rmap.c 40.0 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
/*
 * mm/rmap.c - physical to virtual reverse mappings
 *
 * Copyright 2001, Rik van Riel <riel@conectiva.com.br>
 * Released under the General Public License (GPL).
 *
 * Simple, low overhead reverse mapping scheme.
 * Please try to keep this thing as modular as possible.
 *
 * Provides methods for unmapping each kind of mapped page:
 * the anon methods track anonymous pages, and
 * the file methods track pages belonging to an inode.
 *
 * Original design by Rik van Riel <riel@conectiva.com.br> 2001
 * File methods by Dave McCracken <dmccr@us.ibm.com> 2003, 2004
 * Anonymous methods by Andrea Arcangeli <andrea@suse.de> 2004
H
Hugh Dickins 已提交
17
 * Contributions by Hugh Dickins 2003, 2004
L
Linus Torvalds 已提交
18 19 20 21 22
 */

/*
 * Lock ordering in mm:
 *
23
 * inode->i_mutex	(while writing or truncating, not reading or faulting)
24 25 26 27 28 29 30 31 32 33 34 35 36 37 38
 *   inode->i_alloc_sem (vmtruncate_range)
 *   mm->mmap_sem
 *     page->flags PG_locked (lock_page)
 *       mapping->i_mmap_lock
 *         anon_vma->lock
 *           mm->page_table_lock or pte_lock
 *             zone->lru_lock (in mark_page_accessed, isolate_lru_page)
 *             swap_lock (in swap_duplicate, swap_info_get)
 *               mmlist_lock (in mmput, drain_mmlist and others)
 *               mapping->private_lock (in __set_page_dirty_buffers)
 *               inode_lock (in set_page_dirty's __mark_inode_dirty)
 *                 sb_lock (within inode_lock in fs/fs-writeback.c)
 *                 mapping->tree_lock (widely used, in set_page_dirty,
 *                           in arch-dependent flush_dcache_mmap_lock,
 *                           within inode_lock in __sync_single_inode)
39 40 41 42 43
 *
 * (code doesn't rely on that order so it could be switched around)
 * ->tasklist_lock
 *   anon_vma->lock      (memory_failure, collect_procs_anon)
 *     pte map lock
L
Linus Torvalds 已提交
44 45 46 47 48 49 50 51
 */

#include <linux/mm.h>
#include <linux/pagemap.h>
#include <linux/swap.h>
#include <linux/swapops.h>
#include <linux/slab.h>
#include <linux/init.h>
H
Hugh Dickins 已提交
52
#include <linux/ksm.h>
L
Linus Torvalds 已提交
53 54
#include <linux/rmap.h>
#include <linux/rcupdate.h>
55
#include <linux/module.h>
56
#include <linux/memcontrol.h>
A
Andrea Arcangeli 已提交
57
#include <linux/mmu_notifier.h>
58
#include <linux/migrate.h>
L
Linus Torvalds 已提交
59 60 61

#include <asm/tlbflush.h>

N
Nick Piggin 已提交
62 63
#include "internal.h"

64
static struct kmem_cache *anon_vma_cachep;
65
static struct kmem_cache *anon_vma_chain_cachep;
66 67 68 69 70 71

static inline struct anon_vma *anon_vma_alloc(void)
{
	return kmem_cache_alloc(anon_vma_cachep, GFP_KERNEL);
}

H
Hugh Dickins 已提交
72
void anon_vma_free(struct anon_vma *anon_vma)
73 74 75
{
	kmem_cache_free(anon_vma_cachep, anon_vma);
}
L
Linus Torvalds 已提交
76

77 78 79 80 81 82 83 84 85 86
static inline struct anon_vma_chain *anon_vma_chain_alloc(void)
{
	return kmem_cache_alloc(anon_vma_chain_cachep, GFP_KERNEL);
}

void anon_vma_chain_free(struct anon_vma_chain *anon_vma_chain)
{
	kmem_cache_free(anon_vma_chain_cachep, anon_vma_chain);
}

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
/**
 * anon_vma_prepare - attach an anon_vma to a memory region
 * @vma: the memory region in question
 *
 * This makes sure the memory mapping described by 'vma' has
 * an 'anon_vma' attached to it, so that we can associate the
 * anonymous pages mapped into it with that anon_vma.
 *
 * The common case will be that we already have one, but if
 * if not we either need to find an adjacent mapping that we
 * can re-use the anon_vma from (very common when the only
 * reason for splitting a vma has been mprotect()), or we
 * allocate a new one.
 *
 * Anon-vma allocations are very subtle, because we may have
 * optimistically looked up an anon_vma in page_lock_anon_vma()
 * and that may actually touch the spinlock even in the newly
 * allocated vma (it depends on RCU to make sure that the
 * anon_vma isn't actually destroyed).
 *
 * As a result, we need to do proper anon_vma locking even
 * for the new allocation. At the same time, we do not want
 * to do any locking for the common case of already having
 * an anon_vma.
 *
 * This must be called with the mmap_sem held for reading.
 */
L
Linus Torvalds 已提交
114 115 116
int anon_vma_prepare(struct vm_area_struct *vma)
{
	struct anon_vma *anon_vma = vma->anon_vma;
117
	struct anon_vma_chain *avc;
L
Linus Torvalds 已提交
118 119 120 121

	might_sleep();
	if (unlikely(!anon_vma)) {
		struct mm_struct *mm = vma->vm_mm;
122
		struct anon_vma *allocated;
L
Linus Torvalds 已提交
123

124 125 126 127
		avc = anon_vma_chain_alloc();
		if (!avc)
			goto out_enomem;

L
Linus Torvalds 已提交
128
		anon_vma = find_mergeable_anon_vma(vma);
129 130
		allocated = NULL;
		if (!anon_vma) {
L
Linus Torvalds 已提交
131 132
			anon_vma = anon_vma_alloc();
			if (unlikely(!anon_vma))
133
				goto out_enomem_free_avc;
L
Linus Torvalds 已提交
134 135 136
			allocated = anon_vma;
		}

137
		spin_lock(&anon_vma->lock);
L
Linus Torvalds 已提交
138 139 140 141
		/* page_table_lock to protect against threads */
		spin_lock(&mm->page_table_lock);
		if (likely(!vma->anon_vma)) {
			vma->anon_vma = anon_vma;
142 143 144 145
			avc->anon_vma = anon_vma;
			avc->vma = vma;
			list_add(&avc->same_vma, &vma->anon_vma_chain);
			list_add(&avc->same_anon_vma, &anon_vma->head);
L
Linus Torvalds 已提交
146
			allocated = NULL;
147
			avc = NULL;
L
Linus Torvalds 已提交
148 149
		}
		spin_unlock(&mm->page_table_lock);
150
		spin_unlock(&anon_vma->lock);
151 152

		if (unlikely(allocated))
L
Linus Torvalds 已提交
153
			anon_vma_free(allocated);
154
		if (unlikely(avc))
155
			anon_vma_chain_free(avc);
L
Linus Torvalds 已提交
156 157
	}
	return 0;
158 159 160 161 162

 out_enomem_free_avc:
	anon_vma_chain_free(avc);
 out_enomem:
	return -ENOMEM;
L
Linus Torvalds 已提交
163 164
}

165 166 167
static void anon_vma_chain_link(struct vm_area_struct *vma,
				struct anon_vma_chain *avc,
				struct anon_vma *anon_vma)
L
Linus Torvalds 已提交
168
{
169 170 171 172 173 174 175
	avc->vma = vma;
	avc->anon_vma = anon_vma;
	list_add(&avc->same_vma, &vma->anon_vma_chain);

	spin_lock(&anon_vma->lock);
	list_add_tail(&avc->same_anon_vma, &anon_vma->head);
	spin_unlock(&anon_vma->lock);
L
Linus Torvalds 已提交
176 177
}

178 179 180 181 182
/*
 * Attach the anon_vmas from src to dst.
 * Returns 0 on success, -ENOMEM on failure.
 */
int anon_vma_clone(struct vm_area_struct *dst, struct vm_area_struct *src)
L
Linus Torvalds 已提交
183
{
184 185
	struct anon_vma_chain *avc, *pavc;

186
	list_for_each_entry_reverse(pavc, &src->anon_vma_chain, same_vma) {
187 188 189 190 191 192
		avc = anon_vma_chain_alloc();
		if (!avc)
			goto enomem_failure;
		anon_vma_chain_link(dst, avc, pavc->anon_vma);
	}
	return 0;
L
Linus Torvalds 已提交
193

194 195 196
 enomem_failure:
	unlink_anon_vmas(dst);
	return -ENOMEM;
L
Linus Torvalds 已提交
197 198
}

199 200 201 202 203 204
/*
 * Attach vma to its own anon_vma, as well as to the anon_vmas that
 * the corresponding VMA in the parent process is attached to.
 * Returns 0 on success, non-zero on failure.
 */
int anon_vma_fork(struct vm_area_struct *vma, struct vm_area_struct *pvma)
L
Linus Torvalds 已提交
205
{
206 207
	struct anon_vma_chain *avc;
	struct anon_vma *anon_vma;
L
Linus Torvalds 已提交
208

209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235
	/* Don't bother if the parent process has no anon_vma here. */
	if (!pvma->anon_vma)
		return 0;

	/*
	 * First, attach the new VMA to the parent VMA's anon_vmas,
	 * so rmap can find non-COWed pages in child processes.
	 */
	if (anon_vma_clone(vma, pvma))
		return -ENOMEM;

	/* Then add our own anon_vma. */
	anon_vma = anon_vma_alloc();
	if (!anon_vma)
		goto out_error;
	avc = anon_vma_chain_alloc();
	if (!avc)
		goto out_error_free_anon_vma;
	anon_vma_chain_link(vma, avc, anon_vma);
	/* Mark this anon_vma as the one where our new (COWed) pages go. */
	vma->anon_vma = anon_vma;

	return 0;

 out_error_free_anon_vma:
	anon_vma_free(anon_vma);
 out_error:
236
	unlink_anon_vmas(vma);
237
	return -ENOMEM;
L
Linus Torvalds 已提交
238 239
}

240
static void anon_vma_unlink(struct anon_vma_chain *anon_vma_chain)
L
Linus Torvalds 已提交
241
{
242
	struct anon_vma *anon_vma = anon_vma_chain->anon_vma;
L
Linus Torvalds 已提交
243 244
	int empty;

245
	/* If anon_vma_fork fails, we can get an empty anon_vma_chain. */
L
Linus Torvalds 已提交
246 247 248 249
	if (!anon_vma)
		return;

	spin_lock(&anon_vma->lock);
250
	list_del(&anon_vma_chain->same_anon_vma);
L
Linus Torvalds 已提交
251 252

	/* We must garbage collect the anon_vma if it's empty */
253
	empty = list_empty(&anon_vma->head) && !anonvma_external_refcount(anon_vma);
L
Linus Torvalds 已提交
254 255 256 257 258 259
	spin_unlock(&anon_vma->lock);

	if (empty)
		anon_vma_free(anon_vma);
}

260 261 262 263 264 265 266 267 268 269 270 271
void unlink_anon_vmas(struct vm_area_struct *vma)
{
	struct anon_vma_chain *avc, *next;

	/* Unlink each anon_vma chained to the VMA. */
	list_for_each_entry_safe(avc, next, &vma->anon_vma_chain, same_vma) {
		anon_vma_unlink(avc);
		list_del(&avc->same_vma);
		anon_vma_chain_free(avc);
	}
}

272
static void anon_vma_ctor(void *data)
L
Linus Torvalds 已提交
273
{
C
Christoph Lameter 已提交
274
	struct anon_vma *anon_vma = data;
L
Linus Torvalds 已提交
275

C
Christoph Lameter 已提交
276
	spin_lock_init(&anon_vma->lock);
277
	anonvma_external_refcount_init(anon_vma);
C
Christoph Lameter 已提交
278
	INIT_LIST_HEAD(&anon_vma->head);
L
Linus Torvalds 已提交
279 280 281 282 283
}

void __init anon_vma_init(void)
{
	anon_vma_cachep = kmem_cache_create("anon_vma", sizeof(struct anon_vma),
284
			0, SLAB_DESTROY_BY_RCU|SLAB_PANIC, anon_vma_ctor);
285
	anon_vma_chain_cachep = KMEM_CACHE(anon_vma_chain, SLAB_PANIC);
L
Linus Torvalds 已提交
286 287 288 289 290 291
}

/*
 * Getting a lock on a stable anon_vma from a page off the LRU is
 * tricky: page_lock_anon_vma rely on RCU to guard against the races.
 */
292
struct anon_vma *page_lock_anon_vma(struct page *page)
L
Linus Torvalds 已提交
293
{
294
	struct anon_vma *anon_vma;
L
Linus Torvalds 已提交
295 296 297
	unsigned long anon_mapping;

	rcu_read_lock();
298
	anon_mapping = (unsigned long) ACCESS_ONCE(page->mapping);
H
Hugh Dickins 已提交
299
	if ((anon_mapping & PAGE_MAPPING_FLAGS) != PAGE_MAPPING_ANON)
L
Linus Torvalds 已提交
300 301 302 303 304 305
		goto out;
	if (!page_mapped(page))
		goto out;

	anon_vma = (struct anon_vma *) (anon_mapping - PAGE_MAPPING_ANON);
	spin_lock(&anon_vma->lock);
306
	return anon_vma;
L
Linus Torvalds 已提交
307 308
out:
	rcu_read_unlock();
309 310 311
	return NULL;
}

312
void page_unlock_anon_vma(struct anon_vma *anon_vma)
313 314 315
{
	spin_unlock(&anon_vma->lock);
	rcu_read_unlock();
L
Linus Torvalds 已提交
316 317 318
}

/*
319 320 321
 * At what user virtual address is page expected in @vma?
 * Returns virtual address or -EFAULT if page's index/offset is not
 * within the range mapped the @vma.
L
Linus Torvalds 已提交
322 323 324 325 326 327 328 329 330
 */
static inline unsigned long
vma_address(struct page *page, struct vm_area_struct *vma)
{
	pgoff_t pgoff = page->index << (PAGE_CACHE_SHIFT - PAGE_SHIFT);
	unsigned long address;

	address = vma->vm_start + ((pgoff - vma->vm_pgoff) << PAGE_SHIFT);
	if (unlikely(address < vma->vm_start || address >= vma->vm_end)) {
331
		/* page should be within @vma mapping range */
L
Linus Torvalds 已提交
332 333 334 335 336 337
		return -EFAULT;
	}
	return address;
}

/*
H
Huang Shijie 已提交
338
 * At what user virtual address is page expected in vma?
339
 * Caller should check the page is actually part of the vma.
L
Linus Torvalds 已提交
340 341 342
 */
unsigned long page_address_in_vma(struct page *page, struct vm_area_struct *vma)
{
343 344 345
	if (PageAnon(page))
		;
	else if (page->mapping && !(vma->vm_flags & VM_NONLINEAR)) {
H
Hugh Dickins 已提交
346 347
		if (!vma->vm_file ||
		    vma->vm_file->f_mapping != page->mapping)
L
Linus Torvalds 已提交
348 349 350 351 352 353
			return -EFAULT;
	} else
		return -EFAULT;
	return vma_address(page, vma);
}

N
Nikita Danilov 已提交
354 355 356
/*
 * Check that @page is mapped at @address into @mm.
 *
N
Nick Piggin 已提交
357 358 359 360
 * If @sync is false, page_check_address may perform a racy check to avoid
 * the page table lock when the pte is not present (helpful when reclaiming
 * highly shared pages).
 *
361
 * On success returns with pte mapped and locked.
N
Nikita Danilov 已提交
362
 */
363
pte_t *page_check_address(struct page *page, struct mm_struct *mm,
N
Nick Piggin 已提交
364
			  unsigned long address, spinlock_t **ptlp, int sync)
N
Nikita Danilov 已提交
365 366 367 368 369
{
	pgd_t *pgd;
	pud_t *pud;
	pmd_t *pmd;
	pte_t *pte;
H
Hugh Dickins 已提交
370
	spinlock_t *ptl;
N
Nikita Danilov 已提交
371 372

	pgd = pgd_offset(mm, address);
H
Hugh Dickins 已提交
373 374 375 376 377 378 379 380 381 382 383 384 385
	if (!pgd_present(*pgd))
		return NULL;

	pud = pud_offset(pgd, address);
	if (!pud_present(*pud))
		return NULL;

	pmd = pmd_offset(pud, address);
	if (!pmd_present(*pmd))
		return NULL;

	pte = pte_offset_map(pmd, address);
	/* Make a quick check before getting the lock */
N
Nick Piggin 已提交
386
	if (!sync && !pte_present(*pte)) {
H
Hugh Dickins 已提交
387 388 389 390
		pte_unmap(pte);
		return NULL;
	}

H
Hugh Dickins 已提交
391
	ptl = pte_lockptr(mm, pmd);
H
Hugh Dickins 已提交
392 393 394 395
	spin_lock(ptl);
	if (pte_present(*pte) && page_to_pfn(page) == pte_pfn(*pte)) {
		*ptlp = ptl;
		return pte;
N
Nikita Danilov 已提交
396
	}
H
Hugh Dickins 已提交
397 398
	pte_unmap_unlock(pte, ptl);
	return NULL;
N
Nikita Danilov 已提交
399 400
}

N
Nick Piggin 已提交
401 402 403 404 405 406 407 408 409
/**
 * page_mapped_in_vma - check whether a page is really mapped in a VMA
 * @page: the page to test
 * @vma: the VMA to test
 *
 * Returns 1 if the page is mapped into the page tables of the VMA, 0
 * if the page is not mapped into the page tables of this VMA.  Only
 * valid for normal file or anonymous VMAs.
 */
410
int page_mapped_in_vma(struct page *page, struct vm_area_struct *vma)
N
Nick Piggin 已提交
411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426
{
	unsigned long address;
	pte_t *pte;
	spinlock_t *ptl;

	address = vma_address(page, vma);
	if (address == -EFAULT)		/* out of vma range */
		return 0;
	pte = page_check_address(page, vma->vm_mm, address, &ptl, 1);
	if (!pte)			/* the page is not in this mm */
		return 0;
	pte_unmap_unlock(pte, ptl);

	return 1;
}

L
Linus Torvalds 已提交
427 428 429 430
/*
 * Subfunctions of page_referenced: page_referenced_one called
 * repeatedly from either page_referenced_anon or page_referenced_file.
 */
H
Hugh Dickins 已提交
431 432 433
int page_referenced_one(struct page *page, struct vm_area_struct *vma,
			unsigned long address, unsigned int *mapcount,
			unsigned long *vm_flags)
L
Linus Torvalds 已提交
434 435 436
{
	struct mm_struct *mm = vma->vm_mm;
	pte_t *pte;
H
Hugh Dickins 已提交
437
	spinlock_t *ptl;
L
Linus Torvalds 已提交
438 439
	int referenced = 0;

N
Nick Piggin 已提交
440
	pte = page_check_address(page, mm, address, &ptl, 0);
H
Hugh Dickins 已提交
441 442
	if (!pte)
		goto out;
L
Linus Torvalds 已提交
443

N
Nick Piggin 已提交
444 445 446 447 448
	/*
	 * Don't want to elevate referenced for mlocked page that gets this far,
	 * in order that it progresses to try_to_unmap and is moved to the
	 * unevictable list.
	 */
449 450
	if (vma->vm_flags & VM_LOCKED) {
		*mapcount = 1;	/* break early from loop */
451
		*vm_flags |= VM_LOCKED;
N
Nick Piggin 已提交
452 453 454
		goto out_unmap;
	}

455 456 457 458 459 460 461 462 463 464 465
	if (ptep_clear_flush_young_notify(vma, address, pte)) {
		/*
		 * Don't treat a reference through a sequentially read
		 * mapping as such.  If the page has been used in
		 * another mapping, we will catch it; if this other
		 * mapping is already gone, the unmap path will have
		 * set PG_referenced or activated the page.
		 */
		if (likely(!VM_SequentialReadHint(vma)))
			referenced++;
	}
L
Linus Torvalds 已提交
466

H
Hugh Dickins 已提交
467 468
	/* Pretend the page is referenced if the task has the
	   swap token and is in the middle of a page fault. */
469
	if (mm != current->mm && has_swap_token(mm) &&
H
Hugh Dickins 已提交
470 471 472
			rwsem_is_locked(&mm->mmap_sem))
		referenced++;

N
Nick Piggin 已提交
473
out_unmap:
H
Hugh Dickins 已提交
474 475
	(*mapcount)--;
	pte_unmap_unlock(pte, ptl);
476

477 478
	if (referenced)
		*vm_flags |= vma->vm_flags;
479
out:
L
Linus Torvalds 已提交
480 481 482
	return referenced;
}

483
static int page_referenced_anon(struct page *page,
484 485
				struct mem_cgroup *mem_cont,
				unsigned long *vm_flags)
L
Linus Torvalds 已提交
486 487 488
{
	unsigned int mapcount;
	struct anon_vma *anon_vma;
489
	struct anon_vma_chain *avc;
L
Linus Torvalds 已提交
490 491 492 493 494 495 496
	int referenced = 0;

	anon_vma = page_lock_anon_vma(page);
	if (!anon_vma)
		return referenced;

	mapcount = page_mapcount(page);
497 498
	list_for_each_entry(avc, &anon_vma->head, same_anon_vma) {
		struct vm_area_struct *vma = avc->vma;
H
Hugh Dickins 已提交
499 500 501
		unsigned long address = vma_address(page, vma);
		if (address == -EFAULT)
			continue;
502 503 504 505 506
		/*
		 * If we are reclaiming on behalf of a cgroup, skip
		 * counting on behalf of references from different
		 * cgroups
		 */
507
		if (mem_cont && !mm_match_cgroup(vma->vm_mm, mem_cont))
508
			continue;
H
Hugh Dickins 已提交
509
		referenced += page_referenced_one(page, vma, address,
510
						  &mapcount, vm_flags);
L
Linus Torvalds 已提交
511 512 513
		if (!mapcount)
			break;
	}
514 515

	page_unlock_anon_vma(anon_vma);
L
Linus Torvalds 已提交
516 517 518 519 520 521
	return referenced;
}

/**
 * page_referenced_file - referenced check for object-based rmap
 * @page: the page we're checking references on.
R
Randy Dunlap 已提交
522
 * @mem_cont: target memory controller
523
 * @vm_flags: collect encountered vma->vm_flags who actually referenced the page
L
Linus Torvalds 已提交
524 525 526 527 528 529 530 531
 *
 * For an object-based mapped page, find all the places it is mapped and
 * check/clear the referenced flag.  This is done by following the page->mapping
 * pointer, then walking the chain of vmas it holds.  It returns the number
 * of references it found.
 *
 * This function is only called from page_referenced for object-based pages.
 */
532
static int page_referenced_file(struct page *page,
533 534
				struct mem_cgroup *mem_cont,
				unsigned long *vm_flags)
L
Linus Torvalds 已提交
535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566
{
	unsigned int mapcount;
	struct address_space *mapping = page->mapping;
	pgoff_t pgoff = page->index << (PAGE_CACHE_SHIFT - PAGE_SHIFT);
	struct vm_area_struct *vma;
	struct prio_tree_iter iter;
	int referenced = 0;

	/*
	 * The caller's checks on page->mapping and !PageAnon have made
	 * sure that this is a file page: the check for page->mapping
	 * excludes the case just before it gets set on an anon page.
	 */
	BUG_ON(PageAnon(page));

	/*
	 * The page lock not only makes sure that page->mapping cannot
	 * suddenly be NULLified by truncation, it makes sure that the
	 * structure at mapping cannot be freed and reused yet,
	 * so we can safely take mapping->i_mmap_lock.
	 */
	BUG_ON(!PageLocked(page));

	spin_lock(&mapping->i_mmap_lock);

	/*
	 * i_mmap_lock does not stabilize mapcount at all, but mapcount
	 * is more likely to be accurate if we note it after spinning.
	 */
	mapcount = page_mapcount(page);

	vma_prio_tree_foreach(vma, &iter, &mapping->i_mmap, pgoff, pgoff) {
H
Hugh Dickins 已提交
567 568 569
		unsigned long address = vma_address(page, vma);
		if (address == -EFAULT)
			continue;
570 571 572 573 574
		/*
		 * If we are reclaiming on behalf of a cgroup, skip
		 * counting on behalf of references from different
		 * cgroups
		 */
575
		if (mem_cont && !mm_match_cgroup(vma->vm_mm, mem_cont))
576
			continue;
H
Hugh Dickins 已提交
577
		referenced += page_referenced_one(page, vma, address,
578
						  &mapcount, vm_flags);
L
Linus Torvalds 已提交
579 580 581 582 583 584 585 586 587 588 589 590
		if (!mapcount)
			break;
	}

	spin_unlock(&mapping->i_mmap_lock);
	return referenced;
}

/**
 * page_referenced - test if the page was referenced
 * @page: the page to test
 * @is_locked: caller holds lock on the page
R
Randy Dunlap 已提交
591
 * @mem_cont: target memory controller
592
 * @vm_flags: collect encountered vma->vm_flags who actually referenced the page
L
Linus Torvalds 已提交
593 594 595 596
 *
 * Quick test_and_clear_referenced for all mappings to a page,
 * returns the number of ptes which referenced the page.
 */
597 598 599 600
int page_referenced(struct page *page,
		    int is_locked,
		    struct mem_cgroup *mem_cont,
		    unsigned long *vm_flags)
L
Linus Torvalds 已提交
601 602
{
	int referenced = 0;
H
Hugh Dickins 已提交
603
	int we_locked = 0;
L
Linus Torvalds 已提交
604

605
	*vm_flags = 0;
H
Hugh Dickins 已提交
606
	if (page_mapped(page) && page_rmapping(page)) {
H
Hugh Dickins 已提交
607 608 609 610 611 612 613 614 615 616 617
		if (!is_locked && (!PageAnon(page) || PageKsm(page))) {
			we_locked = trylock_page(page);
			if (!we_locked) {
				referenced++;
				goto out;
			}
		}
		if (unlikely(PageKsm(page)))
			referenced += page_referenced_ksm(page, mem_cont,
								vm_flags);
		else if (PageAnon(page))
618 619
			referenced += page_referenced_anon(page, mem_cont,
								vm_flags);
H
Hugh Dickins 已提交
620
		else if (page->mapping)
621 622
			referenced += page_referenced_file(page, mem_cont,
								vm_flags);
H
Hugh Dickins 已提交
623
		if (we_locked)
L
Linus Torvalds 已提交
624 625
			unlock_page(page);
	}
H
Hugh Dickins 已提交
626
out:
627 628 629
	if (page_test_and_clear_young(page))
		referenced++;

L
Linus Torvalds 已提交
630 631 632
	return referenced;
}

H
Hugh Dickins 已提交
633 634
static int page_mkclean_one(struct page *page, struct vm_area_struct *vma,
			    unsigned long address)
635 636
{
	struct mm_struct *mm = vma->vm_mm;
637
	pte_t *pte;
638 639 640
	spinlock_t *ptl;
	int ret = 0;

N
Nick Piggin 已提交
641
	pte = page_check_address(page, mm, address, &ptl, 1);
642 643 644
	if (!pte)
		goto out;

645 646
	if (pte_dirty(*pte) || pte_write(*pte)) {
		pte_t entry;
647

648
		flush_cache_page(vma, address, pte_pfn(*pte));
A
Andrea Arcangeli 已提交
649
		entry = ptep_clear_flush_notify(vma, address, pte);
650 651
		entry = pte_wrprotect(entry);
		entry = pte_mkclean(entry);
652
		set_pte_at(mm, address, pte, entry);
653 654
		ret = 1;
	}
655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671

	pte_unmap_unlock(pte, ptl);
out:
	return ret;
}

static int page_mkclean_file(struct address_space *mapping, struct page *page)
{
	pgoff_t pgoff = page->index << (PAGE_CACHE_SHIFT - PAGE_SHIFT);
	struct vm_area_struct *vma;
	struct prio_tree_iter iter;
	int ret = 0;

	BUG_ON(PageAnon(page));

	spin_lock(&mapping->i_mmap_lock);
	vma_prio_tree_foreach(vma, &iter, &mapping->i_mmap, pgoff, pgoff) {
H
Hugh Dickins 已提交
672 673 674 675 676 677
		if (vma->vm_flags & VM_SHARED) {
			unsigned long address = vma_address(page, vma);
			if (address == -EFAULT)
				continue;
			ret += page_mkclean_one(page, vma, address);
		}
678 679 680 681 682 683 684 685 686 687 688 689 690
	}
	spin_unlock(&mapping->i_mmap_lock);
	return ret;
}

int page_mkclean(struct page *page)
{
	int ret = 0;

	BUG_ON(!PageLocked(page));

	if (page_mapped(page)) {
		struct address_space *mapping = page_mapping(page);
691
		if (mapping) {
692
			ret = page_mkclean_file(mapping, page);
693 694 695 696
			if (page_test_dirty(page)) {
				page_clear_dirty(page);
				ret = 1;
			}
697
		}
698 699 700 701
	}

	return ret;
}
J
Jaya Kumar 已提交
702
EXPORT_SYMBOL_GPL(page_mkclean);
703

704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727
/**
 * page_move_anon_rmap - move a page to our anon_vma
 * @page:	the page to move to our anon_vma
 * @vma:	the vma the page belongs to
 * @address:	the user virtual address mapped
 *
 * When a page belongs exclusively to one process after a COW event,
 * that page can be moved into the anon_vma that belongs to just that
 * process, so the rmap code will not search the parent or sibling
 * processes.
 */
void page_move_anon_rmap(struct page *page,
	struct vm_area_struct *vma, unsigned long address)
{
	struct anon_vma *anon_vma = vma->anon_vma;

	VM_BUG_ON(!PageLocked(page));
	VM_BUG_ON(!anon_vma);
	VM_BUG_ON(page->index != linear_page_index(vma, address));

	anon_vma = (void *) anon_vma + PAGE_MAPPING_ANON;
	page->mapping = (struct address_space *) anon_vma;
}

N
Nick Piggin 已提交
728
/**
R
Randy Dunlap 已提交
729
 * __page_set_anon_rmap - setup new anonymous rmap
N
Nick Piggin 已提交
730 731 732
 * @page:	the page to add the mapping to
 * @vma:	the vm area in which the mapping is added
 * @address:	the user virtual address mapped
733
 * @exclusive:	the page is exclusively owned by the current process
N
Nick Piggin 已提交
734 735
 */
static void __page_set_anon_rmap(struct page *page,
736
	struct vm_area_struct *vma, unsigned long address, int exclusive)
N
Nick Piggin 已提交
737
{
738
	struct anon_vma *anon_vma = vma->anon_vma;
739

740
	BUG_ON(!anon_vma);
741 742

	/*
743 744 745
	 * If the page isn't exclusively mapped into this vma,
	 * we must use the _oldest_ possible anon_vma for the
	 * page mapping!
746
	 *
747 748
	 * So take the last AVC chain entry in the vma, which is
	 * the deepest ancestor, and use the anon_vma from that.
749
	 */
750 751 752 753 754
	if (!exclusive) {
		struct anon_vma_chain *avc;
		avc = list_entry(vma->anon_vma_chain.prev, struct anon_vma_chain, same_vma);
		anon_vma = avc->anon_vma;
	}
N
Nick Piggin 已提交
755 756 757 758 759 760

	anon_vma = (void *) anon_vma + PAGE_MAPPING_ANON;
	page->mapping = (struct address_space *) anon_vma;
	page->index = linear_page_index(vma, address);
}

N
Nick Piggin 已提交
761
/**
R
Randy Dunlap 已提交
762
 * __page_check_anon_rmap - sanity check anonymous rmap addition
N
Nick Piggin 已提交
763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786
 * @page:	the page to add the mapping to
 * @vma:	the vm area in which the mapping is added
 * @address:	the user virtual address mapped
 */
static void __page_check_anon_rmap(struct page *page,
	struct vm_area_struct *vma, unsigned long address)
{
#ifdef CONFIG_DEBUG_VM
	/*
	 * The page's anon-rmap details (mapping and index) are guaranteed to
	 * be set up correctly at this point.
	 *
	 * We have exclusion against page_add_anon_rmap because the caller
	 * always holds the page locked, except if called from page_dup_rmap,
	 * in which case the page is already known to be setup.
	 *
	 * We have exclusion against page_add_new_anon_rmap because those pages
	 * are initially only visible via the pagetables, and the pte is locked
	 * over the call to page_add_new_anon_rmap.
	 */
	BUG_ON(page->index != linear_page_index(vma, address));
#endif
}

L
Linus Torvalds 已提交
787 788 789 790 791 792
/**
 * page_add_anon_rmap - add pte mapping to an anonymous page
 * @page:	the page to add the mapping to
 * @vma:	the vm area in which the mapping is added
 * @address:	the user virtual address mapped
 *
H
Hugh Dickins 已提交
793
 * The caller needs to hold the pte lock, and the page must be locked in
794 795 796
 * the anon_vma case: to serialize mapping,index checking after setting,
 * and to ensure that PageAnon is not being upgraded racily to PageKsm
 * (but PageKsm is never downgraded to PageAnon).
L
Linus Torvalds 已提交
797 798 799 800
 */
void page_add_anon_rmap(struct page *page,
	struct vm_area_struct *vma, unsigned long address)
{
H
Hugh Dickins 已提交
801 802 803 804 805 806
	int first = atomic_inc_and_test(&page->_mapcount);
	if (first)
		__inc_zone_page_state(page, NR_ANON_PAGES);
	if (unlikely(PageKsm(page)))
		return;

N
Nick Piggin 已提交
807 808
	VM_BUG_ON(!PageLocked(page));
	VM_BUG_ON(address < vma->vm_start || address >= vma->vm_end);
H
Hugh Dickins 已提交
809
	if (first)
810
		__page_set_anon_rmap(page, vma, address, 0);
811
	else
N
Nick Piggin 已提交
812
		__page_check_anon_rmap(page, vma, address);
L
Linus Torvalds 已提交
813 814
}

R
Randy Dunlap 已提交
815
/**
N
Nick Piggin 已提交
816 817 818 819 820 821 822
 * page_add_new_anon_rmap - add pte mapping to a new anonymous page
 * @page:	the page to add the mapping to
 * @vma:	the vm area in which the mapping is added
 * @address:	the user virtual address mapped
 *
 * Same as page_add_anon_rmap but must only be called on *new* pages.
 * This means the inc-and-test can be bypassed.
N
Nick Piggin 已提交
823
 * Page does not have to be locked.
N
Nick Piggin 已提交
824 825 826 827
 */
void page_add_new_anon_rmap(struct page *page,
	struct vm_area_struct *vma, unsigned long address)
{
828
	VM_BUG_ON(address < vma->vm_start || address >= vma->vm_end);
829 830
	SetPageSwapBacked(page);
	atomic_set(&page->_mapcount, 0); /* increment count (starts at -1) */
H
Hugh Dickins 已提交
831
	__inc_zone_page_state(page, NR_ANON_PAGES);
832
	__page_set_anon_rmap(page, vma, address, 1);
833
	if (page_evictable(page, vma))
834
		lru_cache_add_lru(page, LRU_ACTIVE_ANON);
835 836
	else
		add_page_to_unevictable_list(page);
N
Nick Piggin 已提交
837 838
}

L
Linus Torvalds 已提交
839 840 841 842
/**
 * page_add_file_rmap - add pte mapping to a file page
 * @page: the page to add the mapping to
 *
843
 * The caller needs to hold the pte lock.
L
Linus Torvalds 已提交
844 845 846
 */
void page_add_file_rmap(struct page *page)
{
847
	if (atomic_inc_and_test(&page->_mapcount)) {
848
		__inc_zone_page_state(page, NR_FILE_MAPPED);
849
		mem_cgroup_update_file_mapped(page, 1);
850
	}
L
Linus Torvalds 已提交
851 852 853 854 855 856
}

/**
 * page_remove_rmap - take down pte mapping from a page
 * @page: page to remove mapping from
 *
857
 * The caller needs to hold the pte lock.
L
Linus Torvalds 已提交
858
 */
859
void page_remove_rmap(struct page *page)
L
Linus Torvalds 已提交
860
{
861 862 863 864 865 866 867 868 869 870 871 872 873 874
	/* page still mapped by someone else? */
	if (!atomic_add_negative(-1, &page->_mapcount))
		return;

	/*
	 * Now that the last pte has gone, s390 must transfer dirty
	 * flag from storage key to struct page.  We can usually skip
	 * this if the page is anon, so about to be freed; but perhaps
	 * not if it's in swapcache - there might be another pte slot
	 * containing the swap entry, but page not yet written to swap.
	 */
	if ((!PageAnon(page) || PageSwapCache(page)) && page_test_dirty(page)) {
		page_clear_dirty(page);
		set_page_dirty(page);
L
Linus Torvalds 已提交
875
	}
876 877 878 879 880
	if (PageAnon(page)) {
		mem_cgroup_uncharge_page(page);
		__dec_zone_page_state(page, NR_ANON_PAGES);
	} else {
		__dec_zone_page_state(page, NR_FILE_MAPPED);
881
		mem_cgroup_update_file_mapped(page, -1);
882 883 884 885 886 887 888 889 890 891
	}
	/*
	 * It would be tidy to reset the PageAnon mapping here,
	 * but that might overwrite a racing page_add_anon_rmap
	 * which increments mapcount after us but sets mapping
	 * before us: so leave the reset to free_hot_cold_page,
	 * and remember that it's only reliable while mapped.
	 * Leaving it set also helps swapoff to reinstate ptes
	 * faster for those pages still in swapcache.
	 */
L
Linus Torvalds 已提交
892 893 894 895 896 897
}

/*
 * Subfunctions of try_to_unmap: try_to_unmap_one called
 * repeatedly from either try_to_unmap_anon or try_to_unmap_file.
 */
H
Hugh Dickins 已提交
898 899
int try_to_unmap_one(struct page *page, struct vm_area_struct *vma,
		     unsigned long address, enum ttu_flags flags)
L
Linus Torvalds 已提交
900 901 902 903
{
	struct mm_struct *mm = vma->vm_mm;
	pte_t *pte;
	pte_t pteval;
H
Hugh Dickins 已提交
904
	spinlock_t *ptl;
L
Linus Torvalds 已提交
905 906
	int ret = SWAP_AGAIN;

N
Nick Piggin 已提交
907
	pte = page_check_address(page, mm, address, &ptl, 0);
H
Hugh Dickins 已提交
908
	if (!pte)
N
Nikita Danilov 已提交
909
		goto out;
L
Linus Torvalds 已提交
910 911 912 913 914 915

	/*
	 * If the page is mlock()d, we cannot swap it out.
	 * If it's recently referenced (perhaps page_referenced
	 * skipped over this mm) then we should reactivate it.
	 */
916
	if (!(flags & TTU_IGNORE_MLOCK)) {
K
KOSAKI Motohiro 已提交
917 918 919
		if (vma->vm_flags & VM_LOCKED)
			goto out_mlock;

H
Hugh Dickins 已提交
920
		if (TTU_ACTION(flags) == TTU_MUNLOCK)
H
Hugh Dickins 已提交
921
			goto out_unmap;
922 923
	}
	if (!(flags & TTU_IGNORE_ACCESS)) {
N
Nick Piggin 已提交
924 925 926 927 928
		if (ptep_clear_flush_young_notify(vma, address, pte)) {
			ret = SWAP_FAIL;
			goto out_unmap;
		}
  	}
L
Linus Torvalds 已提交
929 930 931

	/* Nuke the page table entry. */
	flush_cache_page(vma, address, page_to_pfn(page));
A
Andrea Arcangeli 已提交
932
	pteval = ptep_clear_flush_notify(vma, address, pte);
L
Linus Torvalds 已提交
933 934 935 936 937

	/* Move the dirty bit to the physical page now the pte is gone. */
	if (pte_dirty(pteval))
		set_page_dirty(page);

938 939 940
	/* Update high watermark before we lower rss */
	update_hiwater_rss(mm);

941 942
	if (PageHWPoison(page) && !(flags & TTU_IGNORE_HWPOISON)) {
		if (PageAnon(page))
K
KAMEZAWA Hiroyuki 已提交
943
			dec_mm_counter(mm, MM_ANONPAGES);
944
		else
K
KAMEZAWA Hiroyuki 已提交
945
			dec_mm_counter(mm, MM_FILEPAGES);
946 947 948
		set_pte_at(mm, address, pte,
				swp_entry_to_pte(make_hwpoison_entry(page)));
	} else if (PageAnon(page)) {
H
Hugh Dickins 已提交
949
		swp_entry_t entry = { .val = page_private(page) };
950 951 952 953 954 955

		if (PageSwapCache(page)) {
			/*
			 * Store the swap location in the pte.
			 * See handle_pte_fault() ...
			 */
H
Hugh Dickins 已提交
956 957 958 959 960
			if (swap_duplicate(entry) < 0) {
				set_pte_at(mm, address, pte, pteval);
				ret = SWAP_FAIL;
				goto out_unmap;
			}
961 962 963 964 965 966
			if (list_empty(&mm->mmlist)) {
				spin_lock(&mmlist_lock);
				if (list_empty(&mm->mmlist))
					list_add(&mm->mmlist, &init_mm.mmlist);
				spin_unlock(&mmlist_lock);
			}
K
KAMEZAWA Hiroyuki 已提交
967
			dec_mm_counter(mm, MM_ANONPAGES);
K
KAMEZAWA Hiroyuki 已提交
968
			inc_mm_counter(mm, MM_SWAPENTS);
969
		} else if (PAGE_MIGRATION) {
970 971 972 973 974
			/*
			 * Store the pfn of the page in a special migration
			 * pte. do_swap_page() will wait until the migration
			 * pte is removed and then restart fault handling.
			 */
975
			BUG_ON(TTU_ACTION(flags) != TTU_MIGRATION);
976
			entry = make_migration_entry(page, pte_write(pteval));
L
Linus Torvalds 已提交
977 978 979
		}
		set_pte_at(mm, address, pte, swp_entry_to_pte(entry));
		BUG_ON(pte_file(*pte));
980
	} else if (PAGE_MIGRATION && (TTU_ACTION(flags) == TTU_MIGRATION)) {
981 982 983 984 985
		/* Establish migration entry for a file page */
		swp_entry_t entry;
		entry = make_migration_entry(page, pte_write(pteval));
		set_pte_at(mm, address, pte, swp_entry_to_pte(entry));
	} else
K
KAMEZAWA Hiroyuki 已提交
986
		dec_mm_counter(mm, MM_FILEPAGES);
L
Linus Torvalds 已提交
987

988
	page_remove_rmap(page);
L
Linus Torvalds 已提交
989 990 991
	page_cache_release(page);

out_unmap:
H
Hugh Dickins 已提交
992
	pte_unmap_unlock(pte, ptl);
K
KOSAKI Motohiro 已提交
993 994
out:
	return ret;
H
Hugh Dickins 已提交
995

K
KOSAKI Motohiro 已提交
996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011
out_mlock:
	pte_unmap_unlock(pte, ptl);


	/*
	 * We need mmap_sem locking, Otherwise VM_LOCKED check makes
	 * unstable result and race. Plus, We can't wait here because
	 * we now hold anon_vma->lock or mapping->i_mmap_lock.
	 * if trylock failed, the page remain in evictable lru and later
	 * vmscan could retry to move the page to unevictable lru if the
	 * page is actually mlocked.
	 */
	if (down_read_trylock(&vma->vm_mm->mmap_sem)) {
		if (vma->vm_flags & VM_LOCKED) {
			mlock_vma_page(page);
			ret = SWAP_MLOCK;
H
Hugh Dickins 已提交
1012
		}
K
KOSAKI Motohiro 已提交
1013
		up_read(&vma->vm_mm->mmap_sem);
H
Hugh Dickins 已提交
1014
	}
L
Linus Torvalds 已提交
1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035
	return ret;
}

/*
 * objrmap doesn't work for nonlinear VMAs because the assumption that
 * offset-into-file correlates with offset-into-virtual-addresses does not hold.
 * Consequently, given a particular page and its ->index, we cannot locate the
 * ptes which are mapping that page without an exhaustive linear search.
 *
 * So what this code does is a mini "virtual scan" of each nonlinear VMA which
 * maps the file to which the target page belongs.  The ->vm_private_data field
 * holds the current cursor into that scan.  Successive searches will circulate
 * around the vma's virtual address space.
 *
 * So as more replacement pressure is applied to the pages in a nonlinear VMA,
 * more scanning pressure is placed against them as well.   Eventually pages
 * will become fully unmapped and are eligible for eviction.
 *
 * For very sparsely populated VMAs this is a little inefficient - chances are
 * there there won't be many ptes located within the scan cluster.  In this case
 * maybe we could scan further - to the end of the pte page, perhaps.
N
Nick Piggin 已提交
1036 1037 1038 1039 1040
 *
 * Mlocked pages:  check VM_LOCKED under mmap_sem held for read, if we can
 * acquire it without blocking.  If vma locked, mlock the pages in the cluster,
 * rather than unmapping them.  If we encounter the "check_page" that vmscan is
 * trying to unmap, return SWAP_MLOCK, else default SWAP_AGAIN.
L
Linus Torvalds 已提交
1041 1042 1043 1044
 */
#define CLUSTER_SIZE	min(32*PAGE_SIZE, PMD_SIZE)
#define CLUSTER_MASK	(~(CLUSTER_SIZE - 1))

N
Nick Piggin 已提交
1045 1046
static int try_to_unmap_cluster(unsigned long cursor, unsigned int *mapcount,
		struct vm_area_struct *vma, struct page *check_page)
L
Linus Torvalds 已提交
1047 1048 1049 1050 1051
{
	struct mm_struct *mm = vma->vm_mm;
	pgd_t *pgd;
	pud_t *pud;
	pmd_t *pmd;
H
Hugh Dickins 已提交
1052
	pte_t *pte;
L
Linus Torvalds 已提交
1053
	pte_t pteval;
H
Hugh Dickins 已提交
1054
	spinlock_t *ptl;
L
Linus Torvalds 已提交
1055 1056 1057
	struct page *page;
	unsigned long address;
	unsigned long end;
N
Nick Piggin 已提交
1058 1059
	int ret = SWAP_AGAIN;
	int locked_vma = 0;
L
Linus Torvalds 已提交
1060 1061 1062 1063 1064 1065 1066 1067 1068 1069

	address = (vma->vm_start + cursor) & CLUSTER_MASK;
	end = address + CLUSTER_SIZE;
	if (address < vma->vm_start)
		address = vma->vm_start;
	if (end > vma->vm_end)
		end = vma->vm_end;

	pgd = pgd_offset(mm, address);
	if (!pgd_present(*pgd))
N
Nick Piggin 已提交
1070
		return ret;
L
Linus Torvalds 已提交
1071 1072 1073

	pud = pud_offset(pgd, address);
	if (!pud_present(*pud))
N
Nick Piggin 已提交
1074
		return ret;
L
Linus Torvalds 已提交
1075 1076 1077

	pmd = pmd_offset(pud, address);
	if (!pmd_present(*pmd))
N
Nick Piggin 已提交
1078 1079 1080
		return ret;

	/*
H
Hugh Dickins 已提交
1081
	 * If we can acquire the mmap_sem for read, and vma is VM_LOCKED,
N
Nick Piggin 已提交
1082 1083
	 * keep the sem while scanning the cluster for mlocking pages.
	 */
H
Hugh Dickins 已提交
1084
	if (down_read_trylock(&vma->vm_mm->mmap_sem)) {
N
Nick Piggin 已提交
1085 1086 1087 1088
		locked_vma = (vma->vm_flags & VM_LOCKED);
		if (!locked_vma)
			up_read(&vma->vm_mm->mmap_sem); /* don't need it */
	}
H
Hugh Dickins 已提交
1089 1090

	pte = pte_offset_map_lock(mm, pmd, address, &ptl);
L
Linus Torvalds 已提交
1091

1092 1093 1094
	/* Update high watermark before we lower rss */
	update_hiwater_rss(mm);

H
Hugh Dickins 已提交
1095
	for (; address < end; pte++, address += PAGE_SIZE) {
L
Linus Torvalds 已提交
1096 1097
		if (!pte_present(*pte))
			continue;
1098 1099
		page = vm_normal_page(vma, address, *pte);
		BUG_ON(!page || PageAnon(page));
L
Linus Torvalds 已提交
1100

N
Nick Piggin 已提交
1101 1102 1103 1104 1105 1106 1107
		if (locked_vma) {
			mlock_vma_page(page);   /* no-op if already mlocked */
			if (page == check_page)
				ret = SWAP_MLOCK;
			continue;	/* don't unmap */
		}

A
Andrea Arcangeli 已提交
1108
		if (ptep_clear_flush_young_notify(vma, address, pte))
L
Linus Torvalds 已提交
1109 1110 1111
			continue;

		/* Nuke the page table entry. */
1112
		flush_cache_page(vma, address, pte_pfn(*pte));
A
Andrea Arcangeli 已提交
1113
		pteval = ptep_clear_flush_notify(vma, address, pte);
L
Linus Torvalds 已提交
1114 1115 1116 1117 1118 1119 1120 1121 1122

		/* If nonlinear, store the file page offset in the pte. */
		if (page->index != linear_page_index(vma, address))
			set_pte_at(mm, address, pte, pgoff_to_pte(page->index));

		/* Move the dirty bit to the physical page now the pte is gone. */
		if (pte_dirty(pteval))
			set_page_dirty(page);

1123
		page_remove_rmap(page);
L
Linus Torvalds 已提交
1124
		page_cache_release(page);
K
KAMEZAWA Hiroyuki 已提交
1125
		dec_mm_counter(mm, MM_FILEPAGES);
L
Linus Torvalds 已提交
1126 1127
		(*mapcount)--;
	}
H
Hugh Dickins 已提交
1128
	pte_unmap_unlock(pte - 1, ptl);
N
Nick Piggin 已提交
1129 1130 1131
	if (locked_vma)
		up_read(&vma->vm_mm->mmap_sem);
	return ret;
L
Linus Torvalds 已提交
1132 1133
}

N
Nick Piggin 已提交
1134 1135 1136 1137
/**
 * try_to_unmap_anon - unmap or unlock anonymous page using the object-based
 * rmap method
 * @page: the page to unmap/unlock
1138
 * @flags: action and flags
N
Nick Piggin 已提交
1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149
 *
 * Find all the mappings of a page using the mapping pointer and the vma chains
 * contained in the anon_vma struct it points to.
 *
 * This function is only called from try_to_unmap/try_to_munlock for
 * anonymous pages.
 * When called from try_to_munlock(), the mmap_sem of the mm containing the vma
 * where the page was found will be held for write.  So, we won't recheck
 * vm_flags for that VMA.  That should be OK, because that vma shouldn't be
 * 'LOCKED.
 */
1150
static int try_to_unmap_anon(struct page *page, enum ttu_flags flags)
L
Linus Torvalds 已提交
1151 1152
{
	struct anon_vma *anon_vma;
1153
	struct anon_vma_chain *avc;
L
Linus Torvalds 已提交
1154
	int ret = SWAP_AGAIN;
N
Nick Piggin 已提交
1155

L
Linus Torvalds 已提交
1156 1157 1158 1159
	anon_vma = page_lock_anon_vma(page);
	if (!anon_vma)
		return ret;

1160 1161
	list_for_each_entry(avc, &anon_vma->head, same_anon_vma) {
		struct vm_area_struct *vma = avc->vma;
H
Hugh Dickins 已提交
1162 1163 1164 1165
		unsigned long address = vma_address(page, vma);
		if (address == -EFAULT)
			continue;
		ret = try_to_unmap_one(page, vma, address, flags);
H
Hugh Dickins 已提交
1166 1167
		if (ret != SWAP_AGAIN || !page_mapped(page))
			break;
L
Linus Torvalds 已提交
1168
	}
1169 1170

	page_unlock_anon_vma(anon_vma);
L
Linus Torvalds 已提交
1171 1172 1173 1174
	return ret;
}

/**
N
Nick Piggin 已提交
1175 1176
 * try_to_unmap_file - unmap/unlock file page using the object-based rmap method
 * @page: the page to unmap/unlock
1177
 * @flags: action and flags
L
Linus Torvalds 已提交
1178 1179 1180 1181
 *
 * Find all the mappings of a page using the mapping pointer and the vma chains
 * contained in the address_space struct it points to.
 *
N
Nick Piggin 已提交
1182 1183 1184 1185 1186 1187
 * This function is only called from try_to_unmap/try_to_munlock for
 * object-based pages.
 * When called from try_to_munlock(), the mmap_sem of the mm containing the vma
 * where the page was found will be held for write.  So, we won't recheck
 * vm_flags for that VMA.  That should be OK, because that vma shouldn't be
 * 'LOCKED.
L
Linus Torvalds 已提交
1188
 */
1189
static int try_to_unmap_file(struct page *page, enum ttu_flags flags)
L
Linus Torvalds 已提交
1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202
{
	struct address_space *mapping = page->mapping;
	pgoff_t pgoff = page->index << (PAGE_CACHE_SHIFT - PAGE_SHIFT);
	struct vm_area_struct *vma;
	struct prio_tree_iter iter;
	int ret = SWAP_AGAIN;
	unsigned long cursor;
	unsigned long max_nl_cursor = 0;
	unsigned long max_nl_size = 0;
	unsigned int mapcount;

	spin_lock(&mapping->i_mmap_lock);
	vma_prio_tree_foreach(vma, &iter, &mapping->i_mmap, pgoff, pgoff) {
H
Hugh Dickins 已提交
1203 1204 1205 1206
		unsigned long address = vma_address(page, vma);
		if (address == -EFAULT)
			continue;
		ret = try_to_unmap_one(page, vma, address, flags);
H
Hugh Dickins 已提交
1207 1208
		if (ret != SWAP_AGAIN || !page_mapped(page))
			goto out;
L
Linus Torvalds 已提交
1209 1210 1211 1212 1213
	}

	if (list_empty(&mapping->i_mmap_nonlinear))
		goto out;

H
Hugh Dickins 已提交
1214 1215 1216 1217 1218 1219 1220 1221
	/*
	 * We don't bother to try to find the munlocked page in nonlinears.
	 * It's costly. Instead, later, page reclaim logic may call
	 * try_to_unmap(TTU_MUNLOCK) and recover PG_mlocked lazily.
	 */
	if (TTU_ACTION(flags) == TTU_MUNLOCK)
		goto out;

L
Linus Torvalds 已提交
1222 1223 1224 1225 1226 1227 1228 1229 1230 1231
	list_for_each_entry(vma, &mapping->i_mmap_nonlinear,
						shared.vm_set.list) {
		cursor = (unsigned long) vma->vm_private_data;
		if (cursor > max_nl_cursor)
			max_nl_cursor = cursor;
		cursor = vma->vm_end - vma->vm_start;
		if (cursor > max_nl_size)
			max_nl_size = cursor;
	}

N
Nick Piggin 已提交
1232
	if (max_nl_size == 0) {	/* all nonlinears locked or reserved ? */
L
Linus Torvalds 已提交
1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256
		ret = SWAP_FAIL;
		goto out;
	}

	/*
	 * We don't try to search for this page in the nonlinear vmas,
	 * and page_referenced wouldn't have found it anyway.  Instead
	 * just walk the nonlinear vmas trying to age and unmap some.
	 * The mapcount of the page we came in with is irrelevant,
	 * but even so use it as a guide to how hard we should try?
	 */
	mapcount = page_mapcount(page);
	if (!mapcount)
		goto out;
	cond_resched_lock(&mapping->i_mmap_lock);

	max_nl_size = (max_nl_size + CLUSTER_SIZE - 1) & CLUSTER_MASK;
	if (max_nl_cursor == 0)
		max_nl_cursor = CLUSTER_SIZE;

	do {
		list_for_each_entry(vma, &mapping->i_mmap_nonlinear,
						shared.vm_set.list) {
			cursor = (unsigned long) vma->vm_private_data;
H
Hugh Dickins 已提交
1257
			while ( cursor < max_nl_cursor &&
L
Linus Torvalds 已提交
1258
				cursor < vma->vm_end - vma->vm_start) {
H
Hugh Dickins 已提交
1259 1260 1261
				if (try_to_unmap_cluster(cursor, &mapcount,
						vma, page) == SWAP_MLOCK)
					ret = SWAP_MLOCK;
L
Linus Torvalds 已提交
1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277
				cursor += CLUSTER_SIZE;
				vma->vm_private_data = (void *) cursor;
				if ((int)mapcount <= 0)
					goto out;
			}
			vma->vm_private_data = (void *) max_nl_cursor;
		}
		cond_resched_lock(&mapping->i_mmap_lock);
		max_nl_cursor += CLUSTER_SIZE;
	} while (max_nl_cursor <= max_nl_size);

	/*
	 * Don't loop forever (perhaps all the remaining pages are
	 * in locked vmas).  Reset cursor on all unreserved nonlinear
	 * vmas, now forgetting on which ones it had fallen behind.
	 */
1278 1279
	list_for_each_entry(vma, &mapping->i_mmap_nonlinear, shared.vm_set.list)
		vma->vm_private_data = NULL;
L
Linus Torvalds 已提交
1280 1281 1282 1283 1284 1285 1286 1287
out:
	spin_unlock(&mapping->i_mmap_lock);
	return ret;
}

/**
 * try_to_unmap - try to remove all page table mappings to a page
 * @page: the page to get unmapped
1288
 * @flags: action and flags
L
Linus Torvalds 已提交
1289 1290 1291 1292 1293 1294 1295 1296
 *
 * Tries to remove all the page table entries which are mapping this
 * page, used in the pageout path.  Caller must hold the page lock.
 * Return values are:
 *
 * SWAP_SUCCESS	- we succeeded in removing all mappings
 * SWAP_AGAIN	- we missed a mapping, try again later
 * SWAP_FAIL	- the page is unswappable
N
Nick Piggin 已提交
1297
 * SWAP_MLOCK	- page is mlocked.
L
Linus Torvalds 已提交
1298
 */
1299
int try_to_unmap(struct page *page, enum ttu_flags flags)
L
Linus Torvalds 已提交
1300 1301 1302 1303 1304
{
	int ret;

	BUG_ON(!PageLocked(page));

H
Hugh Dickins 已提交
1305 1306 1307
	if (unlikely(PageKsm(page)))
		ret = try_to_unmap_ksm(page, flags);
	else if (PageAnon(page))
1308
		ret = try_to_unmap_anon(page, flags);
L
Linus Torvalds 已提交
1309
	else
1310
		ret = try_to_unmap_file(page, flags);
N
Nick Piggin 已提交
1311
	if (ret != SWAP_MLOCK && !page_mapped(page))
L
Linus Torvalds 已提交
1312 1313 1314
		ret = SWAP_SUCCESS;
	return ret;
}
N
Nikita Danilov 已提交
1315

N
Nick Piggin 已提交
1316 1317 1318 1319 1320 1321 1322 1323 1324 1325
/**
 * try_to_munlock - try to munlock a page
 * @page: the page to be munlocked
 *
 * Called from munlock code.  Checks all of the VMAs mapping the page
 * to make sure nobody else has this page mlocked. The page will be
 * returned with PG_mlocked cleared if no other vmas have it mlocked.
 *
 * Return values are:
 *
H
Hugh Dickins 已提交
1326
 * SWAP_AGAIN	- no vma is holding page mlocked, or,
N
Nick Piggin 已提交
1327
 * SWAP_AGAIN	- page mapped in mlocked vma -- couldn't acquire mmap sem
H
Hugh Dickins 已提交
1328
 * SWAP_FAIL	- page cannot be located at present
N
Nick Piggin 已提交
1329 1330 1331 1332 1333 1334
 * SWAP_MLOCK	- page is now mlocked.
 */
int try_to_munlock(struct page *page)
{
	VM_BUG_ON(!PageLocked(page) || PageLRU(page));

H
Hugh Dickins 已提交
1335 1336 1337
	if (unlikely(PageKsm(page)))
		return try_to_unmap_ksm(page, TTU_MUNLOCK);
	else if (PageAnon(page))
1338
		return try_to_unmap_anon(page, TTU_MUNLOCK);
N
Nick Piggin 已提交
1339
	else
1340
		return try_to_unmap_file(page, TTU_MUNLOCK);
N
Nick Piggin 已提交
1341
}
1342 1343 1344 1345 1346 1347 1348 1349 1350 1351

#ifdef CONFIG_MIGRATION
/*
 * rmap_walk() and its helpers rmap_walk_anon() and rmap_walk_file():
 * Called by migrate.c to remove migration ptes, but might be used more later.
 */
static int rmap_walk_anon(struct page *page, int (*rmap_one)(struct page *,
		struct vm_area_struct *, unsigned long, void *), void *arg)
{
	struct anon_vma *anon_vma;
1352
	struct anon_vma_chain *avc;
1353 1354 1355 1356 1357
	int ret = SWAP_AGAIN;

	/*
	 * Note: remove_migration_ptes() cannot use page_lock_anon_vma()
	 * because that depends on page_mapped(); but not all its usages
1358 1359
	 * are holding mmap_sem. Users without mmap_sem are required to
	 * take a reference count to prevent the anon_vma disappearing
1360 1361 1362 1363 1364
	 */
	anon_vma = page_anon_vma(page);
	if (!anon_vma)
		return ret;
	spin_lock(&anon_vma->lock);
1365 1366
	list_for_each_entry(avc, &anon_vma->head, same_anon_vma) {
		struct vm_area_struct *vma = avc->vma;
1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419
		unsigned long address = vma_address(page, vma);
		if (address == -EFAULT)
			continue;
		ret = rmap_one(page, vma, address, arg);
		if (ret != SWAP_AGAIN)
			break;
	}
	spin_unlock(&anon_vma->lock);
	return ret;
}

static int rmap_walk_file(struct page *page, int (*rmap_one)(struct page *,
		struct vm_area_struct *, unsigned long, void *), void *arg)
{
	struct address_space *mapping = page->mapping;
	pgoff_t pgoff = page->index << (PAGE_CACHE_SHIFT - PAGE_SHIFT);
	struct vm_area_struct *vma;
	struct prio_tree_iter iter;
	int ret = SWAP_AGAIN;

	if (!mapping)
		return ret;
	spin_lock(&mapping->i_mmap_lock);
	vma_prio_tree_foreach(vma, &iter, &mapping->i_mmap, pgoff, pgoff) {
		unsigned long address = vma_address(page, vma);
		if (address == -EFAULT)
			continue;
		ret = rmap_one(page, vma, address, arg);
		if (ret != SWAP_AGAIN)
			break;
	}
	/*
	 * No nonlinear handling: being always shared, nonlinear vmas
	 * never contain migration ptes.  Decide what to do about this
	 * limitation to linear when we need rmap_walk() on nonlinear.
	 */
	spin_unlock(&mapping->i_mmap_lock);
	return ret;
}

int rmap_walk(struct page *page, int (*rmap_one)(struct page *,
		struct vm_area_struct *, unsigned long, void *), void *arg)
{
	VM_BUG_ON(!PageLocked(page));

	if (unlikely(PageKsm(page)))
		return rmap_walk_ksm(page, rmap_one, arg);
	else if (PageAnon(page))
		return rmap_walk_anon(page, rmap_one, arg);
	else
		return rmap_walk_file(page, rmap_one, arg);
}
#endif /* CONFIG_MIGRATION */