rmap.c 27.3 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 17 18 19 20 21 22
/*
 * 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
 * Contributions by Hugh Dickins <hugh@veritas.com> 2003, 2004
 */

/*
 * 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>
N
Nick Piggin 已提交
50
#include <linux/kallsyms.h>
51
#include <linux/memcontrol.h>
A
Andrea Arcangeli 已提交
52
#include <linux/mmu_notifier.h>
L
Linus Torvalds 已提交
53 54 55

#include <asm/tlbflush.h>

56
struct kmem_cache *anon_vma_cachep;
L
Linus Torvalds 已提交
57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84

/* This must be called under the mmap_sem. */
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;
		struct anon_vma *allocated, *locked;

		anon_vma = find_mergeable_anon_vma(vma);
		if (anon_vma) {
			allocated = NULL;
			locked = anon_vma;
			spin_lock(&locked->lock);
		} else {
			anon_vma = anon_vma_alloc();
			if (unlikely(!anon_vma))
				return -ENOMEM;
			allocated = anon_vma;
			locked = NULL;
		}

		/* page_table_lock to protect against threads */
		spin_lock(&mm->page_table_lock);
		if (likely(!vma->anon_vma)) {
			vma->anon_vma = anon_vma;
85
			list_add_tail(&vma->anon_vma_node, &anon_vma->head);
L
Linus Torvalds 已提交
86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107
			allocated = NULL;
		}
		spin_unlock(&mm->page_table_lock);

		if (locked)
			spin_unlock(&locked->lock);
		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;

108
	if (anon_vma)
109
		list_add_tail(&vma->anon_vma_node, &anon_vma->head);
L
Linus Torvalds 已提交
110 111 112 113 114 115 116 117
}

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);
118
		list_add_tail(&vma->anon_vma_node, &anon_vma->head);
L
Linus Torvalds 已提交
119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141
		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);
}

142
static void anon_vma_ctor(void *data)
L
Linus Torvalds 已提交
143
{
C
Christoph Lameter 已提交
144
	struct anon_vma *anon_vma = data;
L
Linus Torvalds 已提交
145

C
Christoph Lameter 已提交
146 147
	spin_lock_init(&anon_vma->lock);
	INIT_LIST_HEAD(&anon_vma->head);
L
Linus Torvalds 已提交
148 149 150 151 152
}

void __init anon_vma_init(void)
{
	anon_vma_cachep = kmem_cache_create("anon_vma", sizeof(struct anon_vma),
153
			0, SLAB_DESTROY_BY_RCU|SLAB_PANIC, anon_vma_ctor);
L
Linus Torvalds 已提交
154 155 156 157 158 159 160 161
}

/*
 * 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.
 */
static struct anon_vma *page_lock_anon_vma(struct page *page)
{
162
	struct anon_vma *anon_vma;
L
Linus Torvalds 已提交
163 164 165 166 167 168 169 170 171 172 173
	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);
174
	return anon_vma;
L
Linus Torvalds 已提交
175 176
out:
	rcu_read_unlock();
177 178 179 180 181 182 183
	return NULL;
}

static void page_unlock_anon_vma(struct anon_vma *anon_vma)
{
	spin_unlock(&anon_vma->lock);
	rcu_read_unlock();
L
Linus Torvalds 已提交
184 185 186
}

/*
187 188 189
 * 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
190 191 192 193 194 195 196 197 198
 */
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)) {
199
		/* page should be within @vma mapping range */
L
Linus Torvalds 已提交
200 201 202 203 204 205 206
		return -EFAULT;
	}
	return address;
}

/*
 * At what user virtual address is page expected in vma? checking that the
H
Hugh Dickins 已提交
207
 * page matches the vma: currently only used on anon pages, by unuse_vma;
L
Linus Torvalds 已提交
208 209 210 211 212 213 214 215
 */
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
216 217
		if (!vma->vm_file ||
		    vma->vm_file->f_mapping != page->mapping)
L
Linus Torvalds 已提交
218 219 220 221 222 223
			return -EFAULT;
	} else
		return -EFAULT;
	return vma_address(page, vma);
}

N
Nikita Danilov 已提交
224 225 226
/*
 * Check that @page is mapped at @address into @mm.
 *
227
 * On success returns with pte mapped and locked.
N
Nikita Danilov 已提交
228
 */
229
pte_t *page_check_address(struct page *page, struct mm_struct *mm,
H
Hugh Dickins 已提交
230
			  unsigned long address, spinlock_t **ptlp)
N
Nikita Danilov 已提交
231 232 233 234 235
{
	pgd_t *pgd;
	pud_t *pud;
	pmd_t *pmd;
	pte_t *pte;
H
Hugh Dickins 已提交
236
	spinlock_t *ptl;
N
Nikita Danilov 已提交
237 238

	pgd = pgd_offset(mm, address);
H
Hugh Dickins 已提交
239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256
	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 */
	if (!pte_present(*pte)) {
		pte_unmap(pte);
		return NULL;
	}

H
Hugh Dickins 已提交
257
	ptl = pte_lockptr(mm, pmd);
H
Hugh Dickins 已提交
258 259 260 261
	spin_lock(ptl);
	if (pte_present(*pte) && page_to_pfn(page) == pte_pfn(*pte)) {
		*ptlp = ptl;
		return pte;
N
Nikita Danilov 已提交
262
	}
H
Hugh Dickins 已提交
263 264
	pte_unmap_unlock(pte, ptl);
	return NULL;
N
Nikita Danilov 已提交
265 266
}

L
Linus Torvalds 已提交
267 268 269 270 271
/*
 * 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,
272
	struct vm_area_struct *vma, unsigned int *mapcount)
L
Linus Torvalds 已提交
273 274 275 276
{
	struct mm_struct *mm = vma->vm_mm;
	unsigned long address;
	pte_t *pte;
H
Hugh Dickins 已提交
277
	spinlock_t *ptl;
L
Linus Torvalds 已提交
278 279 280 281 282 283
	int referenced = 0;

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

H
Hugh Dickins 已提交
284 285 286
	pte = page_check_address(page, mm, address, &ptl);
	if (!pte)
		goto out;
L
Linus Torvalds 已提交
287

288 289 290
	if (vma->vm_flags & VM_LOCKED) {
		referenced++;
		*mapcount = 1;	/* break early from loop */
A
Andrea Arcangeli 已提交
291
	} else if (ptep_clear_flush_young_notify(vma, address, pte))
H
Hugh Dickins 已提交
292
		referenced++;
L
Linus Torvalds 已提交
293

H
Hugh Dickins 已提交
294 295
	/* Pretend the page is referenced if the task has the
	   swap token and is in the middle of a page fault. */
296
	if (mm != current->mm && has_swap_token(mm) &&
H
Hugh Dickins 已提交
297 298 299 300 301
			rwsem_is_locked(&mm->mmap_sem))
		referenced++;

	(*mapcount)--;
	pte_unmap_unlock(pte, ptl);
L
Linus Torvalds 已提交
302 303 304 305
out:
	return referenced;
}

306 307
static int page_referenced_anon(struct page *page,
				struct mem_cgroup *mem_cont)
L
Linus Torvalds 已提交
308 309 310 311 312 313 314 315 316 317 318 319
{
	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) {
320 321 322 323 324
		/*
		 * If we are reclaiming on behalf of a cgroup, skip
		 * counting on behalf of references from different
		 * cgroups
		 */
325
		if (mem_cont && !mm_match_cgroup(vma->vm_mm, mem_cont))
326
			continue;
327
		referenced += page_referenced_one(page, vma, &mapcount);
L
Linus Torvalds 已提交
328 329 330
		if (!mapcount)
			break;
	}
331 332

	page_unlock_anon_vma(anon_vma);
L
Linus Torvalds 已提交
333 334 335 336 337 338
	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
339
 * @mem_cont: target memory controller
L
Linus Torvalds 已提交
340 341 342 343 344 345 346 347
 *
 * 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.
 */
348 349
static int page_referenced_file(struct page *page,
				struct mem_cgroup *mem_cont)
L
Linus Torvalds 已提交
350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381
{
	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) {
382 383 384 385 386
		/*
		 * If we are reclaiming on behalf of a cgroup, skip
		 * counting on behalf of references from different
		 * cgroups
		 */
387
		if (mem_cont && !mm_match_cgroup(vma->vm_mm, mem_cont))
388
			continue;
L
Linus Torvalds 已提交
389 390 391 392 393
		if ((vma->vm_flags & (VM_LOCKED|VM_MAYSHARE))
				  == (VM_LOCKED|VM_MAYSHARE)) {
			referenced++;
			break;
		}
394
		referenced += page_referenced_one(page, vma, &mapcount);
L
Linus Torvalds 已提交
395 396 397 398 399 400 401 402 403 404 405 406
		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
407
 * @mem_cont: target memory controller
L
Linus Torvalds 已提交
408 409 410 411
 *
 * Quick test_and_clear_referenced for all mappings to a page,
 * returns the number of ptes which referenced the page.
 */
412 413
int page_referenced(struct page *page, int is_locked,
			struct mem_cgroup *mem_cont)
L
Linus Torvalds 已提交
414 415 416 417 418 419 420 421
{
	int referenced = 0;

	if (TestClearPageReferenced(page))
		referenced++;

	if (page_mapped(page) && page->mapping) {
		if (PageAnon(page))
422
			referenced += page_referenced_anon(page, mem_cont);
L
Linus Torvalds 已提交
423
		else if (is_locked)
424
			referenced += page_referenced_file(page, mem_cont);
N
Nick Piggin 已提交
425
		else if (!trylock_page(page))
L
Linus Torvalds 已提交
426 427 428
			referenced++;
		else {
			if (page->mapping)
429 430
				referenced +=
					page_referenced_file(page, mem_cont);
L
Linus Torvalds 已提交
431 432 433
			unlock_page(page);
		}
	}
434 435 436 437

	if (page_test_and_clear_young(page))
		referenced++;

L
Linus Torvalds 已提交
438 439 440
	return referenced;
}

441 442 443 444
static int page_mkclean_one(struct page *page, struct vm_area_struct *vma)
{
	struct mm_struct *mm = vma->vm_mm;
	unsigned long address;
445
	pte_t *pte;
446 447 448 449 450 451 452 453 454 455 456
	spinlock_t *ptl;
	int ret = 0;

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

	pte = page_check_address(page, mm, address, &ptl);
	if (!pte)
		goto out;

457 458
	if (pte_dirty(*pte) || pte_write(*pte)) {
		pte_t entry;
459

460
		flush_cache_page(vma, address, pte_pfn(*pte));
A
Andrea Arcangeli 已提交
461
		entry = ptep_clear_flush_notify(vma, address, pte);
462 463
		entry = pte_wrprotect(entry);
		entry = pte_mkclean(entry);
464
		set_pte_at(mm, address, pte, entry);
465 466
		ret = 1;
	}
467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498

	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);
499
		if (mapping) {
500
			ret = page_mkclean_file(mapping, page);
501 502 503 504
			if (page_test_dirty(page)) {
				page_clear_dirty(page);
				ret = 1;
			}
505
		}
506 507 508 509
	}

	return ret;
}
J
Jaya Kumar 已提交
510
EXPORT_SYMBOL_GPL(page_mkclean);
511

N
Nick Piggin 已提交
512
/**
R
Randy Dunlap 已提交
513
 * __page_set_anon_rmap - setup new anonymous rmap
N
Nick Piggin 已提交
514 515 516 517 518 519 520 521 522 523 524 525 526 527 528
 * @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
529 530 531 532
	/*
	 * nr_mapped state can be updated without turning off
	 * interrupts because it is not modified via interrupt.
	 */
533
	__inc_zone_page_state(page, NR_ANON_PAGES);
N
Nick Piggin 已提交
534 535
}

N
Nick Piggin 已提交
536
/**
R
Randy Dunlap 已提交
537
 * __page_check_anon_rmap - sanity check anonymous rmap addition
N
Nick Piggin 已提交
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
 * @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
565 566 567 568 569 570
/**
 * 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
571
 * The caller needs to hold the pte lock and the page must be locked.
L
Linus Torvalds 已提交
572 573 574 575
 */
void page_add_anon_rmap(struct page *page,
	struct vm_area_struct *vma, unsigned long address)
{
N
Nick Piggin 已提交
576 577
	VM_BUG_ON(!PageLocked(page));
	VM_BUG_ON(address < vma->vm_start || address >= vma->vm_end);
N
Nick Piggin 已提交
578 579
	if (atomic_inc_and_test(&page->_mapcount))
		__page_set_anon_rmap(page, vma, address);
580
	else
N
Nick Piggin 已提交
581
		__page_check_anon_rmap(page, vma, address);
L
Linus Torvalds 已提交
582 583
}

R
Randy Dunlap 已提交
584
/**
N
Nick Piggin 已提交
585 586 587 588 589 590 591
 * 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
592
 * Page does not have to be locked.
N
Nick Piggin 已提交
593 594 595 596
 */
void page_add_new_anon_rmap(struct page *page,
	struct vm_area_struct *vma, unsigned long address)
{
N
Nick Piggin 已提交
597
	BUG_ON(address < vma->vm_start || address >= vma->vm_end);
N
Nick Piggin 已提交
598 599 600 601
	atomic_set(&page->_mapcount, 0); /* elevate count by 1 (starts at -1) */
	__page_set_anon_rmap(page, vma, address);
}

L
Linus Torvalds 已提交
602 603 604 605
/**
 * page_add_file_rmap - add pte mapping to a file page
 * @page: the page to add the mapping to
 *
606
 * The caller needs to hold the pte lock.
L
Linus Torvalds 已提交
607 608 609 610
 */
void page_add_file_rmap(struct page *page)
{
	if (atomic_inc_and_test(&page->_mapcount))
611
		__inc_zone_page_state(page, NR_FILE_MAPPED);
L
Linus Torvalds 已提交
612 613
}

N
Nick Piggin 已提交
614 615 616 617
#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
618 619
 * @vma:	the vm area being duplicated
 * @address:	the user virtual address mapped
N
Nick Piggin 已提交
620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635
 *
 * 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)
{
	BUG_ON(page_mapcount(page) == 0);
	if (PageAnon(page))
		__page_check_anon_rmap(page, vma, address);
	atomic_inc(&page->_mapcount);
}
#endif

L
Linus Torvalds 已提交
636 637 638
/**
 * page_remove_rmap - take down pte mapping from a page
 * @page: page to remove mapping from
R
Randy Dunlap 已提交
639
 * @vma: the vm area in which the mapping is removed
L
Linus Torvalds 已提交
640
 *
641
 * The caller needs to hold the pte lock.
L
Linus Torvalds 已提交
642
 */
N
Nick Piggin 已提交
643
void page_remove_rmap(struct page *page, struct vm_area_struct *vma)
L
Linus Torvalds 已提交
644 645
{
	if (atomic_add_negative(-1, &page->_mapcount)) {
N
Nick Piggin 已提交
646
		if (unlikely(page_mapcount(page) < 0)) {
647
			printk (KERN_EMERG "Eeek! page_mapcount(page) went negative! (%d)\n", page_mapcount(page));
N
Nick Piggin 已提交
648
			printk (KERN_EMERG "  page pfn = %lx\n", page_to_pfn(page));
649 650 651
			printk (KERN_EMERG "  page->flags = %lx\n", page->flags);
			printk (KERN_EMERG "  page->count = %x\n", page_count(page));
			printk (KERN_EMERG "  page->mapping = %p\n", page->mapping);
N
Nick Piggin 已提交
652
			print_symbol (KERN_EMERG "  vma->vm_ops = %s\n", (unsigned long)vma->vm_ops);
653 654 655
			if (vma->vm_ops) {
				print_symbol (KERN_EMERG "  vma->vm_ops->fault = %s\n", (unsigned long)vma->vm_ops->fault);
			}
N
Nick Piggin 已提交
656 657
			if (vma->vm_file && vma->vm_file->f_op)
				print_symbol (KERN_EMERG "  vma->vm_file->f_op->mmap = %s\n", (unsigned long)vma->vm_file->f_op->mmap);
658
			BUG();
659
		}
660

L
Linus Torvalds 已提交
661 662 663 664 665 666 667 668 669
		/*
		 * 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.
		 */
670 671
		if ((!PageAnon(page) || PageSwapCache(page)) &&
		    page_test_dirty(page)) {
672
			page_clear_dirty(page);
L
Linus Torvalds 已提交
673
			set_page_dirty(page);
674
		}
675 676
		mem_cgroup_uncharge_page(page);

677 678
		__dec_zone_page_state(page,
				PageAnon(page) ? NR_ANON_PAGES : NR_FILE_MAPPED);
L
Linus Torvalds 已提交
679 680 681 682 683 684 685
	}
}

/*
 * Subfunctions of try_to_unmap: try_to_unmap_one called
 * repeatedly from either try_to_unmap_anon or try_to_unmap_file.
 */
686
static int try_to_unmap_one(struct page *page, struct vm_area_struct *vma,
687
				int migration)
L
Linus Torvalds 已提交
688 689 690 691 692
{
	struct mm_struct *mm = vma->vm_mm;
	unsigned long address;
	pte_t *pte;
	pte_t pteval;
H
Hugh Dickins 已提交
693
	spinlock_t *ptl;
L
Linus Torvalds 已提交
694 695 696 697 698 699
	int ret = SWAP_AGAIN;

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

H
Hugh Dickins 已提交
700 701
	pte = page_check_address(page, mm, address, &ptl);
	if (!pte)
N
Nikita Danilov 已提交
702
		goto out;
L
Linus Torvalds 已提交
703 704 705 706 707 708

	/*
	 * 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.
	 */
709
	if (!migration && ((vma->vm_flags & VM_LOCKED) ||
A
Andrea Arcangeli 已提交
710
			(ptep_clear_flush_young_notify(vma, address, pte)))) {
L
Linus Torvalds 已提交
711 712 713 714 715 716
		ret = SWAP_FAIL;
		goto out_unmap;
	}

	/* Nuke the page table entry. */
	flush_cache_page(vma, address, page_to_pfn(page));
A
Andrea Arcangeli 已提交
717
	pteval = ptep_clear_flush_notify(vma, address, pte);
L
Linus Torvalds 已提交
718 719 720 721 722

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

723 724 725
	/* Update high watermark before we lower rss */
	update_hiwater_rss(mm);

L
Linus Torvalds 已提交
726
	if (PageAnon(page)) {
H
Hugh Dickins 已提交
727
		swp_entry_t entry = { .val = page_private(page) };
728 729 730 731 732 733 734 735 736 737 738 739 740

		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);
			}
741
			dec_mm_counter(mm, anon_rss);
742
#ifdef CONFIG_MIGRATION
743 744 745 746 747 748 749 750
		} else {
			/*
			 * 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.
			 */
			BUG_ON(!migration);
			entry = make_migration_entry(page, pte_write(pteval));
751
#endif
L
Linus Torvalds 已提交
752 753 754
		}
		set_pte_at(mm, address, pte, swp_entry_to_pte(entry));
		BUG_ON(pte_file(*pte));
755
	} else
756 757 758 759 760 761 762 763
#ifdef CONFIG_MIGRATION
	if (migration) {
		/* 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
#endif
764
		dec_mm_counter(mm, file_rss);
L
Linus Torvalds 已提交
765

766

N
Nick Piggin 已提交
767
	page_remove_rmap(page, vma);
L
Linus Torvalds 已提交
768 769 770
	page_cache_release(page);

out_unmap:
H
Hugh Dickins 已提交
771
	pte_unmap_unlock(pte, ptl);
L
Linus Torvalds 已提交
772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804
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.
 */
#define CLUSTER_SIZE	min(32*PAGE_SIZE, PMD_SIZE)
#define CLUSTER_MASK	(~(CLUSTER_SIZE - 1))

static void try_to_unmap_cluster(unsigned long cursor,
	unsigned int *mapcount, struct vm_area_struct *vma)
{
	struct mm_struct *mm = vma->vm_mm;
	pgd_t *pgd;
	pud_t *pud;
	pmd_t *pmd;
H
Hugh Dickins 已提交
805
	pte_t *pte;
L
Linus Torvalds 已提交
806
	pte_t pteval;
H
Hugh Dickins 已提交
807
	spinlock_t *ptl;
L
Linus Torvalds 已提交
808 809 810 811 812 813 814 815 816 817 818 819 820
	struct page *page;
	unsigned long address;
	unsigned long end;

	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))
H
Hugh Dickins 已提交
821
		return;
L
Linus Torvalds 已提交
822 823 824

	pud = pud_offset(pgd, address);
	if (!pud_present(*pud))
H
Hugh Dickins 已提交
825
		return;
L
Linus Torvalds 已提交
826 827 828

	pmd = pmd_offset(pud, address);
	if (!pmd_present(*pmd))
H
Hugh Dickins 已提交
829 830 831
		return;

	pte = pte_offset_map_lock(mm, pmd, address, &ptl);
L
Linus Torvalds 已提交
832

833 834 835
	/* Update high watermark before we lower rss */
	update_hiwater_rss(mm);

H
Hugh Dickins 已提交
836
	for (; address < end; pte++, address += PAGE_SIZE) {
L
Linus Torvalds 已提交
837 838
		if (!pte_present(*pte))
			continue;
839 840
		page = vm_normal_page(vma, address, *pte);
		BUG_ON(!page || PageAnon(page));
L
Linus Torvalds 已提交
841

A
Andrea Arcangeli 已提交
842
		if (ptep_clear_flush_young_notify(vma, address, pte))
L
Linus Torvalds 已提交
843 844 845
			continue;

		/* Nuke the page table entry. */
846
		flush_cache_page(vma, address, pte_pfn(*pte));
A
Andrea Arcangeli 已提交
847
		pteval = ptep_clear_flush_notify(vma, address, pte);
L
Linus Torvalds 已提交
848 849 850 851 852 853 854 855 856

		/* 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);

N
Nick Piggin 已提交
857
		page_remove_rmap(page, vma);
L
Linus Torvalds 已提交
858
		page_cache_release(page);
859
		dec_mm_counter(mm, file_rss);
L
Linus Torvalds 已提交
860 861
		(*mapcount)--;
	}
H
Hugh Dickins 已提交
862
	pte_unmap_unlock(pte - 1, ptl);
L
Linus Torvalds 已提交
863 864
}

865
static int try_to_unmap_anon(struct page *page, int migration)
L
Linus Torvalds 已提交
866 867 868 869 870 871 872 873 874 875
{
	struct anon_vma *anon_vma;
	struct vm_area_struct *vma;
	int ret = SWAP_AGAIN;

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

	list_for_each_entry(vma, &anon_vma->head, anon_vma_node) {
876
		ret = try_to_unmap_one(page, vma, migration);
L
Linus Torvalds 已提交
877 878 879
		if (ret == SWAP_FAIL || !page_mapped(page))
			break;
	}
880 881

	page_unlock_anon_vma(anon_vma);
L
Linus Torvalds 已提交
882 883 884 885 886 887
	return ret;
}

/**
 * try_to_unmap_file - unmap file page using the object-based rmap method
 * @page: the page to unmap
R
Randy Dunlap 已提交
888
 * @migration: migration flag
L
Linus Torvalds 已提交
889 890 891 892 893 894
 *
 * Find all the mappings of a page using the mapping pointer and the vma chains
 * contained in the address_space struct it points to.
 *
 * This function is only called from try_to_unmap for object-based pages.
 */
895
static int try_to_unmap_file(struct page *page, int migration)
L
Linus Torvalds 已提交
896 897 898 899 900 901 902 903 904 905 906 907 908
{
	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) {
909
		ret = try_to_unmap_one(page, vma, migration);
L
Linus Torvalds 已提交
910 911 912 913 914 915 916 917 918
		if (ret == SWAP_FAIL || !page_mapped(page))
			goto out;
	}

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

	list_for_each_entry(vma, &mapping->i_mmap_nonlinear,
						shared.vm_set.list) {
919
		if ((vma->vm_flags & VM_LOCKED) && !migration)
L
Linus Torvalds 已提交
920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952
			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;
	}

	if (max_nl_size == 0) {	/* any nonlinears locked or reserved */
		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) {
953
			if ((vma->vm_flags & VM_LOCKED) && !migration)
L
Linus Torvalds 已提交
954 955
				continue;
			cursor = (unsigned long) vma->vm_private_data;
H
Hugh Dickins 已提交
956
			while ( cursor < max_nl_cursor &&
L
Linus Torvalds 已提交
957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974
				cursor < vma->vm_end - vma->vm_start) {
				try_to_unmap_cluster(cursor, &mapcount, vma);
				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.
	 */
975 976
	list_for_each_entry(vma, &mapping->i_mmap_nonlinear, shared.vm_set.list)
		vma->vm_private_data = NULL;
L
Linus Torvalds 已提交
977 978 979 980 981 982 983 984
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
R
Randy Dunlap 已提交
985
 * @migration: migration flag
L
Linus Torvalds 已提交
986 987 988 989 990 991 992 993 994
 *
 * 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
 */
995
int try_to_unmap(struct page *page, int migration)
L
Linus Torvalds 已提交
996 997 998 999 1000 1001
{
	int ret;

	BUG_ON(!PageLocked(page));

	if (PageAnon(page))
1002
		ret = try_to_unmap_anon(page, migration);
L
Linus Torvalds 已提交
1003
	else
1004
		ret = try_to_unmap_file(page, migration);
L
Linus Torvalds 已提交
1005 1006 1007 1008 1009

	if (!page_mapped(page))
		ret = SWAP_SUCCESS;
	return ret;
}
N
Nikita Danilov 已提交
1010