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 TTU_LZFREE = 8, /* lazy free mode */ 89 90 TTU_IGNORE_MLOCK = (1 << 8), /* ignore mlock */ 91 TTU_IGNORE_ACCESS = (1 << 9), /* don't age */ 92 TTU_IGNORE_HWPOISON = (1 << 10),/* corrupted page is recoverable */ 93 TTU_BATCH_FLUSH = (1 << 11), /* Batch TLB flushes where possible 94 * and caller guarantees they will 95 * do a final flush if necessary */ 96 }; 97 98 #ifdef CONFIG_MMU 99 static inline void get_anon_vma(struct anon_vma *anon_vma) 100 { 101 atomic_inc(&anon_vma->refcount); 102 } 103 104 void __put_anon_vma(struct anon_vma *anon_vma); 105 106 static inline void put_anon_vma(struct anon_vma *anon_vma) 107 { 108 if (atomic_dec_and_test(&anon_vma->refcount)) 109 __put_anon_vma(anon_vma); 110 } 111 112 static inline void anon_vma_lock_write(struct anon_vma *anon_vma) 113 { 114 down_write(&anon_vma->root->rwsem); 115 } 116 117 static inline void anon_vma_unlock_write(struct anon_vma *anon_vma) 118 { 119 up_write(&anon_vma->root->rwsem); 120 } 121 122 static inline void anon_vma_lock_read(struct anon_vma *anon_vma) 123 { 124 down_read(&anon_vma->root->rwsem); 125 } 126 127 static inline void anon_vma_unlock_read(struct anon_vma *anon_vma) 128 { 129 up_read(&anon_vma->root->rwsem); 130 } 131 132 133 /* 134 * anon_vma helper functions. 135 */ 136 void anon_vma_init(void); /* create anon_vma_cachep */ 137 int anon_vma_prepare(struct vm_area_struct *); 138 void unlink_anon_vmas(struct vm_area_struct *); 139 int anon_vma_clone(struct vm_area_struct *, struct vm_area_struct *); 140 int anon_vma_fork(struct vm_area_struct *, struct vm_area_struct *); 141 142 static inline void anon_vma_merge(struct vm_area_struct *vma, 143 struct vm_area_struct *next) 144 { 145 VM_BUG_ON_VMA(vma->anon_vma != next->anon_vma, vma); 146 unlink_anon_vmas(next); 147 } 148 149 struct anon_vma *page_get_anon_vma(struct page *page); 150 151 /* bitflags for do_page_add_anon_rmap() */ 152 #define RMAP_EXCLUSIVE 0x01 153 #define RMAP_COMPOUND 0x02 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 *, 160 unsigned long, bool); 161 void do_page_add_anon_rmap(struct page *, struct vm_area_struct *, 162 unsigned long, int); 163 void page_add_new_anon_rmap(struct page *, struct vm_area_struct *, 164 unsigned long, bool); 165 void page_add_file_rmap(struct page *); 166 void page_remove_rmap(struct page *, bool); 167 168 void hugepage_add_anon_rmap(struct page *, struct vm_area_struct *, 169 unsigned long); 170 void hugepage_add_new_anon_rmap(struct page *, struct vm_area_struct *, 171 unsigned long); 172 173 static inline void page_dup_rmap(struct page *page, bool compound) 174 { 175 atomic_inc(compound ? compound_mapcount_ptr(page) : &page->_mapcount); 176 } 177 178 /* 179 * Called from mm/vmscan.c to handle paging out 180 */ 181 int page_referenced(struct page *, int is_locked, 182 struct mem_cgroup *memcg, unsigned long *vm_flags); 183 184 #define TTU_ACTION(x) ((x) & TTU_ACTION_MASK) 185 186 int try_to_unmap(struct page *, enum ttu_flags flags); 187 188 /* 189 * Used by uprobes to replace a userspace page safely 190 */ 191 pte_t *__page_check_address(struct page *, struct mm_struct *, 192 unsigned long, spinlock_t **, int); 193 194 static inline pte_t *page_check_address(struct page *page, struct mm_struct *mm, 195 unsigned long address, 196 spinlock_t **ptlp, int sync) 197 { 198 pte_t *ptep; 199 200 __cond_lock(*ptlp, ptep = __page_check_address(page, mm, address, 201 ptlp, sync)); 202 return ptep; 203 } 204 205 /* 206 * Used by idle page tracking to check if a page was referenced via page 207 * tables. 208 */ 209 #ifdef CONFIG_TRANSPARENT_HUGEPAGE 210 bool page_check_address_transhuge(struct page *page, struct mm_struct *mm, 211 unsigned long address, pmd_t **pmdp, 212 pte_t **ptep, spinlock_t **ptlp); 213 #else 214 static inline bool page_check_address_transhuge(struct page *page, 215 struct mm_struct *mm, unsigned long address, 216 pmd_t **pmdp, pte_t **ptep, spinlock_t **ptlp) 217 { 218 *ptep = page_check_address(page, mm, address, ptlp, 0); 219 *pmdp = NULL; 220 return !!*ptep; 221 } 222 #endif 223 224 /* 225 * Used by swapoff to help locate where page is expected in vma. 226 */ 227 unsigned long page_address_in_vma(struct page *, struct vm_area_struct *); 228 229 /* 230 * Cleans the PTEs of shared mappings. 231 * (and since clean PTEs should also be readonly, write protects them too) 232 * 233 * returns the number of cleaned PTEs. 234 */ 235 int page_mkclean(struct page *); 236 237 /* 238 * called in munlock()/munmap() path to check for other vmas holding 239 * the page mlocked. 240 */ 241 int try_to_munlock(struct page *); 242 243 /* 244 * Called by memory-failure.c to kill processes. 245 */ 246 struct anon_vma *page_lock_anon_vma_read(struct page *page); 247 void page_unlock_anon_vma_read(struct anon_vma *anon_vma); 248 int page_mapped_in_vma(struct page *page, struct vm_area_struct *vma); 249 250 /* 251 * rmap_walk_control: To control rmap traversing for specific needs 252 * 253 * arg: passed to rmap_one() and invalid_vma() 254 * rmap_one: executed on each vma where page is mapped 255 * done: for checking traversing termination condition 256 * anon_lock: for getting anon_lock by optimized way rather than default 257 * invalid_vma: for skipping uninterested vma 258 */ 259 struct rmap_walk_control { 260 void *arg; 261 int (*rmap_one)(struct page *page, struct vm_area_struct *vma, 262 unsigned long addr, void *arg); 263 int (*done)(struct page *page); 264 struct anon_vma *(*anon_lock)(struct page *page); 265 bool (*invalid_vma)(struct vm_area_struct *vma, void *arg); 266 }; 267 268 int rmap_walk(struct page *page, struct rmap_walk_control *rwc); 269 270 #else /* !CONFIG_MMU */ 271 272 #define anon_vma_init() do {} while (0) 273 #define anon_vma_prepare(vma) (0) 274 #define anon_vma_link(vma) do {} while (0) 275 276 static inline int page_referenced(struct page *page, int is_locked, 277 struct mem_cgroup *memcg, 278 unsigned long *vm_flags) 279 { 280 *vm_flags = 0; 281 return 0; 282 } 283 284 #define try_to_unmap(page, refs) SWAP_FAIL 285 286 static inline int page_mkclean(struct page *page) 287 { 288 return 0; 289 } 290 291 292 #endif /* CONFIG_MMU */ 293 294 /* 295 * Return values of try_to_unmap 296 */ 297 #define SWAP_SUCCESS 0 298 #define SWAP_AGAIN 1 299 #define SWAP_FAIL 2 300 #define SWAP_MLOCK 3 301 #define SWAP_LZFREE 4 302 303 #endif /* _LINUX_RMAP_H */ 304