rmap.c 50.1 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->mutex
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
 *               inode_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 inode_wb_list_lock in __sync_single_inode)
39
 *
40 41
 * anon_vma->mutex,mapping->i_mutex      (memory_failure, collect_procs_anon)
 *   ->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/module.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>
L
Linus Torvalds 已提交
59 60 61

#include <asm/tlbflush.h>

N
Nick Piggin 已提交
62 63
#include "internal.h"

64
static struct kmem_cache *anon_vma_cachep;
65
static struct kmem_cache *anon_vma_chain_cachep;
66 67 68

static inline struct anon_vma *anon_vma_alloc(void)
{
P
Peter Zijlstra 已提交
69 70 71 72 73 74 75 76 77 78 79 80 81
	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;
82 83
}

P
Peter Zijlstra 已提交
84
static inline void anon_vma_free(struct anon_vma *anon_vma)
85
{
P
Peter Zijlstra 已提交
86
	VM_BUG_ON(atomic_read(&anon_vma->refcount));
87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109

	/*
	 * Synchronize against page_lock_anon_vma() such that
	 * 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
	 * mutex_trylock() from page_lock_anon_vma(). This orders:
	 *
	 * page_lock_anon_vma()		VS	put_anon_vma()
	 *   mutex_trylock()			  atomic_dec_and_test()
	 *   LOCK				  MB
	 *   atomic_read()			  mutex_is_locked()
	 *
	 * LOCK should suffice since the actual taking of the lock must
	 * happen _before_ what follows.
	 */
	if (mutex_is_locked(&anon_vma->root->mutex)) {
		anon_vma_lock(anon_vma);
		anon_vma_unlock(anon_vma);
	}

110 111
	kmem_cache_free(anon_vma_cachep, anon_vma);
}
L
Linus Torvalds 已提交
112

113
static inline struct anon_vma_chain *anon_vma_chain_alloc(gfp_t gfp)
114
{
115
	return kmem_cache_alloc(anon_vma_chain_cachep, gfp);
116 117
}

118
static void anon_vma_chain_free(struct anon_vma_chain *anon_vma_chain)
119 120 121 122
{
	kmem_cache_free(anon_vma_chain_cachep, anon_vma_chain);
}

123 124 125 126 127 128 129 130 131
/**
 * 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
132
 * not we either need to find an adjacent mapping that we
133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149
 * can re-use the anon_vma from (very common when the only
 * reason for splitting a vma has been mprotect()), or we
 * allocate a new one.
 *
 * Anon-vma allocations are very subtle, because we may have
 * optimistically looked up an anon_vma in page_lock_anon_vma()
 * 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
150 151 152
int anon_vma_prepare(struct vm_area_struct *vma)
{
	struct anon_vma *anon_vma = vma->anon_vma;
153
	struct anon_vma_chain *avc;
L
Linus Torvalds 已提交
154 155 156 157

	might_sleep();
	if (unlikely(!anon_vma)) {
		struct mm_struct *mm = vma->vm_mm;
158
		struct anon_vma *allocated;
L
Linus Torvalds 已提交
159

160
		avc = anon_vma_chain_alloc(GFP_KERNEL);
161 162 163
		if (!avc)
			goto out_enomem;

L
Linus Torvalds 已提交
164
		anon_vma = find_mergeable_anon_vma(vma);
165 166
		allocated = NULL;
		if (!anon_vma) {
L
Linus Torvalds 已提交
167 168
			anon_vma = anon_vma_alloc();
			if (unlikely(!anon_vma))
169
				goto out_enomem_free_avc;
L
Linus Torvalds 已提交
170 171 172
			allocated = anon_vma;
		}

173
		anon_vma_lock(anon_vma);
L
Linus Torvalds 已提交
174 175 176 177
		/* page_table_lock to protect against threads */
		spin_lock(&mm->page_table_lock);
		if (likely(!vma->anon_vma)) {
			vma->anon_vma = anon_vma;
178 179 180
			avc->anon_vma = anon_vma;
			avc->vma = vma;
			list_add(&avc->same_vma, &vma->anon_vma_chain);
181
			list_add_tail(&avc->same_anon_vma, &anon_vma->head);
L
Linus Torvalds 已提交
182
			allocated = NULL;
183
			avc = NULL;
L
Linus Torvalds 已提交
184 185
		}
		spin_unlock(&mm->page_table_lock);
186
		anon_vma_unlock(anon_vma);
187 188

		if (unlikely(allocated))
P
Peter Zijlstra 已提交
189
			put_anon_vma(allocated);
190
		if (unlikely(avc))
191
			anon_vma_chain_free(avc);
L
Linus Torvalds 已提交
192 193
	}
	return 0;
194 195 196 197 198

 out_enomem_free_avc:
	anon_vma_chain_free(avc);
 out_enomem:
	return -ENOMEM;
L
Linus Torvalds 已提交
199 200
}

201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226
/*
 * 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))
			mutex_unlock(&root->mutex);
		root = new_root;
		mutex_lock(&root->mutex);
	}
	return root;
}

static inline void unlock_anon_vma_root(struct anon_vma *root)
{
	if (root)
		mutex_unlock(&root->mutex);
}

227 228 229
static void anon_vma_chain_link(struct vm_area_struct *vma,
				struct anon_vma_chain *avc,
				struct anon_vma *anon_vma)
L
Linus Torvalds 已提交
230
{
231 232 233 234
	avc->vma = vma;
	avc->anon_vma = anon_vma;
	list_add(&avc->same_vma, &vma->anon_vma_chain);

235 236 237 238
	/*
	 * It's critical to add new vmas to the tail of the anon_vma,
	 * see comment in huge_memory.c:__split_huge_page().
	 */
239
	list_add_tail(&avc->same_anon_vma, &anon_vma->head);
L
Linus Torvalds 已提交
240 241
}

242 243 244 245 246
/*
 * 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
247
{
248
	struct anon_vma_chain *avc, *pavc;
249
	struct anon_vma *root = NULL;
250

251
	list_for_each_entry_reverse(pavc, &src->anon_vma_chain, same_vma) {
252 253
		struct anon_vma *anon_vma;

254 255 256 257 258 259 260 261
		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;
		}
262 263 264
		anon_vma = pavc->anon_vma;
		root = lock_anon_vma_root(root, anon_vma);
		anon_vma_chain_link(dst, avc, anon_vma);
265
	}
266
	unlock_anon_vma_root(root);
267
	return 0;
L
Linus Torvalds 已提交
268

269 270 271
 enomem_failure:
	unlink_anon_vmas(dst);
	return -ENOMEM;
L
Linus Torvalds 已提交
272 273
}

274 275 276 277 278 279
/*
 * 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
280
{
281 282
	struct anon_vma_chain *avc;
	struct anon_vma *anon_vma;
L
Linus Torvalds 已提交
283

284 285 286 287 288 289 290 291 292 293 294 295 296 297 298
	/* 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;
299
	avc = anon_vma_chain_alloc(GFP_KERNEL);
300 301
	if (!avc)
		goto out_error_free_anon_vma;
302 303 304 305 306 307

	/*
	 * 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;
308
	/*
P
Peter Zijlstra 已提交
309 310 311
	 * 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.
312 313
	 */
	get_anon_vma(anon_vma->root);
314 315
	/* Mark this anon_vma as the one where our new (COWed) pages go. */
	vma->anon_vma = anon_vma;
316
	anon_vma_lock(anon_vma);
317
	anon_vma_chain_link(vma, avc, anon_vma);
318
	anon_vma_unlock(anon_vma);
319 320 321 322

	return 0;

 out_error_free_anon_vma:
P
Peter Zijlstra 已提交
323
	put_anon_vma(anon_vma);
324
 out_error:
325
	unlink_anon_vmas(vma);
326
	return -ENOMEM;
L
Linus Torvalds 已提交
327 328
}

329 330 331
void unlink_anon_vmas(struct vm_area_struct *vma)
{
	struct anon_vma_chain *avc, *next;
332
	struct anon_vma *root = NULL;
333

334 335 336 337
	/*
	 * Unlink each anon_vma chained to the VMA.  This list is ordered
	 * from newest to oldest, ensuring the root anon_vma gets freed last.
	 */
338
	list_for_each_entry_safe(avc, next, &vma->anon_vma_chain, same_vma) {
339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365
		struct anon_vma *anon_vma = avc->anon_vma;

		root = lock_anon_vma_root(root, anon_vma);
		list_del(&avc->same_anon_vma);

		/*
		 * Leave empty anon_vmas on the list - we'll need
		 * to free them outside the lock.
		 */
		if (list_empty(&anon_vma->head))
			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()
	 * needing to acquire the anon_vma->root->mutex.
	 */
	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);

366 367 368 369 370
		list_del(&avc->same_vma);
		anon_vma_chain_free(avc);
	}
}

371
static void anon_vma_ctor(void *data)
L
Linus Torvalds 已提交
372
{
C
Christoph Lameter 已提交
373
	struct anon_vma *anon_vma = data;
L
Linus Torvalds 已提交
374

375
	mutex_init(&anon_vma->mutex);
376
	atomic_set(&anon_vma->refcount, 0);
C
Christoph Lameter 已提交
377
	INIT_LIST_HEAD(&anon_vma->head);
L
Linus Torvalds 已提交
378 379 380 381 382
}

void __init anon_vma_init(void)
{
	anon_vma_cachep = kmem_cache_create("anon_vma", sizeof(struct anon_vma),
383
			0, SLAB_DESTROY_BY_RCU|SLAB_PANIC, anon_vma_ctor);
384
	anon_vma_chain_cachep = KMEM_CACHE(anon_vma_chain, SLAB_PANIC);
L
Linus Torvalds 已提交
385 386 387
}

/*
388 389 390 391 392 393 394 395 396
 * 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).
 *
397 398 399 400 401
 * 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 ].
 *
402 403 404 405 406 407 408
 * 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
409
 */
410
struct anon_vma *page_get_anon_vma(struct page *page)
L
Linus Torvalds 已提交
411
{
412
	struct anon_vma *anon_vma = NULL;
L
Linus Torvalds 已提交
413 414 415
	unsigned long anon_mapping;

	rcu_read_lock();
416
	anon_mapping = (unsigned long) ACCESS_ONCE(page->mapping);
H
Hugh Dickins 已提交
417
	if ((anon_mapping & PAGE_MAPPING_FLAGS) != PAGE_MAPPING_ANON)
L
Linus Torvalds 已提交
418 419 420 421 422
		goto out;
	if (!page_mapped(page))
		goto out;

	anon_vma = (struct anon_vma *) (anon_mapping - PAGE_MAPPING_ANON);
423 424 425 426
	if (!atomic_inc_not_zero(&anon_vma->refcount)) {
		anon_vma = NULL;
		goto out;
	}
427 428 429

	/*
	 * If this page is still mapped, then its anon_vma cannot have been
430 431 432 433
	 * 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).
434
	 */
435 436 437 438
	if (!page_mapped(page)) {
		put_anon_vma(anon_vma);
		anon_vma = NULL;
	}
L
Linus Torvalds 已提交
439 440
out:
	rcu_read_unlock();
441 442 443 444

	return anon_vma;
}

445 446 447 448 449 450 451
/*
 * 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.
 */
452 453
struct anon_vma *page_lock_anon_vma(struct page *page)
{
454
	struct anon_vma *anon_vma = NULL;
455
	struct anon_vma *root_anon_vma;
456
	unsigned long anon_mapping;
457

458 459 460 461 462 463 464 465
	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);
466 467
	root_anon_vma = ACCESS_ONCE(anon_vma->root);
	if (mutex_trylock(&root_anon_vma->mutex)) {
468
		/*
469 470
		 * If the page is still mapped, then this anon_vma is still
		 * its anon_vma, and holding the mutex ensures that it will
471
		 * not go away, see anon_vma_free().
472
		 */
473 474
		if (!page_mapped(page)) {
			mutex_unlock(&root_anon_vma->mutex);
475 476 477 478
			anon_vma = NULL;
		}
		goto out;
	}
479

480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510
	/* 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();
	anon_vma_lock(anon_vma);

	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
		 * we'll deadlock on the anon_vma_lock() recursion.
		 */
		anon_vma_unlock(anon_vma);
		__put_anon_vma(anon_vma);
		anon_vma = NULL;
	}

	return anon_vma;

out:
	rcu_read_unlock();
511
	return anon_vma;
512 513
}

514
void page_unlock_anon_vma(struct anon_vma *anon_vma)
515
{
516
	anon_vma_unlock(anon_vma);
L
Linus Torvalds 已提交
517 518 519
}

/*
520 521 522
 * At what user virtual address is page expected in @vma?
 * Returns virtual address or -EFAULT if page's index/offset is not
 * within the range mapped the @vma.
L
Linus Torvalds 已提交
523
 */
524
inline unsigned long
L
Linus Torvalds 已提交
525 526 527 528 529
vma_address(struct page *page, struct vm_area_struct *vma)
{
	pgoff_t pgoff = page->index << (PAGE_CACHE_SHIFT - PAGE_SHIFT);
	unsigned long address;

530 531
	if (unlikely(is_vm_hugetlb_page(vma)))
		pgoff = page->index << huge_page_order(page_hstate(page));
L
Linus Torvalds 已提交
532 533
	address = vma->vm_start + ((pgoff - vma->vm_pgoff) << PAGE_SHIFT);
	if (unlikely(address < vma->vm_start || address >= vma->vm_end)) {
534
		/* page should be within @vma mapping range */
L
Linus Torvalds 已提交
535 536 537 538 539 540
		return -EFAULT;
	}
	return address;
}

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

N
Nikita Danilov 已提交
564 565 566
/*
 * Check that @page is mapped at @address into @mm.
 *
N
Nick Piggin 已提交
567 568 569 570
 * 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).
 *
571
 * On success returns with pte mapped and locked.
N
Nikita Danilov 已提交
572
 */
573
pte_t *__page_check_address(struct page *page, struct mm_struct *mm,
N
Nick Piggin 已提交
574
			  unsigned long address, spinlock_t **ptlp, int sync)
N
Nikita Danilov 已提交
575 576 577 578 579
{
	pgd_t *pgd;
	pud_t *pud;
	pmd_t *pmd;
	pte_t *pte;
H
Hugh Dickins 已提交
580
	spinlock_t *ptl;
N
Nikita Danilov 已提交
581

582 583 584 585 586 587
	if (unlikely(PageHuge(page))) {
		pte = huge_pte_offset(mm, address);
		ptl = &mm->page_table_lock;
		goto check;
	}

N
Nikita Danilov 已提交
588
	pgd = pgd_offset(mm, address);
H
Hugh Dickins 已提交
589 590 591 592 593 594 595 596 597 598
	if (!pgd_present(*pgd))
		return NULL;

	pud = pud_offset(pgd, address);
	if (!pud_present(*pud))
		return NULL;

	pmd = pmd_offset(pud, address);
	if (!pmd_present(*pmd))
		return NULL;
599 600
	if (pmd_trans_huge(*pmd))
		return NULL;
H
Hugh Dickins 已提交
601 602 603

	pte = pte_offset_map(pmd, address);
	/* Make a quick check before getting the lock */
N
Nick Piggin 已提交
604
	if (!sync && !pte_present(*pte)) {
H
Hugh Dickins 已提交
605 606 607 608
		pte_unmap(pte);
		return NULL;
	}

H
Hugh Dickins 已提交
609
	ptl = pte_lockptr(mm, pmd);
610
check:
H
Hugh Dickins 已提交
611 612 613 614
	spin_lock(ptl);
	if (pte_present(*pte) && page_to_pfn(page) == pte_pfn(*pte)) {
		*ptlp = ptl;
		return pte;
N
Nikita Danilov 已提交
615
	}
H
Hugh Dickins 已提交
616 617
	pte_unmap_unlock(pte, ptl);
	return NULL;
N
Nikita Danilov 已提交
618 619
}

N
Nick Piggin 已提交
620 621 622 623 624 625 626 627 628
/**
 * 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.
 */
629
int page_mapped_in_vma(struct page *page, struct vm_area_struct *vma)
N
Nick Piggin 已提交
630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645
{
	unsigned long address;
	pte_t *pte;
	spinlock_t *ptl;

	address = vma_address(page, vma);
	if (address == -EFAULT)		/* out of vma range */
		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
646 647 648 649
/*
 * Subfunctions of page_referenced: page_referenced_one called
 * repeatedly from either page_referenced_anon or page_referenced_file.
 */
H
Hugh Dickins 已提交
650 651 652
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
653 654 655 656
{
	struct mm_struct *mm = vma->vm_mm;
	int referenced = 0;

657 658 659 660
	if (unlikely(PageTransHuge(page))) {
		pmd_t *pmd;

		spin_lock(&mm->page_table_lock);
661 662 663 664
		/*
		 * rmap might return false positives; we must filter
		 * these out using page_check_address_pmd().
		 */
665 666
		pmd = page_check_address_pmd(page, mm, address,
					     PAGE_CHECK_ADDRESS_PMD_FLAG);
667 668 669 670 671 672 673 674 675 676 677 678 679 680
		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))
681 682 683 684 685 686
			referenced++;
		spin_unlock(&mm->page_table_lock);
	} else {
		pte_t *pte;
		spinlock_t *ptl;

687 688 689 690
		/*
		 * rmap might return false positives; we must filter
		 * these out using page_check_address().
		 */
691 692 693 694
		pte = page_check_address(page, mm, address, &ptl, 0);
		if (!pte)
			goto out;

695 696 697 698 699 700 701
		if (vma->vm_flags & VM_LOCKED) {
			pte_unmap_unlock(pte, ptl);
			*mapcount = 0;	/* break early from loop */
			*vm_flags |= VM_LOCKED;
			goto out;
		}

702 703 704 705 706 707 708 709 710 711 712 713 714 715
		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.
			 */
			if (likely(!VM_SequentialReadHint(vma)))
				referenced++;
		}
		pte_unmap_unlock(pte, ptl);
	}

716 717 718 719 720 721
	/* Pretend the page is referenced if the task has the
	   swap token and is in the middle of a page fault. */
	if (mm != current->mm && has_swap_token(mm) &&
			rwsem_is_locked(&mm->mmap_sem))
		referenced++;

H
Hugh Dickins 已提交
722
	(*mapcount)--;
723

724 725
	if (referenced)
		*vm_flags |= vma->vm_flags;
726
out:
L
Linus Torvalds 已提交
727 728 729
	return referenced;
}

730
static int page_referenced_anon(struct page *page,
731 732
				struct mem_cgroup *mem_cont,
				unsigned long *vm_flags)
L
Linus Torvalds 已提交
733 734 735
{
	unsigned int mapcount;
	struct anon_vma *anon_vma;
736
	struct anon_vma_chain *avc;
L
Linus Torvalds 已提交
737 738 739 740 741 742 743
	int referenced = 0;

	anon_vma = page_lock_anon_vma(page);
	if (!anon_vma)
		return referenced;

	mapcount = page_mapcount(page);
744 745
	list_for_each_entry(avc, &anon_vma->head, same_anon_vma) {
		struct vm_area_struct *vma = avc->vma;
H
Hugh Dickins 已提交
746 747 748
		unsigned long address = vma_address(page, vma);
		if (address == -EFAULT)
			continue;
749 750 751 752 753
		/*
		 * If we are reclaiming on behalf of a cgroup, skip
		 * counting on behalf of references from different
		 * cgroups
		 */
754
		if (mem_cont && !mm_match_cgroup(vma->vm_mm, mem_cont))
755
			continue;
H
Hugh Dickins 已提交
756
		referenced += page_referenced_one(page, vma, address,
757
						  &mapcount, vm_flags);
L
Linus Torvalds 已提交
758 759 760
		if (!mapcount)
			break;
	}
761 762

	page_unlock_anon_vma(anon_vma);
L
Linus Torvalds 已提交
763 764 765 766 767 768
	return referenced;
}

/**
 * page_referenced_file - referenced check for object-based rmap
 * @page: the page we're checking references on.
R
Randy Dunlap 已提交
769
 * @mem_cont: target memory controller
770
 * @vm_flags: collect encountered vma->vm_flags who actually referenced the page
L
Linus Torvalds 已提交
771 772 773 774 775 776 777 778
 *
 * 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.
 */
779
static int page_referenced_file(struct page *page,
780 781
				struct mem_cgroup *mem_cont,
				unsigned long *vm_flags)
L
Linus Torvalds 已提交
782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800
{
	unsigned int mapcount;
	struct address_space *mapping = page->mapping;
	pgoff_t pgoff = page->index << (PAGE_CACHE_SHIFT - PAGE_SHIFT);
	struct vm_area_struct *vma;
	struct prio_tree_iter iter;
	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,
801
	 * so we can safely take mapping->i_mmap_mutex.
L
Linus Torvalds 已提交
802 803 804
	 */
	BUG_ON(!PageLocked(page));

805
	mutex_lock(&mapping->i_mmap_mutex);
L
Linus Torvalds 已提交
806 807

	/*
808
	 * i_mmap_mutex does not stabilize mapcount at all, but mapcount
L
Linus Torvalds 已提交
809 810 811 812 813
	 * is more likely to be accurate if we note it after spinning.
	 */
	mapcount = page_mapcount(page);

	vma_prio_tree_foreach(vma, &iter, &mapping->i_mmap, pgoff, pgoff) {
H
Hugh Dickins 已提交
814 815 816
		unsigned long address = vma_address(page, vma);
		if (address == -EFAULT)
			continue;
817 818 819 820 821
		/*
		 * If we are reclaiming on behalf of a cgroup, skip
		 * counting on behalf of references from different
		 * cgroups
		 */
822
		if (mem_cont && !mm_match_cgroup(vma->vm_mm, mem_cont))
823
			continue;
H
Hugh Dickins 已提交
824
		referenced += page_referenced_one(page, vma, address,
825
						  &mapcount, vm_flags);
L
Linus Torvalds 已提交
826 827 828 829
		if (!mapcount)
			break;
	}

830
	mutex_unlock(&mapping->i_mmap_mutex);
L
Linus Torvalds 已提交
831 832 833 834 835 836 837
	return referenced;
}

/**
 * page_referenced - test if the page was referenced
 * @page: the page to test
 * @is_locked: caller holds lock on the page
R
Randy Dunlap 已提交
838
 * @mem_cont: target memory controller
839
 * @vm_flags: collect encountered vma->vm_flags who actually referenced the page
L
Linus Torvalds 已提交
840 841 842 843
 *
 * Quick test_and_clear_referenced for all mappings to a page,
 * returns the number of ptes which referenced the page.
 */
844 845 846 847
int page_referenced(struct page *page,
		    int is_locked,
		    struct mem_cgroup *mem_cont,
		    unsigned long *vm_flags)
L
Linus Torvalds 已提交
848 849
{
	int referenced = 0;
H
Hugh Dickins 已提交
850
	int we_locked = 0;
L
Linus Torvalds 已提交
851

852
	*vm_flags = 0;
H
Hugh Dickins 已提交
853
	if (page_mapped(page) && page_rmapping(page)) {
H
Hugh Dickins 已提交
854 855 856 857 858 859 860 861 862 863 864
		if (!is_locked && (!PageAnon(page) || PageKsm(page))) {
			we_locked = trylock_page(page);
			if (!we_locked) {
				referenced++;
				goto out;
			}
		}
		if (unlikely(PageKsm(page)))
			referenced += page_referenced_ksm(page, mem_cont,
								vm_flags);
		else if (PageAnon(page))
865 866
			referenced += page_referenced_anon(page, mem_cont,
								vm_flags);
H
Hugh Dickins 已提交
867
		else if (page->mapping)
868 869
			referenced += page_referenced_file(page, mem_cont,
								vm_flags);
H
Hugh Dickins 已提交
870
		if (we_locked)
L
Linus Torvalds 已提交
871 872
			unlock_page(page);
	}
H
Hugh Dickins 已提交
873
out:
874
	if (page_test_and_clear_young(page_to_pfn(page)))
875 876
		referenced++;

L
Linus Torvalds 已提交
877 878 879
	return referenced;
}

H
Hugh Dickins 已提交
880 881
static int page_mkclean_one(struct page *page, struct vm_area_struct *vma,
			    unsigned long address)
882 883
{
	struct mm_struct *mm = vma->vm_mm;
884
	pte_t *pte;
885 886 887
	spinlock_t *ptl;
	int ret = 0;

N
Nick Piggin 已提交
888
	pte = page_check_address(page, mm, address, &ptl, 1);
889 890 891
	if (!pte)
		goto out;

892 893
	if (pte_dirty(*pte) || pte_write(*pte)) {
		pte_t entry;
894

895
		flush_cache_page(vma, address, pte_pfn(*pte));
A
Andrea Arcangeli 已提交
896
		entry = ptep_clear_flush_notify(vma, address, pte);
897 898
		entry = pte_wrprotect(entry);
		entry = pte_mkclean(entry);
899
		set_pte_at(mm, address, pte, entry);
900 901
		ret = 1;
	}
902 903 904 905 906 907 908 909 910 911 912 913 914 915 916

	pte_unmap_unlock(pte, ptl);
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;
	struct prio_tree_iter iter;
	int ret = 0;

	BUG_ON(PageAnon(page));

917
	mutex_lock(&mapping->i_mmap_mutex);
918
	vma_prio_tree_foreach(vma, &iter, &mapping->i_mmap, pgoff, pgoff) {
H
Hugh Dickins 已提交
919 920 921 922 923 924
		if (vma->vm_flags & VM_SHARED) {
			unsigned long address = vma_address(page, vma);
			if (address == -EFAULT)
				continue;
			ret += page_mkclean_one(page, vma, address);
		}
925
	}
926
	mutex_unlock(&mapping->i_mmap_mutex);
927 928 929 930 931 932 933 934 935 936 937
	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);
938
		if (mapping) {
939
			ret = page_mkclean_file(mapping, page);
940
			if (page_test_and_clear_dirty(page_to_pfn(page), 1))
941
				ret = 1;
942
		}
943 944 945 946
	}

	return ret;
}
J
Jaya Kumar 已提交
947
EXPORT_SYMBOL_GPL(page_mkclean);
948

949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972
/**
 * 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
973
/**
A
Andi Kleen 已提交
974 975 976 977
 * __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	
978
 * @exclusive:	the page is exclusively owned by the current process
N
Nick Piggin 已提交
979 980
 */
static void __page_set_anon_rmap(struct page *page,
981
	struct vm_area_struct *vma, unsigned long address, int exclusive)
N
Nick Piggin 已提交
982
{
983
	struct anon_vma *anon_vma = vma->anon_vma;
984

985
	BUG_ON(!anon_vma);
986

A
Andi Kleen 已提交
987 988 989
	if (PageAnon(page))
		return;

990
	/*
991 992 993
	 * If the page isn't exclusively mapped into this vma,
	 * we must use the _oldest_ possible anon_vma for the
	 * page mapping!
994
	 */
A
Andi Kleen 已提交
995
	if (!exclusive)
996
		anon_vma = anon_vma->root;
N
Nick Piggin 已提交
997 998 999 1000 1001 1002

	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
1003
/**
R
Randy Dunlap 已提交
1004
 * __page_check_anon_rmap - sanity check anonymous rmap addition
N
Nick Piggin 已提交
1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024
 * @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
1025
	BUG_ON(page_anon_vma(page)->root != vma->anon_vma->root);
N
Nick Piggin 已提交
1026 1027 1028 1029
	BUG_ON(page->index != linear_page_index(vma, address));
#endif
}

L
Linus Torvalds 已提交
1030 1031 1032 1033 1034 1035
/**
 * 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
1036
 * The caller needs to hold the pte lock, and the page must be locked in
1037 1038 1039
 * 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
1040 1041 1042
 */
void page_add_anon_rmap(struct page *page,
	struct vm_area_struct *vma, unsigned long address)
1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053
{
	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
1054
{
H
Hugh Dickins 已提交
1055
	int first = atomic_inc_and_test(&page->_mapcount);
1056 1057 1058 1059 1060 1061 1062
	if (first) {
		if (!PageTransHuge(page))
			__inc_zone_page_state(page, NR_ANON_PAGES);
		else
			__inc_zone_page_state(page,
					      NR_ANON_TRANSPARENT_HUGEPAGES);
	}
H
Hugh Dickins 已提交
1063 1064 1065
	if (unlikely(PageKsm(page)))
		return;

N
Nick Piggin 已提交
1066
	VM_BUG_ON(!PageLocked(page));
1067
	/* address might be in next vma when migration races vma_adjust */
H
Hugh Dickins 已提交
1068
	if (first)
1069
		__page_set_anon_rmap(page, vma, address, exclusive);
1070
	else
N
Nick Piggin 已提交
1071
		__page_check_anon_rmap(page, vma, address);
L
Linus Torvalds 已提交
1072 1073
}

R
Randy Dunlap 已提交
1074
/**
N
Nick Piggin 已提交
1075 1076 1077 1078 1079 1080 1081
 * 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
1082
 * Page does not have to be locked.
N
Nick Piggin 已提交
1083 1084 1085 1086
 */
void page_add_new_anon_rmap(struct page *page,
	struct vm_area_struct *vma, unsigned long address)
{
1087
	VM_BUG_ON(address < vma->vm_start || address >= vma->vm_end);
1088 1089
	SetPageSwapBacked(page);
	atomic_set(&page->_mapcount, 0); /* increment count (starts at -1) */
1090 1091 1092 1093
	if (!PageTransHuge(page))
		__inc_zone_page_state(page, NR_ANON_PAGES);
	else
		__inc_zone_page_state(page, NR_ANON_TRANSPARENT_HUGEPAGES);
1094
	__page_set_anon_rmap(page, vma, address, 1);
1095
	if (page_evictable(page, vma))
1096
		lru_cache_add_lru(page, LRU_ACTIVE_ANON);
1097 1098
	else
		add_page_to_unevictable_list(page);
N
Nick Piggin 已提交
1099 1100
}

L
Linus Torvalds 已提交
1101 1102 1103 1104
/**
 * page_add_file_rmap - add pte mapping to a file page
 * @page: the page to add the mapping to
 *
1105
 * The caller needs to hold the pte lock.
L
Linus Torvalds 已提交
1106 1107 1108
 */
void page_add_file_rmap(struct page *page)
{
1109
	if (atomic_inc_and_test(&page->_mapcount)) {
1110
		__inc_zone_page_state(page, NR_FILE_MAPPED);
1111
		mem_cgroup_inc_page_stat(page, MEMCG_NR_FILE_MAPPED);
1112
	}
L
Linus Torvalds 已提交
1113 1114 1115 1116 1117 1118
}

/**
 * page_remove_rmap - take down pte mapping from a page
 * @page: page to remove mapping from
 *
1119
 * The caller needs to hold the pte lock.
L
Linus Torvalds 已提交
1120
 */
1121
void page_remove_rmap(struct page *page)
L
Linus Torvalds 已提交
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))
		return;

	/*
	 * Now that the last pte has gone, s390 must transfer dirty
	 * flag from storage key to struct page.  We can usually skip
	 * this if the page is anon, so about to be freed; but perhaps
	 * not if it's in swapcache - there might be another pte slot
	 * containing the swap entry, but page not yet written to swap.
	 */
1134 1135
	if ((!PageAnon(page) || PageSwapCache(page)) &&
	    page_test_and_clear_dirty(page_to_pfn(page), 1))
1136
		set_page_dirty(page);
1137 1138 1139 1140 1141 1142
	/*
	 * Hugepages are not counted in NR_ANON_PAGES nor NR_FILE_MAPPED
	 * and not charged by memcg for now.
	 */
	if (unlikely(PageHuge(page)))
		return;
1143 1144
	if (PageAnon(page)) {
		mem_cgroup_uncharge_page(page);
1145 1146 1147 1148 1149
		if (!PageTransHuge(page))
			__dec_zone_page_state(page, NR_ANON_PAGES);
		else
			__dec_zone_page_state(page,
					      NR_ANON_TRANSPARENT_HUGEPAGES);
1150 1151
	} else {
		__dec_zone_page_state(page, NR_FILE_MAPPED);
1152
		mem_cgroup_dec_page_stat(page, MEMCG_NR_FILE_MAPPED);
1153 1154 1155 1156 1157 1158 1159 1160 1161 1162
	}
	/*
	 * 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
1163 1164 1165 1166 1167 1168
}

/*
 * Subfunctions of try_to_unmap: try_to_unmap_one called
 * repeatedly from either try_to_unmap_anon or try_to_unmap_file.
 */
H
Hugh Dickins 已提交
1169 1170
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
1171 1172 1173 1174
{
	struct mm_struct *mm = vma->vm_mm;
	pte_t *pte;
	pte_t pteval;
H
Hugh Dickins 已提交
1175
	spinlock_t *ptl;
L
Linus Torvalds 已提交
1176 1177
	int ret = SWAP_AGAIN;

N
Nick Piggin 已提交
1178
	pte = page_check_address(page, mm, address, &ptl, 0);
H
Hugh Dickins 已提交
1179
	if (!pte)
N
Nikita Danilov 已提交
1180
		goto out;
L
Linus Torvalds 已提交
1181 1182 1183 1184 1185 1186

	/*
	 * 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.
	 */
1187
	if (!(flags & TTU_IGNORE_MLOCK)) {
K
KOSAKI Motohiro 已提交
1188 1189 1190
		if (vma->vm_flags & VM_LOCKED)
			goto out_mlock;

H
Hugh Dickins 已提交
1191
		if (TTU_ACTION(flags) == TTU_MUNLOCK)
H
Hugh Dickins 已提交
1192
			goto out_unmap;
1193 1194
	}
	if (!(flags & TTU_IGNORE_ACCESS)) {
N
Nick Piggin 已提交
1195 1196 1197 1198 1199
		if (ptep_clear_flush_young_notify(vma, address, pte)) {
			ret = SWAP_FAIL;
			goto out_unmap;
		}
  	}
L
Linus Torvalds 已提交
1200 1201 1202

	/* Nuke the page table entry. */
	flush_cache_page(vma, address, page_to_pfn(page));
A
Andrea Arcangeli 已提交
1203
	pteval = ptep_clear_flush_notify(vma, address, pte);
L
Linus Torvalds 已提交
1204 1205 1206 1207 1208

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

1209 1210 1211
	/* Update high watermark before we lower rss */
	update_hiwater_rss(mm);

1212 1213
	if (PageHWPoison(page) && !(flags & TTU_IGNORE_HWPOISON)) {
		if (PageAnon(page))
K
KAMEZAWA Hiroyuki 已提交
1214
			dec_mm_counter(mm, MM_ANONPAGES);
1215
		else
K
KAMEZAWA Hiroyuki 已提交
1216
			dec_mm_counter(mm, MM_FILEPAGES);
1217 1218 1219
		set_pte_at(mm, address, pte,
				swp_entry_to_pte(make_hwpoison_entry(page)));
	} else if (PageAnon(page)) {
H
Hugh Dickins 已提交
1220
		swp_entry_t entry = { .val = page_private(page) };
1221 1222 1223 1224 1225 1226

		if (PageSwapCache(page)) {
			/*
			 * Store the swap location in the pte.
			 * See handle_pte_fault() ...
			 */
H
Hugh Dickins 已提交
1227 1228 1229 1230 1231
			if (swap_duplicate(entry) < 0) {
				set_pte_at(mm, address, pte, pteval);
				ret = SWAP_FAIL;
				goto out_unmap;
			}
1232 1233 1234 1235 1236 1237
			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
1238
			dec_mm_counter(mm, MM_ANONPAGES);
K
KAMEZAWA Hiroyuki 已提交
1239
			inc_mm_counter(mm, MM_SWAPENTS);
1240
		} else if (PAGE_MIGRATION) {
1241 1242 1243 1244 1245
			/*
			 * 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.
			 */
1246
			BUG_ON(TTU_ACTION(flags) != TTU_MIGRATION);
1247
			entry = make_migration_entry(page, pte_write(pteval));
L
Linus Torvalds 已提交
1248 1249 1250
		}
		set_pte_at(mm, address, pte, swp_entry_to_pte(entry));
		BUG_ON(pte_file(*pte));
1251
	} else if (PAGE_MIGRATION && (TTU_ACTION(flags) == TTU_MIGRATION)) {
1252 1253 1254 1255 1256
		/* 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
1257
		dec_mm_counter(mm, MM_FILEPAGES);
L
Linus Torvalds 已提交
1258

1259
	page_remove_rmap(page);
L
Linus Torvalds 已提交
1260 1261 1262
	page_cache_release(page);

out_unmap:
H
Hugh Dickins 已提交
1263
	pte_unmap_unlock(pte, ptl);
K
KOSAKI Motohiro 已提交
1264 1265
out:
	return ret;
H
Hugh Dickins 已提交
1266

K
KOSAKI Motohiro 已提交
1267 1268 1269 1270 1271 1272 1273
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
1274
	 * we now hold anon_vma->mutex or mapping->i_mmap_mutex.
K
KOSAKI Motohiro 已提交
1275 1276 1277 1278 1279 1280 1281 1282
	 * 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
1283
		}
K
KOSAKI Motohiro 已提交
1284
		up_read(&vma->vm_mm->mmap_sem);
H
Hugh Dickins 已提交
1285
	}
L
Linus Torvalds 已提交
1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306
	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
1307 1308 1309 1310 1311
 *
 * 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
1312 1313 1314 1315
 */
#define CLUSTER_SIZE	min(32*PAGE_SIZE, PMD_SIZE)
#define CLUSTER_MASK	(~(CLUSTER_SIZE - 1))

N
Nick Piggin 已提交
1316 1317
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
1318 1319 1320 1321 1322
{
	struct mm_struct *mm = vma->vm_mm;
	pgd_t *pgd;
	pud_t *pud;
	pmd_t *pmd;
H
Hugh Dickins 已提交
1323
	pte_t *pte;
L
Linus Torvalds 已提交
1324
	pte_t pteval;
H
Hugh Dickins 已提交
1325
	spinlock_t *ptl;
L
Linus Torvalds 已提交
1326 1327 1328
	struct page *page;
	unsigned long address;
	unsigned long end;
N
Nick Piggin 已提交
1329 1330
	int ret = SWAP_AGAIN;
	int locked_vma = 0;
L
Linus Torvalds 已提交
1331 1332 1333 1334 1335 1336 1337 1338 1339 1340

	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;

	pgd = pgd_offset(mm, address);
	if (!pgd_present(*pgd))
N
Nick Piggin 已提交
1341
		return ret;
L
Linus Torvalds 已提交
1342 1343 1344

	pud = pud_offset(pgd, address);
	if (!pud_present(*pud))
N
Nick Piggin 已提交
1345
		return ret;
L
Linus Torvalds 已提交
1346 1347 1348

	pmd = pmd_offset(pud, address);
	if (!pmd_present(*pmd))
N
Nick Piggin 已提交
1349 1350 1351
		return ret;

	/*
H
Hugh Dickins 已提交
1352
	 * If we can acquire the mmap_sem for read, and vma is VM_LOCKED,
N
Nick Piggin 已提交
1353 1354
	 * keep the sem while scanning the cluster for mlocking pages.
	 */
H
Hugh Dickins 已提交
1355
	if (down_read_trylock(&vma->vm_mm->mmap_sem)) {
N
Nick Piggin 已提交
1356 1357 1358 1359
		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
1360 1361

	pte = pte_offset_map_lock(mm, pmd, address, &ptl);
L
Linus Torvalds 已提交
1362

1363 1364 1365
	/* Update high watermark before we lower rss */
	update_hiwater_rss(mm);

H
Hugh Dickins 已提交
1366
	for (; address < end; pte++, address += PAGE_SIZE) {
L
Linus Torvalds 已提交
1367 1368
		if (!pte_present(*pte))
			continue;
1369 1370
		page = vm_normal_page(vma, address, *pte);
		BUG_ON(!page || PageAnon(page));
L
Linus Torvalds 已提交
1371

N
Nick Piggin 已提交
1372 1373 1374 1375 1376 1377 1378
		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
1379
		if (ptep_clear_flush_young_notify(vma, address, pte))
L
Linus Torvalds 已提交
1380 1381 1382
			continue;

		/* Nuke the page table entry. */
1383
		flush_cache_page(vma, address, pte_pfn(*pte));
A
Andrea Arcangeli 已提交
1384
		pteval = ptep_clear_flush_notify(vma, address, pte);
L
Linus Torvalds 已提交
1385 1386 1387 1388 1389 1390 1391 1392 1393

		/* If nonlinear, store the file page offset in the pte. */
		if (page->index != linear_page_index(vma, address))
			set_pte_at(mm, address, pte, pgoff_to_pte(page->index));

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

1394
		page_remove_rmap(page);
L
Linus Torvalds 已提交
1395
		page_cache_release(page);
K
KAMEZAWA Hiroyuki 已提交
1396
		dec_mm_counter(mm, MM_FILEPAGES);
L
Linus Torvalds 已提交
1397 1398
		(*mapcount)--;
	}
H
Hugh Dickins 已提交
1399
	pte_unmap_unlock(pte - 1, ptl);
N
Nick Piggin 已提交
1400 1401 1402
	if (locked_vma)
		up_read(&vma->vm_mm->mmap_sem);
	return ret;
L
Linus Torvalds 已提交
1403 1404
}

1405
bool is_vma_temporary_stack(struct vm_area_struct *vma)
1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418
{
	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
1419 1420 1421 1422
/**
 * try_to_unmap_anon - unmap or unlock anonymous page using the object-based
 * rmap method
 * @page: the page to unmap/unlock
1423
 * @flags: action and flags
N
Nick Piggin 已提交
1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434
 *
 * 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.
 */
1435
static int try_to_unmap_anon(struct page *page, enum ttu_flags flags)
L
Linus Torvalds 已提交
1436 1437
{
	struct anon_vma *anon_vma;
1438
	struct anon_vma_chain *avc;
L
Linus Torvalds 已提交
1439
	int ret = SWAP_AGAIN;
N
Nick Piggin 已提交
1440

L
Linus Torvalds 已提交
1441 1442 1443 1444
	anon_vma = page_lock_anon_vma(page);
	if (!anon_vma)
		return ret;

1445 1446
	list_for_each_entry(avc, &anon_vma->head, same_anon_vma) {
		struct vm_area_struct *vma = avc->vma;
1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461
		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.
		 */
		if (PAGE_MIGRATION && (flags & TTU_MIGRATION) &&
				is_vma_temporary_stack(vma))
			continue;

		address = vma_address(page, vma);
H
Hugh Dickins 已提交
1462 1463 1464
		if (address == -EFAULT)
			continue;
		ret = try_to_unmap_one(page, vma, address, flags);
H
Hugh Dickins 已提交
1465 1466
		if (ret != SWAP_AGAIN || !page_mapped(page))
			break;
L
Linus Torvalds 已提交
1467
	}
1468 1469

	page_unlock_anon_vma(anon_vma);
L
Linus Torvalds 已提交
1470 1471 1472 1473
	return ret;
}

/**
N
Nick Piggin 已提交
1474 1475
 * try_to_unmap_file - unmap/unlock file page using the object-based rmap method
 * @page: the page to unmap/unlock
1476
 * @flags: action and flags
L
Linus Torvalds 已提交
1477 1478 1479 1480
 *
 * 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
1481 1482 1483 1484 1485 1486
 * 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
1487
 */
1488
static int try_to_unmap_file(struct page *page, enum ttu_flags flags)
L
Linus Torvalds 已提交
1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499
{
	struct address_space *mapping = page->mapping;
	pgoff_t pgoff = page->index << (PAGE_CACHE_SHIFT - PAGE_SHIFT);
	struct vm_area_struct *vma;
	struct prio_tree_iter iter;
	int ret = SWAP_AGAIN;
	unsigned long cursor;
	unsigned long max_nl_cursor = 0;
	unsigned long max_nl_size = 0;
	unsigned int mapcount;

1500
	mutex_lock(&mapping->i_mmap_mutex);
L
Linus Torvalds 已提交
1501
	vma_prio_tree_foreach(vma, &iter, &mapping->i_mmap, pgoff, pgoff) {
H
Hugh Dickins 已提交
1502 1503 1504 1505
		unsigned long address = vma_address(page, vma);
		if (address == -EFAULT)
			continue;
		ret = try_to_unmap_one(page, vma, address, flags);
H
Hugh Dickins 已提交
1506 1507
		if (ret != SWAP_AGAIN || !page_mapped(page))
			goto out;
L
Linus Torvalds 已提交
1508 1509 1510 1511 1512
	}

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

H
Hugh Dickins 已提交
1513 1514 1515 1516 1517 1518 1519 1520
	/*
	 * 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
1521 1522 1523 1524 1525 1526 1527 1528 1529 1530
	list_for_each_entry(vma, &mapping->i_mmap_nonlinear,
						shared.vm_set.list) {
		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
1531
	if (max_nl_size == 0) {	/* all nonlinears locked or reserved ? */
L
Linus Torvalds 已提交
1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545
		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;
1546
	cond_resched();
L
Linus Torvalds 已提交
1547 1548 1549 1550 1551 1552 1553 1554 1555

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

	do {
		list_for_each_entry(vma, &mapping->i_mmap_nonlinear,
						shared.vm_set.list) {
			cursor = (unsigned long) vma->vm_private_data;
H
Hugh Dickins 已提交
1556
			while ( cursor < max_nl_cursor &&
L
Linus Torvalds 已提交
1557
				cursor < vma->vm_end - vma->vm_start) {
H
Hugh Dickins 已提交
1558 1559 1560
				if (try_to_unmap_cluster(cursor, &mapcount,
						vma, page) == SWAP_MLOCK)
					ret = SWAP_MLOCK;
L
Linus Torvalds 已提交
1561 1562 1563 1564 1565 1566 1567
				cursor += CLUSTER_SIZE;
				vma->vm_private_data = (void *) cursor;
				if ((int)mapcount <= 0)
					goto out;
			}
			vma->vm_private_data = (void *) max_nl_cursor;
		}
1568
		cond_resched();
L
Linus Torvalds 已提交
1569 1570 1571 1572 1573 1574 1575 1576
		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.
	 */
1577 1578
	list_for_each_entry(vma, &mapping->i_mmap_nonlinear, shared.vm_set.list)
		vma->vm_private_data = NULL;
L
Linus Torvalds 已提交
1579
out:
1580
	mutex_unlock(&mapping->i_mmap_mutex);
L
Linus Torvalds 已提交
1581 1582 1583 1584 1585 1586
	return ret;
}

/**
 * try_to_unmap - try to remove all page table mappings to a page
 * @page: the page to get unmapped
1587
 * @flags: action and flags
L
Linus Torvalds 已提交
1588 1589 1590 1591 1592 1593 1594 1595
 *
 * 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
1596
 * SWAP_MLOCK	- page is mlocked.
L
Linus Torvalds 已提交
1597
 */
1598
int try_to_unmap(struct page *page, enum ttu_flags flags)
L
Linus Torvalds 已提交
1599 1600 1601 1602
{
	int ret;

	BUG_ON(!PageLocked(page));
1603
	VM_BUG_ON(!PageHuge(page) && PageTransHuge(page));
L
Linus Torvalds 已提交
1604

H
Hugh Dickins 已提交
1605 1606 1607
	if (unlikely(PageKsm(page)))
		ret = try_to_unmap_ksm(page, flags);
	else if (PageAnon(page))
1608
		ret = try_to_unmap_anon(page, flags);
L
Linus Torvalds 已提交
1609
	else
1610
		ret = try_to_unmap_file(page, flags);
N
Nick Piggin 已提交
1611
	if (ret != SWAP_MLOCK && !page_mapped(page))
L
Linus Torvalds 已提交
1612 1613 1614
		ret = SWAP_SUCCESS;
	return ret;
}
N
Nikita Danilov 已提交
1615

N
Nick Piggin 已提交
1616 1617 1618 1619 1620 1621 1622 1623 1624 1625
/**
 * 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
1626
 * SWAP_AGAIN	- no vma is holding page mlocked, or,
N
Nick Piggin 已提交
1627
 * SWAP_AGAIN	- page mapped in mlocked vma -- couldn't acquire mmap sem
H
Hugh Dickins 已提交
1628
 * SWAP_FAIL	- page cannot be located at present
N
Nick Piggin 已提交
1629 1630 1631 1632 1633 1634
 * 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
1635 1636 1637
	if (unlikely(PageKsm(page)))
		return try_to_unmap_ksm(page, TTU_MUNLOCK);
	else if (PageAnon(page))
1638
		return try_to_unmap_anon(page, TTU_MUNLOCK);
N
Nick Piggin 已提交
1639
	else
1640
		return try_to_unmap_file(page, TTU_MUNLOCK);
N
Nick Piggin 已提交
1641
}
1642

P
Peter Zijlstra 已提交
1643
void __put_anon_vma(struct anon_vma *anon_vma)
1644
{
P
Peter Zijlstra 已提交
1645
	struct anon_vma *root = anon_vma->root;
1646

P
Peter Zijlstra 已提交
1647 1648
	if (root != anon_vma && atomic_dec_and_test(&root->refcount))
		anon_vma_free(root);
1649

P
Peter Zijlstra 已提交
1650
	anon_vma_free(anon_vma);
1651 1652
}

1653 1654 1655 1656 1657 1658 1659 1660 1661
#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;
1662
	struct anon_vma_chain *avc;
1663 1664 1665 1666 1667
	int ret = SWAP_AGAIN;

	/*
	 * Note: remove_migration_ptes() cannot use page_lock_anon_vma()
	 * because that depends on page_mapped(); but not all its usages
1668 1669
	 * are holding mmap_sem. Users without mmap_sem are required to
	 * take a reference count to prevent the anon_vma disappearing
1670 1671 1672 1673
	 */
	anon_vma = page_anon_vma(page);
	if (!anon_vma)
		return ret;
1674
	anon_vma_lock(anon_vma);
1675 1676
	list_for_each_entry(avc, &anon_vma->head, same_anon_vma) {
		struct vm_area_struct *vma = avc->vma;
1677 1678 1679 1680 1681 1682 1683
		unsigned long address = vma_address(page, vma);
		if (address == -EFAULT)
			continue;
		ret = rmap_one(page, vma, address, arg);
		if (ret != SWAP_AGAIN)
			break;
	}
1684
	anon_vma_unlock(anon_vma);
1685 1686 1687 1688 1689 1690 1691 1692 1693 1694 1695 1696 1697 1698
	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;
	struct prio_tree_iter iter;
	int ret = SWAP_AGAIN;

	if (!mapping)
		return ret;
1699
	mutex_lock(&mapping->i_mmap_mutex);
1700 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712
	vma_prio_tree_foreach(vma, &iter, &mapping->i_mmap, pgoff, pgoff) {
		unsigned long address = vma_address(page, vma);
		if (address == -EFAULT)
			continue;
		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.
	 */
1713
	mutex_unlock(&mapping->i_mmap_mutex);
1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724 1725 1726 1727 1728 1729
	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 */
1730

N
Naoya Horiguchi 已提交
1731
#ifdef CONFIG_HUGETLB_PAGE
1732 1733 1734 1735 1736 1737 1738 1739 1740
/*
 * 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;
1741

1742
	BUG_ON(!anon_vma);
1743 1744 1745 1746 1747 1748

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

1749 1750 1751 1752 1753 1754 1755 1756 1757 1758
	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;
1759 1760

	BUG_ON(!PageLocked(page));
1761
	BUG_ON(!anon_vma);
1762
	/* address might be in next vma when migration races vma_adjust */
1763 1764 1765 1766 1767 1768 1769 1770 1771 1772 1773 1774
	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
1775
#endif /* CONFIG_HUGETLB_PAGE */