hugetlb.c 157.8 KB
Newer Older
1
// SPDX-License-Identifier: GPL-2.0-only
L
Linus Torvalds 已提交
2 3
/*
 * Generic hugetlb support.
4
 * (C) Nadia Yvette Chambers, April 2004
L
Linus Torvalds 已提交
5 6 7 8
 */
#include <linux/list.h>
#include <linux/init.h>
#include <linux/mm.h>
9
#include <linux/seq_file.h>
L
Linus Torvalds 已提交
10 11
#include <linux/sysctl.h>
#include <linux/highmem.h>
A
Andrea Arcangeli 已提交
12
#include <linux/mmu_notifier.h>
L
Linus Torvalds 已提交
13
#include <linux/nodemask.h>
D
David Gibson 已提交
14
#include <linux/pagemap.h>
15
#include <linux/mempolicy.h>
16
#include <linux/compiler.h>
17
#include <linux/cpuset.h>
18
#include <linux/mutex.h>
19
#include <linux/memblock.h>
20
#include <linux/sysfs.h>
21
#include <linux/slab.h>
22
#include <linux/sched/mm.h>
23
#include <linux/mmdebug.h>
24
#include <linux/sched/signal.h>
25
#include <linux/rmap.h>
26
#include <linux/string_helpers.h>
27 28
#include <linux/swap.h>
#include <linux/swapops.h>
29
#include <linux/jhash.h>
30
#include <linux/numa.h>
31
#include <linux/llist.h>
32
#include <linux/cma.h>
33

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

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

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

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

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

61 62
__initdata LIST_HEAD(huge_boot_pages);

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

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

75 76 77 78 79
/*
 * Serializes faults on the same logical page.  This is used to
 * prevent spurious OOMs when the hugepage pool is fully utilized.
 */
static int num_fault_mutexes;
80
struct mutex *hugetlb_fault_mutex_table ____cacheline_aligned_in_smp;
81

82 83 84
/* Forward declaration */
static int hugetlb_acct_memory(struct hstate *h, long delta);

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

	return true;
}
96

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

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

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

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

	spin_lock_init(&spool->lock);
	spool->count = 1;
123 124 125 126 127 128 129 130 131
	spool->max_hpages = max_hpages;
	spool->hstate = h;
	spool->min_hpages = min_hpages;

	if (min_hpages != -1 && hugetlb_acct_memory(h, min_hpages)) {
		kfree(spool);
		return NULL;
	}
	spool->rsv_hpages = min_hpages;
132 133 134 135 136 137 138 139 140 141 142 143

	return spool;
}

void hugepage_put_subpool(struct hugepage_subpool *spool)
{
	spin_lock(&spool->lock);
	BUG_ON(!spool->count);
	spool->count--;
	unlock_or_release_subpool(spool);
}

144 145 146
/*
 * Subpool accounting for allocating and reserving pages.
 * Return -ENOMEM if there are not enough resources to satisfy the
147
 * request.  Otherwise, return the number of pages by which the
148 149
 * global pools must be adjusted (upward).  The returned value may
 * only be different than the passed value (delta) in the case where
E
Ethon Paul 已提交
150
 * a subpool minimum size must be maintained.
151 152
 */
static long hugepage_subpool_get_pages(struct hugepage_subpool *spool,
153 154
				      long delta)
{
155
	long ret = delta;
156 157

	if (!spool)
158
		return ret;
159 160

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

	if (spool->max_hpages != -1) {		/* maximum size accounting */
		if ((spool->used_hpages + delta) <= spool->max_hpages)
			spool->used_hpages += delta;
		else {
			ret = -ENOMEM;
			goto unlock_ret;
		}
169 170
	}

171 172
	/* minimum size accounting */
	if (spool->min_hpages != -1 && spool->rsv_hpages) {
173 174 175 176 177 178 179 180 181 182 183 184 185 186 187
		if (delta > spool->rsv_hpages) {
			/*
			 * Asking for more reserves than those already taken on
			 * behalf of subpool.  Return difference.
			 */
			ret = delta - spool->rsv_hpages;
			spool->rsv_hpages = 0;
		} else {
			ret = 0;	/* reserves already accounted for */
			spool->rsv_hpages -= delta;
		}
	}

unlock_ret:
	spin_unlock(&spool->lock);
188 189 190
	return ret;
}

191 192 193 194 195 196 197
/*
 * Subpool accounting for freeing and unreserving pages.
 * Return the number of global page reservations that must be dropped.
 * The return value may only be different than the passed value (delta)
 * in the case where a subpool minimum size must be maintained.
 */
static long hugepage_subpool_put_pages(struct hugepage_subpool *spool,
198 199
				       long delta)
{
200 201
	long ret = delta;

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

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

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

210 211
	 /* minimum size accounting */
	if (spool->min_hpages != -1 && spool->used_hpages < spool->min_hpages) {
212 213 214 215 216 217 218 219 220 221 222 223 224 225
		if (spool->rsv_hpages + delta <= spool->min_hpages)
			ret = 0;
		else
			ret = spool->rsv_hpages + delta - spool->min_hpages;

		spool->rsv_hpages += delta;
		if (spool->rsv_hpages > spool->min_hpages)
			spool->rsv_hpages = spool->min_hpages;
	}

	/*
	 * If hugetlbfs_put_super couldn't free spool due to an outstanding
	 * quota reference, free it now.
	 */
226
	unlock_or_release_subpool(spool);
227 228

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

static inline struct hugepage_subpool *subpool_inode(struct inode *inode)
{
	return HUGETLBFS_SB(inode->i_sb)->spool;
}

static inline struct hugepage_subpool *subpool_vma(struct vm_area_struct *vma)
{
A
Al Viro 已提交
238
	return subpool_inode(file_inode(vma->vm_file));
239 240
}

241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260
/* Helper that removes a struct file_region from the resv_map cache and returns
 * it for use.
 */
static struct file_region *
get_file_region_entry_from_cache(struct resv_map *resv, long from, long to)
{
	struct file_region *nrg = NULL;

	VM_BUG_ON(resv->region_cache_count <= 0);

	resv->region_cache_count--;
	nrg = list_first_entry(&resv->region_cache, struct file_region, link);
	list_del(&nrg->link);

	nrg->from = from;
	nrg->to = to;

	return nrg;
}

261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282
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;
283 284 285 286 287 288 289 290 291 292 293
		/*
		 * 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);
294 295 296 297 298 299 300 301 302 303 304 305 306
		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
}

307 308 309 310 311 312 313 314
static void put_uncharge_info(struct file_region *rg)
{
#ifdef CONFIG_CGROUP_HUGETLB
	if (rg->css)
		css_put(rg->css);
#endif
}

315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337
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);
338
		put_uncharge_info(rg);
339 340
		kfree(rg);

341
		rg = prg;
342 343 344 345 346 347 348 349
	}

	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);
350
		put_uncharge_info(rg);
351 352 353 354
		kfree(rg);
	}
}

355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372
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;
}

373 374 375 376 377 378 379
/*
 * 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
380 381
 */
static long add_reservation_in_range(struct resv_map *resv, long f, long t,
382
				     struct hugetlb_cgroup *h_cg,
383
				     struct hstate *h, long *regions_needed)
M
Mina Almasry 已提交
384
{
385
	long add = 0;
M
Mina Almasry 已提交
386
	struct list_head *head = &resv->regions;
387
	long last_accounted_offset = f;
388
	struct file_region *rg = NULL, *trg = NULL;
M
Mina Almasry 已提交
389

390 391
	if (regions_needed)
		*regions_needed = 0;
M
Mina Almasry 已提交
392

393 394 395 396 397 398 399 400 401 402 403 404 405 406
	/* 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
407

408 409 410
		/* When we find a region that starts beyond our range, we've
		 * finished.
		 */
411
		if (rg->from >= t)
M
Mina Almasry 已提交
412 413
			break;

414 415 416
		/* Add an entry for last_accounted_offset -> rg->from, and
		 * update last_accounted_offset.
		 */
417 418 419 420 421
		if (rg->from > last_accounted_offset)
			add += hugetlb_resv_map_add(resv, rg,
						    last_accounted_offset,
						    rg->from, h, h_cg,
						    regions_needed);
422 423 424 425 426 427 428

		last_accounted_offset = rg->to;
	}

	/* Handle the case where our range extends beyond
	 * last_accounted_offset.
	 */
429 430 431
	if (last_accounted_offset < t)
		add += hugetlb_resv_map_add(resv, rg, last_accounted_offset,
					    t, h, h_cg, regions_needed);
432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467

	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
468
		 */
469 470 471 472 473 474 475 476
		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
477 478
		}

479 480
		spin_lock(&resv->lock);

481 482
		list_splice(&allocated_regions, &resv->region_cache);
		resv->region_cache_count += to_allocate;
M
Mina Almasry 已提交
483 484
	}

485
	return 0;
M
Mina Almasry 已提交
486

487 488 489 490 491 492
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
493 494
}

495 496
/*
 * Add the huge page range represented by [f, t) to the reserve
497 498 499 500 501
 * 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.
502
 *
503 504 505 506
 * 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
507
 * this operation and we were not able to allocate, it returns -ENOMEM.
508 509 510
 * 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.
511
 */
512
static long region_add(struct resv_map *resv, long f, long t,
513 514
		       long in_regions_needed, struct hstate *h,
		       struct hugetlb_cgroup *h_cg)
515
{
516
	long add = 0, actual_regions_needed = 0;
517

518
	spin_lock(&resv->lock);
519 520 521
retry:

	/* Count how many regions are actually needed to execute this add. */
522 523
	add_reservation_in_range(resv, f, t, NULL, NULL,
				 &actual_regions_needed);
524

525
	/*
526 527 528 529 530 531 532
	 * 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.
533
	 */
534 535 536 537 538 539 540 541
	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);
542

543 544 545 546
		if (allocate_file_region_entries(
			    resv, actual_regions_needed - in_regions_needed)) {
			return -ENOMEM;
		}
547

548
		goto retry;
549 550
	}

551
	add = add_reservation_in_range(resv, f, t, h_cg, h, NULL);
552 553

	resv->adds_in_progress -= in_regions_needed;
554

555
	spin_unlock(&resv->lock);
556 557
	VM_BUG_ON(add < 0);
	return add;
558 559
}

560 561 562 563 564 565 566
/*
 * 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
567 568 569 570 571 572 573
 * 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.
574 575 576 577 578
 *
 * 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.
579
 */
580 581
static long region_chg(struct resv_map *resv, long f, long t,
		       long *out_regions_needed)
582 583 584
{
	long chg = 0;

585
	spin_lock(&resv->lock);
586

587
	/* Count how many hugepages in this range are NOT represented. */
588
	chg = add_reservation_in_range(resv, f, t, NULL, NULL,
589
				       out_regions_needed);
590

591 592
	if (*out_regions_needed == 0)
		*out_regions_needed = 1;
593

594 595
	if (allocate_file_region_entries(resv, *out_regions_needed))
		return -ENOMEM;
596

597
	resv->adds_in_progress += *out_regions_needed;
598 599

	spin_unlock(&resv->lock);
600 601 602
	return chg;
}

603 604 605 606 607
/*
 * 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
608 609 610
 * 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.
611 612 613 614 615
 *
 * 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.
 */
616 617
static void region_abort(struct resv_map *resv, long f, long t,
			 long regions_needed)
618 619 620
{
	spin_lock(&resv->lock);
	VM_BUG_ON(!resv->region_cache_count);
621
	resv->adds_in_progress -= regions_needed;
622 623 624
	spin_unlock(&resv->lock);
}

625
/*
626 627 628 629 630 631 632 633 634 635 636 637
 * 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.
638
 */
639
static long region_del(struct resv_map *resv, long f, long t)
640
{
641
	struct list_head *head = &resv->regions;
642
	struct file_region *rg, *trg;
643 644
	struct file_region *nrg = NULL;
	long del = 0;
645

646
retry:
647
	spin_lock(&resv->lock);
648
	list_for_each_entry_safe(rg, trg, head, link) {
649 650 651 652 653 654 655 656
		/*
		 * 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))
657
			continue;
658

659
		if (rg->from >= t)
660 661
			break;

662 663 664 665 666 667 668 669 670 671 672 673 674
		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--;
			}
675

676 677 678 679 680 681 682 683 684
			if (!nrg) {
				spin_unlock(&resv->lock);
				nrg = kmalloc(sizeof(*nrg), GFP_KERNEL);
				if (!nrg)
					return -ENOMEM;
				goto retry;
			}

			del += t - f;
685
			hugetlb_cgroup_uncharge_file_region(
686
				resv, rg, t - f, false);
687 688 689 690

			/* New entry for end of split region */
			nrg->from = t;
			nrg->to = rg->to;
691 692 693

			copy_hugetlb_cgroup_uncharge_info(nrg, rg);

694 695 696 697 698 699 700
			INIT_LIST_HEAD(&nrg->link);

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

			list_add(&nrg->link, &rg->link);
			nrg = NULL;
701
			break;
702 703 704 705
		}

		if (f <= rg->from && t >= rg->to) { /* Remove entire region */
			del += rg->to - rg->from;
706
			hugetlb_cgroup_uncharge_file_region(resv, rg,
707
							    rg->to - rg->from, true);
708 709 710 711 712 713
			list_del(&rg->link);
			kfree(rg);
			continue;
		}

		if (f <= rg->from) {	/* Trim beginning of region */
714
			hugetlb_cgroup_uncharge_file_region(resv, rg,
715
							    t - rg->from, false);
716

717 718 719
			del += t - rg->from;
			rg->from = t;
		} else {		/* Trim end of region */
720
			hugetlb_cgroup_uncharge_file_region(resv, rg,
721
							    rg->to - f, false);
722 723 724

			del += rg->to - f;
			rg->to = f;
725
		}
726
	}
727 728

	spin_unlock(&resv->lock);
729 730
	kfree(nrg);
	return del;
731 732
}

733 734 735 736 737 738 739 740 741
/*
 * 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.
 */
742
void hugetlb_fix_reserve_counts(struct inode *inode)
743 744 745 746 747
{
	struct hugepage_subpool *spool = subpool_inode(inode);
	long rsv_adjust;

	rsv_adjust = hugepage_subpool_get_pages(spool, 1);
748
	if (rsv_adjust) {
749 750 751 752 753 754
		struct hstate *h = hstate_inode(inode);

		hugetlb_acct_memory(h, 1);
	}
}

755 756 757 758
/*
 * Count and return the number of huge pages in the reserve map
 * that intersect with the range [f, t).
 */
759
static long region_count(struct resv_map *resv, long f, long t)
760
{
761
	struct list_head *head = &resv->regions;
762 763 764
	struct file_region *rg;
	long chg = 0;

765
	spin_lock(&resv->lock);
766 767
	/* Locate each segment we overlap with, and count that overlap. */
	list_for_each_entry(rg, head, link) {
768 769
		long seg_from;
		long seg_to;
770 771 772 773 774 775 776 777 778 779 780

		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;
	}
781
	spin_unlock(&resv->lock);
782 783 784 785

	return chg;
}

786 787 788 789
/*
 * Convert the address within this vma to the page offset within
 * the mapping, in pagecache page units; huge pages here.
 */
790 791
static pgoff_t vma_hugecache_offset(struct hstate *h,
			struct vm_area_struct *vma, unsigned long address)
792
{
793 794
	return ((address - vma->vm_start) >> huge_page_shift(h)) +
			(vma->vm_pgoff >> huge_page_order(h));
795 796
}

797 798 799 800 801
pgoff_t linear_hugepage_index(struct vm_area_struct *vma,
				     unsigned long address)
{
	return vma_hugecache_offset(hstate_vma(vma), vma, address);
}
802
EXPORT_SYMBOL_GPL(linear_hugepage_index);
803

804 805 806 807 808 809
/*
 * 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)
{
810 811 812
	if (vma->vm_ops && vma->vm_ops->pagesize)
		return vma->vm_ops->pagesize(vma);
	return PAGE_SIZE;
813
}
814
EXPORT_SYMBOL_GPL(vma_kernel_pagesize);
815

816 817 818
/*
 * 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
819 820
 * architectures where it differs, an architecture-specific 'strong'
 * version of this symbol is required.
821
 */
822
__weak unsigned long vma_mmu_pagesize(struct vm_area_struct *vma)
823 824 825 826
{
	return vma_kernel_pagesize(vma);
}

827 828 829 830 831 832 833
/*
 * 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)
834
#define HPAGE_RESV_MASK (HPAGE_RESV_OWNER | HPAGE_RESV_UNMAPPED)
835

836 837 838 839 840 841 842 843 844
/*
 * 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.
845 846 847 848 849 850 851 852 853
 *
 * 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.
854
 */
855 856 857 858 859 860 861 862 863 864 865
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;
}

866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884
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
}

885
struct resv_map *resv_map_alloc(void)
886 887
{
	struct resv_map *resv_map = kmalloc(sizeof(*resv_map), GFP_KERNEL);
888 889 890 891 892
	struct file_region *rg = kmalloc(sizeof(*rg), GFP_KERNEL);

	if (!resv_map || !rg) {
		kfree(resv_map);
		kfree(rg);
893
		return NULL;
894
	}
895 896

	kref_init(&resv_map->refs);
897
	spin_lock_init(&resv_map->lock);
898 899
	INIT_LIST_HEAD(&resv_map->regions);

900
	resv_map->adds_in_progress = 0;
901 902 903 904 905 906 907
	/*
	 * 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);
908 909 910 911 912

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

913 914 915
	return resv_map;
}

916
void resv_map_release(struct kref *ref)
917 918
{
	struct resv_map *resv_map = container_of(ref, struct resv_map, refs);
919 920
	struct list_head *head = &resv_map->region_cache;
	struct file_region *rg, *trg;
921 922

	/* Clear out any active regions before we release the map. */
923
	region_del(resv_map, 0, LONG_MAX);
924 925 926 927 928 929 930 931 932

	/* ... 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);

933 934 935
	kfree(resv_map);
}

936 937
static inline struct resv_map *inode_resv_map(struct inode *inode)
{
938 939 940 941 942 943 944 945 946
	/*
	 * 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;
947 948
}

949
static struct resv_map *vma_resv_map(struct vm_area_struct *vma)
950
{
951
	VM_BUG_ON_VMA(!is_vm_hugetlb_page(vma), vma);
952 953 954 955 956 957 958
	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 {
959 960
		return (struct resv_map *)(get_vma_private_data(vma) &
							~HPAGE_RESV_MASK);
961
	}
962 963
}

964
static void set_vma_resv_map(struct vm_area_struct *vma, struct resv_map *map)
965
{
966 967
	VM_BUG_ON_VMA(!is_vm_hugetlb_page(vma), vma);
	VM_BUG_ON_VMA(vma->vm_flags & VM_MAYSHARE, vma);
968

969 970
	set_vma_private_data(vma, (get_vma_private_data(vma) &
				HPAGE_RESV_MASK) | (unsigned long)map);
971 972 973 974
}

static void set_vma_resv_flags(struct vm_area_struct *vma, unsigned long flags)
{
975 976
	VM_BUG_ON_VMA(!is_vm_hugetlb_page(vma), vma);
	VM_BUG_ON_VMA(vma->vm_flags & VM_MAYSHARE, vma);
977 978

	set_vma_private_data(vma, get_vma_private_data(vma) | flags);
979 980 981 982
}

static int is_vma_resv_set(struct vm_area_struct *vma, unsigned long flag)
{
983
	VM_BUG_ON_VMA(!is_vm_hugetlb_page(vma), vma);
984 985

	return (get_vma_private_data(vma) & flag) != 0;
986 987
}

988
/* Reset counters to 0 and clear all HPAGE_RESV_* flags */
989 990
void reset_vma_resv_huge_pages(struct vm_area_struct *vma)
{
991
	VM_BUG_ON_VMA(!is_vm_hugetlb_page(vma), vma);
992
	if (!(vma->vm_flags & VM_MAYSHARE))
993 994 995 996
		vma->vm_private_data = (void *)0;
}

/* Returns true if the VMA has associated reserve pages */
997
static bool vma_has_reserves(struct vm_area_struct *vma, long chg)
998
{
999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009
	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)
1010
			return true;
1011
		else
1012
			return false;
1013
	}
1014 1015

	/* Shared mappings always use reserves */
1016 1017 1018 1019 1020
	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
1021
		 * fallocate.  In this case, there really are no reserves to
1022 1023 1024 1025 1026 1027 1028
		 * use.  This situation is indicated if chg != 0.
		 */
		if (chg)
			return false;
		else
			return true;
	}
1029 1030 1031 1032 1033

	/*
	 * Only the process that called mmap() has reserves for
	 * private mappings.
	 */
1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054
	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;
	}
1055

1056
	return false;
1057 1058
}

1059
static void enqueue_huge_page(struct hstate *h, struct page *page)
L
Linus Torvalds 已提交
1060 1061
{
	int nid = page_to_nid(page);
1062
	list_move(&page->lru, &h->hugepage_freelists[nid]);
1063 1064
	h->free_huge_pages++;
	h->free_huge_pages_node[nid]++;
1065
	SetHPageFreed(page);
L
Linus Torvalds 已提交
1066 1067
}

1068
static struct page *dequeue_huge_page_node_exact(struct hstate *h, int nid)
1069 1070
{
	struct page *page;
1071 1072 1073 1074 1075
	bool nocma = !!(current->flags & PF_MEMALLOC_NOCMA);

	list_for_each_entry(page, &h->hugepage_freelists[nid], lru) {
		if (nocma && is_migrate_cma_page(page))
			continue;
1076

1077 1078 1079 1080 1081
		if (PageHWPoison(page))
			continue;

		list_move(&page->lru, &h->hugepage_activelist);
		set_page_refcounted(page);
1082
		ClearHPageFreed(page);
1083 1084 1085
		h->free_huge_pages--;
		h->free_huge_pages_node[nid]--;
		return page;
1086 1087
	}

1088
	return NULL;
1089 1090
}

1091 1092
static struct page *dequeue_huge_page_nodemask(struct hstate *h, gfp_t gfp_mask, int nid,
		nodemask_t *nmask)
1093
{
1094 1095 1096 1097
	unsigned int cpuset_mems_cookie;
	struct zonelist *zonelist;
	struct zone *zone;
	struct zoneref *z;
1098
	int node = NUMA_NO_NODE;
1099

1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115
	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);
1116 1117 1118 1119 1120

		page = dequeue_huge_page_node_exact(h, node);
		if (page)
			return page;
	}
1121 1122 1123
	if (unlikely(read_mems_allowed_retry(cpuset_mems_cookie)))
		goto retry_cpuset;

1124 1125 1126
	return NULL;
}

1127 1128
static struct page *dequeue_huge_page_vma(struct hstate *h,
				struct vm_area_struct *vma,
1129 1130
				unsigned long address, int avoid_reserve,
				long chg)
L
Linus Torvalds 已提交
1131
{
1132
	struct page *page;
1133
	struct mempolicy *mpol;
1134
	gfp_t gfp_mask;
1135
	nodemask_t *nodemask;
1136
	int nid;
L
Linus Torvalds 已提交
1137

1138 1139 1140 1141 1142
	/*
	 * 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
	 */
1143
	if (!vma_has_reserves(vma, chg) &&
1144
			h->free_huge_pages - h->resv_huge_pages == 0)
1145
		goto err;
1146

1147
	/* If reserves cannot be used, ensure enough pages are in the pool */
1148
	if (avoid_reserve && h->free_huge_pages - h->resv_huge_pages == 0)
1149
		goto err;
1150

1151 1152
	gfp_mask = htlb_alloc_mask(h);
	nid = huge_node(vma, address, gfp_mask, &mpol, &nodemask);
1153 1154
	page = dequeue_huge_page_nodemask(h, gfp_mask, nid, nodemask);
	if (page && !avoid_reserve && vma_has_reserves(vma, chg)) {
1155
		SetHPageRestoreReserve(page);
1156
		h->resv_huge_pages--;
L
Linus Torvalds 已提交
1157
	}
1158

1159
	mpol_cond_put(mpol);
L
Linus Torvalds 已提交
1160
	return page;
1161 1162 1163

err:
	return NULL;
L
Linus Torvalds 已提交
1164 1165
}

1166 1167 1168 1169 1170 1171 1172 1173 1174
/*
 * 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)
{
1175
	nid = next_node_in(nid, *nodes_allowed);
1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236
	VM_BUG_ON(nid >= MAX_NUMNODES);

	return nid;
}

static int get_valid_node_allowed(int nid, nodemask_t *nodes_allowed)
{
	if (!node_isset(nid, *nodes_allowed))
		nid = next_node_allowed(nid, nodes_allowed);
	return nid;
}

/*
 * returns the previously saved node ["this node"] from which to
 * allocate a persistent huge page for the pool and advance the
 * next node from which to allocate, handling wrap at end of node
 * mask.
 */
static int hstate_next_node_to_alloc(struct hstate *h,
					nodemask_t *nodes_allowed)
{
	int nid;

	VM_BUG_ON(!nodes_allowed);

	nid = get_valid_node_allowed(h->next_nid_to_alloc, nodes_allowed);
	h->next_nid_to_alloc = next_node_allowed(nid, nodes_allowed);

	return nid;
}

/*
 * helper for free_pool_huge_page() - return the previously saved
 * node ["this node"] from which to free a huge page.  Advance the
 * next node id whether or not we find a free huge page to free so
 * that the next attempt to free addresses the next node.
 */
static int hstate_next_node_to_free(struct hstate *h, nodemask_t *nodes_allowed)
{
	int nid;

	VM_BUG_ON(!nodes_allowed);

	nid = get_valid_node_allowed(h->next_nid_to_free, nodes_allowed);
	h->next_nid_to_free = next_node_allowed(nid, nodes_allowed);

	return nid;
}

#define for_each_node_mask_to_alloc(hs, nr_nodes, node, mask)		\
	for (nr_nodes = nodes_weight(*mask);				\
		nr_nodes > 0 &&						\
		((node = hstate_next_node_to_alloc(hs, mask)) || 1);	\
		nr_nodes--)

#define for_each_node_mask_to_free(hs, nr_nodes, node, mask)		\
	for (nr_nodes = nodes_weight(*mask);				\
		nr_nodes > 0 &&						\
		((node = hstate_next_node_to_free(hs, mask)) || 1);	\
		nr_nodes--)

1237
#ifdef CONFIG_ARCH_HAS_GIGANTIC_PAGE
1238
static void destroy_compound_gigantic_page(struct page *page,
1239
					unsigned int order)
1240 1241 1242 1243 1244
{
	int i;
	int nr_pages = 1 << order;
	struct page *p = page + 1;

1245
	atomic_set(compound_mapcount_ptr(page), 0);
1246
	atomic_set(compound_pincount_ptr(page), 0);
1247

1248
	for (i = 1; i < nr_pages; i++, p = mem_map_next(p, page, i)) {
1249
		clear_compound_head(p);
1250 1251 1252 1253
		set_page_refcounted(p);
	}

	set_compound_order(page, 0);
1254
	page[1].compound_nr = 0;
1255 1256 1257
	__ClearPageHead(page);
}

1258
static void free_gigantic_page(struct page *page, unsigned int order)
1259
{
1260 1261 1262 1263
	/*
	 * If the page isn't allocated using the cma allocator,
	 * cma_release() returns false.
	 */
1264 1265
#ifdef CONFIG_CMA
	if (cma_release(hugetlb_cma[page_to_nid(page)], page, 1 << order))
1266
		return;
1267
#endif
1268

1269 1270 1271
	free_contig_range(page_to_pfn(page), 1 << order);
}

1272
#ifdef CONFIG_CONTIG_ALLOC
1273 1274
static struct page *alloc_gigantic_page(struct hstate *h, gfp_t gfp_mask,
		int nid, nodemask_t *nodemask)
1275
{
1276
	unsigned long nr_pages = 1UL << huge_page_order(h);
1277 1278
	if (nid == NUMA_NO_NODE)
		nid = numa_mem_id();
1279

1280 1281
#ifdef CONFIG_CMA
	{
1282 1283 1284
		struct page *page;
		int node;

1285 1286 1287
		if (hugetlb_cma[nid]) {
			page = cma_alloc(hugetlb_cma[nid], nr_pages,
					huge_page_order(h), true);
1288 1289 1290
			if (page)
				return page;
		}
1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302

		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;
			}
		}
1303
	}
1304
#endif
1305

1306
	return alloc_contig_pages(nr_pages, gfp_mask, nid, nodemask);
1307 1308 1309
}

static void prep_new_huge_page(struct hstate *h, struct page *page, int nid);
1310
static void prep_compound_gigantic_page(struct page *page, unsigned int order);
1311 1312 1313 1314 1315 1316 1317
#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 */
1318

1319
#else /* !CONFIG_ARCH_HAS_GIGANTIC_PAGE */
1320
static struct page *alloc_gigantic_page(struct hstate *h, gfp_t gfp_mask,
1321 1322 1323 1324
					int nid, nodemask_t *nodemask)
{
	return NULL;
}
1325
static inline void free_gigantic_page(struct page *page, unsigned int order) { }
1326
static inline void destroy_compound_gigantic_page(struct page *page,
1327
						unsigned int order) { }
1328 1329
#endif

1330
static void update_and_free_page(struct hstate *h, struct page *page)
A
Adam Litke 已提交
1331 1332
{
	int i;
1333
	struct page *subpage = page;
1334

1335
	if (hstate_is_gigantic(h) && !gigantic_page_runtime_supported())
1336
		return;
1337

1338 1339
	h->nr_huge_pages--;
	h->nr_huge_pages_node[page_to_nid(page)]--;
1340 1341 1342
	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 |
1343
				1 << PG_referenced | 1 << PG_dirty |
1344 1345
				1 << PG_active | 1 << PG_private |
				1 << PG_writeback);
A
Adam Litke 已提交
1346
	}
1347
	VM_BUG_ON_PAGE(hugetlb_cgroup_from_page(page), page);
1348
	VM_BUG_ON_PAGE(hugetlb_cgroup_from_page_rsvd(page), page);
1349
	set_compound_page_dtor(page, NULL_COMPOUND_DTOR);
A
Adam Litke 已提交
1350
	set_page_refcounted(page);
1351
	if (hstate_is_gigantic(h)) {
1352 1353 1354 1355 1356
		/*
		 * Temporarily drop the hugetlb_lock, because
		 * we might block in free_gigantic_page().
		 */
		spin_unlock(&hugetlb_lock);
1357 1358
		destroy_compound_gigantic_page(page, huge_page_order(h));
		free_gigantic_page(page, huge_page_order(h));
1359
		spin_lock(&hugetlb_lock);
1360 1361 1362
	} else {
		__free_pages(page, huge_page_order(h));
	}
A
Adam Litke 已提交
1363 1364
}

1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375
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;
}

1376
static void __free_huge_page(struct page *page)
1377
{
1378 1379 1380 1381
	/*
	 * Can't pass hstate in here because it is called from the
	 * compound page destructor.
	 */
1382
	struct hstate *h = page_hstate(page);
1383
	int nid = page_to_nid(page);
1384
	struct hugepage_subpool *spool = hugetlb_page_subpool(page);
1385
	bool restore_reserve;
1386

1387 1388
	VM_BUG_ON_PAGE(page_count(page), page);
	VM_BUG_ON_PAGE(page_mapcount(page), page);
1389

1390
	hugetlb_set_page_subpool(page, NULL);
1391
	page->mapping = NULL;
1392 1393
	restore_reserve = HPageRestoreReserve(page);
	ClearHPageRestoreReserve(page);
1394

1395
	/*
1396
	 * If HPageRestoreReserve was set on page, page allocation consumed a
1397 1398 1399
	 * 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
1400
	 * reservation, do not call hugepage_subpool_put_pages() as this will
1401
	 * remove the reserved page from the subpool.
1402
	 */
1403 1404 1405 1406 1407 1408 1409 1410 1411 1412
	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;
	}
1413

1414
	spin_lock(&hugetlb_lock);
1415
	ClearHPageMigratable(page);
1416 1417
	hugetlb_cgroup_uncharge_page(hstate_index(h),
				     pages_per_huge_page(h), page);
1418 1419
	hugetlb_cgroup_uncharge_page_rsvd(hstate_index(h),
					  pages_per_huge_page(h), page);
1420 1421 1422
	if (restore_reserve)
		h->resv_huge_pages++;

1423
	if (HPageTemporary(page)) {
1424
		list_del(&page->lru);
1425
		ClearHPageTemporary(page);
1426 1427
		update_and_free_page(h, page);
	} else if (h->surplus_huge_pages_node[nid]) {
1428 1429
		/* remove the page from active list */
		list_del(&page->lru);
1430 1431 1432
		update_and_free_page(h, page);
		h->surplus_huge_pages--;
		h->surplus_huge_pages_node[nid]--;
1433
	} else {
1434
		arch_clear_hugepage_flags(page);
1435
		enqueue_huge_page(h, page);
1436
	}
1437 1438 1439
	spin_unlock(&hugetlb_lock);
}

1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487
/*
 * As free_huge_page() can be called from a non-task context, we have
 * to defer the actual freeing in a workqueue to prevent potential
 * hugetlb_lock deadlock.
 *
 * free_hpage_workfn() locklessly retrieves the linked list of pages to
 * be freed and frees them one-by-one. As the page->mapping pointer is
 * going to be cleared in __free_huge_page() anyway, it is reused as the
 * llist_node structure of a lockless linked list of huge pages to be freed.
 */
static LLIST_HEAD(hpage_freelist);

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

	node = llist_del_all(&hpage_freelist);

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

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

	__free_huge_page(page);
}

1488
static void prep_new_huge_page(struct hstate *h, struct page *page, int nid)
1489
{
1490
	INIT_LIST_HEAD(&page->lru);
1491
	set_compound_page_dtor(page, HUGETLB_PAGE_DTOR);
1492
	hugetlb_set_page_subpool(page, NULL);
1493
	set_hugetlb_cgroup(page, NULL);
1494
	set_hugetlb_cgroup_rsvd(page, NULL);
1495
	spin_lock(&hugetlb_lock);
1496 1497
	h->nr_huge_pages++;
	h->nr_huge_pages_node[nid]++;
1498
	ClearHPageFreed(page);
1499 1500 1501
	spin_unlock(&hugetlb_lock);
}

1502
static void prep_compound_gigantic_page(struct page *page, unsigned int order)
1503 1504 1505 1506 1507 1508 1509
{
	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);
1510
	__ClearPageReserved(page);
1511
	__SetPageHead(page);
1512
	for (i = 1; i < nr_pages; i++, p = mem_map_next(p, page, i)) {
1513 1514 1515 1516
		/*
		 * 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
1517
		 * too.  Otherwise drivers using get_user_pages() to access tail
1518 1519 1520 1521 1522 1523 1524 1525
		 * 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);
1526
		set_page_count(p, 0);
1527
		set_compound_head(p, page);
1528
	}
1529
	atomic_set(compound_mapcount_ptr(page), -1);
1530
	atomic_set(compound_pincount_ptr(page), 0);
1531 1532
}

A
Andrew Morton 已提交
1533 1534 1535 1536 1537
/*
 * PageHuge() only returns true for hugetlbfs pages, but not for normal or
 * transparent huge pages.  See the PageTransHuge() documentation for more
 * details.
 */
1538 1539 1540 1541 1542 1543
int PageHuge(struct page *page)
{
	if (!PageCompound(page))
		return 0;

	page = compound_head(page);
1544
	return page[1].compound_dtor == HUGETLB_PAGE_DTOR;
1545
}
1546 1547
EXPORT_SYMBOL_GPL(PageHuge);

1548 1549 1550 1551 1552 1553 1554 1555 1556
/*
 * 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;

1557
	return page_head[1].compound_dtor == HUGETLB_PAGE_DTOR;
1558 1559
}

1560 1561 1562
/*
 * Find and lock address space (mapping) in write mode.
 *
1563 1564 1565
 * 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.
1566 1567 1568
 */
struct address_space *hugetlb_page_mapping_lock_write(struct page *hpage)
{
1569
	struct address_space *mapping = page_mapping(hpage);
1570 1571 1572 1573 1574 1575 1576

	if (!mapping)
		return mapping;

	if (i_mmap_trylock_write(mapping))
		return mapping;

1577
	return NULL;
1578 1579
}

1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596
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;
}

1597
static struct page *alloc_buddy_huge_page(struct hstate *h,
1598 1599
		gfp_t gfp_mask, int nid, nodemask_t *nmask,
		nodemask_t *node_alloc_noretry)
L
Linus Torvalds 已提交
1600
{
1601
	int order = huge_page_order(h);
L
Linus Torvalds 已提交
1602
	struct page *page;
1603
	bool alloc_try_hard = true;
1604

1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616
	/*
	 * 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;
1617 1618
	if (nid == NUMA_NO_NODE)
		nid = numa_mem_id();
1619
	page = __alloc_pages(gfp_mask, order, nid, nmask);
1620 1621 1622 1623
	if (page)
		__count_vm_event(HTLB_BUDDY_PGALLOC);
	else
		__count_vm_event(HTLB_BUDDY_PGALLOC_FAIL);
1624

1625 1626 1627 1628 1629 1630 1631 1632 1633 1634 1635 1636 1637 1638 1639 1640
	/*
	 * 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);

1641 1642 1643
	return page;
}

1644 1645 1646 1647 1648
/*
 * 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,
1649 1650
		gfp_t gfp_mask, int nid, nodemask_t *nmask,
		nodemask_t *node_alloc_noretry)
1651 1652 1653 1654 1655 1656 1657
{
	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,
1658
				nid, nmask, node_alloc_noretry);
1659 1660 1661 1662 1663 1664 1665 1666 1667 1668
	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;
}

1669 1670 1671 1672
/*
 * Allocates a fresh page to the hugetlb allocator pool in the node interleaved
 * manner.
 */
1673 1674
static int alloc_pool_huge_page(struct hstate *h, nodemask_t *nodes_allowed,
				nodemask_t *node_alloc_noretry)
1675 1676 1677
{
	struct page *page;
	int nr_nodes, node;
1678
	gfp_t gfp_mask = htlb_alloc_mask(h) | __GFP_THISNODE;
1679 1680

	for_each_node_mask_to_alloc(h, nr_nodes, node, nodes_allowed) {
1681 1682
		page = alloc_fresh_huge_page(h, gfp_mask, node, nodes_allowed,
						node_alloc_noretry);
1683
		if (page)
1684 1685 1686
			break;
	}

1687 1688
	if (!page)
		return 0;
1689

1690 1691 1692
	put_page(page); /* free it into the hugepage allocator */

	return 1;
1693 1694
}

1695 1696 1697 1698 1699 1700
/*
 * Free huge page from pool from next node to free.
 * Attempt to keep persistent huge pages more or less
 * balanced over allowed nodes.
 * Called with hugetlb_lock locked.
 */
1701 1702
static int free_pool_huge_page(struct hstate *h, nodemask_t *nodes_allowed,
							 bool acct_surplus)
1703
{
1704
	int nr_nodes, node;
1705 1706
	int ret = 0;

1707
	for_each_node_mask_to_free(h, nr_nodes, node, nodes_allowed) {
1708 1709 1710 1711
		/*
		 * If we're returning unused surplus pages, only examine
		 * nodes with surplus pages.
		 */
1712 1713
		if ((!acct_surplus || h->surplus_huge_pages_node[node]) &&
		    !list_empty(&h->hugepage_freelists[node])) {
1714
			struct page *page =
1715
				list_entry(h->hugepage_freelists[node].next,
1716 1717 1718
					  struct page, lru);
			list_del(&page->lru);
			h->free_huge_pages--;
1719
			h->free_huge_pages_node[node]--;
1720 1721
			if (acct_surplus) {
				h->surplus_huge_pages--;
1722
				h->surplus_huge_pages_node[node]--;
1723
			}
1724 1725
			update_and_free_page(h, page);
			ret = 1;
1726
			break;
1727
		}
1728
	}
1729 1730 1731 1732

	return ret;
}

1733 1734
/*
 * Dissolve a given free hugepage into free buddy pages. This function does
1735 1736 1737 1738 1739 1740 1741
 * 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)
1742
 */
1743
int dissolve_free_huge_page(struct page *page)
1744
{
1745
	int rc = -EBUSY;
1746

1747
retry:
1748 1749 1750 1751
	/* Not to disrupt normal path by vainly holding hugetlb_lock */
	if (!PageHuge(page))
		return 0;

1752
	spin_lock(&hugetlb_lock);
1753 1754 1755 1756 1757 1758
	if (!PageHuge(page)) {
		rc = 0;
		goto out;
	}

	if (!page_count(page)) {
1759 1760 1761
		struct page *head = compound_head(page);
		struct hstate *h = page_hstate(head);
		int nid = page_to_nid(head);
1762
		if (h->free_huge_pages - h->resv_huge_pages == 0)
1763
			goto out;
1764 1765 1766 1767 1768

		/*
		 * We should make sure that the page is already on the free list
		 * when it is dissolved.
		 */
1769
		if (unlikely(!HPageFreed(head))) {
1770 1771 1772 1773 1774 1775 1776 1777 1778 1779 1780 1781 1782 1783
			spin_unlock(&hugetlb_lock);
			cond_resched();

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

1784 1785 1786 1787 1788 1789 1790 1791
		/*
		 * 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);
		}
1792
		list_del(&head->lru);
1793 1794
		h->free_huge_pages--;
		h->free_huge_pages_node[nid]--;
1795
		h->max_huge_pages--;
1796
		update_and_free_page(h, head);
1797
		rc = 0;
1798
	}
1799
out:
1800
	spin_unlock(&hugetlb_lock);
1801
	return rc;
1802 1803 1804 1805 1806
}

/*
 * Dissolve free hugepages in a given pfn range. Used by memory hotplug to
 * make specified memory blocks removable from the system.
1807 1808
 * Note that this will dissolve a free gigantic hugepage completely, if any
 * part of it lies within the given range.
1809 1810
 * Also note that if dissolve_free_huge_page() returns with an error, all
 * free hugepages that were dissolved before that error are lost.
1811
 */
1812
int dissolve_free_huge_pages(unsigned long start_pfn, unsigned long end_pfn)
1813 1814
{
	unsigned long pfn;
1815
	struct page *page;
1816
	int rc = 0;
1817

1818
	if (!hugepages_supported())
1819
		return rc;
1820

1821 1822
	for (pfn = start_pfn; pfn < end_pfn; pfn += 1 << minimum_order) {
		page = pfn_to_page(pfn);
1823 1824 1825
		rc = dissolve_free_huge_page(page);
		if (rc)
			break;
1826
	}
1827 1828

	return rc;
1829 1830
}

1831 1832 1833
/*
 * Allocates a fresh surplus page from the page allocator.
 */
1834
static struct page *alloc_surplus_huge_page(struct hstate *h, gfp_t gfp_mask,
1835
		int nid, nodemask_t *nmask)
1836
{
1837
	struct page *page = NULL;
1838

1839
	if (hstate_is_gigantic(h))
1840 1841
		return NULL;

1842
	spin_lock(&hugetlb_lock);
1843 1844
	if (h->surplus_huge_pages >= h->nr_overcommit_huge_pages)
		goto out_unlock;
1845 1846
	spin_unlock(&hugetlb_lock);

1847
	page = alloc_fresh_huge_page(h, gfp_mask, nid, nmask, NULL);
1848
	if (!page)
1849
		return NULL;
1850 1851

	spin_lock(&hugetlb_lock);
1852 1853 1854 1855 1856 1857 1858 1859
	/*
	 * 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) {
1860
		SetHPageTemporary(page);
1861
		spin_unlock(&hugetlb_lock);
1862
		put_page(page);
1863
		return NULL;
1864 1865
	} else {
		h->surplus_huge_pages++;
1866
		h->surplus_huge_pages_node[page_to_nid(page)]++;
1867
	}
1868 1869

out_unlock:
1870
	spin_unlock(&hugetlb_lock);
1871 1872 1873 1874

	return page;
}

1875
static struct page *alloc_migrate_huge_page(struct hstate *h, gfp_t gfp_mask,
1876
				     int nid, nodemask_t *nmask)
1877 1878 1879 1880 1881 1882
{
	struct page *page;

	if (hstate_is_gigantic(h))
		return NULL;

1883
	page = alloc_fresh_huge_page(h, gfp_mask, nid, nmask, NULL);
1884 1885 1886 1887 1888 1889 1890
	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
	 */
1891
	SetHPageTemporary(page);
1892 1893 1894 1895

	return page;
}

1896 1897 1898
/*
 * Use the VMA's mpolicy to allocate a huge page from the buddy.
 */
D
Dave Hansen 已提交
1899
static
1900
struct page *alloc_buddy_huge_page_with_mpol(struct hstate *h,
1901 1902
		struct vm_area_struct *vma, unsigned long addr)
{
1903 1904 1905 1906 1907 1908 1909
	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);
1910
	page = alloc_surplus_huge_page(h, gfp_mask, nid, nodemask);
1911 1912 1913
	mpol_cond_put(mpol);

	return page;
1914 1915
}

1916
/* page migration callback function */
1917
struct page *alloc_huge_page_nodemask(struct hstate *h, int preferred_nid,
1918
		nodemask_t *nmask, gfp_t gfp_mask)
1919 1920 1921
{
	spin_lock(&hugetlb_lock);
	if (h->free_huge_pages - h->resv_huge_pages > 0) {
1922 1923 1924 1925 1926 1927
		struct page *page;

		page = dequeue_huge_page_nodemask(h, gfp_mask, preferred_nid, nmask);
		if (page) {
			spin_unlock(&hugetlb_lock);
			return page;
1928 1929 1930 1931
		}
	}
	spin_unlock(&hugetlb_lock);

1932
	return alloc_migrate_huge_page(h, gfp_mask, preferred_nid, nmask);
1933 1934
}

1935
/* mempolicy aware migration callback */
1936 1937
struct page *alloc_huge_page_vma(struct hstate *h, struct vm_area_struct *vma,
		unsigned long address)
1938 1939 1940 1941 1942 1943 1944 1945 1946
{
	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);
1947
	page = alloc_huge_page_nodemask(h, node, nodemask, gfp_mask);
1948 1949 1950 1951 1952
	mpol_cond_put(mpol);

	return page;
}

1953
/*
L
Lucas De Marchi 已提交
1954
 * Increase the hugetlb pool such that it can accommodate a reservation
1955 1956
 * of size 'delta'.
 */
1957
static int gather_surplus_pages(struct hstate *h, long delta)
1958
	__must_hold(&hugetlb_lock)
1959 1960 1961
{
	struct list_head surplus_list;
	struct page *page, *tmp;
1962 1963 1964
	int ret;
	long i;
	long needed, allocated;
1965
	bool alloc_ok = true;
1966

1967
	needed = (h->resv_huge_pages + delta) - h->free_huge_pages;
1968
	if (needed <= 0) {
1969
		h->resv_huge_pages += delta;
1970
		return 0;
1971
	}
1972 1973 1974 1975 1976 1977 1978 1979

	allocated = 0;
	INIT_LIST_HEAD(&surplus_list);

	ret = -ENOMEM;
retry:
	spin_unlock(&hugetlb_lock);
	for (i = 0; i < needed; i++) {
1980
		page = alloc_surplus_huge_page(h, htlb_alloc_mask(h),
1981
				NUMA_NO_NODE, NULL);
1982 1983 1984 1985
		if (!page) {
			alloc_ok = false;
			break;
		}
1986
		list_add(&page->lru, &surplus_list);
1987
		cond_resched();
1988
	}
1989
	allocated += i;
1990 1991 1992 1993 1994 1995

	/*
	 * After retaking hugetlb_lock, we need to recalculate 'needed'
	 * because either resv_huge_pages or free_huge_pages may have changed.
	 */
	spin_lock(&hugetlb_lock);
1996 1997
	needed = (h->resv_huge_pages + delta) -
			(h->free_huge_pages + allocated);
1998 1999 2000 2001 2002 2003 2004 2005 2006 2007
	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;
	}
2008 2009
	/*
	 * The surplus_list now contains _at_least_ the number of extra pages
L
Lucas De Marchi 已提交
2010
	 * needed to accommodate the reservation.  Add the appropriate number
2011
	 * of pages to the hugetlb pool and free the extras back to the buddy
2012 2013 2014
	 * 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.
2015 2016
	 */
	needed += allocated;
2017
	h->resv_huge_pages += delta;
2018
	ret = 0;
2019

2020
	/* Free the needed pages to the hugetlb pool */
2021
	list_for_each_entry_safe(page, tmp, &surplus_list, lru) {
2022 2023
		int zeroed;

2024 2025
		if ((--needed) < 0)
			break;
2026 2027 2028 2029
		/*
		 * This page is now managed by the hugetlb allocator and has
		 * no users -- drop the buddy allocator's reference.
		 */
2030 2031
		zeroed = put_page_testzero(page);
		VM_BUG_ON_PAGE(!zeroed, page);
2032
		enqueue_huge_page(h, page);
2033
	}
2034
free:
2035
	spin_unlock(&hugetlb_lock);
2036 2037

	/* Free unnecessary surplus pages to the buddy allocator */
2038 2039
	list_for_each_entry_safe(page, tmp, &surplus_list, lru)
		put_page(page);
2040
	spin_lock(&hugetlb_lock);
2041 2042 2043 2044 2045

	return ret;
}

/*
2046 2047 2048 2049 2050 2051 2052 2053 2054 2055 2056 2057
 * This routine has two main purposes:
 * 1) Decrement the reservation count (resv_huge_pages) by the value passed
 *    in unused_resv_pages.  This corresponds to the prior adjustments made
 *    to the associated reservation map.
 * 2) Free any unused surplus pages that may have been allocated to satisfy
 *    the reservation.  As many as unused_resv_pages may be freed.
 *
 * Called with hugetlb_lock held.  However, the lock could be dropped (and
 * reacquired) during calls to cond_resched_lock.  Whenever dropping the lock,
 * we must make sure nobody else can claim pages we are in the process of
 * freeing.  Do this by ensuring resv_huge_page always is greater than the
 * number of huge pages we plan to free when dropping the lock.
2058
 */
2059 2060
static void return_unused_surplus_pages(struct hstate *h,
					unsigned long unused_resv_pages)
2061 2062 2063
{
	unsigned long nr_pages;

2064
	/* Cannot return gigantic pages currently */
2065
	if (hstate_is_gigantic(h))
2066
		goto out;
2067

2068 2069 2070 2071
	/*
	 * Part (or even all) of the reservation could have been backed
	 * by pre-allocated pages. Only free surplus pages.
	 */
2072
	nr_pages = min(unused_resv_pages, h->surplus_huge_pages);
2073

2074 2075
	/*
	 * We want to release as many surplus pages as possible, spread
2076 2077 2078
	 * 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.
2079
	 * free_pool_huge_page() will balance the freed pages across the
2080
	 * on-line nodes with memory and will handle the hstate accounting.
2081 2082 2083 2084
	 *
	 * Note that we decrement resv_huge_pages as we free the pages.  If
	 * we drop the lock, resv_huge_pages will still be sufficiently large
	 * to cover subsequent pages we may free.
2085 2086
	 */
	while (nr_pages--) {
2087 2088
		h->resv_huge_pages--;
		unused_resv_pages--;
2089
		if (!free_pool_huge_page(h, &node_states[N_MEMORY], 1))
2090
			goto out;
2091
		cond_resched_lock(&hugetlb_lock);
2092
	}
2093 2094 2095 2096

out:
	/* Fully uncommit the reservation */
	h->resv_huge_pages -= unused_resv_pages;
2097 2098
}

2099

2100
/*
2101
 * vma_needs_reservation, vma_commit_reservation and vma_end_reservation
2102
 * are used by the huge page allocation routines to manage reservations.
2103 2104 2105 2106 2107 2108
 *
 * 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
2109 2110 2111
 * 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.
2112 2113 2114 2115 2116 2117
 *
 * 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.
2118 2119 2120 2121 2122
 *
 * 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.
2123
 */
2124 2125 2126
enum vma_resv_mode {
	VMA_NEEDS_RESV,
	VMA_COMMIT_RESV,
2127
	VMA_END_RESV,
2128
	VMA_ADD_RESV,
2129
};
2130 2131
static long __vma_reservation_common(struct hstate *h,
				struct vm_area_struct *vma, unsigned long addr,
2132
				enum vma_resv_mode mode)
2133
{
2134 2135
	struct resv_map *resv;
	pgoff_t idx;
2136
	long ret;
2137
	long dummy_out_regions_needed;
2138

2139 2140
	resv = vma_resv_map(vma);
	if (!resv)
2141
		return 1;
2142

2143
	idx = vma_hugecache_offset(h, vma, addr);
2144 2145
	switch (mode) {
	case VMA_NEEDS_RESV:
2146 2147 2148 2149 2150 2151
		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);
2152 2153
		break;
	case VMA_COMMIT_RESV:
2154
		ret = region_add(resv, idx, idx + 1, 1, NULL, NULL);
2155 2156
		/* region_add calls of range 1 should never fail. */
		VM_BUG_ON(ret < 0);
2157
		break;
2158
	case VMA_END_RESV:
2159
		region_abort(resv, idx, idx + 1, 1);
2160 2161
		ret = 0;
		break;
2162
	case VMA_ADD_RESV:
2163
		if (vma->vm_flags & VM_MAYSHARE) {
2164
			ret = region_add(resv, idx, idx + 1, 1, NULL, NULL);
2165 2166 2167 2168
			/* region_add calls of range 1 should never fail. */
			VM_BUG_ON(ret < 0);
		} else {
			region_abort(resv, idx, idx + 1, 1);
2169 2170 2171
			ret = region_del(resv, idx, idx + 1);
		}
		break;
2172 2173 2174
	default:
		BUG();
	}
2175

2176
	if (vma->vm_flags & VM_MAYSHARE)
2177
		return ret;
2178 2179 2180 2181 2182 2183 2184 2185 2186 2187 2188 2189 2190 2191 2192 2193 2194 2195 2196
	else if (is_vma_resv_set(vma, HPAGE_RESV_OWNER) && ret >= 0) {
		/*
		 * In most cases, reserves always exist for private mappings.
		 * However, a file associated with mapping could have been
		 * hole punched or truncated after reserves were consumed.
		 * As subsequent fault on such a range will not use reserves.
		 * Subtle - The reserve map for private mappings has the
		 * opposite meaning than that of shared mappings.  If NO
		 * entry is in the reserve map, it means a reservation exists.
		 * If an entry exists in the reserve map, it means the
		 * reservation has already been consumed.  As a result, the
		 * return value of this routine is the opposite of the
		 * value returned from reserve map manipulation routines above.
		 */
		if (ret)
			return 0;
		else
			return 1;
	}
2197
	else
2198
		return ret < 0 ? ret : 0;
2199
}
2200 2201

static long vma_needs_reservation(struct hstate *h,
2202
			struct vm_area_struct *vma, unsigned long addr)
2203
{
2204
	return __vma_reservation_common(h, vma, addr, VMA_NEEDS_RESV);
2205
}
2206

2207 2208 2209
static long vma_commit_reservation(struct hstate *h,
			struct vm_area_struct *vma, unsigned long addr)
{
2210 2211 2212
	return __vma_reservation_common(h, vma, addr, VMA_COMMIT_RESV);
}

2213
static void vma_end_reservation(struct hstate *h,
2214 2215
			struct vm_area_struct *vma, unsigned long addr)
{
2216
	(void)__vma_reservation_common(h, vma, addr, VMA_END_RESV);
2217 2218
}

2219 2220 2221 2222 2223 2224 2225 2226 2227 2228
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,
2229 2230 2231 2232 2233 2234
 * 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.
2235 2236 2237 2238 2239
 */
static void restore_reserve_on_error(struct hstate *h,
			struct vm_area_struct *vma, unsigned long address,
			struct page *page)
{
2240
	if (unlikely(HPageRestoreReserve(page))) {
2241 2242 2243 2244 2245
		long rc = vma_needs_reservation(h, vma, address);

		if (unlikely(rc < 0)) {
			/*
			 * Rare out of memory condition in reserve map
2246
			 * manipulation.  Clear HPageRestoreReserve so that
2247 2248 2249 2250 2251 2252 2253 2254
			 * 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.
			 */
2255
			ClearHPageRestoreReserve(page);
2256 2257 2258 2259 2260 2261 2262
		} else if (rc) {
			rc = vma_add_reservation(h, vma, address);
			if (unlikely(rc < 0))
				/*
				 * See above comment about rare out of
				 * memory condition.
				 */
2263
				ClearHPageRestoreReserve(page);
2264 2265 2266 2267 2268
		} else
			vma_end_reservation(h, vma, address);
	}
}

2269
struct page *alloc_huge_page(struct vm_area_struct *vma,
2270
				    unsigned long addr, int avoid_reserve)
L
Linus Torvalds 已提交
2271
{
2272
	struct hugepage_subpool *spool = subpool_vma(vma);
2273
	struct hstate *h = hstate_vma(vma);
2274
	struct page *page;
2275 2276
	long map_chg, map_commit;
	long gbl_chg;
2277 2278
	int ret, idx;
	struct hugetlb_cgroup *h_cg;
2279
	bool deferred_reserve;
2280

2281
	idx = hstate_index(h);
2282
	/*
2283 2284 2285
	 * 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).
2286
	 */
2287 2288
	map_chg = gbl_chg = vma_needs_reservation(h, vma, addr);
	if (map_chg < 0)
2289
		return ERR_PTR(-ENOMEM);
2290 2291 2292 2293 2294 2295 2296 2297 2298 2299 2300

	/*
	 * 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) {
2301
			vma_end_reservation(h, vma, addr);
2302
			return ERR_PTR(-ENOSPC);
2303
		}
L
Linus Torvalds 已提交
2304

2305 2306 2307 2308 2309 2310 2311 2312 2313 2314 2315 2316
		/*
		 * 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;
	}

2317 2318 2319 2320 2321 2322 2323 2324 2325 2326
	/* If this allocation is not consuming a reservation, charge it now.
	 */
	deferred_reserve = map_chg || avoid_reserve || !vma_resv_map(vma);
	if (deferred_reserve) {
		ret = hugetlb_cgroup_charge_cgroup_rsvd(
			idx, pages_per_huge_page(h), &h_cg);
		if (ret)
			goto out_subpool_put;
	}

2327
	ret = hugetlb_cgroup_charge_cgroup(idx, pages_per_huge_page(h), &h_cg);
2328
	if (ret)
2329
		goto out_uncharge_cgroup_reservation;
2330

L
Linus Torvalds 已提交
2331
	spin_lock(&hugetlb_lock);
2332 2333 2334 2335 2336 2337
	/*
	 * 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);
2338
	if (!page) {
2339
		spin_unlock(&hugetlb_lock);
2340
		page = alloc_buddy_huge_page_with_mpol(h, vma, addr);
2341 2342
		if (!page)
			goto out_uncharge_cgroup;
2343
		if (!avoid_reserve && vma_has_reserves(vma, gbl_chg)) {
2344
			SetHPageRestoreReserve(page);
2345 2346
			h->resv_huge_pages--;
		}
2347
		spin_lock(&hugetlb_lock);
2348
		list_add(&page->lru, &h->hugepage_activelist);
2349
		/* Fall through */
K
Ken Chen 已提交
2350
	}
2351
	hugetlb_cgroup_commit_charge(idx, pages_per_huge_page(h), h_cg, page);
2352 2353 2354 2355 2356 2357 2358 2359
	/* 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);
	}

2360
	spin_unlock(&hugetlb_lock);
2361

2362
	hugetlb_set_page_subpool(page, spool);
2363

2364 2365
	map_commit = vma_commit_reservation(h, vma, addr);
	if (unlikely(map_chg > map_commit)) {
2366 2367 2368 2369 2370 2371 2372 2373 2374 2375 2376 2377 2378
		/*
		 * 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);
2379 2380 2381
		if (deferred_reserve)
			hugetlb_cgroup_uncharge_page_rsvd(hstate_index(h),
					pages_per_huge_page(h), page);
2382
	}
2383
	return page;
2384 2385 2386

out_uncharge_cgroup:
	hugetlb_cgroup_uncharge_cgroup(idx, pages_per_huge_page(h), h_cg);
2387 2388 2389 2390
out_uncharge_cgroup_reservation:
	if (deferred_reserve)
		hugetlb_cgroup_uncharge_cgroup_rsvd(idx, pages_per_huge_page(h),
						    h_cg);
2391
out_subpool_put:
2392
	if (map_chg || avoid_reserve)
2393
		hugepage_subpool_put_pages(spool, 1);
2394
	vma_end_reservation(h, vma, addr);
2395
	return ERR_PTR(-ENOSPC);
2396 2397
}

2398 2399 2400
int alloc_bootmem_huge_page(struct hstate *h)
	__attribute__ ((weak, alias("__alloc_bootmem_huge_page")));
int __alloc_bootmem_huge_page(struct hstate *h)
2401 2402
{
	struct huge_bootmem_page *m;
2403
	int nr_nodes, node;
2404

2405
	for_each_node_mask_to_alloc(h, nr_nodes, node, &node_states[N_MEMORY]) {
2406 2407
		void *addr;

2408
		addr = memblock_alloc_try_nid_raw(
2409
				huge_page_size(h), huge_page_size(h),
2410
				0, MEMBLOCK_ALLOC_ACCESSIBLE, node);
2411 2412 2413 2414 2415 2416 2417
		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;
2418
			goto found;
2419 2420 2421 2422 2423
		}
	}
	return 0;

found:
2424
	BUG_ON(!IS_ALIGNED(virt_to_phys(m), huge_page_size(h)));
2425
	/* Put them into a private list first because mem_map is not up yet */
2426
	INIT_LIST_HEAD(&m->list);
2427 2428 2429 2430 2431
	list_add(&m->list, &huge_boot_pages);
	m->hstate = h;
	return 1;
}

2432 2433
static void __init prep_compound_huge_page(struct page *page,
		unsigned int order)
2434 2435 2436 2437 2438 2439 2440
{
	if (unlikely(order > (MAX_ORDER - 1)))
		prep_compound_gigantic_page(page, order);
	else
		prep_compound_page(page, order);
}

2441 2442 2443 2444 2445 2446
/* 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) {
2447
		struct page *page = virt_to_page(m);
2448
		struct hstate *h = m->hstate;
2449

2450
		WARN_ON(page_count(page) != 1);
2451
		prep_compound_huge_page(page, huge_page_order(h));
2452
		WARN_ON(PageReserved(page));
2453
		prep_new_huge_page(h, page, page_to_nid(page));
2454 2455
		put_page(page); /* free it into the hugepage allocator */

2456 2457 2458 2459 2460 2461
		/*
		 * 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.
		 */
2462
		if (hstate_is_gigantic(h))
2463
			adjust_managed_page_count(page, pages_per_huge_page(h));
2464
		cond_resched();
2465 2466 2467
	}
}

2468
static void __init hugetlb_hstate_alloc_pages(struct hstate *h)
L
Linus Torvalds 已提交
2469 2470
{
	unsigned long i;
2471 2472 2473 2474 2475 2476 2477 2478 2479 2480 2481 2482 2483 2484 2485 2486 2487 2488 2489
	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);
2490

2491
	for (i = 0; i < h->max_huge_pages; ++i) {
2492
		if (hstate_is_gigantic(h)) {
2493
			if (hugetlb_cma_size) {
2494
				pr_warn_once("HugeTLB: hugetlb_cma is enabled, skip boot time allocation\n");
2495
				goto free;
2496
			}
2497 2498
			if (!alloc_bootmem_huge_page(h))
				break;
2499
		} else if (!alloc_pool_huge_page(h,
2500 2501
					 &node_states[N_MEMORY],
					 node_alloc_noretry))
L
Linus Torvalds 已提交
2502
			break;
2503
		cond_resched();
L
Linus Torvalds 已提交
2504
	}
2505 2506 2507
	if (i < h->max_huge_pages) {
		char buf[32];

2508
		string_get_size(huge_page_size(h), 1, STRING_UNITS_2, buf, 32);
2509 2510 2511 2512
		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;
	}
2513
free:
2514
	kfree(node_alloc_noretry);
2515 2516 2517 2518 2519 2520 2521
}

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

	for_each_hstate(h) {
2522 2523 2524
		if (minimum_order > huge_page_order(h))
			minimum_order = huge_page_order(h);

2525
		/* oversize hugepages were init'ed in early boot */
2526
		if (!hstate_is_gigantic(h))
2527
			hugetlb_hstate_alloc_pages(h);
2528
	}
2529
	VM_BUG_ON(minimum_order == UINT_MAX);
2530 2531 2532 2533 2534 2535 2536
}

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

	for_each_hstate(h) {
A
Andi Kleen 已提交
2537
		char buf[32];
2538 2539

		string_get_size(huge_page_size(h), 1, STRING_UNITS_2, buf, 32);
2540
		pr_info("HugeTLB registered %s page size, pre-allocated %ld pages\n",
2541
			buf, h->free_huge_pages);
2542 2543 2544
	}
}

L
Linus Torvalds 已提交
2545
#ifdef CONFIG_HIGHMEM
2546 2547
static void try_to_free_low(struct hstate *h, unsigned long count,
						nodemask_t *nodes_allowed)
L
Linus Torvalds 已提交
2548
{
2549 2550
	int i;

2551
	if (hstate_is_gigantic(h))
2552 2553
		return;

2554
	for_each_node_mask(i, *nodes_allowed) {
L
Linus Torvalds 已提交
2555
		struct page *page, *next;
2556 2557 2558
		struct list_head *freel = &h->hugepage_freelists[i];
		list_for_each_entry_safe(page, next, freel, lru) {
			if (count >= h->nr_huge_pages)
2559
				return;
L
Linus Torvalds 已提交
2560 2561 2562
			if (PageHighMem(page))
				continue;
			list_del(&page->lru);
2563
			update_and_free_page(h, page);
2564 2565
			h->free_huge_pages--;
			h->free_huge_pages_node[page_to_nid(page)]--;
L
Linus Torvalds 已提交
2566 2567 2568 2569
		}
	}
}
#else
2570 2571
static inline void try_to_free_low(struct hstate *h, unsigned long count,
						nodemask_t *nodes_allowed)
L
Linus Torvalds 已提交
2572 2573 2574 2575
{
}
#endif

2576 2577 2578 2579 2580
/*
 * 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.
 */
2581 2582
static int adjust_pool_surplus(struct hstate *h, nodemask_t *nodes_allowed,
				int delta)
2583
{
2584
	int nr_nodes, node;
2585 2586 2587

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

2588 2589 2590 2591
	if (delta < 0) {
		for_each_node_mask_to_alloc(h, nr_nodes, node, nodes_allowed) {
			if (h->surplus_huge_pages_node[node])
				goto found;
2592
		}
2593 2594 2595 2596 2597
	} 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;
2598
		}
2599 2600
	}
	return 0;
2601

2602 2603 2604 2605
found:
	h->surplus_huge_pages += delta;
	h->surplus_huge_pages_node[node] += delta;
	return 1;
2606 2607
}

2608
#define persistent_huge_pages(h) (h->nr_huge_pages - h->surplus_huge_pages)
2609
static int set_max_huge_pages(struct hstate *h, unsigned long count, int nid,
2610
			      nodemask_t *nodes_allowed)
L
Linus Torvalds 已提交
2611
{
2612
	unsigned long min_count, ret;
2613 2614 2615 2616 2617 2618 2619 2620 2621 2622 2623
	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
2624

2625 2626
	spin_lock(&hugetlb_lock);

2627 2628 2629 2630 2631 2632 2633 2634 2635 2636 2637 2638 2639 2640 2641 2642 2643 2644 2645 2646
	/*
	 * 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;
	}

2647 2648 2649 2650 2651 2652 2653 2654 2655 2656
	/*
	 * Gigantic pages runtime allocation depend on the capability for large
	 * page range allocation.
	 * If the system does not provide this feature, return an error when
	 * the user tries to allocate gigantic pages but let the user free the
	 * boottime allocated gigantic pages.
	 */
	if (hstate_is_gigantic(h) && !IS_ENABLED(CONFIG_CONTIG_ALLOC)) {
		if (count > persistent_huge_pages(h)) {
			spin_unlock(&hugetlb_lock);
2657
			NODEMASK_FREE(node_alloc_noretry);
2658 2659 2660 2661
			return -EINVAL;
		}
		/* Fall through to decrease pool */
	}
2662

2663 2664 2665 2666
	/*
	 * Increase the pool size
	 * First take pages out of surplus state.  Then make up the
	 * remaining difference by allocating fresh huge pages.
2667
	 *
2668
	 * We might race with alloc_surplus_huge_page() here and be unable
2669 2670 2671 2672
	 * 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.
2673
	 */
2674
	while (h->surplus_huge_pages && count > persistent_huge_pages(h)) {
2675
		if (!adjust_pool_surplus(h, nodes_allowed, -1))
2676 2677 2678
			break;
	}

2679
	while (count > persistent_huge_pages(h)) {
2680 2681 2682 2683 2684 2685
		/*
		 * If this allocation races such that we no longer need the
		 * page, free_huge_page will handle it by freeing the page
		 * and reducing the surplus.
		 */
		spin_unlock(&hugetlb_lock);
2686 2687 2688 2689

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

2690 2691
		ret = alloc_pool_huge_page(h, nodes_allowed,
						node_alloc_noretry);
2692 2693 2694 2695
		spin_lock(&hugetlb_lock);
		if (!ret)
			goto out;

2696 2697 2698
		/* Bail for signals. Probably ctrl-c from user */
		if (signal_pending(current))
			goto out;
2699 2700 2701 2702 2703 2704 2705 2706
	}

	/*
	 * 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.
2707 2708 2709 2710
	 *
	 * 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
2711
	 * alloc_surplus_huge_page() is checking the global counter,
2712 2713 2714
	 * 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.
2715
	 */
2716
	min_count = h->resv_huge_pages + h->nr_huge_pages - h->free_huge_pages;
2717
	min_count = max(count, min_count);
2718
	try_to_free_low(h, min_count, nodes_allowed);
2719
	while (min_count < persistent_huge_pages(h)) {
2720
		if (!free_pool_huge_page(h, nodes_allowed, 0))
L
Linus Torvalds 已提交
2721
			break;
2722
		cond_resched_lock(&hugetlb_lock);
L
Linus Torvalds 已提交
2723
	}
2724
	while (count < persistent_huge_pages(h)) {
2725
		if (!adjust_pool_surplus(h, nodes_allowed, 1))
2726 2727 2728
			break;
	}
out:
2729
	h->max_huge_pages = persistent_huge_pages(h);
L
Linus Torvalds 已提交
2730
	spin_unlock(&hugetlb_lock);
2731

2732 2733
	NODEMASK_FREE(node_alloc_noretry);

2734
	return 0;
L
Linus Torvalds 已提交
2735 2736
}

2737 2738 2739 2740 2741 2742 2743 2744 2745 2746
#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];

2747 2748 2749
static struct hstate *kobj_to_node_hstate(struct kobject *kobj, int *nidp);

static struct hstate *kobj_to_hstate(struct kobject *kobj, int *nidp)
2750 2751
{
	int i;
2752

2753
	for (i = 0; i < HUGE_MAX_HSTATE; i++)
2754 2755 2756
		if (hstate_kobjs[i] == kobj) {
			if (nidp)
				*nidp = NUMA_NO_NODE;
2757
			return &hstates[i];
2758 2759 2760
		}

	return kobj_to_node_hstate(kobj, nidp);
2761 2762
}

2763
static ssize_t nr_hugepages_show_common(struct kobject *kobj,
2764 2765
					struct kobj_attribute *attr, char *buf)
{
2766 2767 2768 2769 2770 2771 2772 2773 2774 2775
	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];

2776
	return sysfs_emit(buf, "%lu\n", nr_huge_pages);
2777
}
2778

2779 2780 2781
static ssize_t __nr_hugepages_store_common(bool obey_mempolicy,
					   struct hstate *h, int nid,
					   unsigned long count, size_t len)
2782 2783
{
	int err;
2784
	nodemask_t nodes_allowed, *n_mask;
2785

2786 2787
	if (hstate_is_gigantic(h) && !gigantic_page_runtime_supported())
		return -EINVAL;
2788

2789 2790 2791 2792 2793
	if (nid == NUMA_NO_NODE) {
		/*
		 * global hstate attribute
		 */
		if (!(obey_mempolicy &&
2794 2795 2796 2797 2798
				init_nodemask_of_mempolicy(&nodes_allowed)))
			n_mask = &node_states[N_MEMORY];
		else
			n_mask = &nodes_allowed;
	} else {
2799
		/*
2800 2801
		 * Node specific request.  count adjustment happens in
		 * set_max_huge_pages() after acquiring hugetlb_lock.
2802
		 */
2803 2804
		init_nodemask_of_node(&nodes_allowed, nid);
		n_mask = &nodes_allowed;
2805
	}
2806

2807
	err = set_max_huge_pages(h, count, nid, n_mask);
2808

2809
	return err ? err : len;
2810 2811
}

2812 2813 2814 2815 2816 2817 2818 2819 2820 2821 2822 2823 2824 2825 2826 2827 2828
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);
}

2829 2830 2831 2832 2833 2834 2835 2836 2837
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)
{
2838
	return nr_hugepages_store_common(false, kobj, buf, len);
2839 2840 2841
}
HSTATE_ATTR(nr_hugepages);

2842 2843 2844 2845 2846 2847 2848
#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,
2849 2850
					   struct kobj_attribute *attr,
					   char *buf)
2851 2852 2853 2854 2855 2856 2857
{
	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)
{
2858
	return nr_hugepages_store_common(true, kobj, buf, len);
2859 2860 2861 2862 2863
}
HSTATE_ATTR(nr_hugepages_mempolicy);
#endif


2864 2865 2866
static ssize_t nr_overcommit_hugepages_show(struct kobject *kobj,
					struct kobj_attribute *attr, char *buf)
{
2867
	struct hstate *h = kobj_to_hstate(kobj, NULL);
2868
	return sysfs_emit(buf, "%lu\n", h->nr_overcommit_huge_pages);
2869
}
2870

2871 2872 2873 2874 2875
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;
2876
	struct hstate *h = kobj_to_hstate(kobj, NULL);
2877

2878
	if (hstate_is_gigantic(h))
2879 2880
		return -EINVAL;

2881
	err = kstrtoul(buf, 10, &input);
2882
	if (err)
2883
		return err;
2884 2885 2886 2887 2888 2889 2890 2891 2892 2893 2894 2895

	spin_lock(&hugetlb_lock);
	h->nr_overcommit_huge_pages = input;
	spin_unlock(&hugetlb_lock);

	return count;
}
HSTATE_ATTR(nr_overcommit_hugepages);

static ssize_t free_hugepages_show(struct kobject *kobj,
					struct kobj_attribute *attr, char *buf)
{
2896 2897 2898 2899 2900 2901 2902 2903 2904 2905
	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];

2906
	return sysfs_emit(buf, "%lu\n", free_huge_pages);
2907 2908 2909 2910 2911 2912
}
HSTATE_ATTR_RO(free_hugepages);

static ssize_t resv_hugepages_show(struct kobject *kobj,
					struct kobj_attribute *attr, char *buf)
{
2913
	struct hstate *h = kobj_to_hstate(kobj, NULL);
2914
	return sysfs_emit(buf, "%lu\n", h->resv_huge_pages);
2915 2916 2917 2918 2919 2920
}
HSTATE_ATTR_RO(resv_hugepages);

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

2931
	return sysfs_emit(buf, "%lu\n", surplus_huge_pages);
2932 2933 2934 2935 2936 2937 2938 2939 2940
}
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,
2941 2942 2943
#ifdef CONFIG_NUMA
	&nr_hugepages_mempolicy_attr.attr,
#endif
2944 2945 2946
	NULL,
};

2947
static const struct attribute_group hstate_attr_group = {
2948 2949 2950
	.attrs = hstate_attrs,
};

J
Jeff Mahoney 已提交
2951 2952
static int hugetlb_sysfs_add_hstate(struct hstate *h, struct kobject *parent,
				    struct kobject **hstate_kobjs,
2953
				    const struct attribute_group *hstate_attr_group)
2954 2955
{
	int retval;
2956
	int hi = hstate_index(h);
2957

2958 2959
	hstate_kobjs[hi] = kobject_create_and_add(h->name, parent);
	if (!hstate_kobjs[hi])
2960 2961
		return -ENOMEM;

2962
	retval = sysfs_create_group(hstate_kobjs[hi], hstate_attr_group);
2963
	if (retval) {
2964
		kobject_put(hstate_kobjs[hi]);
2965 2966
		hstate_kobjs[hi] = NULL;
	}
2967 2968 2969 2970 2971 2972 2973 2974 2975 2976 2977 2978 2979 2980

	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) {
2981 2982
		err = hugetlb_sysfs_add_hstate(h, hugepages_kobj,
					 hstate_kobjs, &hstate_attr_group);
2983
		if (err)
2984
			pr_err("HugeTLB: Unable to add hstate %s", h->name);
2985 2986 2987
	}
}

2988 2989 2990 2991
#ifdef CONFIG_NUMA

/*
 * node_hstate/s - associate per node hstate attributes, via their kobjects,
2992 2993 2994
 * 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
2995 2996 2997 2998 2999 3000
 * the base kernel, on the hugetlb module.
 */
struct node_hstate {
	struct kobject		*hugepages_kobj;
	struct kobject		*hstate_kobjs[HUGE_MAX_HSTATE];
};
3001
static struct node_hstate node_hstates[MAX_NUMNODES];
3002 3003

/*
3004
 * A subset of global hstate attributes for node devices
3005 3006 3007 3008 3009 3010 3011 3012
 */
static struct attribute *per_node_hstate_attrs[] = {
	&nr_hugepages_attr.attr,
	&free_hugepages_attr.attr,
	&surplus_hugepages_attr.attr,
	NULL,
};

3013
static const struct attribute_group per_node_hstate_attr_group = {
3014 3015 3016 3017
	.attrs = per_node_hstate_attrs,
};

/*
3018
 * kobj_to_node_hstate - lookup global hstate for node device hstate attr kobj.
3019 3020 3021 3022 3023 3024 3025 3026 3027 3028 3029 3030 3031 3032 3033 3034 3035 3036 3037 3038 3039 3040
 * 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;
}

/*
3041
 * Unregister hstate attributes from a single node device.
3042 3043
 * No-op if no hstate attributes attached.
 */
3044
static void hugetlb_unregister_node(struct node *node)
3045 3046
{
	struct hstate *h;
3047
	struct node_hstate *nhs = &node_hstates[node->dev.id];
3048 3049

	if (!nhs->hugepages_kobj)
3050
		return;		/* no hstate attributes */
3051

3052 3053 3054 3055 3056
	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;
3057
		}
3058
	}
3059 3060 3061 3062 3063 3064 3065

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


/*
3066
 * Register hstate attributes for a single node device.
3067 3068
 * No-op if attributes already registered.
 */
3069
static void hugetlb_register_node(struct node *node)
3070 3071
{
	struct hstate *h;
3072
	struct node_hstate *nhs = &node_hstates[node->dev.id];
3073 3074 3075 3076 3077 3078
	int err;

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

	nhs->hugepages_kobj = kobject_create_and_add("hugepages",
3079
							&node->dev.kobj);
3080 3081 3082 3083 3084 3085 3086 3087
	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) {
3088
			pr_err("HugeTLB: Unable to add hstate %s for node %d\n",
3089
				h->name, node->dev.id);
3090 3091 3092 3093 3094 3095 3096
			hugetlb_unregister_node(node);
			break;
		}
	}
}

/*
3097
 * hugetlb init time:  register hstate attributes for all registered node
3098 3099
 * devices of nodes that have memory.  All on-line nodes should have
 * registered their associated device by this time.
3100
 */
3101
static void __init hugetlb_register_all_nodes(void)
3102 3103 3104
{
	int nid;

3105
	for_each_node_state(nid, N_MEMORY) {
3106
		struct node *node = node_devices[nid];
3107
		if (node->dev.id == nid)
3108 3109 3110 3111
			hugetlb_register_node(node);
	}

	/*
3112
	 * Let the node device driver know we're here so it can
3113 3114 3115 3116 3117 3118 3119 3120 3121 3122 3123 3124 3125 3126 3127 3128 3129 3130 3131
	 * [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

3132 3133
static int __init hugetlb_init(void)
{
3134 3135
	int i;

3136 3137 3138
	BUILD_BUG_ON(sizeof_field(struct page, private) * BITS_PER_BYTE <
			__NR_HPAGEFLAGS);

3139 3140 3141
	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");
3142
		return 0;
3143
	}
3144

3145 3146 3147 3148 3149 3150 3151 3152 3153 3154 3155 3156 3157 3158 3159 3160 3161 3162 3163 3164 3165 3166 3167 3168 3169 3170 3171 3172
	/*
	 * 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;
3173
		}
3174
	}
3175

3176
	hugetlb_cma_check();
3177
	hugetlb_init_hstates();
3178
	gather_bootmem_prealloc();
3179 3180 3181
	report_hugepages();

	hugetlb_sysfs_init();
3182
	hugetlb_register_all_nodes();
3183
	hugetlb_cgroup_file_init();
3184

3185 3186 3187 3188 3189
#ifdef CONFIG_SMP
	num_fault_mutexes = roundup_pow_of_two(8 * num_possible_cpus());
#else
	num_fault_mutexes = 1;
#endif
3190
	hugetlb_fault_mutex_table =
3191 3192
		kmalloc_array(num_fault_mutexes, sizeof(struct mutex),
			      GFP_KERNEL);
3193
	BUG_ON(!hugetlb_fault_mutex_table);
3194 3195

	for (i = 0; i < num_fault_mutexes; i++)
3196
		mutex_init(&hugetlb_fault_mutex_table[i]);
3197 3198
	return 0;
}
3199
subsys_initcall(hugetlb_init);
3200

3201 3202
/* Overwritten by architectures with more huge page sizes */
bool __init __attribute((weak)) arch_hugetlb_valid_size(unsigned long size)
3203
{
3204
	return size == HPAGE_SIZE;
3205 3206
}

3207
void __init hugetlb_add_hstate(unsigned int order)
3208 3209
{
	struct hstate *h;
3210 3211
	unsigned long i;

3212 3213 3214
	if (size_to_hstate(PAGE_SIZE << order)) {
		return;
	}
3215
	BUG_ON(hugetlb_max_hstate >= HUGE_MAX_HSTATE);
3216
	BUG_ON(order == 0);
3217
	h = &hstates[hugetlb_max_hstate++];
3218
	h->order = order;
3219
	h->mask = ~(huge_page_size(h) - 1);
3220 3221
	for (i = 0; i < MAX_NUMNODES; ++i)
		INIT_LIST_HEAD(&h->hugepage_freelists[i]);
3222
	INIT_LIST_HEAD(&h->hugepage_activelist);
3223 3224
	h->next_nid_to_alloc = first_memory_node;
	h->next_nid_to_free = first_memory_node;
3225 3226
	snprintf(h->name, HSTATE_NAME_LEN, "hugepages-%lukB",
					huge_page_size(h)/1024);
3227

3228 3229 3230
	parsed_hstate = h;
}

3231 3232 3233 3234 3235 3236 3237 3238
/*
 * 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)
3239 3240
{
	unsigned long *mhp;
3241
	static unsigned long *last_mhp;
3242

3243
	if (!parsed_valid_hugepagesz) {
3244
		pr_warn("HugeTLB: hugepages=%s does not follow a valid hugepagesz, ignoring\n", s);
3245
		parsed_valid_hugepagesz = true;
3246
		return 0;
3247
	}
3248

3249
	/*
3250 3251 3252 3253
	 * !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.
3254
	 */
3255
	else if (!hugetlb_max_hstate)
3256 3257 3258 3259
		mhp = &default_hstate_max_huge_pages;
	else
		mhp = &parsed_hstate->max_huge_pages;

3260
	if (mhp == last_mhp) {
3261 3262
		pr_warn("HugeTLB: hugepages= specified twice without interleaving hugepagesz=, ignoring hugepages=%s\n", s);
		return 0;
3263 3264
	}

3265 3266 3267
	if (sscanf(s, "%lu", mhp) <= 0)
		*mhp = 0;

3268 3269 3270 3271 3272
	/*
	 * Global state is always initialized later in hugetlb_init.
	 * But we need to allocate >= MAX_ORDER hstates here early to still
	 * use the bootmem allocator.
	 */
3273
	if (hugetlb_max_hstate && parsed_hstate->order >= MAX_ORDER)
3274 3275 3276 3277
		hugetlb_hstate_alloc_pages(parsed_hstate);

	last_mhp = mhp;

3278 3279
	return 1;
}
3280
__setup("hugepages=", hugepages_setup);
3281

3282 3283 3284 3285 3286 3287 3288
/*
 * 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.
 */
3289
static int __init hugepagesz_setup(char *s)
3290
{
3291
	unsigned long size;
3292 3293 3294
	struct hstate *h;

	parsed_valid_hugepagesz = false;
3295 3296 3297
	size = (unsigned long)memparse(s, NULL);

	if (!arch_hugetlb_valid_size(size)) {
3298
		pr_err("HugeTLB: unsupported hugepagesz=%s\n", s);
3299 3300 3301
		return 0;
	}

3302 3303 3304 3305 3306 3307 3308 3309 3310 3311 3312 3313 3314 3315 3316 3317 3318 3319 3320 3321 3322 3323 3324
	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;
3325 3326
	}

3327
	hugetlb_add_hstate(ilog2(size) - PAGE_SHIFT);
3328
	parsed_valid_hugepagesz = true;
3329 3330
	return 1;
}
3331 3332
__setup("hugepagesz=", hugepagesz_setup);

3333 3334 3335 3336
/*
 * default_hugepagesz command line input
 * Only one instance of default_hugepagesz allowed on command line.
 */
3337
static int __init default_hugepagesz_setup(char *s)
3338
{
3339 3340
	unsigned long size;

3341 3342 3343 3344 3345 3346
	parsed_valid_hugepagesz = false;
	if (parsed_default_hugepagesz) {
		pr_err("HugeTLB: default_hugepagesz previously specified, ignoring %s\n", s);
		return 0;
	}

3347 3348 3349
	size = (unsigned long)memparse(s, NULL);

	if (!arch_hugetlb_valid_size(size)) {
3350
		pr_err("HugeTLB: unsupported default_hugepagesz=%s\n", s);
3351 3352 3353
		return 0;
	}

3354 3355 3356 3357 3358 3359 3360 3361 3362 3363 3364 3365 3366 3367 3368 3369 3370 3371 3372
	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;
	}

3373 3374
	return 1;
}
3375
__setup("default_hugepagesz=", default_hugepagesz_setup);
3376

3377
static unsigned int allowed_mems_nr(struct hstate *h)
3378 3379 3380
{
	int node;
	unsigned int nr = 0;
3381 3382 3383 3384 3385
	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);
3386

3387
	for_each_node_mask(node, cpuset_current_mems_allowed) {
3388
		if (!mpol_allowed || node_isset(node, *mpol_allowed))
3389 3390
			nr += array[node];
	}
3391 3392 3393 3394 3395

	return nr;
}

#ifdef CONFIG_SYSCTL
3396 3397 3398 3399 3400 3401 3402 3403 3404 3405 3406 3407 3408 3409 3410 3411
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);
}

3412 3413
static int hugetlb_sysctl_handler_common(bool obey_mempolicy,
			 struct ctl_table *table, int write,
3414
			 void *buffer, size_t *length, loff_t *ppos)
L
Linus Torvalds 已提交
3415
{
3416
	struct hstate *h = &default_hstate;
3417
	unsigned long tmp = h->max_huge_pages;
3418
	int ret;
3419

3420
	if (!hugepages_supported())
3421
		return -EOPNOTSUPP;
3422

3423 3424
	ret = proc_hugetlb_doulongvec_minmax(table, write, buffer, length, ppos,
					     &tmp);
3425 3426
	if (ret)
		goto out;
3427

3428 3429 3430
	if (write)
		ret = __nr_hugepages_store_common(obey_mempolicy, h,
						  NUMA_NO_NODE, tmp, *length);
3431 3432
out:
	return ret;
L
Linus Torvalds 已提交
3433
}
3434

3435
int hugetlb_sysctl_handler(struct ctl_table *table, int write,
3436
			  void *buffer, size_t *length, loff_t *ppos)
3437 3438 3439 3440 3441 3442 3443 3444
{

	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,
3445
			  void *buffer, size_t *length, loff_t *ppos)
3446 3447 3448 3449 3450 3451
{
	return hugetlb_sysctl_handler_common(true, table, write,
							buffer, length, ppos);
}
#endif /* CONFIG_NUMA */

3452
int hugetlb_overcommit_handler(struct ctl_table *table, int write,
3453
		void *buffer, size_t *length, loff_t *ppos)
3454
{
3455
	struct hstate *h = &default_hstate;
3456
	unsigned long tmp;
3457
	int ret;
3458

3459
	if (!hugepages_supported())
3460
		return -EOPNOTSUPP;
3461

3462
	tmp = h->nr_overcommit_huge_pages;
3463

3464
	if (write && hstate_is_gigantic(h))
3465 3466
		return -EINVAL;

3467 3468
	ret = proc_hugetlb_doulongvec_minmax(table, write, buffer, length, ppos,
					     &tmp);
3469 3470
	if (ret)
		goto out;
3471 3472 3473 3474 3475 3476

	if (write) {
		spin_lock(&hugetlb_lock);
		h->nr_overcommit_huge_pages = tmp;
		spin_unlock(&hugetlb_lock);
	}
3477 3478
out:
	return ret;
3479 3480
}

L
Linus Torvalds 已提交
3481 3482
#endif /* CONFIG_SYSCTL */

3483
void hugetlb_report_meminfo(struct seq_file *m)
L
Linus Torvalds 已提交
3484
{
3485 3486 3487
	struct hstate *h;
	unsigned long total = 0;

3488 3489
	if (!hugepages_supported())
		return;
3490 3491 3492 3493

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

3494
		total += huge_page_size(h) * count;
3495 3496 3497 3498 3499 3500 3501 3502 3503 3504 3505 3506

		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,
3507
				   huge_page_size(h) / SZ_1K);
3508 3509
	}

3510
	seq_printf(m, "Hugetlb:        %8lu kB\n", total / SZ_1K);
L
Linus Torvalds 已提交
3511 3512
}

3513
int hugetlb_report_node_meminfo(char *buf, int len, int nid)
L
Linus Torvalds 已提交
3514
{
3515
	struct hstate *h = &default_hstate;
3516

3517 3518
	if (!hugepages_supported())
		return 0;
3519 3520 3521 3522 3523 3524 3525 3526

	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
3527 3528
}

3529 3530 3531 3532 3533
void hugetlb_show_meminfo(void)
{
	struct hstate *h;
	int nid;

3534 3535 3536
	if (!hugepages_supported())
		return;

3537 3538 3539 3540 3541 3542 3543
	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],
3544
				huge_page_size(h) / SZ_1K);
3545 3546
}

3547 3548 3549 3550 3551 3552
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
3553 3554 3555
/* Return the number pages of memory we physically have, in PAGE_SIZE units. */
unsigned long hugetlb_total_pages(void)
{
3556 3557 3558 3559 3560 3561
	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
3562 3563
}

3564
static int hugetlb_acct_memory(struct hstate *h, long delta)
M
Mel Gorman 已提交
3565 3566 3567
{
	int ret = -ENOMEM;

3568 3569 3570
	if (!delta)
		return 0;

M
Mel Gorman 已提交
3571 3572 3573 3574 3575 3576 3577 3578 3579 3580 3581 3582 3583 3584 3585 3586 3587
	spin_lock(&hugetlb_lock);
	/*
	 * When cpuset is configured, it breaks the strict hugetlb page
	 * reservation as the accounting is done on a global variable. Such
	 * reservation is completely rubbish in the presence of cpuset because
	 * the reservation is not checked against page availability for the
	 * current cpuset. Application can still potentially OOM'ed by kernel
	 * with lack of free htlb page in cpuset that the task is in.
	 * Attempt to enforce strict accounting with cpuset is almost
	 * impossible (or too ugly) because cpuset is too fluid that
	 * task or memory node can be dynamically moved between cpusets.
	 *
	 * The change of semantics for shared hugetlb mapping with cpuset is
	 * undesirable. However, in order to preserve some of the semantics,
	 * we fall back to check against current free page availability as
	 * a best attempt and hopefully to minimize the impact of changing
	 * semantics that cpuset has.
3588 3589 3590 3591 3592 3593
	 *
	 * 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
3594 3595
	 */
	if (delta > 0) {
3596
		if (gather_surplus_pages(h, delta) < 0)
M
Mel Gorman 已提交
3597 3598
			goto out;

3599
		if (delta > allowed_mems_nr(h)) {
3600
			return_unused_surplus_pages(h, delta);
M
Mel Gorman 已提交
3601 3602 3603 3604 3605 3606
			goto out;
		}
	}

	ret = 0;
	if (delta < 0)
3607
		return_unused_surplus_pages(h, (unsigned long) -delta);
M
Mel Gorman 已提交
3608 3609 3610 3611 3612 3613

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

3614 3615
static void hugetlb_vm_op_open(struct vm_area_struct *vma)
{
3616
	struct resv_map *resv = vma_resv_map(vma);
3617 3618 3619 3620 3621

	/*
	 * 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
3622
	 * has a reference to the reservation map it cannot disappear until
3623 3624 3625
	 * after this open call completes.  It is therefore safe to take a
	 * new reference here without additional locking.
	 */
3626
	if (resv && is_vma_resv_set(vma, HPAGE_RESV_OWNER))
3627
		kref_get(&resv->refs);
3628 3629
}

3630 3631
static void hugetlb_vm_op_close(struct vm_area_struct *vma)
{
3632
	struct hstate *h = hstate_vma(vma);
3633
	struct resv_map *resv = vma_resv_map(vma);
3634
	struct hugepage_subpool *spool = subpool_vma(vma);
3635
	unsigned long reserve, start, end;
3636
	long gbl_reserve;
3637

3638 3639
	if (!resv || !is_vma_resv_set(vma, HPAGE_RESV_OWNER))
		return;
3640

3641 3642
	start = vma_hugecache_offset(h, vma, vma->vm_start);
	end = vma_hugecache_offset(h, vma, vma->vm_end);
3643

3644
	reserve = (end - start) - region_count(resv, start, end);
3645
	hugetlb_cgroup_uncharge_counter(resv, start, end);
3646
	if (reserve) {
3647 3648 3649 3650 3651 3652
		/*
		 * 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);
3653
	}
3654 3655

	kref_put(&resv->refs, resv_map_release);
3656 3657
}

3658 3659 3660 3661 3662 3663 3664
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;
}

3665 3666
static unsigned long hugetlb_vm_op_pagesize(struct vm_area_struct *vma)
{
3667
	return huge_page_size(hstate_vma(vma));
3668 3669
}

L
Linus Torvalds 已提交
3670 3671 3672
/*
 * 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
3673
 * hugepage VMA.  do_page_fault() is supposed to trap this, so BUG is we get
L
Linus Torvalds 已提交
3674 3675
 * this far.
 */
3676
static vm_fault_t hugetlb_vm_op_fault(struct vm_fault *vmf)
L
Linus Torvalds 已提交
3677 3678
{
	BUG();
N
Nick Piggin 已提交
3679
	return 0;
L
Linus Torvalds 已提交
3680 3681
}

3682 3683 3684 3685 3686 3687 3688
/*
 * 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.
 */
3689
const struct vm_operations_struct hugetlb_vm_ops = {
N
Nick Piggin 已提交
3690
	.fault = hugetlb_vm_op_fault,
3691
	.open = hugetlb_vm_op_open,
3692
	.close = hugetlb_vm_op_close,
3693
	.may_split = hugetlb_vm_op_split,
3694
	.pagesize = hugetlb_vm_op_pagesize,
L
Linus Torvalds 已提交
3695 3696
};

3697 3698
static pte_t make_huge_pte(struct vm_area_struct *vma, struct page *page,
				int writable)
D
David Gibson 已提交
3699 3700 3701
{
	pte_t entry;

3702
	if (writable) {
3703 3704
		entry = huge_pte_mkwrite(huge_pte_mkdirty(mk_huge_pte(page,
					 vma->vm_page_prot)));
D
David Gibson 已提交
3705
	} else {
3706 3707
		entry = huge_pte_wrprotect(mk_huge_pte(page,
					   vma->vm_page_prot));
D
David Gibson 已提交
3708 3709 3710
	}
	entry = pte_mkyoung(entry);
	entry = pte_mkhuge(entry);
3711
	entry = arch_make_huge_pte(entry, vma, page, writable);
D
David Gibson 已提交
3712 3713 3714 3715

	return entry;
}

3716 3717 3718 3719 3720
static void set_huge_ptep_writable(struct vm_area_struct *vma,
				   unsigned long address, pte_t *ptep)
{
	pte_t entry;

3721
	entry = huge_pte_mkwrite(huge_pte_mkdirty(huge_ptep_get(ptep)));
3722
	if (huge_ptep_set_access_flags(vma, address, ptep, entry, 1))
3723
		update_mmu_cache(vma, address, ptep);
3724 3725
}

3726
bool is_hugetlb_entry_migration(pte_t pte)
3727 3728 3729 3730
{
	swp_entry_t swp;

	if (huge_pte_none(pte) || pte_present(pte))
3731
		return false;
3732
	swp = pte_to_swp_entry(pte);
3733
	if (is_migration_entry(swp))
3734
		return true;
3735
	else
3736
		return false;
3737 3738
}

3739
static bool is_hugetlb_entry_hwpoisoned(pte_t pte)
3740 3741 3742 3743
{
	swp_entry_t swp;

	if (huge_pte_none(pte) || pte_present(pte))
3744
		return false;
3745
	swp = pte_to_swp_entry(pte);
3746
	if (is_hwpoison_entry(swp))
3747
		return true;
3748
	else
3749
		return false;
3750
}
3751

3752 3753 3754 3755 3756 3757 3758 3759 3760 3761 3762 3763
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
3764 3765 3766
int copy_hugetlb_page_range(struct mm_struct *dst, struct mm_struct *src,
			    struct vm_area_struct *vma)
{
3767
	pte_t *src_pte, *dst_pte, entry, dst_entry;
D
David Gibson 已提交
3768
	struct page *ptepage;
3769
	unsigned long addr;
3770
	bool cow = is_cow_mapping(vma->vm_flags);
3771 3772
	struct hstate *h = hstate_vma(vma);
	unsigned long sz = huge_page_size(h);
3773
	unsigned long npages = pages_per_huge_page(h);
3774
	struct address_space *mapping = vma->vm_file->f_mapping;
3775
	struct mmu_notifier_range range;
3776
	int ret = 0;
3777

3778
	if (cow) {
3779
		mmu_notifier_range_init(&range, MMU_NOTIFY_CLEAR, 0, vma, src,
3780
					vma->vm_start,
3781 3782
					vma->vm_end);
		mmu_notifier_invalidate_range_start(&range);
3783 3784 3785 3786 3787 3788 3789 3790
	} 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);
3791
	}
3792

3793
	for (addr = vma->vm_start; addr < vma->vm_end; addr += sz) {
3794
		spinlock_t *src_ptl, *dst_ptl;
3795
		src_pte = huge_pte_offset(src, addr, sz);
H
Hugh Dickins 已提交
3796 3797
		if (!src_pte)
			continue;
3798
		dst_pte = huge_pte_alloc(dst, vma, addr, sz);
3799 3800 3801 3802
		if (!dst_pte) {
			ret = -ENOMEM;
			break;
		}
3803

3804 3805 3806 3807 3808 3809 3810 3811 3812 3813 3814
		/*
		 * 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))
3815 3816
			continue;

3817 3818 3819
		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);
3820
		entry = huge_ptep_get(src_pte);
3821
		dst_entry = huge_ptep_get(dst_pte);
3822
again:
3823 3824 3825 3826 3827 3828
		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.
			 */
3829 3830 3831 3832 3833 3834 3835 3836 3837 3838 3839 3840
			;
		} 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);
3841 3842
				set_huge_swap_pte_at(src, addr, src_pte,
						     entry, sz);
3843
			}
3844
			set_huge_swap_pte_at(dst, addr, dst_pte, entry, sz);
3845
		} else {
3846 3847 3848 3849 3850 3851 3852 3853 3854 3855 3856 3857 3858 3859 3860 3861 3862 3863 3864 3865 3866 3867 3868 3869 3870 3871 3872 3873 3874 3875 3876 3877 3878 3879 3880 3881 3882 3883 3884 3885 3886 3887 3888 3889 3890 3891
			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;
			}

3892
			if (cow) {
3893 3894 3895 3896 3897
				/*
				 * No need to notify as we are downgrading page
				 * table protection not changing it to point
				 * to a new page.
				 *
3898
				 * See Documentation/vm/mmu_notifier.rst
3899
				 */
3900
				huge_ptep_set_wrprotect(src, addr, src_pte);
3901
			}
3902

3903
			page_dup_rmap(ptepage, true);
3904
			set_huge_pte_at(dst, addr, dst_pte, entry);
3905
			hugetlb_count_add(npages, dst);
3906
		}
3907 3908
		spin_unlock(src_ptl);
		spin_unlock(dst_ptl);
D
David Gibson 已提交
3909 3910
	}

3911
	if (cow)
3912
		mmu_notifier_invalidate_range_end(&range);
3913 3914
	else
		i_mmap_unlock_read(mapping);
3915 3916

	return ret;
D
David Gibson 已提交
3917 3918
}

3919 3920 3921
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
3922 3923 3924
{
	struct mm_struct *mm = vma->vm_mm;
	unsigned long address;
3925
	pte_t *ptep;
D
David Gibson 已提交
3926
	pte_t pte;
3927
	spinlock_t *ptl;
D
David Gibson 已提交
3928
	struct page *page;
3929 3930
	struct hstate *h = hstate_vma(vma);
	unsigned long sz = huge_page_size(h);
3931
	struct mmu_notifier_range range;
3932

D
David Gibson 已提交
3933
	WARN_ON(!is_vm_hugetlb_page(vma));
3934 3935
	BUG_ON(start & ~huge_page_mask(h));
	BUG_ON(end & ~huge_page_mask(h));
D
David Gibson 已提交
3936

3937 3938 3939 3940
	/*
	 * This is a hugetlb vma, all the pte entries should point
	 * to huge page.
	 */
3941
	tlb_change_page_size(tlb, sz);
3942
	tlb_start_vma(tlb, vma);
3943 3944 3945 3946

	/*
	 * If sharing possible, alert mmu notifiers of worst case.
	 */
3947 3948
	mmu_notifier_range_init(&range, MMU_NOTIFY_UNMAP, 0, vma, mm, start,
				end);
3949 3950
	adjust_range_if_pmd_sharing_possible(vma, &range.start, &range.end);
	mmu_notifier_invalidate_range_start(&range);
3951 3952
	address = start;
	for (; address < end; address += sz) {
3953
		ptep = huge_pte_offset(mm, address, sz);
A
Adam Litke 已提交
3954
		if (!ptep)
3955 3956
			continue;

3957
		ptl = huge_pte_lock(h, mm, ptep);
3958
		if (huge_pmd_unshare(mm, vma, &address, ptep)) {
3959
			spin_unlock(ptl);
3960 3961 3962 3963
			/*
			 * We just unmapped a page of PMDs by clearing a PUD.
			 * The caller's TLB flush range should cover this area.
			 */
3964 3965
			continue;
		}
3966

3967
		pte = huge_ptep_get(ptep);
3968 3969 3970 3971
		if (huge_pte_none(pte)) {
			spin_unlock(ptl);
			continue;
		}
3972 3973

		/*
3974 3975
		 * Migrating hugepage or HWPoisoned hugepage is already
		 * unmapped and its refcount is dropped, so just clear pte here.
3976
		 */
3977
		if (unlikely(!pte_present(pte))) {
3978
			huge_pte_clear(mm, address, ptep, sz);
3979 3980
			spin_unlock(ptl);
			continue;
3981
		}
3982 3983

		page = pte_page(pte);
3984 3985 3986 3987 3988 3989
		/*
		 * 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) {
3990 3991 3992 3993
			if (page != ref_page) {
				spin_unlock(ptl);
				continue;
			}
3994 3995 3996 3997 3998 3999 4000 4001
			/*
			 * 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);
		}

4002
		pte = huge_ptep_get_and_clear(mm, address, ptep);
4003
		tlb_remove_huge_tlb_entry(h, tlb, ptep, address);
4004
		if (huge_pte_dirty(pte))
4005
			set_page_dirty(page);
4006

4007
		hugetlb_count_sub(pages_per_huge_page(h), mm);
4008
		page_remove_rmap(page, true);
4009

4010
		spin_unlock(ptl);
4011
		tlb_remove_page_size(tlb, page, huge_page_size(h));
4012 4013 4014 4015 4016
		/*
		 * Bail out after unmapping reference page if supplied
		 */
		if (ref_page)
			break;
4017
	}
4018
	mmu_notifier_invalidate_range_end(&range);
4019
	tlb_end_vma(tlb, vma);
L
Linus Torvalds 已提交
4020
}
D
David Gibson 已提交
4021

4022 4023 4024 4025 4026 4027 4028 4029 4030 4031 4032 4033
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
4034
	 * is to clear it before releasing the i_mmap_rwsem. This works
4035
	 * because in the context this is called, the VMA is about to be
4036
	 * destroyed and the i_mmap_rwsem is held.
4037 4038 4039 4040
	 */
	vma->vm_flags &= ~VM_MAYSHARE;
}

4041
void unmap_hugepage_range(struct vm_area_struct *vma, unsigned long start,
4042
			  unsigned long end, struct page *ref_page)
4043
{
4044
	struct mmu_gather tlb;
4045

4046
	tlb_gather_mmu(&tlb, vma->vm_mm);
4047
	__unmap_hugepage_range(&tlb, vma, start, end, ref_page);
4048
	tlb_finish_mmu(&tlb);
4049 4050
}

4051 4052
/*
 * This is called when the original mapper is failing to COW a MAP_PRIVATE
4053
 * mapping it owns the reserve page for. The intention is to unmap the page
4054 4055 4056
 * from other VMAs and let the children be SIGKILLed if they are faulting the
 * same region.
 */
4057 4058
static void unmap_ref_private(struct mm_struct *mm, struct vm_area_struct *vma,
			      struct page *page, unsigned long address)
4059
{
4060
	struct hstate *h = hstate_vma(vma);
4061 4062 4063 4064 4065 4066 4067 4068
	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.
	 */
4069
	address = address & huge_page_mask(h);
4070 4071
	pgoff = ((address - vma->vm_start) >> PAGE_SHIFT) +
			vma->vm_pgoff;
4072
	mapping = vma->vm_file->f_mapping;
4073

4074 4075 4076 4077 4078
	/*
	 * 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
	 */
4079
	i_mmap_lock_write(mapping);
4080
	vma_interval_tree_foreach(iter_vma, &mapping->i_mmap, pgoff, pgoff) {
4081 4082 4083 4084
		/* Do not unmap the current VMA */
		if (iter_vma == vma)
			continue;

4085 4086 4087 4088 4089 4090 4091 4092
		/*
		 * 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;

4093 4094 4095 4096 4097 4098 4099 4100
		/*
		 * 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))
4101 4102
			unmap_hugepage_range(iter_vma, address,
					     address + huge_page_size(h), page);
4103
	}
4104
	i_mmap_unlock_write(mapping);
4105 4106
}

4107 4108
/*
 * Hugetlb_cow() should be called with page lock of the original hugepage held.
4109 4110 4111
 * 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.
4112
 */
4113
static vm_fault_t hugetlb_cow(struct mm_struct *mm, struct vm_area_struct *vma,
4114
		       unsigned long address, pte_t *ptep,
4115
		       struct page *pagecache_page, spinlock_t *ptl)
4116
{
4117
	pte_t pte;
4118
	struct hstate *h = hstate_vma(vma);
4119
	struct page *old_page, *new_page;
4120 4121
	int outside_reserve = 0;
	vm_fault_t ret = 0;
4122
	unsigned long haddr = address & huge_page_mask(h);
4123
	struct mmu_notifier_range range;
4124

4125
	pte = huge_ptep_get(ptep);
4126 4127
	old_page = pte_page(pte);

4128
retry_avoidcopy:
4129 4130
	/* If no-one else is actually using this page, avoid the copy
	 * and just make the page writable */
4131
	if (page_mapcount(old_page) == 1 && PageAnon(old_page)) {
4132
		page_move_anon_rmap(old_page, vma);
4133
		set_huge_ptep_writable(vma, haddr, ptep);
N
Nick Piggin 已提交
4134
		return 0;
4135 4136
	}

4137 4138 4139 4140 4141 4142 4143 4144 4145
	/*
	 * 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.
	 */
4146
	if (is_vma_resv_set(vma, HPAGE_RESV_OWNER) &&
4147 4148 4149
			old_page != pagecache_page)
		outside_reserve = 1;

4150
	get_page(old_page);
4151

4152 4153 4154 4155
	/*
	 * Drop page table lock as buddy allocator may be called. It will
	 * be acquired again before returning to the caller, as expected.
	 */
4156
	spin_unlock(ptl);
4157
	new_page = alloc_huge_page(vma, haddr, outside_reserve);
4158

4159
	if (IS_ERR(new_page)) {
4160 4161 4162 4163 4164 4165 4166 4167
		/*
		 * 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) {
4168 4169 4170 4171
			struct address_space *mapping = vma->vm_file->f_mapping;
			pgoff_t idx;
			u32 hash;

4172
			put_page(old_page);
4173
			BUG_ON(huge_pte_none(pte));
4174 4175 4176 4177 4178 4179 4180 4181 4182 4183 4184 4185 4186 4187
			/*
			 * 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);

4188
			unmap_ref_private(mm, vma, old_page, haddr);
4189 4190 4191

			i_mmap_lock_read(mapping);
			mutex_lock(&hugetlb_fault_mutex_table[hash]);
4192
			spin_lock(ptl);
4193
			ptep = huge_pte_offset(mm, haddr, huge_page_size(h));
4194 4195 4196 4197 4198 4199 4200 4201
			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;
4202 4203
		}

4204
		ret = vmf_error(PTR_ERR(new_page));
4205
		goto out_release_old;
4206 4207
	}

4208 4209 4210 4211
	/*
	 * When the original hugepage is shared one, it does not have
	 * anon_vma prepared.
	 */
4212
	if (unlikely(anon_vma_prepare(vma))) {
4213 4214
		ret = VM_FAULT_OOM;
		goto out_release_all;
4215
	}
4216

4217
	copy_user_huge_page(new_page, old_page, address, vma,
A
Andrea Arcangeli 已提交
4218
			    pages_per_huge_page(h));
N
Nick Piggin 已提交
4219
	__SetPageUptodate(new_page);
4220

4221
	mmu_notifier_range_init(&range, MMU_NOTIFY_CLEAR, 0, vma, mm, haddr,
4222
				haddr + huge_page_size(h));
4223
	mmu_notifier_invalidate_range_start(&range);
4224

4225
	/*
4226
	 * Retake the page table lock to check for racing updates
4227 4228
	 * before the page tables are altered
	 */
4229
	spin_lock(ptl);
4230
	ptep = huge_pte_offset(mm, haddr, huge_page_size(h));
4231
	if (likely(ptep && pte_same(huge_ptep_get(ptep), pte))) {
4232
		ClearHPageRestoreReserve(new_page);
4233

4234
		/* Break COW */
4235
		huge_ptep_clear_flush(vma, haddr, ptep);
4236
		mmu_notifier_invalidate_range(mm, range.start, range.end);
4237
		set_huge_pte_at(mm, haddr, ptep,
4238
				make_huge_pte(vma, new_page, 1));
4239
		page_remove_rmap(old_page, true);
4240
		hugepage_add_new_anon_rmap(new_page, vma, haddr);
4241
		SetHPageMigratable(new_page);
4242 4243 4244
		/* Make the old page be freed below */
		new_page = old_page;
	}
4245
	spin_unlock(ptl);
4246
	mmu_notifier_invalidate_range_end(&range);
4247
out_release_all:
4248
	restore_reserve_on_error(h, vma, haddr, new_page);
4249
	put_page(new_page);
4250
out_release_old:
4251
	put_page(old_page);
4252

4253 4254
	spin_lock(ptl); /* Caller expects lock to be held */
	return ret;
4255 4256
}

4257
/* Return the pagecache page at a given address within a VMA */
4258 4259
static struct page *hugetlbfs_pagecache_page(struct hstate *h,
			struct vm_area_struct *vma, unsigned long address)
4260 4261
{
	struct address_space *mapping;
4262
	pgoff_t idx;
4263 4264

	mapping = vma->vm_file->f_mapping;
4265
	idx = vma_hugecache_offset(h, vma, address);
4266 4267 4268 4269

	return find_lock_page(mapping, idx);
}

H
Hugh Dickins 已提交
4270 4271 4272 4273 4274
/*
 * 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
4275 4276 4277 4278 4279 4280 4281 4282 4283 4284 4285 4286 4287 4288 4289
			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;
}

4290 4291 4292 4293 4294 4295 4296 4297 4298
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;
4299
	ClearHPageRestoreReserve(page);
4300

4301 4302 4303 4304 4305 4306
	/*
	 * set page dirty so that it will not be removed from cache/file
	 * by non-hugetlbfs specific code paths.
	 */
	set_page_dirty(page);

4307 4308 4309 4310 4311 4312
	spin_lock(&inode->i_lock);
	inode->i_blocks += blocks_per_huge_page(h);
	spin_unlock(&inode->i_lock);
	return 0;
}

4313 4314 4315 4316
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)
4317
{
4318
	struct hstate *h = hstate_vma(vma);
4319
	vm_fault_t ret = VM_FAULT_SIGBUS;
4320
	int anon_rmap = 0;
A
Adam Litke 已提交
4321 4322
	unsigned long size;
	struct page *page;
4323
	pte_t new_pte;
4324
	spinlock_t *ptl;
4325
	unsigned long haddr = address & huge_page_mask(h);
4326
	bool new_page = false;
A
Adam Litke 已提交
4327

4328 4329 4330
	/*
	 * 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
4331
	 * COW. Warn that such a situation has occurred as it may not be obvious
4332 4333
	 */
	if (is_vma_resv_set(vma, HPAGE_RESV_UNMAPPED)) {
4334
		pr_warn_ratelimited("PID %d killed due to inadequate hugepage pool\n",
4335
			   current->pid);
4336 4337 4338
		return ret;
	}

A
Adam Litke 已提交
4339
	/*
4340 4341 4342
	 * 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
4343
	 */
4344 4345 4346 4347
	size = i_size_read(mapping->host) >> huge_page_shift(h);
	if (idx >= size)
		goto out;

4348 4349 4350
retry:
	page = find_lock_page(mapping, idx);
	if (!page) {
4351 4352 4353 4354 4355 4356 4357
		/*
		 * Check for page in userfault range
		 */
		if (userfaultfd_missing(vma)) {
			u32 hash;
			struct vm_fault vmf = {
				.vma = vma,
4358
				.address = haddr,
4359 4360 4361 4362 4363 4364 4365 4366 4367 4368 4369
				.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
				 */
			};

			/*
4370 4371 4372
			 * hugetlb_fault_mutex and i_mmap_rwsem must be
			 * dropped before handling userfault.  Reacquire
			 * after handling fault to make calling code simpler.
4373
			 */
4374
			hash = hugetlb_fault_mutex_hash(mapping, idx);
4375
			mutex_unlock(&hugetlb_fault_mutex_table[hash]);
4376
			i_mmap_unlock_read(mapping);
4377
			ret = handle_userfault(&vmf, VM_UFFD_MISSING);
4378
			i_mmap_lock_read(mapping);
4379 4380 4381 4382
			mutex_lock(&hugetlb_fault_mutex_table[hash]);
			goto out;
		}

4383
		page = alloc_huge_page(vma, haddr, 0);
4384
		if (IS_ERR(page)) {
4385 4386 4387 4388 4389 4390 4391 4392 4393 4394 4395 4396 4397 4398 4399 4400 4401 4402 4403
			/*
			 * Returning error will result in faulting task being
			 * sent SIGBUS.  The hugetlb fault mutex prevents two
			 * tasks from racing to fault in the same page which
			 * could result in false unable to allocate errors.
			 * Page migration does not take the fault mutex, but
			 * does a clear then write of pte's under page table
			 * lock.  Page fault code could race with migration,
			 * notice the clear pte and try to allocate a page
			 * here.  Before returning error, get ptl and make
			 * sure there really is no pte entry.
			 */
			ptl = huge_pte_lock(h, mm, ptep);
			if (!huge_pte_none(huge_ptep_get(ptep))) {
				ret = 0;
				spin_unlock(ptl);
				goto out;
			}
			spin_unlock(ptl);
4404
			ret = vmf_error(PTR_ERR(page));
4405 4406
			goto out;
		}
A
Andrea Arcangeli 已提交
4407
		clear_huge_page(page, address, pages_per_huge_page(h));
N
Nick Piggin 已提交
4408
		__SetPageUptodate(page);
4409
		new_page = true;
4410

4411
		if (vma->vm_flags & VM_MAYSHARE) {
4412
			int err = huge_add_to_page_cache(page, mapping, idx);
4413 4414 4415 4416 4417 4418
			if (err) {
				put_page(page);
				if (err == -EEXIST)
					goto retry;
				goto out;
			}
4419
		} else {
4420
			lock_page(page);
4421 4422 4423 4424
			if (unlikely(anon_vma_prepare(vma))) {
				ret = VM_FAULT_OOM;
				goto backout_unlocked;
			}
4425
			anon_rmap = 1;
4426
		}
4427
	} else {
4428 4429 4430 4431 4432 4433
		/*
		 * 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))) {
4434
			ret = VM_FAULT_HWPOISON_LARGE |
4435
				VM_FAULT_SET_HINDEX(hstate_index(h));
4436 4437
			goto backout_unlocked;
		}
4438
	}
4439

4440 4441 4442 4443 4444 4445
	/*
	 * 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.
	 */
4446
	if ((flags & FAULT_FLAG_WRITE) && !(vma->vm_flags & VM_SHARED)) {
4447
		if (vma_needs_reservation(h, vma, haddr) < 0) {
4448 4449 4450
			ret = VM_FAULT_OOM;
			goto backout_unlocked;
		}
4451
		/* Just decrements count, does not deallocate */
4452
		vma_end_reservation(h, vma, haddr);
4453
	}
4454

4455
	ptl = huge_pte_lock(h, mm, ptep);
N
Nick Piggin 已提交
4456
	ret = 0;
4457
	if (!huge_pte_none(huge_ptep_get(ptep)))
A
Adam Litke 已提交
4458 4459
		goto backout;

4460
	if (anon_rmap) {
4461
		ClearHPageRestoreReserve(page);
4462
		hugepage_add_new_anon_rmap(page, vma, haddr);
4463
	} else
4464
		page_dup_rmap(page, true);
4465 4466
	new_pte = make_huge_pte(vma, page, ((vma->vm_flags & VM_WRITE)
				&& (vma->vm_flags & VM_SHARED)));
4467
	set_huge_pte_at(mm, haddr, ptep, new_pte);
4468

4469
	hugetlb_count_add(pages_per_huge_page(h), mm);
4470
	if ((flags & FAULT_FLAG_WRITE) && !(vma->vm_flags & VM_SHARED)) {
4471
		/* Optimization, do the COW without a second fault */
4472
		ret = hugetlb_cow(mm, vma, address, ptep, page, ptl);
4473 4474
	}

4475
	spin_unlock(ptl);
4476 4477

	/*
4478 4479 4480
	 * Only set HPageMigratable in newly allocated pages.  Existing pages
	 * found in the pagecache may not have HPageMigratableset if they have
	 * been isolated for migration.
4481 4482
	 */
	if (new_page)
4483
		SetHPageMigratable(page);
4484

A
Adam Litke 已提交
4485 4486
	unlock_page(page);
out:
4487
	return ret;
A
Adam Litke 已提交
4488 4489

backout:
4490
	spin_unlock(ptl);
4491
backout_unlocked:
A
Adam Litke 已提交
4492
	unlock_page(page);
4493
	restore_reserve_on_error(h, vma, haddr, page);
A
Adam Litke 已提交
4494 4495
	put_page(page);
	goto out;
4496 4497
}

4498
#ifdef CONFIG_SMP
4499
u32 hugetlb_fault_mutex_hash(struct address_space *mapping, pgoff_t idx)
4500 4501 4502 4503
{
	unsigned long key[2];
	u32 hash;

4504 4505
	key[0] = (unsigned long) mapping;
	key[1] = idx;
4506

4507
	hash = jhash2((u32 *)&key, sizeof(key)/(sizeof(u32)), 0);
4508 4509 4510 4511 4512

	return hash & (num_fault_mutexes - 1);
}
#else
/*
M
Miaohe Lin 已提交
4513
 * For uniprocessor systems we always use a single mutex, so just
4514 4515
 * return 0 and avoid the hashing overhead.
 */
4516
u32 hugetlb_fault_mutex_hash(struct address_space *mapping, pgoff_t idx)
4517 4518 4519 4520 4521
{
	return 0;
}
#endif

4522
vm_fault_t hugetlb_fault(struct mm_struct *mm, struct vm_area_struct *vma,
4523
			unsigned long address, unsigned int flags)
4524
{
4525
	pte_t *ptep, entry;
4526
	spinlock_t *ptl;
4527
	vm_fault_t ret;
4528 4529
	u32 hash;
	pgoff_t idx;
4530
	struct page *page = NULL;
4531
	struct page *pagecache_page = NULL;
4532
	struct hstate *h = hstate_vma(vma);
4533
	struct address_space *mapping;
4534
	int need_wait_lock = 0;
4535
	unsigned long haddr = address & huge_page_mask(h);
4536

4537
	ptep = huge_pte_offset(mm, haddr, huge_page_size(h));
4538
	if (ptep) {
4539 4540 4541 4542 4543
		/*
		 * 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.
		 */
4544
		entry = huge_ptep_get(ptep);
N
Naoya Horiguchi 已提交
4545
		if (unlikely(is_hugetlb_entry_migration(entry))) {
4546
			migration_entry_wait_huge(vma, mm, ptep);
N
Naoya Horiguchi 已提交
4547 4548
			return 0;
		} else if (unlikely(is_hugetlb_entry_hwpoisoned(entry)))
4549
			return VM_FAULT_HWPOISON_LARGE |
4550
				VM_FAULT_SET_HINDEX(hstate_index(h));
4551 4552
	}

4553 4554
	/*
	 * Acquire i_mmap_rwsem before calling huge_pte_alloc and hold
4555 4556 4557 4558
	 * 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.
4559 4560 4561 4562 4563
	 *
	 * 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.
	 */
4564
	mapping = vma->vm_file->f_mapping;
4565
	i_mmap_lock_read(mapping);
4566
	ptep = huge_pte_alloc(mm, vma, haddr, huge_page_size(h));
4567 4568 4569 4570
	if (!ptep) {
		i_mmap_unlock_read(mapping);
		return VM_FAULT_OOM;
	}
4571

4572 4573 4574 4575 4576
	/*
	 * 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.
	 */
4577
	idx = vma_hugecache_offset(h, vma, haddr);
4578
	hash = hugetlb_fault_mutex_hash(mapping, idx);
4579
	mutex_lock(&hugetlb_fault_mutex_table[hash]);
4580

4581 4582
	entry = huge_ptep_get(ptep);
	if (huge_pte_none(entry)) {
4583
		ret = hugetlb_no_page(mm, vma, mapping, idx, address, ptep, flags);
4584
		goto out_mutex;
4585
	}
4586

N
Nick Piggin 已提交
4587
	ret = 0;
4588

4589 4590 4591
	/*
	 * 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
4592 4593 4594
	 * an active hugepage in pagecache. This goto expects the 2nd page
	 * fault, and is_hugetlb_entry_(migration|hwpoisoned) check will
	 * properly handle it.
4595 4596 4597 4598
	 */
	if (!pte_present(entry))
		goto out_mutex;

4599 4600 4601 4602 4603 4604 4605 4606
	/*
	 * 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.
	 */
4607
	if ((flags & FAULT_FLAG_WRITE) && !huge_pte_write(entry)) {
4608
		if (vma_needs_reservation(h, vma, haddr) < 0) {
4609
			ret = VM_FAULT_OOM;
4610
			goto out_mutex;
4611
		}
4612
		/* Just decrements count, does not deallocate */
4613
		vma_end_reservation(h, vma, haddr);
4614

4615
		if (!(vma->vm_flags & VM_MAYSHARE))
4616
			pagecache_page = hugetlbfs_pagecache_page(h,
4617
								vma, haddr);
4618 4619
	}

4620 4621 4622 4623 4624 4625
	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;

4626 4627 4628 4629 4630 4631 4632
	/*
	 * 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)
4633 4634 4635 4636
		if (!trylock_page(page)) {
			need_wait_lock = 1;
			goto out_ptl;
		}
4637

4638
	get_page(page);
4639

4640
	if (flags & FAULT_FLAG_WRITE) {
4641
		if (!huge_pte_write(entry)) {
4642
			ret = hugetlb_cow(mm, vma, address, ptep,
4643
					  pagecache_page, ptl);
4644
			goto out_put_page;
4645
		}
4646
		entry = huge_pte_mkdirty(entry);
4647 4648
	}
	entry = pte_mkyoung(entry);
4649
	if (huge_ptep_set_access_flags(vma, haddr, ptep, entry,
4650
						flags & FAULT_FLAG_WRITE))
4651
		update_mmu_cache(vma, haddr, ptep);
4652 4653 4654 4655
out_put_page:
	if (page != pagecache_page)
		unlock_page(page);
	put_page(page);
4656 4657
out_ptl:
	spin_unlock(ptl);
4658 4659 4660 4661 4662

	if (pagecache_page) {
		unlock_page(pagecache_page);
		put_page(pagecache_page);
	}
4663
out_mutex:
4664
	mutex_unlock(&hugetlb_fault_mutex_table[hash]);
4665
	i_mmap_unlock_read(mapping);
4666 4667 4668 4669 4670 4671 4672 4673 4674
	/*
	 * 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);
4675
	return ret;
4676 4677
}

4678 4679 4680 4681 4682 4683 4684 4685 4686 4687 4688
/*
 * 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)
{
4689 4690 4691
	struct address_space *mapping;
	pgoff_t idx;
	unsigned long size;
4692
	int vm_shared = dst_vma->vm_flags & VM_SHARED;
4693 4694 4695 4696 4697 4698 4699 4700 4701 4702 4703 4704 4705 4706
	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,
4707
						pages_per_huge_page(h), false);
4708

4709
		/* fallback to copy_from_user outside mmap_lock */
4710
		if (unlikely(ret)) {
4711
			ret = -ENOENT;
4712 4713 4714 4715 4716 4717 4718 4719 4720 4721 4722 4723 4724 4725 4726 4727
			*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);

4728 4729 4730
	mapping = dst_vma->vm_file->f_mapping;
	idx = vma_hugecache_offset(h, dst_vma, dst_addr);

4731 4732 4733 4734
	/*
	 * If shared, add to page cache
	 */
	if (vm_shared) {
4735 4736 4737 4738
		size = i_size_read(mapping->host) >> huge_page_shift(h);
		ret = -EFAULT;
		if (idx >= size)
			goto out_release_nounlock;
4739

4740 4741 4742 4743 4744 4745
		/*
		 * 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.
		 */
4746 4747 4748 4749 4750
		ret = huge_add_to_page_cache(page, mapping, idx);
		if (ret)
			goto out_release_nounlock;
	}

4751 4752 4753
	ptl = huge_pte_lockptr(h, dst_mm, dst_pte);
	spin_lock(ptl);

4754 4755 4756 4757 4758 4759 4760 4761 4762 4763 4764 4765 4766 4767
	/*
	 * 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;

4768 4769 4770 4771
	ret = -EEXIST;
	if (!huge_pte_none(huge_ptep_get(dst_pte)))
		goto out_release_unlock;

4772 4773 4774
	if (vm_shared) {
		page_dup_rmap(page, true);
	} else {
4775
		ClearHPageRestoreReserve(page);
4776 4777
		hugepage_add_new_anon_rmap(page, dst_vma, dst_addr);
	}
4778 4779 4780 4781 4782 4783 4784 4785 4786 4787 4788 4789 4790 4791 4792 4793

	_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);
4794
	SetHPageMigratable(page);
4795 4796
	if (vm_shared)
		unlock_page(page);
4797 4798 4799 4800 4801
	ret = 0;
out:
	return ret;
out_release_unlock:
	spin_unlock(ptl);
4802 4803
	if (vm_shared)
		unlock_page(page);
4804
out_release_nounlock:
4805 4806 4807 4808
	put_page(page);
	goto out;
}

4809 4810 4811 4812 4813 4814 4815 4816 4817 4818 4819 4820 4821 4822
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;
	}
}

4823 4824 4825
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,
4826
			 long i, unsigned int flags, int *locked)
D
David Gibson 已提交
4827
{
4828 4829
	unsigned long pfn_offset;
	unsigned long vaddr = *position;
4830
	unsigned long remainder = *nr_pages;
4831
	struct hstate *h = hstate_vma(vma);
4832
	int err = -EFAULT, refs;
D
David Gibson 已提交
4833 4834

	while (vaddr < vma->vm_end && remainder) {
A
Adam Litke 已提交
4835
		pte_t *pte;
4836
		spinlock_t *ptl = NULL;
H
Hugh Dickins 已提交
4837
		int absent;
A
Adam Litke 已提交
4838
		struct page *page;
D
David Gibson 已提交
4839

4840 4841 4842 4843
		/*
		 * If we have a pending SIGKILL, don't keep faulting pages and
		 * potentially allocating memory.
		 */
4844
		if (fatal_signal_pending(current)) {
4845 4846 4847 4848
			remainder = 0;
			break;
		}

A
Adam Litke 已提交
4849 4850
		/*
		 * Some archs (sparc64, sh*) have multiple pte_ts to
H
Hugh Dickins 已提交
4851
		 * each hugepage.  We have to make sure we get the
A
Adam Litke 已提交
4852
		 * first, for the page indexing below to work.
4853 4854
		 *
		 * Note that page table lock is not held when pte is null.
A
Adam Litke 已提交
4855
		 */
4856 4857
		pte = huge_pte_offset(mm, vaddr & huge_page_mask(h),
				      huge_page_size(h));
4858 4859
		if (pte)
			ptl = huge_pte_lock(h, mm, pte);
H
Hugh Dickins 已提交
4860 4861 4862 4863
		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
4864 4865 4866 4867
		 * 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
4868
		 */
H
Hugh Dickins 已提交
4869 4870
		if (absent && (flags & FOLL_DUMP) &&
		    !hugetlbfs_pagecache_present(h, vma, vaddr)) {
4871 4872
			if (pte)
				spin_unlock(ptl);
H
Hugh Dickins 已提交
4873 4874 4875
			remainder = 0;
			break;
		}
D
David Gibson 已提交
4876

4877 4878 4879 4880 4881 4882 4883 4884 4885 4886 4887
		/*
		 * 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)) ||
4888 4889
		    ((flags & FOLL_WRITE) &&
		      !huge_pte_write(huge_ptep_get(pte)))) {
4890
			vm_fault_t ret;
4891
			unsigned int fault_flags = 0;
D
David Gibson 已提交
4892

4893 4894
			if (pte)
				spin_unlock(ptl);
4895 4896
			if (flags & FOLL_WRITE)
				fault_flags |= FAULT_FLAG_WRITE;
4897
			if (locked)
4898 4899
				fault_flags |= FAULT_FLAG_ALLOW_RETRY |
					FAULT_FLAG_KILLABLE;
4900 4901 4902 4903
			if (flags & FOLL_NOWAIT)
				fault_flags |= FAULT_FLAG_ALLOW_RETRY |
					FAULT_FLAG_RETRY_NOWAIT;
			if (flags & FOLL_TRIED) {
4904 4905 4906 4907
				/*
				 * Note: FAULT_FLAG_ALLOW_RETRY and
				 * FAULT_FLAG_TRIED can co-exist
				 */
4908 4909 4910 4911
				fault_flags |= FAULT_FLAG_TRIED;
			}
			ret = hugetlb_fault(mm, vma, vaddr, fault_flags);
			if (ret & VM_FAULT_ERROR) {
4912
				err = vm_fault_to_errno(ret, flags);
4913 4914 4915 4916
				remainder = 0;
				break;
			}
			if (ret & VM_FAULT_RETRY) {
4917
				if (locked &&
4918
				    !(fault_flags & FAULT_FLAG_RETRY_NOWAIT))
4919
					*locked = 0;
4920 4921 4922 4923 4924 4925 4926 4927 4928 4929 4930 4931 4932
				*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
4933 4934
		}

4935
		pfn_offset = (vaddr & ~huge_page_mask(h)) >> PAGE_SHIFT;
4936
		page = pte_page(huge_ptep_get(pte));
4937

4938 4939 4940 4941 4942 4943 4944 4945 4946 4947 4948 4949 4950 4951
		/*
		 * 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;
		}

4952 4953
		refs = min3(pages_per_huge_page(h) - pfn_offset,
			    (vma->vm_end - vaddr) >> PAGE_SHIFT, remainder);
4954

4955 4956 4957 4958 4959
		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
4960

4961
		if (pages) {
4962 4963 4964 4965 4966 4967 4968 4969 4970 4971
			/*
			 * 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:
			 */
4972
			if (WARN_ON_ONCE(!try_grab_compound_head(pages[i],
4973 4974 4975 4976 4977 4978 4979
								 refs,
								 flags))) {
				spin_unlock(ptl);
				remainder = 0;
				err = -ENOMEM;
				break;
			}
4980
		}
4981 4982 4983 4984 4985

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

4986
		spin_unlock(ptl);
D
David Gibson 已提交
4987
	}
4988
	*nr_pages = remainder;
4989 4990 4991 4992 4993
	/*
	 * 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
4994 4995
	*position = vaddr;

4996
	return i ? i : err;
D
David Gibson 已提交
4997
}
4998

4999 5000 5001 5002 5003 5004 5005 5006
#ifndef __HAVE_ARCH_FLUSH_HUGETLB_TLB_RANGE
/*
 * ARCHes with special requirements for evicting HUGETLB backing TLB entries can
 * implement this.
 */
#define flush_hugetlb_tlb_range(vma, addr, end)	flush_tlb_range(vma, addr, end)
#endif

5007
unsigned long hugetlb_change_protection(struct vm_area_struct *vma,
5008 5009 5010 5011 5012 5013
		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;
5014
	struct hstate *h = hstate_vma(vma);
5015
	unsigned long pages = 0;
5016
	bool shared_pmd = false;
5017
	struct mmu_notifier_range range;
5018 5019 5020

	/*
	 * In the case of shared PMDs, the area to flush could be beyond
5021
	 * start/end.  Set range.start/range.end to cover the maximum possible
5022 5023
	 * range if PMD sharing is possible.
	 */
5024 5025
	mmu_notifier_range_init(&range, MMU_NOTIFY_PROTECTION_VMA,
				0, vma, mm, start, end);
5026
	adjust_range_if_pmd_sharing_possible(vma, &range.start, &range.end);
5027 5028

	BUG_ON(address >= end);
5029
	flush_cache_range(vma, range.start, range.end);
5030

5031
	mmu_notifier_invalidate_range_start(&range);
5032
	i_mmap_lock_write(vma->vm_file->f_mapping);
5033
	for (; address < end; address += huge_page_size(h)) {
5034
		spinlock_t *ptl;
5035
		ptep = huge_pte_offset(mm, address, huge_page_size(h));
5036 5037
		if (!ptep)
			continue;
5038
		ptl = huge_pte_lock(h, mm, ptep);
5039
		if (huge_pmd_unshare(mm, vma, &address, ptep)) {
5040
			pages++;
5041
			spin_unlock(ptl);
5042
			shared_pmd = true;
5043
			continue;
5044
		}
5045 5046 5047 5048 5049 5050 5051 5052 5053 5054 5055 5056 5057
		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);
5058 5059
				set_huge_swap_pte_at(mm, address, ptep,
						     newpte, huge_page_size(h));
5060 5061 5062 5063 5064 5065
				pages++;
			}
			spin_unlock(ptl);
			continue;
		}
		if (!huge_pte_none(pte)) {
5066 5067 5068 5069
			pte_t old_pte;

			old_pte = huge_ptep_modify_prot_start(vma, address, ptep);
			pte = pte_mkhuge(huge_pte_modify(old_pte, newprot));
5070
			pte = arch_make_huge_pte(pte, vma, NULL, 0);
5071
			huge_ptep_modify_prot_commit(vma, address, ptep, old_pte, pte);
5072
			pages++;
5073
		}
5074
		spin_unlock(ptl);
5075
	}
5076
	/*
5077
	 * Must flush TLB before releasing i_mmap_rwsem: x86's huge_pmd_unshare
5078
	 * may have cleared our pud entry and done put_page on the page table:
5079
	 * once we release i_mmap_rwsem, another task can do the final put_page
5080 5081
	 * 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.
5082
	 */
5083
	if (shared_pmd)
5084
		flush_hugetlb_tlb_range(vma, range.start, range.end);
5085 5086
	else
		flush_hugetlb_tlb_range(vma, start, end);
5087 5088 5089 5090
	/*
	 * No need to call mmu_notifier_invalidate_range() we are downgrading
	 * page table protection not changing it to point to a new page.
	 *
5091
	 * See Documentation/vm/mmu_notifier.rst
5092
	 */
5093
	i_mmap_unlock_write(vma->vm_file->f_mapping);
5094
	mmu_notifier_invalidate_range_end(&range);
5095 5096

	return pages << h->order;
5097 5098
}

5099 5100
/* Return true if reservation was successful, false otherwise.  */
bool hugetlb_reserve_pages(struct inode *inode,
5101
					long from, long to,
5102
					struct vm_area_struct *vma,
5103
					vm_flags_t vm_flags)
5104
{
5105
	long chg, add = -1;
5106
	struct hstate *h = hstate_inode(inode);
5107
	struct hugepage_subpool *spool = subpool_inode(inode);
5108
	struct resv_map *resv_map;
5109
	struct hugetlb_cgroup *h_cg = NULL;
5110
	long gbl_reserve, regions_needed = 0;
5111

5112 5113 5114
	/* This should never happen */
	if (from > to) {
		VM_WARN(1, "%s called with a negative range\n", __func__);
5115
		return false;
5116 5117
	}

5118 5119 5120
	/*
	 * Only apply hugepage reservation if asked. At fault time, an
	 * attempt will be made for VM_NORESERVE to allocate a page
5121
	 * without using reserves
5122
	 */
5123
	if (vm_flags & VM_NORESERVE)
5124
		return true;
5125

5126 5127 5128 5129 5130 5131
	/*
	 * 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
	 */
5132
	if (!vma || vma->vm_flags & VM_MAYSHARE) {
5133 5134 5135 5136 5137
		/*
		 * 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).
		 */
5138
		resv_map = inode_resv_map(inode);
5139

5140
		chg = region_chg(resv_map, from, to, &regions_needed);
5141 5142

	} else {
5143
		/* Private mapping. */
5144
		resv_map = resv_map_alloc();
5145
		if (!resv_map)
5146
			return false;
5147

5148
		chg = to - from;
5149

5150 5151 5152 5153
		set_vma_resv_map(vma, resv_map);
		set_vma_resv_flags(vma, HPAGE_RESV_OWNER);
	}

5154
	if (chg < 0)
5155
		goto out_err;
5156

5157 5158
	if (hugetlb_cgroup_charge_cgroup_rsvd(hstate_index(h),
				chg * pages_per_huge_page(h), &h_cg) < 0)
5159 5160 5161 5162 5163 5164 5165 5166 5167
		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);
	}

5168 5169 5170 5171 5172 5173
	/*
	 * 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);
5174
	if (gbl_reserve < 0)
5175
		goto out_uncharge_cgroup;
5176 5177

	/*
5178
	 * Check enough hugepages are available for the reservation.
5179
	 * Hand the pages back to the subpool if there are not
5180
	 */
5181
	if (hugetlb_acct_memory(h, gbl_reserve) < 0)
5182
		goto out_put_pages;
5183 5184 5185 5186 5187 5188 5189 5190 5191 5192 5193 5194

	/*
	 * 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
	 */
5195
	if (!vma || vma->vm_flags & VM_MAYSHARE) {
5196
		add = region_add(resv_map, from, to, regions_needed, h, h_cg);
5197 5198 5199

		if (unlikely(add < 0)) {
			hugetlb_acct_memory(h, -gbl_reserve);
5200
			goto out_put_pages;
5201
		} else if (unlikely(chg > add)) {
5202 5203 5204 5205 5206 5207 5208 5209 5210
			/*
			 * 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;

5211 5212 5213 5214
			/*
			 * hugetlb_cgroup_uncharge_cgroup_rsvd() will put the
			 * reference to h_cg->css. See comment below for detail.
			 */
5215 5216 5217 5218
			hugetlb_cgroup_uncharge_cgroup_rsvd(
				hstate_index(h),
				(chg - add) * pages_per_huge_page(h), h_cg);

5219 5220 5221
			rsv_adjust = hugepage_subpool_put_pages(spool,
								chg - add);
			hugetlb_acct_memory(h, -rsv_adjust);
5222 5223 5224 5225 5226 5227 5228 5229
		} 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);
5230 5231
		}
	}
5232 5233
	return true;

5234 5235 5236 5237 5238 5239
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);
5240
out_err:
5241
	if (!vma || vma->vm_flags & VM_MAYSHARE)
5242 5243 5244 5245 5246
		/* 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
5247 5248
	if (vma && is_vma_resv_set(vma, HPAGE_RESV_OWNER))
		kref_put(&resv_map->refs, resv_map_release);
5249
	return false;
5250 5251
}

5252 5253
long hugetlb_unreserve_pages(struct inode *inode, long start, long end,
								long freed)
5254
{
5255
	struct hstate *h = hstate_inode(inode);
5256
	struct resv_map *resv_map = inode_resv_map(inode);
5257
	long chg = 0;
5258
	struct hugepage_subpool *spool = subpool_inode(inode);
5259
	long gbl_reserve;
K
Ken Chen 已提交
5260

5261 5262 5263 5264
	/*
	 * Since this routine can be called in the evict inode path for all
	 * hugetlbfs inodes, resv_map could be NULL.
	 */
5265 5266 5267 5268 5269 5270 5271 5272 5273 5274 5275
	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
5276
	spin_lock(&inode->i_lock);
5277
	inode->i_blocks -= (blocks_per_huge_page(h) * freed);
K
Ken Chen 已提交
5278 5279
	spin_unlock(&inode->i_lock);

5280 5281 5282 5283 5284 5285
	/*
	 * If the subpool has a minimum size, the number of global
	 * reservations to be released may be adjusted.
	 */
	gbl_reserve = hugepage_subpool_put_pages(spool, (chg - freed));
	hugetlb_acct_memory(h, -gbl_reserve);
5286 5287

	return 0;
5288
}
5289

5290 5291 5292 5293 5294 5295 5296 5297 5298 5299 5300
#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
5301 5302
	unsigned long vm_flags = vma->vm_flags & VM_LOCKED_CLEAR_MASK;
	unsigned long svm_flags = svma->vm_flags & VM_LOCKED_CLEAR_MASK;
5303 5304 5305 5306 5307 5308 5309

	/*
	 * match the virtual addresses, permission and the alignment of the
	 * page table page.
	 */
	if (pmd_index(addr) != pmd_index(saddr) ||
	    vm_flags != svm_flags ||
5310
	    !range_in_vma(svma, sbase, s_end))
5311 5312 5313 5314 5315
		return 0;

	return saddr;
}

5316
static bool vma_shareable(struct vm_area_struct *vma, unsigned long addr)
5317 5318 5319 5320 5321 5322 5323
{
	unsigned long base = addr & PUD_MASK;
	unsigned long end = base + PUD_SIZE;

	/*
	 * check on proper vm_flags and page table alignment
	 */
5324
	if (vma->vm_flags & VM_MAYSHARE && range_in_vma(vma, base, end))
5325 5326
		return true;
	return false;
5327 5328
}

5329 5330 5331 5332 5333 5334 5335 5336
/*
 * 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)
{
5337 5338
	unsigned long v_start = ALIGN(vma->vm_start, PUD_SIZE),
		v_end = ALIGN_DOWN(vma->vm_end, PUD_SIZE);
5339

5340 5341 5342 5343 5344 5345
	/*
	 * 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))
5346 5347
		return;

5348
	/* Extend the range to be PUD aligned for a worst case scenario */
5349 5350
	if (*start > v_start)
		*start = ALIGN_DOWN(*start, PUD_SIZE);
5351

5352 5353
	if (*end < v_end)
		*end = ALIGN(*end, PUD_SIZE);
5354 5355
}

5356 5357 5358 5359
/*
 * 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
5360 5361
 * code much cleaner.
 *
5362 5363 5364 5365 5366 5367 5368 5369 5370 5371
 * 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.
5372
 */
5373 5374
pte_t *huge_pmd_share(struct mm_struct *mm, struct vm_area_struct *vma,
		      unsigned long addr, pud_t *pud)
5375 5376 5377 5378 5379 5380 5381 5382
{
	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;
5383
	spinlock_t *ptl;
5384 5385 5386 5387

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

5388
	i_mmap_assert_locked(mapping);
5389 5390 5391 5392 5393 5394
	vma_interval_tree_foreach(svma, &mapping->i_mmap, idx, idx) {
		if (svma == vma)
			continue;

		saddr = page_table_shareable(svma, vma, addr, idx);
		if (saddr) {
5395 5396
			spte = huge_pte_offset(svma->vm_mm, saddr,
					       vma_mmu_pagesize(svma));
5397 5398 5399 5400 5401 5402 5403 5404 5405 5406
			if (spte) {
				get_page(virt_to_page(spte));
				break;
			}
		}
	}

	if (!spte)
		goto out;

5407
	ptl = huge_pte_lock(hstate_vma(vma), mm, spte);
5408
	if (pud_none(*pud)) {
5409 5410
		pud_populate(mm, pud,
				(pmd_t *)((unsigned long)spte & PAGE_MASK));
5411
		mm_inc_nr_pmds(mm);
5412
	} else {
5413
		put_page(virt_to_page(spte));
5414
	}
5415
	spin_unlock(ptl);
5416 5417 5418 5419 5420 5421 5422 5423 5424 5425 5426 5427
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.
 *
5428
 * Called with page table lock held and i_mmap_rwsem held in write mode.
5429 5430 5431 5432
 *
 * returns: 1 successfully unmapped a shared pte page
 *	    0 the underlying pte page is not shared, or it is the last user
 */
5433 5434
int huge_pmd_unshare(struct mm_struct *mm, struct vm_area_struct *vma,
					unsigned long *addr, pte_t *ptep)
5435 5436
{
	pgd_t *pgd = pgd_offset(mm, *addr);
5437 5438
	p4d_t *p4d = p4d_offset(pgd, *addr);
	pud_t *pud = pud_offset(p4d, *addr);
5439

5440
	i_mmap_assert_write_locked(vma->vm_file->f_mapping);
5441 5442 5443 5444 5445 5446
	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));
5447
	mm_dec_nr_pmds(mm);
5448 5449 5450
	*addr = ALIGN(*addr, HPAGE_SIZE * PTRS_PER_PTE) - HPAGE_SIZE;
	return 1;
}
5451 5452
#define want_pmd_share()	(1)
#else /* !CONFIG_ARCH_WANT_HUGE_PMD_SHARE */
5453 5454
pte_t *huge_pmd_share(struct mm_struct *mm, struct vm_area_struct *vma,
		      unsigned long addr, pud_t *pud)
5455 5456 5457
{
	return NULL;
}
5458

5459 5460
int huge_pmd_unshare(struct mm_struct *mm, struct vm_area_struct *vma,
				unsigned long *addr, pte_t *ptep)
5461 5462 5463
{
	return 0;
}
5464 5465 5466 5467 5468

void adjust_range_if_pmd_sharing_possible(struct vm_area_struct *vma,
				unsigned long *start, unsigned long *end)
{
}
5469
#define want_pmd_share()	(0)
5470 5471
#endif /* CONFIG_ARCH_WANT_HUGE_PMD_SHARE */

5472
#ifdef CONFIG_ARCH_WANT_GENERAL_HUGETLB
5473
pte_t *huge_pte_alloc(struct mm_struct *mm, struct vm_area_struct *vma,
5474 5475 5476
			unsigned long addr, unsigned long sz)
{
	pgd_t *pgd;
5477
	p4d_t *p4d;
5478 5479 5480 5481
	pud_t *pud;
	pte_t *pte = NULL;

	pgd = pgd_offset(mm, addr);
5482 5483 5484
	p4d = p4d_alloc(mm, pgd, addr);
	if (!p4d)
		return NULL;
5485
	pud = pud_alloc(mm, p4d, addr);
5486 5487 5488 5489 5490 5491
	if (pud) {
		if (sz == PUD_SIZE) {
			pte = (pte_t *)pud;
		} else {
			BUG_ON(sz != PMD_SIZE);
			if (want_pmd_share() && pud_none(*pud))
5492
				pte = huge_pmd_share(mm, vma, addr, pud);
5493 5494 5495 5496
			else
				pte = (pte_t *)pmd_alloc(mm, pud, addr);
		}
	}
5497
	BUG_ON(pte && pte_present(*pte) && !pte_huge(*pte));
5498 5499 5500 5501

	return pte;
}

5502 5503 5504 5505
/*
 * huge_pte_offset() - Walk the page table to resolve the hugepage
 * entry at address @addr
 *
5506 5507
 * Return: Pointer to page table entry (PUD or PMD) for
 * address @addr, or NULL if a !p*d_present() entry is encountered and the
5508 5509 5510
 * size @sz doesn't match the hugepage size at this level of the page
 * table.
 */
5511 5512
pte_t *huge_pte_offset(struct mm_struct *mm,
		       unsigned long addr, unsigned long sz)
5513 5514
{
	pgd_t *pgd;
5515
	p4d_t *p4d;
5516 5517
	pud_t *pud;
	pmd_t *pmd;
5518 5519

	pgd = pgd_offset(mm, addr);
5520 5521 5522 5523 5524
	if (!pgd_present(*pgd))
		return NULL;
	p4d = p4d_offset(pgd, addr);
	if (!p4d_present(*p4d))
		return NULL;
5525

5526
	pud = pud_offset(p4d, addr);
5527 5528
	if (sz == PUD_SIZE)
		/* must be pud huge, non-present or none */
5529
		return (pte_t *)pud;
5530
	if (!pud_present(*pud))
5531
		return NULL;
5532
	/* must have a valid entry and size to go further */
5533

5534 5535 5536
	pmd = pmd_offset(pud, addr);
	/* must be pmd huge, non-present or none */
	return (pte_t *)pmd;
5537 5538
}

5539 5540 5541 5542 5543 5544 5545 5546 5547 5548 5549 5550 5551
#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);
}

5552 5553 5554 5555 5556 5557 5558 5559
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;
}

5560
struct page * __weak
5561
follow_huge_pmd(struct mm_struct *mm, unsigned long address,
5562
		pmd_t *pmd, int flags)
5563
{
5564 5565
	struct page *page = NULL;
	spinlock_t *ptl;
5566
	pte_t pte;
J
John Hubbard 已提交
5567 5568 5569 5570 5571 5572

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

5573 5574 5575 5576 5577 5578 5579 5580 5581
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;
5582 5583
	pte = huge_ptep_get((pte_t *)pmd);
	if (pte_present(pte)) {
5584
		page = pmd_page(*pmd) + ((address & ~PMD_MASK) >> PAGE_SHIFT);
J
John Hubbard 已提交
5585 5586 5587 5588 5589 5590 5591 5592 5593 5594 5595 5596
		/*
		 * 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;
		}
5597
	} else {
5598
		if (is_hugetlb_entry_migration(pte)) {
5599 5600 5601 5602 5603 5604 5605 5606 5607 5608 5609
			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);
5610 5611 5612
	return page;
}

5613
struct page * __weak
5614
follow_huge_pud(struct mm_struct *mm, unsigned long address,
5615
		pud_t *pud, int flags)
5616
{
J
John Hubbard 已提交
5617
	if (flags & (FOLL_GET | FOLL_PIN))
5618
		return NULL;
5619

5620
	return pte_page(*(pte_t *)pud) + ((address & ~PUD_MASK) >> PAGE_SHIFT);
5621 5622
}

5623 5624 5625
struct page * __weak
follow_huge_pgd(struct mm_struct *mm, unsigned long address, pgd_t *pgd, int flags)
{
J
John Hubbard 已提交
5626
	if (flags & (FOLL_GET | FOLL_PIN))
5627 5628 5629 5630 5631
		return NULL;

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

5632 5633
bool isolate_huge_page(struct page *page, struct list_head *list)
{
5634 5635
	bool ret = true;

5636
	spin_lock(&hugetlb_lock);
5637 5638
	if (!PageHeadHuge(page) ||
	    !HPageMigratable(page) ||
5639
	    !get_page_unless_zero(page)) {
5640 5641 5642
		ret = false;
		goto unlock;
	}
5643
	ClearHPageMigratable(page);
5644
	list_move_tail(&page->lru, list);
5645
unlock:
5646
	spin_unlock(&hugetlb_lock);
5647
	return ret;
5648 5649 5650 5651 5652
}

void putback_active_hugepage(struct page *page)
{
	spin_lock(&hugetlb_lock);
5653
	SetHPageMigratable(page);
5654 5655 5656 5657
	list_move_tail(&page->lru, &(page_hstate(page))->hugepage_activelist);
	spin_unlock(&hugetlb_lock);
	put_page(page);
}
5658 5659 5660 5661 5662 5663 5664 5665 5666 5667 5668 5669 5670 5671 5672 5673 5674 5675

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.
	 */
5676
	if (HPageTemporary(newpage)) {
5677 5678 5679
		int old_nid = page_to_nid(oldpage);
		int new_nid = page_to_nid(newpage);

5680 5681
		SetHPageTemporary(oldpage);
		ClearHPageTemporary(newpage);
5682 5683 5684 5685 5686 5687 5688 5689 5690

		spin_lock(&hugetlb_lock);
		if (h->surplus_huge_pages_node[old_nid]) {
			h->surplus_huge_pages_node[old_nid]--;
			h->surplus_huge_pages_node[new_nid]++;
		}
		spin_unlock(&hugetlb_lock);
	}
}
5691 5692 5693 5694 5695 5696 5697 5698 5699 5700 5701 5702 5703 5704 5705 5706 5707 5708 5709 5710 5711 5712 5713 5714 5715 5716 5717 5718 5719 5720 5721 5722 5723 5724 5725 5726 5727 5728 5729

#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;
5730
		char name[CMA_MAX_NAME];
5731 5732 5733 5734

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

5735
		snprintf(name, sizeof(name), "hugetlb%d", nid);
5736
		res = cma_declare_contiguous_nid(0, size, 0, PAGE_SIZE << order,
5737
						 0, false, name,
5738 5739 5740 5741 5742 5743 5744 5745 5746 5747 5748 5749 5750 5751 5752 5753 5754 5755 5756 5757 5758 5759 5760 5761 5762
						 &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 */