rmap.c 52.8 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 27 28 29
 *       hugetlbfs_i_mmap_rwsem_key (in huge_pmd_share)
 *         mapping->i_mmap_rwsem
 *           anon_vma->rwsem
 *             mm->page_table_lock or pte_lock
30
 *               zone_lru_lock (in mark_page_accessed, isolate_lru_page)
31 32 33 34 35 36 37 38 39 40 41
 *               swap_lock (in swap_duplicate, swap_info_get)
 *                 mmlist_lock (in mmput, drain_mmlist and others)
 *                 mapping->private_lock (in __set_page_dirty_buffers)
 *                   mem_cgroup_{begin,end}_page_stat (memcg->move_lock)
 *                     mapping->tree_lock (widely used)
 *                 inode->i_lock (in set_page_dirty's __mark_inode_dirty)
 *                 bdi.wb->list_lock (in set_page_dirty's __mark_inode_dirty)
 *                   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,
 *                             within bdi.wb->list_lock in __sync_single_inode)
42
 *
43
 * anon_vma->rwsem,mapping->i_mutex      (memory_failure, collect_procs_anon)
44
 *   ->tasklist_lock
45
 *     pte map lock
L
Linus Torvalds 已提交
46 47 48
 */

#include <linux/mm.h>
49
#include <linux/sched/mm.h>
50
#include <linux/sched/task.h>
L
Linus Torvalds 已提交
51 52 53 54 55
#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
56
#include <linux/ksm.h>
L
Linus Torvalds 已提交
57 58
#include <linux/rmap.h>
#include <linux/rcupdate.h>
59
#include <linux/export.h>
60
#include <linux/memcontrol.h>
A
Andrea Arcangeli 已提交
61
#include <linux/mmu_notifier.h>
62
#include <linux/migrate.h>
63
#include <linux/hugetlb.h>
64
#include <linux/backing-dev.h>
65
#include <linux/page_idle.h>
66
#include <linux/memremap.h>
L
Linus Torvalds 已提交
67 68 69

#include <asm/tlbflush.h>

70 71
#include <trace/events/tlb.h>

N
Nick Piggin 已提交
72 73
#include "internal.h"

74
static struct kmem_cache *anon_vma_cachep;
75
static struct kmem_cache *anon_vma_chain_cachep;
76 77 78

static inline struct anon_vma *anon_vma_alloc(void)
{
P
Peter Zijlstra 已提交
79 80 81 82 83
	struct anon_vma *anon_vma;

	anon_vma = kmem_cache_alloc(anon_vma_cachep, GFP_KERNEL);
	if (anon_vma) {
		atomic_set(&anon_vma->refcount, 1);
84 85
		anon_vma->degree = 1;	/* Reference for first vma */
		anon_vma->parent = anon_vma;
P
Peter Zijlstra 已提交
86 87 88 89 90 91 92 93
		/*
		 * 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;
94 95
}

P
Peter Zijlstra 已提交
96
static inline void anon_vma_free(struct anon_vma *anon_vma)
97
{
P
Peter Zijlstra 已提交
98
	VM_BUG_ON(atomic_read(&anon_vma->refcount));
99 100

	/*
101
	 * Synchronize against page_lock_anon_vma_read() such that
102 103 104 105 106
	 * 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
107
	 * down_read_trylock() from page_lock_anon_vma_read(). This orders:
108
	 *
109 110
	 * page_lock_anon_vma_read()	VS	put_anon_vma()
	 *   down_read_trylock()		  atomic_dec_and_test()
111
	 *   LOCK				  MB
112
	 *   atomic_read()			  rwsem_is_locked()
113 114 115 116
	 *
	 * LOCK should suffice since the actual taking of the lock must
	 * happen _before_ what follows.
	 */
117
	might_sleep();
118
	if (rwsem_is_locked(&anon_vma->root->rwsem)) {
119
		anon_vma_lock_write(anon_vma);
120
		anon_vma_unlock_write(anon_vma);
121 122
	}

123 124
	kmem_cache_free(anon_vma_cachep, anon_vma);
}
L
Linus Torvalds 已提交
125

126
static inline struct anon_vma_chain *anon_vma_chain_alloc(gfp_t gfp)
127
{
128
	return kmem_cache_alloc(anon_vma_chain_cachep, gfp);
129 130
}

131
static void anon_vma_chain_free(struct anon_vma_chain *anon_vma_chain)
132 133 134 135
{
	kmem_cache_free(anon_vma_chain_cachep, anon_vma_chain);
}

136 137 138 139 140 141 142
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);
143
	anon_vma_interval_tree_insert(avc, &anon_vma->rb_root);
144 145
}

146
/**
147
 * __anon_vma_prepare - attach an anon_vma to a memory region
148 149 150 151 152 153
 * @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.
 *
154 155
 * The common case will be that we already have one, which
 * is handled inline by anon_vma_prepare(). But if
F
Figo.zhang 已提交
156
 * not we either need to find an adjacent mapping that we
157 158 159 160 161
 * 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
162
 * optimistically looked up an anon_vma in page_lock_anon_vma_read()
163 164 165 166 167 168 169 170 171 172 173
 * 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.
 */
174
int __anon_vma_prepare(struct vm_area_struct *vma)
L
Linus Torvalds 已提交
175
{
176 177
	struct mm_struct *mm = vma->vm_mm;
	struct anon_vma *anon_vma, *allocated;
178
	struct anon_vma_chain *avc;
L
Linus Torvalds 已提交
179 180 181

	might_sleep();

182 183 184 185 186 187 188 189 190 191 192 193
	avc = anon_vma_chain_alloc(GFP_KERNEL);
	if (!avc)
		goto out_enomem;

	anon_vma = find_mergeable_anon_vma(vma);
	allocated = NULL;
	if (!anon_vma) {
		anon_vma = anon_vma_alloc();
		if (unlikely(!anon_vma))
			goto out_enomem_free_avc;
		allocated = anon_vma;
	}
194

195 196 197 198 199 200 201 202
	anon_vma_lock_write(anon_vma);
	/* page_table_lock to protect against threads */
	spin_lock(&mm->page_table_lock);
	if (likely(!vma->anon_vma)) {
		vma->anon_vma = anon_vma;
		anon_vma_chain_link(vma, avc, anon_vma);
		/* vma reference or self-parent link for new root */
		anon_vma->degree++;
203
		allocated = NULL;
204 205 206 207
		avc = NULL;
	}
	spin_unlock(&mm->page_table_lock);
	anon_vma_unlock_write(anon_vma);
L
Linus Torvalds 已提交
208

209 210 211 212
	if (unlikely(allocated))
		put_anon_vma(allocated);
	if (unlikely(avc))
		anon_vma_chain_free(avc);
213

L
Linus Torvalds 已提交
214
	return 0;
215 216 217 218 219

 out_enomem_free_avc:
	anon_vma_chain_free(avc);
 out_enomem:
	return -ENOMEM;
L
Linus Torvalds 已提交
220 221
}

222 223 224 225 226 227 228 229 230 231 232 233 234
/*
 * 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))
235
			up_write(&root->rwsem);
236
		root = new_root;
237
		down_write(&root->rwsem);
238 239 240 241 242 243 244
	}
	return root;
}

static inline void unlock_anon_vma_root(struct anon_vma *root)
{
	if (root)
245
		up_write(&root->rwsem);
246 247
}

248 249 250
/*
 * Attach the anon_vmas from src to dst.
 * Returns 0 on success, -ENOMEM on failure.
251 252 253 254 255 256 257 258
 *
 * 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.
259 260
 */
int anon_vma_clone(struct vm_area_struct *dst, struct vm_area_struct *src)
L
Linus Torvalds 已提交
261
{
262
	struct anon_vma_chain *avc, *pavc;
263
	struct anon_vma *root = NULL;
264

265
	list_for_each_entry_reverse(pavc, &src->anon_vma_chain, same_vma) {
266 267
		struct anon_vma *anon_vma;

268 269 270 271 272 273 274 275
		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;
		}
276 277 278
		anon_vma = pavc->anon_vma;
		root = lock_anon_vma_root(root, anon_vma);
		anon_vma_chain_link(dst, avc, anon_vma);
279 280 281 282 283 284 285 286 287 288 289 290

		/*
		 * 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;
291
	}
292 293
	if (dst->anon_vma)
		dst->anon_vma->degree++;
294
	unlock_anon_vma_root(root);
295
	return 0;
L
Linus Torvalds 已提交
296

297
 enomem_failure:
298 299 300 301 302 303 304
	/*
	 * 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;
305 306
	unlink_anon_vmas(dst);
	return -ENOMEM;
L
Linus Torvalds 已提交
307 308
}

309 310 311 312 313 314
/*
 * 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
315
{
316 317
	struct anon_vma_chain *avc;
	struct anon_vma *anon_vma;
318
	int error;
L
Linus Torvalds 已提交
319

320 321 322 323
	/* Don't bother if the parent process has no anon_vma here. */
	if (!pvma->anon_vma)
		return 0;

324 325 326
	/* Drop inherited anon_vma, we'll reuse existing or allocate new. */
	vma->anon_vma = NULL;

327 328 329 330
	/*
	 * First, attach the new VMA to the parent VMA's anon_vmas,
	 * so rmap can find non-COWed pages in child processes.
	 */
331 332 333
	error = anon_vma_clone(vma, pvma);
	if (error)
		return error;
334

335 336 337 338
	/* An existing anon_vma has been reused, all done then. */
	if (vma->anon_vma)
		return 0;

339 340 341 342
	/* Then add our own anon_vma. */
	anon_vma = anon_vma_alloc();
	if (!anon_vma)
		goto out_error;
343
	avc = anon_vma_chain_alloc(GFP_KERNEL);
344 345
	if (!avc)
		goto out_error_free_anon_vma;
346 347 348 349 350 351

	/*
	 * 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;
352
	anon_vma->parent = pvma->anon_vma;
353
	/*
P
Peter Zijlstra 已提交
354 355 356
	 * 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.
357 358
	 */
	get_anon_vma(anon_vma->root);
359 360
	/* Mark this anon_vma as the one where our new (COWed) pages go. */
	vma->anon_vma = anon_vma;
361
	anon_vma_lock_write(anon_vma);
362
	anon_vma_chain_link(vma, avc, anon_vma);
363
	anon_vma->parent->degree++;
364
	anon_vma_unlock_write(anon_vma);
365 366 367 368

	return 0;

 out_error_free_anon_vma:
P
Peter Zijlstra 已提交
369
	put_anon_vma(anon_vma);
370
 out_error:
371
	unlink_anon_vmas(vma);
372
	return -ENOMEM;
L
Linus Torvalds 已提交
373 374
}

375 376 377
void unlink_anon_vmas(struct vm_area_struct *vma)
{
	struct anon_vma_chain *avc, *next;
378
	struct anon_vma *root = NULL;
379

380 381 382 383
	/*
	 * Unlink each anon_vma chained to the VMA.  This list is ordered
	 * from newest to oldest, ensuring the root anon_vma gets freed last.
	 */
384
	list_for_each_entry_safe(avc, next, &vma->anon_vma_chain, same_vma) {
385 386 387
		struct anon_vma *anon_vma = avc->anon_vma;

		root = lock_anon_vma_root(root, anon_vma);
388
		anon_vma_interval_tree_remove(avc, &anon_vma->rb_root);
389 390 391 392 393

		/*
		 * Leave empty anon_vmas on the list - we'll need
		 * to free them outside the lock.
		 */
394
		if (RB_EMPTY_ROOT(&anon_vma->rb_root.rb_root)) {
395
			anon_vma->parent->degree--;
396
			continue;
397
		}
398 399 400 401

		list_del(&avc->same_vma);
		anon_vma_chain_free(avc);
	}
402 403
	if (vma->anon_vma)
		vma->anon_vma->degree--;
404 405 406 407 408
	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()
409
	 * needing to write-acquire the anon_vma->root->rwsem.
410 411 412 413
	 */
	list_for_each_entry_safe(avc, next, &vma->anon_vma_chain, same_vma) {
		struct anon_vma *anon_vma = avc->anon_vma;

414
		VM_WARN_ON(anon_vma->degree);
415 416
		put_anon_vma(anon_vma);

417 418 419 420 421
		list_del(&avc->same_vma);
		anon_vma_chain_free(avc);
	}
}

422
static void anon_vma_ctor(void *data)
L
Linus Torvalds 已提交
423
{
C
Christoph Lameter 已提交
424
	struct anon_vma *anon_vma = data;
L
Linus Torvalds 已提交
425

426
	init_rwsem(&anon_vma->rwsem);
427
	atomic_set(&anon_vma->refcount, 0);
428
	anon_vma->rb_root = RB_ROOT_CACHED;
L
Linus Torvalds 已提交
429 430 431 432 433
}

void __init anon_vma_init(void)
{
	anon_vma_cachep = kmem_cache_create("anon_vma", sizeof(struct anon_vma),
434
			0, SLAB_TYPESAFE_BY_RCU|SLAB_PANIC|SLAB_ACCOUNT,
435 436 437
			anon_vma_ctor);
	anon_vma_chain_cachep = KMEM_CACHE(anon_vma_chain,
			SLAB_PANIC|SLAB_ACCOUNT);
L
Linus Torvalds 已提交
438 439 440
}

/*
441 442 443 444 445 446 447 448 449
 * 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).
 *
450 451 452 453 454
 * 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 ].
 *
455 456 457 458 459 460 461
 * 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
462
 */
463
struct anon_vma *page_get_anon_vma(struct page *page)
L
Linus Torvalds 已提交
464
{
465
	struct anon_vma *anon_vma = NULL;
L
Linus Torvalds 已提交
466 467 468
	unsigned long anon_mapping;

	rcu_read_lock();
469
	anon_mapping = (unsigned long)READ_ONCE(page->mapping);
H
Hugh Dickins 已提交
470
	if ((anon_mapping & PAGE_MAPPING_FLAGS) != PAGE_MAPPING_ANON)
L
Linus Torvalds 已提交
471 472 473 474 475
		goto out;
	if (!page_mapped(page))
		goto out;

	anon_vma = (struct anon_vma *) (anon_mapping - PAGE_MAPPING_ANON);
476 477 478 479
	if (!atomic_inc_not_zero(&anon_vma->refcount)) {
		anon_vma = NULL;
		goto out;
	}
480 481 482

	/*
	 * If this page is still mapped, then its anon_vma cannot have been
483 484
	 * freed.  But if it has been unmapped, we have no security against the
	 * anon_vma structure being freed and reused (for another anon_vma:
485
	 * SLAB_TYPESAFE_BY_RCU guarantees that - so the atomic_inc_not_zero()
486
	 * above cannot corrupt).
487
	 */
488
	if (!page_mapped(page)) {
489
		rcu_read_unlock();
490
		put_anon_vma(anon_vma);
491
		return NULL;
492
	}
L
Linus Torvalds 已提交
493 494
out:
	rcu_read_unlock();
495 496 497 498

	return anon_vma;
}

499 500 501 502 503 504 505
/*
 * 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.
 */
506
struct anon_vma *page_lock_anon_vma_read(struct page *page)
507
{
508
	struct anon_vma *anon_vma = NULL;
509
	struct anon_vma *root_anon_vma;
510
	unsigned long anon_mapping;
511

512
	rcu_read_lock();
513
	anon_mapping = (unsigned long)READ_ONCE(page->mapping);
514 515 516 517 518 519
	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);
520
	root_anon_vma = READ_ONCE(anon_vma->root);
521
	if (down_read_trylock(&root_anon_vma->rwsem)) {
522
		/*
523 524
		 * If the page is still mapped, then this anon_vma is still
		 * its anon_vma, and holding the mutex ensures that it will
525
		 * not go away, see anon_vma_free().
526
		 */
527
		if (!page_mapped(page)) {
528
			up_read(&root_anon_vma->rwsem);
529 530 531 532
			anon_vma = NULL;
		}
		goto out;
	}
533

534 535 536 537 538 539 540
	/* trylock failed, we got to sleep */
	if (!atomic_inc_not_zero(&anon_vma->refcount)) {
		anon_vma = NULL;
		goto out;
	}

	if (!page_mapped(page)) {
541
		rcu_read_unlock();
542
		put_anon_vma(anon_vma);
543
		return NULL;
544 545 546 547
	}

	/* we pinned the anon_vma, its safe to sleep */
	rcu_read_unlock();
548
	anon_vma_lock_read(anon_vma);
549 550 551 552 553

	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
554
		 * we'll deadlock on the anon_vma_lock_write() recursion.
555
		 */
556
		anon_vma_unlock_read(anon_vma);
557 558 559 560 561 562 563 564
		__put_anon_vma(anon_vma);
		anon_vma = NULL;
	}

	return anon_vma;

out:
	rcu_read_unlock();
565
	return anon_vma;
566 567
}

568
void page_unlock_anon_vma_read(struct anon_vma *anon_vma)
569
{
570
	anon_vma_unlock_read(anon_vma);
L
Linus Torvalds 已提交
571 572
}

573 574 575 576 577 578 579 580 581 582 583 584 585 586
#ifdef CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH
/*
 * 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;

	if (!tlb_ubc->flush_required)
		return;

587
	arch_tlbbatch_flush(&tlb_ubc->arch);
588
	tlb_ubc->flush_required = false;
589
	tlb_ubc->writable = false;
590 591
}

592 593 594 595 596 597 598 599 600
/* 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();
}

601
static void set_tlb_ubc_flush_pending(struct mm_struct *mm, bool writable)
602 603 604
{
	struct tlbflush_unmap_batch *tlb_ubc = &current->tlb_ubc;

605
	arch_tlbbatch_add_mm(&tlb_ubc->arch, mm);
606
	tlb_ubc->flush_required = true;
607

608 609 610 611 612 613 614
	/*
	 * Ensure compiler does not re-order the setting of tlb_flush_batched
	 * before the PTE is cleared.
	 */
	barrier();
	mm->tlb_flush_batched = true;

615 616 617 618 619 620 621
	/*
	 * 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;
622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641
}

/*
 * 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;
}
642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670

/*
 * Reclaim unmaps pages under the PTL but do not flush the TLB prior to
 * releasing the PTL if TLB flushes are batched. It's possible for a parallel
 * operation such as mprotect or munmap to race between reclaim unmapping
 * the page and flushing the page. If this race occurs, it potentially allows
 * access to data via a stale TLB entry. Tracking all mm's that have TLB
 * batching in flight would be expensive during reclaim so instead track
 * whether TLB batching occurred in the past and if so then do a flush here
 * if required. This will cost one additional flush per reclaim cycle paid
 * by the first operation at risk such as mprotect and mumap.
 *
 * This must be called under the PTL so that an access to tlb_flush_batched
 * that is potentially a "reclaim vs mprotect/munmap/etc" race will synchronise
 * via the PTL.
 */
void flush_tlb_batched_pending(struct mm_struct *mm)
{
	if (mm->tlb_flush_batched) {
		flush_tlb_mm(mm);

		/*
		 * Do not allow the compiler to re-order the clearing of
		 * tlb_flush_batched before the tlb is flushed.
		 */
		barrier();
		mm->tlb_flush_batched = false;
	}
}
671
#else
672
static void set_tlb_ubc_flush_pending(struct mm_struct *mm, bool writable)
673 674 675 676 677 678 679 680 681
{
}

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
682
/*
H
Huang Shijie 已提交
683
 * At what user virtual address is page expected in vma?
684
 * Caller should check the page is actually part of the vma.
L
Linus Torvalds 已提交
685 686 687
 */
unsigned long page_address_in_vma(struct page *page, struct vm_area_struct *vma)
{
688
	unsigned long address;
689
	if (PageAnon(page)) {
690 691 692 693 694 695 696
		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)
697
			return -EFAULT;
698 699
	} else if (page->mapping) {
		if (!vma->vm_file || vma->vm_file->f_mapping != page->mapping)
L
Linus Torvalds 已提交
700 701 702
			return -EFAULT;
	} else
		return -EFAULT;
703 704 705 706
	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
707 708
}

B
Bob Liu 已提交
709 710 711
pmd_t *mm_find_pmd(struct mm_struct *mm, unsigned long address)
{
	pgd_t *pgd;
712
	p4d_t *p4d;
B
Bob Liu 已提交
713 714
	pud_t *pud;
	pmd_t *pmd = NULL;
715
	pmd_t pmde;
B
Bob Liu 已提交
716 717 718 719 720

	pgd = pgd_offset(mm, address);
	if (!pgd_present(*pgd))
		goto out;

721 722 723 724 725
	p4d = p4d_offset(pgd, address);
	if (!p4d_present(*p4d))
		goto out;

	pud = pud_offset(p4d, address);
B
Bob Liu 已提交
726 727 728 729
	if (!pud_present(*pud))
		goto out;

	pmd = pmd_offset(pud, address);
730
	/*
731
	 * Some THP functions use the sequence pmdp_huge_clear_flush(), set_pmd_at()
732 733 734
	 * without holding anon_vma lock for write.  So when looking for a
	 * genuine pmde (in which to find pte), test present and !THP together.
	 */
735 736
	pmde = *pmd;
	barrier();
737
	if (!pmd_present(pmde) || pmd_trans_huge(pmde))
B
Bob Liu 已提交
738 739 740 741 742
		pmd = NULL;
out:
	return pmd;
}

743 744 745 746 747 748 749 750 751
struct page_referenced_arg {
	int mapcount;
	int referenced;
	unsigned long vm_flags;
	struct mem_cgroup *memcg;
};
/*
 * arg: page_referenced_arg will be passed
 */
M
Minchan Kim 已提交
752
static bool page_referenced_one(struct page *page, struct vm_area_struct *vma,
753 754 755
			unsigned long address, void *arg)
{
	struct page_referenced_arg *pra = arg;
756 757 758 759 760
	struct page_vma_mapped_walk pvmw = {
		.page = page,
		.vma = vma,
		.address = address,
	};
761 762
	int referenced = 0;

763 764
	while (page_vma_mapped_walk(&pvmw)) {
		address = pvmw.address;
765

766 767 768
		if (vma->vm_flags & VM_LOCKED) {
			page_vma_mapped_walk_done(&pvmw);
			pra->vm_flags |= VM_LOCKED;
M
Minchan Kim 已提交
769
			return false; /* To break the loop */
770
		}
771

772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788
		if (pvmw.pte) {
			if (ptep_clear_flush_young_notify(vma, address,
						pvmw.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(!(vma->vm_flags & VM_SEQ_READ)))
					referenced++;
			}
		} else if (IS_ENABLED(CONFIG_TRANSPARENT_HUGEPAGE)) {
			if (pmdp_clear_flush_young_notify(vma, address,
						pvmw.pmd))
789
				referenced++;
790 791 792
		} else {
			/* unexpected pmd-mapped page? */
			WARN_ON_ONCE(1);
793
		}
794 795

		pra->mapcount--;
796 797
	}

798 799 800 801 802
	if (referenced)
		clear_page_idle(page);
	if (test_and_clear_page_young(page))
		referenced++;

803 804 805
	if (referenced) {
		pra->referenced++;
		pra->vm_flags |= vma->vm_flags;
L
Linus Torvalds 已提交
806
	}
807

808
	if (!pra->mapcount)
M
Minchan Kim 已提交
809
		return false; /* To break the loop */
810

M
Minchan Kim 已提交
811
	return true;
L
Linus Torvalds 已提交
812 813
}

814
static bool invalid_page_referenced_vma(struct vm_area_struct *vma, void *arg)
L
Linus Torvalds 已提交
815
{
816 817
	struct page_referenced_arg *pra = arg;
	struct mem_cgroup *memcg = pra->memcg;
L
Linus Torvalds 已提交
818

819 820
	if (!mm_match_cgroup(vma->vm_mm, memcg))
		return true;
L
Linus Torvalds 已提交
821

822
	return false;
L
Linus Torvalds 已提交
823 824 825 826 827 828
}

/**
 * page_referenced - test if the page was referenced
 * @page: the page to test
 * @is_locked: caller holds lock on the page
829
 * @memcg: target memory cgroup
830
 * @vm_flags: collect encountered vma->vm_flags who actually referenced the page
L
Linus Torvalds 已提交
831 832 833 834
 *
 * Quick test_and_clear_referenced for all mappings to a page,
 * returns the number of ptes which referenced the page.
 */
835 836
int page_referenced(struct page *page,
		    int is_locked,
837
		    struct mem_cgroup *memcg,
838
		    unsigned long *vm_flags)
L
Linus Torvalds 已提交
839
{
H
Hugh Dickins 已提交
840
	int we_locked = 0;
841
	struct page_referenced_arg pra = {
842
		.mapcount = total_mapcount(page),
843 844 845 846 847 848 849
		.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
850

851
	*vm_flags = 0;
852 853 854 855 856 857 858 859 860 861
	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
862
	}
863 864 865 866 867 868 869 870 871 872

	/*
	 * 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;
	}

873
	rmap_walk(page, &rwc);
874 875 876 877 878 879
	*vm_flags = pra.vm_flags;

	if (we_locked)
		unlock_page(page);

	return pra.referenced;
L
Linus Torvalds 已提交
880 881
}

M
Minchan Kim 已提交
882
static bool page_mkclean_one(struct page *page, struct vm_area_struct *vma,
883
			    unsigned long address, void *arg)
884
{
885 886 887 888 889 890
	struct page_vma_mapped_walk pvmw = {
		.page = page,
		.vma = vma,
		.address = address,
		.flags = PVMW_SYNC,
	};
891
	unsigned long start = address, end;
892
	int *cleaned = arg;
893

894 895 896 897 898 899 900
	/*
	 * We have to assume the worse case ie pmd for invalidation. Note that
	 * the page can not be free from this function.
	 */
	end = min(vma->vm_end, start + (PAGE_SIZE << compound_order(page)));
	mmu_notifier_invalidate_range_start(vma->vm_mm, start, end);

901
	while (page_vma_mapped_walk(&pvmw)) {
902
		unsigned long cstart;
903
		int ret = 0;
904 905

		cstart = address = pvmw.address;
906 907 908 909 910 911 912
		if (pvmw.pte) {
			pte_t entry;
			pte_t *pte = pvmw.pte;

			if (!pte_dirty(*pte) && !pte_write(*pte))
				continue;

913 914
			flush_cache_page(vma, address, pte_pfn(*pte));
			entry = ptep_clear_flush(vma, address, pte);
915 916
			entry = pte_wrprotect(entry);
			entry = pte_mkclean(entry);
917
			set_pte_at(vma->vm_mm, address, pte, entry);
918 919 920 921 922 923 924 925 926
			ret = 1;
		} else {
#ifdef CONFIG_TRANSPARENT_HUGE_PAGECACHE
			pmd_t *pmd = pvmw.pmd;
			pmd_t entry;

			if (!pmd_dirty(*pmd) && !pmd_write(*pmd))
				continue;

927 928
			flush_cache_page(vma, address, page_to_pfn(page));
			entry = pmdp_huge_clear_flush(vma, address, pmd);
929 930
			entry = pmd_wrprotect(entry);
			entry = pmd_mkclean(entry);
931
			set_pmd_at(vma->vm_mm, address, pmd, entry);
932
			cstart &= PMD_MASK;
933 934 935 936 937 938
			ret = 1;
#else
			/* unexpected pmd-mapped page? */
			WARN_ON_ONCE(1);
#endif
		}
939

940 941 942 943 944 945 946 947
		/*
		 * No need to call mmu_notifier_invalidate_range() as we are
		 * downgrading page table protection not changing it to point
		 * to a new page.
		 *
		 * See Documentation/vm/mmu_notifier.txt
		 */
		if (ret)
948
			(*cleaned)++;
949
	}
950

951 952
	mmu_notifier_invalidate_range_end(vma->vm_mm, start, end);

M
Minchan Kim 已提交
953
	return true;
954 955
}

956
static bool invalid_mkclean_vma(struct vm_area_struct *vma, void *arg)
957
{
958
	if (vma->vm_flags & VM_SHARED)
F
Fengguang Wu 已提交
959
		return false;
960

F
Fengguang Wu 已提交
961
	return true;
962 963 964 965
}

int page_mkclean(struct page *page)
{
966 967 968 969 970 971 972
	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,
	};
973 974 975

	BUG_ON(!PageLocked(page));

976 977 978 979 980 981 982 983
	if (!page_mapped(page))
		return 0;

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

	rmap_walk(page, &rwc);
984

985
	return cleaned;
986
}
J
Jaya Kumar 已提交
987
EXPORT_SYMBOL_GPL(page_mkclean);
988

989 990 991 992 993 994 995 996 997 998
/**
 * 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
 *
 * 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.
 */
999
void page_move_anon_rmap(struct page *page, struct vm_area_struct *vma)
1000 1001 1002
{
	struct anon_vma *anon_vma = vma->anon_vma;

1003 1004
	page = compound_head(page);

1005
	VM_BUG_ON_PAGE(!PageLocked(page), page);
1006
	VM_BUG_ON_VMA(!anon_vma, vma);
1007 1008

	anon_vma = (void *) anon_vma + PAGE_MAPPING_ANON;
1009 1010 1011 1012 1013 1014
	/*
	 * 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);
1015 1016
}

N
Nick Piggin 已提交
1017
/**
A
Andi Kleen 已提交
1018 1019 1020 1021
 * __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	
1022
 * @exclusive:	the page is exclusively owned by the current process
N
Nick Piggin 已提交
1023 1024
 */
static void __page_set_anon_rmap(struct page *page,
1025
	struct vm_area_struct *vma, unsigned long address, int exclusive)
N
Nick Piggin 已提交
1026
{
1027
	struct anon_vma *anon_vma = vma->anon_vma;
1028

1029
	BUG_ON(!anon_vma);
1030

A
Andi Kleen 已提交
1031 1032 1033
	if (PageAnon(page))
		return;

1034
	/*
1035 1036 1037
	 * If the page isn't exclusively mapped into this vma,
	 * we must use the _oldest_ possible anon_vma for the
	 * page mapping!
1038
	 */
A
Andi Kleen 已提交
1039
	if (!exclusive)
1040
		anon_vma = anon_vma->root;
N
Nick Piggin 已提交
1041 1042 1043 1044 1045 1046

	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
1047
/**
R
Randy Dunlap 已提交
1048
 * __page_check_anon_rmap - sanity check anonymous rmap addition
N
Nick Piggin 已提交
1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068
 * @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
1069
	BUG_ON(page_anon_vma(page)->root != vma->anon_vma->root);
1070
	BUG_ON(page_to_pgoff(page) != linear_page_index(vma, address));
N
Nick Piggin 已提交
1071 1072 1073
#endif
}

L
Linus Torvalds 已提交
1074 1075 1076 1077 1078
/**
 * 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
1079
 * @compound:	charge the page as compound or small page
L
Linus Torvalds 已提交
1080
 *
H
Hugh Dickins 已提交
1081
 * The caller needs to hold the pte lock, and the page must be locked in
1082 1083 1084
 * 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
1085 1086
 */
void page_add_anon_rmap(struct page *page,
1087
	struct vm_area_struct *vma, unsigned long address, bool compound)
1088
{
1089
	do_page_add_anon_rmap(page, vma, address, compound ? RMAP_COMPOUND : 0);
1090 1091 1092 1093 1094 1095 1096 1097
}

/*
 * 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,
1098
	struct vm_area_struct *vma, unsigned long address, int flags)
L
Linus Torvalds 已提交
1099
{
1100 1101 1102
	bool compound = flags & RMAP_COMPOUND;
	bool first;

1103 1104
	if (compound) {
		atomic_t *mapcount;
1105
		VM_BUG_ON_PAGE(!PageLocked(page), page);
1106 1107 1108
		VM_BUG_ON_PAGE(!PageTransHuge(page), page);
		mapcount = compound_mapcount_ptr(page);
		first = atomic_inc_and_test(mapcount);
1109 1110 1111 1112
	} else {
		first = atomic_inc_and_test(&page->_mapcount);
	}

1113
	if (first) {
1114
		int nr = compound ? hpage_nr_pages(page) : 1;
1115 1116 1117 1118 1119 1120
		/*
		 * 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.
		 */
1121
		if (compound)
1122
			__inc_node_page_state(page, NR_ANON_THPS);
1123
		__mod_node_page_state(page_pgdat(page), NR_ANON_MAPPED, nr);
1124
	}
H
Hugh Dickins 已提交
1125 1126 1127
	if (unlikely(PageKsm(page)))
		return;

1128
	VM_BUG_ON_PAGE(!PageLocked(page), page);
1129

1130
	/* address might be in next vma when migration races vma_adjust */
H
Hugh Dickins 已提交
1131
	if (first)
1132 1133
		__page_set_anon_rmap(page, vma, address,
				flags & RMAP_EXCLUSIVE);
1134
	else
N
Nick Piggin 已提交
1135
		__page_check_anon_rmap(page, vma, address);
L
Linus Torvalds 已提交
1136 1137
}

R
Randy Dunlap 已提交
1138
/**
N
Nick Piggin 已提交
1139 1140 1141 1142
 * 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
1143
 * @compound:	charge the page as compound or small page
N
Nick Piggin 已提交
1144 1145 1146
 *
 * 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
1147
 * Page does not have to be locked.
N
Nick Piggin 已提交
1148 1149
 */
void page_add_new_anon_rmap(struct page *page,
1150
	struct vm_area_struct *vma, unsigned long address, bool compound)
N
Nick Piggin 已提交
1151
{
1152 1153
	int nr = compound ? hpage_nr_pages(page) : 1;

1154
	VM_BUG_ON_VMA(address < vma->vm_start || address >= vma->vm_end, vma);
1155
	__SetPageSwapBacked(page);
1156 1157
	if (compound) {
		VM_BUG_ON_PAGE(!PageTransHuge(page), page);
1158 1159
		/* increment count (starts at -1) */
		atomic_set(compound_mapcount_ptr(page), 0);
1160
		__inc_node_page_state(page, NR_ANON_THPS);
1161 1162 1163 1164 1165
	} else {
		/* Anon THP always mapped first with PMD */
		VM_BUG_ON_PAGE(PageTransCompound(page), page);
		/* increment count (starts at -1) */
		atomic_set(&page->_mapcount, 0);
1166
	}
1167
	__mod_node_page_state(page_pgdat(page), NR_ANON_MAPPED, nr);
1168
	__page_set_anon_rmap(page, vma, address, 1);
N
Nick Piggin 已提交
1169 1170
}

L
Linus Torvalds 已提交
1171 1172 1173 1174
/**
 * page_add_file_rmap - add pte mapping to a file page
 * @page: the page to add the mapping to
 *
1175
 * The caller needs to hold the pte lock.
L
Linus Torvalds 已提交
1176
 */
K
Kirill A. Shutemov 已提交
1177
void page_add_file_rmap(struct page *page, bool compound)
L
Linus Torvalds 已提交
1178
{
K
Kirill A. Shutemov 已提交
1179 1180 1181
	int i, nr = 1;

	VM_BUG_ON_PAGE(compound && !PageTransHuge(page), page);
J
Johannes Weiner 已提交
1182
	lock_page_memcg(page);
K
Kirill A. Shutemov 已提交
1183 1184 1185 1186 1187 1188 1189
	if (compound && PageTransHuge(page)) {
		for (i = 0, nr = 0; i < HPAGE_PMD_NR; i++) {
			if (atomic_inc_and_test(&page[i]._mapcount))
				nr++;
		}
		if (!atomic_inc_and_test(compound_mapcount_ptr(page)))
			goto out;
1190
		VM_BUG_ON_PAGE(!PageSwapBacked(page), page);
1191
		__inc_node_page_state(page, NR_SHMEM_PMDMAPPED);
K
Kirill A. Shutemov 已提交
1192
	} else {
1193 1194 1195
		if (PageTransCompound(page) && page_mapping(page)) {
			VM_WARN_ON_ONCE(!PageLocked(page));

1196 1197 1198 1199
			SetPageDoubleMap(compound_head(page));
			if (PageMlocked(page))
				clear_page_mlock(compound_head(page));
		}
K
Kirill A. Shutemov 已提交
1200 1201
		if (!atomic_inc_and_test(&page->_mapcount))
			goto out;
1202
	}
1203
	__mod_lruvec_page_state(page, NR_FILE_MAPPED, nr);
K
Kirill A. Shutemov 已提交
1204
out:
J
Johannes Weiner 已提交
1205
	unlock_page_memcg(page);
L
Linus Torvalds 已提交
1206 1207
}

K
Kirill A. Shutemov 已提交
1208
static void page_remove_file_rmap(struct page *page, bool compound)
1209
{
K
Kirill A. Shutemov 已提交
1210 1211
	int i, nr = 1;

1212
	VM_BUG_ON_PAGE(compound && !PageHead(page), page);
J
Johannes Weiner 已提交
1213
	lock_page_memcg(page);
1214

1215 1216 1217 1218
	/* Hugepages are not counted in NR_FILE_MAPPED for now. */
	if (unlikely(PageHuge(page))) {
		/* hugetlb pages are always mapped with pmds */
		atomic_dec(compound_mapcount_ptr(page));
1219
		goto out;
1220
	}
1221

1222
	/* page still mapped by someone else? */
K
Kirill A. Shutemov 已提交
1223 1224 1225 1226 1227 1228 1229
	if (compound && PageTransHuge(page)) {
		for (i = 0, nr = 0; i < HPAGE_PMD_NR; i++) {
			if (atomic_add_negative(-1, &page[i]._mapcount))
				nr++;
		}
		if (!atomic_add_negative(-1, compound_mapcount_ptr(page)))
			goto out;
1230
		VM_BUG_ON_PAGE(!PageSwapBacked(page), page);
1231
		__dec_node_page_state(page, NR_SHMEM_PMDMAPPED);
K
Kirill A. Shutemov 已提交
1232 1233 1234 1235
	} else {
		if (!atomic_add_negative(-1, &page->_mapcount))
			goto out;
	}
1236 1237

	/*
1238
	 * We use the irq-unsafe __{inc|mod}_lruvec_page_state because
1239 1240 1241
	 * these counters are not modified in interrupt context, and
	 * pte lock(a spinlock) is held, which implies preemption disabled.
	 */
1242
	__mod_lruvec_page_state(page, NR_FILE_MAPPED, -nr);
1243 1244 1245 1246

	if (unlikely(PageMlocked(page)))
		clear_page_mlock(page);
out:
J
Johannes Weiner 已提交
1247
	unlock_page_memcg(page);
1248 1249
}

1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263
static void page_remove_anon_compound_rmap(struct page *page)
{
	int i, nr;

	if (!atomic_add_negative(-1, compound_mapcount_ptr(page)))
		return;

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

	if (!IS_ENABLED(CONFIG_TRANSPARENT_HUGEPAGE))
		return;

1264
	__dec_node_page_state(page, NR_ANON_THPS);
1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278

	if (TestClearPageDoubleMap(page)) {
		/*
		 * Subpages can be mapped with PTEs too. Check how many of
		 * themi are still mapped.
		 */
		for (i = 0, nr = 0; i < HPAGE_PMD_NR; i++) {
			if (atomic_add_negative(-1, &page[i]._mapcount))
				nr++;
		}
	} else {
		nr = HPAGE_PMD_NR;
	}

1279 1280 1281
	if (unlikely(PageMlocked(page)))
		clear_page_mlock(page);

1282
	if (nr) {
1283
		__mod_node_page_state(page_pgdat(page), NR_ANON_MAPPED, -nr);
1284 1285
		deferred_split_huge_page(page);
	}
1286 1287
}

L
Linus Torvalds 已提交
1288 1289
/**
 * page_remove_rmap - take down pte mapping from a page
1290 1291
 * @page:	page to remove mapping from
 * @compound:	uncharge the page as compound or small page
L
Linus Torvalds 已提交
1292
 *
1293
 * The caller needs to hold the pte lock.
L
Linus Torvalds 已提交
1294
 */
1295
void page_remove_rmap(struct page *page, bool compound)
L
Linus Torvalds 已提交
1296
{
K
Kirill A. Shutemov 已提交
1297 1298
	if (!PageAnon(page))
		return page_remove_file_rmap(page, compound);
1299

1300 1301 1302
	if (compound)
		return page_remove_anon_compound_rmap(page);

1303 1304
	/* page still mapped by someone else? */
	if (!atomic_add_negative(-1, &page->_mapcount))
1305 1306
		return;

1307
	/*
1308 1309 1310
	 * 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.
1311
	 */
1312
	__dec_node_page_state(page, NR_ANON_MAPPED);
1313

1314 1315
	if (unlikely(PageMlocked(page)))
		clear_page_mlock(page);
1316

1317 1318 1319
	if (PageTransCompound(page))
		deferred_split_huge_page(compound_head(page));

1320 1321 1322 1323
	/*
	 * 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
1324
	 * before us: so leave the reset to free_unref_page,
1325 1326 1327 1328
	 * 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
1329 1330 1331
}

/*
1332
 * @arg: enum ttu_flags will be passed to this argument
L
Linus Torvalds 已提交
1333
 */
M
Minchan Kim 已提交
1334
static bool try_to_unmap_one(struct page *page, struct vm_area_struct *vma,
1335
		     unsigned long address, void *arg)
L
Linus Torvalds 已提交
1336 1337
{
	struct mm_struct *mm = vma->vm_mm;
1338 1339 1340 1341 1342
	struct page_vma_mapped_walk pvmw = {
		.page = page,
		.vma = vma,
		.address = address,
	};
L
Linus Torvalds 已提交
1343
	pte_t pteval;
1344
	struct page *subpage;
1345
	bool ret = true;
1346
	unsigned long start = address, end;
S
Shaohua Li 已提交
1347
	enum ttu_flags flags = (enum ttu_flags)arg;
L
Linus Torvalds 已提交
1348

1349 1350
	/* munlock has nothing to gain from examining un-locked vmas */
	if ((flags & TTU_MUNLOCK) && !(vma->vm_flags & VM_LOCKED))
M
Minchan Kim 已提交
1351
		return true;
1352

1353 1354 1355 1356
	if (IS_ENABLED(CONFIG_MIGRATION) && (flags & TTU_MIGRATION) &&
	    is_zone_device_page(page) && !is_device_private_page(page))
		return true;

1357 1358
	if (flags & TTU_SPLIT_HUGE_PMD) {
		split_huge_pmd_address(vma, address,
1359
				flags & TTU_SPLIT_FREEZE, page);
1360 1361
	}

1362 1363 1364 1365 1366 1367 1368 1369
	/*
	 * We have to assume the worse case ie pmd for invalidation. Note that
	 * the page can not be free in this function as call of try_to_unmap()
	 * must hold a reference on the page.
	 */
	end = min(vma->vm_end, start + (PAGE_SIZE << compound_order(page)));
	mmu_notifier_invalidate_range_start(vma->vm_mm, start, end);

1370
	while (page_vma_mapped_walk(&pvmw)) {
1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383
#ifdef CONFIG_ARCH_ENABLE_THP_MIGRATION
		/* PMD-mapped THP migration entry */
		if (!pvmw.pte && (flags & TTU_MIGRATION)) {
			VM_BUG_ON_PAGE(PageHuge(page) || !PageTransCompound(page), page);

			if (!PageAnon(page))
				continue;

			set_pmd_migration_entry(&pvmw, page);
			continue;
		}
#endif

1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398
		/*
		 * 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.
		 */
		if (!(flags & TTU_IGNORE_MLOCK)) {
			if (vma->vm_flags & VM_LOCKED) {
				/* PTE-mapped THP are never mlocked */
				if (!PageTransCompound(page)) {
					/*
					 * Holding pte lock, we do *not* need
					 * mmap_sem here
					 */
					mlock_vma_page(page);
				}
M
Minchan Kim 已提交
1399
				ret = false;
1400 1401
				page_vma_mapped_walk_done(&pvmw);
				break;
1402
			}
1403 1404
			if (flags & TTU_MUNLOCK)
				continue;
1405
		}
1406

1407 1408 1409 1410
		/* Unexpected PMD-mapped THP? */
		VM_BUG_ON_PAGE(!pvmw.pte, page);

		subpage = page - page_to_pfn(page) + pte_pfn(*pvmw.pte);
1411 1412
		address = pvmw.address;

1413

1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431
		if (IS_ENABLED(CONFIG_MIGRATION) &&
		    (flags & TTU_MIGRATION) &&
		    is_zone_device_page(page)) {
			swp_entry_t entry;
			pte_t swp_pte;

			pteval = ptep_get_and_clear(mm, pvmw.address, pvmw.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, 0);
			swp_pte = swp_entry_to_pte(entry);
			if (pte_soft_dirty(pteval))
				swp_pte = pte_swp_mksoft_dirty(swp_pte);
			set_pte_at(mm, pvmw.address, pvmw.pte, swp_pte);
1432 1433 1434 1435
			/*
			 * No need to invalidate here it will synchronize on
			 * against the special swap migration pte.
			 */
1436 1437 1438
			goto discard;
		}

1439
		if (!(flags & TTU_IGNORE_ACCESS)) {
1440
			if (ptep_clear_flush_young_notify(vma, address,
1441
						pvmw.pte)) {
M
Minchan Kim 已提交
1442
				ret = false;
1443 1444 1445
				page_vma_mapped_walk_done(&pvmw);
				break;
			}
N
Nick Piggin 已提交
1446
		}
L
Linus Torvalds 已提交
1447

1448
		/* Nuke the page table entry. */
1449
		flush_cache_page(vma, address, pte_pfn(*pvmw.pte));
1450 1451 1452 1453 1454 1455 1456 1457 1458
		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.
			 */
1459
			pteval = ptep_get_and_clear(mm, address, pvmw.pte);
1460 1461 1462

			set_tlb_ubc_flush_pending(mm, pte_dirty(pteval));
		} else {
1463
			pteval = ptep_clear_flush(vma, address, pvmw.pte);
1464
		}
1465

1466 1467 1468
		/* Move the dirty bit to the page. Now the pte is gone. */
		if (pte_dirty(pteval))
			set_page_dirty(page);
L
Linus Torvalds 已提交
1469

1470 1471
		/* Update high watermark before we lower rss */
		update_hiwater_rss(mm);
L
Linus Torvalds 已提交
1472

1473
		if (PageHWPoison(page) && !(flags & TTU_IGNORE_HWPOISON)) {
1474
			pteval = swp_entry_to_pte(make_hwpoison_entry(subpage));
1475 1476 1477
			if (PageHuge(page)) {
				int nr = 1 << compound_order(page);
				hugetlb_count_sub(nr, mm);
1478
				set_huge_swap_pte_at(mm, address,
1479 1480
						     pvmw.pte, pteval,
						     vma_mmu_pagesize(vma));
1481 1482
			} else {
				dec_mm_counter(mm, mm_counter(page));
1483
				set_pte_at(mm, address, pvmw.pte, pteval);
1484
			}
1485

1486 1487 1488 1489 1490 1491
		} 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.
			 */
1492
			dec_mm_counter(mm, mm_counter(page));
1493 1494 1495
			/* We have to invalidate as we cleared the pte */
			mmu_notifier_invalidate_range(mm, address,
						      address + PAGE_SIZE);
1496
		} else if (IS_ENABLED(CONFIG_MIGRATION) &&
1497
				(flags & (TTU_MIGRATION|TTU_SPLIT_FREEZE))) {
1498 1499
			swp_entry_t entry;
			pte_t swp_pte;
1500 1501 1502 1503 1504 1505 1506 1507

			if (arch_unmap_one(mm, vma, address, pteval) < 0) {
				set_pte_at(mm, address, pvmw.pte, pteval);
				ret = false;
				page_vma_mapped_walk_done(&pvmw);
				break;
			}

1508 1509 1510 1511 1512 1513 1514 1515 1516 1517
			/*
			 * 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(subpage,
					pte_write(pteval));
			swp_pte = swp_entry_to_pte(entry);
			if (pte_soft_dirty(pteval))
				swp_pte = pte_swp_mksoft_dirty(swp_pte);
1518
			set_pte_at(mm, address, pvmw.pte, swp_pte);
1519 1520 1521 1522
			/*
			 * No need to invalidate here it will synchronize on
			 * against the special swap migration pte.
			 */
1523 1524 1525 1526 1527 1528 1529
		} else if (PageAnon(page)) {
			swp_entry_t entry = { .val = page_private(subpage) };
			pte_t swp_pte;
			/*
			 * Store the swap location in the pte.
			 * See handle_pte_fault() ...
			 */
1530 1531
			if (unlikely(PageSwapBacked(page) != PageSwapCache(page))) {
				WARN_ON_ONCE(1);
1532
				ret = false;
1533
				/* We have to invalidate as we cleared the pte */
1534 1535
				mmu_notifier_invalidate_range(mm, address,
							address + PAGE_SIZE);
1536 1537 1538
				page_vma_mapped_walk_done(&pvmw);
				break;
			}
1539

S
Shaohua Li 已提交
1540 1541 1542
			/* MADV_FREE page check */
			if (!PageSwapBacked(page)) {
				if (!PageDirty(page)) {
1543 1544 1545
					/* Invalidate as we cleared the pte */
					mmu_notifier_invalidate_range(mm,
						address, address + PAGE_SIZE);
S
Shaohua Li 已提交
1546 1547 1548 1549 1550 1551 1552 1553
					dec_mm_counter(mm, MM_ANONPAGES);
					goto discard;
				}

				/*
				 * If the page was redirtied, it cannot be
				 * discarded. Remap the page to page table.
				 */
1554
				set_pte_at(mm, address, pvmw.pte, pteval);
M
Minchan Kim 已提交
1555
				SetPageSwapBacked(page);
M
Minchan Kim 已提交
1556
				ret = false;
S
Shaohua Li 已提交
1557 1558
				page_vma_mapped_walk_done(&pvmw);
				break;
1559
			}
M
Minchan Kim 已提交
1560

1561
			if (swap_duplicate(entry) < 0) {
1562
				set_pte_at(mm, address, pvmw.pte, pteval);
M
Minchan Kim 已提交
1563
				ret = false;
1564 1565 1566
				page_vma_mapped_walk_done(&pvmw);
				break;
			}
1567 1568 1569 1570 1571 1572
			if (arch_unmap_one(mm, vma, address, pteval) < 0) {
				set_pte_at(mm, address, pvmw.pte, pteval);
				ret = false;
				page_vma_mapped_walk_done(&pvmw);
				break;
			}
1573 1574 1575 1576 1577 1578
			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);
			}
M
Minchan Kim 已提交
1579
			dec_mm_counter(mm, MM_ANONPAGES);
1580 1581 1582 1583
			inc_mm_counter(mm, MM_SWAPENTS);
			swp_pte = swp_entry_to_pte(entry);
			if (pte_soft_dirty(pteval))
				swp_pte = pte_swp_mksoft_dirty(swp_pte);
1584
			set_pte_at(mm, address, pvmw.pte, swp_pte);
1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601 1602 1603 1604 1605
			/* Invalidate as we cleared the pte */
			mmu_notifier_invalidate_range(mm, address,
						      address + PAGE_SIZE);
		} else {
			/*
			 * We should not need to notify here as we reach this
			 * case only from freeze_page() itself only call from
			 * split_huge_page_to_list() so everything below must
			 * be true:
			 *   - page is not anonymous
			 *   - page is locked
			 *
			 * So as it is a locked file back page thus it can not
			 * be remove from the page cache and replace by a new
			 * page before mmu_notifier_invalidate_range_end so no
			 * concurrent thread might update its page table to
			 * point at new page while a device still is using this
			 * page.
			 *
			 * See Documentation/vm/mmu_notifier.txt
			 */
1606
			dec_mm_counter(mm, mm_counter_file(page));
1607
		}
M
Minchan Kim 已提交
1608
discard:
1609 1610 1611 1612 1613 1614 1615
		/*
		 * No need to call mmu_notifier_invalidate_range() it has be
		 * done above for all cases requiring it to happen under page
		 * table lock before mmu_notifier_invalidate_range_end()
		 *
		 * See Documentation/vm/mmu_notifier.txt
		 */
1616 1617 1618
		page_remove_rmap(subpage, PageHuge(page));
		put_page(page);
	}
1619 1620 1621

	mmu_notifier_invalidate_range_end(vma->vm_mm, start, end);

K
KOSAKI Motohiro 已提交
1622
	return ret;
L
Linus Torvalds 已提交
1623 1624
}

1625
bool is_vma_temporary_stack(struct vm_area_struct *vma)
1626 1627 1628 1629 1630 1631 1632 1633 1634 1635 1636 1637 1638
{
	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;
}

1639 1640 1641 1642 1643
static bool invalid_migration_vma(struct vm_area_struct *vma, void *arg)
{
	return is_vma_temporary_stack(vma);
}

1644
static int page_mapcount_is_zero(struct page *page)
1645
{
1646
	return !total_mapcount(page);
1647
}
1648

L
Linus Torvalds 已提交
1649 1650 1651
/**
 * try_to_unmap - try to remove all page table mappings to a page
 * @page: the page to get unmapped
1652
 * @flags: action and flags
L
Linus Torvalds 已提交
1653 1654 1655 1656
 *
 * Tries to remove all the page table entries which are mapping this
 * page, used in the pageout path.  Caller must hold the page lock.
 *
M
Minchan Kim 已提交
1657
 * If unmap is successful, return true. Otherwise, false.
L
Linus Torvalds 已提交
1658
 */
M
Minchan Kim 已提交
1659
bool try_to_unmap(struct page *page, enum ttu_flags flags)
L
Linus Torvalds 已提交
1660
{
1661 1662
	struct rmap_walk_control rwc = {
		.rmap_one = try_to_unmap_one,
S
Shaohua Li 已提交
1663
		.arg = (void *)flags,
1664
		.done = page_mapcount_is_zero,
1665 1666
		.anon_lock = page_lock_anon_vma_read,
	};
L
Linus Torvalds 已提交
1667

1668 1669 1670 1671 1672 1673 1674 1675
	/*
	 * 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.
	 */
1676 1677
	if ((flags & (TTU_MIGRATION|TTU_SPLIT_FREEZE))
	    && !PageKsm(page) && PageAnon(page))
1678 1679
		rwc.invalid_vma = invalid_migration_vma;

1680
	if (flags & TTU_RMAP_LOCKED)
M
Minchan Kim 已提交
1681
		rmap_walk_locked(page, &rwc);
1682
	else
M
Minchan Kim 已提交
1683
		rmap_walk(page, &rwc);
1684

M
Minchan Kim 已提交
1685
	return !page_mapcount(page) ? true : false;
L
Linus Torvalds 已提交
1686
}
N
Nikita Danilov 已提交
1687

1688 1689 1690 1691 1692
static int page_not_mapped(struct page *page)
{
	return !page_mapped(page);
};

N
Nick Piggin 已提交
1693 1694 1695 1696 1697 1698 1699 1700
/**
 * 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.
 */
M
Minchan Kim 已提交
1701

1702 1703
void try_to_munlock(struct page *page)
{
1704 1705
	struct rmap_walk_control rwc = {
		.rmap_one = try_to_unmap_one,
S
Shaohua Li 已提交
1706
		.arg = (void *)TTU_MUNLOCK,
1707 1708 1709 1710 1711
		.done = page_not_mapped,
		.anon_lock = page_lock_anon_vma_read,

	};

1712
	VM_BUG_ON_PAGE(!PageLocked(page) || PageLRU(page), page);
1713
	VM_BUG_ON_PAGE(PageCompound(page) && PageDoubleMap(page), page);
N
Nick Piggin 已提交
1714

1715
	rmap_walk(page, &rwc);
N
Nick Piggin 已提交
1716
}
1717

P
Peter Zijlstra 已提交
1718
void __put_anon_vma(struct anon_vma *anon_vma)
1719
{
P
Peter Zijlstra 已提交
1720
	struct anon_vma *root = anon_vma->root;
1721

1722
	anon_vma_free(anon_vma);
P
Peter Zijlstra 已提交
1723 1724
	if (root != anon_vma && atomic_dec_and_test(&root->refcount))
		anon_vma_free(root);
1725 1726
}

1727 1728
static struct anon_vma *rmap_walk_anon_lock(struct page *page,
					struct rmap_walk_control *rwc)
1729 1730 1731
{
	struct anon_vma *anon_vma;

1732 1733 1734
	if (rwc->anon_lock)
		return rwc->anon_lock(page);

1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746 1747 1748
	/*
	 * 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;
}

1749
/*
1750 1751 1752 1753 1754 1755 1756 1757 1758 1759 1760 1761
 * 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.
1762
 */
M
Minchan Kim 已提交
1763
static void rmap_walk_anon(struct page *page, struct rmap_walk_control *rwc,
1764
		bool locked)
1765 1766
{
	struct anon_vma *anon_vma;
1767
	pgoff_t pgoff_start, pgoff_end;
1768
	struct anon_vma_chain *avc;
1769

1770 1771 1772 1773 1774 1775 1776
	if (locked) {
		anon_vma = page_anon_vma(page);
		/* anon_vma disappear under us? */
		VM_BUG_ON_PAGE(!anon_vma, page);
	} else {
		anon_vma = rmap_walk_anon_lock(page, rwc);
	}
1777
	if (!anon_vma)
M
Minchan Kim 已提交
1778
		return;
1779

1780 1781 1782 1783
	pgoff_start = page_to_pgoff(page);
	pgoff_end = pgoff_start + hpage_nr_pages(page) - 1;
	anon_vma_interval_tree_foreach(avc, &anon_vma->rb_root,
			pgoff_start, pgoff_end) {
1784
		struct vm_area_struct *vma = avc->vma;
1785
		unsigned long address = vma_address(page, vma);
1786

1787 1788
		cond_resched();

1789 1790 1791
		if (rwc->invalid_vma && rwc->invalid_vma(vma, rwc->arg))
			continue;

M
Minchan Kim 已提交
1792
		if (!rwc->rmap_one(page, vma, address, rwc->arg))
1793
			break;
1794 1795
		if (rwc->done && rwc->done(page))
			break;
1796
	}
1797 1798 1799

	if (!locked)
		anon_vma_unlock_read(anon_vma);
1800 1801
}

1802 1803 1804 1805 1806 1807 1808 1809 1810 1811 1812 1813 1814
/*
 * 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.
 */
M
Minchan Kim 已提交
1815
static void rmap_walk_file(struct page *page, struct rmap_walk_control *rwc,
1816
		bool locked)
1817
{
1818
	struct address_space *mapping = page_mapping(page);
1819
	pgoff_t pgoff_start, pgoff_end;
1820 1821
	struct vm_area_struct *vma;

1822 1823 1824 1825
	/*
	 * 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,
1826
	 * so we can safely take mapping->i_mmap_rwsem.
1827
	 */
1828
	VM_BUG_ON_PAGE(!PageLocked(page), page);
1829

1830
	if (!mapping)
M
Minchan Kim 已提交
1831
		return;
D
Davidlohr Bueso 已提交
1832

1833 1834
	pgoff_start = page_to_pgoff(page);
	pgoff_end = pgoff_start + hpage_nr_pages(page) - 1;
1835 1836
	if (!locked)
		i_mmap_lock_read(mapping);
1837 1838
	vma_interval_tree_foreach(vma, &mapping->i_mmap,
			pgoff_start, pgoff_end) {
1839
		unsigned long address = vma_address(page, vma);
1840

1841 1842
		cond_resched();

1843 1844 1845
		if (rwc->invalid_vma && rwc->invalid_vma(vma, rwc->arg))
			continue;

M
Minchan Kim 已提交
1846
		if (!rwc->rmap_one(page, vma, address, rwc->arg))
1847 1848 1849
			goto done;
		if (rwc->done && rwc->done(page))
			goto done;
1850
	}
1851 1852

done:
1853 1854
	if (!locked)
		i_mmap_unlock_read(mapping);
1855 1856
}

M
Minchan Kim 已提交
1857
void rmap_walk(struct page *page, struct rmap_walk_control *rwc)
1858 1859
{
	if (unlikely(PageKsm(page)))
M
Minchan Kim 已提交
1860
		rmap_walk_ksm(page, rwc);
1861
	else if (PageAnon(page))
M
Minchan Kim 已提交
1862
		rmap_walk_anon(page, rwc, false);
1863
	else
M
Minchan Kim 已提交
1864
		rmap_walk_file(page, rwc, false);
1865 1866 1867
}

/* Like rmap_walk, but caller holds relevant rmap lock */
M
Minchan Kim 已提交
1868
void rmap_walk_locked(struct page *page, struct rmap_walk_control *rwc)
1869 1870 1871 1872
{
	/* no ksm support for now */
	VM_BUG_ON_PAGE(PageKsm(page), page);
	if (PageAnon(page))
M
Minchan Kim 已提交
1873
		rmap_walk_anon(page, rwc, true);
1874
	else
M
Minchan Kim 已提交
1875
		rmap_walk_file(page, rwc, true);
1876
}
1877

N
Naoya Horiguchi 已提交
1878
#ifdef CONFIG_HUGETLB_PAGE
1879 1880 1881 1882 1883 1884 1885 1886 1887
/*
 * 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;
1888

1889
	BUG_ON(!anon_vma);
1890 1891 1892 1893 1894 1895

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

1896 1897 1898 1899 1900 1901 1902 1903 1904 1905
	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;
1906 1907

	BUG_ON(!PageLocked(page));
1908
	BUG_ON(!anon_vma);
1909
	/* address might be in next vma when migration races vma_adjust */
1910
	first = atomic_inc_and_test(compound_mapcount_ptr(page));
1911 1912 1913 1914 1915 1916 1917 1918
	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);
1919
	atomic_set(compound_mapcount_ptr(page), 0);
1920 1921
	__hugepage_set_anon_rmap(page, vma, address, 1);
}
N
Naoya Horiguchi 已提交
1922
#endif /* CONFIG_HUGETLB_PAGE */