hugetlb.c 192.5 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
#include <linux/migrate.h>
34

D
David Gibson 已提交
35
#include <asm/page.h>
36
#include <asm/pgalloc.h>
37
#include <asm/tlb.h>
D
David Gibson 已提交
38

39
#include <linux/io.h>
D
David Gibson 已提交
40
#include <linux/hugetlb.h>
41
#include <linux/hugetlb_cgroup.h>
42
#include <linux/node.h>
43
#include <linux/page_owner.h>
44
#include "internal.h"
45
#include "hugetlb_vmemmap.h"
L
Linus Torvalds 已提交
46

47
int hugetlb_max_hstate __read_mostly;
48 49
unsigned int default_hstate_idx;
struct hstate hstates[HUGE_MAX_HSTATE];
50

51
#ifdef CONFIG_CMA
52
static struct cma *hugetlb_cma[MAX_NUMNODES];
53
static unsigned long hugetlb_cma_size_in_node[MAX_NUMNODES] __initdata;
54 55 56 57 58 59 60 61 62 63
static bool hugetlb_cma_page(struct page *page, unsigned int order)
{
	return cma_pages_valid(hugetlb_cma[page_to_nid(page)], page,
				1 << order);
}
#else
static bool hugetlb_cma_page(struct page *page, unsigned int order)
{
	return false;
}
64 65
#endif
static unsigned long hugetlb_cma_size __initdata;
66

67 68 69 70 71
/*
 * Minimum page order among possible hugepage sizes, set to a proper value
 * at boot time.
 */
static unsigned int minimum_order __read_mostly = UINT_MAX;
72

73 74
__initdata LIST_HEAD(huge_boot_pages);

75 76 77
/* for command line parsing */
static struct hstate * __initdata parsed_hstate;
static unsigned long __initdata default_hstate_max_huge_pages;
78
static bool __initdata parsed_valid_hugepagesz = true;
79
static bool __initdata parsed_default_hugepagesz;
80
static unsigned int default_hugepages_in_node[MAX_NUMNODES] __initdata;
81

82
/*
83 84
 * Protects updates to hugepage_freelists, hugepage_activelist, nr_huge_pages,
 * free_huge_pages, and surplus_huge_pages.
85
 */
86
DEFINE_SPINLOCK(hugetlb_lock);
87

88 89 90 91 92
/*
 * 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;
93
struct mutex *hugetlb_fault_mutex_table ____cacheline_aligned_in_smp;
94

95 96 97
/* Forward declaration */
static int hugetlb_acct_memory(struct hstate *h, long delta);

98
static inline bool subpool_is_free(struct hugepage_subpool *spool)
99
{
100 101 102 103 104 105 106 107 108
	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;
}
109

110 111
static inline void unlock_or_release_subpool(struct hugepage_subpool *spool,
						unsigned long irq_flags)
112
{
113
	spin_unlock_irqrestore(&spool->lock, irq_flags);
114 115

	/* If no pages are used, and no other handles to the subpool
E
Ethon Paul 已提交
116
	 * remain, give up any reservations based on minimum size and
117
	 * free the subpool */
118
	if (subpool_is_free(spool)) {
119 120 121
		if (spool->min_hpages != -1)
			hugetlb_acct_memory(spool->hstate,
						-spool->min_hpages);
122
		kfree(spool);
123
	}
124 125
}

126 127
struct hugepage_subpool *hugepage_new_subpool(struct hstate *h, long max_hpages,
						long min_hpages)
128 129 130
{
	struct hugepage_subpool *spool;

131
	spool = kzalloc(sizeof(*spool), GFP_KERNEL);
132 133 134 135 136
	if (!spool)
		return NULL;

	spin_lock_init(&spool->lock);
	spool->count = 1;
137 138 139 140 141 142 143 144 145
	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;
146 147 148 149 150 151

	return spool;
}

void hugepage_put_subpool(struct hugepage_subpool *spool)
{
152 153 154
	unsigned long flags;

	spin_lock_irqsave(&spool->lock, flags);
155 156
	BUG_ON(!spool->count);
	spool->count--;
157
	unlock_or_release_subpool(spool, flags);
158 159
}

160 161 162
/*
 * Subpool accounting for allocating and reserving pages.
 * Return -ENOMEM if there are not enough resources to satisfy the
163
 * request.  Otherwise, return the number of pages by which the
164 165
 * 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
166
 * a subpool minimum size must be maintained.
167 168
 */
static long hugepage_subpool_get_pages(struct hugepage_subpool *spool,
169 170
				      long delta)
{
171
	long ret = delta;
172 173

	if (!spool)
174
		return ret;
175

176
	spin_lock_irq(&spool->lock);
177 178 179 180 181 182 183 184

	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;
		}
185 186
	}

187 188
	/* minimum size accounting */
	if (spool->min_hpages != -1 && spool->rsv_hpages) {
189 190 191 192 193 194 195 196 197 198 199 200 201 202
		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:
203
	spin_unlock_irq(&spool->lock);
204 205 206
	return ret;
}

207 208 209 210 211 212 213
/*
 * 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,
214 215
				       long delta)
{
216
	long ret = delta;
217
	unsigned long flags;
218

219
	if (!spool)
220
		return delta;
221

222
	spin_lock_irqsave(&spool->lock, flags);
223 224 225 226

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

227 228
	 /* minimum size accounting */
	if (spool->min_hpages != -1 && spool->used_hpages < spool->min_hpages) {
229 230 231 232 233 234 235 236 237 238 239 240 241 242
		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.
	 */
243
	unlock_or_release_subpool(spool, flags);
244 245

	return ret;
246 247 248 249 250 251 252 253 254
}

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
255
	return subpool_inode(file_inode(vma->vm_file));
256 257
}

258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277
/* 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;
}

278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299
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;
300 301 302 303 304 305 306 307 308 309 310
		/*
		 * 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);
311 312 313 314 315 316 317 318 319 320 321 322 323
		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
}

324 325 326 327 328 329 330 331
static void put_uncharge_info(struct file_region *rg)
{
#ifdef CONFIG_CGROUP_HUGETLB
	if (rg->css)
		css_put(rg->css);
#endif
}

332 333 334 335
static bool has_same_uncharge_info(struct file_region *rg,
				   struct file_region *org)
{
#ifdef CONFIG_CGROUP_HUGETLB
336
	return rg->reservation_counter == org->reservation_counter &&
337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353
	       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);
354
		put_uncharge_info(rg);
355 356
		kfree(rg);

357
		rg = prg;
358 359 360 361 362 363 364 365
	}

	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);
366
		put_uncharge_info(rg);
367 368 369 370
		kfree(rg);
	}
}

371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388
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;
}

389 390 391 392 393 394 395
/*
 * 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
396 397
 */
static long add_reservation_in_range(struct resv_map *resv, long f, long t,
398
				     struct hugetlb_cgroup *h_cg,
399
				     struct hstate *h, long *regions_needed)
M
Mina Almasry 已提交
400
{
401
	long add = 0;
M
Mina Almasry 已提交
402
	struct list_head *head = &resv->regions;
403
	long last_accounted_offset = f;
404
	struct file_region *rg = NULL, *trg = NULL;
M
Mina Almasry 已提交
405

406 407
	if (regions_needed)
		*regions_needed = 0;
M
Mina Almasry 已提交
408

409 410 411 412 413 414 415 416 417 418 419 420 421 422
	/* 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
423

424 425 426
		/* When we find a region that starts beyond our range, we've
		 * finished.
		 */
427
		if (rg->from >= t)
M
Mina Almasry 已提交
428 429
			break;

430 431 432
		/* Add an entry for last_accounted_offset -> rg->from, and
		 * update last_accounted_offset.
		 */
433 434 435 436 437
		if (rg->from > last_accounted_offset)
			add += hugetlb_resv_map_add(resv, rg,
						    last_accounted_offset,
						    rg->from, h, h_cg,
						    regions_needed);
438 439 440 441 442 443 444

		last_accounted_offset = rg->to;
	}

	/* Handle the case where our range extends beyond
	 * last_accounted_offset.
	 */
445 446 447
	if (last_accounted_offset < t)
		add += hugetlb_resv_map_add(resv, rg, last_accounted_offset,
					    t, h, h_cg, regions_needed);
448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480

	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
I
Ingo Molnar 已提交
481
		 * for all the existing adds_in_progress. We should only be
482
		 * needing to allocate for regions_needed.
M
Mina Almasry 已提交
483
		 */
484 485 486 487 488 489 490 491
		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
492 493
		}

494 495
		spin_lock(&resv->lock);

496 497
		list_splice(&allocated_regions, &resv->region_cache);
		resv->region_cache_count += to_allocate;
M
Mina Almasry 已提交
498 499
	}

500
	return 0;
M
Mina Almasry 已提交
501

502 503 504 505 506 507
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
508 509
}

510 511
/*
 * Add the huge page range represented by [f, t) to the reserve
512 513 514 515 516
 * 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.
517
 *
518 519 520 521
 * 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
522
 * this operation and we were not able to allocate, it returns -ENOMEM.
523 524 525
 * 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.
526
 */
527
static long region_add(struct resv_map *resv, long f, long t,
528 529
		       long in_regions_needed, struct hstate *h,
		       struct hugetlb_cgroup *h_cg)
530
{
531
	long add = 0, actual_regions_needed = 0;
532

533
	spin_lock(&resv->lock);
534 535 536
retry:

	/* Count how many regions are actually needed to execute this add. */
537 538
	add_reservation_in_range(resv, f, t, NULL, NULL,
				 &actual_regions_needed);
539

540
	/*
541 542 543 544 545 546 547
	 * 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.
548
	 */
549 550 551 552 553 554 555 556
	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);
557

558 559 560 561
		if (allocate_file_region_entries(
			    resv, actual_regions_needed - in_regions_needed)) {
			return -ENOMEM;
		}
562

563
		goto retry;
564 565
	}

566
	add = add_reservation_in_range(resv, f, t, h_cg, h, NULL);
567 568

	resv->adds_in_progress -= in_regions_needed;
569

570
	spin_unlock(&resv->lock);
571
	return add;
572 573
}

574 575 576 577 578 579 580
/*
 * 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
581 582 583 584 585 586 587
 * 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.
588 589 590 591 592
 *
 * 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.
593
 */
594 595
static long region_chg(struct resv_map *resv, long f, long t,
		       long *out_regions_needed)
596 597 598
{
	long chg = 0;

599
	spin_lock(&resv->lock);
600

601
	/* Count how many hugepages in this range are NOT represented. */
602
	chg = add_reservation_in_range(resv, f, t, NULL, NULL,
603
				       out_regions_needed);
604

605 606
	if (*out_regions_needed == 0)
		*out_regions_needed = 1;
607

608 609
	if (allocate_file_region_entries(resv, *out_regions_needed))
		return -ENOMEM;
610

611
	resv->adds_in_progress += *out_regions_needed;
612 613

	spin_unlock(&resv->lock);
614 615 616
	return chg;
}

617 618 619 620 621
/*
 * 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
622 623 624
 * 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.
625 626 627 628 629
 *
 * 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.
 */
630 631
static void region_abort(struct resv_map *resv, long f, long t,
			 long regions_needed)
632 633 634
{
	spin_lock(&resv->lock);
	VM_BUG_ON(!resv->region_cache_count);
635
	resv->adds_in_progress -= regions_needed;
636 637 638
	spin_unlock(&resv->lock);
}

639
/*
640 641 642 643 644 645 646 647 648 649 650 651
 * 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.
652
 */
653
static long region_del(struct resv_map *resv, long f, long t)
654
{
655
	struct list_head *head = &resv->regions;
656
	struct file_region *rg, *trg;
657 658
	struct file_region *nrg = NULL;
	long del = 0;
659

660
retry:
661
	spin_lock(&resv->lock);
662
	list_for_each_entry_safe(rg, trg, head, link) {
663 664 665 666 667 668 669 670
		/*
		 * 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))
671
			continue;
672

673
		if (rg->from >= t)
674 675
			break;

676 677 678 679 680 681 682 683 684 685 686 687 688
		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--;
			}
689

690 691 692 693 694 695 696 697 698
			if (!nrg) {
				spin_unlock(&resv->lock);
				nrg = kmalloc(sizeof(*nrg), GFP_KERNEL);
				if (!nrg)
					return -ENOMEM;
				goto retry;
			}

			del += t - f;
699
			hugetlb_cgroup_uncharge_file_region(
700
				resv, rg, t - f, false);
701 702 703 704

			/* New entry for end of split region */
			nrg->from = t;
			nrg->to = rg->to;
705 706 707

			copy_hugetlb_cgroup_uncharge_info(nrg, rg);

708 709 710 711 712 713 714
			INIT_LIST_HEAD(&nrg->link);

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

			list_add(&nrg->link, &rg->link);
			nrg = NULL;
715
			break;
716 717 718 719
		}

		if (f <= rg->from && t >= rg->to) { /* Remove entire region */
			del += rg->to - rg->from;
720
			hugetlb_cgroup_uncharge_file_region(resv, rg,
721
							    rg->to - rg->from, true);
722 723 724 725 726 727
			list_del(&rg->link);
			kfree(rg);
			continue;
		}

		if (f <= rg->from) {	/* Trim beginning of region */
728
			hugetlb_cgroup_uncharge_file_region(resv, rg,
729
							    t - rg->from, false);
730

731 732 733
			del += t - rg->from;
			rg->from = t;
		} else {		/* Trim end of region */
734
			hugetlb_cgroup_uncharge_file_region(resv, rg,
735
							    rg->to - f, false);
736 737 738

			del += rg->to - f;
			rg->to = f;
739
		}
740
	}
741 742

	spin_unlock(&resv->lock);
743 744
	kfree(nrg);
	return del;
745 746
}

747 748 749 750 751 752 753 754 755
/*
 * 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.
 */
756
void hugetlb_fix_reserve_counts(struct inode *inode)
757 758 759
{
	struct hugepage_subpool *spool = subpool_inode(inode);
	long rsv_adjust;
760
	bool reserved = false;
761 762

	rsv_adjust = hugepage_subpool_get_pages(spool, 1);
763
	if (rsv_adjust > 0) {
764 765
		struct hstate *h = hstate_inode(inode);

766 767 768 769
		if (!hugetlb_acct_memory(h, 1))
			reserved = true;
	} else if (!rsv_adjust) {
		reserved = true;
770
	}
771 772 773

	if (!reserved)
		pr_warn("hugetlb: Huge Page Reserved count may go negative.\n");
774 775
}

776 777 778 779
/*
 * Count and return the number of huge pages in the reserve map
 * that intersect with the range [f, t).
 */
780
static long region_count(struct resv_map *resv, long f, long t)
781
{
782
	struct list_head *head = &resv->regions;
783 784 785
	struct file_region *rg;
	long chg = 0;

786
	spin_lock(&resv->lock);
787 788
	/* Locate each segment we overlap with, and count that overlap. */
	list_for_each_entry(rg, head, link) {
789 790
		long seg_from;
		long seg_to;
791 792 793 794 795 796 797 798 799 800 801

		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;
	}
802
	spin_unlock(&resv->lock);
803 804 805 806

	return chg;
}

807 808 809 810
/*
 * Convert the address within this vma to the page offset within
 * the mapping, in pagecache page units; huge pages here.
 */
811 812
static pgoff_t vma_hugecache_offset(struct hstate *h,
			struct vm_area_struct *vma, unsigned long address)
813
{
814 815
	return ((address - vma->vm_start) >> huge_page_shift(h)) +
			(vma->vm_pgoff >> huge_page_order(h));
816 817
}

818 819 820 821 822
pgoff_t linear_hugepage_index(struct vm_area_struct *vma,
				     unsigned long address)
{
	return vma_hugecache_offset(hstate_vma(vma), vma, address);
}
823
EXPORT_SYMBOL_GPL(linear_hugepage_index);
824

825 826 827 828 829 830
/*
 * 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)
{
831 832 833
	if (vma->vm_ops && vma->vm_ops->pagesize)
		return vma->vm_ops->pagesize(vma);
	return PAGE_SIZE;
834
}
835
EXPORT_SYMBOL_GPL(vma_kernel_pagesize);
836

837 838 839
/*
 * 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
840 841
 * architectures where it differs, an architecture-specific 'strong'
 * version of this symbol is required.
842
 */
843
__weak unsigned long vma_mmu_pagesize(struct vm_area_struct *vma)
844 845 846 847
{
	return vma_kernel_pagesize(vma);
}

848 849 850 851 852 853 854
/*
 * 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)
855
#define HPAGE_RESV_MASK (HPAGE_RESV_OWNER | HPAGE_RESV_UNMAPPED)
856

857 858 859 860 861 862 863 864 865
/*
 * 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.
866 867 868 869 870 871 872 873 874
 *
 * 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.
875
 */
876 877 878 879 880 881 882 883 884 885 886
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;
}

887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905
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
}

906
struct resv_map *resv_map_alloc(void)
907 908
{
	struct resv_map *resv_map = kmalloc(sizeof(*resv_map), GFP_KERNEL);
909 910 911 912 913
	struct file_region *rg = kmalloc(sizeof(*rg), GFP_KERNEL);

	if (!resv_map || !rg) {
		kfree(resv_map);
		kfree(rg);
914
		return NULL;
915
	}
916 917

	kref_init(&resv_map->refs);
918
	spin_lock_init(&resv_map->lock);
919 920
	INIT_LIST_HEAD(&resv_map->regions);

921
	resv_map->adds_in_progress = 0;
922 923 924 925 926 927 928
	/*
	 * 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);
929 930 931 932 933

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

934 935 936
	return resv_map;
}

937
void resv_map_release(struct kref *ref)
938 939
{
	struct resv_map *resv_map = container_of(ref, struct resv_map, refs);
940 941
	struct list_head *head = &resv_map->region_cache;
	struct file_region *rg, *trg;
942 943

	/* Clear out any active regions before we release the map. */
944
	region_del(resv_map, 0, LONG_MAX);
945 946 947 948 949 950 951 952 953

	/* ... 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);

954 955 956
	kfree(resv_map);
}

957 958
static inline struct resv_map *inode_resv_map(struct inode *inode)
{
959 960 961 962 963 964 965 966 967
	/*
	 * 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;
968 969
}

970
static struct resv_map *vma_resv_map(struct vm_area_struct *vma)
971
{
972
	VM_BUG_ON_VMA(!is_vm_hugetlb_page(vma), vma);
973 974 975 976 977 978 979
	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 {
980 981
		return (struct resv_map *)(get_vma_private_data(vma) &
							~HPAGE_RESV_MASK);
982
	}
983 984
}

985
static void set_vma_resv_map(struct vm_area_struct *vma, struct resv_map *map)
986
{
987 988
	VM_BUG_ON_VMA(!is_vm_hugetlb_page(vma), vma);
	VM_BUG_ON_VMA(vma->vm_flags & VM_MAYSHARE, vma);
989

990 991
	set_vma_private_data(vma, (get_vma_private_data(vma) &
				HPAGE_RESV_MASK) | (unsigned long)map);
992 993 994 995
}

static void set_vma_resv_flags(struct vm_area_struct *vma, unsigned long flags)
{
996 997
	VM_BUG_ON_VMA(!is_vm_hugetlb_page(vma), vma);
	VM_BUG_ON_VMA(vma->vm_flags & VM_MAYSHARE, vma);
998 999

	set_vma_private_data(vma, get_vma_private_data(vma) | flags);
1000 1001 1002 1003
}

static int is_vma_resv_set(struct vm_area_struct *vma, unsigned long flag)
{
1004
	VM_BUG_ON_VMA(!is_vm_hugetlb_page(vma), vma);
1005 1006

	return (get_vma_private_data(vma) & flag) != 0;
1007 1008
}

1009
/* Reset counters to 0 and clear all HPAGE_RESV_* flags */
1010 1011
void reset_vma_resv_huge_pages(struct vm_area_struct *vma)
{
1012
	VM_BUG_ON_VMA(!is_vm_hugetlb_page(vma), vma);
1013
	if (!(vma->vm_flags & VM_MAYSHARE))
1014 1015 1016
		vma->vm_private_data = (void *)0;
}

1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039
/*
 * Reset and decrement one ref on hugepage private reservation.
 * Called with mm->mmap_sem writer semaphore held.
 * This function should be only used by move_vma() and operate on
 * same sized vma. It should never come here with last ref on the
 * reservation.
 */
void clear_vma_resv_huge_pages(struct vm_area_struct *vma)
{
	/*
	 * Clear the old hugetlb private page reservation.
	 * It has already been transferred to new_vma.
	 *
	 * During a mremap() operation of a hugetlb vma we call move_vma()
	 * which copies vma into new_vma and unmaps vma. After the copy
	 * operation both new_vma and vma share a reference to the resv_map
	 * struct, and at that point vma is about to be unmapped. We don't
	 * want to return the reservation to the pool at unmap of vma because
	 * the reservation still lives on in new_vma, so simply decrement the
	 * ref here and remove the resv_map reference from this vma.
	 */
	struct resv_map *reservations = vma_resv_map(vma);

1040 1041
	if (reservations && is_vma_resv_set(vma, HPAGE_RESV_OWNER)) {
		resv_map_put_hugetlb_cgroup_uncharge_info(reservations);
1042
		kref_put(&reservations->refs, resv_map_release);
1043
	}
1044 1045 1046 1047

	reset_vma_resv_huge_pages(vma);
}

1048
/* Returns true if the VMA has associated reserve pages */
1049
static bool vma_has_reserves(struct vm_area_struct *vma, long chg)
1050
{
1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061
	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)
1062
			return true;
1063
		else
1064
			return false;
1065
	}
1066 1067

	/* Shared mappings always use reserves */
1068 1069 1070 1071 1072
	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
1073
		 * fallocate.  In this case, there really are no reserves to
1074 1075 1076 1077 1078 1079 1080
		 * use.  This situation is indicated if chg != 0.
		 */
		if (chg)
			return false;
		else
			return true;
	}
1081 1082 1083 1084 1085

	/*
	 * Only the process that called mmap() has reserves for
	 * private mappings.
	 */
1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106
	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;
	}
1107

1108
	return false;
1109 1110
}

1111
static void enqueue_huge_page(struct hstate *h, struct page *page)
L
Linus Torvalds 已提交
1112 1113
{
	int nid = page_to_nid(page);
1114 1115

	lockdep_assert_held(&hugetlb_lock);
1116 1117
	VM_BUG_ON_PAGE(page_count(page), page);

1118
	list_move(&page->lru, &h->hugepage_freelists[nid]);
1119 1120
	h->free_huge_pages++;
	h->free_huge_pages_node[nid]++;
1121
	SetHPageFreed(page);
L
Linus Torvalds 已提交
1122 1123
}

1124
static struct page *dequeue_huge_page_node_exact(struct hstate *h, int nid)
1125 1126
{
	struct page *page;
1127
	bool pin = !!(current->flags & PF_MEMALLOC_PIN);
1128

1129
	lockdep_assert_held(&hugetlb_lock);
1130
	list_for_each_entry(page, &h->hugepage_freelists[nid], lru) {
1131
		if (pin && !is_pinnable_page(page))
1132
			continue;
1133

1134 1135 1136 1137 1138
		if (PageHWPoison(page))
			continue;

		list_move(&page->lru, &h->hugepage_activelist);
		set_page_refcounted(page);
1139
		ClearHPageFreed(page);
1140 1141 1142
		h->free_huge_pages--;
		h->free_huge_pages_node[nid]--;
		return page;
1143 1144
	}

1145
	return NULL;
1146 1147
}

1148 1149
static struct page *dequeue_huge_page_nodemask(struct hstate *h, gfp_t gfp_mask, int nid,
		nodemask_t *nmask)
1150
{
1151 1152 1153 1154
	unsigned int cpuset_mems_cookie;
	struct zonelist *zonelist;
	struct zone *zone;
	struct zoneref *z;
1155
	int node = NUMA_NO_NODE;
1156

1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172
	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);
1173 1174 1175 1176 1177

		page = dequeue_huge_page_node_exact(h, node);
		if (page)
			return page;
	}
1178 1179 1180
	if (unlikely(read_mems_allowed_retry(cpuset_mems_cookie)))
		goto retry_cpuset;

1181 1182 1183
	return NULL;
}

1184 1185
static struct page *dequeue_huge_page_vma(struct hstate *h,
				struct vm_area_struct *vma,
1186 1187
				unsigned long address, int avoid_reserve,
				long chg)
L
Linus Torvalds 已提交
1188
{
1189
	struct page *page = NULL;
1190
	struct mempolicy *mpol;
1191
	gfp_t gfp_mask;
1192
	nodemask_t *nodemask;
1193
	int nid;
L
Linus Torvalds 已提交
1194

1195 1196 1197 1198 1199
	/*
	 * 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
	 */
1200
	if (!vma_has_reserves(vma, chg) &&
1201
			h->free_huge_pages - h->resv_huge_pages == 0)
1202
		goto err;
1203

1204
	/* If reserves cannot be used, ensure enough pages are in the pool */
1205
	if (avoid_reserve && h->free_huge_pages - h->resv_huge_pages == 0)
1206
		goto err;
1207

1208 1209
	gfp_mask = htlb_alloc_mask(h);
	nid = huge_node(vma, address, gfp_mask, &mpol, &nodemask);
1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220

	if (mpol_is_preferred_many(mpol)) {
		page = dequeue_huge_page_nodemask(h, gfp_mask, nid, nodemask);

		/* Fallback to all nodes if page==NULL */
		nodemask = NULL;
	}

	if (!page)
		page = dequeue_huge_page_nodemask(h, gfp_mask, nid, nodemask);

1221
	if (page && !avoid_reserve && vma_has_reserves(vma, chg)) {
1222
		SetHPageRestoreReserve(page);
1223
		h->resv_huge_pages--;
L
Linus Torvalds 已提交
1224
	}
1225

1226
	mpol_cond_put(mpol);
L
Linus Torvalds 已提交
1227
	return page;
1228 1229 1230

err:
	return NULL;
L
Linus Torvalds 已提交
1231 1232
}

1233 1234 1235 1236 1237 1238 1239 1240 1241
/*
 * 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)
{
1242
	nid = next_node_in(nid, *nodes_allowed);
1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274
	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;
}

/*
1275
 * helper for remove_pool_huge_page() - return the previously saved
1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303
 * 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--)

1304
/* used to demote non-gigantic_huge pages as well */
1305 1306
static void __destroy_compound_gigantic_page(struct page *page,
					unsigned int order, bool demote)
1307 1308 1309 1310 1311
{
	int i;
	int nr_pages = 1 << order;
	struct page *p = page + 1;

1312
	atomic_set(compound_mapcount_ptr(page), 0);
1313
	atomic_set(compound_pincount_ptr(page), 0);
1314

1315
	for (i = 1; i < nr_pages; i++, p = mem_map_next(p, page, i)) {
1316
		p->mapping = NULL;
1317
		clear_compound_head(p);
1318 1319
		if (!demote)
			set_page_refcounted(p);
1320 1321 1322
	}

	set_compound_order(page, 0);
1323
	page[1].compound_nr = 0;
1324 1325 1326
	__ClearPageHead(page);
}

1327 1328 1329 1330 1331 1332 1333
static void destroy_compound_hugetlb_page_for_demote(struct page *page,
					unsigned int order)
{
	__destroy_compound_gigantic_page(page, order, true);
}

#ifdef CONFIG_ARCH_HAS_GIGANTIC_PAGE
1334 1335 1336 1337 1338 1339
static void destroy_compound_gigantic_page(struct page *page,
					unsigned int order)
{
	__destroy_compound_gigantic_page(page, order, false);
}

1340
static void free_gigantic_page(struct page *page, unsigned int order)
1341
{
1342 1343 1344 1345
	/*
	 * If the page isn't allocated using the cma allocator,
	 * cma_release() returns false.
	 */
1346 1347
#ifdef CONFIG_CMA
	if (cma_release(hugetlb_cma[page_to_nid(page)], page, 1 << order))
1348
		return;
1349
#endif
1350

1351 1352 1353
	free_contig_range(page_to_pfn(page), 1 << order);
}

1354
#ifdef CONFIG_CONTIG_ALLOC
1355 1356
static struct page *alloc_gigantic_page(struct hstate *h, gfp_t gfp_mask,
		int nid, nodemask_t *nodemask)
1357
{
1358
	unsigned long nr_pages = pages_per_huge_page(h);
1359 1360
	if (nid == NUMA_NO_NODE)
		nid = numa_mem_id();
1361

1362 1363
#ifdef CONFIG_CMA
	{
1364 1365 1366
		struct page *page;
		int node;

1367 1368 1369
		if (hugetlb_cma[nid]) {
			page = cma_alloc(hugetlb_cma[nid], nr_pages,
					huge_page_order(h), true);
1370 1371 1372
			if (page)
				return page;
		}
1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384

		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;
			}
		}
1385
	}
1386
#endif
1387

1388
	return alloc_contig_pages(nr_pages, gfp_mask, nid, nodemask);
1389 1390
}

1391 1392 1393 1394 1395 1396 1397
#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 */
1398

1399
#else /* !CONFIG_ARCH_HAS_GIGANTIC_PAGE */
1400
static struct page *alloc_gigantic_page(struct hstate *h, gfp_t gfp_mask,
1401 1402 1403 1404
					int nid, nodemask_t *nodemask)
{
	return NULL;
}
1405
static inline void free_gigantic_page(struct page *page, unsigned int order) { }
1406
static inline void destroy_compound_gigantic_page(struct page *page,
1407
						unsigned int order) { }
1408 1409
#endif

1410 1411
/*
 * Remove hugetlb page from lists, and update dtor so that page appears
1412 1413 1414
 * as just a compound page.
 *
 * A reference is held on the page, except in the case of demote.
1415 1416 1417
 *
 * Must be called with hugetlb lock held.
 */
1418 1419 1420
static void __remove_hugetlb_page(struct hstate *h, struct page *page,
							bool adjust_surplus,
							bool demote)
1421 1422 1423 1424 1425 1426
{
	int nid = page_to_nid(page);

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

1427
	lockdep_assert_held(&hugetlb_lock);
1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441
	if (hstate_is_gigantic(h) && !gigantic_page_runtime_supported())
		return;

	list_del(&page->lru);

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

1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457
	/*
	 * Very subtle
	 *
	 * For non-gigantic pages set the destructor to the normal compound
	 * page dtor.  This is needed in case someone takes an additional
	 * temporary ref to the page, and freeing is delayed until they drop
	 * their reference.
	 *
	 * For gigantic pages set the destructor to the null dtor.  This
	 * destructor will never be called.  Before freeing the gigantic
	 * page destroy_compound_gigantic_page will turn the compound page
	 * into a simple group of pages.  After this the destructor does not
	 * apply.
	 *
	 * This handles the case where more than one ref is held when and
	 * after update_and_free_page is called.
1458 1459 1460
	 *
	 * In the case of demote we do not ref count the page as it will soon
	 * be turned into a page of smaller size.
1461
	 */
1462 1463
	if (!demote)
		set_page_refcounted(page);
1464 1465 1466 1467
	if (hstate_is_gigantic(h))
		set_compound_page_dtor(page, NULL_COMPOUND_DTOR);
	else
		set_compound_page_dtor(page, COMPOUND_PAGE_DTOR);
1468 1469 1470 1471 1472

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

1473 1474 1475 1476 1477 1478
static void remove_hugetlb_page(struct hstate *h, struct page *page,
							bool adjust_surplus)
{
	__remove_hugetlb_page(h, page, adjust_surplus, false);
}

1479 1480 1481 1482 1483 1484
static void remove_hugetlb_page_for_demote(struct hstate *h, struct page *page,
							bool adjust_surplus)
{
	__remove_hugetlb_page(h, page, adjust_surplus, true);
}

1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508
static void add_hugetlb_page(struct hstate *h, struct page *page,
			     bool adjust_surplus)
{
	int zeroed;
	int nid = page_to_nid(page);

	VM_BUG_ON_PAGE(!HPageVmemmapOptimized(page), page);

	lockdep_assert_held(&hugetlb_lock);

	INIT_LIST_HEAD(&page->lru);
	h->nr_huge_pages++;
	h->nr_huge_pages_node[nid]++;

	if (adjust_surplus) {
		h->surplus_huge_pages++;
		h->surplus_huge_pages_node[nid]++;
	}

	set_compound_page_dtor(page, HUGETLB_PAGE_DTOR);
	set_page_private(page, 0);
	SetHPageVmemmapOptimized(page);

	/*
1509 1510 1511
	 * This page is about to be managed by the hugetlb allocator and
	 * should have no users.  Drop our reference, and check for others
	 * just in case.
1512 1513
	 */
	zeroed = put_page_testzero(page);
1514 1515 1516 1517 1518 1519 1520 1521 1522
	if (!zeroed)
		/*
		 * It is VERY unlikely soneone else has taken a ref on
		 * the page.  In this case, we simply return as the
		 * hugetlb destructor (free_huge_page) will be called
		 * when this other ref is dropped.
		 */
		return;

1523 1524 1525 1526
	arch_clear_hugepage_flags(page);
	enqueue_huge_page(h, page);
}

1527
static void __update_and_free_page(struct hstate *h, struct page *page)
A
Adam Litke 已提交
1528 1529
{
	int i;
1530
	struct page *subpage = page;
1531

1532
	if (hstate_is_gigantic(h) && !gigantic_page_runtime_supported())
1533
		return;
1534

1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546
	if (alloc_huge_page_vmemmap(h, page)) {
		spin_lock_irq(&hugetlb_lock);
		/*
		 * If we cannot allocate vmemmap pages, just refuse to free the
		 * page and put the page back on the hugetlb free list and treat
		 * as a surplus page.
		 */
		add_hugetlb_page(h, page, true);
		spin_unlock_irq(&hugetlb_lock);
		return;
	}

1547 1548 1549
	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 |
1550
				1 << PG_referenced | 1 << PG_dirty |
1551 1552
				1 << PG_active | 1 << PG_private |
				1 << PG_writeback);
A
Adam Litke 已提交
1553
	}
1554 1555 1556 1557 1558 1559 1560

	/*
	 * Non-gigantic pages demoted from CMA allocated gigantic pages
	 * need to be given back to CMA in free_gigantic_page.
	 */
	if (hstate_is_gigantic(h) ||
	    hugetlb_cma_page(page, huge_page_order(h))) {
1561 1562 1563 1564 1565
		destroy_compound_gigantic_page(page, huge_page_order(h));
		free_gigantic_page(page, huge_page_order(h));
	} else {
		__free_pages(page, huge_page_order(h));
	}
A
Adam Litke 已提交
1566 1567
}

1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618
/*
 * As update_and_free_page() can be called under any context, so we cannot
 * use GFP_KERNEL to allocate vmemmap pages. However, we can defer the
 * actual freeing in a workqueue to prevent from using GFP_ATOMIC to allocate
 * the vmemmap pages.
 *
 * 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_hpage_workfn() 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;

	node = llist_del_all(&hpage_freelist);

	while (node) {
		struct page *page;
		struct hstate *h;

		page = container_of((struct address_space **)node,
				     struct page, mapping);
		node = node->next;
		page->mapping = NULL;
		/*
		 * The VM_BUG_ON_PAGE(!PageHuge(page), page) in page_hstate()
		 * is going to trigger because a previous call to
		 * remove_hugetlb_page() will set_compound_page_dtor(page,
		 * NULL_COMPOUND_DTOR), so do not use page_hstate() directly.
		 */
		h = size_to_hstate(page_size(page));

		__update_and_free_page(h, page);

		cond_resched();
	}
}
static DECLARE_WORK(free_hpage_work, free_hpage_workfn);

static inline void flush_free_hpage_work(struct hstate *h)
{
	if (free_vmemmap_pages_per_hpage(h))
		flush_work(&free_hpage_work);
}

static void update_and_free_page(struct hstate *h, struct page *page,
				 bool atomic)
{
1619
	if (!HPageVmemmapOptimized(page) || !atomic) {
1620 1621 1622 1623 1624 1625 1626 1627 1628 1629 1630 1631 1632 1633 1634
		__update_and_free_page(h, page);
		return;
	}

	/*
	 * Defer freeing to avoid using GFP_ATOMIC to allocate vmemmap pages.
	 *
	 * 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);
}

1635 1636 1637 1638 1639
static void update_and_free_pages_bulk(struct hstate *h, struct list_head *list)
{
	struct page *page, *t_page;

	list_for_each_entry_safe(page, t_page, list, lru) {
1640
		update_and_free_page(h, page, false);
1641 1642 1643 1644
		cond_resched();
	}
}

1645 1646 1647 1648 1649 1650 1651 1652 1653 1654 1655
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;
}

1656
void free_huge_page(struct page *page)
1657
{
1658 1659 1660 1661
	/*
	 * Can't pass hstate in here because it is called from the
	 * compound page destructor.
	 */
1662
	struct hstate *h = page_hstate(page);
1663
	int nid = page_to_nid(page);
1664
	struct hugepage_subpool *spool = hugetlb_page_subpool(page);
1665
	bool restore_reserve;
1666
	unsigned long flags;
1667

1668 1669
	VM_BUG_ON_PAGE(page_count(page), page);
	VM_BUG_ON_PAGE(page_mapcount(page), page);
1670

1671
	hugetlb_set_page_subpool(page, NULL);
1672
	page->mapping = NULL;
1673 1674
	restore_reserve = HPageRestoreReserve(page);
	ClearHPageRestoreReserve(page);
1675

1676
	/*
1677
	 * If HPageRestoreReserve was set on page, page allocation consumed a
1678 1679 1680
	 * 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
1681
	 * reservation, do not call hugepage_subpool_put_pages() as this will
1682
	 * remove the reserved page from the subpool.
1683
	 */
1684 1685 1686 1687 1688 1689 1690 1691 1692 1693
	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;
	}
1694

1695
	spin_lock_irqsave(&hugetlb_lock, flags);
1696
	ClearHPageMigratable(page);
1697 1698
	hugetlb_cgroup_uncharge_page(hstate_index(h),
				     pages_per_huge_page(h), page);
1699 1700
	hugetlb_cgroup_uncharge_page_rsvd(hstate_index(h),
					  pages_per_huge_page(h), page);
1701 1702 1703
	if (restore_reserve)
		h->resv_huge_pages++;

1704
	if (HPageTemporary(page)) {
1705
		remove_hugetlb_page(h, page, false);
1706
		spin_unlock_irqrestore(&hugetlb_lock, flags);
1707
		update_and_free_page(h, page, true);
1708
	} else if (h->surplus_huge_pages_node[nid]) {
1709
		/* remove the page from active list */
1710
		remove_hugetlb_page(h, page, true);
1711
		spin_unlock_irqrestore(&hugetlb_lock, flags);
1712
		update_and_free_page(h, page, true);
1713
	} else {
1714
		arch_clear_hugepage_flags(page);
1715
		enqueue_huge_page(h, page);
1716
		spin_unlock_irqrestore(&hugetlb_lock, flags);
1717 1718 1719
	}
}

1720 1721 1722 1723 1724 1725 1726 1727 1728 1729
/*
 * Must be called with the hugetlb lock held
 */
static void __prep_account_new_huge_page(struct hstate *h, int nid)
{
	lockdep_assert_held(&hugetlb_lock);
	h->nr_huge_pages++;
	h->nr_huge_pages_node[nid]++;
}

1730
static void __prep_new_huge_page(struct hstate *h, struct page *page)
1731
{
1732
	free_huge_page_vmemmap(h, page);
1733
	INIT_LIST_HEAD(&page->lru);
1734
	set_compound_page_dtor(page, HUGETLB_PAGE_DTOR);
1735
	hugetlb_set_page_subpool(page, NULL);
1736
	set_hugetlb_cgroup(page, NULL);
1737
	set_hugetlb_cgroup_rsvd(page, NULL);
1738 1739 1740 1741
}

static void prep_new_huge_page(struct hstate *h, struct page *page, int nid)
{
1742
	__prep_new_huge_page(h, page);
1743
	spin_lock_irq(&hugetlb_lock);
1744
	__prep_account_new_huge_page(h, nid);
1745
	spin_unlock_irq(&hugetlb_lock);
1746 1747
}

1748 1749
static bool __prep_compound_gigantic_page(struct page *page, unsigned int order,
								bool demote)
1750
{
1751
	int i, j;
1752 1753 1754 1755 1756
	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);
1757
	__ClearPageReserved(page);
1758
	__SetPageHead(page);
1759
	for (i = 1; i < nr_pages; i++, p = mem_map_next(p, page, i)) {
1760 1761 1762 1763
		/*
		 * 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
1764
		 * too.  Otherwise drivers using get_user_pages() to access tail
1765 1766 1767 1768 1769 1770 1771 1772
		 * 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);
1773 1774 1775 1776 1777 1778 1779 1780 1781 1782
		/*
		 * Subtle and very unlikely
		 *
		 * Gigantic 'page allocators' such as memblock or cma will
		 * return a set of pages with each page ref counted.  We need
		 * to turn this set of pages into a compound page with tail
		 * page ref counts set to zero.  Code such as speculative page
		 * cache adding could take a ref on a 'to be' tail page.
		 * We need to respect any increased ref count, and only set
		 * the ref count to zero if count is currently 1.  If count
1783 1784 1785 1786
		 * is not 1, we return an error.  An error return indicates
		 * the set of pages can not be converted to a gigantic page.
		 * The caller who allocated the pages should then discard the
		 * pages using the appropriate free interface.
1787 1788
		 *
		 * In the case of demote, the ref count will be zero.
1789
		 */
1790 1791 1792 1793 1794 1795 1796
		if (!demote) {
			if (!page_ref_freeze(p, 1)) {
				pr_warn("HugeTLB page can not be used due to unexpected inflated ref count\n");
				goto out_error;
			}
		} else {
			VM_BUG_ON_PAGE(page_count(p), p);
1797
		}
1798
		set_compound_head(p, page);
1799
	}
1800
	atomic_set(compound_mapcount_ptr(page), -1);
1801
	atomic_set(compound_pincount_ptr(page), 0);
1802 1803 1804 1805 1806 1807 1808 1809 1810 1811 1812 1813 1814 1815 1816 1817
	return true;

out_error:
	/* undo tail page modifications made above */
	p = page + 1;
	for (j = 1; j < i; j++, p = mem_map_next(p, page, j)) {
		clear_compound_head(p);
		set_page_refcounted(p);
	}
	/* need to clear PG_reserved on remaining tail pages  */
	for (; j < nr_pages; j++, p = mem_map_next(p, page, j))
		__ClearPageReserved(p);
	set_compound_order(page, 0);
	page[1].compound_nr = 0;
	__ClearPageHead(page);
	return false;
1818 1819
}

1820 1821 1822 1823 1824
static bool prep_compound_gigantic_page(struct page *page, unsigned int order)
{
	return __prep_compound_gigantic_page(page, order, false);
}

1825 1826 1827 1828 1829 1830
static bool prep_compound_gigantic_page_for_demote(struct page *page,
							unsigned int order)
{
	return __prep_compound_gigantic_page(page, order, true);
}

A
Andrew Morton 已提交
1831 1832 1833 1834 1835
/*
 * PageHuge() only returns true for hugetlbfs pages, but not for normal or
 * transparent huge pages.  See the PageTransHuge() documentation for more
 * details.
 */
1836 1837 1838 1839 1840 1841
int PageHuge(struct page *page)
{
	if (!PageCompound(page))
		return 0;

	page = compound_head(page);
1842
	return page[1].compound_dtor == HUGETLB_PAGE_DTOR;
1843
}
1844 1845
EXPORT_SYMBOL_GPL(PageHuge);

1846 1847 1848 1849 1850 1851 1852 1853 1854
/*
 * 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;

1855
	return page_head[1].compound_dtor == HUGETLB_PAGE_DTOR;
1856 1857
}

1858 1859 1860
/*
 * Find and lock address space (mapping) in write mode.
 *
1861 1862 1863
 * 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.
1864 1865 1866
 */
struct address_space *hugetlb_page_mapping_lock_write(struct page *hpage)
{
1867
	struct address_space *mapping = page_mapping(hpage);
1868 1869 1870 1871 1872 1873 1874

	if (!mapping)
		return mapping;

	if (i_mmap_trylock_write(mapping))
		return mapping;

1875
	return NULL;
1876 1877
}

1878
pgoff_t hugetlb_basepage_index(struct page *page)
1879 1880 1881 1882 1883 1884 1885 1886 1887 1888 1889 1890 1891
{
	struct page *page_head = compound_head(page);
	pgoff_t index = page_index(page_head);
	unsigned long compound_idx;

	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;
}

1892
static struct page *alloc_buddy_huge_page(struct hstate *h,
1893 1894
		gfp_t gfp_mask, int nid, nodemask_t *nmask,
		nodemask_t *node_alloc_noretry)
L
Linus Torvalds 已提交
1895
{
1896
	int order = huge_page_order(h);
L
Linus Torvalds 已提交
1897
	struct page *page;
1898
	bool alloc_try_hard = true;
1899

1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911
	/*
	 * 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;
1912 1913
	if (nid == NUMA_NO_NODE)
		nid = numa_mem_id();
1914
	page = __alloc_pages(gfp_mask, order, nid, nmask);
1915 1916 1917 1918
	if (page)
		__count_vm_event(HTLB_BUDDY_PGALLOC);
	else
		__count_vm_event(HTLB_BUDDY_PGALLOC_FAIL);
1919

1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931 1932 1933 1934 1935
	/*
	 * 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);

1936 1937 1938
	return page;
}

1939 1940 1941 1942 1943
/*
 * 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,
1944 1945
		gfp_t gfp_mask, int nid, nodemask_t *nmask,
		nodemask_t *node_alloc_noretry)
1946 1947
{
	struct page *page;
1948
	bool retry = false;
1949

1950
retry:
1951 1952 1953 1954
	if (hstate_is_gigantic(h))
		page = alloc_gigantic_page(h, gfp_mask, nid, nmask);
	else
		page = alloc_buddy_huge_page(h, gfp_mask,
1955
				nid, nmask, node_alloc_noretry);
1956 1957 1958
	if (!page)
		return NULL;

1959 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972
	if (hstate_is_gigantic(h)) {
		if (!prep_compound_gigantic_page(page, huge_page_order(h))) {
			/*
			 * Rare failure to convert pages to compound page.
			 * Free pages and try again - ONCE!
			 */
			free_gigantic_page(page, huge_page_order(h));
			if (!retry) {
				retry = true;
				goto retry;
			}
			return NULL;
		}
	}
1973 1974 1975 1976 1977
	prep_new_huge_page(h, page, page_to_nid(page));

	return page;
}

1978 1979 1980 1981
/*
 * Allocates a fresh page to the hugetlb allocator pool in the node interleaved
 * manner.
 */
1982 1983
static int alloc_pool_huge_page(struct hstate *h, nodemask_t *nodes_allowed,
				nodemask_t *node_alloc_noretry)
1984 1985 1986
{
	struct page *page;
	int nr_nodes, node;
1987
	gfp_t gfp_mask = htlb_alloc_mask(h) | __GFP_THISNODE;
1988 1989

	for_each_node_mask_to_alloc(h, nr_nodes, node, nodes_allowed) {
1990 1991
		page = alloc_fresh_huge_page(h, gfp_mask, node, nodes_allowed,
						node_alloc_noretry);
1992
		if (page)
1993 1994 1995
			break;
	}

1996 1997
	if (!page)
		return 0;
1998

1999 2000 2001
	put_page(page); /* free it into the hugepage allocator */

	return 1;
2002 2003
}

2004
/*
2005 2006 2007 2008
 * Remove huge page from pool from next node to free.  Attempt to keep
 * persistent huge pages more or less balanced over allowed nodes.
 * This routine only 'removes' the hugetlb page.  The caller must make
 * an additional call to free the page to low level allocators.
2009 2010
 * Called with hugetlb_lock locked.
 */
2011 2012 2013
static struct page *remove_pool_huge_page(struct hstate *h,
						nodemask_t *nodes_allowed,
						 bool acct_surplus)
2014
{
2015
	int nr_nodes, node;
2016
	struct page *page = NULL;
2017

2018
	lockdep_assert_held(&hugetlb_lock);
2019
	for_each_node_mask_to_free(h, nr_nodes, node, nodes_allowed) {
2020 2021 2022 2023
		/*
		 * If we're returning unused surplus pages, only examine
		 * nodes with surplus pages.
		 */
2024 2025
		if ((!acct_surplus || h->surplus_huge_pages_node[node]) &&
		    !list_empty(&h->hugepage_freelists[node])) {
2026
			page = list_entry(h->hugepage_freelists[node].next,
2027
					  struct page, lru);
2028
			remove_hugetlb_page(h, page, acct_surplus);
2029
			break;
2030
		}
2031
	}
2032

2033
	return page;
2034 2035
}

2036 2037
/*
 * Dissolve a given free hugepage into free buddy pages. This function does
2038 2039 2040
 * nothing for in-use hugepages and non-hugepages.
 * This function returns values like below:
 *
2041 2042 2043 2044 2045 2046 2047 2048
 *  -ENOMEM: failed to allocate vmemmap pages to free the freed hugepages
 *           when the system is under memory pressure and the feature of
 *           freeing unused vmemmap pages associated with each hugetlb page
 *           is enabled.
 *  -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)
2049
 */
2050
int dissolve_free_huge_page(struct page *page)
2051
{
2052
	int rc = -EBUSY;
2053

2054
retry:
2055 2056 2057 2058
	/* Not to disrupt normal path by vainly holding hugetlb_lock */
	if (!PageHuge(page))
		return 0;

2059
	spin_lock_irq(&hugetlb_lock);
2060 2061 2062 2063 2064 2065
	if (!PageHuge(page)) {
		rc = 0;
		goto out;
	}

	if (!page_count(page)) {
2066 2067
		struct page *head = compound_head(page);
		struct hstate *h = page_hstate(head);
2068
		if (h->free_huge_pages - h->resv_huge_pages == 0)
2069
			goto out;
2070 2071 2072 2073 2074

		/*
		 * We should make sure that the page is already on the free list
		 * when it is dissolved.
		 */
2075
		if (unlikely(!HPageFreed(head))) {
2076
			spin_unlock_irq(&hugetlb_lock);
2077 2078 2079 2080 2081 2082 2083 2084 2085 2086 2087 2088 2089
			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;
		}

2090
		remove_hugetlb_page(h, head, false);
2091
		h->max_huge_pages--;
2092
		spin_unlock_irq(&hugetlb_lock);
2093 2094 2095 2096 2097 2098 2099 2100 2101 2102 2103 2104 2105 2106 2107 2108 2109 2110 2111 2112 2113 2114 2115 2116 2117 2118 2119 2120 2121

		/*
		 * Normally update_and_free_page will allocate required vmemmmap
		 * before freeing the page.  update_and_free_page will fail to
		 * free the page if it can not allocate required vmemmap.  We
		 * need to adjust max_huge_pages if the page is not freed.
		 * Attempt to allocate vmemmmap here so that we can take
		 * appropriate action on failure.
		 */
		rc = alloc_huge_page_vmemmap(h, head);
		if (!rc) {
			/*
			 * 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);
			}
			update_and_free_page(h, head, false);
		} else {
			spin_lock_irq(&hugetlb_lock);
			add_hugetlb_page(h, head, false);
			h->max_huge_pages++;
			spin_unlock_irq(&hugetlb_lock);
		}

		return rc;
2122
	}
2123
out:
2124
	spin_unlock_irq(&hugetlb_lock);
2125
	return rc;
2126 2127 2128 2129 2130
}

/*
 * Dissolve free hugepages in a given pfn range. Used by memory hotplug to
 * make specified memory blocks removable from the system.
2131 2132
 * Note that this will dissolve a free gigantic hugepage completely, if any
 * part of it lies within the given range.
2133 2134
 * Also note that if dissolve_free_huge_page() returns with an error, all
 * free hugepages that were dissolved before that error are lost.
2135
 */
2136
int dissolve_free_huge_pages(unsigned long start_pfn, unsigned long end_pfn)
2137 2138
{
	unsigned long pfn;
2139
	struct page *page;
2140
	int rc = 0;
2141

2142
	if (!hugepages_supported())
2143
		return rc;
2144

2145 2146
	for (pfn = start_pfn; pfn < end_pfn; pfn += 1 << minimum_order) {
		page = pfn_to_page(pfn);
2147 2148 2149
		rc = dissolve_free_huge_page(page);
		if (rc)
			break;
2150
	}
2151 2152

	return rc;
2153 2154
}

2155 2156 2157
/*
 * Allocates a fresh surplus page from the page allocator.
 */
2158
static struct page *alloc_surplus_huge_page(struct hstate *h, gfp_t gfp_mask,
2159
		int nid, nodemask_t *nmask, bool zero_ref)
2160
{
2161
	struct page *page = NULL;
2162
	bool retry = false;
2163

2164
	if (hstate_is_gigantic(h))
2165 2166
		return NULL;

2167
	spin_lock_irq(&hugetlb_lock);
2168 2169
	if (h->surplus_huge_pages >= h->nr_overcommit_huge_pages)
		goto out_unlock;
2170
	spin_unlock_irq(&hugetlb_lock);
2171

2172
retry:
2173
	page = alloc_fresh_huge_page(h, gfp_mask, nid, nmask, NULL);
2174
	if (!page)
2175
		return NULL;
2176

2177
	spin_lock_irq(&hugetlb_lock);
2178 2179 2180 2181 2182 2183 2184 2185
	/*
	 * 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) {
2186
		SetHPageTemporary(page);
2187
		spin_unlock_irq(&hugetlb_lock);
2188
		put_page(page);
2189
		return NULL;
2190
	}
2191

2192 2193 2194 2195 2196 2197 2198 2199 2200 2201 2202 2203 2204 2205 2206 2207 2208 2209 2210 2211 2212 2213 2214 2215 2216 2217 2218
	if (zero_ref) {
		/*
		 * Caller requires a page with zero ref count.
		 * We will drop ref count here.  If someone else is holding
		 * a ref, the page will be freed when they drop it.  Abuse
		 * temporary page flag to accomplish this.
		 */
		SetHPageTemporary(page);
		if (!put_page_testzero(page)) {
			/*
			 * Unexpected inflated ref count on freshly allocated
			 * huge.  Retry once.
			 */
			pr_info("HugeTLB unexpected inflated ref count on freshly allocated page\n");
			spin_unlock_irq(&hugetlb_lock);
			if (retry)
				return NULL;

			retry = true;
			goto retry;
		}
		ClearHPageTemporary(page);
	}

	h->surplus_huge_pages++;
	h->surplus_huge_pages_node[page_to_nid(page)]++;

2219
out_unlock:
2220
	spin_unlock_irq(&hugetlb_lock);
2221 2222 2223 2224

	return page;
}

2225
static struct page *alloc_migrate_huge_page(struct hstate *h, gfp_t gfp_mask,
2226
				     int nid, nodemask_t *nmask)
2227 2228 2229 2230 2231 2232
{
	struct page *page;

	if (hstate_is_gigantic(h))
		return NULL;

2233
	page = alloc_fresh_huge_page(h, gfp_mask, nid, nmask, NULL);
2234 2235 2236 2237 2238 2239 2240
	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
	 */
2241
	SetHPageTemporary(page);
2242 2243 2244 2245

	return page;
}

2246 2247 2248
/*
 * Use the VMA's mpolicy to allocate a huge page from the buddy.
 */
D
Dave Hansen 已提交
2249
static
2250
struct page *alloc_buddy_huge_page_with_mpol(struct hstate *h,
2251 2252
		struct vm_area_struct *vma, unsigned long addr)
{
2253
	struct page *page = NULL;
2254 2255 2256 2257 2258 2259
	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);
2260 2261 2262 2263 2264
	if (mpol_is_preferred_many(mpol)) {
		gfp_t gfp = gfp_mask | __GFP_NOWARN;

		gfp &=  ~(__GFP_DIRECT_RECLAIM | __GFP_NOFAIL);
		page = alloc_surplus_huge_page(h, gfp, nid, nodemask, false);
2265

2266 2267 2268 2269 2270 2271 2272
		/* Fallback to all nodes if page==NULL */
		nodemask = NULL;
	}

	if (!page)
		page = alloc_surplus_huge_page(h, gfp_mask, nid, nodemask, false);
	mpol_cond_put(mpol);
2273
	return page;
2274 2275
}

2276
/* page migration callback function */
2277
struct page *alloc_huge_page_nodemask(struct hstate *h, int preferred_nid,
2278
		nodemask_t *nmask, gfp_t gfp_mask)
2279
{
2280
	spin_lock_irq(&hugetlb_lock);
2281
	if (h->free_huge_pages - h->resv_huge_pages > 0) {
2282 2283 2284 2285
		struct page *page;

		page = dequeue_huge_page_nodemask(h, gfp_mask, preferred_nid, nmask);
		if (page) {
2286
			spin_unlock_irq(&hugetlb_lock);
2287
			return page;
2288 2289
		}
	}
2290
	spin_unlock_irq(&hugetlb_lock);
2291

2292
	return alloc_migrate_huge_page(h, gfp_mask, preferred_nid, nmask);
2293 2294
}

2295
/* mempolicy aware migration callback */
2296 2297
struct page *alloc_huge_page_vma(struct hstate *h, struct vm_area_struct *vma,
		unsigned long address)
2298 2299 2300 2301 2302 2303 2304 2305 2306
{
	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);
2307
	page = alloc_huge_page_nodemask(h, node, nodemask, gfp_mask);
2308 2309 2310 2311 2312
	mpol_cond_put(mpol);

	return page;
}

2313
/*
L
Lucas De Marchi 已提交
2314
 * Increase the hugetlb pool such that it can accommodate a reservation
2315 2316
 * of size 'delta'.
 */
2317
static int gather_surplus_pages(struct hstate *h, long delta)
2318
	__must_hold(&hugetlb_lock)
2319 2320 2321
{
	struct list_head surplus_list;
	struct page *page, *tmp;
2322 2323 2324
	int ret;
	long i;
	long needed, allocated;
2325
	bool alloc_ok = true;
2326

2327
	lockdep_assert_held(&hugetlb_lock);
2328
	needed = (h->resv_huge_pages + delta) - h->free_huge_pages;
2329
	if (needed <= 0) {
2330
		h->resv_huge_pages += delta;
2331
		return 0;
2332
	}
2333 2334 2335 2336 2337 2338

	allocated = 0;
	INIT_LIST_HEAD(&surplus_list);

	ret = -ENOMEM;
retry:
2339
	spin_unlock_irq(&hugetlb_lock);
2340
	for (i = 0; i < needed; i++) {
2341
		page = alloc_surplus_huge_page(h, htlb_alloc_mask(h),
2342
				NUMA_NO_NODE, NULL, true);
2343 2344 2345 2346
		if (!page) {
			alloc_ok = false;
			break;
		}
2347
		list_add(&page->lru, &surplus_list);
2348
		cond_resched();
2349
	}
2350
	allocated += i;
2351 2352 2353 2354 2355

	/*
	 * After retaking hugetlb_lock, we need to recalculate 'needed'
	 * because either resv_huge_pages or free_huge_pages may have changed.
	 */
2356
	spin_lock_irq(&hugetlb_lock);
2357 2358
	needed = (h->resv_huge_pages + delta) -
			(h->free_huge_pages + allocated);
2359 2360 2361 2362 2363 2364 2365 2366 2367 2368
	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;
	}
2369 2370
	/*
	 * The surplus_list now contains _at_least_ the number of extra pages
L
Lucas De Marchi 已提交
2371
	 * needed to accommodate the reservation.  Add the appropriate number
2372
	 * of pages to the hugetlb pool and free the extras back to the buddy
2373 2374 2375
	 * 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.
2376 2377
	 */
	needed += allocated;
2378
	h->resv_huge_pages += delta;
2379
	ret = 0;
2380

2381
	/* Free the needed pages to the hugetlb pool */
2382
	list_for_each_entry_safe(page, tmp, &surplus_list, lru) {
2383 2384
		if ((--needed) < 0)
			break;
2385
		/* Add the page to the hugetlb allocator */
2386
		enqueue_huge_page(h, page);
2387
	}
2388
free:
2389
	spin_unlock_irq(&hugetlb_lock);
2390

2391 2392 2393 2394
	/*
	 * Free unnecessary surplus pages to the buddy allocator.
	 * Pages have no ref count, call free_huge_page directly.
	 */
2395
	list_for_each_entry_safe(page, tmp, &surplus_list, lru)
2396
		free_huge_page(page);
2397
	spin_lock_irq(&hugetlb_lock);
2398 2399 2400 2401 2402

	return ret;
}

/*
2403 2404 2405 2406 2407 2408
 * 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.
2409
 */
2410 2411
static void return_unused_surplus_pages(struct hstate *h,
					unsigned long unused_resv_pages)
2412 2413
{
	unsigned long nr_pages;
2414 2415 2416
	struct page *page;
	LIST_HEAD(page_list);

2417
	lockdep_assert_held(&hugetlb_lock);
2418 2419
	/* Uncommit the reservation */
	h->resv_huge_pages -= unused_resv_pages;
2420

2421
	/* Cannot return gigantic pages currently */
2422
	if (hstate_is_gigantic(h))
2423
		goto out;
2424

2425 2426 2427 2428
	/*
	 * Part (or even all) of the reservation could have been backed
	 * by pre-allocated pages. Only free surplus pages.
	 */
2429
	nr_pages = min(unused_resv_pages, h->surplus_huge_pages);
2430

2431 2432
	/*
	 * We want to release as many surplus pages as possible, spread
2433 2434 2435
	 * 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.
2436
	 * remove_pool_huge_page() will balance the freed pages across the
2437
	 * on-line nodes with memory and will handle the hstate accounting.
2438 2439
	 */
	while (nr_pages--) {
2440 2441
		page = remove_pool_huge_page(h, &node_states[N_MEMORY], 1);
		if (!page)
2442
			goto out;
2443 2444

		list_add(&page->lru, &page_list);
2445
	}
2446 2447

out:
2448
	spin_unlock_irq(&hugetlb_lock);
2449
	update_and_free_pages_bulk(h, &page_list);
2450
	spin_lock_irq(&hugetlb_lock);
2451 2452
}

2453

2454
/*
2455
 * vma_needs_reservation, vma_commit_reservation and vma_end_reservation
2456
 * are used by the huge page allocation routines to manage reservations.
2457 2458 2459 2460 2461 2462
 *
 * 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
2463 2464 2465
 * 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.
2466 2467 2468 2469 2470 2471
 *
 * 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.
2472 2473 2474 2475 2476
 *
 * 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.
2477 2478 2479 2480 2481
 *
 * vma_del_reservation is used in error paths where an entry in the reserve
 * map was created during huge page allocation and must be removed.  It is to
 * be called after calling vma_needs_reservation to determine if a reservation
 * exists.
2482
 */
2483 2484 2485
enum vma_resv_mode {
	VMA_NEEDS_RESV,
	VMA_COMMIT_RESV,
2486
	VMA_END_RESV,
2487
	VMA_ADD_RESV,
2488
	VMA_DEL_RESV,
2489
};
2490 2491
static long __vma_reservation_common(struct hstate *h,
				struct vm_area_struct *vma, unsigned long addr,
2492
				enum vma_resv_mode mode)
2493
{
2494 2495
	struct resv_map *resv;
	pgoff_t idx;
2496
	long ret;
2497
	long dummy_out_regions_needed;
2498

2499 2500
	resv = vma_resv_map(vma);
	if (!resv)
2501
		return 1;
2502

2503
	idx = vma_hugecache_offset(h, vma, addr);
2504 2505
	switch (mode) {
	case VMA_NEEDS_RESV:
2506 2507 2508 2509 2510 2511
		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);
2512 2513
		break;
	case VMA_COMMIT_RESV:
2514
		ret = region_add(resv, idx, idx + 1, 1, NULL, NULL);
2515 2516
		/* region_add calls of range 1 should never fail. */
		VM_BUG_ON(ret < 0);
2517
		break;
2518
	case VMA_END_RESV:
2519
		region_abort(resv, idx, idx + 1, 1);
2520 2521
		ret = 0;
		break;
2522
	case VMA_ADD_RESV:
2523
		if (vma->vm_flags & VM_MAYSHARE) {
2524
			ret = region_add(resv, idx, idx + 1, 1, NULL, NULL);
2525 2526 2527 2528
			/* region_add calls of range 1 should never fail. */
			VM_BUG_ON(ret < 0);
		} else {
			region_abort(resv, idx, idx + 1, 1);
2529 2530 2531
			ret = region_del(resv, idx, idx + 1);
		}
		break;
2532 2533 2534 2535 2536 2537 2538 2539 2540 2541
	case VMA_DEL_RESV:
		if (vma->vm_flags & VM_MAYSHARE) {
			region_abort(resv, idx, idx + 1, 1);
			ret = region_del(resv, idx, idx + 1);
		} else {
			ret = region_add(resv, idx, idx + 1, 1, NULL, NULL);
			/* region_add calls of range 1 should never fail. */
			VM_BUG_ON(ret < 0);
		}
		break;
2542 2543 2544
	default:
		BUG();
	}
2545

2546
	if (vma->vm_flags & VM_MAYSHARE || mode == VMA_DEL_RESV)
2547
		return ret;
2548 2549 2550 2551 2552 2553 2554 2555 2556 2557 2558 2559 2560 2561 2562 2563 2564 2565 2566 2567
	/*
	 * We know private mapping must have HPAGE_RESV_OWNER set.
	 *
	 * In most cases, reserves always exist for private mappings.
	 * However, a file associated with mapping could have been
	 * hole punched or truncated after reserves were consumed.
	 * As subsequent fault on such a range will not use reserves.
	 * Subtle - The reserve map for private mappings has the
	 * opposite meaning than that of shared mappings.  If NO
	 * entry is in the reserve map, it means a reservation exists.
	 * If an entry exists in the reserve map, it means the
	 * reservation has already been consumed.  As a result, the
	 * return value of this routine is the opposite of the
	 * value returned from reserve map manipulation routines above.
	 */
	if (ret > 0)
		return 0;
	if (ret == 0)
		return 1;
	return ret;
2568
}
2569 2570

static long vma_needs_reservation(struct hstate *h,
2571
			struct vm_area_struct *vma, unsigned long addr)
2572
{
2573
	return __vma_reservation_common(h, vma, addr, VMA_NEEDS_RESV);
2574
}
2575

2576 2577 2578
static long vma_commit_reservation(struct hstate *h,
			struct vm_area_struct *vma, unsigned long addr)
{
2579 2580 2581
	return __vma_reservation_common(h, vma, addr, VMA_COMMIT_RESV);
}

2582
static void vma_end_reservation(struct hstate *h,
2583 2584
			struct vm_area_struct *vma, unsigned long addr)
{
2585
	(void)__vma_reservation_common(h, vma, addr, VMA_END_RESV);
2586 2587
}

2588 2589 2590 2591 2592 2593
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);
}

2594 2595 2596 2597 2598 2599
static long vma_del_reservation(struct hstate *h,
			struct vm_area_struct *vma, unsigned long addr)
{
	return __vma_reservation_common(h, vma, addr, VMA_DEL_RESV);
}

2600
/*
2601 2602 2603 2604 2605 2606 2607 2608 2609 2610 2611 2612 2613 2614 2615 2616 2617 2618
 * This routine is called to restore reservation information on error paths.
 * It should ONLY be called for pages allocated via alloc_huge_page(), and
 * the hugetlb mutex should remain held when calling this routine.
 *
 * It handles two specific cases:
 * 1) A reservation was in place and the page consumed the reservation.
 *    HPageRestoreReserve is set in the page.
 * 2) No reservation was in place for the page, so HPageRestoreReserve is
 *    not set.  However, alloc_huge_page always updates the reserve map.
 *
 * In case 1, free_huge_page later in the error path will increment the
 * global reserve count.  But, free_huge_page does not have enough context
 * to adjust the reservation map.  This case deals primarily with private
 * mappings.  Adjust the reserve map here to be consistent with global
 * reserve count adjustments to be made by free_huge_page.  Make sure the
 * reserve map indicates there is a reservation present.
 *
 * In case 2, simply undo reserve map modifications done by alloc_huge_page.
2619
 */
2620 2621
void restore_reserve_on_error(struct hstate *h, struct vm_area_struct *vma,
			unsigned long address, struct page *page)
2622
{
2623
	long rc = vma_needs_reservation(h, vma, address);
2624

2625 2626
	if (HPageRestoreReserve(page)) {
		if (unlikely(rc < 0))
2627 2628
			/*
			 * Rare out of memory condition in reserve map
2629
			 * manipulation.  Clear HPageRestoreReserve so that
2630 2631 2632 2633 2634 2635 2636 2637
			 * 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.
			 */
2638
			ClearHPageRestoreReserve(page);
2639 2640 2641 2642 2643 2644 2645 2646
		else if (rc)
			(void)vma_add_reservation(h, vma, address);
		else
			vma_end_reservation(h, vma, address);
	} else {
		if (!rc) {
			/*
			 * This indicates there is an entry in the reserve map
2647
			 * not added by alloc_huge_page.  We know it was added
2648 2649 2650 2651 2652 2653 2654
			 * before the alloc_huge_page call, otherwise
			 * HPageRestoreReserve would be set on the page.
			 * Remove the entry so that a subsequent allocation
			 * does not consume a reservation.
			 */
			rc = vma_del_reservation(h, vma, address);
			if (rc < 0)
2655
				/*
2656 2657 2658 2659 2660 2661
				 * VERY rare out of memory condition.  Since
				 * we can not delete the entry, set
				 * HPageRestoreReserve so that the reserve
				 * count will be incremented when the page
				 * is freed.  This reserve will be consumed
				 * on a subsequent allocation.
2662
				 */
2663 2664 2665 2666 2667 2668 2669 2670 2671 2672 2673 2674 2675 2676 2677 2678 2679 2680 2681 2682 2683 2684
				SetHPageRestoreReserve(page);
		} else if (rc < 0) {
			/*
			 * Rare out of memory condition from
			 * vma_needs_reservation call.  Memory allocation is
			 * only attempted if a new entry is needed.  Therefore,
			 * this implies there is not an entry in the
			 * reserve map.
			 *
			 * For shared mappings, no entry in the map indicates
			 * no reservation.  We are done.
			 */
			if (!(vma->vm_flags & VM_MAYSHARE))
				/*
				 * For private mappings, no entry indicates
				 * a reservation is present.  Since we can
				 * not add an entry, set SetHPageRestoreReserve
				 * on the page so reserve count will be
				 * incremented when freed.  This reserve will
				 * be consumed on a subsequent allocation.
				 */
				SetHPageRestoreReserve(page);
2685
		} else
2686 2687 2688 2689
			/*
			 * No reservation present, do nothing
			 */
			 vma_end_reservation(h, vma, address);
2690 2691 2692
	}
}

2693 2694 2695 2696
/*
 * alloc_and_dissolve_huge_page - Allocate a new page and dissolve the old one
 * @h: struct hstate old page belongs to
 * @old_page: Old page to dissolve
2697
 * @list: List to isolate the page in case we need to
2698 2699
 * Returns 0 on success, otherwise negated error.
 */
2700 2701
static int alloc_and_dissolve_huge_page(struct hstate *h, struct page *old_page,
					struct list_head *list)
2702 2703 2704
{
	gfp_t gfp_mask = htlb_alloc_mask(h) | __GFP_THISNODE;
	int nid = page_to_nid(old_page);
2705
	bool alloc_retry = false;
2706 2707 2708 2709 2710
	struct page *new_page;
	int ret = 0;

	/*
	 * Before dissolving the page, we need to allocate a new one for the
2711 2712 2713 2714
	 * pool to remain stable.  Here, we allocate the page and 'prep' it
	 * by doing everything but actually updating counters and adding to
	 * the pool.  This simplifies and let us do most of the processing
	 * under the lock.
2715
	 */
2716
alloc_retry:
2717 2718 2719
	new_page = alloc_buddy_huge_page(h, gfp_mask, nid, NULL, NULL);
	if (!new_page)
		return -ENOMEM;
2720 2721 2722 2723 2724 2725 2726 2727 2728 2729 2730 2731 2732 2733 2734 2735 2736 2737 2738 2739
	/*
	 * If all goes well, this page will be directly added to the free
	 * list in the pool.  For this the ref count needs to be zero.
	 * Attempt to drop now, and retry once if needed.  It is VERY
	 * unlikely there is another ref on the page.
	 *
	 * If someone else has a reference to the page, it will be freed
	 * when they drop their ref.  Abuse temporary page flag to accomplish
	 * this.  Retry once if there is an inflated ref count.
	 */
	SetHPageTemporary(new_page);
	if (!put_page_testzero(new_page)) {
		if (alloc_retry)
			return -EBUSY;

		alloc_retry = true;
		goto alloc_retry;
	}
	ClearHPageTemporary(new_page);

2740
	__prep_new_huge_page(h, new_page);
2741 2742 2743 2744 2745 2746 2747 2748 2749 2750

retry:
	spin_lock_irq(&hugetlb_lock);
	if (!PageHuge(old_page)) {
		/*
		 * Freed from under us. Drop new_page too.
		 */
		goto free_new;
	} else if (page_count(old_page)) {
		/*
2751 2752
		 * Someone has grabbed the page, try to isolate it here.
		 * Fail with -EBUSY if not possible.
2753
		 */
2754 2755 2756 2757
		spin_unlock_irq(&hugetlb_lock);
		if (!isolate_huge_page(old_page, list))
			ret = -EBUSY;
		spin_lock_irq(&hugetlb_lock);
2758 2759 2760 2761 2762 2763 2764 2765 2766 2767 2768 2769 2770 2771 2772 2773 2774 2775 2776 2777 2778
		goto free_new;
	} else if (!HPageFreed(old_page)) {
		/*
		 * Page's refcount is 0 but it has not been enqueued in the
		 * freelist yet. Race window is small, so we can succeed here if
		 * we retry.
		 */
		spin_unlock_irq(&hugetlb_lock);
		cond_resched();
		goto retry;
	} else {
		/*
		 * Ok, old_page is still a genuine free hugepage. Remove it from
		 * the freelist and decrease the counters. These will be
		 * incremented again when calling __prep_account_new_huge_page()
		 * and enqueue_huge_page() for new_page. The counters will remain
		 * stable since this happens under the lock.
		 */
		remove_hugetlb_page(h, old_page, false);

		/*
2779 2780
		 * Ref count on new page is already zero as it was dropped
		 * earlier.  It can be directly added to the pool free list.
2781 2782 2783 2784 2785 2786 2787 2788
		 */
		__prep_account_new_huge_page(h, nid);
		enqueue_huge_page(h, new_page);

		/*
		 * Pages have been replaced, we can safely free the old one.
		 */
		spin_unlock_irq(&hugetlb_lock);
2789
		update_and_free_page(h, old_page, false);
2790 2791 2792 2793 2794 2795
	}

	return ret;

free_new:
	spin_unlock_irq(&hugetlb_lock);
2796 2797
	/* Page has a zero ref count, but needs a ref to be freed */
	set_page_refcounted(new_page);
2798
	update_and_free_page(h, new_page, false);
2799 2800 2801 2802

	return ret;
}

2803
int isolate_or_dissolve_huge_page(struct page *page, struct list_head *list)
2804 2805 2806
{
	struct hstate *h;
	struct page *head;
2807
	int ret = -EBUSY;
2808 2809 2810 2811 2812 2813 2814 2815 2816 2817 2818 2819 2820 2821 2822 2823 2824 2825 2826 2827 2828 2829 2830 2831

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

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

2832 2833 2834 2835 2836 2837
	if (page_count(head) && isolate_huge_page(head, list))
		ret = 0;
	else if (!page_count(head))
		ret = alloc_and_dissolve_huge_page(h, head, list);

	return ret;
2838 2839
}

2840
struct page *alloc_huge_page(struct vm_area_struct *vma,
2841
				    unsigned long addr, int avoid_reserve)
L
Linus Torvalds 已提交
2842
{
2843
	struct hugepage_subpool *spool = subpool_vma(vma);
2844
	struct hstate *h = hstate_vma(vma);
2845
	struct page *page;
2846 2847
	long map_chg, map_commit;
	long gbl_chg;
2848 2849
	int ret, idx;
	struct hugetlb_cgroup *h_cg;
2850
	bool deferred_reserve;
2851

2852
	idx = hstate_index(h);
2853
	/*
2854 2855 2856
	 * 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).
2857
	 */
2858 2859
	map_chg = gbl_chg = vma_needs_reservation(h, vma, addr);
	if (map_chg < 0)
2860
		return ERR_PTR(-ENOMEM);
2861 2862 2863 2864 2865 2866 2867 2868 2869 2870 2871

	/*
	 * 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) {
2872
			vma_end_reservation(h, vma, addr);
2873
			return ERR_PTR(-ENOSPC);
2874
		}
L
Linus Torvalds 已提交
2875

2876 2877 2878 2879 2880 2881 2882 2883 2884 2885 2886 2887
		/*
		 * 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;
	}

2888 2889
	/* If this allocation is not consuming a reservation, charge it now.
	 */
2890
	deferred_reserve = map_chg || avoid_reserve;
2891 2892 2893 2894 2895 2896 2897
	if (deferred_reserve) {
		ret = hugetlb_cgroup_charge_cgroup_rsvd(
			idx, pages_per_huge_page(h), &h_cg);
		if (ret)
			goto out_subpool_put;
	}

2898
	ret = hugetlb_cgroup_charge_cgroup(idx, pages_per_huge_page(h), &h_cg);
2899
	if (ret)
2900
		goto out_uncharge_cgroup_reservation;
2901

2902
	spin_lock_irq(&hugetlb_lock);
2903 2904 2905 2906 2907 2908
	/*
	 * 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);
2909
	if (!page) {
2910
		spin_unlock_irq(&hugetlb_lock);
2911
		page = alloc_buddy_huge_page_with_mpol(h, vma, addr);
2912 2913
		if (!page)
			goto out_uncharge_cgroup;
2914
		if (!avoid_reserve && vma_has_reserves(vma, gbl_chg)) {
2915
			SetHPageRestoreReserve(page);
2916 2917
			h->resv_huge_pages--;
		}
2918
		spin_lock_irq(&hugetlb_lock);
2919
		list_add(&page->lru, &h->hugepage_activelist);
2920
		/* Fall through */
K
Ken Chen 已提交
2921
	}
2922
	hugetlb_cgroup_commit_charge(idx, pages_per_huge_page(h), h_cg, page);
2923 2924 2925 2926 2927 2928 2929 2930
	/* 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);
	}

2931
	spin_unlock_irq(&hugetlb_lock);
2932

2933
	hugetlb_set_page_subpool(page, spool);
2934

2935 2936
	map_commit = vma_commit_reservation(h, vma, addr);
	if (unlikely(map_chg > map_commit)) {
2937 2938 2939 2940 2941 2942 2943 2944 2945 2946 2947 2948 2949
		/*
		 * 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);
2950 2951 2952
		if (deferred_reserve)
			hugetlb_cgroup_uncharge_page_rsvd(hstate_index(h),
					pages_per_huge_page(h), page);
2953
	}
2954
	return page;
2955 2956 2957

out_uncharge_cgroup:
	hugetlb_cgroup_uncharge_cgroup(idx, pages_per_huge_page(h), h_cg);
2958 2959 2960 2961
out_uncharge_cgroup_reservation:
	if (deferred_reserve)
		hugetlb_cgroup_uncharge_cgroup_rsvd(idx, pages_per_huge_page(h),
						    h_cg);
2962
out_subpool_put:
2963
	if (map_chg || avoid_reserve)
2964
		hugepage_subpool_put_pages(spool, 1);
2965
	vma_end_reservation(h, vma, addr);
2966
	return ERR_PTR(-ENOSPC);
2967 2968
}

2969
int alloc_bootmem_huge_page(struct hstate *h, int nid)
2970
	__attribute__ ((weak, alias("__alloc_bootmem_huge_page")));
2971
int __alloc_bootmem_huge_page(struct hstate *h, int nid)
2972
{
2973
	struct huge_bootmem_page *m = NULL; /* initialize for clang */
2974
	int nr_nodes, node;
2975

2976
	if (nid != NUMA_NO_NODE && nid >= nr_online_nodes)
2977 2978 2979 2980 2981 2982 2983 2984 2985 2986
		return 0;
	/* do node specific alloc */
	if (nid != NUMA_NO_NODE) {
		m = memblock_alloc_try_nid_raw(huge_page_size(h), huge_page_size(h),
				0, MEMBLOCK_ALLOC_ACCESSIBLE, nid);
		if (!m)
			return 0;
		goto found;
	}
	/* allocate from next node when distributing huge pages */
2987
	for_each_node_mask_to_alloc(h, nr_nodes, node, &node_states[N_MEMORY]) {
2988
		m = memblock_alloc_try_nid_raw(
2989
				huge_page_size(h), huge_page_size(h),
2990
				0, MEMBLOCK_ALLOC_ACCESSIBLE, node);
2991 2992 2993 2994 2995 2996 2997 2998
		/*
		 * Use the beginning of the huge page to store the
		 * huge_bootmem_page struct (until gather_bootmem
		 * puts them into the mem_map).
		 */
		if (!m)
			return 0;
		goto found;
2999 3000 3001 3002
	}

found:
	/* Put them into a private list first because mem_map is not up yet */
3003
	INIT_LIST_HEAD(&m->list);
3004 3005 3006 3007 3008
	list_add(&m->list, &huge_boot_pages);
	m->hstate = h;
	return 1;
}

3009 3010 3011 3012
/*
 * Put bootmem huge pages into the standard lists after mem_map is up.
 * Note: This only applies to gigantic (order > MAX_ORDER) pages.
 */
3013 3014 3015 3016 3017
static void __init gather_bootmem_prealloc(void)
{
	struct huge_bootmem_page *m;

	list_for_each_entry(m, &huge_boot_pages, list) {
3018
		struct page *page = virt_to_page(m);
3019
		struct hstate *h = m->hstate;
3020

3021
		VM_BUG_ON(!hstate_is_gigantic(h));
3022
		WARN_ON(page_count(page) != 1);
3023 3024 3025 3026 3027
		if (prep_compound_gigantic_page(page, huge_page_order(h))) {
			WARN_ON(PageReserved(page));
			prep_new_huge_page(h, page, page_to_nid(page));
			put_page(page); /* add to the hugepage allocator */
		} else {
3028
			/* VERY unlikely inflated ref count on a tail page */
3029 3030
			free_gigantic_page(page, huge_page_order(h));
		}
3031

3032
		/*
3033 3034 3035
		 * We need to restore the 'stolen' pages to totalram_pages
		 * in order to fix confusing memory reports from free(1) and
		 * other side-effects, like CommitLimit going negative.
3036
		 */
3037
		adjust_managed_page_count(page, pages_per_huge_page(h));
3038
		cond_resched();
3039 3040
	}
}
3041 3042 3043 3044 3045 3046 3047 3048 3049 3050 3051 3052 3053 3054 3055 3056 3057 3058 3059 3060 3061 3062 3063 3064 3065 3066 3067 3068 3069 3070
static void __init hugetlb_hstate_alloc_pages_onenode(struct hstate *h, int nid)
{
	unsigned long i;
	char buf[32];

	for (i = 0; i < h->max_huge_pages_node[nid]; ++i) {
		if (hstate_is_gigantic(h)) {
			if (!alloc_bootmem_huge_page(h, nid))
				break;
		} else {
			struct page *page;
			gfp_t gfp_mask = htlb_alloc_mask(h) | __GFP_THISNODE;

			page = alloc_fresh_huge_page(h, gfp_mask, nid,
					&node_states[N_MEMORY], NULL);
			if (!page)
				break;
			put_page(page); /* free it into the hugepage allocator */
		}
		cond_resched();
	}
	if (i == h->max_huge_pages_node[nid])
		return;

	string_get_size(huge_page_size(h), 1, STRING_UNITS_2, buf, 32);
	pr_warn("HugeTLB: allocating %u of page size %s failed node%d.  Only allocated %lu hugepages.\n",
		h->max_huge_pages_node[nid], buf, nid, i);
	h->max_huge_pages -= (h->max_huge_pages_node[nid] - i);
	h->max_huge_pages_node[nid] = i;
}
3071

3072
static void __init hugetlb_hstate_alloc_pages(struct hstate *h)
L
Linus Torvalds 已提交
3073 3074
{
	unsigned long i;
3075
	nodemask_t *node_alloc_noretry;
3076 3077 3078 3079 3080 3081 3082 3083 3084 3085 3086 3087 3088 3089 3090
	bool node_specific_alloc = false;

	/* skip gigantic hugepages allocation if hugetlb_cma enabled */
	if (hstate_is_gigantic(h) && hugetlb_cma_size) {
		pr_warn_once("HugeTLB: hugetlb_cma is enabled, skip boot time allocation\n");
		return;
	}

	/* do node specific alloc */
	for (i = 0; i < nr_online_nodes; i++) {
		if (h->max_huge_pages_node[i] > 0) {
			hugetlb_hstate_alloc_pages_onenode(h, i);
			node_specific_alloc = true;
		}
	}
3091

3092 3093 3094 3095
	if (node_specific_alloc)
		return;

	/* below will do all node balanced alloc */
3096 3097 3098 3099 3100 3101 3102 3103 3104 3105 3106 3107 3108 3109 3110 3111 3112
	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);
3113

3114
	for (i = 0; i < h->max_huge_pages; ++i) {
3115
		if (hstate_is_gigantic(h)) {
3116
			if (!alloc_bootmem_huge_page(h, NUMA_NO_NODE))
3117
				break;
3118
		} else if (!alloc_pool_huge_page(h,
3119 3120
					 &node_states[N_MEMORY],
					 node_alloc_noretry))
L
Linus Torvalds 已提交
3121
			break;
3122
		cond_resched();
L
Linus Torvalds 已提交
3123
	}
3124 3125 3126
	if (i < h->max_huge_pages) {
		char buf[32];

3127
		string_get_size(huge_page_size(h), 1, STRING_UNITS_2, buf, 32);
3128 3129 3130 3131
		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;
	}
3132
	kfree(node_alloc_noretry);
3133 3134 3135 3136
}

static void __init hugetlb_init_hstates(void)
{
3137
	struct hstate *h, *h2;
3138 3139

	for_each_hstate(h) {
3140 3141 3142
		if (minimum_order > huge_page_order(h))
			minimum_order = huge_page_order(h);

3143
		/* oversize hugepages were init'ed in early boot */
3144
		if (!hstate_is_gigantic(h))
3145
			hugetlb_hstate_alloc_pages(h);
3146 3147 3148 3149 3150 3151

		/*
		 * Set demote order for each hstate.  Note that
		 * h->demote_order is initially 0.
		 * - We can not demote gigantic pages if runtime freeing
		 *   is not supported, so skip this.
3152 3153
		 * - If CMA allocation is possible, we can not demote
		 *   HUGETLB_PAGE_ORDER or smaller size pages.
3154 3155 3156
		 */
		if (hstate_is_gigantic(h) && !gigantic_page_runtime_supported())
			continue;
3157 3158
		if (hugetlb_cma_size && h->order <= HUGETLB_PAGE_ORDER)
			continue;
3159 3160 3161 3162 3163 3164 3165
		for_each_hstate(h2) {
			if (h2 == h)
				continue;
			if (h2->order < h->order &&
			    h2->order > h->demote_order)
				h->demote_order = h2->order;
		}
3166
	}
3167
	VM_BUG_ON(minimum_order == UINT_MAX);
3168 3169 3170 3171 3172 3173 3174
}

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

	for_each_hstate(h) {
A
Andi Kleen 已提交
3175
		char buf[32];
3176 3177

		string_get_size(huge_page_size(h), 1, STRING_UNITS_2, buf, 32);
3178
		pr_info("HugeTLB registered %s page size, pre-allocated %ld pages\n",
3179
			buf, h->free_huge_pages);
3180 3181 3182
	}
}

L
Linus Torvalds 已提交
3183
#ifdef CONFIG_HIGHMEM
3184 3185
static void try_to_free_low(struct hstate *h, unsigned long count,
						nodemask_t *nodes_allowed)
L
Linus Torvalds 已提交
3186
{
3187
	int i;
3188
	LIST_HEAD(page_list);
3189

3190
	lockdep_assert_held(&hugetlb_lock);
3191
	if (hstate_is_gigantic(h))
3192 3193
		return;

3194 3195 3196
	/*
	 * Collect pages to be freed on a list, and free after dropping lock
	 */
3197
	for_each_node_mask(i, *nodes_allowed) {
3198
		struct page *page, *next;
3199 3200 3201
		struct list_head *freel = &h->hugepage_freelists[i];
		list_for_each_entry_safe(page, next, freel, lru) {
			if (count >= h->nr_huge_pages)
3202
				goto out;
L
Linus Torvalds 已提交
3203 3204
			if (PageHighMem(page))
				continue;
3205
			remove_hugetlb_page(h, page, false);
3206
			list_add(&page->lru, &page_list);
L
Linus Torvalds 已提交
3207 3208
		}
	}
3209 3210

out:
3211
	spin_unlock_irq(&hugetlb_lock);
3212
	update_and_free_pages_bulk(h, &page_list);
3213
	spin_lock_irq(&hugetlb_lock);
L
Linus Torvalds 已提交
3214 3215
}
#else
3216 3217
static inline void try_to_free_low(struct hstate *h, unsigned long count,
						nodemask_t *nodes_allowed)
L
Linus Torvalds 已提交
3218 3219 3220 3221
{
}
#endif

3222 3223 3224 3225 3226
/*
 * 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.
 */
3227 3228
static int adjust_pool_surplus(struct hstate *h, nodemask_t *nodes_allowed,
				int delta)
3229
{
3230
	int nr_nodes, node;
3231

3232
	lockdep_assert_held(&hugetlb_lock);
3233 3234
	VM_BUG_ON(delta != -1 && delta != 1);

3235 3236 3237 3238
	if (delta < 0) {
		for_each_node_mask_to_alloc(h, nr_nodes, node, nodes_allowed) {
			if (h->surplus_huge_pages_node[node])
				goto found;
3239
		}
3240 3241 3242 3243 3244
	} 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;
3245
		}
3246 3247
	}
	return 0;
3248

3249 3250 3251 3252
found:
	h->surplus_huge_pages += delta;
	h->surplus_huge_pages_node[node] += delta;
	return 1;
3253 3254
}

3255
#define persistent_huge_pages(h) (h->nr_huge_pages - h->surplus_huge_pages)
3256
static int set_max_huge_pages(struct hstate *h, unsigned long count, int nid,
3257
			      nodemask_t *nodes_allowed)
L
Linus Torvalds 已提交
3258
{
3259
	unsigned long min_count, ret;
3260 3261
	struct page *page;
	LIST_HEAD(page_list);
3262 3263 3264 3265 3266 3267 3268 3269 3270 3271 3272
	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
3273

3274 3275 3276 3277 3278
	/*
	 * resize_lock mutex prevents concurrent adjustments to number of
	 * pages in hstate via the proc/sysfs interfaces.
	 */
	mutex_lock(&h->resize_lock);
3279
	flush_free_hpage_work(h);
3280
	spin_lock_irq(&hugetlb_lock);
3281

3282 3283 3284 3285 3286 3287 3288 3289 3290 3291 3292 3293 3294 3295 3296 3297 3298 3299 3300 3301
	/*
	 * 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;
	}

3302 3303 3304 3305 3306 3307 3308 3309 3310
	/*
	 * 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)) {
3311
			spin_unlock_irq(&hugetlb_lock);
3312
			mutex_unlock(&h->resize_lock);
3313
			NODEMASK_FREE(node_alloc_noretry);
3314 3315 3316 3317
			return -EINVAL;
		}
		/* Fall through to decrease pool */
	}
3318

3319 3320 3321 3322
	/*
	 * Increase the pool size
	 * First take pages out of surplus state.  Then make up the
	 * remaining difference by allocating fresh huge pages.
3323
	 *
3324
	 * We might race with alloc_surplus_huge_page() here and be unable
3325 3326 3327 3328
	 * 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.
3329
	 */
3330
	while (h->surplus_huge_pages && count > persistent_huge_pages(h)) {
3331
		if (!adjust_pool_surplus(h, nodes_allowed, -1))
3332 3333 3334
			break;
	}

3335
	while (count > persistent_huge_pages(h)) {
3336 3337 3338 3339 3340
		/*
		 * 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.
		 */
3341
		spin_unlock_irq(&hugetlb_lock);
3342 3343 3344 3345

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

3346 3347
		ret = alloc_pool_huge_page(h, nodes_allowed,
						node_alloc_noretry);
3348
		spin_lock_irq(&hugetlb_lock);
3349 3350 3351
		if (!ret)
			goto out;

3352 3353 3354
		/* Bail for signals. Probably ctrl-c from user */
		if (signal_pending(current))
			goto out;
3355 3356 3357 3358 3359 3360 3361 3362
	}

	/*
	 * 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.
3363 3364 3365 3366
	 *
	 * 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
3367
	 * alloc_surplus_huge_page() is checking the global counter,
3368 3369 3370
	 * 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.
3371
	 */
3372
	min_count = h->resv_huge_pages + h->nr_huge_pages - h->free_huge_pages;
3373
	min_count = max(count, min_count);
3374
	try_to_free_low(h, min_count, nodes_allowed);
3375 3376 3377 3378

	/*
	 * Collect pages to be removed on list without dropping lock
	 */
3379
	while (min_count < persistent_huge_pages(h)) {
3380 3381
		page = remove_pool_huge_page(h, nodes_allowed, 0);
		if (!page)
L
Linus Torvalds 已提交
3382
			break;
3383 3384

		list_add(&page->lru, &page_list);
L
Linus Torvalds 已提交
3385
	}
3386
	/* free the pages after dropping lock */
3387
	spin_unlock_irq(&hugetlb_lock);
3388
	update_and_free_pages_bulk(h, &page_list);
3389
	flush_free_hpage_work(h);
3390
	spin_lock_irq(&hugetlb_lock);
3391

3392
	while (count < persistent_huge_pages(h)) {
3393
		if (!adjust_pool_surplus(h, nodes_allowed, 1))
3394 3395 3396
			break;
	}
out:
3397
	h->max_huge_pages = persistent_huge_pages(h);
3398
	spin_unlock_irq(&hugetlb_lock);
3399
	mutex_unlock(&h->resize_lock);
3400

3401 3402
	NODEMASK_FREE(node_alloc_noretry);

3403
	return 0;
L
Linus Torvalds 已提交
3404 3405
}

3406 3407 3408 3409 3410 3411 3412 3413 3414 3415 3416 3417 3418 3419 3420 3421 3422 3423 3424 3425 3426 3427 3428 3429 3430 3431 3432 3433 3434 3435 3436 3437 3438 3439 3440 3441 3442 3443 3444 3445 3446 3447 3448 3449 3450 3451 3452 3453 3454 3455 3456 3457 3458 3459 3460 3461 3462 3463 3464 3465 3466
static int demote_free_huge_page(struct hstate *h, struct page *page)
{
	int i, nid = page_to_nid(page);
	struct hstate *target_hstate;
	int rc = 0;

	target_hstate = size_to_hstate(PAGE_SIZE << h->demote_order);

	remove_hugetlb_page_for_demote(h, page, false);
	spin_unlock_irq(&hugetlb_lock);

	rc = alloc_huge_page_vmemmap(h, page);
	if (rc) {
		/* Allocation of vmemmmap failed, we can not demote page */
		spin_lock_irq(&hugetlb_lock);
		set_page_refcounted(page);
		add_hugetlb_page(h, page, false);
		return rc;
	}

	/*
	 * Use destroy_compound_hugetlb_page_for_demote for all huge page
	 * sizes as it will not ref count pages.
	 */
	destroy_compound_hugetlb_page_for_demote(page, huge_page_order(h));

	/*
	 * Taking target hstate mutex synchronizes with set_max_huge_pages.
	 * Without the mutex, pages added to target hstate could be marked
	 * as surplus.
	 *
	 * Note that we already hold h->resize_lock.  To prevent deadlock,
	 * use the convention of always taking larger size hstate mutex first.
	 */
	mutex_lock(&target_hstate->resize_lock);
	for (i = 0; i < pages_per_huge_page(h);
				i += pages_per_huge_page(target_hstate)) {
		if (hstate_is_gigantic(target_hstate))
			prep_compound_gigantic_page_for_demote(page + i,
							target_hstate->order);
		else
			prep_compound_page(page + i, target_hstate->order);
		set_page_private(page + i, 0);
		set_page_refcounted(page + i);
		prep_new_huge_page(target_hstate, page + i, nid);
		put_page(page + i);
	}
	mutex_unlock(&target_hstate->resize_lock);

	spin_lock_irq(&hugetlb_lock);

	/*
	 * Not absolutely necessary, but for consistency update max_huge_pages
	 * based on pool changes for the demoted page.
	 */
	h->max_huge_pages--;
	target_hstate->max_huge_pages += pages_per_huge_page(h);

	return rc;
}

3467 3468 3469
static int demote_pool_huge_page(struct hstate *h, nodemask_t *nodes_allowed)
	__must_hold(&hugetlb_lock)
{
3470 3471
	int nr_nodes, node;
	struct page *page;
3472 3473 3474 3475 3476 3477 3478 3479 3480 3481
	int rc = 0;

	lockdep_assert_held(&hugetlb_lock);

	/* We should never get here if no demote order */
	if (!h->demote_order) {
		pr_warn("HugeTLB: NULL demote order passed to demote_pool_huge_page.\n");
		return -EINVAL;		/* internal error */
	}

3482 3483 3484 3485 3486 3487 3488 3489 3490
	for_each_node_mask_to_free(h, nr_nodes, node, nodes_allowed) {
		if (!list_empty(&h->hugepage_freelists[node])) {
			page = list_entry(h->hugepage_freelists[node].next,
					struct page, lru);
			rc = demote_free_huge_page(h, page);
			break;
		}
	}

3491 3492 3493
	return rc;
}

3494 3495 3496
#define HSTATE_ATTR_RO(_name) \
	static struct kobj_attribute _name##_attr = __ATTR_RO(_name)

3497 3498 3499
#define HSTATE_ATTR_WO(_name) \
	static struct kobj_attribute _name##_attr = __ATTR_WO(_name)

3500 3501 3502 3503 3504 3505 3506
#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];

3507 3508 3509
static struct hstate *kobj_to_node_hstate(struct kobject *kobj, int *nidp);

static struct hstate *kobj_to_hstate(struct kobject *kobj, int *nidp)
3510 3511
{
	int i;
3512

3513
	for (i = 0; i < HUGE_MAX_HSTATE; i++)
3514 3515 3516
		if (hstate_kobjs[i] == kobj) {
			if (nidp)
				*nidp = NUMA_NO_NODE;
3517
			return &hstates[i];
3518 3519 3520
		}

	return kobj_to_node_hstate(kobj, nidp);
3521 3522
}

3523
static ssize_t nr_hugepages_show_common(struct kobject *kobj,
3524 3525
					struct kobj_attribute *attr, char *buf)
{
3526 3527 3528 3529 3530 3531 3532 3533 3534 3535
	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];

3536
	return sysfs_emit(buf, "%lu\n", nr_huge_pages);
3537
}
3538

3539 3540 3541
static ssize_t __nr_hugepages_store_common(bool obey_mempolicy,
					   struct hstate *h, int nid,
					   unsigned long count, size_t len)
3542 3543
{
	int err;
3544
	nodemask_t nodes_allowed, *n_mask;
3545

3546 3547
	if (hstate_is_gigantic(h) && !gigantic_page_runtime_supported())
		return -EINVAL;
3548

3549 3550 3551 3552 3553
	if (nid == NUMA_NO_NODE) {
		/*
		 * global hstate attribute
		 */
		if (!(obey_mempolicy &&
3554 3555 3556 3557 3558
				init_nodemask_of_mempolicy(&nodes_allowed)))
			n_mask = &node_states[N_MEMORY];
		else
			n_mask = &nodes_allowed;
	} else {
3559
		/*
3560 3561
		 * Node specific request.  count adjustment happens in
		 * set_max_huge_pages() after acquiring hugetlb_lock.
3562
		 */
3563 3564
		init_nodemask_of_node(&nodes_allowed, nid);
		n_mask = &nodes_allowed;
3565
	}
3566

3567
	err = set_max_huge_pages(h, count, nid, n_mask);
3568

3569
	return err ? err : len;
3570 3571
}

3572 3573 3574 3575 3576 3577 3578 3579 3580 3581 3582 3583 3584 3585 3586 3587 3588
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);
}

3589 3590 3591 3592 3593 3594 3595 3596 3597
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)
{
3598
	return nr_hugepages_store_common(false, kobj, buf, len);
3599 3600 3601
}
HSTATE_ATTR(nr_hugepages);

3602 3603 3604 3605 3606 3607 3608
#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,
3609 3610
					   struct kobj_attribute *attr,
					   char *buf)
3611 3612 3613 3614 3615 3616 3617
{
	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)
{
3618
	return nr_hugepages_store_common(true, kobj, buf, len);
3619 3620 3621 3622 3623
}
HSTATE_ATTR(nr_hugepages_mempolicy);
#endif


3624 3625 3626
static ssize_t nr_overcommit_hugepages_show(struct kobject *kobj,
					struct kobj_attribute *attr, char *buf)
{
3627
	struct hstate *h = kobj_to_hstate(kobj, NULL);
3628
	return sysfs_emit(buf, "%lu\n", h->nr_overcommit_huge_pages);
3629
}
3630

3631 3632 3633 3634 3635
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;
3636
	struct hstate *h = kobj_to_hstate(kobj, NULL);
3637

3638
	if (hstate_is_gigantic(h))
3639 3640
		return -EINVAL;

3641
	err = kstrtoul(buf, 10, &input);
3642
	if (err)
3643
		return err;
3644

3645
	spin_lock_irq(&hugetlb_lock);
3646
	h->nr_overcommit_huge_pages = input;
3647
	spin_unlock_irq(&hugetlb_lock);
3648 3649 3650 3651 3652 3653 3654 3655

	return count;
}
HSTATE_ATTR(nr_overcommit_hugepages);

static ssize_t free_hugepages_show(struct kobject *kobj,
					struct kobj_attribute *attr, char *buf)
{
3656 3657 3658 3659 3660 3661 3662 3663 3664 3665
	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];

3666
	return sysfs_emit(buf, "%lu\n", free_huge_pages);
3667 3668 3669 3670 3671 3672
}
HSTATE_ATTR_RO(free_hugepages);

static ssize_t resv_hugepages_show(struct kobject *kobj,
					struct kobj_attribute *attr, char *buf)
{
3673
	struct hstate *h = kobj_to_hstate(kobj, NULL);
3674
	return sysfs_emit(buf, "%lu\n", h->resv_huge_pages);
3675 3676 3677 3678 3679 3680
}
HSTATE_ATTR_RO(resv_hugepages);

static ssize_t surplus_hugepages_show(struct kobject *kobj,
					struct kobj_attribute *attr, char *buf)
{
3681 3682 3683 3684 3685 3686 3687 3688 3689 3690
	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];

3691
	return sysfs_emit(buf, "%lu\n", surplus_huge_pages);
3692 3693 3694
}
HSTATE_ATTR_RO(surplus_hugepages);

3695 3696 3697 3698 3699 3700 3701 3702 3703 3704 3705 3706 3707 3708 3709 3710 3711 3712 3713 3714 3715 3716 3717 3718 3719 3720 3721 3722 3723 3724 3725 3726 3727 3728 3729 3730 3731 3732 3733 3734 3735 3736 3737 3738 3739 3740 3741 3742 3743 3744 3745 3746 3747 3748 3749 3750 3751 3752 3753 3754 3755 3756 3757 3758 3759 3760 3761 3762 3763 3764 3765 3766 3767 3768 3769 3770 3771 3772 3773 3774
static ssize_t demote_store(struct kobject *kobj,
	       struct kobj_attribute *attr, const char *buf, size_t len)
{
	unsigned long nr_demote;
	unsigned long nr_available;
	nodemask_t nodes_allowed, *n_mask;
	struct hstate *h;
	int err = 0;
	int nid;

	err = kstrtoul(buf, 10, &nr_demote);
	if (err)
		return err;
	h = kobj_to_hstate(kobj, &nid);

	if (nid != NUMA_NO_NODE) {
		init_nodemask_of_node(&nodes_allowed, nid);
		n_mask = &nodes_allowed;
	} else {
		n_mask = &node_states[N_MEMORY];
	}

	/* Synchronize with other sysfs operations modifying huge pages */
	mutex_lock(&h->resize_lock);
	spin_lock_irq(&hugetlb_lock);

	while (nr_demote) {
		/*
		 * Check for available pages to demote each time thorough the
		 * loop as demote_pool_huge_page will drop hugetlb_lock.
		 */
		if (nid != NUMA_NO_NODE)
			nr_available = h->free_huge_pages_node[nid];
		else
			nr_available = h->free_huge_pages;
		nr_available -= h->resv_huge_pages;
		if (!nr_available)
			break;

		err = demote_pool_huge_page(h, n_mask);
		if (err)
			break;

		nr_demote--;
	}

	spin_unlock_irq(&hugetlb_lock);
	mutex_unlock(&h->resize_lock);

	if (err)
		return err;
	return len;
}
HSTATE_ATTR_WO(demote);

static ssize_t demote_size_show(struct kobject *kobj,
					struct kobj_attribute *attr, char *buf)
{
	int nid;
	struct hstate *h = kobj_to_hstate(kobj, &nid);
	unsigned long demote_size = (PAGE_SIZE << h->demote_order) / SZ_1K;

	return sysfs_emit(buf, "%lukB\n", demote_size);
}

static ssize_t demote_size_store(struct kobject *kobj,
					struct kobj_attribute *attr,
					const char *buf, size_t count)
{
	struct hstate *h, *demote_hstate;
	unsigned long demote_size;
	unsigned int demote_order;
	int nid;

	demote_size = (unsigned long)memparse(buf, NULL);

	demote_hstate = size_to_hstate(demote_size);
	if (!demote_hstate)
		return -EINVAL;
	demote_order = demote_hstate->order;
3775 3776
	if (demote_order < HUGETLB_PAGE_ORDER)
		return -EINVAL;
3777 3778 3779 3780 3781 3782 3783 3784 3785 3786 3787 3788 3789 3790 3791

	/* demote order must be smaller than hstate order */
	h = kobj_to_hstate(kobj, &nid);
	if (demote_order >= h->order)
		return -EINVAL;

	/* resize_lock synchronizes access to demote size and writes */
	mutex_lock(&h->resize_lock);
	h->demote_order = demote_order;
	mutex_unlock(&h->resize_lock);

	return count;
}
HSTATE_ATTR(demote_size);

3792 3793 3794 3795 3796 3797
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,
3798 3799 3800
#ifdef CONFIG_NUMA
	&nr_hugepages_mempolicy_attr.attr,
#endif
3801 3802 3803
	NULL,
};

3804
static const struct attribute_group hstate_attr_group = {
3805 3806 3807
	.attrs = hstate_attrs,
};

3808 3809 3810 3811 3812 3813 3814 3815 3816 3817
static struct attribute *hstate_demote_attrs[] = {
	&demote_size_attr.attr,
	&demote_attr.attr,
	NULL,
};

static const struct attribute_group hstate_demote_attr_group = {
	.attrs = hstate_demote_attrs,
};

J
Jeff Mahoney 已提交
3818 3819
static int hugetlb_sysfs_add_hstate(struct hstate *h, struct kobject *parent,
				    struct kobject **hstate_kobjs,
3820
				    const struct attribute_group *hstate_attr_group)
3821 3822
{
	int retval;
3823
	int hi = hstate_index(h);
3824

3825 3826
	hstate_kobjs[hi] = kobject_create_and_add(h->name, parent);
	if (!hstate_kobjs[hi])
3827 3828
		return -ENOMEM;

3829
	retval = sysfs_create_group(hstate_kobjs[hi], hstate_attr_group);
3830
	if (retval) {
3831
		kobject_put(hstate_kobjs[hi]);
3832 3833
		hstate_kobjs[hi] = NULL;
	}
3834

3835 3836 3837 3838 3839 3840
	if (h->demote_order) {
		if (sysfs_create_group(hstate_kobjs[hi],
					&hstate_demote_attr_group))
			pr_warn("HugeTLB unable to create demote interfaces for %s\n", h->name);
	}

3841 3842 3843 3844 3845 3846 3847 3848 3849 3850 3851 3852 3853
	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) {
3854 3855
		err = hugetlb_sysfs_add_hstate(h, hugepages_kobj,
					 hstate_kobjs, &hstate_attr_group);
3856
		if (err)
3857
			pr_err("HugeTLB: Unable to add hstate %s", h->name);
3858 3859 3860
	}
}

3861 3862 3863 3864
#ifdef CONFIG_NUMA

/*
 * node_hstate/s - associate per node hstate attributes, via their kobjects,
3865 3866 3867
 * 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
3868 3869 3870 3871 3872 3873
 * the base kernel, on the hugetlb module.
 */
struct node_hstate {
	struct kobject		*hugepages_kobj;
	struct kobject		*hstate_kobjs[HUGE_MAX_HSTATE];
};
3874
static struct node_hstate node_hstates[MAX_NUMNODES];
3875 3876

/*
3877
 * A subset of global hstate attributes for node devices
3878 3879 3880 3881 3882 3883 3884 3885
 */
static struct attribute *per_node_hstate_attrs[] = {
	&nr_hugepages_attr.attr,
	&free_hugepages_attr.attr,
	&surplus_hugepages_attr.attr,
	NULL,
};

3886
static const struct attribute_group per_node_hstate_attr_group = {
3887 3888 3889 3890
	.attrs = per_node_hstate_attrs,
};

/*
3891
 * kobj_to_node_hstate - lookup global hstate for node device hstate attr kobj.
3892 3893 3894 3895 3896 3897 3898 3899 3900 3901 3902 3903 3904 3905 3906 3907 3908 3909 3910 3911 3912 3913
 * 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;
}

/*
3914
 * Unregister hstate attributes from a single node device.
3915 3916
 * No-op if no hstate attributes attached.
 */
3917
static void hugetlb_unregister_node(struct node *node)
3918 3919
{
	struct hstate *h;
3920
	struct node_hstate *nhs = &node_hstates[node->dev.id];
3921 3922

	if (!nhs->hugepages_kobj)
3923
		return;		/* no hstate attributes */
3924

3925 3926 3927 3928 3929
	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;
3930
		}
3931
	}
3932 3933 3934 3935 3936 3937 3938

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


/*
3939
 * Register hstate attributes for a single node device.
3940 3941
 * No-op if attributes already registered.
 */
3942
static void hugetlb_register_node(struct node *node)
3943 3944
{
	struct hstate *h;
3945
	struct node_hstate *nhs = &node_hstates[node->dev.id];
3946 3947 3948 3949 3950 3951
	int err;

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

	nhs->hugepages_kobj = kobject_create_and_add("hugepages",
3952
							&node->dev.kobj);
3953 3954 3955 3956 3957 3958 3959 3960
	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) {
3961
			pr_err("HugeTLB: Unable to add hstate %s for node %d\n",
3962
				h->name, node->dev.id);
3963 3964 3965 3966 3967 3968 3969
			hugetlb_unregister_node(node);
			break;
		}
	}
}

/*
3970
 * hugetlb init time:  register hstate attributes for all registered node
3971 3972
 * devices of nodes that have memory.  All on-line nodes should have
 * registered their associated device by this time.
3973
 */
3974
static void __init hugetlb_register_all_nodes(void)
3975 3976 3977
{
	int nid;

3978
	for_each_node_state(nid, N_MEMORY) {
3979
		struct node *node = node_devices[nid];
3980
		if (node->dev.id == nid)
3981 3982 3983 3984
			hugetlb_register_node(node);
	}

	/*
3985
	 * Let the node device driver know we're here so it can
3986 3987 3988 3989 3990 3991 3992 3993 3994 3995 3996 3997 3998 3999 4000 4001 4002 4003 4004
	 * [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

4005 4006
static int __init hugetlb_init(void)
{
4007 4008
	int i;

4009 4010 4011
	BUILD_BUG_ON(sizeof_field(struct page, private) * BITS_PER_BYTE <
			__NR_HPAGEFLAGS);

4012 4013 4014
	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");
4015
		return 0;
4016
	}
4017

4018 4019 4020 4021 4022 4023 4024 4025 4026 4027 4028 4029 4030 4031 4032 4033 4034 4035 4036 4037 4038 4039 4040 4041 4042 4043 4044 4045
	/*
	 * 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;
4046 4047 4048 4049

			for (i = 0; i < nr_online_nodes; i++)
				default_hstate.max_huge_pages_node[i] =
					default_hugepages_in_node[i];
4050
		}
4051
	}
4052

4053
	hugetlb_cma_check();
4054
	hugetlb_init_hstates();
4055
	gather_bootmem_prealloc();
4056 4057 4058
	report_hugepages();

	hugetlb_sysfs_init();
4059
	hugetlb_register_all_nodes();
4060
	hugetlb_cgroup_file_init();
4061

4062 4063 4064 4065 4066
#ifdef CONFIG_SMP
	num_fault_mutexes = roundup_pow_of_two(8 * num_possible_cpus());
#else
	num_fault_mutexes = 1;
#endif
4067
	hugetlb_fault_mutex_table =
4068 4069
		kmalloc_array(num_fault_mutexes, sizeof(struct mutex),
			      GFP_KERNEL);
4070
	BUG_ON(!hugetlb_fault_mutex_table);
4071 4072

	for (i = 0; i < num_fault_mutexes; i++)
4073
		mutex_init(&hugetlb_fault_mutex_table[i]);
4074 4075
	return 0;
}
4076
subsys_initcall(hugetlb_init);
4077

4078 4079
/* Overwritten by architectures with more huge page sizes */
bool __init __attribute((weak)) arch_hugetlb_valid_size(unsigned long size)
4080
{
4081
	return size == HPAGE_SIZE;
4082 4083
}

4084
void __init hugetlb_add_hstate(unsigned int order)
4085 4086
{
	struct hstate *h;
4087 4088
	unsigned long i;

4089 4090 4091
	if (size_to_hstate(PAGE_SIZE << order)) {
		return;
	}
4092
	BUG_ON(hugetlb_max_hstate >= HUGE_MAX_HSTATE);
4093
	BUG_ON(order == 0);
4094
	h = &hstates[hugetlb_max_hstate++];
4095
	mutex_init(&h->resize_lock);
4096
	h->order = order;
4097
	h->mask = ~(huge_page_size(h) - 1);
4098 4099
	for (i = 0; i < MAX_NUMNODES; ++i)
		INIT_LIST_HEAD(&h->hugepage_freelists[i]);
4100
	INIT_LIST_HEAD(&h->hugepage_activelist);
4101 4102
	h->next_nid_to_alloc = first_memory_node;
	h->next_nid_to_free = first_memory_node;
4103 4104
	snprintf(h->name, HSTATE_NAME_LEN, "hugepages-%lukB",
					huge_page_size(h)/1024);
4105
	hugetlb_vmemmap_init(h);
4106

4107 4108 4109
	parsed_hstate = h;
}

4110 4111 4112 4113
bool __init __weak hugetlb_node_alloc_supported(void)
{
	return true;
}
4114 4115 4116 4117 4118 4119 4120 4121
/*
 * 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)
4122 4123
{
	unsigned long *mhp;
4124
	static unsigned long *last_mhp;
4125 4126 4127 4128
	int node = NUMA_NO_NODE;
	int count;
	unsigned long tmp;
	char *p = s;
4129

4130
	if (!parsed_valid_hugepagesz) {
4131
		pr_warn("HugeTLB: hugepages=%s does not follow a valid hugepagesz, ignoring\n", s);
4132
		parsed_valid_hugepagesz = true;
4133
		return 0;
4134
	}
4135

4136
	/*
4137 4138 4139 4140
	 * !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.
4141
	 */
4142
	else if (!hugetlb_max_hstate)
4143 4144 4145 4146
		mhp = &default_hstate_max_huge_pages;
	else
		mhp = &parsed_hstate->max_huge_pages;

4147
	if (mhp == last_mhp) {
4148 4149
		pr_warn("HugeTLB: hugepages= specified twice without interleaving hugepagesz=, ignoring hugepages=%s\n", s);
		return 0;
4150 4151
	}

4152 4153 4154 4155 4156 4157 4158 4159 4160 4161
	while (*p) {
		count = 0;
		if (sscanf(p, "%lu%n", &tmp, &count) != 1)
			goto invalid;
		/* Parameter is node format */
		if (p[count] == ':') {
			if (!hugetlb_node_alloc_supported()) {
				pr_warn("HugeTLB: architecture can't support node specific alloc, ignoring!\n");
				return 0;
			}
4162 4163
			if (tmp >= nr_online_nodes)
				goto invalid;
4164 4165 4166 4167 4168 4169 4170 4171 4172 4173 4174 4175 4176 4177 4178 4179 4180 4181 4182 4183 4184 4185
			node = tmp;
			p += count + 1;
			/* Parse hugepages */
			if (sscanf(p, "%lu%n", &tmp, &count) != 1)
				goto invalid;
			if (!hugetlb_max_hstate)
				default_hugepages_in_node[node] = tmp;
			else
				parsed_hstate->max_huge_pages_node[node] = tmp;
			*mhp += tmp;
			/* Go to parse next node*/
			if (p[count] == ',')
				p += count + 1;
			else
				break;
		} else {
			if (p != s)
				goto invalid;
			*mhp = tmp;
			break;
		}
	}
4186

4187 4188
	/*
	 * Global state is always initialized later in hugetlb_init.
4189
	 * But we need to allocate gigantic hstates here early to still
4190 4191
	 * use the bootmem allocator.
	 */
4192
	if (hugetlb_max_hstate && hstate_is_gigantic(parsed_hstate))
4193 4194 4195 4196
		hugetlb_hstate_alloc_pages(parsed_hstate);

	last_mhp = mhp;

4197
	return 1;
4198 4199 4200 4201

invalid:
	pr_warn("HugeTLB: Invalid hugepages parameter %s\n", p);
	return 0;
4202
}
4203
__setup("hugepages=", hugepages_setup);
4204

4205 4206 4207 4208 4209 4210 4211
/*
 * 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.
 */
4212
static int __init hugepagesz_setup(char *s)
4213
{
4214
	unsigned long size;
4215 4216 4217
	struct hstate *h;

	parsed_valid_hugepagesz = false;
4218 4219 4220
	size = (unsigned long)memparse(s, NULL);

	if (!arch_hugetlb_valid_size(size)) {
4221
		pr_err("HugeTLB: unsupported hugepagesz=%s\n", s);
4222 4223 4224
		return 0;
	}

4225 4226 4227 4228 4229 4230 4231 4232 4233 4234 4235 4236 4237 4238 4239 4240 4241 4242 4243 4244 4245 4246 4247
	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;
4248 4249
	}

4250
	hugetlb_add_hstate(ilog2(size) - PAGE_SHIFT);
4251
	parsed_valid_hugepagesz = true;
4252 4253
	return 1;
}
4254 4255
__setup("hugepagesz=", hugepagesz_setup);

4256 4257 4258 4259
/*
 * default_hugepagesz command line input
 * Only one instance of default_hugepagesz allowed on command line.
 */
4260
static int __init default_hugepagesz_setup(char *s)
4261
{
4262
	unsigned long size;
4263
	int i;
4264

4265 4266 4267 4268 4269 4270
	parsed_valid_hugepagesz = false;
	if (parsed_default_hugepagesz) {
		pr_err("HugeTLB: default_hugepagesz previously specified, ignoring %s\n", s);
		return 0;
	}

4271 4272 4273
	size = (unsigned long)memparse(s, NULL);

	if (!arch_hugetlb_valid_size(size)) {
4274
		pr_err("HugeTLB: unsupported default_hugepagesz=%s\n", s);
4275 4276 4277
		return 0;
	}

4278 4279 4280 4281 4282 4283 4284 4285 4286 4287 4288 4289 4290 4291
	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;
4292 4293 4294
		for (i = 0; i < nr_online_nodes; i++)
			default_hstate.max_huge_pages_node[i] =
				default_hugepages_in_node[i];
4295 4296 4297 4298 4299
		if (hstate_is_gigantic(&default_hstate))
			hugetlb_hstate_alloc_pages(&default_hstate);
		default_hstate_max_huge_pages = 0;
	}

4300 4301
	return 1;
}
4302
__setup("default_hugepagesz=", default_hugepagesz_setup);
4303

4304
static unsigned int allowed_mems_nr(struct hstate *h)
4305 4306 4307
{
	int node;
	unsigned int nr = 0;
4308 4309 4310 4311 4312
	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);
4313

4314
	for_each_node_mask(node, cpuset_current_mems_allowed) {
4315
		if (!mpol_allowed || node_isset(node, *mpol_allowed))
4316 4317
			nr += array[node];
	}
4318 4319 4320 4321 4322

	return nr;
}

#ifdef CONFIG_SYSCTL
4323 4324 4325 4326 4327 4328 4329 4330 4331 4332 4333 4334 4335 4336 4337 4338
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);
}

4339 4340
static int hugetlb_sysctl_handler_common(bool obey_mempolicy,
			 struct ctl_table *table, int write,
4341
			 void *buffer, size_t *length, loff_t *ppos)
L
Linus Torvalds 已提交
4342
{
4343
	struct hstate *h = &default_hstate;
4344
	unsigned long tmp = h->max_huge_pages;
4345
	int ret;
4346

4347
	if (!hugepages_supported())
4348
		return -EOPNOTSUPP;
4349

4350 4351
	ret = proc_hugetlb_doulongvec_minmax(table, write, buffer, length, ppos,
					     &tmp);
4352 4353
	if (ret)
		goto out;
4354

4355 4356 4357
	if (write)
		ret = __nr_hugepages_store_common(obey_mempolicy, h,
						  NUMA_NO_NODE, tmp, *length);
4358 4359
out:
	return ret;
L
Linus Torvalds 已提交
4360
}
4361

4362
int hugetlb_sysctl_handler(struct ctl_table *table, int write,
4363
			  void *buffer, size_t *length, loff_t *ppos)
4364 4365 4366 4367 4368 4369 4370 4371
{

	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,
4372
			  void *buffer, size_t *length, loff_t *ppos)
4373 4374 4375 4376 4377 4378
{
	return hugetlb_sysctl_handler_common(true, table, write,
							buffer, length, ppos);
}
#endif /* CONFIG_NUMA */

4379
int hugetlb_overcommit_handler(struct ctl_table *table, int write,
4380
		void *buffer, size_t *length, loff_t *ppos)
4381
{
4382
	struct hstate *h = &default_hstate;
4383
	unsigned long tmp;
4384
	int ret;
4385

4386
	if (!hugepages_supported())
4387
		return -EOPNOTSUPP;
4388

4389
	tmp = h->nr_overcommit_huge_pages;
4390

4391
	if (write && hstate_is_gigantic(h))
4392 4393
		return -EINVAL;

4394 4395
	ret = proc_hugetlb_doulongvec_minmax(table, write, buffer, length, ppos,
					     &tmp);
4396 4397
	if (ret)
		goto out;
4398 4399

	if (write) {
4400
		spin_lock_irq(&hugetlb_lock);
4401
		h->nr_overcommit_huge_pages = tmp;
4402
		spin_unlock_irq(&hugetlb_lock);
4403
	}
4404 4405
out:
	return ret;
4406 4407
}

L
Linus Torvalds 已提交
4408 4409
#endif /* CONFIG_SYSCTL */

4410
void hugetlb_report_meminfo(struct seq_file *m)
L
Linus Torvalds 已提交
4411
{
4412 4413 4414
	struct hstate *h;
	unsigned long total = 0;

4415 4416
	if (!hugepages_supported())
		return;
4417 4418 4419 4420

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

4421
		total += huge_page_size(h) * count;
4422 4423 4424 4425 4426 4427 4428 4429 4430 4431 4432 4433

		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,
4434
				   huge_page_size(h) / SZ_1K);
4435 4436
	}

4437
	seq_printf(m, "Hugetlb:        %8lu kB\n", total / SZ_1K);
L
Linus Torvalds 已提交
4438 4439
}

4440
int hugetlb_report_node_meminfo(char *buf, int len, int nid)
L
Linus Torvalds 已提交
4441
{
4442
	struct hstate *h = &default_hstate;
4443

4444 4445
	if (!hugepages_supported())
		return 0;
4446 4447 4448 4449 4450 4451 4452 4453

	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
4454 4455
}

4456 4457 4458 4459 4460
void hugetlb_show_meminfo(void)
{
	struct hstate *h;
	int nid;

4461 4462 4463
	if (!hugepages_supported())
		return;

4464 4465 4466 4467 4468 4469 4470
	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],
4471
				huge_page_size(h) / SZ_1K);
4472 4473
}

4474 4475 4476 4477 4478 4479
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
4480 4481 4482
/* Return the number pages of memory we physically have, in PAGE_SIZE units. */
unsigned long hugetlb_total_pages(void)
{
4483 4484 4485 4486 4487 4488
	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
4489 4490
}

4491
static int hugetlb_acct_memory(struct hstate *h, long delta)
M
Mel Gorman 已提交
4492 4493 4494
{
	int ret = -ENOMEM;

4495 4496 4497
	if (!delta)
		return 0;

4498
	spin_lock_irq(&hugetlb_lock);
M
Mel Gorman 已提交
4499 4500 4501 4502 4503 4504 4505 4506 4507 4508 4509 4510 4511 4512 4513 4514
	/*
	 * 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.
4515 4516 4517 4518 4519 4520
	 *
	 * 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
4521 4522
	 */
	if (delta > 0) {
4523
		if (gather_surplus_pages(h, delta) < 0)
M
Mel Gorman 已提交
4524 4525
			goto out;

4526
		if (delta > allowed_mems_nr(h)) {
4527
			return_unused_surplus_pages(h, delta);
M
Mel Gorman 已提交
4528 4529 4530 4531 4532 4533
			goto out;
		}
	}

	ret = 0;
	if (delta < 0)
4534
		return_unused_surplus_pages(h, (unsigned long) -delta);
M
Mel Gorman 已提交
4535 4536

out:
4537
	spin_unlock_irq(&hugetlb_lock);
M
Mel Gorman 已提交
4538 4539 4540
	return ret;
}

4541 4542
static void hugetlb_vm_op_open(struct vm_area_struct *vma)
{
4543
	struct resv_map *resv = vma_resv_map(vma);
4544 4545 4546 4547 4548

	/*
	 * 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
4549
	 * has a reference to the reservation map it cannot disappear until
4550 4551 4552
	 * after this open call completes.  It is therefore safe to take a
	 * new reference here without additional locking.
	 */
4553 4554
	if (resv && is_vma_resv_set(vma, HPAGE_RESV_OWNER)) {
		resv_map_dup_hugetlb_cgroup_uncharge_info(resv);
4555
		kref_get(&resv->refs);
4556
	}
4557 4558
}

4559 4560
static void hugetlb_vm_op_close(struct vm_area_struct *vma)
{
4561
	struct hstate *h = hstate_vma(vma);
4562
	struct resv_map *resv = vma_resv_map(vma);
4563
	struct hugepage_subpool *spool = subpool_vma(vma);
4564
	unsigned long reserve, start, end;
4565
	long gbl_reserve;
4566

4567 4568
	if (!resv || !is_vma_resv_set(vma, HPAGE_RESV_OWNER))
		return;
4569

4570 4571
	start = vma_hugecache_offset(h, vma, vma->vm_start);
	end = vma_hugecache_offset(h, vma, vma->vm_end);
4572

4573
	reserve = (end - start) - region_count(resv, start, end);
4574
	hugetlb_cgroup_uncharge_counter(resv, start, end);
4575
	if (reserve) {
4576 4577 4578 4579 4580 4581
		/*
		 * 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);
4582
	}
4583 4584

	kref_put(&resv->refs, resv_map_release);
4585 4586
}

4587 4588 4589 4590 4591 4592 4593
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;
}

4594 4595
static unsigned long hugetlb_vm_op_pagesize(struct vm_area_struct *vma)
{
4596
	return huge_page_size(hstate_vma(vma));
4597 4598
}

L
Linus Torvalds 已提交
4599 4600 4601
/*
 * 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
4602
 * hugepage VMA.  do_page_fault() is supposed to trap this, so BUG is we get
L
Linus Torvalds 已提交
4603 4604
 * this far.
 */
4605
static vm_fault_t hugetlb_vm_op_fault(struct vm_fault *vmf)
L
Linus Torvalds 已提交
4606 4607
{
	BUG();
N
Nick Piggin 已提交
4608
	return 0;
L
Linus Torvalds 已提交
4609 4610
}

4611 4612 4613 4614 4615 4616 4617
/*
 * 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.
 */
4618
const struct vm_operations_struct hugetlb_vm_ops = {
N
Nick Piggin 已提交
4619
	.fault = hugetlb_vm_op_fault,
4620
	.open = hugetlb_vm_op_open,
4621
	.close = hugetlb_vm_op_close,
4622
	.may_split = hugetlb_vm_op_split,
4623
	.pagesize = hugetlb_vm_op_pagesize,
L
Linus Torvalds 已提交
4624 4625
};

4626 4627
static pte_t make_huge_pte(struct vm_area_struct *vma, struct page *page,
				int writable)
D
David Gibson 已提交
4628 4629
{
	pte_t entry;
4630
	unsigned int shift = huge_page_shift(hstate_vma(vma));
D
David Gibson 已提交
4631

4632
	if (writable) {
4633 4634
		entry = huge_pte_mkwrite(huge_pte_mkdirty(mk_huge_pte(page,
					 vma->vm_page_prot)));
D
David Gibson 已提交
4635
	} else {
4636 4637
		entry = huge_pte_wrprotect(mk_huge_pte(page,
					   vma->vm_page_prot));
D
David Gibson 已提交
4638 4639
	}
	entry = pte_mkyoung(entry);
4640
	entry = arch_make_huge_pte(entry, shift, vma->vm_flags);
D
David Gibson 已提交
4641 4642 4643 4644

	return entry;
}

4645 4646 4647 4648 4649
static void set_huge_ptep_writable(struct vm_area_struct *vma,
				   unsigned long address, pte_t *ptep)
{
	pte_t entry;

4650
	entry = huge_pte_mkwrite(huge_pte_mkdirty(huge_ptep_get(ptep)));
4651
	if (huge_ptep_set_access_flags(vma, address, ptep, entry, 1))
4652
		update_mmu_cache(vma, address, ptep);
4653 4654
}

4655
bool is_hugetlb_entry_migration(pte_t pte)
4656 4657 4658 4659
{
	swp_entry_t swp;

	if (huge_pte_none(pte) || pte_present(pte))
4660
		return false;
4661
	swp = pte_to_swp_entry(pte);
4662
	if (is_migration_entry(swp))
4663
		return true;
4664
	else
4665
		return false;
4666 4667
}

4668
static bool is_hugetlb_entry_hwpoisoned(pte_t pte)
4669 4670 4671 4672
{
	swp_entry_t swp;

	if (huge_pte_none(pte) || pte_present(pte))
4673
		return false;
4674
	swp = pte_to_swp_entry(pte);
4675
	if (is_hwpoison_entry(swp))
4676
		return true;
4677
	else
4678
		return false;
4679
}
4680

4681 4682 4683 4684 4685 4686
static void
hugetlb_install_page(struct vm_area_struct *vma, pte_t *ptep, unsigned long addr,
		     struct page *new_page)
{
	__SetPageUptodate(new_page);
	hugepage_add_new_anon_rmap(new_page, vma, addr);
4687
	set_huge_pte_at(vma->vm_mm, addr, ptep, make_huge_pte(vma, new_page, 1));
4688 4689 4690 4691 4692
	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
4693 4694 4695
int copy_hugetlb_page_range(struct mm_struct *dst, struct mm_struct *src,
			    struct vm_area_struct *vma)
{
4696
	pte_t *src_pte, *dst_pte, entry, dst_entry;
D
David Gibson 已提交
4697
	struct page *ptepage;
4698
	unsigned long addr;
4699
	bool cow = is_cow_mapping(vma->vm_flags);
4700 4701
	struct hstate *h = hstate_vma(vma);
	unsigned long sz = huge_page_size(h);
4702
	unsigned long npages = pages_per_huge_page(h);
4703
	struct address_space *mapping = vma->vm_file->f_mapping;
4704
	struct mmu_notifier_range range;
4705
	int ret = 0;
4706

4707
	if (cow) {
4708
		mmu_notifier_range_init(&range, MMU_NOTIFY_CLEAR, 0, vma, src,
4709
					vma->vm_start,
4710 4711
					vma->vm_end);
		mmu_notifier_invalidate_range_start(&range);
4712 4713 4714 4715 4716 4717 4718 4719
	} 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);
4720
	}
4721

4722
	for (addr = vma->vm_start; addr < vma->vm_end; addr += sz) {
4723
		spinlock_t *src_ptl, *dst_ptl;
4724
		src_pte = huge_pte_offset(src, addr, sz);
H
Hugh Dickins 已提交
4725 4726
		if (!src_pte)
			continue;
4727
		dst_pte = huge_pte_alloc(dst, vma, addr, sz);
4728 4729 4730 4731
		if (!dst_pte) {
			ret = -ENOMEM;
			break;
		}
4732

4733 4734 4735 4736 4737 4738 4739 4740 4741 4742 4743
		/*
		 * 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))
4744 4745
			continue;

4746 4747 4748
		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);
4749
		entry = huge_ptep_get(src_pte);
4750
		dst_entry = huge_ptep_get(dst_pte);
4751
again:
4752 4753 4754 4755 4756 4757
		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.
			 */
4758 4759 4760 4761 4762
			;
		} else if (unlikely(is_hugetlb_entry_migration(entry) ||
				    is_hugetlb_entry_hwpoisoned(entry))) {
			swp_entry_t swp_entry = pte_to_swp_entry(entry);

4763
			if (is_writable_migration_entry(swp_entry) && cow) {
4764 4765 4766 4767
				/*
				 * COW mappings require pages in both
				 * parent and child to be set to read.
				 */
4768 4769
				swp_entry = make_readable_migration_entry(
							swp_offset(swp_entry));
4770
				entry = swp_entry_to_pte(swp_entry);
4771 4772
				set_huge_swap_pte_at(src, addr, src_pte,
						     entry, sz);
4773
			}
4774
			set_huge_swap_pte_at(dst, addr, dst_pte, entry, sz);
4775
		} else {
4776 4777 4778 4779 4780 4781 4782 4783 4784 4785 4786 4787 4788 4789 4790 4791 4792 4793 4794 4795 4796 4797 4798 4799 4800 4801 4802 4803 4804 4805 4806 4807 4808 4809 4810 4811
			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)) {
4812 4813
					restore_reserve_on_error(h, vma, addr,
								new);
4814 4815 4816 4817 4818 4819 4820 4821 4822 4823
					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;
			}

4824
			if (cow) {
4825 4826 4827 4828 4829
				/*
				 * No need to notify as we are downgrading page
				 * table protection not changing it to point
				 * to a new page.
				 *
4830
				 * See Documentation/vm/mmu_notifier.rst
4831
				 */
4832
				huge_ptep_set_wrprotect(src, addr, src_pte);
4833
				entry = huge_pte_wrprotect(entry);
4834
			}
4835

4836
			page_dup_rmap(ptepage, true);
4837
			set_huge_pte_at(dst, addr, dst_pte, entry);
4838
			hugetlb_count_add(npages, dst);
4839
		}
4840 4841
		spin_unlock(src_ptl);
		spin_unlock(dst_ptl);
D
David Gibson 已提交
4842 4843
	}

4844
	if (cow)
4845
		mmu_notifier_invalidate_range_end(&range);
4846 4847
	else
		i_mmap_unlock_read(mapping);
4848 4849

	return ret;
D
David Gibson 已提交
4850 4851
}

4852
static void move_huge_pte(struct vm_area_struct *vma, unsigned long old_addr,
4853
			  unsigned long new_addr, pte_t *src_pte, pte_t *dst_pte)
4854 4855 4856 4857
{
	struct hstate *h = hstate_vma(vma);
	struct mm_struct *mm = vma->vm_mm;
	spinlock_t *src_ptl, *dst_ptl;
4858
	pte_t pte;
4859 4860 4861 4862 4863 4864 4865 4866 4867 4868 4869 4870 4871 4872 4873 4874 4875 4876 4877 4878 4879 4880 4881 4882 4883 4884 4885 4886 4887 4888 4889 4890 4891 4892 4893 4894 4895 4896 4897 4898 4899 4900 4901 4902 4903 4904 4905 4906 4907 4908 4909 4910 4911 4912 4913 4914 4915 4916 4917

	dst_ptl = huge_pte_lock(h, mm, dst_pte);
	src_ptl = huge_pte_lockptr(h, mm, src_pte);

	/*
	 * We don't have to worry about the ordering of src and dst ptlocks
	 * because exclusive mmap_sem (or the i_mmap_lock) prevents deadlock.
	 */
	if (src_ptl != dst_ptl)
		spin_lock_nested(src_ptl, SINGLE_DEPTH_NESTING);

	pte = huge_ptep_get_and_clear(mm, old_addr, src_pte);
	set_huge_pte_at(mm, new_addr, dst_pte, pte);

	if (src_ptl != dst_ptl)
		spin_unlock(src_ptl);
	spin_unlock(dst_ptl);
}

int move_hugetlb_page_tables(struct vm_area_struct *vma,
			     struct vm_area_struct *new_vma,
			     unsigned long old_addr, unsigned long new_addr,
			     unsigned long len)
{
	struct hstate *h = hstate_vma(vma);
	struct address_space *mapping = vma->vm_file->f_mapping;
	unsigned long sz = huge_page_size(h);
	struct mm_struct *mm = vma->vm_mm;
	unsigned long old_end = old_addr + len;
	unsigned long old_addr_copy;
	pte_t *src_pte, *dst_pte;
	struct mmu_notifier_range range;

	mmu_notifier_range_init(&range, MMU_NOTIFY_CLEAR, 0, vma, mm, old_addr,
				old_end);
	adjust_range_if_pmd_sharing_possible(vma, &range.start, &range.end);
	mmu_notifier_invalidate_range_start(&range);
	/* Prevent race with file truncation */
	i_mmap_lock_write(mapping);
	for (; old_addr < old_end; old_addr += sz, new_addr += sz) {
		src_pte = huge_pte_offset(mm, old_addr, sz);
		if (!src_pte)
			continue;
		if (huge_pte_none(huge_ptep_get(src_pte)))
			continue;

		/* old_addr arg to huge_pmd_unshare() is a pointer and so the
		 * arg may be modified. Pass a copy instead to preserve the
		 * value in old_addr.
		 */
		old_addr_copy = old_addr;

		if (huge_pmd_unshare(mm, vma, &old_addr_copy, src_pte))
			continue;

		dst_pte = huge_pte_alloc(mm, new_vma, new_addr, sz);
		if (!dst_pte)
			break;

4918
		move_huge_pte(vma, old_addr, new_addr, src_pte, dst_pte);
4919 4920 4921
	}
	flush_tlb_range(vma, old_end - len, old_end);
	mmu_notifier_invalidate_range_end(&range);
4922
	i_mmap_unlock_write(mapping);
4923 4924 4925 4926

	return len + old_addr - old_end;
}

4927 4928 4929
static 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
4930 4931 4932
{
	struct mm_struct *mm = vma->vm_mm;
	unsigned long address;
4933
	pte_t *ptep;
D
David Gibson 已提交
4934
	pte_t pte;
4935
	spinlock_t *ptl;
D
David Gibson 已提交
4936
	struct page *page;
4937 4938
	struct hstate *h = hstate_vma(vma);
	unsigned long sz = huge_page_size(h);
4939
	struct mmu_notifier_range range;
4940
	bool force_flush = false;
4941

D
David Gibson 已提交
4942
	WARN_ON(!is_vm_hugetlb_page(vma));
4943 4944
	BUG_ON(start & ~huge_page_mask(h));
	BUG_ON(end & ~huge_page_mask(h));
D
David Gibson 已提交
4945

4946 4947 4948 4949
	/*
	 * This is a hugetlb vma, all the pte entries should point
	 * to huge page.
	 */
4950
	tlb_change_page_size(tlb, sz);
4951
	tlb_start_vma(tlb, vma);
4952 4953 4954 4955

	/*
	 * If sharing possible, alert mmu notifiers of worst case.
	 */
4956 4957
	mmu_notifier_range_init(&range, MMU_NOTIFY_UNMAP, 0, vma, mm, start,
				end);
4958 4959
	adjust_range_if_pmd_sharing_possible(vma, &range.start, &range.end);
	mmu_notifier_invalidate_range_start(&range);
4960 4961
	address = start;
	for (; address < end; address += sz) {
4962
		ptep = huge_pte_offset(mm, address, sz);
A
Adam Litke 已提交
4963
		if (!ptep)
4964 4965
			continue;

4966
		ptl = huge_pte_lock(h, mm, ptep);
4967
		if (huge_pmd_unshare(mm, vma, &address, ptep)) {
4968
			spin_unlock(ptl);
4969 4970
			tlb_flush_pmd_range(tlb, address & PUD_MASK, PUD_SIZE);
			force_flush = true;
4971 4972
			continue;
		}
4973

4974
		pte = huge_ptep_get(ptep);
4975 4976 4977 4978
		if (huge_pte_none(pte)) {
			spin_unlock(ptl);
			continue;
		}
4979 4980

		/*
4981 4982
		 * Migrating hugepage or HWPoisoned hugepage is already
		 * unmapped and its refcount is dropped, so just clear pte here.
4983
		 */
4984
		if (unlikely(!pte_present(pte))) {
4985
			huge_pte_clear(mm, address, ptep, sz);
4986 4987
			spin_unlock(ptl);
			continue;
4988
		}
4989 4990

		page = pte_page(pte);
4991 4992 4993 4994 4995 4996
		/*
		 * 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) {
4997 4998 4999 5000
			if (page != ref_page) {
				spin_unlock(ptl);
				continue;
			}
5001 5002 5003 5004 5005 5006 5007 5008
			/*
			 * 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);
		}

5009
		pte = huge_ptep_get_and_clear(mm, address, ptep);
5010
		tlb_remove_huge_tlb_entry(h, tlb, ptep, address);
5011
		if (huge_pte_dirty(pte))
5012
			set_page_dirty(page);
5013

5014
		hugetlb_count_sub(pages_per_huge_page(h), mm);
5015
		page_remove_rmap(page, true);
5016

5017
		spin_unlock(ptl);
5018
		tlb_remove_page_size(tlb, page, huge_page_size(h));
5019 5020 5021 5022 5023
		/*
		 * Bail out after unmapping reference page if supplied
		 */
		if (ref_page)
			break;
5024
	}
5025
	mmu_notifier_invalidate_range_end(&range);
5026
	tlb_end_vma(tlb, vma);
5027 5028 5029 5030 5031 5032 5033 5034 5035 5036 5037 5038 5039 5040 5041 5042

	/*
	 * If we unshared PMDs, the TLB flush was not recorded in mmu_gather. We
	 * could defer the flush until now, since by holding i_mmap_rwsem we
	 * guaranteed that the last refernece would not be dropped. But we must
	 * do the flushing before we return, as otherwise i_mmap_rwsem will be
	 * dropped and the last reference to the shared PMDs page might be
	 * dropped as well.
	 *
	 * In theory we could defer the freeing of the PMD pages as well, but
	 * huge_pmd_unshare() relies on the exact page_count for the PMD page to
	 * detect sharing, so we cannot defer the release of the page either.
	 * Instead, do flush now.
	 */
	if (force_flush)
		tlb_flush_mmu_tlbonly(tlb);
L
Linus Torvalds 已提交
5043
}
D
David Gibson 已提交
5044

5045 5046 5047 5048 5049 5050 5051 5052 5053 5054 5055 5056
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
5057
	 * is to clear it before releasing the i_mmap_rwsem. This works
5058
	 * because in the context this is called, the VMA is about to be
5059
	 * destroyed and the i_mmap_rwsem is held.
5060 5061 5062 5063
	 */
	vma->vm_flags &= ~VM_MAYSHARE;
}

5064
void unmap_hugepage_range(struct vm_area_struct *vma, unsigned long start,
5065
			  unsigned long end, struct page *ref_page)
5066
{
5067
	struct mmu_gather tlb;
5068

5069
	tlb_gather_mmu(&tlb, vma->vm_mm);
5070
	__unmap_hugepage_range(&tlb, vma, start, end, ref_page);
5071
	tlb_finish_mmu(&tlb);
5072 5073
}

5074 5075
/*
 * This is called when the original mapper is failing to COW a MAP_PRIVATE
5076
 * mapping it owns the reserve page for. The intention is to unmap the page
5077 5078 5079
 * from other VMAs and let the children be SIGKILLed if they are faulting the
 * same region.
 */
5080 5081
static void unmap_ref_private(struct mm_struct *mm, struct vm_area_struct *vma,
			      struct page *page, unsigned long address)
5082
{
5083
	struct hstate *h = hstate_vma(vma);
5084 5085 5086 5087 5088 5089 5090 5091
	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.
	 */
5092
	address = address & huge_page_mask(h);
5093 5094
	pgoff = ((address - vma->vm_start) >> PAGE_SHIFT) +
			vma->vm_pgoff;
5095
	mapping = vma->vm_file->f_mapping;
5096

5097 5098 5099 5100 5101
	/*
	 * 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
	 */
5102
	i_mmap_lock_write(mapping);
5103
	vma_interval_tree_foreach(iter_vma, &mapping->i_mmap, pgoff, pgoff) {
5104 5105 5106 5107
		/* Do not unmap the current VMA */
		if (iter_vma == vma)
			continue;

5108 5109 5110 5111 5112 5113 5114 5115
		/*
		 * 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;

5116 5117 5118 5119 5120 5121 5122 5123
		/*
		 * 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))
5124 5125
			unmap_hugepage_range(iter_vma, address,
					     address + huge_page_size(h), page);
5126
	}
5127
	i_mmap_unlock_write(mapping);
5128 5129
}

5130 5131
/*
 * Hugetlb_cow() should be called with page lock of the original hugepage held.
5132
 * Called with hugetlb_fault_mutex_table held and pte_page locked so we
5133 5134
 * cannot race with other handlers or page migration.
 * Keep the pte_same checks anyway to make transition from the mutex easier.
5135
 */
5136
static vm_fault_t hugetlb_cow(struct mm_struct *mm, struct vm_area_struct *vma,
5137
		       unsigned long address, pte_t *ptep,
5138
		       struct page *pagecache_page, spinlock_t *ptl)
5139
{
5140
	pte_t pte;
5141
	struct hstate *h = hstate_vma(vma);
5142
	struct page *old_page, *new_page;
5143 5144
	int outside_reserve = 0;
	vm_fault_t ret = 0;
5145
	unsigned long haddr = address & huge_page_mask(h);
5146
	struct mmu_notifier_range range;
5147

5148
	pte = huge_ptep_get(ptep);
5149 5150
	old_page = pte_page(pte);

5151
retry_avoidcopy:
5152 5153
	/* If no-one else is actually using this page, avoid the copy
	 * and just make the page writable */
5154
	if (page_mapcount(old_page) == 1 && PageAnon(old_page)) {
5155
		page_move_anon_rmap(old_page, vma);
5156
		set_huge_ptep_writable(vma, haddr, ptep);
N
Nick Piggin 已提交
5157
		return 0;
5158 5159
	}

5160 5161 5162 5163 5164 5165 5166 5167 5168
	/*
	 * 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.
	 */
5169
	if (is_vma_resv_set(vma, HPAGE_RESV_OWNER) &&
5170 5171 5172
			old_page != pagecache_page)
		outside_reserve = 1;

5173
	get_page(old_page);
5174

5175 5176 5177 5178
	/*
	 * Drop page table lock as buddy allocator may be called. It will
	 * be acquired again before returning to the caller, as expected.
	 */
5179
	spin_unlock(ptl);
5180
	new_page = alloc_huge_page(vma, haddr, outside_reserve);
5181

5182
	if (IS_ERR(new_page)) {
5183 5184 5185 5186 5187 5188 5189 5190
		/*
		 * 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) {
5191 5192 5193 5194
			struct address_space *mapping = vma->vm_file->f_mapping;
			pgoff_t idx;
			u32 hash;

5195
			put_page(old_page);
5196
			BUG_ON(huge_pte_none(pte));
5197 5198 5199 5200 5201 5202 5203 5204 5205 5206 5207 5208 5209 5210
			/*
			 * 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);

5211
			unmap_ref_private(mm, vma, old_page, haddr);
5212 5213 5214

			i_mmap_lock_read(mapping);
			mutex_lock(&hugetlb_fault_mutex_table[hash]);
5215
			spin_lock(ptl);
5216
			ptep = huge_pte_offset(mm, haddr, huge_page_size(h));
5217 5218 5219 5220 5221 5222 5223 5224
			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;
5225 5226
		}

5227
		ret = vmf_error(PTR_ERR(new_page));
5228
		goto out_release_old;
5229 5230
	}

5231 5232 5233 5234
	/*
	 * When the original hugepage is shared one, it does not have
	 * anon_vma prepared.
	 */
5235
	if (unlikely(anon_vma_prepare(vma))) {
5236 5237
		ret = VM_FAULT_OOM;
		goto out_release_all;
5238
	}
5239

5240
	copy_user_huge_page(new_page, old_page, address, vma,
A
Andrea Arcangeli 已提交
5241
			    pages_per_huge_page(h));
N
Nick Piggin 已提交
5242
	__SetPageUptodate(new_page);
5243

5244
	mmu_notifier_range_init(&range, MMU_NOTIFY_CLEAR, 0, vma, mm, haddr,
5245
				haddr + huge_page_size(h));
5246
	mmu_notifier_invalidate_range_start(&range);
5247

5248
	/*
5249
	 * Retake the page table lock to check for racing updates
5250 5251
	 * before the page tables are altered
	 */
5252
	spin_lock(ptl);
5253
	ptep = huge_pte_offset(mm, haddr, huge_page_size(h));
5254
	if (likely(ptep && pte_same(huge_ptep_get(ptep), pte))) {
5255
		ClearHPageRestoreReserve(new_page);
5256

5257
		/* Break COW */
5258
		huge_ptep_clear_flush(vma, haddr, ptep);
5259
		mmu_notifier_invalidate_range(mm, range.start, range.end);
5260
		page_remove_rmap(old_page, true);
5261
		hugepage_add_new_anon_rmap(new_page, vma, haddr);
5262 5263
		set_huge_pte_at(mm, haddr, ptep,
				make_huge_pte(vma, new_page, 1));
5264
		SetHPageMigratable(new_page);
5265 5266 5267
		/* Make the old page be freed below */
		new_page = old_page;
	}
5268
	spin_unlock(ptl);
5269
	mmu_notifier_invalidate_range_end(&range);
5270
out_release_all:
5271 5272 5273
	/* No restore in case of successful pagetable update (Break COW) */
	if (new_page != old_page)
		restore_reserve_on_error(h, vma, haddr, new_page);
5274
	put_page(new_page);
5275
out_release_old:
5276
	put_page(old_page);
5277

5278 5279
	spin_lock(ptl); /* Caller expects lock to be held */
	return ret;
5280 5281
}

5282
/* Return the pagecache page at a given address within a VMA */
5283 5284
static struct page *hugetlbfs_pagecache_page(struct hstate *h,
			struct vm_area_struct *vma, unsigned long address)
5285 5286
{
	struct address_space *mapping;
5287
	pgoff_t idx;
5288 5289

	mapping = vma->vm_file->f_mapping;
5290
	idx = vma_hugecache_offset(h, vma, address);
5291 5292 5293 5294

	return find_lock_page(mapping, idx);
}

H
Hugh Dickins 已提交
5295 5296 5297 5298 5299
/*
 * 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
5300 5301 5302 5303 5304 5305 5306 5307 5308 5309 5310 5311 5312 5313 5314
			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;
}

5315 5316 5317 5318 5319 5320 5321 5322 5323
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;
5324
	ClearHPageRestoreReserve(page);
5325

5326 5327 5328 5329 5330 5331
	/*
	 * set page dirty so that it will not be removed from cache/file
	 * by non-hugetlbfs specific code paths.
	 */
	set_page_dirty(page);

5332 5333 5334 5335 5336 5337
	spin_lock(&inode->i_lock);
	inode->i_blocks += blocks_per_huge_page(h);
	spin_unlock(&inode->i_lock);
	return 0;
}

5338 5339 5340 5341 5342 5343 5344 5345 5346 5347 5348 5349 5350 5351 5352 5353 5354 5355 5356 5357 5358 5359 5360 5361 5362 5363 5364 5365 5366 5367 5368 5369 5370 5371 5372 5373 5374 5375
static inline vm_fault_t hugetlb_handle_userfault(struct vm_area_struct *vma,
						  struct address_space *mapping,
						  pgoff_t idx,
						  unsigned int flags,
						  unsigned long haddr,
						  unsigned long reason)
{
	vm_fault_t ret;
	u32 hash;
	struct vm_fault vmf = {
		.vma = vma,
		.address = haddr,
		.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
		 */
	};

	/*
	 * hugetlb_fault_mutex and i_mmap_rwsem must be
	 * dropped before handling userfault.  Reacquire
	 * after handling fault to make calling code simpler.
	 */
	hash = hugetlb_fault_mutex_hash(mapping, idx);
	mutex_unlock(&hugetlb_fault_mutex_table[hash]);
	i_mmap_unlock_read(mapping);
	ret = handle_userfault(&vmf, reason);
	i_mmap_lock_read(mapping);
	mutex_lock(&hugetlb_fault_mutex_table[hash]);

	return ret;
}

5376 5377 5378 5379
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)
5380
{
5381
	struct hstate *h = hstate_vma(vma);
5382
	vm_fault_t ret = VM_FAULT_SIGBUS;
5383
	int anon_rmap = 0;
A
Adam Litke 已提交
5384 5385
	unsigned long size;
	struct page *page;
5386
	pte_t new_pte;
5387
	spinlock_t *ptl;
5388
	unsigned long haddr = address & huge_page_mask(h);
5389
	bool new_page, new_pagecache_page = false;
A
Adam Litke 已提交
5390

5391 5392 5393
	/*
	 * 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
5394
	 * COW. Warn that such a situation has occurred as it may not be obvious
5395 5396
	 */
	if (is_vma_resv_set(vma, HPAGE_RESV_UNMAPPED)) {
5397
		pr_warn_ratelimited("PID %d killed due to inadequate hugepage pool\n",
5398
			   current->pid);
5399 5400 5401
		return ret;
	}

A
Adam Litke 已提交
5402
	/*
5403 5404 5405
	 * 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
5406
	 */
5407 5408 5409 5410
	size = i_size_read(mapping->host) >> huge_page_shift(h);
	if (idx >= size)
		goto out;

5411
retry:
5412
	new_page = false;
5413 5414
	page = find_lock_page(mapping, idx);
	if (!page) {
5415
		/* Check for page in userfault range */
5416
		if (userfaultfd_missing(vma)) {
5417 5418 5419
			ret = hugetlb_handle_userfault(vma, mapping, idx,
						       flags, haddr,
						       VM_UFFD_MISSING);
5420 5421 5422
			goto out;
		}

5423
		page = alloc_huge_page(vma, haddr, 0);
5424
		if (IS_ERR(page)) {
5425 5426 5427 5428 5429 5430 5431 5432 5433 5434 5435 5436 5437
			/*
			 * 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);
5438 5439 5440
			ret = 0;
			if (huge_pte_none(huge_ptep_get(ptep)))
				ret = vmf_error(PTR_ERR(page));
5441
			spin_unlock(ptl);
5442 5443
			goto out;
		}
A
Andrea Arcangeli 已提交
5444
		clear_huge_page(page, address, pages_per_huge_page(h));
N
Nick Piggin 已提交
5445
		__SetPageUptodate(page);
5446
		new_page = true;
5447

5448
		if (vma->vm_flags & VM_MAYSHARE) {
5449
			int err = huge_add_to_page_cache(page, mapping, idx);
5450 5451 5452 5453 5454 5455
			if (err) {
				put_page(page);
				if (err == -EEXIST)
					goto retry;
				goto out;
			}
5456
			new_pagecache_page = true;
5457
		} else {
5458
			lock_page(page);
5459 5460 5461 5462
			if (unlikely(anon_vma_prepare(vma))) {
				ret = VM_FAULT_OOM;
				goto backout_unlocked;
			}
5463
			anon_rmap = 1;
5464
		}
5465
	} else {
5466 5467 5468 5469 5470 5471
		/*
		 * 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))) {
5472
			ret = VM_FAULT_HWPOISON_LARGE |
5473
				VM_FAULT_SET_HINDEX(hstate_index(h));
5474 5475
			goto backout_unlocked;
		}
5476 5477 5478 5479 5480 5481 5482 5483 5484 5485

		/* Check for page in userfault range. */
		if (userfaultfd_minor(vma)) {
			unlock_page(page);
			put_page(page);
			ret = hugetlb_handle_userfault(vma, mapping, idx,
						       flags, haddr,
						       VM_UFFD_MINOR);
			goto out;
		}
5486
	}
5487

5488 5489 5490 5491 5492 5493
	/*
	 * 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.
	 */
5494
	if ((flags & FAULT_FLAG_WRITE) && !(vma->vm_flags & VM_SHARED)) {
5495
		if (vma_needs_reservation(h, vma, haddr) < 0) {
5496 5497 5498
			ret = VM_FAULT_OOM;
			goto backout_unlocked;
		}
5499
		/* Just decrements count, does not deallocate */
5500
		vma_end_reservation(h, vma, haddr);
5501
	}
5502

5503
	ptl = huge_pte_lock(h, mm, ptep);
N
Nick Piggin 已提交
5504
	ret = 0;
5505
	if (!huge_pte_none(huge_ptep_get(ptep)))
A
Adam Litke 已提交
5506 5507
		goto backout;

5508
	if (anon_rmap) {
5509
		ClearHPageRestoreReserve(page);
5510
		hugepage_add_new_anon_rmap(page, vma, haddr);
5511
	} else
5512
		page_dup_rmap(page, true);
5513 5514
	new_pte = make_huge_pte(vma, page, ((vma->vm_flags & VM_WRITE)
				&& (vma->vm_flags & VM_SHARED)));
5515
	set_huge_pte_at(mm, haddr, ptep, new_pte);
5516

5517
	hugetlb_count_add(pages_per_huge_page(h), mm);
5518
	if ((flags & FAULT_FLAG_WRITE) && !(vma->vm_flags & VM_SHARED)) {
5519
		/* Optimization, do the COW without a second fault */
5520
		ret = hugetlb_cow(mm, vma, address, ptep, page, ptl);
5521 5522
	}

5523
	spin_unlock(ptl);
5524 5525

	/*
5526 5527 5528
	 * Only set HPageMigratable in newly allocated pages.  Existing pages
	 * found in the pagecache may not have HPageMigratableset if they have
	 * been isolated for migration.
5529 5530
	 */
	if (new_page)
5531
		SetHPageMigratable(page);
5532

A
Adam Litke 已提交
5533 5534
	unlock_page(page);
out:
5535
	return ret;
A
Adam Litke 已提交
5536 5537

backout:
5538
	spin_unlock(ptl);
5539
backout_unlocked:
A
Adam Litke 已提交
5540
	unlock_page(page);
5541 5542 5543
	/* restore reserve for newly allocated pages not in page cache */
	if (new_page && !new_pagecache_page)
		restore_reserve_on_error(h, vma, haddr, page);
A
Adam Litke 已提交
5544 5545
	put_page(page);
	goto out;
5546 5547
}

5548
#ifdef CONFIG_SMP
5549
u32 hugetlb_fault_mutex_hash(struct address_space *mapping, pgoff_t idx)
5550 5551 5552 5553
{
	unsigned long key[2];
	u32 hash;

5554 5555
	key[0] = (unsigned long) mapping;
	key[1] = idx;
5556

5557
	hash = jhash2((u32 *)&key, sizeof(key)/(sizeof(u32)), 0);
5558 5559 5560 5561 5562

	return hash & (num_fault_mutexes - 1);
}
#else
/*
M
Miaohe Lin 已提交
5563
 * For uniprocessor systems we always use a single mutex, so just
5564 5565
 * return 0 and avoid the hashing overhead.
 */
5566
u32 hugetlb_fault_mutex_hash(struct address_space *mapping, pgoff_t idx)
5567 5568 5569 5570 5571
{
	return 0;
}
#endif

5572
vm_fault_t hugetlb_fault(struct mm_struct *mm, struct vm_area_struct *vma,
5573
			unsigned long address, unsigned int flags)
5574
{
5575
	pte_t *ptep, entry;
5576
	spinlock_t *ptl;
5577
	vm_fault_t ret;
5578 5579
	u32 hash;
	pgoff_t idx;
5580
	struct page *page = NULL;
5581
	struct page *pagecache_page = NULL;
5582
	struct hstate *h = hstate_vma(vma);
5583
	struct address_space *mapping;
5584
	int need_wait_lock = 0;
5585
	unsigned long haddr = address & huge_page_mask(h);
5586

5587
	ptep = huge_pte_offset(mm, haddr, huge_page_size(h));
5588
	if (ptep) {
5589 5590 5591 5592 5593
		/*
		 * 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.
		 */
5594
		entry = huge_ptep_get(ptep);
N
Naoya Horiguchi 已提交
5595
		if (unlikely(is_hugetlb_entry_migration(entry))) {
5596
			migration_entry_wait_huge(vma, mm, ptep);
N
Naoya Horiguchi 已提交
5597 5598
			return 0;
		} else if (unlikely(is_hugetlb_entry_hwpoisoned(entry)))
5599
			return VM_FAULT_HWPOISON_LARGE |
5600
				VM_FAULT_SET_HINDEX(hstate_index(h));
5601 5602
	}

5603 5604
	/*
	 * Acquire i_mmap_rwsem before calling huge_pte_alloc and hold
5605 5606 5607 5608
	 * 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.
5609 5610 5611 5612 5613
	 *
	 * 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.
	 */
5614
	mapping = vma->vm_file->f_mapping;
5615
	i_mmap_lock_read(mapping);
5616
	ptep = huge_pte_alloc(mm, vma, haddr, huge_page_size(h));
5617 5618 5619 5620
	if (!ptep) {
		i_mmap_unlock_read(mapping);
		return VM_FAULT_OOM;
	}
5621

5622 5623 5624 5625 5626
	/*
	 * 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.
	 */
5627
	idx = vma_hugecache_offset(h, vma, haddr);
5628
	hash = hugetlb_fault_mutex_hash(mapping, idx);
5629
	mutex_lock(&hugetlb_fault_mutex_table[hash]);
5630

5631 5632
	entry = huge_ptep_get(ptep);
	if (huge_pte_none(entry)) {
5633
		ret = hugetlb_no_page(mm, vma, mapping, idx, address, ptep, flags);
5634
		goto out_mutex;
5635
	}
5636

N
Nick Piggin 已提交
5637
	ret = 0;
5638

5639 5640 5641
	/*
	 * 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
5642 5643 5644
	 * an active hugepage in pagecache. This goto expects the 2nd page
	 * fault, and is_hugetlb_entry_(migration|hwpoisoned) check will
	 * properly handle it.
5645 5646 5647 5648
	 */
	if (!pte_present(entry))
		goto out_mutex;

5649 5650 5651 5652 5653 5654 5655 5656
	/*
	 * 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.
	 */
5657
	if ((flags & FAULT_FLAG_WRITE) && !huge_pte_write(entry)) {
5658
		if (vma_needs_reservation(h, vma, haddr) < 0) {
5659
			ret = VM_FAULT_OOM;
5660
			goto out_mutex;
5661
		}
5662
		/* Just decrements count, does not deallocate */
5663
		vma_end_reservation(h, vma, haddr);
5664

5665
		if (!(vma->vm_flags & VM_MAYSHARE))
5666
			pagecache_page = hugetlbfs_pagecache_page(h,
5667
								vma, haddr);
5668 5669
	}

5670 5671 5672 5673 5674 5675
	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;

5676 5677 5678 5679 5680 5681 5682
	/*
	 * 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)
5683 5684 5685 5686
		if (!trylock_page(page)) {
			need_wait_lock = 1;
			goto out_ptl;
		}
5687

5688
	get_page(page);
5689

5690
	if (flags & FAULT_FLAG_WRITE) {
5691
		if (!huge_pte_write(entry)) {
5692
			ret = hugetlb_cow(mm, vma, address, ptep,
5693
					  pagecache_page, ptl);
5694
			goto out_put_page;
5695
		}
5696
		entry = huge_pte_mkdirty(entry);
5697 5698
	}
	entry = pte_mkyoung(entry);
5699
	if (huge_ptep_set_access_flags(vma, haddr, ptep, entry,
5700
						flags & FAULT_FLAG_WRITE))
5701
		update_mmu_cache(vma, haddr, ptep);
5702 5703 5704 5705
out_put_page:
	if (page != pagecache_page)
		unlock_page(page);
	put_page(page);
5706 5707
out_ptl:
	spin_unlock(ptl);
5708 5709 5710 5711 5712

	if (pagecache_page) {
		unlock_page(pagecache_page);
		put_page(pagecache_page);
	}
5713
out_mutex:
5714
	mutex_unlock(&hugetlb_fault_mutex_table[hash]);
5715
	i_mmap_unlock_read(mapping);
5716 5717 5718 5719 5720 5721 5722 5723 5724
	/*
	 * 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);
5725
	return ret;
5726 5727
}

5728
#ifdef CONFIG_USERFAULTFD
5729 5730 5731 5732 5733 5734 5735 5736 5737
/*
 * 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,
5738
			    enum mcopy_atomic_mode mode,
5739 5740
			    struct page **pagep)
{
5741
	bool is_continue = (mode == MCOPY_ATOMIC_CONTINUE);
5742 5743 5744
	struct hstate *h = hstate_vma(dst_vma);
	struct address_space *mapping = dst_vma->vm_file->f_mapping;
	pgoff_t idx = vma_hugecache_offset(h, dst_vma, dst_addr);
5745
	unsigned long size;
5746
	int vm_shared = dst_vma->vm_flags & VM_SHARED;
5747 5748
	pte_t _dst_pte;
	spinlock_t *ptl;
5749
	int ret = -ENOMEM;
5750
	struct page *page;
5751
	int writable;
5752
	bool page_in_pagecache = false;
5753

5754 5755 5756 5757 5758
	if (is_continue) {
		ret = -EFAULT;
		page = find_lock_page(mapping, idx);
		if (!page)
			goto out;
5759
		page_in_pagecache = true;
5760
	} else if (!*pagep) {
5761 5762 5763 5764 5765 5766 5767 5768 5769
		/* If a page already exists, then it's UFFDIO_COPY for
		 * a non-missing case. Return -EEXIST.
		 */
		if (vm_shared &&
		    hugetlbfs_pagecache_present(h, dst_vma, dst_addr)) {
			ret = -EEXIST;
			goto out;
		}

5770
		page = alloc_huge_page(dst_vma, dst_addr, 0);
5771 5772
		if (IS_ERR(page)) {
			ret = -ENOMEM;
5773
			goto out;
5774
		}
5775 5776 5777

		ret = copy_huge_page_from_user(page,
						(const void __user *) src_addr,
5778
						pages_per_huge_page(h), false);
5779

5780
		/* fallback to copy_from_user outside mmap_lock */
5781
		if (unlikely(ret)) {
5782
			ret = -ENOENT;
5783 5784 5785 5786 5787 5788 5789 5790 5791 5792 5793 5794 5795 5796
			/* Free the allocated page which may have
			 * consumed a reservation.
			 */
			restore_reserve_on_error(h, dst_vma, dst_addr, page);
			put_page(page);

			/* Allocate a temporary page to hold the copied
			 * contents.
			 */
			page = alloc_huge_page_vma(h, dst_vma, dst_addr);
			if (!page) {
				ret = -ENOMEM;
				goto out;
			}
5797
			*pagep = page;
5798 5799 5800 5801
			/* Set the outparam pagep and return to the caller to
			 * copy the contents outside the lock. Don't free the
			 * page.
			 */
5802 5803 5804
			goto out;
		}
	} else {
5805 5806 5807 5808 5809 5810 5811 5812 5813 5814 5815 5816 5817 5818
		if (vm_shared &&
		    hugetlbfs_pagecache_present(h, dst_vma, dst_addr)) {
			put_page(*pagep);
			ret = -EEXIST;
			*pagep = NULL;
			goto out;
		}

		page = alloc_huge_page(dst_vma, dst_addr, 0);
		if (IS_ERR(page)) {
			ret = -ENOMEM;
			*pagep = NULL;
			goto out;
		}
5819 5820
		copy_user_huge_page(page, *pagep, dst_addr, dst_vma,
				    pages_per_huge_page(h));
5821
		put_page(*pagep);
5822 5823 5824 5825 5826 5827 5828 5829 5830 5831
		*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);

5832 5833
	/* Add shared, newly allocated pages to the page cache. */
	if (vm_shared && !is_continue) {
5834 5835 5836 5837
		size = i_size_read(mapping->host) >> huge_page_shift(h);
		ret = -EFAULT;
		if (idx >= size)
			goto out_release_nounlock;
5838

5839 5840 5841 5842 5843 5844
		/*
		 * 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.
		 */
5845 5846 5847
		ret = huge_add_to_page_cache(page, mapping, idx);
		if (ret)
			goto out_release_nounlock;
5848
		page_in_pagecache = true;
5849 5850
	}

5851 5852 5853
	ptl = huge_pte_lockptr(h, dst_mm, dst_pte);
	spin_lock(ptl);

5854 5855 5856 5857 5858 5859 5860 5861 5862 5863 5864 5865 5866 5867
	/*
	 * 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;

5868 5869 5870 5871
	ret = -EEXIST;
	if (!huge_pte_none(huge_ptep_get(dst_pte)))
		goto out_release_unlock;

5872 5873 5874
	if (vm_shared) {
		page_dup_rmap(page, true);
	} else {
5875
		ClearHPageRestoreReserve(page);
5876 5877
		hugepage_add_new_anon_rmap(page, dst_vma, dst_addr);
	}
5878

5879 5880 5881 5882 5883 5884 5885 5886
	/* For CONTINUE on a non-shared VMA, don't set VM_WRITE for CoW. */
	if (is_continue && !vm_shared)
		writable = 0;
	else
		writable = dst_vma->vm_flags & VM_WRITE;

	_dst_pte = make_huge_pte(dst_vma, page, writable);
	if (writable)
5887 5888 5889 5890 5891 5892 5893 5894 5895 5896 5897 5898 5899
		_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);
5900 5901 5902
	if (!is_continue)
		SetHPageMigratable(page);
	if (vm_shared || is_continue)
5903
		unlock_page(page);
5904 5905 5906 5907 5908
	ret = 0;
out:
	return ret;
out_release_unlock:
	spin_unlock(ptl);
5909
	if (vm_shared || is_continue)
5910
		unlock_page(page);
5911
out_release_nounlock:
5912
	if (!page_in_pagecache)
5913
		restore_reserve_on_error(h, dst_vma, dst_addr, page);
5914 5915 5916
	put_page(page);
	goto out;
}
5917
#endif /* CONFIG_USERFAULTFD */
5918

5919 5920 5921 5922 5923 5924 5925 5926 5927 5928 5929 5930 5931 5932
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;
	}
}

5933 5934 5935
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,
5936
			 long i, unsigned int flags, int *locked)
D
David Gibson 已提交
5937
{
5938 5939
	unsigned long pfn_offset;
	unsigned long vaddr = *position;
5940
	unsigned long remainder = *nr_pages;
5941
	struct hstate *h = hstate_vma(vma);
5942
	int err = -EFAULT, refs;
D
David Gibson 已提交
5943 5944

	while (vaddr < vma->vm_end && remainder) {
A
Adam Litke 已提交
5945
		pte_t *pte;
5946
		spinlock_t *ptl = NULL;
H
Hugh Dickins 已提交
5947
		int absent;
A
Adam Litke 已提交
5948
		struct page *page;
D
David Gibson 已提交
5949

5950 5951 5952 5953
		/*
		 * If we have a pending SIGKILL, don't keep faulting pages and
		 * potentially allocating memory.
		 */
5954
		if (fatal_signal_pending(current)) {
5955 5956 5957 5958
			remainder = 0;
			break;
		}

A
Adam Litke 已提交
5959 5960
		/*
		 * Some archs (sparc64, sh*) have multiple pte_ts to
H
Hugh Dickins 已提交
5961
		 * each hugepage.  We have to make sure we get the
A
Adam Litke 已提交
5962
		 * first, for the page indexing below to work.
5963 5964
		 *
		 * Note that page table lock is not held when pte is null.
A
Adam Litke 已提交
5965
		 */
5966 5967
		pte = huge_pte_offset(mm, vaddr & huge_page_mask(h),
				      huge_page_size(h));
5968 5969
		if (pte)
			ptl = huge_pte_lock(h, mm, pte);
H
Hugh Dickins 已提交
5970 5971 5972 5973
		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
5974 5975 5976 5977
		 * 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
5978
		 */
H
Hugh Dickins 已提交
5979 5980
		if (absent && (flags & FOLL_DUMP) &&
		    !hugetlbfs_pagecache_present(h, vma, vaddr)) {
5981 5982
			if (pte)
				spin_unlock(ptl);
H
Hugh Dickins 已提交
5983 5984 5985
			remainder = 0;
			break;
		}
D
David Gibson 已提交
5986

5987 5988 5989 5990 5991 5992 5993 5994 5995 5996 5997
		/*
		 * 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)) ||
5998 5999
		    ((flags & FOLL_WRITE) &&
		      !huge_pte_write(huge_ptep_get(pte)))) {
6000
			vm_fault_t ret;
6001
			unsigned int fault_flags = 0;
D
David Gibson 已提交
6002

6003 6004
			if (pte)
				spin_unlock(ptl);
6005 6006
			if (flags & FOLL_WRITE)
				fault_flags |= FAULT_FLAG_WRITE;
6007
			if (locked)
6008 6009
				fault_flags |= FAULT_FLAG_ALLOW_RETRY |
					FAULT_FLAG_KILLABLE;
6010 6011 6012 6013
			if (flags & FOLL_NOWAIT)
				fault_flags |= FAULT_FLAG_ALLOW_RETRY |
					FAULT_FLAG_RETRY_NOWAIT;
			if (flags & FOLL_TRIED) {
6014 6015 6016 6017
				/*
				 * Note: FAULT_FLAG_ALLOW_RETRY and
				 * FAULT_FLAG_TRIED can co-exist
				 */
6018 6019 6020 6021
				fault_flags |= FAULT_FLAG_TRIED;
			}
			ret = hugetlb_fault(mm, vma, vaddr, fault_flags);
			if (ret & VM_FAULT_ERROR) {
6022
				err = vm_fault_to_errno(ret, flags);
6023 6024 6025 6026
				remainder = 0;
				break;
			}
			if (ret & VM_FAULT_RETRY) {
6027
				if (locked &&
6028
				    !(fault_flags & FAULT_FLAG_RETRY_NOWAIT))
6029
					*locked = 0;
6030 6031 6032 6033 6034 6035 6036 6037 6038 6039 6040 6041 6042
				*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
6043 6044
		}

6045
		pfn_offset = (vaddr & ~huge_page_mask(h)) >> PAGE_SHIFT;
6046
		page = pte_page(huge_ptep_get(pte));
6047

6048 6049 6050 6051 6052 6053 6054 6055 6056 6057 6058 6059 6060 6061
		/*
		 * 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;
		}

6062 6063 6064
		/* vaddr may not be aligned to PAGE_SIZE */
		refs = min3(pages_per_huge_page(h) - pfn_offset, remainder,
		    (vma->vm_end - ALIGN_DOWN(vaddr, PAGE_SIZE)) >> PAGE_SHIFT);
6065

6066 6067 6068 6069 6070
		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
6071

6072
		if (pages) {
6073 6074 6075 6076 6077 6078 6079 6080 6081 6082
			/*
			 * 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:
			 */
6083
			if (WARN_ON_ONCE(!try_grab_compound_head(pages[i],
6084 6085 6086 6087 6088 6089 6090
								 refs,
								 flags))) {
				spin_unlock(ptl);
				remainder = 0;
				err = -ENOMEM;
				break;
			}
6091
		}
6092 6093 6094 6095 6096

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

6097
		spin_unlock(ptl);
D
David Gibson 已提交
6098
	}
6099
	*nr_pages = remainder;
6100 6101 6102 6103 6104
	/*
	 * 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
6105 6106
	*position = vaddr;

6107
	return i ? i : err;
D
David Gibson 已提交
6108
}
6109

6110
unsigned long hugetlb_change_protection(struct vm_area_struct *vma,
6111 6112 6113 6114 6115 6116
		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;
6117
	struct hstate *h = hstate_vma(vma);
6118
	unsigned long pages = 0;
6119
	bool shared_pmd = false;
6120
	struct mmu_notifier_range range;
6121 6122 6123

	/*
	 * In the case of shared PMDs, the area to flush could be beyond
6124
	 * start/end.  Set range.start/range.end to cover the maximum possible
6125 6126
	 * range if PMD sharing is possible.
	 */
6127 6128
	mmu_notifier_range_init(&range, MMU_NOTIFY_PROTECTION_VMA,
				0, vma, mm, start, end);
6129
	adjust_range_if_pmd_sharing_possible(vma, &range.start, &range.end);
6130 6131

	BUG_ON(address >= end);
6132
	flush_cache_range(vma, range.start, range.end);
6133

6134
	mmu_notifier_invalidate_range_start(&range);
6135
	i_mmap_lock_write(vma->vm_file->f_mapping);
6136
	for (; address < end; address += huge_page_size(h)) {
6137
		spinlock_t *ptl;
6138
		ptep = huge_pte_offset(mm, address, huge_page_size(h));
6139 6140
		if (!ptep)
			continue;
6141
		ptl = huge_pte_lock(h, mm, ptep);
6142
		if (huge_pmd_unshare(mm, vma, &address, ptep)) {
6143
			pages++;
6144
			spin_unlock(ptl);
6145
			shared_pmd = true;
6146
			continue;
6147
		}
6148 6149 6150 6151 6152 6153 6154 6155
		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);

6156
			if (is_writable_migration_entry(entry)) {
6157 6158
				pte_t newpte;

6159 6160
				entry = make_readable_migration_entry(
							swp_offset(entry));
6161
				newpte = swp_entry_to_pte(entry);
6162 6163
				set_huge_swap_pte_at(mm, address, ptep,
						     newpte, huge_page_size(h));
6164 6165 6166 6167 6168 6169
				pages++;
			}
			spin_unlock(ptl);
			continue;
		}
		if (!huge_pte_none(pte)) {
6170
			pte_t old_pte;
6171
			unsigned int shift = huge_page_shift(hstate_vma(vma));
6172 6173

			old_pte = huge_ptep_modify_prot_start(vma, address, ptep);
6174
			pte = huge_pte_modify(old_pte, newprot);
6175
			pte = arch_make_huge_pte(pte, shift, vma->vm_flags);
6176
			huge_ptep_modify_prot_commit(vma, address, ptep, old_pte, pte);
6177
			pages++;
6178
		}
6179
		spin_unlock(ptl);
6180
	}
6181
	/*
6182
	 * Must flush TLB before releasing i_mmap_rwsem: x86's huge_pmd_unshare
6183
	 * may have cleared our pud entry and done put_page on the page table:
6184
	 * once we release i_mmap_rwsem, another task can do the final put_page
6185 6186
	 * 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.
6187
	 */
6188
	if (shared_pmd)
6189
		flush_hugetlb_tlb_range(vma, range.start, range.end);
6190 6191
	else
		flush_hugetlb_tlb_range(vma, start, end);
6192 6193 6194 6195
	/*
	 * No need to call mmu_notifier_invalidate_range() we are downgrading
	 * page table protection not changing it to point to a new page.
	 *
6196
	 * See Documentation/vm/mmu_notifier.rst
6197
	 */
6198
	i_mmap_unlock_write(vma->vm_file->f_mapping);
6199
	mmu_notifier_invalidate_range_end(&range);
6200 6201

	return pages << h->order;
6202 6203
}

6204 6205
/* Return true if reservation was successful, false otherwise.  */
bool hugetlb_reserve_pages(struct inode *inode,
6206
					long from, long to,
6207
					struct vm_area_struct *vma,
6208
					vm_flags_t vm_flags)
6209
{
6210
	long chg, add = -1;
6211
	struct hstate *h = hstate_inode(inode);
6212
	struct hugepage_subpool *spool = subpool_inode(inode);
6213
	struct resv_map *resv_map;
6214
	struct hugetlb_cgroup *h_cg = NULL;
6215
	long gbl_reserve, regions_needed = 0;
6216

6217 6218 6219
	/* This should never happen */
	if (from > to) {
		VM_WARN(1, "%s called with a negative range\n", __func__);
6220
		return false;
6221 6222
	}

6223 6224 6225
	/*
	 * Only apply hugepage reservation if asked. At fault time, an
	 * attempt will be made for VM_NORESERVE to allocate a page
6226
	 * without using reserves
6227
	 */
6228
	if (vm_flags & VM_NORESERVE)
6229
		return true;
6230

6231 6232 6233 6234 6235 6236
	/*
	 * 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
	 */
6237
	if (!vma || vma->vm_flags & VM_MAYSHARE) {
6238 6239 6240 6241 6242
		/*
		 * 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).
		 */
6243
		resv_map = inode_resv_map(inode);
6244

6245
		chg = region_chg(resv_map, from, to, &regions_needed);
6246 6247

	} else {
6248
		/* Private mapping. */
6249
		resv_map = resv_map_alloc();
6250
		if (!resv_map)
6251
			return false;
6252

6253
		chg = to - from;
6254

6255 6256 6257 6258
		set_vma_resv_map(vma, resv_map);
		set_vma_resv_flags(vma, HPAGE_RESV_OWNER);
	}

6259
	if (chg < 0)
6260
		goto out_err;
6261

6262 6263
	if (hugetlb_cgroup_charge_cgroup_rsvd(hstate_index(h),
				chg * pages_per_huge_page(h), &h_cg) < 0)
6264 6265 6266 6267 6268 6269 6270 6271 6272
		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);
	}

6273 6274 6275 6276 6277 6278
	/*
	 * 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);
6279
	if (gbl_reserve < 0)
6280
		goto out_uncharge_cgroup;
6281 6282

	/*
6283
	 * Check enough hugepages are available for the reservation.
6284
	 * Hand the pages back to the subpool if there are not
6285
	 */
6286
	if (hugetlb_acct_memory(h, gbl_reserve) < 0)
6287
		goto out_put_pages;
6288 6289 6290 6291 6292 6293 6294 6295 6296 6297 6298 6299

	/*
	 * 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
	 */
6300
	if (!vma || vma->vm_flags & VM_MAYSHARE) {
6301
		add = region_add(resv_map, from, to, regions_needed, h, h_cg);
6302 6303 6304

		if (unlikely(add < 0)) {
			hugetlb_acct_memory(h, -gbl_reserve);
6305
			goto out_put_pages;
6306
		} else if (unlikely(chg > add)) {
6307 6308 6309 6310 6311 6312 6313 6314 6315
			/*
			 * 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;

6316 6317 6318 6319
			/*
			 * hugetlb_cgroup_uncharge_cgroup_rsvd() will put the
			 * reference to h_cg->css. See comment below for detail.
			 */
6320 6321 6322 6323
			hugetlb_cgroup_uncharge_cgroup_rsvd(
				hstate_index(h),
				(chg - add) * pages_per_huge_page(h), h_cg);

6324 6325 6326
			rsv_adjust = hugepage_subpool_put_pages(spool,
								chg - add);
			hugetlb_acct_memory(h, -rsv_adjust);
6327 6328 6329 6330 6331 6332 6333 6334
		} 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);
6335 6336
		}
	}
6337 6338
	return true;

6339 6340 6341 6342 6343 6344
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);
6345
out_err:
6346
	if (!vma || vma->vm_flags & VM_MAYSHARE)
6347 6348 6349 6350 6351
		/* 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
6352 6353
	if (vma && is_vma_resv_set(vma, HPAGE_RESV_OWNER))
		kref_put(&resv_map->refs, resv_map_release);
6354
	return false;
6355 6356
}

6357 6358
long hugetlb_unreserve_pages(struct inode *inode, long start, long end,
								long freed)
6359
{
6360
	struct hstate *h = hstate_inode(inode);
6361
	struct resv_map *resv_map = inode_resv_map(inode);
6362
	long chg = 0;
6363
	struct hugepage_subpool *spool = subpool_inode(inode);
6364
	long gbl_reserve;
K
Ken Chen 已提交
6365

6366 6367 6368 6369
	/*
	 * Since this routine can be called in the evict inode path for all
	 * hugetlbfs inodes, resv_map could be NULL.
	 */
6370 6371 6372 6373 6374 6375 6376 6377 6378 6379 6380
	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
6381
	spin_lock(&inode->i_lock);
6382
	inode->i_blocks -= (blocks_per_huge_page(h) * freed);
K
Ken Chen 已提交
6383 6384
	spin_unlock(&inode->i_lock);

6385 6386 6387
	/*
	 * If the subpool has a minimum size, the number of global
	 * reservations to be released may be adjusted.
6388 6389 6390
	 *
	 * Note that !resv_map implies freed == 0. So (chg - freed)
	 * won't go negative.
6391 6392 6393
	 */
	gbl_reserve = hugepage_subpool_put_pages(spool, (chg - freed));
	hugetlb_acct_memory(h, -gbl_reserve);
6394 6395

	return 0;
6396
}
6397

6398 6399 6400 6401 6402 6403 6404 6405 6406 6407 6408
#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
6409 6410
	unsigned long vm_flags = vma->vm_flags & VM_LOCKED_CLEAR_MASK;
	unsigned long svm_flags = svma->vm_flags & VM_LOCKED_CLEAR_MASK;
6411 6412 6413 6414 6415 6416 6417

	/*
	 * match the virtual addresses, permission and the alignment of the
	 * page table page.
	 */
	if (pmd_index(addr) != pmd_index(saddr) ||
	    vm_flags != svm_flags ||
6418
	    !range_in_vma(svma, sbase, s_end))
6419 6420 6421 6422 6423
		return 0;

	return saddr;
}

6424
static bool vma_shareable(struct vm_area_struct *vma, unsigned long addr)
6425 6426 6427 6428 6429 6430 6431
{
	unsigned long base = addr & PUD_MASK;
	unsigned long end = base + PUD_SIZE;

	/*
	 * check on proper vm_flags and page table alignment
	 */
6432
	if (vma->vm_flags & VM_MAYSHARE && range_in_vma(vma, base, end))
6433 6434
		return true;
	return false;
6435 6436
}

6437 6438 6439 6440 6441 6442 6443 6444 6445
bool want_pmd_share(struct vm_area_struct *vma, unsigned long addr)
{
#ifdef CONFIG_USERFAULTFD
	if (uffd_disable_huge_pmd_share(vma))
		return false;
#endif
	return vma_shareable(vma, addr);
}

6446 6447 6448 6449 6450 6451 6452 6453
/*
 * 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)
{
6454 6455
	unsigned long v_start = ALIGN(vma->vm_start, PUD_SIZE),
		v_end = ALIGN_DOWN(vma->vm_end, PUD_SIZE);
6456

6457
	/*
I
Ingo Molnar 已提交
6458 6459
	 * vma needs to span at least one aligned PUD size, and the range
	 * must be at least partially within in.
6460 6461 6462
	 */
	if (!(vma->vm_flags & VM_MAYSHARE) || !(v_end > v_start) ||
		(*end <= v_start) || (*start >= v_end))
6463 6464
		return;

6465
	/* Extend the range to be PUD aligned for a worst case scenario */
6466 6467
	if (*start > v_start)
		*start = ALIGN_DOWN(*start, PUD_SIZE);
6468

6469 6470
	if (*end < v_end)
		*end = ALIGN(*end, PUD_SIZE);
6471 6472
}

6473 6474 6475 6476
/*
 * 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
6477 6478
 * code much cleaner.
 *
6479 6480 6481 6482
 * 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).
6483
 */
6484 6485
pte_t *huge_pmd_share(struct mm_struct *mm, struct vm_area_struct *vma,
		      unsigned long addr, pud_t *pud)
6486 6487 6488 6489 6490 6491 6492 6493
{
	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;
6494
	spinlock_t *ptl;
6495

6496
	i_mmap_assert_locked(mapping);
6497 6498 6499 6500 6501 6502
	vma_interval_tree_foreach(svma, &mapping->i_mmap, idx, idx) {
		if (svma == vma)
			continue;

		saddr = page_table_shareable(svma, vma, addr, idx);
		if (saddr) {
6503 6504
			spte = huge_pte_offset(svma->vm_mm, saddr,
					       vma_mmu_pagesize(svma));
6505 6506 6507 6508 6509 6510 6511 6512 6513 6514
			if (spte) {
				get_page(virt_to_page(spte));
				break;
			}
		}
	}

	if (!spte)
		goto out;

6515
	ptl = huge_pte_lock(hstate_vma(vma), mm, spte);
6516
	if (pud_none(*pud)) {
6517 6518
		pud_populate(mm, pud,
				(pmd_t *)((unsigned long)spte & PAGE_MASK));
6519
		mm_inc_nr_pmds(mm);
6520
	} else {
6521
		put_page(virt_to_page(spte));
6522
	}
6523
	spin_unlock(ptl);
6524 6525 6526 6527 6528 6529 6530 6531 6532 6533 6534 6535
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.
 *
6536
 * Called with page table lock held and i_mmap_rwsem held in write mode.
6537 6538 6539 6540
 *
 * returns: 1 successfully unmapped a shared pte page
 *	    0 the underlying pte page is not shared, or it is the last user
 */
6541 6542
int huge_pmd_unshare(struct mm_struct *mm, struct vm_area_struct *vma,
					unsigned long *addr, pte_t *ptep)
6543 6544
{
	pgd_t *pgd = pgd_offset(mm, *addr);
6545 6546
	p4d_t *p4d = p4d_offset(pgd, *addr);
	pud_t *pud = pud_offset(p4d, *addr);
6547

6548
	i_mmap_assert_write_locked(vma->vm_file->f_mapping);
6549 6550 6551 6552 6553 6554
	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));
6555
	mm_dec_nr_pmds(mm);
6556 6557 6558
	*addr = ALIGN(*addr, HPAGE_SIZE * PTRS_PER_PTE) - HPAGE_SIZE;
	return 1;
}
6559

6560
#else /* !CONFIG_ARCH_WANT_HUGE_PMD_SHARE */
6561 6562
pte_t *huge_pmd_share(struct mm_struct *mm, struct vm_area_struct *vma,
		      unsigned long addr, pud_t *pud)
6563 6564 6565
{
	return NULL;
}
6566

6567 6568
int huge_pmd_unshare(struct mm_struct *mm, struct vm_area_struct *vma,
				unsigned long *addr, pte_t *ptep)
6569 6570 6571
{
	return 0;
}
6572 6573 6574 6575 6576

void adjust_range_if_pmd_sharing_possible(struct vm_area_struct *vma,
				unsigned long *start, unsigned long *end)
{
}
6577 6578 6579 6580 6581

bool want_pmd_share(struct vm_area_struct *vma, unsigned long addr)
{
	return false;
}
6582 6583
#endif /* CONFIG_ARCH_WANT_HUGE_PMD_SHARE */

6584
#ifdef CONFIG_ARCH_WANT_GENERAL_HUGETLB
6585
pte_t *huge_pte_alloc(struct mm_struct *mm, struct vm_area_struct *vma,
6586 6587 6588
			unsigned long addr, unsigned long sz)
{
	pgd_t *pgd;
6589
	p4d_t *p4d;
6590 6591 6592 6593
	pud_t *pud;
	pte_t *pte = NULL;

	pgd = pgd_offset(mm, addr);
6594 6595 6596
	p4d = p4d_alloc(mm, pgd, addr);
	if (!p4d)
		return NULL;
6597
	pud = pud_alloc(mm, p4d, addr);
6598 6599 6600 6601 6602
	if (pud) {
		if (sz == PUD_SIZE) {
			pte = (pte_t *)pud;
		} else {
			BUG_ON(sz != PMD_SIZE);
6603
			if (want_pmd_share(vma, addr) && pud_none(*pud))
6604
				pte = huge_pmd_share(mm, vma, addr, pud);
6605 6606 6607 6608
			else
				pte = (pte_t *)pmd_alloc(mm, pud, addr);
		}
	}
6609
	BUG_ON(pte && pte_present(*pte) && !pte_huge(*pte));
6610 6611 6612 6613

	return pte;
}

6614 6615 6616 6617
/*
 * huge_pte_offset() - Walk the page table to resolve the hugepage
 * entry at address @addr
 *
6618 6619
 * Return: Pointer to page table entry (PUD or PMD) for
 * address @addr, or NULL if a !p*d_present() entry is encountered and the
6620 6621 6622
 * size @sz doesn't match the hugepage size at this level of the page
 * table.
 */
6623 6624
pte_t *huge_pte_offset(struct mm_struct *mm,
		       unsigned long addr, unsigned long sz)
6625 6626
{
	pgd_t *pgd;
6627
	p4d_t *p4d;
6628 6629
	pud_t *pud;
	pmd_t *pmd;
6630 6631

	pgd = pgd_offset(mm, addr);
6632 6633 6634 6635 6636
	if (!pgd_present(*pgd))
		return NULL;
	p4d = p4d_offset(pgd, addr);
	if (!p4d_present(*p4d))
		return NULL;
6637

6638
	pud = pud_offset(p4d, addr);
6639 6640
	if (sz == PUD_SIZE)
		/* must be pud huge, non-present or none */
6641
		return (pte_t *)pud;
6642
	if (!pud_present(*pud))
6643
		return NULL;
6644
	/* must have a valid entry and size to go further */
6645

6646 6647 6648
	pmd = pmd_offset(pud, addr);
	/* must be pmd huge, non-present or none */
	return (pte_t *)pmd;
6649 6650
}

6651 6652 6653 6654 6655 6656 6657 6658 6659 6660 6661 6662 6663
#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);
}

6664 6665 6666 6667 6668 6669 6670 6671
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;
}

6672
struct page * __weak
6673
follow_huge_pmd(struct mm_struct *mm, unsigned long address,
6674
		pmd_t *pmd, int flags)
6675
{
6676 6677
	struct page *page = NULL;
	spinlock_t *ptl;
6678
	pte_t pte;
J
John Hubbard 已提交
6679 6680 6681 6682 6683 6684

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

6685 6686 6687 6688 6689 6690 6691 6692 6693
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;
6694 6695
	pte = huge_ptep_get((pte_t *)pmd);
	if (pte_present(pte)) {
6696
		page = pmd_page(*pmd) + ((address & ~PMD_MASK) >> PAGE_SHIFT);
J
John Hubbard 已提交
6697 6698 6699 6700 6701 6702 6703 6704 6705 6706 6707 6708
		/*
		 * 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;
		}
6709
	} else {
6710
		if (is_hugetlb_entry_migration(pte)) {
6711 6712 6713 6714 6715 6716 6717 6718 6719 6720 6721
			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);
6722 6723 6724
	return page;
}

6725
struct page * __weak
6726
follow_huge_pud(struct mm_struct *mm, unsigned long address,
6727
		pud_t *pud, int flags)
6728
{
J
John Hubbard 已提交
6729
	if (flags & (FOLL_GET | FOLL_PIN))
6730
		return NULL;
6731

6732
	return pte_page(*(pte_t *)pud) + ((address & ~PUD_MASK) >> PAGE_SHIFT);
6733 6734
}

6735 6736 6737
struct page * __weak
follow_huge_pgd(struct mm_struct *mm, unsigned long address, pgd_t *pgd, int flags)
{
J
John Hubbard 已提交
6738
	if (flags & (FOLL_GET | FOLL_PIN))
6739 6740 6741 6742 6743
		return NULL;

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

6744 6745
bool isolate_huge_page(struct page *page, struct list_head *list)
{
6746 6747
	bool ret = true;

6748
	spin_lock_irq(&hugetlb_lock);
6749 6750
	if (!PageHeadHuge(page) ||
	    !HPageMigratable(page) ||
6751
	    !get_page_unless_zero(page)) {
6752 6753 6754
		ret = false;
		goto unlock;
	}
6755
	ClearHPageMigratable(page);
6756
	list_move_tail(&page->lru, list);
6757
unlock:
6758
	spin_unlock_irq(&hugetlb_lock);
6759
	return ret;
6760 6761
}

6762 6763 6764 6765 6766 6767 6768 6769 6770 6771
int get_hwpoison_huge_page(struct page *page, bool *hugetlb)
{
	int ret = 0;

	*hugetlb = false;
	spin_lock_irq(&hugetlb_lock);
	if (PageHeadHuge(page)) {
		*hugetlb = true;
		if (HPageFreed(page) || HPageMigratable(page))
			ret = get_page_unless_zero(page);
6772 6773
		else
			ret = -EBUSY;
6774 6775 6776 6777 6778
	}
	spin_unlock_irq(&hugetlb_lock);
	return ret;
}

6779 6780
void putback_active_hugepage(struct page *page)
{
6781
	spin_lock_irq(&hugetlb_lock);
6782
	SetHPageMigratable(page);
6783
	list_move_tail(&page->lru, &(page_hstate(page))->hugepage_activelist);
6784
	spin_unlock_irq(&hugetlb_lock);
6785 6786
	put_page(page);
}
6787 6788 6789 6790 6791 6792 6793 6794 6795 6796 6797 6798 6799 6800 6801 6802 6803 6804

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.
	 */
6805
	if (HPageTemporary(newpage)) {
6806 6807 6808
		int old_nid = page_to_nid(oldpage);
		int new_nid = page_to_nid(newpage);

6809 6810
		SetHPageTemporary(oldpage);
		ClearHPageTemporary(newpage);
6811

6812 6813 6814 6815 6816 6817
		/*
		 * There is no need to transfer the per-node surplus state
		 * when we do not cross the node.
		 */
		if (new_nid == old_nid)
			return;
6818
		spin_lock_irq(&hugetlb_lock);
6819 6820 6821 6822
		if (h->surplus_huge_pages_node[old_nid]) {
			h->surplus_huge_pages_node[old_nid]--;
			h->surplus_huge_pages_node[new_nid]++;
		}
6823
		spin_unlock_irq(&hugetlb_lock);
6824 6825
	}
}
6826

6827 6828 6829 6830 6831 6832 6833 6834 6835 6836 6837 6838 6839 6840 6841 6842 6843 6844 6845 6846 6847 6848 6849 6850 6851 6852 6853 6854 6855 6856 6857 6858 6859 6860 6861 6862 6863 6864 6865 6866 6867 6868 6869 6870 6871 6872 6873 6874 6875 6876 6877
/*
 * This function will unconditionally remove all the shared pmd pgtable entries
 * within the specific vma for a hugetlbfs memory range.
 */
void hugetlb_unshare_all_pmds(struct vm_area_struct *vma)
{
	struct hstate *h = hstate_vma(vma);
	unsigned long sz = huge_page_size(h);
	struct mm_struct *mm = vma->vm_mm;
	struct mmu_notifier_range range;
	unsigned long address, start, end;
	spinlock_t *ptl;
	pte_t *ptep;

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

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

	if (start >= end)
		return;

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

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

6878 6879 6880 6881 6882
#ifdef CONFIG_CMA
static bool cma_reserve_called __initdata;

static int __init cmdline_parse_hugetlb_cma(char *p)
{
6883 6884 6885 6886 6887 6888 6889 6890 6891 6892 6893 6894 6895 6896 6897 6898 6899 6900 6901 6902 6903 6904 6905 6906 6907 6908 6909 6910 6911 6912 6913 6914
	int nid, count = 0;
	unsigned long tmp;
	char *s = p;

	while (*s) {
		if (sscanf(s, "%lu%n", &tmp, &count) != 1)
			break;

		if (s[count] == ':') {
			nid = tmp;
			if (nid < 0 || nid >= MAX_NUMNODES)
				break;

			s += count + 1;
			tmp = memparse(s, &s);
			hugetlb_cma_size_in_node[nid] = tmp;
			hugetlb_cma_size += tmp;

			/*
			 * Skip the separator if have one, otherwise
			 * break the parsing.
			 */
			if (*s == ',')
				s++;
			else
				break;
		} else {
			hugetlb_cma_size = memparse(p, &p);
			break;
		}
	}

6915 6916 6917 6918 6919 6920 6921 6922
	return 0;
}

early_param("hugetlb_cma", cmdline_parse_hugetlb_cma);

void __init hugetlb_cma_reserve(int order)
{
	unsigned long size, reserved, per_node;
6923
	bool node_specific_cma_alloc = false;
6924 6925 6926 6927
	int nid;

	cma_reserve_called = true;

6928 6929 6930 6931 6932 6933 6934 6935 6936 6937 6938 6939 6940 6941 6942 6943 6944 6945 6946 6947 6948 6949 6950 6951 6952
	if (!hugetlb_cma_size)
		return;

	for (nid = 0; nid < MAX_NUMNODES; nid++) {
		if (hugetlb_cma_size_in_node[nid] == 0)
			continue;

		if (!node_state(nid, N_ONLINE)) {
			pr_warn("hugetlb_cma: invalid node %d specified\n", nid);
			hugetlb_cma_size -= hugetlb_cma_size_in_node[nid];
			hugetlb_cma_size_in_node[nid] = 0;
			continue;
		}

		if (hugetlb_cma_size_in_node[nid] < (PAGE_SIZE << order)) {
			pr_warn("hugetlb_cma: cma area of node %d should be at least %lu MiB\n",
				nid, (PAGE_SIZE << order) / SZ_1M);
			hugetlb_cma_size -= hugetlb_cma_size_in_node[nid];
			hugetlb_cma_size_in_node[nid] = 0;
		} else {
			node_specific_cma_alloc = true;
		}
	}

	/* Validate the CMA size again in case some invalid nodes specified. */
6953 6954 6955 6956 6957 6958
	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);
6959
		hugetlb_cma_size = 0;
6960 6961 6962
		return;
	}

6963 6964 6965 6966 6967 6968 6969 6970 6971
	if (!node_specific_cma_alloc) {
		/*
		 * 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);
	}
6972 6973 6974 6975

	reserved = 0;
	for_each_node_state(nid, N_ONLINE) {
		int res;
6976
		char name[CMA_MAX_NAME];
6977

6978 6979 6980 6981 6982 6983 6984 6985 6986
		if (node_specific_cma_alloc) {
			if (hugetlb_cma_size_in_node[nid] == 0)
				continue;

			size = hugetlb_cma_size_in_node[nid];
		} else {
			size = min(per_node, hugetlb_cma_size - reserved);
		}

6987 6988
		size = round_up(size, PAGE_SIZE << order);

6989
		snprintf(name, sizeof(name), "hugetlb%d", nid);
6990 6991 6992 6993 6994 6995 6996
		/*
		 * Note that 'order per bit' is based on smallest size that
		 * may be returned to CMA allocator in the case of
		 * huge page demotion.
		 */
		res = cma_declare_contiguous_nid(0, size, 0,
						PAGE_SIZE << HUGETLB_PAGE_ORDER,
6997
						 0, false, name,
6998 6999 7000 7001 7002 7003 7004 7005 7006 7007 7008 7009 7010 7011
						 &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;
	}
7012 7013 7014 7015 7016 7017 7018

	if (!reserved)
		/*
		 * hugetlb_cma_size is used to determine if allocations from
		 * cma are possible.  Set to zero if no cma regions are set up.
		 */
		hugetlb_cma_size = 0;
7019 7020 7021 7022 7023 7024 7025 7026 7027 7028 7029
}

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 */