hugetlb.c 88.6 KB
Newer Older
L
Linus Torvalds 已提交
1 2
/*
 * Generic hugetlb support.
3
 * (C) Nadia Yvette Chambers, April 2004
L
Linus Torvalds 已提交
4 5 6 7 8
 */
#include <linux/list.h>
#include <linux/init.h>
#include <linux/module.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/cpuset.h>
17
#include <linux/mutex.h>
18
#include <linux/bootmem.h>
19
#include <linux/sysfs.h>
20
#include <linux/slab.h>
21
#include <linux/rmap.h>
22 23
#include <linux/swap.h>
#include <linux/swapops.h>
24

D
David Gibson 已提交
25 26
#include <asm/page.h>
#include <asm/pgtable.h>
27
#include <asm/tlb.h>
D
David Gibson 已提交
28

29
#include <linux/io.h>
D
David Gibson 已提交
30
#include <linux/hugetlb.h>
31
#include <linux/hugetlb_cgroup.h>
32
#include <linux/node.h>
33
#include "internal.h"
L
Linus Torvalds 已提交
34 35

const unsigned long hugetlb_zero = 0, hugetlb_infinity = ~0UL;
36 37
static gfp_t htlb_alloc_mask = GFP_HIGHUSER;
unsigned long hugepages_treat_as_movable;
38

39
int hugetlb_max_hstate __read_mostly;
40 41 42
unsigned int default_hstate_idx;
struct hstate hstates[HUGE_MAX_HSTATE];

43 44
__initdata LIST_HEAD(huge_boot_pages);

45 46 47
/* for command line parsing */
static struct hstate * __initdata parsed_hstate;
static unsigned long __initdata default_hstate_max_huge_pages;
48
static unsigned long __initdata default_hstate_size;
49

50 51 52
/*
 * Protects updates to hugepage_freelists, nr_huge_pages, and free_huge_pages
 */
53
DEFINE_SPINLOCK(hugetlb_lock);
54

55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129
static inline void unlock_or_release_subpool(struct hugepage_subpool *spool)
{
	bool free = (spool->count == 0) && (spool->used_hpages == 0);

	spin_unlock(&spool->lock);

	/* If no pages are used, and no other handles to the subpool
	 * remain, free the subpool the subpool remain */
	if (free)
		kfree(spool);
}

struct hugepage_subpool *hugepage_new_subpool(long nr_blocks)
{
	struct hugepage_subpool *spool;

	spool = kmalloc(sizeof(*spool), GFP_KERNEL);
	if (!spool)
		return NULL;

	spin_lock_init(&spool->lock);
	spool->count = 1;
	spool->max_hpages = nr_blocks;
	spool->used_hpages = 0;

	return spool;
}

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

static int hugepage_subpool_get_pages(struct hugepage_subpool *spool,
				      long delta)
{
	int ret = 0;

	if (!spool)
		return 0;

	spin_lock(&spool->lock);
	if ((spool->used_hpages + delta) <= spool->max_hpages) {
		spool->used_hpages += delta;
	} else {
		ret = -ENOMEM;
	}
	spin_unlock(&spool->lock);

	return ret;
}

static void hugepage_subpool_put_pages(struct hugepage_subpool *spool,
				       long delta)
{
	if (!spool)
		return;

	spin_lock(&spool->lock);
	spool->used_hpages -= delta;
	/* If hugetlbfs_put_super couldn't free spool due to
	* an outstanding quota reference, free it now. */
	unlock_or_release_subpool(spool);
}

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
130
	return subpool_inode(file_inode(vma->vm_file));
131 132
}

133 134 135
/*
 * Region tracking -- allows tracking of reservations and instantiated pages
 *                    across the pages in a mapping.
136 137
 *
 * The region data structures are protected by a combination of the mmap_sem
138
 * and the hugetlb_instantiation_mutex.  To access or modify a region the caller
139
 * must either hold the mmap_sem for write, or the mmap_sem for read and
140
 * the hugetlb_instantiation_mutex:
141
 *
142
 *	down_write(&mm->mmap_sem);
143
 * or
144 145
 *	down_read(&mm->mmap_sem);
 *	mutex_lock(&hugetlb_instantiation_mutex);
146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225
 */
struct file_region {
	struct list_head link;
	long from;
	long to;
};

static long region_add(struct list_head *head, long f, long t)
{
	struct file_region *rg, *nrg, *trg;

	/* Locate the region we are either in or before. */
	list_for_each_entry(rg, head, link)
		if (f <= rg->to)
			break;

	/* Round our left edge to the current segment if it encloses us. */
	if (f > rg->from)
		f = rg->from;

	/* Check for and consume any regions we now overlap with. */
	nrg = rg;
	list_for_each_entry_safe(rg, trg, rg->link.prev, link) {
		if (&rg->link == head)
			break;
		if (rg->from > t)
			break;

		/* If this area reaches higher then extend our area to
		 * include it completely.  If this is not the first area
		 * which we intend to reuse, free it. */
		if (rg->to > t)
			t = rg->to;
		if (rg != nrg) {
			list_del(&rg->link);
			kfree(rg);
		}
	}
	nrg->from = f;
	nrg->to = t;
	return 0;
}

static long region_chg(struct list_head *head, long f, long t)
{
	struct file_region *rg, *nrg;
	long chg = 0;

	/* Locate the region we are before or in. */
	list_for_each_entry(rg, head, link)
		if (f <= rg->to)
			break;

	/* If we are below the current region then a new region is required.
	 * Subtle, allocate a new region at the position but make it zero
	 * size such that we can guarantee to record the reservation. */
	if (&rg->link == head || t < rg->from) {
		nrg = kmalloc(sizeof(*nrg), GFP_KERNEL);
		if (!nrg)
			return -ENOMEM;
		nrg->from = f;
		nrg->to   = f;
		INIT_LIST_HEAD(&nrg->link);
		list_add(&nrg->link, rg->link.prev);

		return t - f;
	}

	/* Round our left edge to the current segment if it encloses us. */
	if (f > rg->from)
		f = rg->from;
	chg = t - f;

	/* Check for and consume any regions we now overlap with. */
	list_for_each_entry(rg, rg->link.prev, link) {
		if (&rg->link == head)
			break;
		if (rg->from > t)
			return chg;

L
Lucas De Marchi 已提交
226
		/* We overlap with this area, if it extends further than
227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267
		 * us then we must extend ourselves.  Account for its
		 * existing reservation. */
		if (rg->to > t) {
			chg += rg->to - t;
			t = rg->to;
		}
		chg -= rg->to - rg->from;
	}
	return chg;
}

static long region_truncate(struct list_head *head, long end)
{
	struct file_region *rg, *trg;
	long chg = 0;

	/* Locate the region we are either in or before. */
	list_for_each_entry(rg, head, link)
		if (end <= rg->to)
			break;
	if (&rg->link == head)
		return 0;

	/* If we are in the middle of a region then adjust it. */
	if (end > rg->from) {
		chg = rg->to - end;
		rg->to = end;
		rg = list_entry(rg->link.next, typeof(*rg), link);
	}

	/* Drop any remaining regions. */
	list_for_each_entry_safe(rg, trg, rg->link.prev, link) {
		if (&rg->link == head)
			break;
		chg += rg->to - rg->from;
		list_del(&rg->link);
		kfree(rg);
	}
	return chg;
}

268 269 270 271 272 273 274
static long region_count(struct list_head *head, long f, long t)
{
	struct file_region *rg;
	long chg = 0;

	/* Locate each segment we overlap with, and count that overlap. */
	list_for_each_entry(rg, head, link) {
275 276
		long seg_from;
		long seg_to;
277 278 279 280 281 282 283 284 285 286 287 288 289 290 291

		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;
	}

	return chg;
}

292 293 294 295
/*
 * Convert the address within this vma to the page offset within
 * the mapping, in pagecache page units; huge pages here.
 */
296 297
static pgoff_t vma_hugecache_offset(struct hstate *h,
			struct vm_area_struct *vma, unsigned long address)
298
{
299 300
	return ((address - vma->vm_start) >> huge_page_shift(h)) +
			(vma->vm_pgoff >> huge_page_order(h));
301 302
}

303 304 305 306 307 308
pgoff_t linear_hugepage_index(struct vm_area_struct *vma,
				     unsigned long address)
{
	return vma_hugecache_offset(hstate_vma(vma), vma, address);
}

309 310 311 312 313 314 315 316 317 318 319 320 321
/*
 * 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)
{
	struct hstate *hstate;

	if (!is_vm_hugetlb_page(vma))
		return PAGE_SIZE;

	hstate = hstate_vma(vma);

322
	return 1UL << huge_page_shift(hstate);
323
}
324
EXPORT_SYMBOL_GPL(vma_kernel_pagesize);
325

326 327 328 329 330 331 332 333 334 335 336 337 338
/*
 * 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
 * architectures where it differs, an architecture-specific version of this
 * function is required.
 */
#ifndef vma_mmu_pagesize
unsigned long vma_mmu_pagesize(struct vm_area_struct *vma)
{
	return vma_kernel_pagesize(vma);
}
#endif

339 340 341 342 343 344 345
/*
 * 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)
346
#define HPAGE_RESV_MASK (HPAGE_RESV_OWNER | HPAGE_RESV_UNMAPPED)
347

348 349 350 351 352 353 354 355 356
/*
 * 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.
357 358 359 360 361 362 363 364 365
 *
 * 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.
366
 */
367 368 369 370 371 372 373 374 375 376 377
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;
}

378 379 380 381 382
struct resv_map {
	struct kref refs;
	struct list_head regions;
};

383
static struct resv_map *resv_map_alloc(void)
384 385 386 387 388 389 390 391 392 393 394
{
	struct resv_map *resv_map = kmalloc(sizeof(*resv_map), GFP_KERNEL);
	if (!resv_map)
		return NULL;

	kref_init(&resv_map->refs);
	INIT_LIST_HEAD(&resv_map->regions);

	return resv_map;
}

395
static void resv_map_release(struct kref *ref)
396 397 398 399 400 401 402 403 404
{
	struct resv_map *resv_map = container_of(ref, struct resv_map, refs);

	/* Clear out any active regions before we release the map. */
	region_truncate(&resv_map->regions, 0);
	kfree(resv_map);
}

static struct resv_map *vma_resv_map(struct vm_area_struct *vma)
405 406
{
	VM_BUG_ON(!is_vm_hugetlb_page(vma));
407
	if (!(vma->vm_flags & VM_MAYSHARE))
408 409
		return (struct resv_map *)(get_vma_private_data(vma) &
							~HPAGE_RESV_MASK);
410
	return NULL;
411 412
}

413
static void set_vma_resv_map(struct vm_area_struct *vma, struct resv_map *map)
414 415
{
	VM_BUG_ON(!is_vm_hugetlb_page(vma));
416
	VM_BUG_ON(vma->vm_flags & VM_MAYSHARE);
417

418 419
	set_vma_private_data(vma, (get_vma_private_data(vma) &
				HPAGE_RESV_MASK) | (unsigned long)map);
420 421 422 423 424
}

static void set_vma_resv_flags(struct vm_area_struct *vma, unsigned long flags)
{
	VM_BUG_ON(!is_vm_hugetlb_page(vma));
425
	VM_BUG_ON(vma->vm_flags & VM_MAYSHARE);
426 427

	set_vma_private_data(vma, get_vma_private_data(vma) | flags);
428 429 430 431 432
}

static int is_vma_resv_set(struct vm_area_struct *vma, unsigned long flag)
{
	VM_BUG_ON(!is_vm_hugetlb_page(vma));
433 434

	return (get_vma_private_data(vma) & flag) != 0;
435 436
}

437
/* Reset counters to 0 and clear all HPAGE_RESV_* flags */
438 439 440
void reset_vma_resv_huge_pages(struct vm_area_struct *vma)
{
	VM_BUG_ON(!is_vm_hugetlb_page(vma));
441
	if (!(vma->vm_flags & VM_MAYSHARE))
442 443 444 445
		vma->vm_private_data = (void *)0;
}

/* Returns true if the VMA has associated reserve pages */
446
static int vma_has_reserves(struct vm_area_struct *vma, long chg)
447
{
448 449 450 451 452 453 454 455 456 457 458 459 460 461 462
	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)
			return 1;
		else
			return 0;
	}
463 464

	/* Shared mappings always use reserves */
465
	if (vma->vm_flags & VM_MAYSHARE)
466
		return 1;
467 468 469 470 471

	/*
	 * Only the process that called mmap() has reserves for
	 * private mappings.
	 */
472 473
	if (is_vma_resv_set(vma, HPAGE_RESV_OWNER))
		return 1;
474

475
	return 0;
476 477
}

478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511
static void copy_gigantic_page(struct page *dst, struct page *src)
{
	int i;
	struct hstate *h = page_hstate(src);
	struct page *dst_base = dst;
	struct page *src_base = src;

	for (i = 0; i < pages_per_huge_page(h); ) {
		cond_resched();
		copy_highpage(dst, src);

		i++;
		dst = mem_map_next(dst, dst_base, i);
		src = mem_map_next(src, src_base, i);
	}
}

void copy_huge_page(struct page *dst, struct page *src)
{
	int i;
	struct hstate *h = page_hstate(src);

	if (unlikely(pages_per_huge_page(h) > MAX_ORDER_NR_PAGES)) {
		copy_gigantic_page(dst, src);
		return;
	}

	might_sleep();
	for (i = 0; i < pages_per_huge_page(h); i++) {
		cond_resched();
		copy_highpage(dst + i, src + i);
	}
}

512
static void enqueue_huge_page(struct hstate *h, struct page *page)
L
Linus Torvalds 已提交
513 514
{
	int nid = page_to_nid(page);
515
	list_move(&page->lru, &h->hugepage_freelists[nid]);
516 517
	h->free_huge_pages++;
	h->free_huge_pages_node[nid]++;
L
Linus Torvalds 已提交
518 519
}

520 521 522 523 524 525 526
static struct page *dequeue_huge_page_node(struct hstate *h, int nid)
{
	struct page *page;

	if (list_empty(&h->hugepage_freelists[nid]))
		return NULL;
	page = list_entry(h->hugepage_freelists[nid].next, struct page, lru);
527
	list_move(&page->lru, &h->hugepage_activelist);
528
	set_page_refcounted(page);
529 530 531 532 533
	h->free_huge_pages--;
	h->free_huge_pages_node[nid]--;
	return page;
}

534 535
static struct page *dequeue_huge_page_vma(struct hstate *h,
				struct vm_area_struct *vma,
536 537
				unsigned long address, int avoid_reserve,
				long chg)
L
Linus Torvalds 已提交
538
{
539
	struct page *page = NULL;
540
	struct mempolicy *mpol;
541
	nodemask_t *nodemask;
542
	struct zonelist *zonelist;
543 544
	struct zone *zone;
	struct zoneref *z;
545
	unsigned int cpuset_mems_cookie;
L
Linus Torvalds 已提交
546

547 548 549 550 551
	/*
	 * 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
	 */
552
	if (!vma_has_reserves(vma, chg) &&
553
			h->free_huge_pages - h->resv_huge_pages == 0)
554
		goto err;
555

556
	/* If reserves cannot be used, ensure enough pages are in the pool */
557
	if (avoid_reserve && h->free_huge_pages - h->resv_huge_pages == 0)
558
		goto err;
559

560 561 562 563 564
retry_cpuset:
	cpuset_mems_cookie = get_mems_allowed();
	zonelist = huge_zonelist(vma, address,
					htlb_alloc_mask, &mpol, &nodemask);

565 566
	for_each_zone_zonelist_nodemask(zone, z, zonelist,
						MAX_NR_ZONES - 1, nodemask) {
567 568 569
		if (cpuset_zone_allowed_softwall(zone, htlb_alloc_mask)) {
			page = dequeue_huge_page_node(h, zone_to_nid(zone));
			if (page) {
570 571 572 573 574 575
				if (avoid_reserve)
					break;
				if (!vma_has_reserves(vma, chg))
					break;

				h->resv_huge_pages--;
576 577
				break;
			}
A
Andrew Morton 已提交
578
		}
L
Linus Torvalds 已提交
579
	}
580

581
	mpol_cond_put(mpol);
582 583
	if (unlikely(!put_mems_allowed(cpuset_mems_cookie) && !page))
		goto retry_cpuset;
L
Linus Torvalds 已提交
584
	return page;
585 586 587

err:
	return NULL;
L
Linus Torvalds 已提交
588 589
}

590
static void update_and_free_page(struct hstate *h, struct page *page)
A
Adam Litke 已提交
591 592
{
	int i;
593

594 595
	VM_BUG_ON(h->order >= MAX_ORDER);

596 597 598
	h->nr_huge_pages--;
	h->nr_huge_pages_node[page_to_nid(page)]--;
	for (i = 0; i < pages_per_huge_page(h); i++) {
599 600 601 602
		page[i].flags &= ~(1 << PG_locked | 1 << PG_error |
				1 << PG_referenced | 1 << PG_dirty |
				1 << PG_active | 1 << PG_reserved |
				1 << PG_private | 1 << PG_writeback);
A
Adam Litke 已提交
603
	}
604
	VM_BUG_ON(hugetlb_cgroup_from_page(page));
A
Adam Litke 已提交
605 606
	set_compound_page_dtor(page, NULL);
	set_page_refcounted(page);
607
	arch_release_hugepage(page);
608
	__free_pages(page, huge_page_order(h));
A
Adam Litke 已提交
609 610
}

611 612 613 614 615 616 617 618 619 620 621
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;
}

622 623
static void free_huge_page(struct page *page)
{
624 625 626 627
	/*
	 * Can't pass hstate in here because it is called from the
	 * compound page destructor.
	 */
628
	struct hstate *h = page_hstate(page);
629
	int nid = page_to_nid(page);
630 631
	struct hugepage_subpool *spool =
		(struct hugepage_subpool *)page_private(page);
632

633
	set_page_private(page, 0);
634
	page->mapping = NULL;
635
	BUG_ON(page_count(page));
636
	BUG_ON(page_mapcount(page));
637 638

	spin_lock(&hugetlb_lock);
639 640
	hugetlb_cgroup_uncharge_page(hstate_index(h),
				     pages_per_huge_page(h), page);
641
	if (h->surplus_huge_pages_node[nid] && huge_page_order(h) < MAX_ORDER) {
642 643
		/* remove the page from active list */
		list_del(&page->lru);
644 645 646
		update_and_free_page(h, page);
		h->surplus_huge_pages--;
		h->surplus_huge_pages_node[nid]--;
647
	} else {
648
		arch_clear_hugepage_flags(page);
649
		enqueue_huge_page(h, page);
650
	}
651
	spin_unlock(&hugetlb_lock);
652
	hugepage_subpool_put_pages(spool, 1);
653 654
}

655
static void prep_new_huge_page(struct hstate *h, struct page *page, int nid)
656
{
657
	INIT_LIST_HEAD(&page->lru);
658 659
	set_compound_page_dtor(page, free_huge_page);
	spin_lock(&hugetlb_lock);
660
	set_hugetlb_cgroup(page, NULL);
661 662
	h->nr_huge_pages++;
	h->nr_huge_pages_node[nid]++;
663 664 665 666
	spin_unlock(&hugetlb_lock);
	put_page(page); /* free it into the hugepage allocator */
}

667 668 669 670 671 672 673 674 675 676 677
static void prep_compound_gigantic_page(struct page *page, unsigned long order)
{
	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);
	__SetPageHead(page);
	for (i = 1; i < nr_pages; i++, p = mem_map_next(p, page, i)) {
		__SetPageTail(p);
678
		set_page_count(p, 0);
679 680 681 682
		p->first_page = page;
	}
}

A
Andrew Morton 已提交
683 684 685 686 687
/*
 * PageHuge() only returns true for hugetlbfs pages, but not for normal or
 * transparent huge pages.  See the PageTransHuge() documentation for more
 * details.
 */
688 689 690 691 692 693 694 695 696 697 698 699
int PageHuge(struct page *page)
{
	compound_page_dtor *dtor;

	if (!PageCompound(page))
		return 0;

	page = compound_head(page);
	dtor = get_compound_page_dtor(page);

	return dtor == free_huge_page;
}
700 701
EXPORT_SYMBOL_GPL(PageHuge);

702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718
pgoff_t __basepage_index(struct page *page)
{
	struct page *page_head = compound_head(page);
	pgoff_t index = page_index(page_head);
	unsigned long compound_idx;

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

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

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

719
static struct page *alloc_fresh_huge_page_node(struct hstate *h, int nid)
L
Linus Torvalds 已提交
720 721
{
	struct page *page;
722

723 724 725
	if (h->order >= MAX_ORDER)
		return NULL;

726
	page = alloc_pages_exact_node(nid,
727 728
		htlb_alloc_mask|__GFP_COMP|__GFP_THISNODE|
						__GFP_REPEAT|__GFP_NOWARN,
729
		huge_page_order(h));
L
Linus Torvalds 已提交
730
	if (page) {
731
		if (arch_prepare_hugepage(page)) {
732
			__free_pages(page, huge_page_order(h));
733
			return NULL;
734
		}
735
		prep_new_huge_page(h, page, nid);
L
Linus Torvalds 已提交
736
	}
737 738 739 740

	return page;
}

741
/*
742 743 744 745 746
 * 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.
747
 */
748
static int next_node_allowed(int nid, nodemask_t *nodes_allowed)
749
{
750
	nid = next_node(nid, *nodes_allowed);
751
	if (nid == MAX_NUMNODES)
752
		nid = first_node(*nodes_allowed);
753 754 755 756 757
	VM_BUG_ON(nid >= MAX_NUMNODES);

	return nid;
}

758 759 760 761 762 763 764
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;
}

765
/*
766 767 768 769
 * 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.
770
 */
771 772
static int hstate_next_node_to_alloc(struct hstate *h,
					nodemask_t *nodes_allowed)
773
{
774 775 776 777 778 779
	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);
780 781

	return nid;
782 783
}

784
/*
785 786 787 788
 * helper for free_pool_huge_page() - return the previously saved
 * node ["this node"] from which to free a huge page.  Advance the
 * next node id whether or not we find a free huge page to free so
 * that the next attempt to free addresses the next node.
789
 */
790
static int hstate_next_node_to_free(struct hstate *h, nodemask_t *nodes_allowed)
791
{
792 793 794 795 796 797
	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);
798 799

	return nid;
800 801
}

802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835
#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--)

static int alloc_fresh_huge_page(struct hstate *h, nodemask_t *nodes_allowed)
{
	struct page *page;
	int nr_nodes, node;
	int ret = 0;

	for_each_node_mask_to_alloc(h, nr_nodes, node, nodes_allowed) {
		page = alloc_fresh_huge_page_node(h, node);
		if (page) {
			ret = 1;
			break;
		}
	}

	if (ret)
		count_vm_event(HTLB_BUDDY_PGALLOC);
	else
		count_vm_event(HTLB_BUDDY_PGALLOC_FAIL);

	return ret;
}

836 837 838 839 840 841
/*
 * Free huge page from pool from next node to free.
 * Attempt to keep persistent huge pages more or less
 * balanced over allowed nodes.
 * Called with hugetlb_lock locked.
 */
842 843
static int free_pool_huge_page(struct hstate *h, nodemask_t *nodes_allowed,
							 bool acct_surplus)
844
{
845
	int nr_nodes, node;
846 847
	int ret = 0;

848
	for_each_node_mask_to_free(h, nr_nodes, node, nodes_allowed) {
849 850 851 852
		/*
		 * If we're returning unused surplus pages, only examine
		 * nodes with surplus pages.
		 */
853 854
		if ((!acct_surplus || h->surplus_huge_pages_node[node]) &&
		    !list_empty(&h->hugepage_freelists[node])) {
855
			struct page *page =
856
				list_entry(h->hugepage_freelists[node].next,
857 858 859
					  struct page, lru);
			list_del(&page->lru);
			h->free_huge_pages--;
860
			h->free_huge_pages_node[node]--;
861 862
			if (acct_surplus) {
				h->surplus_huge_pages--;
863
				h->surplus_huge_pages_node[node]--;
864
			}
865 866
			update_and_free_page(h, page);
			ret = 1;
867
			break;
868
		}
869
	}
870 871 872 873

	return ret;
}

874
static struct page *alloc_buddy_huge_page(struct hstate *h, int nid)
875 876
{
	struct page *page;
877
	unsigned int r_nid;
878

879 880 881
	if (h->order >= MAX_ORDER)
		return NULL;

882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905
	/*
	 * Assume we will successfully allocate the surplus page to
	 * prevent racing processes from causing the surplus to exceed
	 * overcommit
	 *
	 * This however introduces a different race, where a process B
	 * tries to grow the static hugepage pool while alloc_pages() is
	 * called by process A. B will only examine the per-node
	 * counters in determining if surplus huge pages can be
	 * converted to normal huge pages in adjust_pool_surplus(). A
	 * won't be able to increment the per-node counter, until the
	 * lock is dropped by B, but B doesn't drop hugetlb_lock until
	 * no more huge pages can be converted from surplus to normal
	 * state (and doesn't try to convert again). Thus, we have a
	 * case where a surplus huge page exists, the pool is grown, and
	 * the surplus huge page still exists after, even though it
	 * should just have been converted to a normal huge page. This
	 * does not leak memory, though, as the hugepage will be freed
	 * once it is out of use. It also does not allow the counters to
	 * go out of whack in adjust_pool_surplus() as we don't modify
	 * the node values until we've gotten the hugepage and only the
	 * per-node value is checked there.
	 */
	spin_lock(&hugetlb_lock);
906
	if (h->surplus_huge_pages >= h->nr_overcommit_huge_pages) {
907 908 909
		spin_unlock(&hugetlb_lock);
		return NULL;
	} else {
910 911
		h->nr_huge_pages++;
		h->surplus_huge_pages++;
912 913 914
	}
	spin_unlock(&hugetlb_lock);

915 916 917 918 919 920 921 922
	if (nid == NUMA_NO_NODE)
		page = alloc_pages(htlb_alloc_mask|__GFP_COMP|
				   __GFP_REPEAT|__GFP_NOWARN,
				   huge_page_order(h));
	else
		page = alloc_pages_exact_node(nid,
			htlb_alloc_mask|__GFP_COMP|__GFP_THISNODE|
			__GFP_REPEAT|__GFP_NOWARN, huge_page_order(h));
923

924 925
	if (page && arch_prepare_hugepage(page)) {
		__free_pages(page, huge_page_order(h));
926
		page = NULL;
927 928
	}

929
	spin_lock(&hugetlb_lock);
930
	if (page) {
931
		INIT_LIST_HEAD(&page->lru);
932
		r_nid = page_to_nid(page);
933
		set_compound_page_dtor(page, free_huge_page);
934
		set_hugetlb_cgroup(page, NULL);
935 936 937
		/*
		 * We incremented the global counters already
		 */
938 939
		h->nr_huge_pages_node[r_nid]++;
		h->surplus_huge_pages_node[r_nid]++;
940
		__count_vm_event(HTLB_BUDDY_PGALLOC);
941
	} else {
942 943
		h->nr_huge_pages--;
		h->surplus_huge_pages--;
944
		__count_vm_event(HTLB_BUDDY_PGALLOC_FAIL);
945
	}
946
	spin_unlock(&hugetlb_lock);
947 948 949 950

	return page;
}

951 952 953 954 955 956 957 958 959 960 961 962 963
/*
 * This allocation function is useful in the context where vma is irrelevant.
 * E.g. soft-offlining uses this function because it only cares physical
 * address of error page.
 */
struct page *alloc_huge_page_node(struct hstate *h, int nid)
{
	struct page *page;

	spin_lock(&hugetlb_lock);
	page = dequeue_huge_page_node(h, nid);
	spin_unlock(&hugetlb_lock);

964
	if (!page)
965 966 967 968 969
		page = alloc_buddy_huge_page(h, nid);

	return page;
}

970
/*
L
Lucas De Marchi 已提交
971
 * Increase the hugetlb pool such that it can accommodate a reservation
972 973
 * of size 'delta'.
 */
974
static int gather_surplus_pages(struct hstate *h, int delta)
975 976 977 978 979
{
	struct list_head surplus_list;
	struct page *page, *tmp;
	int ret, i;
	int needed, allocated;
980
	bool alloc_ok = true;
981

982
	needed = (h->resv_huge_pages + delta) - h->free_huge_pages;
983
	if (needed <= 0) {
984
		h->resv_huge_pages += delta;
985
		return 0;
986
	}
987 988 989 990 991 992 993 994

	allocated = 0;
	INIT_LIST_HEAD(&surplus_list);

	ret = -ENOMEM;
retry:
	spin_unlock(&hugetlb_lock);
	for (i = 0; i < needed; i++) {
995
		page = alloc_buddy_huge_page(h, NUMA_NO_NODE);
996 997 998 999
		if (!page) {
			alloc_ok = false;
			break;
		}
1000 1001
		list_add(&page->lru, &surplus_list);
	}
1002
	allocated += i;
1003 1004 1005 1006 1007 1008

	/*
	 * After retaking hugetlb_lock, we need to recalculate 'needed'
	 * because either resv_huge_pages or free_huge_pages may have changed.
	 */
	spin_lock(&hugetlb_lock);
1009 1010
	needed = (h->resv_huge_pages + delta) -
			(h->free_huge_pages + allocated);
1011 1012 1013 1014 1015 1016 1017 1018 1019 1020
	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;
	}
1021 1022
	/*
	 * The surplus_list now contains _at_least_ the number of extra pages
L
Lucas De Marchi 已提交
1023
	 * needed to accommodate the reservation.  Add the appropriate number
1024
	 * of pages to the hugetlb pool and free the extras back to the buddy
1025 1026 1027
	 * 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.
1028 1029
	 */
	needed += allocated;
1030
	h->resv_huge_pages += delta;
1031
	ret = 0;
1032

1033
	/* Free the needed pages to the hugetlb pool */
1034
	list_for_each_entry_safe(page, tmp, &surplus_list, lru) {
1035 1036
		if ((--needed) < 0)
			break;
1037 1038 1039 1040 1041 1042
		/*
		 * This page is now managed by the hugetlb allocator and has
		 * no users -- drop the buddy allocator's reference.
		 */
		put_page_testzero(page);
		VM_BUG_ON(page_count(page));
1043
		enqueue_huge_page(h, page);
1044
	}
1045
free:
1046
	spin_unlock(&hugetlb_lock);
1047 1048

	/* Free unnecessary surplus pages to the buddy allocator */
1049 1050
	list_for_each_entry_safe(page, tmp, &surplus_list, lru)
		put_page(page);
1051
	spin_lock(&hugetlb_lock);
1052 1053 1054 1055 1056 1057 1058 1059

	return ret;
}

/*
 * When releasing a hugetlb pool reservation, any surplus pages that were
 * allocated to satisfy the reservation must be explicitly freed if they were
 * never used.
1060
 * Called with hugetlb_lock held.
1061
 */
1062 1063
static void return_unused_surplus_pages(struct hstate *h,
					unsigned long unused_resv_pages)
1064 1065 1066
{
	unsigned long nr_pages;

1067
	/* Uncommit the reservation */
1068
	h->resv_huge_pages -= unused_resv_pages;
1069

1070 1071 1072 1073
	/* Cannot return gigantic pages currently */
	if (h->order >= MAX_ORDER)
		return;

1074
	nr_pages = min(unused_resv_pages, h->surplus_huge_pages);
1075

1076 1077
	/*
	 * We want to release as many surplus pages as possible, spread
1078 1079 1080 1081 1082
	 * 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.
	 * free_pool_huge_page() will balance the the freed pages across the
	 * on-line nodes with memory and will handle the hstate accounting.
1083 1084
	 */
	while (nr_pages--) {
1085
		if (!free_pool_huge_page(h, &node_states[N_MEMORY], 1))
1086
			break;
1087 1088 1089
	}
}

1090 1091 1092
/*
 * Determine if the huge page at addr within the vma has an associated
 * reservation.  Where it does not we will need to logically increase
1093 1094 1095 1096 1097 1098
 * reservation and actually increase subpool usage before an allocation
 * can occur.  Where any new reservation would be required the
 * reservation change is prepared, but not committed.  Once the page
 * has been allocated from the subpool and instantiated the change should
 * be committed via vma_commit_reservation.  No action is required on
 * failure.
1099
 */
1100
static long vma_needs_reservation(struct hstate *h,
1101
			struct vm_area_struct *vma, unsigned long addr)
1102 1103 1104 1105
{
	struct address_space *mapping = vma->vm_file->f_mapping;
	struct inode *inode = mapping->host;

1106
	if (vma->vm_flags & VM_MAYSHARE) {
1107
		pgoff_t idx = vma_hugecache_offset(h, vma, addr);
1108 1109 1110
		return region_chg(&inode->i_mapping->private_list,
							idx, idx + 1);

1111 1112
	} else if (!is_vma_resv_set(vma, HPAGE_RESV_OWNER)) {
		return 1;
1113

1114
	} else  {
1115
		long err;
1116
		pgoff_t idx = vma_hugecache_offset(h, vma, addr);
1117 1118 1119 1120 1121 1122 1123
		struct resv_map *reservations = vma_resv_map(vma);

		err = region_chg(&reservations->regions, idx, idx + 1);
		if (err < 0)
			return err;
		return 0;
	}
1124
}
1125 1126
static void vma_commit_reservation(struct hstate *h,
			struct vm_area_struct *vma, unsigned long addr)
1127 1128 1129 1130
{
	struct address_space *mapping = vma->vm_file->f_mapping;
	struct inode *inode = mapping->host;

1131
	if (vma->vm_flags & VM_MAYSHARE) {
1132
		pgoff_t idx = vma_hugecache_offset(h, vma, addr);
1133
		region_add(&inode->i_mapping->private_list, idx, idx + 1);
1134 1135

	} else if (is_vma_resv_set(vma, HPAGE_RESV_OWNER)) {
1136
		pgoff_t idx = vma_hugecache_offset(h, vma, addr);
1137 1138 1139 1140
		struct resv_map *reservations = vma_resv_map(vma);

		/* Mark this page used in the map. */
		region_add(&reservations->regions, idx, idx + 1);
1141 1142 1143
	}
}

1144
static struct page *alloc_huge_page(struct vm_area_struct *vma,
1145
				    unsigned long addr, int avoid_reserve)
L
Linus Torvalds 已提交
1146
{
1147
	struct hugepage_subpool *spool = subpool_vma(vma);
1148
	struct hstate *h = hstate_vma(vma);
1149
	struct page *page;
1150
	long chg;
1151 1152
	int ret, idx;
	struct hugetlb_cgroup *h_cg;
1153

1154
	idx = hstate_index(h);
1155
	/*
1156 1157 1158 1159 1160 1161
	 * Processes that did not create the mapping will have no
	 * reserves and will not have accounted against subpool
	 * limit. Check that the subpool limit can be made before
	 * satisfying the allocation MAP_NORESERVE mappings may also
	 * need pages and subpool limit allocated allocated if no reserve
	 * mapping overlaps.
1162
	 */
1163
	chg = vma_needs_reservation(h, vma, addr);
1164
	if (chg < 0)
1165
		return ERR_PTR(-ENOMEM);
1166
	if (chg)
1167
		if (hugepage_subpool_get_pages(spool, chg))
1168
			return ERR_PTR(-ENOSPC);
L
Linus Torvalds 已提交
1169

1170 1171 1172 1173 1174
	ret = hugetlb_cgroup_charge_cgroup(idx, pages_per_huge_page(h), &h_cg);
	if (ret) {
		hugepage_subpool_put_pages(spool, chg);
		return ERR_PTR(-ENOSPC);
	}
L
Linus Torvalds 已提交
1175
	spin_lock(&hugetlb_lock);
1176
	page = dequeue_huge_page_vma(h, vma, addr, avoid_reserve, chg);
1177
	if (!page) {
1178
		spin_unlock(&hugetlb_lock);
1179
		page = alloc_buddy_huge_page(h, NUMA_NO_NODE);
K
Ken Chen 已提交
1180
		if (!page) {
1181 1182 1183
			hugetlb_cgroup_uncharge_cgroup(idx,
						       pages_per_huge_page(h),
						       h_cg);
1184
			hugepage_subpool_put_pages(spool, chg);
1185
			return ERR_PTR(-ENOSPC);
K
Ken Chen 已提交
1186
		}
1187 1188
		spin_lock(&hugetlb_lock);
		list_move(&page->lru, &h->hugepage_activelist);
1189
		/* Fall through */
K
Ken Chen 已提交
1190
	}
1191 1192
	hugetlb_cgroup_commit_charge(idx, pages_per_huge_page(h), h_cg, page);
	spin_unlock(&hugetlb_lock);
1193

1194
	set_page_private(page, (unsigned long)spool);
1195

1196
	vma_commit_reservation(h, vma, addr);
1197
	return page;
1198 1199
}

1200
int __weak alloc_bootmem_huge_page(struct hstate *h)
1201 1202
{
	struct huge_bootmem_page *m;
1203
	int nr_nodes, node;
1204

1205
	for_each_node_mask_to_alloc(h, nr_nodes, node, &node_states[N_MEMORY]) {
1206 1207
		void *addr;

1208
		addr = __alloc_bootmem_node_nopanic(NODE_DATA(node),
1209 1210 1211 1212 1213 1214 1215 1216 1217
				huge_page_size(h), huge_page_size(h), 0);

		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;
1218
			goto found;
1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230
		}
	}
	return 0;

found:
	BUG_ON((unsigned long)virt_to_phys(m) & (huge_page_size(h) - 1));
	/* Put them into a private list first because mem_map is not up yet */
	list_add(&m->list, &huge_boot_pages);
	m->hstate = h;
	return 1;
}

1231 1232 1233 1234 1235 1236 1237 1238
static void prep_compound_huge_page(struct page *page, int order)
{
	if (unlikely(order > (MAX_ORDER - 1)))
		prep_compound_gigantic_page(page, order);
	else
		prep_compound_page(page, order);
}

1239 1240 1241 1242 1243 1244 1245
/* 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) {
		struct hstate *h = m->hstate;
1246 1247 1248 1249 1250 1251 1252 1253 1254
		struct page *page;

#ifdef CONFIG_HIGHMEM
		page = pfn_to_page(m->phys >> PAGE_SHIFT);
		free_bootmem_late((unsigned long)m,
				  sizeof(struct huge_bootmem_page));
#else
		page = virt_to_page(m);
#endif
1255 1256
		__ClearPageReserved(page);
		WARN_ON(page_count(page) != 1);
1257
		prep_compound_huge_page(page, h->order);
1258
		prep_new_huge_page(h, page, page_to_nid(page));
1259 1260 1261 1262 1263 1264 1265
		/*
		 * 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.
		 */
		if (h->order > (MAX_ORDER - 1))
1266
			adjust_managed_page_count(page, 1 << h->order);
1267 1268 1269
	}
}

1270
static void __init hugetlb_hstate_alloc_pages(struct hstate *h)
L
Linus Torvalds 已提交
1271 1272
{
	unsigned long i;
1273

1274
	for (i = 0; i < h->max_huge_pages; ++i) {
1275 1276 1277
		if (h->order >= MAX_ORDER) {
			if (!alloc_bootmem_huge_page(h))
				break;
1278
		} else if (!alloc_fresh_huge_page(h,
1279
					 &node_states[N_MEMORY]))
L
Linus Torvalds 已提交
1280 1281
			break;
	}
1282
	h->max_huge_pages = i;
1283 1284 1285 1286 1287 1288 1289
}

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

	for_each_hstate(h) {
1290 1291 1292
		/* oversize hugepages were init'ed in early boot */
		if (h->order < MAX_ORDER)
			hugetlb_hstate_alloc_pages(h);
1293 1294 1295
	}
}

A
Andi Kleen 已提交
1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306
static char * __init memfmt(char *buf, unsigned long n)
{
	if (n >= (1UL << 30))
		sprintf(buf, "%lu GB", n >> 30);
	else if (n >= (1UL << 20))
		sprintf(buf, "%lu MB", n >> 20);
	else
		sprintf(buf, "%lu KB", n >> 10);
	return buf;
}

1307 1308 1309 1310 1311
static void __init report_hugepages(void)
{
	struct hstate *h;

	for_each_hstate(h) {
A
Andi Kleen 已提交
1312
		char buf[32];
1313
		pr_info("HugeTLB registered %s page size, pre-allocated %ld pages\n",
A
Andi Kleen 已提交
1314 1315
			memfmt(buf, huge_page_size(h)),
			h->free_huge_pages);
1316 1317 1318
	}
}

L
Linus Torvalds 已提交
1319
#ifdef CONFIG_HIGHMEM
1320 1321
static void try_to_free_low(struct hstate *h, unsigned long count,
						nodemask_t *nodes_allowed)
L
Linus Torvalds 已提交
1322
{
1323 1324
	int i;

1325 1326 1327
	if (h->order >= MAX_ORDER)
		return;

1328
	for_each_node_mask(i, *nodes_allowed) {
L
Linus Torvalds 已提交
1329
		struct page *page, *next;
1330 1331 1332
		struct list_head *freel = &h->hugepage_freelists[i];
		list_for_each_entry_safe(page, next, freel, lru) {
			if (count >= h->nr_huge_pages)
1333
				return;
L
Linus Torvalds 已提交
1334 1335 1336
			if (PageHighMem(page))
				continue;
			list_del(&page->lru);
1337
			update_and_free_page(h, page);
1338 1339
			h->free_huge_pages--;
			h->free_huge_pages_node[page_to_nid(page)]--;
L
Linus Torvalds 已提交
1340 1341 1342 1343
		}
	}
}
#else
1344 1345
static inline void try_to_free_low(struct hstate *h, unsigned long count,
						nodemask_t *nodes_allowed)
L
Linus Torvalds 已提交
1346 1347 1348 1349
{
}
#endif

1350 1351 1352 1353 1354
/*
 * 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.
 */
1355 1356
static int adjust_pool_surplus(struct hstate *h, nodemask_t *nodes_allowed,
				int delta)
1357
{
1358
	int nr_nodes, node;
1359 1360 1361

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

1362 1363 1364 1365
	if (delta < 0) {
		for_each_node_mask_to_alloc(h, nr_nodes, node, nodes_allowed) {
			if (h->surplus_huge_pages_node[node])
				goto found;
1366
		}
1367 1368 1369 1370 1371
	} 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;
1372
		}
1373 1374
	}
	return 0;
1375

1376 1377 1378 1379
found:
	h->surplus_huge_pages += delta;
	h->surplus_huge_pages_node[node] += delta;
	return 1;
1380 1381
}

1382
#define persistent_huge_pages(h) (h->nr_huge_pages - h->surplus_huge_pages)
1383 1384
static unsigned long set_max_huge_pages(struct hstate *h, unsigned long count,
						nodemask_t *nodes_allowed)
L
Linus Torvalds 已提交
1385
{
1386
	unsigned long min_count, ret;
L
Linus Torvalds 已提交
1387

1388 1389 1390
	if (h->order >= MAX_ORDER)
		return h->max_huge_pages;

1391 1392 1393 1394
	/*
	 * Increase the pool size
	 * First take pages out of surplus state.  Then make up the
	 * remaining difference by allocating fresh huge pages.
1395 1396 1397 1398 1399 1400
	 *
	 * We might race with alloc_buddy_huge_page() here and be unable
	 * 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.
1401
	 */
L
Linus Torvalds 已提交
1402
	spin_lock(&hugetlb_lock);
1403
	while (h->surplus_huge_pages && count > persistent_huge_pages(h)) {
1404
		if (!adjust_pool_surplus(h, nodes_allowed, -1))
1405 1406 1407
			break;
	}

1408
	while (count > persistent_huge_pages(h)) {
1409 1410 1411 1412 1413 1414
		/*
		 * If this allocation races such that we no longer need the
		 * page, free_huge_page will handle it by freeing the page
		 * and reducing the surplus.
		 */
		spin_unlock(&hugetlb_lock);
1415
		ret = alloc_fresh_huge_page(h, nodes_allowed);
1416 1417 1418 1419
		spin_lock(&hugetlb_lock);
		if (!ret)
			goto out;

1420 1421 1422
		/* Bail for signals. Probably ctrl-c from user */
		if (signal_pending(current))
			goto out;
1423 1424 1425 1426 1427 1428 1429 1430
	}

	/*
	 * 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.
1431 1432 1433 1434 1435 1436 1437 1438
	 *
	 * 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
	 * alloc_buddy_huge_page() is checking the global counter,
	 * 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.
1439
	 */
1440
	min_count = h->resv_huge_pages + h->nr_huge_pages - h->free_huge_pages;
1441
	min_count = max(count, min_count);
1442
	try_to_free_low(h, min_count, nodes_allowed);
1443
	while (min_count < persistent_huge_pages(h)) {
1444
		if (!free_pool_huge_page(h, nodes_allowed, 0))
L
Linus Torvalds 已提交
1445 1446
			break;
	}
1447
	while (count < persistent_huge_pages(h)) {
1448
		if (!adjust_pool_surplus(h, nodes_allowed, 1))
1449 1450 1451
			break;
	}
out:
1452
	ret = persistent_huge_pages(h);
L
Linus Torvalds 已提交
1453
	spin_unlock(&hugetlb_lock);
1454
	return ret;
L
Linus Torvalds 已提交
1455 1456
}

1457 1458 1459 1460 1461 1462 1463 1464 1465 1466
#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];

1467 1468 1469
static struct hstate *kobj_to_node_hstate(struct kobject *kobj, int *nidp);

static struct hstate *kobj_to_hstate(struct kobject *kobj, int *nidp)
1470 1471
{
	int i;
1472

1473
	for (i = 0; i < HUGE_MAX_HSTATE; i++)
1474 1475 1476
		if (hstate_kobjs[i] == kobj) {
			if (nidp)
				*nidp = NUMA_NO_NODE;
1477
			return &hstates[i];
1478 1479 1480
		}

	return kobj_to_node_hstate(kobj, nidp);
1481 1482
}

1483
static ssize_t nr_hugepages_show_common(struct kobject *kobj,
1484 1485
					struct kobj_attribute *attr, char *buf)
{
1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496
	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);
1497
}
1498

1499 1500 1501
static ssize_t nr_hugepages_store_common(bool obey_mempolicy,
			struct kobject *kobj, struct kobj_attribute *attr,
			const char *buf, size_t len)
1502 1503
{
	int err;
1504
	int nid;
1505
	unsigned long count;
1506
	struct hstate *h;
1507
	NODEMASK_ALLOC(nodemask_t, nodes_allowed, GFP_KERNEL | __GFP_NORETRY);
1508

1509
	err = kstrtoul(buf, 10, &count);
1510
	if (err)
1511
		goto out;
1512

1513
	h = kobj_to_hstate(kobj, &nid);
1514 1515 1516 1517 1518
	if (h->order >= MAX_ORDER) {
		err = -EINVAL;
		goto out;
	}

1519 1520 1521 1522 1523 1524 1525
	if (nid == NUMA_NO_NODE) {
		/*
		 * global hstate attribute
		 */
		if (!(obey_mempolicy &&
				init_nodemask_of_mempolicy(nodes_allowed))) {
			NODEMASK_FREE(nodes_allowed);
1526
			nodes_allowed = &node_states[N_MEMORY];
1527 1528 1529 1530 1531 1532 1533 1534 1535
		}
	} else if (nodes_allowed) {
		/*
		 * per node hstate attribute: adjust count to global,
		 * but restrict alloc/free to the specified node.
		 */
		count += h->nr_huge_pages - h->nr_huge_pages_node[nid];
		init_nodemask_of_node(nodes_allowed, nid);
	} else
1536
		nodes_allowed = &node_states[N_MEMORY];
1537

1538
	h->max_huge_pages = set_max_huge_pages(h, count, nodes_allowed);
1539

1540
	if (nodes_allowed != &node_states[N_MEMORY])
1541 1542 1543
		NODEMASK_FREE(nodes_allowed);

	return len;
1544 1545 1546
out:
	NODEMASK_FREE(nodes_allowed);
	return err;
1547 1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558
}

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)
{
	return nr_hugepages_store_common(false, kobj, attr, buf, len);
1559 1560 1561
}
HSTATE_ATTR(nr_hugepages);

1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582
#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)
{
	return nr_hugepages_store_common(true, kobj, attr, buf, len);
}
HSTATE_ATTR(nr_hugepages_mempolicy);
#endif


1583 1584 1585
static ssize_t nr_overcommit_hugepages_show(struct kobject *kobj,
					struct kobj_attribute *attr, char *buf)
{
1586
	struct hstate *h = kobj_to_hstate(kobj, NULL);
1587 1588
	return sprintf(buf, "%lu\n", h->nr_overcommit_huge_pages);
}
1589

1590 1591 1592 1593 1594
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;
1595
	struct hstate *h = kobj_to_hstate(kobj, NULL);
1596

1597 1598 1599
	if (h->order >= MAX_ORDER)
		return -EINVAL;

1600
	err = kstrtoul(buf, 10, &input);
1601
	if (err)
1602
		return err;
1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614

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

	return count;
}
HSTATE_ATTR(nr_overcommit_hugepages);

static ssize_t free_hugepages_show(struct kobject *kobj,
					struct kobj_attribute *attr, char *buf)
{
1615 1616 1617 1618 1619 1620 1621 1622 1623 1624 1625
	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);
1626 1627 1628 1629 1630 1631
}
HSTATE_ATTR_RO(free_hugepages);

static ssize_t resv_hugepages_show(struct kobject *kobj,
					struct kobj_attribute *attr, char *buf)
{
1632
	struct hstate *h = kobj_to_hstate(kobj, NULL);
1633 1634 1635 1636 1637 1638 1639
	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)
{
1640 1641 1642 1643 1644 1645 1646 1647 1648 1649 1650
	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);
1651 1652 1653 1654 1655 1656 1657 1658 1659
}
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,
1660 1661 1662
#ifdef CONFIG_NUMA
	&nr_hugepages_mempolicy_attr.attr,
#endif
1663 1664 1665 1666 1667 1668 1669
	NULL,
};

static struct attribute_group hstate_attr_group = {
	.attrs = hstate_attrs,
};

J
Jeff Mahoney 已提交
1670 1671 1672
static int hugetlb_sysfs_add_hstate(struct hstate *h, struct kobject *parent,
				    struct kobject **hstate_kobjs,
				    struct attribute_group *hstate_attr_group)
1673 1674
{
	int retval;
1675
	int hi = hstate_index(h);
1676

1677 1678
	hstate_kobjs[hi] = kobject_create_and_add(h->name, parent);
	if (!hstate_kobjs[hi])
1679 1680
		return -ENOMEM;

1681
	retval = sysfs_create_group(hstate_kobjs[hi], hstate_attr_group);
1682
	if (retval)
1683
		kobject_put(hstate_kobjs[hi]);
1684 1685 1686 1687 1688 1689 1690 1691 1692 1693 1694 1695 1696 1697

	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) {
1698 1699
		err = hugetlb_sysfs_add_hstate(h, hugepages_kobj,
					 hstate_kobjs, &hstate_attr_group);
1700
		if (err)
1701
			pr_err("Hugetlb: Unable to add hstate %s", h->name);
1702 1703 1704
	}
}

1705 1706 1707 1708
#ifdef CONFIG_NUMA

/*
 * node_hstate/s - associate per node hstate attributes, via their kobjects,
1709 1710 1711
 * 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
1712 1713 1714 1715 1716 1717 1718 1719 1720
 * the base kernel, on the hugetlb module.
 */
struct node_hstate {
	struct kobject		*hugepages_kobj;
	struct kobject		*hstate_kobjs[HUGE_MAX_HSTATE];
};
struct node_hstate node_hstates[MAX_NUMNODES];

/*
1721
 * A subset of global hstate attributes for node devices
1722 1723 1724 1725 1726 1727 1728 1729 1730 1731 1732 1733 1734
 */
static struct attribute *per_node_hstate_attrs[] = {
	&nr_hugepages_attr.attr,
	&free_hugepages_attr.attr,
	&surplus_hugepages_attr.attr,
	NULL,
};

static struct attribute_group per_node_hstate_attr_group = {
	.attrs = per_node_hstate_attrs,
};

/*
1735
 * kobj_to_node_hstate - lookup global hstate for node device hstate attr kobj.
1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746 1747 1748 1749 1750 1751 1752 1753 1754 1755 1756 1757
 * 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;
}

/*
1758
 * Unregister hstate attributes from a single node device.
1759 1760
 * No-op if no hstate attributes attached.
 */
1761
static void hugetlb_unregister_node(struct node *node)
1762 1763
{
	struct hstate *h;
1764
	struct node_hstate *nhs = &node_hstates[node->dev.id];
1765 1766

	if (!nhs->hugepages_kobj)
1767
		return;		/* no hstate attributes */
1768

1769 1770 1771 1772 1773
	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;
1774
		}
1775
	}
1776 1777 1778 1779 1780 1781

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

/*
1782
 * hugetlb module exit:  unregister hstate attributes from node devices
1783 1784 1785 1786 1787 1788 1789
 * that have them.
 */
static void hugetlb_unregister_all_nodes(void)
{
	int nid;

	/*
1790
	 * disable node device registrations.
1791 1792 1793 1794 1795 1796 1797
	 */
	register_hugetlbfs_with_node(NULL, NULL);

	/*
	 * remove hstate attributes from any nodes that have them.
	 */
	for (nid = 0; nid < nr_node_ids; nid++)
1798
		hugetlb_unregister_node(node_devices[nid]);
1799 1800 1801
}

/*
1802
 * Register hstate attributes for a single node device.
1803 1804
 * No-op if attributes already registered.
 */
1805
static void hugetlb_register_node(struct node *node)
1806 1807
{
	struct hstate *h;
1808
	struct node_hstate *nhs = &node_hstates[node->dev.id];
1809 1810 1811 1812 1813 1814
	int err;

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

	nhs->hugepages_kobj = kobject_create_and_add("hugepages",
1815
							&node->dev.kobj);
1816 1817 1818 1819 1820 1821 1822 1823
	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) {
1824 1825
			pr_err("Hugetlb: Unable to add hstate %s for node %d\n",
				h->name, node->dev.id);
1826 1827 1828 1829 1830 1831 1832
			hugetlb_unregister_node(node);
			break;
		}
	}
}

/*
1833
 * hugetlb init time:  register hstate attributes for all registered node
1834 1835
 * devices of nodes that have memory.  All on-line nodes should have
 * registered their associated device by this time.
1836 1837 1838 1839 1840
 */
static void hugetlb_register_all_nodes(void)
{
	int nid;

1841
	for_each_node_state(nid, N_MEMORY) {
1842
		struct node *node = node_devices[nid];
1843
		if (node->dev.id == nid)
1844 1845 1846 1847
			hugetlb_register_node(node);
	}

	/*
1848
	 * Let the node device driver know we're here so it can
1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866 1867 1868 1869
	 * [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_unregister_all_nodes(void) { }

static void hugetlb_register_all_nodes(void) { }

#endif

1870 1871 1872 1873
static void __exit hugetlb_exit(void)
{
	struct hstate *h;

1874 1875
	hugetlb_unregister_all_nodes();

1876
	for_each_hstate(h) {
1877
		kobject_put(hstate_kobjs[hstate_index(h)]);
1878 1879 1880 1881 1882 1883 1884 1885
	}

	kobject_put(hugepages_kobj);
}
module_exit(hugetlb_exit);

static int __init hugetlb_init(void)
{
1886 1887 1888 1889 1890 1891
	/* Some platform decide whether they support huge pages at boot
	 * time. On these, such as powerpc, HPAGE_SHIFT is set to 0 when
	 * there is no such support
	 */
	if (HPAGE_SHIFT == 0)
		return 0;
1892

1893 1894 1895 1896
	if (!size_to_hstate(default_hstate_size)) {
		default_hstate_size = HPAGE_SIZE;
		if (!size_to_hstate(default_hstate_size))
			hugetlb_add_hstate(HUGETLB_PAGE_ORDER);
1897
	}
1898
	default_hstate_idx = hstate_index(size_to_hstate(default_hstate_size));
1899 1900
	if (default_hstate_max_huge_pages)
		default_hstate.max_huge_pages = default_hstate_max_huge_pages;
1901 1902

	hugetlb_init_hstates();
1903
	gather_bootmem_prealloc();
1904 1905 1906
	report_hugepages();

	hugetlb_sysfs_init();
1907
	hugetlb_register_all_nodes();
1908
	hugetlb_cgroup_file_init();
1909

1910 1911 1912 1913 1914 1915 1916 1917
	return 0;
}
module_init(hugetlb_init);

/* Should be called on processing a hugepagesz=... option */
void __init hugetlb_add_hstate(unsigned order)
{
	struct hstate *h;
1918 1919
	unsigned long i;

1920
	if (size_to_hstate(PAGE_SIZE << order)) {
1921
		pr_warning("hugepagesz= specified twice, ignoring\n");
1922 1923
		return;
	}
1924
	BUG_ON(hugetlb_max_hstate >= HUGE_MAX_HSTATE);
1925
	BUG_ON(order == 0);
1926
	h = &hstates[hugetlb_max_hstate++];
1927 1928
	h->order = order;
	h->mask = ~((1ULL << (order + PAGE_SHIFT)) - 1);
1929 1930 1931 1932
	h->nr_huge_pages = 0;
	h->free_huge_pages = 0;
	for (i = 0; i < MAX_NUMNODES; ++i)
		INIT_LIST_HEAD(&h->hugepage_freelists[i]);
1933
	INIT_LIST_HEAD(&h->hugepage_activelist);
1934 1935
	h->next_nid_to_alloc = first_node(node_states[N_MEMORY]);
	h->next_nid_to_free = first_node(node_states[N_MEMORY]);
1936 1937
	snprintf(h->name, HSTATE_NAME_LEN, "hugepages-%lukB",
					huge_page_size(h)/1024);
1938

1939 1940 1941
	parsed_hstate = h;
}

1942
static int __init hugetlb_nrpages_setup(char *s)
1943 1944
{
	unsigned long *mhp;
1945
	static unsigned long *last_mhp;
1946 1947

	/*
1948
	 * !hugetlb_max_hstate means we haven't parsed a hugepagesz= parameter yet,
1949 1950
	 * so this hugepages= parameter goes to the "default hstate".
	 */
1951
	if (!hugetlb_max_hstate)
1952 1953 1954 1955
		mhp = &default_hstate_max_huge_pages;
	else
		mhp = &parsed_hstate->max_huge_pages;

1956
	if (mhp == last_mhp) {
1957 1958
		pr_warning("hugepages= specified twice without "
			   "interleaving hugepagesz=, ignoring\n");
1959 1960 1961
		return 1;
	}

1962 1963 1964
	if (sscanf(s, "%lu", mhp) <= 0)
		*mhp = 0;

1965 1966 1967 1968 1969
	/*
	 * 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.
	 */
1970
	if (hugetlb_max_hstate && parsed_hstate->order >= MAX_ORDER)
1971 1972 1973 1974
		hugetlb_hstate_alloc_pages(parsed_hstate);

	last_mhp = mhp;

1975 1976
	return 1;
}
1977 1978 1979 1980 1981 1982 1983 1984
__setup("hugepages=", hugetlb_nrpages_setup);

static int __init hugetlb_default_setup(char *s)
{
	default_hstate_size = memparse(s, &s);
	return 1;
}
__setup("default_hugepagesz=", hugetlb_default_setup);
1985

1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997
static unsigned int cpuset_mems_nr(unsigned int *array)
{
	int node;
	unsigned int nr = 0;

	for_each_node_mask(node, cpuset_current_mems_allowed)
		nr += array[node];

	return nr;
}

#ifdef CONFIG_SYSCTL
1998 1999 2000
static int hugetlb_sysctl_handler_common(bool obey_mempolicy,
			 struct ctl_table *table, int write,
			 void __user *buffer, size_t *length, loff_t *ppos)
L
Linus Torvalds 已提交
2001
{
2002 2003
	struct hstate *h = &default_hstate;
	unsigned long tmp;
2004
	int ret;
2005

2006
	tmp = h->max_huge_pages;
2007

2008 2009 2010
	if (write && h->order >= MAX_ORDER)
		return -EINVAL;

2011 2012
	table->data = &tmp;
	table->maxlen = sizeof(unsigned long);
2013 2014 2015
	ret = proc_doulongvec_minmax(table, write, buffer, length, ppos);
	if (ret)
		goto out;
2016

2017
	if (write) {
2018 2019
		NODEMASK_ALLOC(nodemask_t, nodes_allowed,
						GFP_KERNEL | __GFP_NORETRY);
2020 2021 2022
		if (!(obey_mempolicy &&
			       init_nodemask_of_mempolicy(nodes_allowed))) {
			NODEMASK_FREE(nodes_allowed);
2023
			nodes_allowed = &node_states[N_MEMORY];
2024 2025 2026
		}
		h->max_huge_pages = set_max_huge_pages(h, tmp, nodes_allowed);

2027
		if (nodes_allowed != &node_states[N_MEMORY])
2028 2029
			NODEMASK_FREE(nodes_allowed);
	}
2030 2031
out:
	return ret;
L
Linus Torvalds 已提交
2032
}
2033

2034 2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046 2047 2048 2049 2050
int hugetlb_sysctl_handler(struct ctl_table *table, int write,
			  void __user *buffer, size_t *length, loff_t *ppos)
{

	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,
			  void __user *buffer, size_t *length, loff_t *ppos)
{
	return hugetlb_sysctl_handler_common(true, table, write,
							buffer, length, ppos);
}
#endif /* CONFIG_NUMA */

2051
int hugetlb_treat_movable_handler(struct ctl_table *table, int write,
2052
			void __user *buffer,
2053 2054
			size_t *length, loff_t *ppos)
{
2055
	proc_dointvec(table, write, buffer, length, ppos);
2056 2057 2058 2059 2060 2061 2062
	if (hugepages_treat_as_movable)
		htlb_alloc_mask = GFP_HIGHUSER_MOVABLE;
	else
		htlb_alloc_mask = GFP_HIGHUSER;
	return 0;
}

2063
int hugetlb_overcommit_handler(struct ctl_table *table, int write,
2064
			void __user *buffer,
2065 2066
			size_t *length, loff_t *ppos)
{
2067
	struct hstate *h = &default_hstate;
2068
	unsigned long tmp;
2069
	int ret;
2070

2071
	tmp = h->nr_overcommit_huge_pages;
2072

2073 2074 2075
	if (write && h->order >= MAX_ORDER)
		return -EINVAL;

2076 2077
	table->data = &tmp;
	table->maxlen = sizeof(unsigned long);
2078 2079 2080
	ret = proc_doulongvec_minmax(table, write, buffer, length, ppos);
	if (ret)
		goto out;
2081 2082 2083 2084 2085 2086

	if (write) {
		spin_lock(&hugetlb_lock);
		h->nr_overcommit_huge_pages = tmp;
		spin_unlock(&hugetlb_lock);
	}
2087 2088
out:
	return ret;
2089 2090
}

L
Linus Torvalds 已提交
2091 2092
#endif /* CONFIG_SYSCTL */

2093
void hugetlb_report_meminfo(struct seq_file *m)
L
Linus Torvalds 已提交
2094
{
2095
	struct hstate *h = &default_hstate;
2096
	seq_printf(m,
2097 2098 2099 2100 2101
			"HugePages_Total:   %5lu\n"
			"HugePages_Free:    %5lu\n"
			"HugePages_Rsvd:    %5lu\n"
			"HugePages_Surp:    %5lu\n"
			"Hugepagesize:   %8lu kB\n",
2102 2103 2104 2105 2106
			h->nr_huge_pages,
			h->free_huge_pages,
			h->resv_huge_pages,
			h->surplus_huge_pages,
			1UL << (huge_page_order(h) + PAGE_SHIFT - 10));
L
Linus Torvalds 已提交
2107 2108 2109 2110
}

int hugetlb_report_node_meminfo(int nid, char *buf)
{
2111
	struct hstate *h = &default_hstate;
L
Linus Torvalds 已提交
2112 2113
	return sprintf(buf,
		"Node %d HugePages_Total: %5u\n"
2114 2115
		"Node %d HugePages_Free:  %5u\n"
		"Node %d HugePages_Surp:  %5u\n",
2116 2117 2118
		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
2119 2120
}

2121 2122 2123 2124 2125 2126 2127 2128 2129 2130 2131 2132 2133 2134 2135
void hugetlb_show_meminfo(void)
{
	struct hstate *h;
	int nid;

	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));
}

L
Linus Torvalds 已提交
2136 2137 2138
/* Return the number pages of memory we physically have, in PAGE_SIZE units. */
unsigned long hugetlb_total_pages(void)
{
2139 2140 2141 2142 2143 2144
	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
2145 2146
}

2147
static int hugetlb_acct_memory(struct hstate *h, long delta)
M
Mel Gorman 已提交
2148 2149 2150 2151 2152 2153 2154 2155 2156 2157 2158 2159 2160 2161 2162 2163 2164 2165 2166 2167 2168 2169
{
	int ret = -ENOMEM;

	spin_lock(&hugetlb_lock);
	/*
	 * When cpuset is configured, it breaks the strict hugetlb page
	 * reservation as the accounting is done on a global variable. Such
	 * reservation is completely rubbish in the presence of cpuset because
	 * the reservation is not checked against page availability for the
	 * current cpuset. Application can still potentially OOM'ed by kernel
	 * with lack of free htlb page in cpuset that the task is in.
	 * Attempt to enforce strict accounting with cpuset is almost
	 * impossible (or too ugly) because cpuset is too fluid that
	 * task or memory node can be dynamically moved between cpusets.
	 *
	 * The change of semantics for shared hugetlb mapping with cpuset is
	 * undesirable. However, in order to preserve some of the semantics,
	 * we fall back to check against current free page availability as
	 * a best attempt and hopefully to minimize the impact of changing
	 * semantics that cpuset has.
	 */
	if (delta > 0) {
2170
		if (gather_surplus_pages(h, delta) < 0)
M
Mel Gorman 已提交
2171 2172
			goto out;

2173 2174
		if (delta > cpuset_mems_nr(h->free_huge_pages_node)) {
			return_unused_surplus_pages(h, delta);
M
Mel Gorman 已提交
2175 2176 2177 2178 2179 2180
			goto out;
		}
	}

	ret = 0;
	if (delta < 0)
2181
		return_unused_surplus_pages(h, (unsigned long) -delta);
M
Mel Gorman 已提交
2182 2183 2184 2185 2186 2187

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

2188 2189 2190 2191 2192 2193 2194 2195
static void hugetlb_vm_op_open(struct vm_area_struct *vma)
{
	struct resv_map *reservations = vma_resv_map(vma);

	/*
	 * 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
2196
	 * has a reference to the reservation map it cannot disappear until
2197 2198 2199 2200 2201 2202 2203
	 * after this open call completes.  It is therefore safe to take a
	 * new reference here without additional locking.
	 */
	if (reservations)
		kref_get(&reservations->refs);
}

2204 2205 2206 2207 2208 2209 2210 2211 2212
static void resv_map_put(struct vm_area_struct *vma)
{
	struct resv_map *reservations = vma_resv_map(vma);

	if (!reservations)
		return;
	kref_put(&reservations->refs, resv_map_release);
}

2213 2214
static void hugetlb_vm_op_close(struct vm_area_struct *vma)
{
2215
	struct hstate *h = hstate_vma(vma);
2216
	struct resv_map *reservations = vma_resv_map(vma);
2217
	struct hugepage_subpool *spool = subpool_vma(vma);
2218 2219 2220 2221 2222
	unsigned long reserve;
	unsigned long start;
	unsigned long end;

	if (reservations) {
2223 2224
		start = vma_hugecache_offset(h, vma, vma->vm_start);
		end = vma_hugecache_offset(h, vma, vma->vm_end);
2225 2226 2227 2228

		reserve = (end - start) -
			region_count(&reservations->regions, start, end);

2229
		resv_map_put(vma);
2230

2231
		if (reserve) {
2232
			hugetlb_acct_memory(h, -reserve);
2233
			hugepage_subpool_put_pages(spool, reserve);
2234
		}
2235
	}
2236 2237
}

L
Linus Torvalds 已提交
2238 2239 2240 2241 2242 2243
/*
 * 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.
 */
N
Nick Piggin 已提交
2244
static int hugetlb_vm_op_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
L
Linus Torvalds 已提交
2245 2246
{
	BUG();
N
Nick Piggin 已提交
2247
	return 0;
L
Linus Torvalds 已提交
2248 2249
}

2250
const struct vm_operations_struct hugetlb_vm_ops = {
N
Nick Piggin 已提交
2251
	.fault = hugetlb_vm_op_fault,
2252
	.open = hugetlb_vm_op_open,
2253
	.close = hugetlb_vm_op_close,
L
Linus Torvalds 已提交
2254 2255
};

2256 2257
static pte_t make_huge_pte(struct vm_area_struct *vma, struct page *page,
				int writable)
D
David Gibson 已提交
2258 2259 2260
{
	pte_t entry;

2261
	if (writable) {
2262 2263
		entry = huge_pte_mkwrite(huge_pte_mkdirty(mk_huge_pte(page,
					 vma->vm_page_prot)));
D
David Gibson 已提交
2264
	} else {
2265 2266
		entry = huge_pte_wrprotect(mk_huge_pte(page,
					   vma->vm_page_prot));
D
David Gibson 已提交
2267 2268 2269
	}
	entry = pte_mkyoung(entry);
	entry = pte_mkhuge(entry);
2270
	entry = arch_make_huge_pte(entry, vma, page, writable);
D
David Gibson 已提交
2271 2272 2273 2274

	return entry;
}

2275 2276 2277 2278 2279
static void set_huge_ptep_writable(struct vm_area_struct *vma,
				   unsigned long address, pte_t *ptep)
{
	pte_t entry;

2280
	entry = huge_pte_mkwrite(huge_pte_mkdirty(huge_ptep_get(ptep)));
2281
	if (huge_ptep_set_access_flags(vma, address, ptep, entry, 1))
2282
		update_mmu_cache(vma, address, ptep);
2283 2284 2285
}


D
David Gibson 已提交
2286 2287 2288 2289 2290
int copy_hugetlb_page_range(struct mm_struct *dst, struct mm_struct *src,
			    struct vm_area_struct *vma)
{
	pte_t *src_pte, *dst_pte, entry;
	struct page *ptepage;
2291
	unsigned long addr;
2292
	int cow;
2293 2294
	struct hstate *h = hstate_vma(vma);
	unsigned long sz = huge_page_size(h);
2295 2296

	cow = (vma->vm_flags & (VM_SHARED | VM_MAYWRITE)) == VM_MAYWRITE;
D
David Gibson 已提交
2297

2298
	for (addr = vma->vm_start; addr < vma->vm_end; addr += sz) {
H
Hugh Dickins 已提交
2299 2300 2301
		src_pte = huge_pte_offset(src, addr);
		if (!src_pte)
			continue;
2302
		dst_pte = huge_pte_alloc(dst, addr, sz);
D
David Gibson 已提交
2303 2304
		if (!dst_pte)
			goto nomem;
2305 2306 2307 2308 2309

		/* If the pagetables are shared don't copy or take references */
		if (dst_pte == src_pte)
			continue;

H
Hugh Dickins 已提交
2310
		spin_lock(&dst->page_table_lock);
N
Nick Piggin 已提交
2311
		spin_lock_nested(&src->page_table_lock, SINGLE_DEPTH_NESTING);
2312
		if (!huge_pte_none(huge_ptep_get(src_pte))) {
2313
			if (cow)
2314 2315
				huge_ptep_set_wrprotect(src, addr, src_pte);
			entry = huge_ptep_get(src_pte);
2316 2317
			ptepage = pte_page(entry);
			get_page(ptepage);
2318
			page_dup_rmap(ptepage);
2319 2320 2321
			set_huge_pte_at(dst, addr, dst_pte, entry);
		}
		spin_unlock(&src->page_table_lock);
H
Hugh Dickins 已提交
2322
		spin_unlock(&dst->page_table_lock);
D
David Gibson 已提交
2323 2324 2325 2326 2327 2328 2329
	}
	return 0;

nomem:
	return -ENOMEM;
}

N
Naoya Horiguchi 已提交
2330 2331 2332 2333 2334 2335 2336
static int is_hugetlb_entry_migration(pte_t pte)
{
	swp_entry_t swp;

	if (huge_pte_none(pte) || pte_present(pte))
		return 0;
	swp = pte_to_swp_entry(pte);
2337
	if (non_swap_entry(swp) && is_migration_entry(swp))
N
Naoya Horiguchi 已提交
2338
		return 1;
2339
	else
N
Naoya Horiguchi 已提交
2340 2341 2342
		return 0;
}

2343 2344 2345 2346 2347 2348 2349
static int is_hugetlb_entry_hwpoisoned(pte_t pte)
{
	swp_entry_t swp;

	if (huge_pte_none(pte) || pte_present(pte))
		return 0;
	swp = pte_to_swp_entry(pte);
2350
	if (non_swap_entry(swp) && is_hwpoison_entry(swp))
2351
		return 1;
2352
	else
2353 2354 2355
		return 0;
}

2356 2357 2358
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
2359
{
2360
	int force_flush = 0;
D
David Gibson 已提交
2361 2362
	struct mm_struct *mm = vma->vm_mm;
	unsigned long address;
2363
	pte_t *ptep;
D
David Gibson 已提交
2364 2365
	pte_t pte;
	struct page *page;
2366 2367
	struct hstate *h = hstate_vma(vma);
	unsigned long sz = huge_page_size(h);
2368 2369
	const unsigned long mmun_start = start;	/* For mmu_notifiers */
	const unsigned long mmun_end   = end;	/* For mmu_notifiers */
2370

D
David Gibson 已提交
2371
	WARN_ON(!is_vm_hugetlb_page(vma));
2372 2373
	BUG_ON(start & ~huge_page_mask(h));
	BUG_ON(end & ~huge_page_mask(h));
D
David Gibson 已提交
2374

2375
	tlb_start_vma(tlb, vma);
2376
	mmu_notifier_invalidate_range_start(mm, mmun_start, mmun_end);
2377
again:
2378
	spin_lock(&mm->page_table_lock);
2379
	for (address = start; address < end; address += sz) {
2380
		ptep = huge_pte_offset(mm, address);
A
Adam Litke 已提交
2381
		if (!ptep)
2382 2383
			continue;

2384 2385 2386
		if (huge_pmd_unshare(mm, &address, ptep))
			continue;

2387 2388 2389 2390 2391 2392 2393
		pte = huge_ptep_get(ptep);
		if (huge_pte_none(pte))
			continue;

		/*
		 * HWPoisoned hugepage is already unmapped and dropped reference
		 */
2394
		if (unlikely(is_hugetlb_entry_hwpoisoned(pte))) {
2395
			huge_pte_clear(mm, address, ptep);
2396
			continue;
2397
		}
2398 2399

		page = pte_page(pte);
2400 2401 2402 2403 2404 2405 2406 2407 2408 2409 2410 2411 2412 2413 2414 2415 2416
		/*
		 * 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) {
			if (page != ref_page)
				continue;

			/*
			 * 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);
		}

2417
		pte = huge_ptep_get_and_clear(mm, address, ptep);
2418
		tlb_remove_tlb_entry(tlb, ptep, address);
2419
		if (huge_pte_dirty(pte))
2420
			set_page_dirty(page);
2421

2422 2423 2424 2425
		page_remove_rmap(page);
		force_flush = !__tlb_remove_page(tlb, page);
		if (force_flush)
			break;
2426 2427 2428
		/* Bail out after unmapping reference page if supplied */
		if (ref_page)
			break;
D
David Gibson 已提交
2429
	}
2430
	spin_unlock(&mm->page_table_lock);
2431 2432 2433 2434 2435 2436 2437 2438 2439 2440
	/*
	 * mmu_gather ran out of room to batch pages, we break out of
	 * the PTE lock to avoid doing the potential expensive TLB invalidate
	 * and page-free while holding it.
	 */
	if (force_flush) {
		force_flush = 0;
		tlb_flush_mmu(tlb);
		if (address < end && !ref_page)
			goto again;
2441
	}
2442
	mmu_notifier_invalidate_range_end(mm, mmun_start, mmun_end);
2443
	tlb_end_vma(tlb, vma);
L
Linus Torvalds 已提交
2444
}
D
David Gibson 已提交
2445

2446 2447 2448 2449 2450 2451 2452 2453 2454 2455 2456 2457 2458 2459 2460 2461 2462 2463 2464
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
	 * is to clear it before releasing the i_mmap_mutex. This works
	 * because in the context this is called, the VMA is about to be
	 * destroyed and the i_mmap_mutex is held.
	 */
	vma->vm_flags &= ~VM_MAYSHARE;
}

2465
void unmap_hugepage_range(struct vm_area_struct *vma, unsigned long start,
2466
			  unsigned long end, struct page *ref_page)
2467
{
2468 2469 2470 2471 2472
	struct mm_struct *mm;
	struct mmu_gather tlb;

	mm = vma->vm_mm;

2473
	tlb_gather_mmu(&tlb, mm, start, end);
2474 2475
	__unmap_hugepage_range(&tlb, vma, start, end, ref_page);
	tlb_finish_mmu(&tlb, start, end);
2476 2477
}

2478 2479 2480 2481 2482 2483
/*
 * 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.
 */
2484 2485
static int unmap_ref_private(struct mm_struct *mm, struct vm_area_struct *vma,
				struct page *page, unsigned long address)
2486
{
2487
	struct hstate *h = hstate_vma(vma);
2488 2489 2490 2491 2492 2493 2494 2495
	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.
	 */
2496
	address = address & huge_page_mask(h);
2497 2498
	pgoff = ((address - vma->vm_start) >> PAGE_SHIFT) +
			vma->vm_pgoff;
A
Al Viro 已提交
2499
	mapping = file_inode(vma->vm_file)->i_mapping;
2500

2501 2502 2503 2504 2505
	/*
	 * 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
	 */
2506
	mutex_lock(&mapping->i_mmap_mutex);
2507
	vma_interval_tree_foreach(iter_vma, &mapping->i_mmap, pgoff, pgoff) {
2508 2509 2510 2511 2512 2513 2514 2515 2516 2517 2518 2519
		/* Do not unmap the current VMA */
		if (iter_vma == vma)
			continue;

		/*
		 * 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))
2520 2521
			unmap_hugepage_range(iter_vma, address,
					     address + huge_page_size(h), page);
2522
	}
2523
	mutex_unlock(&mapping->i_mmap_mutex);
2524 2525 2526 2527

	return 1;
}

2528 2529
/*
 * Hugetlb_cow() should be called with page lock of the original hugepage held.
2530 2531 2532
 * 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.
2533
 */
2534
static int hugetlb_cow(struct mm_struct *mm, struct vm_area_struct *vma,
2535 2536
			unsigned long address, pte_t *ptep, pte_t pte,
			struct page *pagecache_page)
2537
{
2538
	struct hstate *h = hstate_vma(vma);
2539
	struct page *old_page, *new_page;
2540
	int outside_reserve = 0;
2541 2542
	unsigned long mmun_start;	/* For mmu_notifiers */
	unsigned long mmun_end;		/* For mmu_notifiers */
2543 2544 2545

	old_page = pte_page(pte);

2546
retry_avoidcopy:
2547 2548
	/* If no-one else is actually using this page, avoid the copy
	 * and just make the page writable */
2549 2550
	if (page_mapcount(old_page) == 1 && PageAnon(old_page)) {
		page_move_anon_rmap(old_page, vma, address);
2551
		set_huge_ptep_writable(vma, address, ptep);
N
Nick Piggin 已提交
2552
		return 0;
2553 2554
	}

2555 2556 2557 2558 2559 2560 2561 2562 2563
	/*
	 * 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.
	 */
2564
	if (!(vma->vm_flags & VM_MAYSHARE) &&
2565 2566 2567 2568
			is_vma_resv_set(vma, HPAGE_RESV_OWNER) &&
			old_page != pagecache_page)
		outside_reserve = 1;

2569
	page_cache_get(old_page);
2570 2571 2572

	/* Drop page_table_lock as buddy allocator may be called */
	spin_unlock(&mm->page_table_lock);
2573
	new_page = alloc_huge_page(vma, address, outside_reserve);
2574

2575
	if (IS_ERR(new_page)) {
2576
		long err = PTR_ERR(new_page);
2577
		page_cache_release(old_page);
2578 2579 2580 2581 2582 2583 2584 2585 2586 2587 2588 2589

		/*
		 * 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) {
			BUG_ON(huge_pte_none(pte));
			if (unmap_ref_private(mm, vma, old_page, address)) {
				BUG_ON(huge_pte_none(pte));
2590
				spin_lock(&mm->page_table_lock);
2591 2592 2593 2594 2595 2596 2597 2598
				ptep = huge_pte_offset(mm, address & huge_page_mask(h));
				if (likely(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;
2599 2600 2601 2602
			}
			WARN_ON_ONCE(1);
		}

2603 2604
		/* Caller expects lock to be held */
		spin_lock(&mm->page_table_lock);
2605 2606 2607 2608
		if (err == -ENOMEM)
			return VM_FAULT_OOM;
		else
			return VM_FAULT_SIGBUS;
2609 2610
	}

2611 2612 2613 2614
	/*
	 * When the original hugepage is shared one, it does not have
	 * anon_vma prepared.
	 */
2615
	if (unlikely(anon_vma_prepare(vma))) {
2616 2617
		page_cache_release(new_page);
		page_cache_release(old_page);
2618 2619
		/* Caller expects lock to be held */
		spin_lock(&mm->page_table_lock);
2620
		return VM_FAULT_OOM;
2621
	}
2622

A
Andrea Arcangeli 已提交
2623 2624
	copy_user_huge_page(new_page, old_page, address, vma,
			    pages_per_huge_page(h));
N
Nick Piggin 已提交
2625
	__SetPageUptodate(new_page);
2626

2627 2628 2629
	mmun_start = address & huge_page_mask(h);
	mmun_end = mmun_start + huge_page_size(h);
	mmu_notifier_invalidate_range_start(mm, mmun_start, mmun_end);
2630 2631 2632 2633 2634
	/*
	 * Retake the page_table_lock to check for racing updates
	 * before the page tables are altered
	 */
	spin_lock(&mm->page_table_lock);
2635
	ptep = huge_pte_offset(mm, address & huge_page_mask(h));
2636
	if (likely(pte_same(huge_ptep_get(ptep), pte))) {
2637
		/* Break COW */
2638
		huge_ptep_clear_flush(vma, address, ptep);
2639 2640
		set_huge_pte_at(mm, address, ptep,
				make_huge_pte(vma, new_page, 1));
2641
		page_remove_rmap(old_page);
2642
		hugepage_add_new_anon_rmap(new_page, vma, address);
2643 2644 2645
		/* Make the old page be freed below */
		new_page = old_page;
	}
2646 2647 2648 2649
	spin_unlock(&mm->page_table_lock);
	mmu_notifier_invalidate_range_end(mm, mmun_start, mmun_end);
	/* Caller expects lock to be held */
	spin_lock(&mm->page_table_lock);
2650 2651
	page_cache_release(new_page);
	page_cache_release(old_page);
N
Nick Piggin 已提交
2652
	return 0;
2653 2654
}

2655
/* Return the pagecache page at a given address within a VMA */
2656 2657
static struct page *hugetlbfs_pagecache_page(struct hstate *h,
			struct vm_area_struct *vma, unsigned long address)
2658 2659
{
	struct address_space *mapping;
2660
	pgoff_t idx;
2661 2662

	mapping = vma->vm_file->f_mapping;
2663
	idx = vma_hugecache_offset(h, vma, address);
2664 2665 2666 2667

	return find_lock_page(mapping, idx);
}

H
Hugh Dickins 已提交
2668 2669 2670 2671 2672
/*
 * 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
2673 2674 2675 2676 2677 2678 2679 2680 2681 2682 2683 2684 2685 2686 2687
			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;
}

2688
static int hugetlb_no_page(struct mm_struct *mm, struct vm_area_struct *vma,
2689
			unsigned long address, pte_t *ptep, unsigned int flags)
2690
{
2691
	struct hstate *h = hstate_vma(vma);
2692
	int ret = VM_FAULT_SIGBUS;
2693
	int anon_rmap = 0;
2694
	pgoff_t idx;
A
Adam Litke 已提交
2695 2696 2697
	unsigned long size;
	struct page *page;
	struct address_space *mapping;
2698
	pte_t new_pte;
A
Adam Litke 已提交
2699

2700 2701 2702
	/*
	 * 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
2703
	 * COW. Warn that such a situation has occurred as it may not be obvious
2704 2705
	 */
	if (is_vma_resv_set(vma, HPAGE_RESV_UNMAPPED)) {
2706 2707
		pr_warning("PID %d killed due to inadequate hugepage pool\n",
			   current->pid);
2708 2709 2710
		return ret;
	}

A
Adam Litke 已提交
2711
	mapping = vma->vm_file->f_mapping;
2712
	idx = vma_hugecache_offset(h, vma, address);
A
Adam Litke 已提交
2713 2714 2715 2716 2717

	/*
	 * Use page lock to guard against racing truncation
	 * before we get page_table_lock.
	 */
2718 2719 2720
retry:
	page = find_lock_page(mapping, idx);
	if (!page) {
2721
		size = i_size_read(mapping->host) >> huge_page_shift(h);
2722 2723
		if (idx >= size)
			goto out;
2724
		page = alloc_huge_page(vma, address, 0);
2725
		if (IS_ERR(page)) {
2726 2727 2728 2729 2730
			ret = PTR_ERR(page);
			if (ret == -ENOMEM)
				ret = VM_FAULT_OOM;
			else
				ret = VM_FAULT_SIGBUS;
2731 2732
			goto out;
		}
A
Andrea Arcangeli 已提交
2733
		clear_huge_page(page, address, pages_per_huge_page(h));
N
Nick Piggin 已提交
2734
		__SetPageUptodate(page);
2735

2736
		if (vma->vm_flags & VM_MAYSHARE) {
2737
			int err;
K
Ken Chen 已提交
2738
			struct inode *inode = mapping->host;
2739 2740 2741 2742 2743 2744 2745 2746

			err = add_to_page_cache(page, mapping, idx, GFP_KERNEL);
			if (err) {
				put_page(page);
				if (err == -EEXIST)
					goto retry;
				goto out;
			}
K
Ken Chen 已提交
2747 2748

			spin_lock(&inode->i_lock);
2749
			inode->i_blocks += blocks_per_huge_page(h);
K
Ken Chen 已提交
2750
			spin_unlock(&inode->i_lock);
2751
		} else {
2752
			lock_page(page);
2753 2754 2755 2756
			if (unlikely(anon_vma_prepare(vma))) {
				ret = VM_FAULT_OOM;
				goto backout_unlocked;
			}
2757
			anon_rmap = 1;
2758
		}
2759
	} else {
2760 2761 2762 2763 2764 2765
		/*
		 * 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))) {
2766
			ret = VM_FAULT_HWPOISON |
2767
				VM_FAULT_SET_HINDEX(hstate_index(h));
2768 2769
			goto backout_unlocked;
		}
2770
	}
2771

2772 2773 2774 2775 2776 2777
	/*
	 * 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.
	 */
2778
	if ((flags & FAULT_FLAG_WRITE) && !(vma->vm_flags & VM_SHARED))
2779 2780 2781 2782
		if (vma_needs_reservation(h, vma, address) < 0) {
			ret = VM_FAULT_OOM;
			goto backout_unlocked;
		}
2783

2784
	spin_lock(&mm->page_table_lock);
2785
	size = i_size_read(mapping->host) >> huge_page_shift(h);
A
Adam Litke 已提交
2786 2787 2788
	if (idx >= size)
		goto backout;

N
Nick Piggin 已提交
2789
	ret = 0;
2790
	if (!huge_pte_none(huge_ptep_get(ptep)))
A
Adam Litke 已提交
2791 2792
		goto backout;

2793 2794 2795 2796
	if (anon_rmap)
		hugepage_add_new_anon_rmap(page, vma, address);
	else
		page_dup_rmap(page);
2797 2798 2799 2800
	new_pte = make_huge_pte(vma, page, ((vma->vm_flags & VM_WRITE)
				&& (vma->vm_flags & VM_SHARED)));
	set_huge_pte_at(mm, address, ptep, new_pte);

2801
	if ((flags & FAULT_FLAG_WRITE) && !(vma->vm_flags & VM_SHARED)) {
2802
		/* Optimization, do the COW without a second fault */
2803
		ret = hugetlb_cow(mm, vma, address, ptep, new_pte, page);
2804 2805
	}

2806
	spin_unlock(&mm->page_table_lock);
A
Adam Litke 已提交
2807 2808
	unlock_page(page);
out:
2809
	return ret;
A
Adam Litke 已提交
2810 2811 2812

backout:
	spin_unlock(&mm->page_table_lock);
2813
backout_unlocked:
A
Adam Litke 已提交
2814 2815 2816
	unlock_page(page);
	put_page(page);
	goto out;
2817 2818
}

2819
int hugetlb_fault(struct mm_struct *mm, struct vm_area_struct *vma,
2820
			unsigned long address, unsigned int flags)
2821 2822 2823
{
	pte_t *ptep;
	pte_t entry;
2824
	int ret;
2825
	struct page *page = NULL;
2826
	struct page *pagecache_page = NULL;
2827
	static DEFINE_MUTEX(hugetlb_instantiation_mutex);
2828
	struct hstate *h = hstate_vma(vma);
2829

2830 2831
	address &= huge_page_mask(h);

2832 2833 2834
	ptep = huge_pte_offset(mm, address);
	if (ptep) {
		entry = huge_ptep_get(ptep);
N
Naoya Horiguchi 已提交
2835
		if (unlikely(is_hugetlb_entry_migration(entry))) {
2836
			migration_entry_wait_huge(mm, ptep);
N
Naoya Horiguchi 已提交
2837 2838
			return 0;
		} else if (unlikely(is_hugetlb_entry_hwpoisoned(entry)))
2839
			return VM_FAULT_HWPOISON_LARGE |
2840
				VM_FAULT_SET_HINDEX(hstate_index(h));
2841 2842
	}

2843
	ptep = huge_pte_alloc(mm, address, huge_page_size(h));
2844 2845 2846
	if (!ptep)
		return VM_FAULT_OOM;

2847 2848 2849 2850 2851 2852
	/*
	 * 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.
	 */
	mutex_lock(&hugetlb_instantiation_mutex);
2853 2854
	entry = huge_ptep_get(ptep);
	if (huge_pte_none(entry)) {
2855
		ret = hugetlb_no_page(mm, vma, address, ptep, flags);
2856
		goto out_mutex;
2857
	}
2858

N
Nick Piggin 已提交
2859
	ret = 0;
2860

2861 2862 2863 2864 2865 2866 2867 2868
	/*
	 * 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.
	 */
2869
	if ((flags & FAULT_FLAG_WRITE) && !huge_pte_write(entry)) {
2870 2871
		if (vma_needs_reservation(h, vma, address) < 0) {
			ret = VM_FAULT_OOM;
2872
			goto out_mutex;
2873
		}
2874

2875
		if (!(vma->vm_flags & VM_MAYSHARE))
2876 2877 2878 2879
			pagecache_page = hugetlbfs_pagecache_page(h,
								vma, address);
	}

2880 2881 2882 2883 2884 2885 2886 2887
	/*
	 * 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.
	 * Note that locking order is always pagecache_page -> page,
	 * so no worry about deadlock.
	 */
	page = pte_page(entry);
2888
	get_page(page);
2889
	if (page != pagecache_page)
2890 2891
		lock_page(page);

2892 2893
	spin_lock(&mm->page_table_lock);
	/* Check for a racing update before calling hugetlb_cow */
2894 2895 2896 2897
	if (unlikely(!pte_same(entry, huge_ptep_get(ptep))))
		goto out_page_table_lock;


2898
	if (flags & FAULT_FLAG_WRITE) {
2899
		if (!huge_pte_write(entry)) {
2900 2901
			ret = hugetlb_cow(mm, vma, address, ptep, entry,
							pagecache_page);
2902 2903
			goto out_page_table_lock;
		}
2904
		entry = huge_pte_mkdirty(entry);
2905 2906
	}
	entry = pte_mkyoung(entry);
2907 2908
	if (huge_ptep_set_access_flags(vma, address, ptep, entry,
						flags & FAULT_FLAG_WRITE))
2909
		update_mmu_cache(vma, address, ptep);
2910 2911

out_page_table_lock:
2912
	spin_unlock(&mm->page_table_lock);
2913 2914 2915 2916 2917

	if (pagecache_page) {
		unlock_page(pagecache_page);
		put_page(pagecache_page);
	}
2918 2919
	if (page != pagecache_page)
		unlock_page(page);
2920
	put_page(page);
2921

2922
out_mutex:
2923
	mutex_unlock(&hugetlb_instantiation_mutex);
2924 2925

	return ret;
2926 2927
}

2928 2929 2930 2931
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,
			 long i, unsigned int flags)
D
David Gibson 已提交
2932
{
2933 2934
	unsigned long pfn_offset;
	unsigned long vaddr = *position;
2935
	unsigned long remainder = *nr_pages;
2936
	struct hstate *h = hstate_vma(vma);
D
David Gibson 已提交
2937

2938
	spin_lock(&mm->page_table_lock);
D
David Gibson 已提交
2939
	while (vaddr < vma->vm_end && remainder) {
A
Adam Litke 已提交
2940
		pte_t *pte;
H
Hugh Dickins 已提交
2941
		int absent;
A
Adam Litke 已提交
2942
		struct page *page;
D
David Gibson 已提交
2943

A
Adam Litke 已提交
2944 2945
		/*
		 * Some archs (sparc64, sh*) have multiple pte_ts to
H
Hugh Dickins 已提交
2946
		 * each hugepage.  We have to make sure we get the
A
Adam Litke 已提交
2947 2948
		 * first, for the page indexing below to work.
		 */
2949
		pte = huge_pte_offset(mm, vaddr & huge_page_mask(h));
H
Hugh Dickins 已提交
2950 2951 2952 2953
		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
2954 2955 2956 2957
		 * 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
2958
		 */
H
Hugh Dickins 已提交
2959 2960
		if (absent && (flags & FOLL_DUMP) &&
		    !hugetlbfs_pagecache_present(h, vma, vaddr)) {
H
Hugh Dickins 已提交
2961 2962 2963
			remainder = 0;
			break;
		}
D
David Gibson 已提交
2964

2965 2966 2967 2968 2969 2970 2971 2972 2973 2974 2975
		/*
		 * 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)) ||
2976 2977
		    ((flags & FOLL_WRITE) &&
		      !huge_pte_write(huge_ptep_get(pte)))) {
A
Adam Litke 已提交
2978
			int ret;
D
David Gibson 已提交
2979

A
Adam Litke 已提交
2980
			spin_unlock(&mm->page_table_lock);
H
Hugh Dickins 已提交
2981 2982
			ret = hugetlb_fault(mm, vma, vaddr,
				(flags & FOLL_WRITE) ? FAULT_FLAG_WRITE : 0);
A
Adam Litke 已提交
2983
			spin_lock(&mm->page_table_lock);
2984
			if (!(ret & VM_FAULT_ERROR))
A
Adam Litke 已提交
2985
				continue;
D
David Gibson 已提交
2986

A
Adam Litke 已提交
2987 2988 2989 2990
			remainder = 0;
			break;
		}

2991
		pfn_offset = (vaddr & ~huge_page_mask(h)) >> PAGE_SHIFT;
2992
		page = pte_page(huge_ptep_get(pte));
2993
same_page:
2994
		if (pages) {
H
Hugh Dickins 已提交
2995
			pages[i] = mem_map_offset(page, pfn_offset);
K
KOSAKI Motohiro 已提交
2996
			get_page(pages[i]);
2997
		}
D
David Gibson 已提交
2998 2999 3000 3001 3002

		if (vmas)
			vmas[i] = vma;

		vaddr += PAGE_SIZE;
3003
		++pfn_offset;
D
David Gibson 已提交
3004 3005
		--remainder;
		++i;
3006
		if (vaddr < vma->vm_end && remainder &&
3007
				pfn_offset < pages_per_huge_page(h)) {
3008 3009 3010 3011 3012 3013
			/*
			 * We use pfn_offset to avoid touching the pageframes
			 * of this compound page.
			 */
			goto same_page;
		}
D
David Gibson 已提交
3014
	}
3015
	spin_unlock(&mm->page_table_lock);
3016
	*nr_pages = remainder;
D
David Gibson 已提交
3017 3018
	*position = vaddr;

H
Hugh Dickins 已提交
3019
	return i ? i : -EFAULT;
D
David Gibson 已提交
3020
}
3021

3022
unsigned long hugetlb_change_protection(struct vm_area_struct *vma,
3023 3024 3025 3026 3027 3028
		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;
3029
	struct hstate *h = hstate_vma(vma);
3030
	unsigned long pages = 0;
3031 3032 3033 3034

	BUG_ON(address >= end);
	flush_cache_range(vma, address, end);

3035
	mutex_lock(&vma->vm_file->f_mapping->i_mmap_mutex);
3036
	spin_lock(&mm->page_table_lock);
3037
	for (; address < end; address += huge_page_size(h)) {
3038 3039 3040
		ptep = huge_pte_offset(mm, address);
		if (!ptep)
			continue;
3041 3042
		if (huge_pmd_unshare(mm, &address, ptep)) {
			pages++;
3043
			continue;
3044
		}
3045
		if (!huge_pte_none(huge_ptep_get(ptep))) {
3046
			pte = huge_ptep_get_and_clear(mm, address, ptep);
3047
			pte = pte_mkhuge(huge_pte_modify(pte, newprot));
3048
			pte = arch_make_huge_pte(pte, vma, NULL, 0);
3049
			set_huge_pte_at(mm, address, ptep, pte);
3050
			pages++;
3051 3052 3053
		}
	}
	spin_unlock(&mm->page_table_lock);
3054 3055 3056 3057 3058 3059
	/*
	 * Must flush TLB before releasing i_mmap_mutex: x86's huge_pmd_unshare
	 * may have cleared our pud entry and done put_page on the page table:
	 * once we release i_mmap_mutex, another task can do the final put_page
	 * and that page table be reused and filled with junk.
	 */
3060
	flush_tlb_range(vma, start, end);
3061
	mutex_unlock(&vma->vm_file->f_mapping->i_mmap_mutex);
3062 3063

	return pages << h->order;
3064 3065
}

3066 3067
int hugetlb_reserve_pages(struct inode *inode,
					long from, long to,
3068
					struct vm_area_struct *vma,
3069
					vm_flags_t vm_flags)
3070
{
3071
	long ret, chg;
3072
	struct hstate *h = hstate_inode(inode);
3073
	struct hugepage_subpool *spool = subpool_inode(inode);
3074

3075 3076 3077
	/*
	 * Only apply hugepage reservation if asked. At fault time, an
	 * attempt will be made for VM_NORESERVE to allocate a page
3078
	 * without using reserves
3079
	 */
3080
	if (vm_flags & VM_NORESERVE)
3081 3082
		return 0;

3083 3084 3085 3086 3087 3088
	/*
	 * 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
	 */
3089
	if (!vma || vma->vm_flags & VM_MAYSHARE)
3090
		chg = region_chg(&inode->i_mapping->private_list, from, to);
3091 3092 3093 3094 3095
	else {
		struct resv_map *resv_map = resv_map_alloc();
		if (!resv_map)
			return -ENOMEM;

3096
		chg = to - from;
3097

3098 3099 3100 3101
		set_vma_resv_map(vma, resv_map);
		set_vma_resv_flags(vma, HPAGE_RESV_OWNER);
	}

3102 3103 3104 3105
	if (chg < 0) {
		ret = chg;
		goto out_err;
	}
3106

3107
	/* There must be enough pages in the subpool for the mapping */
3108 3109 3110 3111
	if (hugepage_subpool_get_pages(spool, chg)) {
		ret = -ENOSPC;
		goto out_err;
	}
3112 3113

	/*
3114
	 * Check enough hugepages are available for the reservation.
3115
	 * Hand the pages back to the subpool if there are not
3116
	 */
3117
	ret = hugetlb_acct_memory(h, chg);
K
Ken Chen 已提交
3118
	if (ret < 0) {
3119
		hugepage_subpool_put_pages(spool, chg);
3120
		goto out_err;
K
Ken Chen 已提交
3121
	}
3122 3123 3124 3125 3126 3127 3128 3129 3130 3131 3132 3133

	/*
	 * 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
	 */
3134
	if (!vma || vma->vm_flags & VM_MAYSHARE)
3135
		region_add(&inode->i_mapping->private_list, from, to);
3136
	return 0;
3137
out_err:
3138 3139
	if (vma)
		resv_map_put(vma);
3140
	return ret;
3141 3142 3143 3144
}

void hugetlb_unreserve_pages(struct inode *inode, long offset, long freed)
{
3145
	struct hstate *h = hstate_inode(inode);
3146
	long chg = region_truncate(&inode->i_mapping->private_list, offset);
3147
	struct hugepage_subpool *spool = subpool_inode(inode);
K
Ken Chen 已提交
3148 3149

	spin_lock(&inode->i_lock);
3150
	inode->i_blocks -= (blocks_per_huge_page(h) * freed);
K
Ken Chen 已提交
3151 3152
	spin_unlock(&inode->i_lock);

3153
	hugepage_subpool_put_pages(spool, (chg - freed));
3154
	hugetlb_acct_memory(h, -(chg - freed));
3155
}
3156

3157 3158 3159 3160 3161 3162 3163 3164 3165 3166 3167 3168 3169 3170 3171 3172 3173 3174 3175 3176 3177 3178 3179 3180 3181 3182 3183 3184 3185 3186 3187 3188 3189 3190 3191 3192 3193 3194 3195 3196 3197 3198 3199 3200 3201 3202 3203 3204 3205 3206 3207 3208 3209 3210 3211 3212 3213 3214 3215 3216 3217 3218 3219 3220 3221 3222 3223 3224 3225 3226 3227 3228 3229 3230 3231 3232 3233 3234 3235 3236 3237 3238 3239 3240 3241 3242 3243 3244 3245 3246 3247 3248 3249 3250 3251 3252 3253 3254 3255 3256 3257 3258 3259 3260 3261 3262 3263 3264 3265 3266 3267 3268 3269 3270 3271 3272 3273 3274 3275 3276
#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 */
	unsigned long vm_flags = vma->vm_flags & ~VM_LOCKED;
	unsigned long svm_flags = svma->vm_flags & ~VM_LOCKED;

	/*
	 * 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;
}

static int vma_shareable(struct vm_area_struct *vma, unsigned long addr)
{
	unsigned long base = addr & PUD_MASK;
	unsigned long end = base + PUD_SIZE;

	/*
	 * check on proper vm_flags and page table alignment
	 */
	if (vma->vm_flags & VM_MAYSHARE &&
	    vma->vm_start <= base && end <= vma->vm_end)
		return 1;
	return 0;
}

/*
 * 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
 * code much cleaner. pmd allocation is essential for the shared case because
 * pud has to be populated inside the same i_mmap_mutex section - otherwise
 * racing tasks could either miss the sharing (see huge_pte_offset) or select a
 * bad pmd for sharing.
 */
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;

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

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

		saddr = page_table_shareable(svma, vma, addr, idx);
		if (saddr) {
			spte = huge_pte_offset(svma->vm_mm, saddr);
			if (spte) {
				get_page(virt_to_page(spte));
				break;
			}
		}
	}

	if (!spte)
		goto out;

	spin_lock(&mm->page_table_lock);
	if (pud_none(*pud))
		pud_populate(mm, pud,
				(pmd_t *)((unsigned long)spte & PAGE_MASK));
	else
		put_page(virt_to_page(spte));
	spin_unlock(&mm->page_table_lock);
out:
	pte = (pte_t *)pmd_alloc(mm, pud, addr);
	mutex_unlock(&mapping->i_mmap_mutex);
	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.
 *
 * called with vma->vm_mm->page_table_lock held.
 *
 * returns: 1 successfully unmapped a shared pte page
 *	    0 the underlying pte page is not shared, or it is the last user
 */
int huge_pmd_unshare(struct mm_struct *mm, unsigned long *addr, pte_t *ptep)
{
	pgd_t *pgd = pgd_offset(mm, *addr);
	pud_t *pud = pud_offset(pgd, *addr);

	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));
	*addr = ALIGN(*addr, HPAGE_SIZE * PTRS_PER_PTE) - HPAGE_SIZE;
	return 1;
}
3277 3278 3279 3280 3281 3282 3283
#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;
}
#define want_pmd_share()	(0)
3284 3285
#endif /* CONFIG_ARCH_WANT_HUGE_PMD_SHARE */

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 3314 3315 3316 3317 3318 3319 3320 3321 3322 3323 3324 3325 3326 3327 3328 3329 3330 3331 3332 3333 3334 3335 3336 3337 3338 3339 3340 3341 3342 3343 3344 3345 3346 3347 3348 3349 3350 3351 3352 3353 3354 3355 3356 3357 3358 3359 3360 3361 3362 3363 3364 3365 3366
#ifdef CONFIG_ARCH_WANT_GENERAL_HUGETLB
pte_t *huge_pte_alloc(struct mm_struct *mm,
			unsigned long addr, unsigned long sz)
{
	pgd_t *pgd;
	pud_t *pud;
	pte_t *pte = NULL;

	pgd = pgd_offset(mm, addr);
	pud = pud_alloc(mm, pgd, addr);
	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);
		}
	}
	BUG_ON(pte && !pte_none(*pte) && !pte_huge(*pte));

	return pte;
}

pte_t *huge_pte_offset(struct mm_struct *mm, unsigned long addr)
{
	pgd_t *pgd;
	pud_t *pud;
	pmd_t *pmd = NULL;

	pgd = pgd_offset(mm, addr);
	if (pgd_present(*pgd)) {
		pud = pud_offset(pgd, addr);
		if (pud_present(*pud)) {
			if (pud_huge(*pud))
				return (pte_t *)pud;
			pmd = pmd_offset(pud, addr);
		}
	}
	return (pte_t *) pmd;
}

struct page *
follow_huge_pmd(struct mm_struct *mm, unsigned long address,
		pmd_t *pmd, int write)
{
	struct page *page;

	page = pte_page(*(pte_t *)pmd);
	if (page)
		page += ((address & ~PMD_MASK) >> PAGE_SHIFT);
	return page;
}

struct page *
follow_huge_pud(struct mm_struct *mm, unsigned long address,
		pud_t *pud, int write)
{
	struct page *page;

	page = pte_page(*(pte_t *)pud);
	if (page)
		page += ((address & ~PUD_MASK) >> PAGE_SHIFT);
	return page;
}

#else /* !CONFIG_ARCH_WANT_GENERAL_HUGETLB */

/* Can be overriden by architectures */
__attribute__((weak)) struct page *
follow_huge_pud(struct mm_struct *mm, unsigned long address,
	       pud_t *pud, int write)
{
	BUG();
	return NULL;
}

#endif /* CONFIG_ARCH_WANT_GENERAL_HUGETLB */

3367 3368
#ifdef CONFIG_MEMORY_FAILURE

3369 3370 3371 3372 3373 3374 3375 3376 3377 3378 3379 3380 3381 3382
/* Should be called in hugetlb_lock */
static int is_hugepage_on_freelist(struct page *hpage)
{
	struct page *page;
	struct page *tmp;
	struct hstate *h = page_hstate(hpage);
	int nid = page_to_nid(hpage);

	list_for_each_entry_safe(page, tmp, &h->hugepage_freelists[nid], lru)
		if (page == hpage)
			return 1;
	return 0;
}

3383 3384 3385 3386
/*
 * This function is called from memory failure code.
 * Assume the caller holds page lock of the head page.
 */
3387
int dequeue_hwpoisoned_huge_page(struct page *hpage)
3388 3389 3390
{
	struct hstate *h = page_hstate(hpage);
	int nid = page_to_nid(hpage);
3391
	int ret = -EBUSY;
3392 3393

	spin_lock(&hugetlb_lock);
3394
	if (is_hugepage_on_freelist(hpage)) {
3395 3396 3397 3398 3399 3400 3401
		/*
		 * Hwpoisoned hugepage isn't linked to activelist or freelist,
		 * but dangling hpage->lru can trigger list-debug warnings
		 * (this happens when we call unpoison_memory() on it),
		 * so let it point to itself with list_del_init().
		 */
		list_del_init(&hpage->lru);
3402
		set_page_refcounted(hpage);
3403 3404 3405 3406
		h->free_huge_pages--;
		h->free_huge_pages_node[nid]--;
		ret = 0;
	}
3407
	spin_unlock(&hugetlb_lock);
3408
	return ret;
3409
}
3410
#endif