstate >= HUGE_MAX_HSTATE);
3374
	BUG_ON(order == 0);
3375
	h = &hstates[hugetlb_max_hstate++];
3376
	mutex_init(&h->resize_lock);
3377
	h->order = order;
3378
	h->mask = ~(huge_page_size(h) - 1);
3379 3380
	for (i = 0; i < MAX_NUMNODES; ++i)
		INIT_LIST_HEAD(&h->hugepage_freelists[i]);
3381
	INIT_LIST_HEAD(&h->hugepage_activelist);
3382 3383
	h->next_nid_to_alloc = first_memory_node;
	h->next_nid_to_free = first_memory_node;
3384 3385
	snprintf(h->name, HSTATE_NAME_LEN, "hugepages-%lukB",
					huge_page_size(h)/1024);
3386

3387 3388 3389
	parsed_hstate = h;
}

3390 3391 3392 3393 3394 3395 3396 3397
/*
 * 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)
3398 3399
{
	unsigned long *mhp;
3400
	static unsigned long *last_mhp;
3401

3402
	if (!parsed_valid_hugepagesz) {
3403
		pr_warn("HugeTLB: hugepages=%s does not follow a valid hugepagesz, ignoring\n", s);
3404
		parsed_valid_hugepagesz = true;
3405
		return 0;
3406
	}
3407

3408
	/*
3409 3410 3411 3412
	 * !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.
3413
	 */
3414
	else if (!hugetlb_max_hstate)
3415 3416 3417 3418
		mhp = &default_hstate_max_huge_pages;
	else
		mhp = &parsed_hstate->max_huge_pages;

3419
	if (mhp == last_mhp) {
3420 3421
		pr_warn("HugeTLB: hugepages= specified twice without interleaving hugepagesz=, ignoring hugepages=%s\n", s);
		return 0;
3422 3423
	}

3424 3425 3426
	if (sscanf(s, "%lu", mhp) <= 0)
		*mhp = 0;

3427 3428
	/*
	 * Global state is always initialized later in hugetlb_init.
3429
	 * But we need to allocate gigantic hstates here early to still
3430 3431
	 * use the bootmem allocator.
	 */
3432
	if (hugetlb_max_hstate && hstate_is_gigantic(parsed_hstate))
3433 3434 3435 3436
		hugetlb_hstate_alloc_pages(parsed_hstate);

	last_mhp = mhp;

3437 3438
	return 1;
}
3439
__setup("hugepages=", hugepages_setup);
3440

3441 3442 3443 3444 3445 3446 3447
/*
 * 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.
 */
3448
static int __init hugepagesz_setup(char *s)
3449
{
3450
	unsigned long size;
3451 3452 3453
	struct hstate *h;

	parsed_valid_hugepagesz = false;
3454 3455 3456
	size = (unsigned long)memparse(s, NULL);

	if (!arch_hugetlb_valid_size(size)) {
3457
		pr_err("HugeTLB: unsupported hugepagesz=%s\n", s);
3458 3459 3460
		return 0;
	}

3461 3462 3463 3464 3465 3466 3467 3468 3469 3470 3471 3472 3473 3474 3475 3476 3477 3478 3479 3480 3481 3482 3483
	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;
3484 3485
	}

3486
	hugetlb_add_hstate(ilog2(size) - PAGE_SHIFT);
3487
	parsed_valid_hugepagesz = true;
3488 3489
	return 1;
}
3490 3491
__setup("hugepagesz=", hugepagesz_setup);

3492 3493 3494 3495
/*
 * default_hugepagesz command line input
 * Only one instance of default_hugepagesz allowed on command line.
 */
3496
static int __init default_hugepagesz_setup(char *s)
3497
{
3498 3499
	unsigned long size;

3500 3501 3502 3503 3504 3505
	parsed_valid_hugepagesz = false;
	if (parsed_default_hugepagesz) {
		pr_err("HugeTLB: default_hugepagesz previously specified, ignoring %s\n", s);
		return 0;
	}

3506 3507 3508
	size = (unsigned long)memparse(s, NULL);

	if (!arch_hugetlb_valid_size(size)) {
3509
		pr_err("HugeTLB: unsupported default_hugepagesz=%s\n", s);
3510 3511 3512
		return 0;
	}

3513 3514 3515 3516 3517 3518 3519 3520 3521 3522 3523 3524 3525 3526 3527 3528 3529 3530 3531
	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;
	}

3532 3533
	return 1;
}
3534
__setup("default_hugepagesz=", default_hugepagesz_setup);
3535

3536
static unsigned int allowed_mems_nr(struct hstate *h)
3537 3538 3539
{
	int node;
	unsigned int nr = 0;
3540 3541 3542 3543 3544
	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);
3545

3546
	for_each_node_mask(node, cpuset_current_mems_allowed) {
3547
		if (!mpol_allowed || node_isset(node, *mpol_allowed))
3548 3549
			nr += array[node];
	}
3550 3551 3552 3553 3554

	return nr;
}

#ifdef CONFIG_SYSCTL
3555 3556 3557 3558 3559 3560 3561 3562 3563 3564 3565 3566 3567 3568 3569 3570
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);
}

3571 3572
static int hugetlb_sysctl_handler_common(bool obey_mempolicy,
			 struct ctl_table *table, int write,
3573
			 void *buffer, size_t *length, loff_t *ppos)
L
Linus Torvalds 已提交
3574
{
3575
	struct hstate *h = &default_hstate;
3576
	unsigned long tmp = h->max_huge_pages;
3577
	int ret;
3578

3579
	if (!hugepages_supported())
3580
		return -EOPNOTSUPP;
3581

3582 3583
	ret = proc_hugetlb_doulongvec_minmax(table, write, buffer, length, ppos,
					     &tmp);
3584 3585
	if (ret)
		goto out;
3586

3587 3588 3589
	if (write)
		ret = __nr_hugepages_store_common(obey_mempolicy, h,
						  NUMA_NO_NODE, tmp, *length);
3590 3591
out:
	return ret;
L
Linus Torvalds 已提交
3592
}
3593

3594
int hugetlb_sysctl_handler(struct ctl_table *table, int write,
3595
			  void *buffer, size_t *length, loff_t *ppos)
3596 3597 3598 3599 3600 3601 3602 3603
{

	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,
3604
			  void *buffer, size_t *length, loff_t *ppos)
3605 3606 3607 3608 3609 3610
{
	return hugetlb_sysctl_handler_common(true, table, write,
							buffer, length, ppos);
}
#endif /* CONFIG_NUMA */

3611
int hugetlb_overcommit_handler(struct ctl_table *table, int write,
3612
		void *buffer, size_t *length, loff_t *ppos)
3613
{
3614
	struct hstate *h = &default_hstate;
3615
	unsigned long tmp;
3616
	int ret;
3617

3618
	if (!hugepages_supported())
3619
		return -EOPNOTSUPP;
3620

3621
	tmp = h->nr_overcommit_huge_pages;
3622

3623
	if (write && hstate_is_gigantic(h))
3624 3625
		return -EINVAL;

3626 3627
	ret = proc_hugetlb_doulongvec_minmax(table, write, buffer, length, ppos,
					     &tmp);
3628 3629
	if (ret)
		goto out;
3630 3631

	if (write) {
3632
		spin_lock_irq(&hugetlb_lock);
3633
		h->nr_overcommit_huge_pages = tmp;
3634
		spin_unlock_irq(&hugetlb_lock);
3635
	}
3636 3637
out:
	return ret;
3638 3639
}

L
Linus Torvalds 已提交
3640 3641
#endif /* CONFIG_SYSCTL */

3642
void hugetlb_report_meminfo(struct seq_file *m)
L
Linus Torvalds 已提交
3643
{
3644 3645 3646
	struct hstate *h;
	unsigned long total = 0;

3647 3648
	if (!hugepages_supported())
		return;
3649 3650 3651 3652

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

3653
		total += huge_page_size(h) * count;
3654 3655 3656 3657 3658 3659 3660 3661 3662 3663 3664 3665

		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,
3666
				   huge_page_size(h) / SZ_1K);
3667 3668
	}

3669
	seq_printf(m, "Hugetlb:        %8lu kB\n", total / SZ_1K);
L
Linus Torvalds 已提交
3670 3671
}

3672
int hugetlb_report_node_meminfo(char *buf, int len, int nid)
L
Linus Torvalds 已提交
3673
{
3674
	struct hstate *h = &default_hstate;
3675

3676 3677
	if (!hugepages_supported())
		return 0;
3678 3679 3680 3681 3682 3683 3684 3685

	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
3686 3687
}

3688 3689 3690 3691 3692
void hugetlb_show_meminfo(void)
{
	struct hstate *h;
	int nid;

3693 3694 3695
	if (!hugepages_supported())
		return;

3696 3697 3698 3699 3700 3701 3702
	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],
3703
				huge_page_size(h) / SZ_1K);
3704 3705
}

3706 3707 3708 3709 3710 3711
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
3712 3713 3714
/* Return the number pages of memory we physically have, in PAGE_SIZE units. */
unsigned long hugetlb_total_pages(void)
{
3715 3716 3717 3718 3719 3720
	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
3721 3722
}

3723
static int hugetlb_acct_memory(struct hstate *h, long delta)
M
Mel Gorman 已提交
3724 3725 3726
{
	int ret = -ENOMEM;

3727 3728 3729
	if (!delta)
		return 0;

3730
	spin_lock_irq(&hugetlb_lock);
M
Mel Gorman 已提交
3731 3732 3733 3734 3735 3736 3737 3738 3739 3740 3741 3742 3743 3744 3745 3746
	/*
	 * 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.
3747 3748 3749 3750 3751 3752
	 *
	 * 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
3753 3754
	 */
	if (delta > 0) {
3755
		if (gather_surplus_pages(h, delta) < 0)
M
Mel Gorman 已提交
3756 3757
			goto out;

3758
		if (delta > allowed_mems_nr(h)) {
3759
			return_unused_surplus_pages(h, delta);
M
Mel Gorman 已提交
3760 3761 3762 3763 3764 3765
			goto out;
		}
	}

	ret = 0;
	if (delta < 0)
3766
		return_unused_surplus_pages(h, (unsigned long) -delta);
M
Mel Gorman 已提交
3767 3768

out:
3769
	spin_unlock_irq(&hugetlb_lock);
M
Mel Gorman 已提交
3770 3771 3772
	return ret;
}

3773 3774
static void hugetlb_vm_op_open(struct vm_area_struct *vma)
{
3775
	struct resv_map *resv = vma_resv_map(vma);
3776 3777 3778 3779 3780

	/*
	 * 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
3781
	 * has a reference to the reservation map it cannot disappear until
3782 3783 3784
	 * after this open call completes.  It is therefore safe to take a
	 * new reference here without additional locking.
	 */
3785
	if (resv && is_vma_resv_set(vma, HPAGE_RESV_OWNER))
3786
		kref_get(&resv->refs);
3787 3788
}

3789 3790
static void hugetlb_vm_op_close(struct vm_area_struct *vma)
{
3791
	struct hstate *h = hstate_vma(vma);
3792
	struct resv_map *resv = vma_resv_map(vma);
3793
	struct hugepage_subpool *spool = subpool_vma(vma);
3794
	unsigned long reserve, start, end;
3795
	long gbl_reserve;
3796

3797 3798
	if (!resv || !is_vma_resv_set(vma, HPAGE_RESV_OWNER))
		return;
3799

3800 3801
	start = vma_hugecache_offset(h, vma, vma->vm_start);
	end = vma_hugecache_offset(h, vma, vma->vm_end);
3802

3803
	reserve = (end - start) - region_count(resv, start, end);
3804
	hugetlb_cgroup_uncharge_counter(resv, start, end);
3805
	if (reserve) {
3806 3807 3808 3809 3810 3811
		/*
		 * 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);
3812
	}
3813 3814

	kref_put(&resv->refs, resv_map_release);
3815 3816
}

3817 3818 3819 3820 3821 3822 3823
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;
}

3824 3825
static unsigned long hugetlb_vm_op_pagesize(struct vm_area_struct *vma)
{
3826
	return huge_page_size(hstate_vma(vma));
3827 3828
}

L
Linus Torvalds 已提交
3829 3830 3831
/*
 * 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
3832
 * hugepage VMA.  do_page_fault() is supposed to trap this, so BUG is we get
L
Linus Torvalds 已提交
3833 3834
 * this far.
 */
3835
static vm_fault_t hugetlb_vm_op_fault(struct vm_fault *vmf)
L
Linus Torvalds 已提交
3836 3837
{
	BUG();
N
Nick Piggin 已提交
3838
	return 0;
L
Linus Torvalds 已提交
3839 3840
}

3841 3842 3843 3844 3845 3846 3847
/*
 * 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.
 */
3848
const struct vm_operations_struct hugetlb_vm_ops = {
N
Nick Piggin 已提交
3849
	.fault = hugetlb_vm_op_fault,
3850
	.open = hugetlb_vm_op_open,
3851
	.close = hugetlb_vm_op_close,
3852
	.may_split = hugetlb_vm_op_split,
3853
	.pagesize = hugetlb_vm_op_pagesize,
L
Linus Torvalds 已提交
3854 3855
};

3856 3857
static pte_t make_huge_pte(struct vm_area_struct *vma, struct page *page,
				int writable)
D
David Gibson 已提交
3858 3859 3860
{
	pte_t entry;

3861
	if (writable) {
3862 3863
		entry = huge_pte_mkwrite(huge_pte_mkdirty(mk_huge_pte(page,
					 vma->vm_page_prot)));
D
David Gibson 已提交
3864
	} else {
3865 3866
		entry = huge_pte_wrprotect(mk_huge_pte(page,
					   vma->vm_page_prot));
D
David Gibson 已提交
3867 3868 3869
	}
	entry = pte_mkyoung(entry);
	entry = pte_mkhuge(entry);
3870
	entry = arch_make_huge_pte(entry, vma, page, writable);
D
David Gibson 已提交
3871 3872 3873 3874

	return entry;
}

3875 3876 3877 3878 3879
static void set_huge_ptep_writable(struct vm_area_struct *vma,
				   unsigned long address, pte_t *ptep)
{
	pte_t entry;

3880
	entry = huge_pte_mkwrite(huge_pte_mkdirty(huge_ptep_get(ptep)));
3881
	if (huge_ptep_set_access_flags(vma, address, ptep, entry, 1))
3882
		update_mmu_cache(vma, address, ptep);
3883 3884
}

3885
bool is_hugetlb_entry_migration(pte_t pte)
3886 3887 3888 3889
{
	swp_entry_t swp;

	if (huge_pte_none(pte) || pte_present(pte))
3890
		return false;
3891
	swp = pte_to_swp_entry(pte);
3892
	if (is_migration_entry(swp))
3893
		return true;
3894
	else
3895
		return false;
3896 3897
}

3898
static bool is_hugetlb_entry_hwpoisoned(pte_t pte)
3899 3900 3901 3902
{
	swp_entry_t swp;

	if (huge_pte_none(pte) || pte_present(pte))
3903
		return false;
3904
	swp = pte_to_swp_entry(pte);
3905
	if (is_hwpoison_entry(swp))
3906
		return true;
3907
	else
3908
		return false;
3909
}
3910

3911 3912 3913 3914 3915 3916 3917 3918 3919 3920 3921 3922
static void
hugetlb_install_page(struct vm_area_struct *vma, pte_t *ptep, unsigned long addr,
		     struct page *new_page)
{
	__SetPageUptodate(new_page);
	set_huge_pte_at(vma->vm_mm, addr, ptep, make_huge_pte(vma, new_page, 1));
	hugepage_add_new_anon_rmap(new_page, vma, addr);
	hugetlb_count_add(pages_per_huge_page(hstate_vma(vma)), vma->vm_mm);
	ClearHPageRestoreReserve(new_page);
	SetHPageMigratable(new_page);
}

D
David Gibson 已提交
3923 3924 3925
int copy_hugetlb_page_range(struct mm_struct *dst, struct mm_struct *src,
			    struct vm_area_struct *vma)
{
3926
	pte_t *src_pte, *dst_pte, entry, dst_entry;
D
David Gibson 已提交
3927
	struct page *ptepage;
3928
	unsigned long addr;
3929
	bool cow = is_cow_mapping(vma->vm_flags);
3930 3931
	struct hstate *h = hstate_vma(vma);
	unsigned long sz = huge_page_size(h);
3932
	unsigned long npages = pages_per_huge_page(h);
3933
	struct address_space *mapping = vma->vm_file->f_mapping;
3934
	struct mmu_notifier_range range;
3935
	int ret = 0;
3936

3937
	if (cow) {
3938
		mmu_notifier_range_init(&range, MMU_NOTIFY_CLEAR, 0, vma, src,
3939
					vma->vm_start,
3940 3941
					vma->vm_end);
		mmu_notifier_invalidate_range_start(&range);
3942 3943 3944 3945 3946 3947 3948 3949
	} 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);
3950
	}
3951

3952
	for (addr = vma->vm_start; addr < vma->vm_end; addr += sz) {
3953
		spinlock_t *src_ptl, *dst_ptl;
3954
		src_pte = huge_pte_offset(src, addr, sz);
H
Hugh Dickins 已提交
3955 3956
		if (!src_pte)
			continue;
3957
		dst_pte = huge_pte_alloc(dst, vma, addr, sz);
3958 3959 3960 3961
		if (!dst_pte) {
			ret = -ENOMEM;
			break;
		}
3962

3963 3964 3965 3966 3967 3968 3969 3970 3971 3972 3973
		/*
		 * 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))
3974 3975
			continue;

3976 3977 3978
		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);
3979
		entry = huge_ptep_get(src_pte);
3980
		dst_entry = huge_ptep_get(dst_pte);
3981
again:
3982 3983 3984 3985 3986 3987
		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.
			 */
3988 3989 3990 3991 3992 3993 3994 3995 3996 3997 3998 3999
			;
		} 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);
4000 4001
				set_huge_swap_pte_at(src, addr, src_pte,
						     entry, sz);
4002
			}
4003
			set_huge_swap_pte_at(dst, addr, dst_pte, entry, sz);
4004
		} else {
4005 4006 4007 4008 4009 4010 4011 4012 4013 4014 4015 4016 4017 4018 4019 4020 4021 4022 4023 4024 4025 4026 4027 4028 4029 4030 4031 4032 4033 4034 4035 4036 4037 4038 4039 4040 4041 4042 4043 4044 4045 4046 4047 4048 4049 4050
			entry = huge_ptep_get(src_pte);
			ptepage = pte_page(entry);
			get_page(ptepage);

			/*
			 * This is a rare case where we see pinned hugetlb
			 * pages while they're prone to COW.  We need to do the
			 * COW earlier during fork.
			 *
			 * When pre-allocating the page or copying data, we
			 * need to be without the pgtable locks since we could
			 * sleep during the process.
			 */
			if (unlikely(page_needs_cow_for_dma(vma, ptepage))) {
				pte_t src_pte_old = entry;
				struct page *new;

				spin_unlock(src_ptl);
				spin_unlock(dst_ptl);
				/* Do not use reserve as it's private owned */
				new = alloc_huge_page(vma, addr, 1);
				if (IS_ERR(new)) {
					put_page(ptepage);
					ret = PTR_ERR(new);
					break;
				}
				copy_user_huge_page(new, ptepage, addr, vma,
						    npages);
				put_page(ptepage);

				/* Install the new huge page if src pte stable */
				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);
				entry = huge_ptep_get(src_pte);
				if (!pte_same(src_pte_old, entry)) {
					put_page(new);
					/* dst_entry won't change as in child */
					goto again;
				}
				hugetlb_install_page(vma, dst_pte, addr, new);
				spin_unlock(src_ptl);
				spin_unlock(dst_ptl);
				continue;
			}

4051
			if (cow) {
4052 4053 4054 4055 4056
				/*
				 * No need to notify as we are downgrading page
				 * table protection not changing it to point
				 * to a new page.
				 *
4057
				 * See Documentation/vm/mmu_notifier.rst
4058
				 */
4059
				huge_ptep_set_wrprotect(src, addr, src_pte);
4060
			}
4061

4062
			page_dup_rmap(ptepage, true);
4063
			set_huge_pte_at(dst, addr, dst_pte, entry);
4064
			hugetlb_count_add(npages, dst);
4065
		}
4066 4067
		spin_unlock(src_ptl);
		spin_unlock(dst_ptl);
D
David Gibson 已提交
4068 4069
	}

4070
	if (cow)
4071
		mmu_notifier_invalidate_range_end(&range);
4072 4073
	else
		i_mmap_unlock_read(mapping);
4074 4075

	return ret;
D
David Gibson 已提交
4076 4077
}

4078 4079 4080
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
4081 4082 4083
{
	struct mm_struct *mm = vma->vm_mm;
	unsigned long address;
4084
	pte_t *ptep;
D
David Gibson 已提交
4085
	pte_t pte;
4086
	spinlock_t *ptl;
D
David Gibson 已提交
4087
	struct page *page;
4088 4089
	struct hstate *h = hstate_vma(vma);
	unsigned long sz = huge_page_size(h);
4090
	struct mmu_notifier_range range;
4091

D
David Gibson 已提交
4092
	WARN_ON(!is_vm_hugetlb_page(vma));
4093 4094
	BUG_ON(start & ~huge_page_mask(h));
	BUG_ON(end & ~huge_page_mask(h));
D
David Gibson 已提交
4095

4096 4097 4098 4099
	/*
	 * This is a hugetlb vma, all the pte entries should point
	 * to huge page.
	 */
4100
	tlb_change_page_size(tlb, sz);
4101
	tlb_start_vma(tlb, vma);
4102 4103 4104 4105

	/*
	 * If sharing possible, alert mmu notifiers of worst case.
	 */
4106 4107
	mmu_notifier_range_init(&range, MMU_NOTIFY_UNMAP, 0, vma, mm, start,
				end);
4108 4109
	adjust_range_if_pmd_sharing_possible(vma, &range.start, &range.end);
	mmu_notifier_invalidate_range_start(&range);
4110 4111
	address = start;
	for (; address < end; address += sz) {
4112
		ptep = huge_pte_offset(mm, address, sz);
A
Adam Litke 已提交
4113
		if (!ptep)
4114 4115
			continue;

4116
		ptl = huge_pte_lock(h, mm, ptep);
4117
		if (huge_pmd_unshare(mm, vma, &address, ptep)) {
4118
			spin_unlock(ptl);
4119 4120 4121 4122
			/*
			 * We just unmapped a page of PMDs by clearing a PUD.
			 * The caller's TLB flush range should cover this area.
			 */
4123 4124
			continue;
		}
4125

4126
		pte = huge_ptep_get(ptep);
4127 4128 4129 4130
		if (huge_pte_none(pte)) {
			spin_unlock(ptl);
			continue;
		}
4131 4132

		/*
4133 4134
		 * Migrating hugepage or HWPoisoned hugepage is already
		 * unmapped and its refcount is dropped, so just clear pte here.
4135
		 */
4136
		if (unlikely(!pte_present(pte))) {
4137
			huge_pte_clear(mm, address, ptep, sz);
4138 4139
			spin_unlock(ptl);
			continue;
4140
		}
4141 4142

		page = pte_page(pte);
4143 4144 4145 4146 4147 4148
		/*
		 * 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) {
4149 4150 4151 4152
			if (page != ref_page) {
				spin_unlock(ptl);
				continue;
			}
4153 4154 4155 4156 4157 4158 4159 4160
			/*
			 * 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);
		}

4161
		pte = huge_ptep_get_and_clear(mm, address, ptep);
4162
		tlb_remove_huge_tlb_entry(h, tlb, ptep, address);
4163
		if (huge_pte_dirty(pte))
4164
			set_page_dirty(page);
4165

4166
		hugetlb_count_sub(pages_per_huge_page(h), mm);
4167
		page_remove_rmap(page, true);
4168

4169
		spin_unlock(ptl);
4170
		tlb_remove_page_size(tlb, page, huge_page_size(h));
4171 4172 4173 4174 4175
		/*
		 * Bail out after unmapping reference page if supplied
		 */
		if (ref_page)
			break;
4176
	}
4177
	mmu_notifier_invalidate_range_end(&range);
4178
	tlb_end_vma(tlb, vma);
L
Linus Torvalds 已提交
4179
}
D
David Gibson 已提交
4180

4181 4182 4183 4184 4185 4186 4187 4188 4189 4190 4191 4192
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
4193
	 * is to clear it before releasing the i_mmap_rwsem. This works
4194
	 * because in the context this is called, the VMA is about to be
4195
	 * destroyed and the i_mmap_rwsem is held.
4196 4197 4198 4199
	 */
	vma->vm_flags &= ~VM_MAYSHARE;
}

4200
void unmap_hugepage_range(struct vm_area_struct *vma, unsigned long start,
4201
			  unsigned long end, struct page *ref_page)
4202
{
4203
	struct mmu_gather tlb;
4204

4205
	tlb_gather_mmu(&tlb, vma->vm_mm);
4206
	__unmap_hugepage_range(&tlb, vma, start, end, ref_page);
4207
	tlb_finish_mmu(&tlb);
4208 4209
}

4210 4211
/*
 * This is called when the original mapper is failing to COW a MAP_PRIVATE
4212
 * mapping it owns the reserve page for. The intention is to unmap the page
4213 4214 4215
 * from other VMAs and let the children be SIGKILLed if they are faulting the
 * same region.
 */
4216 4217
static void unmap_ref_private(struct mm_struct *mm, struct vm_area_struct *vma,
			      struct page *page, unsigned long address)
4218
{
4219
	struct hstate *h = hstate_vma(vma);
4220 4221 4222 4223 4224 4225 4226 4227
	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.
	 */
4228
	address = address & huge_page_mask(h);
4229 4230
	pgoff = ((address - vma->vm_start) >> PAGE_SHIFT) +
			vma->vm_pgoff;
4231
	mapping = vma->vm_file->f_mapping;
4232

4233 4234 4235 4236 4237
	/*
	 * 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
	 */
4238
	i_mmap_lock_write(mapping);
4239
	vma_interval_tree_foreach(iter_vma, &mapping->i_mmap, pgoff, pgoff) {
4240 4241 4242 4243
		/* Do not unmap the current VMA */
		if (iter_vma == vma)
			continue;

4244 4245 4246 4247 4248 4249 4250 4251
		/*
		 * 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;

4252 4253 4254 4255 4256 4257 4258 4259
		/*
		 * 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))
4260 4261
			unmap_hugepage_range(iter_vma, address,
					     address + huge_page_size(h), page);
4262
	}
4263
	i_mmap_unlock_write(mapping);
4264 4265
}

4266 4267
/*
 * Hugetlb_cow() should be called with page lock of the original hugepage held.
4268 4269 4270
 * 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.
4271
 */
4272
static vm_fault_t hugetlb_cow(struct mm_struct *mm, struct vm_area_struct *vma,
4273
		       unsigned long address, pte_t *ptep,
4274
		       struct page *pagecache_page, spinlock_t *ptl)
4275
{
4276
	pte_t pte;
4277
	struct hstate *h = hstate_vma(vma);
4278
	struct page *old_page, *new_page;
4279 4280
	int outside_reserve = 0;
	vm_fault_t ret = 0;
4281
	unsigned long haddr = address & huge_page_mask(h);
4282
	struct mmu_notifier_range range;
4283

4284
	pte = huge_ptep_get(ptep);
4285 4286
	old_page = pte_page(pte);

4287
retry_avoidcopy:
4288 4289
	/* If no-one else is actually using this page, avoid the copy
	 * and just make the page writable */
4290
	if (page_mapcount(old_page) == 1 && PageAnon(old_page)) {
4291
		page_move_anon_rmap(old_page, vma);
4292
		set_huge_ptep_writable(vma, haddr, ptep);
N
Nick Piggin 已提交
4293
		return 0;
4294 4295
	}

4296 4297 4298 4299 4300 4301 4302 4303 4304
	/*
	 * 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.
	 */
4305
	if (is_vma_resv_set(vma, HPAGE_RESV_OWNER) &&
4306 4307 4308
			old_page != pagecache_page)
		outside_reserve = 1;

4309
	get_page(old_page);
4310

4311 4312 4313 4314
	/*
	 * Drop page table lock as buddy allocator may be called. It will
	 * be acquired again before returning to the caller, as expected.
	 */
4315
	spin_unlock(ptl);
4316
	new_page = alloc_huge_page(vma, haddr, outside_reserve);
4317

4318
	if (IS_ERR(new_page)) {
4319 4320 4321 4322 4323 4324 4325 4326
		/*
		 * 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) {
4327 4328 4329 4330
			struct address_space *mapping = vma->vm_file->f_mapping;
			pgoff_t idx;
			u32 hash;

4331
			put_page(old_page);
4332
			BUG_ON(huge_pte_none(pte));
4333 4334 4335 4336 4337 4338 4339 4340 4341 4342 4343 4344 4345 4346
			/*
			 * 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);

4347
			unmap_ref_private(mm, vma, old_page, haddr);
4348 4349 4350

			i_mmap_lock_read(mapping);
			mutex_lock(&hugetlb_fault_mutex_table[hash]);
4351
			spin_lock(ptl);
4352
			ptep = huge_pte_offset(mm, haddr, huge_page_size(h));
4353 4354 4355 4356 4357 4358 4359 4360
			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;
4361 4362
		}

4363
		ret = vmf_error(PTR_ERR(new_page));
4364
		goto out_release_old;
4365 4366
	}

4367 4368 4369 4370
	/*
	 * When the original hugepage is shared one, it does not have
	 * anon_vma prepared.
	 */
4371
	if (unlikely(anon_vma_prepare(vma))) {
4372 4373
		ret = VM_FAULT_OOM;
		goto out_release_all;
4374
	}
4375

4376
	copy_user_huge_page(new_page, old_page, address, vma,
A
Andrea Arcangeli 已提交
4377
			    pages_per_huge_page(h));
N
Nick Piggin 已提交
4378
	__SetPageUptodate(new_page);
4379

4380
	mmu_notifier_range_init(&range, MMU_NOTIFY_CLEAR, 0, vma, mm, haddr,
4381
				haddr + huge_page_size(h));
4382
	mmu_notifier_invalidate_range_start(&range);
4383

4384
	/*
4385
	 * Retake the page table lock to check for racing updates
4386 4387
	 * before the page tables are altered
	 */
4388
	spin_lock(ptl);
4389
	ptep = huge_pte_offset(mm, haddr, huge_page_size(h));
4390
	if (likely(ptep && pte_same(huge_ptep_get(ptep), pte))) {
4391
		ClearHPageRestoreReserve(new_page);
4392

4393
		/* Break COW */
4394
		huge_ptep_clear_flush(vma, haddr, ptep);
4395
		mmu_notifier_invalidate_range(mm, range.start, range.end);
4396
		set_huge_pte_at(mm, haddr, ptep,
4397
				make_huge_pte(vma, new_page, 1));
4398
		page_remove_rmap(old_page, true);
4399
		hugepage_add_new_anon_rmap(new_page, vma, haddr);
4400
		SetHPageMigratable(new_page);
4401 4402 4403
		/* Make the old page be freed below */
		new_page = old_page;
	}
4404
	spin_unlock(ptl);
4405
	mmu_notifier_invalidate_range_end(&range);
4406
out_release_all:
4407
	restore_reserve_on_error(h, vma, haddr, new_page);
4408
	put_page(new_page);
4409
out_release_old:
4410
	put_page(old_page);
4411

4412 4413
	spin_lock(ptl); /* Caller expects lock to be held */
	return ret;
4414 4415
}

4416
/* Return the pagecache page at a given address within a VMA */
4417 4418
static struct page *hugetlbfs_pagecache_page(struct hstate *h,
			struct vm_area_struct *vma, unsigned long address)
4419 4420
{
	struct address_space *mapping;
4421
	pgoff_t idx;
4422 4423

	mapping = vma->vm_file->f_mapping;
4424
	idx = vma_hugecache_offset(h, vma, address);
4425 4426 4427 4428

	return find_lock_page(mapping, idx);
}

H
Hugh Dickins 已提交
4429 4430 4431 4432 4433
/*
 * 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
4434 4435 4436 4437 4438 4439 4440 4441 4442 4443 4444 4445 4446 4447 4448
			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;
}

4449 4450 4451 4452 4453 4454 4455 4456 4457
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;
4458
	ClearHPageRestoreReserve(page);
4459

4460 4461 4462 4463 4464 4465
	/*
	 * set page dirty so that it will not be removed from cache/file
	 * by non-hugetlbfs specific code paths.
	 */
	set_page_dirty(page);

4466 4467 4468 4469 4470 4471
	spin_lock(&inode->i_lock);
	inode->i_blocks += blocks_per_huge_page(h);
	spin_unlock(&inode->i_lock);
	return 0;
}

4472 4473 4474 4475
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)
4476
{
4477
	struct hstate *h = hstate_vma(vma);
4478
	vm_fault_t ret = VM_FAULT_SIGBUS;
4479
	int anon_rmap = 0;
A
Adam Litke 已提交
4480 4481
	unsigned long size;
	struct page *page;
4482
	pte_t new_pte;
4483
	spinlock_t *ptl;
4484
	unsigned long haddr = address & huge_page_mask(h);
4485
	bool new_page = false;
A
Adam Litke 已提交
4486

4487 4488 4489
	/*
	 * 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
4490
	 * COW. Warn that such a situation has occurred as it may not be obvious
4491 4492
	 */
	if (is_vma_resv_set(vma, HPAGE_RESV_UNMAPPED)) {
4493
		pr_warn_ratelimited("PID %d killed due to inadequate hugepage pool\n",
4494
			   current->pid);
4495 4496 4497
		return ret;
	}

A
Adam Litke 已提交
4498
	/*
4499 4500 4501
	 * 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
4502
	 */
4503 4504 4505 4506
	size = i_size_read(mapping->host) >> huge_page_shift(h);
	if (idx >= size)
		goto out;

4507 4508 4509
retry:
	page = find_lock_page(mapping, idx);
	if (!page) {
4510 4511 4512 4513 4514 4515 4516
		/*
		 * Check for page in userfault range
		 */
		if (userfaultfd_missing(vma)) {
			u32 hash;
			struct vm_fault vmf = {
				.vma = vma,
4517
				.address = haddr,
4518 4519 4520 4521 4522 4523 4524 4525 4526 4527 4528
				.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
				 */
			};

			/*
4529 4530 4531
			 * hugetlb_fault_mutex and i_mmap_rwsem must be
			 * dropped before handling userfault.  Reacquire
			 * after handling fault to make calling code simpler.
4532
			 */
4533
			hash = hugetlb_fault_mutex_hash(mapping, idx);
4534
			mutex_unlock(&hugetlb_fault_mutex_table[hash]);
4535
			i_mmap_unlock_read(mapping);
4536
			ret = handle_userfault(&vmf, VM_UFFD_MISSING);
4537
			i_mmap_lock_read(mapping);
4538 4539 4540 4541
			mutex_lock(&hugetlb_fault_mutex_table[hash]);
			goto out;
		}

4542
		page = alloc_huge_page(vma, haddr, 0);
4543
		if (IS_ERR(page)) {
4544 4545 4546 4547 4548 4549 4550 4551 4552 4553 4554 4555 4556
			/*
			 * 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);
4557 4558 4559
			ret = 0;
			if (huge_pte_none(huge_ptep_get(ptep)))
				ret = vmf_error(PTR_ERR(page));
4560
			spin_unlock(ptl);
4561 4562
			goto out;
		}
A
Andrea Arcangeli 已提交
4563
		clear_huge_page(page, address, pages_per_huge_page(h));
N
Nick Piggin 已提交
4564
		__SetPageUptodate(page);
4565
		new_page = true;
4566

4567
		if (vma->vm_flags & VM_MAYSHARE) {
4568
			int err = huge_add_to_page_cache(page, mapping, idx);
4569 4570 4571 4572 4573 4574
			if (err) {
				put_page(page);
				if (err == -EEXIST)
					goto retry;
				goto out;
			}
4575
		} else {
4576
			lock_page(page);
4577 4578 4579 4580
			if (unlikely(anon_vma_prepare(vma))) {
				ret = VM_FAULT_OOM;
				goto backout_unlocked;
			}
4581
			anon_rmap = 1;
4582
		}
4583
	} else {
4584 4585 4586 4587 4588 4589
		/*
		 * 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))) {
4590
			ret = VM_FAULT_HWPOISON_LARGE |
4591
				VM_FAULT_SET_HINDEX(hstate_index(h));
4592 4593
			goto backout_unlocked;
		}
4594
	}
4595

4596 4597 4598 4599 4600 4601
	/*
	 * 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.
	 */
4602
	if ((flags & FAULT_FLAG_WRITE) && !(vma->vm_flags & VM_SHARED)) {
4603
		if (vma_needs_reservation(h, vma, haddr) < 0) {
4604 4605 4606
			ret = VM_FAULT_OOM;
			goto backout_unlocked;
		}
4607
		/* Just decrements count, does not deallocate */
4608
		vma_end_reservation(h, vma, haddr);
4609
	}
4610

4611
	ptl = huge_pte_lock(h, mm, ptep);
N
Nick Piggin 已提交
4612
	ret = 0;
4613
	if (!huge_pte_none(huge_ptep_get(ptep)))
A
Adam Litke 已提交
4614 4615
		goto backout;

4616
	if (anon_rmap) {
4617
		ClearHPageRestoreReserve(page);
4618
		hugepage_add_new_anon_rmap(page, vma, haddr);
4619
	} else
4620
		page_dup_rmap(page, true);
4621 4622
	new_pte = make_huge_pte(vma, page, ((vma->vm_flags & VM_WRITE)
				&& (vma->vm_flags & VM_SHARED)));
4623
	set_huge_pte_at(mm, haddr, ptep, new_pte);
4624

4625
	hugetlb_count_add(pages_per_huge_page(h), mm);
4626
	if ((flags & FAULT_FLAG_WRITE) && !(vma->vm_flags & VM_SHARED)) {
4627
		/* Optimization, do the COW without a second fault */
4628
		ret = hugetlb_cow(mm, vma, address, ptep, page, ptl);
4629 4630
	}

4631
	spin_unlock(ptl);
4632 4633

	/*
4634 4635 4636
	 * Only set HPageMigratable in newly allocated pages.  Existing pages
	 * found in the pagecache may not have HPageMigratableset if they have
	 * been isolated for migration.
4637 4638
	 */
	if (new_page)
4639
		SetHPageMigratable(page);
4640

A
Adam Litke 已提交
4641 4642
	unlock_page(page);
out:
4643
	return ret;
A
Adam Litke 已提交
4644 4645

backout:
4646
	spin_unlock(ptl);
4647
backout_unlocked:
A
Adam Litke 已提交
4648
	unlock_page(page);
4649
	restore_reserve_on_error(h, vma, haddr, page);
A
Adam Litke 已提交
4650 4651
	put_page(page);
	goto out;
4652 4653
}

4654
#ifdef CONFIG_SMP
4655
u32 hugetlb_fault_mutex_hash(struct address_space *mapping, pgoff_t idx)
4656 4657 4658 4659
{
	unsigned long key[2];
	u32 hash;

4660 4661
	key[0] = (unsigned long) mapping;
	key[1] = idx;
4662

4663
	hash = jhash2((u32 *)&key, sizeof(key)/(sizeof(u32)), 0);
4664 4665 4666 4667 4668

	return hash & (num_fault_mutexes - 1);
}
#else
/*
M
Miaohe Lin 已提交
4669
 * For uniprocessor systems we always use a single mutex, so just
4670 4671
 * return 0 and avoid the hashing overhead.
 */
4672
u32 hugetlb_fault_mutex_hash(struct address_space *mapping, pgoff_t idx)
4673 4674 4675 4676 4677
{
	return 0;
}
#endif

4678
vm_fault_t hugetlb_fault(struct mm_struct *mm, struct vm_area_struct *vma,
4679
			unsigned long address, unsigned int flags)
4680
{
4681
	pte_t *ptep, entry;
4682
	spinlock_t *ptl;
4683
	vm_fault_t ret;
4684 4685
	u32 hash;
	pgoff_t idx;
4686
	struct page *page = NULL;
4687
	struct page *pagecache_page = NULL;
4688
	struct hstate *h = hstate_vma(vma);
4689
	struct address_space *mapping;
4690
	int need_wait_lock = 0;
4691
	unsigned long haddr = address & huge_page_mask(h);
4692

4693
	ptep = huge_pte_offset(mm, haddr, huge_page_size(h));
4694
	if (ptep) {
4695 4696 4697 4698 4699
		/*
		 * 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.
		 */
4700
		entry = huge_ptep_get(ptep);
N
Naoya Horiguchi 已提交
4701
		if (unlikely(is_hugetlb_entry_migration(entry))) {
4702
			migration_entry_wait_huge(vma, mm, ptep);
N
Naoya Horiguchi 已提交
4703 4704
			return 0;
		} else if (unlikely(is_hugetlb_entry_hwpoisoned(entry)))
4705
			return VM_FAULT_HWPOISON_LARGE |
4706
				VM_FAULT_SET_HINDEX(hstate_index(h));
4707 4708
	}

4709 4710
	/*
	 * Acquire i_mmap_rwsem before calling huge_pte_alloc and hold
4711 4712 4713 4714
	 * 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.
4715 4716 4717 4718 4719
	 *
	 * 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.
	 */
4720
	mapping = vma->vm_file->f_mapping;
4721
	i_mmap_lock_read(mapping);
4722
	ptep = huge_pte_alloc(mm, vma, haddr, huge_page_size(h));
4723 4724 4725 4726
	if (!ptep) {
		i_mmap_unlock_read(mapping);
		return VM_FAULT_OOM;
	}
4727

4728 4729 4730 4731 4732
	/*
	 * 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.
	 */
4733
	idx = vma_hugecache_offset(h, vma, haddr);
4734
	hash = hugetlb_fault_mutex_hash(mapping, idx);
4735
	mutex_lock(&hugetlb_fault_mutex_table[hash]);
4736

4737 4738
	entry = huge_ptep_get(ptep);
	if (huge_pte_none(entry)) {
4739
		ret = hugetlb_no_page(mm, vma, mapping, idx, address, ptep, flags);
4740
		goto out_mutex;
4741
	}
4742

N
Nick Piggin 已提交
4743
	ret = 0;
4744

4745 4746 4747
	/*
	 * 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
4748 4749 4750
	 * an active hugepage in pagecache. This goto expects the 2nd page
	 * fault, and is_hugetlb_entry_(migration|hwpoisoned) check will
	 * properly handle it.
4751 4752 4753 4754
	 */
	if (!pte_present(entry))
		goto out_mutex;

4755 4756 4757 4758 4759 4760 4761 4762
	/*
	 * 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.
	 */
4763
	if ((flags & FAULT_FLAG_WRITE) && !huge_pte_write(entry)) {
4764
		if (vma_needs_reservation(h, vma, haddr) < 0) {
4765
			ret = VM_FAULT_OOM;
4766
			goto out_mutex;
4767
		}
4768
		/* Just decrements count, does not deallocate */
4769
		vma_end_reservation(h, vma, haddr);
4770

4771
		if (!(vma->vm_flags & VM_MAYSHARE))
4772
			pagecache_page = hugetlbfs_pagecache_page(h,
4773
								vma, haddr);
4774 4775
	}

4776 4777 4778 4779 4780 4781
	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;

4782 4783 4784 4785 4786 4787 4788
	/*
	 * 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)
4789 4790 4791 4792
		if (!trylock_page(page)) {
			need_wait_lock = 1;
			goto out_ptl;
		}
4793

4794
	get_page(page);
4795

4796
	if (flags & FAULT_FLAG_WRITE) {
4797
		if (!huge_pte_write(entry)) {
4798
			ret = hugetlb_cow(mm, vma, address, ptep,
4799
					  pagecache_page, ptl);
4800
			goto out_put_page;
4801
		}
4802
		entry = huge_pte_mkdirty(entry);
4803 4804
	}
	entry = pte_mkyoung(entry);
4805
	if (huge_ptep_set_access_flags(vma, haddr, ptep, entry,
4806
						flags & FAULT_FLAG_WRITE))
4807
		update_mmu_cache(vma, haddr, ptep);
4808 4809 4810 4811
out_put_page:
	if (page != pagecache_page)
		unlock_page(page);
	put_page(page);
4812 4813
out_ptl:
	spin_unlock(ptl);
4814 4815 4816 4817 4818

	if (pagecache_page) {
		unlock_page(pagecache_page);
		put_page(pagecache_page);
	}
4819
out_mutex:
4820
	mutex_unlock(&hugetlb_fault_mutex_table[hash]);
4821
	i_mmap_unlock_read(mapping);
4822 4823 4824 4825 4826 4827 4828 4829 4830
	/*
	 * 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);
4831
	return ret;
4832 4833
}

4834 4835 4836 4837 4838 4839 4840 4841 4842 4843 4844
/*
 * 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)
{
4845 4846 4847
	struct address_space *mapping;
	pgoff_t idx;
	unsigned long size;
4848
	int vm_shared = dst_vma->vm_flags & VM_SHARED;
4849 4850 4851 4852 4853 4854 4855 4856 4857 4858 4859 4860 4861 4862
	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,
4863
						pages_per_huge_page(h), false);
4864

4865
		/* fallback to copy_from_user outside mmap_lock */
4866
		if (unlikely(ret)) {
4867
			ret = -ENOENT;
4868 4869 4870 4871 4872 4873 4874 4875 4876 4877 4878 4879 4880 4881 4882 4883
			*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);

4884 4885 4886
	mapping = dst_vma->vm_file->f_mapping;
	idx = vma_hugecache_offset(h, dst_vma, dst_addr);

4887 4888 4889 4890
	/*
	 * If shared, add to page cache
	 */
	if (vm_shared) {
4891 4892 4893 4894
		size = i_size_read(mapping->host) >> huge_page_shift(h);
		ret = -EFAULT;
		if (idx >= size)
			goto out_release_nounlock;
4895

4896 4897 4898 4899 4900 4901
		/*
		 * 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.
		 */
4902 4903 4904 4905 4906
		ret = huge_add_to_page_cache(page, mapping, idx);
		if (ret)
			goto out_release_nounlock;
	}

4907 4908 4909
	ptl = huge_pte_lockptr(h, dst_mm, dst_pte);
	spin_lock(ptl);

4910 4911 4912 4913 4914 4915 4916 4917 4918 4919 4920 4921 4922 4923
	/*
	 * 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;

4924 4925 4926 4927
	ret = -EEXIST;
	if (!huge_pte_none(huge_ptep_get(dst_pte)))
		goto out_release_unlock;

4928 4929 4930
	if (vm_shared) {
		page_dup_rmap(page, true);
	} else {
4931
		ClearHPageRestoreReserve(page);
4932 4933
		hugepage_add_new_anon_rmap(page, dst_vma, dst_addr);
	}
4934 4935 4936 4937 4938 4939 4940 4941 4942 4943 4944 4945 4946 4947 4948 4949

	_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);
4950
	SetHPageMigratable(page);
4951 4952
	if (vm_shared)
		unlock_page(page);
4953 4954 4955 4956 4957
	ret = 0;
out:
	return ret;
out_release_unlock:
	spin_unlock(ptl);
4958 4959
	if (vm_shared)
		unlock_page(page);
4960
out_release_nounlock:
4961 4962 4963 4964
	put_page(page);
	goto out;
}

4965 4966 4967 4968 4969 4970 4971 4972 4973 4974 4975 4976 4977 4978
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;
	}
}

4979 4980 4981
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,
4982
			 long i, unsigned int flags, int *locked)
D
David Gibson 已提交
4983
{
4984 4985
	unsigned long pfn_offset;
	unsigned long vaddr = *position;
4986
	unsigned long remainder = *nr_pages;
4987
	struct hstate *h = hstate_vma(vma);
4988
	int err = -EFAULT, refs;
D
David Gibson 已提交
4989 4990

	while (vaddr < vma->vm_end && remainder) {
A
Adam Litke 已提交
4991
		pte_t *pte;
4992
		spinlock_t *ptl = NULL;
H
Hugh Dickins 已提交
4993
		int absent;
A
Adam Litke 已提交
4994
		struct page *page;
D
David Gibson 已提交
4995

4996 4997 4998 4999
		/*
		 * If we have a pending SIGKILL, don't keep faulting pages and
		 * potentially allocating memory.
		 */
5000
		if (fatal_signal_pending(current)) {
5001 5002 5003 5004
			remainder = 0;
			break;
		}

A
Adam Litke 已提交
5005 5006
		/*
		 * Some archs (sparc64, sh*) have multiple pte_ts to
H
Hugh Dickins 已提交
5007
		 * each hugepage.  We have to make sure we get the
A
Adam Litke 已提交
5008
		 * first, for the page indexing below to work.
5009 5010
		 *
		 * Note that page table lock is not held when pte is null.
A
Adam Litke 已提交
5011
		 */
5012 5013
		pte = huge_pte_offset(mm, vaddr & huge_page_mask(h),
				      huge_page_size(h));
5014 5015
		if (pte)
			ptl = huge_pte_lock(h, mm, pte);
H
Hugh Dickins 已提交
5016 5017 5018 5019
		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
5020 5021 5022 5023
		 * 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
5024
		 */
H
Hugh Dickins 已提交
5025 5026
		if (absent && (flags & FOLL_DUMP) &&
		    !hugetlbfs_pagecache_present(h, vma, vaddr)) {
5027 5028
			if (pte)
				spin_unlock(ptl);
H
Hugh Dickins 已提交
5029 5030 5031
			remainder = 0;
			break;
		}
D
David Gibson 已提交
5032

5033 5034 5035 5036 5037 5038 5039 5040 5041 5042 5043
		/*
		 * 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)) ||
5044 5045
		    ((flags & FOLL_WRITE) &&
		      !huge_pte_write(huge_ptep_get(pte)))) {
5046
			vm_fault_t ret;
5047
			unsigned int fault_flags = 0;
D
David Gibson 已提交
5048

5049 5050
			if (pte)
				spin_unlock(ptl);
5051 5052
			if (flags & FOLL_WRITE)
				fault_flags |= FAULT_FLAG_WRITE;
5053
			if (locked)
5054 5055
				fault_flags |= FAULT_FLAG_ALLOW_RETRY |
					FAULT_FLAG_KILLABLE;
5056 5057 5058 5059
			if (flags & FOLL_NOWAIT)
				fault_flags |= FAULT_FLAG_ALLOW_RETRY |
					FAULT_FLAG_RETRY_NOWAIT;
			if (flags & FOLL_TRIED) {
5060 5061 5062 5063
				/*
				 * Note: FAULT_FLAG_ALLOW_RETRY and
				 * FAULT_FLAG_TRIED can co-exist
				 */
5064 5065 5066 5067
				fault_flags |= FAULT_FLAG_TRIED;
			}
			ret = hugetlb_fault(mm, vma, vaddr, fault_flags);
			if (ret & VM_FAULT_ERROR) {
5068
				err = vm_fault_to_errno(ret, flags);
5069 5070 5071 5072
				remainder = 0;
				break;
			}
			if (ret & VM_FAULT_RETRY) {
5073
				if (locked &&
5074
				    !(fault_flags & FAULT_FLAG_RETRY_NOWAIT))
5075
					*locked = 0;
5076 5077 5078 5079 5080 5081 5082 5083 5084 5085 5086 5087 5088
				*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
5089 5090
		}

5091
		pfn_offset = (vaddr & ~huge_page_mask(h)) >> PAGE_SHIFT;
5092
		page = pte_page(huge_ptep_get(pte));
5093

5094 5095 5096 5097 5098 5099 5100 5101 5102 5103 5104 5105 5106 5107
		/*
		 * 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;
		}

5108 5109
		refs = min3(pages_per_huge_page(h) - pfn_offset,
			    (vma->vm_end - vaddr) >> PAGE_SHIFT, remainder);
5110

5111 5112 5113 5114 5115
		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
5116

5117
		if (pages) {
5118 5119 5120 5121 5122 5123 5124 5125 5126 5127
			/*
			 * 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:
			 */
5128
			if (WARN_ON_ONCE(!try_grab_compound_head(pages[i],
5129 5130 5131 5132 5133 5134 5135
								 refs,
								 flags))) {
				spin_unlock(ptl);
				remainder = 0;
				err = -ENOMEM;
				break;
			}
5136
		}
5137 5138 5139 5140 5141

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

5142
		spin_unlock(ptl);
D
David Gibson 已提交
5143
	}
5144
	*nr_pages = remainder;
5145 5146 5147 5148 5149
	/*
	 * 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
5150 5151
	*position = vaddr;

5152
	return i ? i : err;
D
David Gibson 已提交
5153
}
5154

5155
unsigned long hugetlb_change_protection(struct vm_area_struct *vma,
5156 5157 5158 5159 5160 5161
		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;
5162
	struct hstate *h = hstate_vma(vma);
5163
	unsigned long pages = 0;
5164
	bool shared_pmd = false;
5165
	struct mmu_notifier_range range;
5166 5167 5168

	/*
	 * In the case of shared PMDs, the area to flush could be beyond
5169
	 * start/end.  Set range.start/range.end to cover the maximum possible
5170 5171
	 * range if PMD sharing is possible.
	 */
5172 5173
	mmu_notifier_range_init(&range, MMU_NOTIFY_PROTECTION_VMA,
				0, vma, mm, start, end);
5174
	adjust_range_if_pmd_sharing_possible(vma, &range.start, &range.end);
5175 5176

	BUG_ON(address >= end);
5177
	flush_cache_range(vma, range.start, range.end);
5178

5179
	mmu_notifier_invalidate_range_start(&range);
5180
	i_mmap_lock_write(vma->vm_file->f_mapping);
5181
	for (; address < end; address += huge_page_size(h)) {
5182
		spinlock_t *ptl;
5183
		ptep = huge_pte_offset(mm, address, huge_page_size(h));
5184 5185
		if (!ptep)
			continue;
5186
		ptl = huge_pte_lock(h, mm, ptep);
5187
		if (huge_pmd_unshare(mm, vma, &address, ptep)) {
5188
			pages++;
5189
			spin_unlock(ptl);
5190
			shared_pmd = true;
5191
			continue;
5192
		}
5193 5194 5195 5196 5197 5198 5199 5200 5201 5202 5203 5204 5205
		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);
5206 5207
				set_huge_swap_pte_at(mm, address, ptep,
						     newpte, huge_page_size(h));
5208 5209 5210 5211 5212 5213
				pages++;
			}
			spin_unlock(ptl);
			continue;
		}
		if (!huge_pte_none(pte)) {
5214 5215 5216 5217
			pte_t old_pte;

			old_pte = huge_ptep_modify_prot_start(vma, address, ptep);
			pte = pte_mkhuge(huge_pte_modify(old_pte, newprot));
5218
			pte = arch_make_huge_pte(pte, vma, NULL, 0);
5219
			huge_ptep_modify_prot_commit(vma, address, ptep, old_pte, pte);
5220
			pages++;
5221
		}
5222
		spin_unlock(ptl);
5223
	}
5224
	/*
5225
	 * Must flush TLB before releasing i_mmap_rwsem: x86's huge_pmd_unshare
5226
	 * may have cleared our pud entry and done put_page on the page table:
5227
	 * once we release i_mmap_rwsem, another task can do the final put_page
5228 5229
	 * 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.
5230
	 */
5231
	if (shared_pmd)
5232
		flush_hugetlb_tlb_range(vma, range.start, range.end);
5233 5234
	else
		flush_hugetlb_tlb_range(vma, start, end);
5235 5236 5237 5238
	/*
	 * No need to call mmu_notifier_invalidate_range() we are downgrading
	 * page table protection not changing it to point to a new page.
	 *
5239
	 * See Documentation/vm/mmu_notifier.rst
5240
	 */
5241
	i_mmap_unlock_write(vma->vm_file->f_mapping);
5242
	mmu_notifier_invalidate_range_end(&range);
5243 5244

	return pages << h->order;
5245 5246
}

5247 5248
/* Return true if reservation was successful, false otherwise.  */
bool hugetlb_reserve_pages(struct inode *inode,
5249
					long from, long to,
5250
					struct vm_area_struct *vma,
5251
					vm_flags_t vm_flags)
5252
{
5253
	long chg, add = -1;
5254
	struct hstate *h = hstate_inode(inode);
5255
	struct hugepage_subpool *spool = subpool_inode(inode);
5256
	struct resv_map *resv_map;
5257
	struct hugetlb_cgroup *h_cg = NULL;
5258
	long gbl_reserve, regions_needed = 0;
5259

5260 5261 5262
	/* This should never happen */
	if (from > to) {
		VM_WARN(1, "%s called with a negative range\n", __func__);
5263
		return false;
5264 5265
	}

5266 5267 5268
	/*
	 * Only apply hugepage reservation if asked. At fault time, an
	 * attempt will be made for VM_NORESERVE to allocate a page
5269
	 * without using reserves
5270
	 */
5271
	if (vm_flags & VM_NORESERVE)
5272
		return true;
5273

5274 5275 5276 5277 5278 5279
	/*
	 * 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
	 */
5280
	if (!vma || vma->vm_flags & VM_MAYSHARE) {
5281 5282 5283 5284 5285
		/*
		 * 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).
		 */
5286
		resv_map = inode_resv_map(inode);
5287

5288
		chg = region_chg(resv_map, from, to, &regions_needed);
5289 5290

	} else {
5291
		/* Private mapping. */
5292
		resv_map = resv_map_alloc();
5293
		if (!resv_map)
5294
			return false;
5295

5296
		chg = to - from;
5297

5298 5299 5300 5301
		set_vma_resv_map(vma, resv_map);
		set_vma_resv_flags(vma, HPAGE_RESV_OWNER);
	}

5302
	if (chg < 0)
5303
		goto out_err;
5304

5305 5306
	if (hugetlb_cgroup_charge_cgroup_rsvd(hstate_index(h),
				chg * pages_per_huge_page(h), &h_cg) < 0)
5307 5308 5309 5310 5311 5312 5313 5314 5315
		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);
	}

5316 5317 5318 5319 5320 5321
	/*
	 * 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);
5322
	if (gbl_reserve < 0)
5323
		goto out_uncharge_cgroup;
5324 5325

	/*
5326
	 * Check enough hugepages are available for the reservation.
5327
	 * Hand the pages back to the subpool if there are not
5328
	 */
5329
	if (hugetlb_acct_memory(h, gbl_reserve) < 0)
5330
		goto out_put_pages;
5331 5332 5333 5334 5335 5336 5337 5338 5339 5340 5341 5342

	/*
	 * 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
	 */
5343
	if (!vma || vma->vm_flags & VM_MAYSHARE) {
5344
		add = region_add(resv_map, from, to, regions_needed, h, h_cg);
5345 5346 5347

		if (unlikely(add < 0)) {
			hugetlb_acct_memory(h, -gbl_reserve);
5348
			goto out_put_pages;
5349
		} else if (unlikely(chg > add)) {
5350 5351 5352 5353 5354 5355 5356 5357 5358
			/*
			 * 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;

5359 5360 5361 5362
			/*
			 * hugetlb_cgroup_uncharge_cgroup_rsvd() will put the
			 * reference to h_cg->css. See comment below for detail.
			 */
5363 5364 5365 5366
			hugetlb_cgroup_uncharge_cgroup_rsvd(
				hstate_index(h),
				(chg - add) * pages_per_huge_page(h), h_cg);

5367 5368 5369
			rsv_adjust = hugepage_subpool_put_pages(spool,
								chg - add);
			hugetlb_acct_memory(h, -rsv_adjust);
5370 5371 5372 5373 5374 5375 5376 5377
		} else if (h_cg) {
			/*
			 * The file_regions will hold their own reference to
			 * h_cg->css. So we should release the reference held
			 * via hugetlb_cgroup_charge_cgroup_rsvd() when we are
			 * done.
			 */
			hugetlb_cgroup_put_rsvd_cgroup(h_cg);
5378 5379
		}
	}
5380 5381
	return true;

5382 5383 5384 5385 5386 5387
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);
5388
out_err:
5389
	if (!vma || vma->vm_flags & VM_MAYSHARE)
5390 5391 5392 5393 5394
		/* 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
5395 5396
	if (vma && is_vma_resv_set(vma, HPAGE_RESV_OWNER))
		kref_put(&resv_map->refs, resv_map_release);
5397
	return false;
5398 5399
}

5400 5401
long hugetlb_unreserve_pages(struct inode *inode, long start, long end,
								long freed)
5402
{
5403
	struct hstate *h = hstate_inode(inode);
5404
	struct resv_map *resv_map = inode_resv_map(inode);
5405
	long chg = 0;
5406
	struct hugepage_subpool *spool = subpool_inode(inode);
5407
	long gbl_reserve;
K
Ken Chen 已提交
5408

5409 5410 5411 5412
	/*
	 * Since this routine can be called in the evict inode path for all
	 * hugetlbfs inodes, resv_map could be NULL.
	 */
5413 5414 5415 5416 5417 5418 5419 5420 5421 5422 5423
	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
5424
	spin_lock(&inode->i_lock);
5425
	inode->i_blocks -= (blocks_per_huge_page(h) * freed);
K
Ken Chen 已提交
5426 5427
	spin_unlock(&inode->i_lock);

5428 5429 5430
	/*
	 * If the subpool has a minimum size, the number of global
	 * reservations to be released may be adjusted.
5431 5432 5433
	 *
	 * Note that !resv_map implies freed == 0. So (chg - freed)
	 * won't go negative.
5434 5435 5436
	 */
	gbl_reserve = hugepage_subpool_put_pages(spool, (chg - freed));
	hugetlb_acct_memory(h, -gbl_reserve);
5437 5438

	return 0;
5439
}
5440

5441 5442 5443 5444 5445 5446 5447 5448 5449 5450 5451
#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
5452 5453
	unsigned long vm_flags = vma->vm_flags & VM_LOCKED_CLEAR_MASK;
	unsigned long svm_flags = svma->vm_flags & VM_LOCKED_CLEAR_MASK;
5454 5455 5456 5457 5458 5459 5460

	/*
	 * match the virtual addresses, permission and the alignment of the
	 * page table page.
	 */
	if (pmd_index(addr) != pmd_index(saddr) ||
	    vm_flags != svm_flags ||
5461
	    !range_in_vma(svma, sbase, s_end))
5462 5463 5464 5465 5466
		return 0;

	return saddr;
}

5467
static bool vma_shareable(struct vm_area_struct *vma, unsigned long addr)
5468 5469 5470 5471 5472 5473 5474
{
	unsigned long base = addr & PUD_MASK;
	unsigned long end = base + PUD_SIZE;

	/*
	 * check on proper vm_flags and page table alignment
	 */
5475
	if (vma->vm_flags & VM_MAYSHARE && range_in_vma(vma, base, end))
5476 5477
		return true;
	return false;
5478 5479
}

5480 5481 5482 5483 5484 5485 5486 5487 5488
bool want_pmd_share(struct vm_area_struct *vma, unsigned long addr)
{
#ifdef CONFIG_USERFAULTFD
	if (uffd_disable_huge_pmd_share(vma))
		return false;
#endif
	return vma_shareable(vma, addr);
}

5489 5490 5491 5492 5493 5494 5495 5496
/*
 * 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)
{
5497 5498
	unsigned long v_start = ALIGN(vma->vm_start, PUD_SIZE),
		v_end = ALIGN_DOWN(vma->vm_end, PUD_SIZE);
5499

5500 5501 5502 5503 5504 5505
	/*
	 * 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))
5506 5507
		return;

5508
	/* Extend the range to be PUD aligned for a worst case scenario */
5509 5510
	if (*start > v_start)
		*start = ALIGN_DOWN(*start, PUD_SIZE);
5511

5512 5513
	if (*end < v_end)
		*end = ALIGN(*end, PUD_SIZE);
5514 5515
}

5516 5517 5518 5519
/*
 * 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
5520 5521
 * code much cleaner.
 *
5522 5523 5524 5525 5526 5527 5528 5529 5530 5531
 * 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.
5532
 */
5533 5534
pte_t *huge_pmd_share(struct mm_struct *mm, struct vm_area_struct *vma,
		      unsigned long addr, pud_t *pud)
5535 5536 5537 5538 5539 5540 5541 5542
{
	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;
5543
	spinlock_t *ptl;
5544

5545
	i_mmap_assert_locked(mapping);
5546 5547 5548 5549 5550 5551
	vma_interval_tree_foreach(svma, &mapping->i_mmap, idx, idx) {
		if (svma == vma)
			continue;

		saddr = page_table_shareable(svma, vma, addr, idx);
		if (saddr) {
5552 5553
			spte = huge_pte_offset(svma->vm_mm, saddr,
					       vma_mmu_pagesize(svma));
5554 5555 5556 5557 5558 5559 5560 5561 5562 5563
			if (spte) {
				get_page(virt_to_page(spte));
				break;
			}
		}
	}

	if (!spte)
		goto out;

5564
	ptl = huge_pte_lock(hstate_vma(vma), mm, spte);
5565
	if (pud_none(*pud)) {
5566 5567
		pud_populate(mm, pud,
				(pmd_t *)((unsigned long)spte & PAGE_MASK));
5568
		mm_inc_nr_pmds(mm);
5569
	} else {
5570
		put_page(virt_to_page(spte));
5571
	}
5572
	spin_unlock(ptl);
5573 5574 5575 5576 5577 5578 5579 5580 5581 5582 5583 5584
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.
 *
5585
 * Called with page table lock held and i_mmap_rwsem held in write mode.
5586 5587 5588 5589
 *
 * returns: 1 successfully unmapped a shared pte page
 *	    0 the underlying pte page is not shared, or it is the last user
 */
5590 5591
int huge_pmd_unshare(struct mm_struct *mm, struct vm_area_struct *vma,
					unsigned long *addr, pte_t *ptep)
5592 5593
{
	pgd_t *pgd = pgd_offset(mm, *addr);
5594 5595
	p4d_t *p4d = p4d_offset(pgd, *addr);
	pud_t *pud = pud_offset(p4d, *addr);
5596

5597
	i_mmap_assert_write_locked(vma->vm_file->f_mapping);
5598 5599 5600 5601 5602 5603
	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));
5604
	mm_dec_nr_pmds(mm);
5605 5606 5607
	*addr = ALIGN(*addr, HPAGE_SIZE * PTRS_PER_PTE) - HPAGE_SIZE;
	return 1;
}
5608

5609
#else /* !CONFIG_ARCH_WANT_HUGE_PMD_SHARE */
5610 5611
pte_t *huge_pmd_share(struct mm_struct *mm, struct vm_area_struct *vma,
		      unsigned long addr, pud_t *pud)
5612 5613 5614
{
	return NULL;
}
5615

5616 5617
int huge_pmd_unshare(struct mm_struct *mm, struct vm_area_struct *vma,
				unsigned long *addr, pte_t *ptep)
5618 5619 5620
{
	return 0;
}
5621 5622 5623 5624 5625

void adjust_range_if_pmd_sharing_possible(struct vm_area_struct *vma,
				unsigned long *start, unsigned long *end)
{
}
5626 5627 5628 5629 5630

bool want_pmd_share(struct vm_area_struct *vma, unsigned long addr)
{
	return false;
}
5631 5632
#endif /* CONFIG_ARCH_WANT_HUGE_PMD_SHARE */

5633
#ifdef CONFIG_ARCH_WANT_GENERAL_HUGETLB
5634
pte_t *huge_pte_alloc(struct mm_struct *mm, struct vm_area_struct *vma,
5635 5636 5637
			unsigned long addr, unsigned long sz)
{
	pgd_t *pgd;
5638
	p4d_t *p4d;
5639 5640 5641 5642
	pud_t *pud;
	pte_t *pte = NULL;

	pgd = pgd_offset(mm, addr);
5643 5644 5645
	p4d = p4d_alloc(mm, pgd, addr);
	if (!p4d)
		return NULL;
5646
	pud = pud_alloc(mm, p4d, addr);
5647 5648 5649 5650 5651
	if (pud) {
		if (sz == PUD_SIZE) {
			pte = (pte_t *)pud;
		} else {
			BUG_ON(sz != PMD_SIZE);
5652
			if (want_pmd_share(vma, addr) && pud_none(*pud))
5653
				pte = huge_pmd_share(mm, vma, addr, pud);
5654 5655 5656 5657
			else
				pte = (pte_t *)pmd_alloc(mm, pud, addr);
		}
	}
5658
	BUG_ON(pte && pte_present(*pte) && !pte_huge(*pte));
5659 5660 5661 5662

	return pte;
}

5663 5664 5665 5666
/*
 * huge_pte_offset() - Walk the page table to resolve the hugepage
 * entry at address @addr
 *
5667 5668
 * Return: Pointer to page table entry (PUD or PMD) for
 * address @addr, or NULL if a !p*d_present() entry is encountered and the
5669 5670 5671
 * size @sz doesn't match the hugepage size at this level of the page
 * table.
 */
5672 5673
pte_t *huge_pte_offset(struct mm_struct *mm,
		       unsigned long addr, unsigned long sz)
5674 5675
{
	pgd_t *pgd;
5676
	p4d_t *p4d;
5677 5678
	pud_t *pud;
	pmd_t *pmd;
5679 5680

	pgd = pgd_offset(mm, addr);
5681 5682 5683 5684 5685
	if (!pgd_present(*pgd))
		return NULL;
	p4d = p4d_offset(pgd, addr);
	if (!p4d_present(*p4d))
		return NULL;
5686

5687
	pud = pud_offset(p4d, addr);
5688 5689
	if (sz == PUD_SIZE)
		/* must be pud huge, non-present or none */
5690
		return (pte_t *)pud;
5691
	if (!pud_present(*pud))
5692
		return NULL;
5693
	/* must have a valid entry and size to go further */
5694

5695 5696 5697
	pmd = pmd_offset(pud, addr);
	/* must be pmd huge, non-present or none */
	return (pte_t *)pmd;
5698 5699
}

5700 5701 5702 5703 5704 5705 5706 5707 5708 5709 5710 5711 5712
#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);
}

5713 5714 5715 5716 5717 5718 5719 5720
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;
}

5721
struct page * __weak
5722
follow_huge_pmd(struct mm_struct *mm, unsigned long address,
5723
		pmd_t *pmd, int flags)
5724
{
5725 5726
	struct page *page = NULL;
	spinlock_t *ptl;
5727
	pte_t pte;
J
John Hubbard 已提交
5728 5729 5730 5731 5732 5733

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

5734 5735 5736 5737 5738 5739 5740 5741 5742
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;
5743 5744
	pte = huge_ptep_get((pte_t *)pmd);
	if (pte_present(pte)) {
5745
		page = pmd_page(*pmd) + ((address & ~PMD_MASK) >> PAGE_SHIFT);
J
John Hubbard 已提交
5746 5747 5748 5749 5750 5751 5752 5753 5754 5755 5756 5757
		/*
		 * 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;
		}
5758
	} else {
5759
		if (is_hugetlb_entry_migration(pte)) {
5760 5761 5762 5763 5764 5765 5766 5767 5768 5769 5770
			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);
5771 5772 5773
	return page;
}

5774
struct page * __weak
5775
follow_huge_pud(struct mm_struct *mm, unsigned long address,
5776
		pud_t *pud, int flags)
5777
{
J
John Hubbard 已提交
5778
	if (flags & (FOLL_GET | FOLL_PIN))
5779
		return NULL;
5780

5781
	return pte_page(*(pte_t *)pud) + ((address & ~PUD_MASK) >> PAGE_SHIFT);
5782 5783
}

5784 5785 5786
struct page * __weak
follow_huge_pgd(struct mm_struct *mm, unsigned long address, pgd_t *pgd, int flags)
{
J
John Hubbard 已提交
5787
	if (flags & (FOLL_GET | FOLL_PIN))
5788 5789 5790 5791 5792
		return NULL;

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

5793 5794
bool isolate_huge_page(struct page *page, struct list_head *list)
{
5795 5796
	bool ret = true;

5797
	spin_lock_irq(&hugetlb_lock);
5798 5799
	if (!PageHeadHuge(page) ||
	    !HPageMigratable(page) ||
5800
	    !get_page_unless_zero(page)) {
5801 5802 5803
		ret = false;
		goto unlock;
	}
5804
	ClearHPageMigratable(page);
5805
	list_move_tail(&page->lru, list);
5806
unlock:
5807
	spin_unlock_irq(&hugetlb_lock);
5808
	return ret;
5809 5810 5811 5812
}

void putback_active_hugepage(struct page *page)
{
5813
	spin_lock_irq(&hugetlb_lock);
5814
	SetHPageMigratable(page);
5815
	list_move_tail(&page->lru, &(page_hstate(page))->hugepage_activelist);
5816
	spin_unlock_irq(&hugetlb_lock);
5817 5818
	put_page(page);
}
5819 5820 5821 5822 5823 5824 5825 5826 5827 5828 5829 5830 5831 5832 5833 5834 5835 5836

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.
	 */
5837
	if (HPageTemporary(newpage)) {
5838 5839 5840
		int old_nid = page_to_nid(oldpage);
		int new_nid = page_to_nid(newpage);

5841 5842
		SetHPageTemporary(oldpage);
		ClearHPageTemporary(newpage);
5843

5844 5845 5846 5847 5848 5849
		/*
		 * There is no need to transfer the per-node surplus state
		 * when we do not cross the node.
		 */
		if (new_nid == old_nid)
			return;
5850
		spin_lock_irq(&hugetlb_lock);
5851 5852 5853 5854
		if (h->surplus_huge_pages_node[old_nid]) {
			h->surplus_huge_pages_node[old_nid]--;
			h->surplus_huge_pages_node[new_nid]++;
		}
5855
		spin_unlock_irq(&hugetlb_lock);
5856 5857
	}
}
5858

5859 5860 5861 5862 5863 5864 5865 5866 5867 5868 5869 5870 5871 5872 5873 5874 5875 5876 5877 5878 5879 5880 5881 5882 5883 5884 5885 5886 5887 5888 5889 5890 5891 5892 5893 5894 5895 5896 5897 5898 5899 5900 5901 5902 5903 5904 5905 5906 5907 5908 5909
/*
 * This function will unconditionally remove all the shared pmd pgtable entries
 * within the specific vma for a hugetlbfs memory range.
 */
void hugetlb_unshare_all_pmds(struct vm_area_struct *vma)
{
	struct hstate *h = hstate_vma(vma);
	unsigned long sz = huge_page_size(h);
	struct mm_struct *mm = vma->vm_mm;
	struct mmu_notifier_range range;
	unsigned long address, start, end;
	spinlock_t *ptl;
	pte_t *ptep;

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

	start = ALIGN(vma->vm_start, PUD_SIZE);
	end = ALIGN_DOWN(vma->vm_end, PUD_SIZE);

	if (start >= end)
		return;

	/*
	 * No need to call adjust_range_if_pmd_sharing_possible(), because
	 * we have already done the PUD_SIZE alignment.
	 */
	mmu_notifier_range_init(&range, MMU_NOTIFY_CLEAR, 0, vma, mm,
				start, end);
	mmu_notifier_invalidate_range_start(&range);
	i_mmap_lock_write(vma->vm_file->f_mapping);
	for (address = start; address < end; address += PUD_SIZE) {
		unsigned long tmp = address;

		ptep = huge_pte_offset(mm, address, sz);
		if (!ptep)
			continue;
		ptl = huge_pte_lock(h, mm, ptep);
		/* We don't want 'address' to be changed */
		huge_pmd_unshare(mm, vma, &tmp, ptep);
		spin_unlock(ptl);
	}
	flush_hugetlb_tlb_range(vma, start, end);
	i_mmap_unlock_write(vma->vm_file->f_mapping);
	/*
	 * No need to call mmu_notifier_invalidate_range(), see
	 * Documentation/vm/mmu_notifier.rst.
	 */
	mmu_notifier_invalidate_range_end(&range);
}

5910 5911 5912 5913 5914 5915 5916 5917 5918 5919 5920 5921 5922 5923 5924 5925 5926 5927 5928 5929 5930 5931 5932 5933 5934 5935 5936 5937 5938 5939 5940 5941 5942 5943 5944 5945 5946 5947
#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;
5948
		char name[CMA_MAX_NAME];
5949 5950 5951 5952

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

5953
		snprintf(name, sizeof(name), "hugetlb%d", nid);
5954
		res = cma_declare_contiguous_nid(0, size, 0, PAGE_SIZE << order,
5955
						 0, false, name,
5956 5957 5958 5959 5960 5961 5962 5963 5964 5965 5966 5967 5968 5969 5970 5971 5972 5973 5974 5975 5976 5977 5978 5979 5980
						 &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 */