rmap.c 46.9 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
/*
 * mm/rmap.c - physical to virtual reverse mappings
 *
 * Copyright 2001, Rik van Riel <riel@conectiva.com.br>
 * Released under the General Public License (GPL).
 *
 * Simple, low overhead reverse mapping scheme.
 * Please try to keep this thing as modular as possible.
 *
 * Provides methods for unmapping each kind of mapped page:
 * the anon methods track anonymous pages, and
 * the file methods track pages belonging to an inode.
 *
 * Original design by Rik van Riel <riel@conectiva.com.br> 2001
 * File methods by Dave McCracken <dmccr@us.ibm.com> 2003, 2004
 * Anonymous methods by Andrea Arcangeli <andrea@suse.de> 2004
H
Hugh Dickins 已提交
17
 * Contributions by Hugh Dickins 2003, 2004
L
Linus Torvalds 已提交
18 19 20 21 22
 */

/*
 * Lock ordering in mm:
 *
23
 * inode->i_mutex	(while writing or truncating, not reading or faulting)
24 25
 *   mm->mmap_sem
 *     page->flags PG_locked (lock_page)
26
 *       mapping->i_mmap_rwsem
27
 *         anon_vma->rwsem
28 29 30 31 32
 *           mm->page_table_lock or pte_lock
 *             zone->lru_lock (in mark_page_accessed, isolate_lru_page)
 *             swap_lock (in swap_duplicate, swap_info_get)
 *               mmlist_lock (in mmput, drain_mmlist and others)
 *               mapping->private_lock (in __set_page_dirty_buffers)
33 34
 *                 mem_cgroup_{begin,end}_page_stat (memcg->move_lock)
 *                   mapping->tree_lock (widely used)
35
 *               inode->i_lock (in set_page_dirty's __mark_inode_dirty)
36
 *               bdi.wb->list_lock (in set_page_dirty's __mark_inode_dirty)
37 38 39
 *                 sb_lock (within inode_lock in fs/fs-writeback.c)
 *                 mapping->tree_lock (widely used, in set_page_dirty,
 *                           in arch-dependent flush_dcache_mmap_lock,
40
 *                           within bdi.wb->list_lock in __sync_single_inode)
41
 *
42
 * anon_vma->rwsem,mapping->i_mutex      (memory_failure, collect_procs_anon)
43
 *   ->tasklist_lock
44
 *     pte map lock
L
Linus Torvalds 已提交
45 46 47 48 49 50 51 52
 */

#include <linux/mm.h>
#include <linux/pagemap.h>
#include <linux/swap.h>
#include <linux/swapops.h>
#include <linux/slab.h>
#include <linux/init.h>
H
Hugh Dickins 已提交
53
#include <linux/ksm.h>
L
Linus Torvalds 已提交
54 55
#include <linux/rmap.h>
#include <linux/rcupdate.h>
56
#include <linux/export.h>
57
#include <linux/memcontrol.h>
A
Andrea Arcangeli 已提交
58
#include <linux/mmu_notifier.h>
59
#include <linux/migrate.h>
60
#include <linux/hugetlb.h>
61
#include <linux/backing-dev.h>
62
#include <linux/page_idle.h>
L
Linus Torvalds 已提交
63 64 65

#include <asm/tlbflush.h>

66 67
#include <trace/events/tlb.h>

N
Nick Piggin 已提交
68 69
#include "internal.h"

70
static struct kmem_cache *anon_vma_cachep;
71
static struct kmem_cache *anon_vma_chain_cachep;
72 73 74

static inline struct anon_vma *anon_vma_alloc(void)
{
P
Peter Zijlstra 已提交
75 76 77 78 79
	struct anon_vma *anon_vma;

	anon_vma = kmem_cache_alloc(anon_vma_cachep, GFP_KERNEL);
	if (anon_vma) {
		atomic_set(&anon_vma->refcount, 1);
80 81
		anon_vma->degree = 1;	/* Reference for first vma */
		anon_vma->parent = anon_vma;
P
Peter Zijlstra 已提交
82 83 84 85 86 87 88 89
		/*
		 * 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;
90 91
}

P
Peter Zijlstra 已提交
92
static inline void anon_vma_free(struct anon_vma *anon_vma)
93
{
P
Peter Zijlstra 已提交
94
	VM_BUG_ON(atomic_read(&anon_vma->refcount));
95 96

	/*
97
	 * Synchronize against page_lock_anon_vma_read() such that
98 99 100 101 102
	 * 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
103
	 * down_read_trylock() from page_lock_anon_vma_read(). This orders:
104
	 *
105 106
	 * page_lock_anon_vma_read()	VS	put_anon_vma()
	 *   down_read_trylock()		  atomic_dec_and_test()
107
	 *   LOCK				  MB
108
	 *   atomic_read()			  rwsem_is_locked()
109 110 111 112
	 *
	 * LOCK should suffice since the actual taking of the lock must
	 * happen _before_ what follows.
	 */
113
	might_sleep();
114
	if (rwsem_is_locked(&anon_vma->root->rwsem)) {
115
		anon_vma_lock_write(anon_vma);
116
		anon_vma_unlock_write(anon_vma);
117 118
	}

119 120
	kmem_cache_free(anon_vma_cachep, anon_vma);
}
L
Linus Torvalds 已提交
121

122
static inline struct anon_vma_chain *anon_vma_chain_alloc(gfp_t gfp)
123
{
124
	return kmem_cache_alloc(anon_vma_chain_cachep, gfp);
125 126
}

127
static void anon_vma_chain_free(struct anon_vma_chain *anon_vma_chain)
128 129 130 131
{
	kmem_cache_free(anon_vma_chain_cachep, anon_vma_chain);
}

132 133 134 135 136 137 138
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);
139
	anon_vma_interval_tree_insert(avc, &anon_vma->rb_root);
140 141
}

142 143 144 145 146 147 148 149 150
/**
 * 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
151
 * not we either need to find an adjacent mapping that we
152 153 154 155 156
 * 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
157
 * optimistically looked up an anon_vma in page_lock_anon_vma_read()
158 159 160 161 162 163 164 165 166 167 168
 * 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
169 170 171
int anon_vma_prepare(struct vm_area_struct *vma)
{
	struct anon_vma *anon_vma = vma->anon_vma;
172
	struct anon_vma_chain *avc;
L
Linus Torvalds 已提交
173 174 175 176

	might_sleep();
	if (unlikely(!anon_vma)) {
		struct mm_struct *mm = vma->vm_mm;
177
		struct anon_vma *allocated;
L
Linus Torvalds 已提交
178

179
		avc = anon_vma_chain_alloc(GFP_KERNEL);
180 181 182
		if (!avc)
			goto out_enomem;

L
Linus Torvalds 已提交
183
		anon_vma = find_mergeable_anon_vma(vma);
184 185
		allocated = NULL;
		if (!anon_vma) {
L
Linus Torvalds 已提交
186 187
			anon_vma = anon_vma_alloc();
			if (unlikely(!anon_vma))
188
				goto out_enomem_free_avc;
L
Linus Torvalds 已提交
189 190 191
			allocated = anon_vma;
		}

192
		anon_vma_lock_write(anon_vma);
L
Linus Torvalds 已提交
193 194 195 196
		/* page_table_lock to protect against threads */
		spin_lock(&mm->page_table_lock);
		if (likely(!vma->anon_vma)) {
			vma->anon_vma = anon_vma;
197
			anon_vma_chain_link(vma, avc, anon_vma);
198 199
			/* vma reference or self-parent link for new root */
			anon_vma->degree++;
L
Linus Torvalds 已提交
200
			allocated = NULL;
201
			avc = NULL;
L
Linus Torvalds 已提交
202 203
		}
		spin_unlock(&mm->page_table_lock);
204
		anon_vma_unlock_write(anon_vma);
205 206

		if (unlikely(allocated))
P
Peter Zijlstra 已提交
207
			put_anon_vma(allocated);
208
		if (unlikely(avc))
209
			anon_vma_chain_free(avc);
L
Linus Torvalds 已提交
210 211
	}
	return 0;
212 213 214 215 216

 out_enomem_free_avc:
	anon_vma_chain_free(avc);
 out_enomem:
	return -ENOMEM;
L
Linus Torvalds 已提交
217 218
}

219 220 221 222 223 224 225 226 227 228 229 230 231
/*
 * 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))
232
			up_write(&root->rwsem);
233
		root = new_root;
234
		down_write(&root->rwsem);
235 236 237 238 239 240 241
	}
	return root;
}

static inline void unlock_anon_vma_root(struct anon_vma *root)
{
	if (root)
242
		up_write(&root->rwsem);
243 244
}

245 246 247
/*
 * Attach the anon_vmas from src to dst.
 * Returns 0 on success, -ENOMEM on failure.
248 249 250 251 252 253 254 255
 *
 * If dst->anon_vma is NULL this function tries to find and reuse existing
 * anon_vma which has no vmas and only one child anon_vma. This prevents
 * degradation of anon_vma hierarchy to endless linear chain in case of
 * constantly forking task. On the other hand, an anon_vma with more than one
 * child isn't reused even if there was no alive vma, thus rmap walker has a
 * good chance of avoiding scanning the whole hierarchy when it searches where
 * page is mapped.
256 257
 */
int anon_vma_clone(struct vm_area_struct *dst, struct vm_area_struct *src)
L
Linus Torvalds 已提交
258
{
259
	struct anon_vma_chain *avc, *pavc;
260
	struct anon_vma *root = NULL;
261

262
	list_for_each_entry_reverse(pavc, &src->anon_vma_chain, same_vma) {
263 264
		struct anon_vma *anon_vma;

265 266 267 268 269 270 271 272
		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;
		}
273 274 275
		anon_vma = pavc->anon_vma;
		root = lock_anon_vma_root(root, anon_vma);
		anon_vma_chain_link(dst, avc, anon_vma);
276 277 278 279 280 281 282 283 284 285 286 287

		/*
		 * Reuse existing anon_vma if its degree lower than two,
		 * that means it has no vma and only one anon_vma child.
		 *
		 * Do not chose parent anon_vma, otherwise first child
		 * will always reuse it. Root anon_vma is never reused:
		 * it has self-parent reference and at least one child.
		 */
		if (!dst->anon_vma && anon_vma != src->anon_vma &&
				anon_vma->degree < 2)
			dst->anon_vma = anon_vma;
288
	}
289 290
	if (dst->anon_vma)
		dst->anon_vma->degree++;
291
	unlock_anon_vma_root(root);
292
	return 0;
L
Linus Torvalds 已提交
293

294
 enomem_failure:
295 296 297 298 299 300 301
	/*
	 * dst->anon_vma is dropped here otherwise its degree can be incorrectly
	 * decremented in unlink_anon_vmas().
	 * We can safely do this because callers of anon_vma_clone() don't care
	 * about dst->anon_vma if anon_vma_clone() failed.
	 */
	dst->anon_vma = NULL;
302 303
	unlink_anon_vmas(dst);
	return -ENOMEM;
L
Linus Torvalds 已提交
304 305
}

306 307 308 309 310 311
/*
 * 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
312
{
313 314
	struct anon_vma_chain *avc;
	struct anon_vma *anon_vma;
315
	int error;
L
Linus Torvalds 已提交
316

317 318 319 320
	/* Don't bother if the parent process has no anon_vma here. */
	if (!pvma->anon_vma)
		return 0;

321 322 323
	/* Drop inherited anon_vma, we'll reuse existing or allocate new. */
	vma->anon_vma = NULL;

324 325 326 327
	/*
	 * First, attach the new VMA to the parent VMA's anon_vmas,
	 * so rmap can find non-COWed pages in child processes.
	 */
328 329 330
	error = anon_vma_clone(vma, pvma);
	if (error)
		return error;
331

332 333 334 335
	/* An existing anon_vma has been reused, all done then. */
	if (vma->anon_vma)
		return 0;

336 337 338 339
	/* Then add our own anon_vma. */
	anon_vma = anon_vma_alloc();
	if (!anon_vma)
		goto out_error;
340
	avc = anon_vma_chain_alloc(GFP_KERNEL);
341 342
	if (!avc)
		goto out_error_free_anon_vma;
343 344 345 346 347 348

	/*
	 * 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;
349
	anon_vma->parent = pvma->anon_vma;
350
	/*
P
Peter Zijlstra 已提交
351 352 353
	 * 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.
354 355
	 */
	get_anon_vma(anon_vma->root);
356 357
	/* Mark this anon_vma as the one where our new (COWed) pages go. */
	vma->anon_vma = anon_vma;
358
	anon_vma_lock_write(anon_vma);
359
	anon_vma_chain_link(vma, avc, anon_vma);
360
	anon_vma->parent->degree++;
361
	anon_vma_unlock_write(anon_vma);
362 363 364 365

	return 0;

 out_error_free_anon_vma:
P
Peter Zijlstra 已提交
366
	put_anon_vma(anon_vma);
367
 out_error:
368
	unlink_anon_vmas(vma);
369
	return -ENOMEM;
L
Linus Torvalds 已提交
370 371
}

372 373 374
void unlink_anon_vmas(struct vm_area_struct *vma)
{
	struct anon_vma_chain *avc, *next;
375
	struct anon_vma *root = NULL;
376

377 378 379 380
	/*
	 * Unlink each anon_vma chained to the VMA.  This list is ordered
	 * from newest to oldest, ensuring the root anon_vma gets freed last.
	 */
381
	list_for_each_entry_safe(avc, next, &vma->anon_vma_chain, same_vma) {
382 383 384
		struct anon_vma *anon_vma = avc->anon_vma;

		root = lock_anon_vma_root(root, anon_vma);
385
		anon_vma_interval_tree_remove(avc, &anon_vma->rb_root);
386 387 388 389 390

		/*
		 * Leave empty anon_vmas on the list - we'll need
		 * to free them outside the lock.
		 */
391 392
		if (RB_EMPTY_ROOT(&anon_vma->rb_root)) {
			anon_vma->parent->degree--;
393
			continue;
394
		}
395 396 397 398

		list_del(&avc->same_vma);
		anon_vma_chain_free(avc);
	}
399 400
	if (vma->anon_vma)
		vma->anon_vma->degree--;
401 402 403 404 405
	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()
406
	 * needing to write-acquire the anon_vma->root->rwsem.
407 408 409 410
	 */
	list_for_each_entry_safe(avc, next, &vma->anon_vma_chain, same_vma) {
		struct anon_vma *anon_vma = avc->anon_vma;

411
		BUG_ON(anon_vma->degree);
412 413
		put_anon_vma(anon_vma);

414 415 416 417 418
		list_del(&avc->same_vma);
		anon_vma_chain_free(avc);
	}
}

419
static void anon_vma_ctor(void *data)
L
Linus Torvalds 已提交
420
{
C
Christoph Lameter 已提交
421
	struct anon_vma *anon_vma = data;
L
Linus Torvalds 已提交
422

423
	init_rwsem(&anon_vma->rwsem);
424
	atomic_set(&anon_vma->refcount, 0);
425
	anon_vma->rb_root = RB_ROOT;
L
Linus Torvalds 已提交
426 427 428 429 430
}

void __init anon_vma_init(void)
{
	anon_vma_cachep = kmem_cache_create("anon_vma", sizeof(struct anon_vma),
431
			0, SLAB_DESTROY_BY_RCU|SLAB_PANIC, anon_vma_ctor);
432
	anon_vma_chain_cachep = KMEM_CACHE(anon_vma_chain, SLAB_PANIC);
L
Linus Torvalds 已提交
433 434 435
}

/*
436 437 438 439 440 441 442 443 444
 * 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).
 *
445 446 447 448 449
 * 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 ].
 *
450 451 452 453 454 455 456
 * 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
457
 */
458
struct anon_vma *page_get_anon_vma(struct page *page)
L
Linus Torvalds 已提交
459
{
460
	struct anon_vma *anon_vma = NULL;
L
Linus Torvalds 已提交
461 462 463
	unsigned long anon_mapping;

	rcu_read_lock();
464
	anon_mapping = (unsigned long)READ_ONCE(page->mapping);
H
Hugh Dickins 已提交
465
	if ((anon_mapping & PAGE_MAPPING_FLAGS) != PAGE_MAPPING_ANON)
L
Linus Torvalds 已提交
466 467 468 469 470
		goto out;
	if (!page_mapped(page))
		goto out;

	anon_vma = (struct anon_vma *) (anon_mapping - PAGE_MAPPING_ANON);
471 472 473 474
	if (!atomic_inc_not_zero(&anon_vma->refcount)) {
		anon_vma = NULL;
		goto out;
	}
475 476 477

	/*
	 * If this page is still mapped, then its anon_vma cannot have been
478 479 480 481
	 * 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).
482
	 */
483
	if (!page_mapped(page)) {
484
		rcu_read_unlock();
485
		put_anon_vma(anon_vma);
486
		return NULL;
487
	}
L
Linus Torvalds 已提交
488 489
out:
	rcu_read_unlock();
490 491 492 493

	return anon_vma;
}

494 495 496 497 498 499 500
/*
 * 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.
 */
501
struct anon_vma *page_lock_anon_vma_read(struct page *page)
502
{
503
	struct anon_vma *anon_vma = NULL;
504
	struct anon_vma *root_anon_vma;
505
	unsigned long anon_mapping;
506

507
	rcu_read_lock();
508
	anon_mapping = (unsigned long)READ_ONCE(page->mapping);
509 510 511 512 513 514
	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);
515
	root_anon_vma = READ_ONCE(anon_vma->root);
516
	if (down_read_trylock(&root_anon_vma->rwsem)) {
517
		/*
518 519
		 * If the page is still mapped, then this anon_vma is still
		 * its anon_vma, and holding the mutex ensures that it will
520
		 * not go away, see anon_vma_free().
521
		 */
522
		if (!page_mapped(page)) {
523
			up_read(&root_anon_vma->rwsem);
524 525 526 527
			anon_vma = NULL;
		}
		goto out;
	}
528

529 530 531 532 533 534 535
	/* trylock failed, we got to sleep */
	if (!atomic_inc_not_zero(&anon_vma->refcount)) {
		anon_vma = NULL;
		goto out;
	}

	if (!page_mapped(page)) {
536
		rcu_read_unlock();
537
		put_anon_vma(anon_vma);
538
		return NULL;
539 540 541 542
	}

	/* we pinned the anon_vma, its safe to sleep */
	rcu_read_unlock();
543
	anon_vma_lock_read(anon_vma);
544 545 546 547 548

	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
549
		 * we'll deadlock on the anon_vma_lock_write() recursion.
550
		 */
551
		anon_vma_unlock_read(anon_vma);
552 553 554 555 556 557 558 559
		__put_anon_vma(anon_vma);
		anon_vma = NULL;
	}

	return anon_vma;

out:
	rcu_read_unlock();
560
	return anon_vma;
561 562
}

563
void page_unlock_anon_vma_read(struct anon_vma *anon_vma)
564
{
565
	anon_vma_unlock_read(anon_vma);
L
Linus Torvalds 已提交
566 567 568
}

/*
569
 * At what user virtual address is page expected in @vma?
L
Linus Torvalds 已提交
570
 */
571 572
static inline unsigned long
__vma_address(struct page *page, struct vm_area_struct *vma)
L
Linus Torvalds 已提交
573
{
574
	pgoff_t pgoff = page_to_pgoff(page);
575 576 577 578 579 580 581 582 583
	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 */
584
	VM_BUG_ON_VMA(address < vma->vm_start || address >= vma->vm_end, vma);
585

L
Linus Torvalds 已提交
586 587 588
	return address;
}

589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629
#ifdef CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH
static void percpu_flush_tlb_batch_pages(void *data)
{
	/*
	 * All TLB entries are flushed on the assumption that it is
	 * cheaper to flush all TLBs and let them be refilled than
	 * flushing individual PFNs. Note that we do not track mm's
	 * to flush as that might simply be multiple full TLB flushes
	 * for no gain.
	 */
	count_vm_tlb_event(NR_TLB_REMOTE_FLUSH_RECEIVED);
	flush_tlb_local();
}

/*
 * Flush TLB entries for recently unmapped pages from remote CPUs. It is
 * important if a PTE was dirty when it was unmapped that it's flushed
 * before any IO is initiated on the page to prevent lost writes. Similarly,
 * it must be flushed before freeing to prevent data leakage.
 */
void try_to_unmap_flush(void)
{
	struct tlbflush_unmap_batch *tlb_ubc = &current->tlb_ubc;
	int cpu;

	if (!tlb_ubc->flush_required)
		return;

	cpu = get_cpu();

	trace_tlb_flush(TLB_REMOTE_SHOOTDOWN, -1UL);

	if (cpumask_test_cpu(cpu, &tlb_ubc->cpumask))
		percpu_flush_tlb_batch_pages(&tlb_ubc->cpumask);

	if (cpumask_any_but(&tlb_ubc->cpumask, cpu) < nr_cpu_ids) {
		smp_call_function_many(&tlb_ubc->cpumask,
			percpu_flush_tlb_batch_pages, (void *)tlb_ubc, true);
	}
	cpumask_clear(&tlb_ubc->cpumask);
	tlb_ubc->flush_required = false;
630
	tlb_ubc->writable = false;
631 632 633
	put_cpu();
}

634 635 636 637 638 639 640 641 642
/* Flush iff there are potentially writable TLB entries that can race with IO */
void try_to_unmap_flush_dirty(void)
{
	struct tlbflush_unmap_batch *tlb_ubc = &current->tlb_ubc;

	if (tlb_ubc->writable)
		try_to_unmap_flush();
}

643
static void set_tlb_ubc_flush_pending(struct mm_struct *mm,
644
		struct page *page, bool writable)
645 646 647 648 649
{
	struct tlbflush_unmap_batch *tlb_ubc = &current->tlb_ubc;

	cpumask_or(&tlb_ubc->cpumask, &tlb_ubc->cpumask, mm_cpumask(mm));
	tlb_ubc->flush_required = true;
650 651 652 653 654 655 656 657

	/*
	 * If the PTE was dirty then it's best to assume it's writable. The
	 * caller must use try_to_unmap_flush_dirty() or try_to_unmap_flush()
	 * before the page is queued for IO.
	 */
	if (writable)
		tlb_ubc->writable = true;
658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679
}

/*
 * Returns true if the TLB flush should be deferred to the end of a batch of
 * unmap operations to reduce IPIs.
 */
static bool should_defer_flush(struct mm_struct *mm, enum ttu_flags flags)
{
	bool should_defer = false;

	if (!(flags & TTU_BATCH_FLUSH))
		return false;

	/* If remote CPUs need to be flushed then defer batch the flush */
	if (cpumask_any_but(mm_cpumask(mm), get_cpu()) < nr_cpu_ids)
		should_defer = true;
	put_cpu();

	return should_defer;
}
#else
static void set_tlb_ubc_flush_pending(struct mm_struct *mm,
680
		struct page *page, bool writable)
681 682 683 684 685 686 687 688 689
{
}

static bool should_defer_flush(struct mm_struct *mm, enum ttu_flags flags)
{
	return false;
}
#endif /* CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH */

L
Linus Torvalds 已提交
690
/*
H
Huang Shijie 已提交
691
 * At what user virtual address is page expected in vma?
692
 * Caller should check the page is actually part of the vma.
L
Linus Torvalds 已提交
693 694 695
 */
unsigned long page_address_in_vma(struct page *page, struct vm_area_struct *vma)
{
696
	unsigned long address;
697
	if (PageAnon(page)) {
698 699 700 701 702 703 704
		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)
705
			return -EFAULT;
706 707
	} else if (page->mapping) {
		if (!vma->vm_file || vma->vm_file->f_mapping != page->mapping)
L
Linus Torvalds 已提交
708 709 710
			return -EFAULT;
	} else
		return -EFAULT;
711 712 713 714
	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
715 716
}

B
Bob Liu 已提交
717 718 719 720 721
pmd_t *mm_find_pmd(struct mm_struct *mm, unsigned long address)
{
	pgd_t *pgd;
	pud_t *pud;
	pmd_t *pmd = NULL;
722
	pmd_t pmde;
B
Bob Liu 已提交
723 724 725 726 727 728 729 730 731 732

	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);
733
	/*
734
	 * Some THP functions use the sequence pmdp_huge_clear_flush(), set_pmd_at()
735 736 737
	 * without holding anon_vma lock for write.  So when looking for a
	 * genuine pmde (in which to find pte), test present and !THP together.
	 */
738 739
	pmde = *pmd;
	barrier();
740
	if (!pmd_present(pmde) || pmd_trans_huge(pmde))
B
Bob Liu 已提交
741 742 743 744 745
		pmd = NULL;
out:
	return pmd;
}

N
Nikita Danilov 已提交
746 747 748
/*
 * Check that @page is mapped at @address into @mm.
 *
N
Nick Piggin 已提交
749 750 751 752
 * 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).
 *
753
 * On success returns with pte mapped and locked.
N
Nikita Danilov 已提交
754
 */
755
pte_t *__page_check_address(struct page *page, struct mm_struct *mm,
N
Nick Piggin 已提交
756
			  unsigned long address, spinlock_t **ptlp, int sync)
N
Nikita Danilov 已提交
757 758 759
{
	pmd_t *pmd;
	pte_t *pte;
H
Hugh Dickins 已提交
760
	spinlock_t *ptl;
N
Nikita Danilov 已提交
761

762
	if (unlikely(PageHuge(page))) {
763
		/* when pud is not present, pte will be NULL */
764
		pte = huge_pte_offset(mm, address);
765 766 767
		if (!pte)
			return NULL;

768
		ptl = huge_pte_lockptr(page_hstate(page), mm, pte);
769 770 771
		goto check;
	}

B
Bob Liu 已提交
772 773
	pmd = mm_find_pmd(mm, address);
	if (!pmd)
H
Hugh Dickins 已提交
774 775 776 777
		return NULL;

	pte = pte_offset_map(pmd, address);
	/* Make a quick check before getting the lock */
N
Nick Piggin 已提交
778
	if (!sync && !pte_present(*pte)) {
H
Hugh Dickins 已提交
779 780 781 782
		pte_unmap(pte);
		return NULL;
	}

H
Hugh Dickins 已提交
783
	ptl = pte_lockptr(mm, pmd);
784
check:
H
Hugh Dickins 已提交
785 786 787 788
	spin_lock(ptl);
	if (pte_present(*pte) && page_to_pfn(page) == pte_pfn(*pte)) {
		*ptlp = ptl;
		return pte;
N
Nikita Danilov 已提交
789
	}
H
Hugh Dickins 已提交
790 791
	pte_unmap_unlock(pte, ptl);
	return NULL;
N
Nikita Danilov 已提交
792 793
}

N
Nick Piggin 已提交
794 795 796 797 798 799 800 801 802
/**
 * 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.
 */
803
int page_mapped_in_vma(struct page *page, struct vm_area_struct *vma)
N
Nick Piggin 已提交
804 805 806 807 808
{
	unsigned long address;
	pte_t *pte;
	spinlock_t *ptl;

809 810
	address = __vma_address(page, vma);
	if (unlikely(address < vma->vm_start || address >= vma->vm_end))
N
Nick Piggin 已提交
811 812 813 814 815 816 817 818 819
		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;
}

820 821 822 823 824 825
struct page_referenced_arg {
	int mapcount;
	int referenced;
	unsigned long vm_flags;
	struct mem_cgroup *memcg;
};
L
Linus Torvalds 已提交
826
/*
827
 * arg: page_referenced_arg will be passed
L
Linus Torvalds 已提交
828
 */
829
static int page_referenced_one(struct page *page, struct vm_area_struct *vma,
830
			unsigned long address, void *arg)
L
Linus Torvalds 已提交
831 832
{
	struct mm_struct *mm = vma->vm_mm;
833
	spinlock_t *ptl;
L
Linus Torvalds 已提交
834
	int referenced = 0;
835
	struct page_referenced_arg *pra = arg;
L
Linus Torvalds 已提交
836

837 838 839
	if (unlikely(PageTransHuge(page))) {
		pmd_t *pmd;

840 841 842 843
		/*
		 * rmap might return false positives; we must filter
		 * these out using page_check_address_pmd().
		 */
844
		pmd = page_check_address_pmd(page, mm, address,
845 846
					     PAGE_CHECK_ADDRESS_PMD_FLAG, &ptl);
		if (!pmd)
847
			return SWAP_AGAIN;
848 849

		if (vma->vm_flags & VM_LOCKED) {
850
			spin_unlock(ptl);
851 852
			pra->vm_flags |= VM_LOCKED;
			return SWAP_FAIL; /* To break the loop */
853 854 855 856
		}

		/* go ahead even if the pmd is pmd_trans_splitting() */
		if (pmdp_clear_flush_young_notify(vma, address, pmd))
857
			referenced++;
858
		spin_unlock(ptl);
859 860 861
	} else {
		pte_t *pte;

862 863 864 865
		/*
		 * rmap might return false positives; we must filter
		 * these out using page_check_address().
		 */
866 867
		pte = page_check_address(page, mm, address, &ptl, 0);
		if (!pte)
868
			return SWAP_AGAIN;
869

870 871
		if (vma->vm_flags & VM_LOCKED) {
			pte_unmap_unlock(pte, ptl);
872 873
			pra->vm_flags |= VM_LOCKED;
			return SWAP_FAIL; /* To break the loop */
874 875
		}

876 877 878 879 880 881 882 883
		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.
			 */
884
			if (likely(!(vma->vm_flags & VM_SEQ_READ)))
885 886 887 888 889
				referenced++;
		}
		pte_unmap_unlock(pte, ptl);
	}

890 891 892 893 894
	if (referenced)
		clear_page_idle(page);
	if (test_and_clear_page_young(page))
		referenced++;

895 896 897
	if (referenced) {
		pra->referenced++;
		pra->vm_flags |= vma->vm_flags;
L
Linus Torvalds 已提交
898
	}
899

900 901 902 903 904
	pra->mapcount--;
	if (!pra->mapcount)
		return SWAP_SUCCESS; /* To break the loop */

	return SWAP_AGAIN;
L
Linus Torvalds 已提交
905 906
}

907
static bool invalid_page_referenced_vma(struct vm_area_struct *vma, void *arg)
L
Linus Torvalds 已提交
908
{
909 910
	struct page_referenced_arg *pra = arg;
	struct mem_cgroup *memcg = pra->memcg;
L
Linus Torvalds 已提交
911

912 913
	if (!mm_match_cgroup(vma->vm_mm, memcg))
		return true;
L
Linus Torvalds 已提交
914

915
	return false;
L
Linus Torvalds 已提交
916 917 918 919 920 921
}

/**
 * page_referenced - test if the page was referenced
 * @page: the page to test
 * @is_locked: caller holds lock on the page
922
 * @memcg: target memory cgroup
923
 * @vm_flags: collect encountered vma->vm_flags who actually referenced the page
L
Linus Torvalds 已提交
924 925 926 927
 *
 * Quick test_and_clear_referenced for all mappings to a page,
 * returns the number of ptes which referenced the page.
 */
928 929
int page_referenced(struct page *page,
		    int is_locked,
930
		    struct mem_cgroup *memcg,
931
		    unsigned long *vm_flags)
L
Linus Torvalds 已提交
932
{
933
	int ret;
H
Hugh Dickins 已提交
934
	int we_locked = 0;
935 936 937 938 939 940 941 942 943
	struct page_referenced_arg pra = {
		.mapcount = page_mapcount(page),
		.memcg = memcg,
	};
	struct rmap_walk_control rwc = {
		.rmap_one = page_referenced_one,
		.arg = (void *)&pra,
		.anon_lock = page_lock_anon_vma_read,
	};
L
Linus Torvalds 已提交
944

945
	*vm_flags = 0;
946 947 948 949 950 951 952 953 954 955
	if (!page_mapped(page))
		return 0;

	if (!page_rmapping(page))
		return 0;

	if (!is_locked && (!PageAnon(page) || PageKsm(page))) {
		we_locked = trylock_page(page);
		if (!we_locked)
			return 1;
L
Linus Torvalds 已提交
956
	}
957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973

	/*
	 * If we are reclaiming on behalf of a cgroup, skip
	 * counting on behalf of references from different
	 * cgroups
	 */
	if (memcg) {
		rwc.invalid_vma = invalid_page_referenced_vma;
	}

	ret = rmap_walk(page, &rwc);
	*vm_flags = pra.vm_flags;

	if (we_locked)
		unlock_page(page);

	return pra.referenced;
L
Linus Torvalds 已提交
974 975
}

H
Hugh Dickins 已提交
976
static int page_mkclean_one(struct page *page, struct vm_area_struct *vma,
977
			    unsigned long address, void *arg)
978 979
{
	struct mm_struct *mm = vma->vm_mm;
980
	pte_t *pte;
981 982
	spinlock_t *ptl;
	int ret = 0;
983
	int *cleaned = arg;
984

N
Nick Piggin 已提交
985
	pte = page_check_address(page, mm, address, &ptl, 1);
986 987 988
	if (!pte)
		goto out;

989 990
	if (pte_dirty(*pte) || pte_write(*pte)) {
		pte_t entry;
991

992
		flush_cache_page(vma, address, pte_pfn(*pte));
993
		entry = ptep_clear_flush(vma, address, pte);
994 995
		entry = pte_wrprotect(entry);
		entry = pte_mkclean(entry);
996
		set_pte_at(mm, address, pte, entry);
997 998
		ret = 1;
	}
999 1000

	pte_unmap_unlock(pte, ptl);
1001

1002
	if (ret) {
1003
		mmu_notifier_invalidate_page(mm, address);
1004 1005
		(*cleaned)++;
	}
1006
out:
1007
	return SWAP_AGAIN;
1008 1009
}

1010
static bool invalid_mkclean_vma(struct vm_area_struct *vma, void *arg)
1011
{
1012
	if (vma->vm_flags & VM_SHARED)
F
Fengguang Wu 已提交
1013
		return false;
1014

F
Fengguang Wu 已提交
1015
	return true;
1016 1017 1018 1019
}

int page_mkclean(struct page *page)
{
1020 1021 1022 1023 1024 1025 1026
	int cleaned = 0;
	struct address_space *mapping;
	struct rmap_walk_control rwc = {
		.arg = (void *)&cleaned,
		.rmap_one = page_mkclean_one,
		.invalid_vma = invalid_mkclean_vma,
	};
1027 1028 1029

	BUG_ON(!PageLocked(page));

1030 1031 1032 1033 1034 1035 1036 1037
	if (!page_mapped(page))
		return 0;

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

	rmap_walk(page, &rwc);
1038

1039
	return cleaned;
1040
}
J
Jaya Kumar 已提交
1041
EXPORT_SYMBOL_GPL(page_mkclean);
1042

1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058
/**
 * 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;

1059
	VM_BUG_ON_PAGE(!PageLocked(page), page);
1060
	VM_BUG_ON_VMA(!anon_vma, vma);
1061
	VM_BUG_ON_PAGE(page->index != linear_page_index(vma, address), page);
1062 1063

	anon_vma = (void *) anon_vma + PAGE_MAPPING_ANON;
1064 1065 1066 1067 1068 1069
	/*
	 * Ensure that anon_vma and the PAGE_MAPPING_ANON bit are written
	 * simultaneously, so a concurrent reader (eg page_referenced()'s
	 * PageAnon()) will not see one without the other.
	 */
	WRITE_ONCE(page->mapping, (struct address_space *) anon_vma);
1070 1071
}

N
Nick Piggin 已提交
1072
/**
A
Andi Kleen 已提交
1073 1074 1075 1076
 * __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	
1077
 * @exclusive:	the page is exclusively owned by the current process
N
Nick Piggin 已提交
1078 1079
 */
static void __page_set_anon_rmap(struct page *page,
1080
	struct vm_area_struct *vma, unsigned long address, int exclusive)
N
Nick Piggin 已提交
1081
{
1082
	struct anon_vma *anon_vma = vma->anon_vma;
1083

1084
	BUG_ON(!anon_vma);
1085

A
Andi Kleen 已提交
1086 1087 1088
	if (PageAnon(page))
		return;

1089
	/*
1090 1091 1092
	 * If the page isn't exclusively mapped into this vma,
	 * we must use the _oldest_ possible anon_vma for the
	 * page mapping!
1093
	 */
A
Andi Kleen 已提交
1094
	if (!exclusive)
1095
		anon_vma = anon_vma->root;
N
Nick Piggin 已提交
1096 1097 1098 1099 1100 1101

	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
1102
/**
R
Randy Dunlap 已提交
1103
 * __page_check_anon_rmap - sanity check anonymous rmap addition
N
Nick Piggin 已提交
1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123
 * @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
1124
	BUG_ON(page_anon_vma(page)->root != vma->anon_vma->root);
N
Nick Piggin 已提交
1125 1126 1127 1128
	BUG_ON(page->index != linear_page_index(vma, address));
#endif
}

L
Linus Torvalds 已提交
1129 1130 1131 1132 1133 1134
/**
 * 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
1135
 * The caller needs to hold the pte lock, and the page must be locked in
1136 1137 1138
 * 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
1139 1140 1141
 */
void page_add_anon_rmap(struct page *page,
	struct vm_area_struct *vma, unsigned long address)
1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152
{
	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
1153
{
H
Hugh Dickins 已提交
1154
	int first = atomic_inc_and_test(&page->_mapcount);
1155
	if (first) {
1156 1157 1158 1159 1160 1161
		/*
		 * We use the irq-unsafe __{inc|mod}_zone_page_stat because
		 * these counters are not modified in interrupt context, and
		 * pte lock(a spinlock) is held, which implies preemption
		 * disabled.
		 */
1162
		if (PageTransHuge(page))
1163 1164
			__inc_zone_page_state(page,
					      NR_ANON_TRANSPARENT_HUGEPAGES);
1165 1166
		__mod_zone_page_state(page_zone(page), NR_ANON_PAGES,
				hpage_nr_pages(page));
1167
	}
H
Hugh Dickins 已提交
1168 1169 1170
	if (unlikely(PageKsm(page)))
		return;

1171
	VM_BUG_ON_PAGE(!PageLocked(page), page);
1172
	/* address might be in next vma when migration races vma_adjust */
H
Hugh Dickins 已提交
1173
	if (first)
1174
		__page_set_anon_rmap(page, vma, address, exclusive);
1175
	else
N
Nick Piggin 已提交
1176
		__page_check_anon_rmap(page, vma, address);
L
Linus Torvalds 已提交
1177 1178
}

R
Randy Dunlap 已提交
1179
/**
N
Nick Piggin 已提交
1180 1181 1182 1183 1184 1185 1186
 * 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
1187
 * Page does not have to be locked.
N
Nick Piggin 已提交
1188 1189 1190 1191
 */
void page_add_new_anon_rmap(struct page *page,
	struct vm_area_struct *vma, unsigned long address)
{
1192
	VM_BUG_ON_VMA(address < vma->vm_start || address >= vma->vm_end, vma);
1193 1194
	SetPageSwapBacked(page);
	atomic_set(&page->_mapcount, 0); /* increment count (starts at -1) */
1195
	if (PageTransHuge(page))
1196
		__inc_zone_page_state(page, NR_ANON_TRANSPARENT_HUGEPAGES);
1197 1198
	__mod_zone_page_state(page_zone(page), NR_ANON_PAGES,
			hpage_nr_pages(page));
1199
	__page_set_anon_rmap(page, vma, address, 1);
N
Nick Piggin 已提交
1200 1201
}

L
Linus Torvalds 已提交
1202 1203 1204 1205
/**
 * page_add_file_rmap - add pte mapping to a file page
 * @page: the page to add the mapping to
 *
1206
 * The caller needs to hold the pte lock.
L
Linus Torvalds 已提交
1207 1208 1209
 */
void page_add_file_rmap(struct page *page)
{
1210
	struct mem_cgroup *memcg;
1211

1212
	memcg = mem_cgroup_begin_page_stat(page);
1213
	if (atomic_inc_and_test(&page->_mapcount)) {
1214
		__inc_zone_page_state(page, NR_FILE_MAPPED);
1215
		mem_cgroup_inc_page_stat(memcg, MEM_CGROUP_STAT_FILE_MAPPED);
1216
	}
1217
	mem_cgroup_end_page_stat(memcg);
L
Linus Torvalds 已提交
1218 1219
}

1220 1221 1222 1223
static void page_remove_file_rmap(struct page *page)
{
	struct mem_cgroup *memcg;

1224
	memcg = mem_cgroup_begin_page_stat(page);
1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244

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

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

	/*
	 * We use the irq-unsafe __{inc|mod}_zone_page_stat because
	 * these counters are not modified in interrupt context, and
	 * pte lock(a spinlock) is held, which implies preemption disabled.
	 */
	__dec_zone_page_state(page, NR_FILE_MAPPED);
	mem_cgroup_dec_page_stat(memcg, MEM_CGROUP_STAT_FILE_MAPPED);

	if (unlikely(PageMlocked(page)))
		clear_page_mlock(page);
out:
1245
	mem_cgroup_end_page_stat(memcg);
1246 1247
}

L
Linus Torvalds 已提交
1248 1249 1250 1251
/**
 * page_remove_rmap - take down pte mapping from a page
 * @page: page to remove mapping from
 *
1252
 * The caller needs to hold the pte lock.
L
Linus Torvalds 已提交
1253
 */
1254
void page_remove_rmap(struct page *page)
L
Linus Torvalds 已提交
1255
{
1256 1257 1258 1259
	if (!PageAnon(page)) {
		page_remove_file_rmap(page);
		return;
	}
1260

1261 1262
	/* page still mapped by someone else? */
	if (!atomic_add_negative(-1, &page->_mapcount))
1263 1264 1265 1266 1267
		return;

	/* Hugepages are not counted in NR_ANON_PAGES for now. */
	if (unlikely(PageHuge(page)))
		return;
1268

1269
	/*
1270 1271 1272
	 * We use the irq-unsafe __{inc|mod}_zone_page_stat because
	 * these counters are not modified in interrupt context, and
	 * pte lock(a spinlock) is held, which implies preemption disabled.
1273
	 */
1274 1275 1276 1277 1278 1279
	if (PageTransHuge(page))
		__dec_zone_page_state(page, NR_ANON_TRANSPARENT_HUGEPAGES);

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

1280 1281
	if (unlikely(PageMlocked(page)))
		clear_page_mlock(page);
1282

1283 1284 1285 1286 1287 1288 1289 1290 1291
	/*
	 * 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
1292 1293 1294
}

/*
1295
 * @arg: enum ttu_flags will be passed to this argument
L
Linus Torvalds 已提交
1296
 */
1297
static int try_to_unmap_one(struct page *page, struct vm_area_struct *vma,
1298
		     unsigned long address, void *arg)
L
Linus Torvalds 已提交
1299 1300 1301 1302
{
	struct mm_struct *mm = vma->vm_mm;
	pte_t *pte;
	pte_t pteval;
H
Hugh Dickins 已提交
1303
	spinlock_t *ptl;
L
Linus Torvalds 已提交
1304
	int ret = SWAP_AGAIN;
1305
	enum ttu_flags flags = (enum ttu_flags)arg;
L
Linus Torvalds 已提交
1306

1307 1308 1309 1310
	/* munlock has nothing to gain from examining un-locked vmas */
	if ((flags & TTU_MUNLOCK) && !(vma->vm_flags & VM_LOCKED))
		goto out;

N
Nick Piggin 已提交
1311
	pte = page_check_address(page, mm, address, &ptl, 0);
H
Hugh Dickins 已提交
1312
	if (!pte)
N
Nikita Danilov 已提交
1313
		goto out;
L
Linus Torvalds 已提交
1314 1315 1316 1317 1318 1319

	/*
	 * 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.
	 */
1320
	if (!(flags & TTU_IGNORE_MLOCK)) {
1321 1322 1323 1324 1325 1326
		if (vma->vm_flags & VM_LOCKED) {
			/* Holding pte lock, we do *not* need mmap_sem here */
			mlock_vma_page(page);
			ret = SWAP_MLOCK;
			goto out_unmap;
		}
1327
		if (flags & TTU_MUNLOCK)
H
Hugh Dickins 已提交
1328
			goto out_unmap;
1329 1330
	}
	if (!(flags & TTU_IGNORE_ACCESS)) {
N
Nick Piggin 已提交
1331 1332 1333 1334 1335
		if (ptep_clear_flush_young_notify(vma, address, pte)) {
			ret = SWAP_FAIL;
			goto out_unmap;
		}
  	}
L
Linus Torvalds 已提交
1336 1337 1338

	/* Nuke the page table entry. */
	flush_cache_page(vma, address, page_to_pfn(page));
1339 1340 1341 1342 1343 1344 1345 1346 1347 1348
	if (should_defer_flush(mm, flags)) {
		/*
		 * We clear the PTE but do not flush so potentially a remote
		 * CPU could still be writing to the page. If the entry was
		 * previously clean then the architecture must guarantee that
		 * a clear->dirty transition on a cached TLB entry is written
		 * through and traps if the PTE is unmapped.
		 */
		pteval = ptep_get_and_clear(mm, address, pte);

1349
		set_tlb_ubc_flush_pending(mm, page, pte_dirty(pteval));
1350 1351 1352
	} else {
		pteval = ptep_clear_flush(vma, address, pte);
	}
L
Linus Torvalds 已提交
1353 1354 1355 1356 1357

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

1358 1359 1360
	/* Update high watermark before we lower rss */
	update_hiwater_rss(mm);

1361
	if (PageHWPoison(page) && !(flags & TTU_IGNORE_HWPOISON)) {
1362 1363 1364
		if (PageHuge(page)) {
			hugetlb_count_sub(1 << compound_order(page), mm);
		} else {
1365 1366 1367 1368 1369
			if (PageAnon(page))
				dec_mm_counter(mm, MM_ANONPAGES);
			else
				dec_mm_counter(mm, MM_FILEPAGES);
		}
1370
		set_pte_at(mm, address, pte,
1371
			   swp_entry_to_pte(make_hwpoison_entry(page)));
1372 1373 1374 1375 1376 1377 1378 1379 1380 1381
	} else if (pte_unused(pteval)) {
		/*
		 * The guest indicated that the page content is of no
		 * interest anymore. Simply discard the pte, vmscan
		 * will take care of the rest.
		 */
		if (PageAnon(page))
			dec_mm_counter(mm, MM_ANONPAGES);
		else
			dec_mm_counter(mm, MM_FILEPAGES);
1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394
	} else if (IS_ENABLED(CONFIG_MIGRATION) && (flags & TTU_MIGRATION)) {
		swp_entry_t entry;
		pte_t swp_pte;
		/*
		 * 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.
		 */
		entry = make_migration_entry(page, pte_write(pteval));
		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);
1395
	} else if (PageAnon(page)) {
H
Hugh Dickins 已提交
1396
		swp_entry_t entry = { .val = page_private(page) };
1397
		pte_t swp_pte;
1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412
		/*
		 * Store the swap location in the pte.
		 * See handle_pte_fault() ...
		 */
		VM_BUG_ON_PAGE(!PageSwapCache(page), page);
		if (swap_duplicate(entry) < 0) {
			set_pte_at(mm, address, pte, pteval);
			ret = SWAP_FAIL;
			goto out_unmap;
		}
		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);
L
Linus Torvalds 已提交
1413
		}
1414 1415
		dec_mm_counter(mm, MM_ANONPAGES);
		inc_mm_counter(mm, MM_SWAPENTS);
1416 1417 1418 1419
		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);
1420
	} else
K
KAMEZAWA Hiroyuki 已提交
1421
		dec_mm_counter(mm, MM_FILEPAGES);
L
Linus Torvalds 已提交
1422

1423
	page_remove_rmap(page);
L
Linus Torvalds 已提交
1424 1425 1426
	page_cache_release(page);

out_unmap:
H
Hugh Dickins 已提交
1427
	pte_unmap_unlock(pte, ptl);
1428
	if (ret != SWAP_FAIL && ret != SWAP_MLOCK && !(flags & TTU_MUNLOCK))
1429
		mmu_notifier_invalidate_page(mm, address);
K
KOSAKI Motohiro 已提交
1430 1431
out:
	return ret;
L
Linus Torvalds 已提交
1432 1433
}

1434
bool is_vma_temporary_stack(struct vm_area_struct *vma)
1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447
{
	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;
}

1448 1449 1450 1451 1452 1453 1454 1455 1456 1457
static bool invalid_migration_vma(struct vm_area_struct *vma, void *arg)
{
	return is_vma_temporary_stack(vma);
}

static int page_not_mapped(struct page *page)
{
	return !page_mapped(page);
};

L
Linus Torvalds 已提交
1458 1459 1460
/**
 * try_to_unmap - try to remove all page table mappings to a page
 * @page: the page to get unmapped
1461
 * @flags: action and flags
L
Linus Torvalds 已提交
1462 1463 1464 1465 1466 1467 1468 1469
 *
 * 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
1470
 * SWAP_MLOCK	- page is mlocked.
L
Linus Torvalds 已提交
1471
 */
1472
int try_to_unmap(struct page *page, enum ttu_flags flags)
L
Linus Torvalds 已提交
1473 1474
{
	int ret;
1475 1476 1477 1478 1479 1480
	struct rmap_walk_control rwc = {
		.rmap_one = try_to_unmap_one,
		.arg = (void *)flags,
		.done = page_not_mapped,
		.anon_lock = page_lock_anon_vma_read,
	};
L
Linus Torvalds 已提交
1481

1482
	VM_BUG_ON_PAGE(!PageHuge(page) && PageTransHuge(page), page);
L
Linus Torvalds 已提交
1483

1484 1485 1486 1487 1488 1489 1490 1491
	/*
	 * 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.
	 */
1492
	if ((flags & TTU_MIGRATION) && !PageKsm(page) && PageAnon(page))
1493 1494 1495 1496
		rwc.invalid_vma = invalid_migration_vma;

	ret = rmap_walk(page, &rwc);

N
Nick Piggin 已提交
1497
	if (ret != SWAP_MLOCK && !page_mapped(page))
L
Linus Torvalds 已提交
1498 1499 1500
		ret = SWAP_SUCCESS;
	return ret;
}
N
Nikita Danilov 已提交
1501

N
Nick Piggin 已提交
1502 1503 1504 1505 1506 1507 1508 1509 1510 1511
/**
 * 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
1512
 * SWAP_AGAIN	- no vma is holding page mlocked, or,
N
Nick Piggin 已提交
1513
 * SWAP_AGAIN	- page mapped in mlocked vma -- couldn't acquire mmap sem
H
Hugh Dickins 已提交
1514
 * SWAP_FAIL	- page cannot be located at present
N
Nick Piggin 已提交
1515 1516 1517 1518
 * SWAP_MLOCK	- page is now mlocked.
 */
int try_to_munlock(struct page *page)
{
1519 1520 1521 1522 1523 1524 1525 1526 1527
	int ret;
	struct rmap_walk_control rwc = {
		.rmap_one = try_to_unmap_one,
		.arg = (void *)TTU_MUNLOCK,
		.done = page_not_mapped,
		.anon_lock = page_lock_anon_vma_read,

	};

1528
	VM_BUG_ON_PAGE(!PageLocked(page) || PageLRU(page), page);
N
Nick Piggin 已提交
1529

1530 1531
	ret = rmap_walk(page, &rwc);
	return ret;
N
Nick Piggin 已提交
1532
}
1533

P
Peter Zijlstra 已提交
1534
void __put_anon_vma(struct anon_vma *anon_vma)
1535
{
P
Peter Zijlstra 已提交
1536
	struct anon_vma *root = anon_vma->root;
1537

1538
	anon_vma_free(anon_vma);
P
Peter Zijlstra 已提交
1539 1540
	if (root != anon_vma && atomic_dec_and_test(&root->refcount))
		anon_vma_free(root);
1541 1542
}

1543 1544
static struct anon_vma *rmap_walk_anon_lock(struct page *page,
					struct rmap_walk_control *rwc)
1545 1546 1547
{
	struct anon_vma *anon_vma;

1548 1549 1550
	if (rwc->anon_lock)
		return rwc->anon_lock(page);

1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564
	/*
	 * Note: remove_migration_ptes() cannot use page_lock_anon_vma_read()
	 * because that depends on page_mapped(); but not all its usages
	 * are holding mmap_sem. Users without mmap_sem are required to
	 * take a reference count to prevent the anon_vma disappearing
	 */
	anon_vma = page_anon_vma(page);
	if (!anon_vma)
		return NULL;

	anon_vma_lock_read(anon_vma);
	return anon_vma;
}

1565
/*
1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577
 * rmap_walk_anon - do something to anonymous page using the object-based
 * rmap method
 * @page: the page to be handled
 * @rwc: control variable according to each walk type
 *
 * Find all the mappings of a page using the mapping pointer and the vma chains
 * contained in the anon_vma struct it points to.
 *
 * When called from try_to_munlock(), the mmap_sem of the mm containing the vma
 * where the page was found will be held for write.  So, we won't recheck
 * vm_flags for that VMA.  That should be OK, because that vma shouldn't be
 * LOCKED.
1578
 */
1579
static int rmap_walk_anon(struct page *page, struct rmap_walk_control *rwc)
1580 1581
{
	struct anon_vma *anon_vma;
1582
	pgoff_t pgoff;
1583
	struct anon_vma_chain *avc;
1584 1585
	int ret = SWAP_AGAIN;

1586
	anon_vma = rmap_walk_anon_lock(page, rwc);
1587 1588
	if (!anon_vma)
		return ret;
1589

1590
	pgoff = page_to_pgoff(page);
1591
	anon_vma_interval_tree_foreach(avc, &anon_vma->rb_root, pgoff, pgoff) {
1592
		struct vm_area_struct *vma = avc->vma;
1593
		unsigned long address = vma_address(page, vma);
1594

1595 1596
		cond_resched();

1597 1598 1599
		if (rwc->invalid_vma && rwc->invalid_vma(vma, rwc->arg))
			continue;

1600
		ret = rwc->rmap_one(page, vma, address, rwc->arg);
1601 1602
		if (ret != SWAP_AGAIN)
			break;
1603 1604
		if (rwc->done && rwc->done(page))
			break;
1605
	}
1606
	anon_vma_unlock_read(anon_vma);
1607 1608 1609
	return ret;
}

1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622
/*
 * rmap_walk_file - do something to file page using the object-based rmap method
 * @page: the page to be handled
 * @rwc: control variable according to each walk type
 *
 * Find all the mappings of a page using the mapping pointer and the vma chains
 * contained in the address_space struct it points to.
 *
 * When called from try_to_munlock(), the mmap_sem of the mm containing the vma
 * where the page was found will be held for write.  So, we won't recheck
 * vm_flags for that VMA.  That should be OK, because that vma shouldn't be
 * LOCKED.
 */
1623
static int rmap_walk_file(struct page *page, struct rmap_walk_control *rwc)
1624 1625
{
	struct address_space *mapping = page->mapping;
1626
	pgoff_t pgoff;
1627 1628 1629
	struct vm_area_struct *vma;
	int ret = SWAP_AGAIN;

1630 1631 1632 1633
	/*
	 * 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,
1634
	 * so we can safely take mapping->i_mmap_rwsem.
1635
	 */
1636
	VM_BUG_ON_PAGE(!PageLocked(page), page);
1637

1638 1639
	if (!mapping)
		return ret;
D
Davidlohr Bueso 已提交
1640

1641
	pgoff = page_to_pgoff(page);
D
Davidlohr Bueso 已提交
1642
	i_mmap_lock_read(mapping);
1643
	vma_interval_tree_foreach(vma, &mapping->i_mmap, pgoff, pgoff) {
1644
		unsigned long address = vma_address(page, vma);
1645

1646 1647
		cond_resched();

1648 1649 1650
		if (rwc->invalid_vma && rwc->invalid_vma(vma, rwc->arg))
			continue;

1651
		ret = rwc->rmap_one(page, vma, address, rwc->arg);
1652
		if (ret != SWAP_AGAIN)
1653 1654 1655
			goto done;
		if (rwc->done && rwc->done(page))
			goto done;
1656
	}
1657 1658

done:
D
Davidlohr Bueso 已提交
1659
	i_mmap_unlock_read(mapping);
1660 1661 1662
	return ret;
}

1663
int rmap_walk(struct page *page, struct rmap_walk_control *rwc)
1664 1665
{
	if (unlikely(PageKsm(page)))
1666
		return rmap_walk_ksm(page, rwc);
1667
	else if (PageAnon(page))
1668
		return rmap_walk_anon(page, rwc);
1669
	else
1670
		return rmap_walk_file(page, rwc);
1671
}
1672

N
Naoya Horiguchi 已提交
1673
#ifdef CONFIG_HUGETLB_PAGE
1674 1675 1676 1677 1678 1679 1680 1681 1682
/*
 * 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;
1683

1684
	BUG_ON(!anon_vma);
1685 1686 1687 1688 1689 1690

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

1691 1692 1693 1694 1695 1696 1697 1698 1699 1700
	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;
1701 1702

	BUG_ON(!PageLocked(page));
1703
	BUG_ON(!anon_vma);
1704
	/* address might be in next vma when migration races vma_adjust */
1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716
	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
1717
#endif /* CONFIG_HUGETLB_PAGE */