rmap.c 35.1 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)
L
Linus Torvalds 已提交
39 40 41 42 43 44 45 46 47 48
 */

#include <linux/mm.h>
#include <linux/pagemap.h>
#include <linux/swap.h>
#include <linux/swapops.h>
#include <linux/slab.h>
#include <linux/init.h>
#include <linux/rmap.h>
#include <linux/rcupdate.h>
49
#include <linux/module.h>
50
#include <linux/memcontrol.h>
A
Andrea Arcangeli 已提交
51
#include <linux/mmu_notifier.h>
52
#include <linux/migrate.h>
L
Linus Torvalds 已提交
53 54 55

#include <asm/tlbflush.h>

N
Nick Piggin 已提交
56 57
#include "internal.h"

58 59 60 61 62 63 64 65 66 67 68
static struct kmem_cache *anon_vma_cachep;

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

static inline void anon_vma_free(struct anon_vma *anon_vma)
{
	kmem_cache_free(anon_vma_cachep, anon_vma);
}
L
Linus Torvalds 已提交
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
/**
 * 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
97 98 99 100 101 102 103
int anon_vma_prepare(struct vm_area_struct *vma)
{
	struct anon_vma *anon_vma = vma->anon_vma;

	might_sleep();
	if (unlikely(!anon_vma)) {
		struct mm_struct *mm = vma->vm_mm;
104
		struct anon_vma *allocated;
L
Linus Torvalds 已提交
105 106

		anon_vma = find_mergeable_anon_vma(vma);
107 108
		allocated = NULL;
		if (!anon_vma) {
L
Linus Torvalds 已提交
109 110 111 112 113
			anon_vma = anon_vma_alloc();
			if (unlikely(!anon_vma))
				return -ENOMEM;
			allocated = anon_vma;
		}
114
		spin_lock(&anon_vma->lock);
L
Linus Torvalds 已提交
115 116 117 118 119

		/* page_table_lock to protect against threads */
		spin_lock(&mm->page_table_lock);
		if (likely(!vma->anon_vma)) {
			vma->anon_vma = anon_vma;
120
			list_add_tail(&vma->anon_vma_node, &anon_vma->head);
L
Linus Torvalds 已提交
121 122 123 124
			allocated = NULL;
		}
		spin_unlock(&mm->page_table_lock);

125
		spin_unlock(&anon_vma->lock);
L
Linus Torvalds 已提交
126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141
		if (unlikely(allocated))
			anon_vma_free(allocated);
	}
	return 0;
}

void __anon_vma_merge(struct vm_area_struct *vma, struct vm_area_struct *next)
{
	BUG_ON(vma->anon_vma != next->anon_vma);
	list_del(&next->anon_vma_node);
}

void __anon_vma_link(struct vm_area_struct *vma)
{
	struct anon_vma *anon_vma = vma->anon_vma;

142
	if (anon_vma)
143
		list_add_tail(&vma->anon_vma_node, &anon_vma->head);
L
Linus Torvalds 已提交
144 145 146 147 148 149 150 151
}

void anon_vma_link(struct vm_area_struct *vma)
{
	struct anon_vma *anon_vma = vma->anon_vma;

	if (anon_vma) {
		spin_lock(&anon_vma->lock);
152
		list_add_tail(&vma->anon_vma_node, &anon_vma->head);
L
Linus Torvalds 已提交
153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175
		spin_unlock(&anon_vma->lock);
	}
}

void anon_vma_unlink(struct vm_area_struct *vma)
{
	struct anon_vma *anon_vma = vma->anon_vma;
	int empty;

	if (!anon_vma)
		return;

	spin_lock(&anon_vma->lock);
	list_del(&vma->anon_vma_node);

	/* We must garbage collect the anon_vma if it's empty */
	empty = list_empty(&anon_vma->head);
	spin_unlock(&anon_vma->lock);

	if (empty)
		anon_vma_free(anon_vma);
}

176
static void anon_vma_ctor(void *data)
L
Linus Torvalds 已提交
177
{
C
Christoph Lameter 已提交
178
	struct anon_vma *anon_vma = data;
L
Linus Torvalds 已提交
179

C
Christoph Lameter 已提交
180 181
	spin_lock_init(&anon_vma->lock);
	INIT_LIST_HEAD(&anon_vma->head);
L
Linus Torvalds 已提交
182 183 184 185 186
}

void __init anon_vma_init(void)
{
	anon_vma_cachep = kmem_cache_create("anon_vma", sizeof(struct anon_vma),
187
			0, SLAB_DESTROY_BY_RCU|SLAB_PANIC, anon_vma_ctor);
L
Linus Torvalds 已提交
188 189 190 191 192 193
}

/*
 * 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.
 */
194
struct anon_vma *page_lock_anon_vma(struct page *page)
L
Linus Torvalds 已提交
195
{
196
	struct anon_vma *anon_vma;
L
Linus Torvalds 已提交
197 198 199 200 201 202 203 204 205 206 207
	unsigned long anon_mapping;

	rcu_read_lock();
	anon_mapping = (unsigned long) page->mapping;
	if (!(anon_mapping & PAGE_MAPPING_ANON))
		goto out;
	if (!page_mapped(page))
		goto out;

	anon_vma = (struct anon_vma *) (anon_mapping - PAGE_MAPPING_ANON);
	spin_lock(&anon_vma->lock);
208
	return anon_vma;
L
Linus Torvalds 已提交
209 210
out:
	rcu_read_unlock();
211 212 213
	return NULL;
}

214
void page_unlock_anon_vma(struct anon_vma *anon_vma)
215 216 217
{
	spin_unlock(&anon_vma->lock);
	rcu_read_unlock();
L
Linus Torvalds 已提交
218 219 220
}

/*
221 222 223
 * 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
224 225 226 227 228 229 230 231 232
 */
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)) {
233
		/* page should be within @vma mapping range */
L
Linus Torvalds 已提交
234 235 236 237 238 239 240
		return -EFAULT;
	}
	return address;
}

/*
 * At what user virtual address is page expected in vma? checking that the
H
Hugh Dickins 已提交
241
 * page matches the vma: currently only used on anon pages, by unuse_vma;
L
Linus Torvalds 已提交
242 243 244 245 246 247 248 249
 */
unsigned long page_address_in_vma(struct page *page, struct vm_area_struct *vma)
{
	if (PageAnon(page)) {
		if ((void *)vma->anon_vma !=
		    (void *)page->mapping - PAGE_MAPPING_ANON)
			return -EFAULT;
	} else if (page->mapping && !(vma->vm_flags & VM_NONLINEAR)) {
H
Hugh Dickins 已提交
250 251
		if (!vma->vm_file ||
		    vma->vm_file->f_mapping != page->mapping)
L
Linus Torvalds 已提交
252 253 254 255 256 257
			return -EFAULT;
	} else
		return -EFAULT;
	return vma_address(page, vma);
}

N
Nikita Danilov 已提交
258 259 260
/*
 * Check that @page is mapped at @address into @mm.
 *
N
Nick Piggin 已提交
261 262 263 264
 * 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).
 *
265
 * On success returns with pte mapped and locked.
N
Nikita Danilov 已提交
266
 */
267
pte_t *page_check_address(struct page *page, struct mm_struct *mm,
N
Nick Piggin 已提交
268
			  unsigned long address, spinlock_t **ptlp, int sync)
N
Nikita Danilov 已提交
269 270 271 272 273
{
	pgd_t *pgd;
	pud_t *pud;
	pmd_t *pmd;
	pte_t *pte;
H
Hugh Dickins 已提交
274
	spinlock_t *ptl;
N
Nikita Danilov 已提交
275 276

	pgd = pgd_offset(mm, address);
H
Hugh Dickins 已提交
277 278 279 280 281 282 283 284 285 286 287 288 289
	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
290
	if (!sync && !pte_present(*pte)) {
H
Hugh Dickins 已提交
291 292 293 294
		pte_unmap(pte);
		return NULL;
	}

H
Hugh Dickins 已提交
295
	ptl = pte_lockptr(mm, pmd);
H
Hugh Dickins 已提交
296 297 298 299
	spin_lock(ptl);
	if (pte_present(*pte) && page_to_pfn(page) == pte_pfn(*pte)) {
		*ptlp = ptl;
		return pte;
N
Nikita Danilov 已提交
300
	}
H
Hugh Dickins 已提交
301 302
	pte_unmap_unlock(pte, ptl);
	return NULL;
N
Nikita Danilov 已提交
303 304
}

N
Nick Piggin 已提交
305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330
/**
 * 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.
 */
static int page_mapped_in_vma(struct page *page, struct vm_area_struct *vma)
{
	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
331 332 333 334 335
/*
 * Subfunctions of page_referenced: page_referenced_one called
 * repeatedly from either page_referenced_anon or page_referenced_file.
 */
static int page_referenced_one(struct page *page,
336 337 338
			       struct vm_area_struct *vma,
			       unsigned int *mapcount,
			       unsigned long *vm_flags)
L
Linus Torvalds 已提交
339 340 341 342
{
	struct mm_struct *mm = vma->vm_mm;
	unsigned long address;
	pte_t *pte;
H
Hugh Dickins 已提交
343
	spinlock_t *ptl;
L
Linus Torvalds 已提交
344 345 346 347 348 349
	int referenced = 0;

	address = vma_address(page, vma);
	if (address == -EFAULT)
		goto out;

N
Nick Piggin 已提交
350
	pte = page_check_address(page, mm, address, &ptl, 0);
H
Hugh Dickins 已提交
351 352
	if (!pte)
		goto out;
L
Linus Torvalds 已提交
353

N
Nick Piggin 已提交
354 355 356 357 358
	/*
	 * 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.
	 */
359 360
	if (vma->vm_flags & VM_LOCKED) {
		*mapcount = 1;	/* break early from loop */
361
		*vm_flags |= VM_LOCKED;
N
Nick Piggin 已提交
362 363 364
		goto out_unmap;
	}

365 366 367 368 369 370 371 372 373 374 375
	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
376

H
Hugh Dickins 已提交
377 378
	/* Pretend the page is referenced if the task has the
	   swap token and is in the middle of a page fault. */
379
	if (mm != current->mm && has_swap_token(mm) &&
H
Hugh Dickins 已提交
380 381 382
			rwsem_is_locked(&mm->mmap_sem))
		referenced++;

N
Nick Piggin 已提交
383
out_unmap:
H
Hugh Dickins 已提交
384 385
	(*mapcount)--;
	pte_unmap_unlock(pte, ptl);
L
Linus Torvalds 已提交
386
out:
387 388
	if (referenced)
		*vm_flags |= vma->vm_flags;
L
Linus Torvalds 已提交
389 390 391
	return referenced;
}

392
static int page_referenced_anon(struct page *page,
393 394
				struct mem_cgroup *mem_cont,
				unsigned long *vm_flags)
L
Linus Torvalds 已提交
395 396 397 398 399 400 401 402 403 404 405 406
{
	unsigned int mapcount;
	struct anon_vma *anon_vma;
	struct vm_area_struct *vma;
	int referenced = 0;

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

	mapcount = page_mapcount(page);
	list_for_each_entry(vma, &anon_vma->head, anon_vma_node) {
407 408 409 410 411
		/*
		 * If we are reclaiming on behalf of a cgroup, skip
		 * counting on behalf of references from different
		 * cgroups
		 */
412
		if (mem_cont && !mm_match_cgroup(vma->vm_mm, mem_cont))
413
			continue;
414 415
		referenced += page_referenced_one(page, vma,
						  &mapcount, vm_flags);
L
Linus Torvalds 已提交
416 417 418
		if (!mapcount)
			break;
	}
419 420

	page_unlock_anon_vma(anon_vma);
L
Linus Torvalds 已提交
421 422 423 424 425 426
	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
427
 * @mem_cont: target memory controller
428
 * @vm_flags: collect encountered vma->vm_flags who actually referenced the page
L
Linus Torvalds 已提交
429 430 431 432 433 434 435 436
 *
 * 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.
 */
437
static int page_referenced_file(struct page *page,
438 439
				struct mem_cgroup *mem_cont,
				unsigned long *vm_flags)
L
Linus Torvalds 已提交
440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471
{
	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) {
472 473 474 475 476
		/*
		 * If we are reclaiming on behalf of a cgroup, skip
		 * counting on behalf of references from different
		 * cgroups
		 */
477
		if (mem_cont && !mm_match_cgroup(vma->vm_mm, mem_cont))
478
			continue;
479 480
		referenced += page_referenced_one(page, vma,
						  &mapcount, vm_flags);
L
Linus Torvalds 已提交
481 482 483 484 485 486 487 488 489 490 491 492
		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
493
 * @mem_cont: target memory controller
494
 * @vm_flags: collect encountered vma->vm_flags who actually referenced the page
L
Linus Torvalds 已提交
495 496 497 498
 *
 * Quick test_and_clear_referenced for all mappings to a page,
 * returns the number of ptes which referenced the page.
 */
499 500 501 502
int page_referenced(struct page *page,
		    int is_locked,
		    struct mem_cgroup *mem_cont,
		    unsigned long *vm_flags)
L
Linus Torvalds 已提交
503 504 505 506 507 508
{
	int referenced = 0;

	if (TestClearPageReferenced(page))
		referenced++;

509
	*vm_flags = 0;
L
Linus Torvalds 已提交
510 511
	if (page_mapped(page) && page->mapping) {
		if (PageAnon(page))
512 513
			referenced += page_referenced_anon(page, mem_cont,
								vm_flags);
L
Linus Torvalds 已提交
514
		else if (is_locked)
515 516
			referenced += page_referenced_file(page, mem_cont,
								vm_flags);
N
Nick Piggin 已提交
517
		else if (!trylock_page(page))
L
Linus Torvalds 已提交
518 519 520
			referenced++;
		else {
			if (page->mapping)
521 522
				referenced += page_referenced_file(page,
							mem_cont, vm_flags);
L
Linus Torvalds 已提交
523 524 525
			unlock_page(page);
		}
	}
526 527 528 529

	if (page_test_and_clear_young(page))
		referenced++;

L
Linus Torvalds 已提交
530 531 532
	return referenced;
}

533 534 535 536
static int page_mkclean_one(struct page *page, struct vm_area_struct *vma)
{
	struct mm_struct *mm = vma->vm_mm;
	unsigned long address;
537
	pte_t *pte;
538 539 540 541 542 543 544
	spinlock_t *ptl;
	int ret = 0;

	address = vma_address(page, vma);
	if (address == -EFAULT)
		goto out;

N
Nick Piggin 已提交
545
	pte = page_check_address(page, mm, address, &ptl, 1);
546 547 548
	if (!pte)
		goto out;

549 550
	if (pte_dirty(*pte) || pte_write(*pte)) {
		pte_t entry;
551

552
		flush_cache_page(vma, address, pte_pfn(*pte));
A
Andrea Arcangeli 已提交
553
		entry = ptep_clear_flush_notify(vma, address, pte);
554 555
		entry = pte_wrprotect(entry);
		entry = pte_mkclean(entry);
556
		set_pte_at(mm, address, pte, entry);
557 558
		ret = 1;
	}
559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590

	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) {
		if (vma->vm_flags & VM_SHARED)
			ret += page_mkclean_one(page, vma);
	}
	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);
591
		if (mapping) {
592
			ret = page_mkclean_file(mapping, page);
593 594 595 596
			if (page_test_dirty(page)) {
				page_clear_dirty(page);
				ret = 1;
			}
597
		}
598 599 600 601
	}

	return ret;
}
J
Jaya Kumar 已提交
602
EXPORT_SYMBOL_GPL(page_mkclean);
603

N
Nick Piggin 已提交
604
/**
R
Randy Dunlap 已提交
605
 * __page_set_anon_rmap - setup new anonymous rmap
N
Nick Piggin 已提交
606 607 608 609 610 611 612 613 614 615 616 617 618 619 620
 * @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_set_anon_rmap(struct page *page,
	struct vm_area_struct *vma, unsigned long address)
{
	struct anon_vma *anon_vma = vma->anon_vma;

	BUG_ON(!anon_vma);
	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
621 622 623 624
	/*
	 * nr_mapped state can be updated without turning off
	 * interrupts because it is not modified via interrupt.
	 */
625
	__inc_zone_page_state(page, NR_ANON_PAGES);
N
Nick Piggin 已提交
626 627
}

N
Nick Piggin 已提交
628
/**
R
Randy Dunlap 已提交
629
 * __page_check_anon_rmap - sanity check anonymous rmap addition
N
Nick Piggin 已提交
630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656
 * @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.
	 */
	struct anon_vma *anon_vma = vma->anon_vma;
	anon_vma = (void *) anon_vma + PAGE_MAPPING_ANON;
	BUG_ON(page->mapping != (struct address_space *)anon_vma);
	BUG_ON(page->index != linear_page_index(vma, address));
#endif
}

L
Linus Torvalds 已提交
657 658 659 660 661 662
/**
 * 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
 *
N
Nick Piggin 已提交
663
 * The caller needs to hold the pte lock and the page must be locked.
L
Linus Torvalds 已提交
664 665 666 667
 */
void page_add_anon_rmap(struct page *page,
	struct vm_area_struct *vma, unsigned long address)
{
N
Nick Piggin 已提交
668 669
	VM_BUG_ON(!PageLocked(page));
	VM_BUG_ON(address < vma->vm_start || address >= vma->vm_end);
N
Nick Piggin 已提交
670 671
	if (atomic_inc_and_test(&page->_mapcount))
		__page_set_anon_rmap(page, vma, address);
672
	else
N
Nick Piggin 已提交
673
		__page_check_anon_rmap(page, vma, address);
L
Linus Torvalds 已提交
674 675
}

R
Randy Dunlap 已提交
676
/**
N
Nick Piggin 已提交
677 678 679 680 681 682 683
 * 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
684
 * Page does not have to be locked.
N
Nick Piggin 已提交
685 686 687 688
 */
void page_add_new_anon_rmap(struct page *page,
	struct vm_area_struct *vma, unsigned long address)
{
689
	VM_BUG_ON(address < vma->vm_start || address >= vma->vm_end);
690 691
	SetPageSwapBacked(page);
	atomic_set(&page->_mapcount, 0); /* increment count (starts at -1) */
N
Nick Piggin 已提交
692
	__page_set_anon_rmap(page, vma, address);
693
	if (page_evictable(page, vma))
694
		lru_cache_add_lru(page, LRU_ACTIVE_ANON);
695 696
	else
		add_page_to_unevictable_list(page);
N
Nick Piggin 已提交
697 698
}

L
Linus Torvalds 已提交
699 700 701 702
/**
 * page_add_file_rmap - add pte mapping to a file page
 * @page: the page to add the mapping to
 *
703
 * The caller needs to hold the pte lock.
L
Linus Torvalds 已提交
704 705 706
 */
void page_add_file_rmap(struct page *page)
{
707
	if (atomic_inc_and_test(&page->_mapcount)) {
708
		__inc_zone_page_state(page, NR_FILE_MAPPED);
709 710
		mem_cgroup_update_mapped_file_stat(page, 1);
	}
L
Linus Torvalds 已提交
711 712
}

N
Nick Piggin 已提交
713 714 715 716
#ifdef CONFIG_DEBUG_VM
/**
 * page_dup_rmap - duplicate pte mapping to a page
 * @page:	the page to add the mapping to
R
Randy Dunlap 已提交
717 718
 * @vma:	the vm area being duplicated
 * @address:	the user virtual address mapped
N
Nick Piggin 已提交
719 720 721 722 723 724 725 726 727 728 729 730 731 732 733
 *
 * For copy_page_range only: minimal extract from page_add_file_rmap /
 * page_add_anon_rmap, avoiding unnecessary tests (already checked) so it's
 * quicker.
 *
 * The caller needs to hold the pte lock.
 */
void page_dup_rmap(struct page *page, struct vm_area_struct *vma, unsigned long address)
{
	if (PageAnon(page))
		__page_check_anon_rmap(page, vma, address);
	atomic_inc(&page->_mapcount);
}
#endif

L
Linus Torvalds 已提交
734 735 736 737
/**
 * page_remove_rmap - take down pte mapping from a page
 * @page: page to remove mapping from
 *
738
 * The caller needs to hold the pte lock.
L
Linus Torvalds 已提交
739
 */
740
void page_remove_rmap(struct page *page)
L
Linus Torvalds 已提交
741 742 743
{
	if (atomic_add_negative(-1, &page->_mapcount)) {
		/*
744 745 746 747 748
		 * 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.
L
Linus Torvalds 已提交
749
		 */
750 751
		if ((!PageAnon(page) || PageSwapCache(page)) &&
		    page_test_dirty(page)) {
752
			page_clear_dirty(page);
L
Linus Torvalds 已提交
753
			set_page_dirty(page);
754
		}
755 756
		if (PageAnon(page))
			mem_cgroup_uncharge_page(page);
757
		__dec_zone_page_state(page,
758
			PageAnon(page) ? NR_ANON_PAGES : NR_FILE_MAPPED);
759
		mem_cgroup_update_mapped_file_stat(page, -1);
760 761 762 763 764 765 766 767 768
		/*
		 * 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
769 770 771 772 773 774 775
	}
}

/*
 * Subfunctions of try_to_unmap: try_to_unmap_one called
 * repeatedly from either try_to_unmap_anon or try_to_unmap_file.
 */
776
static int try_to_unmap_one(struct page *page, struct vm_area_struct *vma,
777
				enum ttu_flags flags)
L
Linus Torvalds 已提交
778 779 780 781 782
{
	struct mm_struct *mm = vma->vm_mm;
	unsigned long address;
	pte_t *pte;
	pte_t pteval;
H
Hugh Dickins 已提交
783
	spinlock_t *ptl;
L
Linus Torvalds 已提交
784 785 786 787 788 789
	int ret = SWAP_AGAIN;

	address = vma_address(page, vma);
	if (address == -EFAULT)
		goto out;

N
Nick Piggin 已提交
790
	pte = page_check_address(page, mm, address, &ptl, 0);
H
Hugh Dickins 已提交
791
	if (!pte)
N
Nikita Danilov 已提交
792
		goto out;
L
Linus Torvalds 已提交
793 794 795 796 797 798

	/*
	 * 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.
	 */
799
	if (!(flags & TTU_IGNORE_MLOCK)) {
N
Nick Piggin 已提交
800 801 802 803
		if (vma->vm_flags & VM_LOCKED) {
			ret = SWAP_MLOCK;
			goto out_unmap;
		}
804 805
	}
	if (!(flags & TTU_IGNORE_ACCESS)) {
N
Nick Piggin 已提交
806 807 808 809 810
		if (ptep_clear_flush_young_notify(vma, address, pte)) {
			ret = SWAP_FAIL;
			goto out_unmap;
		}
  	}
L
Linus Torvalds 已提交
811 812 813

	/* Nuke the page table entry. */
	flush_cache_page(vma, address, page_to_pfn(page));
A
Andrea Arcangeli 已提交
814
	pteval = ptep_clear_flush_notify(vma, address, pte);
L
Linus Torvalds 已提交
815 816 817 818 819

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

820 821 822
	/* Update high watermark before we lower rss */
	update_hiwater_rss(mm);

823 824 825 826 827 828 829 830
	if (PageHWPoison(page) && !(flags & TTU_IGNORE_HWPOISON)) {
		if (PageAnon(page))
			dec_mm_counter(mm, anon_rss);
		else
			dec_mm_counter(mm, file_rss);
		set_pte_at(mm, address, pte,
				swp_entry_to_pte(make_hwpoison_entry(page)));
	} else if (PageAnon(page)) {
H
Hugh Dickins 已提交
831
		swp_entry_t entry = { .val = page_private(page) };
832 833 834 835 836 837 838 839 840 841 842 843 844

		if (PageSwapCache(page)) {
			/*
			 * Store the swap location in the pte.
			 * See handle_pte_fault() ...
			 */
			swap_duplicate(entry);
			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);
			}
845
			dec_mm_counter(mm, anon_rss);
846
		} else if (PAGE_MIGRATION) {
847 848 849 850 851
			/*
			 * 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.
			 */
852
			BUG_ON(TTU_ACTION(flags) != TTU_MIGRATION);
853
			entry = make_migration_entry(page, pte_write(pteval));
L
Linus Torvalds 已提交
854 855 856
		}
		set_pte_at(mm, address, pte, swp_entry_to_pte(entry));
		BUG_ON(pte_file(*pte));
857
	} else if (PAGE_MIGRATION && (TTU_ACTION(flags) == TTU_MIGRATION)) {
858 859 860 861 862
		/* 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
863
		dec_mm_counter(mm, file_rss);
L
Linus Torvalds 已提交
864

865

866
	page_remove_rmap(page);
L
Linus Torvalds 已提交
867 868 869
	page_cache_release(page);

out_unmap:
H
Hugh Dickins 已提交
870
	pte_unmap_unlock(pte, ptl);
L
Linus Torvalds 已提交
871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892
out:
	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
893 894 895 896 897
 *
 * 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
898 899 900 901
 */
#define CLUSTER_SIZE	min(32*PAGE_SIZE, PMD_SIZE)
#define CLUSTER_MASK	(~(CLUSTER_SIZE - 1))

N
Nick Piggin 已提交
902 903
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
904 905 906 907 908
{
	struct mm_struct *mm = vma->vm_mm;
	pgd_t *pgd;
	pud_t *pud;
	pmd_t *pmd;
H
Hugh Dickins 已提交
909
	pte_t *pte;
L
Linus Torvalds 已提交
910
	pte_t pteval;
H
Hugh Dickins 已提交
911
	spinlock_t *ptl;
L
Linus Torvalds 已提交
912 913 914
	struct page *page;
	unsigned long address;
	unsigned long end;
N
Nick Piggin 已提交
915 916
	int ret = SWAP_AGAIN;
	int locked_vma = 0;
L
Linus Torvalds 已提交
917 918 919 920 921 922 923 924 925 926

	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
927
		return ret;
L
Linus Torvalds 已提交
928 929 930

	pud = pud_offset(pgd, address);
	if (!pud_present(*pud))
N
Nick Piggin 已提交
931
		return ret;
L
Linus Torvalds 已提交
932 933 934

	pmd = pmd_offset(pud, address);
	if (!pmd_present(*pmd))
N
Nick Piggin 已提交
935 936 937 938 939 940 941 942 943 944 945 946
		return ret;

	/*
	 * MLOCK_PAGES => feature is configured.
	 * if we can acquire the mmap_sem for read, and vma is VM_LOCKED,
	 * keep the sem while scanning the cluster for mlocking pages.
	 */
	if (MLOCK_PAGES && down_read_trylock(&vma->vm_mm->mmap_sem)) {
		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
947 948

	pte = pte_offset_map_lock(mm, pmd, address, &ptl);
L
Linus Torvalds 已提交
949

950 951 952
	/* Update high watermark before we lower rss */
	update_hiwater_rss(mm);

H
Hugh Dickins 已提交
953
	for (; address < end; pte++, address += PAGE_SIZE) {
L
Linus Torvalds 已提交
954 955
		if (!pte_present(*pte))
			continue;
956 957
		page = vm_normal_page(vma, address, *pte);
		BUG_ON(!page || PageAnon(page));
L
Linus Torvalds 已提交
958

N
Nick Piggin 已提交
959 960 961 962 963 964 965
		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
966
		if (ptep_clear_flush_young_notify(vma, address, pte))
L
Linus Torvalds 已提交
967 968 969
			continue;

		/* Nuke the page table entry. */
970
		flush_cache_page(vma, address, pte_pfn(*pte));
A
Andrea Arcangeli 已提交
971
		pteval = ptep_clear_flush_notify(vma, address, pte);
L
Linus Torvalds 已提交
972 973 974 975 976 977 978 979 980

		/* 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);

981
		page_remove_rmap(page);
L
Linus Torvalds 已提交
982
		page_cache_release(page);
983
		dec_mm_counter(mm, file_rss);
L
Linus Torvalds 已提交
984 985
		(*mapcount)--;
	}
H
Hugh Dickins 已提交
986
	pte_unmap_unlock(pte - 1, ptl);
N
Nick Piggin 已提交
987 988 989
	if (locked_vma)
		up_read(&vma->vm_mm->mmap_sem);
	return ret;
L
Linus Torvalds 已提交
990 991
}

N
Nick Piggin 已提交
992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025
/*
 * common handling for pages mapped in VM_LOCKED vmas
 */
static int try_to_mlock_page(struct page *page, struct vm_area_struct *vma)
{
	int mlocked = 0;

	if (down_read_trylock(&vma->vm_mm->mmap_sem)) {
		if (vma->vm_flags & VM_LOCKED) {
			mlock_vma_page(page);
			mlocked++;	/* really mlocked the page */
		}
		up_read(&vma->vm_mm->mmap_sem);
	}
	return mlocked;
}

/**
 * try_to_unmap_anon - unmap or unlock anonymous page using the object-based
 * rmap method
 * @page: the page to unmap/unlock
 * @unlock:  request for unlock rather than unmap [unlikely]
 * @migration:  unmapping for migration - ignored if @unlock
 *
 * 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.
 */
1026
static int try_to_unmap_anon(struct page *page, enum ttu_flags flags)
L
Linus Torvalds 已提交
1027 1028 1029
{
	struct anon_vma *anon_vma;
	struct vm_area_struct *vma;
N
Nick Piggin 已提交
1030
	unsigned int mlocked = 0;
L
Linus Torvalds 已提交
1031
	int ret = SWAP_AGAIN;
1032
	int unlock = TTU_ACTION(flags) == TTU_MUNLOCK;
L
Linus Torvalds 已提交
1033

N
Nick Piggin 已提交
1034 1035 1036
	if (MLOCK_PAGES && unlikely(unlock))
		ret = SWAP_SUCCESS;	/* default for try_to_munlock() */

L
Linus Torvalds 已提交
1037 1038 1039 1040 1041
	anon_vma = page_lock_anon_vma(page);
	if (!anon_vma)
		return ret;

	list_for_each_entry(vma, &anon_vma->head, anon_vma_node) {
N
Nick Piggin 已提交
1042 1043 1044 1045 1046 1047
		if (MLOCK_PAGES && unlikely(unlock)) {
			if (!((vma->vm_flags & VM_LOCKED) &&
			      page_mapped_in_vma(page, vma)))
				continue;  /* must visit all unlocked vmas */
			ret = SWAP_MLOCK;  /* saw at least one mlocked vma */
		} else {
1048
			ret = try_to_unmap_one(page, vma, flags);
N
Nick Piggin 已提交
1049 1050 1051 1052 1053 1054 1055 1056
			if (ret == SWAP_FAIL || !page_mapped(page))
				break;
		}
		if (ret == SWAP_MLOCK) {
			mlocked = try_to_mlock_page(page, vma);
			if (mlocked)
				break;	/* stop if actually mlocked page */
		}
L
Linus Torvalds 已提交
1057
	}
1058 1059

	page_unlock_anon_vma(anon_vma);
N
Nick Piggin 已提交
1060 1061 1062 1063 1064 1065

	if (mlocked)
		ret = SWAP_MLOCK;	/* actually mlocked the page */
	else if (ret == SWAP_MLOCK)
		ret = SWAP_AGAIN;	/* saw VM_LOCKED vma */

L
Linus Torvalds 已提交
1066 1067 1068 1069
	return ret;
}

/**
N
Nick Piggin 已提交
1070 1071
 * try_to_unmap_file - unmap/unlock file page using the object-based rmap method
 * @page: the page to unmap/unlock
1072
 * @flags: action and flags
L
Linus Torvalds 已提交
1073 1074 1075 1076
 *
 * 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
1077 1078 1079 1080 1081 1082
 * 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
1083
 */
1084
static int try_to_unmap_file(struct page *page, enum ttu_flags flags)
L
Linus Torvalds 已提交
1085 1086 1087 1088 1089 1090 1091 1092 1093 1094
{
	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;
N
Nick Piggin 已提交
1095
	unsigned int mlocked = 0;
1096
	int unlock = TTU_ACTION(flags) == TTU_MUNLOCK;
N
Nick Piggin 已提交
1097 1098 1099

	if (MLOCK_PAGES && unlikely(unlock))
		ret = SWAP_SUCCESS;	/* default for try_to_munlock() */
L
Linus Torvalds 已提交
1100 1101 1102

	spin_lock(&mapping->i_mmap_lock);
	vma_prio_tree_foreach(vma, &iter, &mapping->i_mmap, pgoff, pgoff) {
N
Nick Piggin 已提交
1103
		if (MLOCK_PAGES && unlikely(unlock)) {
1104 1105
			if (!((vma->vm_flags & VM_LOCKED) &&
						page_mapped_in_vma(page, vma)))
N
Nick Piggin 已提交
1106 1107 1108
				continue;	/* must visit all vmas */
			ret = SWAP_MLOCK;
		} else {
1109
			ret = try_to_unmap_one(page, vma, flags);
N
Nick Piggin 已提交
1110 1111 1112 1113 1114 1115 1116 1117
			if (ret == SWAP_FAIL || !page_mapped(page))
				goto out;
		}
		if (ret == SWAP_MLOCK) {
			mlocked = try_to_mlock_page(page, vma);
			if (mlocked)
				break;  /* stop if actually mlocked page */
		}
L
Linus Torvalds 已提交
1118 1119
	}

N
Nick Piggin 已提交
1120 1121 1122
	if (mlocked)
		goto out;

L
Linus Torvalds 已提交
1123 1124 1125 1126 1127
	if (list_empty(&mapping->i_mmap_nonlinear))
		goto out;

	list_for_each_entry(vma, &mapping->i_mmap_nonlinear,
						shared.vm_set.list) {
N
Nick Piggin 已提交
1128 1129 1130 1131 1132 1133
		if (MLOCK_PAGES && unlikely(unlock)) {
			if (!(vma->vm_flags & VM_LOCKED))
				continue;	/* must visit all vmas */
			ret = SWAP_MLOCK;	/* leave mlocked == 0 */
			goto out;		/* no need to look further */
		}
1134 1135
		if (!MLOCK_PAGES && !(flags & TTU_IGNORE_MLOCK) &&
			(vma->vm_flags & VM_LOCKED))
L
Linus Torvalds 已提交
1136 1137 1138 1139 1140 1141 1142 1143 1144
			continue;
		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
1145
	if (max_nl_size == 0) {	/* all nonlinears locked or reserved ? */
L
Linus Torvalds 已提交
1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168
		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) {
1169
			if (!MLOCK_PAGES && !(flags & TTU_IGNORE_MLOCK) &&
N
Nick Piggin 已提交
1170
			    (vma->vm_flags & VM_LOCKED))
L
Linus Torvalds 已提交
1171 1172
				continue;
			cursor = (unsigned long) vma->vm_private_data;
H
Hugh Dickins 已提交
1173
			while ( cursor < max_nl_cursor &&
L
Linus Torvalds 已提交
1174
				cursor < vma->vm_end - vma->vm_start) {
N
Nick Piggin 已提交
1175 1176 1177 1178
				ret = try_to_unmap_cluster(cursor, &mapcount,
								vma, page);
				if (ret == SWAP_MLOCK)
					mlocked = 2;	/* to return below */
L
Linus Torvalds 已提交
1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194
				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.
	 */
1195 1196
	list_for_each_entry(vma, &mapping->i_mmap_nonlinear, shared.vm_set.list)
		vma->vm_private_data = NULL;
L
Linus Torvalds 已提交
1197 1198
out:
	spin_unlock(&mapping->i_mmap_lock);
N
Nick Piggin 已提交
1199 1200 1201 1202
	if (mlocked)
		ret = SWAP_MLOCK;	/* actually mlocked the page */
	else if (ret == SWAP_MLOCK)
		ret = SWAP_AGAIN;	/* saw VM_LOCKED vma */
L
Linus Torvalds 已提交
1203 1204 1205 1206 1207 1208
	return ret;
}

/**
 * try_to_unmap - try to remove all page table mappings to a page
 * @page: the page to get unmapped
1209
 * @flags: action and flags
L
Linus Torvalds 已提交
1210 1211 1212 1213 1214 1215 1216 1217
 *
 * 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
1218
 * SWAP_MLOCK	- page is mlocked.
L
Linus Torvalds 已提交
1219
 */
1220
int try_to_unmap(struct page *page, enum ttu_flags flags)
L
Linus Torvalds 已提交
1221 1222 1223 1224 1225 1226
{
	int ret;

	BUG_ON(!PageLocked(page));

	if (PageAnon(page))
1227
		ret = try_to_unmap_anon(page, flags);
L
Linus Torvalds 已提交
1228
	else
1229
		ret = try_to_unmap_file(page, flags);
N
Nick Piggin 已提交
1230
	if (ret != SWAP_MLOCK && !page_mapped(page))
L
Linus Torvalds 已提交
1231 1232 1233
		ret = SWAP_SUCCESS;
	return ret;
}
N
Nikita Danilov 已提交
1234

N
Nick Piggin 已提交
1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253
/**
 * 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:
 *
 * SWAP_SUCCESS	- no vma's holding page mlocked.
 * SWAP_AGAIN	- page mapped in mlocked vma -- couldn't acquire mmap sem
 * SWAP_MLOCK	- page is now mlocked.
 */
int try_to_munlock(struct page *page)
{
	VM_BUG_ON(!PageLocked(page) || PageLRU(page));

	if (PageAnon(page))
1254
		return try_to_unmap_anon(page, TTU_MUNLOCK);
N
Nick Piggin 已提交
1255
	else
1256
		return try_to_unmap_file(page, TTU_MUNLOCK);
N
Nick Piggin 已提交
1257
}
1258