rmap.c 50.7 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_mutex
27
 *         anon_vma->rwsem
28 29 30 31 32
 *           mm->page_table_lock or pte_lock
 *             zone->lru_lock (in mark_page_accessed, isolate_lru_page)
 *             swap_lock (in swap_duplicate, swap_info_get)
 *               mmlist_lock (in mmput, drain_mmlist and others)
 *               mapping->private_lock (in __set_page_dirty_buffers)
33
 *               inode->i_lock (in set_page_dirty's __mark_inode_dirty)
34
 *               bdi.wb->list_lock (in set_page_dirty's __mark_inode_dirty)
35 36 37
 *                 sb_lock (within inode_lock in fs/fs-writeback.c)
 *                 mapping->tree_lock (widely used, in set_page_dirty,
 *                           in arch-dependent flush_dcache_mmap_lock,
38
 *                           within bdi.wb->list_lock in __sync_single_inode)
39
 *
40
 * anon_vma->rwsem,mapping->i_mutex      (memory_failure, collect_procs_anon)
41
 *   ->tasklist_lock
42
 *     pte map lock
L
Linus Torvalds 已提交
43 44 45 46 47 48 49 50
 */

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

#include <asm/tlbflush.h>

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

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

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

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

	return anon_vma;
83 84
}

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

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

111 112
	kmem_cache_free(anon_vma_cachep, anon_vma);
}
L
Linus Torvalds 已提交
113

114
static inline struct anon_vma_chain *anon_vma_chain_alloc(gfp_t gfp)
115
{
116
	return kmem_cache_alloc(anon_vma_chain_cachep, gfp);
117 118
}

119
static void anon_vma_chain_free(struct anon_vma_chain *anon_vma_chain)
120 121 122 123
{
	kmem_cache_free(anon_vma_chain_cachep, anon_vma_chain);
}

124 125 126 127 128 129 130
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);
131
	anon_vma_interval_tree_insert(avc, &anon_vma->rb_root);
132 133
}

134 135 136 137 138 139 140 141 142
/**
 * 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
143
 * not we either need to find an adjacent mapping that we
144 145 146 147 148
 * 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
149
 * optimistically looked up an anon_vma in page_lock_anon_vma_read()
150 151 152 153 154 155 156 157 158 159 160
 * 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
161 162 163
int anon_vma_prepare(struct vm_area_struct *vma)
{
	struct anon_vma *anon_vma = vma->anon_vma;
164
	struct anon_vma_chain *avc;
L
Linus Torvalds 已提交
165 166 167 168

	might_sleep();
	if (unlikely(!anon_vma)) {
		struct mm_struct *mm = vma->vm_mm;
169
		struct anon_vma *allocated;
L
Linus Torvalds 已提交
170

171
		avc = anon_vma_chain_alloc(GFP_KERNEL);
172 173 174
		if (!avc)
			goto out_enomem;

L
Linus Torvalds 已提交
175
		anon_vma = find_mergeable_anon_vma(vma);
176 177
		allocated = NULL;
		if (!anon_vma) {
L
Linus Torvalds 已提交
178 179
			anon_vma = anon_vma_alloc();
			if (unlikely(!anon_vma))
180
				goto out_enomem_free_avc;
L
Linus Torvalds 已提交
181 182 183
			allocated = anon_vma;
		}

184
		anon_vma_lock_write(anon_vma);
L
Linus Torvalds 已提交
185 186 187 188
		/* page_table_lock to protect against threads */
		spin_lock(&mm->page_table_lock);
		if (likely(!vma->anon_vma)) {
			vma->anon_vma = anon_vma;
189
			anon_vma_chain_link(vma, avc, anon_vma);
L
Linus Torvalds 已提交
190
			allocated = NULL;
191
			avc = NULL;
L
Linus Torvalds 已提交
192 193
		}
		spin_unlock(&mm->page_table_lock);
194
		anon_vma_unlock_write(anon_vma);
195 196

		if (unlikely(allocated))
P
Peter Zijlstra 已提交
197
			put_anon_vma(allocated);
198
		if (unlikely(avc))
199
			anon_vma_chain_free(avc);
L
Linus Torvalds 已提交
200 201
	}
	return 0;
202 203 204 205 206

 out_enomem_free_avc:
	anon_vma_chain_free(avc);
 out_enomem:
	return -ENOMEM;
L
Linus Torvalds 已提交
207 208
}

209 210 211 212 213 214 215 216 217 218 219 220 221
/*
 * 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))
222
			up_write(&root->rwsem);
223
		root = new_root;
224
		down_write(&root->rwsem);
225 226 227 228 229 230 231
	}
	return root;
}

static inline void unlock_anon_vma_root(struct anon_vma *root)
{
	if (root)
232
		up_write(&root->rwsem);
233 234
}

235 236 237 238 239
/*
 * Attach the anon_vmas from src to dst.
 * Returns 0 on success, -ENOMEM on failure.
 */
int anon_vma_clone(struct vm_area_struct *dst, struct vm_area_struct *src)
L
Linus Torvalds 已提交
240
{
241
	struct anon_vma_chain *avc, *pavc;
242
	struct anon_vma *root = NULL;
243

244
	list_for_each_entry_reverse(pavc, &src->anon_vma_chain, same_vma) {
245 246
		struct anon_vma *anon_vma;

247 248 249 250 251 252 253 254
		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;
		}
255 256 257
		anon_vma = pavc->anon_vma;
		root = lock_anon_vma_root(root, anon_vma);
		anon_vma_chain_link(dst, avc, anon_vma);
258
	}
259
	unlock_anon_vma_root(root);
260
	return 0;
L
Linus Torvalds 已提交
261

262 263 264
 enomem_failure:
	unlink_anon_vmas(dst);
	return -ENOMEM;
L
Linus Torvalds 已提交
265 266
}

267 268 269 270 271 272
/*
 * 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
273
{
274 275
	struct anon_vma_chain *avc;
	struct anon_vma *anon_vma;
L
Linus Torvalds 已提交
276

277 278 279 280 281 282 283 284 285 286 287 288 289 290 291
	/* Don't bother if the parent process has no anon_vma here. */
	if (!pvma->anon_vma)
		return 0;

	/*
	 * First, attach the new VMA to the parent VMA's anon_vmas,
	 * so rmap can find non-COWed pages in child processes.
	 */
	if (anon_vma_clone(vma, pvma))
		return -ENOMEM;

	/* Then add our own anon_vma. */
	anon_vma = anon_vma_alloc();
	if (!anon_vma)
		goto out_error;
292
	avc = anon_vma_chain_alloc(GFP_KERNEL);
293 294
	if (!avc)
		goto out_error_free_anon_vma;
295 296 297 298 299 300

	/*
	 * 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;
301
	/*
P
Peter Zijlstra 已提交
302 303 304
	 * 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.
305 306
	 */
	get_anon_vma(anon_vma->root);
307 308
	/* Mark this anon_vma as the one where our new (COWed) pages go. */
	vma->anon_vma = anon_vma;
309
	anon_vma_lock_write(anon_vma);
310
	anon_vma_chain_link(vma, avc, anon_vma);
311
	anon_vma_unlock_write(anon_vma);
312 313 314 315

	return 0;

 out_error_free_anon_vma:
P
Peter Zijlstra 已提交
316
	put_anon_vma(anon_vma);
317
 out_error:
318
	unlink_anon_vmas(vma);
319
	return -ENOMEM;
L
Linus Torvalds 已提交
320 321
}

322 323 324
void unlink_anon_vmas(struct vm_area_struct *vma)
{
	struct anon_vma_chain *avc, *next;
325
	struct anon_vma *root = NULL;
326

327 328 329 330
	/*
	 * Unlink each anon_vma chained to the VMA.  This list is ordered
	 * from newest to oldest, ensuring the root anon_vma gets freed last.
	 */
331
	list_for_each_entry_safe(avc, next, &vma->anon_vma_chain, same_vma) {
332 333 334
		struct anon_vma *anon_vma = avc->anon_vma;

		root = lock_anon_vma_root(root, anon_vma);
335
		anon_vma_interval_tree_remove(avc, &anon_vma->rb_root);
336 337 338 339 340

		/*
		 * Leave empty anon_vmas on the list - we'll need
		 * to free them outside the lock.
		 */
341
		if (RB_EMPTY_ROOT(&anon_vma->rb_root))
342 343 344 345 346 347 348 349 350 351
			continue;

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

	/*
	 * Iterate the list once more, it now only contains empty and unlinked
	 * anon_vmas, destroy them. Could not do before due to __put_anon_vma()
352
	 * needing to write-acquire the anon_vma->root->rwsem.
353 354 355 356 357 358
	 */
	list_for_each_entry_safe(avc, next, &vma->anon_vma_chain, same_vma) {
		struct anon_vma *anon_vma = avc->anon_vma;

		put_anon_vma(anon_vma);

359 360 361 362 363
		list_del(&avc->same_vma);
		anon_vma_chain_free(avc);
	}
}

364
static void anon_vma_ctor(void *data)
L
Linus Torvalds 已提交
365
{
C
Christoph Lameter 已提交
366
	struct anon_vma *anon_vma = data;
L
Linus Torvalds 已提交
367

368
	init_rwsem(&anon_vma->rwsem);
369
	atomic_set(&anon_vma->refcount, 0);
370
	anon_vma->rb_root = RB_ROOT;
L
Linus Torvalds 已提交
371 372 373 374 375
}

void __init anon_vma_init(void)
{
	anon_vma_cachep = kmem_cache_create("anon_vma", sizeof(struct anon_vma),
376
			0, SLAB_DESTROY_BY_RCU|SLAB_PANIC, anon_vma_ctor);
377
	anon_vma_chain_cachep = KMEM_CACHE(anon_vma_chain, SLAB_PANIC);
L
Linus Torvalds 已提交
378 379 380
}

/*
381 382 383 384 385 386 387 388 389
 * 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).
 *
390 391 392 393 394
 * 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 ].
 *
395 396 397 398 399 400 401
 * 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
402
 */
403
struct anon_vma *page_get_anon_vma(struct page *page)
L
Linus Torvalds 已提交
404
{
405
	struct anon_vma *anon_vma = NULL;
L
Linus Torvalds 已提交
406 407 408
	unsigned long anon_mapping;

	rcu_read_lock();
409
	anon_mapping = (unsigned long) ACCESS_ONCE(page->mapping);
H
Hugh Dickins 已提交
410
	if ((anon_mapping & PAGE_MAPPING_FLAGS) != PAGE_MAPPING_ANON)
L
Linus Torvalds 已提交
411 412 413 414 415
		goto out;
	if (!page_mapped(page))
		goto out;

	anon_vma = (struct anon_vma *) (anon_mapping - PAGE_MAPPING_ANON);
416 417 418 419
	if (!atomic_inc_not_zero(&anon_vma->refcount)) {
		anon_vma = NULL;
		goto out;
	}
420 421 422

	/*
	 * If this page is still mapped, then its anon_vma cannot have been
423 424 425 426
	 * 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).
427
	 */
428 429 430 431
	if (!page_mapped(page)) {
		put_anon_vma(anon_vma);
		anon_vma = NULL;
	}
L
Linus Torvalds 已提交
432 433
out:
	rcu_read_unlock();
434 435 436 437

	return anon_vma;
}

438 439 440 441 442 443 444
/*
 * 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.
 */
445
struct anon_vma *page_lock_anon_vma_read(struct page *page)
446
{
447
	struct anon_vma *anon_vma = NULL;
448
	struct anon_vma *root_anon_vma;
449
	unsigned long anon_mapping;
450

451 452 453 454 455 456 457 458
	rcu_read_lock();
	anon_mapping = (unsigned long) ACCESS_ONCE(page->mapping);
	if ((anon_mapping & PAGE_MAPPING_FLAGS) != PAGE_MAPPING_ANON)
		goto out;
	if (!page_mapped(page))
		goto out;

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

473 474 475 476 477 478 479 480 481 482 483 484 485 486
	/* trylock failed, we got to sleep */
	if (!atomic_inc_not_zero(&anon_vma->refcount)) {
		anon_vma = NULL;
		goto out;
	}

	if (!page_mapped(page)) {
		put_anon_vma(anon_vma);
		anon_vma = NULL;
		goto out;
	}

	/* we pinned the anon_vma, its safe to sleep */
	rcu_read_unlock();
487
	anon_vma_lock_read(anon_vma);
488 489 490 491 492

	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
493
		 * we'll deadlock on the anon_vma_lock_write() recursion.
494
		 */
495
		anon_vma_unlock_read(anon_vma);
496 497 498 499 500 501 502 503
		__put_anon_vma(anon_vma);
		anon_vma = NULL;
	}

	return anon_vma;

out:
	rcu_read_unlock();
504
	return anon_vma;
505 506
}

507
void page_unlock_anon_vma_read(struct anon_vma *anon_vma)
508
{
509
	anon_vma_unlock_read(anon_vma);
L
Linus Torvalds 已提交
510 511 512
}

/*
513
 * At what user virtual address is page expected in @vma?
L
Linus Torvalds 已提交
514
 */
515 516
static inline unsigned long
__vma_address(struct page *page, struct vm_area_struct *vma)
L
Linus Torvalds 已提交
517 518 519
{
	pgoff_t pgoff = page->index << (PAGE_CACHE_SHIFT - PAGE_SHIFT);

520 521
	if (unlikely(is_vm_hugetlb_page(vma)))
		pgoff = page->index << huge_page_order(page_hstate(page));
522 523 524 525 526 527 528 529 530 531 532 533

	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 */
	VM_BUG_ON(address < vma->vm_start || address >= vma->vm_end);

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

/*
H
Huang Shijie 已提交
538
 * At what user virtual address is page expected in vma?
539
 * Caller should check the page is actually part of the vma.
L
Linus Torvalds 已提交
540 541 542
 */
unsigned long page_address_in_vma(struct page *page, struct vm_area_struct *vma)
{
543
	unsigned long address;
544
	if (PageAnon(page)) {
545 546 547 548 549 550 551
		struct anon_vma *page__anon_vma = page_anon_vma(page);
		/*
		 * Note: swapoff's unuse_vma() is more efficient with this
		 * check, and needs it to match anon_vma when KSM is active.
		 */
		if (!vma->anon_vma || !page__anon_vma ||
		    vma->anon_vma->root != page__anon_vma->root)
552 553
			return -EFAULT;
	} else if (page->mapping && !(vma->vm_flags & VM_NONLINEAR)) {
H
Hugh Dickins 已提交
554 555
		if (!vma->vm_file ||
		    vma->vm_file->f_mapping != page->mapping)
L
Linus Torvalds 已提交
556 557 558
			return -EFAULT;
	} else
		return -EFAULT;
559 560 561 562
	address = __vma_address(page, vma);
	if (unlikely(address < vma->vm_start || address >= vma->vm_end))
		return -EFAULT;
	return address;
L
Linus Torvalds 已提交
563 564
}

B
Bob Liu 已提交
565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585
pmd_t *mm_find_pmd(struct mm_struct *mm, unsigned long address)
{
	pgd_t *pgd;
	pud_t *pud;
	pmd_t *pmd = NULL;

	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);
	if (!pmd_present(*pmd))
		pmd = NULL;
out:
	return pmd;
}

N
Nikita Danilov 已提交
586 587 588
/*
 * Check that @page is mapped at @address into @mm.
 *
N
Nick Piggin 已提交
589 590 591 592
 * 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).
 *
593
 * On success returns with pte mapped and locked.
N
Nikita Danilov 已提交
594
 */
595
pte_t *__page_check_address(struct page *page, struct mm_struct *mm,
N
Nick Piggin 已提交
596
			  unsigned long address, spinlock_t **ptlp, int sync)
N
Nikita Danilov 已提交
597 598 599
{
	pmd_t *pmd;
	pte_t *pte;
H
Hugh Dickins 已提交
600
	spinlock_t *ptl;
N
Nikita Danilov 已提交
601

602 603 604 605 606 607
	if (unlikely(PageHuge(page))) {
		pte = huge_pte_offset(mm, address);
		ptl = &mm->page_table_lock;
		goto check;
	}

B
Bob Liu 已提交
608 609
	pmd = mm_find_pmd(mm, address);
	if (!pmd)
H
Hugh Dickins 已提交
610 611
		return NULL;

612 613
	if (pmd_trans_huge(*pmd))
		return NULL;
H
Hugh Dickins 已提交
614 615 616

	pte = pte_offset_map(pmd, address);
	/* Make a quick check before getting the lock */
N
Nick Piggin 已提交
617
	if (!sync && !pte_present(*pte)) {
H
Hugh Dickins 已提交
618 619 620 621
		pte_unmap(pte);
		return NULL;
	}

H
Hugh Dickins 已提交
622
	ptl = pte_lockptr(mm, pmd);
623
check:
H
Hugh Dickins 已提交
624 625 626 627
	spin_lock(ptl);
	if (pte_present(*pte) && page_to_pfn(page) == pte_pfn(*pte)) {
		*ptlp = ptl;
		return pte;
N
Nikita Danilov 已提交
628
	}
H
Hugh Dickins 已提交
629 630
	pte_unmap_unlock(pte, ptl);
	return NULL;
N
Nikita Danilov 已提交
631 632
}

N
Nick Piggin 已提交
633 634 635 636 637 638 639 640 641
/**
 * 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.
 */
642
int page_mapped_in_vma(struct page *page, struct vm_area_struct *vma)
N
Nick Piggin 已提交
643 644 645 646 647
{
	unsigned long address;
	pte_t *pte;
	spinlock_t *ptl;

648 649
	address = __vma_address(page, vma);
	if (unlikely(address < vma->vm_start || address >= vma->vm_end))
N
Nick Piggin 已提交
650 651 652 653 654 655 656 657 658
		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;
}

L
Linus Torvalds 已提交
659 660 661 662
/*
 * Subfunctions of page_referenced: page_referenced_one called
 * repeatedly from either page_referenced_anon or page_referenced_file.
 */
H
Hugh Dickins 已提交
663 664 665
int page_referenced_one(struct page *page, struct vm_area_struct *vma,
			unsigned long address, unsigned int *mapcount,
			unsigned long *vm_flags)
L
Linus Torvalds 已提交
666 667 668 669
{
	struct mm_struct *mm = vma->vm_mm;
	int referenced = 0;

670 671 672 673
	if (unlikely(PageTransHuge(page))) {
		pmd_t *pmd;

		spin_lock(&mm->page_table_lock);
674 675 676 677
		/*
		 * rmap might return false positives; we must filter
		 * these out using page_check_address_pmd().
		 */
678 679
		pmd = page_check_address_pmd(page, mm, address,
					     PAGE_CHECK_ADDRESS_PMD_FLAG);
680 681 682 683 684 685 686 687 688 689 690 691 692 693
		if (!pmd) {
			spin_unlock(&mm->page_table_lock);
			goto out;
		}

		if (vma->vm_flags & VM_LOCKED) {
			spin_unlock(&mm->page_table_lock);
			*mapcount = 0;	/* break early from loop */
			*vm_flags |= VM_LOCKED;
			goto out;
		}

		/* go ahead even if the pmd is pmd_trans_splitting() */
		if (pmdp_clear_flush_young_notify(vma, address, pmd))
694 695 696 697 698 699
			referenced++;
		spin_unlock(&mm->page_table_lock);
	} else {
		pte_t *pte;
		spinlock_t *ptl;

700 701 702 703
		/*
		 * rmap might return false positives; we must filter
		 * these out using page_check_address().
		 */
704 705 706 707
		pte = page_check_address(page, mm, address, &ptl, 0);
		if (!pte)
			goto out;

708 709 710 711 712 713 714
		if (vma->vm_flags & VM_LOCKED) {
			pte_unmap_unlock(pte, ptl);
			*mapcount = 0;	/* break early from loop */
			*vm_flags |= VM_LOCKED;
			goto out;
		}

715 716 717 718 719 720 721 722
		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.
			 */
723
			if (likely(!(vma->vm_flags & VM_SEQ_READ)))
724 725 726 727 728
				referenced++;
		}
		pte_unmap_unlock(pte, ptl);
	}

H
Hugh Dickins 已提交
729
	(*mapcount)--;
730

731 732
	if (referenced)
		*vm_flags |= vma->vm_flags;
733
out:
L
Linus Torvalds 已提交
734 735 736
	return referenced;
}

737
static int page_referenced_anon(struct page *page,
738
				struct mem_cgroup *memcg,
739
				unsigned long *vm_flags)
L
Linus Torvalds 已提交
740 741 742
{
	unsigned int mapcount;
	struct anon_vma *anon_vma;
743
	pgoff_t pgoff;
744
	struct anon_vma_chain *avc;
L
Linus Torvalds 已提交
745 746
	int referenced = 0;

747
	anon_vma = page_lock_anon_vma_read(page);
L
Linus Torvalds 已提交
748 749 750 751
	if (!anon_vma)
		return referenced;

	mapcount = page_mapcount(page);
752 753
	pgoff = page->index << (PAGE_CACHE_SHIFT - PAGE_SHIFT);
	anon_vma_interval_tree_foreach(avc, &anon_vma->rb_root, pgoff, pgoff) {
754
		struct vm_area_struct *vma = avc->vma;
H
Hugh Dickins 已提交
755
		unsigned long address = vma_address(page, vma);
756 757 758 759 760
		/*
		 * If we are reclaiming on behalf of a cgroup, skip
		 * counting on behalf of references from different
		 * cgroups
		 */
761
		if (memcg && !mm_match_cgroup(vma->vm_mm, memcg))
762
			continue;
H
Hugh Dickins 已提交
763
		referenced += page_referenced_one(page, vma, address,
764
						  &mapcount, vm_flags);
L
Linus Torvalds 已提交
765 766 767
		if (!mapcount)
			break;
	}
768

769
	page_unlock_anon_vma_read(anon_vma);
L
Linus Torvalds 已提交
770 771 772 773 774 775
	return referenced;
}

/**
 * page_referenced_file - referenced check for object-based rmap
 * @page: the page we're checking references on.
776
 * @memcg: target memory control group
777
 * @vm_flags: collect encountered vma->vm_flags who actually referenced the page
L
Linus Torvalds 已提交
778 779 780 781 782 783 784 785
 *
 * For an object-based mapped page, find all the places it is mapped and
 * check/clear the referenced flag.  This is done by following the page->mapping
 * pointer, then walking the chain of vmas it holds.  It returns the number
 * of references it found.
 *
 * This function is only called from page_referenced for object-based pages.
 */
786
static int page_referenced_file(struct page *page,
787
				struct mem_cgroup *memcg,
788
				unsigned long *vm_flags)
L
Linus Torvalds 已提交
789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806
{
	unsigned int mapcount;
	struct address_space *mapping = page->mapping;
	pgoff_t pgoff = page->index << (PAGE_CACHE_SHIFT - PAGE_SHIFT);
	struct vm_area_struct *vma;
	int referenced = 0;

	/*
	 * The caller's checks on page->mapping and !PageAnon have made
	 * sure that this is a file page: the check for page->mapping
	 * excludes the case just before it gets set on an anon page.
	 */
	BUG_ON(PageAnon(page));

	/*
	 * The page lock not only makes sure that page->mapping cannot
	 * suddenly be NULLified by truncation, it makes sure that the
	 * structure at mapping cannot be freed and reused yet,
807
	 * so we can safely take mapping->i_mmap_mutex.
L
Linus Torvalds 已提交
808 809 810
	 */
	BUG_ON(!PageLocked(page));

811
	mutex_lock(&mapping->i_mmap_mutex);
L
Linus Torvalds 已提交
812 813

	/*
814
	 * i_mmap_mutex does not stabilize mapcount at all, but mapcount
L
Linus Torvalds 已提交
815 816 817 818
	 * is more likely to be accurate if we note it after spinning.
	 */
	mapcount = page_mapcount(page);

819
	vma_interval_tree_foreach(vma, &mapping->i_mmap, pgoff, pgoff) {
H
Hugh Dickins 已提交
820
		unsigned long address = vma_address(page, vma);
821 822 823 824 825
		/*
		 * If we are reclaiming on behalf of a cgroup, skip
		 * counting on behalf of references from different
		 * cgroups
		 */
826
		if (memcg && !mm_match_cgroup(vma->vm_mm, memcg))
827
			continue;
H
Hugh Dickins 已提交
828
		referenced += page_referenced_one(page, vma, address,
829
						  &mapcount, vm_flags);
L
Linus Torvalds 已提交
830 831 832 833
		if (!mapcount)
			break;
	}

834
	mutex_unlock(&mapping->i_mmap_mutex);
L
Linus Torvalds 已提交
835 836 837 838 839 840 841
	return referenced;
}

/**
 * page_referenced - test if the page was referenced
 * @page: the page to test
 * @is_locked: caller holds lock on the page
842
 * @memcg: target memory cgroup
843
 * @vm_flags: collect encountered vma->vm_flags who actually referenced the page
L
Linus Torvalds 已提交
844 845 846 847
 *
 * Quick test_and_clear_referenced for all mappings to a page,
 * returns the number of ptes which referenced the page.
 */
848 849
int page_referenced(struct page *page,
		    int is_locked,
850
		    struct mem_cgroup *memcg,
851
		    unsigned long *vm_flags)
L
Linus Torvalds 已提交
852 853
{
	int referenced = 0;
H
Hugh Dickins 已提交
854
	int we_locked = 0;
L
Linus Torvalds 已提交
855

856
	*vm_flags = 0;
H
Hugh Dickins 已提交
857
	if (page_mapped(page) && page_rmapping(page)) {
H
Hugh Dickins 已提交
858 859 860 861 862 863 864 865
		if (!is_locked && (!PageAnon(page) || PageKsm(page))) {
			we_locked = trylock_page(page);
			if (!we_locked) {
				referenced++;
				goto out;
			}
		}
		if (unlikely(PageKsm(page)))
866
			referenced += page_referenced_ksm(page, memcg,
H
Hugh Dickins 已提交
867 868
								vm_flags);
		else if (PageAnon(page))
869
			referenced += page_referenced_anon(page, memcg,
870
								vm_flags);
H
Hugh Dickins 已提交
871
		else if (page->mapping)
872
			referenced += page_referenced_file(page, memcg,
873
								vm_flags);
H
Hugh Dickins 已提交
874
		if (we_locked)
L
Linus Torvalds 已提交
875 876
			unlock_page(page);
	}
H
Hugh Dickins 已提交
877
out:
L
Linus Torvalds 已提交
878 879 880
	return referenced;
}

H
Hugh Dickins 已提交
881 882
static int page_mkclean_one(struct page *page, struct vm_area_struct *vma,
			    unsigned long address)
883 884
{
	struct mm_struct *mm = vma->vm_mm;
885
	pte_t *pte;
886 887 888
	spinlock_t *ptl;
	int ret = 0;

N
Nick Piggin 已提交
889
	pte = page_check_address(page, mm, address, &ptl, 1);
890 891 892
	if (!pte)
		goto out;

893 894
	if (pte_dirty(*pte) || pte_write(*pte)) {
		pte_t entry;
895

896
		flush_cache_page(vma, address, pte_pfn(*pte));
897
		entry = ptep_clear_flush(vma, address, pte);
898 899
		entry = pte_wrprotect(entry);
		entry = pte_mkclean(entry);
900
		set_pte_at(mm, address, pte, entry);
901 902
		ret = 1;
	}
903 904

	pte_unmap_unlock(pte, ptl);
905 906 907

	if (ret)
		mmu_notifier_invalidate_page(mm, address);
908 909 910 911 912 913 914 915 916 917 918 919
out:
	return ret;
}

static int page_mkclean_file(struct address_space *mapping, struct page *page)
{
	pgoff_t pgoff = page->index << (PAGE_CACHE_SHIFT - PAGE_SHIFT);
	struct vm_area_struct *vma;
	int ret = 0;

	BUG_ON(PageAnon(page));

920
	mutex_lock(&mapping->i_mmap_mutex);
921
	vma_interval_tree_foreach(vma, &mapping->i_mmap, pgoff, pgoff) {
H
Hugh Dickins 已提交
922 923 924 925
		if (vma->vm_flags & VM_SHARED) {
			unsigned long address = vma_address(page, vma);
			ret += page_mkclean_one(page, vma, address);
		}
926
	}
927
	mutex_unlock(&mapping->i_mmap_mutex);
928 929 930 931 932 933 934 935 936 937 938
	return ret;
}

int page_mkclean(struct page *page)
{
	int ret = 0;

	BUG_ON(!PageLocked(page));

	if (page_mapped(page)) {
		struct address_space *mapping = page_mapping(page);
939
		if (mapping)
940 941 942 943 944
			ret = page_mkclean_file(mapping, page);
	}

	return ret;
}
J
Jaya Kumar 已提交
945
EXPORT_SYMBOL_GPL(page_mkclean);
946

947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970
/**
 * 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;

	VM_BUG_ON(!PageLocked(page));
	VM_BUG_ON(!anon_vma);
	VM_BUG_ON(page->index != linear_page_index(vma, address));

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

N
Nick Piggin 已提交
971
/**
A
Andi Kleen 已提交
972 973 974 975
 * __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	
976
 * @exclusive:	the page is exclusively owned by the current process
N
Nick Piggin 已提交
977 978
 */
static void __page_set_anon_rmap(struct page *page,
979
	struct vm_area_struct *vma, unsigned long address, int exclusive)
N
Nick Piggin 已提交
980
{
981
	struct anon_vma *anon_vma = vma->anon_vma;
982

983
	BUG_ON(!anon_vma);
984

A
Andi Kleen 已提交
985 986 987
	if (PageAnon(page))
		return;

988
	/*
989 990 991
	 * If the page isn't exclusively mapped into this vma,
	 * we must use the _oldest_ possible anon_vma for the
	 * page mapping!
992
	 */
A
Andi Kleen 已提交
993
	if (!exclusive)
994
		anon_vma = anon_vma->root;
N
Nick Piggin 已提交
995 996 997 998 999 1000

	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
1001
/**
R
Randy Dunlap 已提交
1002
 * __page_check_anon_rmap - sanity check anonymous rmap addition
N
Nick Piggin 已提交
1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022
 * @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
1023
	BUG_ON(page_anon_vma(page)->root != vma->anon_vma->root);
N
Nick Piggin 已提交
1024 1025 1026 1027
	BUG_ON(page->index != linear_page_index(vma, address));
#endif
}

L
Linus Torvalds 已提交
1028 1029 1030 1031 1032 1033
/**
 * 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
1034
 * The caller needs to hold the pte lock, and the page must be locked in
1035 1036 1037
 * 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
1038 1039 1040
 */
void page_add_anon_rmap(struct page *page,
	struct vm_area_struct *vma, unsigned long address)
1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051
{
	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
1052
{
H
Hugh Dickins 已提交
1053
	int first = atomic_inc_and_test(&page->_mapcount);
1054
	if (first) {
1055
		if (PageTransHuge(page))
1056 1057
			__inc_zone_page_state(page,
					      NR_ANON_TRANSPARENT_HUGEPAGES);
1058 1059
		__mod_zone_page_state(page_zone(page), NR_ANON_PAGES,
				hpage_nr_pages(page));
1060
	}
H
Hugh Dickins 已提交
1061 1062 1063
	if (unlikely(PageKsm(page)))
		return;

N
Nick Piggin 已提交
1064
	VM_BUG_ON(!PageLocked(page));
1065
	/* address might be in next vma when migration races vma_adjust */
H
Hugh Dickins 已提交
1066
	if (first)
1067
		__page_set_anon_rmap(page, vma, address, exclusive);
1068
	else
N
Nick Piggin 已提交
1069
		__page_check_anon_rmap(page, vma, address);
L
Linus Torvalds 已提交
1070 1071
}

R
Randy Dunlap 已提交
1072
/**
N
Nick Piggin 已提交
1073 1074 1075 1076 1077 1078 1079
 * 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
1080
 * Page does not have to be locked.
N
Nick Piggin 已提交
1081 1082 1083 1084
 */
void page_add_new_anon_rmap(struct page *page,
	struct vm_area_struct *vma, unsigned long address)
{
1085
	VM_BUG_ON(address < vma->vm_start || address >= vma->vm_end);
1086 1087
	SetPageSwapBacked(page);
	atomic_set(&page->_mapcount, 0); /* increment count (starts at -1) */
1088
	if (PageTransHuge(page))
1089
		__inc_zone_page_state(page, NR_ANON_TRANSPARENT_HUGEPAGES);
1090 1091
	__mod_zone_page_state(page_zone(page), NR_ANON_PAGES,
			hpage_nr_pages(page));
1092
	__page_set_anon_rmap(page, vma, address, 1);
1093 1094 1095 1096
	if (!mlocked_vma_newpage(vma, page)) {
		SetPageActive(page);
		lru_cache_add(page);
	} else
1097
		add_page_to_unevictable_list(page);
N
Nick Piggin 已提交
1098 1099
}

L
Linus Torvalds 已提交
1100 1101 1102 1103
/**
 * page_add_file_rmap - add pte mapping to a file page
 * @page: the page to add the mapping to
 *
1104
 * The caller needs to hold the pte lock.
L
Linus Torvalds 已提交
1105 1106 1107
 */
void page_add_file_rmap(struct page *page)
{
1108 1109 1110 1111
	bool locked;
	unsigned long flags;

	mem_cgroup_begin_update_page_stat(page, &locked, &flags);
1112
	if (atomic_inc_and_test(&page->_mapcount)) {
1113
		__inc_zone_page_state(page, NR_FILE_MAPPED);
S
Sha Zhengju 已提交
1114
		mem_cgroup_inc_page_stat(page, MEM_CGROUP_STAT_FILE_MAPPED);
1115
	}
1116
	mem_cgroup_end_update_page_stat(page, &locked, &flags);
L
Linus Torvalds 已提交
1117 1118 1119 1120 1121 1122
}

/**
 * page_remove_rmap - take down pte mapping from a page
 * @page: page to remove mapping from
 *
1123
 * The caller needs to hold the pte lock.
L
Linus Torvalds 已提交
1124
 */
1125
void page_remove_rmap(struct page *page)
L
Linus Torvalds 已提交
1126
{
1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138
	bool anon = PageAnon(page);
	bool locked;
	unsigned long flags;

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

1139 1140
	/* page still mapped by someone else? */
	if (!atomic_add_negative(-1, &page->_mapcount))
1141
		goto out;
1142

1143 1144 1145 1146 1147
	/*
	 * Hugepages are not counted in NR_ANON_PAGES nor NR_FILE_MAPPED
	 * and not charged by memcg for now.
	 */
	if (unlikely(PageHuge(page)))
1148 1149
		goto out;
	if (anon) {
1150
		mem_cgroup_uncharge_page(page);
1151
		if (PageTransHuge(page))
1152 1153
			__dec_zone_page_state(page,
					      NR_ANON_TRANSPARENT_HUGEPAGES);
1154 1155
		__mod_zone_page_state(page_zone(page), NR_ANON_PAGES,
				-hpage_nr_pages(page));
1156 1157
	} else {
		__dec_zone_page_state(page, NR_FILE_MAPPED);
S
Sha Zhengju 已提交
1158
		mem_cgroup_dec_page_stat(page, MEM_CGROUP_STAT_FILE_MAPPED);
1159
		mem_cgroup_end_update_page_stat(page, &locked, &flags);
1160
	}
1161 1162
	if (unlikely(PageMlocked(page)))
		clear_page_mlock(page);
1163 1164 1165 1166 1167 1168 1169 1170 1171
	/*
	 * 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.
	 */
1172
	return;
1173 1174 1175
out:
	if (!anon)
		mem_cgroup_end_update_page_stat(page, &locked, &flags);
L
Linus Torvalds 已提交
1176 1177 1178 1179
}

/*
 * Subfunctions of try_to_unmap: try_to_unmap_one called
1180
 * repeatedly from try_to_unmap_ksm, try_to_unmap_anon or try_to_unmap_file.
L
Linus Torvalds 已提交
1181
 */
H
Hugh Dickins 已提交
1182 1183
int try_to_unmap_one(struct page *page, struct vm_area_struct *vma,
		     unsigned long address, enum ttu_flags flags)
L
Linus Torvalds 已提交
1184 1185 1186 1187
{
	struct mm_struct *mm = vma->vm_mm;
	pte_t *pte;
	pte_t pteval;
H
Hugh Dickins 已提交
1188
	spinlock_t *ptl;
L
Linus Torvalds 已提交
1189 1190
	int ret = SWAP_AGAIN;

N
Nick Piggin 已提交
1191
	pte = page_check_address(page, mm, address, &ptl, 0);
H
Hugh Dickins 已提交
1192
	if (!pte)
N
Nikita Danilov 已提交
1193
		goto out;
L
Linus Torvalds 已提交
1194 1195 1196 1197 1198 1199

	/*
	 * 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.
	 */
1200
	if (!(flags & TTU_IGNORE_MLOCK)) {
K
KOSAKI Motohiro 已提交
1201 1202 1203
		if (vma->vm_flags & VM_LOCKED)
			goto out_mlock;

H
Hugh Dickins 已提交
1204
		if (TTU_ACTION(flags) == TTU_MUNLOCK)
H
Hugh Dickins 已提交
1205
			goto out_unmap;
1206 1207
	}
	if (!(flags & TTU_IGNORE_ACCESS)) {
N
Nick Piggin 已提交
1208 1209 1210 1211 1212
		if (ptep_clear_flush_young_notify(vma, address, pte)) {
			ret = SWAP_FAIL;
			goto out_unmap;
		}
  	}
L
Linus Torvalds 已提交
1213 1214 1215

	/* Nuke the page table entry. */
	flush_cache_page(vma, address, page_to_pfn(page));
1216
	pteval = ptep_clear_flush(vma, address, pte);
L
Linus Torvalds 已提交
1217 1218 1219 1220 1221

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

1222 1223 1224
	/* Update high watermark before we lower rss */
	update_hiwater_rss(mm);

1225
	if (PageHWPoison(page) && !(flags & TTU_IGNORE_HWPOISON)) {
1226 1227 1228 1229 1230 1231
		if (!PageHuge(page)) {
			if (PageAnon(page))
				dec_mm_counter(mm, MM_ANONPAGES);
			else
				dec_mm_counter(mm, MM_FILEPAGES);
		}
1232
		set_pte_at(mm, address, pte,
1233
			   swp_entry_to_pte(make_hwpoison_entry(page)));
1234
	} else if (PageAnon(page)) {
H
Hugh Dickins 已提交
1235
		swp_entry_t entry = { .val = page_private(page) };
1236
		pte_t swp_pte;
1237 1238 1239 1240 1241 1242

		if (PageSwapCache(page)) {
			/*
			 * Store the swap location in the pte.
			 * See handle_pte_fault() ...
			 */
H
Hugh Dickins 已提交
1243 1244 1245 1246 1247
			if (swap_duplicate(entry) < 0) {
				set_pte_at(mm, address, pte, pteval);
				ret = SWAP_FAIL;
				goto out_unmap;
			}
1248 1249 1250 1251 1252 1253
			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
1254
			dec_mm_counter(mm, MM_ANONPAGES);
K
KAMEZAWA Hiroyuki 已提交
1255
			inc_mm_counter(mm, MM_SWAPENTS);
1256
		} else if (IS_ENABLED(CONFIG_MIGRATION)) {
1257 1258 1259 1260 1261
			/*
			 * 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.
			 */
1262
			BUG_ON(TTU_ACTION(flags) != TTU_MIGRATION);
1263
			entry = make_migration_entry(page, pte_write(pteval));
L
Linus Torvalds 已提交
1264
		}
1265 1266 1267 1268
		swp_pte = swp_entry_to_pte(entry);
		if (pte_soft_dirty(pteval))
			swp_pte = pte_swp_mksoft_dirty(swp_pte);
		set_pte_at(mm, address, pte, swp_pte);
L
Linus Torvalds 已提交
1269
		BUG_ON(pte_file(*pte));
1270 1271
	} else if (IS_ENABLED(CONFIG_MIGRATION) &&
		   (TTU_ACTION(flags) == TTU_MIGRATION)) {
1272 1273 1274 1275 1276
		/* 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
1277
		dec_mm_counter(mm, MM_FILEPAGES);
L
Linus Torvalds 已提交
1278

1279
	page_remove_rmap(page);
L
Linus Torvalds 已提交
1280 1281 1282
	page_cache_release(page);

out_unmap:
H
Hugh Dickins 已提交
1283
	pte_unmap_unlock(pte, ptl);
1284 1285
	if (ret != SWAP_FAIL)
		mmu_notifier_invalidate_page(mm, address);
K
KOSAKI Motohiro 已提交
1286 1287
out:
	return ret;
H
Hugh Dickins 已提交
1288

K
KOSAKI Motohiro 已提交
1289 1290 1291 1292 1293 1294 1295
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
1296
	 * we now hold anon_vma->rwsem or mapping->i_mmap_mutex.
K
KOSAKI Motohiro 已提交
1297 1298 1299 1300 1301 1302 1303 1304
	 * 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
1305
		}
K
KOSAKI Motohiro 已提交
1306
		up_read(&vma->vm_mm->mmap_sem);
H
Hugh Dickins 已提交
1307
	}
L
Linus Torvalds 已提交
1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328
	return ret;
}

/*
 * objrmap doesn't work for nonlinear VMAs because the assumption that
 * offset-into-file correlates with offset-into-virtual-addresses does not hold.
 * Consequently, given a particular page and its ->index, we cannot locate the
 * ptes which are mapping that page without an exhaustive linear search.
 *
 * So what this code does is a mini "virtual scan" of each nonlinear VMA which
 * maps the file to which the target page belongs.  The ->vm_private_data field
 * holds the current cursor into that scan.  Successive searches will circulate
 * around the vma's virtual address space.
 *
 * So as more replacement pressure is applied to the pages in a nonlinear VMA,
 * more scanning pressure is placed against them as well.   Eventually pages
 * will become fully unmapped and are eligible for eviction.
 *
 * For very sparsely populated VMAs this is a little inefficient - chances are
 * there there won't be many ptes located within the scan cluster.  In this case
 * maybe we could scan further - to the end of the pte page, perhaps.
N
Nick Piggin 已提交
1329 1330 1331 1332 1333
 *
 * Mlocked pages:  check VM_LOCKED under mmap_sem held for read, if we can
 * acquire it without blocking.  If vma locked, mlock the pages in the cluster,
 * rather than unmapping them.  If we encounter the "check_page" that vmscan is
 * trying to unmap, return SWAP_MLOCK, else default SWAP_AGAIN.
L
Linus Torvalds 已提交
1334 1335 1336 1337
 */
#define CLUSTER_SIZE	min(32*PAGE_SIZE, PMD_SIZE)
#define CLUSTER_MASK	(~(CLUSTER_SIZE - 1))

N
Nick Piggin 已提交
1338 1339
static int try_to_unmap_cluster(unsigned long cursor, unsigned int *mapcount,
		struct vm_area_struct *vma, struct page *check_page)
L
Linus Torvalds 已提交
1340 1341 1342
{
	struct mm_struct *mm = vma->vm_mm;
	pmd_t *pmd;
H
Hugh Dickins 已提交
1343
	pte_t *pte;
L
Linus Torvalds 已提交
1344
	pte_t pteval;
H
Hugh Dickins 已提交
1345
	spinlock_t *ptl;
L
Linus Torvalds 已提交
1346 1347
	struct page *page;
	unsigned long address;
1348 1349
	unsigned long mmun_start;	/* For mmu_notifiers */
	unsigned long mmun_end;		/* For mmu_notifiers */
L
Linus Torvalds 已提交
1350
	unsigned long end;
N
Nick Piggin 已提交
1351 1352
	int ret = SWAP_AGAIN;
	int locked_vma = 0;
L
Linus Torvalds 已提交
1353 1354 1355 1356 1357 1358 1359 1360

	address = (vma->vm_start + cursor) & CLUSTER_MASK;
	end = address + CLUSTER_SIZE;
	if (address < vma->vm_start)
		address = vma->vm_start;
	if (end > vma->vm_end)
		end = vma->vm_end;

B
Bob Liu 已提交
1361 1362
	pmd = mm_find_pmd(mm, address);
	if (!pmd)
N
Nick Piggin 已提交
1363 1364
		return ret;

1365 1366 1367 1368
	mmun_start = address;
	mmun_end   = end;
	mmu_notifier_invalidate_range_start(mm, mmun_start, mmun_end);

N
Nick Piggin 已提交
1369
	/*
H
Hugh Dickins 已提交
1370
	 * If we can acquire the mmap_sem for read, and vma is VM_LOCKED,
N
Nick Piggin 已提交
1371 1372
	 * keep the sem while scanning the cluster for mlocking pages.
	 */
H
Hugh Dickins 已提交
1373
	if (down_read_trylock(&vma->vm_mm->mmap_sem)) {
N
Nick Piggin 已提交
1374 1375 1376 1377
		locked_vma = (vma->vm_flags & VM_LOCKED);
		if (!locked_vma)
			up_read(&vma->vm_mm->mmap_sem); /* don't need it */
	}
H
Hugh Dickins 已提交
1378 1379

	pte = pte_offset_map_lock(mm, pmd, address, &ptl);
L
Linus Torvalds 已提交
1380

1381 1382 1383
	/* Update high watermark before we lower rss */
	update_hiwater_rss(mm);

H
Hugh Dickins 已提交
1384
	for (; address < end; pte++, address += PAGE_SIZE) {
L
Linus Torvalds 已提交
1385 1386
		if (!pte_present(*pte))
			continue;
1387 1388
		page = vm_normal_page(vma, address, *pte);
		BUG_ON(!page || PageAnon(page));
L
Linus Torvalds 已提交
1389

N
Nick Piggin 已提交
1390 1391 1392 1393 1394 1395 1396
		if (locked_vma) {
			mlock_vma_page(page);   /* no-op if already mlocked */
			if (page == check_page)
				ret = SWAP_MLOCK;
			continue;	/* don't unmap */
		}

A
Andrea Arcangeli 已提交
1397
		if (ptep_clear_flush_young_notify(vma, address, pte))
L
Linus Torvalds 已提交
1398 1399 1400
			continue;

		/* Nuke the page table entry. */
1401
		flush_cache_page(vma, address, pte_pfn(*pte));
1402
		pteval = ptep_clear_flush(vma, address, pte);
L
Linus Torvalds 已提交
1403 1404

		/* If nonlinear, store the file page offset in the pte. */
1405 1406 1407 1408 1409 1410
		if (page->index != linear_page_index(vma, address)) {
			pte_t ptfile = pgoff_to_pte(page->index);
			if (pte_soft_dirty(pteval))
				pte_file_mksoft_dirty(ptfile);
			set_pte_at(mm, address, pte, ptfile);
		}
L
Linus Torvalds 已提交
1411 1412 1413 1414 1415

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

1416
		page_remove_rmap(page);
L
Linus Torvalds 已提交
1417
		page_cache_release(page);
K
KAMEZAWA Hiroyuki 已提交
1418
		dec_mm_counter(mm, MM_FILEPAGES);
L
Linus Torvalds 已提交
1419 1420
		(*mapcount)--;
	}
H
Hugh Dickins 已提交
1421
	pte_unmap_unlock(pte - 1, ptl);
1422
	mmu_notifier_invalidate_range_end(mm, mmun_start, mmun_end);
N
Nick Piggin 已提交
1423 1424 1425
	if (locked_vma)
		up_read(&vma->vm_mm->mmap_sem);
	return ret;
L
Linus Torvalds 已提交
1426 1427
}

1428
bool is_vma_temporary_stack(struct vm_area_struct *vma)
1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441
{
	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;
}

N
Nick Piggin 已提交
1442 1443 1444 1445
/**
 * try_to_unmap_anon - unmap or unlock anonymous page using the object-based
 * rmap method
 * @page: the page to unmap/unlock
1446
 * @flags: action and flags
N
Nick Piggin 已提交
1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457
 *
 * Find all the mappings of a page using the mapping pointer and the vma chains
 * contained in the anon_vma struct it points to.
 *
 * This function is only called from try_to_unmap/try_to_munlock for
 * anonymous pages.
 * 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.
 */
1458
static int try_to_unmap_anon(struct page *page, enum ttu_flags flags)
L
Linus Torvalds 已提交
1459 1460
{
	struct anon_vma *anon_vma;
1461
	pgoff_t pgoff;
1462
	struct anon_vma_chain *avc;
L
Linus Torvalds 已提交
1463
	int ret = SWAP_AGAIN;
N
Nick Piggin 已提交
1464

1465
	anon_vma = page_lock_anon_vma_read(page);
L
Linus Torvalds 已提交
1466 1467 1468
	if (!anon_vma)
		return ret;

1469 1470
	pgoff = page->index << (PAGE_CACHE_SHIFT - PAGE_SHIFT);
	anon_vma_interval_tree_foreach(avc, &anon_vma->rb_root, pgoff, pgoff) {
1471
		struct vm_area_struct *vma = avc->vma;
1472 1473 1474 1475 1476 1477 1478 1479 1480 1481
		unsigned long address;

		/*
		 * 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.
		 */
1482
		if (IS_ENABLED(CONFIG_MIGRATION) && (flags & TTU_MIGRATION) &&
1483 1484 1485 1486
				is_vma_temporary_stack(vma))
			continue;

		address = vma_address(page, vma);
H
Hugh Dickins 已提交
1487
		ret = try_to_unmap_one(page, vma, address, flags);
H
Hugh Dickins 已提交
1488 1489
		if (ret != SWAP_AGAIN || !page_mapped(page))
			break;
L
Linus Torvalds 已提交
1490
	}
1491

1492
	page_unlock_anon_vma_read(anon_vma);
L
Linus Torvalds 已提交
1493 1494 1495 1496
	return ret;
}

/**
N
Nick Piggin 已提交
1497 1498
 * try_to_unmap_file - unmap/unlock file page using the object-based rmap method
 * @page: the page to unmap/unlock
1499
 * @flags: action and flags
L
Linus Torvalds 已提交
1500 1501 1502 1503
 *
 * Find all the mappings of a page using the mapping pointer and the vma chains
 * contained in the address_space struct it points to.
 *
N
Nick Piggin 已提交
1504 1505 1506 1507 1508 1509
 * This function is only called from try_to_unmap/try_to_munlock for
 * object-based pages.
 * 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.
L
Linus Torvalds 已提交
1510
 */
1511
static int try_to_unmap_file(struct page *page, enum ttu_flags flags)
L
Linus Torvalds 已提交
1512 1513 1514 1515 1516 1517 1518 1519 1520 1521
{
	struct address_space *mapping = page->mapping;
	pgoff_t pgoff = page->index << (PAGE_CACHE_SHIFT - PAGE_SHIFT);
	struct vm_area_struct *vma;
	int ret = SWAP_AGAIN;
	unsigned long cursor;
	unsigned long max_nl_cursor = 0;
	unsigned long max_nl_size = 0;
	unsigned int mapcount;

1522 1523 1524
	if (PageHuge(page))
		pgoff = page->index << compound_order(page);

1525
	mutex_lock(&mapping->i_mmap_mutex);
1526
	vma_interval_tree_foreach(vma, &mapping->i_mmap, pgoff, pgoff) {
H
Hugh Dickins 已提交
1527 1528
		unsigned long address = vma_address(page, vma);
		ret = try_to_unmap_one(page, vma, address, flags);
H
Hugh Dickins 已提交
1529 1530
		if (ret != SWAP_AGAIN || !page_mapped(page))
			goto out;
L
Linus Torvalds 已提交
1531 1532 1533 1534 1535
	}

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

H
Hugh Dickins 已提交
1536 1537 1538 1539 1540 1541 1542 1543
	/*
	 * We don't bother to try to find the munlocked page in nonlinears.
	 * It's costly. Instead, later, page reclaim logic may call
	 * try_to_unmap(TTU_MUNLOCK) and recover PG_mlocked lazily.
	 */
	if (TTU_ACTION(flags) == TTU_MUNLOCK)
		goto out;

L
Linus Torvalds 已提交
1544
	list_for_each_entry(vma, &mapping->i_mmap_nonlinear,
1545
							shared.nonlinear) {
L
Linus Torvalds 已提交
1546 1547 1548 1549 1550 1551 1552 1553
		cursor = (unsigned long) vma->vm_private_data;
		if (cursor > max_nl_cursor)
			max_nl_cursor = cursor;
		cursor = vma->vm_end - vma->vm_start;
		if (cursor > max_nl_size)
			max_nl_size = cursor;
	}

N
Nick Piggin 已提交
1554
	if (max_nl_size == 0) {	/* all nonlinears locked or reserved ? */
L
Linus Torvalds 已提交
1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568
		ret = SWAP_FAIL;
		goto out;
	}

	/*
	 * We don't try to search for this page in the nonlinear vmas,
	 * and page_referenced wouldn't have found it anyway.  Instead
	 * just walk the nonlinear vmas trying to age and unmap some.
	 * The mapcount of the page we came in with is irrelevant,
	 * but even so use it as a guide to how hard we should try?
	 */
	mapcount = page_mapcount(page);
	if (!mapcount)
		goto out;
1569
	cond_resched();
L
Linus Torvalds 已提交
1570 1571 1572 1573 1574 1575 1576

	max_nl_size = (max_nl_size + CLUSTER_SIZE - 1) & CLUSTER_MASK;
	if (max_nl_cursor == 0)
		max_nl_cursor = CLUSTER_SIZE;

	do {
		list_for_each_entry(vma, &mapping->i_mmap_nonlinear,
1577
							shared.nonlinear) {
L
Linus Torvalds 已提交
1578
			cursor = (unsigned long) vma->vm_private_data;
H
Hugh Dickins 已提交
1579
			while ( cursor < max_nl_cursor &&
L
Linus Torvalds 已提交
1580
				cursor < vma->vm_end - vma->vm_start) {
H
Hugh Dickins 已提交
1581 1582 1583
				if (try_to_unmap_cluster(cursor, &mapcount,
						vma, page) == SWAP_MLOCK)
					ret = SWAP_MLOCK;
L
Linus Torvalds 已提交
1584 1585 1586 1587 1588 1589 1590
				cursor += CLUSTER_SIZE;
				vma->vm_private_data = (void *) cursor;
				if ((int)mapcount <= 0)
					goto out;
			}
			vma->vm_private_data = (void *) max_nl_cursor;
		}
1591
		cond_resched();
L
Linus Torvalds 已提交
1592 1593 1594 1595 1596 1597 1598 1599
		max_nl_cursor += CLUSTER_SIZE;
	} while (max_nl_cursor <= max_nl_size);

	/*
	 * Don't loop forever (perhaps all the remaining pages are
	 * in locked vmas).  Reset cursor on all unreserved nonlinear
	 * vmas, now forgetting on which ones it had fallen behind.
	 */
1600
	list_for_each_entry(vma, &mapping->i_mmap_nonlinear, shared.nonlinear)
1601
		vma->vm_private_data = NULL;
L
Linus Torvalds 已提交
1602
out:
1603
	mutex_unlock(&mapping->i_mmap_mutex);
L
Linus Torvalds 已提交
1604 1605 1606 1607 1608 1609
	return ret;
}

/**
 * try_to_unmap - try to remove all page table mappings to a page
 * @page: the page to get unmapped
1610
 * @flags: action and flags
L
Linus Torvalds 已提交
1611 1612 1613 1614 1615 1616 1617 1618
 *
 * 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
1619
 * SWAP_MLOCK	- page is mlocked.
L
Linus Torvalds 已提交
1620
 */
1621
int try_to_unmap(struct page *page, enum ttu_flags flags)
L
Linus Torvalds 已提交
1622 1623 1624 1625
{
	int ret;

	BUG_ON(!PageLocked(page));
1626
	VM_BUG_ON(!PageHuge(page) && PageTransHuge(page));
L
Linus Torvalds 已提交
1627

H
Hugh Dickins 已提交
1628 1629 1630
	if (unlikely(PageKsm(page)))
		ret = try_to_unmap_ksm(page, flags);
	else if (PageAnon(page))
1631
		ret = try_to_unmap_anon(page, flags);
L
Linus Torvalds 已提交
1632
	else
1633
		ret = try_to_unmap_file(page, flags);
N
Nick Piggin 已提交
1634
	if (ret != SWAP_MLOCK && !page_mapped(page))
L
Linus Torvalds 已提交
1635 1636 1637
		ret = SWAP_SUCCESS;
	return ret;
}
N
Nikita Danilov 已提交
1638

N
Nick Piggin 已提交
1639 1640 1641 1642 1643 1644 1645 1646 1647 1648
/**
 * 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
1649
 * SWAP_AGAIN	- no vma is holding page mlocked, or,
N
Nick Piggin 已提交
1650
 * SWAP_AGAIN	- page mapped in mlocked vma -- couldn't acquire mmap sem
H
Hugh Dickins 已提交
1651
 * SWAP_FAIL	- page cannot be located at present
N
Nick Piggin 已提交
1652 1653 1654 1655 1656 1657
 * SWAP_MLOCK	- page is now mlocked.
 */
int try_to_munlock(struct page *page)
{
	VM_BUG_ON(!PageLocked(page) || PageLRU(page));

H
Hugh Dickins 已提交
1658 1659 1660
	if (unlikely(PageKsm(page)))
		return try_to_unmap_ksm(page, TTU_MUNLOCK);
	else if (PageAnon(page))
1661
		return try_to_unmap_anon(page, TTU_MUNLOCK);
N
Nick Piggin 已提交
1662
	else
1663
		return try_to_unmap_file(page, TTU_MUNLOCK);
N
Nick Piggin 已提交
1664
}
1665

P
Peter Zijlstra 已提交
1666
void __put_anon_vma(struct anon_vma *anon_vma)
1667
{
P
Peter Zijlstra 已提交
1668
	struct anon_vma *root = anon_vma->root;
1669

P
Peter Zijlstra 已提交
1670 1671
	if (root != anon_vma && atomic_dec_and_test(&root->refcount))
		anon_vma_free(root);
1672

P
Peter Zijlstra 已提交
1673
	anon_vma_free(anon_vma);
1674 1675
}

1676 1677 1678 1679 1680 1681 1682 1683 1684
#ifdef CONFIG_MIGRATION
/*
 * rmap_walk() and its helpers rmap_walk_anon() and rmap_walk_file():
 * Called by migrate.c to remove migration ptes, but might be used more later.
 */
static int rmap_walk_anon(struct page *page, int (*rmap_one)(struct page *,
		struct vm_area_struct *, unsigned long, void *), void *arg)
{
	struct anon_vma *anon_vma;
1685
	pgoff_t pgoff = page->index << (PAGE_CACHE_SHIFT - PAGE_SHIFT);
1686
	struct anon_vma_chain *avc;
1687 1688 1689
	int ret = SWAP_AGAIN;

	/*
1690
	 * Note: remove_migration_ptes() cannot use page_lock_anon_vma_read()
1691
	 * because that depends on page_mapped(); but not all its usages
1692 1693
	 * are holding mmap_sem. Users without mmap_sem are required to
	 * take a reference count to prevent the anon_vma disappearing
1694 1695 1696 1697
	 */
	anon_vma = page_anon_vma(page);
	if (!anon_vma)
		return ret;
1698
	anon_vma_lock_read(anon_vma);
1699
	anon_vma_interval_tree_foreach(avc, &anon_vma->rb_root, pgoff, pgoff) {
1700
		struct vm_area_struct *vma = avc->vma;
1701 1702 1703 1704 1705
		unsigned long address = vma_address(page, vma);
		ret = rmap_one(page, vma, address, arg);
		if (ret != SWAP_AGAIN)
			break;
	}
1706
	anon_vma_unlock_read(anon_vma);
1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719
	return ret;
}

static int rmap_walk_file(struct page *page, int (*rmap_one)(struct page *,
		struct vm_area_struct *, unsigned long, void *), void *arg)
{
	struct address_space *mapping = page->mapping;
	pgoff_t pgoff = page->index << (PAGE_CACHE_SHIFT - PAGE_SHIFT);
	struct vm_area_struct *vma;
	int ret = SWAP_AGAIN;

	if (!mapping)
		return ret;
1720
	mutex_lock(&mapping->i_mmap_mutex);
1721
	vma_interval_tree_foreach(vma, &mapping->i_mmap, pgoff, pgoff) {
1722 1723 1724 1725 1726 1727 1728 1729 1730 1731
		unsigned long address = vma_address(page, vma);
		ret = rmap_one(page, vma, address, arg);
		if (ret != SWAP_AGAIN)
			break;
	}
	/*
	 * No nonlinear handling: being always shared, nonlinear vmas
	 * never contain migration ptes.  Decide what to do about this
	 * limitation to linear when we need rmap_walk() on nonlinear.
	 */
1732
	mutex_unlock(&mapping->i_mmap_mutex);
1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746 1747 1748
	return ret;
}

int rmap_walk(struct page *page, int (*rmap_one)(struct page *,
		struct vm_area_struct *, unsigned long, void *), void *arg)
{
	VM_BUG_ON(!PageLocked(page));

	if (unlikely(PageKsm(page)))
		return rmap_walk_ksm(page, rmap_one, arg);
	else if (PageAnon(page))
		return rmap_walk_anon(page, rmap_one, arg);
	else
		return rmap_walk_file(page, rmap_one, arg);
}
#endif /* CONFIG_MIGRATION */
1749

N
Naoya Horiguchi 已提交
1750
#ifdef CONFIG_HUGETLB_PAGE
1751 1752 1753 1754 1755 1756 1757 1758 1759
/*
 * 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;
1760

1761
	BUG_ON(!anon_vma);
1762 1763 1764 1765 1766 1767

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

1768 1769 1770 1771 1772 1773 1774 1775 1776 1777
	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;
1778 1779

	BUG_ON(!PageLocked(page));
1780
	BUG_ON(!anon_vma);
1781
	/* address might be in next vma when migration races vma_adjust */
1782 1783 1784 1785 1786 1787 1788 1789 1790 1791 1792 1793
	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
1794
#endif /* CONFIG_HUGETLB_PAGE */