hugetlb.c 156.7 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/mmdebug.h>
23
#include <linux/sched/signal.h>
24
#include <linux/rmap.h>
25
#include <linux/string_helpers.h>
26 27
#include <linux/swap.h>
#include <linux/swapops.h>
28
#include <linux/jhash.h>
29
#include <linux/numa.h>
30
#include <linux/llist.h>
31
#include <linux/cma.h>
32

D
David Gibson 已提交
33
#include <asm/page.h>
34
#include <asm/pgalloc.h>
35
#include <asm/tlb.h>
D
David Gibson 已提交
36

37
#include <linux/io.h>
D
David Gibson 已提交
38
#include <linux/hugetlb.h>
39
#include <linux/hugetlb_cgroup.h>
40
#include <linux/node.h>
41
#include <linux/userfaultfd_k.h>
42
#include <linux/page_owner.h>
43
#include "internal.h"
L
Linus Torvalds 已提交
44

45
int hugetlb_max_hstate __read_mostly;
46 47
unsigned int default_hstate_idx;
struct hstate hstates[HUGE_MAX_HSTATE];
48

49
#ifdef CONFIG_CMA
50
static struct cma *hugetlb_cma[MAX_NUMNODES];
51 52
#endif
static unsigned long hugetlb_cma_size __initdata;
53

54 55 56 57 58
/*
 * Minimum page order among possible hugepage sizes, set to a proper value
 * at boot time.
 */
static unsigned int minimum_order __read_mostly = UINT_MAX;
59

60 61
__initdata LIST_HEAD(huge_boot_pages);

62 63 64
/* for command line parsing */
static struct hstate * __initdata parsed_hstate;
static unsigned long __initdata default_hstate_max_huge_pages;
65
static bool __initdata parsed_valid_hugepagesz = true;
66
static bool __initdata parsed_default_hugepagesz;
67

68
/*
69 70
 * Protects updates to hugepage_freelists, hugepage_activelist, nr_huge_pages,
 * free_huge_pages, and surplus_huge_pages.
71
 */
72
DEFINE_SPINLOCK(hugetlb_lock);
73

74 75 76 77 78
/*
 * 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;
79
struct mutex *hugetlb_fault_mutex_table ____cacheline_aligned_in_smp;
80

81 82 83
/* Forward declaration */
static int hugetlb_acct_memory(struct hstate *h, long delta);

84 85 86 87 88 89 90
static inline void unlock_or_release_subpool(struct hugepage_subpool *spool)
{
	bool free = (spool->count == 0) && (spool->used_hpages == 0);

	spin_unlock(&spool->lock);

	/* If no pages are used, and no other handles to the subpool
E
Ethon Paul 已提交
91
	 * remain, give up any reservations based on minimum size and
92 93 94 95 96
	 * free the subpool */
	if (free) {
		if (spool->min_hpages != -1)
			hugetlb_acct_memory(spool->hstate,
						-spool->min_hpages);
97
		kfree(spool);
98
	}
99 100
}

101 102
struct hugepage_subpool *hugepage_new_subpool(struct hstate *h, long max_hpages,
						long min_hpages)
103 104 105
{
	struct hugepage_subpool *spool;

106
	spool = kzalloc(sizeof(*spool), GFP_KERNEL);
107 108 109 110 111
	if (!spool)
		return NULL;

	spin_lock_init(&spool->lock);
	spool->count = 1;
112 113 114 115 116 117 118 119 120
	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;
121 122 123 124 125 126 127 128 129 130 131 132

	return spool;
}

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

133 134 135
/*
 * Subpool accounting for allocating and reserving pages.
 * Return -ENOMEM if there are not enough resources to satisfy the
136
 * request.  Otherwise, return the number of pages by which the
137 138
 * 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
139
 * a subpool minimum size must be maintained.
140 141
 */
static long hugepage_subpool_get_pages(struct hugepage_subpool *spool,
142 143
				      long delta)
{
144
	long ret = delta;
145 146

	if (!spool)
147
		return ret;
148 149

	spin_lock(&spool->lock);
150 151 152 153 154 155 156 157

	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;
		}
158 159
	}

160 161
	/* minimum size accounting */
	if (spool->min_hpages != -1 && spool->rsv_hpages) {
162 163 164 165 166 167 168 169 170 171 172 173 174 175 176
		if (delta > spool->rsv_hpages) {
			/*
			 * Asking for more reserves than those already taken on
			 * behalf of subpool.  Return difference.
			 */
			ret = delta - spool->rsv_hpages;
			spool->rsv_hpages = 0;
		} else {
			ret = 0;	/* reserves already accounted for */
			spool->rsv_hpages -= delta;
		}
	}

unlock_ret:
	spin_unlock(&spool->lock);
177 178 179
	return ret;
}

180 181 182 183 184 185 186
/*
 * 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,
187 188
				       long delta)
{
189 190
	long ret = delta;

191
	if (!spool)
192
		return delta;
193 194

	spin_lock(&spool->lock);
195 196 197 198

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

199 200
	 /* minimum size accounting */
	if (spool->min_hpages != -1 && spool->used_hpages < spool->min_hpages) {
201 202 203 204 205 206 207 208 209 210 211 212 213 214
		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.
	 */
215
	unlock_or_release_subpool(spool);
216 217

	return ret;
218 219 220 221 222 223 224 225 226
}

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
227
	return subpool_inode(file_inode(vma->vm_file));
228 229
}

230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250
/* 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);
	VM_BUG_ON(!nrg);
	list_del(&nrg->link);

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

	return nrg;
}

251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285
static void copy_hugetlb_cgroup_uncharge_info(struct file_region *nrg,
					      struct file_region *rg)
{
#ifdef CONFIG_CGROUP_HUGETLB
	nrg->reservation_counter = rg->reservation_counter;
	nrg->css = rg->css;
	if (rg->css)
		css_get(rg->css);
#endif
}

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

286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327
static bool has_same_uncharge_info(struct file_region *rg,
				   struct file_region *org)
{
#ifdef CONFIG_CGROUP_HUGETLB
	return rg && org &&
	       rg->reservation_counter == org->reservation_counter &&
	       rg->css == org->css;

#else
	return true;
#endif
}

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

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

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

		coalesce_file_region(resv, prg);
		return;
	}

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

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

		coalesce_file_region(resv, nrg);
		return;
	}
}

M
Mina Almasry 已提交
328 329
/* Must be called with resv->lock held. Calling this with count_only == true
 * will count the number of pages to be added but will not modify the linked
330 331 332
 * list. If regions_needed != NULL and count_only == true, then 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
333 334
 */
static long add_reservation_in_range(struct resv_map *resv, long f, long t,
335 336 337
				     struct hugetlb_cgroup *h_cg,
				     struct hstate *h, long *regions_needed,
				     bool count_only)
M
Mina Almasry 已提交
338
{
339
	long add = 0;
M
Mina Almasry 已提交
340
	struct list_head *head = &resv->regions;
341
	long last_accounted_offset = f;
M
Mina Almasry 已提交
342 343
	struct file_region *rg = NULL, *trg = NULL, *nrg = NULL;

344 345
	if (regions_needed)
		*regions_needed = 0;
M
Mina Almasry 已提交
346

347 348 349 350 351 352 353 354 355 356 357 358 359 360
	/* 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
361

362 363 364
		/* When we find a region that starts beyond our range, we've
		 * finished.
		 */
M
Mina Almasry 已提交
365 366 367
		if (rg->from > t)
			break;

368 369 370 371 372 373 374 375
		/* Add an entry for last_accounted_offset -> rg->from, and
		 * update last_accounted_offset.
		 */
		if (rg->from > last_accounted_offset) {
			add += rg->from - last_accounted_offset;
			if (!count_only) {
				nrg = get_file_region_entry_from_cache(
					resv, last_accounted_offset, rg->from);
376 377
				record_hugetlb_cgroup_uncharge_info(h_cg, h,
								    resv, nrg);
378
				list_add(&nrg->link, rg->link.prev);
379
				coalesce_file_region(resv, nrg);
380 381 382 383 384 385 386 387 388 389 390 391 392 393 394
			} else if (regions_needed)
				*regions_needed += 1;
		}

		last_accounted_offset = rg->to;
	}

	/* Handle the case where our range extends beyond
	 * last_accounted_offset.
	 */
	if (last_accounted_offset < t) {
		add += t - last_accounted_offset;
		if (!count_only) {
			nrg = get_file_region_entry_from_cache(
				resv, last_accounted_offset, t);
395
			record_hugetlb_cgroup_uncharge_info(h_cg, h, resv, nrg);
396
			list_add(&nrg->link, rg->link.prev);
397
			coalesce_file_region(resv, nrg);
398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436
		} else if (regions_needed)
			*regions_needed += 1;
	}

	VM_BUG_ON(add < 0);
	return add;
}

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

	VM_BUG_ON(regions_needed < 0);

	INIT_LIST_HEAD(&allocated_regions);

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

		/* At this point, we should have enough entries in the cache
		 * for all the existings adds_in_progress. We should only be
		 * needing to allocate for regions_needed.
M
Mina Almasry 已提交
437
		 */
438 439 440 441 442 443 444 445
		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
446 447
		}

448 449 450
		spin_lock(&resv->lock);

		list_for_each_entry_safe(rg, trg, &allocated_regions, link) {
M
Mina Almasry 已提交
451
			list_del(&rg->link);
452 453
			list_add(&rg->link, &resv->region_cache);
			resv->region_cache_count++;
M
Mina Almasry 已提交
454 455 456
		}
	}

457
	return 0;
M
Mina Almasry 已提交
458

459 460 461 462 463 464
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
465 466
}

467 468
/*
 * Add the huge page range represented by [f, t) to the reserve
469 470 471 472 473
 * 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.
474
 *
475 476 477 478
 * 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
479
 * this operation and we were not able to allocate, it returns -ENOMEM.
480 481 482
 * 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.
483
 */
484
static long region_add(struct resv_map *resv, long f, long t,
485 486
		       long in_regions_needed, struct hstate *h,
		       struct hugetlb_cgroup *h_cg)
487
{
488
	long add = 0, actual_regions_needed = 0;
489

490
	spin_lock(&resv->lock);
491 492 493
retry:

	/* Count how many regions are actually needed to execute this add. */
494 495
	add_reservation_in_range(resv, f, t, NULL, NULL, &actual_regions_needed,
				 true);
496

497
	/*
498 499 500 501 502 503 504
	 * 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.
505
	 */
506 507 508 509 510 511 512 513
	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);
514

515 516 517 518
		if (allocate_file_region_entries(
			    resv, actual_regions_needed - in_regions_needed)) {
			return -ENOMEM;
		}
519

520
		goto retry;
521 522
	}

523
	add = add_reservation_in_range(resv, f, t, h_cg, h, NULL, false);
524 525

	resv->adds_in_progress -= in_regions_needed;
526

527
	spin_unlock(&resv->lock);
528 529
	VM_BUG_ON(add < 0);
	return add;
530 531
}

532 533 534 535 536 537 538
/*
 * 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
539 540 541 542 543 544 545
 * 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.
546 547 548 549 550
 *
 * 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.
551
 */
552 553
static long region_chg(struct resv_map *resv, long f, long t,
		       long *out_regions_needed)
554 555 556
{
	long chg = 0;

557
	spin_lock(&resv->lock);
558

559
	/* Count how many hugepages in this range are NOT respresented. */
560 561
	chg = add_reservation_in_range(resv, f, t, NULL, NULL,
				       out_regions_needed, true);
562

563 564
	if (*out_regions_needed == 0)
		*out_regions_needed = 1;
565

566 567
	if (allocate_file_region_entries(resv, *out_regions_needed))
		return -ENOMEM;
568

569
	resv->adds_in_progress += *out_regions_needed;
570 571

	spin_unlock(&resv->lock);
572 573 574
	return chg;
}

575 576 577 578 579
/*
 * 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
580 581 582
 * 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.
583 584 585 586 587
 *
 * 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.
 */
588 589
static void region_abort(struct resv_map *resv, long f, long t,
			 long regions_needed)
590 591 592
{
	spin_lock(&resv->lock);
	VM_BUG_ON(!resv->region_cache_count);
593
	resv->adds_in_progress -= regions_needed;
594 595 596
	spin_unlock(&resv->lock);
}

597
/*
598 599 600 601 602 603 604 605 606 607 608 609
 * 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.
610
 */
611
static long region_del(struct resv_map *resv, long f, long t)
612
{
613
	struct list_head *head = &resv->regions;
614
	struct file_region *rg, *trg;
615 616
	struct file_region *nrg = NULL;
	long del = 0;
617

618
retry:
619
	spin_lock(&resv->lock);
620
	list_for_each_entry_safe(rg, trg, head, link) {
621 622 623 624 625 626 627 628
		/*
		 * 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))
629
			continue;
630

631
		if (rg->from >= t)
632 633
			break;

634 635 636 637 638 639 640 641 642 643 644 645 646
		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--;
			}
647

648 649 650 651 652 653 654 655 656 657 658 659 660
			if (!nrg) {
				spin_unlock(&resv->lock);
				nrg = kmalloc(sizeof(*nrg), GFP_KERNEL);
				if (!nrg)
					return -ENOMEM;
				goto retry;
			}

			del += t - f;

			/* New entry for end of split region */
			nrg->from = t;
			nrg->to = rg->to;
661 662 663

			copy_hugetlb_cgroup_uncharge_info(nrg, rg);

664 665 666 667 668
			INIT_LIST_HEAD(&nrg->link);

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

669 670 671
			hugetlb_cgroup_uncharge_file_region(
				resv, rg, nrg->to - nrg->from);

672 673
			list_add(&nrg->link, &rg->link);
			nrg = NULL;
674
			break;
675 676 677 678
		}

		if (f <= rg->from && t >= rg->to) { /* Remove entire region */
			del += rg->to - rg->from;
679 680
			hugetlb_cgroup_uncharge_file_region(resv, rg,
							    rg->to - rg->from);
681 682 683 684 685 686 687 688
			list_del(&rg->link);
			kfree(rg);
			continue;
		}

		if (f <= rg->from) {	/* Trim beginning of region */
			del += t - rg->from;
			rg->from = t;
689 690 691

			hugetlb_cgroup_uncharge_file_region(resv, rg,
							    t - rg->from);
692 693 694
		} else {		/* Trim end of region */
			del += rg->to - f;
			rg->to = f;
695 696 697

			hugetlb_cgroup_uncharge_file_region(resv, rg,
							    rg->to - f);
698
		}
699
	}
700 701

	spin_unlock(&resv->lock);
702 703
	kfree(nrg);
	return del;
704 705
}

706 707 708 709 710 711 712 713 714
/*
 * 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.
 */
715
void hugetlb_fix_reserve_counts(struct inode *inode)
716 717 718 719 720
{
	struct hugepage_subpool *spool = subpool_inode(inode);
	long rsv_adjust;

	rsv_adjust = hugepage_subpool_get_pages(spool, 1);
721
	if (rsv_adjust) {
722 723 724 725 726 727
		struct hstate *h = hstate_inode(inode);

		hugetlb_acct_memory(h, 1);
	}
}

728 729 730 731
/*
 * Count and return the number of huge pages in the reserve map
 * that intersect with the range [f, t).
 */
732
static long region_count(struct resv_map *resv, long f, long t)
733
{
734
	struct list_head *head = &resv->regions;
735 736 737
	struct file_region *rg;
	long chg = 0;

738
	spin_lock(&resv->lock);
739 740
	/* Locate each segment we overlap with, and count that overlap. */
	list_for_each_entry(rg, head, link) {
741 742
		long seg_from;
		long seg_to;
743 744 745 746 747 748 749 750 751 752 753

		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;
	}
754
	spin_unlock(&resv->lock);
755 756 757 758

	return chg;
}

759 760 761 762
/*
 * Convert the address within this vma to the page offset within
 * the mapping, in pagecache page units; huge pages here.
 */
763 764
static pgoff_t vma_hugecache_offset(struct hstate *h,
			struct vm_area_struct *vma, unsigned long address)
765
{
766 767
	return ((address - vma->vm_start) >> huge_page_shift(h)) +
			(vma->vm_pgoff >> huge_page_order(h));
768 769
}

770 771 772 773 774
pgoff_t linear_hugepage_index(struct vm_area_struct *vma,
				     unsigned long address)
{
	return vma_hugecache_offset(hstate_vma(vma), vma, address);
}
775
EXPORT_SYMBOL_GPL(linear_hugepage_index);
776

777 778 779 780 781 782
/*
 * 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)
{
783 784 785
	if (vma->vm_ops && vma->vm_ops->pagesize)
		return vma->vm_ops->pagesize(vma);
	return PAGE_SIZE;
786
}
787
EXPORT_SYMBOL_GPL(vma_kernel_pagesize);
788

789 790 791
/*
 * 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
792 793
 * architectures where it differs, an architecture-specific 'strong'
 * version of this symbol is required.
794
 */
795
__weak unsigned long vma_mmu_pagesize(struct vm_area_struct *vma)
796 797 798 799
{
	return vma_kernel_pagesize(vma);
}

800 801 802 803 804 805 806
/*
 * 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)
807
#define HPAGE_RESV_MASK (HPAGE_RESV_OWNER | HPAGE_RESV_UNMAPPED)
808

809 810 811 812 813 814 815 816 817
/*
 * 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.
818 819 820 821 822 823 824 825 826
 *
 * 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.
827
 */
828 829 830 831 832 833 834 835 836 837 838
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;
}

839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857
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
}

858
struct resv_map *resv_map_alloc(void)
859 860
{
	struct resv_map *resv_map = kmalloc(sizeof(*resv_map), GFP_KERNEL);
861 862 863 864 865
	struct file_region *rg = kmalloc(sizeof(*rg), GFP_KERNEL);

	if (!resv_map || !rg) {
		kfree(resv_map);
		kfree(rg);
866
		return NULL;
867
	}
868 869

	kref_init(&resv_map->refs);
870
	spin_lock_init(&resv_map->lock);
871 872
	INIT_LIST_HEAD(&resv_map->regions);

873
	resv_map->adds_in_progress = 0;
874 875 876 877 878 879 880
	/*
	 * 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);
881 882 883 884 885

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

886 887 888
	return resv_map;
}

889
void resv_map_release(struct kref *ref)
890 891
{
	struct resv_map *resv_map = container_of(ref, struct resv_map, refs);
892 893
	struct list_head *head = &resv_map->region_cache;
	struct file_region *rg, *trg;
894 895

	/* Clear out any active regions before we release the map. */
896
	region_del(resv_map, 0, LONG_MAX);
897 898 899 900 901 902 903 904 905

	/* ... 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);

906 907 908
	kfree(resv_map);
}

909 910
static inline struct resv_map *inode_resv_map(struct inode *inode)
{
911 912 913 914 915 916 917 918 919
	/*
	 * 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;
920 921
}

922
static struct resv_map *vma_resv_map(struct vm_area_struct *vma)
923
{
924
	VM_BUG_ON_VMA(!is_vm_hugetlb_page(vma), vma);
925 926 927 928 929 930 931
	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 {
932 933
		return (struct resv_map *)(get_vma_private_data(vma) &
							~HPAGE_RESV_MASK);
934
	}
935 936
}

937
static void set_vma_resv_map(struct vm_area_struct *vma, struct resv_map *map)
938
{
939 940
	VM_BUG_ON_VMA(!is_vm_hugetlb_page(vma), vma);
	VM_BUG_ON_VMA(vma->vm_flags & VM_MAYSHARE, vma);
941

942 943
	set_vma_private_data(vma, (get_vma_private_data(vma) &
				HPAGE_RESV_MASK) | (unsigned long)map);
944 945 946 947
}

static void set_vma_resv_flags(struct vm_area_struct *vma, unsigned long flags)
{
948 949
	VM_BUG_ON_VMA(!is_vm_hugetlb_page(vma), vma);
	VM_BUG_ON_VMA(vma->vm_flags & VM_MAYSHARE, vma);
950 951

	set_vma_private_data(vma, get_vma_private_data(vma) | flags);
952 953 954 955
}

static int is_vma_resv_set(struct vm_area_struct *vma, unsigned long flag)
{
956
	VM_BUG_ON_VMA(!is_vm_hugetlb_page(vma), vma);
957 958

	return (get_vma_private_data(vma) & flag) != 0;
959 960
}

961
/* Reset counters to 0 and clear all HPAGE_RESV_* flags */
962 963
void reset_vma_resv_huge_pages(struct vm_area_struct *vma)
{
964
	VM_BUG_ON_VMA(!is_vm_hugetlb_page(vma), vma);
965
	if (!(vma->vm_flags & VM_MAYSHARE))
966 967 968 969
		vma->vm_private_data = (void *)0;
}

/* Returns true if the VMA has associated reserve pages */
970
static bool vma_has_reserves(struct vm_area_struct *vma, long chg)
971
{
972 973 974 975 976 977 978 979 980 981 982
	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)
983
			return true;
984
		else
985
			return false;
986
	}
987 988

	/* Shared mappings always use reserves */
989 990 991 992 993
	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
994
		 * fallocate.  In this case, there really are no reserves to
995 996 997 998 999 1000 1001
		 * use.  This situation is indicated if chg != 0.
		 */
		if (chg)
			return false;
		else
			return true;
	}
1002 1003 1004 1005 1006

	/*
	 * Only the process that called mmap() has reserves for
	 * private mappings.
	 */
1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027
	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;
	}
1028

1029
	return false;
1030 1031
}

1032
static void enqueue_huge_page(struct hstate *h, struct page *page)
L
Linus Torvalds 已提交
1033 1034
{
	int nid = page_to_nid(page);
1035
	list_move(&page->lru, &h->hugepage_freelists[nid]);
1036 1037
	h->free_huge_pages++;
	h->free_huge_pages_node[nid]++;
L
Linus Torvalds 已提交
1038 1039
}

1040
static struct page *dequeue_huge_page_node_exact(struct hstate *h, int nid)
1041 1042 1043
{
	struct page *page;

1044
	list_for_each_entry(page, &h->hugepage_freelists[nid], lru)
1045
		if (!PageHWPoison(page))
1046 1047 1048 1049 1050 1051
			break;
	/*
	 * if 'non-isolated free hugepage' not found on the list,
	 * the allocation fails.
	 */
	if (&h->hugepage_freelists[nid] == &page->lru)
1052
		return NULL;
1053
	list_move(&page->lru, &h->hugepage_activelist);
1054
	set_page_refcounted(page);
1055 1056 1057 1058 1059
	h->free_huge_pages--;
	h->free_huge_pages_node[nid]--;
	return page;
}

1060 1061
static struct page *dequeue_huge_page_nodemask(struct hstate *h, gfp_t gfp_mask, int nid,
		nodemask_t *nmask)
1062
{
1063 1064 1065 1066
	unsigned int cpuset_mems_cookie;
	struct zonelist *zonelist;
	struct zone *zone;
	struct zoneref *z;
1067
	int node = NUMA_NO_NODE;
1068

1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084
	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);
1085 1086 1087 1088 1089

		page = dequeue_huge_page_node_exact(h, node);
		if (page)
			return page;
	}
1090 1091 1092
	if (unlikely(read_mems_allowed_retry(cpuset_mems_cookie)))
		goto retry_cpuset;

1093 1094 1095
	return NULL;
}

1096 1097 1098
/* Movability of hugepages depends on migration support. */
static inline gfp_t htlb_alloc_mask(struct hstate *h)
{
1099
	if (hugepage_movable_supported(h))
1100 1101 1102 1103 1104
		return GFP_HIGHUSER_MOVABLE;
	else
		return GFP_HIGHUSER;
}

1105 1106
static struct page *dequeue_huge_page_vma(struct hstate *h,
				struct vm_area_struct *vma,
1107 1108
				unsigned long address, int avoid_reserve,
				long chg)
L
Linus Torvalds 已提交
1109
{
1110
	struct page *page;
1111
	struct mempolicy *mpol;
1112
	gfp_t gfp_mask;
1113
	nodemask_t *nodemask;
1114
	int nid;
L
Linus Torvalds 已提交
1115

1116 1117 1118 1119 1120
	/*
	 * 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
	 */
1121
	if (!vma_has_reserves(vma, chg) &&
1122
			h->free_huge_pages - h->resv_huge_pages == 0)
1123
		goto err;
1124

1125
	/* If reserves cannot be used, ensure enough pages are in the pool */
1126
	if (avoid_reserve && h->free_huge_pages - h->resv_huge_pages == 0)
1127
		goto err;
1128

1129 1130
	gfp_mask = htlb_alloc_mask(h);
	nid = huge_node(vma, address, gfp_mask, &mpol, &nodemask);
1131 1132 1133 1134
	page = dequeue_huge_page_nodemask(h, gfp_mask, nid, nodemask);
	if (page && !avoid_reserve && vma_has_reserves(vma, chg)) {
		SetPagePrivate(page);
		h->resv_huge_pages--;
L
Linus Torvalds 已提交
1135
	}
1136

1137
	mpol_cond_put(mpol);
L
Linus Torvalds 已提交
1138
	return page;
1139 1140 1141

err:
	return NULL;
L
Linus Torvalds 已提交
1142 1143
}

1144 1145 1146 1147 1148 1149 1150 1151 1152
/*
 * 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)
{
1153
	nid = next_node_in(nid, *nodes_allowed);
1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214
	VM_BUG_ON(nid >= MAX_NUMNODES);

	return nid;
}

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

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

	VM_BUG_ON(!nodes_allowed);

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

	return nid;
}

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

	VM_BUG_ON(!nodes_allowed);

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

	return nid;
}

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

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

1215
#ifdef CONFIG_ARCH_HAS_GIGANTIC_PAGE
1216
static void destroy_compound_gigantic_page(struct page *page,
1217
					unsigned int order)
1218 1219 1220 1221 1222
{
	int i;
	int nr_pages = 1 << order;
	struct page *p = page + 1;

1223
	atomic_set(compound_mapcount_ptr(page), 0);
1224 1225 1226
	if (hpage_pincount_available(page))
		atomic_set(compound_pincount_ptr(page), 0);

1227
	for (i = 1; i < nr_pages; i++, p = mem_map_next(p, page, i)) {
1228
		clear_compound_head(p);
1229 1230 1231 1232 1233 1234 1235
		set_page_refcounted(p);
	}

	set_compound_order(page, 0);
	__ClearPageHead(page);
}

1236
static void free_gigantic_page(struct page *page, unsigned int order)
1237
{
1238 1239 1240 1241
	/*
	 * If the page isn't allocated using the cma allocator,
	 * cma_release() returns false.
	 */
1242 1243
#ifdef CONFIG_CMA
	if (cma_release(hugetlb_cma[page_to_nid(page)], page, 1 << order))
1244
		return;
1245
#endif
1246

1247 1248 1249
	free_contig_range(page_to_pfn(page), 1 << order);
}

1250
#ifdef CONFIG_CONTIG_ALLOC
1251 1252
static struct page *alloc_gigantic_page(struct hstate *h, gfp_t gfp_mask,
		int nid, nodemask_t *nodemask)
1253
{
1254
	unsigned long nr_pages = 1UL << huge_page_order(h);
1255

1256 1257
#ifdef CONFIG_CMA
	{
1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270
		struct page *page;
		int node;

		for_each_node_mask(node, *nodemask) {
			if (!hugetlb_cma[node])
				continue;

			page = cma_alloc(hugetlb_cma[node], nr_pages,
					 huge_page_order(h), true);
			if (page)
				return page;
		}
	}
1271
#endif
1272

1273
	return alloc_contig_pages(nr_pages, gfp_mask, nid, nodemask);
1274 1275 1276
}

static void prep_new_huge_page(struct hstate *h, struct page *page, int nid);
1277
static void prep_compound_gigantic_page(struct page *page, unsigned int order);
1278 1279 1280 1281 1282 1283 1284
#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 */
1285

1286
#else /* !CONFIG_ARCH_HAS_GIGANTIC_PAGE */
1287
static struct page *alloc_gigantic_page(struct hstate *h, gfp_t gfp_mask,
1288 1289 1290 1291
					int nid, nodemask_t *nodemask)
{
	return NULL;
}
1292
static inline void free_gigantic_page(struct page *page, unsigned int order) { }
1293
static inline void destroy_compound_gigantic_page(struct page *page,
1294
						unsigned int order) { }
1295 1296
#endif

1297
static void update_and_free_page(struct hstate *h, struct page *page)
A
Adam Litke 已提交
1298 1299
{
	int i;
1300

1301
	if (hstate_is_gigantic(h) && !gigantic_page_runtime_supported())
1302
		return;
1303

1304 1305 1306
	h->nr_huge_pages--;
	h->nr_huge_pages_node[page_to_nid(page)]--;
	for (i = 0; i < pages_per_huge_page(h); i++) {
1307 1308
		page[i].flags &= ~(1 << PG_locked | 1 << PG_error |
				1 << PG_referenced | 1 << PG_dirty |
1309 1310
				1 << PG_active | 1 << PG_private |
				1 << PG_writeback);
A
Adam Litke 已提交
1311
	}
1312
	VM_BUG_ON_PAGE(hugetlb_cgroup_from_page(page), page);
1313
	VM_BUG_ON_PAGE(hugetlb_cgroup_from_page_rsvd(page), page);
1314
	set_compound_page_dtor(page, NULL_COMPOUND_DTOR);
A
Adam Litke 已提交
1315
	set_page_refcounted(page);
1316
	if (hstate_is_gigantic(h)) {
1317 1318 1319 1320 1321
		/*
		 * Temporarily drop the hugetlb_lock, because
		 * we might block in free_gigantic_page().
		 */
		spin_unlock(&hugetlb_lock);
1322 1323
		destroy_compound_gigantic_page(page, huge_page_order(h));
		free_gigantic_page(page, huge_page_order(h));
1324
		spin_lock(&hugetlb_lock);
1325 1326 1327
	} else {
		__free_pages(page, huge_page_order(h));
	}
A
Adam Litke 已提交
1328 1329
}

1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340
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;
}

1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365
/*
 * Test to determine whether the hugepage is "active/in-use" (i.e. being linked
 * to hstate->hugepage_activelist.)
 *
 * This function can be called for tail pages, but never returns true for them.
 */
bool page_huge_active(struct page *page)
{
	VM_BUG_ON_PAGE(!PageHuge(page), page);
	return PageHead(page) && PagePrivate(&page[1]);
}

/* never called for tail page */
static void set_page_huge_active(struct page *page)
{
	VM_BUG_ON_PAGE(!PageHeadHuge(page), page);
	SetPagePrivate(&page[1]);
}

static void clear_page_huge_active(struct page *page)
{
	VM_BUG_ON_PAGE(!PageHeadHuge(page), page);
	ClearPagePrivate(&page[1]);
}

1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387
/*
 * Internal hugetlb specific page flag. Do not use outside of the hugetlb
 * code
 */
static inline bool PageHugeTemporary(struct page *page)
{
	if (!PageHuge(page))
		return false;

	return (unsigned long)page[2].mapping == -1U;
}

static inline void SetPageHugeTemporary(struct page *page)
{
	page[2].mapping = (void *)-1U;
}

static inline void ClearPageHugeTemporary(struct page *page)
{
	page[2].mapping = NULL;
}

1388
static void __free_huge_page(struct page *page)
1389
{
1390 1391 1392 1393
	/*
	 * Can't pass hstate in here because it is called from the
	 * compound page destructor.
	 */
1394
	struct hstate *h = page_hstate(page);
1395
	int nid = page_to_nid(page);
1396 1397
	struct hugepage_subpool *spool =
		(struct hugepage_subpool *)page_private(page);
1398
	bool restore_reserve;
1399

1400 1401
	VM_BUG_ON_PAGE(page_count(page), page);
	VM_BUG_ON_PAGE(page_mapcount(page), page);
1402 1403 1404

	set_page_private(page, 0);
	page->mapping = NULL;
1405
	restore_reserve = PagePrivate(page);
1406
	ClearPagePrivate(page);
1407

1408
	/*
1409 1410 1411 1412 1413 1414
	 * If PagePrivate() was set on page, page allocation consumed a
	 * reservation.  If the page was associated with a subpool, there
	 * would have been a page reserved in the subpool before allocation
	 * via hugepage_subpool_get_pages().  Since we are 'restoring' the
	 * reservtion, do not call hugepage_subpool_put_pages() as this will
	 * remove the reserved page from the subpool.
1415
	 */
1416 1417 1418 1419 1420 1421 1422 1423 1424 1425
	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;
	}
1426

1427
	spin_lock(&hugetlb_lock);
1428
	clear_page_huge_active(page);
1429 1430
	hugetlb_cgroup_uncharge_page(hstate_index(h),
				     pages_per_huge_page(h), page);
1431 1432
	hugetlb_cgroup_uncharge_page_rsvd(hstate_index(h),
					  pages_per_huge_page(h), page);
1433 1434 1435
	if (restore_reserve)
		h->resv_huge_pages++;

1436 1437 1438 1439 1440
	if (PageHugeTemporary(page)) {
		list_del(&page->lru);
		ClearPageHugeTemporary(page);
		update_and_free_page(h, page);
	} else if (h->surplus_huge_pages_node[nid]) {
1441 1442
		/* remove the page from active list */
		list_del(&page->lru);
1443 1444 1445
		update_and_free_page(h, page);
		h->surplus_huge_pages--;
		h->surplus_huge_pages_node[nid]--;
1446
	} else {
1447
		arch_clear_hugepage_flags(page);
1448
		enqueue_huge_page(h, page);
1449
	}
1450 1451 1452
	spin_unlock(&hugetlb_lock);
}

1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500
/*
 * As free_huge_page() can be called from a non-task context, we have
 * to defer the actual freeing in a workqueue to prevent potential
 * hugetlb_lock deadlock.
 *
 * free_hpage_workfn() locklessly retrieves the linked list of pages to
 * be freed and frees them one-by-one. As the page->mapping pointer is
 * going to be cleared in __free_huge_page() anyway, it is reused as the
 * llist_node structure of a lockless linked list of huge pages to be freed.
 */
static LLIST_HEAD(hpage_freelist);

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

	node = llist_del_all(&hpage_freelist);

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

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

	__free_huge_page(page);
}

1501
static void prep_new_huge_page(struct hstate *h, struct page *page, int nid)
1502
{
1503
	INIT_LIST_HEAD(&page->lru);
1504
	set_compound_page_dtor(page, HUGETLB_PAGE_DTOR);
1505
	spin_lock(&hugetlb_lock);
1506
	set_hugetlb_cgroup(page, NULL);
1507
	set_hugetlb_cgroup_rsvd(page, NULL);
1508 1509
	h->nr_huge_pages++;
	h->nr_huge_pages_node[nid]++;
1510 1511 1512
	spin_unlock(&hugetlb_lock);
}

1513
static void prep_compound_gigantic_page(struct page *page, unsigned int order)
1514 1515 1516 1517 1518 1519 1520
{
	int i;
	int nr_pages = 1 << order;
	struct page *p = page + 1;

	/* we rely on prep_new_huge_page to set the destructor */
	set_compound_order(page, order);
1521
	__ClearPageReserved(page);
1522
	__SetPageHead(page);
1523
	for (i = 1; i < nr_pages; i++, p = mem_map_next(p, page, i)) {
1524 1525 1526 1527
		/*
		 * 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
1528
		 * too.  Otherwise drivers using get_user_pages() to access tail
1529 1530 1531 1532 1533 1534 1535 1536
		 * 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);
1537
		set_page_count(p, 0);
1538
		set_compound_head(p, page);
1539
	}
1540
	atomic_set(compound_mapcount_ptr(page), -1);
1541 1542 1543

	if (hpage_pincount_available(page))
		atomic_set(compound_pincount_ptr(page), 0);
1544 1545
}

A
Andrew Morton 已提交
1546 1547 1548 1549 1550
/*
 * PageHuge() only returns true for hugetlbfs pages, but not for normal or
 * transparent huge pages.  See the PageTransHuge() documentation for more
 * details.
 */
1551 1552 1553 1554 1555 1556
int PageHuge(struct page *page)
{
	if (!PageCompound(page))
		return 0;

	page = compound_head(page);
1557
	return page[1].compound_dtor == HUGETLB_PAGE_DTOR;
1558
}
1559 1560
EXPORT_SYMBOL_GPL(PageHuge);

1561 1562 1563 1564 1565 1566 1567 1568 1569
/*
 * 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;

1570
	return page_head[1].compound_dtor == HUGETLB_PAGE_DTOR;
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
/*
 * Find address_space associated with hugetlbfs page.
 * Upon entry page is locked and page 'was' mapped although mapped state
 * could change.  If necessary, use anon_vma to find vma and associated
 * address space.  The returned mapping may be stale, but it can not be
 * invalid as page lock (which is held) is required to destroy mapping.
 */
static struct address_space *_get_hugetlb_page_mapping(struct page *hpage)
{
	struct anon_vma *anon_vma;
	pgoff_t pgoff_start, pgoff_end;
	struct anon_vma_chain *avc;
	struct address_space *mapping = page_mapping(hpage);

	/* Simple file based mapping */
	if (mapping)
		return mapping;

	/*
	 * Even anonymous hugetlbfs mappings are associated with an
	 * underlying hugetlbfs file (see hugetlb_file_setup in mmap
	 * code).  Find a vma associated with the anonymous vma, and
	 * use the file pointer to get address_space.
	 */
	anon_vma = page_lock_anon_vma_read(hpage);
	if (!anon_vma)
		return mapping;  /* NULL */

	/* Use first found vma */
	pgoff_start = page_to_pgoff(hpage);
1603
	pgoff_end = pgoff_start + pages_per_huge_page(page_hstate(hpage)) - 1;
1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622 1623 1624 1625 1626 1627 1628 1629 1630 1631 1632 1633 1634 1635 1636 1637 1638 1639 1640 1641 1642 1643 1644 1645 1646 1647 1648 1649 1650 1651 1652 1653 1654 1655 1656 1657 1658 1659 1660 1661 1662 1663 1664 1665 1666 1667 1668 1669 1670 1671 1672
	anon_vma_interval_tree_foreach(avc, &anon_vma->rb_root,
					pgoff_start, pgoff_end) {
		struct vm_area_struct *vma = avc->vma;

		mapping = vma->vm_file->f_mapping;
		break;
	}

	anon_vma_unlock_read(anon_vma);
	return mapping;
}

/*
 * Find and lock address space (mapping) in write mode.
 *
 * Upon entry, the page is locked which allows us to find the mapping
 * even in the case of an anon page.  However, locking order dictates
 * the i_mmap_rwsem be acquired BEFORE the page lock.  This is hugetlbfs
 * specific.  So, we first try to lock the sema while still holding the
 * page lock.  If this works, great!  If not, then we need to drop the
 * page lock and then acquire i_mmap_rwsem and reacquire page lock.  Of
 * course, need to revalidate state along the way.
 */
struct address_space *hugetlb_page_mapping_lock_write(struct page *hpage)
{
	struct address_space *mapping, *mapping2;

	mapping = _get_hugetlb_page_mapping(hpage);
retry:
	if (!mapping)
		return mapping;

	/*
	 * If no contention, take lock and return
	 */
	if (i_mmap_trylock_write(mapping))
		return mapping;

	/*
	 * Must drop page lock and wait on mapping sema.
	 * Note:  Once page lock is dropped, mapping could become invalid.
	 * As a hack, increase map count until we lock page again.
	 */
	atomic_inc(&hpage->_mapcount);
	unlock_page(hpage);
	i_mmap_lock_write(mapping);
	lock_page(hpage);
	atomic_add_negative(-1, &hpage->_mapcount);

	/* verify page is still mapped */
	if (!page_mapped(hpage)) {
		i_mmap_unlock_write(mapping);
		return NULL;
	}

	/*
	 * Get address space again and verify it is the same one
	 * we locked.  If not, drop lock and retry.
	 */
	mapping2 = _get_hugetlb_page_mapping(hpage);
	if (mapping2 != mapping) {
		i_mmap_unlock_write(mapping);
		mapping = mapping2;
		goto retry;
	}

	return mapping;
}

1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684 1685 1686 1687 1688 1689
pgoff_t __basepage_index(struct page *page)
{
	struct page *page_head = compound_head(page);
	pgoff_t index = page_index(page_head);
	unsigned long compound_idx;

	if (!PageHuge(page_head))
		return page_index(page);

	if (compound_order(page_head) >= MAX_ORDER)
		compound_idx = page_to_pfn(page) - page_to_pfn(page_head);
	else
		compound_idx = page - page_head;

	return (index << compound_order(page_head)) + compound_idx;
}

1690
static struct page *alloc_buddy_huge_page(struct hstate *h,
1691 1692
		gfp_t gfp_mask, int nid, nodemask_t *nmask,
		nodemask_t *node_alloc_noretry)
L
Linus Torvalds 已提交
1693
{
1694
	int order = huge_page_order(h);
L
Linus Torvalds 已提交
1695
	struct page *page;
1696
	bool alloc_try_hard = true;
1697

1698 1699 1700 1701 1702 1703 1704 1705 1706 1707 1708 1709
	/*
	 * 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;
1710 1711 1712 1713 1714 1715 1716
	if (nid == NUMA_NO_NODE)
		nid = numa_mem_id();
	page = __alloc_pages_nodemask(gfp_mask, order, nid, nmask);
	if (page)
		__count_vm_event(HTLB_BUDDY_PGALLOC);
	else
		__count_vm_event(HTLB_BUDDY_PGALLOC_FAIL);
1717

1718 1719 1720 1721 1722 1723 1724 1725 1726 1727 1728 1729 1730 1731 1732 1733
	/*
	 * 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);

1734 1735 1736
	return page;
}

1737 1738 1739 1740 1741
/*
 * 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,
1742 1743
		gfp_t gfp_mask, int nid, nodemask_t *nmask,
		nodemask_t *node_alloc_noretry)
1744 1745 1746 1747 1748 1749 1750
{
	struct page *page;

	if (hstate_is_gigantic(h))
		page = alloc_gigantic_page(h, gfp_mask, nid, nmask);
	else
		page = alloc_buddy_huge_page(h, gfp_mask,
1751
				nid, nmask, node_alloc_noretry);
1752 1753 1754 1755 1756 1757 1758 1759 1760 1761
	if (!page)
		return NULL;

	if (hstate_is_gigantic(h))
		prep_compound_gigantic_page(page, huge_page_order(h));
	prep_new_huge_page(h, page, page_to_nid(page));

	return page;
}

1762 1763 1764 1765
/*
 * Allocates a fresh page to the hugetlb allocator pool in the node interleaved
 * manner.
 */
1766 1767
static int alloc_pool_huge_page(struct hstate *h, nodemask_t *nodes_allowed,
				nodemask_t *node_alloc_noretry)
1768 1769 1770
{
	struct page *page;
	int nr_nodes, node;
1771
	gfp_t gfp_mask = htlb_alloc_mask(h) | __GFP_THISNODE;
1772 1773

	for_each_node_mask_to_alloc(h, nr_nodes, node, nodes_allowed) {
1774 1775
		page = alloc_fresh_huge_page(h, gfp_mask, node, nodes_allowed,
						node_alloc_noretry);
1776
		if (page)
1777 1778 1779
			break;
	}

1780 1781
	if (!page)
		return 0;
1782

1783 1784 1785
	put_page(page); /* free it into the hugepage allocator */

	return 1;
1786 1787
}

1788 1789 1790 1791 1792 1793
/*
 * Free huge page from pool from next node to free.
 * Attempt to keep persistent huge pages more or less
 * balanced over allowed nodes.
 * Called with hugetlb_lock locked.
 */
1794 1795
static int free_pool_huge_page(struct hstate *h, nodemask_t *nodes_allowed,
							 bool acct_surplus)
1796
{
1797
	int nr_nodes, node;
1798 1799
	int ret = 0;

1800
	for_each_node_mask_to_free(h, nr_nodes, node, nodes_allowed) {
1801 1802 1803 1804
		/*
		 * If we're returning unused surplus pages, only examine
		 * nodes with surplus pages.
		 */
1805 1806
		if ((!acct_surplus || h->surplus_huge_pages_node[node]) &&
		    !list_empty(&h->hugepage_freelists[node])) {
1807
			struct page *page =
1808
				list_entry(h->hugepage_freelists[node].next,
1809 1810 1811
					  struct page, lru);
			list_del(&page->lru);
			h->free_huge_pages--;
1812
			h->free_huge_pages_node[node]--;
1813 1814
			if (acct_surplus) {
				h->surplus_huge_pages--;
1815
				h->surplus_huge_pages_node[node]--;
1816
			}
1817 1818
			update_and_free_page(h, page);
			ret = 1;
1819
			break;
1820
		}
1821
	}
1822 1823 1824 1825

	return ret;
}

1826 1827
/*
 * Dissolve a given free hugepage into free buddy pages. This function does
1828 1829 1830 1831 1832 1833 1834
 * nothing for in-use hugepages and non-hugepages.
 * This function returns values like below:
 *
 *  -EBUSY: failed to dissolved free hugepages or the hugepage is in-use
 *          (allocated or reserved.)
 *       0: successfully dissolved free hugepages or the page is not a
 *          hugepage (considered as already dissolved)
1835
 */
1836
int dissolve_free_huge_page(struct page *page)
1837
{
1838
	int rc = -EBUSY;
1839

1840 1841 1842 1843
	/* Not to disrupt normal path by vainly holding hugetlb_lock */
	if (!PageHuge(page))
		return 0;

1844
	spin_lock(&hugetlb_lock);
1845 1846 1847 1848 1849 1850
	if (!PageHuge(page)) {
		rc = 0;
		goto out;
	}

	if (!page_count(page)) {
1851 1852 1853
		struct page *head = compound_head(page);
		struct hstate *h = page_hstate(head);
		int nid = page_to_nid(head);
1854
		if (h->free_huge_pages - h->resv_huge_pages == 0)
1855
			goto out;
1856 1857 1858 1859 1860 1861 1862 1863
		/*
		 * 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);
		}
1864
		list_del(&head->lru);
1865 1866
		h->free_huge_pages--;
		h->free_huge_pages_node[nid]--;
1867
		h->max_huge_pages--;
1868
		update_and_free_page(h, head);
1869
		rc = 0;
1870
	}
1871
out:
1872
	spin_unlock(&hugetlb_lock);
1873
	return rc;
1874 1875 1876 1877 1878
}

/*
 * Dissolve free hugepages in a given pfn range. Used by memory hotplug to
 * make specified memory blocks removable from the system.
1879 1880
 * Note that this will dissolve a free gigantic hugepage completely, if any
 * part of it lies within the given range.
1881 1882
 * Also note that if dissolve_free_huge_page() returns with an error, all
 * free hugepages that were dissolved before that error are lost.
1883
 */
1884
int dissolve_free_huge_pages(unsigned long start_pfn, unsigned long end_pfn)
1885 1886
{
	unsigned long pfn;
1887
	struct page *page;
1888
	int rc = 0;
1889

1890
	if (!hugepages_supported())
1891
		return rc;
1892

1893 1894
	for (pfn = start_pfn; pfn < end_pfn; pfn += 1 << minimum_order) {
		page = pfn_to_page(pfn);
1895 1896 1897
		rc = dissolve_free_huge_page(page);
		if (rc)
			break;
1898
	}
1899 1900

	return rc;
1901 1902
}

1903 1904 1905
/*
 * Allocates a fresh surplus page from the page allocator.
 */
1906
static struct page *alloc_surplus_huge_page(struct hstate *h, gfp_t gfp_mask,
1907
		int nid, nodemask_t *nmask)
1908
{
1909
	struct page *page = NULL;
1910

1911
	if (hstate_is_gigantic(h))
1912 1913
		return NULL;

1914
	spin_lock(&hugetlb_lock);
1915 1916
	if (h->surplus_huge_pages >= h->nr_overcommit_huge_pages)
		goto out_unlock;
1917 1918
	spin_unlock(&hugetlb_lock);

1919
	page = alloc_fresh_huge_page(h, gfp_mask, nid, nmask, NULL);
1920
	if (!page)
1921
		return NULL;
1922 1923

	spin_lock(&hugetlb_lock);
1924 1925 1926 1927 1928 1929 1930 1931 1932
	/*
	 * We could have raced with the pool size change.
	 * Double check that and simply deallocate the new page
	 * if we would end up overcommiting the surpluses. Abuse
	 * temporary page to workaround the nasty free_huge_page
	 * codeflow
	 */
	if (h->surplus_huge_pages >= h->nr_overcommit_huge_pages) {
		SetPageHugeTemporary(page);
1933
		spin_unlock(&hugetlb_lock);
1934
		put_page(page);
1935
		return NULL;
1936 1937
	} else {
		h->surplus_huge_pages++;
1938
		h->surplus_huge_pages_node[page_to_nid(page)]++;
1939
	}
1940 1941

out_unlock:
1942
	spin_unlock(&hugetlb_lock);
1943 1944 1945 1946

	return page;
}

1947 1948
struct page *alloc_migrate_huge_page(struct hstate *h, gfp_t gfp_mask,
				     int nid, nodemask_t *nmask)
1949 1950 1951 1952 1953 1954
{
	struct page *page;

	if (hstate_is_gigantic(h))
		return NULL;

1955
	page = alloc_fresh_huge_page(h, gfp_mask, nid, nmask, NULL);
1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 1966 1967
	if (!page)
		return NULL;

	/*
	 * We do not account these pages as surplus because they are only
	 * temporary and will be released properly on the last reference
	 */
	SetPageHugeTemporary(page);

	return page;
}

1968 1969 1970
/*
 * Use the VMA's mpolicy to allocate a huge page from the buddy.
 */
D
Dave Hansen 已提交
1971
static
1972
struct page *alloc_buddy_huge_page_with_mpol(struct hstate *h,
1973 1974
		struct vm_area_struct *vma, unsigned long addr)
{
1975 1976 1977 1978 1979 1980 1981
	struct page *page;
	struct mempolicy *mpol;
	gfp_t gfp_mask = htlb_alloc_mask(h);
	int nid;
	nodemask_t *nodemask;

	nid = huge_node(vma, addr, gfp_mask, &mpol, &nodemask);
1982
	page = alloc_surplus_huge_page(h, gfp_mask, nid, nodemask);
1983 1984 1985
	mpol_cond_put(mpol);

	return page;
1986 1987
}

1988
/* page migration callback function */
1989 1990
struct page *alloc_huge_page_node(struct hstate *h, int nid)
{
1991
	gfp_t gfp_mask = htlb_alloc_mask(h);
1992
	struct page *page = NULL;
1993

1994 1995 1996
	if (nid != NUMA_NO_NODE)
		gfp_mask |= __GFP_THISNODE;

1997
	spin_lock(&hugetlb_lock);
1998
	if (h->free_huge_pages - h->resv_huge_pages > 0)
1999
		page = dequeue_huge_page_nodemask(h, gfp_mask, nid, NULL);
2000 2001
	spin_unlock(&hugetlb_lock);

2002
	if (!page)
2003
		page = alloc_migrate_huge_page(h, gfp_mask, nid, NULL);
2004 2005 2006 2007

	return page;
}

2008
/* page migration callback function */
2009 2010
struct page *alloc_huge_page_nodemask(struct hstate *h, int preferred_nid,
		nodemask_t *nmask)
2011
{
2012
	gfp_t gfp_mask = htlb_alloc_mask(h);
2013 2014 2015

	spin_lock(&hugetlb_lock);
	if (h->free_huge_pages - h->resv_huge_pages > 0) {
2016 2017 2018 2019 2020 2021
		struct page *page;

		page = dequeue_huge_page_nodemask(h, gfp_mask, preferred_nid, nmask);
		if (page) {
			spin_unlock(&hugetlb_lock);
			return page;
2022 2023 2024 2025
		}
	}
	spin_unlock(&hugetlb_lock);

2026
	return alloc_migrate_huge_page(h, gfp_mask, preferred_nid, nmask);
2027 2028
}

2029
/* mempolicy aware migration callback */
2030 2031
struct page *alloc_huge_page_vma(struct hstate *h, struct vm_area_struct *vma,
		unsigned long address)
2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046
{
	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);
	page = alloc_huge_page_nodemask(h, node, nodemask);
	mpol_cond_put(mpol);

	return page;
}

2047
/*
L
Lucas De Marchi 已提交
2048
 * Increase the hugetlb pool such that it can accommodate a reservation
2049 2050
 * of size 'delta'.
 */
2051
static int gather_surplus_pages(struct hstate *h, int delta)
2052
	__must_hold(&hugetlb_lock)
2053 2054 2055 2056 2057
{
	struct list_head surplus_list;
	struct page *page, *tmp;
	int ret, i;
	int needed, allocated;
2058
	bool alloc_ok = true;
2059

2060
	needed = (h->resv_huge_pages + delta) - h->free_huge_pages;
2061
	if (needed <= 0) {
2062
		h->resv_huge_pages += delta;
2063
		return 0;
2064
	}
2065 2066 2067 2068 2069 2070 2071 2072

	allocated = 0;
	INIT_LIST_HEAD(&surplus_list);

	ret = -ENOMEM;
retry:
	spin_unlock(&hugetlb_lock);
	for (i = 0; i < needed; i++) {
2073
		page = alloc_surplus_huge_page(h, htlb_alloc_mask(h),
2074
				NUMA_NO_NODE, NULL);
2075 2076 2077 2078
		if (!page) {
			alloc_ok = false;
			break;
		}
2079
		list_add(&page->lru, &surplus_list);
2080
		cond_resched();
2081
	}
2082
	allocated += i;
2083 2084 2085 2086 2087 2088

	/*
	 * After retaking hugetlb_lock, we need to recalculate 'needed'
	 * because either resv_huge_pages or free_huge_pages may have changed.
	 */
	spin_lock(&hugetlb_lock);
2089 2090
	needed = (h->resv_huge_pages + delta) -
			(h->free_huge_pages + allocated);
2091 2092 2093 2094 2095 2096 2097 2098 2099 2100
	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;
	}
2101 2102
	/*
	 * The surplus_list now contains _at_least_ the number of extra pages
L
Lucas De Marchi 已提交
2103
	 * needed to accommodate the reservation.  Add the appropriate number
2104
	 * of pages to the hugetlb pool and free the extras back to the buddy
2105 2106 2107
	 * 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.
2108 2109
	 */
	needed += allocated;
2110
	h->resv_huge_pages += delta;
2111
	ret = 0;
2112

2113
	/* Free the needed pages to the hugetlb pool */
2114
	list_for_each_entry_safe(page, tmp, &surplus_list, lru) {
2115 2116
		if ((--needed) < 0)
			break;
2117 2118 2119 2120 2121
		/*
		 * This page is now managed by the hugetlb allocator and has
		 * no users -- drop the buddy allocator's reference.
		 */
		put_page_testzero(page);
2122
		VM_BUG_ON_PAGE(page_count(page), page);
2123
		enqueue_huge_page(h, page);
2124
	}
2125
free:
2126
	spin_unlock(&hugetlb_lock);
2127 2128

	/* Free unnecessary surplus pages to the buddy allocator */
2129 2130
	list_for_each_entry_safe(page, tmp, &surplus_list, lru)
		put_page(page);
2131
	spin_lock(&hugetlb_lock);
2132 2133 2134 2135 2136

	return ret;
}

/*
2137 2138 2139 2140 2141 2142 2143 2144 2145 2146 2147 2148
 * This routine has two main purposes:
 * 1) Decrement the reservation count (resv_huge_pages) by the value passed
 *    in unused_resv_pages.  This corresponds to the prior adjustments made
 *    to the associated reservation map.
 * 2) Free any unused surplus pages that may have been allocated to satisfy
 *    the reservation.  As many as unused_resv_pages may be freed.
 *
 * Called with hugetlb_lock held.  However, the lock could be dropped (and
 * reacquired) during calls to cond_resched_lock.  Whenever dropping the lock,
 * we must make sure nobody else can claim pages we are in the process of
 * freeing.  Do this by ensuring resv_huge_page always is greater than the
 * number of huge pages we plan to free when dropping the lock.
2149
 */
2150 2151
static void return_unused_surplus_pages(struct hstate *h,
					unsigned long unused_resv_pages)
2152 2153 2154
{
	unsigned long nr_pages;

2155
	/* Cannot return gigantic pages currently */
2156
	if (hstate_is_gigantic(h))
2157
		goto out;
2158

2159 2160 2161 2162
	/*
	 * Part (or even all) of the reservation could have been backed
	 * by pre-allocated pages. Only free surplus pages.
	 */
2163
	nr_pages = min(unused_resv_pages, h->surplus_huge_pages);
2164

2165 2166
	/*
	 * We want to release as many surplus pages as possible, spread
2167 2168 2169
	 * 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.
2170
	 * free_pool_huge_page() will balance the freed pages across the
2171
	 * on-line nodes with memory and will handle the hstate accounting.
2172 2173 2174 2175
	 *
	 * Note that we decrement resv_huge_pages as we free the pages.  If
	 * we drop the lock, resv_huge_pages will still be sufficiently large
	 * to cover subsequent pages we may free.
2176 2177
	 */
	while (nr_pages--) {
2178 2179
		h->resv_huge_pages--;
		unused_resv_pages--;
2180
		if (!free_pool_huge_page(h, &node_states[N_MEMORY], 1))
2181
			goto out;
2182
		cond_resched_lock(&hugetlb_lock);
2183
	}
2184 2185 2186 2187

out:
	/* Fully uncommit the reservation */
	h->resv_huge_pages -= unused_resv_pages;
2188 2189
}

2190

2191
/*
2192
 * vma_needs_reservation, vma_commit_reservation and vma_end_reservation
2193
 * are used by the huge page allocation routines to manage reservations.
2194 2195 2196 2197 2198 2199
 *
 * 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
2200 2201 2202
 * 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.
2203 2204 2205 2206 2207 2208
 *
 * 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.
2209 2210 2211 2212 2213
 *
 * 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.
2214
 */
2215 2216 2217
enum vma_resv_mode {
	VMA_NEEDS_RESV,
	VMA_COMMIT_RESV,
2218
	VMA_END_RESV,
2219
	VMA_ADD_RESV,
2220
};
2221 2222
static long __vma_reservation_common(struct hstate *h,
				struct vm_area_struct *vma, unsigned long addr,
2223
				enum vma_resv_mode mode)
2224
{
2225 2226
	struct resv_map *resv;
	pgoff_t idx;
2227
	long ret;
2228
	long dummy_out_regions_needed;
2229

2230 2231
	resv = vma_resv_map(vma);
	if (!resv)
2232
		return 1;
2233

2234
	idx = vma_hugecache_offset(h, vma, addr);
2235 2236
	switch (mode) {
	case VMA_NEEDS_RESV:
2237 2238 2239 2240 2241 2242
		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);
2243 2244
		break;
	case VMA_COMMIT_RESV:
2245
		ret = region_add(resv, idx, idx + 1, 1, NULL, NULL);
2246 2247
		/* region_add calls of range 1 should never fail. */
		VM_BUG_ON(ret < 0);
2248
		break;
2249
	case VMA_END_RESV:
2250
		region_abort(resv, idx, idx + 1, 1);
2251 2252
		ret = 0;
		break;
2253
	case VMA_ADD_RESV:
2254
		if (vma->vm_flags & VM_MAYSHARE) {
2255
			ret = region_add(resv, idx, idx + 1, 1, NULL, NULL);
2256 2257 2258 2259
			/* region_add calls of range 1 should never fail. */
			VM_BUG_ON(ret < 0);
		} else {
			region_abort(resv, idx, idx + 1, 1);
2260 2261 2262
			ret = region_del(resv, idx, idx + 1);
		}
		break;
2263 2264 2265
	default:
		BUG();
	}
2266

2267
	if (vma->vm_flags & VM_MAYSHARE)
2268
		return ret;
2269 2270 2271 2272 2273 2274 2275 2276 2277 2278 2279 2280 2281 2282 2283 2284 2285 2286 2287
	else if (is_vma_resv_set(vma, HPAGE_RESV_OWNER) && ret >= 0) {
		/*
		 * In most cases, reserves always exist for private mappings.
		 * However, a file associated with mapping could have been
		 * hole punched or truncated after reserves were consumed.
		 * As subsequent fault on such a range will not use reserves.
		 * Subtle - The reserve map for private mappings has the
		 * opposite meaning than that of shared mappings.  If NO
		 * entry is in the reserve map, it means a reservation exists.
		 * If an entry exists in the reserve map, it means the
		 * reservation has already been consumed.  As a result, the
		 * return value of this routine is the opposite of the
		 * value returned from reserve map manipulation routines above.
		 */
		if (ret)
			return 0;
		else
			return 1;
	}
2288
	else
2289
		return ret < 0 ? ret : 0;
2290
}
2291 2292

static long vma_needs_reservation(struct hstate *h,
2293
			struct vm_area_struct *vma, unsigned long addr)
2294
{
2295
	return __vma_reservation_common(h, vma, addr, VMA_NEEDS_RESV);
2296
}
2297

2298 2299 2300
static long vma_commit_reservation(struct hstate *h,
			struct vm_area_struct *vma, unsigned long addr)
{
2301 2302 2303
	return __vma_reservation_common(h, vma, addr, VMA_COMMIT_RESV);
}

2304
static void vma_end_reservation(struct hstate *h,
2305 2306
			struct vm_area_struct *vma, unsigned long addr)
{
2307
	(void)__vma_reservation_common(h, vma, addr, VMA_END_RESV);
2308 2309
}

2310 2311 2312 2313 2314 2315 2316 2317 2318 2319 2320 2321 2322 2323 2324 2325 2326 2327 2328 2329 2330 2331 2332 2333 2334 2335 2336 2337 2338 2339 2340 2341 2342 2343 2344 2345 2346 2347 2348 2349 2350 2351 2352 2353 2354 2355 2356 2357 2358 2359
static long vma_add_reservation(struct hstate *h,
			struct vm_area_struct *vma, unsigned long addr)
{
	return __vma_reservation_common(h, vma, addr, VMA_ADD_RESV);
}

/*
 * This routine is called to restore a reservation on error paths.  In the
 * specific error paths, a huge page was allocated (via alloc_huge_page)
 * and is about to be freed.  If a reservation for the page existed,
 * alloc_huge_page would have consumed the reservation and set PagePrivate
 * in the newly allocated page.  When the page is freed via free_huge_page,
 * the global reservation count will be incremented if PagePrivate is set.
 * However, free_huge_page can not adjust the reserve map.  Adjust the
 * reserve map here to be consistent with global reserve count adjustments
 * to be made by free_huge_page.
 */
static void restore_reserve_on_error(struct hstate *h,
			struct vm_area_struct *vma, unsigned long address,
			struct page *page)
{
	if (unlikely(PagePrivate(page))) {
		long rc = vma_needs_reservation(h, vma, address);

		if (unlikely(rc < 0)) {
			/*
			 * Rare out of memory condition in reserve map
			 * manipulation.  Clear PagePrivate so that
			 * global reserve count will not be incremented
			 * by free_huge_page.  This will make it appear
			 * as though the reservation for this page was
			 * consumed.  This may prevent the task from
			 * faulting in the page at a later time.  This
			 * is better than inconsistent global huge page
			 * accounting of reserve counts.
			 */
			ClearPagePrivate(page);
		} else if (rc) {
			rc = vma_add_reservation(h, vma, address);
			if (unlikely(rc < 0))
				/*
				 * See above comment about rare out of
				 * memory condition.
				 */
				ClearPagePrivate(page);
		} else
			vma_end_reservation(h, vma, address);
	}
}

2360
struct page *alloc_huge_page(struct vm_area_struct *vma,
2361
				    unsigned long addr, int avoid_reserve)
L
Linus Torvalds 已提交
2362
{
2363
	struct hugepage_subpool *spool = subpool_vma(vma);
2364
	struct hstate *h = hstate_vma(vma);
2365
	struct page *page;
2366 2367
	long map_chg, map_commit;
	long gbl_chg;
2368 2369
	int ret, idx;
	struct hugetlb_cgroup *h_cg;
2370
	bool deferred_reserve;
2371

2372
	idx = hstate_index(h);
2373
	/*
2374 2375 2376
	 * 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).
2377
	 */
2378 2379
	map_chg = gbl_chg = vma_needs_reservation(h, vma, addr);
	if (map_chg < 0)
2380
		return ERR_PTR(-ENOMEM);
2381 2382 2383 2384 2385 2386 2387 2388 2389 2390 2391

	/*
	 * 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) {
2392
			vma_end_reservation(h, vma, addr);
2393
			return ERR_PTR(-ENOSPC);
2394
		}
L
Linus Torvalds 已提交
2395

2396 2397 2398 2399 2400 2401 2402 2403 2404 2405 2406 2407
		/*
		 * 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;
	}

2408 2409 2410 2411 2412 2413 2414 2415 2416 2417
	/* If this allocation is not consuming a reservation, charge it now.
	 */
	deferred_reserve = map_chg || avoid_reserve || !vma_resv_map(vma);
	if (deferred_reserve) {
		ret = hugetlb_cgroup_charge_cgroup_rsvd(
			idx, pages_per_huge_page(h), &h_cg);
		if (ret)
			goto out_subpool_put;
	}

2418
	ret = hugetlb_cgroup_charge_cgroup(idx, pages_per_huge_page(h), &h_cg);
2419
	if (ret)
2420
		goto out_uncharge_cgroup_reservation;
2421

L
Linus Torvalds 已提交
2422
	spin_lock(&hugetlb_lock);
2423 2424 2425 2426 2427 2428
	/*
	 * 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);
2429
	if (!page) {
2430
		spin_unlock(&hugetlb_lock);
2431
		page = alloc_buddy_huge_page_with_mpol(h, vma, addr);
2432 2433
		if (!page)
			goto out_uncharge_cgroup;
2434 2435 2436 2437
		if (!avoid_reserve && vma_has_reserves(vma, gbl_chg)) {
			SetPagePrivate(page);
			h->resv_huge_pages--;
		}
2438 2439
		spin_lock(&hugetlb_lock);
		list_move(&page->lru, &h->hugepage_activelist);
2440
		/* Fall through */
K
Ken Chen 已提交
2441
	}
2442
	hugetlb_cgroup_commit_charge(idx, pages_per_huge_page(h), h_cg, page);
2443 2444 2445 2446 2447 2448 2449 2450
	/* 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);
	}

2451
	spin_unlock(&hugetlb_lock);
2452

2453
	set_page_private(page, (unsigned long)spool);
2454

2455 2456
	map_commit = vma_commit_reservation(h, vma, addr);
	if (unlikely(map_chg > map_commit)) {
2457 2458 2459 2460 2461 2462 2463 2464 2465 2466 2467 2468 2469 2470
		/*
		 * 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);
	}
2471
	return page;
2472 2473 2474

out_uncharge_cgroup:
	hugetlb_cgroup_uncharge_cgroup(idx, pages_per_huge_page(h), h_cg);
2475 2476 2477 2478
out_uncharge_cgroup_reservation:
	if (deferred_reserve)
		hugetlb_cgroup_uncharge_cgroup_rsvd(idx, pages_per_huge_page(h),
						    h_cg);
2479
out_subpool_put:
2480
	if (map_chg || avoid_reserve)
2481
		hugepage_subpool_put_pages(spool, 1);
2482
	vma_end_reservation(h, vma, addr);
2483
	return ERR_PTR(-ENOSPC);
2484 2485
}

2486 2487 2488
int alloc_bootmem_huge_page(struct hstate *h)
	__attribute__ ((weak, alias("__alloc_bootmem_huge_page")));
int __alloc_bootmem_huge_page(struct hstate *h)
2489 2490
{
	struct huge_bootmem_page *m;
2491
	int nr_nodes, node;
2492

2493
	for_each_node_mask_to_alloc(h, nr_nodes, node, &node_states[N_MEMORY]) {
2494 2495
		void *addr;

2496
		addr = memblock_alloc_try_nid_raw(
2497
				huge_page_size(h), huge_page_size(h),
2498
				0, MEMBLOCK_ALLOC_ACCESSIBLE, node);
2499 2500 2501 2502 2503 2504 2505
		if (addr) {
			/*
			 * Use the beginning of the huge page to store the
			 * huge_bootmem_page struct (until gather_bootmem
			 * puts them into the mem_map).
			 */
			m = addr;
2506
			goto found;
2507 2508 2509 2510 2511
		}
	}
	return 0;

found:
2512
	BUG_ON(!IS_ALIGNED(virt_to_phys(m), huge_page_size(h)));
2513
	/* Put them into a private list first because mem_map is not up yet */
2514
	INIT_LIST_HEAD(&m->list);
2515 2516 2517 2518 2519
	list_add(&m->list, &huge_boot_pages);
	m->hstate = h;
	return 1;
}

2520 2521
static void __init prep_compound_huge_page(struct page *page,
		unsigned int order)
2522 2523 2524 2525 2526 2527 2528
{
	if (unlikely(order > (MAX_ORDER - 1)))
		prep_compound_gigantic_page(page, order);
	else
		prep_compound_page(page, order);
}

2529 2530 2531 2532 2533 2534
/* Put bootmem huge pages into the standard lists after mem_map is up */
static void __init gather_bootmem_prealloc(void)
{
	struct huge_bootmem_page *m;

	list_for_each_entry(m, &huge_boot_pages, list) {
2535
		struct page *page = virt_to_page(m);
2536
		struct hstate *h = m->hstate;
2537

2538
		WARN_ON(page_count(page) != 1);
2539
		prep_compound_huge_page(page, h->order);
2540
		WARN_ON(PageReserved(page));
2541
		prep_new_huge_page(h, page, page_to_nid(page));
2542 2543
		put_page(page); /* free it into the hugepage allocator */

2544 2545 2546 2547 2548 2549
		/*
		 * If we had gigantic hugepages allocated at boot time, we need
		 * to restore the 'stolen' pages to totalram_pages in order to
		 * fix confusing memory reports from free(1) and another
		 * side-effects, like CommitLimit going negative.
		 */
2550
		if (hstate_is_gigantic(h))
2551
			adjust_managed_page_count(page, 1 << h->order);
2552
		cond_resched();
2553 2554 2555
	}
}

2556
static void __init hugetlb_hstate_alloc_pages(struct hstate *h)
L
Linus Torvalds 已提交
2557 2558
{
	unsigned long i;
2559 2560 2561 2562 2563 2564 2565 2566 2567 2568 2569 2570 2571 2572 2573 2574 2575 2576 2577
	nodemask_t *node_alloc_noretry;

	if (!hstate_is_gigantic(h)) {
		/*
		 * Bit mask controlling how hard we retry per-node allocations.
		 * Ignore errors as lower level routines can deal with
		 * node_alloc_noretry == NULL.  If this kmalloc fails at boot
		 * time, we are likely in bigger trouble.
		 */
		node_alloc_noretry = kmalloc(sizeof(*node_alloc_noretry),
						GFP_KERNEL);
	} else {
		/* allocations done at boot time */
		node_alloc_noretry = NULL;
	}

	/* bit mask controlling how hard we retry per-node allocations */
	if (node_alloc_noretry)
		nodes_clear(*node_alloc_noretry);
2578

2579
	for (i = 0; i < h->max_huge_pages; ++i) {
2580
		if (hstate_is_gigantic(h)) {
2581
			if (hugetlb_cma_size) {
2582 2583 2584
				pr_warn_once("HugeTLB: hugetlb_cma is enabled, skip boot time allocation\n");
				break;
			}
2585 2586
			if (!alloc_bootmem_huge_page(h))
				break;
2587
		} else if (!alloc_pool_huge_page(h,
2588 2589
					 &node_states[N_MEMORY],
					 node_alloc_noretry))
L
Linus Torvalds 已提交
2590
			break;
2591
		cond_resched();
L
Linus Torvalds 已提交
2592
	}
2593 2594 2595
	if (i < h->max_huge_pages) {
		char buf[32];

2596
		string_get_size(huge_page_size(h), 1, STRING_UNITS_2, buf, 32);
2597 2598 2599 2600
		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;
	}
2601 2602

	kfree(node_alloc_noretry);
2603 2604 2605 2606 2607 2608 2609
}

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

	for_each_hstate(h) {
2610 2611 2612
		if (minimum_order > huge_page_order(h))
			minimum_order = huge_page_order(h);

2613
		/* oversize hugepages were init'ed in early boot */
2614
		if (!hstate_is_gigantic(h))
2615
			hugetlb_hstate_alloc_pages(h);
2616
	}
2617
	VM_BUG_ON(minimum_order == UINT_MAX);
2618 2619 2620 2621 2622 2623 2624
}

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

	for_each_hstate(h) {
A
Andi Kleen 已提交
2625
		char buf[32];
2626 2627

		string_get_size(huge_page_size(h), 1, STRING_UNITS_2, buf, 32);
2628
		pr_info("HugeTLB registered %s page size, pre-allocated %ld pages\n",
2629
			buf, h->free_huge_pages);
2630 2631 2632
	}
}

L
Linus Torvalds 已提交
2633
#ifdef CONFIG_HIGHMEM
2634 2635
static void try_to_free_low(struct hstate *h, unsigned long count,
						nodemask_t *nodes_allowed)
L
Linus Torvalds 已提交
2636
{
2637 2638
	int i;

2639
	if (hstate_is_gigantic(h))
2640 2641
		return;

2642
	for_each_node_mask(i, *nodes_allowed) {
L
Linus Torvalds 已提交
2643
		struct page *page, *next;
2644 2645 2646
		struct list_head *freel = &h->hugepage_freelists[i];
		list_for_each_entry_safe(page, next, freel, lru) {
			if (count >= h->nr_huge_pages)
2647
				return;
L
Linus Torvalds 已提交
2648 2649 2650
			if (PageHighMem(page))
				continue;
			list_del(&page->lru);
2651
			update_and_free_page(h, page);
2652 2653
			h->free_huge_pages--;
			h->free_huge_pages_node[page_to_nid(page)]--;
L
Linus Torvalds 已提交
2654 2655 2656 2657
		}
	}
}
#else
2658 2659
static inline void try_to_free_low(struct hstate *h, unsigned long count,
						nodemask_t *nodes_allowed)
L
Linus Torvalds 已提交
2660 2661 2662 2663
{
}
#endif

2664 2665 2666 2667 2668
/*
 * 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.
 */
2669 2670
static int adjust_pool_surplus(struct hstate *h, nodemask_t *nodes_allowed,
				int delta)
2671
{
2672
	int nr_nodes, node;
2673 2674 2675

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

2676 2677 2678 2679
	if (delta < 0) {
		for_each_node_mask_to_alloc(h, nr_nodes, node, nodes_allowed) {
			if (h->surplus_huge_pages_node[node])
				goto found;
2680
		}
2681 2682 2683 2684 2685
	} 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;
2686
		}
2687 2688
	}
	return 0;
2689

2690 2691 2692 2693
found:
	h->surplus_huge_pages += delta;
	h->surplus_huge_pages_node[node] += delta;
	return 1;
2694 2695
}

2696
#define persistent_huge_pages(h) (h->nr_huge_pages - h->surplus_huge_pages)
2697
static int set_max_huge_pages(struct hstate *h, unsigned long count, int nid,
2698
			      nodemask_t *nodes_allowed)
L
Linus Torvalds 已提交
2699
{
2700
	unsigned long min_count, ret;
2701 2702 2703 2704 2705 2706 2707 2708 2709 2710 2711
	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
2712

2713 2714
	spin_lock(&hugetlb_lock);

2715 2716 2717 2718 2719 2720 2721 2722 2723 2724 2725 2726 2727 2728 2729 2730 2731 2732 2733 2734
	/*
	 * 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;
	}

2735 2736 2737 2738 2739 2740 2741 2742 2743 2744
	/*
	 * Gigantic pages runtime allocation depend on the capability for large
	 * page range allocation.
	 * If the system does not provide this feature, return an error when
	 * the user tries to allocate gigantic pages but let the user free the
	 * boottime allocated gigantic pages.
	 */
	if (hstate_is_gigantic(h) && !IS_ENABLED(CONFIG_CONTIG_ALLOC)) {
		if (count > persistent_huge_pages(h)) {
			spin_unlock(&hugetlb_lock);
2745
			NODEMASK_FREE(node_alloc_noretry);
2746 2747 2748 2749
			return -EINVAL;
		}
		/* Fall through to decrease pool */
	}
2750

2751 2752 2753 2754
	/*
	 * Increase the pool size
	 * First take pages out of surplus state.  Then make up the
	 * remaining difference by allocating fresh huge pages.
2755
	 *
2756
	 * We might race with alloc_surplus_huge_page() here and be unable
2757 2758 2759 2760
	 * 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.
2761
	 */
2762
	while (h->surplus_huge_pages && count > persistent_huge_pages(h)) {
2763
		if (!adjust_pool_surplus(h, nodes_allowed, -1))
2764 2765 2766
			break;
	}

2767
	while (count > persistent_huge_pages(h)) {
2768 2769 2770 2771 2772 2773
		/*
		 * If this allocation races such that we no longer need the
		 * page, free_huge_page will handle it by freeing the page
		 * and reducing the surplus.
		 */
		spin_unlock(&hugetlb_lock);
2774 2775 2776 2777

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

2778 2779
		ret = alloc_pool_huge_page(h, nodes_allowed,
						node_alloc_noretry);
2780 2781 2782 2783
		spin_lock(&hugetlb_lock);
		if (!ret)
			goto out;

2784 2785 2786
		/* Bail for signals. Probably ctrl-c from user */
		if (signal_pending(current))
			goto out;
2787 2788 2789 2790 2791 2792 2793 2794
	}

	/*
	 * 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.
2795 2796 2797 2798
	 *
	 * 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
2799
	 * alloc_surplus_huge_page() is checking the global counter,
2800 2801 2802
	 * 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.
2803
	 */
2804
	min_count = h->resv_huge_pages + h->nr_huge_pages - h->free_huge_pages;
2805
	min_count = max(count, min_count);
2806
	try_to_free_low(h, min_count, nodes_allowed);
2807
	while (min_count < persistent_huge_pages(h)) {
2808
		if (!free_pool_huge_page(h, nodes_allowed, 0))
L
Linus Torvalds 已提交
2809
			break;
2810
		cond_resched_lock(&hugetlb_lock);
L
Linus Torvalds 已提交
2811
	}
2812
	while (count < persistent_huge_pages(h)) {
2813
		if (!adjust_pool_surplus(h, nodes_allowed, 1))
2814 2815 2816
			break;
	}
out:
2817
	h->max_huge_pages = persistent_huge_pages(h);
L
Linus Torvalds 已提交
2818
	spin_unlock(&hugetlb_lock);
2819

2820 2821
	NODEMASK_FREE(node_alloc_noretry);

2822
	return 0;
L
Linus Torvalds 已提交
2823 2824
}

2825 2826 2827 2828 2829 2830 2831 2832 2833 2834
#define HSTATE_ATTR_RO(_name) \
	static struct kobj_attribute _name##_attr = __ATTR_RO(_name)

#define HSTATE_ATTR(_name) \
	static struct kobj_attribute _name##_attr = \
		__ATTR(_name, 0644, _name##_show, _name##_store)

static struct kobject *hugepages_kobj;
static struct kobject *hstate_kobjs[HUGE_MAX_HSTATE];

2835 2836 2837
static struct hstate *kobj_to_node_hstate(struct kobject *kobj, int *nidp);

static struct hstate *kobj_to_hstate(struct kobject *kobj, int *nidp)
2838 2839
{
	int i;
2840

2841
	for (i = 0; i < HUGE_MAX_HSTATE; i++)
2842 2843 2844
		if (hstate_kobjs[i] == kobj) {
			if (nidp)
				*nidp = NUMA_NO_NODE;
2845
			return &hstates[i];
2846 2847 2848
		}

	return kobj_to_node_hstate(kobj, nidp);
2849 2850
}

2851
static ssize_t nr_hugepages_show_common(struct kobject *kobj,
2852 2853
					struct kobj_attribute *attr, char *buf)
{
2854 2855 2856 2857 2858 2859 2860 2861 2862 2863 2864
	struct hstate *h;
	unsigned long nr_huge_pages;
	int nid;

	h = kobj_to_hstate(kobj, &nid);
	if (nid == NUMA_NO_NODE)
		nr_huge_pages = h->nr_huge_pages;
	else
		nr_huge_pages = h->nr_huge_pages_node[nid];

	return sprintf(buf, "%lu\n", nr_huge_pages);
2865
}
2866

2867 2868 2869
static ssize_t __nr_hugepages_store_common(bool obey_mempolicy,
					   struct hstate *h, int nid,
					   unsigned long count, size_t len)
2870 2871
{
	int err;
2872
	nodemask_t nodes_allowed, *n_mask;
2873

2874 2875
	if (hstate_is_gigantic(h) && !gigantic_page_runtime_supported())
		return -EINVAL;
2876

2877 2878 2879 2880 2881
	if (nid == NUMA_NO_NODE) {
		/*
		 * global hstate attribute
		 */
		if (!(obey_mempolicy &&
2882 2883 2884 2885 2886
				init_nodemask_of_mempolicy(&nodes_allowed)))
			n_mask = &node_states[N_MEMORY];
		else
			n_mask = &nodes_allowed;
	} else {
2887
		/*
2888 2889
		 * Node specific request.  count adjustment happens in
		 * set_max_huge_pages() after acquiring hugetlb_lock.
2890
		 */
2891 2892
		init_nodemask_of_node(&nodes_allowed, nid);
		n_mask = &nodes_allowed;
2893
	}
2894

2895
	err = set_max_huge_pages(h, count, nid, n_mask);
2896

2897
	return err ? err : len;
2898 2899
}

2900 2901 2902 2903 2904 2905 2906 2907 2908 2909 2910 2911 2912 2913 2914 2915 2916
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);
}

2917 2918 2919 2920 2921 2922 2923 2924 2925
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)
{
2926
	return nr_hugepages_store_common(false, kobj, buf, len);
2927 2928 2929
}
HSTATE_ATTR(nr_hugepages);

2930 2931 2932 2933 2934 2935 2936 2937 2938 2939 2940 2941 2942 2943 2944
#ifdef CONFIG_NUMA

/*
 * hstate attribute for optionally mempolicy-based constraint on persistent
 * huge page alloc/free.
 */
static ssize_t nr_hugepages_mempolicy_show(struct kobject *kobj,
				       struct kobj_attribute *attr, char *buf)
{
	return nr_hugepages_show_common(kobj, attr, buf);
}

static ssize_t nr_hugepages_mempolicy_store(struct kobject *kobj,
	       struct kobj_attribute *attr, const char *buf, size_t len)
{
2945
	return nr_hugepages_store_common(true, kobj, buf, len);
2946 2947 2948 2949 2950
}
HSTATE_ATTR(nr_hugepages_mempolicy);
#endif


2951 2952 2953
static ssize_t nr_overcommit_hugepages_show(struct kobject *kobj,
					struct kobj_attribute *attr, char *buf)
{
2954
	struct hstate *h = kobj_to_hstate(kobj, NULL);
2955 2956
	return sprintf(buf, "%lu\n", h->nr_overcommit_huge_pages);
}
2957

2958 2959 2960 2961 2962
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;
2963
	struct hstate *h = kobj_to_hstate(kobj, NULL);
2964

2965
	if (hstate_is_gigantic(h))
2966 2967
		return -EINVAL;

2968
	err = kstrtoul(buf, 10, &input);
2969
	if (err)
2970
		return err;
2971 2972 2973 2974 2975 2976 2977 2978 2979 2980 2981 2982

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

	return count;
}
HSTATE_ATTR(nr_overcommit_hugepages);

static ssize_t free_hugepages_show(struct kobject *kobj,
					struct kobj_attribute *attr, char *buf)
{
2983 2984 2985 2986 2987 2988 2989 2990 2991 2992 2993
	struct hstate *h;
	unsigned long free_huge_pages;
	int nid;

	h = kobj_to_hstate(kobj, &nid);
	if (nid == NUMA_NO_NODE)
		free_huge_pages = h->free_huge_pages;
	else
		free_huge_pages = h->free_huge_pages_node[nid];

	return sprintf(buf, "%lu\n", free_huge_pages);
2994 2995 2996 2997 2998 2999
}
HSTATE_ATTR_RO(free_hugepages);

static ssize_t resv_hugepages_show(struct kobject *kobj,
					struct kobj_attribute *attr, char *buf)
{
3000
	struct hstate *h = kobj_to_hstate(kobj, NULL);
3001 3002 3003 3004 3005 3006 3007
	return sprintf(buf, "%lu\n", h->resv_huge_pages);
}
HSTATE_ATTR_RO(resv_hugepages);

static ssize_t surplus_hugepages_show(struct kobject *kobj,
					struct kobj_attribute *attr, char *buf)
{
3008 3009 3010 3011 3012 3013 3014 3015 3016 3017 3018
	struct hstate *h;
	unsigned long surplus_huge_pages;
	int nid;

	h = kobj_to_hstate(kobj, &nid);
	if (nid == NUMA_NO_NODE)
		surplus_huge_pages = h->surplus_huge_pages;
	else
		surplus_huge_pages = h->surplus_huge_pages_node[nid];

	return sprintf(buf, "%lu\n", surplus_huge_pages);
3019 3020 3021 3022 3023 3024 3025 3026 3027
}
HSTATE_ATTR_RO(surplus_hugepages);

static struct attribute *hstate_attrs[] = {
	&nr_hugepages_attr.attr,
	&nr_overcommit_hugepages_attr.attr,
	&free_hugepages_attr.attr,
	&resv_hugepages_attr.attr,
	&surplus_hugepages_attr.attr,
3028 3029 3030
#ifdef CONFIG_NUMA
	&nr_hugepages_mempolicy_attr.attr,
#endif
3031 3032 3033
	NULL,
};

3034
static const struct attribute_group hstate_attr_group = {
3035 3036 3037
	.attrs = hstate_attrs,
};

J
Jeff Mahoney 已提交
3038 3039
static int hugetlb_sysfs_add_hstate(struct hstate *h, struct kobject *parent,
				    struct kobject **hstate_kobjs,
3040
				    const struct attribute_group *hstate_attr_group)
3041 3042
{
	int retval;
3043
	int hi = hstate_index(h);
3044

3045 3046
	hstate_kobjs[hi] = kobject_create_and_add(h->name, parent);
	if (!hstate_kobjs[hi])
3047 3048
		return -ENOMEM;

3049
	retval = sysfs_create_group(hstate_kobjs[hi], hstate_attr_group);
3050
	if (retval)
3051
		kobject_put(hstate_kobjs[hi]);
3052 3053 3054 3055 3056 3057 3058 3059 3060 3061 3062 3063 3064 3065

	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) {
3066 3067
		err = hugetlb_sysfs_add_hstate(h, hugepages_kobj,
					 hstate_kobjs, &hstate_attr_group);
3068
		if (err)
3069
			pr_err("HugeTLB: Unable to add hstate %s", h->name);
3070 3071 3072
	}
}

3073 3074 3075 3076
#ifdef CONFIG_NUMA

/*
 * node_hstate/s - associate per node hstate attributes, via their kobjects,
3077 3078 3079
 * 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
3080 3081 3082 3083 3084 3085
 * the base kernel, on the hugetlb module.
 */
struct node_hstate {
	struct kobject		*hugepages_kobj;
	struct kobject		*hstate_kobjs[HUGE_MAX_HSTATE];
};
3086
static struct node_hstate node_hstates[MAX_NUMNODES];
3087 3088

/*
3089
 * A subset of global hstate attributes for node devices
3090 3091 3092 3093 3094 3095 3096 3097
 */
static struct attribute *per_node_hstate_attrs[] = {
	&nr_hugepages_attr.attr,
	&free_hugepages_attr.attr,
	&surplus_hugepages_attr.attr,
	NULL,
};

3098
static const struct attribute_group per_node_hstate_attr_group = {
3099 3100 3101 3102
	.attrs = per_node_hstate_attrs,
};

/*
3103
 * kobj_to_node_hstate - lookup global hstate for node device hstate attr kobj.
3104 3105 3106 3107 3108 3109 3110 3111 3112 3113 3114 3115 3116 3117 3118 3119 3120 3121 3122 3123 3124 3125
 * 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;
}

/*
3126
 * Unregister hstate attributes from a single node device.
3127 3128
 * No-op if no hstate attributes attached.
 */
3129
static void hugetlb_unregister_node(struct node *node)
3130 3131
{
	struct hstate *h;
3132
	struct node_hstate *nhs = &node_hstates[node->dev.id];
3133 3134

	if (!nhs->hugepages_kobj)
3135
		return;		/* no hstate attributes */
3136

3137 3138 3139 3140 3141
	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;
3142
		}
3143
	}
3144 3145 3146 3147 3148 3149 3150

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


/*
3151
 * Register hstate attributes for a single node device.
3152 3153
 * No-op if attributes already registered.
 */
3154
static void hugetlb_register_node(struct node *node)
3155 3156
{
	struct hstate *h;
3157
	struct node_hstate *nhs = &node_hstates[node->dev.id];
3158 3159 3160 3161 3162 3163
	int err;

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

	nhs->hugepages_kobj = kobject_create_and_add("hugepages",
3164
							&node->dev.kobj);
3165 3166 3167 3168 3169 3170 3171 3172
	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) {
3173
			pr_err("HugeTLB: Unable to add hstate %s for node %d\n",
3174
				h->name, node->dev.id);
3175 3176 3177 3178 3179 3180 3181
			hugetlb_unregister_node(node);
			break;
		}
	}
}

/*
3182
 * hugetlb init time:  register hstate attributes for all registered node
3183 3184
 * devices of nodes that have memory.  All on-line nodes should have
 * registered their associated device by this time.
3185
 */
3186
static void __init hugetlb_register_all_nodes(void)
3187 3188 3189
{
	int nid;

3190
	for_each_node_state(nid, N_MEMORY) {
3191
		struct node *node = node_devices[nid];
3192
		if (node->dev.id == nid)
3193 3194 3195 3196
			hugetlb_register_node(node);
	}

	/*
3197
	 * Let the node device driver know we're here so it can
3198 3199 3200 3201 3202 3203 3204 3205 3206 3207 3208 3209 3210 3211 3212 3213 3214 3215 3216
	 * [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

3217 3218
static int __init hugetlb_init(void)
{
3219 3220
	int i;

3221 3222 3223
	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");
3224
		return 0;
3225
	}
3226

3227 3228 3229 3230 3231 3232 3233 3234 3235 3236 3237 3238 3239 3240 3241 3242 3243 3244 3245 3246 3247 3248 3249 3250 3251 3252 3253 3254
	/*
	 * 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;
3255
		}
3256
	}
3257

3258
	hugetlb_cma_check();
3259
	hugetlb_init_hstates();
3260
	gather_bootmem_prealloc();
3261 3262 3263
	report_hugepages();

	hugetlb_sysfs_init();
3264
	hugetlb_register_all_nodes();
3265
	hugetlb_cgroup_file_init();
3266

3267 3268 3269 3270 3271
#ifdef CONFIG_SMP
	num_fault_mutexes = roundup_pow_of_two(8 * num_possible_cpus());
#else
	num_fault_mutexes = 1;
#endif
3272
	hugetlb_fault_mutex_table =
3273 3274
		kmalloc_array(num_fault_mutexes, sizeof(struct mutex),
			      GFP_KERNEL);
3275
	BUG_ON(!hugetlb_fault_mutex_table);
3276 3277

	for (i = 0; i < num_fault_mutexes; i++)
3278
		mutex_init(&hugetlb_fault_mutex_table[i]);
3279 3280
	return 0;
}
3281
subsys_initcall(hugetlb_init);
3282

3283 3284
/* Overwritten by architectures with more huge page sizes */
bool __init __attribute((weak)) arch_hugetlb_valid_size(unsigned long size)
3285
{
3286
	return size == HPAGE_SIZE;
3287 3288
}

3289
void __init hugetlb_add_hstate(unsigned int order)
3290 3291
{
	struct hstate *h;
3292 3293
	unsigned long i;

3294 3295 3296
	if (size_to_hstate(PAGE_SIZE << order)) {
		return;
	}
3297
	BUG_ON(hugetlb_max_hstate >= HUGE_MAX_HSTATE);
3298
	BUG_ON(order == 0);
3299
	h = &hstates[hugetlb_max_hstate++];
3300 3301
	h->order = order;
	h->mask = ~((1ULL << (order + PAGE_SHIFT)) - 1);
3302 3303 3304 3305
	h->nr_huge_pages = 0;
	h->free_huge_pages = 0;
	for (i = 0; i < MAX_NUMNODES; ++i)
		INIT_LIST_HEAD(&h->hugepage_freelists[i]);
3306
	INIT_LIST_HEAD(&h->hugepage_activelist);
3307 3308
	h->next_nid_to_alloc = first_memory_node;
	h->next_nid_to_free = first_memory_node;
3309 3310
	snprintf(h->name, HSTATE_NAME_LEN, "hugepages-%lukB",
					huge_page_size(h)/1024);
3311

3312 3313 3314
	parsed_hstate = h;
}

3315 3316 3317 3318 3319 3320 3321 3322
/*
 * 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)
3323 3324
{
	unsigned long *mhp;
3325
	static unsigned long *last_mhp;
3326

3327
	if (!parsed_valid_hugepagesz) {
3328
		pr_warn("HugeTLB: hugepages=%s does not follow a valid hugepagesz, ignoring\n", s);
3329
		parsed_valid_hugepagesz = true;
3330
		return 0;
3331
	}
3332

3333
	/*
3334 3335 3336 3337
	 * !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.
3338
	 */
3339
	else if (!hugetlb_max_hstate)
3340 3341 3342 3343
		mhp = &default_hstate_max_huge_pages;
	else
		mhp = &parsed_hstate->max_huge_pages;

3344
	if (mhp == last_mhp) {
3345 3346
		pr_warn("HugeTLB: hugepages= specified twice without interleaving hugepagesz=, ignoring hugepages=%s\n", s);
		return 0;
3347 3348
	}

3349 3350 3351
	if (sscanf(s, "%lu", mhp) <= 0)
		*mhp = 0;

3352 3353 3354 3355 3356
	/*
	 * Global state is always initialized later in hugetlb_init.
	 * But we need to allocate >= MAX_ORDER hstates here early to still
	 * use the bootmem allocator.
	 */
3357
	if (hugetlb_max_hstate && parsed_hstate->order >= MAX_ORDER)
3358 3359 3360 3361
		hugetlb_hstate_alloc_pages(parsed_hstate);

	last_mhp = mhp;

3362 3363
	return 1;
}
3364
__setup("hugepages=", hugepages_setup);
3365

3366 3367 3368 3369 3370 3371 3372
/*
 * 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.
 */
3373
static int __init hugepagesz_setup(char *s)
3374
{
3375
	unsigned long size;
3376 3377 3378
	struct hstate *h;

	parsed_valid_hugepagesz = false;
3379 3380 3381
	size = (unsigned long)memparse(s, NULL);

	if (!arch_hugetlb_valid_size(size)) {
3382
		pr_err("HugeTLB: unsupported hugepagesz=%s\n", s);
3383 3384 3385
		return 0;
	}

3386 3387 3388 3389 3390 3391 3392 3393 3394 3395 3396 3397 3398 3399 3400 3401 3402 3403 3404 3405 3406 3407 3408
	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;
3409 3410
	}

3411
	hugetlb_add_hstate(ilog2(size) - PAGE_SHIFT);
3412
	parsed_valid_hugepagesz = true;
3413 3414
	return 1;
}
3415 3416
__setup("hugepagesz=", hugepagesz_setup);

3417 3418 3419 3420
/*
 * default_hugepagesz command line input
 * Only one instance of default_hugepagesz allowed on command line.
 */
3421
static int __init default_hugepagesz_setup(char *s)
3422
{
3423 3424
	unsigned long size;

3425 3426 3427 3428 3429 3430
	parsed_valid_hugepagesz = false;
	if (parsed_default_hugepagesz) {
		pr_err("HugeTLB: default_hugepagesz previously specified, ignoring %s\n", s);
		return 0;
	}

3431 3432 3433
	size = (unsigned long)memparse(s, NULL);

	if (!arch_hugetlb_valid_size(size)) {
3434
		pr_err("HugeTLB: unsupported default_hugepagesz=%s\n", s);
3435 3436 3437
		return 0;
	}

3438 3439 3440 3441 3442 3443 3444 3445 3446 3447 3448 3449 3450 3451 3452 3453 3454 3455 3456
	hugetlb_add_hstate(ilog2(size) - PAGE_SHIFT);
	parsed_valid_hugepagesz = true;
	parsed_default_hugepagesz = true;
	default_hstate_idx = hstate_index(size_to_hstate(size));

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

3457 3458
	return 1;
}
3459
__setup("default_hugepagesz=", default_hugepagesz_setup);
3460

3461
static unsigned int allowed_mems_nr(struct hstate *h)
3462 3463 3464
{
	int node;
	unsigned int nr = 0;
3465 3466 3467 3468 3469
	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);
3470

3471 3472 3473 3474 3475
	for_each_node_mask(node, cpuset_current_mems_allowed) {
		if (!mpol_allowed ||
		    (mpol_allowed && node_isset(node, *mpol_allowed)))
			nr += array[node];
	}
3476 3477 3478 3479 3480

	return nr;
}

#ifdef CONFIG_SYSCTL
3481 3482
static int hugetlb_sysctl_handler_common(bool obey_mempolicy,
			 struct ctl_table *table, int write,
3483
			 void *buffer, size_t *length, loff_t *ppos)
L
Linus Torvalds 已提交
3484
{
3485
	struct hstate *h = &default_hstate;
3486
	unsigned long tmp = h->max_huge_pages;
3487
	int ret;
3488

3489
	if (!hugepages_supported())
3490
		return -EOPNOTSUPP;
3491

3492 3493
	table->data = &tmp;
	table->maxlen = sizeof(unsigned long);
3494 3495 3496
	ret = proc_doulongvec_minmax(table, write, buffer, length, ppos);
	if (ret)
		goto out;
3497

3498 3499 3500
	if (write)
		ret = __nr_hugepages_store_common(obey_mempolicy, h,
						  NUMA_NO_NODE, tmp, *length);
3501 3502
out:
	return ret;
L
Linus Torvalds 已提交
3503
}
3504

3505
int hugetlb_sysctl_handler(struct ctl_table *table, int write,
3506
			  void *buffer, size_t *length, loff_t *ppos)
3507 3508 3509 3510 3511 3512 3513 3514
{

	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,
3515
			  void *buffer, size_t *length, loff_t *ppos)
3516 3517 3518 3519 3520 3521
{
	return hugetlb_sysctl_handler_common(true, table, write,
							buffer, length, ppos);
}
#endif /* CONFIG_NUMA */

3522
int hugetlb_overcommit_handler(struct ctl_table *table, int write,
3523
		void *buffer, size_t *length, loff_t *ppos)
3524
{
3525
	struct hstate *h = &default_hstate;
3526
	unsigned long tmp;
3527
	int ret;
3528

3529
	if (!hugepages_supported())
3530
		return -EOPNOTSUPP;
3531

3532
	tmp = h->nr_overcommit_huge_pages;
3533

3534
	if (write && hstate_is_gigantic(h))
3535 3536
		return -EINVAL;

3537 3538
	table->data = &tmp;
	table->maxlen = sizeof(unsigned long);
3539 3540 3541
	ret = proc_doulongvec_minmax(table, write, buffer, length, ppos);
	if (ret)
		goto out;
3542 3543 3544 3545 3546 3547

	if (write) {
		spin_lock(&hugetlb_lock);
		h->nr_overcommit_huge_pages = tmp;
		spin_unlock(&hugetlb_lock);
	}
3548 3549
out:
	return ret;
3550 3551
}

L
Linus Torvalds 已提交
3552 3553
#endif /* CONFIG_SYSCTL */

3554
void hugetlb_report_meminfo(struct seq_file *m)
L
Linus Torvalds 已提交
3555
{
3556 3557 3558
	struct hstate *h;
	unsigned long total = 0;

3559 3560
	if (!hugepages_supported())
		return;
3561 3562 3563 3564 3565 3566 3567 3568 3569 3570 3571 3572 3573 3574 3575 3576 3577 3578 3579 3580 3581

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

		total += (PAGE_SIZE << huge_page_order(h)) * count;

		if (h == &default_hstate)
			seq_printf(m,
				   "HugePages_Total:   %5lu\n"
				   "HugePages_Free:    %5lu\n"
				   "HugePages_Rsvd:    %5lu\n"
				   "HugePages_Surp:    %5lu\n"
				   "Hugepagesize:   %8lu kB\n",
				   count,
				   h->free_huge_pages,
				   h->resv_huge_pages,
				   h->surplus_huge_pages,
				   (PAGE_SIZE << huge_page_order(h)) / 1024);
	}

	seq_printf(m, "Hugetlb:        %8lu kB\n", total / 1024);
L
Linus Torvalds 已提交
3582 3583 3584 3585
}

int hugetlb_report_node_meminfo(int nid, char *buf)
{
3586
	struct hstate *h = &default_hstate;
3587 3588
	if (!hugepages_supported())
		return 0;
L
Linus Torvalds 已提交
3589 3590
	return sprintf(buf,
		"Node %d HugePages_Total: %5u\n"
3591 3592
		"Node %d HugePages_Free:  %5u\n"
		"Node %d HugePages_Surp:  %5u\n",
3593 3594 3595
		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
3596 3597
}

3598 3599 3600 3601 3602
void hugetlb_show_meminfo(void)
{
	struct hstate *h;
	int nid;

3603 3604 3605
	if (!hugepages_supported())
		return;

3606 3607 3608 3609 3610 3611 3612 3613 3614 3615
	for_each_node_state(nid, N_MEMORY)
		for_each_hstate(h)
			pr_info("Node %d hugepages_total=%u hugepages_free=%u hugepages_surp=%u hugepages_size=%lukB\n",
				nid,
				h->nr_huge_pages_node[nid],
				h->free_huge_pages_node[nid],
				h->surplus_huge_pages_node[nid],
				1UL << (huge_page_order(h) + PAGE_SHIFT - 10));
}

3616 3617 3618 3619 3620 3621
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
3622 3623 3624
/* Return the number pages of memory we physically have, in PAGE_SIZE units. */
unsigned long hugetlb_total_pages(void)
{
3625 3626 3627 3628 3629 3630
	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
3631 3632
}

3633
static int hugetlb_acct_memory(struct hstate *h, long delta)
M
Mel Gorman 已提交
3634 3635 3636 3637 3638 3639 3640 3641 3642 3643 3644 3645 3646 3647 3648 3649 3650 3651 3652 3653
{
	int ret = -ENOMEM;

	spin_lock(&hugetlb_lock);
	/*
	 * When cpuset is configured, it breaks the strict hugetlb page
	 * reservation as the accounting is done on a global variable. Such
	 * reservation is completely rubbish in the presence of cpuset because
	 * the reservation is not checked against page availability for the
	 * current cpuset. Application can still potentially OOM'ed by kernel
	 * with lack of free htlb page in cpuset that the task is in.
	 * Attempt to enforce strict accounting with cpuset is almost
	 * impossible (or too ugly) because cpuset is too fluid that
	 * task or memory node can be dynamically moved between cpusets.
	 *
	 * The change of semantics for shared hugetlb mapping with cpuset is
	 * undesirable. However, in order to preserve some of the semantics,
	 * we fall back to check against current free page availability as
	 * a best attempt and hopefully to minimize the impact of changing
	 * semantics that cpuset has.
3654 3655 3656 3657 3658 3659
	 *
	 * 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
3660 3661
	 */
	if (delta > 0) {
3662
		if (gather_surplus_pages(h, delta) < 0)
M
Mel Gorman 已提交
3663 3664
			goto out;

3665
		if (delta > allowed_mems_nr(h)) {
3666
			return_unused_surplus_pages(h, delta);
M
Mel Gorman 已提交
3667 3668 3669 3670 3671 3672
			goto out;
		}
	}

	ret = 0;
	if (delta < 0)
3673
		return_unused_surplus_pages(h, (unsigned long) -delta);
M
Mel Gorman 已提交
3674 3675 3676 3677 3678 3679

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

3680 3681
static void hugetlb_vm_op_open(struct vm_area_struct *vma)
{
3682
	struct resv_map *resv = vma_resv_map(vma);
3683 3684 3685 3686 3687

	/*
	 * 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
3688
	 * has a reference to the reservation map it cannot disappear until
3689 3690 3691
	 * after this open call completes.  It is therefore safe to take a
	 * new reference here without additional locking.
	 */
3692
	if (resv && is_vma_resv_set(vma, HPAGE_RESV_OWNER))
3693
		kref_get(&resv->refs);
3694 3695
}

3696 3697
static void hugetlb_vm_op_close(struct vm_area_struct *vma)
{
3698
	struct hstate *h = hstate_vma(vma);
3699
	struct resv_map *resv = vma_resv_map(vma);
3700
	struct hugepage_subpool *spool = subpool_vma(vma);
3701
	unsigned long reserve, start, end;
3702
	long gbl_reserve;
3703

3704 3705
	if (!resv || !is_vma_resv_set(vma, HPAGE_RESV_OWNER))
		return;
3706

3707 3708
	start = vma_hugecache_offset(h, vma, vma->vm_start);
	end = vma_hugecache_offset(h, vma, vma->vm_end);
3709

3710
	reserve = (end - start) - region_count(resv, start, end);
3711
	hugetlb_cgroup_uncharge_counter(resv, start, end);
3712
	if (reserve) {
3713 3714 3715 3716 3717 3718
		/*
		 * 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);
3719
	}
3720 3721

	kref_put(&resv->refs, resv_map_release);
3722 3723
}

3724 3725 3726 3727 3728 3729 3730
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;
}

3731 3732 3733 3734 3735 3736 3737
static unsigned long hugetlb_vm_op_pagesize(struct vm_area_struct *vma)
{
	struct hstate *hstate = hstate_vma(vma);

	return 1UL << huge_page_shift(hstate);
}

L
Linus Torvalds 已提交
3738 3739 3740 3741 3742 3743
/*
 * We cannot handle pagefaults against hugetlb pages at all.  They cause
 * handle_mm_fault() to try to instantiate regular-sized pages in the
 * hugegpage VMA.  do_page_fault() is supposed to trap this, so BUG is we get
 * this far.
 */
3744
static vm_fault_t hugetlb_vm_op_fault(struct vm_fault *vmf)
L
Linus Torvalds 已提交
3745 3746
{
	BUG();
N
Nick Piggin 已提交
3747
	return 0;
L
Linus Torvalds 已提交
3748 3749
}

3750 3751 3752 3753 3754 3755 3756
/*
 * 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.
 */
3757
const struct vm_operations_struct hugetlb_vm_ops = {
N
Nick Piggin 已提交
3758
	.fault = hugetlb_vm_op_fault,
3759
	.open = hugetlb_vm_op_open,
3760
	.close = hugetlb_vm_op_close,
3761
	.split = hugetlb_vm_op_split,
3762
	.pagesize = hugetlb_vm_op_pagesize,
L
Linus Torvalds 已提交
3763 3764
};

3765 3766
static pte_t make_huge_pte(struct vm_area_struct *vma, struct page *page,
				int writable)
D
David Gibson 已提交
3767 3768 3769
{
	pte_t entry;

3770
	if (writable) {
3771 3772
		entry = huge_pte_mkwrite(huge_pte_mkdirty(mk_huge_pte(page,
					 vma->vm_page_prot)));
D
David Gibson 已提交
3773
	} else {
3774 3775
		entry = huge_pte_wrprotect(mk_huge_pte(page,
					   vma->vm_page_prot));
D
David Gibson 已提交
3776 3777 3778
	}
	entry = pte_mkyoung(entry);
	entry = pte_mkhuge(entry);
3779
	entry = arch_make_huge_pte(entry, vma, page, writable);
D
David Gibson 已提交
3780 3781 3782 3783

	return entry;
}

3784 3785 3786 3787 3788
static void set_huge_ptep_writable(struct vm_area_struct *vma,
				   unsigned long address, pte_t *ptep)
{
	pte_t entry;

3789
	entry = huge_pte_mkwrite(huge_pte_mkdirty(huge_ptep_get(ptep)));
3790
	if (huge_ptep_set_access_flags(vma, address, ptep, entry, 1))
3791
		update_mmu_cache(vma, address, ptep);
3792 3793
}

3794
bool is_hugetlb_entry_migration(pte_t pte)
3795 3796 3797 3798
{
	swp_entry_t swp;

	if (huge_pte_none(pte) || pte_present(pte))
3799
		return false;
3800 3801
	swp = pte_to_swp_entry(pte);
	if (non_swap_entry(swp) && is_migration_entry(swp))
3802
		return true;
3803
	else
3804
		return false;
3805 3806 3807 3808 3809 3810 3811 3812 3813 3814 3815 3816 3817 3818
}

static int is_hugetlb_entry_hwpoisoned(pte_t pte)
{
	swp_entry_t swp;

	if (huge_pte_none(pte) || pte_present(pte))
		return 0;
	swp = pte_to_swp_entry(pte);
	if (non_swap_entry(swp) && is_hwpoison_entry(swp))
		return 1;
	else
		return 0;
}
3819

D
David Gibson 已提交
3820 3821 3822
int copy_hugetlb_page_range(struct mm_struct *dst, struct mm_struct *src,
			    struct vm_area_struct *vma)
{
3823
	pte_t *src_pte, *dst_pte, entry, dst_entry;
D
David Gibson 已提交
3824
	struct page *ptepage;
3825
	unsigned long addr;
3826
	int cow;
3827 3828
	struct hstate *h = hstate_vma(vma);
	unsigned long sz = huge_page_size(h);
3829
	struct address_space *mapping = vma->vm_file->f_mapping;
3830
	struct mmu_notifier_range range;
3831
	int ret = 0;
3832 3833

	cow = (vma->vm_flags & (VM_SHARED | VM_MAYWRITE)) == VM_MAYWRITE;
D
David Gibson 已提交
3834

3835
	if (cow) {
3836
		mmu_notifier_range_init(&range, MMU_NOTIFY_CLEAR, 0, vma, src,
3837
					vma->vm_start,
3838 3839
					vma->vm_end);
		mmu_notifier_invalidate_range_start(&range);
3840 3841 3842 3843 3844 3845 3846 3847
	} 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);
3848
	}
3849

3850
	for (addr = vma->vm_start; addr < vma->vm_end; addr += sz) {
3851
		spinlock_t *src_ptl, *dst_ptl;
3852
		src_pte = huge_pte_offset(src, addr, sz);
H
Hugh Dickins 已提交
3853 3854
		if (!src_pte)
			continue;
3855
		dst_pte = huge_pte_alloc(dst, addr, sz);
3856 3857 3858 3859
		if (!dst_pte) {
			ret = -ENOMEM;
			break;
		}
3860

3861 3862 3863 3864 3865 3866 3867 3868 3869 3870 3871
		/*
		 * 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))
3872 3873
			continue;

3874 3875 3876
		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);
3877
		entry = huge_ptep_get(src_pte);
3878 3879 3880 3881 3882 3883 3884
		dst_entry = huge_ptep_get(dst_pte);
		if (huge_pte_none(entry) || !huge_pte_none(dst_entry)) {
			/*
			 * Skip if src entry none.  Also, skip in the
			 * unlikely case dst entry !none as this implies
			 * sharing with another vma.
			 */
3885 3886 3887 3888 3889 3890 3891 3892 3893 3894 3895 3896
			;
		} else if (unlikely(is_hugetlb_entry_migration(entry) ||
				    is_hugetlb_entry_hwpoisoned(entry))) {
			swp_entry_t swp_entry = pte_to_swp_entry(entry);

			if (is_write_migration_entry(swp_entry) && cow) {
				/*
				 * COW mappings require pages in both
				 * parent and child to be set to read.
				 */
				make_migration_entry_read(&swp_entry);
				entry = swp_entry_to_pte(swp_entry);
3897 3898
				set_huge_swap_pte_at(src, addr, src_pte,
						     entry, sz);
3899
			}
3900
			set_huge_swap_pte_at(dst, addr, dst_pte, entry, sz);
3901
		} else {
3902
			if (cow) {
3903 3904 3905 3906 3907
				/*
				 * No need to notify as we are downgrading page
				 * table protection not changing it to point
				 * to a new page.
				 *
3908
				 * See Documentation/vm/mmu_notifier.rst
3909
				 */
3910
				huge_ptep_set_wrprotect(src, addr, src_pte);
3911
			}
3912
			entry = huge_ptep_get(src_pte);
3913 3914
			ptepage = pte_page(entry);
			get_page(ptepage);
3915
			page_dup_rmap(ptepage, true);
3916
			set_huge_pte_at(dst, addr, dst_pte, entry);
3917
			hugetlb_count_add(pages_per_huge_page(h), dst);
3918
		}
3919 3920
		spin_unlock(src_ptl);
		spin_unlock(dst_ptl);
D
David Gibson 已提交
3921 3922
	}

3923
	if (cow)
3924
		mmu_notifier_invalidate_range_end(&range);
3925 3926
	else
		i_mmap_unlock_read(mapping);
3927 3928

	return ret;
D
David Gibson 已提交
3929 3930
}

3931 3932 3933
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
3934 3935 3936
{
	struct mm_struct *mm = vma->vm_mm;
	unsigned long address;
3937
	pte_t *ptep;
D
David Gibson 已提交
3938
	pte_t pte;
3939
	spinlock_t *ptl;
D
David Gibson 已提交
3940
	struct page *page;
3941 3942
	struct hstate *h = hstate_vma(vma);
	unsigned long sz = huge_page_size(h);
3943
	struct mmu_notifier_range range;
3944

D
David Gibson 已提交
3945
	WARN_ON(!is_vm_hugetlb_page(vma));
3946 3947
	BUG_ON(start & ~huge_page_mask(h));
	BUG_ON(end & ~huge_page_mask(h));
D
David Gibson 已提交
3948

3949 3950 3951 3952
	/*
	 * This is a hugetlb vma, all the pte entries should point
	 * to huge page.
	 */
3953
	tlb_change_page_size(tlb, sz);
3954
	tlb_start_vma(tlb, vma);
3955 3956 3957 3958

	/*
	 * If sharing possible, alert mmu notifiers of worst case.
	 */
3959 3960
	mmu_notifier_range_init(&range, MMU_NOTIFY_UNMAP, 0, vma, mm, start,
				end);
3961 3962
	adjust_range_if_pmd_sharing_possible(vma, &range.start, &range.end);
	mmu_notifier_invalidate_range_start(&range);
3963 3964
	address = start;
	for (; address < end; address += sz) {
3965
		ptep = huge_pte_offset(mm, address, sz);
A
Adam Litke 已提交
3966
		if (!ptep)
3967 3968
			continue;

3969
		ptl = huge_pte_lock(h, mm, ptep);
3970
		if (huge_pmd_unshare(mm, vma, &address, ptep)) {
3971
			spin_unlock(ptl);
3972 3973 3974 3975
			/*
			 * We just unmapped a page of PMDs by clearing a PUD.
			 * The caller's TLB flush range should cover this area.
			 */
3976 3977
			continue;
		}
3978

3979
		pte = huge_ptep_get(ptep);
3980 3981 3982 3983
		if (huge_pte_none(pte)) {
			spin_unlock(ptl);
			continue;
		}
3984 3985

		/*
3986 3987
		 * Migrating hugepage or HWPoisoned hugepage is already
		 * unmapped and its refcount is dropped, so just clear pte here.
3988
		 */
3989
		if (unlikely(!pte_present(pte))) {
3990
			huge_pte_clear(mm, address, ptep, sz);
3991 3992
			spin_unlock(ptl);
			continue;
3993
		}
3994 3995

		page = pte_page(pte);
3996 3997 3998 3999 4000 4001
		/*
		 * 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) {
4002 4003 4004 4005
			if (page != ref_page) {
				spin_unlock(ptl);
				continue;
			}
4006 4007 4008 4009 4010 4011 4012 4013
			/*
			 * 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);
		}

4014
		pte = huge_ptep_get_and_clear(mm, address, ptep);
4015
		tlb_remove_huge_tlb_entry(h, tlb, ptep, address);
4016
		if (huge_pte_dirty(pte))
4017
			set_page_dirty(page);
4018

4019
		hugetlb_count_sub(pages_per_huge_page(h), mm);
4020
		page_remove_rmap(page, true);
4021

4022
		spin_unlock(ptl);
4023
		tlb_remove_page_size(tlb, page, huge_page_size(h));
4024 4025 4026 4027 4028
		/*
		 * Bail out after unmapping reference page if supplied
		 */
		if (ref_page)
			break;
4029
	}
4030
	mmu_notifier_invalidate_range_end(&range);
4031
	tlb_end_vma(tlb, vma);
L
Linus Torvalds 已提交
4032
}
D
David Gibson 已提交
4033

4034 4035 4036 4037 4038 4039 4040 4041 4042 4043 4044 4045
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
4046
	 * is to clear it before releasing the i_mmap_rwsem. This works
4047
	 * because in the context this is called, the VMA is about to be
4048
	 * destroyed and the i_mmap_rwsem is held.
4049 4050 4051 4052
	 */
	vma->vm_flags &= ~VM_MAYSHARE;
}

4053
void unmap_hugepage_range(struct vm_area_struct *vma, unsigned long start,
4054
			  unsigned long end, struct page *ref_page)
4055
{
4056 4057
	struct mm_struct *mm;
	struct mmu_gather tlb;
4058 4059 4060 4061 4062 4063 4064 4065 4066 4067 4068
	unsigned long tlb_start = start;
	unsigned long tlb_end = end;

	/*
	 * If shared PMDs were possibly used within this vma range, adjust
	 * start/end for worst case tlb flushing.
	 * Note that we can not be sure if PMDs are shared until we try to
	 * unmap pages.  However, we want to make sure TLB flushing covers
	 * the largest possible range.
	 */
	adjust_range_if_pmd_sharing_possible(vma, &tlb_start, &tlb_end);
4069 4070 4071

	mm = vma->vm_mm;

4072
	tlb_gather_mmu(&tlb, mm, tlb_start, tlb_end);
4073
	__unmap_hugepage_range(&tlb, vma, start, end, ref_page);
4074
	tlb_finish_mmu(&tlb, tlb_start, tlb_end);
4075 4076
}

4077 4078 4079 4080 4081 4082
/*
 * This is called when the original mapper is failing to COW a MAP_PRIVATE
 * mappping it owns the reserve page for. The intention is to unmap the page
 * from other VMAs and let the children be SIGKILLed if they are faulting the
 * same region.
 */
4083 4084
static void unmap_ref_private(struct mm_struct *mm, struct vm_area_struct *vma,
			      struct page *page, unsigned long address)
4085
{
4086
	struct hstate *h = hstate_vma(vma);
4087 4088 4089 4090 4091 4092 4093 4094
	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.
	 */
4095
	address = address & huge_page_mask(h);
4096 4097
	pgoff = ((address - vma->vm_start) >> PAGE_SHIFT) +
			vma->vm_pgoff;
4098
	mapping = vma->vm_file->f_mapping;
4099

4100 4101 4102 4103 4104
	/*
	 * 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
	 */
4105
	i_mmap_lock_write(mapping);
4106
	vma_interval_tree_foreach(iter_vma, &mapping->i_mmap, pgoff, pgoff) {
4107 4108 4109 4110
		/* Do not unmap the current VMA */
		if (iter_vma == vma)
			continue;

4111 4112 4113 4114 4115 4116 4117 4118
		/*
		 * 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;

4119 4120 4121 4122 4123 4124 4125 4126
		/*
		 * 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))
4127 4128
			unmap_hugepage_range(iter_vma, address,
					     address + huge_page_size(h), page);
4129
	}
4130
	i_mmap_unlock_write(mapping);
4131 4132
}

4133 4134
/*
 * Hugetlb_cow() should be called with page lock of the original hugepage held.
4135 4136 4137
 * Called with hugetlb_instantiation_mutex held and pte_page locked so we
 * cannot race with other handlers or page migration.
 * Keep the pte_same checks anyway to make transition from the mutex easier.
4138
 */
4139
static vm_fault_t hugetlb_cow(struct mm_struct *mm, struct vm_area_struct *vma,
4140
		       unsigned long address, pte_t *ptep,
4141
		       struct page *pagecache_page, spinlock_t *ptl)
4142
{
4143
	pte_t pte;
4144
	struct hstate *h = hstate_vma(vma);
4145
	struct page *old_page, *new_page;
4146 4147
	int outside_reserve = 0;
	vm_fault_t ret = 0;
4148
	unsigned long haddr = address & huge_page_mask(h);
4149
	struct mmu_notifier_range range;
4150

4151
	pte = huge_ptep_get(ptep);
4152 4153
	old_page = pte_page(pte);

4154
retry_avoidcopy:
4155 4156
	/* If no-one else is actually using this page, avoid the copy
	 * and just make the page writable */
4157
	if (page_mapcount(old_page) == 1 && PageAnon(old_page)) {
4158
		page_move_anon_rmap(old_page, vma);
4159
		set_huge_ptep_writable(vma, haddr, ptep);
N
Nick Piggin 已提交
4160
		return 0;
4161 4162
	}

4163 4164 4165 4166 4167 4168 4169 4170 4171
	/*
	 * 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.
	 */
4172
	if (is_vma_resv_set(vma, HPAGE_RESV_OWNER) &&
4173 4174 4175
			old_page != pagecache_page)
		outside_reserve = 1;

4176
	get_page(old_page);
4177

4178 4179 4180 4181
	/*
	 * Drop page table lock as buddy allocator may be called. It will
	 * be acquired again before returning to the caller, as expected.
	 */
4182
	spin_unlock(ptl);
4183
	new_page = alloc_huge_page(vma, haddr, outside_reserve);
4184

4185
	if (IS_ERR(new_page)) {
4186 4187 4188 4189 4190 4191 4192 4193
		/*
		 * 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) {
4194
			put_page(old_page);
4195
			BUG_ON(huge_pte_none(pte));
4196
			unmap_ref_private(mm, vma, old_page, haddr);
4197 4198
			BUG_ON(huge_pte_none(pte));
			spin_lock(ptl);
4199
			ptep = huge_pte_offset(mm, haddr, huge_page_size(h));
4200 4201 4202 4203 4204 4205 4206 4207
			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;
4208 4209
		}

4210
		ret = vmf_error(PTR_ERR(new_page));
4211
		goto out_release_old;
4212 4213
	}

4214 4215 4216 4217
	/*
	 * When the original hugepage is shared one, it does not have
	 * anon_vma prepared.
	 */
4218
	if (unlikely(anon_vma_prepare(vma))) {
4219 4220
		ret = VM_FAULT_OOM;
		goto out_release_all;
4221
	}
4222

4223
	copy_user_huge_page(new_page, old_page, address, vma,
A
Andrea Arcangeli 已提交
4224
			    pages_per_huge_page(h));
N
Nick Piggin 已提交
4225
	__SetPageUptodate(new_page);
4226

4227
	mmu_notifier_range_init(&range, MMU_NOTIFY_CLEAR, 0, vma, mm, haddr,
4228
				haddr + huge_page_size(h));
4229
	mmu_notifier_invalidate_range_start(&range);
4230

4231
	/*
4232
	 * Retake the page table lock to check for racing updates
4233 4234
	 * before the page tables are altered
	 */
4235
	spin_lock(ptl);
4236
	ptep = huge_pte_offset(mm, haddr, huge_page_size(h));
4237
	if (likely(ptep && pte_same(huge_ptep_get(ptep), pte))) {
4238 4239
		ClearPagePrivate(new_page);

4240
		/* Break COW */
4241
		huge_ptep_clear_flush(vma, haddr, ptep);
4242
		mmu_notifier_invalidate_range(mm, range.start, range.end);
4243
		set_huge_pte_at(mm, haddr, ptep,
4244
				make_huge_pte(vma, new_page, 1));
4245
		page_remove_rmap(old_page, true);
4246
		hugepage_add_new_anon_rmap(new_page, vma, haddr);
4247
		set_page_huge_active(new_page);
4248 4249 4250
		/* Make the old page be freed below */
		new_page = old_page;
	}
4251
	spin_unlock(ptl);
4252
	mmu_notifier_invalidate_range_end(&range);
4253
out_release_all:
4254
	restore_reserve_on_error(h, vma, haddr, new_page);
4255
	put_page(new_page);
4256
out_release_old:
4257
	put_page(old_page);
4258

4259 4260
	spin_lock(ptl); /* Caller expects lock to be held */
	return ret;
4261 4262
}

4263
/* Return the pagecache page at a given address within a VMA */
4264 4265
static struct page *hugetlbfs_pagecache_page(struct hstate *h,
			struct vm_area_struct *vma, unsigned long address)
4266 4267
{
	struct address_space *mapping;
4268
	pgoff_t idx;
4269 4270

	mapping = vma->vm_file->f_mapping;
4271
	idx = vma_hugecache_offset(h, vma, address);
4272 4273 4274 4275

	return find_lock_page(mapping, idx);
}

H
Hugh Dickins 已提交
4276 4277 4278 4279 4280
/*
 * 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
4281 4282 4283 4284 4285 4286 4287 4288 4289 4290 4291 4292 4293 4294 4295
			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;
}

4296 4297 4298 4299 4300 4301 4302 4303 4304 4305 4306
int huge_add_to_page_cache(struct page *page, struct address_space *mapping,
			   pgoff_t idx)
{
	struct inode *inode = mapping->host;
	struct hstate *h = hstate_inode(inode);
	int err = add_to_page_cache(page, mapping, idx, GFP_KERNEL);

	if (err)
		return err;
	ClearPagePrivate(page);

4307 4308 4309 4310 4311 4312
	/*
	 * set page dirty so that it will not be removed from cache/file
	 * by non-hugetlbfs specific code paths.
	 */
	set_page_dirty(page);

4313 4314 4315 4316 4317 4318
	spin_lock(&inode->i_lock);
	inode->i_blocks += blocks_per_huge_page(h);
	spin_unlock(&inode->i_lock);
	return 0;
}

4319 4320 4321 4322
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)
4323
{
4324
	struct hstate *h = hstate_vma(vma);
4325
	vm_fault_t ret = VM_FAULT_SIGBUS;
4326
	int anon_rmap = 0;
A
Adam Litke 已提交
4327 4328
	unsigned long size;
	struct page *page;
4329
	pte_t new_pte;
4330
	spinlock_t *ptl;
4331
	unsigned long haddr = address & huge_page_mask(h);
4332
	bool new_page = false;
A
Adam Litke 已提交
4333

4334 4335 4336
	/*
	 * 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
4337
	 * COW. Warn that such a situation has occurred as it may not be obvious
4338 4339
	 */
	if (is_vma_resv_set(vma, HPAGE_RESV_UNMAPPED)) {
4340
		pr_warn_ratelimited("PID %d killed due to inadequate hugepage pool\n",
4341
			   current->pid);
4342 4343 4344
		return ret;
	}

A
Adam Litke 已提交
4345
	/*
4346 4347 4348
	 * 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
4349
	 */
4350 4351 4352 4353
	size = i_size_read(mapping->host) >> huge_page_shift(h);
	if (idx >= size)
		goto out;

4354 4355 4356
retry:
	page = find_lock_page(mapping, idx);
	if (!page) {
4357 4358 4359 4360 4361 4362 4363
		/*
		 * Check for page in userfault range
		 */
		if (userfaultfd_missing(vma)) {
			u32 hash;
			struct vm_fault vmf = {
				.vma = vma,
4364
				.address = haddr,
4365 4366 4367 4368 4369 4370 4371 4372 4373 4374 4375
				.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
				 */
			};

			/*
4376 4377 4378
			 * hugetlb_fault_mutex and i_mmap_rwsem must be
			 * dropped before handling userfault.  Reacquire
			 * after handling fault to make calling code simpler.
4379
			 */
4380
			hash = hugetlb_fault_mutex_hash(mapping, idx);
4381
			mutex_unlock(&hugetlb_fault_mutex_table[hash]);
4382
			i_mmap_unlock_read(mapping);
4383
			ret = handle_userfault(&vmf, VM_UFFD_MISSING);
4384
			i_mmap_lock_read(mapping);
4385 4386 4387 4388
			mutex_lock(&hugetlb_fault_mutex_table[hash]);
			goto out;
		}

4389
		page = alloc_huge_page(vma, haddr, 0);
4390
		if (IS_ERR(page)) {
4391 4392 4393 4394 4395 4396 4397 4398 4399 4400 4401 4402 4403 4404 4405 4406 4407 4408 4409
			/*
			 * Returning error will result in faulting task being
			 * sent SIGBUS.  The hugetlb fault mutex prevents two
			 * tasks from racing to fault in the same page which
			 * could result in false unable to allocate errors.
			 * Page migration does not take the fault mutex, but
			 * does a clear then write of pte's under page table
			 * lock.  Page fault code could race with migration,
			 * notice the clear pte and try to allocate a page
			 * here.  Before returning error, get ptl and make
			 * sure there really is no pte entry.
			 */
			ptl = huge_pte_lock(h, mm, ptep);
			if (!huge_pte_none(huge_ptep_get(ptep))) {
				ret = 0;
				spin_unlock(ptl);
				goto out;
			}
			spin_unlock(ptl);
4410
			ret = vmf_error(PTR_ERR(page));
4411 4412
			goto out;
		}
A
Andrea Arcangeli 已提交
4413
		clear_huge_page(page, address, pages_per_huge_page(h));
N
Nick Piggin 已提交
4414
		__SetPageUptodate(page);
4415
		new_page = true;
4416

4417
		if (vma->vm_flags & VM_MAYSHARE) {
4418
			int err = huge_add_to_page_cache(page, mapping, idx);
4419 4420 4421 4422 4423 4424
			if (err) {
				put_page(page);
				if (err == -EEXIST)
					goto retry;
				goto out;
			}
4425
		} else {
4426
			lock_page(page);
4427 4428 4429 4430
			if (unlikely(anon_vma_prepare(vma))) {
				ret = VM_FAULT_OOM;
				goto backout_unlocked;
			}
4431
			anon_rmap = 1;
4432
		}
4433
	} else {
4434 4435 4436 4437 4438 4439
		/*
		 * 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))) {
4440
			ret = VM_FAULT_HWPOISON |
4441
				VM_FAULT_SET_HINDEX(hstate_index(h));
4442 4443
			goto backout_unlocked;
		}
4444
	}
4445

4446 4447 4448 4449 4450 4451
	/*
	 * 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.
	 */
4452
	if ((flags & FAULT_FLAG_WRITE) && !(vma->vm_flags & VM_SHARED)) {
4453
		if (vma_needs_reservation(h, vma, haddr) < 0) {
4454 4455 4456
			ret = VM_FAULT_OOM;
			goto backout_unlocked;
		}
4457
		/* Just decrements count, does not deallocate */
4458
		vma_end_reservation(h, vma, haddr);
4459
	}
4460

4461
	ptl = huge_pte_lock(h, mm, ptep);
N
Nick Piggin 已提交
4462
	ret = 0;
4463
	if (!huge_pte_none(huge_ptep_get(ptep)))
A
Adam Litke 已提交
4464 4465
		goto backout;

4466 4467
	if (anon_rmap) {
		ClearPagePrivate(page);
4468
		hugepage_add_new_anon_rmap(page, vma, haddr);
4469
	} else
4470
		page_dup_rmap(page, true);
4471 4472
	new_pte = make_huge_pte(vma, page, ((vma->vm_flags & VM_WRITE)
				&& (vma->vm_flags & VM_SHARED)));
4473
	set_huge_pte_at(mm, haddr, ptep, new_pte);
4474

4475
	hugetlb_count_add(pages_per_huge_page(h), mm);
4476
	if ((flags & FAULT_FLAG_WRITE) && !(vma->vm_flags & VM_SHARED)) {
4477
		/* Optimization, do the COW without a second fault */
4478
		ret = hugetlb_cow(mm, vma, address, ptep, page, ptl);
4479 4480
	}

4481
	spin_unlock(ptl);
4482 4483 4484 4485 4486 4487 4488 4489 4490

	/*
	 * Only make newly allocated pages active.  Existing pages found
	 * in the pagecache could be !page_huge_active() if they have been
	 * isolated for migration.
	 */
	if (new_page)
		set_page_huge_active(page);

A
Adam Litke 已提交
4491 4492
	unlock_page(page);
out:
4493
	return ret;
A
Adam Litke 已提交
4494 4495

backout:
4496
	spin_unlock(ptl);
4497
backout_unlocked:
A
Adam Litke 已提交
4498
	unlock_page(page);
4499
	restore_reserve_on_error(h, vma, haddr, page);
A
Adam Litke 已提交
4500 4501
	put_page(page);
	goto out;
4502 4503
}

4504
#ifdef CONFIG_SMP
4505
u32 hugetlb_fault_mutex_hash(struct address_space *mapping, pgoff_t idx)
4506 4507 4508 4509
{
	unsigned long key[2];
	u32 hash;

4510 4511
	key[0] = (unsigned long) mapping;
	key[1] = idx;
4512

4513
	hash = jhash2((u32 *)&key, sizeof(key)/(sizeof(u32)), 0);
4514 4515 4516 4517 4518 4519 4520 4521

	return hash & (num_fault_mutexes - 1);
}
#else
/*
 * For uniprocesor systems we always use a single mutex, so just
 * return 0 and avoid the hashing overhead.
 */
4522
u32 hugetlb_fault_mutex_hash(struct address_space *mapping, pgoff_t idx)
4523 4524 4525 4526 4527
{
	return 0;
}
#endif

4528
vm_fault_t hugetlb_fault(struct mm_struct *mm, struct vm_area_struct *vma,
4529
			unsigned long address, unsigned int flags)
4530
{
4531
	pte_t *ptep, entry;
4532
	spinlock_t *ptl;
4533
	vm_fault_t ret;
4534 4535
	u32 hash;
	pgoff_t idx;
4536
	struct page *page = NULL;
4537
	struct page *pagecache_page = NULL;
4538
	struct hstate *h = hstate_vma(vma);
4539
	struct address_space *mapping;
4540
	int need_wait_lock = 0;
4541
	unsigned long haddr = address & huge_page_mask(h);
4542

4543
	ptep = huge_pte_offset(mm, haddr, huge_page_size(h));
4544
	if (ptep) {
4545 4546 4547 4548 4549
		/*
		 * 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.
		 */
4550
		entry = huge_ptep_get(ptep);
N
Naoya Horiguchi 已提交
4551
		if (unlikely(is_hugetlb_entry_migration(entry))) {
4552
			migration_entry_wait_huge(vma, mm, ptep);
N
Naoya Horiguchi 已提交
4553 4554
			return 0;
		} else if (unlikely(is_hugetlb_entry_hwpoisoned(entry)))
4555
			return VM_FAULT_HWPOISON_LARGE |
4556
				VM_FAULT_SET_HINDEX(hstate_index(h));
4557 4558
	}

4559 4560
	/*
	 * Acquire i_mmap_rwsem before calling huge_pte_alloc and hold
4561 4562 4563 4564
	 * 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.
4565 4566 4567 4568 4569
	 *
	 * 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.
	 */
4570
	mapping = vma->vm_file->f_mapping;
4571 4572 4573 4574 4575 4576
	i_mmap_lock_read(mapping);
	ptep = huge_pte_alloc(mm, haddr, huge_page_size(h));
	if (!ptep) {
		i_mmap_unlock_read(mapping);
		return VM_FAULT_OOM;
	}
4577

4578 4579 4580 4581 4582
	/*
	 * 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.
	 */
4583
	idx = vma_hugecache_offset(h, vma, haddr);
4584
	hash = hugetlb_fault_mutex_hash(mapping, idx);
4585
	mutex_lock(&hugetlb_fault_mutex_table[hash]);
4586

4587 4588
	entry = huge_ptep_get(ptep);
	if (huge_pte_none(entry)) {
4589
		ret = hugetlb_no_page(mm, vma, mapping, idx, address, ptep, flags);
4590
		goto out_mutex;
4591
	}
4592

N
Nick Piggin 已提交
4593
	ret = 0;
4594

4595 4596 4597
	/*
	 * 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
4598 4599 4600
	 * an active hugepage in pagecache. This goto expects the 2nd page
	 * fault, and is_hugetlb_entry_(migration|hwpoisoned) check will
	 * properly handle it.
4601 4602 4603 4604
	 */
	if (!pte_present(entry))
		goto out_mutex;

4605 4606 4607 4608 4609 4610 4611 4612
	/*
	 * 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.
	 */
4613
	if ((flags & FAULT_FLAG_WRITE) && !huge_pte_write(entry)) {
4614
		if (vma_needs_reservation(h, vma, haddr) < 0) {
4615
			ret = VM_FAULT_OOM;
4616
			goto out_mutex;
4617
		}
4618
		/* Just decrements count, does not deallocate */
4619
		vma_end_reservation(h, vma, haddr);
4620

4621
		if (!(vma->vm_flags & VM_MAYSHARE))
4622
			pagecache_page = hugetlbfs_pagecache_page(h,
4623
								vma, haddr);
4624 4625
	}

4626 4627 4628 4629 4630 4631
	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;

4632 4633 4634 4635 4636 4637 4638
	/*
	 * 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)
4639 4640 4641 4642
		if (!trylock_page(page)) {
			need_wait_lock = 1;
			goto out_ptl;
		}
4643

4644
	get_page(page);
4645

4646
	if (flags & FAULT_FLAG_WRITE) {
4647
		if (!huge_pte_write(entry)) {
4648
			ret = hugetlb_cow(mm, vma, address, ptep,
4649
					  pagecache_page, ptl);
4650
			goto out_put_page;
4651
		}
4652
		entry = huge_pte_mkdirty(entry);
4653 4654
	}
	entry = pte_mkyoung(entry);
4655
	if (huge_ptep_set_access_flags(vma, haddr, ptep, entry,
4656
						flags & FAULT_FLAG_WRITE))
4657
		update_mmu_cache(vma, haddr, ptep);
4658 4659 4660 4661
out_put_page:
	if (page != pagecache_page)
		unlock_page(page);
	put_page(page);
4662 4663
out_ptl:
	spin_unlock(ptl);
4664 4665 4666 4667 4668

	if (pagecache_page) {
		unlock_page(pagecache_page);
		put_page(pagecache_page);
	}
4669
out_mutex:
4670
	mutex_unlock(&hugetlb_fault_mutex_table[hash]);
4671
	i_mmap_unlock_read(mapping);
4672 4673 4674 4675 4676 4677 4678 4679 4680
	/*
	 * 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);
4681
	return ret;
4682 4683
}

4684 4685 4686 4687 4688 4689 4690 4691 4692 4693 4694
/*
 * Used by userfaultfd UFFDIO_COPY.  Based on mcopy_atomic_pte with
 * modifications for huge pages.
 */
int hugetlb_mcopy_atomic_pte(struct mm_struct *dst_mm,
			    pte_t *dst_pte,
			    struct vm_area_struct *dst_vma,
			    unsigned long dst_addr,
			    unsigned long src_addr,
			    struct page **pagep)
{
4695 4696 4697
	struct address_space *mapping;
	pgoff_t idx;
	unsigned long size;
4698
	int vm_shared = dst_vma->vm_flags & VM_SHARED;
4699 4700 4701 4702 4703 4704 4705 4706 4707 4708 4709 4710 4711 4712
	struct hstate *h = hstate_vma(dst_vma);
	pte_t _dst_pte;
	spinlock_t *ptl;
	int ret;
	struct page *page;

	if (!*pagep) {
		ret = -ENOMEM;
		page = alloc_huge_page(dst_vma, dst_addr, 0);
		if (IS_ERR(page))
			goto out;

		ret = copy_huge_page_from_user(page,
						(const void __user *) src_addr,
4713
						pages_per_huge_page(h), false);
4714

4715
		/* fallback to copy_from_user outside mmap_lock */
4716
		if (unlikely(ret)) {
4717
			ret = -ENOENT;
4718 4719 4720 4721 4722 4723 4724 4725 4726 4727 4728 4729 4730 4731 4732 4733
			*pagep = page;
			/* don't free the page */
			goto out;
		}
	} else {
		page = *pagep;
		*pagep = NULL;
	}

	/*
	 * The memory barrier inside __SetPageUptodate makes sure that
	 * preceding stores to the page contents become visible before
	 * the set_pte_at() write.
	 */
	__SetPageUptodate(page);

4734 4735 4736
	mapping = dst_vma->vm_file->f_mapping;
	idx = vma_hugecache_offset(h, dst_vma, dst_addr);

4737 4738 4739 4740
	/*
	 * If shared, add to page cache
	 */
	if (vm_shared) {
4741 4742 4743 4744
		size = i_size_read(mapping->host) >> huge_page_shift(h);
		ret = -EFAULT;
		if (idx >= size)
			goto out_release_nounlock;
4745

4746 4747 4748 4749 4750 4751
		/*
		 * 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.
		 */
4752 4753 4754 4755 4756
		ret = huge_add_to_page_cache(page, mapping, idx);
		if (ret)
			goto out_release_nounlock;
	}

4757 4758 4759
	ptl = huge_pte_lockptr(h, dst_mm, dst_pte);
	spin_lock(ptl);

4760 4761 4762 4763 4764 4765 4766 4767 4768 4769 4770 4771 4772 4773
	/*
	 * 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;

4774 4775 4776 4777
	ret = -EEXIST;
	if (!huge_pte_none(huge_ptep_get(dst_pte)))
		goto out_release_unlock;

4778 4779 4780 4781 4782 4783
	if (vm_shared) {
		page_dup_rmap(page, true);
	} else {
		ClearPagePrivate(page);
		hugepage_add_new_anon_rmap(page, dst_vma, dst_addr);
	}
4784 4785 4786 4787 4788 4789 4790 4791 4792 4793 4794 4795 4796 4797 4798 4799

	_dst_pte = make_huge_pte(dst_vma, page, dst_vma->vm_flags & VM_WRITE);
	if (dst_vma->vm_flags & VM_WRITE)
		_dst_pte = huge_pte_mkdirty(_dst_pte);
	_dst_pte = pte_mkyoung(_dst_pte);

	set_huge_pte_at(dst_mm, dst_addr, dst_pte, _dst_pte);

	(void)huge_ptep_set_access_flags(dst_vma, dst_addr, dst_pte, _dst_pte,
					dst_vma->vm_flags & VM_WRITE);
	hugetlb_count_add(pages_per_huge_page(h), dst_mm);

	/* No need to invalidate - it was non-present before */
	update_mmu_cache(dst_vma, dst_addr, dst_pte);

	spin_unlock(ptl);
4800
	set_page_huge_active(page);
4801 4802
	if (vm_shared)
		unlock_page(page);
4803 4804 4805 4806 4807
	ret = 0;
out:
	return ret;
out_release_unlock:
	spin_unlock(ptl);
4808 4809
	if (vm_shared)
		unlock_page(page);
4810
out_release_nounlock:
4811 4812 4813 4814
	put_page(page);
	goto out;
}

4815 4816 4817
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,
4818
			 long i, unsigned int flags, int *locked)
D
David Gibson 已提交
4819
{
4820 4821
	unsigned long pfn_offset;
	unsigned long vaddr = *position;
4822
	unsigned long remainder = *nr_pages;
4823
	struct hstate *h = hstate_vma(vma);
4824
	int err = -EFAULT;
D
David Gibson 已提交
4825 4826

	while (vaddr < vma->vm_end && remainder) {
A
Adam Litke 已提交
4827
		pte_t *pte;
4828
		spinlock_t *ptl = NULL;
H
Hugh Dickins 已提交
4829
		int absent;
A
Adam Litke 已提交
4830
		struct page *page;
D
David Gibson 已提交
4831

4832 4833 4834 4835
		/*
		 * If we have a pending SIGKILL, don't keep faulting pages and
		 * potentially allocating memory.
		 */
4836
		if (fatal_signal_pending(current)) {
4837 4838 4839 4840
			remainder = 0;
			break;
		}

A
Adam Litke 已提交
4841 4842
		/*
		 * Some archs (sparc64, sh*) have multiple pte_ts to
H
Hugh Dickins 已提交
4843
		 * each hugepage.  We have to make sure we get the
A
Adam Litke 已提交
4844
		 * first, for the page indexing below to work.
4845 4846
		 *
		 * Note that page table lock is not held when pte is null.
A
Adam Litke 已提交
4847
		 */
4848 4849
		pte = huge_pte_offset(mm, vaddr & huge_page_mask(h),
				      huge_page_size(h));
4850 4851
		if (pte)
			ptl = huge_pte_lock(h, mm, pte);
H
Hugh Dickins 已提交
4852 4853 4854 4855
		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
4856 4857 4858 4859
		 * 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
4860
		 */
H
Hugh Dickins 已提交
4861 4862
		if (absent && (flags & FOLL_DUMP) &&
		    !hugetlbfs_pagecache_present(h, vma, vaddr)) {
4863 4864
			if (pte)
				spin_unlock(ptl);
H
Hugh Dickins 已提交
4865 4866 4867
			remainder = 0;
			break;
		}
D
David Gibson 已提交
4868

4869 4870 4871 4872 4873 4874 4875 4876 4877 4878 4879
		/*
		 * 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)) ||
4880 4881
		    ((flags & FOLL_WRITE) &&
		      !huge_pte_write(huge_ptep_get(pte)))) {
4882
			vm_fault_t ret;
4883
			unsigned int fault_flags = 0;
D
David Gibson 已提交
4884

4885 4886
			if (pte)
				spin_unlock(ptl);
4887 4888
			if (flags & FOLL_WRITE)
				fault_flags |= FAULT_FLAG_WRITE;
4889
			if (locked)
4890 4891
				fault_flags |= FAULT_FLAG_ALLOW_RETRY |
					FAULT_FLAG_KILLABLE;
4892 4893 4894 4895
			if (flags & FOLL_NOWAIT)
				fault_flags |= FAULT_FLAG_ALLOW_RETRY |
					FAULT_FLAG_RETRY_NOWAIT;
			if (flags & FOLL_TRIED) {
4896 4897 4898 4899
				/*
				 * Note: FAULT_FLAG_ALLOW_RETRY and
				 * FAULT_FLAG_TRIED can co-exist
				 */
4900 4901 4902 4903
				fault_flags |= FAULT_FLAG_TRIED;
			}
			ret = hugetlb_fault(mm, vma, vaddr, fault_flags);
			if (ret & VM_FAULT_ERROR) {
4904
				err = vm_fault_to_errno(ret, flags);
4905 4906 4907 4908
				remainder = 0;
				break;
			}
			if (ret & VM_FAULT_RETRY) {
4909
				if (locked &&
4910
				    !(fault_flags & FAULT_FLAG_RETRY_NOWAIT))
4911
					*locked = 0;
4912 4913 4914 4915 4916 4917 4918 4919 4920 4921 4922 4923 4924
				*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
4925 4926
		}

4927
		pfn_offset = (vaddr & ~huge_page_mask(h)) >> PAGE_SHIFT;
4928
		page = pte_page(huge_ptep_get(pte));
4929

4930 4931 4932 4933 4934 4935 4936 4937 4938 4939 4940 4941 4942 4943
		/*
		 * 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;
		}

4944
same_page:
4945
		if (pages) {
H
Hugh Dickins 已提交
4946
			pages[i] = mem_map_offset(page, pfn_offset);
J
John Hubbard 已提交
4947 4948 4949 4950 4951 4952 4953 4954 4955 4956 4957 4958 4959 4960 4961 4962
			/*
			 * try_grab_page() should always succeed here, because:
			 * a) we hold the ptl lock, and b) we've just checked
			 * that the huge page is present in the page tables. If
			 * the huge page is present, then the tail pages must
			 * also be present. The ptl prevents the head page and
			 * tail pages from being rearranged in any way. So this
			 * page must be available at this point, unless the page
			 * refcount overflowed:
			 */
			if (WARN_ON_ONCE(!try_grab_page(pages[i], flags))) {
				spin_unlock(ptl);
				remainder = 0;
				err = -ENOMEM;
				break;
			}
4963
		}
D
David Gibson 已提交
4964 4965 4966 4967 4968

		if (vmas)
			vmas[i] = vma;

		vaddr += PAGE_SIZE;
4969
		++pfn_offset;
D
David Gibson 已提交
4970 4971
		--remainder;
		++i;
4972
		if (vaddr < vma->vm_end && remainder &&
4973
				pfn_offset < pages_per_huge_page(h)) {
4974 4975 4976 4977 4978 4979
			/*
			 * We use pfn_offset to avoid touching the pageframes
			 * of this compound page.
			 */
			goto same_page;
		}
4980
		spin_unlock(ptl);
D
David Gibson 已提交
4981
	}
4982
	*nr_pages = remainder;
4983 4984 4985 4986 4987
	/*
	 * 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
4988 4989
	*position = vaddr;

4990
	return i ? i : err;
D
David Gibson 已提交
4991
}
4992

4993 4994 4995 4996 4997 4998 4999 5000
#ifndef __HAVE_ARCH_FLUSH_HUGETLB_TLB_RANGE
/*
 * ARCHes with special requirements for evicting HUGETLB backing TLB entries can
 * implement this.
 */
#define flush_hugetlb_tlb_range(vma, addr, end)	flush_tlb_range(vma, addr, end)
#endif

5001
unsigned long hugetlb_change_protection(struct vm_area_struct *vma,
5002 5003 5004 5005 5006 5007
		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;
5008
	struct hstate *h = hstate_vma(vma);
5009
	unsigned long pages = 0;
5010
	bool shared_pmd = false;
5011
	struct mmu_notifier_range range;
5012 5013 5014

	/*
	 * In the case of shared PMDs, the area to flush could be beyond
5015
	 * start/end.  Set range.start/range.end to cover the maximum possible
5016 5017
	 * range if PMD sharing is possible.
	 */
5018 5019
	mmu_notifier_range_init(&range, MMU_NOTIFY_PROTECTION_VMA,
				0, vma, mm, start, end);
5020
	adjust_range_if_pmd_sharing_possible(vma, &range.start, &range.end);
5021 5022

	BUG_ON(address >= end);
5023
	flush_cache_range(vma, range.start, range.end);
5024

5025
	mmu_notifier_invalidate_range_start(&range);
5026
	i_mmap_lock_write(vma->vm_file->f_mapping);
5027
	for (; address < end; address += huge_page_size(h)) {
5028
		spinlock_t *ptl;
5029
		ptep = huge_pte_offset(mm, address, huge_page_size(h));
5030 5031
		if (!ptep)
			continue;
5032
		ptl = huge_pte_lock(h, mm, ptep);
5033
		if (huge_pmd_unshare(mm, vma, &address, ptep)) {
5034
			pages++;
5035
			spin_unlock(ptl);
5036
			shared_pmd = true;
5037
			continue;
5038
		}
5039 5040 5041 5042 5043 5044 5045 5046 5047 5048 5049 5050 5051
		pte = huge_ptep_get(ptep);
		if (unlikely(is_hugetlb_entry_hwpoisoned(pte))) {
			spin_unlock(ptl);
			continue;
		}
		if (unlikely(is_hugetlb_entry_migration(pte))) {
			swp_entry_t entry = pte_to_swp_entry(pte);

			if (is_write_migration_entry(entry)) {
				pte_t newpte;

				make_migration_entry_read(&entry);
				newpte = swp_entry_to_pte(entry);
5052 5053
				set_huge_swap_pte_at(mm, address, ptep,
						     newpte, huge_page_size(h));
5054 5055 5056 5057 5058 5059
				pages++;
			}
			spin_unlock(ptl);
			continue;
		}
		if (!huge_pte_none(pte)) {
5060 5061 5062 5063
			pte_t old_pte;

			old_pte = huge_ptep_modify_prot_start(vma, address, ptep);
			pte = pte_mkhuge(huge_pte_modify(old_pte, newprot));
5064
			pte = arch_make_huge_pte(pte, vma, NULL, 0);
5065
			huge_ptep_modify_prot_commit(vma, address, ptep, old_pte, pte);
5066
			pages++;
5067
		}
5068
		spin_unlock(ptl);
5069
	}
5070
	/*
5071
	 * Must flush TLB before releasing i_mmap_rwsem: x86's huge_pmd_unshare
5072
	 * may have cleared our pud entry and done put_page on the page table:
5073
	 * once we release i_mmap_rwsem, another task can do the final put_page
5074 5075
	 * 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.
5076
	 */
5077
	if (shared_pmd)
5078
		flush_hugetlb_tlb_range(vma, range.start, range.end);
5079 5080
	else
		flush_hugetlb_tlb_range(vma, start, end);
5081 5082 5083 5084
	/*
	 * No need to call mmu_notifier_invalidate_range() we are downgrading
	 * page table protection not changing it to point to a new page.
	 *
5085
	 * See Documentation/vm/mmu_notifier.rst
5086
	 */
5087
	i_mmap_unlock_write(vma->vm_file->f_mapping);
5088
	mmu_notifier_invalidate_range_end(&range);
5089 5090

	return pages << h->order;
5091 5092
}

5093 5094
int hugetlb_reserve_pages(struct inode *inode,
					long from, long to,
5095
					struct vm_area_struct *vma,
5096
					vm_flags_t vm_flags)
5097
{
5098
	long ret, chg, add = -1;
5099
	struct hstate *h = hstate_inode(inode);
5100
	struct hugepage_subpool *spool = subpool_inode(inode);
5101
	struct resv_map *resv_map;
5102
	struct hugetlb_cgroup *h_cg = NULL;
5103
	long gbl_reserve, regions_needed = 0;
5104

5105 5106 5107 5108 5109 5110
	/* This should never happen */
	if (from > to) {
		VM_WARN(1, "%s called with a negative range\n", __func__);
		return -EINVAL;
	}

5111 5112 5113
	/*
	 * Only apply hugepage reservation if asked. At fault time, an
	 * attempt will be made for VM_NORESERVE to allocate a page
5114
	 * without using reserves
5115
	 */
5116
	if (vm_flags & VM_NORESERVE)
5117 5118
		return 0;

5119 5120 5121 5122 5123 5124
	/*
	 * 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
	 */
5125
	if (!vma || vma->vm_flags & VM_MAYSHARE) {
5126 5127 5128 5129 5130
		/*
		 * 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).
		 */
5131
		resv_map = inode_resv_map(inode);
5132

5133
		chg = region_chg(resv_map, from, to, &regions_needed);
5134 5135

	} else {
5136
		/* Private mapping. */
5137
		resv_map = resv_map_alloc();
5138 5139 5140
		if (!resv_map)
			return -ENOMEM;

5141
		chg = to - from;
5142

5143 5144 5145 5146
		set_vma_resv_map(vma, resv_map);
		set_vma_resv_flags(vma, HPAGE_RESV_OWNER);
	}

5147 5148 5149 5150
	if (chg < 0) {
		ret = chg;
		goto out_err;
	}
5151

5152 5153 5154 5155 5156 5157 5158 5159 5160 5161 5162 5163 5164 5165 5166
	ret = hugetlb_cgroup_charge_cgroup_rsvd(
		hstate_index(h), chg * pages_per_huge_page(h), &h_cg);

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

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

5167 5168 5169 5170 5171 5172 5173
	/*
	 * There must be enough pages in the subpool for the mapping. If
	 * the subpool has a minimum size, there may be some global
	 * reservations already in place (gbl_reserve).
	 */
	gbl_reserve = hugepage_subpool_get_pages(spool, chg);
	if (gbl_reserve < 0) {
5174
		ret = -ENOSPC;
5175
		goto out_uncharge_cgroup;
5176
	}
5177 5178

	/*
5179
	 * Check enough hugepages are available for the reservation.
5180
	 * Hand the pages back to the subpool if there are not
5181
	 */
5182
	ret = hugetlb_acct_memory(h, gbl_reserve);
K
Ken Chen 已提交
5183
	if (ret < 0) {
5184
		goto out_put_pages;
K
Ken Chen 已提交
5185
	}
5186 5187 5188 5189 5190 5191 5192 5193 5194 5195 5196 5197

	/*
	 * 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
	 */
5198
	if (!vma || vma->vm_flags & VM_MAYSHARE) {
5199
		add = region_add(resv_map, from, to, regions_needed, h, h_cg);
5200 5201 5202

		if (unlikely(add < 0)) {
			hugetlb_acct_memory(h, -gbl_reserve);
5203
			goto out_put_pages;
5204
		} else if (unlikely(chg > add)) {
5205 5206 5207 5208 5209 5210 5211 5212 5213
			/*
			 * 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;

5214 5215 5216 5217
			hugetlb_cgroup_uncharge_cgroup_rsvd(
				hstate_index(h),
				(chg - add) * pages_per_huge_page(h), h_cg);

5218 5219 5220 5221 5222
			rsv_adjust = hugepage_subpool_put_pages(spool,
								chg - add);
			hugetlb_acct_memory(h, -rsv_adjust);
		}
	}
5223
	return 0;
5224 5225 5226 5227 5228 5229
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);
5230
out_err:
5231
	if (!vma || vma->vm_flags & VM_MAYSHARE)
5232 5233 5234 5235 5236
		/* 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
5237 5238
	if (vma && is_vma_resv_set(vma, HPAGE_RESV_OWNER))
		kref_put(&resv_map->refs, resv_map_release);
5239
	return ret;
5240 5241
}

5242 5243
long hugetlb_unreserve_pages(struct inode *inode, long start, long end,
								long freed)
5244
{
5245
	struct hstate *h = hstate_inode(inode);
5246
	struct resv_map *resv_map = inode_resv_map(inode);
5247
	long chg = 0;
5248
	struct hugepage_subpool *spool = subpool_inode(inode);
5249
	long gbl_reserve;
K
Ken Chen 已提交
5250

5251 5252 5253 5254
	/*
	 * Since this routine can be called in the evict inode path for all
	 * hugetlbfs inodes, resv_map could be NULL.
	 */
5255 5256 5257 5258 5259 5260 5261 5262 5263 5264 5265
	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
5266
	spin_lock(&inode->i_lock);
5267
	inode->i_blocks -= (blocks_per_huge_page(h) * freed);
K
Ken Chen 已提交
5268 5269
	spin_unlock(&inode->i_lock);

5270 5271 5272 5273 5274 5275
	/*
	 * If the subpool has a minimum size, the number of global
	 * reservations to be released may be adjusted.
	 */
	gbl_reserve = hugepage_subpool_put_pages(spool, (chg - freed));
	hugetlb_acct_memory(h, -gbl_reserve);
5276 5277

	return 0;
5278
}
5279

5280 5281 5282 5283 5284 5285 5286 5287 5288 5289 5290
#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
5291 5292
	unsigned long vm_flags = vma->vm_flags & VM_LOCKED_CLEAR_MASK;
	unsigned long svm_flags = svma->vm_flags & VM_LOCKED_CLEAR_MASK;
5293 5294 5295 5296 5297 5298 5299 5300 5301 5302 5303 5304 5305

	/*
	 * match the virtual addresses, permission and the alignment of the
	 * page table page.
	 */
	if (pmd_index(addr) != pmd_index(saddr) ||
	    vm_flags != svm_flags ||
	    sbase < svma->vm_start || svma->vm_end < s_end)
		return 0;

	return saddr;
}

5306
static bool vma_shareable(struct vm_area_struct *vma, unsigned long addr)
5307 5308 5309 5310 5311 5312 5313
{
	unsigned long base = addr & PUD_MASK;
	unsigned long end = base + PUD_SIZE;

	/*
	 * check on proper vm_flags and page table alignment
	 */
5314
	if (vma->vm_flags & VM_MAYSHARE && range_in_vma(vma, base, end))
5315 5316
		return true;
	return false;
5317 5318
}

5319 5320 5321 5322 5323 5324 5325 5326
/*
 * 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)
{
5327
	unsigned long a_start, a_end;
5328 5329 5330 5331

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

5332 5333 5334
	/* Extend the range to be PUD aligned for a worst case scenario */
	a_start = ALIGN_DOWN(*start, PUD_SIZE);
	a_end = ALIGN(*end, PUD_SIZE);
5335

5336 5337 5338 5339 5340 5341
	/*
	 * Intersect the range with the vma range, since pmd sharing won't be
	 * across vma after all
	 */
	*start = max(vma->vm_start, a_start);
	*end = min(vma->vm_end, a_end);
5342 5343
}

5344 5345 5346 5347
/*
 * 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
5348 5349 5350 5351 5352 5353
 * code much cleaner.
 *
 * This routine must be called with i_mmap_rwsem held in at least read mode.
 * 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).
5354 5355 5356 5357 5358 5359 5360 5361 5362 5363 5364
 */
pte_t *huge_pmd_share(struct mm_struct *mm, unsigned long addr, pud_t *pud)
{
	struct vm_area_struct *vma = find_vma(mm, addr);
	struct address_space *mapping = vma->vm_file->f_mapping;
	pgoff_t idx = ((addr - vma->vm_start) >> PAGE_SHIFT) +
			vma->vm_pgoff;
	struct vm_area_struct *svma;
	unsigned long saddr;
	pte_t *spte = NULL;
	pte_t *pte;
5365
	spinlock_t *ptl;
5366 5367 5368 5369 5370 5371 5372 5373 5374 5375

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

	vma_interval_tree_foreach(svma, &mapping->i_mmap, idx, idx) {
		if (svma == vma)
			continue;

		saddr = page_table_shareable(svma, vma, addr, idx);
		if (saddr) {
5376 5377
			spte = huge_pte_offset(svma->vm_mm, saddr,
					       vma_mmu_pagesize(svma));
5378 5379 5380 5381 5382 5383 5384 5385 5386 5387
			if (spte) {
				get_page(virt_to_page(spte));
				break;
			}
		}
	}

	if (!spte)
		goto out;

5388
	ptl = huge_pte_lock(hstate_vma(vma), mm, spte);
5389
	if (pud_none(*pud)) {
5390 5391
		pud_populate(mm, pud,
				(pmd_t *)((unsigned long)spte & PAGE_MASK));
5392
		mm_inc_nr_pmds(mm);
5393
	} else {
5394
		put_page(virt_to_page(spte));
5395
	}
5396
	spin_unlock(ptl);
5397 5398 5399 5400 5401 5402 5403 5404 5405 5406 5407 5408
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.
 *
5409
 * Called with page table lock held and i_mmap_rwsem held in write mode.
5410 5411 5412 5413
 *
 * returns: 1 successfully unmapped a shared pte page
 *	    0 the underlying pte page is not shared, or it is the last user
 */
5414 5415
int huge_pmd_unshare(struct mm_struct *mm, struct vm_area_struct *vma,
					unsigned long *addr, pte_t *ptep)
5416 5417
{
	pgd_t *pgd = pgd_offset(mm, *addr);
5418 5419
	p4d_t *p4d = p4d_offset(pgd, *addr);
	pud_t *pud = pud_offset(p4d, *addr);
5420

5421
	i_mmap_assert_write_locked(vma->vm_file->f_mapping);
5422 5423 5424 5425 5426 5427
	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));
5428
	mm_dec_nr_pmds(mm);
5429 5430 5431
	*addr = ALIGN(*addr, HPAGE_SIZE * PTRS_PER_PTE) - HPAGE_SIZE;
	return 1;
}
5432 5433 5434 5435 5436 5437
#define want_pmd_share()	(1)
#else /* !CONFIG_ARCH_WANT_HUGE_PMD_SHARE */
pte_t *huge_pmd_share(struct mm_struct *mm, unsigned long addr, pud_t *pud)
{
	return NULL;
}
5438

5439 5440
int huge_pmd_unshare(struct mm_struct *mm, struct vm_area_struct *vma,
				unsigned long *addr, pte_t *ptep)
5441 5442 5443
{
	return 0;
}
5444 5445 5446 5447 5448

void adjust_range_if_pmd_sharing_possible(struct vm_area_struct *vma,
				unsigned long *start, unsigned long *end)
{
}
5449
#define want_pmd_share()	(0)
5450 5451
#endif /* CONFIG_ARCH_WANT_HUGE_PMD_SHARE */

5452 5453 5454 5455 5456
#ifdef CONFIG_ARCH_WANT_GENERAL_HUGETLB
pte_t *huge_pte_alloc(struct mm_struct *mm,
			unsigned long addr, unsigned long sz)
{
	pgd_t *pgd;
5457
	p4d_t *p4d;
5458 5459 5460 5461
	pud_t *pud;
	pte_t *pte = NULL;

	pgd = pgd_offset(mm, addr);
5462 5463 5464
	p4d = p4d_alloc(mm, pgd, addr);
	if (!p4d)
		return NULL;
5465
	pud = pud_alloc(mm, p4d, addr);
5466 5467 5468 5469 5470 5471 5472 5473 5474 5475 5476
	if (pud) {
		if (sz == PUD_SIZE) {
			pte = (pte_t *)pud;
		} else {
			BUG_ON(sz != PMD_SIZE);
			if (want_pmd_share() && pud_none(*pud))
				pte = huge_pmd_share(mm, addr, pud);
			else
				pte = (pte_t *)pmd_alloc(mm, pud, addr);
		}
	}
5477
	BUG_ON(pte && pte_present(*pte) && !pte_huge(*pte));
5478 5479 5480 5481

	return pte;
}

5482 5483 5484 5485
/*
 * huge_pte_offset() - Walk the page table to resolve the hugepage
 * entry at address @addr
 *
5486 5487
 * Return: Pointer to page table entry (PUD or PMD) for
 * address @addr, or NULL if a !p*d_present() entry is encountered and the
5488 5489 5490
 * size @sz doesn't match the hugepage size at this level of the page
 * table.
 */
5491 5492
pte_t *huge_pte_offset(struct mm_struct *mm,
		       unsigned long addr, unsigned long sz)
5493 5494
{
	pgd_t *pgd;
5495
	p4d_t *p4d;
5496 5497
	pud_t *pud;
	pmd_t *pmd;
5498 5499

	pgd = pgd_offset(mm, addr);
5500 5501 5502 5503 5504
	if (!pgd_present(*pgd))
		return NULL;
	p4d = p4d_offset(pgd, addr);
	if (!p4d_present(*p4d))
		return NULL;
5505

5506
	pud = pud_offset(p4d, addr);
5507 5508
	if (sz == PUD_SIZE)
		/* must be pud huge, non-present or none */
5509
		return (pte_t *)pud;
5510
	if (!pud_present(*pud))
5511
		return NULL;
5512
	/* must have a valid entry and size to go further */
5513

5514 5515 5516
	pmd = pmd_offset(pud, addr);
	/* must be pmd huge, non-present or none */
	return (pte_t *)pmd;
5517 5518
}

5519 5520 5521 5522 5523 5524 5525 5526 5527 5528 5529 5530 5531
#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);
}

5532 5533 5534 5535 5536 5537 5538 5539
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;
}

5540
struct page * __weak
5541
follow_huge_pmd(struct mm_struct *mm, unsigned long address,
5542
		pmd_t *pmd, int flags)
5543
{
5544 5545
	struct page *page = NULL;
	spinlock_t *ptl;
5546
	pte_t pte;
J
John Hubbard 已提交
5547 5548 5549 5550 5551 5552

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

5553 5554 5555 5556 5557 5558 5559 5560 5561
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;
5562 5563
	pte = huge_ptep_get((pte_t *)pmd);
	if (pte_present(pte)) {
5564
		page = pmd_page(*pmd) + ((address & ~PMD_MASK) >> PAGE_SHIFT);
J
John Hubbard 已提交
5565 5566 5567 5568 5569 5570 5571 5572 5573 5574 5575 5576
		/*
		 * 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;
		}
5577
	} else {
5578
		if (is_hugetlb_entry_migration(pte)) {
5579 5580 5581 5582 5583 5584 5585 5586 5587 5588 5589
			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);
5590 5591 5592
	return page;
}

5593
struct page * __weak
5594
follow_huge_pud(struct mm_struct *mm, unsigned long address,
5595
		pud_t *pud, int flags)
5596
{
J
John Hubbard 已提交
5597
	if (flags & (FOLL_GET | FOLL_PIN))
5598
		return NULL;
5599

5600
	return pte_page(*(pte_t *)pud) + ((address & ~PUD_MASK) >> PAGE_SHIFT);
5601 5602
}

5603 5604 5605
struct page * __weak
follow_huge_pgd(struct mm_struct *mm, unsigned long address, pgd_t *pgd, int flags)
{
J
John Hubbard 已提交
5606
	if (flags & (FOLL_GET | FOLL_PIN))
5607 5608 5609 5610 5611
		return NULL;

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

5612 5613
bool isolate_huge_page(struct page *page, struct list_head *list)
{
5614 5615
	bool ret = true;

5616
	VM_BUG_ON_PAGE(!PageHead(page), page);
5617
	spin_lock(&hugetlb_lock);
5618 5619 5620 5621 5622
	if (!page_huge_active(page) || !get_page_unless_zero(page)) {
		ret = false;
		goto unlock;
	}
	clear_page_huge_active(page);
5623
	list_move_tail(&page->lru, list);
5624
unlock:
5625
	spin_unlock(&hugetlb_lock);
5626
	return ret;
5627 5628 5629 5630
}

void putback_active_hugepage(struct page *page)
{
5631
	VM_BUG_ON_PAGE(!PageHead(page), page);
5632
	spin_lock(&hugetlb_lock);
5633
	set_page_huge_active(page);
5634 5635 5636 5637
	list_move_tail(&page->lru, &(page_hstate(page))->hugepage_activelist);
	spin_unlock(&hugetlb_lock);
	put_page(page);
}
5638 5639 5640 5641 5642 5643 5644 5645 5646 5647 5648 5649 5650 5651 5652 5653 5654 5655 5656 5657 5658 5659 5660 5661 5662 5663 5664 5665 5666 5667 5668 5669 5670

void move_hugetlb_state(struct page *oldpage, struct page *newpage, int reason)
{
	struct hstate *h = page_hstate(oldpage);

	hugetlb_cgroup_migrate(oldpage, newpage);
	set_page_owner_migrate_reason(newpage, reason);

	/*
	 * transfer temporary state of the new huge page. This is
	 * reverse to other transitions because the newpage is going to
	 * be final while the old one will be freed so it takes over
	 * the temporary status.
	 *
	 * Also note that we have to transfer the per-node surplus state
	 * here as well otherwise the global surplus count will not match
	 * the per-node's.
	 */
	if (PageHugeTemporary(newpage)) {
		int old_nid = page_to_nid(oldpage);
		int new_nid = page_to_nid(newpage);

		SetPageHugeTemporary(oldpage);
		ClearPageHugeTemporary(newpage);

		spin_lock(&hugetlb_lock);
		if (h->surplus_huge_pages_node[old_nid]) {
			h->surplus_huge_pages_node[old_nid]--;
			h->surplus_huge_pages_node[new_nid]++;
		}
		spin_unlock(&hugetlb_lock);
	}
}
5671 5672 5673 5674 5675 5676 5677 5678 5679 5680 5681 5682 5683 5684 5685 5686 5687 5688 5689 5690 5691 5692 5693 5694 5695 5696 5697 5698 5699 5700 5701 5702 5703 5704 5705 5706 5707 5708 5709

#ifdef CONFIG_CMA
static bool cma_reserve_called __initdata;

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

early_param("hugetlb_cma", cmdline_parse_hugetlb_cma);

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

	cma_reserve_called = true;

	if (!hugetlb_cma_size)
		return;

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

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

	reserved = 0;
	for_each_node_state(nid, N_ONLINE) {
		int res;
5710
		char name[20];
5711 5712 5713 5714

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

5715
		snprintf(name, 20, "hugetlb%d", nid);
5716
		res = cma_declare_contiguous_nid(0, size, 0, PAGE_SIZE << order,
5717
						 0, false, name,
5718 5719 5720 5721 5722 5723 5724 5725 5726 5727 5728 5729 5730 5731 5732 5733 5734 5735 5736 5737 5738 5739 5740 5741 5742
						 &hugetlb_cma[nid], nid);
		if (res) {
			pr_warn("hugetlb_cma: reservation failed: err %d, node %d",
				res, nid);
			continue;
		}

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

		if (reserved >= hugetlb_cma_size)
			break;
	}
}

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

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

#endif /* CONFIG_CMA */