hugetlb.c 160.7 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/sched/mm.h>
23
#include <linux/mmdebug.h>
24
#include <linux/sched/signal.h>
25
#include <linux/rmap.h>
26
#include <linux/string_helpers.h>
27 28
#include <linux/swap.h>
#include <linux/swapops.h>
29
#include <linux/jhash.h>
30
#include <linux/numa.h>
31
#include <linux/llist.h>
32
#include <linux/cma.h>
33

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

38
#include <linux/io.h>
D
David Gibson 已提交
39
#include <linux/hugetlb.h>
40
#include <linux/hugetlb_cgroup.h>
41
#include <linux/node.h>
42
#include <linux/userfaultfd_k.h>
43
#include <linux/page_owner.h>
44
#include "internal.h"
45
#include "hugetlb_vmemmap.h"
L
Linus Torvalds 已提交
46

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

51
#ifdef CONFIG_CMA
52
static struct cma *hugetlb_cma[MAX_NUMNODES];
53 54
#endif
static unsigned long hugetlb_cma_size __initdata;
55

56 57 58 59 60
/*
 * Minimum page order among possible hugepage sizes, set to a proper value
 * at boot time.
 */
static unsigned int minimum_order __read_mostly = UINT_MAX;
61

62 63
__initdata LIST_HEAD(huge_boot_pages);

64 65 66
/* for command line parsing */
static struct hstate * __initdata parsed_hstate;
static unsigned long __initdata default_hstate_max_huge_pages;
67
static bool __initdata parsed_valid_hugepagesz = true;
68
static bool __initdata parsed_default_hugepagesz;
69

70
/*
71 72
 * Protects updates to hugepage_freelists, hugepage_activelist, nr_huge_pages,
 * free_huge_pages, and surplus_huge_pages.
73
 */
74
DEFINE_SPINLOCK(hugetlb_lock);
75

76 77 78 79 80
/*
 * 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;
81
struct mutex *hugetlb_fault_mutex_table ____cacheline_aligned_in_smp;
82

83 84 85
/* Forward declaration */
static int hugetlb_acct_memory(struct hstate *h, long delta);

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

91
	spin_unlock_irqrestore(&spool->lock, irq_flags);
92 93

	/* If no pages are used, and no other handles to the subpool
E
Ethon Paul 已提交
94
	 * remain, give up any reservations based on minimum size and
95 96 97 98 99
	 * free the subpool */
	if (free) {
		if (spool->min_hpages != -1)
			hugetlb_acct_memory(spool->hstate,
						-spool->min_hpages);
100
		kfree(spool);
101
	}
102 103
}

104 105
struct hugepage_subpool *hugepage_new_subpool(struct hstate *h, long max_hpages,
						long min_hpages)
106 107 108
{
	struct hugepage_subpool *spool;

109
	spool = kzalloc(sizeof(*spool), GFP_KERNEL);
110 111 112 113 114
	if (!spool)
		return NULL;

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

	return spool;
}

void hugepage_put_subpool(struct hugepage_subpool *spool)
{
130 131 132
	unsigned long flags;

	spin_lock_irqsave(&spool->lock, flags);
133 134
	BUG_ON(!spool->count);
	spool->count--;
135
	unlock_or_release_subpool(spool, flags);
136 137
}

138 139 140
/*
 * Subpool accounting for allocating and reserving pages.
 * Return -ENOMEM if there are not enough resources to satisfy the
141
 * request.  Otherwise, return the number of pages by which the
142 143
 * 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
144
 * a subpool minimum size must be maintained.
145 146
 */
static long hugepage_subpool_get_pages(struct hugepage_subpool *spool,
147 148
				      long delta)
{
149
	long ret = delta;
150 151

	if (!spool)
152
		return ret;
153

154
	spin_lock_irq(&spool->lock);
155 156 157 158 159 160 161 162

	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;
		}
163 164
	}

165 166
	/* minimum size accounting */
	if (spool->min_hpages != -1 && spool->rsv_hpages) {
167 168 169 170 171 172 173 174 175 176 177 178 179 180
		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:
181
	spin_unlock_irq(&spool->lock);
182 183 184
	return ret;
}

185 186 187 188 189 190 191
/*
 * 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,
192 193
				       long delta)
{
194
	long ret = delta;
195
	unsigned long flags;
196

197
	if (!spool)
198
		return delta;
199

200
	spin_lock_irqsave(&spool->lock, flags);
201 202 203 204

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

205 206
	 /* minimum size accounting */
	if (spool->min_hpages != -1 && spool->used_hpages < spool->min_hpages) {
207 208 209 210 211 212 213 214 215 216 217 218 219 220
		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.
	 */
221
	unlock_or_release_subpool(spool, flags);
222 223

	return ret;
224 225 226 227 228 229 230 231 232
}

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
233
	return subpool_inode(file_inode(vma->vm_file));
234 235
}

236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255
/* Helper that removes a struct file_region from the resv_map cache and returns
 * it for use.
 */
static struct file_region *
get_file_region_entry_from_cache(struct resv_map *resv, long from, long to)
{
	struct file_region *nrg = NULL;

	VM_BUG_ON(resv->region_cache_count <= 0);

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

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

	return nrg;
}

256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277
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;
278 279 280 281 282 283 284 285 286 287 288
		/*
		 * The caller will hold exactly one h_cg->css reference for the
		 * whole contiguous reservation region. But this area might be
		 * scattered when there are already some file_regions reside in
		 * it. As a result, many file_regions may share only one css
		 * reference. In order to ensure that one file_region must hold
		 * exactly one h_cg->css reference, we should do css_get for
		 * each file_region and leave the reference held by caller
		 * untouched.
		 */
		css_get(&h_cg->css);
289 290 291 292 293 294 295 296 297 298 299 300 301
		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
}

302 303 304 305 306 307 308 309
static void put_uncharge_info(struct file_region *rg)
{
#ifdef CONFIG_CGROUP_HUGETLB
	if (rg->css)
		css_put(rg->css);
#endif
}

310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332
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);
333
		put_uncharge_info(rg);
334 335
		kfree(rg);

336
		rg = prg;
337 338 339 340 341 342 343 344
	}

	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);
345
		put_uncharge_info(rg);
346 347 348 349
		kfree(rg);
	}
}

350 351 352 353 354 355 356
/*
 * Must be called with resv->lock held.
 *
 * Calling this with regions_needed != NULL will count the number of pages
 * to be added but will not modify the linked list. And regions_needed will
 * indicate the number of file_regions needed in the cache to carry out to add
 * the regions for this range.
M
Mina Almasry 已提交
357 358
 */
static long add_reservation_in_range(struct resv_map *resv, long f, long t,
359
				     struct hugetlb_cgroup *h_cg,
360
				     struct hstate *h, long *regions_needed)
M
Mina Almasry 已提交
361
{
362
	long add = 0;
M
Mina Almasry 已提交
363
	struct list_head *head = &resv->regions;
364
	long last_accounted_offset = f;
M
Mina Almasry 已提交
365 366
	struct file_region *rg = NULL, *trg = NULL, *nrg = NULL;

367 368
	if (regions_needed)
		*regions_needed = 0;
M
Mina Almasry 已提交
369

370 371 372 373 374 375 376 377 378 379 380 381 382 383
	/* 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
384

385 386 387
		/* When we find a region that starts beyond our range, we've
		 * finished.
		 */
M
Mina Almasry 已提交
388 389 390
		if (rg->from > t)
			break;

391 392 393 394 395
		/* 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;
396
			if (!regions_needed) {
397 398
				nrg = get_file_region_entry_from_cache(
					resv, last_accounted_offset, rg->from);
399 400
				record_hugetlb_cgroup_uncharge_info(h_cg, h,
								    resv, nrg);
401
				list_add(&nrg->link, rg->link.prev);
402
				coalesce_file_region(resv, nrg);
403
			} else
404 405 406 407 408 409 410 411 412 413 414
				*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;
415
		if (!regions_needed) {
416 417
			nrg = get_file_region_entry_from_cache(
				resv, last_accounted_offset, t);
418
			record_hugetlb_cgroup_uncharge_info(h_cg, h, resv, nrg);
419
			list_add(&nrg->link, rg->link.prev);
420
			coalesce_file_region(resv, nrg);
421
		} else
422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459
			*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
460
		 */
461 462 463 464 465 466 467 468
		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
469 470
		}

471 472
		spin_lock(&resv->lock);

473 474
		list_splice(&allocated_regions, &resv->region_cache);
		resv->region_cache_count += to_allocate;
M
Mina Almasry 已提交
475 476
	}

477
	return 0;
M
Mina Almasry 已提交
478

479 480 481 482 483 484
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
485 486
}

487 488
/*
 * Add the huge page range represented by [f, t) to the reserve
489 490 491 492 493
 * 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.
494
 *
495 496 497 498
 * 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
499
 * this operation and we were not able to allocate, it returns -ENOMEM.
500 501 502
 * 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.
503
 */
504
static long region_add(struct resv_map *resv, long f, long t,
505 506
		       long in_regions_needed, struct hstate *h,
		       struct hugetlb_cgroup *h_cg)
507
{
508
	long add = 0, actual_regions_needed = 0;
509

510
	spin_lock(&resv->lock);
511 512 513
retry:

	/* Count how many regions are actually needed to execute this add. */
514 515
	add_reservation_in_range(resv, f, t, NULL, NULL,
				 &actual_regions_needed);
516

517
	/*
518 519 520 521 522 523 524
	 * 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.
525
	 */
526 527 528 529 530 531 532 533
	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);
534

535 536 537 538
		if (allocate_file_region_entries(
			    resv, actual_regions_needed - in_regions_needed)) {
			return -ENOMEM;
		}
539

540
		goto retry;
541 542
	}

543
	add = add_reservation_in_range(resv, f, t, h_cg, h, NULL);
544 545

	resv->adds_in_progress -= in_regions_needed;
546

547
	spin_unlock(&resv->lock);
548 549
	VM_BUG_ON(add < 0);
	return add;
550 551
}

552 553 554 555 556 557 558
/*
 * 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
559 560 561 562 563 564 565
 * 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.
566 567 568 569 570
 *
 * 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.
571
 */
572 573
static long region_chg(struct resv_map *resv, long f, long t,
		       long *out_regions_needed)
574 575 576
{
	long chg = 0;

577
	spin_lock(&resv->lock);
578

579
	/* Count how many hugepages in this range are NOT represented. */
580
	chg = add_reservation_in_range(resv, f, t, NULL, NULL,
581
				       out_regions_needed);
582

583 584
	if (*out_regions_needed == 0)
		*out_regions_needed = 1;
585

586 587
	if (allocate_file_region_entries(resv, *out_regions_needed))
		return -ENOMEM;
588

589
	resv->adds_in_progress += *out_regions_needed;
590 591

	spin_unlock(&resv->lock);
592 593 594
	return chg;
}

595 596 597 598 599
/*
 * 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
600 601 602
 * 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.
603 604 605 606 607
 *
 * 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.
 */
608 609
static void region_abort(struct resv_map *resv, long f, long t,
			 long regions_needed)
610 611 612
{
	spin_lock(&resv->lock);
	VM_BUG_ON(!resv->region_cache_count);
613
	resv->adds_in_progress -= regions_needed;
614 615 616
	spin_unlock(&resv->lock);
}

617
/*
618 619 620 621 622 623 624 625 626 627 628 629
 * 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.
630
 */
631
static long region_del(struct resv_map *resv, long f, long t)
632
{
633
	struct list_head *head = &resv->regions;
634
	struct file_region *rg, *trg;
635 636
	struct file_region *nrg = NULL;
	long del = 0;
637

638
retry:
639
	spin_lock(&resv->lock);
640
	list_for_each_entry_safe(rg, trg, head, link) {
641 642 643 644 645 646 647 648
		/*
		 * 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))
649
			continue;
650

651
		if (rg->from >= t)
652 653
			break;

654 655 656 657 658 659 660 661 662 663 664 665 666
		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--;
			}
667

668 669 670 671 672 673 674 675 676
			if (!nrg) {
				spin_unlock(&resv->lock);
				nrg = kmalloc(sizeof(*nrg), GFP_KERNEL);
				if (!nrg)
					return -ENOMEM;
				goto retry;
			}

			del += t - f;
677
			hugetlb_cgroup_uncharge_file_region(
678
				resv, rg, t - f, false);
679 680 681 682

			/* New entry for end of split region */
			nrg->from = t;
			nrg->to = rg->to;
683 684 685

			copy_hugetlb_cgroup_uncharge_info(nrg, rg);

686 687 688 689 690 691 692
			INIT_LIST_HEAD(&nrg->link);

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

			list_add(&nrg->link, &rg->link);
			nrg = NULL;
693
			break;
694 695 696 697
		}

		if (f <= rg->from && t >= rg->to) { /* Remove entire region */
			del += rg->to - rg->from;
698
			hugetlb_cgroup_uncharge_file_region(resv, rg,
699
							    rg->to - rg->from, true);
700 701 702 703 704 705
			list_del(&rg->link);
			kfree(rg);
			continue;
		}

		if (f <= rg->from) {	/* Trim beginning of region */
706
			hugetlb_cgroup_uncharge_file_region(resv, rg,
707
							    t - rg->from, false);
708

709 710 711
			del += t - rg->from;
			rg->from = t;
		} else {		/* Trim end of region */
712
			hugetlb_cgroup_uncharge_file_region(resv, rg,
713
							    rg->to - f, false);
714 715 716

			del += rg->to - f;
			rg->to = f;
717
		}
718
	}
719 720

	spin_unlock(&resv->lock);
721 722
	kfree(nrg);
	return del;
723 724
}

725 726 727 728 729 730 731 732 733
/*
 * 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.
 */
734
void hugetlb_fix_reserve_counts(struct inode *inode)
735 736 737
{
	struct hugepage_subpool *spool = subpool_inode(inode);
	long rsv_adjust;
738
	bool reserved = false;
739 740

	rsv_adjust = hugepage_subpool_get_pages(spool, 1);
741
	if (rsv_adjust > 0) {
742 743
		struct hstate *h = hstate_inode(inode);

744 745 746 747
		if (!hugetlb_acct_memory(h, 1))
			reserved = true;
	} else if (!rsv_adjust) {
		reserved = true;
748
	}
749 750 751

	if (!reserved)
		pr_warn("hugetlb: Huge Page Reserved count may go negative.\n");
752 753
}

754 755 756 757
/*
 * Count and return the number of huge pages in the reserve map
 * that intersect with the range [f, t).
 */
758
static long region_count(struct resv_map *resv, long f, long t)
759
{
760
	struct list_head *head = &resv->regions;
761 762 763
	struct file_region *rg;
	long chg = 0;

764
	spin_lock(&resv->lock);
765 766
	/* Locate each segment we overlap with, and count that overlap. */
	list_for_each_entry(rg, head, link) {
767 768
		long seg_from;
		long seg_to;
769 770 771 772 773 774 775 776 777 778 779

		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;
	}
780
	spin_unlock(&resv->lock);
781 782 783 784

	return chg;
}

785 786 787 788
/*
 * Convert the address within this vma to the page offset within
 * the mapping, in pagecache page units; huge pages here.
 */
789 790
static pgoff_t vma_hugecache_offset(struct hstate *h,
			struct vm_area_struct *vma, unsigned long address)
791
{
792 793
	return ((address - vma->vm_start) >> huge_page_shift(h)) +
			(vma->vm_pgoff >> huge_page_order(h));
794 795
}

796 797 798 799 800
pgoff_t linear_hugepage_index(struct vm_area_struct *vma,
				     unsigned long address)
{
	return vma_hugecache_offset(hstate_vma(vma), vma, address);
}
801
EXPORT_SYMBOL_GPL(linear_hugepage_index);
802

803 804 805 806 807 808
/*
 * 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)
{
809 810 811
	if (vma->vm_ops && vma->vm_ops->pagesize)
		return vma->vm_ops->pagesize(vma);
	return PAGE_SIZE;
812
}
813
EXPORT_SYMBOL_GPL(vma_kernel_pagesize);
814

815 816 817
/*
 * 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
818 819
 * architectures where it differs, an architecture-specific 'strong'
 * version of this symbol is required.
820
 */
821
__weak unsigned long vma_mmu_pagesize(struct vm_area_struct *vma)
822 823 824 825
{
	return vma_kernel_pagesize(vma);
}

826 827 828 829 830 831 832
/*
 * 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)
833
#define HPAGE_RESV_MASK (HPAGE_RESV_OWNER | HPAGE_RESV_UNMAPPED)
834

835 836 837 838 839 840 841 842 843
/*
 * 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.
844 845 846 847 848 849 850 851 852
 *
 * 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.
853
 */
854 855 856 857 858 859 860 861 862 863 864
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;
}

865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883
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
}

884
struct resv_map *resv_map_alloc(void)
885 886
{
	struct resv_map *resv_map = kmalloc(sizeof(*resv_map), GFP_KERNEL);
887 888 889 890 891
	struct file_region *rg = kmalloc(sizeof(*rg), GFP_KERNEL);

	if (!resv_map || !rg) {
		kfree(resv_map);
		kfree(rg);
892
		return NULL;
893
	}
894 895

	kref_init(&resv_map->refs);
896
	spin_lock_init(&resv_map->lock);
897 898
	INIT_LIST_HEAD(&resv_map->regions);

899
	resv_map->adds_in_progress = 0;
900 901 902 903 904 905 906
	/*
	 * 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);
907 908 909 910 911

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

912 913 914
	return resv_map;
}

915
void resv_map_release(struct kref *ref)
916 917
{
	struct resv_map *resv_map = container_of(ref, struct resv_map, refs);
918 919
	struct list_head *head = &resv_map->region_cache;
	struct file_region *rg, *trg;
920 921

	/* Clear out any active regions before we release the map. */
922
	region_del(resv_map, 0, LONG_MAX);
923 924 925 926 927 928 929 930 931

	/* ... 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);

932 933 934
	kfree(resv_map);
}

935 936
static inline struct resv_map *inode_resv_map(struct inode *inode)
{
937 938 939 940 941 942 943 944 945
	/*
	 * 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;
946 947
}

948
static struct resv_map *vma_resv_map(struct vm_area_struct *vma)
949
{
950
	VM_BUG_ON_VMA(!is_vm_hugetlb_page(vma), vma);
951 952 953 954 955 956 957
	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 {
958 959
		return (struct resv_map *)(get_vma_private_data(vma) &
							~HPAGE_RESV_MASK);
960
	}
961 962
}

963
static void set_vma_resv_map(struct vm_area_struct *vma, struct resv_map *map)
964
{
965 966
	VM_BUG_ON_VMA(!is_vm_hugetlb_page(vma), vma);
	VM_BUG_ON_VMA(vma->vm_flags & VM_MAYSHARE, vma);
967

968 969
	set_vma_private_data(vma, (get_vma_private_data(vma) &
				HPAGE_RESV_MASK) | (unsigned long)map);
970 971 972 973
}

static void set_vma_resv_flags(struct vm_area_struct *vma, unsigned long flags)
{
974 975
	VM_BUG_ON_VMA(!is_vm_hugetlb_page(vma), vma);
	VM_BUG_ON_VMA(vma->vm_flags & VM_MAYSHARE, vma);
976 977

	set_vma_private_data(vma, get_vma_private_data(vma) | flags);
978 979 980 981
}

static int is_vma_resv_set(struct vm_area_struct *vma, unsigned long flag)
{
982
	VM_BUG_ON_VMA(!is_vm_hugetlb_page(vma), vma);
983 984

	return (get_vma_private_data(vma) & flag) != 0;
985 986
}

987
/* Reset counters to 0 and clear all HPAGE_RESV_* flags */
988 989
void reset_vma_resv_huge_pages(struct vm_area_struct *vma)
{
990
	VM_BUG_ON_VMA(!is_vm_hugetlb_page(vma), vma);
991
	if (!(vma->vm_flags & VM_MAYSHARE))
992 993 994 995
		vma->vm_private_data = (void *)0;
}

/* Returns true if the VMA has associated reserve pages */
996
static bool vma_has_reserves(struct vm_area_struct *vma, long chg)
997
{
998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008
	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)
1009
			return true;
1010
		else
1011
			return false;
1012
	}
1013 1014

	/* Shared mappings always use reserves */
1015 1016 1017 1018 1019
	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
1020
		 * fallocate.  In this case, there really are no reserves to
1021 1022 1023 1024 1025 1026 1027
		 * use.  This situation is indicated if chg != 0.
		 */
		if (chg)
			return false;
		else
			return true;
	}
1028 1029 1030 1031 1032

	/*
	 * Only the process that called mmap() has reserves for
	 * private mappings.
	 */
1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053
	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;
	}
1054

1055
	return false;
1056 1057
}

1058
static void enqueue_huge_page(struct hstate *h, struct page *page)
L
Linus Torvalds 已提交
1059 1060
{
	int nid = page_to_nid(page);
1061 1062

	lockdep_assert_held(&hugetlb_lock);
1063
	list_move(&page->lru, &h->hugepage_freelists[nid]);
1064 1065
	h->free_huge_pages++;
	h->free_huge_pages_node[nid]++;
1066
	SetHPageFreed(page);
L
Linus Torvalds 已提交
1067 1068
}

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

1074
	lockdep_assert_held(&hugetlb_lock);
1075 1076 1077
	list_for_each_entry(page, &h->hugepage_freelists[nid], lru) {
		if (nocma && is_migrate_cma_page(page))
			continue;
1078

1079 1080 1081 1082 1083
		if (PageHWPoison(page))
			continue;

		list_move(&page->lru, &h->hugepage_activelist);
		set_page_refcounted(page);
1084
		ClearHPageFreed(page);
1085 1086 1087
		h->free_huge_pages--;
		h->free_huge_pages_node[nid]--;
		return page;
1088 1089
	}

1090
	return NULL;
1091 1092
}

1093 1094
static struct page *dequeue_huge_page_nodemask(struct hstate *h, gfp_t gfp_mask, int nid,
		nodemask_t *nmask)
1095
{
1096 1097 1098 1099
	unsigned int cpuset_mems_cookie;
	struct zonelist *zonelist;
	struct zone *zone;
	struct zoneref *z;
1100
	int node = NUMA_NO_NODE;
1101

1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117
	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);
1118 1119 1120 1121 1122

		page = dequeue_huge_page_node_exact(h, node);
		if (page)
			return page;
	}
1123 1124 1125
	if (unlikely(read_mems_allowed_retry(cpuset_mems_cookie)))
		goto retry_cpuset;

1126 1127 1128
	return NULL;
}

1129 1130
static struct page *dequeue_huge_page_vma(struct hstate *h,
				struct vm_area_struct *vma,
1131 1132
				unsigned long address, int avoid_reserve,
				long chg)
L
Linus Torvalds 已提交
1133
{
1134
	struct page *page;
1135
	struct mempolicy *mpol;
1136
	gfp_t gfp_mask;
1137
	nodemask_t *nodemask;
1138
	int nid;
L
Linus Torvalds 已提交
1139

1140 1141 1142 1143 1144
	/*
	 * 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
	 */
1145
	if (!vma_has_reserves(vma, chg) &&
1146
			h->free_huge_pages - h->resv_huge_pages == 0)
1147
		goto err;
1148

1149
	/* If reserves cannot be used, ensure enough pages are in the pool */
1150
	if (avoid_reserve && h->free_huge_pages - h->resv_huge_pages == 0)
1151
		goto err;
1152

1153 1154
	gfp_mask = htlb_alloc_mask(h);
	nid = huge_node(vma, address, gfp_mask, &mpol, &nodemask);
1155 1156
	page = dequeue_huge_page_nodemask(h, gfp_mask, nid, nodemask);
	if (page && !avoid_reserve && vma_has_reserves(vma, chg)) {
1157
		SetHPageRestoreReserve(page);
1158
		h->resv_huge_pages--;
L
Linus Torvalds 已提交
1159
	}
1160

1161
	mpol_cond_put(mpol);
L
Linus Torvalds 已提交
1162
	return page;
1163 1164 1165

err:
	return NULL;
L
Linus Torvalds 已提交
1166 1167
}

1168 1169 1170 1171 1172 1173 1174 1175 1176
/*
 * 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)
{
1177
	nid = next_node_in(nid, *nodes_allowed);
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
	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;
}

/*
1210
 * helper for remove_pool_huge_page() - return the previously saved
1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238
 * 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--)

1239
#ifdef CONFIG_ARCH_HAS_GIGANTIC_PAGE
1240
static void destroy_compound_gigantic_page(struct page *page,
1241
					unsigned int order)
1242 1243 1244 1245 1246
{
	int i;
	int nr_pages = 1 << order;
	struct page *p = page + 1;

1247
	atomic_set(compound_mapcount_ptr(page), 0);
1248 1249 1250
	if (hpage_pincount_available(page))
		atomic_set(compound_pincount_ptr(page), 0);

1251
	for (i = 1; i < nr_pages; i++, p = mem_map_next(p, page, i)) {
1252
		clear_compound_head(p);
1253 1254 1255 1256
		set_page_refcounted(p);
	}

	set_compound_order(page, 0);
1257
	page[1].compound_nr = 0;
1258 1259 1260
	__ClearPageHead(page);
}

1261
static void free_gigantic_page(struct page *page, unsigned int order)
1262
{
1263 1264 1265 1266
	/*
	 * If the page isn't allocated using the cma allocator,
	 * cma_release() returns false.
	 */
1267 1268
#ifdef CONFIG_CMA
	if (cma_release(hugetlb_cma[page_to_nid(page)], page, 1 << order))
1269
		return;
1270
#endif
1271

1272 1273 1274
	free_contig_range(page_to_pfn(page), 1 << order);
}

1275
#ifdef CONFIG_CONTIG_ALLOC
1276 1277
static struct page *alloc_gigantic_page(struct hstate *h, gfp_t gfp_mask,
		int nid, nodemask_t *nodemask)
1278
{
1279
	unsigned long nr_pages = 1UL << huge_page_order(h);
1280 1281
	if (nid == NUMA_NO_NODE)
		nid = numa_mem_id();
1282

1283 1284
#ifdef CONFIG_CMA
	{
1285 1286 1287
		struct page *page;
		int node;

1288 1289 1290
		if (hugetlb_cma[nid]) {
			page = cma_alloc(hugetlb_cma[nid], nr_pages,
					huge_page_order(h), true);
1291 1292 1293
			if (page)
				return page;
		}
1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305

		if (!(gfp_mask & __GFP_THISNODE)) {
			for_each_node_mask(node, *nodemask) {
				if (node == nid || !hugetlb_cma[node])
					continue;

				page = cma_alloc(hugetlb_cma[node], nr_pages,
						huge_page_order(h), true);
				if (page)
					return page;
			}
		}
1306
	}
1307
#endif
1308

1309
	return alloc_contig_pages(nr_pages, gfp_mask, nid, nodemask);
1310 1311 1312
}

static void prep_new_huge_page(struct hstate *h, struct page *page, int nid);
1313
static void prep_compound_gigantic_page(struct page *page, unsigned int order);
1314 1315 1316 1317 1318 1319 1320
#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 */
1321

1322
#else /* !CONFIG_ARCH_HAS_GIGANTIC_PAGE */
1323
static struct page *alloc_gigantic_page(struct hstate *h, gfp_t gfp_mask,
1324 1325 1326 1327
					int nid, nodemask_t *nodemask)
{
	return NULL;
}
1328
static inline void free_gigantic_page(struct page *page, unsigned int order) { }
1329
static inline void destroy_compound_gigantic_page(struct page *page,
1330
						unsigned int order) { }
1331 1332
#endif

1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346
/*
 * Remove hugetlb page from lists, and update dtor so that page appears
 * as just a compound page.  A reference is held on the page.
 *
 * Must be called with hugetlb lock held.
 */
static void remove_hugetlb_page(struct hstate *h, struct page *page,
							bool adjust_surplus)
{
	int nid = page_to_nid(page);

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

1347
	lockdep_assert_held(&hugetlb_lock);
1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368
	if (hstate_is_gigantic(h) && !gigantic_page_runtime_supported())
		return;

	list_del(&page->lru);

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

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

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

1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401
static void add_hugetlb_page(struct hstate *h, struct page *page,
			     bool adjust_surplus)
{
	int zeroed;
	int nid = page_to_nid(page);

	VM_BUG_ON_PAGE(!HPageVmemmapOptimized(page), page);

	lockdep_assert_held(&hugetlb_lock);

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

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

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

	/*
	 * This page is now managed by the hugetlb allocator and has
	 * no users -- drop the last reference.
	 */
	zeroed = put_page_testzero(page);
	VM_BUG_ON_PAGE(!zeroed, page);
	arch_clear_hugepage_flags(page);
	enqueue_huge_page(h, page);
}

1402
static void __update_and_free_page(struct hstate *h, struct page *page)
A
Adam Litke 已提交
1403 1404
{
	int i;
1405
	struct page *subpage = page;
1406

1407
	if (hstate_is_gigantic(h) && !gigantic_page_runtime_supported())
1408
		return;
1409

1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421
	if (alloc_huge_page_vmemmap(h, page)) {
		spin_lock_irq(&hugetlb_lock);
		/*
		 * If we cannot allocate vmemmap pages, just refuse to free the
		 * page and put the page back on the hugetlb free list and treat
		 * as a surplus page.
		 */
		add_hugetlb_page(h, page, true);
		spin_unlock_irq(&hugetlb_lock);
		return;
	}

1422 1423 1424
	for (i = 0; i < pages_per_huge_page(h);
	     i++, subpage = mem_map_next(subpage, page, i)) {
		subpage->flags &= ~(1 << PG_locked | 1 << PG_error |
1425
				1 << PG_referenced | 1 << PG_dirty |
1426 1427
				1 << PG_active | 1 << PG_private |
				1 << PG_writeback);
A
Adam Litke 已提交
1428
	}
1429 1430 1431 1432 1433 1434
	if (hstate_is_gigantic(h)) {
		destroy_compound_gigantic_page(page, huge_page_order(h));
		free_gigantic_page(page, huge_page_order(h));
	} else {
		__free_pages(page, huge_page_order(h));
	}
A
Adam Litke 已提交
1435 1436
}

1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487
/*
 * As update_and_free_page() can be called under any context, so we cannot
 * use GFP_KERNEL to allocate vmemmap pages. However, we can defer the
 * actual freeing in a workqueue to prevent from using GFP_ATOMIC to allocate
 * the vmemmap pages.
 *
 * free_hpage_workfn() locklessly retrieves the linked list of pages to be
 * freed and frees them one-by-one. As the page->mapping pointer is going
 * to be cleared in free_hpage_workfn() anyway, it is reused as the llist_node
 * structure of a lockless linked list of huge pages to be freed.
 */
static LLIST_HEAD(hpage_freelist);

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

	node = llist_del_all(&hpage_freelist);

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

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

		__update_and_free_page(h, page);

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

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

static void update_and_free_page(struct hstate *h, struct page *page,
				 bool atomic)
{
1488
	if (!HPageVmemmapOptimized(page) || !atomic) {
1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503
		__update_and_free_page(h, page);
		return;
	}

	/*
	 * Defer freeing to avoid using GFP_ATOMIC to allocate vmemmap pages.
	 *
	 * Only call schedule_work() if hpage_freelist is previously
	 * empty. Otherwise, schedule_work() had been called but the workfn
	 * hasn't retrieved the list yet.
	 */
	if (llist_add((struct llist_node *)&page->mapping, &hpage_freelist))
		schedule_work(&free_hpage_work);
}

1504 1505 1506 1507 1508
static void update_and_free_pages_bulk(struct hstate *h, struct list_head *list)
{
	struct page *page, *t_page;

	list_for_each_entry_safe(page, t_page, list, lru) {
1509
		update_and_free_page(h, page, false);
1510 1511 1512 1513
		cond_resched();
	}
}

1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524
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;
}

1525
void free_huge_page(struct page *page)
1526
{
1527 1528 1529 1530
	/*
	 * Can't pass hstate in here because it is called from the
	 * compound page destructor.
	 */
1531
	struct hstate *h = page_hstate(page);
1532
	int nid = page_to_nid(page);
1533
	struct hugepage_subpool *spool = hugetlb_page_subpool(page);
1534
	bool restore_reserve;
1535
	unsigned long flags;
1536

1537 1538
	VM_BUG_ON_PAGE(page_count(page), page);
	VM_BUG_ON_PAGE(page_mapcount(page), page);
1539

1540
	hugetlb_set_page_subpool(page, NULL);
1541
	page->mapping = NULL;
1542 1543
	restore_reserve = HPageRestoreReserve(page);
	ClearHPageRestoreReserve(page);
1544

1545
	/*
1546
	 * If HPageRestoreReserve was set on page, page allocation consumed a
1547 1548 1549 1550 1551
	 * 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.
1552
	 */
1553 1554 1555 1556 1557 1558 1559 1560 1561 1562
	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;
	}
1563

1564
	spin_lock_irqsave(&hugetlb_lock, flags);
1565
	ClearHPageMigratable(page);
1566 1567
	hugetlb_cgroup_uncharge_page(hstate_index(h),
				     pages_per_huge_page(h), page);
1568 1569
	hugetlb_cgroup_uncharge_page_rsvd(hstate_index(h),
					  pages_per_huge_page(h), page);
1570 1571 1572
	if (restore_reserve)
		h->resv_huge_pages++;

1573
	if (HPageTemporary(page)) {
1574
		remove_hugetlb_page(h, page, false);
1575
		spin_unlock_irqrestore(&hugetlb_lock, flags);
1576
		update_and_free_page(h, page, true);
1577
	} else if (h->surplus_huge_pages_node[nid]) {
1578
		/* remove the page from active list */
1579
		remove_hugetlb_page(h, page, true);
1580
		spin_unlock_irqrestore(&hugetlb_lock, flags);
1581
		update_and_free_page(h, page, true);
1582
	} else {
1583
		arch_clear_hugepage_flags(page);
1584
		enqueue_huge_page(h, page);
1585
		spin_unlock_irqrestore(&hugetlb_lock, flags);
1586 1587 1588
	}
}

1589
static void prep_new_huge_page(struct hstate *h, struct page *page, int nid)
1590
{
1591
	free_huge_page_vmemmap(h, page);
1592
	INIT_LIST_HEAD(&page->lru);
1593
	set_compound_page_dtor(page, HUGETLB_PAGE_DTOR);
1594
	hugetlb_set_page_subpool(page, NULL);
1595
	set_hugetlb_cgroup(page, NULL);
1596
	set_hugetlb_cgroup_rsvd(page, NULL);
1597
	spin_lock_irq(&hugetlb_lock);
1598 1599
	h->nr_huge_pages++;
	h->nr_huge_pages_node[nid]++;
1600
	ClearHPageFreed(page);
1601
	spin_unlock_irq(&hugetlb_lock);
1602 1603
}

1604
static void prep_compound_gigantic_page(struct page *page, unsigned int order)
1605 1606 1607 1608 1609 1610 1611
{
	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);
1612
	__ClearPageReserved(page);
1613
	__SetPageHead(page);
1614
	for (i = 1; i < nr_pages; i++, p = mem_map_next(p, page, i)) {
1615 1616 1617 1618
		/*
		 * 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
1619
		 * too.  Otherwise drivers using get_user_pages() to access tail
1620 1621 1622 1623 1624 1625 1626 1627
		 * 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);
1628
		set_page_count(p, 0);
1629
		set_compound_head(p, page);
1630
	}
1631
	atomic_set(compound_mapcount_ptr(page), -1);
1632 1633 1634

	if (hpage_pincount_available(page))
		atomic_set(compound_pincount_ptr(page), 0);
1635 1636
}

A
Andrew Morton 已提交
1637 1638 1639 1640 1641
/*
 * PageHuge() only returns true for hugetlbfs pages, but not for normal or
 * transparent huge pages.  See the PageTransHuge() documentation for more
 * details.
 */
1642 1643 1644 1645 1646 1647
int PageHuge(struct page *page)
{
	if (!PageCompound(page))
		return 0;

	page = compound_head(page);
1648
	return page[1].compound_dtor == HUGETLB_PAGE_DTOR;
1649
}
1650 1651
EXPORT_SYMBOL_GPL(PageHuge);

1652 1653 1654 1655 1656 1657 1658 1659 1660
/*
 * 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;

1661
	return page_head[1].compound_dtor == HUGETLB_PAGE_DTOR;
1662 1663
}

1664 1665 1666
/*
 * Find and lock address space (mapping) in write mode.
 *
1667 1668 1669
 * Upon entry, the page is locked which means that page_mapping() is
 * stable.  Due to locking order, we can only trylock_write.  If we can
 * not get the lock, simply return NULL to caller.
1670 1671 1672
 */
struct address_space *hugetlb_page_mapping_lock_write(struct page *hpage)
{
1673
	struct address_space *mapping = page_mapping(hpage);
1674 1675 1676 1677 1678 1679 1680

	if (!mapping)
		return mapping;

	if (i_mmap_trylock_write(mapping))
		return mapping;

1681
	return NULL;
1682 1683
}

1684
pgoff_t hugetlb_basepage_index(struct page *page)
1685 1686 1687 1688 1689 1690 1691 1692 1693 1694 1695 1696 1697
{
	struct page *page_head = compound_head(page);
	pgoff_t index = page_index(page_head);
	unsigned long compound_idx;

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

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

1698
static struct page *alloc_buddy_huge_page(struct hstate *h,
1699 1700
		gfp_t gfp_mask, int nid, nodemask_t *nmask,
		nodemask_t *node_alloc_noretry)
L
Linus Torvalds 已提交
1701
{
1702
	int order = huge_page_order(h);
L
Linus Torvalds 已提交
1703
	struct page *page;
1704
	bool alloc_try_hard = true;
1705

1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717
	/*
	 * 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;
1718 1719
	if (nid == NUMA_NO_NODE)
		nid = numa_mem_id();
1720
	page = __alloc_pages(gfp_mask, order, nid, nmask);
1721 1722 1723 1724
	if (page)
		__count_vm_event(HTLB_BUDDY_PGALLOC);
	else
		__count_vm_event(HTLB_BUDDY_PGALLOC_FAIL);
1725

1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736 1737 1738 1739 1740 1741
	/*
	 * 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);

1742 1743 1744
	return page;
}

1745 1746 1747 1748 1749
/*
 * 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,
1750 1751
		gfp_t gfp_mask, int nid, nodemask_t *nmask,
		nodemask_t *node_alloc_noretry)
1752 1753 1754 1755 1756 1757 1758
{
	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,
1759
				nid, nmask, node_alloc_noretry);
1760 1761 1762 1763 1764 1765 1766 1767 1768 1769
	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;
}

1770 1771 1772 1773
/*
 * Allocates a fresh page to the hugetlb allocator pool in the node interleaved
 * manner.
 */
1774 1775
static int alloc_pool_huge_page(struct hstate *h, nodemask_t *nodes_allowed,
				nodemask_t *node_alloc_noretry)
1776 1777 1778
{
	struct page *page;
	int nr_nodes, node;
1779
	gfp_t gfp_mask = htlb_alloc_mask(h) | __GFP_THISNODE;
1780 1781

	for_each_node_mask_to_alloc(h, nr_nodes, node, nodes_allowed) {
1782 1783
		page = alloc_fresh_huge_page(h, gfp_mask, node, nodes_allowed,
						node_alloc_noretry);
1784
		if (page)
1785 1786 1787
			break;
	}

1788 1789
	if (!page)
		return 0;
1790

1791 1792 1793
	put_page(page); /* free it into the hugepage allocator */

	return 1;
1794 1795
}

1796
/*
1797 1798 1799 1800
 * Remove huge page from pool from next node to free.  Attempt to keep
 * persistent huge pages more or less balanced over allowed nodes.
 * This routine only 'removes' the hugetlb page.  The caller must make
 * an additional call to free the page to low level allocators.
1801 1802
 * Called with hugetlb_lock locked.
 */
1803 1804 1805
static struct page *remove_pool_huge_page(struct hstate *h,
						nodemask_t *nodes_allowed,
						 bool acct_surplus)
1806
{
1807
	int nr_nodes, node;
1808
	struct page *page = NULL;
1809

1810
	lockdep_assert_held(&hugetlb_lock);
1811
	for_each_node_mask_to_free(h, nr_nodes, node, nodes_allowed) {
1812 1813 1814 1815
		/*
		 * If we're returning unused surplus pages, only examine
		 * nodes with surplus pages.
		 */
1816 1817
		if ((!acct_surplus || h->surplus_huge_pages_node[node]) &&
		    !list_empty(&h->hugepage_freelists[node])) {
1818
			page = list_entry(h->hugepage_freelists[node].next,
1819
					  struct page, lru);
1820
			remove_hugetlb_page(h, page, acct_surplus);
1821
			break;
1822
		}
1823
	}
1824

1825
	return page;
1826 1827
}

1828 1829
/*
 * Dissolve a given free hugepage into free buddy pages. This function does
1830 1831 1832
 * nothing for in-use hugepages and non-hugepages.
 * This function returns values like below:
 *
1833 1834 1835 1836 1837 1838 1839 1840
 *  -ENOMEM: failed to allocate vmemmap pages to free the freed hugepages
 *           when the system is under memory pressure and the feature of
 *           freeing unused vmemmap pages associated with each hugetlb page
 *           is enabled.
 *  -EBUSY:  failed to dissolved free hugepages or the hugepage is in-use
 *           (allocated or reserved.)
 *       0:  successfully dissolved free hugepages or the page is not a
 *           hugepage (considered as already dissolved)
1841
 */
1842
int dissolve_free_huge_page(struct page *page)
1843
{
1844
	int rc = -EBUSY;
1845

1846
retry:
1847 1848 1849 1850
	/* Not to disrupt normal path by vainly holding hugetlb_lock */
	if (!PageHuge(page))
		return 0;

1851
	spin_lock_irq(&hugetlb_lock);
1852 1853 1854 1855 1856 1857
	if (!PageHuge(page)) {
		rc = 0;
		goto out;
	}

	if (!page_count(page)) {
1858 1859
		struct page *head = compound_head(page);
		struct hstate *h = page_hstate(head);
1860
		if (h->free_huge_pages - h->resv_huge_pages == 0)
1861
			goto out;
1862 1863 1864 1865 1866

		/*
		 * We should make sure that the page is already on the free list
		 * when it is dissolved.
		 */
1867
		if (unlikely(!HPageFreed(head))) {
1868
			spin_unlock_irq(&hugetlb_lock);
1869 1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880 1881
			cond_resched();

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

1882
		remove_hugetlb_page(h, head, false);
1883
		h->max_huge_pages--;
1884
		spin_unlock_irq(&hugetlb_lock);
1885 1886 1887 1888 1889 1890 1891 1892 1893 1894 1895 1896 1897 1898 1899 1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913

		/*
		 * Normally update_and_free_page will allocate required vmemmmap
		 * before freeing the page.  update_and_free_page will fail to
		 * free the page if it can not allocate required vmemmap.  We
		 * need to adjust max_huge_pages if the page is not freed.
		 * Attempt to allocate vmemmmap here so that we can take
		 * appropriate action on failure.
		 */
		rc = alloc_huge_page_vmemmap(h, head);
		if (!rc) {
			/*
			 * Move PageHWPoison flag from head page to the raw
			 * error page, which makes any subpages rather than
			 * the error page reusable.
			 */
			if (PageHWPoison(head) && page != head) {
				SetPageHWPoison(page);
				ClearPageHWPoison(head);
			}
			update_and_free_page(h, head, false);
		} else {
			spin_lock_irq(&hugetlb_lock);
			add_hugetlb_page(h, head, false);
			h->max_huge_pages++;
			spin_unlock_irq(&hugetlb_lock);
		}

		return rc;
1914
	}
1915
out:
1916
	spin_unlock_irq(&hugetlb_lock);
1917
	return rc;
1918 1919 1920 1921 1922
}

/*
 * Dissolve free hugepages in a given pfn range. Used by memory hotplug to
 * make specified memory blocks removable from the system.
1923 1924
 * Note that this will dissolve a free gigantic hugepage completely, if any
 * part of it lies within the given range.
1925 1926
 * Also note that if dissolve_free_huge_page() returns with an error, all
 * free hugepages that were dissolved before that error are lost.
1927
 */
1928
int dissolve_free_huge_pages(unsigned long start_pfn, unsigned long end_pfn)
1929 1930
{
	unsigned long pfn;
1931
	struct page *page;
1932
	int rc = 0;
1933

1934
	if (!hugepages_supported())
1935
		return rc;
1936

1937 1938
	for (pfn = start_pfn; pfn < end_pfn; pfn += 1 << minimum_order) {
		page = pfn_to_page(pfn);
1939 1940 1941
		rc = dissolve_free_huge_page(page);
		if (rc)
			break;
1942
	}
1943 1944

	return rc;
1945 1946
}

1947 1948 1949
/*
 * Allocates a fresh surplus page from the page allocator.
 */
1950
static struct page *alloc_surplus_huge_page(struct hstate *h, gfp_t gfp_mask,
1951
		int nid, nodemask_t *nmask)
1952
{
1953
	struct page *page = NULL;
1954

1955
	if (hstate_is_gigantic(h))
1956 1957
		return NULL;

1958
	spin_lock_irq(&hugetlb_lock);
1959 1960
	if (h->surplus_huge_pages >= h->nr_overcommit_huge_pages)
		goto out_unlock;
1961
	spin_unlock_irq(&hugetlb_lock);
1962

1963
	page = alloc_fresh_huge_page(h, gfp_mask, nid, nmask, NULL);
1964
	if (!page)
1965
		return NULL;
1966

1967
	spin_lock_irq(&hugetlb_lock);
1968 1969 1970 1971 1972 1973 1974 1975
	/*
	 * 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) {
1976
		SetHPageTemporary(page);
1977
		spin_unlock_irq(&hugetlb_lock);
1978
		put_page(page);
1979
		return NULL;
1980 1981
	} else {
		h->surplus_huge_pages++;
1982
		h->surplus_huge_pages_node[page_to_nid(page)]++;
1983
	}
1984 1985

out_unlock:
1986
	spin_unlock_irq(&hugetlb_lock);
1987 1988 1989 1990

	return page;
}

1991
static struct page *alloc_migrate_huge_page(struct hstate *h, gfp_t gfp_mask,
1992
				     int nid, nodemask_t *nmask)
1993 1994 1995 1996 1997 1998
{
	struct page *page;

	if (hstate_is_gigantic(h))
		return NULL;

1999
	page = alloc_fresh_huge_page(h, gfp_mask, nid, nmask, NULL);
2000 2001 2002 2003 2004 2005 2006
	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
	 */
2007
	SetHPageTemporary(page);
2008 2009 2010 2011

	return page;
}

2012 2013 2014
/*
 * Use the VMA's mpolicy to allocate a huge page from the buddy.
 */
D
Dave Hansen 已提交
2015
static
2016
struct page *alloc_buddy_huge_page_with_mpol(struct hstate *h,
2017 2018
		struct vm_area_struct *vma, unsigned long addr)
{
2019 2020 2021 2022 2023 2024 2025
	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);
2026
	page = alloc_surplus_huge_page(h, gfp_mask, nid, nodemask);
2027 2028 2029
	mpol_cond_put(mpol);

	return page;
2030 2031
}

2032
/* page migration callback function */
2033
struct page *alloc_huge_page_nodemask(struct hstate *h, int preferred_nid,
2034
		nodemask_t *nmask, gfp_t gfp_mask)
2035
{
2036
	spin_lock_irq(&hugetlb_lock);
2037
	if (h->free_huge_pages - h->resv_huge_pages > 0) {
2038 2039 2040 2041
		struct page *page;

		page = dequeue_huge_page_nodemask(h, gfp_mask, preferred_nid, nmask);
		if (page) {
2042
			spin_unlock_irq(&hugetlb_lock);
2043
			return page;
2044 2045
		}
	}
2046
	spin_unlock_irq(&hugetlb_lock);
2047

2048
	return alloc_migrate_huge_page(h, gfp_mask, preferred_nid, nmask);
2049 2050
}

2051
/* mempolicy aware migration callback */
2052 2053
struct page *alloc_huge_page_vma(struct hstate *h, struct vm_area_struct *vma,
		unsigned long address)
2054 2055 2056 2057 2058 2059 2060 2061 2062
{
	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);
2063
	page = alloc_huge_page_nodemask(h, node, nodemask, gfp_mask);
2064 2065 2066 2067 2068
	mpol_cond_put(mpol);

	return page;
}

2069
/*
L
Lucas De Marchi 已提交
2070
 * Increase the hugetlb pool such that it can accommodate a reservation
2071 2072
 * of size 'delta'.
 */
2073
static int gather_surplus_pages(struct hstate *h, int delta)
2074
	__must_hold(&hugetlb_lock)
2075 2076 2077 2078 2079
{
	struct list_head surplus_list;
	struct page *page, *tmp;
	int ret, i;
	int needed, allocated;
2080
	bool alloc_ok = true;
2081

2082
	lockdep_assert_held(&hugetlb_lock);
2083
	needed = (h->resv_huge_pages + delta) - h->free_huge_pages;
2084
	if (needed <= 0) {
2085
		h->resv_huge_pages += delta;
2086
		return 0;
2087
	}
2088 2089 2090 2091 2092 2093

	allocated = 0;
	INIT_LIST_HEAD(&surplus_list);

	ret = -ENOMEM;
retry:
2094
	spin_unlock_irq(&hugetlb_lock);
2095
	for (i = 0; i < needed; i++) {
2096
		page = alloc_surplus_huge_page(h, htlb_alloc_mask(h),
2097
				NUMA_NO_NODE, NULL);
2098 2099 2100 2101
		if (!page) {
			alloc_ok = false;
			break;
		}
2102
		list_add(&page->lru, &surplus_list);
2103
		cond_resched();
2104
	}
2105
	allocated += i;
2106 2107 2108 2109 2110

	/*
	 * After retaking hugetlb_lock, we need to recalculate 'needed'
	 * because either resv_huge_pages or free_huge_pages may have changed.
	 */
2111
	spin_lock_irq(&hugetlb_lock);
2112 2113
	needed = (h->resv_huge_pages + delta) -
			(h->free_huge_pages + allocated);
2114 2115 2116 2117 2118 2119 2120 2121 2122 2123
	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;
	}
2124 2125
	/*
	 * The surplus_list now contains _at_least_ the number of extra pages
L
Lucas De Marchi 已提交
2126
	 * needed to accommodate the reservation.  Add the appropriate number
2127
	 * of pages to the hugetlb pool and free the extras back to the buddy
2128 2129 2130
	 * 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.
2131 2132
	 */
	needed += allocated;
2133
	h->resv_huge_pages += delta;
2134
	ret = 0;
2135

2136
	/* Free the needed pages to the hugetlb pool */
2137
	list_for_each_entry_safe(page, tmp, &surplus_list, lru) {
2138 2139
		if ((--needed) < 0)
			break;
2140 2141 2142 2143 2144
		/*
		 * This page is now managed by the hugetlb allocator and has
		 * no users -- drop the buddy allocator's reference.
		 */
		put_page_testzero(page);
2145
		VM_BUG_ON_PAGE(page_count(page), page);
2146
		enqueue_huge_page(h, page);
2147
	}
2148
free:
2149
	spin_unlock_irq(&hugetlb_lock);
2150 2151

	/* Free unnecessary surplus pages to the buddy allocator */
2152 2153
	list_for_each_entry_safe(page, tmp, &surplus_list, lru)
		put_page(page);
2154
	spin_lock_irq(&hugetlb_lock);
2155 2156 2157 2158 2159

	return ret;
}

/*
2160 2161 2162 2163 2164 2165
 * 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.
2166
 */
2167 2168
static void return_unused_surplus_pages(struct hstate *h,
					unsigned long unused_resv_pages)
2169 2170
{
	unsigned long nr_pages;
2171 2172 2173
	struct page *page;
	LIST_HEAD(page_list);

2174
	lockdep_assert_held(&hugetlb_lock);
2175 2176
	/* Uncommit the reservation */
	h->resv_huge_pages -= unused_resv_pages;
2177

2178
	/* Cannot return gigantic pages currently */
2179
	if (hstate_is_gigantic(h))
2180
		goto out;
2181

2182 2183 2184 2185
	/*
	 * Part (or even all) of the reservation could have been backed
	 * by pre-allocated pages. Only free surplus pages.
	 */
2186
	nr_pages = min(unused_resv_pages, h->surplus_huge_pages);
2187

2188 2189
	/*
	 * We want to release as many surplus pages as possible, spread
2190 2191 2192
	 * 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.
2193
	 * remove_pool_huge_page() will balance the freed pages across the
2194
	 * on-line nodes with memory and will handle the hstate accounting.
2195 2196
	 */
	while (nr_pages--) {
2197 2198
		page = remove_pool_huge_page(h, &node_states[N_MEMORY], 1);
		if (!page)
2199
			goto out;
2200 2201

		list_add(&page->lru, &page_list);
2202
	}
2203 2204

out:
2205
	spin_unlock_irq(&hugetlb_lock);
2206
	update_and_free_pages_bulk(h, &page_list);
2207
	spin_lock_irq(&hugetlb_lock);
2208 2209
}

2210

2211
/*
2212
 * vma_needs_reservation, vma_commit_reservation and vma_end_reservation
2213
 * are used by the huge page allocation routines to manage reservations.
2214 2215 2216 2217 2218 2219
 *
 * 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
2220 2221 2222
 * 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.
2223 2224 2225 2226 2227 2228
 *
 * 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.
2229 2230 2231 2232 2233
 *
 * 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.
2234
 */
2235 2236 2237
enum vma_resv_mode {
	VMA_NEEDS_RESV,
	VMA_COMMIT_RESV,
2238
	VMA_END_RESV,
2239
	VMA_ADD_RESV,
2240
};
2241 2242
static long __vma_reservation_common(struct hstate *h,
				struct vm_area_struct *vma, unsigned long addr,
2243
				enum vma_resv_mode mode)
2244
{
2245 2246
	struct resv_map *resv;
	pgoff_t idx;
2247
	long ret;
2248
	long dummy_out_regions_needed;
2249

2250 2251
	resv = vma_resv_map(vma);
	if (!resv)
2252
		return 1;
2253

2254
	idx = vma_hugecache_offset(h, vma, addr);
2255 2256
	switch (mode) {
	case VMA_NEEDS_RESV:
2257 2258 2259 2260 2261 2262
		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);
2263 2264
		break;
	case VMA_COMMIT_RESV:
2265
		ret = region_add(resv, idx, idx + 1, 1, NULL, NULL);
2266 2267
		/* region_add calls of range 1 should never fail. */
		VM_BUG_ON(ret < 0);
2268
		break;
2269
	case VMA_END_RESV:
2270
		region_abort(resv, idx, idx + 1, 1);
2271 2272
		ret = 0;
		break;
2273
	case VMA_ADD_RESV:
2274
		if (vma->vm_flags & VM_MAYSHARE) {
2275
			ret = region_add(resv, idx, idx + 1, 1, NULL, NULL);
2276 2277 2278 2279
			/* region_add calls of range 1 should never fail. */
			VM_BUG_ON(ret < 0);
		} else {
			region_abort(resv, idx, idx + 1, 1);
2280 2281 2282
			ret = region_del(resv, idx, idx + 1);
		}
		break;
2283 2284 2285
	default:
		BUG();
	}
2286

2287
	if (vma->vm_flags & VM_MAYSHARE)
2288
		return ret;
2289 2290 2291 2292 2293 2294 2295 2296 2297 2298 2299 2300 2301 2302 2303 2304 2305 2306 2307
	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;
	}
2308
	else
2309
		return ret < 0 ? ret : 0;
2310
}
2311 2312

static long vma_needs_reservation(struct hstate *h,
2313
			struct vm_area_struct *vma, unsigned long addr)
2314
{
2315
	return __vma_reservation_common(h, vma, addr, VMA_NEEDS_RESV);
2316
}
2317

2318 2319 2320
static long vma_commit_reservation(struct hstate *h,
			struct vm_area_struct *vma, unsigned long addr)
{
2321 2322 2323
	return __vma_reservation_common(h, vma, addr, VMA_COMMIT_RESV);
}

2324
static void vma_end_reservation(struct hstate *h,
2325 2326
			struct vm_area_struct *vma, unsigned long addr)
{
2327
	(void)__vma_reservation_common(h, vma, addr, VMA_END_RESV);
2328 2329
}

2330 2331 2332 2333 2334 2335 2336 2337 2338 2339
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,
2340 2341 2342 2343 2344 2345
 * alloc_huge_page would have consumed the reservation and set
 * HPageRestoreReserve in the newly allocated page.  When the page is freed
 * via free_huge_page, the global reservation count will be incremented if
 * HPageRestoreReserve is set.  However, free_huge_page can not adjust the
 * reserve map.  Adjust the reserve map here to be consistent with global
 * reserve count adjustments to be made by free_huge_page.
2346 2347 2348 2349 2350
 */
static void restore_reserve_on_error(struct hstate *h,
			struct vm_area_struct *vma, unsigned long address,
			struct page *page)
{
2351
	if (unlikely(HPageRestoreReserve(page))) {
2352 2353 2354 2355 2356
		long rc = vma_needs_reservation(h, vma, address);

		if (unlikely(rc < 0)) {
			/*
			 * Rare out of memory condition in reserve map
2357
			 * manipulation.  Clear HPageRestoreReserve so that
2358 2359 2360 2361 2362 2363 2364 2365
			 * 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.
			 */
2366
			ClearHPageRestoreReserve(page);
2367 2368 2369 2370 2371 2372 2373
		} else if (rc) {
			rc = vma_add_reservation(h, vma, address);
			if (unlikely(rc < 0))
				/*
				 * See above comment about rare out of
				 * memory condition.
				 */
2374
				ClearHPageRestoreReserve(page);
2375 2376 2377 2378 2379
		} else
			vma_end_reservation(h, vma, address);
	}
}

2380
struct page *alloc_huge_page(struct vm_area_struct *vma,
2381
				    unsigned long addr, int avoid_reserve)
L
Linus Torvalds 已提交
2382
{
2383
	struct hugepage_subpool *spool = subpool_vma(vma);
2384
	struct hstate *h = hstate_vma(vma);
2385
	struct page *page;
2386 2387
	long map_chg, map_commit;
	long gbl_chg;
2388 2389
	int ret, idx;
	struct hugetlb_cgroup *h_cg;
2390
	bool deferred_reserve;
2391

2392
	idx = hstate_index(h);
2393
	/*
2394 2395 2396
	 * 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).
2397
	 */
2398 2399
	map_chg = gbl_chg = vma_needs_reservation(h, vma, addr);
	if (map_chg < 0)
2400
		return ERR_PTR(-ENOMEM);
2401 2402 2403 2404 2405 2406 2407 2408 2409 2410 2411

	/*
	 * 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) {
2412
			vma_end_reservation(h, vma, addr);
2413
			return ERR_PTR(-ENOSPC);
2414
		}
L
Linus Torvalds 已提交
2415

2416 2417 2418 2419 2420 2421 2422 2423 2424 2425 2426 2427
		/*
		 * 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;
	}

2428 2429 2430 2431 2432 2433 2434 2435 2436 2437
	/* 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;
	}

2438
	ret = hugetlb_cgroup_charge_cgroup(idx, pages_per_huge_page(h), &h_cg);
2439
	if (ret)
2440
		goto out_uncharge_cgroup_reservation;
2441

2442
	spin_lock_irq(&hugetlb_lock);
2443 2444 2445 2446 2447 2448
	/*
	 * 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);
2449
	if (!page) {
2450
		spin_unlock_irq(&hugetlb_lock);
2451
		page = alloc_buddy_huge_page_with_mpol(h, vma, addr);
2452 2453
		if (!page)
			goto out_uncharge_cgroup;
2454
		if (!avoid_reserve && vma_has_reserves(vma, gbl_chg)) {
2455
			SetHPageRestoreReserve(page);
2456 2457
			h->resv_huge_pages--;
		}
2458
		spin_lock_irq(&hugetlb_lock);
2459
		list_add(&page->lru, &h->hugepage_activelist);
2460
		/* Fall through */
K
Ken Chen 已提交
2461
	}
2462
	hugetlb_cgroup_commit_charge(idx, pages_per_huge_page(h), h_cg, page);
2463 2464 2465 2466 2467 2468 2469 2470
	/* 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);
	}

2471
	spin_unlock_irq(&hugetlb_lock);
2472

2473
	hugetlb_set_page_subpool(page, spool);
2474

2475 2476
	map_commit = vma_commit_reservation(h, vma, addr);
	if (unlikely(map_chg > map_commit)) {
2477 2478 2479 2480 2481 2482 2483 2484 2485 2486 2487 2488 2489
		/*
		 * 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);
2490 2491 2492
		if (deferred_reserve)
			hugetlb_cgroup_uncharge_page_rsvd(hstate_index(h),
					pages_per_huge_page(h), page);
2493
	}
2494
	return page;
2495 2496 2497

out_uncharge_cgroup:
	hugetlb_cgroup_uncharge_cgroup(idx, pages_per_huge_page(h), h_cg);
2498 2499 2500 2501
out_uncharge_cgroup_reservation:
	if (deferred_reserve)
		hugetlb_cgroup_uncharge_cgroup_rsvd(idx, pages_per_huge_page(h),
						    h_cg);
2502
out_subpool_put:
2503
	if (map_chg || avoid_reserve)
2504
		hugepage_subpool_put_pages(spool, 1);
2505
	vma_end_reservation(h, vma, addr);
2506
	return ERR_PTR(-ENOSPC);
2507 2508
}

2509 2510 2511
int alloc_bootmem_huge_page(struct hstate *h)
	__attribute__ ((weak, alias("__alloc_bootmem_huge_page")));
int __alloc_bootmem_huge_page(struct hstate *h)
2512 2513
{
	struct huge_bootmem_page *m;
2514
	int nr_nodes, node;
2515

2516
	for_each_node_mask_to_alloc(h, nr_nodes, node, &node_states[N_MEMORY]) {
2517 2518
		void *addr;

2519
		addr = memblock_alloc_try_nid_raw(
2520
				huge_page_size(h), huge_page_size(h),
2521
				0, MEMBLOCK_ALLOC_ACCESSIBLE, node);
2522 2523 2524 2525 2526 2527 2528
		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;
2529
			goto found;
2530 2531 2532 2533 2534
		}
	}
	return 0;

found:
2535
	BUG_ON(!IS_ALIGNED(virt_to_phys(m), huge_page_size(h)));
2536
	/* Put them into a private list first because mem_map is not up yet */
2537
	INIT_LIST_HEAD(&m->list);
2538 2539 2540 2541 2542
	list_add(&m->list, &huge_boot_pages);
	m->hstate = h;
	return 1;
}

2543 2544
static void __init prep_compound_huge_page(struct page *page,
		unsigned int order)
2545 2546 2547 2548 2549 2550 2551
{
	if (unlikely(order > (MAX_ORDER - 1)))
		prep_compound_gigantic_page(page, order);
	else
		prep_compound_page(page, order);
}

2552 2553 2554 2555 2556 2557
/* 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) {
2558
		struct page *page = virt_to_page(m);
2559
		struct hstate *h = m->hstate;
2560

2561
		WARN_ON(page_count(page) != 1);
2562
		prep_compound_huge_page(page, h->order);
2563
		WARN_ON(PageReserved(page));
2564
		prep_new_huge_page(h, page, page_to_nid(page));
2565 2566
		put_page(page); /* free it into the hugepage allocator */

2567 2568 2569 2570 2571 2572
		/*
		 * 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.
		 */
2573
		if (hstate_is_gigantic(h))
2574
			adjust_managed_page_count(page, 1 << h->order);
2575
		cond_resched();
2576 2577 2578
	}
}

2579
static void __init hugetlb_hstate_alloc_pages(struct hstate *h)
L
Linus Torvalds 已提交
2580 2581
{
	unsigned long i;
2582 2583 2584 2585 2586 2587 2588 2589 2590 2591 2592 2593 2594 2595 2596 2597 2598 2599 2600
	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);
2601

2602
	for (i = 0; i < h->max_huge_pages; ++i) {
2603
		if (hstate_is_gigantic(h)) {
2604
			if (hugetlb_cma_size) {
2605
				pr_warn_once("HugeTLB: hugetlb_cma is enabled, skip boot time allocation\n");
2606
				goto free;
2607
			}
2608 2609
			if (!alloc_bootmem_huge_page(h))
				break;
2610
		} else if (!alloc_pool_huge_page(h,
2611 2612
					 &node_states[N_MEMORY],
					 node_alloc_noretry))
L
Linus Torvalds 已提交
2613
			break;
2614
		cond_resched();
L
Linus Torvalds 已提交
2615
	}
2616 2617 2618
	if (i < h->max_huge_pages) {
		char buf[32];

2619
		string_get_size(huge_page_size(h), 1, STRING_UNITS_2, buf, 32);
2620 2621 2622 2623
		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;
	}
2624
free:
2625
	kfree(node_alloc_noretry);
2626 2627 2628 2629 2630 2631 2632
}

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

	for_each_hstate(h) {
2633 2634 2635
		if (minimum_order > huge_page_order(h))
			minimum_order = huge_page_order(h);

2636
		/* oversize hugepages were init'ed in early boot */
2637
		if (!hstate_is_gigantic(h))
2638
			hugetlb_hstate_alloc_pages(h);
2639
	}
2640
	VM_BUG_ON(minimum_order == UINT_MAX);
2641 2642 2643 2644 2645 2646 2647
}

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

	for_each_hstate(h) {
A
Andi Kleen 已提交
2648
		char buf[32];
2649 2650

		string_get_size(huge_page_size(h), 1, STRING_UNITS_2, buf, 32);
2651
		pr_info("HugeTLB registered %s page size, pre-allocated %ld pages\n",
2652
			buf, h->free_huge_pages);
2653 2654 2655
	}
}

L
Linus Torvalds 已提交
2656
#ifdef CONFIG_HIGHMEM
2657 2658
static void try_to_free_low(struct hstate *h, unsigned long count,
						nodemask_t *nodes_allowed)
L
Linus Torvalds 已提交
2659
{
2660
	int i;
2661
	LIST_HEAD(page_list);
2662

2663
	lockdep_assert_held(&hugetlb_lock);
2664
	if (hstate_is_gigantic(h))
2665 2666
		return;

2667 2668 2669
	/*
	 * Collect pages to be freed on a list, and free after dropping lock
	 */
2670
	for_each_node_mask(i, *nodes_allowed) {
2671
		struct page *page, *next;
2672 2673 2674
		struct list_head *freel = &h->hugepage_freelists[i];
		list_for_each_entry_safe(page, next, freel, lru) {
			if (count >= h->nr_huge_pages)
2675
				goto out;
L
Linus Torvalds 已提交
2676 2677
			if (PageHighMem(page))
				continue;
2678
			remove_hugetlb_page(h, page, false);
2679
			list_add(&page->lru, &page_list);
L
Linus Torvalds 已提交
2680 2681
		}
	}
2682 2683

out:
2684
	spin_unlock_irq(&hugetlb_lock);
2685
	update_and_free_pages_bulk(h, &page_list);
2686
	spin_lock_irq(&hugetlb_lock);
L
Linus Torvalds 已提交
2687 2688
}
#else
2689 2690
static inline void try_to_free_low(struct hstate *h, unsigned long count,
						nodemask_t *nodes_allowed)
L
Linus Torvalds 已提交
2691 2692 2693 2694
{
}
#endif

2695 2696 2697 2698 2699
/*
 * 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.
 */
2700 2701
static int adjust_pool_surplus(struct hstate *h, nodemask_t *nodes_allowed,
				int delta)
2702
{
2703
	int nr_nodes, node;
2704

2705
	lockdep_assert_held(&hugetlb_lock);
2706 2707
	VM_BUG_ON(delta != -1 && delta != 1);

2708 2709 2710 2711
	if (delta < 0) {
		for_each_node_mask_to_alloc(h, nr_nodes, node, nodes_allowed) {
			if (h->surplus_huge_pages_node[node])
				goto found;
2712
		}
2713 2714 2715 2716 2717
	} 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;
2718
		}
2719 2720
	}
	return 0;
2721

2722 2723 2724 2725
found:
	h->surplus_huge_pages += delta;
	h->surplus_huge_pages_node[node] += delta;
	return 1;
2726 2727
}

2728
#define persistent_huge_pages(h) (h->nr_huge_pages - h->surplus_huge_pages)
2729
static int set_max_huge_pages(struct hstate *h, unsigned long count, int nid,
2730
			      nodemask_t *nodes_allowed)
L
Linus Torvalds 已提交
2731
{
2732
	unsigned long min_count, ret;
2733 2734
	struct page *page;
	LIST_HEAD(page_list);
2735 2736 2737 2738 2739 2740 2741 2742 2743 2744 2745
	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
2746

2747 2748 2749 2750 2751
	/*
	 * resize_lock mutex prevents concurrent adjustments to number of
	 * pages in hstate via the proc/sysfs interfaces.
	 */
	mutex_lock(&h->resize_lock);
2752
	flush_free_hpage_work(h);
2753
	spin_lock_irq(&hugetlb_lock);
2754

2755 2756 2757 2758 2759 2760 2761 2762 2763 2764 2765 2766 2767 2768 2769 2770 2771 2772 2773 2774
	/*
	 * 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;
	}

2775 2776 2777 2778 2779 2780 2781 2782 2783
	/*
	 * 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)) {
2784
			spin_unlock_irq(&hugetlb_lock);
2785
			mutex_unlock(&h->resize_lock);
2786
			NODEMASK_FREE(node_alloc_noretry);
2787 2788 2789 2790
			return -EINVAL;
		}
		/* Fall through to decrease pool */
	}
2791

2792 2793 2794 2795
	/*
	 * Increase the pool size
	 * First take pages out of surplus state.  Then make up the
	 * remaining difference by allocating fresh huge pages.
2796
	 *
2797
	 * We might race with alloc_surplus_huge_page() here and be unable
2798 2799 2800 2801
	 * 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.
2802
	 */
2803
	while (h->surplus_huge_pages && count > persistent_huge_pages(h)) {
2804
		if (!adjust_pool_surplus(h, nodes_allowed, -1))
2805 2806 2807
			break;
	}

2808
	while (count > persistent_huge_pages(h)) {
2809 2810 2811 2812 2813
		/*
		 * 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.
		 */
2814
		spin_unlock_irq(&hugetlb_lock);
2815 2816 2817 2818

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

2819 2820
		ret = alloc_pool_huge_page(h, nodes_allowed,
						node_alloc_noretry);
2821
		spin_lock_irq(&hugetlb_lock);
2822 2823 2824
		if (!ret)
			goto out;

2825 2826 2827
		/* Bail for signals. Probably ctrl-c from user */
		if (signal_pending(current))
			goto out;
2828 2829 2830 2831 2832 2833 2834 2835
	}

	/*
	 * 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.
2836 2837 2838 2839
	 *
	 * 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
2840
	 * alloc_surplus_huge_page() is checking the global counter,
2841 2842 2843
	 * 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.
2844
	 */
2845
	min_count = h->resv_huge_pages + h->nr_huge_pages - h->free_huge_pages;
2846
	min_count = max(count, min_count);
2847
	try_to_free_low(h, min_count, nodes_allowed);
2848 2849 2850 2851

	/*
	 * Collect pages to be removed on list without dropping lock
	 */
2852
	while (min_count < persistent_huge_pages(h)) {
2853 2854
		page = remove_pool_huge_page(h, nodes_allowed, 0);
		if (!page)
L
Linus Torvalds 已提交
2855
			break;
2856 2857

		list_add(&page->lru, &page_list);
L
Linus Torvalds 已提交
2858
	}
2859
	/* free the pages after dropping lock */
2860
	spin_unlock_irq(&hugetlb_lock);
2861
	update_and_free_pages_bulk(h, &page_list);
2862
	flush_free_hpage_work(h);
2863
	spin_lock_irq(&hugetlb_lock);
2864

2865
	while (count < persistent_huge_pages(h)) {
2866
		if (!adjust_pool_surplus(h, nodes_allowed, 1))
2867 2868 2869
			break;
	}
out:
2870
	h->max_huge_pages = persistent_huge_pages(h);
2871
	spin_unlock_irq(&hugetlb_lock);
2872
	mutex_unlock(&h->resize_lock);
2873

2874 2875
	NODEMASK_FREE(node_alloc_noretry);

2876
	return 0;
L
Linus Torvalds 已提交
2877 2878
}

2879 2880 2881 2882 2883 2884 2885 2886 2887 2888
#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];

2889 2890 2891
static struct hstate *kobj_to_node_hstate(struct kobject *kobj, int *nidp);

static struct hstate *kobj_to_hstate(struct kobject *kobj, int *nidp)
2892 2893
{
	int i;
2894

2895
	for (i = 0; i < HUGE_MAX_HSTATE; i++)
2896 2897 2898
		if (hstate_kobjs[i] == kobj) {
			if (nidp)
				*nidp = NUMA_NO_NODE;
2899
			return &hstates[i];
2900 2901 2902
		}

	return kobj_to_node_hstate(kobj, nidp);
2903 2904
}

2905
static ssize_t nr_hugepages_show_common(struct kobject *kobj,
2906 2907
					struct kobj_attribute *attr, char *buf)
{
2908 2909 2910 2911 2912 2913 2914 2915 2916 2917 2918
	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);
2919
}
2920

2921 2922 2923
static ssize_t __nr_hugepages_store_common(bool obey_mempolicy,
					   struct hstate *h, int nid,
					   unsigned long count, size_t len)
2924 2925
{
	int err;
2926
	nodemask_t nodes_allowed, *n_mask;
2927

2928 2929
	if (hstate_is_gigantic(h) && !gigantic_page_runtime_supported())
		return -EINVAL;
2930

2931 2932 2933 2934 2935
	if (nid == NUMA_NO_NODE) {
		/*
		 * global hstate attribute
		 */
		if (!(obey_mempolicy &&
2936 2937 2938 2939 2940
				init_nodemask_of_mempolicy(&nodes_allowed)))
			n_mask = &node_states[N_MEMORY];
		else
			n_mask = &nodes_allowed;
	} else {
2941
		/*
2942 2943
		 * Node specific request.  count adjustment happens in
		 * set_max_huge_pages() after acquiring hugetlb_lock.
2944
		 */
2945 2946
		init_nodemask_of_node(&nodes_allowed, nid);
		n_mask = &nodes_allowed;
2947
	}
2948

2949
	err = set_max_huge_pages(h, count, nid, n_mask);
2950

2951
	return err ? err : len;
2952 2953
}

2954 2955 2956 2957 2958 2959 2960 2961 2962 2963 2964 2965 2966 2967 2968 2969 2970
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);
}

2971 2972 2973 2974 2975 2976 2977 2978 2979
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)
{
2980
	return nr_hugepages_store_common(false, kobj, buf, len);
2981 2982 2983
}
HSTATE_ATTR(nr_hugepages);

2984 2985 2986 2987 2988 2989 2990 2991 2992 2993 2994 2995 2996 2997 2998
#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)
{
2999
	return nr_hugepages_store_common(true, kobj, buf, len);
3000 3001 3002 3003 3004
}
HSTATE_ATTR(nr_hugepages_mempolicy);
#endif


3005 3006 3007
static ssize_t nr_overcommit_hugepages_show(struct kobject *kobj,
					struct kobj_attribute *attr, char *buf)
{
3008
	struct hstate *h = kobj_to_hstate(kobj, NULL);
3009 3010
	return sprintf(buf, "%lu\n", h->nr_overcommit_huge_pages);
}
3011

3012 3013 3014 3015 3016
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;
3017
	struct hstate *h = kobj_to_hstate(kobj, NULL);
3018

3019
	if (hstate_is_gigantic(h))
3020 3021
		return -EINVAL;

3022
	err = kstrtoul(buf, 10, &input);
3023
	if (err)
3024
		return err;
3025

3026
	spin_lock_irq(&hugetlb_lock);
3027
	h->nr_overcommit_huge_pages = input;
3028
	spin_unlock_irq(&hugetlb_lock);
3029 3030 3031 3032 3033 3034 3035 3036

	return count;
}
HSTATE_ATTR(nr_overcommit_hugepages);

static ssize_t free_hugepages_show(struct kobject *kobj,
					struct kobj_attribute *attr, char *buf)
{
3037 3038 3039 3040 3041 3042 3043 3044 3045 3046 3047
	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);
3048 3049 3050 3051 3052 3053
}
HSTATE_ATTR_RO(free_hugepages);

static ssize_t resv_hugepages_show(struct kobject *kobj,
					struct kobj_attribute *attr, char *buf)
{
3054
	struct hstate *h = kobj_to_hstate(kobj, NULL);
3055 3056 3057 3058 3059 3060 3061
	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)
{
3062 3063 3064 3065 3066 3067 3068 3069 3070 3071 3072
	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);
3073 3074 3075 3076 3077 3078 3079 3080 3081
}
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,
3082 3083 3084
#ifdef CONFIG_NUMA
	&nr_hugepages_mempolicy_attr.attr,
#endif
3085 3086 3087
	NULL,
};

3088
static const struct attribute_group hstate_attr_group = {
3089 3090 3091
	.attrs = hstate_attrs,
};

J
Jeff Mahoney 已提交
3092 3093
static int hugetlb_sysfs_add_hstate(struct hstate *h, struct kobject *parent,
				    struct kobject **hstate_kobjs,
3094
				    const struct attribute_group *hstate_attr_group)
3095 3096
{
	int retval;
3097
	int hi = hstate_index(h);
3098

3099 3100
	hstate_kobjs[hi] = kobject_create_and_add(h->name, parent);
	if (!hstate_kobjs[hi])
3101 3102
		return -ENOMEM;

3103
	retval = sysfs_create_group(hstate_kobjs[hi], hstate_attr_group);
3104
	if (retval) {
3105
		kobject_put(hstate_kobjs[hi]);
3106 3107
		hstate_kobjs[hi] = NULL;
	}
3108 3109 3110 3111 3112 3113 3114 3115 3116 3117 3118 3119 3120 3121

	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) {
3122 3123
		err = hugetlb_sysfs_add_hstate(h, hugepages_kobj,
					 hstate_kobjs, &hstate_attr_group);
3124
		if (err)
3125
			pr_err("HugeTLB: Unable to add hstate %s", h->name);
3126 3127 3128
	}
}

3129 3130 3131 3132
#ifdef CONFIG_NUMA

/*
 * node_hstate/s - associate per node hstate attributes, via their kobjects,
3133 3134 3135
 * 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
3136 3137 3138 3139 3140 3141
 * the base kernel, on the hugetlb module.
 */
struct node_hstate {
	struct kobject		*hugepages_kobj;
	struct kobject		*hstate_kobjs[HUGE_MAX_HSTATE];
};
3142
static struct node_hstate node_hstates[MAX_NUMNODES];
3143 3144

/*
3145
 * A subset of global hstate attributes for node devices
3146 3147 3148 3149 3150 3151 3152 3153
 */
static struct attribute *per_node_hstate_attrs[] = {
	&nr_hugepages_attr.attr,
	&free_hugepages_attr.attr,
	&surplus_hugepages_attr.attr,
	NULL,
};

3154
static const struct attribute_group per_node_hstate_attr_group = {
3155 3156 3157 3158
	.attrs = per_node_hstate_attrs,
};

/*
3159
 * kobj_to_node_hstate - lookup global hstate for node device hstate attr kobj.
3160 3161 3162 3163 3164 3165 3166 3167 3168 3169 3170 3171 3172 3173 3174 3175 3176 3177 3178 3179 3180 3181
 * 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;
}

/*
3182
 * Unregister hstate attributes from a single node device.
3183 3184
 * No-op if no hstate attributes attached.
 */
3185
static void hugetlb_unregister_node(struct node *node)
3186 3187
{
	struct hstate *h;
3188
	struct node_hstate *nhs = &node_hstates[node->dev.id];
3189 3190

	if (!nhs->hugepages_kobj)
3191
		return;		/* no hstate attributes */
3192

3193 3194 3195 3196 3197
	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;
3198
		}
3199
	}
3200 3201 3202 3203 3204 3205 3206

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


/*
3207
 * Register hstate attributes for a single node device.
3208 3209
 * No-op if attributes already registered.
 */
3210
static void hugetlb_register_node(struct node *node)
3211 3212
{
	struct hstate *h;
3213
	struct node_hstate *nhs = &node_hstates[node->dev.id];
3214 3215 3216 3217 3218 3219
	int err;

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

	nhs->hugepages_kobj = kobject_create_and_add("hugepages",
3220
							&node->dev.kobj);
3221 3222 3223 3224 3225 3226 3227 3228
	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) {
3229
			pr_err("HugeTLB: Unable to add hstate %s for node %d\n",
3230
				h->name, node->dev.id);
3231 3232 3233 3234 3235 3236 3237
			hugetlb_unregister_node(node);
			break;
		}
	}
}

/*
3238
 * hugetlb init time:  register hstate attributes for all registered node
3239 3240
 * devices of nodes that have memory.  All on-line nodes should have
 * registered their associated device by this time.
3241
 */
3242
static void __init hugetlb_register_all_nodes(void)
3243 3244 3245
{
	int nid;

3246
	for_each_node_state(nid, N_MEMORY) {
3247
		struct node *node = node_devices[nid];
3248
		if (node->dev.id == nid)
3249 3250 3251 3252
			hugetlb_register_node(node);
	}

	/*
3253
	 * Let the node device driver know we're here so it can
3254 3255 3256 3257 3258 3259 3260 3261 3262 3263 3264 3265 3266 3267 3268 3269 3270 3271 3272
	 * [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

3273 3274
static int __init hugetlb_init(void)
{
3275 3276
	int i;

3277 3278 3279
	BUILD_BUG_ON(sizeof_field(struct page, private) * BITS_PER_BYTE <
			__NR_HPAGEFLAGS);

3280 3281 3282
	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");
3283
		return 0;
3284
	}
3285

3286 3287 3288 3289 3290 3291 3292 3293 3294 3295 3296 3297 3298 3299 3300 3301 3302 3303 3304 3305 3306 3307 3308 3309 3310 3311 3312 3313
	/*
	 * 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;
3314
		}
3315
	}
3316

3317
	hugetlb_cma_check();
3318
	hugetlb_init_hstates();
3319
	gather_bootmem_prealloc();
3320 3321 3322
	report_hugepages();

	hugetlb_sysfs_init();
3323
	hugetlb_register_all_nodes();
3324
	hugetlb_cgroup_file_init();
3325

3326 3327 3328 3329 3330
#ifdef CONFIG_SMP
	num_fault_mutexes = roundup_pow_of_two(8 * num_possible_cpus());
#else
	num_fault_mutexes = 1;
#endif
3331
	hugetlb_fault_mutex_table =
3332 3333
		kmalloc_array(num_fault_mutexes, sizeof(struct mutex),
			      GFP_KERNEL);
3334
	BUG_ON(!hugetlb_fault_mutex_table);
3335 3336

	for (i = 0; i < num_fault_mutexes; i++)
3337
		mutex_init(&hugetlb_fault_mutex_table[i]);
3338 3339
	return 0;
}
3340
subsys_initcall(hugetlb_init);
3341

3342 3343
/* Overwritten by architectures with more huge page sizes */
bool __init __attribute((weak)) arch_hugetlb_valid_size(unsigned long size)
3344
{
3345
	return size == HPAGE_SIZE;
3346 3347
}

3348
void __init hugetlb_add_hstate(unsigned int order)
3349 3350
{
	struct hstate *h;
3351 3352
	unsigned long i;

3353 3354 3355
	if (size_to_hstate(PAGE_SIZE << order)) {
		return;
	}
3356
	BUG_ON(hugetlb_max_hstate >= HUGE_MAX_HSTATE);
3357
	BUG_ON(order == 0);
3358
	h = &hstates[hugetlb_max_hstate++];
3359
	mutex_init(&h->resize_lock);
3360 3361
	h->order = order;
	h->mask = ~((1ULL << (order + PAGE_SHIFT)) - 1);
3362 3363 3364 3365
	h->nr_huge_pages = 0;
	h->free_huge_pages = 0;
	for (i = 0; i < MAX_NUMNODES; ++i)
		INIT_LIST_HEAD(&h->hugepage_freelists[i]);
3366
	INIT_LIST_HEAD(&h->hugepage_activelist);
3367 3368
	h->next_nid_to_alloc = first_memory_node;
	h->next_nid_to_free = first_memory_node;
3369 3370
	snprintf(h->name, HSTATE_NAME_LEN, "hugepages-%lukB",
					huge_page_size(h)/1024);
3371
	hugetlb_vmemmap_init(h);
3372

3373 3374 3375
	parsed_hstate = h;
}

3376 3377 3378 3379 3380 3381 3382 3383
/*
 * 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)
3384 3385
{
	unsigned long *mhp;
3386
	static unsigned long *last_mhp;
3387

3388
	if (!parsed_valid_hugepagesz) {
3389
		pr_warn("HugeTLB: hugepages=%s does not follow a valid hugepagesz, ignoring\n", s);
3390
		parsed_valid_hugepagesz = true;
3391
		return 0;
3392
	}
3393

3394
	/*
3395 3396 3397 3398
	 * !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.
3399
	 */
3400
	else if (!hugetlb_max_hstate)
3401 3402 3403 3404
		mhp = &default_hstate_max_huge_pages;
	else
		mhp = &parsed_hstate->max_huge_pages;

3405
	if (mhp == last_mhp) {
3406 3407
		pr_warn("HugeTLB: hugepages= specified twice without interleaving hugepagesz=, ignoring hugepages=%s\n", s);
		return 0;
3408 3409
	}

3410 3411 3412
	if (sscanf(s, "%lu", mhp) <= 0)
		*mhp = 0;

3413 3414 3415 3416 3417
	/*
	 * 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.
	 */
3418
	if (hugetlb_max_hstate && parsed_hstate->order >= MAX_ORDER)
3419 3420 3421 3422
		hugetlb_hstate_alloc_pages(parsed_hstate);

	last_mhp = mhp;

3423 3424
	return 1;
}
3425
__setup("hugepages=", hugepages_setup);
3426

3427 3428 3429 3430 3431 3432 3433
/*
 * 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.
 */
3434
static int __init hugepagesz_setup(char *s)
3435
{
3436
	unsigned long size;
3437 3438 3439
	struct hstate *h;

	parsed_valid_hugepagesz = false;
3440 3441 3442
	size = (unsigned long)memparse(s, NULL);

	if (!arch_hugetlb_valid_size(size)) {
3443
		pr_err("HugeTLB: unsupported hugepagesz=%s\n", s);
3444 3445 3446
		return 0;
	}

3447 3448 3449 3450 3451 3452 3453 3454 3455 3456 3457 3458 3459 3460 3461 3462 3463 3464 3465 3466 3467 3468 3469
	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;
3470 3471
	}

3472
	hugetlb_add_hstate(ilog2(size) - PAGE_SHIFT);
3473
	parsed_valid_hugepagesz = true;
3474 3475
	return 1;
}
3476 3477
__setup("hugepagesz=", hugepagesz_setup);

3478 3479 3480 3481
/*
 * default_hugepagesz command line input
 * Only one instance of default_hugepagesz allowed on command line.
 */
3482
static int __init default_hugepagesz_setup(char *s)
3483
{
3484 3485
	unsigned long size;

3486 3487 3488 3489 3490 3491
	parsed_valid_hugepagesz = false;
	if (parsed_default_hugepagesz) {
		pr_err("HugeTLB: default_hugepagesz previously specified, ignoring %s\n", s);
		return 0;
	}

3492 3493 3494
	size = (unsigned long)memparse(s, NULL);

	if (!arch_hugetlb_valid_size(size)) {
3495
		pr_err("HugeTLB: unsupported default_hugepagesz=%s\n", s);
3496 3497 3498
		return 0;
	}

3499 3500 3501 3502 3503 3504 3505 3506 3507 3508 3509 3510 3511 3512 3513 3514 3515 3516 3517
	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;
	}

3518 3519
	return 1;
}
3520
__setup("default_hugepagesz=", default_hugepagesz_setup);
3521

3522
static unsigned int allowed_mems_nr(struct hstate *h)
3523 3524 3525
{
	int node;
	unsigned int nr = 0;
3526 3527 3528 3529 3530
	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);
3531

3532 3533 3534 3535 3536
	for_each_node_mask(node, cpuset_current_mems_allowed) {
		if (!mpol_allowed ||
		    (mpol_allowed && node_isset(node, *mpol_allowed)))
			nr += array[node];
	}
3537 3538 3539 3540 3541

	return nr;
}

#ifdef CONFIG_SYSCTL
3542 3543 3544 3545 3546 3547 3548 3549 3550 3551 3552 3553 3554 3555 3556 3557
static int proc_hugetlb_doulongvec_minmax(struct ctl_table *table, int write,
					  void *buffer, size_t *length,
					  loff_t *ppos, unsigned long *out)
{
	struct ctl_table dup_table;

	/*
	 * In order to avoid races with __do_proc_doulongvec_minmax(), we
	 * can duplicate the @table and alter the duplicate of it.
	 */
	dup_table = *table;
	dup_table.data = out;

	return proc_doulongvec_minmax(&dup_table, write, buffer, length, ppos);
}

3558 3559
static int hugetlb_sysctl_handler_common(bool obey_mempolicy,
			 struct ctl_table *table, int write,
3560
			 void *buffer, size_t *length, loff_t *ppos)
L
Linus Torvalds 已提交
3561
{
3562
	struct hstate *h = &default_hstate;
3563
	unsigned long tmp = h->max_huge_pages;
3564
	int ret;
3565

3566
	if (!hugepages_supported())
3567
		return -EOPNOTSUPP;
3568

3569 3570
	ret = proc_hugetlb_doulongvec_minmax(table, write, buffer, length, ppos,
					     &tmp);
3571 3572
	if (ret)
		goto out;
3573

3574 3575 3576
	if (write)
		ret = __nr_hugepages_store_common(obey_mempolicy, h,
						  NUMA_NO_NODE, tmp, *length);
3577 3578
out:
	return ret;
L
Linus Torvalds 已提交
3579
}
3580

3581
int hugetlb_sysctl_handler(struct ctl_table *table, int write,
3582
			  void *buffer, size_t *length, loff_t *ppos)
3583 3584 3585 3586 3587 3588 3589 3590
{

	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,
3591
			  void *buffer, size_t *length, loff_t *ppos)
3592 3593 3594 3595 3596 3597
{
	return hugetlb_sysctl_handler_common(true, table, write,
							buffer, length, ppos);
}
#endif /* CONFIG_NUMA */

3598
int hugetlb_overcommit_handler(struct ctl_table *table, int write,
3599
		void *buffer, size_t *length, loff_t *ppos)
3600
{
3601
	struct hstate *h = &default_hstate;
3602
	unsigned long tmp;
3603
	int ret;
3604

3605
	if (!hugepages_supported())
3606
		return -EOPNOTSUPP;
3607

3608
	tmp = h->nr_overcommit_huge_pages;
3609

3610
	if (write && hstate_is_gigantic(h))
3611 3612
		return -EINVAL;

3613 3614
	ret = proc_hugetlb_doulongvec_minmax(table, write, buffer, length, ppos,
					     &tmp);
3615 3616
	if (ret)
		goto out;
3617 3618

	if (write) {
3619
		spin_lock_irq(&hugetlb_lock);
3620
		h->nr_overcommit_huge_pages = tmp;
3621
		spin_unlock_irq(&hugetlb_lock);
3622
	}
3623 3624
out:
	return ret;
3625 3626
}

L
Linus Torvalds 已提交
3627 3628
#endif /* CONFIG_SYSCTL */

3629
void hugetlb_report_meminfo(struct seq_file *m)
L
Linus Torvalds 已提交
3630
{
3631 3632 3633
	struct hstate *h;
	unsigned long total = 0;

3634 3635
	if (!hugepages_supported())
		return;
3636 3637 3638 3639 3640 3641 3642 3643 3644 3645 3646 3647 3648 3649 3650 3651 3652 3653 3654 3655 3656

	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
3657 3658
}

3659
int hugetlb_report_node_meminfo(char *buf, int len, int nid)
L
Linus Torvalds 已提交
3660
{
3661
	struct hstate *h = &default_hstate;
3662

3663 3664
	if (!hugepages_supported())
		return 0;
3665 3666 3667 3668 3669 3670 3671 3672

	return sysfs_emit_at(buf, len,
			     "Node %d HugePages_Total: %5u\n"
			     "Node %d HugePages_Free:  %5u\n"
			     "Node %d HugePages_Surp:  %5u\n",
			     nid, h->nr_huge_pages_node[nid],
			     nid, h->free_huge_pages_node[nid],
			     nid, h->surplus_huge_pages_node[nid]);
L
Linus Torvalds 已提交
3673 3674
}

3675 3676 3677 3678 3679
void hugetlb_show_meminfo(void)
{
	struct hstate *h;
	int nid;

3680 3681 3682
	if (!hugepages_supported())
		return;

3683 3684 3685 3686 3687 3688 3689 3690 3691 3692
	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));
}

3693 3694 3695 3696 3697 3698
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
3699 3700 3701
/* Return the number pages of memory we physically have, in PAGE_SIZE units. */
unsigned long hugetlb_total_pages(void)
{
3702 3703 3704 3705 3706 3707
	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
3708 3709
}

3710
static int hugetlb_acct_memory(struct hstate *h, long delta)
M
Mel Gorman 已提交
3711 3712 3713
{
	int ret = -ENOMEM;

3714
	spin_lock_irq(&hugetlb_lock);
M
Mel Gorman 已提交
3715 3716 3717 3718 3719 3720 3721 3722 3723 3724 3725 3726 3727 3728 3729 3730
	/*
	 * 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.
3731 3732 3733 3734 3735 3736
	 *
	 * 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
3737 3738
	 */
	if (delta > 0) {
3739
		if (gather_surplus_pages(h, delta) < 0)
M
Mel Gorman 已提交
3740 3741
			goto out;

3742
		if (delta > allowed_mems_nr(h)) {
3743
			return_unused_surplus_pages(h, delta);
M
Mel Gorman 已提交
3744 3745 3746 3747 3748 3749
			goto out;
		}
	}

	ret = 0;
	if (delta < 0)
3750
		return_unused_surplus_pages(h, (unsigned long) -delta);
M
Mel Gorman 已提交
3751 3752

out:
3753
	spin_unlock_irq(&hugetlb_lock);
M
Mel Gorman 已提交
3754 3755 3756
	return ret;
}

3757 3758
static void hugetlb_vm_op_open(struct vm_area_struct *vma)
{
3759
	struct resv_map *resv = vma_resv_map(vma);
3760 3761 3762 3763 3764

	/*
	 * 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
3765
	 * has a reference to the reservation map it cannot disappear until
3766 3767 3768
	 * after this open call completes.  It is therefore safe to take a
	 * new reference here without additional locking.
	 */
3769
	if (resv && is_vma_resv_set(vma, HPAGE_RESV_OWNER))
3770
		kref_get(&resv->refs);
3771 3772
}

3773 3774
static void hugetlb_vm_op_close(struct vm_area_struct *vma)
{
3775
	struct hstate *h = hstate_vma(vma);
3776
	struct resv_map *resv = vma_resv_map(vma);
3777
	struct hugepage_subpool *spool = subpool_vma(vma);
3778
	unsigned long reserve, start, end;
3779
	long gbl_reserve;
3780

3781 3782
	if (!resv || !is_vma_resv_set(vma, HPAGE_RESV_OWNER))
		return;
3783

3784 3785
	start = vma_hugecache_offset(h, vma, vma->vm_start);
	end = vma_hugecache_offset(h, vma, vma->vm_end);
3786

3787
	reserve = (end - start) - region_count(resv, start, end);
3788
	hugetlb_cgroup_uncharge_counter(resv, start, end);
3789
	if (reserve) {
3790 3791 3792 3793 3794 3795
		/*
		 * 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);
3796
	}
3797 3798

	kref_put(&resv->refs, resv_map_release);
3799 3800
}

3801 3802 3803 3804 3805 3806 3807
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;
}

3808 3809 3810 3811 3812 3813 3814
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
3815 3816 3817 3818 3819 3820
/*
 * 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.
 */
3821
static vm_fault_t hugetlb_vm_op_fault(struct vm_fault *vmf)
L
Linus Torvalds 已提交
3822 3823
{
	BUG();
N
Nick Piggin 已提交
3824
	return 0;
L
Linus Torvalds 已提交
3825 3826
}

3827 3828 3829 3830 3831 3832 3833
/*
 * 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.
 */
3834
const struct vm_operations_struct hugetlb_vm_ops = {
N
Nick Piggin 已提交
3835
	.fault = hugetlb_vm_op_fault,
3836
	.open = hugetlb_vm_op_open,
3837
	.close = hugetlb_vm_op_close,
3838
	.split = hugetlb_vm_op_split,
3839
	.pagesize = hugetlb_vm_op_pagesize,
L
Linus Torvalds 已提交
3840 3841
};

3842 3843
static pte_t make_huge_pte(struct vm_area_struct *vma, struct page *page,
				int writable)
D
David Gibson 已提交
3844 3845 3846
{
	pte_t entry;

3847
	if (writable) {
3848 3849
		entry = huge_pte_mkwrite(huge_pte_mkdirty(mk_huge_pte(page,
					 vma->vm_page_prot)));
D
David Gibson 已提交
3850
	} else {
3851 3852
		entry = huge_pte_wrprotect(mk_huge_pte(page,
					   vma->vm_page_prot));
D
David Gibson 已提交
3853 3854 3855
	}
	entry = pte_mkyoung(entry);
	entry = pte_mkhuge(entry);
3856
	entry = arch_make_huge_pte(entry, vma, page, writable);
D
David Gibson 已提交
3857 3858 3859 3860

	return entry;
}

3861 3862 3863 3864 3865
static void set_huge_ptep_writable(struct vm_area_struct *vma,
				   unsigned long address, pte_t *ptep)
{
	pte_t entry;

3866
	entry = huge_pte_mkwrite(huge_pte_mkdirty(huge_ptep_get(ptep)));
3867
	if (huge_ptep_set_access_flags(vma, address, ptep, entry, 1))
3868
		update_mmu_cache(vma, address, ptep);
3869 3870
}

3871
bool is_hugetlb_entry_migration(pte_t pte)
3872 3873 3874 3875
{
	swp_entry_t swp;

	if (huge_pte_none(pte) || pte_present(pte))
3876
		return false;
3877
	swp = pte_to_swp_entry(pte);
3878
	if (is_migration_entry(swp))
3879
		return true;
3880
	else
3881
		return false;
3882 3883
}

3884
static bool is_hugetlb_entry_hwpoisoned(pte_t pte)
3885 3886 3887 3888
{
	swp_entry_t swp;

	if (huge_pte_none(pte) || pte_present(pte))
3889
		return false;
3890
	swp = pte_to_swp_entry(pte);
3891
	if (is_hwpoison_entry(swp))
3892
		return true;
3893
	else
3894
		return false;
3895
}
3896

D
David Gibson 已提交
3897 3898 3899
int copy_hugetlb_page_range(struct mm_struct *dst, struct mm_struct *src,
			    struct vm_area_struct *vma)
{
3900
	pte_t *src_pte, *dst_pte, entry, dst_entry;
D
David Gibson 已提交
3901
	struct page *ptepage;
3902
	unsigned long addr;
3903
	int cow;
3904 3905
	struct hstate *h = hstate_vma(vma);
	unsigned long sz = huge_page_size(h);
3906
	struct address_space *mapping = vma->vm_file->f_mapping;
3907
	struct mmu_notifier_range range;
3908
	int ret = 0;
3909 3910

	cow = (vma->vm_flags & (VM_SHARED | VM_MAYWRITE)) == VM_MAYWRITE;
D
David Gibson 已提交
3911

3912
	if (cow) {
3913
		mmu_notifier_range_init(&range, MMU_NOTIFY_CLEAR, 0, vma, src,
3914
					vma->vm_start,
3915 3916
					vma->vm_end);
		mmu_notifier_invalidate_range_start(&range);
3917 3918 3919 3920 3921 3922 3923 3924
	} 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);
3925
	}
3926

3927
	for (addr = vma->vm_start; addr < vma->vm_end; addr += sz) {
3928
		spinlock_t *src_ptl, *dst_ptl;
3929
		src_pte = huge_pte_offset(src, addr, sz);
H
Hugh Dickins 已提交
3930 3931
		if (!src_pte)
			continue;
3932
		dst_pte = huge_pte_alloc(dst, addr, sz);
3933 3934 3935 3936
		if (!dst_pte) {
			ret = -ENOMEM;
			break;
		}
3937

3938 3939 3940 3941 3942 3943 3944 3945 3946 3947 3948
		/*
		 * 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))
3949 3950
			continue;

3951 3952 3953
		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);
3954
		entry = huge_ptep_get(src_pte);
3955 3956 3957 3958 3959 3960 3961
		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.
			 */
3962 3963 3964 3965 3966 3967 3968 3969 3970 3971 3972 3973
			;
		} 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);
3974 3975
				set_huge_swap_pte_at(src, addr, src_pte,
						     entry, sz);
3976
			}
3977
			set_huge_swap_pte_at(dst, addr, dst_pte, entry, sz);
3978
		} else {
3979
			if (cow) {
3980 3981 3982 3983 3984
				/*
				 * No need to notify as we are downgrading page
				 * table protection not changing it to point
				 * to a new page.
				 *
3985
				 * See Documentation/vm/mmu_notifier.rst
3986
				 */
3987
				huge_ptep_set_wrprotect(src, addr, src_pte);
3988
			}
3989
			entry = huge_ptep_get(src_pte);
3990 3991
			ptepage = pte_page(entry);
			get_page(ptepage);
3992
			page_dup_rmap(ptepage, true);
3993
			set_huge_pte_at(dst, addr, dst_pte, entry);
3994
			hugetlb_count_add(pages_per_huge_page(h), dst);
3995
		}
3996 3997
		spin_unlock(src_ptl);
		spin_unlock(dst_ptl);
D
David Gibson 已提交
3998 3999
	}

4000
	if (cow)
4001
		mmu_notifier_invalidate_range_end(&range);
4002 4003
	else
		i_mmap_unlock_read(mapping);
4004 4005

	return ret;
D
David Gibson 已提交
4006 4007
}

4008 4009 4010
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
4011 4012 4013
{
	struct mm_struct *mm = vma->vm_mm;
	unsigned long address;
4014
	pte_t *ptep;
D
David Gibson 已提交
4015
	pte_t pte;
4016
	spinlock_t *ptl;
D
David Gibson 已提交
4017
	struct page *page;
4018 4019
	struct hstate *h = hstate_vma(vma);
	unsigned long sz = huge_page_size(h);
4020
	struct mmu_notifier_range range;
4021

D
David Gibson 已提交
4022
	WARN_ON(!is_vm_hugetlb_page(vma));
4023 4024
	BUG_ON(start & ~huge_page_mask(h));
	BUG_ON(end & ~huge_page_mask(h));
D
David Gibson 已提交
4025

4026 4027 4028 4029
	/*
	 * This is a hugetlb vma, all the pte entries should point
	 * to huge page.
	 */
4030
	tlb_change_page_size(tlb, sz);
4031
	tlb_start_vma(tlb, vma);
4032 4033 4034 4035

	/*
	 * If sharing possible, alert mmu notifiers of worst case.
	 */
4036 4037
	mmu_notifier_range_init(&range, MMU_NOTIFY_UNMAP, 0, vma, mm, start,
				end);
4038 4039
	adjust_range_if_pmd_sharing_possible(vma, &range.start, &range.end);
	mmu_notifier_invalidate_range_start(&range);
4040 4041
	address = start;
	for (; address < end; address += sz) {
4042
		ptep = huge_pte_offset(mm, address, sz);
A
Adam Litke 已提交
4043
		if (!ptep)
4044 4045
			continue;

4046
		ptl = huge_pte_lock(h, mm, ptep);
4047
		if (huge_pmd_unshare(mm, vma, &address, ptep)) {
4048
			spin_unlock(ptl);
4049 4050 4051 4052
			/*
			 * We just unmapped a page of PMDs by clearing a PUD.
			 * The caller's TLB flush range should cover this area.
			 */
4053 4054
			continue;
		}
4055

4056
		pte = huge_ptep_get(ptep);
4057 4058 4059 4060
		if (huge_pte_none(pte)) {
			spin_unlock(ptl);
			continue;
		}
4061 4062

		/*
4063 4064
		 * Migrating hugepage or HWPoisoned hugepage is already
		 * unmapped and its refcount is dropped, so just clear pte here.
4065
		 */
4066
		if (unlikely(!pte_present(pte))) {
4067
			huge_pte_clear(mm, address, ptep, sz);
4068 4069
			spin_unlock(ptl);
			continue;
4070
		}
4071 4072

		page = pte_page(pte);
4073 4074 4075 4076 4077 4078
		/*
		 * 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) {
4079 4080 4081 4082
			if (page != ref_page) {
				spin_unlock(ptl);
				continue;
			}
4083 4084 4085 4086 4087 4088 4089 4090
			/*
			 * 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);
		}

4091
		pte = huge_ptep_get_and_clear(mm, address, ptep);
4092
		tlb_remove_huge_tlb_entry(h, tlb, ptep, address);
4093
		if (huge_pte_dirty(pte))
4094
			set_page_dirty(page);
4095

4096
		hugetlb_count_sub(pages_per_huge_page(h), mm);
4097
		page_remove_rmap(page, true);
4098

4099
		spin_unlock(ptl);
4100
		tlb_remove_page_size(tlb, page, huge_page_size(h));
4101 4102 4103 4104 4105
		/*
		 * Bail out after unmapping reference page if supplied
		 */
		if (ref_page)
			break;
4106
	}
4107
	mmu_notifier_invalidate_range_end(&range);
4108
	tlb_end_vma(tlb, vma);
L
Linus Torvalds 已提交
4109
}
D
David Gibson 已提交
4110

4111 4112 4113 4114 4115 4116 4117 4118 4119 4120 4121 4122
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
4123
	 * is to clear it before releasing the i_mmap_rwsem. This works
4124
	 * because in the context this is called, the VMA is about to be
4125
	 * destroyed and the i_mmap_rwsem is held.
4126 4127 4128 4129
	 */
	vma->vm_flags &= ~VM_MAYSHARE;
}

4130
void unmap_hugepage_range(struct vm_area_struct *vma, unsigned long start,
4131
			  unsigned long end, struct page *ref_page)
4132
{
4133 4134
	struct mm_struct *mm;
	struct mmu_gather tlb;
4135 4136 4137 4138 4139 4140 4141 4142 4143 4144 4145
	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);
4146 4147 4148

	mm = vma->vm_mm;

4149
	tlb_gather_mmu(&tlb, mm, tlb_start, tlb_end);
4150
	__unmap_hugepage_range(&tlb, vma, start, end, ref_page);
4151
	tlb_finish_mmu(&tlb, tlb_start, tlb_end);
4152 4153
}

4154 4155 4156 4157 4158 4159
/*
 * 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.
 */
4160 4161
static void unmap_ref_private(struct mm_struct *mm, struct vm_area_struct *vma,
			      struct page *page, unsigned long address)
4162
{
4163
	struct hstate *h = hstate_vma(vma);
4164 4165 4166 4167 4168 4169 4170 4171
	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.
	 */
4172
	address = address & huge_page_mask(h);
4173 4174
	pgoff = ((address - vma->vm_start) >> PAGE_SHIFT) +
			vma->vm_pgoff;
4175
	mapping = vma->vm_file->f_mapping;
4176

4177 4178 4179 4180 4181
	/*
	 * 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
	 */
4182
	i_mmap_lock_write(mapping);
4183
	vma_interval_tree_foreach(iter_vma, &mapping->i_mmap, pgoff, pgoff) {
4184 4185 4186 4187
		/* Do not unmap the current VMA */
		if (iter_vma == vma)
			continue;

4188 4189 4190 4191 4192 4193 4194 4195
		/*
		 * 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;

4196 4197 4198 4199 4200 4201 4202 4203
		/*
		 * 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))
4204 4205
			unmap_hugepage_range(iter_vma, address,
					     address + huge_page_size(h), page);
4206
	}
4207
	i_mmap_unlock_write(mapping);
4208 4209
}

4210 4211
/*
 * Hugetlb_cow() should be called with page lock of the original hugepage held.
4212 4213 4214
 * 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.
4215
 */
4216
static vm_fault_t hugetlb_cow(struct mm_struct *mm, struct vm_area_struct *vma,
4217
		       unsigned long address, pte_t *ptep,
4218
		       struct page *pagecache_page, spinlock_t *ptl)
4219
{
4220
	pte_t pte;
4221
	struct hstate *h = hstate_vma(vma);
4222
	struct page *old_page, *new_page;
4223 4224
	int outside_reserve = 0;
	vm_fault_t ret = 0;
4225
	unsigned long haddr = address & huge_page_mask(h);
4226
	struct mmu_notifier_range range;
4227

4228
	pte = huge_ptep_get(ptep);
4229 4230
	old_page = pte_page(pte);

4231
retry_avoidcopy:
4232 4233
	/* If no-one else is actually using this page, avoid the copy
	 * and just make the page writable */
4234
	if (page_mapcount(old_page) == 1 && PageAnon(old_page)) {
4235
		page_move_anon_rmap(old_page, vma);
4236
		set_huge_ptep_writable(vma, haddr, ptep);
N
Nick Piggin 已提交
4237
		return 0;
4238 4239
	}

4240 4241 4242 4243 4244 4245 4246 4247 4248
	/*
	 * 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.
	 */
4249
	if (is_vma_resv_set(vma, HPAGE_RESV_OWNER) &&
4250 4251 4252
			old_page != pagecache_page)
		outside_reserve = 1;

4253
	get_page(old_page);
4254

4255 4256 4257 4258
	/*
	 * Drop page table lock as buddy allocator may be called. It will
	 * be acquired again before returning to the caller, as expected.
	 */
4259
	spin_unlock(ptl);
4260
	new_page = alloc_huge_page(vma, haddr, outside_reserve);
4261

4262
	if (IS_ERR(new_page)) {
4263 4264 4265 4266 4267 4268 4269 4270
		/*
		 * 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) {
4271 4272 4273 4274
			struct address_space *mapping = vma->vm_file->f_mapping;
			pgoff_t idx;
			u32 hash;

4275
			put_page(old_page);
4276
			BUG_ON(huge_pte_none(pte));
4277 4278 4279 4280 4281 4282 4283 4284 4285 4286 4287 4288 4289 4290
			/*
			 * Drop hugetlb_fault_mutex and i_mmap_rwsem before
			 * unmapping.  unmapping needs to hold i_mmap_rwsem
			 * in write mode.  Dropping i_mmap_rwsem in read mode
			 * here is OK as COW mappings do not interact with
			 * PMD sharing.
			 *
			 * Reacquire both after unmap operation.
			 */
			idx = vma_hugecache_offset(h, vma, haddr);
			hash = hugetlb_fault_mutex_hash(mapping, idx);
			mutex_unlock(&hugetlb_fault_mutex_table[hash]);
			i_mmap_unlock_read(mapping);

4291
			unmap_ref_private(mm, vma, old_page, haddr);
4292 4293 4294

			i_mmap_lock_read(mapping);
			mutex_lock(&hugetlb_fault_mutex_table[hash]);
4295
			spin_lock(ptl);
4296
			ptep = huge_pte_offset(mm, haddr, huge_page_size(h));
4297 4298 4299 4300 4301 4302 4303 4304
			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;
4305 4306
		}

4307
		ret = vmf_error(PTR_ERR(new_page));
4308
		goto out_release_old;
4309 4310
	}

4311 4312 4313 4314
	/*
	 * When the original hugepage is shared one, it does not have
	 * anon_vma prepared.
	 */
4315
	if (unlikely(anon_vma_prepare(vma))) {
4316 4317
		ret = VM_FAULT_OOM;
		goto out_release_all;
4318
	}
4319

4320
	copy_user_huge_page(new_page, old_page, address, vma,
A
Andrea Arcangeli 已提交
4321
			    pages_per_huge_page(h));
N
Nick Piggin 已提交
4322
	__SetPageUptodate(new_page);
4323

4324
	mmu_notifier_range_init(&range, MMU_NOTIFY_CLEAR, 0, vma, mm, haddr,
4325
				haddr + huge_page_size(h));
4326
	mmu_notifier_invalidate_range_start(&range);
4327

4328
	/*
4329
	 * Retake the page table lock to check for racing updates
4330 4331
	 * before the page tables are altered
	 */
4332
	spin_lock(ptl);
4333
	ptep = huge_pte_offset(mm, haddr, huge_page_size(h));
4334
	if (likely(ptep && pte_same(huge_ptep_get(ptep), pte))) {
4335
		ClearHPageRestoreReserve(new_page);
4336

4337
		/* Break COW */
4338
		huge_ptep_clear_flush(vma, haddr, ptep);
4339
		mmu_notifier_invalidate_range(mm, range.start, range.end);
4340
		set_huge_pte_at(mm, haddr, ptep,
4341
				make_huge_pte(vma, new_page, 1));
4342
		page_remove_rmap(old_page, true);
4343
		hugepage_add_new_anon_rmap(new_page, vma, haddr);
4344
		SetHPageMigratable(new_page);
4345 4346 4347
		/* Make the old page be freed below */
		new_page = old_page;
	}
4348
	spin_unlock(ptl);
4349
	mmu_notifier_invalidate_range_end(&range);
4350
out_release_all:
4351
	restore_reserve_on_error(h, vma, haddr, new_page);
4352
	put_page(new_page);
4353
out_release_old:
4354
	put_page(old_page);
4355

4356 4357
	spin_lock(ptl); /* Caller expects lock to be held */
	return ret;
4358 4359
}

4360
/* Return the pagecache page at a given address within a VMA */
4361 4362
static struct page *hugetlbfs_pagecache_page(struct hstate *h,
			struct vm_area_struct *vma, unsigned long address)
4363 4364
{
	struct address_space *mapping;
4365
	pgoff_t idx;
4366 4367

	mapping = vma->vm_file->f_mapping;
4368
	idx = vma_hugecache_offset(h, vma, address);
4369 4370 4371 4372

	return find_lock_page(mapping, idx);
}

H
Hugh Dickins 已提交
4373 4374 4375 4376 4377
/*
 * 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
4378 4379 4380 4381 4382 4383 4384 4385 4386 4387 4388 4389 4390 4391 4392
			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;
}

4393 4394 4395 4396 4397 4398 4399 4400 4401
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;
4402
	ClearHPageRestoreReserve(page);
4403

4404 4405 4406 4407 4408 4409
	/*
	 * set page dirty so that it will not be removed from cache/file
	 * by non-hugetlbfs specific code paths.
	 */
	set_page_dirty(page);

4410 4411 4412 4413 4414 4415
	spin_lock(&inode->i_lock);
	inode->i_blocks += blocks_per_huge_page(h);
	spin_unlock(&inode->i_lock);
	return 0;
}

4416 4417 4418 4419
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)
4420
{
4421
	struct hstate *h = hstate_vma(vma);
4422
	vm_fault_t ret = VM_FAULT_SIGBUS;
4423
	int anon_rmap = 0;
A
Adam Litke 已提交
4424 4425
	unsigned long size;
	struct page *page;
4426
	pte_t new_pte;
4427
	spinlock_t *ptl;
4428
	unsigned long haddr = address & huge_page_mask(h);
4429
	bool new_page = false;
A
Adam Litke 已提交
4430

4431 4432 4433
	/*
	 * 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
4434
	 * COW. Warn that such a situation has occurred as it may not be obvious
4435 4436
	 */
	if (is_vma_resv_set(vma, HPAGE_RESV_UNMAPPED)) {
4437
		pr_warn_ratelimited("PID %d killed due to inadequate hugepage pool\n",
4438
			   current->pid);
4439 4440 4441
		return ret;
	}

A
Adam Litke 已提交
4442
	/*
4443 4444 4445
	 * 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
4446
	 */
4447 4448 4449 4450
	size = i_size_read(mapping->host) >> huge_page_shift(h);
	if (idx >= size)
		goto out;

4451 4452 4453
retry:
	page = find_lock_page(mapping, idx);
	if (!page) {
4454 4455 4456 4457 4458 4459 4460
		/*
		 * Check for page in userfault range
		 */
		if (userfaultfd_missing(vma)) {
			u32 hash;
			struct vm_fault vmf = {
				.vma = vma,
4461
				.address = haddr,
4462 4463 4464 4465 4466 4467 4468 4469 4470 4471 4472
				.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
				 */
			};

			/*
4473 4474 4475
			 * hugetlb_fault_mutex and i_mmap_rwsem must be
			 * dropped before handling userfault.  Reacquire
			 * after handling fault to make calling code simpler.
4476
			 */
4477
			hash = hugetlb_fault_mutex_hash(mapping, idx);
4478
			mutex_unlock(&hugetlb_fault_mutex_table[hash]);
4479
			i_mmap_unlock_read(mapping);
4480
			ret = handle_userfault(&vmf, VM_UFFD_MISSING);
4481
			i_mmap_lock_read(mapping);
4482 4483 4484 4485
			mutex_lock(&hugetlb_fault_mutex_table[hash]);
			goto out;
		}

4486
		page = alloc_huge_page(vma, haddr, 0);
4487
		if (IS_ERR(page)) {
4488 4489 4490 4491 4492 4493 4494 4495 4496 4497 4498 4499 4500 4501 4502 4503 4504 4505 4506
			/*
			 * 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);
4507
			ret = vmf_error(PTR_ERR(page));
4508 4509
			goto out;
		}
A
Andrea Arcangeli 已提交
4510
		clear_huge_page(page, address, pages_per_huge_page(h));
N
Nick Piggin 已提交
4511
		__SetPageUptodate(page);
4512
		new_page = true;
4513

4514
		if (vma->vm_flags & VM_MAYSHARE) {
4515
			int err = huge_add_to_page_cache(page, mapping, idx);
4516 4517 4518 4519 4520 4521
			if (err) {
				put_page(page);
				if (err == -EEXIST)
					goto retry;
				goto out;
			}
4522
		} else {
4523
			lock_page(page);
4524 4525 4526 4527
			if (unlikely(anon_vma_prepare(vma))) {
				ret = VM_FAULT_OOM;
				goto backout_unlocked;
			}
4528
			anon_rmap = 1;
4529
		}
4530
	} else {
4531 4532 4533 4534 4535 4536
		/*
		 * 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))) {
4537
			ret = VM_FAULT_HWPOISON_LARGE |
4538
				VM_FAULT_SET_HINDEX(hstate_index(h));
4539 4540
			goto backout_unlocked;
		}
4541
	}
4542

4543 4544 4545 4546 4547 4548
	/*
	 * 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.
	 */
4549
	if ((flags & FAULT_FLAG_WRITE) && !(vma->vm_flags & VM_SHARED)) {
4550
		if (vma_needs_reservation(h, vma, haddr) < 0) {
4551 4552 4553
			ret = VM_FAULT_OOM;
			goto backout_unlocked;
		}
4554
		/* Just decrements count, does not deallocate */
4555
		vma_end_reservation(h, vma, haddr);
4556
	}
4557

4558
	ptl = huge_pte_lock(h, mm, ptep);
N
Nick Piggin 已提交
4559
	ret = 0;
4560
	if (!huge_pte_none(huge_ptep_get(ptep)))
A
Adam Litke 已提交
4561 4562
		goto backout;

4563
	if (anon_rmap) {
4564
		ClearHPageRestoreReserve(page);
4565
		hugepage_add_new_anon_rmap(page, vma, haddr);
4566
	} else
4567
		page_dup_rmap(page, true);
4568 4569
	new_pte = make_huge_pte(vma, page, ((vma->vm_flags & VM_WRITE)
				&& (vma->vm_flags & VM_SHARED)));
4570
	set_huge_pte_at(mm, haddr, ptep, new_pte);
4571

4572
	hugetlb_count_add(pages_per_huge_page(h), mm);
4573
	if ((flags & FAULT_FLAG_WRITE) && !(vma->vm_flags & VM_SHARED)) {
4574
		/* Optimization, do the COW without a second fault */
4575
		ret = hugetlb_cow(mm, vma, address, ptep, page, ptl);
4576 4577
	}

4578
	spin_unlock(ptl);
4579 4580

	/*
4581 4582 4583
	 * Only set HPageMigratable in newly allocated pages.  Existing pages
	 * found in the pagecache may not have HPageMigratableset if they have
	 * been isolated for migration.
4584 4585
	 */
	if (new_page)
4586
		SetHPageMigratable(page);
4587

A
Adam Litke 已提交
4588 4589
	unlock_page(page);
out:
4590
	return ret;
A
Adam Litke 已提交
4591 4592

backout:
4593
	spin_unlock(ptl);
4594
backout_unlocked:
A
Adam Litke 已提交
4595
	unlock_page(page);
4596
	restore_reserve_on_error(h, vma, haddr, page);
A
Adam Litke 已提交
4597 4598
	put_page(page);
	goto out;
4599 4600
}

4601
#ifdef CONFIG_SMP
4602
u32 hugetlb_fault_mutex_hash(struct address_space *mapping, pgoff_t idx)
4603 4604 4605 4606
{
	unsigned long key[2];
	u32 hash;

4607 4608
	key[0] = (unsigned long) mapping;
	key[1] = idx;
4609

4610
	hash = jhash2((u32 *)&key, sizeof(key)/(sizeof(u32)), 0);
4611 4612 4613 4614 4615 4616 4617 4618

	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.
 */
4619
u32 hugetlb_fault_mutex_hash(struct address_space *mapping, pgoff_t idx)
4620 4621 4622 4623 4624
{
	return 0;
}
#endif

4625
vm_fault_t hugetlb_fault(struct mm_struct *mm, struct vm_area_struct *vma,
4626
			unsigned long address, unsigned int flags)
4627
{
4628
	pte_t *ptep, entry;
4629
	spinlock_t *ptl;
4630
	vm_fault_t ret;
4631 4632
	u32 hash;
	pgoff_t idx;
4633
	struct page *page = NULL;
4634
	struct page *pagecache_page = NULL;
4635
	struct hstate *h = hstate_vma(vma);
4636
	struct address_space *mapping;
4637
	int need_wait_lock = 0;
4638
	unsigned long haddr = address & huge_page_mask(h);
4639

4640
	ptep = huge_pte_offset(mm, haddr, huge_page_size(h));
4641
	if (ptep) {
4642 4643 4644 4645 4646
		/*
		 * 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.
		 */
4647
		entry = huge_ptep_get(ptep);
N
Naoya Horiguchi 已提交
4648
		if (unlikely(is_hugetlb_entry_migration(entry))) {
4649
			migration_entry_wait_huge(vma, mm, ptep);
N
Naoya Horiguchi 已提交
4650 4651
			return 0;
		} else if (unlikely(is_hugetlb_entry_hwpoisoned(entry)))
4652
			return VM_FAULT_HWPOISON_LARGE |
4653
				VM_FAULT_SET_HINDEX(hstate_index(h));
4654 4655
	}

4656 4657
	/*
	 * Acquire i_mmap_rwsem before calling huge_pte_alloc and hold
4658 4659 4660 4661
	 * 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.
4662 4663 4664 4665 4666
	 *
	 * 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.
	 */
4667
	mapping = vma->vm_file->f_mapping;
4668 4669 4670 4671 4672 4673
	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;
	}
4674

4675 4676 4677 4678 4679
	/*
	 * 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.
	 */
4680
	idx = vma_hugecache_offset(h, vma, haddr);
4681
	hash = hugetlb_fault_mutex_hash(mapping, idx);
4682
	mutex_lock(&hugetlb_fault_mutex_table[hash]);
4683

4684 4685
	entry = huge_ptep_get(ptep);
	if (huge_pte_none(entry)) {
4686
		ret = hugetlb_no_page(mm, vma, mapping, idx, address, ptep, flags);
4687
		goto out_mutex;
4688
	}
4689

N
Nick Piggin 已提交
4690
	ret = 0;
4691

4692 4693 4694
	/*
	 * 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
4695 4696 4697
	 * an active hugepage in pagecache. This goto expects the 2nd page
	 * fault, and is_hugetlb_entry_(migration|hwpoisoned) check will
	 * properly handle it.
4698 4699 4700 4701
	 */
	if (!pte_present(entry))
		goto out_mutex;

4702 4703 4704 4705 4706 4707 4708 4709
	/*
	 * 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.
	 */
4710
	if ((flags & FAULT_FLAG_WRITE) && !huge_pte_write(entry)) {
4711
		if (vma_needs_reservation(h, vma, haddr) < 0) {
4712
			ret = VM_FAULT_OOM;
4713
			goto out_mutex;
4714
		}
4715
		/* Just decrements count, does not deallocate */
4716
		vma_end_reservation(h, vma, haddr);
4717

4718
		if (!(vma->vm_flags & VM_MAYSHARE))
4719
			pagecache_page = hugetlbfs_pagecache_page(h,
4720
								vma, haddr);
4721 4722
	}

4723 4724 4725 4726 4727 4728
	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;

4729 4730 4731 4732 4733 4734 4735
	/*
	 * 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)
4736 4737 4738 4739
		if (!trylock_page(page)) {
			need_wait_lock = 1;
			goto out_ptl;
		}
4740

4741
	get_page(page);
4742

4743
	if (flags & FAULT_FLAG_WRITE) {
4744
		if (!huge_pte_write(entry)) {
4745
			ret = hugetlb_cow(mm, vma, address, ptep,
4746
					  pagecache_page, ptl);
4747
			goto out_put_page;
4748
		}
4749
		entry = huge_pte_mkdirty(entry);
4750 4751
	}
	entry = pte_mkyoung(entry);
4752
	if (huge_ptep_set_access_flags(vma, haddr, ptep, entry,
4753
						flags & FAULT_FLAG_WRITE))
4754
		update_mmu_cache(vma, haddr, ptep);
4755 4756 4757 4758
out_put_page:
	if (page != pagecache_page)
		unlock_page(page);
	put_page(page);
4759 4760
out_ptl:
	spin_unlock(ptl);
4761 4762 4763 4764 4765

	if (pagecache_page) {
		unlock_page(pagecache_page);
		put_page(pagecache_page);
	}
4766
out_mutex:
4767
	mutex_unlock(&hugetlb_fault_mutex_table[hash]);
4768
	i_mmap_unlock_read(mapping);
4769 4770 4771 4772 4773 4774 4775 4776 4777
	/*
	 * 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);
4778
	return ret;
4779 4780
}

4781 4782 4783 4784 4785 4786 4787 4788 4789 4790 4791
/*
 * 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)
{
4792 4793 4794
	struct address_space *mapping;
	pgoff_t idx;
	unsigned long size;
4795
	int vm_shared = dst_vma->vm_flags & VM_SHARED;
4796 4797 4798 4799 4800 4801 4802
	struct hstate *h = hstate_vma(dst_vma);
	pte_t _dst_pte;
	spinlock_t *ptl;
	int ret;
	struct page *page;

	if (!*pagep) {
4803 4804 4805 4806 4807 4808 4809 4810 4811
		/* If a page already exists, then it's UFFDIO_COPY for
		 * a non-missing case. Return -EEXIST.
		 */
		if (vm_shared &&
		    hugetlbfs_pagecache_present(h, dst_vma, dst_addr)) {
			ret = -EEXIST;
			goto out;
		}

4812
		page = alloc_huge_page(dst_vma, dst_addr, 0);
4813 4814
		if (IS_ERR(page)) {
			ret = -ENOMEM;
4815
			goto out;
4816
		}
4817 4818 4819

		ret = copy_huge_page_from_user(page,
						(const void __user *) src_addr,
4820
						pages_per_huge_page(h), false);
4821

4822
		/* fallback to copy_from_user outside mmap_lock */
4823
		if (unlikely(ret)) {
4824
			ret = -ENOENT;
4825 4826 4827 4828 4829 4830 4831 4832 4833 4834 4835 4836 4837 4838 4839 4840
			*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);

4841 4842 4843
	mapping = dst_vma->vm_file->f_mapping;
	idx = vma_hugecache_offset(h, dst_vma, dst_addr);

4844 4845 4846 4847
	/*
	 * If shared, add to page cache
	 */
	if (vm_shared) {
4848 4849 4850 4851
		size = i_size_read(mapping->host) >> huge_page_shift(h);
		ret = -EFAULT;
		if (idx >= size)
			goto out_release_nounlock;
4852

4853 4854 4855 4856 4857 4858
		/*
		 * 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.
		 */
4859 4860 4861 4862 4863
		ret = huge_add_to_page_cache(page, mapping, idx);
		if (ret)
			goto out_release_nounlock;
	}

4864 4865 4866
	ptl = huge_pte_lockptr(h, dst_mm, dst_pte);
	spin_lock(ptl);

4867 4868 4869 4870 4871 4872 4873 4874 4875 4876 4877 4878 4879 4880
	/*
	 * 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;

4881 4882 4883 4884
	ret = -EEXIST;
	if (!huge_pte_none(huge_ptep_get(dst_pte)))
		goto out_release_unlock;

4885 4886 4887
	if (vm_shared) {
		page_dup_rmap(page, true);
	} else {
4888
		ClearHPageRestoreReserve(page);
4889 4890
		hugepage_add_new_anon_rmap(page, dst_vma, dst_addr);
	}
4891 4892 4893 4894 4895 4896 4897 4898 4899 4900 4901 4902 4903 4904 4905 4906

	_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);
4907
	SetHPageMigratable(page);
4908 4909
	if (vm_shared)
		unlock_page(page);
4910 4911 4912 4913 4914
	ret = 0;
out:
	return ret;
out_release_unlock:
	spin_unlock(ptl);
4915 4916
	if (vm_shared)
		unlock_page(page);
4917
out_release_nounlock:
4918 4919 4920 4921
	put_page(page);
	goto out;
}

4922 4923 4924
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,
4925
			 long i, unsigned int flags, int *locked)
D
David Gibson 已提交
4926
{
4927 4928
	unsigned long pfn_offset;
	unsigned long vaddr = *position;
4929
	unsigned long remainder = *nr_pages;
4930
	struct hstate *h = hstate_vma(vma);
4931
	int err = -EFAULT;
D
David Gibson 已提交
4932 4933

	while (vaddr < vma->vm_end && remainder) {
A
Adam Litke 已提交
4934
		pte_t *pte;
4935
		spinlock_t *ptl = NULL;
H
Hugh Dickins 已提交
4936
		int absent;
A
Adam Litke 已提交
4937
		struct page *page;
D
David Gibson 已提交
4938

4939 4940 4941 4942
		/*
		 * If we have a pending SIGKILL, don't keep faulting pages and
		 * potentially allocating memory.
		 */
4943
		if (fatal_signal_pending(current)) {
4944 4945 4946 4947
			remainder = 0;
			break;
		}

A
Adam Litke 已提交
4948 4949
		/*
		 * Some archs (sparc64, sh*) have multiple pte_ts to
H
Hugh Dickins 已提交
4950
		 * each hugepage.  We have to make sure we get the
A
Adam Litke 已提交
4951
		 * first, for the page indexing below to work.
4952 4953
		 *
		 * Note that page table lock is not held when pte is null.
A
Adam Litke 已提交
4954
		 */
4955 4956
		pte = huge_pte_offset(mm, vaddr & huge_page_mask(h),
				      huge_page_size(h));
4957 4958
		if (pte)
			ptl = huge_pte_lock(h, mm, pte);
H
Hugh Dickins 已提交
4959 4960 4961 4962
		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
4963 4964 4965 4966
		 * 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
4967
		 */
H
Hugh Dickins 已提交
4968 4969
		if (absent && (flags & FOLL_DUMP) &&
		    !hugetlbfs_pagecache_present(h, vma, vaddr)) {
4970 4971
			if (pte)
				spin_unlock(ptl);
H
Hugh Dickins 已提交
4972 4973 4974
			remainder = 0;
			break;
		}
D
David Gibson 已提交
4975

4976 4977 4978 4979 4980 4981 4982 4983 4984 4985 4986
		/*
		 * 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)) ||
4987 4988
		    ((flags & FOLL_WRITE) &&
		      !huge_pte_write(huge_ptep_get(pte)))) {
4989
			vm_fault_t ret;
4990
			unsigned int fault_flags = 0;
D
David Gibson 已提交
4991

4992 4993
			if (pte)
				spin_unlock(ptl);
4994 4995
			if (flags & FOLL_WRITE)
				fault_flags |= FAULT_FLAG_WRITE;
4996
			if (locked)
4997 4998
				fault_flags |= FAULT_FLAG_ALLOW_RETRY |
					FAULT_FLAG_KILLABLE;
4999 5000 5001 5002
			if (flags & FOLL_NOWAIT)
				fault_flags |= FAULT_FLAG_ALLOW_RETRY |
					FAULT_FLAG_RETRY_NOWAIT;
			if (flags & FOLL_TRIED) {
5003 5004 5005 5006
				/*
				 * Note: FAULT_FLAG_ALLOW_RETRY and
				 * FAULT_FLAG_TRIED can co-exist
				 */
5007 5008 5009 5010
				fault_flags |= FAULT_FLAG_TRIED;
			}
			ret = hugetlb_fault(mm, vma, vaddr, fault_flags);
			if (ret & VM_FAULT_ERROR) {
5011
				err = vm_fault_to_errno(ret, flags);
5012 5013 5014 5015
				remainder = 0;
				break;
			}
			if (ret & VM_FAULT_RETRY) {
5016
				if (locked &&
5017
				    !(fault_flags & FAULT_FLAG_RETRY_NOWAIT))
5018
					*locked = 0;
5019 5020 5021 5022 5023 5024 5025 5026 5027 5028 5029 5030 5031
				*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
5032 5033
		}

5034
		pfn_offset = (vaddr & ~huge_page_mask(h)) >> PAGE_SHIFT;
5035
		page = pte_page(huge_ptep_get(pte));
5036

5037 5038 5039 5040 5041 5042 5043 5044 5045 5046 5047 5048 5049 5050
		/*
		 * 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;
		}

5051
same_page:
5052
		if (pages) {
H
Hugh Dickins 已提交
5053
			pages[i] = mem_map_offset(page, pfn_offset);
J
John Hubbard 已提交
5054 5055 5056 5057 5058 5059 5060 5061 5062 5063 5064 5065 5066 5067 5068 5069
			/*
			 * 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;
			}
5070
		}
D
David Gibson 已提交
5071 5072 5073 5074 5075

		if (vmas)
			vmas[i] = vma;

		vaddr += PAGE_SIZE;
5076
		++pfn_offset;
D
David Gibson 已提交
5077 5078
		--remainder;
		++i;
5079
		if (vaddr < vma->vm_end && remainder &&
5080
				pfn_offset < pages_per_huge_page(h)) {
5081 5082 5083 5084 5085 5086
			/*
			 * We use pfn_offset to avoid touching the pageframes
			 * of this compound page.
			 */
			goto same_page;
		}
5087
		spin_unlock(ptl);
D
David Gibson 已提交
5088
	}
5089
	*nr_pages = remainder;
5090 5091 5092 5093 5094
	/*
	 * 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
5095 5096
	*position = vaddr;

5097
	return i ? i : err;
D
David Gibson 已提交
5098
}
5099

5100 5101 5102 5103 5104 5105 5106 5107
#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

5108
unsigned long hugetlb_change_protection(struct vm_area_struct *vma,
5109 5110 5111 5112 5113 5114
		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;
5115
	struct hstate *h = hstate_vma(vma);
5116
	unsigned long pages = 0;
5117
	bool shared_pmd = false;
5118
	struct mmu_notifier_range range;
5119 5120 5121

	/*
	 * In the case of shared PMDs, the area to flush could be beyond
5122
	 * start/end.  Set range.start/range.end to cover the maximum possible
5123 5124
	 * range if PMD sharing is possible.
	 */
5125 5126
	mmu_notifier_range_init(&range, MMU_NOTIFY_PROTECTION_VMA,
				0, vma, mm, start, end);
5127
	adjust_range_if_pmd_sharing_possible(vma, &range.start, &range.end);
5128 5129

	BUG_ON(address >= end);
5130
	flush_cache_range(vma, range.start, range.end);
5131

5132
	mmu_notifier_invalidate_range_start(&range);
5133
	i_mmap_lock_write(vma->vm_file->f_mapping);
5134
	for (; address < end; address += huge_page_size(h)) {
5135
		spinlock_t *ptl;
5136
		ptep = huge_pte_offset(mm, address, huge_page_size(h));
5137 5138
		if (!ptep)
			continue;
5139
		ptl = huge_pte_lock(h, mm, ptep);
5140
		if (huge_pmd_unshare(mm, vma, &address, ptep)) {
5141
			pages++;
5142
			spin_unlock(ptl);
5143
			shared_pmd = true;
5144
			continue;
5145
		}
5146 5147 5148 5149 5150 5151 5152 5153 5154 5155 5156 5157 5158
		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);
5159 5160
				set_huge_swap_pte_at(mm, address, ptep,
						     newpte, huge_page_size(h));
5161 5162 5163 5164 5165 5166
				pages++;
			}
			spin_unlock(ptl);
			continue;
		}
		if (!huge_pte_none(pte)) {
5167 5168 5169 5170
			pte_t old_pte;

			old_pte = huge_ptep_modify_prot_start(vma, address, ptep);
			pte = pte_mkhuge(huge_pte_modify(old_pte, newprot));
5171
			pte = arch_make_huge_pte(pte, vma, NULL, 0);
5172
			huge_ptep_modify_prot_commit(vma, address, ptep, old_pte, pte);
5173
			pages++;
5174
		}
5175
		spin_unlock(ptl);
5176
	}
5177
	/*
5178
	 * Must flush TLB before releasing i_mmap_rwsem: x86's huge_pmd_unshare
5179
	 * may have cleared our pud entry and done put_page on the page table:
5180
	 * once we release i_mmap_rwsem, another task can do the final put_page
5181 5182
	 * 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.
5183
	 */
5184
	if (shared_pmd)
5185
		flush_hugetlb_tlb_range(vma, range.start, range.end);
5186 5187
	else
		flush_hugetlb_tlb_range(vma, start, end);
5188 5189 5190 5191
	/*
	 * No need to call mmu_notifier_invalidate_range() we are downgrading
	 * page table protection not changing it to point to a new page.
	 *
5192
	 * See Documentation/vm/mmu_notifier.rst
5193
	 */
5194
	i_mmap_unlock_write(vma->vm_file->f_mapping);
5195
	mmu_notifier_invalidate_range_end(&range);
5196 5197

	return pages << h->order;
5198 5199
}

5200 5201
int hugetlb_reserve_pages(struct inode *inode,
					long from, long to,
5202
					struct vm_area_struct *vma,
5203
					vm_flags_t vm_flags)
5204
{
5205
	long ret, chg, add = -1;
5206
	struct hstate *h = hstate_inode(inode);
5207
	struct hugepage_subpool *spool = subpool_inode(inode);
5208
	struct resv_map *resv_map;
5209
	struct hugetlb_cgroup *h_cg = NULL;
5210
	long gbl_reserve, regions_needed = 0;
5211

5212 5213 5214 5215 5216 5217
	/* This should never happen */
	if (from > to) {
		VM_WARN(1, "%s called with a negative range\n", __func__);
		return -EINVAL;
	}

5218 5219 5220
	/*
	 * Only apply hugepage reservation if asked. At fault time, an
	 * attempt will be made for VM_NORESERVE to allocate a page
5221
	 * without using reserves
5222
	 */
5223
	if (vm_flags & VM_NORESERVE)
5224 5225
		return 0;

5226 5227 5228 5229 5230 5231
	/*
	 * 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
	 */
5232
	if (!vma || vma->vm_flags & VM_MAYSHARE) {
5233 5234 5235 5236 5237
		/*
		 * 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).
		 */
5238
		resv_map = inode_resv_map(inode);
5239

5240
		chg = region_chg(resv_map, from, to, &regions_needed);
5241 5242

	} else {
5243
		/* Private mapping. */
5244
		resv_map = resv_map_alloc();
5245 5246 5247
		if (!resv_map)
			return -ENOMEM;

5248
		chg = to - from;
5249

5250 5251 5252 5253
		set_vma_resv_map(vma, resv_map);
		set_vma_resv_flags(vma, HPAGE_RESV_OWNER);
	}

5254 5255 5256 5257
	if (chg < 0) {
		ret = chg;
		goto out_err;
	}
5258

5259 5260 5261 5262 5263 5264 5265 5266 5267 5268 5269 5270 5271 5272 5273
	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);
	}

5274 5275 5276 5277 5278 5279 5280
	/*
	 * 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) {
5281
		ret = -ENOSPC;
5282
		goto out_uncharge_cgroup;
5283
	}
5284 5285

	/*
5286
	 * Check enough hugepages are available for the reservation.
5287
	 * Hand the pages back to the subpool if there are not
5288
	 */
5289
	ret = hugetlb_acct_memory(h, gbl_reserve);
K
Ken Chen 已提交
5290
	if (ret < 0) {
5291
		goto out_put_pages;
K
Ken Chen 已提交
5292
	}
5293 5294 5295 5296 5297 5298 5299 5300 5301 5302 5303 5304

	/*
	 * 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
	 */
5305
	if (!vma || vma->vm_flags & VM_MAYSHARE) {
5306
		add = region_add(resv_map, from, to, regions_needed, h, h_cg);
5307 5308 5309

		if (unlikely(add < 0)) {
			hugetlb_acct_memory(h, -gbl_reserve);
5310
			ret = add;
5311
			goto out_put_pages;
5312
		} else if (unlikely(chg > add)) {
5313 5314 5315 5316 5317 5318 5319 5320 5321
			/*
			 * 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;

5322 5323 5324 5325
			/*
			 * hugetlb_cgroup_uncharge_cgroup_rsvd() will put the
			 * reference to h_cg->css. See comment below for detail.
			 */
5326 5327 5328 5329
			hugetlb_cgroup_uncharge_cgroup_rsvd(
				hstate_index(h),
				(chg - add) * pages_per_huge_page(h), h_cg);

5330 5331 5332
			rsv_adjust = hugepage_subpool_put_pages(spool,
								chg - add);
			hugetlb_acct_memory(h, -rsv_adjust);
5333 5334 5335 5336 5337 5338 5339 5340
		} else if (h_cg) {
			/*
			 * The file_regions will hold their own reference to
			 * h_cg->css. So we should release the reference held
			 * via hugetlb_cgroup_charge_cgroup_rsvd() when we are
			 * done.
			 */
			hugetlb_cgroup_put_rsvd_cgroup(h_cg);
5341 5342
		}
	}
5343
	return 0;
5344 5345 5346 5347 5348 5349
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);
5350
out_err:
5351
	if (!vma || vma->vm_flags & VM_MAYSHARE)
5352 5353 5354 5355 5356
		/* 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
5357 5358
	if (vma && is_vma_resv_set(vma, HPAGE_RESV_OWNER))
		kref_put(&resv_map->refs, resv_map_release);
5359
	return ret;
5360 5361
}

5362 5363
long hugetlb_unreserve_pages(struct inode *inode, long start, long end,
								long freed)
5364
{
5365
	struct hstate *h = hstate_inode(inode);
5366
	struct resv_map *resv_map = inode_resv_map(inode);
5367
	long chg = 0;
5368
	struct hugepage_subpool *spool = subpool_inode(inode);
5369
	long gbl_reserve;
K
Ken Chen 已提交
5370

5371 5372 5373 5374
	/*
	 * Since this routine can be called in the evict inode path for all
	 * hugetlbfs inodes, resv_map could be NULL.
	 */
5375 5376 5377 5378 5379 5380 5381 5382 5383 5384 5385
	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
5386
	spin_lock(&inode->i_lock);
5387
	inode->i_blocks -= (blocks_per_huge_page(h) * freed);
K
Ken Chen 已提交
5388 5389
	spin_unlock(&inode->i_lock);

5390 5391 5392 5393 5394 5395
	/*
	 * 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);
5396 5397

	return 0;
5398
}
5399

5400 5401 5402 5403 5404 5405 5406 5407 5408 5409 5410
#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
5411 5412
	unsigned long vm_flags = vma->vm_flags & VM_LOCKED_CLEAR_MASK;
	unsigned long svm_flags = svma->vm_flags & VM_LOCKED_CLEAR_MASK;
5413 5414 5415 5416 5417 5418 5419 5420 5421 5422 5423 5424 5425

	/*
	 * 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;
}

5426
static bool vma_shareable(struct vm_area_struct *vma, unsigned long addr)
5427 5428 5429 5430 5431 5432 5433
{
	unsigned long base = addr & PUD_MASK;
	unsigned long end = base + PUD_SIZE;

	/*
	 * check on proper vm_flags and page table alignment
	 */
5434
	if (vma->vm_flags & VM_MAYSHARE && range_in_vma(vma, base, end))
5435 5436
		return true;
	return false;
5437 5438
}

5439 5440 5441 5442 5443 5444 5445 5446
/*
 * 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)
{
5447 5448
	unsigned long v_start = ALIGN(vma->vm_start, PUD_SIZE),
		v_end = ALIGN_DOWN(vma->vm_end, PUD_SIZE);
5449

5450 5451 5452 5453 5454 5455
	/*
	 * vma need span at least one aligned PUD size and the start,end range
	 * must at least partialy within it.
	 */
	if (!(vma->vm_flags & VM_MAYSHARE) || !(v_end > v_start) ||
		(*end <= v_start) || (*start >= v_end))
5456 5457
		return;

5458
	/* Extend the range to be PUD aligned for a worst case scenario */
5459 5460
	if (*start > v_start)
		*start = ALIGN_DOWN(*start, PUD_SIZE);
5461

5462 5463
	if (*end < v_end)
		*end = ALIGN(*end, PUD_SIZE);
5464 5465
}

5466 5467 5468 5469
/*
 * 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
5470 5471
 * code much cleaner.
 *
5472 5473 5474 5475 5476 5477 5478 5479 5480 5481
 * This routine must be called with i_mmap_rwsem held in at least read mode if
 * sharing is possible.  For hugetlbfs, this prevents removal of any page
 * table entries associated with the address space.  This is important as we
 * are setting up sharing based on existing page table entries (mappings).
 *
 * NOTE: This routine is only called from huge_pte_alloc.  Some callers of
 * huge_pte_alloc know that sharing is not possible and do not take
 * i_mmap_rwsem as a performance optimization.  This is handled by the
 * if !vma_shareable check at the beginning of the routine. i_mmap_rwsem is
 * only required for subsequent processing.
5482 5483 5484 5485 5486 5487 5488 5489 5490 5491 5492
 */
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;
5493
	spinlock_t *ptl;
5494 5495 5496 5497

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

5498
	i_mmap_assert_locked(mapping);
5499 5500 5501 5502 5503 5504
	vma_interval_tree_foreach(svma, &mapping->i_mmap, idx, idx) {
		if (svma == vma)
			continue;

		saddr = page_table_shareable(svma, vma, addr, idx);
		if (saddr) {
5505 5506
			spte = huge_pte_offset(svma->vm_mm, saddr,
					       vma_mmu_pagesize(svma));
5507 5508 5509 5510 5511 5512 5513 5514 5515 5516
			if (spte) {
				get_page(virt_to_page(spte));
				break;
			}
		}
	}

	if (!spte)
		goto out;

5517
	ptl = huge_pte_lock(hstate_vma(vma), mm, spte);
5518
	if (pud_none(*pud)) {
5519 5520
		pud_populate(mm, pud,
				(pmd_t *)((unsigned long)spte & PAGE_MASK));
5521
		mm_inc_nr_pmds(mm);
5522
	} else {
5523
		put_page(virt_to_page(spte));
5524
	}
5525
	spin_unlock(ptl);
5526 5527 5528 5529 5530 5531 5532 5533 5534 5535 5536 5537
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.
 *
5538
 * Called with page table lock held and i_mmap_rwsem held in write mode.
5539 5540 5541 5542
 *
 * returns: 1 successfully unmapped a shared pte page
 *	    0 the underlying pte page is not shared, or it is the last user
 */
5543 5544
int huge_pmd_unshare(struct mm_struct *mm, struct vm_area_struct *vma,
					unsigned long *addr, pte_t *ptep)
5545 5546
{
	pgd_t *pgd = pgd_offset(mm, *addr);
5547 5548
	p4d_t *p4d = p4d_offset(pgd, *addr);
	pud_t *pud = pud_offset(p4d, *addr);
5549

5550
	i_mmap_assert_write_locked(vma->vm_file->f_mapping);
5551 5552 5553 5554 5555 5556
	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));
5557
	mm_dec_nr_pmds(mm);
5558 5559 5560
	*addr = ALIGN(*addr, HPAGE_SIZE * PTRS_PER_PTE) - HPAGE_SIZE;
	return 1;
}
5561 5562 5563 5564 5565 5566
#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;
}
5567

5568 5569
int huge_pmd_unshare(struct mm_struct *mm, struct vm_area_struct *vma,
				unsigned long *addr, pte_t *ptep)
5570 5571 5572
{
	return 0;
}
5573 5574 5575 5576 5577

void adjust_range_if_pmd_sharing_possible(struct vm_area_struct *vma,
				unsigned long *start, unsigned long *end)
{
}
5578
#define want_pmd_share()	(0)
5579 5580
#endif /* CONFIG_ARCH_WANT_HUGE_PMD_SHARE */

5581 5582 5583 5584 5585
#ifdef CONFIG_ARCH_WANT_GENERAL_HUGETLB
pte_t *huge_pte_alloc(struct mm_struct *mm,
			unsigned long addr, unsigned long sz)
{
	pgd_t *pgd;
5586
	p4d_t *p4d;
5587 5588 5589 5590
	pud_t *pud;
	pte_t *pte = NULL;

	pgd = pgd_offset(mm, addr);
5591 5592 5593
	p4d = p4d_alloc(mm, pgd, addr);
	if (!p4d)
		return NULL;
5594
	pud = pud_alloc(mm, p4d, addr);
5595 5596 5597 5598 5599 5600 5601 5602 5603 5604 5605
	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);
		}
	}
5606
	BUG_ON(pte && pte_present(*pte) && !pte_huge(*pte));
5607 5608 5609 5610

	return pte;
}

5611 5612 5613 5614
/*
 * huge_pte_offset() - Walk the page table to resolve the hugepage
 * entry at address @addr
 *
5615 5616
 * Return: Pointer to page table entry (PUD or PMD) for
 * address @addr, or NULL if a !p*d_present() entry is encountered and the
5617 5618 5619
 * size @sz doesn't match the hugepage size at this level of the page
 * table.
 */
5620 5621
pte_t *huge_pte_offset(struct mm_struct *mm,
		       unsigned long addr, unsigned long sz)
5622 5623
{
	pgd_t *pgd;
5624
	p4d_t *p4d;
5625 5626
	pud_t *pud;
	pmd_t *pmd;
5627 5628

	pgd = pgd_offset(mm, addr);
5629 5630 5631 5632 5633
	if (!pgd_present(*pgd))
		return NULL;
	p4d = p4d_offset(pgd, addr);
	if (!p4d_present(*p4d))
		return NULL;
5634

5635
	pud = pud_offset(p4d, addr);
5636 5637
	if (sz == PUD_SIZE)
		/* must be pud huge, non-present or none */
5638
		return (pte_t *)pud;
5639
	if (!pud_present(*pud))
5640
		return NULL;
5641
	/* must have a valid entry and size to go further */
5642

5643 5644 5645
	pmd = pmd_offset(pud, addr);
	/* must be pmd huge, non-present or none */
	return (pte_t *)pmd;
5646 5647
}

5648 5649 5650 5651 5652 5653 5654 5655 5656 5657 5658 5659 5660
#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);
}

5661 5662 5663 5664 5665 5666 5667 5668
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;
}

5669
struct page * __weak
5670
follow_huge_pmd(struct mm_struct *mm, unsigned long address,
5671
		pmd_t *pmd, int flags)
5672
{
5673 5674
	struct page *page = NULL;
	spinlock_t *ptl;
5675
	pte_t pte;
J
John Hubbard 已提交
5676 5677 5678 5679 5680 5681

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

5682 5683 5684 5685 5686 5687 5688 5689 5690
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;
5691 5692
	pte = huge_ptep_get((pte_t *)pmd);
	if (pte_present(pte)) {
5693
		page = pmd_page(*pmd) + ((address & ~PMD_MASK) >> PAGE_SHIFT);
J
John Hubbard 已提交
5694 5695 5696 5697 5698 5699 5700 5701 5702 5703 5704 5705
		/*
		 * 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;
		}
5706
	} else {
5707
		if (is_hugetlb_entry_migration(pte)) {
5708 5709 5710 5711 5712 5713 5714 5715 5716 5717 5718
			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);
5719 5720 5721
	return page;
}

5722
struct page * __weak
5723
follow_huge_pud(struct mm_struct *mm, unsigned long address,
5724
		pud_t *pud, int flags)
5725
{
J
John Hubbard 已提交
5726
	if (flags & (FOLL_GET | FOLL_PIN))
5727
		return NULL;
5728

5729
	return pte_page(*(pte_t *)pud) + ((address & ~PUD_MASK) >> PAGE_SHIFT);
5730 5731
}

5732 5733 5734
struct page * __weak
follow_huge_pgd(struct mm_struct *mm, unsigned long address, pgd_t *pgd, int flags)
{
J
John Hubbard 已提交
5735
	if (flags & (FOLL_GET | FOLL_PIN))
5736 5737 5738 5739 5740
		return NULL;

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

5741 5742
bool isolate_huge_page(struct page *page, struct list_head *list)
{
5743 5744
	bool ret = true;

5745
	spin_lock_irq(&hugetlb_lock);
5746 5747
	if (!PageHeadHuge(page) ||
	    !HPageMigratable(page) ||
5748
	    !get_page_unless_zero(page)) {
5749 5750 5751
		ret = false;
		goto unlock;
	}
5752
	ClearHPageMigratable(page);
5753
	list_move_tail(&page->lru, list);
5754
unlock:
5755
	spin_unlock_irq(&hugetlb_lock);
5756
	return ret;
5757 5758 5759 5760
}

void putback_active_hugepage(struct page *page)
{
5761
	VM_BUG_ON_PAGE(!PageHead(page), page);
5762
	spin_lock_irq(&hugetlb_lock);
5763
	SetHPageMigratable(page);
5764
	list_move_tail(&page->lru, &(page_hstate(page))->hugepage_activelist);
5765
	spin_unlock_irq(&hugetlb_lock);
5766 5767
	put_page(page);
}
5768 5769 5770 5771 5772 5773 5774 5775 5776 5777 5778 5779 5780 5781 5782 5783 5784 5785

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.
	 */
5786
	if (HPageTemporary(newpage)) {
5787 5788 5789
		int old_nid = page_to_nid(oldpage);
		int new_nid = page_to_nid(newpage);

5790 5791
		SetHPageTemporary(oldpage);
		ClearHPageTemporary(newpage);
5792

5793
		spin_lock_irq(&hugetlb_lock);
5794 5795 5796 5797
		if (h->surplus_huge_pages_node[old_nid]) {
			h->surplus_huge_pages_node[old_nid]--;
			h->surplus_huge_pages_node[new_nid]++;
		}
5798
		spin_unlock_irq(&hugetlb_lock);
5799 5800
	}
}
5801 5802 5803 5804 5805 5806 5807 5808 5809 5810 5811 5812 5813 5814 5815 5816 5817 5818 5819 5820 5821 5822 5823 5824 5825 5826 5827 5828 5829 5830 5831 5832 5833 5834 5835 5836 5837 5838 5839

#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;
5840
		char name[CMA_MAX_NAME];
5841 5842 5843 5844

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

5845
		snprintf(name, sizeof(name), "hugetlb%d", nid);
5846
		res = cma_declare_contiguous_nid(0, size, 0, PAGE_SIZE << order,
5847
						 0, false, name,
5848 5849 5850 5851 5852 5853 5854 5855 5856 5857 5858 5859 5860 5861 5862 5863 5864 5865 5866 5867 5868 5869 5870 5871 5872
						 &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 */