rmap.c 43.6 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 34
 *                 mem_cgroup_{begin,end}_page_stat (memcg->move_lock)
 *                   mapping->tree_lock (widely used)
35
 *               inode->i_lock (in set_page_dirty's __mark_inode_dirty)
36
 *               bdi.wb->list_lock (in set_page_dirty's __mark_inode_dirty)
37 38 39
 *                 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,
40
 *                           within bdi.wb->list_lock in __sync_single_inode)
41
 *
42
 * anon_vma->rwsem,mapping->i_mutex      (memory_failure, collect_procs_anon)
43
 *   ->tasklist_lock
44
 *     pte map lock
L
Linus Torvalds 已提交
45 46 47 48 49 50 51 52
 */

#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
53
#include <linux/ksm.h>
L
Linus Torvalds 已提交
54 55
#include <linux/rmap.h>
#include <linux/rcupdate.h>
56
#include <linux/export.h>
57
#include <linux/memcontrol.h>
A
Andrea Arcangeli 已提交
58
#include <linux/mmu_notifier.h>
59
#include <linux/migrate.h>
60
#include <linux/hugetlb.h>
61
#include <linux/backing-dev.h>
L
Linus Torvalds 已提交
62 63 64

#include <asm/tlbflush.h>

N
Nick Piggin 已提交
65 66
#include "internal.h"

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

static inline struct anon_vma *anon_vma_alloc(void)
{
P
Peter Zijlstra 已提交
72 73 74 75 76
	struct anon_vma *anon_vma;

	anon_vma = kmem_cache_alloc(anon_vma_cachep, GFP_KERNEL);
	if (anon_vma) {
		atomic_set(&anon_vma->refcount, 1);
77 78
		anon_vma->degree = 1;	/* Reference for first vma */
		anon_vma->parent = anon_vma;
P
Peter Zijlstra 已提交
79 80 81 82 83 84 85 86
		/*
		 * 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;
87 88
}

P
Peter Zijlstra 已提交
89
static inline void anon_vma_free(struct anon_vma *anon_vma)
90
{
P
Peter Zijlstra 已提交
91
	VM_BUG_ON(atomic_read(&anon_vma->refcount));
92 93

	/*
94
	 * Synchronize against page_lock_anon_vma_read() such that
95 96 97 98 99
	 * 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
100
	 * down_read_trylock() from page_lock_anon_vma_read(). This orders:
101
	 *
102 103
	 * page_lock_anon_vma_read()	VS	put_anon_vma()
	 *   down_read_trylock()		  atomic_dec_and_test()
104
	 *   LOCK				  MB
105
	 *   atomic_read()			  rwsem_is_locked()
106 107 108 109
	 *
	 * LOCK should suffice since the actual taking of the lock must
	 * happen _before_ what follows.
	 */
110
	might_sleep();
111
	if (rwsem_is_locked(&anon_vma->root->rwsem)) {
112
		anon_vma_lock_write(anon_vma);
113
		anon_vma_unlock_write(anon_vma);
114 115
	}

116 117
	kmem_cache_free(anon_vma_cachep, anon_vma);
}
L
Linus Torvalds 已提交
118

119
static inline struct anon_vma_chain *anon_vma_chain_alloc(gfp_t gfp)
120
{
121
	return kmem_cache_alloc(anon_vma_chain_cachep, gfp);
122 123
}

124
static void anon_vma_chain_free(struct anon_vma_chain *anon_vma_chain)
125 126 127 128
{
	kmem_cache_free(anon_vma_chain_cachep, anon_vma_chain);
}

129 130 131 132 133 134 135
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);
136
	anon_vma_interval_tree_insert(avc, &anon_vma->rb_root);
137 138
}

139 140 141 142 143 144 145 146 147
/**
 * 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
148
 * not we either need to find an adjacent mapping that we
149 150 151 152 153
 * 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
154
 * optimistically looked up an anon_vma in page_lock_anon_vma_read()
155 156 157 158 159 160 161 162 163 164 165
 * 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
166 167 168
int anon_vma_prepare(struct vm_area_struct *vma)
{
	struct anon_vma *anon_vma = vma->anon_vma;
169
	struct anon_vma_chain *avc;
L
Linus Torvalds 已提交
170 171 172 173

	might_sleep();
	if (unlikely(!anon_vma)) {
		struct mm_struct *mm = vma->vm_mm;
174
		struct anon_vma *allocated;
L
Linus Torvalds 已提交
175

176
		avc = anon_vma_chain_alloc(GFP_KERNEL);
177 178 179
		if (!avc)
			goto out_enomem;

L
Linus Torvalds 已提交
180
		anon_vma = find_mergeable_anon_vma(vma);
181 182
		allocated = NULL;
		if (!anon_vma) {
L
Linus Torvalds 已提交
183 184
			anon_vma = anon_vma_alloc();
			if (unlikely(!anon_vma))
185
				goto out_enomem_free_avc;
L
Linus Torvalds 已提交
186 187 188
			allocated = anon_vma;
		}

189
		anon_vma_lock_write(anon_vma);
L
Linus Torvalds 已提交
190 191 192 193
		/* page_table_lock to protect against threads */
		spin_lock(&mm->page_table_lock);
		if (likely(!vma->anon_vma)) {
			vma->anon_vma = anon_vma;
194
			anon_vma_chain_link(vma, avc, anon_vma);
195 196
			/* vma reference or self-parent link for new root */
			anon_vma->degree++;
L
Linus Torvalds 已提交
197
			allocated = NULL;
198
			avc = NULL;
L
Linus Torvalds 已提交
199 200
		}
		spin_unlock(&mm->page_table_lock);
201
		anon_vma_unlock_write(anon_vma);
202 203

		if (unlikely(allocated))
P
Peter Zijlstra 已提交
204
			put_anon_vma(allocated);
205
		if (unlikely(avc))
206
			anon_vma_chain_free(avc);
L
Linus Torvalds 已提交
207 208
	}
	return 0;
209 210 211 212 213

 out_enomem_free_avc:
	anon_vma_chain_free(avc);
 out_enomem:
	return -ENOMEM;
L
Linus Torvalds 已提交
214 215
}

216 217 218 219 220 221 222 223 224 225 226 227 228
/*
 * 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))
229
			up_write(&root->rwsem);
230
		root = new_root;
231
		down_write(&root->rwsem);
232 233 234 235 236 237 238
	}
	return root;
}

static inline void unlock_anon_vma_root(struct anon_vma *root)
{
	if (root)
239
		up_write(&root->rwsem);
240 241
}

242 243 244
/*
 * Attach the anon_vmas from src to dst.
 * Returns 0 on success, -ENOMEM on failure.
245 246 247 248 249 250 251 252
 *
 * If dst->anon_vma is NULL this function tries to find and reuse existing
 * anon_vma which has no vmas and only one child anon_vma. This prevents
 * degradation of anon_vma hierarchy to endless linear chain in case of
 * constantly forking task. On the other hand, an anon_vma with more than one
 * child isn't reused even if there was no alive vma, thus rmap walker has a
 * good chance of avoiding scanning the whole hierarchy when it searches where
 * page is mapped.
253 254
 */
int anon_vma_clone(struct vm_area_struct *dst, struct vm_area_struct *src)
L
Linus Torvalds 已提交
255
{
256
	struct anon_vma_chain *avc, *pavc;
257
	struct anon_vma *root = NULL;
258

259
	list_for_each_entry_reverse(pavc, &src->anon_vma_chain, same_vma) {
260 261
		struct anon_vma *anon_vma;

262 263 264 265 266 267 268 269
		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;
		}
270 271 272
		anon_vma = pavc->anon_vma;
		root = lock_anon_vma_root(root, anon_vma);
		anon_vma_chain_link(dst, avc, anon_vma);
273 274 275 276 277 278 279 280 281 282 283 284

		/*
		 * Reuse existing anon_vma if its degree lower than two,
		 * that means it has no vma and only one anon_vma child.
		 *
		 * Do not chose parent anon_vma, otherwise first child
		 * will always reuse it. Root anon_vma is never reused:
		 * it has self-parent reference and at least one child.
		 */
		if (!dst->anon_vma && anon_vma != src->anon_vma &&
				anon_vma->degree < 2)
			dst->anon_vma = anon_vma;
285
	}
286 287
	if (dst->anon_vma)
		dst->anon_vma->degree++;
288
	unlock_anon_vma_root(root);
289
	return 0;
L
Linus Torvalds 已提交
290

291
 enomem_failure:
292 293 294 295 296 297 298
	/*
	 * dst->anon_vma is dropped here otherwise its degree can be incorrectly
	 * decremented in unlink_anon_vmas().
	 * We can safely do this because callers of anon_vma_clone() don't care
	 * about dst->anon_vma if anon_vma_clone() failed.
	 */
	dst->anon_vma = NULL;
299 300
	unlink_anon_vmas(dst);
	return -ENOMEM;
L
Linus Torvalds 已提交
301 302
}

303 304 305 306 307 308
/*
 * 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
309
{
310 311
	struct anon_vma_chain *avc;
	struct anon_vma *anon_vma;
312
	int error;
L
Linus Torvalds 已提交
313

314 315 316 317
	/* Don't bother if the parent process has no anon_vma here. */
	if (!pvma->anon_vma)
		return 0;

318 319 320
	/* Drop inherited anon_vma, we'll reuse existing or allocate new. */
	vma->anon_vma = NULL;

321 322 323 324
	/*
	 * First, attach the new VMA to the parent VMA's anon_vmas,
	 * so rmap can find non-COWed pages in child processes.
	 */
325 326 327
	error = anon_vma_clone(vma, pvma);
	if (error)
		return error;
328

329 330 331 332
	/* An existing anon_vma has been reused, all done then. */
	if (vma->anon_vma)
		return 0;

333 334 335 336
	/* Then add our own anon_vma. */
	anon_vma = anon_vma_alloc();
	if (!anon_vma)
		goto out_error;
337
	avc = anon_vma_chain_alloc(GFP_KERNEL);
338 339
	if (!avc)
		goto out_error_free_anon_vma;
340 341 342 343 344 345

	/*
	 * 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;
346
	anon_vma->parent = pvma->anon_vma;
347
	/*
P
Peter Zijlstra 已提交
348 349 350
	 * 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.
351 352
	 */
	get_anon_vma(anon_vma->root);
353 354
	/* Mark this anon_vma as the one where our new (COWed) pages go. */
	vma->anon_vma = anon_vma;
355
	anon_vma_lock_write(anon_vma);
356
	anon_vma_chain_link(vma, avc, anon_vma);
357
	anon_vma->parent->degree++;
358
	anon_vma_unlock_write(anon_vma);
359 360 361 362

	return 0;

 out_error_free_anon_vma:
P
Peter Zijlstra 已提交
363
	put_anon_vma(anon_vma);
364
 out_error:
365
	unlink_anon_vmas(vma);
366
	return -ENOMEM;
L
Linus Torvalds 已提交
367 368
}

369 370 371
void unlink_anon_vmas(struct vm_area_struct *vma)
{
	struct anon_vma_chain *avc, *next;
372
	struct anon_vma *root = NULL;
373

374 375 376 377
	/*
	 * Unlink each anon_vma chained to the VMA.  This list is ordered
	 * from newest to oldest, ensuring the root anon_vma gets freed last.
	 */
378
	list_for_each_entry_safe(avc, next, &vma->anon_vma_chain, same_vma) {
379 380 381
		struct anon_vma *anon_vma = avc->anon_vma;

		root = lock_anon_vma_root(root, anon_vma);
382
		anon_vma_interval_tree_remove(avc, &anon_vma->rb_root);
383 384 385 386 387

		/*
		 * Leave empty anon_vmas on the list - we'll need
		 * to free them outside the lock.
		 */
388 389
		if (RB_EMPTY_ROOT(&anon_vma->rb_root)) {
			anon_vma->parent->degree--;
390
			continue;
391
		}
392 393 394 395

		list_del(&avc->same_vma);
		anon_vma_chain_free(avc);
	}
396 397
	if (vma->anon_vma)
		vma->anon_vma->degree--;
398 399 400 401 402
	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()
403
	 * needing to write-acquire the anon_vma->root->rwsem.
404 405 406 407
	 */
	list_for_each_entry_safe(avc, next, &vma->anon_vma_chain, same_vma) {
		struct anon_vma *anon_vma = avc->anon_vma;

408
		BUG_ON(anon_vma->degree);
409 410
		put_anon_vma(anon_vma);

411 412 413 414 415
		list_del(&avc->same_vma);
		anon_vma_chain_free(avc);
	}
}

416
static void anon_vma_ctor(void *data)
L
Linus Torvalds 已提交
417
{
C
Christoph Lameter 已提交
418
	struct anon_vma *anon_vma = data;
L
Linus Torvalds 已提交
419

420
	init_rwsem(&anon_vma->rwsem);
421
	atomic_set(&anon_vma->refcount, 0);
422
	anon_vma->rb_root = RB_ROOT;
L
Linus Torvalds 已提交
423 424 425 426 427
}

void __init anon_vma_init(void)
{
	anon_vma_cachep = kmem_cache_create("anon_vma", sizeof(struct anon_vma),
428
			0, SLAB_DESTROY_BY_RCU|SLAB_PANIC, anon_vma_ctor);
429
	anon_vma_chain_cachep = KMEM_CACHE(anon_vma_chain, SLAB_PANIC);
L
Linus Torvalds 已提交
430 431 432
}

/*
433 434 435 436 437 438 439 440 441
 * 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).
 *
442 443 444 445 446
 * 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 ].
 *
447 448 449 450 451 452 453
 * 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
454
 */
455
struct anon_vma *page_get_anon_vma(struct page *page)
L
Linus Torvalds 已提交
456
{
457
	struct anon_vma *anon_vma = NULL;
L
Linus Torvalds 已提交
458 459 460
	unsigned long anon_mapping;

	rcu_read_lock();
461
	anon_mapping = (unsigned long)READ_ONCE(page->mapping);
H
Hugh Dickins 已提交
462
	if ((anon_mapping & PAGE_MAPPING_FLAGS) != PAGE_MAPPING_ANON)
L
Linus Torvalds 已提交
463 464 465 466 467
		goto out;
	if (!page_mapped(page))
		goto out;

	anon_vma = (struct anon_vma *) (anon_mapping - PAGE_MAPPING_ANON);
468 469 470 471
	if (!atomic_inc_not_zero(&anon_vma->refcount)) {
		anon_vma = NULL;
		goto out;
	}
472 473 474

	/*
	 * If this page is still mapped, then its anon_vma cannot have been
475 476 477 478
	 * 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).
479
	 */
480
	if (!page_mapped(page)) {
481
		rcu_read_unlock();
482
		put_anon_vma(anon_vma);
483
		return NULL;
484
	}
L
Linus Torvalds 已提交
485 486
out:
	rcu_read_unlock();
487 488 489 490

	return anon_vma;
}

491 492 493 494 495 496 497
/*
 * 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.
 */
498
struct anon_vma *page_lock_anon_vma_read(struct page *page)
499
{
500
	struct anon_vma *anon_vma = NULL;
501
	struct anon_vma *root_anon_vma;
502
	unsigned long anon_mapping;
503

504
	rcu_read_lock();
505
	anon_mapping = (unsigned long)READ_ONCE(page->mapping);
506 507 508 509 510 511
	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);
512
	root_anon_vma = READ_ONCE(anon_vma->root);
513
	if (down_read_trylock(&root_anon_vma->rwsem)) {
514
		/*
515 516
		 * If the page is still mapped, then this anon_vma is still
		 * its anon_vma, and holding the mutex ensures that it will
517
		 * not go away, see anon_vma_free().
518
		 */
519
		if (!page_mapped(page)) {
520
			up_read(&root_anon_vma->rwsem);
521 522 523 524
			anon_vma = NULL;
		}
		goto out;
	}
525

526 527 528 529 530 531 532
	/* trylock failed, we got to sleep */
	if (!atomic_inc_not_zero(&anon_vma->refcount)) {
		anon_vma = NULL;
		goto out;
	}

	if (!page_mapped(page)) {
533
		rcu_read_unlock();
534
		put_anon_vma(anon_vma);
535
		return NULL;
536 537 538 539
	}

	/* we pinned the anon_vma, its safe to sleep */
	rcu_read_unlock();
540
	anon_vma_lock_read(anon_vma);
541 542 543 544 545

	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
546
		 * we'll deadlock on the anon_vma_lock_write() recursion.
547
		 */
548
		anon_vma_unlock_read(anon_vma);
549 550 551 552 553 554 555 556
		__put_anon_vma(anon_vma);
		anon_vma = NULL;
	}

	return anon_vma;

out:
	rcu_read_unlock();
557
	return anon_vma;
558 559
}

560
void page_unlock_anon_vma_read(struct anon_vma *anon_vma)
561
{
562
	anon_vma_unlock_read(anon_vma);
L
Linus Torvalds 已提交
563 564 565
}

/*
566
 * At what user virtual address is page expected in @vma?
L
Linus Torvalds 已提交
567
 */
568 569
static inline unsigned long
__vma_address(struct page *page, struct vm_area_struct *vma)
L
Linus Torvalds 已提交
570
{
571
	pgoff_t pgoff = page_to_pgoff(page);
572 573 574 575 576 577 578 579 580
	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 */
581
	VM_BUG_ON_VMA(address < vma->vm_start || address >= vma->vm_end, vma);
582

L
Linus Torvalds 已提交
583 584 585 586
	return address;
}

/*
H
Huang Shijie 已提交
587
 * At what user virtual address is page expected in vma?
588
 * Caller should check the page is actually part of the vma.
L
Linus Torvalds 已提交
589 590 591
 */
unsigned long page_address_in_vma(struct page *page, struct vm_area_struct *vma)
{
592
	unsigned long address;
593
	if (PageAnon(page)) {
594 595 596 597 598 599 600
		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)
601
			return -EFAULT;
602 603
	} else if (page->mapping) {
		if (!vma->vm_file || vma->vm_file->f_mapping != page->mapping)
L
Linus Torvalds 已提交
604 605 606
			return -EFAULT;
	} else
		return -EFAULT;
607 608 609 610
	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
611 612
}

B
Bob Liu 已提交
613 614 615 616 617
pmd_t *mm_find_pmd(struct mm_struct *mm, unsigned long address)
{
	pgd_t *pgd;
	pud_t *pud;
	pmd_t *pmd = NULL;
618
	pmd_t pmde;
B
Bob Liu 已提交
619 620 621 622 623 624 625 626 627 628

	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);
629 630 631 632 633
	/*
	 * 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.
	 */
634 635
	pmde = *pmd;
	barrier();
636
	if (!pmd_present(pmde) || pmd_trans_huge(pmde))
B
Bob Liu 已提交
637 638 639 640 641
		pmd = NULL;
out:
	return pmd;
}

N
Nikita Danilov 已提交
642 643 644
/*
 * Check that @page is mapped at @address into @mm.
 *
N
Nick Piggin 已提交
645 646 647 648
 * 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).
 *
649
 * On success returns with pte mapped and locked.
N
Nikita Danilov 已提交
650
 */
651
pte_t *__page_check_address(struct page *page, struct mm_struct *mm,
N
Nick Piggin 已提交
652
			  unsigned long address, spinlock_t **ptlp, int sync)
N
Nikita Danilov 已提交
653 654 655
{
	pmd_t *pmd;
	pte_t *pte;
H
Hugh Dickins 已提交
656
	spinlock_t *ptl;
N
Nikita Danilov 已提交
657

658
	if (unlikely(PageHuge(page))) {
659
		/* when pud is not present, pte will be NULL */
660
		pte = huge_pte_offset(mm, address);
661 662 663
		if (!pte)
			return NULL;

664
		ptl = huge_pte_lockptr(page_hstate(page), mm, pte);
665 666 667
		goto check;
	}

B
Bob Liu 已提交
668 669
	pmd = mm_find_pmd(mm, address);
	if (!pmd)
H
Hugh Dickins 已提交
670 671 672 673
		return NULL;

	pte = pte_offset_map(pmd, address);
	/* Make a quick check before getting the lock */
N
Nick Piggin 已提交
674
	if (!sync && !pte_present(*pte)) {
H
Hugh Dickins 已提交
675 676 677 678
		pte_unmap(pte);
		return NULL;
	}

H
Hugh Dickins 已提交
679
	ptl = pte_lockptr(mm, pmd);
680
check:
H
Hugh Dickins 已提交
681 682 683 684
	spin_lock(ptl);
	if (pte_present(*pte) && page_to_pfn(page) == pte_pfn(*pte)) {
		*ptlp = ptl;
		return pte;
N
Nikita Danilov 已提交
685
	}
H
Hugh Dickins 已提交
686 687
	pte_unmap_unlock(pte, ptl);
	return NULL;
N
Nikita Danilov 已提交
688 689
}

N
Nick Piggin 已提交
690 691 692 693 694 695 696 697 698
/**
 * 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.
 */
699
int page_mapped_in_vma(struct page *page, struct vm_area_struct *vma)
N
Nick Piggin 已提交
700 701 702 703 704
{
	unsigned long address;
	pte_t *pte;
	spinlock_t *ptl;

705 706
	address = __vma_address(page, vma);
	if (unlikely(address < vma->vm_start || address >= vma->vm_end))
N
Nick Piggin 已提交
707 708 709 710 711 712 713 714 715
		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;
}

716 717 718 719 720 721
struct page_referenced_arg {
	int mapcount;
	int referenced;
	unsigned long vm_flags;
	struct mem_cgroup *memcg;
};
L
Linus Torvalds 已提交
722
/*
723
 * arg: page_referenced_arg will be passed
L
Linus Torvalds 已提交
724
 */
725
static int page_referenced_one(struct page *page, struct vm_area_struct *vma,
726
			unsigned long address, void *arg)
L
Linus Torvalds 已提交
727 728
{
	struct mm_struct *mm = vma->vm_mm;
729
	spinlock_t *ptl;
L
Linus Torvalds 已提交
730
	int referenced = 0;
731
	struct page_referenced_arg *pra = arg;
L
Linus Torvalds 已提交
732

733 734 735
	if (unlikely(PageTransHuge(page))) {
		pmd_t *pmd;

736 737 738 739
		/*
		 * rmap might return false positives; we must filter
		 * these out using page_check_address_pmd().
		 */
740
		pmd = page_check_address_pmd(page, mm, address,
741 742
					     PAGE_CHECK_ADDRESS_PMD_FLAG, &ptl);
		if (!pmd)
743
			return SWAP_AGAIN;
744 745

		if (vma->vm_flags & VM_LOCKED) {
746
			spin_unlock(ptl);
747 748
			pra->vm_flags |= VM_LOCKED;
			return SWAP_FAIL; /* To break the loop */
749 750 751 752
		}

		/* go ahead even if the pmd is pmd_trans_splitting() */
		if (pmdp_clear_flush_young_notify(vma, address, pmd))
753
			referenced++;
754
		spin_unlock(ptl);
755 756 757
	} else {
		pte_t *pte;

758 759 760 761
		/*
		 * rmap might return false positives; we must filter
		 * these out using page_check_address().
		 */
762 763
		pte = page_check_address(page, mm, address, &ptl, 0);
		if (!pte)
764
			return SWAP_AGAIN;
765

766 767
		if (vma->vm_flags & VM_LOCKED) {
			pte_unmap_unlock(pte, ptl);
768 769
			pra->vm_flags |= VM_LOCKED;
			return SWAP_FAIL; /* To break the loop */
770 771
		}

772 773 774 775 776 777 778 779
		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.
			 */
780
			if (likely(!(vma->vm_flags & VM_SEQ_READ)))
781 782 783 784 785
				referenced++;
		}
		pte_unmap_unlock(pte, ptl);
	}

786 787 788
	if (referenced) {
		pra->referenced++;
		pra->vm_flags |= vma->vm_flags;
L
Linus Torvalds 已提交
789
	}
790

791 792 793 794 795
	pra->mapcount--;
	if (!pra->mapcount)
		return SWAP_SUCCESS; /* To break the loop */

	return SWAP_AGAIN;
L
Linus Torvalds 已提交
796 797
}

798
static bool invalid_page_referenced_vma(struct vm_area_struct *vma, void *arg)
L
Linus Torvalds 已提交
799
{
800 801
	struct page_referenced_arg *pra = arg;
	struct mem_cgroup *memcg = pra->memcg;
L
Linus Torvalds 已提交
802

803 804
	if (!mm_match_cgroup(vma->vm_mm, memcg))
		return true;
L
Linus Torvalds 已提交
805

806
	return false;
L
Linus Torvalds 已提交
807 808 809 810 811 812
}

/**
 * page_referenced - test if the page was referenced
 * @page: the page to test
 * @is_locked: caller holds lock on the page
813
 * @memcg: target memory cgroup
814
 * @vm_flags: collect encountered vma->vm_flags who actually referenced the page
L
Linus Torvalds 已提交
815 816 817 818
 *
 * Quick test_and_clear_referenced for all mappings to a page,
 * returns the number of ptes which referenced the page.
 */
819 820
int page_referenced(struct page *page,
		    int is_locked,
821
		    struct mem_cgroup *memcg,
822
		    unsigned long *vm_flags)
L
Linus Torvalds 已提交
823
{
824
	int ret;
H
Hugh Dickins 已提交
825
	int we_locked = 0;
826 827 828 829 830 831 832 833 834
	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
835

836
	*vm_flags = 0;
837 838 839 840 841 842 843 844 845 846
	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
847
	}
848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864

	/*
	 * 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
865 866
}

H
Hugh Dickins 已提交
867
static int page_mkclean_one(struct page *page, struct vm_area_struct *vma,
868
			    unsigned long address, void *arg)
869 870
{
	struct mm_struct *mm = vma->vm_mm;
871
	pte_t *pte;
872 873
	spinlock_t *ptl;
	int ret = 0;
874
	int *cleaned = arg;
875

N
Nick Piggin 已提交
876
	pte = page_check_address(page, mm, address, &ptl, 1);
877 878 879
	if (!pte)
		goto out;

880 881
	if (pte_dirty(*pte) || pte_write(*pte)) {
		pte_t entry;
882

883
		flush_cache_page(vma, address, pte_pfn(*pte));
884
		entry = ptep_clear_flush(vma, address, pte);
885 886
		entry = pte_wrprotect(entry);
		entry = pte_mkclean(entry);
887
		set_pte_at(mm, address, pte, entry);
888 889
		ret = 1;
	}
890 891

	pte_unmap_unlock(pte, ptl);
892

893
	if (ret) {
894
		mmu_notifier_invalidate_page(mm, address);
895 896
		(*cleaned)++;
	}
897
out:
898
	return SWAP_AGAIN;
899 900
}

901
static bool invalid_mkclean_vma(struct vm_area_struct *vma, void *arg)
902
{
903
	if (vma->vm_flags & VM_SHARED)
F
Fengguang Wu 已提交
904
		return false;
905

F
Fengguang Wu 已提交
906
	return true;
907 908 909 910
}

int page_mkclean(struct page *page)
{
911 912 913 914 915 916 917
	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,
	};
918 919 920

	BUG_ON(!PageLocked(page));

921 922 923 924 925 926 927 928
	if (!page_mapped(page))
		return 0;

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

	rmap_walk(page, &rwc);
929

930
	return cleaned;
931
}
J
Jaya Kumar 已提交
932
EXPORT_SYMBOL_GPL(page_mkclean);
933

934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949
/**
 * 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;

950
	VM_BUG_ON_PAGE(!PageLocked(page), page);
951
	VM_BUG_ON_VMA(!anon_vma, vma);
952
	VM_BUG_ON_PAGE(page->index != linear_page_index(vma, address), page);
953 954 955 956 957

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

N
Nick Piggin 已提交
958
/**
A
Andi Kleen 已提交
959 960 961 962
 * __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	
963
 * @exclusive:	the page is exclusively owned by the current process
N
Nick Piggin 已提交
964 965
 */
static void __page_set_anon_rmap(struct page *page,
966
	struct vm_area_struct *vma, unsigned long address, int exclusive)
N
Nick Piggin 已提交
967
{
968
	struct anon_vma *anon_vma = vma->anon_vma;
969

970
	BUG_ON(!anon_vma);
971

A
Andi Kleen 已提交
972 973 974
	if (PageAnon(page))
		return;

975
	/*
976 977 978
	 * If the page isn't exclusively mapped into this vma,
	 * we must use the _oldest_ possible anon_vma for the
	 * page mapping!
979
	 */
A
Andi Kleen 已提交
980
	if (!exclusive)
981
		anon_vma = anon_vma->root;
N
Nick Piggin 已提交
982 983 984 985 986 987

	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
988
/**
R
Randy Dunlap 已提交
989
 * __page_check_anon_rmap - sanity check anonymous rmap addition
N
Nick Piggin 已提交
990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009
 * @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
1010
	BUG_ON(page_anon_vma(page)->root != vma->anon_vma->root);
N
Nick Piggin 已提交
1011 1012 1013 1014
	BUG_ON(page->index != linear_page_index(vma, address));
#endif
}

L
Linus Torvalds 已提交
1015 1016 1017 1018 1019 1020
/**
 * 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
1021
 * The caller needs to hold the pte lock, and the page must be locked in
1022 1023 1024
 * 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
1025 1026 1027
 */
void page_add_anon_rmap(struct page *page,
	struct vm_area_struct *vma, unsigned long address)
1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038
{
	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
1039
{
H
Hugh Dickins 已提交
1040
	int first = atomic_inc_and_test(&page->_mapcount);
1041
	if (first) {
1042 1043 1044 1045 1046 1047
		/*
		 * 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.
		 */
1048
		if (PageTransHuge(page))
1049 1050
			__inc_zone_page_state(page,
					      NR_ANON_TRANSPARENT_HUGEPAGES);
1051 1052
		__mod_zone_page_state(page_zone(page), NR_ANON_PAGES,
				hpage_nr_pages(page));
1053
	}
H
Hugh Dickins 已提交
1054 1055 1056
	if (unlikely(PageKsm(page)))
		return;

1057
	VM_BUG_ON_PAGE(!PageLocked(page), page);
1058
	/* address might be in next vma when migration races vma_adjust */
H
Hugh Dickins 已提交
1059
	if (first)
1060
		__page_set_anon_rmap(page, vma, address, exclusive);
1061
	else
N
Nick Piggin 已提交
1062
		__page_check_anon_rmap(page, vma, address);
L
Linus Torvalds 已提交
1063 1064
}

R
Randy Dunlap 已提交
1065
/**
N
Nick Piggin 已提交
1066 1067 1068 1069 1070 1071 1072
 * 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
1073
 * Page does not have to be locked.
N
Nick Piggin 已提交
1074 1075 1076 1077
 */
void page_add_new_anon_rmap(struct page *page,
	struct vm_area_struct *vma, unsigned long address)
{
1078
	VM_BUG_ON_VMA(address < vma->vm_start || address >= vma->vm_end, vma);
1079 1080
	SetPageSwapBacked(page);
	atomic_set(&page->_mapcount, 0); /* increment count (starts at -1) */
1081
	if (PageTransHuge(page))
1082
		__inc_zone_page_state(page, NR_ANON_TRANSPARENT_HUGEPAGES);
1083 1084
	__mod_zone_page_state(page_zone(page), NR_ANON_PAGES,
			hpage_nr_pages(page));
1085
	__page_set_anon_rmap(page, vma, address, 1);
N
Nick Piggin 已提交
1086 1087
}

L
Linus Torvalds 已提交
1088 1089 1090 1091
/**
 * page_add_file_rmap - add pte mapping to a file page
 * @page: the page to add the mapping to
 *
1092
 * The caller needs to hold the pte lock.
L
Linus Torvalds 已提交
1093 1094 1095
 */
void page_add_file_rmap(struct page *page)
{
1096
	struct mem_cgroup *memcg;
1097

1098
	memcg = mem_cgroup_begin_page_stat(page);
1099
	if (atomic_inc_and_test(&page->_mapcount)) {
1100
		__inc_zone_page_state(page, NR_FILE_MAPPED);
1101
		mem_cgroup_inc_page_stat(memcg, MEM_CGROUP_STAT_FILE_MAPPED);
1102
	}
1103
	mem_cgroup_end_page_stat(memcg);
L
Linus Torvalds 已提交
1104 1105
}

1106 1107 1108 1109
static void page_remove_file_rmap(struct page *page)
{
	struct mem_cgroup *memcg;

1110
	memcg = mem_cgroup_begin_page_stat(page);
1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130

	/* 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:
1131
	mem_cgroup_end_page_stat(memcg);
1132 1133
}

L
Linus Torvalds 已提交
1134 1135 1136 1137
/**
 * page_remove_rmap - take down pte mapping from a page
 * @page: page to remove mapping from
 *
1138
 * The caller needs to hold the pte lock.
L
Linus Torvalds 已提交
1139
 */
1140
void page_remove_rmap(struct page *page)
L
Linus Torvalds 已提交
1141
{
1142 1143 1144 1145
	if (!PageAnon(page)) {
		page_remove_file_rmap(page);
		return;
	}
1146

1147 1148
	/* page still mapped by someone else? */
	if (!atomic_add_negative(-1, &page->_mapcount))
1149 1150 1151 1152 1153
		return;

	/* Hugepages are not counted in NR_ANON_PAGES for now. */
	if (unlikely(PageHuge(page)))
		return;
1154

1155
	/*
1156 1157 1158
	 * 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.
1159
	 */
1160 1161 1162 1163 1164 1165
	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));

1166 1167
	if (unlikely(PageMlocked(page)))
		clear_page_mlock(page);
1168

1169 1170 1171 1172 1173 1174 1175 1176 1177
	/*
	 * 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
1178 1179 1180
}

/*
1181
 * @arg: enum ttu_flags will be passed to this argument
L
Linus Torvalds 已提交
1182
 */
1183
static int try_to_unmap_one(struct page *page, struct vm_area_struct *vma,
1184
		     unsigned long address, void *arg)
L
Linus Torvalds 已提交
1185 1186 1187 1188
{
	struct mm_struct *mm = vma->vm_mm;
	pte_t *pte;
	pte_t pteval;
H
Hugh Dickins 已提交
1189
	spinlock_t *ptl;
L
Linus Torvalds 已提交
1190
	int ret = SWAP_AGAIN;
1191
	enum ttu_flags flags = (enum ttu_flags)arg;
L
Linus Torvalds 已提交
1192

N
Nick Piggin 已提交
1193
	pte = page_check_address(page, mm, address, &ptl, 0);
H
Hugh Dickins 已提交
1194
	if (!pte)
N
Nikita Danilov 已提交
1195
		goto out;
L
Linus Torvalds 已提交
1196 1197 1198 1199 1200 1201

	/*
	 * 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.
	 */
1202
	if (!(flags & TTU_IGNORE_MLOCK)) {
K
KOSAKI Motohiro 已提交
1203 1204 1205
		if (vma->vm_flags & VM_LOCKED)
			goto out_mlock;

1206
		if (flags & TTU_MUNLOCK)
H
Hugh Dickins 已提交
1207
			goto out_unmap;
1208 1209
	}
	if (!(flags & TTU_IGNORE_ACCESS)) {
N
Nick Piggin 已提交
1210 1211 1212 1213 1214
		if (ptep_clear_flush_young_notify(vma, address, pte)) {
			ret = SWAP_FAIL;
			goto out_unmap;
		}
  	}
L
Linus Torvalds 已提交
1215 1216 1217

	/* Nuke the page table entry. */
	flush_cache_page(vma, address, page_to_pfn(page));
1218
	pteval = ptep_clear_flush(vma, address, pte);
L
Linus Torvalds 已提交
1219 1220 1221 1222 1223

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

1224 1225 1226
	/* Update high watermark before we lower rss */
	update_hiwater_rss(mm);

1227
	if (PageHWPoison(page) && !(flags & TTU_IGNORE_HWPOISON)) {
1228 1229 1230 1231 1232 1233
		if (!PageHuge(page)) {
			if (PageAnon(page))
				dec_mm_counter(mm, MM_ANONPAGES);
			else
				dec_mm_counter(mm, MM_FILEPAGES);
		}
1234
		set_pte_at(mm, address, pte,
1235
			   swp_entry_to_pte(make_hwpoison_entry(page)));
1236 1237 1238 1239 1240 1241 1242 1243 1244 1245
	} 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);
1246
	} else if (PageAnon(page)) {
H
Hugh Dickins 已提交
1247
		swp_entry_t entry = { .val = page_private(page) };
1248
		pte_t swp_pte;
1249 1250 1251 1252 1253 1254

		if (PageSwapCache(page)) {
			/*
			 * Store the swap location in the pte.
			 * See handle_pte_fault() ...
			 */
H
Hugh Dickins 已提交
1255 1256 1257 1258 1259
			if (swap_duplicate(entry) < 0) {
				set_pte_at(mm, address, pte, pteval);
				ret = SWAP_FAIL;
				goto out_unmap;
			}
1260 1261 1262 1263 1264 1265
			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
1266
			dec_mm_counter(mm, MM_ANONPAGES);
K
KAMEZAWA Hiroyuki 已提交
1267
			inc_mm_counter(mm, MM_SWAPENTS);
1268
		} else if (IS_ENABLED(CONFIG_MIGRATION)) {
1269 1270 1271 1272 1273
			/*
			 * 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.
			 */
1274
			BUG_ON(!(flags & TTU_MIGRATION));
1275
			entry = make_migration_entry(page, pte_write(pteval));
L
Linus Torvalds 已提交
1276
		}
1277 1278 1279 1280
		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);
1281
	} else if (IS_ENABLED(CONFIG_MIGRATION) &&
1282
		   (flags & TTU_MIGRATION)) {
1283 1284 1285 1286 1287
		/* 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
1288
		dec_mm_counter(mm, MM_FILEPAGES);
L
Linus Torvalds 已提交
1289

1290
	page_remove_rmap(page);
L
Linus Torvalds 已提交
1291 1292 1293
	page_cache_release(page);

out_unmap:
H
Hugh Dickins 已提交
1294
	pte_unmap_unlock(pte, ptl);
1295
	if (ret != SWAP_FAIL && !(flags & TTU_MUNLOCK))
1296
		mmu_notifier_invalidate_page(mm, address);
K
KOSAKI Motohiro 已提交
1297 1298
out:
	return ret;
H
Hugh Dickins 已提交
1299

K
KOSAKI Motohiro 已提交
1300 1301 1302 1303 1304 1305 1306
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
1307
	 * we now hold anon_vma->rwsem or mapping->i_mmap_rwsem.
K
KOSAKI Motohiro 已提交
1308 1309 1310 1311 1312 1313 1314 1315
	 * 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
1316
		}
K
KOSAKI Motohiro 已提交
1317
		up_read(&vma->vm_mm->mmap_sem);
H
Hugh Dickins 已提交
1318
	}
L
Linus Torvalds 已提交
1319 1320 1321
	return ret;
}

1322
bool is_vma_temporary_stack(struct vm_area_struct *vma)
1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335
{
	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;
}

1336 1337 1338 1339 1340 1341 1342 1343 1344 1345
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
1346 1347 1348
/**
 * try_to_unmap - try to remove all page table mappings to a page
 * @page: the page to get unmapped
1349
 * @flags: action and flags
L
Linus Torvalds 已提交
1350 1351 1352 1353 1354 1355 1356 1357
 *
 * 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
1358
 * SWAP_MLOCK	- page is mlocked.
L
Linus Torvalds 已提交
1359
 */
1360
int try_to_unmap(struct page *page, enum ttu_flags flags)
L
Linus Torvalds 已提交
1361 1362
{
	int ret;
1363 1364 1365 1366 1367 1368
	struct rmap_walk_control rwc = {
		.rmap_one = try_to_unmap_one,
		.arg = (void *)flags,
		.done = page_not_mapped,
		.anon_lock = page_lock_anon_vma_read,
	};
L
Linus Torvalds 已提交
1369

1370
	VM_BUG_ON_PAGE(!PageHuge(page) && PageTransHuge(page), page);
L
Linus Torvalds 已提交
1371

1372 1373 1374 1375 1376 1377 1378 1379
	/*
	 * 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.
	 */
1380
	if ((flags & TTU_MIGRATION) && !PageKsm(page) && PageAnon(page))
1381 1382 1383 1384
		rwc.invalid_vma = invalid_migration_vma;

	ret = rmap_walk(page, &rwc);

N
Nick Piggin 已提交
1385
	if (ret != SWAP_MLOCK && !page_mapped(page))
L
Linus Torvalds 已提交
1386 1387 1388
		ret = SWAP_SUCCESS;
	return ret;
}
N
Nikita Danilov 已提交
1389

N
Nick Piggin 已提交
1390 1391 1392 1393 1394 1395 1396 1397 1398 1399
/**
 * 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
1400
 * SWAP_AGAIN	- no vma is holding page mlocked, or,
N
Nick Piggin 已提交
1401
 * SWAP_AGAIN	- page mapped in mlocked vma -- couldn't acquire mmap sem
H
Hugh Dickins 已提交
1402
 * SWAP_FAIL	- page cannot be located at present
N
Nick Piggin 已提交
1403 1404 1405 1406
 * SWAP_MLOCK	- page is now mlocked.
 */
int try_to_munlock(struct page *page)
{
1407 1408 1409 1410 1411 1412 1413 1414 1415
	int ret;
	struct rmap_walk_control rwc = {
		.rmap_one = try_to_unmap_one,
		.arg = (void *)TTU_MUNLOCK,
		.done = page_not_mapped,
		.anon_lock = page_lock_anon_vma_read,

	};

1416
	VM_BUG_ON_PAGE(!PageLocked(page) || PageLRU(page), page);
N
Nick Piggin 已提交
1417

1418 1419
	ret = rmap_walk(page, &rwc);
	return ret;
N
Nick Piggin 已提交
1420
}
1421

P
Peter Zijlstra 已提交
1422
void __put_anon_vma(struct anon_vma *anon_vma)
1423
{
P
Peter Zijlstra 已提交
1424
	struct anon_vma *root = anon_vma->root;
1425

1426
	anon_vma_free(anon_vma);
P
Peter Zijlstra 已提交
1427 1428
	if (root != anon_vma && atomic_dec_and_test(&root->refcount))
		anon_vma_free(root);
1429 1430
}

1431 1432
static struct anon_vma *rmap_walk_anon_lock(struct page *page,
					struct rmap_walk_control *rwc)
1433 1434 1435
{
	struct anon_vma *anon_vma;

1436 1437 1438
	if (rwc->anon_lock)
		return rwc->anon_lock(page);

1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452
	/*
	 * 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;
}

1453
/*
1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465
 * 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.
1466
 */
1467
static int rmap_walk_anon(struct page *page, struct rmap_walk_control *rwc)
1468 1469
{
	struct anon_vma *anon_vma;
1470
	pgoff_t pgoff;
1471
	struct anon_vma_chain *avc;
1472 1473
	int ret = SWAP_AGAIN;

1474
	anon_vma = rmap_walk_anon_lock(page, rwc);
1475 1476
	if (!anon_vma)
		return ret;
1477

1478
	pgoff = page_to_pgoff(page);
1479
	anon_vma_interval_tree_foreach(avc, &anon_vma->rb_root, pgoff, pgoff) {
1480
		struct vm_area_struct *vma = avc->vma;
1481
		unsigned long address = vma_address(page, vma);
1482 1483 1484 1485

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

1486
		ret = rwc->rmap_one(page, vma, address, rwc->arg);
1487 1488
		if (ret != SWAP_AGAIN)
			break;
1489 1490
		if (rwc->done && rwc->done(page))
			break;
1491
	}
1492
	anon_vma_unlock_read(anon_vma);
1493 1494 1495
	return ret;
}

1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508
/*
 * 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.
 */
1509
static int rmap_walk_file(struct page *page, struct rmap_walk_control *rwc)
1510 1511
{
	struct address_space *mapping = page->mapping;
1512
	pgoff_t pgoff;
1513 1514 1515
	struct vm_area_struct *vma;
	int ret = SWAP_AGAIN;

1516 1517 1518 1519
	/*
	 * 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,
1520
	 * so we can safely take mapping->i_mmap_rwsem.
1521
	 */
1522
	VM_BUG_ON_PAGE(!PageLocked(page), page);
1523

1524 1525
	if (!mapping)
		return ret;
D
Davidlohr Bueso 已提交
1526

1527
	pgoff = page_to_pgoff(page);
D
Davidlohr Bueso 已提交
1528
	i_mmap_lock_read(mapping);
1529
	vma_interval_tree_foreach(vma, &mapping->i_mmap, pgoff, pgoff) {
1530
		unsigned long address = vma_address(page, vma);
1531 1532 1533 1534

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

1535
		ret = rwc->rmap_one(page, vma, address, rwc->arg);
1536
		if (ret != SWAP_AGAIN)
1537 1538 1539
			goto done;
		if (rwc->done && rwc->done(page))
			goto done;
1540
	}
1541 1542

done:
D
Davidlohr Bueso 已提交
1543
	i_mmap_unlock_read(mapping);
1544 1545 1546
	return ret;
}

1547
int rmap_walk(struct page *page, struct rmap_walk_control *rwc)
1548 1549
{
	if (unlikely(PageKsm(page)))
1550
		return rmap_walk_ksm(page, rwc);
1551
	else if (PageAnon(page))
1552
		return rmap_walk_anon(page, rwc);
1553
	else
1554
		return rmap_walk_file(page, rwc);
1555
}
1556

N
Naoya Horiguchi 已提交
1557
#ifdef CONFIG_HUGETLB_PAGE
1558 1559 1560 1561 1562 1563 1564 1565 1566
/*
 * 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;
1567

1568
	BUG_ON(!anon_vma);
1569 1570 1571 1572 1573 1574

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

1575 1576 1577 1578 1579 1580 1581 1582 1583 1584
	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;
1585 1586

	BUG_ON(!PageLocked(page));
1587
	BUG_ON(!anon_vma);
1588
	/* address might be in next vma when migration races vma_adjust */
1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600
	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
1601
#endif /* CONFIG_HUGETLB_PAGE */