1 #ifndef _LINUX_RMAP_H
2 #define _LINUX_RMAP_H
3 /*
4 * Declarations for Reverse Mapping functions in mm/rmap.c
5 */
6
7 #include <linux/list.h>
8 #include <linux/slab.h>
9 #include <linux/mm.h>
10 #include <linux/rwsem.h>
11 #include <linux/memcontrol.h>
12
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 */
27 struct anon_vma {
28 struct anon_vma *root; /* Root of this anon_vma tree */
29 struct rw_semaphore rwsem; /* W: modification, R: walking the list */
30 /*
31 * The refcount is taken on an anon_vma when there is no
32 * 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 */
37 atomic_t refcount;
38
39 /*
40 * Count of child anon_vmas and VMAs which points to this anon_vma.
41 *
42 * This counter is used for making decision about reusing anon_vma
43 * instead of forking new one. See comments in function anon_vma_clone.
44 */
45 unsigned degree;
46
47 struct anon_vma *parent; /* Parent of this anon_vma */
48
49 /*
50 * NOTE: the LSB of the rb_root.rb_node is set by
51 * mm_take_all_locks() _after_ taking the above lock. So the
52 * rb_root must only be read/written after taking the above lock
53 * to be sure to see a valid next pointer. The LSB bit itself
54 * is serialized by a system wide lock only visible to
55 * mm_take_all_locks() (mm_all_locks_mutex).
56 */
57 struct rb_root rb_root; /* Interval tree of private "related" vmas */
58 };
59
60 /*
61 * The copy-on-write semantics of fork mean that an anon_vma
62 * can become associated with multiple processes. Furthermore,
63 * each child process will have its own anon_vma, where new
64 * pages for that process are instantiated.
65 *
66 * This structure allows us to find the anon_vmas associated
67 * with a VMA, or the VMAs associated with an anon_vma.
68 * The "same_vma" list contains the anon_vma_chains linking
69 * all the anon_vmas associated with this VMA.
70 * The "rb" field indexes on an interval tree the anon_vma_chains
71 * which link all the VMAs associated with this anon_vma.
72 */
73 struct anon_vma_chain {
74 struct vm_area_struct *vma;
75 struct anon_vma *anon_vma;
76 struct list_head same_vma; /* locked by mmap_sem & page_table_lock */
77 struct rb_node rb; /* locked by anon_vma->rwsem */
78 unsigned long rb_subtree_last;
79 #ifdef CONFIG_DEBUG_VM_RB
80 unsigned long cached_vma_start, cached_vma_last;
81 #endif
82 };
83
84 enum ttu_flags {
85 TTU_UNMAP = 1, /* unmap mode */
86 TTU_MIGRATION = 2, /* migration mode */
87 TTU_MUNLOCK = 4, /* munlock mode */
88
89 TTU_IGNORE_MLOCK = (1 << 8), /* ignore mlock */
90 TTU_IGNORE_ACCESS = (1 << 9), /* don't age */
91 TTU_IGNORE_HWPOISON = (1 << 10),/* corrupted page is recoverable */
92 };
93
94 #ifdef CONFIG_MMU
get_anon_vma(struct anon_vma * anon_vma)95 static inline void get_anon_vma(struct anon_vma *anon_vma)
96 {
97 atomic_inc(&anon_vma->refcount);
98 }
99
100 void __put_anon_vma(struct anon_vma *anon_vma);
101
put_anon_vma(struct anon_vma * anon_vma)102 static inline void put_anon_vma(struct anon_vma *anon_vma)
103 {
104 if (atomic_dec_and_test(&anon_vma->refcount))
105 __put_anon_vma(anon_vma);
106 }
107
page_anon_vma(struct page * page)108 static inline struct anon_vma *page_anon_vma(struct page *page)
109 {
110 if (((unsigned long)page->mapping & PAGE_MAPPING_FLAGS) !=
111 PAGE_MAPPING_ANON)
112 return NULL;
113 return page_rmapping(page);
114 }
115
anon_vma_lock_write(struct anon_vma * anon_vma)116 static inline void anon_vma_lock_write(struct anon_vma *anon_vma)
117 {
118 down_write(&anon_vma->root->rwsem);
119 }
120
anon_vma_unlock_write(struct anon_vma * anon_vma)121 static inline void anon_vma_unlock_write(struct anon_vma *anon_vma)
122 {
123 up_write(&anon_vma->root->rwsem);
124 }
125
anon_vma_lock_read(struct anon_vma * anon_vma)126 static inline void anon_vma_lock_read(struct anon_vma *anon_vma)
127 {
128 down_read(&anon_vma->root->rwsem);
129 }
130
anon_vma_unlock_read(struct anon_vma * anon_vma)131 static inline void anon_vma_unlock_read(struct anon_vma *anon_vma)
132 {
133 up_read(&anon_vma->root->rwsem);
134 }
135
136
137 /*
138 * anon_vma helper functions.
139 */
140 void anon_vma_init(void); /* create anon_vma_cachep */
141 int anon_vma_prepare(struct vm_area_struct *);
142 void unlink_anon_vmas(struct vm_area_struct *);
143 int anon_vma_clone(struct vm_area_struct *, struct vm_area_struct *);
144 int anon_vma_fork(struct vm_area_struct *, struct vm_area_struct *);
145
anon_vma_merge(struct vm_area_struct * vma,struct vm_area_struct * next)146 static inline void anon_vma_merge(struct vm_area_struct *vma,
147 struct vm_area_struct *next)
148 {
149 VM_BUG_ON_VMA(vma->anon_vma != next->anon_vma, vma);
150 unlink_anon_vmas(next);
151 }
152
153 struct anon_vma *page_get_anon_vma(struct page *page);
154
155 /*
156 * rmap interfaces called when adding or removing pte of page
157 */
158 void page_move_anon_rmap(struct page *, struct vm_area_struct *, unsigned long);
159 void page_add_anon_rmap(struct page *, struct vm_area_struct *, unsigned long);
160 void do_page_add_anon_rmap(struct page *, struct vm_area_struct *,
161 unsigned long, int);
162 void page_add_new_anon_rmap(struct page *, struct vm_area_struct *, unsigned long);
163 void page_add_file_rmap(struct page *);
164 void page_remove_rmap(struct page *);
165
166 void hugepage_add_anon_rmap(struct page *, struct vm_area_struct *,
167 unsigned long);
168 void hugepage_add_new_anon_rmap(struct page *, struct vm_area_struct *,
169 unsigned long);
170
page_dup_rmap(struct page * page)171 static inline void page_dup_rmap(struct page *page)
172 {
173 atomic_inc(&page->_mapcount);
174 }
175
176 /*
177 * Called from mm/vmscan.c to handle paging out
178 */
179 int page_referenced(struct page *, int is_locked,
180 struct mem_cgroup *memcg, unsigned long *vm_flags);
181
182 #define TTU_ACTION(x) ((x) & TTU_ACTION_MASK)
183
184 int try_to_unmap(struct page *, enum ttu_flags flags);
185
186 /*
187 * Called from mm/filemap_xip.c to unmap empty zero page
188 */
189 pte_t *__page_check_address(struct page *, struct mm_struct *,
190 unsigned long, spinlock_t **, int);
191
page_check_address(struct page * page,struct mm_struct * mm,unsigned long address,spinlock_t ** ptlp,int sync)192 static inline pte_t *page_check_address(struct page *page, struct mm_struct *mm,
193 unsigned long address,
194 spinlock_t **ptlp, int sync)
195 {
196 pte_t *ptep;
197
198 __cond_lock(*ptlp, ptep = __page_check_address(page, mm, address,
199 ptlp, sync));
200 return ptep;
201 }
202
203 /*
204 * Used by swapoff to help locate where page is expected in vma.
205 */
206 unsigned long page_address_in_vma(struct page *, struct vm_area_struct *);
207
208 /*
209 * Cleans the PTEs of shared mappings.
210 * (and since clean PTEs should also be readonly, write protects them too)
211 *
212 * returns the number of cleaned PTEs.
213 */
214 int page_mkclean(struct page *);
215
216 /*
217 * called in munlock()/munmap() path to check for other vmas holding
218 * the page mlocked.
219 */
220 int try_to_munlock(struct page *);
221
222 /*
223 * Called by memory-failure.c to kill processes.
224 */
225 struct anon_vma *page_lock_anon_vma_read(struct page *page);
226 void page_unlock_anon_vma_read(struct anon_vma *anon_vma);
227 int page_mapped_in_vma(struct page *page, struct vm_area_struct *vma);
228
229 /*
230 * rmap_walk_control: To control rmap traversing for specific needs
231 *
232 * arg: passed to rmap_one() and invalid_vma()
233 * rmap_one: executed on each vma where page is mapped
234 * done: for checking traversing termination condition
235 * file_nonlinear: for handling file nonlinear mapping
236 * anon_lock: for getting anon_lock by optimized way rather than default
237 * invalid_vma: for skipping uninterested vma
238 */
239 struct rmap_walk_control {
240 void *arg;
241 int (*rmap_one)(struct page *page, struct vm_area_struct *vma,
242 unsigned long addr, void *arg);
243 int (*done)(struct page *page);
244 int (*file_nonlinear)(struct page *, struct address_space *, void *arg);
245 struct anon_vma *(*anon_lock)(struct page *page);
246 bool (*invalid_vma)(struct vm_area_struct *vma, void *arg);
247 };
248
249 int rmap_walk(struct page *page, struct rmap_walk_control *rwc);
250
251 #else /* !CONFIG_MMU */
252
253 #define anon_vma_init() do {} while (0)
254 #define anon_vma_prepare(vma) (0)
255 #define anon_vma_link(vma) do {} while (0)
256
page_referenced(struct page * page,int is_locked,struct mem_cgroup * memcg,unsigned long * vm_flags)257 static inline int page_referenced(struct page *page, int is_locked,
258 struct mem_cgroup *memcg,
259 unsigned long *vm_flags)
260 {
261 *vm_flags = 0;
262 return 0;
263 }
264
265 #define try_to_unmap(page, refs) SWAP_FAIL
266
page_mkclean(struct page * page)267 static inline int page_mkclean(struct page *page)
268 {
269 return 0;
270 }
271
272
273 #endif /* CONFIG_MMU */
274
275 /*
276 * Return values of try_to_unmap
277 */
278 #define SWAP_SUCCESS 0
279 #define SWAP_AGAIN 1
280 #define SWAP_FAIL 2
281 #define SWAP_MLOCK 3
282
283 #endif /* _LINUX_RMAP_H */
284