rmap.c 48.6 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_rwsem
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;
277
	int error;
L
Linus Torvalds 已提交
278

279 280 281 282 283 284 285 286
	/* 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.
	 */
287 288 289
	error = anon_vma_clone(vma, pvma);
	if (error)
		return error;
290 291 292 293 294

	/* Then add our own anon_vma. */
	anon_vma = anon_vma_alloc();
	if (!anon_vma)
		goto out_error;
295
	avc = anon_vma_chain_alloc(GFP_KERNEL);
296 297
	if (!avc)
		goto out_error_free_anon_vma;
298 299 300 301 302 303

	/*
	 * 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;
304
	/*
P
Peter Zijlstra 已提交
305 306 307
	 * 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.
308 309
	 */
	get_anon_vma(anon_vma->root);
310 311
	/* Mark this anon_vma as the one where our new (COWed) pages go. */
	vma->anon_vma = anon_vma;
312
	anon_vma_lock_write(anon_vma);
313
	anon_vma_chain_link(vma, avc, anon_vma);
314
	anon_vma_unlock_write(anon_vma);
315 316 317 318

	return 0;

 out_error_free_anon_vma:
P
Peter Zijlstra 已提交
319
	put_anon_vma(anon_vma);
320
 out_error:
321
	unlink_anon_vmas(vma);
322
	return -ENOMEM;
L
Linus Torvalds 已提交
323 324
}

325 326 327
void unlink_anon_vmas(struct vm_area_struct *vma)
{
	struct anon_vma_chain *avc, *next;
328
	struct anon_vma *root = NULL;
329

330 331 332 333
	/*
	 * Unlink each anon_vma chained to the VMA.  This list is ordered
	 * from newest to oldest, ensuring the root anon_vma gets freed last.
	 */
334
	list_for_each_entry_safe(avc, next, &vma->anon_vma_chain, same_vma) {
335 336 337
		struct anon_vma *anon_vma = avc->anon_vma;

		root = lock_anon_vma_root(root, anon_vma);
338
		anon_vma_interval_tree_remove(avc, &anon_vma->rb_root);
339 340 341 342 343

		/*
		 * Leave empty anon_vmas on the list - we'll need
		 * to free them outside the lock.
		 */
344
		if (RB_EMPTY_ROOT(&anon_vma->rb_root))
345 346 347 348 349 350 351 352 353 354
			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()
355
	 * needing to write-acquire the anon_vma->root->rwsem.
356 357 358 359 360 361
	 */
	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);

362 363 364 365 366
		list_del(&avc->same_vma);
		anon_vma_chain_free(avc);
	}
}

367
static void anon_vma_ctor(void *data)
L
Linus Torvalds 已提交
368
{
C
Christoph Lameter 已提交
369
	struct anon_vma *anon_vma = data;
L
Linus Torvalds 已提交
370

371
	init_rwsem(&anon_vma->rwsem);
372
	atomic_set(&anon_vma->refcount, 0);
373
	anon_vma->rb_root = RB_ROOT;
L
Linus Torvalds 已提交
374 375 376 377 378
}

void __init anon_vma_init(void)
{
	anon_vma_cachep = kmem_cache_create("anon_vma", sizeof(struct anon_vma),
379
			0, SLAB_DESTROY_BY_RCU|SLAB_PANIC, anon_vma_ctor);
380
	anon_vma_chain_cachep = KMEM_CACHE(anon_vma_chain, SLAB_PANIC);
L
Linus Torvalds 已提交
381 382 383
}

/*
384 385 386 387 388 389 390 391 392
 * 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).
 *
393 394 395 396 397
 * 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 ].
 *
398 399 400 401 402 403 404
 * 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
405
 */
406
struct anon_vma *page_get_anon_vma(struct page *page)
L
Linus Torvalds 已提交
407
{
408
	struct anon_vma *anon_vma = NULL;
L
Linus Torvalds 已提交
409 410 411
	unsigned long anon_mapping;

	rcu_read_lock();
412
	anon_mapping = (unsigned long) ACCESS_ONCE(page->mapping);
H
Hugh Dickins 已提交
413
	if ((anon_mapping & PAGE_MAPPING_FLAGS) != PAGE_MAPPING_ANON)
L
Linus Torvalds 已提交
414 415 416 417 418
		goto out;
	if (!page_mapped(page))
		goto out;

	anon_vma = (struct anon_vma *) (anon_mapping - PAGE_MAPPING_ANON);
419 420 421 422
	if (!atomic_inc_not_zero(&anon_vma->refcount)) {
		anon_vma = NULL;
		goto out;
	}
423 424 425

	/*
	 * If this page is still mapped, then its anon_vma cannot have been
426 427 428 429
	 * 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).
430
	 */
431
	if (!page_mapped(page)) {
432
		rcu_read_unlock();
433
		put_anon_vma(anon_vma);
434
		return NULL;
435
	}
L
Linus Torvalds 已提交
436 437
out:
	rcu_read_unlock();
438 439 440 441

	return anon_vma;
}

442 443 444 445 446 447 448
/*
 * 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.
 */
449
struct anon_vma *page_lock_anon_vma_read(struct page *page)
450
{
451
	struct anon_vma *anon_vma = NULL;
452
	struct anon_vma *root_anon_vma;
453
	unsigned long anon_mapping;
454

455 456 457 458 459 460 461 462
	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);
463
	root_anon_vma = ACCESS_ONCE(anon_vma->root);
464
	if (down_read_trylock(&root_anon_vma->rwsem)) {
465
		/*
466 467
		 * If the page is still mapped, then this anon_vma is still
		 * its anon_vma, and holding the mutex ensures that it will
468
		 * not go away, see anon_vma_free().
469
		 */
470
		if (!page_mapped(page)) {
471
			up_read(&root_anon_vma->rwsem);
472 473 474 475
			anon_vma = NULL;
		}
		goto out;
	}
476

477 478 479 480 481 482 483
	/* trylock failed, we got to sleep */
	if (!atomic_inc_not_zero(&anon_vma->refcount)) {
		anon_vma = NULL;
		goto out;
	}

	if (!page_mapped(page)) {
484
		rcu_read_unlock();
485
		put_anon_vma(anon_vma);
486
		return NULL;
487 488 489 490
	}

	/* we pinned the anon_vma, its safe to sleep */
	rcu_read_unlock();
491
	anon_vma_lock_read(anon_vma);
492 493 494 495 496

	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
497
		 * we'll deadlock on the anon_vma_lock_write() recursion.
498
		 */
499
		anon_vma_unlock_read(anon_vma);
500 501 502 503 504 505 506 507
		__put_anon_vma(anon_vma);
		anon_vma = NULL;
	}

	return anon_vma;

out:
	rcu_read_unlock();
508
	return anon_vma;
509 510
}

511
void page_unlock_anon_vma_read(struct anon_vma *anon_vma)
512
{
513
	anon_vma_unlock_read(anon_vma);
L
Linus Torvalds 已提交
514 515 516
}

/*
517
 * At what user virtual address is page expected in @vma?
L
Linus Torvalds 已提交
518
 */
519 520
static inline unsigned long
__vma_address(struct page *page, struct vm_area_struct *vma)
L
Linus Torvalds 已提交
521
{
522
	pgoff_t pgoff = page_to_pgoff(page);
523 524 525 526 527 528 529 530 531
	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 */
532
	VM_BUG_ON_VMA(address < vma->vm_start || address >= vma->vm_end, vma);
533

L
Linus Torvalds 已提交
534 535 536 537
	return address;
}

/*
H
Huang Shijie 已提交
538
 * At what user virtual address is page expected in vma?
539
 * Caller should check the page is actually part of the vma.
L
Linus Torvalds 已提交
540 541 542
 */
unsigned long page_address_in_vma(struct page *page, struct vm_area_struct *vma)
{
543
	unsigned long address;
544
	if (PageAnon(page)) {
545 546 547 548 549 550 551
		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)
552 553
			return -EFAULT;
	} else if (page->mapping && !(vma->vm_flags & VM_NONLINEAR)) {
H
Hugh Dickins 已提交
554 555
		if (!vma->vm_file ||
		    vma->vm_file->f_mapping != page->mapping)
L
Linus Torvalds 已提交
556 557 558
			return -EFAULT;
	} else
		return -EFAULT;
559 560 561 562
	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
563 564
}

B
Bob Liu 已提交
565 566 567 568 569
pmd_t *mm_find_pmd(struct mm_struct *mm, unsigned long address)
{
	pgd_t *pgd;
	pud_t *pud;
	pmd_t *pmd = NULL;
570
	pmd_t pmde;
B
Bob Liu 已提交
571 572 573 574 575 576 577 578 579 580

	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);
581 582 583 584 585 586 587
	/*
	 * 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
588 589 590 591 592
		pmd = NULL;
out:
	return pmd;
}

N
Nikita Danilov 已提交
593 594 595
/*
 * Check that @page is mapped at @address into @mm.
 *
N
Nick Piggin 已提交
596 597 598 599
 * 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).
 *
600
 * On success returns with pte mapped and locked.
N
Nikita Danilov 已提交
601
 */
602
pte_t *__page_check_address(struct page *page, struct mm_struct *mm,
N
Nick Piggin 已提交
603
			  unsigned long address, spinlock_t **ptlp, int sync)
N
Nikita Danilov 已提交
604 605 606
{
	pmd_t *pmd;
	pte_t *pte;
H
Hugh Dickins 已提交
607
	spinlock_t *ptl;
N
Nikita Danilov 已提交
608

609
	if (unlikely(PageHuge(page))) {
610
		/* when pud is not present, pte will be NULL */
611
		pte = huge_pte_offset(mm, address);
612 613 614
		if (!pte)
			return NULL;

615
		ptl = huge_pte_lockptr(page_hstate(page), mm, pte);
616 617 618
		goto check;
	}

B
Bob Liu 已提交
619 620
	pmd = mm_find_pmd(mm, address);
	if (!pmd)
H
Hugh Dickins 已提交
621 622 623 624
		return NULL;

	pte = pte_offset_map(pmd, address);
	/* Make a quick check before getting the lock */
N
Nick Piggin 已提交
625
	if (!sync && !pte_present(*pte)) {
H
Hugh Dickins 已提交
626 627 628 629
		pte_unmap(pte);
		return NULL;
	}

H
Hugh Dickins 已提交
630
	ptl = pte_lockptr(mm, pmd);
631
check:
H
Hugh Dickins 已提交
632 633 634 635
	spin_lock(ptl);
	if (pte_present(*pte) && page_to_pfn(page) == pte_pfn(*pte)) {
		*ptlp = ptl;
		return pte;
N
Nikita Danilov 已提交
636
	}
H
Hugh Dickins 已提交
637 638
	pte_unmap_unlock(pte, ptl);
	return NULL;
N
Nikita Danilov 已提交
639 640
}

N
Nick Piggin 已提交
641 642 643 644 645 646 647 648 649
/**
 * 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.
 */
650
int page_mapped_in_vma(struct page *page, struct vm_area_struct *vma)
N
Nick Piggin 已提交
651 652 653 654 655
{
	unsigned long address;
	pte_t *pte;
	spinlock_t *ptl;

656 657
	address = __vma_address(page, vma);
	if (unlikely(address < vma->vm_start || address >= vma->vm_end))
N
Nick Piggin 已提交
658 659 660 661 662 663 664 665 666
		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;
}

667 668 669 670 671 672
struct page_referenced_arg {
	int mapcount;
	int referenced;
	unsigned long vm_flags;
	struct mem_cgroup *memcg;
};
L
Linus Torvalds 已提交
673
/*
674
 * arg: page_referenced_arg will be passed
L
Linus Torvalds 已提交
675
 */
676
static int page_referenced_one(struct page *page, struct vm_area_struct *vma,
677
			unsigned long address, void *arg)
L
Linus Torvalds 已提交
678 679
{
	struct mm_struct *mm = vma->vm_mm;
680
	spinlock_t *ptl;
L
Linus Torvalds 已提交
681
	int referenced = 0;
682
	struct page_referenced_arg *pra = arg;
L
Linus Torvalds 已提交
683

684 685 686
	if (unlikely(PageTransHuge(page))) {
		pmd_t *pmd;

687 688 689 690
		/*
		 * rmap might return false positives; we must filter
		 * these out using page_check_address_pmd().
		 */
691
		pmd = page_check_address_pmd(page, mm, address,
692 693
					     PAGE_CHECK_ADDRESS_PMD_FLAG, &ptl);
		if (!pmd)
694
			return SWAP_AGAIN;
695 696

		if (vma->vm_flags & VM_LOCKED) {
697
			spin_unlock(ptl);
698 699
			pra->vm_flags |= VM_LOCKED;
			return SWAP_FAIL; /* To break the loop */
700 701 702 703
		}

		/* go ahead even if the pmd is pmd_trans_splitting() */
		if (pmdp_clear_flush_young_notify(vma, address, pmd))
704
			referenced++;
705
		spin_unlock(ptl);
706 707 708
	} else {
		pte_t *pte;

709 710 711 712
		/*
		 * rmap might return false positives; we must filter
		 * these out using page_check_address().
		 */
713 714
		pte = page_check_address(page, mm, address, &ptl, 0);
		if (!pte)
715
			return SWAP_AGAIN;
716

717 718
		if (vma->vm_flags & VM_LOCKED) {
			pte_unmap_unlock(pte, ptl);
719 720
			pra->vm_flags |= VM_LOCKED;
			return SWAP_FAIL; /* To break the loop */
721 722
		}

723 724 725 726 727 728 729 730
		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.
			 */
731
			if (likely(!(vma->vm_flags & VM_SEQ_READ)))
732 733 734 735 736
				referenced++;
		}
		pte_unmap_unlock(pte, ptl);
	}

737 738 739
	if (referenced) {
		pra->referenced++;
		pra->vm_flags |= vma->vm_flags;
L
Linus Torvalds 已提交
740
	}
741

742 743 744 745 746
	pra->mapcount--;
	if (!pra->mapcount)
		return SWAP_SUCCESS; /* To break the loop */

	return SWAP_AGAIN;
L
Linus Torvalds 已提交
747 748
}

749
static bool invalid_page_referenced_vma(struct vm_area_struct *vma, void *arg)
L
Linus Torvalds 已提交
750
{
751 752
	struct page_referenced_arg *pra = arg;
	struct mem_cgroup *memcg = pra->memcg;
L
Linus Torvalds 已提交
753

754 755
	if (!mm_match_cgroup(vma->vm_mm, memcg))
		return true;
L
Linus Torvalds 已提交
756

757
	return false;
L
Linus Torvalds 已提交
758 759 760 761 762 763
}

/**
 * page_referenced - test if the page was referenced
 * @page: the page to test
 * @is_locked: caller holds lock on the page
764
 * @memcg: target memory cgroup
765
 * @vm_flags: collect encountered vma->vm_flags who actually referenced the page
L
Linus Torvalds 已提交
766 767 768 769
 *
 * Quick test_and_clear_referenced for all mappings to a page,
 * returns the number of ptes which referenced the page.
 */
770 771
int page_referenced(struct page *page,
		    int is_locked,
772
		    struct mem_cgroup *memcg,
773
		    unsigned long *vm_flags)
L
Linus Torvalds 已提交
774
{
775
	int ret;
H
Hugh Dickins 已提交
776
	int we_locked = 0;
777 778 779 780 781 782 783 784 785
	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
786

787
	*vm_flags = 0;
788 789 790 791 792 793 794 795 796 797
	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
798
	}
799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815

	/*
	 * 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
816 817
}

H
Hugh Dickins 已提交
818
static int page_mkclean_one(struct page *page, struct vm_area_struct *vma,
819
			    unsigned long address, void *arg)
820 821
{
	struct mm_struct *mm = vma->vm_mm;
822
	pte_t *pte;
823 824
	spinlock_t *ptl;
	int ret = 0;
825
	int *cleaned = arg;
826

N
Nick Piggin 已提交
827
	pte = page_check_address(page, mm, address, &ptl, 1);
828 829 830
	if (!pte)
		goto out;

831 832
	if (pte_dirty(*pte) || pte_write(*pte)) {
		pte_t entry;
833

834
		flush_cache_page(vma, address, pte_pfn(*pte));
835
		entry = ptep_clear_flush(vma, address, pte);
836 837
		entry = pte_wrprotect(entry);
		entry = pte_mkclean(entry);
838
		set_pte_at(mm, address, pte, entry);
839 840
		ret = 1;
	}
841 842

	pte_unmap_unlock(pte, ptl);
843

844
	if (ret) {
845
		mmu_notifier_invalidate_page(mm, address);
846 847
		(*cleaned)++;
	}
848
out:
849
	return SWAP_AGAIN;
850 851
}

852
static bool invalid_mkclean_vma(struct vm_area_struct *vma, void *arg)
853
{
854
	if (vma->vm_flags & VM_SHARED)
F
Fengguang Wu 已提交
855
		return false;
856

F
Fengguang Wu 已提交
857
	return true;
858 859 860 861
}

int page_mkclean(struct page *page)
{
862 863 864 865 866 867 868
	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,
	};
869 870 871

	BUG_ON(!PageLocked(page));

872 873 874 875 876 877 878 879
	if (!page_mapped(page))
		return 0;

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

	rmap_walk(page, &rwc);
880

881
	return cleaned;
882
}
J
Jaya Kumar 已提交
883
EXPORT_SYMBOL_GPL(page_mkclean);
884

885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900
/**
 * 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;

901
	VM_BUG_ON_PAGE(!PageLocked(page), page);
902
	VM_BUG_ON_VMA(!anon_vma, vma);
903
	VM_BUG_ON_PAGE(page->index != linear_page_index(vma, address), page);
904 905 906 907 908

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

N
Nick Piggin 已提交
909
/**
A
Andi Kleen 已提交
910 911 912 913
 * __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	
914
 * @exclusive:	the page is exclusively owned by the current process
N
Nick Piggin 已提交
915 916
 */
static void __page_set_anon_rmap(struct page *page,
917
	struct vm_area_struct *vma, unsigned long address, int exclusive)
N
Nick Piggin 已提交
918
{
919
	struct anon_vma *anon_vma = vma->anon_vma;
920

921
	BUG_ON(!anon_vma);
922

A
Andi Kleen 已提交
923 924 925
	if (PageAnon(page))
		return;

926
	/*
927 928 929
	 * If the page isn't exclusively mapped into this vma,
	 * we must use the _oldest_ possible anon_vma for the
	 * page mapping!
930
	 */
A
Andi Kleen 已提交
931
	if (!exclusive)
932
		anon_vma = anon_vma->root;
N
Nick Piggin 已提交
933 934 935 936 937 938

	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
939
/**
R
Randy Dunlap 已提交
940
 * __page_check_anon_rmap - sanity check anonymous rmap addition
N
Nick Piggin 已提交
941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960
 * @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
961
	BUG_ON(page_anon_vma(page)->root != vma->anon_vma->root);
N
Nick Piggin 已提交
962 963 964 965
	BUG_ON(page->index != linear_page_index(vma, address));
#endif
}

L
Linus Torvalds 已提交
966 967 968 969 970 971
/**
 * 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
972
 * The caller needs to hold the pte lock, and the page must be locked in
973 974 975
 * 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
976 977 978
 */
void page_add_anon_rmap(struct page *page,
	struct vm_area_struct *vma, unsigned long address)
979 980 981 982 983 984 985 986 987 988 989
{
	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
990
{
H
Hugh Dickins 已提交
991
	int first = atomic_inc_and_test(&page->_mapcount);
992
	if (first) {
993 994 995 996 997 998
		/*
		 * 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.
		 */
999
		if (PageTransHuge(page))
1000 1001
			__inc_zone_page_state(page,
					      NR_ANON_TRANSPARENT_HUGEPAGES);
1002 1003
		__mod_zone_page_state(page_zone(page), NR_ANON_PAGES,
				hpage_nr_pages(page));
1004
	}
H
Hugh Dickins 已提交
1005 1006 1007
	if (unlikely(PageKsm(page)))
		return;

1008
	VM_BUG_ON_PAGE(!PageLocked(page), page);
1009
	/* address might be in next vma when migration races vma_adjust */
H
Hugh Dickins 已提交
1010
	if (first)
1011
		__page_set_anon_rmap(page, vma, address, exclusive);
1012
	else
N
Nick Piggin 已提交
1013
		__page_check_anon_rmap(page, vma, address);
L
Linus Torvalds 已提交
1014 1015
}

R
Randy Dunlap 已提交
1016
/**
N
Nick Piggin 已提交
1017 1018 1019 1020 1021 1022 1023
 * 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
1024
 * Page does not have to be locked.
N
Nick Piggin 已提交
1025 1026 1027 1028
 */
void page_add_new_anon_rmap(struct page *page,
	struct vm_area_struct *vma, unsigned long address)
{
1029
	VM_BUG_ON_VMA(address < vma->vm_start || address >= vma->vm_end, vma);
1030 1031
	SetPageSwapBacked(page);
	atomic_set(&page->_mapcount, 0); /* increment count (starts at -1) */
1032
	if (PageTransHuge(page))
1033
		__inc_zone_page_state(page, NR_ANON_TRANSPARENT_HUGEPAGES);
1034 1035
	__mod_zone_page_state(page_zone(page), NR_ANON_PAGES,
			hpage_nr_pages(page));
1036
	__page_set_anon_rmap(page, vma, address, 1);
N
Nick Piggin 已提交
1037 1038
}

L
Linus Torvalds 已提交
1039 1040 1041 1042
/**
 * page_add_file_rmap - add pte mapping to a file page
 * @page: the page to add the mapping to
 *
1043
 * The caller needs to hold the pte lock.
L
Linus Torvalds 已提交
1044 1045 1046
 */
void page_add_file_rmap(struct page *page)
{
1047
	struct mem_cgroup *memcg;
1048
	unsigned long flags;
1049
	bool locked;
1050

1051
	memcg = mem_cgroup_begin_page_stat(page, &locked, &flags);
1052
	if (atomic_inc_and_test(&page->_mapcount)) {
1053
		__inc_zone_page_state(page, NR_FILE_MAPPED);
1054
		mem_cgroup_inc_page_stat(memcg, MEM_CGROUP_STAT_FILE_MAPPED);
1055
	}
1056
	mem_cgroup_end_page_stat(memcg, &locked, &flags);
L
Linus Torvalds 已提交
1057 1058
}

1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085
static void page_remove_file_rmap(struct page *page)
{
	struct mem_cgroup *memcg;
	unsigned long flags;
	bool locked;

	memcg = mem_cgroup_begin_page_stat(page, &locked, &flags);

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

	/* Hugepages are not counted in NR_FILE_MAPPED for now. */
	if (unlikely(PageHuge(page)))
		goto out;

	/*
	 * 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.
	 */
	__dec_zone_page_state(page, NR_FILE_MAPPED);
	mem_cgroup_dec_page_stat(memcg, MEM_CGROUP_STAT_FILE_MAPPED);

	if (unlikely(PageMlocked(page)))
		clear_page_mlock(page);
out:
1086
	mem_cgroup_end_page_stat(memcg, &locked, &flags);
1087 1088
}

L
Linus Torvalds 已提交
1089 1090 1091 1092
/**
 * page_remove_rmap - take down pte mapping from a page
 * @page: page to remove mapping from
 *
1093
 * The caller needs to hold the pte lock.
L
Linus Torvalds 已提交
1094
 */
1095
void page_remove_rmap(struct page *page)
L
Linus Torvalds 已提交
1096
{
1097 1098 1099 1100
	if (!PageAnon(page)) {
		page_remove_file_rmap(page);
		return;
	}
1101

1102 1103
	/* page still mapped by someone else? */
	if (!atomic_add_negative(-1, &page->_mapcount))
1104 1105 1106 1107 1108
		return;

	/* Hugepages are not counted in NR_ANON_PAGES for now. */
	if (unlikely(PageHuge(page)))
		return;
1109

1110
	/*
1111 1112 1113
	 * 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.
1114
	 */
1115 1116 1117 1118 1119 1120
	if (PageTransHuge(page))
		__dec_zone_page_state(page, NR_ANON_TRANSPARENT_HUGEPAGES);

	__mod_zone_page_state(page_zone(page), NR_ANON_PAGES,
			      -hpage_nr_pages(page));

1121 1122
	if (unlikely(PageMlocked(page)))
		clear_page_mlock(page);
1123

1124 1125 1126 1127 1128 1129 1130 1131 1132
	/*
	 * It would be tidy to reset the PageAnon mapping here,
	 * but that might overwrite a racing page_add_anon_rmap
	 * which increments mapcount after us but sets mapping
	 * before us: so leave the reset to free_hot_cold_page,
	 * and remember that it's only reliable while mapped.
	 * Leaving it set also helps swapoff to reinstate ptes
	 * faster for those pages still in swapcache.
	 */
L
Linus Torvalds 已提交
1133 1134 1135
}

/*
1136
 * @arg: enum ttu_flags will be passed to this argument
L
Linus Torvalds 已提交
1137
 */
1138
static int try_to_unmap_one(struct page *page, struct vm_area_struct *vma,
1139
		     unsigned long address, void *arg)
L
Linus Torvalds 已提交
1140 1141 1142 1143
{
	struct mm_struct *mm = vma->vm_mm;
	pte_t *pte;
	pte_t pteval;
H
Hugh Dickins 已提交
1144
	spinlock_t *ptl;
L
Linus Torvalds 已提交
1145
	int ret = SWAP_AGAIN;
1146
	enum ttu_flags flags = (enum ttu_flags)arg;
L
Linus Torvalds 已提交
1147

N
Nick Piggin 已提交
1148
	pte = page_check_address(page, mm, address, &ptl, 0);
H
Hugh Dickins 已提交
1149
	if (!pte)
N
Nikita Danilov 已提交
1150
		goto out;
L
Linus Torvalds 已提交
1151 1152 1153 1154 1155 1156

	/*
	 * 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.
	 */
1157
	if (!(flags & TTU_IGNORE_MLOCK)) {
K
KOSAKI Motohiro 已提交
1158 1159 1160
		if (vma->vm_flags & VM_LOCKED)
			goto out_mlock;

1161
		if (flags & TTU_MUNLOCK)
H
Hugh Dickins 已提交
1162
			goto out_unmap;
1163 1164
	}
	if (!(flags & TTU_IGNORE_ACCESS)) {
N
Nick Piggin 已提交
1165 1166 1167 1168 1169
		if (ptep_clear_flush_young_notify(vma, address, pte)) {
			ret = SWAP_FAIL;
			goto out_unmap;
		}
  	}
L
Linus Torvalds 已提交
1170 1171 1172

	/* Nuke the page table entry. */
	flush_cache_page(vma, address, page_to_pfn(page));
1173
	pteval = ptep_clear_flush(vma, address, pte);
L
Linus Torvalds 已提交
1174 1175 1176 1177 1178

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

1179 1180 1181
	/* Update high watermark before we lower rss */
	update_hiwater_rss(mm);

1182
	if (PageHWPoison(page) && !(flags & TTU_IGNORE_HWPOISON)) {
1183 1184 1185 1186 1187 1188
		if (!PageHuge(page)) {
			if (PageAnon(page))
				dec_mm_counter(mm, MM_ANONPAGES);
			else
				dec_mm_counter(mm, MM_FILEPAGES);
		}
1189
		set_pte_at(mm, address, pte,
1190
			   swp_entry_to_pte(make_hwpoison_entry(page)));
1191 1192 1193 1194 1195 1196 1197 1198 1199 1200
	} 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);
1201
	} else if (PageAnon(page)) {
H
Hugh Dickins 已提交
1202
		swp_entry_t entry = { .val = page_private(page) };
1203
		pte_t swp_pte;
1204 1205 1206 1207 1208 1209

		if (PageSwapCache(page)) {
			/*
			 * Store the swap location in the pte.
			 * See handle_pte_fault() ...
			 */
H
Hugh Dickins 已提交
1210 1211 1212 1213 1214
			if (swap_duplicate(entry) < 0) {
				set_pte_at(mm, address, pte, pteval);
				ret = SWAP_FAIL;
				goto out_unmap;
			}
1215 1216 1217 1218 1219 1220
			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
1221
			dec_mm_counter(mm, MM_ANONPAGES);
K
KAMEZAWA Hiroyuki 已提交
1222
			inc_mm_counter(mm, MM_SWAPENTS);
1223
		} else if (IS_ENABLED(CONFIG_MIGRATION)) {
1224 1225 1226 1227 1228
			/*
			 * 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.
			 */
1229
			BUG_ON(!(flags & TTU_MIGRATION));
1230
			entry = make_migration_entry(page, pte_write(pteval));
L
Linus Torvalds 已提交
1231
		}
1232 1233 1234 1235
		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
1236
		BUG_ON(pte_file(*pte));
1237
	} else if (IS_ENABLED(CONFIG_MIGRATION) &&
1238
		   (flags & TTU_MIGRATION)) {
1239 1240 1241 1242 1243
		/* 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
1244
		dec_mm_counter(mm, MM_FILEPAGES);
L
Linus Torvalds 已提交
1245

1246
	page_remove_rmap(page);
L
Linus Torvalds 已提交
1247 1248 1249
	page_cache_release(page);

out_unmap:
H
Hugh Dickins 已提交
1250
	pte_unmap_unlock(pte, ptl);
1251
	if (ret != SWAP_FAIL && !(flags & TTU_MUNLOCK))
1252
		mmu_notifier_invalidate_page(mm, address);
K
KOSAKI Motohiro 已提交
1253 1254
out:
	return ret;
H
Hugh Dickins 已提交
1255

K
KOSAKI Motohiro 已提交
1256 1257 1258 1259 1260 1261 1262
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
1263
	 * we now hold anon_vma->rwsem or mapping->i_mmap_rwsem.
K
KOSAKI Motohiro 已提交
1264 1265 1266 1267 1268 1269 1270 1271
	 * 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
1272
		}
K
KOSAKI Motohiro 已提交
1273
		up_read(&vma->vm_mm->mmap_sem);
H
Hugh Dickins 已提交
1274
	}
L
Linus Torvalds 已提交
1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295
	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
1296 1297 1298 1299 1300
 *
 * 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
1301 1302 1303 1304
 */
#define CLUSTER_SIZE	min(32*PAGE_SIZE, PMD_SIZE)
#define CLUSTER_MASK	(~(CLUSTER_SIZE - 1))

N
Nick Piggin 已提交
1305 1306
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
1307 1308 1309
{
	struct mm_struct *mm = vma->vm_mm;
	pmd_t *pmd;
H
Hugh Dickins 已提交
1310
	pte_t *pte;
L
Linus Torvalds 已提交
1311
	pte_t pteval;
H
Hugh Dickins 已提交
1312
	spinlock_t *ptl;
L
Linus Torvalds 已提交
1313 1314
	struct page *page;
	unsigned long address;
1315 1316
	unsigned long mmun_start;	/* For mmu_notifiers */
	unsigned long mmun_end;		/* For mmu_notifiers */
L
Linus Torvalds 已提交
1317
	unsigned long end;
N
Nick Piggin 已提交
1318 1319
	int ret = SWAP_AGAIN;
	int locked_vma = 0;
L
Linus Torvalds 已提交
1320 1321 1322 1323 1324 1325 1326 1327

	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
1328 1329
	pmd = mm_find_pmd(mm, address);
	if (!pmd)
N
Nick Piggin 已提交
1330 1331
		return ret;

1332 1333 1334 1335
	mmun_start = address;
	mmun_end   = end;
	mmu_notifier_invalidate_range_start(mm, mmun_start, mmun_end);

N
Nick Piggin 已提交
1336
	/*
H
Hugh Dickins 已提交
1337
	 * If we can acquire the mmap_sem for read, and vma is VM_LOCKED,
N
Nick Piggin 已提交
1338 1339
	 * keep the sem while scanning the cluster for mlocking pages.
	 */
H
Hugh Dickins 已提交
1340
	if (down_read_trylock(&vma->vm_mm->mmap_sem)) {
N
Nick Piggin 已提交
1341 1342 1343 1344
		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
1345 1346

	pte = pte_offset_map_lock(mm, pmd, address, &ptl);
L
Linus Torvalds 已提交
1347

1348 1349 1350
	/* Update high watermark before we lower rss */
	update_hiwater_rss(mm);

H
Hugh Dickins 已提交
1351
	for (; address < end; pte++, address += PAGE_SIZE) {
L
Linus Torvalds 已提交
1352 1353
		if (!pte_present(*pte))
			continue;
1354 1355
		page = vm_normal_page(vma, address, *pte);
		BUG_ON(!page || PageAnon(page));
L
Linus Torvalds 已提交
1356

N
Nick Piggin 已提交
1357
		if (locked_vma) {
1358 1359 1360
			if (page == check_page) {
				/* we know we have check_page locked */
				mlock_vma_page(page);
N
Nick Piggin 已提交
1361
				ret = SWAP_MLOCK;
1362 1363 1364 1365 1366 1367 1368 1369 1370
			} 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
1371 1372 1373
			continue;	/* don't unmap */
		}

A
Andres Lagar-Cavilla 已提交
1374 1375 1376 1377 1378
		/*
		 * 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
1379 1380 1381
			continue;

		/* Nuke the page table entry. */
1382
		flush_cache_page(vma, address, pte_pfn(*pte));
1383
		pteval = ptep_clear_flush(vma, address, pte);
L
Linus Torvalds 已提交
1384 1385

		/* If nonlinear, store the file page offset in the pte. */
1386 1387 1388
		if (page->index != linear_page_index(vma, address)) {
			pte_t ptfile = pgoff_to_pte(page->index);
			if (pte_soft_dirty(pteval))
1389
				ptfile = pte_file_mksoft_dirty(ptfile);
1390 1391
			set_pte_at(mm, address, pte, ptfile);
		}
L
Linus Torvalds 已提交
1392 1393 1394 1395 1396

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

1397
		page_remove_rmap(page);
L
Linus Torvalds 已提交
1398
		page_cache_release(page);
K
KAMEZAWA Hiroyuki 已提交
1399
		dec_mm_counter(mm, MM_FILEPAGES);
L
Linus Torvalds 已提交
1400 1401
		(*mapcount)--;
	}
H
Hugh Dickins 已提交
1402
	pte_unmap_unlock(pte - 1, ptl);
1403
	mmu_notifier_invalidate_range_end(mm, mmun_start, mmun_end);
N
Nick Piggin 已提交
1404 1405 1406
	if (locked_vma)
		up_read(&vma->vm_mm->mmap_sem);
	return ret;
L
Linus Torvalds 已提交
1407 1408
}

1409
static int try_to_unmap_nonlinear(struct page *page,
1410
		struct address_space *mapping, void *arg)
1411
{
1412
	struct vm_area_struct *vma;
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 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482
	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;
}

1483
bool is_vma_temporary_stack(struct vm_area_struct *vma)
1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496
{
	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;
}

1497 1498 1499 1500 1501 1502 1503 1504 1505 1506
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
1507 1508 1509
/**
 * try_to_unmap - try to remove all page table mappings to a page
 * @page: the page to get unmapped
1510
 * @flags: action and flags
L
Linus Torvalds 已提交
1511 1512 1513 1514 1515 1516 1517 1518
 *
 * 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
1519
 * SWAP_MLOCK	- page is mlocked.
L
Linus Torvalds 已提交
1520
 */
1521
int try_to_unmap(struct page *page, enum ttu_flags flags)
L
Linus Torvalds 已提交
1522 1523
{
	int ret;
1524 1525 1526 1527 1528 1529 1530
	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
1531

1532
	VM_BUG_ON_PAGE(!PageHuge(page) && PageTransHuge(page), page);
L
Linus Torvalds 已提交
1533

1534 1535 1536 1537 1538 1539 1540 1541
	/*
	 * 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.
	 */
1542
	if ((flags & TTU_MIGRATION) && !PageKsm(page) && PageAnon(page))
1543 1544 1545 1546
		rwc.invalid_vma = invalid_migration_vma;

	ret = rmap_walk(page, &rwc);

N
Nick Piggin 已提交
1547
	if (ret != SWAP_MLOCK && !page_mapped(page))
L
Linus Torvalds 已提交
1548 1549 1550
		ret = SWAP_SUCCESS;
	return ret;
}
N
Nikita Danilov 已提交
1551

N
Nick Piggin 已提交
1552 1553 1554 1555 1556 1557 1558 1559 1560 1561
/**
 * 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
1562
 * SWAP_AGAIN	- no vma is holding page mlocked, or,
N
Nick Piggin 已提交
1563
 * SWAP_AGAIN	- page mapped in mlocked vma -- couldn't acquire mmap sem
H
Hugh Dickins 已提交
1564
 * SWAP_FAIL	- page cannot be located at present
N
Nick Piggin 已提交
1565 1566 1567 1568
 * SWAP_MLOCK	- page is now mlocked.
 */
int try_to_munlock(struct page *page)
{
1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583
	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,

	};

1584
	VM_BUG_ON_PAGE(!PageLocked(page) || PageLRU(page), page);
N
Nick Piggin 已提交
1585

1586 1587
	ret = rmap_walk(page, &rwc);
	return ret;
N
Nick Piggin 已提交
1588
}
1589

P
Peter Zijlstra 已提交
1590
void __put_anon_vma(struct anon_vma *anon_vma)
1591
{
P
Peter Zijlstra 已提交
1592
	struct anon_vma *root = anon_vma->root;
1593

1594
	anon_vma_free(anon_vma);
P
Peter Zijlstra 已提交
1595 1596
	if (root != anon_vma && atomic_dec_and_test(&root->refcount))
		anon_vma_free(root);
1597 1598
}

1599 1600
static struct anon_vma *rmap_walk_anon_lock(struct page *page,
					struct rmap_walk_control *rwc)
1601 1602 1603
{
	struct anon_vma *anon_vma;

1604 1605 1606
	if (rwc->anon_lock)
		return rwc->anon_lock(page);

1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620
	/*
	 * 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;
}

1621
/*
1622 1623 1624 1625 1626 1627 1628 1629 1630 1631 1632 1633
 * 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.
1634
 */
1635
static int rmap_walk_anon(struct page *page, struct rmap_walk_control *rwc)
1636 1637
{
	struct anon_vma *anon_vma;
1638
	pgoff_t pgoff = page_to_pgoff(page);
1639
	struct anon_vma_chain *avc;
1640 1641
	int ret = SWAP_AGAIN;

1642
	anon_vma = rmap_walk_anon_lock(page, rwc);
1643 1644
	if (!anon_vma)
		return ret;
1645

1646
	anon_vma_interval_tree_foreach(avc, &anon_vma->rb_root, pgoff, pgoff) {
1647
		struct vm_area_struct *vma = avc->vma;
1648
		unsigned long address = vma_address(page, vma);
1649 1650 1651 1652

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

1653
		ret = rwc->rmap_one(page, vma, address, rwc->arg);
1654 1655
		if (ret != SWAP_AGAIN)
			break;
1656 1657
		if (rwc->done && rwc->done(page))
			break;
1658
	}
1659
	anon_vma_unlock_read(anon_vma);
1660 1661 1662
	return ret;
}

1663 1664 1665 1666 1667 1668 1669 1670 1671 1672 1673 1674 1675
/*
 * 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.
 */
1676
static int rmap_walk_file(struct page *page, struct rmap_walk_control *rwc)
1677 1678
{
	struct address_space *mapping = page->mapping;
1679
	pgoff_t pgoff = page_to_pgoff(page);
1680 1681 1682
	struct vm_area_struct *vma;
	int ret = SWAP_AGAIN;

1683 1684 1685 1686
	/*
	 * 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,
1687
	 * so we can safely take mapping->i_mmap_rwsem.
1688
	 */
1689
	VM_BUG_ON_PAGE(!PageLocked(page), page);
1690

1691 1692
	if (!mapping)
		return ret;
D
Davidlohr Bueso 已提交
1693 1694

	i_mmap_lock_read(mapping);
1695
	vma_interval_tree_foreach(vma, &mapping->i_mmap, pgoff, pgoff) {
1696
		unsigned long address = vma_address(page, vma);
1697 1698 1699 1700

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

1701
		ret = rwc->rmap_one(page, vma, address, rwc->arg);
1702
		if (ret != SWAP_AGAIN)
1703 1704 1705
			goto done;
		if (rwc->done && rwc->done(page))
			goto done;
1706
	}
1707 1708 1709 1710 1711 1712 1713

	if (!rwc->file_nonlinear)
		goto done;

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

1714
	ret = rwc->file_nonlinear(page, mapping, rwc->arg);
1715
done:
D
Davidlohr Bueso 已提交
1716
	i_mmap_unlock_read(mapping);
1717 1718 1719
	return ret;
}

1720
int rmap_walk(struct page *page, struct rmap_walk_control *rwc)
1721 1722
{
	if (unlikely(PageKsm(page)))
1723
		return rmap_walk_ksm(page, rwc);
1724
	else if (PageAnon(page))
1725
		return rmap_walk_anon(page, rwc);
1726
	else
1727
		return rmap_walk_file(page, rwc);
1728
}
1729

N
Naoya Horiguchi 已提交
1730
#ifdef CONFIG_HUGETLB_PAGE
1731 1732 1733 1734 1735 1736 1737 1738 1739
/*
 * 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;
1740

1741
	BUG_ON(!anon_vma);
1742 1743 1744 1745 1746 1747

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

1748 1749 1750 1751 1752 1753 1754 1755 1756 1757
	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;
1758 1759

	BUG_ON(!PageLocked(page));
1760
	BUG_ON(!anon_vma);
1761
	/* address might be in next vma when migration races vma_adjust */
1762 1763 1764 1765 1766 1767 1768 1769 1770 1771 1772 1773
	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
1774
#endif /* CONFIG_HUGETLB_PAGE */