]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | #ifndef _LINUX_RMAP_H |
2 | #define _LINUX_RMAP_H | |
3 | /* | |
4 | * Declarations for Reverse Mapping functions in mm/rmap.c | |
5 | */ | |
6 | ||
1da177e4 LT |
7 | #include <linux/list.h> |
8 | #include <linux/slab.h> | |
9 | #include <linux/mm.h> | |
5a505085 | 10 | #include <linux/rwsem.h> |
bed7161a | 11 | #include <linux/memcontrol.h> |
ace71a19 | 12 | #include <linux/highmem.h> |
1da177e4 LT |
13 | |
14 | /* | |
15 | * The anon_vma heads a list of private "related" vmas, to scan if | |
16 | * an anonymous page pointing to this anon_vma needs to be unmapped: | |
17 | * the vmas on the list will be related by forking, or by splitting. | |
18 | * | |
19 | * Since vmas come and go as they are split and merged (particularly | |
20 | * in mprotect), the mapping field of an anonymous page cannot point | |
21 | * directly to a vma: instead it points to an anon_vma, on whose list | |
22 | * the related vmas can be easily linked or unlinked. | |
23 | * | |
24 | * After unlinking the last vma on the list, we must garbage collect | |
25 | * the anon_vma object itself: we're guaranteed no page can be | |
26 | * pointing to this anon_vma once its vma list is empty. | |
27 | */ | |
28 | struct anon_vma { | |
5a505085 IM |
29 | struct anon_vma *root; /* Root of this anon_vma tree */ |
30 | struct rw_semaphore rwsem; /* W: modification, R: walking the list */ | |
7f60c214 | 31 | /* |
83813267 | 32 | * The refcount is taken on an anon_vma when there is no |
7f60c214 MG |
33 | * guarantee that the vma of page tables will exist for |
34 | * the duration of the operation. A caller that takes | |
35 | * the reference is responsible for clearing up the | |
36 | * anon_vma if they are the last user on release | |
37 | */ | |
83813267 PZ |
38 | atomic_t refcount; |
39 | ||
7a3ef208 KK |
40 | /* |
41 | * Count of child anon_vmas and VMAs which points to this anon_vma. | |
42 | * | |
43 | * This counter is used for making decision about reusing anon_vma | |
44 | * instead of forking new one. See comments in function anon_vma_clone. | |
45 | */ | |
46 | unsigned degree; | |
47 | ||
48 | struct anon_vma *parent; /* Parent of this anon_vma */ | |
49 | ||
7906d00c | 50 | /* |
bf181b9f | 51 | * NOTE: the LSB of the rb_root.rb_node is set by |
7906d00c | 52 | * mm_take_all_locks() _after_ taking the above lock. So the |
bf181b9f | 53 | * rb_root must only be read/written after taking the above lock |
7906d00c AA |
54 | * to be sure to see a valid next pointer. The LSB bit itself |
55 | * is serialized by a system wide lock only visible to | |
56 | * mm_take_all_locks() (mm_all_locks_mutex). | |
57 | */ | |
bf181b9f | 58 | struct rb_root rb_root; /* Interval tree of private "related" vmas */ |
5beb4930 RR |
59 | }; |
60 | ||
61 | /* | |
62 | * The copy-on-write semantics of fork mean that an anon_vma | |
63 | * can become associated with multiple processes. Furthermore, | |
64 | * each child process will have its own anon_vma, where new | |
65 | * pages for that process are instantiated. | |
66 | * | |
67 | * This structure allows us to find the anon_vmas associated | |
68 | * with a VMA, or the VMAs associated with an anon_vma. | |
69 | * The "same_vma" list contains the anon_vma_chains linking | |
70 | * all the anon_vmas associated with this VMA. | |
bf181b9f | 71 | * The "rb" field indexes on an interval tree the anon_vma_chains |
5beb4930 RR |
72 | * which link all the VMAs associated with this anon_vma. |
73 | */ | |
74 | struct anon_vma_chain { | |
75 | struct vm_area_struct *vma; | |
76 | struct anon_vma *anon_vma; | |
77 | struct list_head same_vma; /* locked by mmap_sem & page_table_lock */ | |
5a505085 | 78 | struct rb_node rb; /* locked by anon_vma->rwsem */ |
bf181b9f | 79 | unsigned long rb_subtree_last; |
ed8ea815 ML |
80 | #ifdef CONFIG_DEBUG_VM_RB |
81 | unsigned long cached_vma_start, cached_vma_last; | |
82 | #endif | |
1da177e4 LT |
83 | }; |
84 | ||
02c6de8d | 85 | enum ttu_flags { |
a128ca71 SL |
86 | TTU_MIGRATION = 0x1, /* migration mode */ |
87 | TTU_MUNLOCK = 0x2, /* munlock mode */ | |
88 | ||
89 | TTU_SPLIT_HUGE_PMD = 0x4, /* split huge PMD if any */ | |
90 | TTU_IGNORE_MLOCK = 0x8, /* ignore mlock */ | |
91 | TTU_IGNORE_ACCESS = 0x10, /* don't age */ | |
92 | TTU_IGNORE_HWPOISON = 0x20, /* corrupted page is recoverable */ | |
93 | TTU_BATCH_FLUSH = 0x40, /* Batch TLB flushes where possible | |
72b252ae MG |
94 | * and caller guarantees they will |
95 | * do a final flush if necessary */ | |
a128ca71 | 96 | TTU_RMAP_LOCKED = 0x80 /* do not grab rmap lock: |
2a52bcbc | 97 | * caller holds it */ |
02c6de8d MK |
98 | }; |
99 | ||
1da177e4 | 100 | #ifdef CONFIG_MMU |
76545066 RR |
101 | static inline void get_anon_vma(struct anon_vma *anon_vma) |
102 | { | |
83813267 | 103 | atomic_inc(&anon_vma->refcount); |
76545066 RR |
104 | } |
105 | ||
01d8b20d PZ |
106 | void __put_anon_vma(struct anon_vma *anon_vma); |
107 | ||
108 | static inline void put_anon_vma(struct anon_vma *anon_vma) | |
109 | { | |
110 | if (atomic_dec_and_test(&anon_vma->refcount)) | |
111 | __put_anon_vma(anon_vma); | |
112 | } | |
1da177e4 | 113 | |
4fc3f1d6 | 114 | static inline void anon_vma_lock_write(struct anon_vma *anon_vma) |
cba48b98 | 115 | { |
5a505085 | 116 | down_write(&anon_vma->root->rwsem); |
cba48b98 RR |
117 | } |
118 | ||
08b52706 | 119 | static inline void anon_vma_unlock_write(struct anon_vma *anon_vma) |
cba48b98 | 120 | { |
5a505085 | 121 | up_write(&anon_vma->root->rwsem); |
cba48b98 RR |
122 | } |
123 | ||
4fc3f1d6 IM |
124 | static inline void anon_vma_lock_read(struct anon_vma *anon_vma) |
125 | { | |
126 | down_read(&anon_vma->root->rwsem); | |
127 | } | |
128 | ||
129 | static inline void anon_vma_unlock_read(struct anon_vma *anon_vma) | |
130 | { | |
131 | up_read(&anon_vma->root->rwsem); | |
132 | } | |
133 | ||
134 | ||
1da177e4 LT |
135 | /* |
136 | * anon_vma helper functions. | |
137 | */ | |
138 | void anon_vma_init(void); /* create anon_vma_cachep */ | |
d5a187da | 139 | int __anon_vma_prepare(struct vm_area_struct *); |
5beb4930 RR |
140 | void unlink_anon_vmas(struct vm_area_struct *); |
141 | int anon_vma_clone(struct vm_area_struct *, struct vm_area_struct *); | |
142 | int anon_vma_fork(struct vm_area_struct *, struct vm_area_struct *); | |
1da177e4 | 143 | |
d5a187da VB |
144 | static inline int anon_vma_prepare(struct vm_area_struct *vma) |
145 | { | |
146 | if (likely(vma->anon_vma)) | |
147 | return 0; | |
148 | ||
149 | return __anon_vma_prepare(vma); | |
150 | } | |
151 | ||
5beb4930 RR |
152 | static inline void anon_vma_merge(struct vm_area_struct *vma, |
153 | struct vm_area_struct *next) | |
154 | { | |
81d1b09c | 155 | VM_BUG_ON_VMA(vma->anon_vma != next->anon_vma, vma); |
5beb4930 RR |
156 | unlink_anon_vmas(next); |
157 | } | |
158 | ||
01d8b20d PZ |
159 | struct anon_vma *page_get_anon_vma(struct page *page); |
160 | ||
d281ee61 KS |
161 | /* bitflags for do_page_add_anon_rmap() */ |
162 | #define RMAP_EXCLUSIVE 0x01 | |
163 | #define RMAP_COMPOUND 0x02 | |
164 | ||
1da177e4 LT |
165 | /* |
166 | * rmap interfaces called when adding or removing pte of page | |
167 | */ | |
5a49973d | 168 | void page_move_anon_rmap(struct page *, struct vm_area_struct *); |
d281ee61 KS |
169 | void page_add_anon_rmap(struct page *, struct vm_area_struct *, |
170 | unsigned long, bool); | |
ad8c2ee8 RR |
171 | void do_page_add_anon_rmap(struct page *, struct vm_area_struct *, |
172 | unsigned long, int); | |
d281ee61 KS |
173 | void page_add_new_anon_rmap(struct page *, struct vm_area_struct *, |
174 | unsigned long, bool); | |
dd78fedd | 175 | void page_add_file_rmap(struct page *, bool); |
d281ee61 | 176 | void page_remove_rmap(struct page *, bool); |
1da177e4 | 177 | |
0fe6e20b NH |
178 | void hugepage_add_anon_rmap(struct page *, struct vm_area_struct *, |
179 | unsigned long); | |
180 | void hugepage_add_new_anon_rmap(struct page *, struct vm_area_struct *, | |
181 | unsigned long); | |
182 | ||
53f9263b | 183 | static inline void page_dup_rmap(struct page *page, bool compound) |
1da177e4 | 184 | { |
53f9263b | 185 | atomic_inc(compound ? compound_mapcount_ptr(page) : &page->_mapcount); |
1da177e4 LT |
186 | } |
187 | ||
188 | /* | |
189 | * Called from mm/vmscan.c to handle paging out | |
190 | */ | |
6fe6b7e3 | 191 | int page_referenced(struct page *, int is_locked, |
72835c86 | 192 | struct mem_cgroup *memcg, unsigned long *vm_flags); |
5ad64688 | 193 | |
666e5a40 | 194 | bool try_to_unmap(struct page *, enum ttu_flags flags); |
1da177e4 | 195 | |
ace71a19 KS |
196 | /* Avoid racy checks */ |
197 | #define PVMW_SYNC (1 << 0) | |
198 | /* Look for migarion entries rather than present PTEs */ | |
199 | #define PVMW_MIGRATION (1 << 1) | |
200 | ||
201 | struct page_vma_mapped_walk { | |
202 | struct page *page; | |
203 | struct vm_area_struct *vma; | |
204 | unsigned long address; | |
205 | pmd_t *pmd; | |
206 | pte_t *pte; | |
207 | spinlock_t *ptl; | |
208 | unsigned int flags; | |
209 | }; | |
210 | ||
211 | static inline void page_vma_mapped_walk_done(struct page_vma_mapped_walk *pvmw) | |
212 | { | |
213 | if (pvmw->pte) | |
214 | pte_unmap(pvmw->pte); | |
215 | if (pvmw->ptl) | |
216 | spin_unlock(pvmw->ptl); | |
217 | } | |
218 | ||
219 | bool page_vma_mapped_walk(struct page_vma_mapped_walk *pvmw); | |
220 | ||
1da177e4 LT |
221 | /* |
222 | * Used by swapoff to help locate where page is expected in vma. | |
223 | */ | |
224 | unsigned long page_address_in_vma(struct page *, struct vm_area_struct *); | |
225 | ||
d08b3851 PZ |
226 | /* |
227 | * Cleans the PTEs of shared mappings. | |
228 | * (and since clean PTEs should also be readonly, write protects them too) | |
229 | * | |
230 | * returns the number of cleaned PTEs. | |
231 | */ | |
232 | int page_mkclean(struct page *); | |
233 | ||
b291f000 NP |
234 | /* |
235 | * called in munlock()/munmap() path to check for other vmas holding | |
236 | * the page mlocked. | |
237 | */ | |
192d7232 | 238 | void try_to_munlock(struct page *); |
b291f000 | 239 | |
e388466d KS |
240 | void remove_migration_ptes(struct page *old, struct page *new, bool locked); |
241 | ||
10be22df AK |
242 | /* |
243 | * Called by memory-failure.c to kill processes. | |
244 | */ | |
4fc3f1d6 IM |
245 | struct anon_vma *page_lock_anon_vma_read(struct page *page); |
246 | void page_unlock_anon_vma_read(struct anon_vma *anon_vma); | |
6a46079c | 247 | int page_mapped_in_vma(struct page *page, struct vm_area_struct *vma); |
10be22df | 248 | |
0dd1c7bb JK |
249 | /* |
250 | * rmap_walk_control: To control rmap traversing for specific needs | |
251 | * | |
252 | * arg: passed to rmap_one() and invalid_vma() | |
253 | * rmap_one: executed on each vma where page is mapped | |
254 | * done: for checking traversing termination condition | |
0dd1c7bb JK |
255 | * anon_lock: for getting anon_lock by optimized way rather than default |
256 | * invalid_vma: for skipping uninterested vma | |
257 | */ | |
051ac83a JK |
258 | struct rmap_walk_control { |
259 | void *arg; | |
e4b82222 MK |
260 | /* |
261 | * Return false if page table scanning in rmap_walk should be stopped. | |
262 | * Otherwise, return true. | |
263 | */ | |
264 | bool (*rmap_one)(struct page *page, struct vm_area_struct *vma, | |
051ac83a | 265 | unsigned long addr, void *arg); |
0dd1c7bb | 266 | int (*done)(struct page *page); |
0dd1c7bb JK |
267 | struct anon_vma *(*anon_lock)(struct page *page); |
268 | bool (*invalid_vma)(struct vm_area_struct *vma, void *arg); | |
051ac83a JK |
269 | }; |
270 | ||
1df631ae MK |
271 | void rmap_walk(struct page *page, struct rmap_walk_control *rwc); |
272 | void rmap_walk_locked(struct page *page, struct rmap_walk_control *rwc); | |
e9995ef9 | 273 | |
1da177e4 LT |
274 | #else /* !CONFIG_MMU */ |
275 | ||
276 | #define anon_vma_init() do {} while (0) | |
277 | #define anon_vma_prepare(vma) (0) | |
278 | #define anon_vma_link(vma) do {} while (0) | |
279 | ||
01ff53f4 | 280 | static inline int page_referenced(struct page *page, int is_locked, |
72835c86 | 281 | struct mem_cgroup *memcg, |
01ff53f4 MF |
282 | unsigned long *vm_flags) |
283 | { | |
284 | *vm_flags = 0; | |
64574746 | 285 | return 0; |
01ff53f4 MF |
286 | } |
287 | ||
666e5a40 | 288 | #define try_to_unmap(page, refs) false |
1da177e4 | 289 | |
d08b3851 PZ |
290 | static inline int page_mkclean(struct page *page) |
291 | { | |
292 | return 0; | |
293 | } | |
294 | ||
295 | ||
1da177e4 LT |
296 | #endif /* CONFIG_MMU */ |
297 | ||
1da177e4 | 298 | #endif /* _LINUX_RMAP_H */ |