rmap.c 48.4 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
 *   mm->mmap_sem
 *     page->flags PG_locked (lock_page)
26
 *       mapping->i_mmap_mutex
27
 *         anon_vma->rwsem
28 29 30 31 32
 *           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)
33
 *               inode->i_lock (in set_page_dirty's __mark_inode_dirty)
34
 *               bdi.wb->list_lock (in set_page_dirty's __mark_inode_dirty)
35 36 37
 *                 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,
38
 *                           within bdi.wb->list_lock in __sync_single_inode)
39
 *
40
 * anon_vma->rwsem,mapping->i_mutex      (memory_failure, collect_procs_anon)
41
 *   ->tasklist_lock
42
 *     pte map lock
L
Linus Torvalds 已提交
43 44 45 46 47 48 49 50
 */

#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
51
#include <linux/ksm.h>
L
Linus Torvalds 已提交
52 53
#include <linux/rmap.h>
#include <linux/rcupdate.h>
54
#include <linux/export.h>
55
#include <linux/memcontrol.h>
A
Andrea Arcangeli 已提交
56
#include <linux/mmu_notifier.h>
57
#include <linux/migrate.h>
58
#include <linux/hugetlb.h>
59
#include <linux/backing-dev.h>
L
Linus Torvalds 已提交
60 61 62

#include <asm/tlbflush.h>

N
Nick Piggin 已提交
63 64
#include "internal.h"

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

static inline struct anon_vma *anon_vma_alloc(void)
{
P
Peter Zijlstra 已提交
70 71 72 73 74 75 76 77 78 79 80 81 82
	struct anon_vma *anon_vma;

	anon_vma = kmem_cache_alloc(anon_vma_cachep, GFP_KERNEL);
	if (anon_vma) {
		atomic_set(&anon_vma->refcount, 1);
		/*
		 * Initialise the anon_vma root to point to itself. If called
		 * from fork, the root will be reset to the parents anon_vma.
		 */
		anon_vma->root = anon_vma;
	}

	return anon_vma;
83 84
}

P
Peter Zijlstra 已提交
85
static inline void anon_vma_free(struct anon_vma *anon_vma)
86
{
P
Peter Zijlstra 已提交
87
	VM_BUG_ON(atomic_read(&anon_vma->refcount));
88 89

	/*
90
	 * Synchronize against page_lock_anon_vma_read() such that
91 92 93 94 95
	 * we can safely hold the lock without the anon_vma getting
	 * freed.
	 *
	 * Relies on the full mb implied by the atomic_dec_and_test() from
	 * put_anon_vma() against the acquire barrier implied by
96
	 * down_read_trylock() from page_lock_anon_vma_read(). This orders:
97
	 *
98 99
	 * page_lock_anon_vma_read()	VS	put_anon_vma()
	 *   down_read_trylock()		  atomic_dec_and_test()
100
	 *   LOCK				  MB
101
	 *   atomic_read()			  rwsem_is_locked()
102 103 104 105
	 *
	 * LOCK should suffice since the actual taking of the lock must
	 * happen _before_ what follows.
	 */
106
	might_sleep();
107
	if (rwsem_is_locked(&anon_vma->root->rwsem)) {
108
		anon_vma_lock_write(anon_vma);
109
		anon_vma_unlock_write(anon_vma);
110 111
	}

112 113
	kmem_cache_free(anon_vma_cachep, anon_vma);
}
L
Linus Torvalds 已提交
114

115
static inline struct anon_vma_chain *anon_vma_chain_alloc(gfp_t gfp)
116
{
117
	return kmem_cache_alloc(anon_vma_chain_cachep, gfp);
118 119
}

120
static void anon_vma_chain_free(struct anon_vma_chain *anon_vma_chain)
121 122 123 124
{
	kmem_cache_free(anon_vma_chain_cachep, anon_vma_chain);
}

125 126 127 128 129 130 131
static void anon_vma_chain_link(struct vm_area_struct *vma,
				struct anon_vma_chain *avc,
				struct anon_vma *anon_vma)
{
	avc->vma = vma;
	avc->anon_vma = anon_vma;
	list_add(&avc->same_vma, &vma->anon_vma_chain);
132
	anon_vma_interval_tree_insert(avc, &anon_vma->rb_root);
133 134
}

135 136 137 138 139 140 141 142 143
/**
 * 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
F
Figo.zhang 已提交
144
 * not we either need to find an adjacent mapping that we
145 146 147 148 149
 * 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
150
 * optimistically looked up an anon_vma in page_lock_anon_vma_read()
151 152 153 154 155 156 157 158 159 160 161
 * 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
162 163 164
int anon_vma_prepare(struct vm_area_struct *vma)
{
	struct anon_vma *anon_vma = vma->anon_vma;
165
	struct anon_vma_chain *avc;
L
Linus Torvalds 已提交
166 167 168 169

	might_sleep();
	if (unlikely(!anon_vma)) {
		struct mm_struct *mm = vma->vm_mm;
170
		struct anon_vma *allocated;
L
Linus Torvalds 已提交
171

172
		avc = anon_vma_chain_alloc(GFP_KERNEL);
173 174 175
		if (!avc)
			goto out_enomem;

L
Linus Torvalds 已提交
176
		anon_vma = find_mergeable_anon_vma(vma);
177 178
		allocated = NULL;
		if (!anon_vma) {
L
Linus Torvalds 已提交
179 180
			anon_vma = anon_vma_alloc();
			if (unlikely(!anon_vma))
181
				goto out_enomem_free_avc;
L
Linus Torvalds 已提交
182 183 184
			allocated = anon_vma;
		}

185
		anon_vma_lock_write(anon_vma);
L
Linus Torvalds 已提交
186 187 188 189
		/* page_table_lock to protect against threads */
		spin_lock(&mm->page_table_lock);
		if (likely(!vma->anon_vma)) {
			vma->anon_vma = anon_vma;
190
			anon_vma_chain_link(vma, avc, anon_vma);
L
Linus Torvalds 已提交
191
			allocated = NULL;
192
			avc = NULL;
L
Linus Torvalds 已提交
193 194
		}
		spin_unlock(&mm->page_table_lock);
195
		anon_vma_unlock_write(anon_vma);
196 197

		if (unlikely(allocated))
P
Peter Zijlstra 已提交
198
			put_anon_vma(allocated);
199
		if (unlikely(avc))
200
			anon_vma_chain_free(avc);
L
Linus Torvalds 已提交
201 202
	}
	return 0;
203 204 205 206 207

 out_enomem_free_avc:
	anon_vma_chain_free(avc);
 out_enomem:
	return -ENOMEM;
L
Linus Torvalds 已提交
208 209
}

210 211 212 213 214 215 216 217 218 219 220 221 222
/*
 * This is a useful helper function for locking the anon_vma root as
 * we traverse the vma->anon_vma_chain, looping over anon_vma's that
 * have the same vma.
 *
 * Such anon_vma's should have the same root, so you'd expect to see
 * just a single mutex_lock for the whole traversal.
 */
static inline struct anon_vma *lock_anon_vma_root(struct anon_vma *root, struct anon_vma *anon_vma)
{
	struct anon_vma *new_root = anon_vma->root;
	if (new_root != root) {
		if (WARN_ON_ONCE(root))
223
			up_write(&root->rwsem);
224
		root = new_root;
225
		down_write(&root->rwsem);
226 227 228 229 230 231 232
	}
	return root;
}

static inline void unlock_anon_vma_root(struct anon_vma *root)
{
	if (root)
233
		up_write(&root->rwsem);
234 235
}

236 237 238 239 240
/*
 * 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
241
{
242
	struct anon_vma_chain *avc, *pavc;
243
	struct anon_vma *root = NULL;
244

245
	list_for_each_entry_reverse(pavc, &src->anon_vma_chain, same_vma) {
246 247
		struct anon_vma *anon_vma;

248 249 250 251 252 253 254 255
		avc = anon_vma_chain_alloc(GFP_NOWAIT | __GFP_NOWARN);
		if (unlikely(!avc)) {
			unlock_anon_vma_root(root);
			root = NULL;
			avc = anon_vma_chain_alloc(GFP_KERNEL);
			if (!avc)
				goto enomem_failure;
		}
256 257 258
		anon_vma = pavc->anon_vma;
		root = lock_anon_vma_root(root, anon_vma);
		anon_vma_chain_link(dst, avc, anon_vma);
259
	}
260
	unlock_anon_vma_root(root);
261
	return 0;
L
Linus Torvalds 已提交
262

263 264 265
 enomem_failure:
	unlink_anon_vmas(dst);
	return -ENOMEM;
L
Linus Torvalds 已提交
266 267
}

268 269 270 271 272 273
/*
 * 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
274
{
275 276
	struct anon_vma_chain *avc;
	struct anon_vma *anon_vma;
L
Linus Torvalds 已提交
277

278 279 280 281 282 283 284 285 286 287 288 289 290 291 292
	/* 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;
293
	avc = anon_vma_chain_alloc(GFP_KERNEL);
294 295
	if (!avc)
		goto out_error_free_anon_vma;
296 297 298 299 300 301

	/*
	 * The root anon_vma's spinlock is the lock actually used when we
	 * lock any of the anon_vmas in this anon_vma tree.
	 */
	anon_vma->root = pvma->anon_vma->root;
302
	/*
P
Peter Zijlstra 已提交
303 304 305
	 * With refcounts, an anon_vma can stay around longer than the
	 * process it belongs to. The root anon_vma needs to be pinned until
	 * this anon_vma is freed, because the lock lives in the root.
306 307
	 */
	get_anon_vma(anon_vma->root);
308 309
	/* Mark this anon_vma as the one where our new (COWed) pages go. */
	vma->anon_vma = anon_vma;
310
	anon_vma_lock_write(anon_vma);
311
	anon_vma_chain_link(vma, avc, anon_vma);
312
	anon_vma_unlock_write(anon_vma);
313 314 315 316

	return 0;

 out_error_free_anon_vma:
P
Peter Zijlstra 已提交
317
	put_anon_vma(anon_vma);
318
 out_error:
319
	unlink_anon_vmas(vma);
320
	return -ENOMEM;
L
Linus Torvalds 已提交
321 322
}

323 324 325
void unlink_anon_vmas(struct vm_area_struct *vma)
{
	struct anon_vma_chain *avc, *next;
326
	struct anon_vma *root = NULL;
327

328 329 330 331
	/*
	 * Unlink each anon_vma chained to the VMA.  This list is ordered
	 * from newest to oldest, ensuring the root anon_vma gets freed last.
	 */
332
	list_for_each_entry_safe(avc, next, &vma->anon_vma_chain, same_vma) {
333 334 335
		struct anon_vma *anon_vma = avc->anon_vma;

		root = lock_anon_vma_root(root, anon_vma);
336
		anon_vma_interval_tree_remove(avc, &anon_vma->rb_root);
337 338 339 340 341

		/*
		 * Leave empty anon_vmas on the list - we'll need
		 * to free them outside the lock.
		 */
342
		if (RB_EMPTY_ROOT(&anon_vma->rb_root))
343 344 345 346 347 348 349 350 351 352
			continue;

		list_del(&avc->same_vma);
		anon_vma_chain_free(avc);
	}
	unlock_anon_vma_root(root);

	/*
	 * Iterate the list once more, it now only contains empty and unlinked
	 * anon_vmas, destroy them. Could not do before due to __put_anon_vma()
353
	 * needing to write-acquire the anon_vma->root->rwsem.
354 355 356 357 358 359
	 */
	list_for_each_entry_safe(avc, next, &vma->anon_vma_chain, same_vma) {
		struct anon_vma *anon_vma = avc->anon_vma;

		put_anon_vma(anon_vma);

360 361 362 363 364
		list_del(&avc->same_vma);
		anon_vma_chain_free(avc);
	}
}

365
static void anon_vma_ctor(void *data)
L
Linus Torvalds 已提交
366
{
C
Christoph Lameter 已提交
367
	struct anon_vma *anon_vma = data;
L
Linus Torvalds 已提交
368

369
	init_rwsem(&anon_vma->rwsem);
370
	atomic_set(&anon_vma->refcount, 0);
371
	anon_vma->rb_root = RB_ROOT;
L
Linus Torvalds 已提交
372 373 374 375 376
}

void __init anon_vma_init(void)
{
	anon_vma_cachep = kmem_cache_create("anon_vma", sizeof(struct anon_vma),
377
			0, SLAB_DESTROY_BY_RCU|SLAB_PANIC, anon_vma_ctor);
378
	anon_vma_chain_cachep = KMEM_CACHE(anon_vma_chain, SLAB_PANIC);
L
Linus Torvalds 已提交
379 380 381
}

/*
382 383 384 385 386 387 388 389 390
 * Getting a lock on a stable anon_vma from a page off the LRU is tricky!
 *
 * Since there is no serialization what so ever against page_remove_rmap()
 * the best this function can do is return a locked anon_vma that might
 * have been relevant to this page.
 *
 * The page might have been remapped to a different anon_vma or the anon_vma
 * returned may already be freed (and even reused).
 *
391 392 393 394 395
 * In case it was remapped to a different anon_vma, the new anon_vma will be a
 * child of the old anon_vma, and the anon_vma lifetime rules will therefore
 * ensure that any anon_vma obtained from the page will still be valid for as
 * long as we observe page_mapped() [ hence all those page_mapped() tests ].
 *
396 397 398 399 400 401 402
 * All users of this function must be very careful when walking the anon_vma
 * chain and verify that the page in question is indeed mapped in it
 * [ something equivalent to page_mapped_in_vma() ].
 *
 * Since anon_vma's slab is DESTROY_BY_RCU and we know from page_remove_rmap()
 * that the anon_vma pointer from page->mapping is valid if there is a
 * mapcount, we can dereference the anon_vma after observing those.
L
Linus Torvalds 已提交
403
 */
404
struct anon_vma *page_get_anon_vma(struct page *page)
L
Linus Torvalds 已提交
405
{
406
	struct anon_vma *anon_vma = NULL;
L
Linus Torvalds 已提交
407 408 409
	unsigned long anon_mapping;

	rcu_read_lock();
410
	anon_mapping = (unsigned long) ACCESS_ONCE(page->mapping);
H
Hugh Dickins 已提交
411
	if ((anon_mapping & PAGE_MAPPING_FLAGS) != PAGE_MAPPING_ANON)
L
Linus Torvalds 已提交
412 413 414 415 416
		goto out;
	if (!page_mapped(page))
		goto out;

	anon_vma = (struct anon_vma *) (anon_mapping - PAGE_MAPPING_ANON);
417 418 419 420
	if (!atomic_inc_not_zero(&anon_vma->refcount)) {
		anon_vma = NULL;
		goto out;
	}
421 422 423

	/*
	 * If this page is still mapped, then its anon_vma cannot have been
424 425 426 427
	 * freed.  But if it has been unmapped, we have no security against the
	 * anon_vma structure being freed and reused (for another anon_vma:
	 * SLAB_DESTROY_BY_RCU guarantees that - so the atomic_inc_not_zero()
	 * above cannot corrupt).
428
	 */
429
	if (!page_mapped(page)) {
430
		rcu_read_unlock();
431
		put_anon_vma(anon_vma);
432
		return NULL;
433
	}
L
Linus Torvalds 已提交
434 435
out:
	rcu_read_unlock();
436 437 438 439

	return anon_vma;
}

440 441 442 443 444 445 446
/*
 * Similar to page_get_anon_vma() except it locks the anon_vma.
 *
 * Its a little more complex as it tries to keep the fast path to a single
 * atomic op -- the trylock. If we fail the trylock, we fall back to getting a
 * reference like with page_get_anon_vma() and then block on the mutex.
 */
447
struct anon_vma *page_lock_anon_vma_read(struct page *page)
448
{
449
	struct anon_vma *anon_vma = NULL;
450
	struct anon_vma *root_anon_vma;
451
	unsigned long anon_mapping;
452

453 454 455 456 457 458 459 460
	rcu_read_lock();
	anon_mapping = (unsigned long) ACCESS_ONCE(page->mapping);
	if ((anon_mapping & PAGE_MAPPING_FLAGS) != PAGE_MAPPING_ANON)
		goto out;
	if (!page_mapped(page))
		goto out;

	anon_vma = (struct anon_vma *) (anon_mapping - PAGE_MAPPING_ANON);
461
	root_anon_vma = ACCESS_ONCE(anon_vma->root);
462
	if (down_read_trylock(&root_anon_vma->rwsem)) {
463
		/*
464 465
		 * If the page is still mapped, then this anon_vma is still
		 * its anon_vma, and holding the mutex ensures that it will
466
		 * not go away, see anon_vma_free().
467
		 */
468
		if (!page_mapped(page)) {
469
			up_read(&root_anon_vma->rwsem);
470 471 472 473
			anon_vma = NULL;
		}
		goto out;
	}
474

475 476 477 478 479 480 481
	/* trylock failed, we got to sleep */
	if (!atomic_inc_not_zero(&anon_vma->refcount)) {
		anon_vma = NULL;
		goto out;
	}

	if (!page_mapped(page)) {
482
		rcu_read_unlock();
483
		put_anon_vma(anon_vma);
484
		return NULL;
485 486 487 488
	}

	/* we pinned the anon_vma, its safe to sleep */
	rcu_read_unlock();
489
	anon_vma_lock_read(anon_vma);
490 491 492 493 494

	if (atomic_dec_and_test(&anon_vma->refcount)) {
		/*
		 * Oops, we held the last refcount, release the lock
		 * and bail -- can't simply use put_anon_vma() because
495
		 * we'll deadlock on the anon_vma_lock_write() recursion.
496
		 */
497
		anon_vma_unlock_read(anon_vma);
498 499 500 501 502 503 504 505
		__put_anon_vma(anon_vma);
		anon_vma = NULL;
	}

	return anon_vma;

out:
	rcu_read_unlock();
506
	return anon_vma;
507 508
}

509
void page_unlock_anon_vma_read(struct anon_vma *anon_vma)
510
{
511
	anon_vma_unlock_read(anon_vma);
L
Linus Torvalds 已提交
512 513 514
}

/*
515
 * At what user virtual address is page expected in @vma?
L
Linus Torvalds 已提交
516
 */
517 518
static inline unsigned long
__vma_address(struct page *page, struct vm_area_struct *vma)
L
Linus Torvalds 已提交
519
{
520
	pgoff_t pgoff = page_to_pgoff(page);
521 522 523 524 525 526 527 528 529
	return vma->vm_start + ((pgoff - vma->vm_pgoff) << PAGE_SHIFT);
}

inline unsigned long
vma_address(struct page *page, struct vm_area_struct *vma)
{
	unsigned long address = __vma_address(page, vma);

	/* page should be within @vma mapping range */
530
	VM_BUG_ON_VMA(address < vma->vm_start || address >= vma->vm_end, vma);
531

L
Linus Torvalds 已提交
532 533 534 535
	return address;
}

/*
H
Huang Shijie 已提交
536
 * At what user virtual address is page expected in vma?
537
 * Caller should check the page is actually part of the vma.
L
Linus Torvalds 已提交
538 539 540
 */
unsigned long page_address_in_vma(struct page *page, struct vm_area_struct *vma)
{
541
	unsigned long address;
542
	if (PageAnon(page)) {
543 544 545 546 547 548 549
		struct anon_vma *page__anon_vma = page_anon_vma(page);
		/*
		 * Note: swapoff's unuse_vma() is more efficient with this
		 * check, and needs it to match anon_vma when KSM is active.
		 */
		if (!vma->anon_vma || !page__anon_vma ||
		    vma->anon_vma->root != page__anon_vma->root)
550 551
			return -EFAULT;
	} else if (page->mapping && !(vma->vm_flags & VM_NONLINEAR)) {
H
Hugh Dickins 已提交
552 553
		if (!vma->vm_file ||
		    vma->vm_file->f_mapping != page->mapping)
L
Linus Torvalds 已提交
554 555 556
			return -EFAULT;
	} else
		return -EFAULT;
557 558 559 560
	address = __vma_address(page, vma);
	if (unlikely(address < vma->vm_start || address >= vma->vm_end))
		return -EFAULT;
	return address;
L
Linus Torvalds 已提交
561 562
}

B
Bob Liu 已提交
563 564 565 566 567
pmd_t *mm_find_pmd(struct mm_struct *mm, unsigned long address)
{
	pgd_t *pgd;
	pud_t *pud;
	pmd_t *pmd = NULL;
568
	pmd_t pmde;
B
Bob Liu 已提交
569 570 571 572 573 574 575 576 577 578

	pgd = pgd_offset(mm, address);
	if (!pgd_present(*pgd))
		goto out;

	pud = pud_offset(pgd, address);
	if (!pud_present(*pud))
		goto out;

	pmd = pmd_offset(pud, address);
579 580 581 582 583 584 585
	/*
	 * Some THP functions use the sequence pmdp_clear_flush(), set_pmd_at()
	 * without holding anon_vma lock for write.  So when looking for a
	 * genuine pmde (in which to find pte), test present and !THP together.
	 */
	pmde = ACCESS_ONCE(*pmd);
	if (!pmd_present(pmde) || pmd_trans_huge(pmde))
B
Bob Liu 已提交
586 587 588 589 590
		pmd = NULL;
out:
	return pmd;
}

N
Nikita Danilov 已提交
591 592 593
/*
 * Check that @page is mapped at @address into @mm.
 *
N
Nick Piggin 已提交
594 595 596 597
 * 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).
 *
598
 * On success returns with pte mapped and locked.
N
Nikita Danilov 已提交
599
 */
600
pte_t *__page_check_address(struct page *page, struct mm_struct *mm,
N
Nick Piggin 已提交
601
			  unsigned long address, spinlock_t **ptlp, int sync)
N
Nikita Danilov 已提交
602 603 604
{
	pmd_t *pmd;
	pte_t *pte;
H
Hugh Dickins 已提交
605
	spinlock_t *ptl;
N
Nikita Danilov 已提交
606

607
	if (unlikely(PageHuge(page))) {
608
		/* when pud is not present, pte will be NULL */
609
		pte = huge_pte_offset(mm, address);
610 611 612
		if (!pte)
			return NULL;

613
		ptl = huge_pte_lockptr(page_hstate(page), mm, pte);
614 615 616
		goto check;
	}

B
Bob Liu 已提交
617 618
	pmd = mm_find_pmd(mm, address);
	if (!pmd)
H
Hugh Dickins 已提交
619 620 621 622
		return NULL;

	pte = pte_offset_map(pmd, address);
	/* Make a quick check before getting the lock */
N
Nick Piggin 已提交
623
	if (!sync && !pte_present(*pte)) {
H
Hugh Dickins 已提交
624 625 626 627
		pte_unmap(pte);
		return NULL;
	}

H
Hugh Dickins 已提交
628
	ptl = pte_lockptr(mm, pmd);
629
check:
H
Hugh Dickins 已提交
630 631 632 633
	spin_lock(ptl);
	if (pte_present(*pte) && page_to_pfn(page) == pte_pfn(*pte)) {
		*ptlp = ptl;
		return pte;
N
Nikita Danilov 已提交
634
	}
H
Hugh Dickins 已提交
635 636
	pte_unmap_unlock(pte, ptl);
	return NULL;
N
Nikita Danilov 已提交
637 638
}

N
Nick Piggin 已提交
639 640 641 642 643 644 645 646 647
/**
 * 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.
 */
648
int page_mapped_in_vma(struct page *page, struct vm_area_struct *vma)
N
Nick Piggin 已提交
649 650 651 652 653
{
	unsigned long address;
	pte_t *pte;
	spinlock_t *ptl;

654 655
	address = __vma_address(page, vma);
	if (unlikely(address < vma->vm_start || address >= vma->vm_end))
N
Nick Piggin 已提交
656 657 658 659 660 661 662 663 664
		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;
}

665 666 667 668 669 670
struct page_referenced_arg {
	int mapcount;
	int referenced;
	unsigned long vm_flags;
	struct mem_cgroup *memcg;
};
L
Linus Torvalds 已提交
671
/*
672
 * arg: page_referenced_arg will be passed
L
Linus Torvalds 已提交
673
 */
674
static int page_referenced_one(struct page *page, struct vm_area_struct *vma,
675
			unsigned long address, void *arg)
L
Linus Torvalds 已提交
676 677
{
	struct mm_struct *mm = vma->vm_mm;
678
	spinlock_t *ptl;
L
Linus Torvalds 已提交
679
	int referenced = 0;
680
	struct page_referenced_arg *pra = arg;
L
Linus Torvalds 已提交
681

682 683 684
	if (unlikely(PageTransHuge(page))) {
		pmd_t *pmd;

685 686 687 688
		/*
		 * rmap might return false positives; we must filter
		 * these out using page_check_address_pmd().
		 */
689
		pmd = page_check_address_pmd(page, mm, address,
690 691
					     PAGE_CHECK_ADDRESS_PMD_FLAG, &ptl);
		if (!pmd)
692
			return SWAP_AGAIN;
693 694

		if (vma->vm_flags & VM_LOCKED) {
695
			spin_unlock(ptl);
696 697
			pra->vm_flags |= VM_LOCKED;
			return SWAP_FAIL; /* To break the loop */
698 699 700 701
		}

		/* go ahead even if the pmd is pmd_trans_splitting() */
		if (pmdp_clear_flush_young_notify(vma, address, pmd))
702
			referenced++;
703
		spin_unlock(ptl);
704 705 706
	} else {
		pte_t *pte;

707 708 709 710
		/*
		 * rmap might return false positives; we must filter
		 * these out using page_check_address().
		 */
711 712
		pte = page_check_address(page, mm, address, &ptl, 0);
		if (!pte)
713
			return SWAP_AGAIN;
714

715 716
		if (vma->vm_flags & VM_LOCKED) {
			pte_unmap_unlock(pte, ptl);
717 718
			pra->vm_flags |= VM_LOCKED;
			return SWAP_FAIL; /* To break the loop */
719 720
		}

721 722 723 724 725 726 727 728
		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.
			 */
729
			if (likely(!(vma->vm_flags & VM_SEQ_READ)))
730 731 732 733 734
				referenced++;
		}
		pte_unmap_unlock(pte, ptl);
	}

735 736 737
	if (referenced) {
		pra->referenced++;
		pra->vm_flags |= vma->vm_flags;
L
Linus Torvalds 已提交
738
	}
739

740 741 742 743 744
	pra->mapcount--;
	if (!pra->mapcount)
		return SWAP_SUCCESS; /* To break the loop */

	return SWAP_AGAIN;
L
Linus Torvalds 已提交
745 746
}

747
static bool invalid_page_referenced_vma(struct vm_area_struct *vma, void *arg)
L
Linus Torvalds 已提交
748
{
749 750
	struct page_referenced_arg *pra = arg;
	struct mem_cgroup *memcg = pra->memcg;
L
Linus Torvalds 已提交
751

752 753
	if (!mm_match_cgroup(vma->vm_mm, memcg))
		return true;
L
Linus Torvalds 已提交
754

755
	return false;
L
Linus Torvalds 已提交
756 757 758 759 760 761
}

/**
 * page_referenced - test if the page was referenced
 * @page: the page to test
 * @is_locked: caller holds lock on the page
762
 * @memcg: target memory cgroup
763
 * @vm_flags: collect encountered vma->vm_flags who actually referenced the page
L
Linus Torvalds 已提交
764 765 766 767
 *
 * Quick test_and_clear_referenced for all mappings to a page,
 * returns the number of ptes which referenced the page.
 */
768 769
int page_referenced(struct page *page,
		    int is_locked,
770
		    struct mem_cgroup *memcg,
771
		    unsigned long *vm_flags)
L
Linus Torvalds 已提交
772
{
773
	int ret;
H
Hugh Dickins 已提交
774
	int we_locked = 0;
775 776 777 778 779 780 781 782 783
	struct page_referenced_arg pra = {
		.mapcount = page_mapcount(page),
		.memcg = memcg,
	};
	struct rmap_walk_control rwc = {
		.rmap_one = page_referenced_one,
		.arg = (void *)&pra,
		.anon_lock = page_lock_anon_vma_read,
	};
L
Linus Torvalds 已提交
784

785
	*vm_flags = 0;
786 787 788 789 790 791 792 793 794 795
	if (!page_mapped(page))
		return 0;

	if (!page_rmapping(page))
		return 0;

	if (!is_locked && (!PageAnon(page) || PageKsm(page))) {
		we_locked = trylock_page(page);
		if (!we_locked)
			return 1;
L
Linus Torvalds 已提交
796
	}
797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813

	/*
	 * If we are reclaiming on behalf of a cgroup, skip
	 * counting on behalf of references from different
	 * cgroups
	 */
	if (memcg) {
		rwc.invalid_vma = invalid_page_referenced_vma;
	}

	ret = rmap_walk(page, &rwc);
	*vm_flags = pra.vm_flags;

	if (we_locked)
		unlock_page(page);

	return pra.referenced;
L
Linus Torvalds 已提交
814 815
}

H
Hugh Dickins 已提交
816
static int page_mkclean_one(struct page *page, struct vm_area_struct *vma,
817
			    unsigned long address, void *arg)
818 819
{
	struct mm_struct *mm = vma->vm_mm;
820
	pte_t *pte;
821 822
	spinlock_t *ptl;
	int ret = 0;
823
	int *cleaned = arg;
824

N
Nick Piggin 已提交
825
	pte = page_check_address(page, mm, address, &ptl, 1);
826 827 828
	if (!pte)
		goto out;

829 830
	if (pte_dirty(*pte) || pte_write(*pte)) {
		pte_t entry;
831

832
		flush_cache_page(vma, address, pte_pfn(*pte));
833
		entry = ptep_clear_flush(vma, address, pte);
834 835
		entry = pte_wrprotect(entry);
		entry = pte_mkclean(entry);
836
		set_pte_at(mm, address, pte, entry);
837 838
		ret = 1;
	}
839 840

	pte_unmap_unlock(pte, ptl);
841

842
	if (ret) {
843
		mmu_notifier_invalidate_page(mm, address);
844 845
		(*cleaned)++;
	}
846
out:
847
	return SWAP_AGAIN;
848 849
}

850
static bool invalid_mkclean_vma(struct vm_area_struct *vma, void *arg)
851
{
852
	if (vma->vm_flags & VM_SHARED)
F
Fengguang Wu 已提交
853
		return false;
854

F
Fengguang Wu 已提交
855
	return true;
856 857 858 859
}

int page_mkclean(struct page *page)
{
860 861 862 863 864 865 866
	int cleaned = 0;
	struct address_space *mapping;
	struct rmap_walk_control rwc = {
		.arg = (void *)&cleaned,
		.rmap_one = page_mkclean_one,
		.invalid_vma = invalid_mkclean_vma,
	};
867 868 869

	BUG_ON(!PageLocked(page));

870 871 872 873 874 875 876 877
	if (!page_mapped(page))
		return 0;

	mapping = page_mapping(page);
	if (!mapping)
		return 0;

	rmap_walk(page, &rwc);
878

879
	return cleaned;
880
}
J
Jaya Kumar 已提交
881
EXPORT_SYMBOL_GPL(page_mkclean);
882

883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898
/**
 * 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;

899
	VM_BUG_ON_PAGE(!PageLocked(page), page);
900
	VM_BUG_ON_VMA(!anon_vma, vma);
901
	VM_BUG_ON_PAGE(page->index != linear_page_index(vma, address), page);
902 903 904 905 906

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

N
Nick Piggin 已提交
907
/**
A
Andi Kleen 已提交
908 909 910 911
 * __page_set_anon_rmap - set up new anonymous rmap
 * @page:	Page to add to rmap	
 * @vma:	VM area to add page to.
 * @address:	User virtual address of the mapping	
912
 * @exclusive:	the page is exclusively owned by the current process
N
Nick Piggin 已提交
913 914
 */
static void __page_set_anon_rmap(struct page *page,
915
	struct vm_area_struct *vma, unsigned long address, int exclusive)
N
Nick Piggin 已提交
916
{
917
	struct anon_vma *anon_vma = vma->anon_vma;
918

919
	BUG_ON(!anon_vma);
920

A
Andi Kleen 已提交
921 922 923
	if (PageAnon(page))
		return;

924
	/*
925 926 927
	 * If the page isn't exclusively mapped into this vma,
	 * we must use the _oldest_ possible anon_vma for the
	 * page mapping!
928
	 */
A
Andi Kleen 已提交
929
	if (!exclusive)
930
		anon_vma = anon_vma->root;
N
Nick Piggin 已提交
931 932 933 934 935 936

	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
937
/**
R
Randy Dunlap 已提交
938
 * __page_check_anon_rmap - sanity check anonymous rmap addition
N
Nick Piggin 已提交
939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958
 * @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.
	 */
A
Andrea Arcangeli 已提交
959
	BUG_ON(page_anon_vma(page)->root != vma->anon_vma->root);
N
Nick Piggin 已提交
960 961 962 963
	BUG_ON(page->index != linear_page_index(vma, address));
#endif
}

L
Linus Torvalds 已提交
964 965 966 967 968 969
/**
 * 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
970
 * The caller needs to hold the pte lock, and the page must be locked in
971 972 973
 * 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
974 975 976
 */
void page_add_anon_rmap(struct page *page,
	struct vm_area_struct *vma, unsigned long address)
977 978 979 980 981 982 983 984 985 986 987
{
	do_page_add_anon_rmap(page, vma, address, 0);
}

/*
 * Special version of the above for do_swap_page, which often runs
 * into pages that are exclusively owned by the current process.
 * Everybody else should continue to use page_add_anon_rmap above.
 */
void do_page_add_anon_rmap(struct page *page,
	struct vm_area_struct *vma, unsigned long address, int exclusive)
L
Linus Torvalds 已提交
988
{
H
Hugh Dickins 已提交
989
	int first = atomic_inc_and_test(&page->_mapcount);
990
	if (first) {
991 992 993 994 995 996
		/*
		 * We use the irq-unsafe __{inc|mod}_zone_page_stat because
		 * these counters are not modified in interrupt context, and
		 * pte lock(a spinlock) is held, which implies preemption
		 * disabled.
		 */
997
		if (PageTransHuge(page))
998 999
			__inc_zone_page_state(page,
					      NR_ANON_TRANSPARENT_HUGEPAGES);
1000 1001
		__mod_zone_page_state(page_zone(page), NR_ANON_PAGES,
				hpage_nr_pages(page));
1002
	}
H
Hugh Dickins 已提交
1003 1004 1005
	if (unlikely(PageKsm(page)))
		return;

1006
	VM_BUG_ON_PAGE(!PageLocked(page), page);
1007
	/* address might be in next vma when migration races vma_adjust */
H
Hugh Dickins 已提交
1008
	if (first)
1009
		__page_set_anon_rmap(page, vma, address, exclusive);
1010
	else
N
Nick Piggin 已提交
1011
		__page_check_anon_rmap(page, vma, address);
L
Linus Torvalds 已提交
1012 1013
}

R
Randy Dunlap 已提交
1014
/**
N
Nick Piggin 已提交
1015 1016 1017 1018 1019 1020 1021
 * 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
1022
 * Page does not have to be locked.
N
Nick Piggin 已提交
1023 1024 1025 1026
 */
void page_add_new_anon_rmap(struct page *page,
	struct vm_area_struct *vma, unsigned long address)
{
1027
	VM_BUG_ON_VMA(address < vma->vm_start || address >= vma->vm_end, vma);
1028 1029
	SetPageSwapBacked(page);
	atomic_set(&page->_mapcount, 0); /* increment count (starts at -1) */
1030
	if (PageTransHuge(page))
1031
		__inc_zone_page_state(page, NR_ANON_TRANSPARENT_HUGEPAGES);
1032 1033
	__mod_zone_page_state(page_zone(page), NR_ANON_PAGES,
			hpage_nr_pages(page));
1034
	__page_set_anon_rmap(page, vma, address, 1);
N
Nick Piggin 已提交
1035 1036
}

L
Linus Torvalds 已提交
1037 1038 1039 1040
/**
 * page_add_file_rmap - add pte mapping to a file page
 * @page: the page to add the mapping to
 *
1041
 * The caller needs to hold the pte lock.
L
Linus Torvalds 已提交
1042 1043 1044
 */
void page_add_file_rmap(struct page *page)
{
1045 1046 1047 1048
	bool locked;
	unsigned long flags;

	mem_cgroup_begin_update_page_stat(page, &locked, &flags);
1049
	if (atomic_inc_and_test(&page->_mapcount)) {
1050
		__inc_zone_page_state(page, NR_FILE_MAPPED);
S
Sha Zhengju 已提交
1051
		mem_cgroup_inc_page_stat(page, MEM_CGROUP_STAT_FILE_MAPPED);
1052
	}
1053
	mem_cgroup_end_update_page_stat(page, &locked, &flags);
L
Linus Torvalds 已提交
1054 1055 1056 1057 1058 1059
}

/**
 * page_remove_rmap - take down pte mapping from a page
 * @page: page to remove mapping from
 *
1060
 * The caller needs to hold the pte lock.
L
Linus Torvalds 已提交
1061
 */
1062
void page_remove_rmap(struct page *page)
L
Linus Torvalds 已提交
1063
{
1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075
	bool anon = PageAnon(page);
	bool locked;
	unsigned long flags;

	/*
	 * The anon case has no mem_cgroup page_stat to update; but may
	 * uncharge_page() below, where the lock ordering can deadlock if
	 * we hold the lock against page_stat move: so avoid it on anon.
	 */
	if (!anon)
		mem_cgroup_begin_update_page_stat(page, &locked, &flags);

1076 1077
	/* page still mapped by someone else? */
	if (!atomic_add_negative(-1, &page->_mapcount))
1078
		goto out;
1079

1080 1081 1082
	/*
	 * Hugepages are not counted in NR_ANON_PAGES nor NR_FILE_MAPPED
	 * and not charged by memcg for now.
1083 1084 1085 1086 1087
	 *
	 * We use the irq-unsafe __{inc|mod}_zone_page_stat because
	 * these counters are not modified in interrupt context, and
	 * these counters are not modified in interrupt context, and
	 * pte lock(a spinlock) is held, which implies preemption disabled.
1088 1089
	 */
	if (unlikely(PageHuge(page)))
1090 1091
		goto out;
	if (anon) {
1092
		if (PageTransHuge(page))
1093 1094
			__dec_zone_page_state(page,
					      NR_ANON_TRANSPARENT_HUGEPAGES);
1095 1096
		__mod_zone_page_state(page_zone(page), NR_ANON_PAGES,
				-hpage_nr_pages(page));
1097 1098
	} else {
		__dec_zone_page_state(page, NR_FILE_MAPPED);
S
Sha Zhengju 已提交
1099
		mem_cgroup_dec_page_stat(page, MEM_CGROUP_STAT_FILE_MAPPED);
1100
		mem_cgroup_end_update_page_stat(page, &locked, &flags);
1101
	}
1102 1103
	if (unlikely(PageMlocked(page)))
		clear_page_mlock(page);
1104 1105 1106 1107 1108 1109 1110 1111 1112
	/*
	 * 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.
	 */
1113
	return;
1114 1115 1116
out:
	if (!anon)
		mem_cgroup_end_update_page_stat(page, &locked, &flags);
L
Linus Torvalds 已提交
1117 1118 1119
}

/*
1120
 * @arg: enum ttu_flags will be passed to this argument
L
Linus Torvalds 已提交
1121
 */
1122
static int try_to_unmap_one(struct page *page, struct vm_area_struct *vma,
1123
		     unsigned long address, void *arg)
L
Linus Torvalds 已提交
1124 1125 1126 1127
{
	struct mm_struct *mm = vma->vm_mm;
	pte_t *pte;
	pte_t pteval;
H
Hugh Dickins 已提交
1128
	spinlock_t *ptl;
L
Linus Torvalds 已提交
1129
	int ret = SWAP_AGAIN;
1130
	enum ttu_flags flags = (enum ttu_flags)arg;
L
Linus Torvalds 已提交
1131

N
Nick Piggin 已提交
1132
	pte = page_check_address(page, mm, address, &ptl, 0);
H
Hugh Dickins 已提交
1133
	if (!pte)
N
Nikita Danilov 已提交
1134
		goto out;
L
Linus Torvalds 已提交
1135 1136 1137 1138 1139 1140

	/*
	 * 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.
	 */
1141
	if (!(flags & TTU_IGNORE_MLOCK)) {
K
KOSAKI Motohiro 已提交
1142 1143 1144
		if (vma->vm_flags & VM_LOCKED)
			goto out_mlock;

1145
		if (flags & TTU_MUNLOCK)
H
Hugh Dickins 已提交
1146
			goto out_unmap;
1147 1148
	}
	if (!(flags & TTU_IGNORE_ACCESS)) {
N
Nick Piggin 已提交
1149 1150 1151 1152 1153
		if (ptep_clear_flush_young_notify(vma, address, pte)) {
			ret = SWAP_FAIL;
			goto out_unmap;
		}
  	}
L
Linus Torvalds 已提交
1154 1155 1156

	/* Nuke the page table entry. */
	flush_cache_page(vma, address, page_to_pfn(page));
1157
	pteval = ptep_clear_flush(vma, address, pte);
L
Linus Torvalds 已提交
1158 1159 1160 1161 1162

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

1163 1164 1165
	/* Update high watermark before we lower rss */
	update_hiwater_rss(mm);

1166
	if (PageHWPoison(page) && !(flags & TTU_IGNORE_HWPOISON)) {
1167 1168 1169 1170 1171 1172
		if (!PageHuge(page)) {
			if (PageAnon(page))
				dec_mm_counter(mm, MM_ANONPAGES);
			else
				dec_mm_counter(mm, MM_FILEPAGES);
		}
1173
		set_pte_at(mm, address, pte,
1174
			   swp_entry_to_pte(make_hwpoison_entry(page)));
1175 1176 1177 1178 1179 1180 1181 1182 1183 1184
	} else if (pte_unused(pteval)) {
		/*
		 * The guest indicated that the page content is of no
		 * interest anymore. Simply discard the pte, vmscan
		 * will take care of the rest.
		 */
		if (PageAnon(page))
			dec_mm_counter(mm, MM_ANONPAGES);
		else
			dec_mm_counter(mm, MM_FILEPAGES);
1185
	} else if (PageAnon(page)) {
H
Hugh Dickins 已提交
1186
		swp_entry_t entry = { .val = page_private(page) };
1187
		pte_t swp_pte;
1188 1189 1190 1191 1192 1193

		if (PageSwapCache(page)) {
			/*
			 * Store the swap location in the pte.
			 * See handle_pte_fault() ...
			 */
H
Hugh Dickins 已提交
1194 1195 1196 1197 1198
			if (swap_duplicate(entry) < 0) {
				set_pte_at(mm, address, pte, pteval);
				ret = SWAP_FAIL;
				goto out_unmap;
			}
1199 1200 1201 1202 1203 1204
			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
1205
			dec_mm_counter(mm, MM_ANONPAGES);
K
KAMEZAWA Hiroyuki 已提交
1206
			inc_mm_counter(mm, MM_SWAPENTS);
1207
		} else if (IS_ENABLED(CONFIG_MIGRATION)) {
1208 1209 1210 1211 1212
			/*
			 * 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.
			 */
1213
			BUG_ON(!(flags & TTU_MIGRATION));
1214
			entry = make_migration_entry(page, pte_write(pteval));
L
Linus Torvalds 已提交
1215
		}
1216 1217 1218 1219
		swp_pte = swp_entry_to_pte(entry);
		if (pte_soft_dirty(pteval))
			swp_pte = pte_swp_mksoft_dirty(swp_pte);
		set_pte_at(mm, address, pte, swp_pte);
L
Linus Torvalds 已提交
1220
		BUG_ON(pte_file(*pte));
1221
	} else if (IS_ENABLED(CONFIG_MIGRATION) &&
1222
		   (flags & TTU_MIGRATION)) {
1223 1224 1225 1226 1227
		/* 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
1228
		dec_mm_counter(mm, MM_FILEPAGES);
L
Linus Torvalds 已提交
1229

1230
	page_remove_rmap(page);
L
Linus Torvalds 已提交
1231 1232 1233
	page_cache_release(page);

out_unmap:
H
Hugh Dickins 已提交
1234
	pte_unmap_unlock(pte, ptl);
1235
	if (ret != SWAP_FAIL && !(flags & TTU_MUNLOCK))
1236
		mmu_notifier_invalidate_page(mm, address);
K
KOSAKI Motohiro 已提交
1237 1238
out:
	return ret;
H
Hugh Dickins 已提交
1239

K
KOSAKI Motohiro 已提交
1240 1241 1242 1243 1244 1245 1246
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
1247
	 * we now hold anon_vma->rwsem or mapping->i_mmap_mutex.
K
KOSAKI Motohiro 已提交
1248 1249 1250 1251 1252 1253 1254 1255
	 * 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
1256
		}
K
KOSAKI Motohiro 已提交
1257
		up_read(&vma->vm_mm->mmap_sem);
H
Hugh Dickins 已提交
1258
	}
L
Linus Torvalds 已提交
1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279
	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
1280 1281 1282 1283 1284
 *
 * 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
1285 1286 1287 1288
 */
#define CLUSTER_SIZE	min(32*PAGE_SIZE, PMD_SIZE)
#define CLUSTER_MASK	(~(CLUSTER_SIZE - 1))

N
Nick Piggin 已提交
1289 1290
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
1291 1292 1293
{
	struct mm_struct *mm = vma->vm_mm;
	pmd_t *pmd;
H
Hugh Dickins 已提交
1294
	pte_t *pte;
L
Linus Torvalds 已提交
1295
	pte_t pteval;
H
Hugh Dickins 已提交
1296
	spinlock_t *ptl;
L
Linus Torvalds 已提交
1297 1298
	struct page *page;
	unsigned long address;
1299 1300
	unsigned long mmun_start;	/* For mmu_notifiers */
	unsigned long mmun_end;		/* For mmu_notifiers */
L
Linus Torvalds 已提交
1301
	unsigned long end;
N
Nick Piggin 已提交
1302 1303
	int ret = SWAP_AGAIN;
	int locked_vma = 0;
L
Linus Torvalds 已提交
1304 1305 1306 1307 1308 1309 1310 1311

	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;

B
Bob Liu 已提交
1312 1313
	pmd = mm_find_pmd(mm, address);
	if (!pmd)
N
Nick Piggin 已提交
1314 1315
		return ret;

1316 1317 1318 1319
	mmun_start = address;
	mmun_end   = end;
	mmu_notifier_invalidate_range_start(mm, mmun_start, mmun_end);

N
Nick Piggin 已提交
1320
	/*
H
Hugh Dickins 已提交
1321
	 * If we can acquire the mmap_sem for read, and vma is VM_LOCKED,
N
Nick Piggin 已提交
1322 1323
	 * keep the sem while scanning the cluster for mlocking pages.
	 */
H
Hugh Dickins 已提交
1324
	if (down_read_trylock(&vma->vm_mm->mmap_sem)) {
N
Nick Piggin 已提交
1325 1326 1327 1328
		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
1329 1330

	pte = pte_offset_map_lock(mm, pmd, address, &ptl);
L
Linus Torvalds 已提交
1331

1332 1333 1334
	/* Update high watermark before we lower rss */
	update_hiwater_rss(mm);

H
Hugh Dickins 已提交
1335
	for (; address < end; pte++, address += PAGE_SIZE) {
L
Linus Torvalds 已提交
1336 1337
		if (!pte_present(*pte))
			continue;
1338 1339
		page = vm_normal_page(vma, address, *pte);
		BUG_ON(!page || PageAnon(page));
L
Linus Torvalds 已提交
1340

N
Nick Piggin 已提交
1341
		if (locked_vma) {
1342 1343 1344
			if (page == check_page) {
				/* we know we have check_page locked */
				mlock_vma_page(page);
N
Nick Piggin 已提交
1345
				ret = SWAP_MLOCK;
1346 1347 1348 1349 1350 1351 1352 1353 1354
			} else if (trylock_page(page)) {
				/*
				 * If we can lock the page, perform mlock.
				 * Otherwise leave the page alone, it will be
				 * eventually encountered again later.
				 */
				mlock_vma_page(page);
				unlock_page(page);
			}
N
Nick Piggin 已提交
1355 1356 1357
			continue;	/* don't unmap */
		}

A
Andres Lagar-Cavilla 已提交
1358 1359 1360 1361 1362
		/*
		 * No need for _notify because we're within an
		 * mmu_notifier_invalidate_range_ {start|end} scope.
		 */
		if (ptep_clear_flush_young(vma, address, pte))
L
Linus Torvalds 已提交
1363 1364 1365
			continue;

		/* Nuke the page table entry. */
1366
		flush_cache_page(vma, address, pte_pfn(*pte));
1367
		pteval = ptep_clear_flush(vma, address, pte);
L
Linus Torvalds 已提交
1368 1369

		/* If nonlinear, store the file page offset in the pte. */
1370 1371 1372
		if (page->index != linear_page_index(vma, address)) {
			pte_t ptfile = pgoff_to_pte(page->index);
			if (pte_soft_dirty(pteval))
1373
				ptfile = pte_file_mksoft_dirty(ptfile);
1374 1375
			set_pte_at(mm, address, pte, ptfile);
		}
L
Linus Torvalds 已提交
1376 1377 1378 1379 1380

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

1381
		page_remove_rmap(page);
L
Linus Torvalds 已提交
1382
		page_cache_release(page);
K
KAMEZAWA Hiroyuki 已提交
1383
		dec_mm_counter(mm, MM_FILEPAGES);
L
Linus Torvalds 已提交
1384 1385
		(*mapcount)--;
	}
H
Hugh Dickins 已提交
1386
	pte_unmap_unlock(pte - 1, ptl);
1387
	mmu_notifier_invalidate_range_end(mm, mmun_start, mmun_end);
N
Nick Piggin 已提交
1388 1389 1390
	if (locked_vma)
		up_read(&vma->vm_mm->mmap_sem);
	return ret;
L
Linus Torvalds 已提交
1391 1392
}

1393
static int try_to_unmap_nonlinear(struct page *page,
1394
		struct address_space *mapping, void *arg)
1395
{
1396
	struct vm_area_struct *vma;
1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466
	int ret = SWAP_AGAIN;
	unsigned long cursor;
	unsigned long max_nl_cursor = 0;
	unsigned long max_nl_size = 0;
	unsigned int mapcount;

	list_for_each_entry(vma,
		&mapping->i_mmap_nonlinear, shared.nonlinear) {

		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) {	/* all nonlinears locked or reserved ? */
		return SWAP_FAIL;
	}

	/*
	 * 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)
		return ret;

	cond_resched();

	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.nonlinear) {

			cursor = (unsigned long) vma->vm_private_data;
			while (cursor < max_nl_cursor &&
				cursor < vma->vm_end - vma->vm_start) {
				if (try_to_unmap_cluster(cursor, &mapcount,
						vma, page) == SWAP_MLOCK)
					ret = SWAP_MLOCK;
				cursor += CLUSTER_SIZE;
				vma->vm_private_data = (void *) cursor;
				if ((int)mapcount <= 0)
					return ret;
			}
			vma->vm_private_data = (void *) max_nl_cursor;
		}
		cond_resched();
		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.
	 */
	list_for_each_entry(vma, &mapping->i_mmap_nonlinear, shared.nonlinear)
		vma->vm_private_data = NULL;

	return ret;
}

1467
bool is_vma_temporary_stack(struct vm_area_struct *vma)
1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480
{
	int maybe_stack = vma->vm_flags & (VM_GROWSDOWN | VM_GROWSUP);

	if (!maybe_stack)
		return false;

	if ((vma->vm_flags & VM_STACK_INCOMPLETE_SETUP) ==
						VM_STACK_INCOMPLETE_SETUP)
		return true;

	return false;
}

1481 1482 1483 1484 1485 1486 1487 1488 1489 1490
static bool invalid_migration_vma(struct vm_area_struct *vma, void *arg)
{
	return is_vma_temporary_stack(vma);
}

static int page_not_mapped(struct page *page)
{
	return !page_mapped(page);
};

L
Linus Torvalds 已提交
1491 1492 1493
/**
 * try_to_unmap - try to remove all page table mappings to a page
 * @page: the page to get unmapped
1494
 * @flags: action and flags
L
Linus Torvalds 已提交
1495 1496 1497 1498 1499 1500 1501 1502
 *
 * 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
1503
 * SWAP_MLOCK	- page is mlocked.
L
Linus Torvalds 已提交
1504
 */
1505
int try_to_unmap(struct page *page, enum ttu_flags flags)
L
Linus Torvalds 已提交
1506 1507
{
	int ret;
1508 1509 1510 1511 1512 1513 1514
	struct rmap_walk_control rwc = {
		.rmap_one = try_to_unmap_one,
		.arg = (void *)flags,
		.done = page_not_mapped,
		.file_nonlinear = try_to_unmap_nonlinear,
		.anon_lock = page_lock_anon_vma_read,
	};
L
Linus Torvalds 已提交
1515

1516
	VM_BUG_ON_PAGE(!PageHuge(page) && PageTransHuge(page), page);
L
Linus Torvalds 已提交
1517

1518 1519 1520 1521 1522 1523 1524 1525
	/*
	 * During exec, a temporary VMA is setup and later moved.
	 * The VMA is moved under the anon_vma lock but not the
	 * page tables leading to a race where migration cannot
	 * find the migration ptes. Rather than increasing the
	 * locking requirements of exec(), migration skips
	 * temporary VMAs until after exec() completes.
	 */
1526
	if ((flags & TTU_MIGRATION) && !PageKsm(page) && PageAnon(page))
1527 1528 1529 1530
		rwc.invalid_vma = invalid_migration_vma;

	ret = rmap_walk(page, &rwc);

N
Nick Piggin 已提交
1531
	if (ret != SWAP_MLOCK && !page_mapped(page))
L
Linus Torvalds 已提交
1532 1533 1534
		ret = SWAP_SUCCESS;
	return ret;
}
N
Nikita Danilov 已提交
1535

N
Nick Piggin 已提交
1536 1537 1538 1539 1540 1541 1542 1543 1544 1545
/**
 * 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
1546
 * SWAP_AGAIN	- no vma is holding page mlocked, or,
N
Nick Piggin 已提交
1547
 * SWAP_AGAIN	- page mapped in mlocked vma -- couldn't acquire mmap sem
H
Hugh Dickins 已提交
1548
 * SWAP_FAIL	- page cannot be located at present
N
Nick Piggin 已提交
1549 1550 1551 1552
 * SWAP_MLOCK	- page is now mlocked.
 */
int try_to_munlock(struct page *page)
{
1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567
	int ret;
	struct rmap_walk_control rwc = {
		.rmap_one = try_to_unmap_one,
		.arg = (void *)TTU_MUNLOCK,
		.done = page_not_mapped,
		/*
		 * 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() and recover PG_mlocked lazily.
		 */
		.file_nonlinear = NULL,
		.anon_lock = page_lock_anon_vma_read,

	};

1568
	VM_BUG_ON_PAGE(!PageLocked(page) || PageLRU(page), page);
N
Nick Piggin 已提交
1569

1570 1571
	ret = rmap_walk(page, &rwc);
	return ret;
N
Nick Piggin 已提交
1572
}
1573

P
Peter Zijlstra 已提交
1574
void __put_anon_vma(struct anon_vma *anon_vma)
1575
{
P
Peter Zijlstra 已提交
1576
	struct anon_vma *root = anon_vma->root;
1577

1578
	anon_vma_free(anon_vma);
P
Peter Zijlstra 已提交
1579 1580
	if (root != anon_vma && atomic_dec_and_test(&root->refcount))
		anon_vma_free(root);
1581 1582
}

1583 1584
static struct anon_vma *rmap_walk_anon_lock(struct page *page,
					struct rmap_walk_control *rwc)
1585 1586 1587
{
	struct anon_vma *anon_vma;

1588 1589 1590
	if (rwc->anon_lock)
		return rwc->anon_lock(page);

1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601 1602 1603 1604
	/*
	 * Note: remove_migration_ptes() cannot use page_lock_anon_vma_read()
	 * because that depends on page_mapped(); but not all its usages
	 * are holding mmap_sem. Users without mmap_sem are required to
	 * take a reference count to prevent the anon_vma disappearing
	 */
	anon_vma = page_anon_vma(page);
	if (!anon_vma)
		return NULL;

	anon_vma_lock_read(anon_vma);
	return anon_vma;
}

1605
/*
1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617
 * rmap_walk_anon - do something to anonymous page using the object-based
 * rmap method
 * @page: the page to be handled
 * @rwc: control variable according to each walk type
 *
 * Find all the mappings of a page using the mapping pointer and the vma chains
 * contained in the anon_vma struct it points to.
 *
 * 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.
1618
 */
1619
static int rmap_walk_anon(struct page *page, struct rmap_walk_control *rwc)
1620 1621
{
	struct anon_vma *anon_vma;
1622
	pgoff_t pgoff = page_to_pgoff(page);
1623
	struct anon_vma_chain *avc;
1624 1625
	int ret = SWAP_AGAIN;

1626
	anon_vma = rmap_walk_anon_lock(page, rwc);
1627 1628
	if (!anon_vma)
		return ret;
1629

1630
	anon_vma_interval_tree_foreach(avc, &anon_vma->rb_root, pgoff, pgoff) {
1631
		struct vm_area_struct *vma = avc->vma;
1632
		unsigned long address = vma_address(page, vma);
1633 1634 1635 1636

		if (rwc->invalid_vma && rwc->invalid_vma(vma, rwc->arg))
			continue;

1637
		ret = rwc->rmap_one(page, vma, address, rwc->arg);
1638 1639
		if (ret != SWAP_AGAIN)
			break;
1640 1641
		if (rwc->done && rwc->done(page))
			break;
1642
	}
1643
	anon_vma_unlock_read(anon_vma);
1644 1645 1646
	return ret;
}

1647 1648 1649 1650 1651 1652 1653 1654 1655 1656 1657 1658 1659
/*
 * rmap_walk_file - do something to file page using the object-based rmap method
 * @page: the page to be handled
 * @rwc: control variable according to each walk type
 *
 * Find all the mappings of a page using the mapping pointer and the vma chains
 * contained in the address_space struct it points to.
 *
 * 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.
 */
1660
static int rmap_walk_file(struct page *page, struct rmap_walk_control *rwc)
1661 1662
{
	struct address_space *mapping = page->mapping;
1663
	pgoff_t pgoff = page_to_pgoff(page);
1664 1665 1666
	struct vm_area_struct *vma;
	int ret = SWAP_AGAIN;

1667 1668 1669 1670 1671 1672
	/*
	 * 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_mutex.
	 */
1673
	VM_BUG_ON_PAGE(!PageLocked(page), page);
1674

1675 1676
	if (!mapping)
		return ret;
1677
	mutex_lock(&mapping->i_mmap_mutex);
1678
	vma_interval_tree_foreach(vma, &mapping->i_mmap, pgoff, pgoff) {
1679
		unsigned long address = vma_address(page, vma);
1680 1681 1682 1683

		if (rwc->invalid_vma && rwc->invalid_vma(vma, rwc->arg))
			continue;

1684
		ret = rwc->rmap_one(page, vma, address, rwc->arg);
1685
		if (ret != SWAP_AGAIN)
1686 1687 1688
			goto done;
		if (rwc->done && rwc->done(page))
			goto done;
1689
	}
1690 1691 1692 1693 1694 1695 1696

	if (!rwc->file_nonlinear)
		goto done;

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

1697
	ret = rwc->file_nonlinear(page, mapping, rwc->arg);
1698 1699

done:
1700
	mutex_unlock(&mapping->i_mmap_mutex);
1701 1702 1703
	return ret;
}

1704
int rmap_walk(struct page *page, struct rmap_walk_control *rwc)
1705 1706
{
	if (unlikely(PageKsm(page)))
1707
		return rmap_walk_ksm(page, rwc);
1708
	else if (PageAnon(page))
1709
		return rmap_walk_anon(page, rwc);
1710
	else
1711
		return rmap_walk_file(page, rwc);
1712
}
1713

N
Naoya Horiguchi 已提交
1714
#ifdef CONFIG_HUGETLB_PAGE
1715 1716 1717 1718 1719 1720 1721 1722 1723
/*
 * The following three functions are for anonymous (private mapped) hugepages.
 * Unlike common anonymous pages, anonymous hugepages have no accounting code
 * and no lru code, because we handle hugepages differently from common pages.
 */
static void __hugepage_set_anon_rmap(struct page *page,
	struct vm_area_struct *vma, unsigned long address, int exclusive)
{
	struct anon_vma *anon_vma = vma->anon_vma;
1724

1725
	BUG_ON(!anon_vma);
1726 1727 1728 1729 1730 1731

	if (PageAnon(page))
		return;
	if (!exclusive)
		anon_vma = anon_vma->root;

1732 1733 1734 1735 1736 1737 1738 1739 1740 1741
	anon_vma = (void *) anon_vma + PAGE_MAPPING_ANON;
	page->mapping = (struct address_space *) anon_vma;
	page->index = linear_page_index(vma, address);
}

void hugepage_add_anon_rmap(struct page *page,
			    struct vm_area_struct *vma, unsigned long address)
{
	struct anon_vma *anon_vma = vma->anon_vma;
	int first;
1742 1743

	BUG_ON(!PageLocked(page));
1744
	BUG_ON(!anon_vma);
1745
	/* address might be in next vma when migration races vma_adjust */
1746 1747 1748 1749 1750 1751 1752 1753 1754 1755 1756 1757
	first = atomic_inc_and_test(&page->_mapcount);
	if (first)
		__hugepage_set_anon_rmap(page, vma, address, 0);
}

void hugepage_add_new_anon_rmap(struct page *page,
			struct vm_area_struct *vma, unsigned long address)
{
	BUG_ON(address < vma->vm_start || address >= vma->vm_end);
	atomic_set(&page->_mapcount, 0);
	__hugepage_set_anon_rmap(page, vma, address, 1);
}
N
Naoya Horiguchi 已提交
1758
#endif /* CONFIG_HUGETLB_PAGE */