include/linux/rmap.h

   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          * NOTE: the LSB of the rb_root.rb_node is set by
  41          * mm_take_all_locks() _after_ taking the above lock. So the
  42          * rb_root must only be read/written after taking the above lock
  43          * 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          */
  47         struct rb_root rb_root; /* Interval tree of private "related" vmas */
  48 };
  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.
  60  * The "rb" field indexes on an interval tree the anon_vma_chains
  61  * which link all the VMAs associated with this anon_vma.
  62  */
  63 struct 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 */
  67         struct rb_node rb;                      /* locked by anon_vma->rwsem */
  68         unsigned long rb_subtree_last;
  69 #ifdef CONFIG_DEBUG_VM_RB
  70         unsigned long cached_vma_start, cached_vma_last;
  71 #endif
  72 };
  73
  74 enum ttu_flags {
  75         TTU_UNMAP = 1,                  /* unmap mode */
  76         TTU_MIGRATION = 2,              /* migration mode */
  77         TTU_MUNLOCK = 4,                /* munlock mode */
  78
  79         TTU_IGNORE_MLOCK = (1 << 8),    /* ignore mlock */
  80         TTU_IGNORE_ACCESS = (1 << 9),   /* don't age */
  81         TTU_IGNORE_HWPOISON = (1 << 10),/* corrupted page is recoverable */
  82 };
  83
  84 #ifdef CONFIG_MMU
  85 static inline void get_anon_vma(struct anon_vma *anon_vma)
  86 {
  87         atomic_inc(&anon_vma->refcount);
  88 }
  89
  90 void __put_anon_vma(struct anon_vma *anon_vma);
  91
  92 static inline void put_anon_vma(struct anon_vma *anon_vma)
  93 {
  94         if (atomic_dec_and_test(&anon_vma->refcount))
  95                 __put_anon_vma(anon_vma);
  96 }
  97
  98 static inline struct anon_vma *page_anon_vma(struct page *page)
  99 {
 100         if (((unsigned long)page->mapping & PAGE_MAPPING_FLAGS) !=
 101                                             PAGE_MAPPING_ANON)
 102                 return NULL;
 103         return page_rmapping(page);
 104 }
 105
 106 static inline void vma_lock_anon_vma(struct vm_area_struct *vma)
 107 {
 108         struct anon_vma *anon_vma = vma->anon_vma;
 109         if (anon_vma)
 110                 down_write(&anon_vma->root->rwsem);
 111 }
 112
 113 static inline void vma_unlock_anon_vma(struct vm_area_struct *vma)
 114 {
 115         struct anon_vma *anon_vma = vma->anon_vma;
 116         if (anon_vma)
 117                 up_write(&anon_vma->root->rwsem);
 118 }
 119
 120 static inline void anon_vma_lock_write(struct anon_vma *anon_vma)
 121 {
 122         down_write(&anon_vma->root->rwsem);
 123 }
 124
 125 static inline void anon_vma_unlock_write(struct anon_vma *anon_vma)
 126 {
 127         up_write(&anon_vma->root->rwsem);
 128 }
 129
 130 static inline void anon_vma_lock_read(struct anon_vma *anon_vma)
 131 {
 132         down_read(&anon_vma->root->rwsem);
 133 }
 134
 135 static inline void anon_vma_unlock_read(struct anon_vma *anon_vma)
 136 {
 137         up_read(&anon_vma->root->rwsem);
 138 }
 139
 140
 141 /*
 142  * anon_vma helper functions.
 143  */
 144 void anon_vma_init(void);       /* create anon_vma_cachep */
 145 int  anon_vma_prepare(struct vm_area_struct *);
 146 void unlink_anon_vmas(struct vm_area_struct *);
 147 int anon_vma_clone(struct vm_area_struct *, struct vm_area_struct *);
 148 int anon_vma_fork(struct vm_area_struct *, struct vm_area_struct *);
 149
 150 static inline void anon_vma_merge(struct vm_area_struct *vma,
 151                                   struct vm_area_struct *next)
 152 {
 153         VM_BUG_ON(vma->anon_vma != next->anon_vma);
 154         unlink_anon_vmas(next);
 155 }
 156
 157 struct anon_vma *page_get_anon_vma(struct page *page);
 158
 159 /*
 160  * rmap interfaces called when adding or removing pte of page
 161  */
 162 void page_move_anon_rmap(struct page *, struct vm_area_struct *, unsigned long);
 163 void page_add_anon_rmap(struct page *, struct vm_area_struct *, unsigned long);
 164 void do_page_add_anon_rmap(struct page *, struct vm_area_struct *,
 165                            unsigned long, int);
 166 void page_add_new_anon_rmap(struct page *, struct vm_area_struct *, unsigned long);
 167 void page_add_file_rmap(struct page *);
 168 void page_remove_rmap(struct page *);
 169
 170 void hugepage_add_anon_rmap(struct page *, struct vm_area_struct *,
 171                             unsigned long);
 172 void hugepage_add_new_anon_rmap(struct page *, struct vm_area_struct *,
 173                                 unsigned long);
 174
 175 static inline void page_dup_rmap(struct page *page)
 176 {
 177         atomic_inc(&page->_mapcount);
 178 }
 179
 180 /*
 181  * Called from mm/vmscan.c to handle paging out
 182  */
 183 int page_referenced(struct page *, int is_locked,
 184                         struct mem_cgroup *memcg, unsigned long *vm_flags);
 185
 186 #define TTU_ACTION(x) ((x) & TTU_ACTION_MASK)
 187
 188 int try_to_unmap(struct page *, enum ttu_flags flags);
 189
 190 /*
 191  * Called from mm/filemap_xip.c to unmap empty zero page
 192  */
 193 pte_t *__page_check_address(struct page *, struct mm_struct *,
 194                                 unsigned long, spinlock_t **, int);
 195
 196 static inline pte_t *page_check_address(struct page *page, struct mm_struct *mm,
 197                                         unsigned long address,
 198                                         spinlock_t **ptlp, int sync)
 199 {
 200         pte_t *ptep;
 201
 202         __cond_lock(*ptlp, ptep = __page_check_address(page, mm, address,
 203                                                        ptlp, sync));
 204         return ptep;
 205 }
 206
 207 /*
 208  * Used by swapoff to help locate where page is expected in vma.
 209  */
 210 unsigned long page_address_in_vma(struct page *, struct vm_area_struct *);
 211
 212 /*
 213  * Cleans the PTEs of shared mappings.
 214  * (and since clean PTEs should also be readonly, write protects them too)
 215  *
 216  * returns the number of cleaned PTEs.
 217  */
 218 int page_mkclean(struct page *);
 219
 220 /*
 221  * called in munlock()/munmap() path to check for other vmas holding
 222  * the page mlocked.
 223  */
 224 int try_to_munlock(struct page *);
 225
 226 /*
 227  * Called by memory-failure.c to kill processes.
 228  */
 229 struct anon_vma *page_lock_anon_vma_read(struct page *page);
 230 void page_unlock_anon_vma_read(struct anon_vma *anon_vma);
 231 int page_mapped_in_vma(struct page *page, struct vm_area_struct *vma);
 232
 233 /*
 234  * rmap_walk_control: To control rmap traversing for specific needs
 235  *
 236  * arg: passed to rmap_one() and invalid_vma()
 237  * rmap_one: executed on each vma where page is mapped
 238  * done: for checking traversing termination condition
 239  * file_nonlinear: for handling file nonlinear mapping
 240  * anon_lock: for getting anon_lock by optimized way rather than default
 241  * invalid_vma: for skipping uninterested vma
 242  */
 243 struct rmap_walk_control {
 244         void *arg;
 245         int (*rmap_one)(struct page *page, struct vm_area_struct *vma,
 246                                         unsigned long addr, void *arg);
 247         int (*done)(struct page *page);
 248         int (*file_nonlinear)(struct page *, struct address_space *, void *arg);
 249         struct anon_vma *(*anon_lock)(struct page *page);
 250         bool (*invalid_vma)(struct vm_area_struct *vma, void *arg);
 251 };
 252
 253 int rmap_walk(struct page *page, struct rmap_walk_control *rwc);
 254
 255 #else   /* !CONFIG_MMU */
 256
 257 #define anon_vma_init()         do {} while (0)
 258 #define anon_vma_prepare(vma)   (0)
 259 #define anon_vma_link(vma)      do {} while (0)
 260
 261 static inline int page_referenced(struct page *page, int is_locked,
 262                                   struct mem_cgroup *memcg,
 263                                   unsigned long *vm_flags)
 264 {
 265         *vm_flags = 0;
 266         return 0;
 267 }
 268
 269 #define try_to_unmap(page, refs) SWAP_FAIL
 270
 271 static inline int page_mkclean(struct page *page)
 272 {
 273         return 0;
 274 }
 275
 276
 277 #endif  /* CONFIG_MMU */
 278
 279 /*
 280  * Return values of try_to_unmap
 281  */
 282 #define SWAP_SUCCESS    0
 283 #define SWAP_AGAIN      1
 284 #define SWAP_FAIL       2
 285 #define SWAP_MLOCK      3
 286
 287 #endif  /* _LINUX_RMAP_H */