rmap.c 43.7 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
	struct anon_vma *anon_vma;

	anon_vma = kmem_cache_alloc(anon_vma_cachep, GFP_KERNEL);
	if (anon_vma) {
		atomic_set(&anon_vma->refcount, 1);
75 76
		anon_vma->degree = 1;	/* Reference for first vma */
		anon_vma->parent = anon_vma;
P
Peter Zijlstra 已提交
77 78 79 80 81 82 83 84
		/*
		 * 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;
85 86
}

P
Peter Zijlstra 已提交
87
static inline void anon_vma_free(struct anon_vma *anon_vma)
88
{
P
Peter Zijlstra 已提交
89
	VM_BUG_ON(atomic_read(&anon_vma->refcount));
90 91

	/*
92
	 * Synchronize against page_lock_anon_vma_read() such that
93 94 95 96 97
	 * 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
98
	 * down_read_trylock() from page_lock_anon_vma_read(). This orders:
99
	 *
100 101
	 * page_lock_anon_vma_read()	VS	put_anon_vma()
	 *   down_read_trylock()		  atomic_dec_and_test()
102
	 *   LOCK				  MB
103
	 *   atomic_read()			  rwsem_is_locked()
104 105 106 107
	 *
	 * LOCK should suffice since the actual taking of the lock must
	 * happen _before_ what follows.
	 */
108
	might_sleep();
109
	if (rwsem_is_locked(&anon_vma->root->rwsem)) {
110
		anon_vma_lock_write(anon_vma);
111
		anon_vma_unlock_write(anon_vma);
112 113
	}

114 115
	kmem_cache_free(anon_vma_cachep, anon_vma);
}
L
Linus Torvalds 已提交
116

117
static inline struct anon_vma_chain *anon_vma_chain_alloc(gfp_t gfp)
118
{
119
	return kmem_cache_alloc(anon_vma_chain_cachep, gfp);
120 121
}

122
static void anon_vma_chain_free(struct anon_vma_chain *anon_vma_chain)
123 124 125 126
{
	kmem_cache_free(anon_vma_chain_cachep, anon_vma_chain);
}

127 128 129 130 131 132 133
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);
134
	anon_vma_interval_tree_insert(avc, &anon_vma->rb_root);
135 136
}

137 138 139 140 141 142 143 144 145
/**
 * 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
146
 * not we either need to find an adjacent mapping that we
147 148 149 150 151
 * 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
152
 * optimistically looked up an anon_vma in page_lock_anon_vma_read()
153 154 155 156 157 158 159 160 161 162 163
 * 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
164 165 166
int anon_vma_prepare(struct vm_area_struct *vma)
{
	struct anon_vma *anon_vma = vma->anon_vma;
167
	struct anon_vma_chain *avc;
L
Linus Torvalds 已提交
168 169 170 171

	might_sleep();
	if (unlikely(!anon_vma)) {
		struct mm_struct *mm = vma->vm_mm;
172
		struct anon_vma *allocated;
L
Linus Torvalds 已提交
173

174
		avc = anon_vma_chain_alloc(GFP_KERNEL);
175 176 177
		if (!avc)
			goto out_enomem;

L
Linus Torvalds 已提交
178
		anon_vma = find_mergeable_anon_vma(vma);
179 180
		allocated = NULL;
		if (!anon_vma) {
L
Linus Torvalds 已提交
181 182
			anon_vma = anon_vma_alloc();
			if (unlikely(!anon_vma))
183
				goto out_enomem_free_avc;
L
Linus Torvalds 已提交
184 185 186
			allocated = anon_vma;
		}

187
		anon_vma_lock_write(anon_vma);
L
Linus Torvalds 已提交
188 189 190 191
		/* page_table_lock to protect against threads */
		spin_lock(&mm->page_table_lock);
		if (likely(!vma->anon_vma)) {
			vma->anon_vma = anon_vma;
192
			anon_vma_chain_link(vma, avc, anon_vma);
193 194
			/* vma reference or self-parent link for new root */
			anon_vma->degree++;
L
Linus Torvalds 已提交
195
			allocated = NULL;
196
			avc = NULL;
L
Linus Torvalds 已提交
197 198
		}
		spin_unlock(&mm->page_table_lock);
199
		anon_vma_unlock_write(anon_vma);
200 201

		if (unlikely(allocated))
P
Peter Zijlstra 已提交
202
			put_anon_vma(allocated);
203
		if (unlikely(avc))
204
			anon_vma_chain_free(avc);
L
Linus Torvalds 已提交
205 206
	}
	return 0;
207 208 209 210 211

 out_enomem_free_avc:
	anon_vma_chain_free(avc);
 out_enomem:
	return -ENOMEM;
L
Linus Torvalds 已提交
212 213
}

214 215 216 217 218 219 220 221 222 223 224 225 226
/*
 * 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))
227
			up_write(&root->rwsem);
228
		root = new_root;
229
		down_write(&root->rwsem);
230 231 232 233 234 235 236
	}
	return root;
}

static inline void unlock_anon_vma_root(struct anon_vma *root)
{
	if (root)
237
		up_write(&root->rwsem);
238 239
}

240 241 242
/*
 * Attach the anon_vmas from src to dst.
 * Returns 0 on success, -ENOMEM on failure.
243 244 245 246 247 248 249 250
 *
 * 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.
251 252
 */
int anon_vma_clone(struct vm_area_struct *dst, struct vm_area_struct *src)
L
Linus Torvalds 已提交
253
{
254
	struct anon_vma_chain *avc, *pavc;
255
	struct anon_vma *root = NULL;
256

257
	list_for_each_entry_reverse(pavc, &src->anon_vma_chain, same_vma) {
258 259
		struct anon_vma *anon_vma;

260 261 262 263 264 265 266 267
		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;
		}
268 269 270
		anon_vma = pavc->anon_vma;
		root = lock_anon_vma_root(root, anon_vma);
		anon_vma_chain_link(dst, avc, anon_vma);
271 272 273 274 275 276 277 278 279 280 281 282

		/*
		 * 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;
283
	}
284 285
	if (dst->anon_vma)
		dst->anon_vma->degree++;
286
	unlock_anon_vma_root(root);
287
	return 0;
L
Linus Torvalds 已提交
288

289
 enomem_failure:
290 291 292 293 294 295 296
	/*
	 * 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;
297 298
	unlink_anon_vmas(dst);
	return -ENOMEM;
L
Linus Torvalds 已提交
299 300
}

301 302 303 304 305 306
/*
 * 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
307
{
308 309
	struct anon_vma_chain *avc;
	struct anon_vma *anon_vma;
310
	int error;
L
Linus Torvalds 已提交
311

312 313 314 315
	/* Don't bother if the parent process has no anon_vma here. */
	if (!pvma->anon_vma)
		return 0;

316 317 318
	/* Drop inherited anon_vma, we'll reuse existing or allocate new. */
	vma->anon_vma = NULL;

319 320 321 322
	/*
	 * First, attach the new VMA to the parent VMA's anon_vmas,
	 * so rmap can find non-COWed pages in child processes.
	 */
323 324 325
	error = anon_vma_clone(vma, pvma);
	if (error)
		return error;
326

327 328 329 330
	/* An existing anon_vma has been reused, all done then. */
	if (vma->anon_vma)
		return 0;

331 332 333 334
	/* Then add our own anon_vma. */
	anon_vma = anon_vma_alloc();
	if (!anon_vma)
		goto out_error;
335
	avc = anon_vma_chain_alloc(GFP_KERNEL);
336 337
	if (!avc)
		goto out_error_free_anon_vma;
338 339 340 341 342 343

	/*
	 * 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;
344
	anon_vma->parent = pvma->anon_vma;
345
	/*
P
Peter Zijlstra 已提交
346 347 348
	 * 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.
349 350
	 */
	get_anon_vma(anon_vma->root);
351 352
	/* Mark this anon_vma as the one where our new (COWed) pages go. */
	vma->anon_vma = anon_vma;
353
	anon_vma_lock_write(anon_vma);
354
	anon_vma_chain_link(vma, avc, anon_vma);
355
	anon_vma->parent->degree++;
356
	anon_vma_unlock_write(anon_vma);
357 358 359 360

	return 0;

 out_error_free_anon_vma:
P
Peter Zijlstra 已提交
361
	put_anon_vma(anon_vma);
362
 out_error:
363
	unlink_anon_vmas(vma);
364
	return -ENOMEM;
L
Linus Torvalds 已提交
365 366
}

367 368 369
void unlink_anon_vmas(struct vm_area_struct *vma)
{
	struct anon_vma_chain *avc, *next;
370
	struct anon_vma *root = NULL;
371

372 373 374 375
	/*
	 * Unlink each anon_vma chained to the VMA.  This list is ordered
	 * from newest to oldest, ensuring the root anon_vma gets freed last.
	 */
376
	list_for_each_entry_safe(avc, next, &vma->anon_vma_chain, same_vma) {
377 378 379
		struct anon_vma *anon_vma = avc->anon_vma;

		root = lock_anon_vma_root(root, anon_vma);
380
		anon_vma_interval_tree_remove(avc, &anon_vma->rb_root);
381 382 383 384 385

		/*
		 * Leave empty anon_vmas on the list - we'll need
		 * to free them outside the lock.
		 */
386 387
		if (RB_EMPTY_ROOT(&anon_vma->rb_root)) {
			anon_vma->parent->degree--;
388
			continue;
389
		}
390 391 392 393

		list_del(&avc->same_vma);
		anon_vma_chain_free(avc);
	}
394 395
	if (vma->anon_vma)
		vma->anon_vma->degree--;
396 397 398 399 400
	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()
401
	 * needing to write-acquire the anon_vma->root->rwsem.
402 403 404 405
	 */
	list_for_each_entry_safe(avc, next, &vma->anon_vma_chain, same_vma) {
		struct anon_vma *anon_vma = avc->anon_vma;

406
		BUG_ON(anon_vma->degree);
407 408
		put_anon_vma(anon_vma);

409 410 411 412 413
		list_del(&avc->same_vma);
		anon_vma_chain_free(avc);
	}
}

414
static void anon_vma_ctor(void *data)
L
Linus Torvalds 已提交
415
{
C
Christoph Lameter 已提交
416
	struct anon_vma *anon_vma = data;
L
Linus Torvalds 已提交
417

418
	init_rwsem(&anon_vma->rwsem);
419
	atomic_set(&anon_vma->refcount, 0);
420
	anon_vma->rb_root = RB_ROOT;
L
Linus Torvalds 已提交
421 422 423 424 425
}

void __init anon_vma_init(void)
{
	anon_vma_cachep = kmem_cache_create("anon_vma", sizeof(struct anon_vma),
426
			0, SLAB_DESTROY_BY_RCU|SLAB_PANIC, anon_vma_ctor);
427
	anon_vma_chain_cachep = KMEM_CACHE(anon_vma_chain, SLAB_PANIC);
L
Linus Torvalds 已提交
428 429 430
}

/*
431 432 433 434 435 436 437 438 439
 * 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).
 *
440 441 442 443 444
 * 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 ].
 *
445 446 447 448 449 450 451
 * 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
452
 */
453
struct anon_vma *page_get_anon_vma(struct page *page)
L
Linus Torvalds 已提交
454
{
455
	struct anon_vma *anon_vma = NULL;
L
Linus Torvalds 已提交
456 457 458
	unsigned long anon_mapping;

	rcu_read_lock();
459
	anon_mapping = (unsigned long)READ_ONCE(page->mapping);
H
Hugh Dickins 已提交
460
	if ((anon_mapping & PAGE_MAPPING_FLAGS) != PAGE_MAPPING_ANON)
L
Linus Torvalds 已提交
461 462 463 464 465
		goto out;
	if (!page_mapped(page))
		goto out;

	anon_vma = (struct anon_vma *) (anon_mapping - PAGE_MAPPING_ANON);
466 467 468 469
	if (!atomic_inc_not_zero(&anon_vma->refcount)) {
		anon_vma = NULL;
		goto out;
	}
470 471 472

	/*
	 * If this page is still mapped, then its anon_vma cannot have been
473 474 475 476
	 * 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).
477
	 */
478
	if (!page_mapped(page)) {
479
		rcu_read_unlock();
480
		put_anon_vma(anon_vma);
481
		return NULL;
482
	}
L
Linus Torvalds 已提交
483 484
out:
	rcu_read_unlock();
485 486 487 488

	return anon_vma;
}

489 490 491 492 493 494 495
/*
 * 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.
 */
496
struct anon_vma *page_lock_anon_vma_read(struct page *page)
497
{
498
	struct anon_vma *anon_vma = NULL;
499
	struct anon_vma *root_anon_vma;
500
	unsigned long anon_mapping;
501

502
	rcu_read_lock();
503
	anon_mapping = (unsigned long)READ_ONCE(page->mapping);
504 505 506 507 508 509
	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);
510
	root_anon_vma = READ_ONCE(anon_vma->root);
511
	if (down_read_trylock(&root_anon_vma->rwsem)) {
512
		/*
513 514
		 * If the page is still mapped, then this anon_vma is still
		 * its anon_vma, and holding the mutex ensures that it will
515
		 * not go away, see anon_vma_free().
516
		 */
517
		if (!page_mapped(page)) {
518
			up_read(&root_anon_vma->rwsem);
519 520 521 522
			anon_vma = NULL;
		}
		goto out;
	}
523

524 525 526 527 528 529 530
	/* trylock failed, we got to sleep */
	if (!atomic_inc_not_zero(&anon_vma->refcount)) {
		anon_vma = NULL;
		goto out;
	}

	if (!page_mapped(page)) {
531
		rcu_read_unlock();
532
		put_anon_vma(anon_vma);
533
		return NULL;
534 535 536 537
	}

	/* we pinned the anon_vma, its safe to sleep */
	rcu_read_unlock();
538
	anon_vma_lock_read(anon_vma);
539 540 541 542 543

	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
544
		 * we'll deadlock on the anon_vma_lock_write() recursion.
545
		 */
546
		anon_vma_unlock_read(anon_vma);
547 548 549 550 551 552 553 554
		__put_anon_vma(anon_vma);
		anon_vma = NULL;
	}

	return anon_vma;

out:
	rcu_read_unlock();
555
	return anon_vma;
556 557
}

558
void page_unlock_anon_vma_read(struct anon_vma *anon_vma)
559
{
560
	anon_vma_unlock_read(anon_vma);
L
Linus Torvalds 已提交
561 562 563
}

/*
564
 * At what user virtual address is page expected in @vma?
L
Linus Torvalds 已提交
565
 */
566 567
static inline unsigned long
__vma_address(struct page *page, struct vm_area_struct *vma)
L
Linus Torvalds 已提交
568
{
569
	pgoff_t pgoff = page_to_pgoff(page);
570 571 572 573 574 575 576 577 578
	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 */
579
	VM_BUG_ON_VMA(address < vma->vm_start || address >= vma->vm_end, vma);
580

L
Linus Torvalds 已提交
581 582 583 584
	return address;
}

/*
H
Huang Shijie 已提交
585
 * At what user virtual address is page expected in vma?
586
 * Caller should check the page is actually part of the vma.
L
Linus Torvalds 已提交
587 588 589
 */
unsigned long page_address_in_vma(struct page *page, struct vm_area_struct *vma)
{
590
	unsigned long address;
591
	if (PageAnon(page)) {
592 593 594 595 596 597 598
		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)
599
			return -EFAULT;
600 601
	} else if (page->mapping) {
		if (!vma->vm_file || vma->vm_file->f_mapping != page->mapping)
L
Linus Torvalds 已提交
602 603 604
			return -EFAULT;
	} else
		return -EFAULT;
605 606 607 608
	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
609 610
}

B
Bob Liu 已提交
611 612 613 614 615
pmd_t *mm_find_pmd(struct mm_struct *mm, unsigned long address)
{
	pgd_t *pgd;
	pud_t *pud;
	pmd_t *pmd = NULL;
616
	pmd_t pmde;
B
Bob Liu 已提交
617 618 619 620 621 622 623 624 625 626

	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);
627 628 629 630 631
	/*
	 * 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.
	 */
632 633
	pmde = *pmd;
	barrier();
634
	if (!pmd_present(pmde) || pmd_trans_huge(pmde))
B
Bob Liu 已提交
635 636 637 638 639
		pmd = NULL;
out:
	return pmd;
}

N
Nikita Danilov 已提交
640 641 642
/*
 * Check that @page is mapped at @address into @mm.
 *
N
Nick Piggin 已提交
643 644 645 646
 * 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).
 *
647
 * On success returns with pte mapped and locked.
N
Nikita Danilov 已提交
648
 */
649
pte_t *__page_check_address(struct page *page, struct mm_struct *mm,
N
Nick Piggin 已提交
650
			  unsigned long address, spinlock_t **ptlp, int sync)
N
Nikita Danilov 已提交
651 652 653
{
	pmd_t *pmd;
	pte_t *pte;
H
Hugh Dickins 已提交
654
	spinlock_t *ptl;
N
Nikita Danilov 已提交
655

656
	if (unlikely(PageHuge(page))) {
657
		/* when pud is not present, pte will be NULL */
658
		pte = huge_pte_offset(mm, address);
659 660 661
		if (!pte)
			return NULL;

662
		ptl = huge_pte_lockptr(page_hstate(page), mm, pte);
663 664 665
		goto check;
	}

B
Bob Liu 已提交
666 667
	pmd = mm_find_pmd(mm, address);
	if (!pmd)
H
Hugh Dickins 已提交
668 669 670 671
		return NULL;

	pte = pte_offset_map(pmd, address);
	/* Make a quick check before getting the lock */
N
Nick Piggin 已提交
672
	if (!sync && !pte_present(*pte)) {
H
Hugh Dickins 已提交
673 674 675 676
		pte_unmap(pte);
		return NULL;
	}

H
Hugh Dickins 已提交
677
	ptl = pte_lockptr(mm, pmd);
678
check:
H
Hugh Dickins 已提交
679 680 681 682
	spin_lock(ptl);
	if (pte_present(*pte) && page_to_pfn(page) == pte_pfn(*pte)) {
		*ptlp = ptl;
		return pte;
N
Nikita Danilov 已提交
683
	}
H
Hugh Dickins 已提交
684 685
	pte_unmap_unlock(pte, ptl);
	return NULL;
N
Nikita Danilov 已提交
686 687
}

N
Nick Piggin 已提交
688 689 690 691 692 693 694 695 696
/**
 * 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.
 */
697
int page_mapped_in_vma(struct page *page, struct vm_area_struct *vma)
N
Nick Piggin 已提交
698 699 700 701 702
{
	unsigned long address;
	pte_t *pte;
	spinlock_t *ptl;

703 704
	address = __vma_address(page, vma);
	if (unlikely(address < vma->vm_start || address >= vma->vm_end))
N
Nick Piggin 已提交
705 706 707 708 709 710 711 712 713
		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;
}

714 715 716 717 718 719
struct page_referenced_arg {
	int mapcount;
	int referenced;
	unsigned long vm_flags;
	struct mem_cgroup *memcg;
};
L
Linus Torvalds 已提交
720
/*
721
 * arg: page_referenced_arg will be passed
L
Linus Torvalds 已提交
722
 */
723
static int page_referenced_one(struct page *page, struct vm_area_struct *vma,
724
			unsigned long address, void *arg)
L
Linus Torvalds 已提交
725 726
{
	struct mm_struct *mm = vma->vm_mm;
727
	spinlock_t *ptl;
L
Linus Torvalds 已提交
728
	int referenced = 0;
729
	struct page_referenced_arg *pra = arg;
L
Linus Torvalds 已提交
730

731 732 733
	if (unlikely(PageTransHuge(page))) {
		pmd_t *pmd;

734 735 736 737
		/*
		 * rmap might return false positives; we must filter
		 * these out using page_check_address_pmd().
		 */
738
		pmd = page_check_address_pmd(page, mm, address,
739 740
					     PAGE_CHECK_ADDRESS_PMD_FLAG, &ptl);
		if (!pmd)
741
			return SWAP_AGAIN;
742 743

		if (vma->vm_flags & VM_LOCKED) {
744
			spin_unlock(ptl);
745 746
			pra->vm_flags |= VM_LOCKED;
			return SWAP_FAIL; /* To break the loop */
747 748 749 750
		}

		/* go ahead even if the pmd is pmd_trans_splitting() */
		if (pmdp_clear_flush_young_notify(vma, address, pmd))
751
			referenced++;
752
		spin_unlock(ptl);
753 754 755
	} else {
		pte_t *pte;

756 757 758 759
		/*
		 * rmap might return false positives; we must filter
		 * these out using page_check_address().
		 */
760 761
		pte = page_check_address(page, mm, address, &ptl, 0);
		if (!pte)
762
			return SWAP_AGAIN;
763

764 765
		if (vma->vm_flags & VM_LOCKED) {
			pte_unmap_unlock(pte, ptl);
766 767
			pra->vm_flags |= VM_LOCKED;
			return SWAP_FAIL; /* To break the loop */
768 769
		}

770 771 772 773 774 775 776 777
		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.
			 */
778
			if (likely(!(vma->vm_flags & VM_SEQ_READ)))
779 780 781 782 783
				referenced++;
		}
		pte_unmap_unlock(pte, ptl);
	}

784 785 786
	if (referenced) {
		pra->referenced++;
		pra->vm_flags |= vma->vm_flags;
L
Linus Torvalds 已提交
787
	}
788

789 790 791 792 793
	pra->mapcount--;
	if (!pra->mapcount)
		return SWAP_SUCCESS; /* To break the loop */

	return SWAP_AGAIN;
L
Linus Torvalds 已提交
794 795
}

796
static bool invalid_page_referenced_vma(struct vm_area_struct *vma, void *arg)
L
Linus Torvalds 已提交
797
{
798 799
	struct page_referenced_arg *pra = arg;
	struct mem_cgroup *memcg = pra->memcg;
L
Linus Torvalds 已提交
800

801 802
	if (!mm_match_cgroup(vma->vm_mm, memcg))
		return true;
L
Linus Torvalds 已提交
803

804
	return false;
L
Linus Torvalds 已提交
805 806 807 808 809 810
}

/**
 * page_referenced - test if the page was referenced
 * @page: the page to test
 * @is_locked: caller holds lock on the page
811
 * @memcg: target memory cgroup
812
 * @vm_flags: collect encountered vma->vm_flags who actually referenced the page
L
Linus Torvalds 已提交
813 814 815 816
 *
 * Quick test_and_clear_referenced for all mappings to a page,
 * returns the number of ptes which referenced the page.
 */
817 818
int page_referenced(struct page *page,
		    int is_locked,
819
		    struct mem_cgroup *memcg,
820
		    unsigned long *vm_flags)
L
Linus Torvalds 已提交
821
{
822
	int ret;
H
Hugh Dickins 已提交
823
	int we_locked = 0;
824 825 826 827 828 829 830 831 832
	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
833

834
	*vm_flags = 0;
835 836 837 838 839 840 841 842 843 844
	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
845
	}
846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862

	/*
	 * 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
863 864
}

H
Hugh Dickins 已提交
865
static int page_mkclean_one(struct page *page, struct vm_area_struct *vma,
866
			    unsigned long address, void *arg)
867 868
{
	struct mm_struct *mm = vma->vm_mm;
869
	pte_t *pte;
870 871
	spinlock_t *ptl;
	int ret = 0;
872
	int *cleaned = arg;
873

N
Nick Piggin 已提交
874
	pte = page_check_address(page, mm, address, &ptl, 1);
875 876 877
	if (!pte)
		goto out;

878 879
	if (pte_dirty(*pte) || pte_write(*pte)) {
		pte_t entry;
880

881
		flush_cache_page(vma, address, pte_pfn(*pte));
882
		entry = ptep_clear_flush(vma, address, pte);
883 884
		entry = pte_wrprotect(entry);
		entry = pte_mkclean(entry);
885
		set_pte_at(mm, address, pte, entry);
886 887
		ret = 1;
	}
888 889

	pte_unmap_unlock(pte, ptl);
890

891
	if (ret) {
892
		mmu_notifier_invalidate_page(mm, address);
893 894
		(*cleaned)++;
	}
895
out:
896
	return SWAP_AGAIN;
897 898
}

899
static bool invalid_mkclean_vma(struct vm_area_struct *vma, void *arg)
900
{
901
	if (vma->vm_flags & VM_SHARED)
F
Fengguang Wu 已提交
902
		return false;
903

F
Fengguang Wu 已提交
904
	return true;
905 906 907 908
}

int page_mkclean(struct page *page)
{
909 910 911 912 913 914 915
	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,
	};
916 917 918

	BUG_ON(!PageLocked(page));

919 920 921 922 923 924 925 926
	if (!page_mapped(page))
		return 0;

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

	rmap_walk(page, &rwc);
927

928
	return cleaned;
929
}
J
Jaya Kumar 已提交
930
EXPORT_SYMBOL_GPL(page_mkclean);
931

932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947
/**
 * 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;

948
	VM_BUG_ON_PAGE(!PageLocked(page), page);
949
	VM_BUG_ON_VMA(!anon_vma, vma);
950
	VM_BUG_ON_PAGE(page->index != linear_page_index(vma, address), page);
951 952

	anon_vma = (void *) anon_vma + PAGE_MAPPING_ANON;
953 954 955 956 957 958
	/*
	 * Ensure that anon_vma and the PAGE_MAPPING_ANON bit are written
	 * simultaneously, so a concurrent reader (eg page_referenced()'s
	 * PageAnon()) will not see one without the other.
	 */
	WRITE_ONCE(page->mapping, (struct address_space *) anon_vma);
959 960
}

N
Nick Piggin 已提交
961
/**
A
Andi Kleen 已提交
962 963 964 965
 * __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	
966
 * @exclusive:	the page is exclusively owned by the current process
N
Nick Piggin 已提交
967 968
 */
static void __page_set_anon_rmap(struct page *page,
969
	struct vm_area_struct *vma, unsigned long address, int exclusive)
N
Nick Piggin 已提交
970
{
971
	struct anon_vma *anon_vma = vma->anon_vma;
972

973
	BUG_ON(!anon_vma);
974

A
Andi Kleen 已提交
975 976 977
	if (PageAnon(page))
		return;

978
	/*
979 980 981
	 * If the page isn't exclusively mapped into this vma,
	 * we must use the _oldest_ possible anon_vma for the
	 * page mapping!
982
	 */
A
Andi Kleen 已提交
983
	if (!exclusive)
984
		anon_vma = anon_vma->root;
N
Nick Piggin 已提交
985 986 987 988 989 990

	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
991
/**
R
Randy Dunlap 已提交
992
 * __page_check_anon_rmap - sanity check anonymous rmap addition
N
Nick Piggin 已提交
993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012
 * @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
1013
	BUG_ON(page_anon_vma(page)->root != vma->anon_vma->root);
N
Nick Piggin 已提交
1014 1015 1016 1017
	BUG_ON(page->index != linear_page_index(vma, address));
#endif
}

L
Linus Torvalds 已提交
1018 1019 1020 1021 1022 1023
/**
 * 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
1024
 * The caller needs to hold the pte lock, and the page must be locked in
1025 1026 1027
 * 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
1028 1029 1030
 */
void page_add_anon_rmap(struct page *page,
	struct vm_area_struct *vma, unsigned long address)
1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041
{
	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
1042
{
H
Hugh Dickins 已提交
1043
	int first = atomic_inc_and_test(&page->_mapcount);
1044
	if (first) {
1045 1046 1047 1048 1049 1050
		/*
		 * 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.
		 */
1051
		if (PageTransHuge(page))
1052 1053
			__inc_zone_page_state(page,
					      NR_ANON_TRANSPARENT_HUGEPAGES);
1054 1055
		__mod_zone_page_state(page_zone(page), NR_ANON_PAGES,
				hpage_nr_pages(page));
1056
	}
H
Hugh Dickins 已提交
1057 1058 1059
	if (unlikely(PageKsm(page)))
		return;

1060
	VM_BUG_ON_PAGE(!PageLocked(page), page);
1061
	/* address might be in next vma when migration races vma_adjust */
H
Hugh Dickins 已提交
1062
	if (first)
1063
		__page_set_anon_rmap(page, vma, address, exclusive);
1064
	else
N
Nick Piggin 已提交
1065
		__page_check_anon_rmap(page, vma, address);
L
Linus Torvalds 已提交
1066 1067
}

R
Randy Dunlap 已提交
1068
/**
N
Nick Piggin 已提交
1069 1070 1071 1072 1073 1074 1075
 * 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
1076
 * Page does not have to be locked.
N
Nick Piggin 已提交
1077 1078 1079 1080
 */
void page_add_new_anon_rmap(struct page *page,
	struct vm_area_struct *vma, unsigned long address)
{
1081
	VM_BUG_ON_VMA(address < vma->vm_start || address >= vma->vm_end, vma);
1082 1083
	SetPageSwapBacked(page);
	atomic_set(&page->_mapcount, 0); /* increment count (starts at -1) */
1084
	if (PageTransHuge(page))
1085
		__inc_zone_page_state(page, NR_ANON_TRANSPARENT_HUGEPAGES);
1086 1087
	__mod_zone_page_state(page_zone(page), NR_ANON_PAGES,
			hpage_nr_pages(page));
1088
	__page_set_anon_rmap(page, vma, address, 1);
N
Nick Piggin 已提交
1089 1090
}

L
Linus Torvalds 已提交
1091 1092 1093 1094
/**
 * page_add_file_rmap - add pte mapping to a file page
 * @page: the page to add the mapping to
 *
1095
 * The caller needs to hold the pte lock.
L
Linus Torvalds 已提交
1096 1097 1098
 */
void page_add_file_rmap(struct page *page)
{
1099
	struct mem_cgroup *memcg;
1100

1101
	memcg = mem_cgroup_begin_page_stat(page);
1102
	if (atomic_inc_and_test(&page->_mapcount)) {
1103
		__inc_zone_page_state(page, NR_FILE_MAPPED);
1104
		mem_cgroup_inc_page_stat(memcg, MEM_CGROUP_STAT_FILE_MAPPED);
1105
	}
1106
	mem_cgroup_end_page_stat(memcg);
L
Linus Torvalds 已提交
1107 1108
}

1109 1110 1111 1112
static void page_remove_file_rmap(struct page *page)
{
	struct mem_cgroup *memcg;

1113
	memcg = mem_cgroup_begin_page_stat(page);
1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133

	/* 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:
1134
	mem_cgroup_end_page_stat(memcg);
1135 1136
}

L
Linus Torvalds 已提交
1137 1138 1139 1140
/**
 * page_remove_rmap - take down pte mapping from a page
 * @page: page to remove mapping from
 *
1141
 * The caller needs to hold the pte lock.
L
Linus Torvalds 已提交
1142
 */
1143
void page_remove_rmap(struct page *page)
L
Linus Torvalds 已提交
1144
{
1145 1146 1147 1148
	if (!PageAnon(page)) {
		page_remove_file_rmap(page);
		return;
	}
1149

1150 1151
	/* page still mapped by someone else? */
	if (!atomic_add_negative(-1, &page->_mapcount))
1152 1153 1154 1155 1156
		return;

	/* Hugepages are not counted in NR_ANON_PAGES for now. */
	if (unlikely(PageHuge(page)))
		return;
1157

1158
	/*
1159 1160 1161
	 * 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.
1162
	 */
1163 1164 1165 1166 1167 1168
	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));

1169 1170
	if (unlikely(PageMlocked(page)))
		clear_page_mlock(page);
1171

1172 1173 1174 1175 1176 1177 1178 1179 1180
	/*
	 * 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
1181 1182 1183
}

/*
1184
 * @arg: enum ttu_flags will be passed to this argument
L
Linus Torvalds 已提交
1185
 */
1186
static int try_to_unmap_one(struct page *page, struct vm_area_struct *vma,
1187
		     unsigned long address, void *arg)
L
Linus Torvalds 已提交
1188 1189 1190 1191
{
	struct mm_struct *mm = vma->vm_mm;
	pte_t *pte;
	pte_t pteval;
H
Hugh Dickins 已提交
1192
	spinlock_t *ptl;
L
Linus Torvalds 已提交
1193
	int ret = SWAP_AGAIN;
1194
	enum ttu_flags flags = (enum ttu_flags)arg;
L
Linus Torvalds 已提交
1195

N
Nick Piggin 已提交
1196
	pte = page_check_address(page, mm, address, &ptl, 0);
H
Hugh Dickins 已提交
1197
	if (!pte)
N
Nikita Danilov 已提交
1198
		goto out;
L
Linus Torvalds 已提交
1199 1200 1201 1202 1203 1204

	/*
	 * 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.
	 */
1205
	if (!(flags & TTU_IGNORE_MLOCK)) {
K
KOSAKI Motohiro 已提交
1206 1207 1208
		if (vma->vm_flags & VM_LOCKED)
			goto out_mlock;

1209
		if (flags & TTU_MUNLOCK)
H
Hugh Dickins 已提交
1210
			goto out_unmap;
1211 1212
	}
	if (!(flags & TTU_IGNORE_ACCESS)) {
N
Nick Piggin 已提交
1213 1214 1215 1216 1217
		if (ptep_clear_flush_young_notify(vma, address, pte)) {
			ret = SWAP_FAIL;
			goto out_unmap;
		}
  	}
L
Linus Torvalds 已提交
1218 1219 1220

	/* Nuke the page table entry. */
	flush_cache_page(vma, address, page_to_pfn(page));
1221
	pteval = ptep_clear_flush(vma, address, pte);
L
Linus Torvalds 已提交
1222 1223 1224 1225 1226

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

1227 1228 1229
	/* Update high watermark before we lower rss */
	update_hiwater_rss(mm);

1230
	if (PageHWPoison(page) && !(flags & TTU_IGNORE_HWPOISON)) {
1231 1232 1233 1234 1235 1236
		if (!PageHuge(page)) {
			if (PageAnon(page))
				dec_mm_counter(mm, MM_ANONPAGES);
			else
				dec_mm_counter(mm, MM_FILEPAGES);
		}
1237
		set_pte_at(mm, address, pte,
1238
			   swp_entry_to_pte(make_hwpoison_entry(page)));
1239 1240 1241 1242 1243 1244 1245 1246 1247 1248
	} 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);
1249
	} else if (PageAnon(page)) {
H
Hugh Dickins 已提交
1250
		swp_entry_t entry = { .val = page_private(page) };
1251
		pte_t swp_pte;
1252 1253 1254 1255 1256 1257

		if (PageSwapCache(page)) {
			/*
			 * Store the swap location in the pte.
			 * See handle_pte_fault() ...
			 */
H
Hugh Dickins 已提交
1258 1259 1260 1261 1262
			if (swap_duplicate(entry) < 0) {
				set_pte_at(mm, address, pte, pteval);
				ret = SWAP_FAIL;
				goto out_unmap;
			}
1263 1264 1265 1266 1267 1268
			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
1269
			dec_mm_counter(mm, MM_ANONPAGES);
K
KAMEZAWA Hiroyuki 已提交
1270
			inc_mm_counter(mm, MM_SWAPENTS);
1271
		} else if (IS_ENABLED(CONFIG_MIGRATION)) {
1272 1273 1274 1275 1276
			/*
			 * 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.
			 */
1277
			BUG_ON(!(flags & TTU_MIGRATION));
1278
			entry = make_migration_entry(page, pte_write(pteval));
L
Linus Torvalds 已提交
1279
		}
1280 1281 1282 1283
		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);
1284
	} else if (IS_ENABLED(CONFIG_MIGRATION) &&
1285
		   (flags & TTU_MIGRATION)) {
1286 1287 1288 1289 1290
		/* 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
1291
		dec_mm_counter(mm, MM_FILEPAGES);
L
Linus Torvalds 已提交
1292

1293
	page_remove_rmap(page);
L
Linus Torvalds 已提交
1294 1295 1296
	page_cache_release(page);

out_unmap:
H
Hugh Dickins 已提交
1297
	pte_unmap_unlock(pte, ptl);
1298
	if (ret != SWAP_FAIL && !(flags & TTU_MUNLOCK))
1299
		mmu_notifier_invalidate_page(mm, address);
K
KOSAKI Motohiro 已提交
1300 1301
out:
	return ret;
H
Hugh Dickins 已提交
1302

K
KOSAKI Motohiro 已提交
1303 1304 1305 1306 1307 1308 1309
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
1310
	 * we now hold anon_vma->rwsem or mapping->i_mmap_rwsem.
K
KOSAKI Motohiro 已提交
1311 1312 1313 1314 1315 1316 1317 1318
	 * 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
1319
		}
K
KOSAKI Motohiro 已提交
1320
		up_read(&vma->vm_mm->mmap_sem);
H
Hugh Dickins 已提交
1321
	}
L
Linus Torvalds 已提交
1322 1323 1324
	return ret;
}

1325
bool is_vma_temporary_stack(struct vm_area_struct *vma)
1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338
{
	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;
}

1339 1340 1341 1342 1343 1344 1345 1346 1347 1348
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
1349 1350 1351
/**
 * try_to_unmap - try to remove all page table mappings to a page
 * @page: the page to get unmapped
1352
 * @flags: action and flags
L
Linus Torvalds 已提交
1353 1354 1355 1356 1357 1358 1359 1360
 *
 * 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
1361
 * SWAP_MLOCK	- page is mlocked.
L
Linus Torvalds 已提交
1362
 */
1363
int try_to_unmap(struct page *page, enum ttu_flags flags)
L
Linus Torvalds 已提交
1364 1365
{
	int ret;
1366 1367 1368 1369 1370 1371
	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
1372

1373
	VM_BUG_ON_PAGE(!PageHuge(page) && PageTransHuge(page), page);
L
Linus Torvalds 已提交
1374

1375 1376 1377 1378 1379 1380 1381 1382
	/*
	 * 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.
	 */
1383
	if ((flags & TTU_MIGRATION) && !PageKsm(page) && PageAnon(page))
1384 1385 1386 1387
		rwc.invalid_vma = invalid_migration_vma;

	ret = rmap_walk(page, &rwc);

N
Nick Piggin 已提交
1388
	if (ret != SWAP_MLOCK && !page_mapped(page))
L
Linus Torvalds 已提交
1389 1390 1391
		ret = SWAP_SUCCESS;
	return ret;
}
N
Nikita Danilov 已提交
1392

N
Nick Piggin 已提交
1393 1394 1395 1396 1397 1398 1399 1400 1401 1402
/**
 * 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
1403
 * SWAP_AGAIN	- no vma is holding page mlocked, or,
N
Nick Piggin 已提交
1404
 * SWAP_AGAIN	- page mapped in mlocked vma -- couldn't acquire mmap sem
H
Hugh Dickins 已提交
1405
 * SWAP_FAIL	- page cannot be located at present
N
Nick Piggin 已提交
1406 1407 1408 1409
 * SWAP_MLOCK	- page is now mlocked.
 */
int try_to_munlock(struct page *page)
{
1410 1411 1412 1413 1414 1415 1416 1417 1418
	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,

	};

1419
	VM_BUG_ON_PAGE(!PageLocked(page) || PageLRU(page), page);
N
Nick Piggin 已提交
1420

1421 1422
	ret = rmap_walk(page, &rwc);
	return ret;
N
Nick Piggin 已提交
1423
}
1424

P
Peter Zijlstra 已提交
1425
void __put_anon_vma(struct anon_vma *anon_vma)
1426
{
P
Peter Zijlstra 已提交
1427
	struct anon_vma *root = anon_vma->root;
1428

1429
	anon_vma_free(anon_vma);
P
Peter Zijlstra 已提交
1430 1431
	if (root != anon_vma && atomic_dec_and_test(&root->refcount))
		anon_vma_free(root);
1432 1433
}

1434 1435
static struct anon_vma *rmap_walk_anon_lock(struct page *page,
					struct rmap_walk_control *rwc)
1436 1437 1438
{
	struct anon_vma *anon_vma;

1439 1440 1441
	if (rwc->anon_lock)
		return rwc->anon_lock(page);

1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455
	/*
	 * 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;
}

1456
/*
1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468
 * 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.
1469
 */
1470
static int rmap_walk_anon(struct page *page, struct rmap_walk_control *rwc)
1471 1472
{
	struct anon_vma *anon_vma;
1473
	pgoff_t pgoff;
1474
	struct anon_vma_chain *avc;
1475 1476
	int ret = SWAP_AGAIN;

1477
	anon_vma = rmap_walk_anon_lock(page, rwc);
1478 1479
	if (!anon_vma)
		return ret;
1480

1481
	pgoff = page_to_pgoff(page);
1482
	anon_vma_interval_tree_foreach(avc, &anon_vma->rb_root, pgoff, pgoff) {
1483
		struct vm_area_struct *vma = avc->vma;
1484
		unsigned long address = vma_address(page, vma);
1485 1486 1487 1488

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

1489
		ret = rwc->rmap_one(page, vma, address, rwc->arg);
1490 1491
		if (ret != SWAP_AGAIN)
			break;
1492 1493
		if (rwc->done && rwc->done(page))
			break;
1494
	}
1495
	anon_vma_unlock_read(anon_vma);
1496 1497 1498
	return ret;
}

1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511
/*
 * 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.
 */
1512
static int rmap_walk_file(struct page *page, struct rmap_walk_control *rwc)
1513 1514
{
	struct address_space *mapping = page->mapping;
1515
	pgoff_t pgoff;
1516 1517 1518
	struct vm_area_struct *vma;
	int ret = SWAP_AGAIN;

1519 1520 1521 1522
	/*
	 * 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,
1523
	 * so we can safely take mapping->i_mmap_rwsem.
1524
	 */
1525
	VM_BUG_ON_PAGE(!PageLocked(page), page);
1526

1527 1528
	if (!mapping)
		return ret;
D
Davidlohr Bueso 已提交
1529

1530
	pgoff = page_to_pgoff(page);
D
Davidlohr Bueso 已提交
1531
	i_mmap_lock_read(mapping);
1532
	vma_interval_tree_foreach(vma, &mapping->i_mmap, pgoff, pgoff) {
1533
		unsigned long address = vma_address(page, vma);
1534 1535 1536 1537

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

1538
		ret = rwc->rmap_one(page, vma, address, rwc->arg);
1539
		if (ret != SWAP_AGAIN)
1540 1541 1542
			goto done;
		if (rwc->done && rwc->done(page))
			goto done;
1543
	}
1544 1545

done:
D
Davidlohr Bueso 已提交
1546
	i_mmap_unlock_read(mapping);
1547 1548 1549
	return ret;
}

1550
int rmap_walk(struct page *page, struct rmap_walk_control *rwc)
1551 1552
{
	if (unlikely(PageKsm(page)))
1553
		return rmap_walk_ksm(page, rwc);
1554
	else if (PageAnon(page))
1555
		return rmap_walk_anon(page, rwc);
1556
	else
1557
		return rmap_walk_file(page, rwc);
1558
}
1559

N
Naoya Horiguchi 已提交
1560
#ifdef CONFIG_HUGETLB_PAGE
1561 1562 1563 1564 1565 1566 1567 1568 1569
/*
 * 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;
1570

1571
	BUG_ON(!anon_vma);
1572 1573 1574 1575 1576 1577

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

1578 1579 1580 1581 1582 1583 1584 1585 1586 1587
	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;
1588 1589

	BUG_ON(!PageLocked(page));
1590
	BUG_ON(!anon_vma);
1591
	/* address might be in next vma when migration races vma_adjust */
1592 1593 1594 1595 1596 1597 1598 1599 1600 1601 1602 1603
	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
1604
#endif /* CONFIG_HUGETLB_PAGE */