hugetlb.c 79.2 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4 5 6 7 8
/*
 * Generic hugetlb support.
 * (C) William Irwin, April 2004
 */
#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 <linux/io.h>
D
David Gibson 已提交
28 29

#include <linux/hugetlb.h>
30
#include <linux/node.h>
31
#include "internal.h"
L
Linus Torvalds 已提交
32 33

const unsigned long hugetlb_zero = 0, hugetlb_infinity = ~0UL;
34 35
static gfp_t htlb_alloc_mask = GFP_HIGHUSER;
unsigned long hugepages_treat_as_movable;
36

37 38 39 40
static int max_hstate;
unsigned int default_hstate_idx;
struct hstate hstates[HUGE_MAX_HSTATE];

41 42
__initdata LIST_HEAD(huge_boot_pages);

43 44 45
/* for command line parsing */
static struct hstate * __initdata parsed_hstate;
static unsigned long __initdata default_hstate_max_huge_pages;
46
static unsigned long __initdata default_hstate_size;
47 48 49

#define for_each_hstate(h) \
	for ((h) = hstates; (h) < &hstates[max_hstate]; (h)++)
50

51 52 53 54
/*
 * Protects updates to hugepage_freelists, nr_huge_pages, and free_huge_pages
 */
static DEFINE_SPINLOCK(hugetlb_lock);
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133
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)
{
	return subpool_inode(vma->vm_file->f_dentry->d_inode);
}

134 135 136
/*
 * Region tracking -- allows tracking of reservations and instantiated pages
 *                    across the pages in a mapping.
137 138 139 140 141 142
 *
 * The region data structures are protected by a combination of the mmap_sem
 * and the hugetlb_instantion_mutex.  To access or modify a region the caller
 * must either hold the mmap_sem for write, or the mmap_sem for read and
 * the hugetlb_instantiation mutex:
 *
143
 *	down_write(&mm->mmap_sem);
144
 * or
145 146
 *	down_read(&mm->mmap_sem);
 *	mutex_lock(&hugetlb_instantiation_mutex);
147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226
 */
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
227
		/* We overlap with this area, if it extends further than
228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268
		 * 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;
}

269 270 271 272 273 274 275
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) {
276 277
		long seg_from;
		long seg_to;
278 279 280 281 282 283 284 285 286 287 288 289 290 291 292

		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;
}

293 294 295 296
/*
 * Convert the address within this vma to the page offset within
 * the mapping, in pagecache page units; huge pages here.
 */
297 298
static pgoff_t vma_hugecache_offset(struct hstate *h,
			struct vm_area_struct *vma, unsigned long address)
299
{
300 301
	return ((address - vma->vm_start) >> huge_page_shift(h)) +
			(vma->vm_pgoff >> huge_page_order(h));
302 303
}

304 305 306 307 308 309
pgoff_t linear_hugepage_index(struct vm_area_struct *vma,
				     unsigned long address)
{
	return vma_hugecache_offset(hstate_vma(vma), vma, address);
}

310 311 312 313 314 315 316 317 318 319 320 321 322 323 324
/*
 * 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);

	return 1UL << (hstate->order + PAGE_SHIFT);
}
325
EXPORT_SYMBOL_GPL(vma_kernel_pagesize);
326

327 328 329 330 331 332 333 334 335 336 337 338 339
/*
 * 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

340 341 342 343 344 345 346
/*
 * 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)
347
#define HPAGE_RESV_MASK (HPAGE_RESV_OWNER | HPAGE_RESV_UNMAPPED)
348

349 350 351 352 353 354 355 356 357
/*
 * 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.
358 359 360 361 362 363 364 365 366
 *
 * 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.
367
 */
368 369 370 371 372 373 374 375 376 377 378
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;
}

379 380 381 382 383
struct resv_map {
	struct kref refs;
	struct list_head regions;
};

384
static struct resv_map *resv_map_alloc(void)
385 386 387 388 389 390 391 392 393 394 395
{
	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;
}

396
static void resv_map_release(struct kref *ref)
397 398 399 400 401 402 403 404 405
{
	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)
406 407
{
	VM_BUG_ON(!is_vm_hugetlb_page(vma));
408
	if (!(vma->vm_flags & VM_MAYSHARE))
409 410
		return (struct resv_map *)(get_vma_private_data(vma) &
							~HPAGE_RESV_MASK);
411
	return NULL;
412 413
}

414
static void set_vma_resv_map(struct vm_area_struct *vma, struct resv_map *map)
415 416
{
	VM_BUG_ON(!is_vm_hugetlb_page(vma));
417
	VM_BUG_ON(vma->vm_flags & VM_MAYSHARE);
418

419 420
	set_vma_private_data(vma, (get_vma_private_data(vma) &
				HPAGE_RESV_MASK) | (unsigned long)map);
421 422 423 424 425
}

static void set_vma_resv_flags(struct vm_area_struct *vma, unsigned long flags)
{
	VM_BUG_ON(!is_vm_hugetlb_page(vma));
426
	VM_BUG_ON(vma->vm_flags & VM_MAYSHARE);
427 428

	set_vma_private_data(vma, get_vma_private_data(vma) | flags);
429 430 431 432 433
}

static int is_vma_resv_set(struct vm_area_struct *vma, unsigned long flag)
{
	VM_BUG_ON(!is_vm_hugetlb_page(vma));
434 435

	return (get_vma_private_data(vma) & flag) != 0;
436 437 438
}

/* Decrement the reserved pages in the hugepage pool by one */
439 440
static void decrement_hugepage_resv_vma(struct hstate *h,
			struct vm_area_struct *vma)
441
{
442 443 444
	if (vma->vm_flags & VM_NORESERVE)
		return;

445
	if (vma->vm_flags & VM_MAYSHARE) {
446
		/* Shared mappings always use reserves */
447
		h->resv_huge_pages--;
448
	} else if (is_vma_resv_set(vma, HPAGE_RESV_OWNER)) {
449 450 451 452
		/*
		 * Only the process that called mmap() has reserves for
		 * private mappings.
		 */
453
		h->resv_huge_pages--;
454 455 456
	}
}

457
/* Reset counters to 0 and clear all HPAGE_RESV_* flags */
458 459 460
void reset_vma_resv_huge_pages(struct vm_area_struct *vma)
{
	VM_BUG_ON(!is_vm_hugetlb_page(vma));
461
	if (!(vma->vm_flags & VM_MAYSHARE))
462 463 464 465
		vma->vm_private_data = (void *)0;
}

/* Returns true if the VMA has associated reserve pages */
466
static int vma_has_reserves(struct vm_area_struct *vma)
467
{
468
	if (vma->vm_flags & VM_MAYSHARE)
469 470 471 472
		return 1;
	if (is_vma_resv_set(vma, HPAGE_RESV_OWNER))
		return 1;
	return 0;
473 474
}

475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508
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);
	}
}

509
static void enqueue_huge_page(struct hstate *h, struct page *page)
L
Linus Torvalds 已提交
510 511
{
	int nid = page_to_nid(page);
512 513 514
	list_add(&page->lru, &h->hugepage_freelists[nid]);
	h->free_huge_pages++;
	h->free_huge_pages_node[nid]++;
L
Linus Torvalds 已提交
515 516
}

517 518 519 520 521 522 523 524
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);
	list_del(&page->lru);
525
	set_page_refcounted(page);
526 527 528 529 530
	h->free_huge_pages--;
	h->free_huge_pages_node[nid]--;
	return page;
}

531 532
static struct page *dequeue_huge_page_vma(struct hstate *h,
				struct vm_area_struct *vma,
533
				unsigned long address, int avoid_reserve)
L
Linus Torvalds 已提交
534
{
535
	struct page *page = NULL;
536
	struct mempolicy *mpol;
537
	nodemask_t *nodemask;
538
	struct zonelist *zonelist;
539 540
	struct zone *zone;
	struct zoneref *z;
541
	unsigned int cpuset_mems_cookie;
L
Linus Torvalds 已提交
542

543 544
retry_cpuset:
	cpuset_mems_cookie = get_mems_allowed();
545 546
	zonelist = huge_zonelist(vma, address,
					htlb_alloc_mask, &mpol, &nodemask);
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) &&
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
	for_each_zone_zonelist_nodemask(zone, z, zonelist,
						MAX_NR_ZONES - 1, nodemask) {
562 563 564 565 566 567 568
		if (cpuset_zone_allowed_softwall(zone, htlb_alloc_mask)) {
			page = dequeue_huge_page_node(h, zone_to_nid(zone));
			if (page) {
				if (!avoid_reserve)
					decrement_hugepage_resv_vma(h, vma);
				break;
			}
A
Andrew Morton 已提交
569
		}
L
Linus Torvalds 已提交
570
	}
571

572
	mpol_cond_put(mpol);
573 574
	if (unlikely(!put_mems_allowed(cpuset_mems_cookie) && !page))
		goto retry_cpuset;
L
Linus Torvalds 已提交
575
	return page;
576 577 578 579

err:
	mpol_cond_put(mpol);
	return NULL;
L
Linus Torvalds 已提交
580 581
}

582
static void update_and_free_page(struct hstate *h, struct page *page)
A
Adam Litke 已提交
583 584
{
	int i;
585

586 587
	VM_BUG_ON(h->order >= MAX_ORDER);

588 589 590
	h->nr_huge_pages--;
	h->nr_huge_pages_node[page_to_nid(page)]--;
	for (i = 0; i < pages_per_huge_page(h); i++) {
591 592 593 594
		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
595 596 597
	}
	set_compound_page_dtor(page, NULL);
	set_page_refcounted(page);
598
	arch_release_hugepage(page);
599
	__free_pages(page, huge_page_order(h));
A
Adam Litke 已提交
600 601
}

602 603 604 605 606 607 608 609 610 611 612
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;
}

613 614
static void free_huge_page(struct page *page)
{
615 616 617 618
	/*
	 * Can't pass hstate in here because it is called from the
	 * compound page destructor.
	 */
619
	struct hstate *h = page_hstate(page);
620
	int nid = page_to_nid(page);
621 622
	struct hugepage_subpool *spool =
		(struct hugepage_subpool *)page_private(page);
623

624
	set_page_private(page, 0);
625
	page->mapping = NULL;
626
	BUG_ON(page_count(page));
627
	BUG_ON(page_mapcount(page));
628 629 630
	INIT_LIST_HEAD(&page->lru);

	spin_lock(&hugetlb_lock);
631
	if (h->surplus_huge_pages_node[nid] && huge_page_order(h) < MAX_ORDER) {
632 633 634
		update_and_free_page(h, page);
		h->surplus_huge_pages--;
		h->surplus_huge_pages_node[nid]--;
635
	} else {
636
		enqueue_huge_page(h, page);
637
	}
638
	spin_unlock(&hugetlb_lock);
639
	hugepage_subpool_put_pages(spool, 1);
640 641
}

642
static void prep_new_huge_page(struct hstate *h, struct page *page, int nid)
643 644 645
{
	set_compound_page_dtor(page, free_huge_page);
	spin_lock(&hugetlb_lock);
646 647
	h->nr_huge_pages++;
	h->nr_huge_pages_node[nid]++;
648 649 650 651
	spin_unlock(&hugetlb_lock);
	put_page(page); /* free it into the hugepage allocator */
}

652 653 654 655 656 657 658 659 660 661 662
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);
663
		set_page_count(p, 0);
664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679
		p->first_page = page;
	}
}

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;
}
680 681
EXPORT_SYMBOL_GPL(PageHuge);

682
static struct page *alloc_fresh_huge_page_node(struct hstate *h, int nid)
L
Linus Torvalds 已提交
683 684
{
	struct page *page;
685

686 687 688
	if (h->order >= MAX_ORDER)
		return NULL;

689
	page = alloc_pages_exact_node(nid,
690 691
		htlb_alloc_mask|__GFP_COMP|__GFP_THISNODE|
						__GFP_REPEAT|__GFP_NOWARN,
692
		huge_page_order(h));
L
Linus Torvalds 已提交
693
	if (page) {
694
		if (arch_prepare_hugepage(page)) {
695
			__free_pages(page, huge_page_order(h));
696
			return NULL;
697
		}
698
		prep_new_huge_page(h, page, nid);
L
Linus Torvalds 已提交
699
	}
700 701 702 703

	return page;
}

704
/*
705 706 707 708 709
 * 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.
710
 */
711
static int next_node_allowed(int nid, nodemask_t *nodes_allowed)
712
{
713
	nid = next_node(nid, *nodes_allowed);
714
	if (nid == MAX_NUMNODES)
715
		nid = first_node(*nodes_allowed);
716 717 718 719 720
	VM_BUG_ON(nid >= MAX_NUMNODES);

	return nid;
}

721 722 723 724 725 726 727
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;
}

728
/*
729 730 731 732
 * 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.
733
 */
734 735
static int hstate_next_node_to_alloc(struct hstate *h,
					nodemask_t *nodes_allowed)
736
{
737 738 739 740 741 742
	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);
743 744

	return nid;
745 746
}

747
static int alloc_fresh_huge_page(struct hstate *h, nodemask_t *nodes_allowed)
748 749 750 751 752 753
{
	struct page *page;
	int start_nid;
	int next_nid;
	int ret = 0;

754
	start_nid = hstate_next_node_to_alloc(h, nodes_allowed);
755
	next_nid = start_nid;
756 757

	do {
758
		page = alloc_fresh_huge_page_node(h, next_nid);
759
		if (page) {
760
			ret = 1;
761 762
			break;
		}
763
		next_nid = hstate_next_node_to_alloc(h, nodes_allowed);
764
	} while (next_nid != start_nid);
765

766 767 768 769 770
	if (ret)
		count_vm_event(HTLB_BUDDY_PGALLOC);
	else
		count_vm_event(HTLB_BUDDY_PGALLOC_FAIL);

771
	return ret;
L
Linus Torvalds 已提交
772 773
}

774
/*
775 776 777 778
 * 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.
779
 */
780
static int hstate_next_node_to_free(struct hstate *h, nodemask_t *nodes_allowed)
781
{
782 783 784 785 786 787
	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);
788 789

	return nid;
790 791 792 793 794 795 796 797
}

/*
 * 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.
 */
798 799
static int free_pool_huge_page(struct hstate *h, nodemask_t *nodes_allowed,
							 bool acct_surplus)
800 801 802 803 804
{
	int start_nid;
	int next_nid;
	int ret = 0;

805
	start_nid = hstate_next_node_to_free(h, nodes_allowed);
806 807 808
	next_nid = start_nid;

	do {
809 810 811 812 813 814
		/*
		 * If we're returning unused surplus pages, only examine
		 * nodes with surplus pages.
		 */
		if ((!acct_surplus || h->surplus_huge_pages_node[next_nid]) &&
		    !list_empty(&h->hugepage_freelists[next_nid])) {
815 816 817 818 819 820
			struct page *page =
				list_entry(h->hugepage_freelists[next_nid].next,
					  struct page, lru);
			list_del(&page->lru);
			h->free_huge_pages--;
			h->free_huge_pages_node[next_nid]--;
821 822 823 824
			if (acct_surplus) {
				h->surplus_huge_pages--;
				h->surplus_huge_pages_node[next_nid]--;
			}
825 826
			update_and_free_page(h, page);
			ret = 1;
827
			break;
828
		}
829
		next_nid = hstate_next_node_to_free(h, nodes_allowed);
830
	} while (next_nid != start_nid);
831 832 833 834

	return ret;
}

835
static struct page *alloc_buddy_huge_page(struct hstate *h, int nid)
836 837
{
	struct page *page;
838
	unsigned int r_nid;
839

840 841 842
	if (h->order >= MAX_ORDER)
		return NULL;

843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866
	/*
	 * 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);
867
	if (h->surplus_huge_pages >= h->nr_overcommit_huge_pages) {
868 869 870
		spin_unlock(&hugetlb_lock);
		return NULL;
	} else {
871 872
		h->nr_huge_pages++;
		h->surplus_huge_pages++;
873 874 875
	}
	spin_unlock(&hugetlb_lock);

876 877 878 879 880 881 882 883
	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));
884

885 886
	if (page && arch_prepare_hugepage(page)) {
		__free_pages(page, huge_page_order(h));
887
		page = NULL;
888 889
	}

890
	spin_lock(&hugetlb_lock);
891
	if (page) {
892
		r_nid = page_to_nid(page);
893
		set_compound_page_dtor(page, free_huge_page);
894 895 896
		/*
		 * We incremented the global counters already
		 */
897 898
		h->nr_huge_pages_node[r_nid]++;
		h->surplus_huge_pages_node[r_nid]++;
899
		__count_vm_event(HTLB_BUDDY_PGALLOC);
900
	} else {
901 902
		h->nr_huge_pages--;
		h->surplus_huge_pages--;
903
		__count_vm_event(HTLB_BUDDY_PGALLOC_FAIL);
904
	}
905
	spin_unlock(&hugetlb_lock);
906 907 908 909

	return page;
}

910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928
/*
 * 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);

	if (!page)
		page = alloc_buddy_huge_page(h, nid);

	return page;
}

929
/*
L
Lucas De Marchi 已提交
930
 * Increase the hugetlb pool such that it can accommodate a reservation
931 932
 * of size 'delta'.
 */
933
static int gather_surplus_pages(struct hstate *h, int delta)
934 935 936 937 938
{
	struct list_head surplus_list;
	struct page *page, *tmp;
	int ret, i;
	int needed, allocated;
939
	bool alloc_ok = true;
940

941
	needed = (h->resv_huge_pages + delta) - h->free_huge_pages;
942
	if (needed <= 0) {
943
		h->resv_huge_pages += delta;
944
		return 0;
945
	}
946 947 948 949 950 951 952 953

	allocated = 0;
	INIT_LIST_HEAD(&surplus_list);

	ret = -ENOMEM;
retry:
	spin_unlock(&hugetlb_lock);
	for (i = 0; i < needed; i++) {
954
		page = alloc_buddy_huge_page(h, NUMA_NO_NODE);
955 956 957 958
		if (!page) {
			alloc_ok = false;
			break;
		}
959 960
		list_add(&page->lru, &surplus_list);
	}
961
	allocated += i;
962 963 964 965 966 967

	/*
	 * 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);
968 969
	needed = (h->resv_huge_pages + delta) -
			(h->free_huge_pages + allocated);
970 971 972 973 974 975 976 977 978 979
	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;
	}
980 981
	/*
	 * The surplus_list now contains _at_least_ the number of extra pages
L
Lucas De Marchi 已提交
982
	 * needed to accommodate the reservation.  Add the appropriate number
983
	 * of pages to the hugetlb pool and free the extras back to the buddy
984 985 986
	 * 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.
987 988
	 */
	needed += allocated;
989
	h->resv_huge_pages += delta;
990
	ret = 0;
991

992
	/* Free the needed pages to the hugetlb pool */
993
	list_for_each_entry_safe(page, tmp, &surplus_list, lru) {
994 995
		if ((--needed) < 0)
			break;
996
		list_del(&page->lru);
997 998 999 1000 1001 1002
		/*
		 * 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));
1003
		enqueue_huge_page(h, page);
1004
	}
1005
free:
1006
	spin_unlock(&hugetlb_lock);
1007 1008 1009 1010 1011

	/* Free unnecessary surplus pages to the buddy allocator */
	if (!list_empty(&surplus_list)) {
		list_for_each_entry_safe(page, tmp, &surplus_list, lru) {
			list_del(&page->lru);
1012
			put_page(page);
1013
		}
1014
	}
1015
	spin_lock(&hugetlb_lock);
1016 1017 1018 1019 1020 1021 1022 1023

	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.
1024
 * Called with hugetlb_lock held.
1025
 */
1026 1027
static void return_unused_surplus_pages(struct hstate *h,
					unsigned long unused_resv_pages)
1028 1029 1030
{
	unsigned long nr_pages;

1031
	/* Uncommit the reservation */
1032
	h->resv_huge_pages -= unused_resv_pages;
1033

1034 1035 1036 1037
	/* Cannot return gigantic pages currently */
	if (h->order >= MAX_ORDER)
		return;

1038
	nr_pages = min(unused_resv_pages, h->surplus_huge_pages);
1039

1040 1041
	/*
	 * We want to release as many surplus pages as possible, spread
1042 1043 1044 1045 1046
	 * 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.
1047 1048
	 */
	while (nr_pages--) {
1049
		if (!free_pool_huge_page(h, &node_states[N_HIGH_MEMORY], 1))
1050
			break;
1051 1052 1053
	}
}

1054 1055 1056
/*
 * Determine if the huge page at addr within the vma has an associated
 * reservation.  Where it does not we will need to logically increase
1057 1058 1059 1060 1061 1062
 * 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.
1063
 */
1064
static long vma_needs_reservation(struct hstate *h,
1065
			struct vm_area_struct *vma, unsigned long addr)
1066 1067 1068 1069
{
	struct address_space *mapping = vma->vm_file->f_mapping;
	struct inode *inode = mapping->host;

1070
	if (vma->vm_flags & VM_MAYSHARE) {
1071
		pgoff_t idx = vma_hugecache_offset(h, vma, addr);
1072 1073 1074
		return region_chg(&inode->i_mapping->private_list,
							idx, idx + 1);

1075 1076
	} else if (!is_vma_resv_set(vma, HPAGE_RESV_OWNER)) {
		return 1;
1077

1078
	} else  {
1079
		long err;
1080
		pgoff_t idx = vma_hugecache_offset(h, vma, addr);
1081 1082 1083 1084 1085 1086 1087
		struct resv_map *reservations = vma_resv_map(vma);

		err = region_chg(&reservations->regions, idx, idx + 1);
		if (err < 0)
			return err;
		return 0;
	}
1088
}
1089 1090
static void vma_commit_reservation(struct hstate *h,
			struct vm_area_struct *vma, unsigned long addr)
1091 1092 1093 1094
{
	struct address_space *mapping = vma->vm_file->f_mapping;
	struct inode *inode = mapping->host;

1095
	if (vma->vm_flags & VM_MAYSHARE) {
1096
		pgoff_t idx = vma_hugecache_offset(h, vma, addr);
1097
		region_add(&inode->i_mapping->private_list, idx, idx + 1);
1098 1099

	} else if (is_vma_resv_set(vma, HPAGE_RESV_OWNER)) {
1100
		pgoff_t idx = vma_hugecache_offset(h, vma, addr);
1101 1102 1103 1104
		struct resv_map *reservations = vma_resv_map(vma);

		/* Mark this page used in the map. */
		region_add(&reservations->regions, idx, idx + 1);
1105 1106 1107
	}
}

1108
static struct page *alloc_huge_page(struct vm_area_struct *vma,
1109
				    unsigned long addr, int avoid_reserve)
L
Linus Torvalds 已提交
1110
{
1111
	struct hugepage_subpool *spool = subpool_vma(vma);
1112
	struct hstate *h = hstate_vma(vma);
1113
	struct page *page;
1114
	long chg;
1115 1116

	/*
1117 1118 1119 1120 1121 1122
	 * 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.
1123
	 */
1124
	chg = vma_needs_reservation(h, vma, addr);
1125
	if (chg < 0)
1126
		return ERR_PTR(-VM_FAULT_OOM);
1127
	if (chg)
1128
		if (hugepage_subpool_get_pages(spool, chg))
1129
			return ERR_PTR(-VM_FAULT_SIGBUS);
L
Linus Torvalds 已提交
1130 1131

	spin_lock(&hugetlb_lock);
1132
	page = dequeue_huge_page_vma(h, vma, addr, avoid_reserve);
L
Linus Torvalds 已提交
1133
	spin_unlock(&hugetlb_lock);
1134

K
Ken Chen 已提交
1135
	if (!page) {
1136
		page = alloc_buddy_huge_page(h, NUMA_NO_NODE);
K
Ken Chen 已提交
1137
		if (!page) {
1138
			hugepage_subpool_put_pages(spool, chg);
1139
			return ERR_PTR(-VM_FAULT_SIGBUS);
K
Ken Chen 已提交
1140 1141
		}
	}
1142

1143
	set_page_private(page, (unsigned long)spool);
1144

1145
	vma_commit_reservation(h, vma, addr);
1146

1147
	return page;
1148 1149
}

1150
int __weak alloc_bootmem_huge_page(struct hstate *h)
1151 1152
{
	struct huge_bootmem_page *m;
1153
	int nr_nodes = nodes_weight(node_states[N_HIGH_MEMORY]);
1154 1155 1156 1157 1158

	while (nr_nodes) {
		void *addr;

		addr = __alloc_bootmem_node_nopanic(
1159
				NODE_DATA(hstate_next_node_to_alloc(h,
1160
						&node_states[N_HIGH_MEMORY])),
1161 1162 1163 1164 1165 1166 1167 1168 1169
				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;
1170
			goto found;
1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183
		}
		nr_nodes--;
	}
	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;
}

1184 1185 1186 1187 1188 1189 1190 1191
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);
}

1192 1193 1194 1195 1196 1197 1198
/* 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;
1199 1200 1201 1202 1203 1204 1205 1206 1207
		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
1208 1209
		__ClearPageReserved(page);
		WARN_ON(page_count(page) != 1);
1210
		prep_compound_huge_page(page, h->order);
1211
		prep_new_huge_page(h, page, page_to_nid(page));
1212 1213 1214 1215 1216 1217 1218 1219
		/*
		 * 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))
			totalram_pages += 1 << h->order;
1220 1221 1222
	}
}

1223
static void __init hugetlb_hstate_alloc_pages(struct hstate *h)
L
Linus Torvalds 已提交
1224 1225
{
	unsigned long i;
1226

1227
	for (i = 0; i < h->max_huge_pages; ++i) {
1228 1229 1230
		if (h->order >= MAX_ORDER) {
			if (!alloc_bootmem_huge_page(h))
				break;
1231 1232
		} else if (!alloc_fresh_huge_page(h,
					 &node_states[N_HIGH_MEMORY]))
L
Linus Torvalds 已提交
1233 1234
			break;
	}
1235
	h->max_huge_pages = i;
1236 1237 1238 1239 1240 1241 1242
}

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

	for_each_hstate(h) {
1243 1244 1245
		/* oversize hugepages were init'ed in early boot */
		if (h->order < MAX_ORDER)
			hugetlb_hstate_alloc_pages(h);
1246 1247 1248
	}
}

A
Andi Kleen 已提交
1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259
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;
}

1260 1261 1262 1263 1264
static void __init report_hugepages(void)
{
	struct hstate *h;

	for_each_hstate(h) {
A
Andi Kleen 已提交
1265 1266 1267 1268 1269
		char buf[32];
		printk(KERN_INFO "HugeTLB registered %s page size, "
				 "pre-allocated %ld pages\n",
			memfmt(buf, huge_page_size(h)),
			h->free_huge_pages);
1270 1271 1272
	}
}

L
Linus Torvalds 已提交
1273
#ifdef CONFIG_HIGHMEM
1274 1275
static void try_to_free_low(struct hstate *h, unsigned long count,
						nodemask_t *nodes_allowed)
L
Linus Torvalds 已提交
1276
{
1277 1278
	int i;

1279 1280 1281
	if (h->order >= MAX_ORDER)
		return;

1282
	for_each_node_mask(i, *nodes_allowed) {
L
Linus Torvalds 已提交
1283
		struct page *page, *next;
1284 1285 1286
		struct list_head *freel = &h->hugepage_freelists[i];
		list_for_each_entry_safe(page, next, freel, lru) {
			if (count >= h->nr_huge_pages)
1287
				return;
L
Linus Torvalds 已提交
1288 1289 1290
			if (PageHighMem(page))
				continue;
			list_del(&page->lru);
1291
			update_and_free_page(h, page);
1292 1293
			h->free_huge_pages--;
			h->free_huge_pages_node[page_to_nid(page)]--;
L
Linus Torvalds 已提交
1294 1295 1296 1297
		}
	}
}
#else
1298 1299
static inline void try_to_free_low(struct hstate *h, unsigned long count,
						nodemask_t *nodes_allowed)
L
Linus Torvalds 已提交
1300 1301 1302 1303
{
}
#endif

1304 1305 1306 1307 1308
/*
 * 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.
 */
1309 1310
static int adjust_pool_surplus(struct hstate *h, nodemask_t *nodes_allowed,
				int delta)
1311
{
1312
	int start_nid, next_nid;
1313 1314 1315 1316
	int ret = 0;

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

1317
	if (delta < 0)
1318
		start_nid = hstate_next_node_to_alloc(h, nodes_allowed);
1319
	else
1320
		start_nid = hstate_next_node_to_free(h, nodes_allowed);
1321 1322 1323 1324 1325 1326 1327 1328
	next_nid = start_nid;

	do {
		int nid = next_nid;
		if (delta < 0)  {
			/*
			 * To shrink on this node, there must be a surplus page
			 */
1329
			if (!h->surplus_huge_pages_node[nid]) {
1330 1331
				next_nid = hstate_next_node_to_alloc(h,
								nodes_allowed);
1332
				continue;
1333
			}
1334 1335 1336 1337 1338 1339
		}
		if (delta > 0) {
			/*
			 * Surplus cannot exceed the total number of pages
			 */
			if (h->surplus_huge_pages_node[nid] >=
1340
						h->nr_huge_pages_node[nid]) {
1341 1342
				next_nid = hstate_next_node_to_free(h,
								nodes_allowed);
1343
				continue;
1344
			}
1345
		}
1346 1347 1348 1349 1350

		h->surplus_huge_pages += delta;
		h->surplus_huge_pages_node[nid] += delta;
		ret = 1;
		break;
1351
	} while (next_nid != start_nid);
1352 1353 1354 1355

	return ret;
}

1356
#define persistent_huge_pages(h) (h->nr_huge_pages - h->surplus_huge_pages)
1357 1358
static unsigned long set_max_huge_pages(struct hstate *h, unsigned long count,
						nodemask_t *nodes_allowed)
L
Linus Torvalds 已提交
1359
{
1360
	unsigned long min_count, ret;
L
Linus Torvalds 已提交
1361

1362 1363 1364
	if (h->order >= MAX_ORDER)
		return h->max_huge_pages;

1365 1366 1367 1368
	/*
	 * Increase the pool size
	 * First take pages out of surplus state.  Then make up the
	 * remaining difference by allocating fresh huge pages.
1369 1370 1371 1372 1373 1374
	 *
	 * 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.
1375
	 */
L
Linus Torvalds 已提交
1376
	spin_lock(&hugetlb_lock);
1377
	while (h->surplus_huge_pages && count > persistent_huge_pages(h)) {
1378
		if (!adjust_pool_surplus(h, nodes_allowed, -1))
1379 1380 1381
			break;
	}

1382
	while (count > persistent_huge_pages(h)) {
1383 1384 1385 1386 1387 1388
		/*
		 * 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);
1389
		ret = alloc_fresh_huge_page(h, nodes_allowed);
1390 1391 1392 1393
		spin_lock(&hugetlb_lock);
		if (!ret)
			goto out;

1394 1395 1396
		/* Bail for signals. Probably ctrl-c from user */
		if (signal_pending(current))
			goto out;
1397 1398 1399 1400 1401 1402 1403 1404
	}

	/*
	 * 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.
1405 1406 1407 1408 1409 1410 1411 1412
	 *
	 * 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.
1413
	 */
1414
	min_count = h->resv_huge_pages + h->nr_huge_pages - h->free_huge_pages;
1415
	min_count = max(count, min_count);
1416
	try_to_free_low(h, min_count, nodes_allowed);
1417
	while (min_count < persistent_huge_pages(h)) {
1418
		if (!free_pool_huge_page(h, nodes_allowed, 0))
L
Linus Torvalds 已提交
1419 1420
			break;
	}
1421
	while (count < persistent_huge_pages(h)) {
1422
		if (!adjust_pool_surplus(h, nodes_allowed, 1))
1423 1424 1425
			break;
	}
out:
1426
	ret = persistent_huge_pages(h);
L
Linus Torvalds 已提交
1427
	spin_unlock(&hugetlb_lock);
1428
	return ret;
L
Linus Torvalds 已提交
1429 1430
}

1431 1432 1433 1434 1435 1436 1437 1438 1439 1440
#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];

1441 1442 1443
static struct hstate *kobj_to_node_hstate(struct kobject *kobj, int *nidp);

static struct hstate *kobj_to_hstate(struct kobject *kobj, int *nidp)
1444 1445
{
	int i;
1446

1447
	for (i = 0; i < HUGE_MAX_HSTATE; i++)
1448 1449 1450
		if (hstate_kobjs[i] == kobj) {
			if (nidp)
				*nidp = NUMA_NO_NODE;
1451
			return &hstates[i];
1452 1453 1454
		}

	return kobj_to_node_hstate(kobj, nidp);
1455 1456
}

1457
static ssize_t nr_hugepages_show_common(struct kobject *kobj,
1458 1459
					struct kobj_attribute *attr, char *buf)
{
1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470
	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);
1471
}
1472

1473 1474 1475
static ssize_t nr_hugepages_store_common(bool obey_mempolicy,
			struct kobject *kobj, struct kobj_attribute *attr,
			const char *buf, size_t len)
1476 1477
{
	int err;
1478
	int nid;
1479
	unsigned long count;
1480
	struct hstate *h;
1481
	NODEMASK_ALLOC(nodemask_t, nodes_allowed, GFP_KERNEL | __GFP_NORETRY);
1482

1483
	err = strict_strtoul(buf, 10, &count);
1484
	if (err)
1485
		goto out;
1486

1487
	h = kobj_to_hstate(kobj, &nid);
1488 1489 1490 1491 1492
	if (h->order >= MAX_ORDER) {
		err = -EINVAL;
		goto out;
	}

1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511
	if (nid == NUMA_NO_NODE) {
		/*
		 * global hstate attribute
		 */
		if (!(obey_mempolicy &&
				init_nodemask_of_mempolicy(nodes_allowed))) {
			NODEMASK_FREE(nodes_allowed);
			nodes_allowed = &node_states[N_HIGH_MEMORY];
		}
	} 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
		nodes_allowed = &node_states[N_HIGH_MEMORY];

1512
	h->max_huge_pages = set_max_huge_pages(h, count, nodes_allowed);
1513

1514
	if (nodes_allowed != &node_states[N_HIGH_MEMORY])
1515 1516 1517
		NODEMASK_FREE(nodes_allowed);

	return len;
1518 1519 1520
out:
	NODEMASK_FREE(nodes_allowed);
	return err;
1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532
}

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);
1533 1534 1535
}
HSTATE_ATTR(nr_hugepages);

1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550 1551 1552 1553 1554 1555 1556
#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


1557 1558 1559
static ssize_t nr_overcommit_hugepages_show(struct kobject *kobj,
					struct kobj_attribute *attr, char *buf)
{
1560
	struct hstate *h = kobj_to_hstate(kobj, NULL);
1561 1562
	return sprintf(buf, "%lu\n", h->nr_overcommit_huge_pages);
}
1563

1564 1565 1566 1567 1568
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;
1569
	struct hstate *h = kobj_to_hstate(kobj, NULL);
1570

1571 1572 1573
	if (h->order >= MAX_ORDER)
		return -EINVAL;

1574 1575
	err = strict_strtoul(buf, 10, &input);
	if (err)
1576
		return err;
1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588

	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)
{
1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599
	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);
1600 1601 1602 1603 1604 1605
}
HSTATE_ATTR_RO(free_hugepages);

static ssize_t resv_hugepages_show(struct kobject *kobj,
					struct kobj_attribute *attr, char *buf)
{
1606
	struct hstate *h = kobj_to_hstate(kobj, NULL);
1607 1608 1609 1610 1611 1612 1613
	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)
{
1614 1615 1616 1617 1618 1619 1620 1621 1622 1623 1624
	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);
1625 1626 1627 1628 1629 1630 1631 1632 1633
}
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,
1634 1635 1636
#ifdef CONFIG_NUMA
	&nr_hugepages_mempolicy_attr.attr,
#endif
1637 1638 1639 1640 1641 1642 1643
	NULL,
};

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

J
Jeff Mahoney 已提交
1644 1645 1646
static int hugetlb_sysfs_add_hstate(struct hstate *h, struct kobject *parent,
				    struct kobject **hstate_kobjs,
				    struct attribute_group *hstate_attr_group)
1647 1648
{
	int retval;
1649
	int hi = h - hstates;
1650

1651 1652
	hstate_kobjs[hi] = kobject_create_and_add(h->name, parent);
	if (!hstate_kobjs[hi])
1653 1654
		return -ENOMEM;

1655
	retval = sysfs_create_group(hstate_kobjs[hi], hstate_attr_group);
1656
	if (retval)
1657
		kobject_put(hstate_kobjs[hi]);
1658 1659 1660 1661 1662 1663 1664 1665 1666 1667 1668 1669 1670 1671

	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) {
1672 1673
		err = hugetlb_sysfs_add_hstate(h, hugepages_kobj,
					 hstate_kobjs, &hstate_attr_group);
1674 1675 1676 1677 1678 1679
		if (err)
			printk(KERN_ERR "Hugetlb: Unable to add hstate %s",
								h->name);
	}
}

1680 1681 1682 1683
#ifdef CONFIG_NUMA

/*
 * node_hstate/s - associate per node hstate attributes, via their kobjects,
1684 1685 1686
 * 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
1687 1688 1689 1690 1691 1692 1693 1694 1695
 * 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];

/*
1696
 * A subset of global hstate attributes for node devices
1697 1698 1699 1700 1701 1702 1703 1704 1705 1706 1707 1708 1709
 */
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,
};

/*
1710
 * kobj_to_node_hstate - lookup global hstate for node device hstate attr kobj.
1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724 1725 1726 1727 1728 1729 1730 1731 1732
 * 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;
}

/*
1733
 * Unregister hstate attributes from a single node device.
1734 1735 1736 1737 1738
 * No-op if no hstate attributes attached.
 */
void hugetlb_unregister_node(struct node *node)
{
	struct hstate *h;
1739
	struct node_hstate *nhs = &node_hstates[node->dev.id];
1740 1741

	if (!nhs->hugepages_kobj)
1742
		return;		/* no hstate attributes */
1743 1744 1745 1746 1747 1748 1749 1750 1751 1752 1753 1754

	for_each_hstate(h)
		if (nhs->hstate_kobjs[h - hstates]) {
			kobject_put(nhs->hstate_kobjs[h - hstates]);
			nhs->hstate_kobjs[h - hstates] = NULL;
		}

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

/*
1755
 * hugetlb module exit:  unregister hstate attributes from node devices
1756 1757 1758 1759 1760 1761 1762
 * that have them.
 */
static void hugetlb_unregister_all_nodes(void)
{
	int nid;

	/*
1763
	 * disable node device registrations.
1764 1765 1766 1767 1768 1769 1770 1771 1772 1773 1774
	 */
	register_hugetlbfs_with_node(NULL, NULL);

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

/*
1775
 * Register hstate attributes for a single node device.
1776 1777 1778 1779 1780
 * No-op if attributes already registered.
 */
void hugetlb_register_node(struct node *node)
{
	struct hstate *h;
1781
	struct node_hstate *nhs = &node_hstates[node->dev.id];
1782 1783 1784 1785 1786 1787
	int err;

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

	nhs->hugepages_kobj = kobject_create_and_add("hugepages",
1788
							&node->dev.kobj);
1789 1790 1791 1792 1793 1794 1795 1796 1797 1798
	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) {
			printk(KERN_ERR "Hugetlb: Unable to add hstate %s"
					" for node %d\n",
1799
						h->name, node->dev.id);
1800 1801 1802 1803 1804 1805 1806
			hugetlb_unregister_node(node);
			break;
		}
	}
}

/*
1807
 * hugetlb init time:  register hstate attributes for all registered node
1808 1809
 * devices of nodes that have memory.  All on-line nodes should have
 * registered their associated device by this time.
1810 1811 1812 1813 1814
 */
static void hugetlb_register_all_nodes(void)
{
	int nid;

1815
	for_each_node_state(nid, N_HIGH_MEMORY) {
1816
		struct node *node = &node_devices[nid];
1817
		if (node->dev.id == nid)
1818 1819 1820 1821
			hugetlb_register_node(node);
	}

	/*
1822
	 * Let the node device driver know we're here so it can
1823 1824 1825 1826 1827 1828 1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843
	 * [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

1844 1845 1846 1847
static void __exit hugetlb_exit(void)
{
	struct hstate *h;

1848 1849
	hugetlb_unregister_all_nodes();

1850 1851 1852 1853 1854 1855 1856 1857 1858 1859
	for_each_hstate(h) {
		kobject_put(hstate_kobjs[h - hstates]);
	}

	kobject_put(hugepages_kobj);
}
module_exit(hugetlb_exit);

static int __init hugetlb_init(void)
{
1860 1861 1862 1863 1864 1865
	/* 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;
1866

1867 1868 1869 1870
	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);
1871
	}
1872 1873 1874
	default_hstate_idx = size_to_hstate(default_hstate_size) - hstates;
	if (default_hstate_max_huge_pages)
		default_hstate.max_huge_pages = default_hstate_max_huge_pages;
1875 1876 1877

	hugetlb_init_hstates();

1878 1879
	gather_bootmem_prealloc();

1880 1881 1882 1883
	report_hugepages();

	hugetlb_sysfs_init();

1884 1885
	hugetlb_register_all_nodes();

1886 1887 1888 1889 1890 1891 1892 1893
	return 0;
}
module_init(hugetlb_init);

/* Should be called on processing a hugepagesz=... option */
void __init hugetlb_add_hstate(unsigned order)
{
	struct hstate *h;
1894 1895
	unsigned long i;

1896 1897 1898 1899 1900 1901 1902 1903 1904
	if (size_to_hstate(PAGE_SIZE << order)) {
		printk(KERN_WARNING "hugepagesz= specified twice, ignoring\n");
		return;
	}
	BUG_ON(max_hstate >= HUGE_MAX_HSTATE);
	BUG_ON(order == 0);
	h = &hstates[max_hstate++];
	h->order = order;
	h->mask = ~((1ULL << (order + PAGE_SHIFT)) - 1);
1905 1906 1907 1908
	h->nr_huge_pages = 0;
	h->free_huge_pages = 0;
	for (i = 0; i < MAX_NUMNODES; ++i)
		INIT_LIST_HEAD(&h->hugepage_freelists[i]);
1909 1910
	h->next_nid_to_alloc = first_node(node_states[N_HIGH_MEMORY]);
	h->next_nid_to_free = first_node(node_states[N_HIGH_MEMORY]);
1911 1912
	snprintf(h->name, HSTATE_NAME_LEN, "hugepages-%lukB",
					huge_page_size(h)/1024);
1913

1914 1915 1916
	parsed_hstate = h;
}

1917
static int __init hugetlb_nrpages_setup(char *s)
1918 1919
{
	unsigned long *mhp;
1920
	static unsigned long *last_mhp;
1921 1922 1923 1924 1925 1926 1927 1928 1929 1930

	/*
	 * !max_hstate means we haven't parsed a hugepagesz= parameter yet,
	 * so this hugepages= parameter goes to the "default hstate".
	 */
	if (!max_hstate)
		mhp = &default_hstate_max_huge_pages;
	else
		mhp = &parsed_hstate->max_huge_pages;

1931 1932 1933 1934 1935 1936
	if (mhp == last_mhp) {
		printk(KERN_WARNING "hugepages= specified twice without "
			"interleaving hugepagesz=, ignoring\n");
		return 1;
	}

1937 1938 1939
	if (sscanf(s, "%lu", mhp) <= 0)
		*mhp = 0;

1940 1941 1942 1943 1944 1945 1946 1947 1948 1949
	/*
	 * 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.
	 */
	if (max_hstate && parsed_hstate->order >= MAX_ORDER)
		hugetlb_hstate_alloc_pages(parsed_hstate);

	last_mhp = mhp;

1950 1951
	return 1;
}
1952 1953 1954 1955 1956 1957 1958 1959
__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);
1960

1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972
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
1973 1974 1975
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
1976
{
1977 1978
	struct hstate *h = &default_hstate;
	unsigned long tmp;
1979
	int ret;
1980

1981
	tmp = h->max_huge_pages;
1982

1983 1984 1985
	if (write && h->order >= MAX_ORDER)
		return -EINVAL;

1986 1987
	table->data = &tmp;
	table->maxlen = sizeof(unsigned long);
1988 1989 1990
	ret = proc_doulongvec_minmax(table, write, buffer, length, ppos);
	if (ret)
		goto out;
1991

1992
	if (write) {
1993 1994
		NODEMASK_ALLOC(nodemask_t, nodes_allowed,
						GFP_KERNEL | __GFP_NORETRY);
1995 1996 1997 1998 1999 2000 2001 2002 2003 2004
		if (!(obey_mempolicy &&
			       init_nodemask_of_mempolicy(nodes_allowed))) {
			NODEMASK_FREE(nodes_allowed);
			nodes_allowed = &node_states[N_HIGH_MEMORY];
		}
		h->max_huge_pages = set_max_huge_pages(h, tmp, nodes_allowed);

		if (nodes_allowed != &node_states[N_HIGH_MEMORY])
			NODEMASK_FREE(nodes_allowed);
	}
2005 2006
out:
	return ret;
L
Linus Torvalds 已提交
2007
}
2008

2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025
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 */

2026
int hugetlb_treat_movable_handler(struct ctl_table *table, int write,
2027
			void __user *buffer,
2028 2029
			size_t *length, loff_t *ppos)
{
2030
	proc_dointvec(table, write, buffer, length, ppos);
2031 2032 2033 2034 2035 2036 2037
	if (hugepages_treat_as_movable)
		htlb_alloc_mask = GFP_HIGHUSER_MOVABLE;
	else
		htlb_alloc_mask = GFP_HIGHUSER;
	return 0;
}

2038
int hugetlb_overcommit_handler(struct ctl_table *table, int write,
2039
			void __user *buffer,
2040 2041
			size_t *length, loff_t *ppos)
{
2042
	struct hstate *h = &default_hstate;
2043
	unsigned long tmp;
2044
	int ret;
2045

2046
	tmp = h->nr_overcommit_huge_pages;
2047

2048 2049 2050
	if (write && h->order >= MAX_ORDER)
		return -EINVAL;

2051 2052
	table->data = &tmp;
	table->maxlen = sizeof(unsigned long);
2053 2054 2055
	ret = proc_doulongvec_minmax(table, write, buffer, length, ppos);
	if (ret)
		goto out;
2056 2057 2058 2059 2060 2061

	if (write) {
		spin_lock(&hugetlb_lock);
		h->nr_overcommit_huge_pages = tmp;
		spin_unlock(&hugetlb_lock);
	}
2062 2063
out:
	return ret;
2064 2065
}

L
Linus Torvalds 已提交
2066 2067
#endif /* CONFIG_SYSCTL */

2068
void hugetlb_report_meminfo(struct seq_file *m)
L
Linus Torvalds 已提交
2069
{
2070
	struct hstate *h = &default_hstate;
2071
	seq_printf(m,
2072 2073 2074 2075 2076
			"HugePages_Total:   %5lu\n"
			"HugePages_Free:    %5lu\n"
			"HugePages_Rsvd:    %5lu\n"
			"HugePages_Surp:    %5lu\n"
			"Hugepagesize:   %8lu kB\n",
2077 2078 2079 2080 2081
			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
2082 2083 2084 2085
}

int hugetlb_report_node_meminfo(int nid, char *buf)
{
2086
	struct hstate *h = &default_hstate;
L
Linus Torvalds 已提交
2087 2088
	return sprintf(buf,
		"Node %d HugePages_Total: %5u\n"
2089 2090
		"Node %d HugePages_Free:  %5u\n"
		"Node %d HugePages_Surp:  %5u\n",
2091 2092 2093
		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
2094 2095 2096 2097 2098
}

/* Return the number pages of memory we physically have, in PAGE_SIZE units. */
unsigned long hugetlb_total_pages(void)
{
2099 2100
	struct hstate *h = &default_hstate;
	return h->nr_huge_pages * pages_per_huge_page(h);
L
Linus Torvalds 已提交
2101 2102
}

2103
static int hugetlb_acct_memory(struct hstate *h, long delta)
M
Mel Gorman 已提交
2104 2105 2106 2107 2108 2109 2110 2111 2112 2113 2114 2115 2116 2117 2118 2119 2120 2121 2122 2123 2124 2125
{
	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) {
2126
		if (gather_surplus_pages(h, delta) < 0)
M
Mel Gorman 已提交
2127 2128
			goto out;

2129 2130
		if (delta > cpuset_mems_nr(h->free_huge_pages_node)) {
			return_unused_surplus_pages(h, delta);
M
Mel Gorman 已提交
2131 2132 2133 2134 2135 2136
			goto out;
		}
	}

	ret = 0;
	if (delta < 0)
2137
		return_unused_surplus_pages(h, (unsigned long) -delta);
M
Mel Gorman 已提交
2138 2139 2140 2141 2142 2143

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

2144 2145 2146 2147 2148 2149 2150 2151
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
2152
	 * has a reference to the reservation map it cannot disappear until
2153 2154 2155 2156 2157 2158 2159
	 * after this open call completes.  It is therefore safe to take a
	 * new reference here without additional locking.
	 */
	if (reservations)
		kref_get(&reservations->refs);
}

2160 2161 2162 2163 2164 2165 2166 2167 2168
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);
}

2169 2170
static void hugetlb_vm_op_close(struct vm_area_struct *vma)
{
2171
	struct hstate *h = hstate_vma(vma);
2172
	struct resv_map *reservations = vma_resv_map(vma);
2173
	struct hugepage_subpool *spool = subpool_vma(vma);
2174 2175 2176 2177 2178
	unsigned long reserve;
	unsigned long start;
	unsigned long end;

	if (reservations) {
2179 2180
		start = vma_hugecache_offset(h, vma, vma->vm_start);
		end = vma_hugecache_offset(h, vma, vma->vm_end);
2181 2182 2183 2184

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

2185
		resv_map_put(vma);
2186

2187
		if (reserve) {
2188
			hugetlb_acct_memory(h, -reserve);
2189
			hugepage_subpool_put_pages(spool, reserve);
2190
		}
2191
	}
2192 2193
}

L
Linus Torvalds 已提交
2194 2195 2196 2197 2198 2199
/*
 * 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
2200
static int hugetlb_vm_op_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
L
Linus Torvalds 已提交
2201 2202
{
	BUG();
N
Nick Piggin 已提交
2203
	return 0;
L
Linus Torvalds 已提交
2204 2205
}

2206
const struct vm_operations_struct hugetlb_vm_ops = {
N
Nick Piggin 已提交
2207
	.fault = hugetlb_vm_op_fault,
2208
	.open = hugetlb_vm_op_open,
2209
	.close = hugetlb_vm_op_close,
L
Linus Torvalds 已提交
2210 2211
};

2212 2213
static pte_t make_huge_pte(struct vm_area_struct *vma, struct page *page,
				int writable)
D
David Gibson 已提交
2214 2215 2216
{
	pte_t entry;

2217
	if (writable) {
D
David Gibson 已提交
2218 2219 2220
		entry =
		    pte_mkwrite(pte_mkdirty(mk_pte(page, vma->vm_page_prot)));
	} else {
2221
		entry = huge_pte_wrprotect(mk_pte(page, vma->vm_page_prot));
D
David Gibson 已提交
2222 2223 2224
	}
	entry = pte_mkyoung(entry);
	entry = pte_mkhuge(entry);
2225
	entry = arch_make_huge_pte(entry, vma, page, writable);
D
David Gibson 已提交
2226 2227 2228 2229

	return entry;
}

2230 2231 2232 2233 2234
static void set_huge_ptep_writable(struct vm_area_struct *vma,
				   unsigned long address, pte_t *ptep)
{
	pte_t entry;

2235
	entry = pte_mkwrite(pte_mkdirty(huge_ptep_get(ptep)));
2236
	if (huge_ptep_set_access_flags(vma, address, ptep, entry, 1))
2237
		update_mmu_cache(vma, address, ptep);
2238 2239 2240
}


D
David Gibson 已提交
2241 2242 2243 2244 2245
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;
2246
	unsigned long addr;
2247
	int cow;
2248 2249
	struct hstate *h = hstate_vma(vma);
	unsigned long sz = huge_page_size(h);
2250 2251

	cow = (vma->vm_flags & (VM_SHARED | VM_MAYWRITE)) == VM_MAYWRITE;
D
David Gibson 已提交
2252

2253
	for (addr = vma->vm_start; addr < vma->vm_end; addr += sz) {
H
Hugh Dickins 已提交
2254 2255 2256
		src_pte = huge_pte_offset(src, addr);
		if (!src_pte)
			continue;
2257
		dst_pte = huge_pte_alloc(dst, addr, sz);
D
David Gibson 已提交
2258 2259
		if (!dst_pte)
			goto nomem;
2260 2261 2262 2263 2264

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

H
Hugh Dickins 已提交
2265
		spin_lock(&dst->page_table_lock);
N
Nick Piggin 已提交
2266
		spin_lock_nested(&src->page_table_lock, SINGLE_DEPTH_NESTING);
2267
		if (!huge_pte_none(huge_ptep_get(src_pte))) {
2268
			if (cow)
2269 2270
				huge_ptep_set_wrprotect(src, addr, src_pte);
			entry = huge_ptep_get(src_pte);
2271 2272
			ptepage = pte_page(entry);
			get_page(ptepage);
2273
			page_dup_rmap(ptepage);
2274 2275 2276
			set_huge_pte_at(dst, addr, dst_pte, entry);
		}
		spin_unlock(&src->page_table_lock);
H
Hugh Dickins 已提交
2277
		spin_unlock(&dst->page_table_lock);
D
David Gibson 已提交
2278 2279 2280 2281 2282 2283 2284
	}
	return 0;

nomem:
	return -ENOMEM;
}

N
Naoya Horiguchi 已提交
2285 2286 2287 2288 2289 2290 2291
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);
2292
	if (non_swap_entry(swp) && is_migration_entry(swp))
N
Naoya Horiguchi 已提交
2293
		return 1;
2294
	else
N
Naoya Horiguchi 已提交
2295 2296 2297
		return 0;
}

2298 2299 2300 2301 2302 2303 2304
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);
2305
	if (non_swap_entry(swp) && is_hwpoison_entry(swp))
2306
		return 1;
2307
	else
2308 2309 2310
		return 0;
}

2311
void __unmap_hugepage_range(struct vm_area_struct *vma, unsigned long start,
2312
			    unsigned long end, struct page *ref_page)
D
David Gibson 已提交
2313 2314 2315
{
	struct mm_struct *mm = vma->vm_mm;
	unsigned long address;
2316
	pte_t *ptep;
D
David Gibson 已提交
2317 2318
	pte_t pte;
	struct page *page;
2319
	struct page *tmp;
2320 2321 2322
	struct hstate *h = hstate_vma(vma);
	unsigned long sz = huge_page_size(h);

2323
	/*
2324
	 * A page gathering list, protected by per file i_mmap_mutex. The
2325 2326 2327
	 * lock is used to avoid list corruption from multiple unmapping
	 * of the same page since we are using page->lru.
	 */
2328
	LIST_HEAD(page_list);
D
David Gibson 已提交
2329 2330

	WARN_ON(!is_vm_hugetlb_page(vma));
2331 2332
	BUG_ON(start & ~huge_page_mask(h));
	BUG_ON(end & ~huge_page_mask(h));
D
David Gibson 已提交
2333

A
Andrea Arcangeli 已提交
2334
	mmu_notifier_invalidate_range_start(mm, start, end);
2335
	spin_lock(&mm->page_table_lock);
2336
	for (address = start; address < end; address += sz) {
2337
		ptep = huge_pte_offset(mm, address);
A
Adam Litke 已提交
2338
		if (!ptep)
2339 2340
			continue;

2341 2342 2343
		if (huge_pmd_unshare(mm, &address, ptep))
			continue;

2344 2345 2346 2347 2348 2349 2350 2351 2352 2353 2354
		pte = huge_ptep_get(ptep);
		if (huge_pte_none(pte))
			continue;

		/*
		 * HWPoisoned hugepage is already unmapped and dropped reference
		 */
		if (unlikely(is_hugetlb_entry_hwpoisoned(pte)))
			continue;

		page = pte_page(pte);
2355 2356 2357 2358 2359 2360 2361 2362 2363 2364 2365 2366 2367 2368 2369 2370 2371
		/*
		 * 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);
		}

2372
		pte = huge_ptep_get_and_clear(mm, address, ptep);
2373 2374
		if (pte_dirty(pte))
			set_page_dirty(page);
2375
		list_add(&page->lru, &page_list);
2376 2377 2378 2379

		/* Bail out after unmapping reference page if supplied */
		if (ref_page)
			break;
D
David Gibson 已提交
2380
	}
2381
	flush_tlb_range(vma, start, end);
2382
	spin_unlock(&mm->page_table_lock);
A
Andrea Arcangeli 已提交
2383
	mmu_notifier_invalidate_range_end(mm, start, end);
2384
	list_for_each_entry_safe(page, tmp, &page_list, lru) {
2385
		page_remove_rmap(page);
2386 2387 2388
		list_del(&page->lru);
		put_page(page);
	}
L
Linus Torvalds 已提交
2389
}
D
David Gibson 已提交
2390

2391
void unmap_hugepage_range(struct vm_area_struct *vma, unsigned long start,
2392
			  unsigned long end, struct page *ref_page)
2393
{
2394
	mutex_lock(&vma->vm_file->f_mapping->i_mmap_mutex);
2395
	__unmap_hugepage_range(vma, start, end, ref_page);
2396
	mutex_unlock(&vma->vm_file->f_mapping->i_mmap_mutex);
2397 2398
}

2399 2400 2401 2402 2403 2404
/*
 * 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.
 */
2405 2406
static int unmap_ref_private(struct mm_struct *mm, struct vm_area_struct *vma,
				struct page *page, unsigned long address)
2407
{
2408
	struct hstate *h = hstate_vma(vma);
2409 2410 2411 2412 2413 2414 2415 2416 2417
	struct vm_area_struct *iter_vma;
	struct address_space *mapping;
	struct prio_tree_iter iter;
	pgoff_t pgoff;

	/*
	 * vm_pgoff is in PAGE_SIZE units, hence the different calculation
	 * from page cache lookup which is in HPAGE_SIZE units.
	 */
2418
	address = address & huge_page_mask(h);
2419
	pgoff = vma_hugecache_offset(h, vma, address);
2420
	mapping = vma->vm_file->f_dentry->d_inode->i_mapping;
2421

2422 2423 2424 2425 2426
	/*
	 * 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
	 */
2427
	mutex_lock(&mapping->i_mmap_mutex);
2428 2429 2430 2431 2432 2433 2434 2435 2436 2437 2438 2439 2440
	vma_prio_tree_foreach(iter_vma, &iter, &mapping->i_mmap, pgoff, pgoff) {
		/* 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))
2441
			__unmap_hugepage_range(iter_vma,
2442
				address, address + huge_page_size(h),
2443 2444
				page);
	}
2445
	mutex_unlock(&mapping->i_mmap_mutex);
2446 2447 2448 2449

	return 1;
}

2450 2451
/*
 * Hugetlb_cow() should be called with page lock of the original hugepage held.
2452 2453 2454
 * 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.
2455
 */
2456
static int hugetlb_cow(struct mm_struct *mm, struct vm_area_struct *vma,
2457 2458
			unsigned long address, pte_t *ptep, pte_t pte,
			struct page *pagecache_page)
2459
{
2460
	struct hstate *h = hstate_vma(vma);
2461
	struct page *old_page, *new_page;
2462
	int avoidcopy;
2463
	int outside_reserve = 0;
2464 2465 2466

	old_page = pte_page(pte);

2467
retry_avoidcopy:
2468 2469
	/* If no-one else is actually using this page, avoid the copy
	 * and just make the page writable */
2470
	avoidcopy = (page_mapcount(old_page) == 1);
2471
	if (avoidcopy) {
2472 2473
		if (PageAnon(old_page))
			page_move_anon_rmap(old_page, vma, address);
2474
		set_huge_ptep_writable(vma, address, ptep);
N
Nick Piggin 已提交
2475
		return 0;
2476 2477
	}

2478 2479 2480 2481 2482 2483 2484 2485 2486
	/*
	 * 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.
	 */
2487
	if (!(vma->vm_flags & VM_MAYSHARE) &&
2488 2489 2490 2491
			is_vma_resv_set(vma, HPAGE_RESV_OWNER) &&
			old_page != pagecache_page)
		outside_reserve = 1;

2492
	page_cache_get(old_page);
2493 2494 2495

	/* Drop page_table_lock as buddy allocator may be called */
	spin_unlock(&mm->page_table_lock);
2496
	new_page = alloc_huge_page(vma, address, outside_reserve);
2497

2498
	if (IS_ERR(new_page)) {
2499
		page_cache_release(old_page);
2500 2501 2502 2503 2504 2505 2506 2507 2508 2509 2510 2511

		/*
		 * 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));
2512
				spin_lock(&mm->page_table_lock);
2513 2514 2515 2516 2517 2518 2519 2520
				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;
2521 2522 2523 2524
			}
			WARN_ON_ONCE(1);
		}

2525 2526
		/* Caller expects lock to be held */
		spin_lock(&mm->page_table_lock);
2527
		return -PTR_ERR(new_page);
2528 2529
	}

2530 2531 2532 2533
	/*
	 * When the original hugepage is shared one, it does not have
	 * anon_vma prepared.
	 */
2534
	if (unlikely(anon_vma_prepare(vma))) {
2535 2536
		page_cache_release(new_page);
		page_cache_release(old_page);
2537 2538
		/* Caller expects lock to be held */
		spin_lock(&mm->page_table_lock);
2539
		return VM_FAULT_OOM;
2540
	}
2541

A
Andrea Arcangeli 已提交
2542 2543
	copy_user_huge_page(new_page, old_page, address, vma,
			    pages_per_huge_page(h));
N
Nick Piggin 已提交
2544
	__SetPageUptodate(new_page);
2545

2546 2547 2548 2549 2550
	/*
	 * Retake the page_table_lock to check for racing updates
	 * before the page tables are altered
	 */
	spin_lock(&mm->page_table_lock);
2551
	ptep = huge_pte_offset(mm, address & huge_page_mask(h));
2552
	if (likely(pte_same(huge_ptep_get(ptep), pte))) {
2553
		/* Break COW */
2554 2555 2556
		mmu_notifier_invalidate_range_start(mm,
			address & huge_page_mask(h),
			(address & huge_page_mask(h)) + huge_page_size(h));
2557
		huge_ptep_clear_flush(vma, address, ptep);
2558 2559
		set_huge_pte_at(mm, address, ptep,
				make_huge_pte(vma, new_page, 1));
2560
		page_remove_rmap(old_page);
2561
		hugepage_add_new_anon_rmap(new_page, vma, address);
2562 2563
		/* Make the old page be freed below */
		new_page = old_page;
2564 2565 2566
		mmu_notifier_invalidate_range_end(mm,
			address & huge_page_mask(h),
			(address & huge_page_mask(h)) + huge_page_size(h));
2567 2568 2569
	}
	page_cache_release(new_page);
	page_cache_release(old_page);
N
Nick Piggin 已提交
2570
	return 0;
2571 2572
}

2573
/* Return the pagecache page at a given address within a VMA */
2574 2575
static struct page *hugetlbfs_pagecache_page(struct hstate *h,
			struct vm_area_struct *vma, unsigned long address)
2576 2577
{
	struct address_space *mapping;
2578
	pgoff_t idx;
2579 2580

	mapping = vma->vm_file->f_mapping;
2581
	idx = vma_hugecache_offset(h, vma, address);
2582 2583 2584 2585

	return find_lock_page(mapping, idx);
}

H
Hugh Dickins 已提交
2586 2587 2588 2589 2590
/*
 * 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
2591 2592 2593 2594 2595 2596 2597 2598 2599 2600 2601 2602 2603 2604 2605
			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;
}

2606
static int hugetlb_no_page(struct mm_struct *mm, struct vm_area_struct *vma,
2607
			unsigned long address, pte_t *ptep, unsigned int flags)
2608
{
2609
	struct hstate *h = hstate_vma(vma);
2610
	int ret = VM_FAULT_SIGBUS;
2611
	int anon_rmap = 0;
2612
	pgoff_t idx;
A
Adam Litke 已提交
2613 2614 2615
	unsigned long size;
	struct page *page;
	struct address_space *mapping;
2616
	pte_t new_pte;
A
Adam Litke 已提交
2617

2618 2619 2620
	/*
	 * 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
2621
	 * COW. Warn that such a situation has occurred as it may not be obvious
2622 2623 2624 2625 2626 2627 2628 2629
	 */
	if (is_vma_resv_set(vma, HPAGE_RESV_UNMAPPED)) {
		printk(KERN_WARNING
			"PID %d killed due to inadequate hugepage pool\n",
			current->pid);
		return ret;
	}

A
Adam Litke 已提交
2630
	mapping = vma->vm_file->f_mapping;
2631
	idx = vma_hugecache_offset(h, vma, address);
A
Adam Litke 已提交
2632 2633 2634 2635 2636

	/*
	 * Use page lock to guard against racing truncation
	 * before we get page_table_lock.
	 */
2637 2638 2639
retry:
	page = find_lock_page(mapping, idx);
	if (!page) {
2640
		size = i_size_read(mapping->host) >> huge_page_shift(h);
2641 2642
		if (idx >= size)
			goto out;
2643
		page = alloc_huge_page(vma, address, 0);
2644 2645
		if (IS_ERR(page)) {
			ret = -PTR_ERR(page);
2646 2647
			goto out;
		}
A
Andrea Arcangeli 已提交
2648
		clear_huge_page(page, address, pages_per_huge_page(h));
N
Nick Piggin 已提交
2649
		__SetPageUptodate(page);
2650

2651
		if (vma->vm_flags & VM_MAYSHARE) {
2652
			int err;
K
Ken Chen 已提交
2653
			struct inode *inode = mapping->host;
2654 2655 2656 2657 2658 2659 2660 2661

			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
2662 2663

			spin_lock(&inode->i_lock);
2664
			inode->i_blocks += blocks_per_huge_page(h);
K
Ken Chen 已提交
2665
			spin_unlock(&inode->i_lock);
2666
		} else {
2667
			lock_page(page);
2668 2669 2670 2671
			if (unlikely(anon_vma_prepare(vma))) {
				ret = VM_FAULT_OOM;
				goto backout_unlocked;
			}
2672
			anon_rmap = 1;
2673
		}
2674
	} else {
2675 2676 2677 2678 2679 2680
		/*
		 * 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))) {
2681
			ret = VM_FAULT_HWPOISON |
2682
			      VM_FAULT_SET_HINDEX(h - hstates);
2683 2684
			goto backout_unlocked;
		}
2685
	}
2686

2687 2688 2689 2690 2691 2692
	/*
	 * 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.
	 */
2693
	if ((flags & FAULT_FLAG_WRITE) && !(vma->vm_flags & VM_SHARED))
2694 2695 2696 2697
		if (vma_needs_reservation(h, vma, address) < 0) {
			ret = VM_FAULT_OOM;
			goto backout_unlocked;
		}
2698

2699
	spin_lock(&mm->page_table_lock);
2700
	size = i_size_read(mapping->host) >> huge_page_shift(h);
A
Adam Litke 已提交
2701 2702 2703
	if (idx >= size)
		goto backout;

N
Nick Piggin 已提交
2704
	ret = 0;
2705
	if (!huge_pte_none(huge_ptep_get(ptep)))
A
Adam Litke 已提交
2706 2707
		goto backout;

2708 2709 2710 2711
	if (anon_rmap)
		hugepage_add_new_anon_rmap(page, vma, address);
	else
		page_dup_rmap(page);
2712 2713 2714 2715
	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);

2716
	if ((flags & FAULT_FLAG_WRITE) && !(vma->vm_flags & VM_SHARED)) {
2717
		/* Optimization, do the COW without a second fault */
2718
		ret = hugetlb_cow(mm, vma, address, ptep, new_pte, page);
2719 2720
	}

2721
	spin_unlock(&mm->page_table_lock);
A
Adam Litke 已提交
2722 2723
	unlock_page(page);
out:
2724
	return ret;
A
Adam Litke 已提交
2725 2726 2727

backout:
	spin_unlock(&mm->page_table_lock);
2728
backout_unlocked:
A
Adam Litke 已提交
2729 2730 2731
	unlock_page(page);
	put_page(page);
	goto out;
2732 2733
}

2734
int hugetlb_fault(struct mm_struct *mm, struct vm_area_struct *vma,
2735
			unsigned long address, unsigned int flags)
2736 2737 2738
{
	pte_t *ptep;
	pte_t entry;
2739
	int ret;
2740
	struct page *page = NULL;
2741
	struct page *pagecache_page = NULL;
2742
	static DEFINE_MUTEX(hugetlb_instantiation_mutex);
2743
	struct hstate *h = hstate_vma(vma);
2744

2745 2746
	address &= huge_page_mask(h);

2747 2748 2749
	ptep = huge_pte_offset(mm, address);
	if (ptep) {
		entry = huge_ptep_get(ptep);
N
Naoya Horiguchi 已提交
2750 2751 2752 2753
		if (unlikely(is_hugetlb_entry_migration(entry))) {
			migration_entry_wait(mm, (pmd_t *)ptep, address);
			return 0;
		} else if (unlikely(is_hugetlb_entry_hwpoisoned(entry)))
2754
			return VM_FAULT_HWPOISON_LARGE |
2755
			       VM_FAULT_SET_HINDEX(h - hstates);
2756 2757
	}

2758
	ptep = huge_pte_alloc(mm, address, huge_page_size(h));
2759 2760 2761
	if (!ptep)
		return VM_FAULT_OOM;

2762 2763 2764 2765 2766 2767
	/*
	 * 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);
2768 2769
	entry = huge_ptep_get(ptep);
	if (huge_pte_none(entry)) {
2770
		ret = hugetlb_no_page(mm, vma, address, ptep, flags);
2771
		goto out_mutex;
2772
	}
2773

N
Nick Piggin 已提交
2774
	ret = 0;
2775

2776 2777 2778 2779 2780 2781 2782 2783
	/*
	 * 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.
	 */
2784
	if ((flags & FAULT_FLAG_WRITE) && !pte_write(entry)) {
2785 2786
		if (vma_needs_reservation(h, vma, address) < 0) {
			ret = VM_FAULT_OOM;
2787
			goto out_mutex;
2788
		}
2789

2790
		if (!(vma->vm_flags & VM_MAYSHARE))
2791 2792 2793 2794
			pagecache_page = hugetlbfs_pagecache_page(h,
								vma, address);
	}

2795 2796 2797 2798 2799 2800 2801 2802
	/*
	 * 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);
2803
	get_page(page);
2804
	if (page != pagecache_page)
2805 2806
		lock_page(page);

2807 2808
	spin_lock(&mm->page_table_lock);
	/* Check for a racing update before calling hugetlb_cow */
2809 2810 2811 2812
	if (unlikely(!pte_same(entry, huge_ptep_get(ptep))))
		goto out_page_table_lock;


2813
	if (flags & FAULT_FLAG_WRITE) {
2814
		if (!pte_write(entry)) {
2815 2816
			ret = hugetlb_cow(mm, vma, address, ptep, entry,
							pagecache_page);
2817 2818 2819 2820 2821
			goto out_page_table_lock;
		}
		entry = pte_mkdirty(entry);
	}
	entry = pte_mkyoung(entry);
2822 2823
	if (huge_ptep_set_access_flags(vma, address, ptep, entry,
						flags & FAULT_FLAG_WRITE))
2824
		update_mmu_cache(vma, address, ptep);
2825 2826

out_page_table_lock:
2827
	spin_unlock(&mm->page_table_lock);
2828 2829 2830 2831 2832

	if (pagecache_page) {
		unlock_page(pagecache_page);
		put_page(pagecache_page);
	}
2833 2834
	if (page != pagecache_page)
		unlock_page(page);
2835
	put_page(page);
2836

2837
out_mutex:
2838
	mutex_unlock(&hugetlb_instantiation_mutex);
2839 2840

	return ret;
2841 2842
}

A
Andi Kleen 已提交
2843 2844 2845 2846 2847 2848 2849 2850 2851
/* 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;
}

D
David Gibson 已提交
2852 2853
int follow_hugetlb_page(struct mm_struct *mm, struct vm_area_struct *vma,
			struct page **pages, struct vm_area_struct **vmas,
2854
			unsigned long *position, int *length, int i,
H
Hugh Dickins 已提交
2855
			unsigned int flags)
D
David Gibson 已提交
2856
{
2857 2858
	unsigned long pfn_offset;
	unsigned long vaddr = *position;
D
David Gibson 已提交
2859
	int remainder = *length;
2860
	struct hstate *h = hstate_vma(vma);
D
David Gibson 已提交
2861

2862
	spin_lock(&mm->page_table_lock);
D
David Gibson 已提交
2863
	while (vaddr < vma->vm_end && remainder) {
A
Adam Litke 已提交
2864
		pte_t *pte;
H
Hugh Dickins 已提交
2865
		int absent;
A
Adam Litke 已提交
2866
		struct page *page;
D
David Gibson 已提交
2867

A
Adam Litke 已提交
2868 2869
		/*
		 * Some archs (sparc64, sh*) have multiple pte_ts to
H
Hugh Dickins 已提交
2870
		 * each hugepage.  We have to make sure we get the
A
Adam Litke 已提交
2871 2872
		 * first, for the page indexing below to work.
		 */
2873
		pte = huge_pte_offset(mm, vaddr & huge_page_mask(h));
H
Hugh Dickins 已提交
2874 2875 2876 2877
		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
2878 2879 2880 2881
		 * 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
2882
		 */
H
Hugh Dickins 已提交
2883 2884
		if (absent && (flags & FOLL_DUMP) &&
		    !hugetlbfs_pagecache_present(h, vma, vaddr)) {
H
Hugh Dickins 已提交
2885 2886 2887
			remainder = 0;
			break;
		}
D
David Gibson 已提交
2888

H
Hugh Dickins 已提交
2889 2890
		if (absent ||
		    ((flags & FOLL_WRITE) && !pte_write(huge_ptep_get(pte)))) {
A
Adam Litke 已提交
2891
			int ret;
D
David Gibson 已提交
2892

A
Adam Litke 已提交
2893
			spin_unlock(&mm->page_table_lock);
H
Hugh Dickins 已提交
2894 2895
			ret = hugetlb_fault(mm, vma, vaddr,
				(flags & FOLL_WRITE) ? FAULT_FLAG_WRITE : 0);
A
Adam Litke 已提交
2896
			spin_lock(&mm->page_table_lock);
2897
			if (!(ret & VM_FAULT_ERROR))
A
Adam Litke 已提交
2898
				continue;
D
David Gibson 已提交
2899

A
Adam Litke 已提交
2900 2901 2902 2903
			remainder = 0;
			break;
		}

2904
		pfn_offset = (vaddr & ~huge_page_mask(h)) >> PAGE_SHIFT;
2905
		page = pte_page(huge_ptep_get(pte));
2906
same_page:
2907
		if (pages) {
H
Hugh Dickins 已提交
2908
			pages[i] = mem_map_offset(page, pfn_offset);
K
KOSAKI Motohiro 已提交
2909
			get_page(pages[i]);
2910
		}
D
David Gibson 已提交
2911 2912 2913 2914 2915

		if (vmas)
			vmas[i] = vma;

		vaddr += PAGE_SIZE;
2916
		++pfn_offset;
D
David Gibson 已提交
2917 2918
		--remainder;
		++i;
2919
		if (vaddr < vma->vm_end && remainder &&
2920
				pfn_offset < pages_per_huge_page(h)) {
2921 2922 2923 2924 2925 2926
			/*
			 * We use pfn_offset to avoid touching the pageframes
			 * of this compound page.
			 */
			goto same_page;
		}
D
David Gibson 已提交
2927
	}
2928
	spin_unlock(&mm->page_table_lock);
D
David Gibson 已提交
2929 2930 2931
	*length = remainder;
	*position = vaddr;

H
Hugh Dickins 已提交
2932
	return i ? i : -EFAULT;
D
David Gibson 已提交
2933
}
2934 2935 2936 2937 2938 2939 2940 2941

void hugetlb_change_protection(struct vm_area_struct *vma,
		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;
2942
	struct hstate *h = hstate_vma(vma);
2943 2944 2945 2946

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

2947
	mutex_lock(&vma->vm_file->f_mapping->i_mmap_mutex);
2948
	spin_lock(&mm->page_table_lock);
2949
	for (; address < end; address += huge_page_size(h)) {
2950 2951 2952
		ptep = huge_pte_offset(mm, address);
		if (!ptep)
			continue;
2953 2954
		if (huge_pmd_unshare(mm, &address, ptep))
			continue;
2955
		if (!huge_pte_none(huge_ptep_get(ptep))) {
2956 2957 2958 2959 2960 2961
			pte = huge_ptep_get_and_clear(mm, address, ptep);
			pte = pte_mkhuge(pte_modify(pte, newprot));
			set_huge_pte_at(mm, address, ptep, pte);
		}
	}
	spin_unlock(&mm->page_table_lock);
2962
	mutex_unlock(&vma->vm_file->f_mapping->i_mmap_mutex);
2963 2964 2965 2966

	flush_tlb_range(vma, start, end);
}

2967 2968
int hugetlb_reserve_pages(struct inode *inode,
					long from, long to,
2969
					struct vm_area_struct *vma,
2970
					vm_flags_t vm_flags)
2971
{
2972
	long ret, chg;
2973
	struct hstate *h = hstate_inode(inode);
2974
	struct hugepage_subpool *spool = subpool_inode(inode);
2975

2976 2977 2978
	/*
	 * Only apply hugepage reservation if asked. At fault time, an
	 * attempt will be made for VM_NORESERVE to allocate a page
2979
	 * without using reserves
2980
	 */
2981
	if (vm_flags & VM_NORESERVE)
2982 2983
		return 0;

2984 2985 2986 2987 2988 2989
	/*
	 * 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
	 */
2990
	if (!vma || vma->vm_flags & VM_MAYSHARE)
2991
		chg = region_chg(&inode->i_mapping->private_list, from, to);
2992 2993 2994 2995 2996
	else {
		struct resv_map *resv_map = resv_map_alloc();
		if (!resv_map)
			return -ENOMEM;

2997
		chg = to - from;
2998

2999 3000 3001 3002
		set_vma_resv_map(vma, resv_map);
		set_vma_resv_flags(vma, HPAGE_RESV_OWNER);
	}

3003 3004 3005 3006
	if (chg < 0) {
		ret = chg;
		goto out_err;
	}
3007

3008
	/* There must be enough pages in the subpool for the mapping */
3009 3010 3011 3012
	if (hugepage_subpool_get_pages(spool, chg)) {
		ret = -ENOSPC;
		goto out_err;
	}
3013 3014

	/*
3015
	 * Check enough hugepages are available for the reservation.
3016
	 * Hand the pages back to the subpool if there are not
3017
	 */
3018
	ret = hugetlb_acct_memory(h, chg);
K
Ken Chen 已提交
3019
	if (ret < 0) {
3020
		hugepage_subpool_put_pages(spool, chg);
3021
		goto out_err;
K
Ken Chen 已提交
3022
	}
3023 3024 3025 3026 3027 3028 3029 3030 3031 3032 3033 3034

	/*
	 * 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
	 */
3035
	if (!vma || vma->vm_flags & VM_MAYSHARE)
3036
		region_add(&inode->i_mapping->private_list, from, to);
3037
	return 0;
3038
out_err:
3039 3040
	if (vma)
		resv_map_put(vma);
3041
	return ret;
3042 3043 3044 3045
}

void hugetlb_unreserve_pages(struct inode *inode, long offset, long freed)
{
3046
	struct hstate *h = hstate_inode(inode);
3047
	long chg = region_truncate(&inode->i_mapping->private_list, offset);
3048
	struct hugepage_subpool *spool = subpool_inode(inode);
K
Ken Chen 已提交
3049 3050

	spin_lock(&inode->i_lock);
3051
	inode->i_blocks -= (blocks_per_huge_page(h) * freed);
K
Ken Chen 已提交
3052 3053
	spin_unlock(&inode->i_lock);

3054
	hugepage_subpool_put_pages(spool, (chg - freed));
3055
	hugetlb_acct_memory(h, -(chg - freed));
3056
}
3057

3058 3059
#ifdef CONFIG_MEMORY_FAILURE

3060 3061 3062 3063 3064 3065 3066 3067 3068 3069 3070 3071 3072 3073
/* 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;
}

3074 3075 3076 3077
/*
 * This function is called from memory failure code.
 * Assume the caller holds page lock of the head page.
 */
3078
int dequeue_hwpoisoned_huge_page(struct page *hpage)
3079 3080 3081
{
	struct hstate *h = page_hstate(hpage);
	int nid = page_to_nid(hpage);
3082
	int ret = -EBUSY;
3083 3084

	spin_lock(&hugetlb_lock);
3085 3086
	if (is_hugepage_on_freelist(hpage)) {
		list_del(&hpage->lru);
3087
		set_page_refcounted(hpage);
3088 3089 3090 3091
		h->free_huge_pages--;
		h->free_huge_pages_node[nid]--;
		ret = 0;
	}
3092
	spin_unlock(&hugetlb_lock);
3093
	return ret;
3094
}
3095
#endif