4 * Declarations for Reverse Mapping functions in mm/rmap.c
7 #include <linux/list.h>
8 #include <linux/slab.h>
10 #include <linux/rwsem.h>
11 #include <linux/memcontrol.h>
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.
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.
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.
28 struct anon_vma *root; /* Root of this anon_vma tree */
29 struct rw_semaphore rwsem; /* W: modification, R: walking the list */
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
40 * Count of child anon_vmas. Equals to the count of all anon_vmas that
41 * have ->parent pointing to this one, including itself.
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.
46 unsigned long num_children;
47 /* Count of VMAs whose ->anon_vma pointer points to this object. */
48 unsigned long num_active_vmas;
50 struct anon_vma *parent; /* Parent of this anon_vma */
53 * NOTE: the LSB of the rb_root.rb_node is set by
54 * mm_take_all_locks() _after_ taking the above lock. So the
55 * rb_root must only be read/written after taking the above lock
56 * to be sure to see a valid next pointer. The LSB bit itself
57 * is serialized by a system wide lock only visible to
58 * mm_take_all_locks() (mm_all_locks_mutex).
60 struct rb_root rb_root; /* Interval tree of private "related" vmas */
64 * The copy-on-write semantics of fork mean that an anon_vma
65 * can become associated with multiple processes. Furthermore,
66 * each child process will have its own anon_vma, where new
67 * pages for that process are instantiated.
69 * This structure allows us to find the anon_vmas associated
70 * with a VMA, or the VMAs associated with an anon_vma.
71 * The "same_vma" list contains the anon_vma_chains linking
72 * all the anon_vmas associated with this VMA.
73 * The "rb" field indexes on an interval tree the anon_vma_chains
74 * which link all the VMAs associated with this anon_vma.
76 struct anon_vma_chain {
77 struct vm_area_struct *vma;
78 struct anon_vma *anon_vma;
79 struct list_head same_vma; /* locked by mmap_sem & page_table_lock */
80 struct rb_node rb; /* locked by anon_vma->rwsem */
81 unsigned long rb_subtree_last;
82 #ifdef CONFIG_DEBUG_VM_RB
83 unsigned long cached_vma_start, cached_vma_last;
88 TTU_UNMAP = 1, /* unmap mode */
89 TTU_MIGRATION = 2, /* migration mode */
90 TTU_MUNLOCK = 4, /* munlock mode */
91 TTU_LZFREE = 8, /* lazy free mode */
92 TTU_SPLIT_HUGE_PMD = 16, /* split huge PMD if any */
94 TTU_IGNORE_MLOCK = (1 << 8), /* ignore mlock */
95 TTU_IGNORE_ACCESS = (1 << 9), /* don't age */
96 TTU_IGNORE_HWPOISON = (1 << 10),/* corrupted page is recoverable */
97 TTU_BATCH_FLUSH = (1 << 11), /* Batch TLB flushes where possible
98 * and caller guarantees they will
99 * do a final flush if necessary */
100 TTU_RMAP_LOCKED = (1 << 12) /* do not grab rmap lock:
105 static inline void get_anon_vma(struct anon_vma *anon_vma)
107 atomic_inc(&anon_vma->refcount);
110 void __put_anon_vma(struct anon_vma *anon_vma);
112 static inline void put_anon_vma(struct anon_vma *anon_vma)
114 if (atomic_dec_and_test(&anon_vma->refcount))
115 __put_anon_vma(anon_vma);
118 static inline void anon_vma_lock_write(struct anon_vma *anon_vma)
120 down_write(&anon_vma->root->rwsem);
123 static inline void anon_vma_unlock_write(struct anon_vma *anon_vma)
125 up_write(&anon_vma->root->rwsem);
128 static inline void anon_vma_lock_read(struct anon_vma *anon_vma)
130 down_read(&anon_vma->root->rwsem);
133 static inline void anon_vma_unlock_read(struct anon_vma *anon_vma)
135 up_read(&anon_vma->root->rwsem);
140 * anon_vma helper functions.
142 void anon_vma_init(void); /* create anon_vma_cachep */
143 int anon_vma_prepare(struct vm_area_struct *);
144 void unlink_anon_vmas(struct vm_area_struct *);
145 int anon_vma_clone(struct vm_area_struct *, struct vm_area_struct *);
146 int anon_vma_fork(struct vm_area_struct *, struct vm_area_struct *);
148 static inline void anon_vma_merge(struct vm_area_struct *vma,
149 struct vm_area_struct *next)
151 VM_BUG_ON_VMA(vma->anon_vma != next->anon_vma, vma);
152 unlink_anon_vmas(next);
155 struct anon_vma *page_get_anon_vma(struct page *page);
157 /* bitflags for do_page_add_anon_rmap() */
158 #define RMAP_EXCLUSIVE 0x01
159 #define RMAP_COMPOUND 0x02
162 * rmap interfaces called when adding or removing pte of page
164 void page_move_anon_rmap(struct page *, struct vm_area_struct *);
165 void page_add_anon_rmap(struct page *, struct vm_area_struct *,
166 unsigned long, bool);
167 void do_page_add_anon_rmap(struct page *, struct vm_area_struct *,
169 void page_add_new_anon_rmap(struct page *, struct vm_area_struct *,
170 unsigned long, bool);
171 void page_add_file_rmap(struct page *, bool);
172 void page_remove_rmap(struct page *, bool);
174 void hugepage_add_anon_rmap(struct page *, struct vm_area_struct *,
176 void hugepage_add_new_anon_rmap(struct page *, struct vm_area_struct *,
179 static inline void page_dup_rmap(struct page *page, bool compound)
181 atomic_inc(compound ? compound_mapcount_ptr(page) : &page->_mapcount);
185 * Called from mm/vmscan.c to handle paging out
187 int page_referenced(struct page *, int is_locked,
188 struct mem_cgroup *memcg, unsigned long *vm_flags);
190 #define TTU_ACTION(x) ((x) & TTU_ACTION_MASK)
192 int try_to_unmap(struct page *, enum ttu_flags flags);
195 * Used by uprobes to replace a userspace page safely
197 pte_t *__page_check_address(struct page *, struct mm_struct *,
198 unsigned long, spinlock_t **, int);
200 static inline pte_t *page_check_address(struct page *page, struct mm_struct *mm,
201 unsigned long address,
202 spinlock_t **ptlp, int sync)
206 __cond_lock(*ptlp, ptep = __page_check_address(page, mm, address,
212 * Used by idle page tracking to check if a page was referenced via page
215 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
216 bool page_check_address_transhuge(struct page *page, struct mm_struct *mm,
217 unsigned long address, pmd_t **pmdp,
218 pte_t **ptep, spinlock_t **ptlp);
220 static inline bool page_check_address_transhuge(struct page *page,
221 struct mm_struct *mm, unsigned long address,
222 pmd_t **pmdp, pte_t **ptep, spinlock_t **ptlp)
224 *ptep = page_check_address(page, mm, address, ptlp, 0);
231 * Used by swapoff to help locate where page is expected in vma.
233 unsigned long page_address_in_vma(struct page *, struct vm_area_struct *);
236 * Cleans the PTEs of shared mappings.
237 * (and since clean PTEs should also be readonly, write protects them too)
239 * returns the number of cleaned PTEs.
241 int page_mkclean(struct page *);
244 * called in munlock()/munmap() path to check for other vmas holding
247 int try_to_munlock(struct page *);
249 void remove_migration_ptes(struct page *old, struct page *new, bool locked);
252 * Called by memory-failure.c to kill processes.
254 struct anon_vma *page_lock_anon_vma_read(struct page *page);
255 void page_unlock_anon_vma_read(struct anon_vma *anon_vma);
256 int page_mapped_in_vma(struct page *page, struct vm_area_struct *vma);
259 * rmap_walk_control: To control rmap traversing for specific needs
261 * arg: passed to rmap_one() and invalid_vma()
262 * rmap_one: executed on each vma where page is mapped
263 * done: for checking traversing termination condition
264 * anon_lock: for getting anon_lock by optimized way rather than default
265 * invalid_vma: for skipping uninterested vma
267 struct rmap_walk_control {
269 int (*rmap_one)(struct page *page, struct vm_area_struct *vma,
270 unsigned long addr, void *arg);
271 int (*done)(struct page *page);
272 struct anon_vma *(*anon_lock)(struct page *page);
273 bool (*invalid_vma)(struct vm_area_struct *vma, void *arg);
276 int rmap_walk(struct page *page, struct rmap_walk_control *rwc);
277 int rmap_walk_locked(struct page *page, struct rmap_walk_control *rwc);
279 #else /* !CONFIG_MMU */
281 #define anon_vma_init() do {} while (0)
282 #define anon_vma_prepare(vma) (0)
283 #define anon_vma_link(vma) do {} while (0)
285 static inline int page_referenced(struct page *page, int is_locked,
286 struct mem_cgroup *memcg,
287 unsigned long *vm_flags)
293 #define try_to_unmap(page, refs) SWAP_FAIL
295 static inline int page_mkclean(struct page *page)
301 #endif /* CONFIG_MMU */
304 * Return values of try_to_unmap
306 #define SWAP_SUCCESS 0
310 #define SWAP_LZFREE 4
312 #endif /* _LINUX_RMAP_H */