blob: bfe1f4780644434cd8901abae42939d661e70a8e [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001#ifndef _LINUX_RMAP_H
2#define _LINUX_RMAP_H
3/*
4 * Declarations for Reverse Mapping functions in mm/rmap.c
5 */
6
Linus Torvalds1da177e2005-04-16 15:20:36 -07007#include <linux/list.h>
8#include <linux/slab.h>
9#include <linux/mm.h>
Peter Zijlstra2b575eb2011-05-24 17:12:11 -070010#include <linux/mutex.h>
Balbir Singhbed71612008-02-07 00:14:01 -080011#include <linux/memcontrol.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070012
13/*
14 * The anon_vma heads a list of private "related" vmas, to scan if
15 * an anonymous page pointing to this anon_vma needs to be unmapped:
16 * the vmas on the list will be related by forking, or by splitting.
17 *
18 * Since vmas come and go as they are split and merged (particularly
19 * in mprotect), the mapping field of an anonymous page cannot point
20 * directly to a vma: instead it points to an anon_vma, on whose list
21 * the related vmas can be easily linked or unlinked.
22 *
23 * After unlinking the last vma on the list, we must garbage collect
24 * the anon_vma object itself: we're guaranteed no page can be
25 * pointing to this anon_vma once its vma list is empty.
26 */
27struct anon_vma {
Rik van Riel5c341ee12010-08-09 17:18:39 -070028 struct anon_vma *root; /* Root of this anon_vma tree */
Peter Zijlstra2b575eb2011-05-24 17:12:11 -070029 struct mutex mutex; /* Serialize access to vma list */
Mel Gorman7f60c212010-05-24 14:32:18 -070030 /*
Peter Zijlstra83813262011-03-22 16:32:48 -070031 * The refcount is taken on an anon_vma when there is no
Mel Gorman7f60c212010-05-24 14:32:18 -070032 * guarantee that the vma of page tables will exist for
33 * the duration of the operation. A caller that takes
34 * the reference is responsible for clearing up the
35 * anon_vma if they are the last user on release
36 */
Peter Zijlstra83813262011-03-22 16:32:48 -070037 atomic_t refcount;
38
Andrea Arcangeli7906d002008-07-28 15:46:26 -070039 /*
Michel Lespinassebf181b92012-10-08 16:31:39 -070040 * NOTE: the LSB of the rb_root.rb_node is set by
Andrea Arcangeli7906d002008-07-28 15:46:26 -070041 * mm_take_all_locks() _after_ taking the above lock. So the
Michel Lespinassebf181b92012-10-08 16:31:39 -070042 * rb_root must only be read/written after taking the above lock
Andrea Arcangeli7906d002008-07-28 15:46:26 -070043 * to be sure to see a valid next pointer. The LSB bit itself
44 * is serialized by a system wide lock only visible to
45 * mm_take_all_locks() (mm_all_locks_mutex).
46 */
Michel Lespinassebf181b92012-10-08 16:31:39 -070047 struct rb_root rb_root; /* Interval tree of private "related" vmas */
Rik van Riel5beb4932010-03-05 13:42:07 -080048};
49
50/*
51 * The copy-on-write semantics of fork mean that an anon_vma
52 * can become associated with multiple processes. Furthermore,
53 * each child process will have its own anon_vma, where new
54 * pages for that process are instantiated.
55 *
56 * This structure allows us to find the anon_vmas associated
57 * with a VMA, or the VMAs associated with an anon_vma.
58 * The "same_vma" list contains the anon_vma_chains linking
59 * all the anon_vmas associated with this VMA.
Michel Lespinassebf181b92012-10-08 16:31:39 -070060 * The "rb" field indexes on an interval tree the anon_vma_chains
Rik van Riel5beb4932010-03-05 13:42:07 -080061 * which link all the VMAs associated with this anon_vma.
62 */
63struct anon_vma_chain {
64 struct vm_area_struct *vma;
65 struct anon_vma *anon_vma;
66 struct list_head same_vma; /* locked by mmap_sem & page_table_lock */
Michel Lespinassebf181b92012-10-08 16:31:39 -070067 struct rb_node rb; /* locked by anon_vma->mutex */
68 unsigned long rb_subtree_last;
Michel Lespinasseed8ea812012-10-08 16:31:45 -070069#ifdef CONFIG_DEBUG_VM_RB
70 unsigned long cached_vma_start, cached_vma_last;
71#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -070072};
73
Minchan Kim02c6de82012-10-08 16:31:55 -070074enum ttu_flags {
75 TTU_UNMAP = 0, /* unmap mode */
76 TTU_MIGRATION = 1, /* migration mode */
77 TTU_MUNLOCK = 2, /* munlock mode */
78 TTU_ACTION_MASK = 0xff,
79
80 TTU_IGNORE_MLOCK = (1 << 8), /* ignore mlock */
81 TTU_IGNORE_ACCESS = (1 << 9), /* don't age */
82 TTU_IGNORE_HWPOISON = (1 << 10),/* corrupted page is recoverable */
83};
84
Linus Torvalds1da177e2005-04-16 15:20:36 -070085#ifdef CONFIG_MMU
Rik van Riel76545062010-08-09 17:18:41 -070086static inline void get_anon_vma(struct anon_vma *anon_vma)
87{
Peter Zijlstra83813262011-03-22 16:32:48 -070088 atomic_inc(&anon_vma->refcount);
Rik van Riel76545062010-08-09 17:18:41 -070089}
90
Peter Zijlstra01d8b202011-03-22 16:32:49 -070091void __put_anon_vma(struct anon_vma *anon_vma);
92
93static inline void put_anon_vma(struct anon_vma *anon_vma)
94{
95 if (atomic_dec_and_test(&anon_vma->refcount))
96 __put_anon_vma(anon_vma);
97}
Linus Torvalds1da177e2005-04-16 15:20:36 -070098
Hugh Dickins3ca7b3c2009-12-14 17:58:57 -080099static inline struct anon_vma *page_anon_vma(struct page *page)
100{
101 if (((unsigned long)page->mapping & PAGE_MAPPING_FLAGS) !=
102 PAGE_MAPPING_ANON)
103 return NULL;
104 return page_rmapping(page);
105}
106
Rik van Rielbb4a3402010-08-09 17:18:37 -0700107static inline void vma_lock_anon_vma(struct vm_area_struct *vma)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700108{
109 struct anon_vma *anon_vma = vma->anon_vma;
110 if (anon_vma)
Peter Zijlstra2b575eb2011-05-24 17:12:11 -0700111 mutex_lock(&anon_vma->root->mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700112}
113
Rik van Rielbb4a3402010-08-09 17:18:37 -0700114static inline void vma_unlock_anon_vma(struct vm_area_struct *vma)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700115{
116 struct anon_vma *anon_vma = vma->anon_vma;
117 if (anon_vma)
Peter Zijlstra2b575eb2011-05-24 17:12:11 -0700118 mutex_unlock(&anon_vma->root->mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700119}
120
Rik van Rielcba48b92010-08-09 17:18:38 -0700121static inline void anon_vma_lock(struct anon_vma *anon_vma)
122{
Peter Zijlstra2b575eb2011-05-24 17:12:11 -0700123 mutex_lock(&anon_vma->root->mutex);
Rik van Rielcba48b92010-08-09 17:18:38 -0700124}
125
126static inline void anon_vma_unlock(struct anon_vma *anon_vma)
127{
Peter Zijlstra2b575eb2011-05-24 17:12:11 -0700128 mutex_unlock(&anon_vma->root->mutex);
Rik van Rielcba48b92010-08-09 17:18:38 -0700129}
130
Linus Torvalds1da177e2005-04-16 15:20:36 -0700131/*
132 * anon_vma helper functions.
133 */
134void anon_vma_init(void); /* create anon_vma_cachep */
135int anon_vma_prepare(struct vm_area_struct *);
Rik van Riel5beb4932010-03-05 13:42:07 -0800136void unlink_anon_vmas(struct vm_area_struct *);
137int anon_vma_clone(struct vm_area_struct *, struct vm_area_struct *);
138int anon_vma_fork(struct vm_area_struct *, struct vm_area_struct *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700139
Rik van Riel5beb4932010-03-05 13:42:07 -0800140static inline void anon_vma_merge(struct vm_area_struct *vma,
141 struct vm_area_struct *next)
142{
143 VM_BUG_ON(vma->anon_vma != next->anon_vma);
144 unlink_anon_vmas(next);
145}
146
Peter Zijlstra01d8b202011-03-22 16:32:49 -0700147struct anon_vma *page_get_anon_vma(struct page *page);
148
Linus Torvalds1da177e2005-04-16 15:20:36 -0700149/*
150 * rmap interfaces called when adding or removing pte of page
151 */
Rik van Rielc44b6742010-03-05 13:42:09 -0800152void page_move_anon_rmap(struct page *, struct vm_area_struct *, unsigned long);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700153void page_add_anon_rmap(struct page *, struct vm_area_struct *, unsigned long);
Rik van Rielad8c2ee2010-08-09 17:19:48 -0700154void do_page_add_anon_rmap(struct page *, struct vm_area_struct *,
155 unsigned long, int);
Nick Piggin9617d952006-01-06 00:11:12 -0800156void page_add_new_anon_rmap(struct page *, struct vm_area_struct *, unsigned long);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700157void page_add_file_rmap(struct page *);
Hugh Dickinsedc315f2009-01-06 14:40:11 -0800158void page_remove_rmap(struct page *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700159
Naoya Horiguchi0fe6e202010-05-28 09:29:16 +0900160void hugepage_add_anon_rmap(struct page *, struct vm_area_struct *,
161 unsigned long);
162void hugepage_add_new_anon_rmap(struct page *, struct vm_area_struct *,
163 unsigned long);
164
Hugh Dickins21333b22009-09-21 17:01:59 -0700165static inline void page_dup_rmap(struct page *page)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700166{
167 atomic_inc(&page->_mapcount);
168}
169
170/*
171 * Called from mm/vmscan.c to handle paging out
172 */
Wu Fengguang6fe6b7e2009-06-16 15:33:05 -0700173int page_referenced(struct page *, int is_locked,
Johannes Weiner72835c82012-01-12 17:18:32 -0800174 struct mem_cgroup *memcg, unsigned long *vm_flags);
Hugh Dickins5ad64682009-12-14 17:59:24 -0800175int page_referenced_one(struct page *, struct vm_area_struct *,
176 unsigned long address, unsigned int *mapcount, unsigned long *vm_flags);
177
Andi Kleen14fa31b2009-09-16 11:50:10 +0200178#define TTU_ACTION(x) ((x) & TTU_ACTION_MASK)
179
180int try_to_unmap(struct page *, enum ttu_flags flags);
Hugh Dickins5ad64682009-12-14 17:59:24 -0800181int try_to_unmap_one(struct page *, struct vm_area_struct *,
182 unsigned long address, enum ttu_flags flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700183
184/*
Carsten Otteceffc072005-06-23 22:05:25 -0700185 * Called from mm/filemap_xip.c to unmap empty zero page
186 */
Namhyung Kime9a81a82010-10-26 14:22:01 -0700187pte_t *__page_check_address(struct page *, struct mm_struct *,
Nick Piggin479db0b2008-08-20 14:09:18 -0700188 unsigned long, spinlock_t **, int);
Carsten Otteceffc072005-06-23 22:05:25 -0700189
Namhyung Kime9a81a82010-10-26 14:22:01 -0700190static inline pte_t *page_check_address(struct page *page, struct mm_struct *mm,
191 unsigned long address,
192 spinlock_t **ptlp, int sync)
193{
194 pte_t *ptep;
195
196 __cond_lock(*ptlp, ptep = __page_check_address(page, mm, address,
197 ptlp, sync));
198 return ptep;
199}
200
Carsten Otteceffc072005-06-23 22:05:25 -0700201/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700202 * Used by swapoff to help locate where page is expected in vma.
203 */
204unsigned long page_address_in_vma(struct page *, struct vm_area_struct *);
205
Peter Zijlstrad08b3852006-09-25 23:30:57 -0700206/*
207 * Cleans the PTEs of shared mappings.
208 * (and since clean PTEs should also be readonly, write protects them too)
209 *
210 * returns the number of cleaned PTEs.
211 */
212int page_mkclean(struct page *);
213
Nick Pigginb291f002008-10-18 20:26:44 -0700214/*
215 * called in munlock()/munmap() path to check for other vmas holding
216 * the page mlocked.
217 */
218int try_to_munlock(struct page *);
Nick Pigginb291f002008-10-18 20:26:44 -0700219
Andi Kleen10be22d2009-09-16 11:50:04 +0200220/*
221 * Called by memory-failure.c to kill processes.
222 */
Peter Zijlstra25aeeb02011-05-24 17:12:07 -0700223struct anon_vma *page_lock_anon_vma(struct page *page);
Andi Kleen10be22d2009-09-16 11:50:04 +0200224void page_unlock_anon_vma(struct anon_vma *anon_vma);
Andi Kleen6a460792009-09-16 11:50:15 +0200225int page_mapped_in_vma(struct page *page, struct vm_area_struct *vma);
Andi Kleen10be22d2009-09-16 11:50:04 +0200226
Hugh Dickinse9995ef2009-12-14 17:59:31 -0800227/*
228 * Called by migrate.c to remove migration ptes, but might be used more later.
229 */
230int rmap_walk(struct page *page, int (*rmap_one)(struct page *,
231 struct vm_area_struct *, unsigned long, void *), void *arg);
232
Linus Torvalds1da177e2005-04-16 15:20:36 -0700233#else /* !CONFIG_MMU */
234
235#define anon_vma_init() do {} while (0)
236#define anon_vma_prepare(vma) (0)
237#define anon_vma_link(vma) do {} while (0)
238
Mike Frysinger01ff53f2009-06-23 12:37:01 -0700239static inline int page_referenced(struct page *page, int is_locked,
Johannes Weiner72835c82012-01-12 17:18:32 -0800240 struct mem_cgroup *memcg,
Mike Frysinger01ff53f2009-06-23 12:37:01 -0700241 unsigned long *vm_flags)
242{
243 *vm_flags = 0;
Johannes Weiner645747462010-03-05 13:42:22 -0800244 return 0;
Mike Frysinger01ff53f2009-06-23 12:37:01 -0700245}
246
Christoph Lametera48d07a2006-02-01 03:05:38 -0800247#define try_to_unmap(page, refs) SWAP_FAIL
Linus Torvalds1da177e2005-04-16 15:20:36 -0700248
Peter Zijlstrad08b3852006-09-25 23:30:57 -0700249static inline int page_mkclean(struct page *page)
250{
251 return 0;
252}
253
254
Linus Torvalds1da177e2005-04-16 15:20:36 -0700255#endif /* CONFIG_MMU */
256
257/*
258 * Return values of try_to_unmap
259 */
260#define SWAP_SUCCESS 0
261#define SWAP_AGAIN 1
262#define SWAP_FAIL 2
Nick Pigginb291f002008-10-18 20:26:44 -0700263#define SWAP_MLOCK 3
Linus Torvalds1da177e2005-04-16 15:20:36 -0700264
265#endif /* _LINUX_RMAP_H */